CDX4 AND RETINOIC ACID INTERACT TO POSITION THE HINDBRAIN-SPINAL CORD TRANSITION

Abstract

The sub-division of the posterior-most territory of the neural plate results in the formation of two distinct neural structures, the hindbrain and the spinal cord. Although many of the molecular signals regulating the development of these individual structures have been elucidated, the mechanisms involved in delineating the boundary between the hindbrain and spinal cord remain elusive. Two molecules, retinoic acid (RA) and the Cdx4 transcription factor have been previously implicated as important regulators of hindbrain and spinal cord development, respectively. Here, we provide evidence that suggests multiple regulatory interactions occur between RA signaling and the Cdx4 transcription factor to establish the anterior-posterior (AP) position of the transition between the hindbrain and spinal cord. Using chemical inhibitors to alter RA concentrations and morpholinos to knock-down Cdx4 function in zebrafish, we show that Cdx4 acts to prevent RA degradation in the presumptive spinal cord domain by suppressing expression of the RA degradation enzyme, Cyp26a1. In the hindbrain, RA signaling modulates its own concentration by activating the expression of cyp26a1 and inhibiting the expansion of cdx4. Therefore, interactions between Cyp26a1 and Cdx4 modulate RA levels along the AP axis to segregate the posterior neural plate into the hindbrain and spinal cord territories.

INTRODUCTION

Following the initial induction of neural tissue, molecular cues act to partition the neural tube along the anterior-posterior (AP) axis to delineate four discrete territories of the nascent nervous system (Stern, 2001, Melton et al., 2004), the forebrain, midbrain, hindbrain and spinal cord. In contrast to the forebrain and midbrain, which are segregated by junctions that also serve as organizing centers (Martinez-Barbera et al., 2001; Rhinn and Brand, 2001), no morphological boundaries have been detected that separate the hindbrain territory from the spinal cord. However, the transition zone between the hindbrain and spinal cord territories can be distinguished by the position of tissue specific neurons and the expression of Hox genes, with the group 1–4 Hox genes expressed within the native hindbrain region and the group 5–13 Hox genes expressed throughout regions of the spinal cord (Gaunt et al., 1989; Gruss and Kressel, 1991; Hunt et al., 1991; Nolte et al., 2006; Prince et al., 1998a and b).

Due to their collinear and nested expression in the hindbrain and spinal cord and their established importance in conferring identity to various tissues, the regulatory mechanisms governing regionalization of the hindbrain and spinal cord have been largely attributed to the function of Hox genes (Reviewed in Krumlauf et al., 1993; Schilling and Knight, 2001). However, alteration in hindbrain and spinal cord size or shifts of the hindbrain-spinal cord transition has so far not been reported in functional studies of Hox genes (Carpenter et al., 1993; Chisaka et al., 1991). Consequently, these findings suggest that additional mechanisms regulate the subdivision of the neural tissue into the domains of the hindbrain and spinal cord.

In contrast, perturbation of Retinoic Acid (RA) signaling, a crucial regulator of Hox gene expression in the hindbrain affects both hindbrain and spinal cord development. This is observed in embryos exposed to excess RA where the rostral expansion of posterior hindbrain Hox gene expression is also accompanied by a rostral expansion of spinal cord motor neurons (Emoto et al., 2005; Hernandez et al., 2007; Marshall et al., 1992; Wood et al., 1994) suggesting that the underlying phenotype of elevating RA signaling could be a rostral shift of the hindbrain-spinal cord (Hb-Sc) transition. Alternatively, these results could also indicate that increased RA levels may induce a transformation of cell fate from hindbrain to spinal cord. Previous findings employing RA reporters have suggested that, in the neural tube, RA signaling is present in the spinal cord and at a higher level than in the hindbrain (Maden et al., 1998; Shimozono et al., 2013; Wagner et al., 1992). Since cells of both the posterior hindbrain and spinal cord are located adjacent to the somitic mesoderm source of RA, the concentration difference between these two regions has been partly attributed to the function of Cyp26, a family of RA degradation enzymes expressed exclusively in the hindbrain (Gu et al., 2005; Hernandez et al., 2007; White et al., 2007). For example, Cyp26-deficient embryos have a smaller hindbrain territory and a rostrally expanded spinal cord (Emoto et al., 2005; Hernandez et al., 2007), a similar phenotype found in embryos treated with excess RA (Durston et al., 1989; Marshall et al., 1992).

In addition to RA signaling, perturbation of CDX/Caudal, a family of transcription factors in vertebrates having at least one family member expressed in the spinal cord (Bel-Vialar et al., 2002; Charite et al., 1998; Isaacs et al., 1998; Shimizu et al., 2006; Skromne et al., 2007), also leads to AP positional shifts of the Hb-Sc transition. For example, cdx4 zebrafish mutants and morphants have caudally expanded hindbrains and reduced spinal cords (Shimizu et al., 2006; Skromne et al., 2007), congruent with a caudal-ward shift of the Hb-Sc transition. In contrast, loss of Cdx1a (Skromne et al., 2007) does not result in the mis-alignment of the Hb-Sc transition. Therefore, this Hb-Sc transition shift observed in Cdx4 deficient embryos suggests that Cdx4 may also be involved in delineating the sizes of the hindbrain versus spinal cord territories; however the regulatory role of Cdx4 and how it interacts with RA signaling in directing this process remains to be established.

Using various chemical inhibitors and genetic manipulations to alter RA signaling and Cdx4 function, we show that coordinated modulation of RA signaling at the Hb-Sc transition is required for proper AP localization of the Hb-Sc transition. Through gene expression analysis, we show that this is achieved through reciprocal regulation between Cdx4, which inhibits the expression of the RA degradation enzyme, cyp26a1, in the spinal cord, and RA, which suppresses cdx4 expression in the hindbrain. Further comparison between the arrangement of hox gene expressions and motor neuron positions in our epistasis experiments indicates that RA signaling and Cdx4 act to set the position of the Hb-Sc transition and do so independently from their roles in regulating hox gene expression.

