Mammalian Brain Morphogenesis and Midline Axon Guidance Require Heparan Sulfate

See allHide authors and affiliations

Science  07 Nov 2003:
Vol. 302, Issue 5647, pp. 1044-1046
DOI: 10.1126/science.1090497


Heparan sulfate (HS) is required for morphogen signaling during Drosophila pattern formation, but little is known about its physiological importance in mammalian development. To define the developmental role of HS in mammalian species, we conditionally disrupted the HS-polymerizing enzyme EXT1 in the embryonic mouse brain. The EXT1-null brain exhibited patterning defects that are composites of those caused by mutations of multiple HS-binding morphogens. Furthermore, the EXT1-null brain displayed severe guidance errors in major commissural tracts, revealing a pivotal role of HS in midline axon guidance. These findings demonstrate that HS is essential for mammalian brain development.

During development, a number of secreted factors play critical roles in morphogenesis, growth regulation, and differentiation. Many of these morphogens and mitogens, including fibroblast growth factors (FGFs), Wingless/Wnt, and Hedgehog/Sonic hedgehog, bind HS (1, 2). Genetic studies in Drosophila have demonstrated specific functions for HS in the signaling of these molecules (3). HS is expressed abundantly and in a developmentally regulated manner in the mammalian central nervous system (CNS), suggesting its functional role in brain development (4). However, the physiological importance and functional specificities of HS in mammalian brain development have not been elucidated.

The most critical step in the HS biosynthetic process is the polymerization of alternating GlcA and GlcNAc sugar residues, catalyzed by the EXT family molecules (5). EXT1 is indispensable for HS synthesis, because cells lacking a functional EXT1 allele do not synthesize HS (6, 7), and EXT1 knockout mice die early as a result of defective gastrulation (7). In the mouse CNS, EXT1 is expressed throughout the neural tube as early as E9.5 (embryonic day 9.5), and its expression persists into adulthood (8) (Fig. 1A). To determine the physiological role of HS in mammalian CNS development, we created a loxP-modified EXT1 allele (9) (fig. S1). Mice carrying the floxed allele were bred to nestin-Cre mice (10) to generate mice in which the EXT1 gene was disrupted selectively in the nervous system. The nestin-Cre line has been shown to delete loxP-modified genes in neural progenitor cells and their progeny as early as E9.5 (11).

Fig. 1.

CNS-targeted inactivation of EXT1. (A) Whole-mount in situ hybridization for EXT1 at E9.5 and E12.5. EXT1 expression in the forebrain (arrows), the midbrain-hindbrain region (open arrows), and the spinal cord (arrowheads) was selectively disrupted in Nes-EXT1–null mice. (B) Immunoblotting analysis for syndecan-3 glycanation. H'ase, heparitinase digestion. (C to E) Gross defects in E18.5 brain. Dorsal views of the entire brain (C) and thionin-stained sagittal sections of the entire brain [(D) and (E)]. Note the defects in the midbrain-hindbrain region (asterisk), the small cerebral cortex (Cx), and the absence of the olfactory bulbs (open arrowheads). WT, control littermate; KO, Nes-EXT1–null. Scale bar, 1 mm.

All the newborns of conditional mutant genotype (EXT1flox/flox;Cre+) died within the first day of life, whereas those of other genotypes were born healthy. Immediately before parturition (E18.5), EXT1flox/flox; Cre+ mice were present approximately in a Mendelian ratio, indicating that the death occurred perinatally. EXT1 mRNA was absent in the E18.5 mutant brain (fig. S1). Expression of EXT1 decreased in the CNS (except for the forebrain) as early as E9.5 and was almost completely eliminated at E12.5 (Fig. 1A). In the brain of EXT1flox/flox;Cre+ mice, the HS proteoglycan syndecan-3 was not glycanated by HS chains (Fig. 1B), indicating that HS synthesis was disrupted. Immunocytochemistry with a monoclonal antibody to HS (10E4) revealed no HS in dissociated cortical neurons from E16.5 mutant mice (12). These analyses confirmed that EXT1flox/flox;Cre+ mice are HS null; they are hereafter referred to as Nes-EXT1–null mice.

Despite the grossly normal appearance of Nes-EXT1–null embryos, specific developmental defects in their CNS were observed. These include malformations in the caudal midbrain-cerebellum region, an abnormally small cerebral cortex, the absence of major commissural tracts, and the absence of the olfactory bulbs (Fig. 1, C to E). These defects were observed with full penetrance in Nes-EXT1–null embryos, whereas none of the mice of control genotypes or constitutive EXT1+/– heterozygotes exhibited these defects. The most notable phenotype was a patterning defect in the midbrain-hindbrain region, characterized by the absence of a discernible inferior colliculus and cerebellum. This phenotype is similar to that caused by a hypomorphic Fgf8 allele (13) and a natural Wnt1 allele called swaying (14). Differentiation of the caudal midbrain and the cerebellum initiated but failed to form the distinct inferior colliculus and cerebellum (Fig. 2, A to D). Immunostaining for calbindin, a marker for Purkinje cells, revealed a disorganized collection of calbindin-positive cells in this region (Fig. 2F). This suggests that the initial specification of the cerebellum occurred in Nes-EXT1–null embryos but that the subsequent developmental steps did not progress to form a fully organized inferior colliculus and the cerebellum.

