Proteoglycan interactions with Sonic Hedgehog specify mitogenic responses

SUMMARY

Sonic Hedgehog (Shh) has dual roles in vertebrate development, as it promotes progenitor cell proliferation and induces tissue patterning. Here we show mitogenic and patterning functions of Shh can be uncoupled from one another. Using a genetic approach to selectively inhibit Shh-proteoglycan interactions in a mouse model, we show binding of Shh to proteoglycans is required for proliferation of neural stem/precursor cells but not for tissue patterning. Shh-proteoglycan interactions regulate both spatial and temporal features of Shh signaling. Proteoglycans localize Shh to specialized mitogenic niches and also act at the single cell level to regulate the duration of Shh signaling, thereby promoting a gene expression program important for cell division. As activation of the Shh pathway is a feature of diverse human cancers, selective stimulation of proliferation by Shh-proteoglycan interactions may also figure prominently in neoplastic growth.

Development of complex tissues requires concomitant growth and cell fate specification. One mechanism for achieving spatial and temporal coordination of size and form is to utilize the same signaling molecules for both processes. The morphogen Hedgehog (Hh), and its mammalian counterparts Sonic, Desert and Indian Hedgehog (Shh, Dhh and Ihh), are critical in the growth and patterning of developing embryos¹. One model for morphogen activity postulates Hedgehog proteins disperse from a localized source and form a gradient that patterns fields of responsive cells ². Controlled ligand distribution may also localize stem cell proliferation to specialized niches^3,4. By influencing patterning and proliferation, Hedgehog proteins could coordinate form and size to allow scaling of developing organisms.

Here we asked whether proteoglycans differentially regulate Shh-dependent proliferation and patterning. Genetic evidence in Drosophila indicates proteoglycans are required for Hh dispersal and gradient formation^5,6. Biochemical evidence indicates proteoglycans bind to Hh and the receptor component Ihog to affect Hh activity^7,8. In Drosophila, proteoglycans critical for Hh functions are heparan sulfate proteoglycans (HSPGs), and the core protein components are glypicans (GPI-linked proteins) ^6,9–12. Dally-like glypicans are required for Hedgehog and Wingless dispersal and responses, modulating signaling of both ligands¹³. However, how HSPGs affect Hedgehog signaling and biological responses are not yet understood.

Proteoglycans (PGs) have also been implicated in Hedgehog-pathway signaling in mammalian systems^11,12,14. Loss of glypican GPC3 causes an overall growth increase, reflecting changes in Shh and/or IGF signaling¹¹, while mutations in HSPG synthesizing enzymes cause dramatic defects that may reflect changes in FGF, Wnt, and/or Shh responses^15,16. To investigate the physiologic role of Shh-proteoglycan interactions, we took the alternative approach of mutating Shh itself. Shh contains an N-terminal Cardin-Weintraub motif that mediates Shh-proteoglycan interactions (KRRHPKK). As mutations in this motif (R34A/K38A, designated Shh^Ala) reduce high-affinity Shh-proteoglycan interactions without altering Shh’s affinity for its receptor Patched (Ptc)¹⁴, we can use this mutation to investigate Shh-HSPG interactions, without confounding effects due to other growth factors that bind proteoglycans.

To identify on an organismal level responses that require Shh-proteoglycan interactions we generated mice in which wild type Shh is replaced with Shh^Ala. While Shh is needed for growth and patterning of diverse tissues, we find proteoglycan interactions selectively affect Shh-induced proliferation rather than Shh-induced patterning. Proteoglycans localize Shh ligand to mitogenic niches in developing brain to promote proliferation of neural stem/precursor cells. Proteoglycans also function at the level of individual responding cells; cell-associated proteoglycans bind Shh and alter ligand perdurance. In this way, Shh-proteoglycan interactions stimulate expression of bmi-1, D-type Cyclins and other genes implicated in proliferation and neoplastic growth^{17, 18}. These studies indicate that HSPGs selectively promote mitogenic responses to Shh.