MATERIALS AND METHODS

Fish husbandry, microinjections and pharmacological treatments

Zebrafish (Danio rerio) were raised and handled using standard protocol. Embryos were collected from *AB stock fish, Tg(isl1:gfp) (Higashijima et al., 2000) or gir^rw716 (Cyp26a1 mutant, Emoto et al., 2005), grown at 28°C in embryo media and staged following Kimmel et al., 1995. Microinjections of cdx4 morpholinos were performed at the 1 cell stage as previously described using standard injection protocol (20ng/ul; Skromne et al., 2007). Uninjected and Cdx4 morphants were incubated in the dark with 10 uM DEAB (inhibitor of RA synthesis; Aldrich) starting at 5.3 hours post-fertilization (hpf) or 10 uM R115866 (Hernandez et al., 2007; Cyp26 inhibitor; JanssenPharmaceutica) starting at 4.3 hpf. Control embryos were incubated with the vehicle, 0.1% DMSO in embryo media.

Whole mount in situ hybridization and antibody staining

Gene expression was detected by standard in-situ protocol using NBT/BCIP and Fast Red as the enzyme substrate. Antisense RNA labeled with DIG or FITC were generated for cdx4 (Skromne et al., 2007); cyp26a1, cyp26b1, cyp26c1 (Hernandez et al., 2007); hoxa2b, hoxb1a, hoxb3a, hoxb4a, hoxd4a, hoxb8a, (Prince et al., 1998a and b); krx20 (Skromne et al., 2007); myoD (Lee and Skromne, 2014); radical fringe (Skromne et al., 2007); mafb (Moens et al. 1998). Antibody staining was performed following previously described method (Skromne et al., 2007). Mouse anti-myosin heavy chain (A4.1025, Developmental Studies, Hybridoma Bank, IA, USA); mouse anti-neurofilament 169K (RMO44, Zymed, CA, USA); mouse zn5 antibody (Trevarrow et al., 1990); polycolonal rabbit anti-GFP (Invitrogen); mouse anti-acetylated tubulin (Sigma-Aldrich); were used at a dilution of 1:100, 1:50, 1:1000, 1:500 and 1:500 respectively. Anti-mouse Alexa-488 and Alexa-546 secondary antibodies were used at a concentration of 1:2000 (Skromne et al., 2007).

Imaging and embryo processing

Embryos processed for in-situ hybridization and antibody staining were deyolked and flat mounted on slides. Fluorescent images were taken using a Zeiss LSM 710 confocal microscope and images were processed using ImageJ. Additional images were collected on a Zeiss compound microscope using a Nikon-5000 camera. Images were compiled using Adobe Photoshop CS5.1.

RESULTS

Cdx4 and RA have opposite activities in the specification and patterning of hindbrain and spinal cord territories

Although previous papers have reported shifts in hox gene expression in Cdx4 and RA deficient embryos, no findings so far have measured and carefully compared the shifts in patterning gene expression to changes in the size of the hindbrain and spinal cord territory. To begin to understand how Cdx4 and RA regulate AP specification and patterning of the hindbrain and spinal cord territories, we first analyzed the spatial distribution of hindbrain and spinal cord hox gene transcripts and motor neuron populations relative to somite landmarks in embryos deficient for Cdx4 or RA signaling.

As previously reported (Davidson and Zon, 2006; Shimizu et al., 2006; Skromne et al., 2007), loss of Cdx4 had no effect on the expression of anterior hox genes (hoxa2b, and hoxb3a; Fig. 1A, B, E, F), but the expression of a posterior hindbrain gene (hoxb4a, Fig. 1C, G) is caudally expanded by 2-somite lengths into what is normally the spinal cord domain. Concurrently, loss of Cdx4 results in caudally shifted spinal cord hoxb8a expression (Fig. 1D, H), suggesting a caudal-ward expansion of the hindbrain with a concurrent caudal reduction of the spinal cord. These early changes in hox expression were mirrored by late changes in motor neuron architecture. For example, loss of Cdx4 caused a 2-somite length caudal expansion in the distribution of isl1-positive hindbrain vagal motor neurons (Fig. 2C), and an equal 2-somite length caudal shift in the position of Zn5-positive spinal motor neurons (Fig. 2D). Thus, Cdx4 is necessary for the anterior placement of the Hb-Sc transition.

Fig. 1. Hindbrain and spinal cord hox gene expression shifts along the AP-axis when Cdx4 or RA functions are perturbed.

Expression of hoxa2b in r2–4 and hoxb3a in r5–6 is the same in Cdx4-deficient embryos (E,F) and wild-types (A,B). hoxb4a expression in r7/8 is caudally expanded in Cdx4-deficient embryos (G arrowhead) compared to wildtypes (C arrowhead). In RA-deficient embryos hoxb3a (J) and hoxb4a (K) expression is lost, while hoxa2b expression in r4 is caudally expanded (I arrowhead). In RA/Cdx4 deficient embryos, hoxb3a expression is rescued (N) but not hoxb4a (O). In both wild type and Cdx4 deficient embryos treated with R115866 hoxa2b (Q,U) expression is lost while hoxb3a (R,V) and hoxb4a (S,W) expression are both rostrally expanded.. The anterior expression limit of spinal cord hoxb8a is located at somite 3 in wildtypes (D arrowhead) and RA deficient embryos (L arrowhead) but its expression is caudally shifted to somite 5 in Cdx4 deficient embryos (H arrowhead) and RA/Cdx4 deficient embryos (P arrowhead). hoxb8a expression expands into the hindbrain in both wild type (T arrowheads) and Cdx4 deficient embryos (X arrowhead) treated with R115866. (A–X) are co-stained with krx20 (r3,r5). (C–D,G–H,K–L,O–P,S–T,W–X) co-stained with myosin heavy chain antibody (somites). (A–X) flat-mounted embryos at 20 somite stage n=45/45 per condition.