Fig. 2.

Midbrain and cerebellar defects in Nes-EXT1–null embryos. (A to D) Thionin-stained sagittal sections at E12.5 and E14.5 show incomplete differentiation of the inferior colliculus (IC) and the cerebellum (Cl) in Nes-EXT1–null embryos [(B) and (D)]. (E and F) Immunostaining reveals the presence of calbindin-immunoreactive cells (arrows) in the defective midbrain-hindbrain region of E18.5 Nes-EXT1–null mice (F). (G to T) Whole-mount in situ hybridization for Fgf8 [(G) to (J)], Wnt1 [(K) to (P)], En1 [(Q) and (R)], and En2 [(S) and (T)]. In Nes-EXT1–null mice, abnormal expansion of expression domain is evident for Wnt1 at E9.5 (L) and, to a lesser degree, for Fgf8 (J) and Wnt1 [(N) and (P)] at E10.5. (U and V) Whole-mount immunohistochemistry for FGF8 at E10.5. Note diffuse FGF8 immunoreactivity in the midbrain-hindbrain region of the Nes-EXT1–null embryo (asterisk). WT, control littermate; KO, Nes-EXT1–null.

If an aberrant distribution of morphogens resulting from the lack of HS is the underlying cause of these defects, expression of genes downstream of HS-binding morphogens would likely be altered. FGF8 is a key mediator of the isthmic organizer to induce the expression of Wnt1, Engrailed 1 (En1), Engrailed 2 (En2), and Fgf8 itself during midbrain-hindbrain patterning (15). In E9.5 Nes-EXT1–null embryos, the Wnt1 expression domain expanded (Fig. 2L, arrowhead). At E10.5, the caudal boundary of the Wnt1 expression domain was disturbed, and small islands of Wnt1-positive cells were ectopically located caudal to the boundary (Fig. 2, N and P, open arrowheads). Expression of En1 and En2 was weaker and more diffuse than in the control littermates (Fig. 2, R and T, open arrows). Although Fgf8 expression showed no major change at E9.5, it was expanded at E10.5 (Fig. 2J). Such abnormal expression patterns of downstream genes suggest that the distribution of FGF8 in the midbrain-hindbrain boundary is abnormal in Nes-EXT1–null embryos. Indeed, the concentrated FGF8 band in the midbrain-hindbrain boundary typically seen in wild-type embryos was not formed in Nes-EXT1–null embryos (Fig. 2V). Together, these results indicate that HS is essential for the proper FGF8 distribution during midbrain-hindbrain patterning.

Another prominent phenotype in Nes-EXT1–null mice is the abnormally small cerebral cortex. The overall size and thickness of the cortex was reduced (Fig. 1E and Fig. 3B). In utero bromodeoxyuridine (BrdU) labeling demonstrated that cell proliferation in the Nes-EXT1–null cortex was reduced by ∼30% compared with that of control littermates (Fig. 3, C to E). Primary cultures prepared from E16.5 Nes-EXT1–null forebrains showed reduced BrdU incorporation in response to FGF2 and FGF8 (Fig. 3, F to J). In contrast, no difference in the frequency of apoptotic cells was observed (fig. S2). These results indicate that reduced proliferation of HS-deficient neurons is likely to be one of the major causes underlying the poor development of the cerebral cortex. Despite the small size, there were no abnormalities in overall cortical lamination patterns (Fig. 3B) or localization of specific neuronal populations (calretinin-positive and Tst-1/SCIP–positive neurons) (fig. S2) in the Nes-EXT1–null cortex, suggesting that neuronal migration is largely unaffected by the elimination of HS.

Fig. 3.

Cortical defects in Nes-EXT1–null mice. (A and B) Thionin-stained coronal sections of E18.5 parietal cortex. CP, cortical plate; IZ, intermediate zone; VZ, ventricular zone. (C to E) The cortical neuroepithelium in E12.5 Nes-EXT1–null embryos shows a marked reduction in BrdU incorporation. Quantification of BrdU-positive cells in the medial and parietal cortex (E). *P = 0.0047, **P = 0.0005. (F to J) Proliferation of cultured cortical progenitors in response to FGF2 and FGF8. [(F) to (I)] BrdU incorporation in FGF-treated cultures. (J) Quantification of BrdU-positive cells. BSA is control treatment. †, P < 0.0001. WT, control littermate; KO, Nes-EXT1–null.