Fig. 2. Hindbrain and spinal cord neural populations shift along the AP-axis when Cdx4 or RA functions are perturbed.

(A–F) Branchial motor neurons marked by Islet1-Gfp and r3,r5 visualized by anti-Epha4. The trigeminals(Va/Vp) in r2–3 and facials(VII) in r6 are unchanged but the vagals(X) in r7/8 are caudally expanded in Cdx4-deficient embryos (C) compared to wildtypes (A). Vagals (r7/8) and Epha4 (r5) are lost in RA-deficient embryos (E). Epha4 in r5 and Islet1-Gfp cells in r5–8 are rescued in RA/Cdx4 deficient embryos (G). Vagals are rostrally shifted while trigeminals and facials are absent in both wildtypes (I) and Cdx4-deficient embryos (K) treated with R115866. (G–L arrowheads) Spinal cord motor neurons marked by Zn5 antibody. The first spinal cord motor neuron arises at the level of somite 3 in wildtypes (B arrow) but at the level of somite 5 in Cdx4-deficient embryos (D arrow). Spinal cord motor neurons are lost in RA-deficient (F) and RA/Cdx4 deficient (H) embryos. Spinal cord motor neurons expand rostrally in both wildtypes (J) and Cdx4-deficient embryos (L) treated with R115866. (A–L) Flat-mounted embryos at 48 hpf. (A–F) n=30/30 per condition and (G–L) n=15/15 per condition.

Next we examined the function of RA on the Hb-Sc transition by exposing embryos to the RA synthesis inhibitor DEAB starting at the beginning of gastrulation (5.3 hpf). Consistent with previous reports (Begemann et al., 2001; Grandel et al., 2002), RA deficient embryos had a slight caudal expansion of anterior hindbrain gene hoxa2b, and the loss of more posterior hindbrain genes hoxb3a and hoxb4a (Fig. 1I–K), indicating the loss of posterior hindbrain. These changes in hindbrain patterning were confirmed by the failure of facial motor neurons to migrate out of r4 and by the loss of vagal motor neurons in r7/8 (Fig. 2E, 3V). RA-inhibition treatments, however, did not change spinal cord hoxb8a expression, which remained at somite 3 (Fig. 1L, S1C,D), suggesting that RA may not be necessary for specifying the AP position of the spinal cord territory. In order to confirm this, we analyzed the position of the spinal motor neurons. However, spinal motor neurons do not develop in these embryos, as RA is essential for spinal motor neuron development (Novitch et al., 2003; Pierani et al., 1999). To circumvent this limitation we restricted DEAB treatments to gastrulation, as this treatment impacts specification of tissues without blocking spinal motor neuron development (Lee and Skromne, 2014). Loss of RA under these conditions likewise failed to change hoxb8a expression at 24 and 48 hpf but shifted the position of the first spinal cord motor neuron rostrally from somite 3 to 2 (Figs. 1L, S1A–D), suggesting that, contrary to the hox expression analysis, the spinal cord territory was indeed shifted rostrally. This phenotype is the opposite to what is seen in embryos lacking Cdx4, indicating that Cdx4 and RA have antagonistic effects in regulating the AP position of the Hb-Sc transition: along the primary axis, Cdx4 functions to position the transition more rostrally, and RA functions to place the transition more caudally.

Fig. 3. Knockdown of Cdx4 in RA-deficient embryos rescues the posterior hindbrain.

hoxb1a expression in r4 is slightly expanded in RA-deficient embryos (B) compared to wildtypes (A) and Cdx4-deficient embryos (C). Expression of krx20 in r5; mafb in r5,r6; hoxd4a and cyp26a1 in r7/8 are lost in RA-deficient embryos (F,N,R) compared to wildtypes ( E,M,Q) and Cdx4-deficient embryos (G,O,S). In RA/Cdx4 deficient embryos hoxb1a (D) expression is reduced, krx20 in r5 (D,H,L,P), mafb in r5,r6 (H) is rescued but hoxd4a (P) and cyp26a1 (T) expression in r7/8 is still absent. Six radical fringe (rf) bands marking rhombomere boundaries are found in wildtypes (I) and Cdx4-deficient embryos (K). Three rf bands are detected in RA-deficient embryos (J), while two additional bands, flanking krx20 expression in r5, return in RA/Cdx4 deficient embryos (L arrows). Prickle1b expression is in facial motor neurons of r4–6 in wildtypes (U bracket), cdx4 morphants (W bracket) and RA deficient embryos (V bracket). Prickle1b expression is expanded into the r7/8 position in Cdx4/RA double deficient embryos (X bracket). (Y–Z,AA,BB) Lateral view of hindbrain motor neurons marked by Islet:gfp and r3,5 marked with epha4. Branchal motor neurons in cdx4 morphants (AA) appear similar to wiltdtypes (Y). Vagal motor neurons of r7/8 are absent in RA deficient embryos (Z). Islet:gfp neurons return to r7/8 in Cdx4/RA double deficient embryos (BB Arrows). RMO44 antibody marks reticulospinal neurons in wild-types (CC) and Cdx4-deficient embryos (EE). Mi and CaD are absent in RA-deficient embryos (DD) but return in Cdx4/RA deficient embryos (FF arrowheads). (A–P) are co-stained with krx20 (r3,r5) and (E–H, Q–T) are co-stained with myosin heavy chain antibody (somites). All embryos are flat-mounted, (A–T) at 20 somite stage and (U–Z,AA–FF) at 48 hpf. (A–D) n=30/30, (E–H) n=15/15, (I–T) n=15/15, (U–X) n=20/20, (Y–Z,AA–BB) n=15/15, (CC–FF) n=13/15 per condition.

Cdx4 and RA signaling interactions specify and pattern the hindbrain territory

Our analyses of Cdx4 and RA deficient embryos indicate that these two factors have opposing effects on hindbrain-spinal cord development and AP axial positioning of the Hb-Sc transition. Therefore, we next addressed whether these two factors genetically interact in regulating the development of these two tissues. To investigate the relationship between Cdx4 and RA, we treated Cdx4-deficient embryos with DEAB and then analyzed the spatial distribution of hox transcripts and neuronal populations. In the hindbrain, loss of Cdx4 partially rescued the defects associated with the loss of RA, in that hoxb3a expression is restored, while hoxb4a is not (Fig. 1I–K, M–O). This was also seen at the morphological level with the reappearance and expansion of several branchial motor neuron populations (Fig. 2G). Importantly, the territory occupied by these hindbrain neuronal populations is expanded caudally into the spinal cord domain, similar to the expansion observed in Cdx4-deficient embryos indicating a caudal shift in the Hb-Sc transition. Consistent with this shift was the observation that in Cdx4/RA deficient embryos, the expression of spinal cord marker hoxb8a was also shifted caudally (Fig. 1P). The observation that the loss of Cdx4 caused a similar posterior expansion of the hindbrain and reduction of the spinal cord irrespective of the presence or absence of RA suggests that Cdx4 is epistatic to RA signaling in establishing the AP position of the Hb-Sc transition. In addition, loss of Cdx4 partially rescues rhombomere patterning defects associated with the loss of RA, suggesting that RA and Cdx4 have antagonistic functions in regulating hindbrain patterning.

In the hindbrain, RA is essential for rhombomere formation and patterning (Begemann et al., 2001; Grandel et at., 2002). Therefore, it was significant to find that loss of Cdx4 could partially rescue the defects induced by loss of RA (Fig. 1M–P). To further characterize the extent of this rescue and better understand Cdx4 and RA interactions during rhombomere formation and patterning, we assayed Cdx4/RA deficient embryos with a repertoire of rhombomere markers. In RA-deficient embryos, hoxb1a expression in r4 expanded caudally while krx20 expression in r5 and mafb in r5–6 were lost (Fig. 3B, F). However, in Cdx4/RA deficient embryos, krx20 expression in r5 and mafb expression in r5–6 was rescued while hoxb1a expression was moderately reduced (Fig. 3D,H). Furthermore, rescued mafb expression was caudally expanded with its expression terminating at the level of somite 5, suggesting that the rescued hindbrain is expanded beyond its normal size (Fig. 3H arrow). Expression analysis of the r7/8 markers hoxd4a and cyp26a1 showed that Cdx4 failed to rescue their transcription in RA-deficient embryos (Fig. 3M,P,Q,T), indicating that the rescue was limited to r5/r6 genes. To corroborate the rhombomere rescue, we analyzed the expression of radical fringe (rfng), which in wild-type embryos is expressed in six distinctive stripes marking boundaries between each rhombomere (Fig. 3I). In RA-deficient embryos, rfng was expressed in three stripes that correspond to the boundaries between r1–4 (Fig. 3J). In Cdx4/RA deficient embryos, two of the three posterior boundaries lost in RA-deficient embryos are rescued (Fig. 3J,L arrows), suggesting the formation of all rhombomeres except r7/r8. This rescue was also observed at the morphological level. Using the Tg(isl:gfp) line we found that the rescued motor neurons, despite occupying the position of r7/8 and the anterior spinal cord, appear to resemble r4–6 facial motor neurons in their morphology and neuron like projections (Fig. 3BB arrows). To confirm that these neurons were facial motor neurons we stained Cdx4/RA deficient embryos with the facial motor neuron specific marker, prickle1b (Rohrschneider et al., 2007), finding that prickle1b expression was ectopically expressed in the rescued hindbrain region, (Fig. 3U–X) further indicating that these rescued hindbrain segments develop r5–6 like properties. Additional motor neuron analysis shows that RMO44-positive Mi3 reticulospinal cord neurons located in wild type r5-r6 (Mendelson, 1986a), but lost in RA-deficient embryos (Fig. 3DD), were rescued by the additional loss of Cdx4 (Fig. 3FF arrowheads). Together, these results show that loss of Cdx4 partially alleviates the hindbrain phenotypes caused by the loss of RA, namely, the mid and some of the posterior hindbrain patterning defects, and consequently leads to the partial restoration of lost rhombomeres. However, despite the absence of the last rhombomere in Cdx4/RA deficient embryos, the hindbrain territory itself is still caudally expanded at the expense of the spinal cord indicating that the interaction between RA signaling and Cdx4 regulates the size of the hindbrain territory and the AP positioning of the Hb-Sc transition, while additional signaling mechanisms are required for directing complete patterning of the hindbrain. Thus, these results suggest that specification and patterning of the hindbrain and spinal cord regions are directed by separable regulatory mechanisms.

Cdx4 and RA signaling pathway components cross regulate each other’s transcription

The opposite activities of Cdx4 and RA in hindbrain spinal cord specification raised the possibility that they cross-regulate each other’s expression or activity. To explore this possibility we examined the impact of eliminating Cdx4 upon components of the RA signaling pathway and the impact of altering RA signaling on cdx4 expression. We first focused on the effect Cdx4 activity has upon the RA signaling pathway by analyzing the expression of enzymes involved in RA synthesis and degradation. RA is synthesized in the paraxial mesoderm by the enzyme aldh2a (Begemann et al., 2001; Grandel et at., 2002) and degraded in the hindbrain by the enzymes cyp26a1, cyp26b1 and cyp26c1 (Emoto et al., 2005; Hernandez et al., 2007; White et al., 2007). In Cdx4-deficient embryos, we corroborated that aldh2a expression domain was at the same axial level as in wild types, with a slight caudal expansion (Shimizu et al., 2006; Skromne et al., 2007; Wingert et al., 2007). In contrast, the expression domain of cyp26a1, but not cy26b1 or cyp26c1, was expanded caudally into what ought to be the spinal cord territory in Cdx4-deficient embryos (Fig. 4A, B, E–H). cyp26a1 is the RA degradation enzyme that is most caudally expressed in r7/8 of the hindbrain and its shift in expression suggests that Cdx4 represses its transcription (Fig. 4A, B). To further test this, we ectopically induced cdx4 expression from a heat inducible transgene (Skromne et al., 2007). Transgenic induction of cdx4 at tailbud stage caused the loss of cyp26a1 expression at 10-somite stage, confirming that Cdx4 represses cyp26a1 expression (Fig. 4C, D). These results suggest that Cdx4 inhibits the expression of the RA inhibitor cyp26a1, thus allowing high levels of RA signaling in the spinal cord.