Nes-EXT1–null mice displayed severe defects in commissural fiber tract development. Three major commissures in the forebrain, namely the corpus callosum, hippocampal commissure, and anterior commissure, were absent in Nes-EXT1–null mice (Fig. 4, A to D; fig. S3). Anterograde tracing revealed that in Nes-EXT1–null mice, DiI-labeled axons from the cortex failed to approach the midline and instead extended ventrally (arrows in Fig. 4F), whereas axons in control mice turned to the midline and crossed toward the contralateral side (Fig. 4E). In the anterior commissure, axons in Nes-EXT1–null mice extended straight toward the ventral surface of the forebrain (arrowhead in Fig. 4H), whereas control axons make a turn toward the midline to form the anterior commissure (Fig. 4G).

Fig. 4.

Abnormal axon guidance in Nes-EXT1–null mice. (A to D) Thionin-stained coronal sections of E18.5 forebrains reveal the absence of the corpus callosum (CC), hippocampal commissure (HC), and anterior commissure (AC) in Nes-EXT1–null brain [(B) and (D)]. Arrows in (B) indicate abnormal Probst bundle. (E and F) DiI tracing of the corpus callosum. In Nes-EXT1–null mice (F), corticocortical axons extend ventrally (arrows) without approaching the midline (broken line). (G and H) DiI tracing of the anterior commissure. In Nes-EXT1–null mice (H), axons emerging from the external capsule (EC) extend straight toward the ventral surface of the brain (arrowheads). (I and J) Guidance defects of retinal axons at the optic chiasm (indicated by asterisk). In Nes-EXT1–null mice, most retinal axons project abnormally to the contralateral optic nerve [arrowheads in (J)]. (K to N) Dosage-sensitive genetic interaction between Slit and EXT1. Note that reduction of one allele of EXT1 in Slit2–/– background causes misprojection of retinal axons into the contralateral optic nerve (N). This genetic interaction was observed with 100% penetrance (n = 8). In those mice, the percentage of axons misguided to the contralateral optic nerve ranged from 59 to 67%, whereas in Nes-EXT1–null mice (L), it ranged from 60 to 67%. Retinal axons project normally in mice with heterozygotic deletion of EXT1 (K) and in Slit2–/– mice (M). Cx, cortex; GP, globus pallidus; IC, internal capsule; LV, lateral ventricle; St, striatum; WT, control littermate; KO, Nes-EXT1–null. Scale bar, 500 μm.

Guidance defects of retinal axons at the optic chiasm were also observed. In Nes-EXT1–null mice, retinal axons projected ectopically into the contralateral optic nerve (Fig. 4, J and L), similar to Slit1/Slit2 double-null mice (16). Slit proteins are HS-binding repulsive guidance cues (17), and in the optic chiasm, Slit1 and Slit2 act cooperatively to guide retinal axons to contralateral sides (16). To examine whether there is genetic interaction between Slit and EXT1, compound mutants were generated. Although few guidance defects were found in Slit2–/– mice (Fig. 4M), a reduction of one EXT1 allele in Slit2–/– background caused profound retinal axon misguidance (Fig. 4N), as was observed in Nes-EXT1–null mice (Fig. 4L) and in Slit1/Slit2 double-null mice (16). Thus, there is a strong dosage-sensitive genetic interaction between Slit and EXT1, indicating that HS plays a physiologically essential role in Slit-mediated retinal axon guidance at the optic chiasm.

Our results demonstrate that the effects of the elimination of HS are surprisingly selective for specific developmental signaling events, rather than generally affecting basic cellular activities such as adhesion and migration. Although multiple developmental signaling pathways appear to be affected in Nes-EXT1–null mice, the overall phenotype suggests that Fgf8 signaling is probably the most critically disrupted pathway. The lack of phenotypes indicative of aberrant Shh signaling, such as holoprosencephaly and cyclopia (18), could be due to the timing of nestin-Cre–mediated gene inactivation, which may not be early enough to affect Shh-dependent morphogenetic events. Another possibility is that HS is required for movement of SHH from the notochord to the neural tube but not for recipient neuroepithelial cells to respond to SHH. Finally, this study revealed a critical role for HS in axon guidance. Although Slit requires HS at the optic chiasm, the molecular mechanism for the forebrain commissure defects remains to be elucidated. The major guidance molecule in the forebrain, netrin-1, binds heparin (19), and the forebrain commissure defects in netrin-1–deficient mice resemble those of Nes-EXT1–null mice (20), suggesting that the defects in Nes-EXT1–null mice are, at least partly, due to aberrant netrin function.

Supporting Online Material

Materials and Methods

Figs. S1 to S3

References and Notes

Stay Connected to Science

Navigate This Article