Fig. 4. Cyp26a1 expression at the Hb-Sc transition is under the regulation of Cdx4.

In wild-types, cyp26a1 is expressed in r7/8, and in the adjacent somites 1–3 (A arrowhead). In Cdx4-deficient embryos (B), cyp26a1 expression is expanded into the spinal cord domain and terminates at the level of somite 5 (B arrowhead). Overexpression of cdx4 using Tg(hsp70:cdx4) line (C) leads to abrogation of cyp26a1 expression (D arrowheads). Cyp26b1 expression (E) in r3–4 and cyp26c1 expression (G) in r2,r4,r6 remain unaffected in Cdx4-deficient embryos (F,H). (A–H) Embryos are co-stained with krx20 (r3,r5), myoD (somites) and flat-mounted at the 10 somite stage. (A–B) n=30/30 (C–D) n=15/15, (E–H) n=30/30, (I–J) n=10/10 per condition.

To determine if RA regulates cdx4 transcription, we examined cdx4 expression in RA-deficient embryos at 10-somite stage, when cdx4 expression is normally restricted to the prospective spinal cord territory between somite 3 to the tailbud (Fig. 5A; Skromne et al., 2007). Loss of RA caused the rostral expansion of cdx4 expression domain into an embryonic region that in the wild type embryo would become hindbrain (Fig. 5B). This result suggests that RA normally inhibits cdx4 expression in the hindbrain. To test this idea, we characterized the expression of cdx4 in embryos overexpressing RA by systematically eliminating cyp26 activities by using previously published combinations of mutant and morpholino approaches (Emoto et al., 2005; Hernandez et al., 2007). Significantly, the individual or combined elimination of Cyp26-a1, -b1, and –c1 did not alter the expression domain of cdx4 (Fig. 5C–G). Similar findings to the individual loss of Cyp26 activities were obtained by exposing embryos to the pan-Cyp26 inhibitor R115866 to increase endogenous RA levels (Fig. 5H, I) or by treating the embryos with exogenous RA (Fig. 5J–L), Together these gain and loss of function experiments suggest a context-dependent regulation of cdx4 by RA whereby RA is necessary for restricting cdx4 expression to the prospective spinal cord territory and preventing its expansion into the hindbrain, but is not sufficient to repress cdx4 expression within the spinal cord.

Fig. 5. Tissue specific regulation of cdx4 expression by RA signaling.

In wildtypes, cdx4 anterior expression limit corresponds to the beginning of the spinal cord located at the level of somite 3 (A,C,H,J, arrowheads). cdx4 expression is rostrally expanded to the level of somite 1 (B arrowhead) in RA-deficient embryos. Expression of cdx4 is unchanged in R115866 treated embryos (I arrowhead), in embryos treated with 500 nM RA (K arrowhead) and 900 nM RA (L arrowhead).. cdx4 expression remains unchanged in cyp26a1 mutants (D), cyp26b1 morphants (E), cyp26c1 morphants (F), and cyp26a1 mutants injected with cyp26b1 and cyp26c1 morpholinos (G). (A–G,J–L) Embryos are co-stained with myoD (somites) and krx20 (r3,r5). (H–J) Embryos co-stained with myoD. (A–L) Embryos are at 10 somite stage and flat-mounted. (A–B,H–I) n=30/30, (C–G) n > 12, (J–L) n=20/20 per condition.

RA, Cyp26a1 and Cdx4 establish a feedback loop to regulate hox gene expression and establish the axial position of the Hb-Sc transition

It has been shown previously that excess RA leads to a posteriorization of the hindbrain by activating posterior hindbrain hox genes in anterior hindbrain regions (Durston et al., 1989; Emoto et al., 2005; Grandel et al., 2002; Hernandez et al., 2007). In this paper, we show that excess RA perturbs the position of the Hb-Sc transition without affecting the cdx4 expression domain in the spinal cord (Fig. 2I, J, 5I, K, L). In contrast, loss of Cdx4 both perturbs cyp26a1 expression (Fig. 4B) and expands the hindbrain region at the expense of the spinal cord. These observations raise the question of what regulatory impact Cdx4 has on RA signaling in establishing hindbrain and spinal cord identities. To answer this question, we compared the spatial distribution of hox gene transcripts and motor neuron populations in wild type and Cdx4-deficient embryos treated with a pan-Cyp26 inhibitor to block RA degradation. We opted to increase RA levels by this method since it directly changes RA levels in the neural tissue while leaving the paraxial mesoderm, a topographically important structure for our analysis, morphologically unaffected. As previously reported (Emoto et al., 2005; Hernandez et al., 2007), increase in RA signaling due to Cyp26 inhibition caused the loss of hoxa2b expression in the anterior hindbrain (Fig. 1Q), and rostrally expanded the expression of posterior hindbrain markers hoxb3a and hoxb4a (Fig. 1R, S). Cyp26 inhibition also caused the anterior expansion of the spinal cord marker hoxb8a (Fig. 1T). These changes at the gene expression level were also observed at the morphological level, with the loss of trigeminal (r2/r3) and facial (r4-r6) motor neurons and the rostral expansion of vagal (r7/8) and spinal motor neurons (Fig. 2I, J). Thus, failure to restrict RA signaling in the hindbrain causes this tissue to develop improperly, suggesting that Cyp26 activity is required for maintaining separate hindbrain and spinal cord territories.

To test if the shift in hoxb8a expression and spinal motor neurons was dependent on Cdx4, we simultaneously elevated RA levels while eliminating Cdx4. Loss of Cdx4 did not change the phenotypes associated with the loss of Cyp26: we observed equivalent changes in hox gene expression and motor neuron distribution in Cyp26 and Cyp26/Cdx4-deficient embryos (Fig. 1Q–X; 2I–L). The observation that loss of Cyp26/Cdx4 mirrors Cyp26 loss of function phenotype suggests that Cyp26 is epistatic to Cdx4 in defining the AP position of the Hb-Sc transition. Furthermore, these results indicate that increased RA levels through inhibition of Cyp26 can compensate for the reduction of the spinal cord territory observed in Cdx4-deficient embryos. As we have shown that Cdx4 normally acts to inhibit cyp26a1 expression in the anterior spinal cord (Fig. 4B, D), these results imply that Cdx4 has a positive modulatory effect on RA levels to establish the anterior spinal cord and limit the extent of the hindbrain territory. These interactions normally play an important role in the proper positioning of the hindbrain-spinal cord boundary.

DISCUSSION

Our findings show that cross-regulatory interactions between RA signaling and Cdx4 are crucial for the correct partitioning of the posterior neural plate into separate domains of the hindbrain and spinal cord. By comparing the AP position of patterning gene expression and motor neuron architecture in RA and Cdx4 single and double deficient embryos, we show that these two molecular pathways interact antagonistically to regulate hindbrain and spinal cord territory size. This is in addition to the individual roles played by RA and Cdx4 during the organization of the hindbrain and spinal cord, suggesting that the processes specifying territory size and identity of the posterior neural plate are under the control of distinct regulatory mechanisms. Significantly, epistatic experiments analyzing RA signaling and Cdx4 interactions reveal that feedback regulatory loops determine the location of the hindbrain-spinal cord (Hb-Sc) transition. This feedback regulation provides a testable model of interactions for the partitioning of large territories into smaller domains (Fig. 6A).

Fig. 6. Model summarizing the genetic interaction between Cyp26a1, Cdx4 and RA signaling with the AP-position of the Hb-Sc transition.

(A) In wild-type embryos, the hindbrain (light blue) ends, and the spinal cord (dark blue) begin at the level of somite 3 (red ovals). In the posterior hindbrain, cyp26a1 expression required for posterior rhombomere formation (Emoto et al., 2005; Hernandez et al., 2007) is under the dual regulation of RA working as an activator, and Cdx4 working as a repressor. RA also represses cdx4 transcription, but only in the posterior hindbrain and not in the spinal cord. Maintenance of cdx4 expression in the spinal cord may be due to the activity of other somite-derived signals (e.g., FGF, Wnts; Lee and Skromne, 2014). (B) Treatment of embryos with the Cyp26 inhibitor R115866 elevates RA levels and shifts the Hb-Sc transition rostrally; however cdx4 expression in the spinal cord remains unchanged suggesting that RA signaling is not sufficient to inhibit cdx4 expression in the spinal cord. (C) Embryos treated with DEAB are RA deficient and both cdx4 expression and the Hb-Sc transition are shifted rostrally to the level of somite 1 in these embryos suggesting that RA inhibits the expansion of cdx4 expression into the hindbrain. (D) In the absence of Cdx4, cyp26a1 expression and the hindbrain domain is expanded and the Hb-Sc transition is shifted caudally. (E) Cyp26/Cdx4 double deficient embryos phenocopy Cyp26 deficient embryos with an expanded spinal cord territory and a rostrally shifted Hs-Sc transition. This indicates that Cyp26 is epistatic to Cdx4. (F) Cdx4/RA double deficient embryos have an expanded hindbrain territory and a caudally shifted Hb-Sc transition. The AP position of the Hb-Sc transition in Cdx4/RA double deficient and Cdx4 deficient embryos are both located at the level of somite 5 suggesting that Cdx4 is epistatic to RA signaling.

Separate processes direct specification versus patterning of the hindbrain and spinal cord

Loss and gain of function studies of RA signaling and Cdx4 have suggested that these factors contribute to hindbrain and spinal cord development, respectively, through the regulation of hox gene expression (Gould et al., 1998; Marshall et al., 1994; Shimizu et al., 2006; Skromne et al., 2007). Here, we provide evidence for a novel interaction of RA signaling and Cdx4 function in establishing hindbrain and spinal cord territories independent of their hox regulatory roles. Of significance is the rostral shift in position of spinal motor neurons without the alteration of hox gene expression in RA-deficient embryos (Fig. 1L and S1B,D), indicating that the size of the spinal cord territory can be modified without shifting hox gene expression. Further support for the idea that territory specification and patterning are independent events, comes from the finding that the elimination of Cdx4 in an RA-deficient embryo rescues the r5–6 cranial motor neurons lost by the lack of RA without rescuing the expression of caudal hindbrain hox genes (Fig. 2G, 3P, T, BB, FF). Thus, the correlated expression of hox patterning genes is not required for the specification of the posterior neural plate into separate spinal cord and hindbrain territories.

The rescue of posterior hindbrain neuronal populations in Cdx4/RA-deficient embryos that do not express characteristic hox genes in the posterior hindbrain is difficult to reconcile with the extensive evidence supporting RA’s role in regulating posterior hindbrain hox gene expression (reviewed in Gavalas and Krumlauf, 2000; Niederreither et al., 2000), and the role of posterior hindbrain hox genes in specifying posterior hindbrain identities (Dolle et al., 1993; Lufkin et al., 1991; Marshall et al., 1992). We propose that loss of Cdx4 function can compensate for the lack of RA signaling required for the activation of r5–6 patterning gene expression (Fig. 1N, 3H) and suggest that activation of these patterning genes by RA signaling could be mediated, at least in part, through the repression of Cdx4. Consistent with this hypothesis, our expression analyses of RA deficient embryos suggest that RA signaling inhibits cdx4 expression in the posterior hindbrain (Fig. 5B). Cdx transcription factors have been shown to inhibit hindbrain development (Shimizu et al., 2006; Skromne et al. 2007) and in mouse there is evidence to suggest that Cdx factors directly limit the expression of the r5–6 patterning gene Mafb (Sturgeon et al., 2011). Therefore, exclusion of Cdx4 from the hindbrain would be necessary for this tissue’s development. Whether RA relieves the repressive effect that Cdx has on hindbrain patterning genes remains to be determined. It is also possible that part of this effect is through Cdx4-dependent regulation of other RA pathway genes, as shown by the ability of Cdx4 protein to bind to these genes’ enhancer regions (Paik, et al. 2013). In this case, attenuation of Cdx4 function in RA-deficient embryos may de-repress factors controlling RA signaling, thus lowering the threshold requirement and making cells more sensitive to RA (Cai et al., 2013). Consequently, cells would still be able to detect and interpret low levels of RA signaling sufficiently to activate some RA response genes. This could explain why expression of genes found in r5–6 cells, which have been shown to require less RA, are rescued, but not that of r7/8 cells which require a higher concentration of RA for rhombomere-specific gene activation (Dupe and Lumsden, 2001; Gould et al., 1998; Maves and Kimmel, 2005).

Alternatively, the rescue and expansion of the hindbrain territory could indirectly influence the return of r5–6 in Cdx4/RA deficient embryos. In this scenario, hindbrain size reduction caused by the loss of RA may limit the development of posterior hindbrain rhombomeres because not enough territory is available for the allocation of all rhombomere fates. By expanding the hindbrain territory through inhibiting Cdx4 function in RA deficient embryos, adequate space is now available to accommodate r5–6 formation. However in this case, even though the hindbrain territory is greatly expanded to beyond wild-type levels (Figure 1N, 2G, 3H, X), we note that the extent of the rescue in hindbrain fate is limited to r5–6 only and does not include the r7/8 region. Previous studies have shown that RARE sites are required to drive expression of r7/8-specific hox genes (Gould et al., 1998; Nolte et al., 2006; Punnamoottil et al., 2010). The requirement for direct signaling by RA may explain why rescue of r7/8 hox gene expression is not observed in Cdx4/RA deficient embryos despite the rescue of hindbrain territory size (Fig. 2G and 3P,T). Alternatively the signaling pathways required for r7/8 formation may continue to be affected in Cdx4/RA deficient embryos. Since no reports to date have shown that RA signaling regulates r5–6 patterning gene expression directly (Hernandez et al., 2004), in this scenario, RA could be indirectly regulating r5–6 development by maintaining the size of the hindbrain territory. In either case, these findings show that a regulatory interaction between Cdx4 and RA signaling is necessary to establish the territory size of the hindbrain, while RA signaling regulates patterning of r7/8 independently of Cdx4. Thus, these data again suggest that there are distinct mechanisms mediating patterning versus specification of the hindbrain and spinal cord territories.

Regulatory feedback loop between RA, Cdx4 and Cyp26 establishes hindbrain and spinal cord territories

Previous reports have demonstrated that RA signaling modulates its own pathway in the hindbrain by activating the expression of its degradation enzyme, cyp26a1 (Dobbs-McAuliffe et al., 2004; Kinkel et al., 2009; Kudoh et al., 2002; White et al., 2007). This regulatory interaction was suggested to regulate the size of the hindbrain-spinal cord territories (Emoto et al., 2005; Lee and Skromne, 2014). Our results support and expand upon previous findings by examining the molecular interactions between RA signaling and Cdx4 function, to demonstrate that cyp26a1 expression is under the dual regulation of Cdx4 and RA signaling, and that these interactions are essential for establishing the position of the Hb-Sc transition.

Through performing genetic interaction experiments, we identify Cyp26a1 as a key RA signaling pathway component whose regulation by Cdx4 is essential for establishing the hindbrain territory and Hb-Sc transition. In the posterior neural plate, Cyp26 gene expression is normally limited to the hindbrain, where it promotes anterior and suppresses posterior neural gene expression (Kudoh et al., 2002). In loss of function experiments, Cyp26 elimination results in the rostral expansion of the spinal cord territory at the expense of the hindbrain (Fig. 1T, 2J; Hernandez et al., 2007). In contrast, knock down of Cdx4 function expands both the caudal expression of cyp26a1 (Fig. 4B) and caudal extent of the hindbrain territory (Fig. 1G, 2C) at the expense of the spinal cord, suggesting that suppression of cyp26a1 expression by Cdx4 is normally necessary to establish the spinal cord territory. Simultaneous inhibition of both Cyp26 and Cdx4 function (Figs. 1X, 2L) results in the rescue of the lost spinal cord territory characteristic of Cdx4-deficient embryos, confirming that Cyp26 is epistatic to Cdx4 in regulating the size of the spinal cord region (Fig. 6E).

In contrast to cyp26, cdx4 expression is confined to the spinal cord territory where it promotes spinal cord formation and suppresses hindbrain fates (Fig. 5A; Shimizu et al., 2006; Skromne et al., 2007; Sturgeon et al., 2011). Here we show that, in the absence of RA signaling, cdx4 expression is expanded rostrally, mirroring the loss of hindbrain territory and the expansion of spinal cord territory in these embryos (Figs. 5B, S1B) suggesting that an inhibitory relationship between RA signaling and Cdx4 is involved in establishing the Hb-Sc boundary. Further confirmation of this interaction is shown through analyses of Cdx4/RA double deficient embryos (Figs. 1N, 2G, 3H,L), which display an expanded hindbrain phenotype similar to Cdx4-deficient embryos, indicating that Cdx4 is epistatic to RA signaling in determining the size of the hindbrain territory. As we have shown that cyp26a1 is under the repression of Cdx4, these data show that RA signaling may additionally be restricting cdx4 expression to ensure proper activation of cyp26a1 transcription in the hindbrain territory.

Our findings suggest that, in the posterior neural tube, RA signaling limits cdx4 expression to the presumptive spinal cord. How RA signaling selectively inhibits cdx4 expression in the hindbrain and not in the spinal cord remains to be determined. RA signaling has been shown to spatially restrict the expression of downstream target genes through acting as both an activator and a repressor. For example, in the case of hoxb1 transcription RA signaling first establishes a broad expression domain, which it then refines by suppression of hoxb1 expression in rhombomeres adjacent to r4 (Studer et al., 1994). Two RAREs in the regulatory region of hoxb1 have been shown to mediate these different responses (Studer et al., 1994). RAREs have also been identified in the regulatory region of Cdx and can drive cdx expression in cell culture assays and mouse embryos (Gaunt et al., 2003; Houle et al., 2000). However, inhibitory factors, activated by RA signaling in the anterior region of the embryo, have also been shown to competitively bind to the RARE suggesting that regulation of cdx expression by RA is likely context dependent (Beland and Lohnes, 2005). Additional analysis of the Cdx4 regulatory region will help clarify the regulatory role of RA signaling on cdx4 expression in the neural tube.

A recent paper (Lee and Skromne, 2014) showed that loss of RA signaling alone or in combination with loss of FGF selectively alters the anterior expression limit of cdx4. Lee and Skromne (2014) additionally showed that expression of cdx4 in the spinal cord can be severely perturbed when RA-deficient embryos are also made Wnt deficient. Their findings suggest that Wnt and Fgf signaling may also be modifying the regulatory effect of RA signaling on cdx4 expression. Multiple signaling pathways, including Wnts, FGFs and RA, have been shown to regulate cdx4 expression (Gaunt et al., 2003; Houle et al., 2000; Isaacs et al., 1998; Shimizu et al., 2005). Our results expand upon these previous findings by characterizing the epistatic relationships between RA, Cyp26 and Cdx4 at the Hb-Sc junction, thereby demonstrating that cross regulation of cdx4 expression by molecular components of the RA signaling pathway is part of a regulatory feed-back loop involved in directing the precise compartmentalization and development of the hindbrain and spinal cord. In addition, we examined the impact of these interactions on the global expression of rhombomere segmentation markers (Fig. 3I–L), patterning genes (Fig. 3M–P) and motor neuron subtype specific genes (Fig. 3U–X); our comprehensive approach has led us to suggest that the initial specification of the hindbrain and spinal cord territories versus the patterning of these two areas may be regulated by distinct processes, albeit likely involving the same signaling molecules utilized at different times of development or with different downstream effectors. In particular, we find that the hindbrain territory size does not necessarily specify the number of formed hindbrain segments. For example, in Cdx4/RA double deficient embryos, we show that the rescued Islet1-Gfp neurons only express r4–6 neuron specific genes and that the r7/8 segment fails to form even within a larger than normal sized hindbrain region (summarized in Fig. 6F).

Having established that RA and Cdx4 have functions that are separate from their regulatory roles on patterning gene expression, we propose a model where the territories of the hindbrain and spinal cord are established by a feedback regulatory loop involving RA, Cyp26a1 and Cdx4 (Fig. 6A). In this model (Fig. 6A), RA signaling excludes cdx4 expression from the future hindbrain territory while Cdx4 limits the expression of the RA degradation enzyme, cyp26a1, from the spinal cord. The cross-regulatory interaction between these factors is necessary to ensure that cells receive the correct levels of RA and Cdx4 required to specify the hindbrain versus spinal cord territories since perturbation of either RA signaling (Fig. 6B,C) or Cdx4 function (Fig. 6D) causes the Hb-Sc transition to be mis-aligned. Through these interactions, Cdx4 and RA function are spatially regulated to distinguish the posterior neural tissue into two separate structures of the hindbrain and spinal cord.

Supplementary Material

supplement. Fig. S1. Spinal cord motor neurons shift rostrally while hox gene expression remains unchanged in embryos treated with DEAB between 50% epiboly and tailbud stage.

Olig2-Rfp:Actin-Gfp double transgenic embryos were treated with DEAB between 50% Epiboly to tailbud stage. Olig2-Rfp marks spinal cord motor neurons (pink) and Actin-Gfp marks somites (green). In wildtypes the first motor neuron (A arrowhead) projects out at the level of somite 3. In DEAB treated embryos the first spinal cord motorneuron (B arrowhead) arises at the level of somite 1. At 48 hpf the anterior expression limit of hoxb8a is at the level of somite 3 in wild-types (C) and this expression limit remains unchanged in embryos treated with DEAB between 50% epiboly to tailbud stage (D). (C,D) Hoxb8a expression, (C′D′) MHC, (C″,D″) merge. (AD) Embryos at 48 hpf with the anterior located to the left. (A) n=6/6, (B) n=10/10, (C,D) n=15/15.

Highlights.

• The RA pathway and Cdx4 function interact to position the boundary between the hindbrain and spinal cord territories.

• RA and Cdx4 both regulate the expression of zebrafish cyp26a1.

• In turn, Cyp26 and Cdx4 act to modulate RA levels along the hindbrain and spinal cord regions.

• The hindbrain territory size does not necessarily specify the number of formed hindbrain segments and vice versa.

• Separate processes direct specification versus patterning of the hindbrain and spinal cord.