RESULTS

HSPG binding motif mutations affect Shh-PG interactions

We identified a conserved Cardin-Weintraub motif in Shh that is responsible for high affinity binding of Shh to HSPGs¹⁴. To determine whether mutations in this motif can be used to identify functions of Shh-proteoglycan interactions, we quantitatively evaluated Shh binding to heparin. We assessed binding of equal amounts of wild type Shh or Shh^Ala to heparin sulfate-coated plates using alkaline phosphatase tagged Shh isoforms. We measured binding in the absence, or presence of increasing amounts of soluble heparin sulfate. Overall binding of mutant Shh was significantly less than wild type. Furthermore, higher concentrations of soluble heparin sulfate were required to abolish binding of wild type Shh to immobilized heparin than concentrations needed to abolish binding of Shh^Ala (Fig. 1a). Binding of Shh^Ala to endogenous proteoglycans was also strikingly impaired. Wild type Shh, but not Shh^Ala, binds endogenous proteoglycans in CNS tissue sections (14, Fig. 1b). Pre-treatment of sections with heparinases, which remove heparan sulfates, or PI-PLC, which removes GPI-linked molecules, prevented binding of wild type Shh to tissue proteoglycans (Fig. 1b, Supplementary Fig. 1). As glypicans are HSPGs with a GPI-linkage, these results suggest that wild type Shh binds glypican proteoglycans in the developing cerebellum and that the Shh^Ala mutation interferes with binding to these endogenous glypicans in developing cerebellum. We next introduced the Shh^Ala mutation into a full length Shh expression vector and produced protein in HEK cells. The Shh^Ala precursor undergoes proteolytic cleavage to generate a mature isoform of the correct size (Fig. 1c) that, like wild type Shh, is palmitoylated (Fig. 1d)¹⁹. We previously demonstrated that Shh^Ala binds Ptc with similar affinity as wild type ligand ¹⁴. Together, these data show Shh^Ala can be used as a specific reagent to determine functions of Shh-proteoglycan interactions.

Figure 1. Shh^Ala specifically alters proteoglycan binding.

(a) Shh^Ala shows reduced binding to heparin-coated plates. Shh-AP (circles), Shh^Ala-AP (squares), or AP (triangles) were incubated with heparin-coated plates in the presence of increasing concentrations of soluble heparin.

(b) Shh binds GPI-linked proteoglycans in cerebellar sections; Shh^Ala does not. P6 sections treated with vehicle control (−), Heparinases or PI-PLC, incubated with Shh-AP, Shh^Ala-AP or vehicle controls, and processed for binding of AP-tagged ligand. Scale bar, 100 μm.

(c) Shh^Ala is processed to mature isoform. Lysates of HEK293 transfected with Shh or Shh^Ala were analyzed by immunoblot with anti-Shh. Immature 45kD (arrowhead) and mature 20kD isoform (arrow) are seen.

(d) Shh^Ala is palmitoylated. HEK293 cells expressing Shh, Shh^Ala or Shh^C24S were labeled with ³H-palmitate, analyzed for palmitoylation, and probed with anti-Shh. Shh and Shh^Ala are palmitoylated. Shh^C24S is not.

Using homologous recombination, we generated knock-in mice in which endogenous Shh was replaced with Shh^Ala (Supplementary Fig. 2). With this targeting strategy, a loxP site remained within the intron between exons 1 and 2. Because intronic elements can affect Shh expression, we also generated a line of control animals wherein the only alteration is this loxP site (Shh^Ctl); Shh^Ctl/Ctl animals are indistinguishable from wild type. Furthermore, the Shh^Ala mutation itself does not alter the expression of Shh protein in vivo: Shh levels are indistinguishable in Shh^+/+, Shh^Ala/+ and Shh^Ala/Ala tissues (Supplementary Fig. 2e, f). This novel genetic approach enables us to determine the role of Shh binding to proteoglycans without interference from proteoglycan-dependent modulation of other growth factor pathways.

Shh-PG interactions affect growth but not patterning

Shh^−/− animals are embryonic lethal and display holoprosencephaly and limb patterning -defects²⁰. Hypomorphic alleles of Shh usually alter size and pattern¹⁹. In contrast, homozygous Shh^Ala/Ala mice only exhibited growth defects and showed no patterning defects. While Shh is critical in generating and patterning diverse organs, mutant animals were viable and fertile, with all organs present and correctly localized. In contrast, Shh^Ala/Ala mice displayed multiple growth abnormalities. Overall size was reduced in mutant animals, with particular differences in size of the brain, spinal cord and eyes. Body weight of Shh^Ala/Ala mice was 11% less than wild type or Shh^Ctl/Ctl animals, and brain weight was reduced by 13% (p=0.0018). There was, however, disproportionate hypoplasia of cerebellum and olfactory bulb (Fig. 2a). In contrast to these clear differences in tissue growth, when we examined areas where Shh is critical for pattern formation^1,21,22, we found Shh^Ala/Ala mice displayed normal digit number and shape, and the spinal cord laminae were correctly formed (Fig. 2a,b, Supplementary Fig. 3). Furthermore, while holoprosencephaly and cyclopia are cardinal features of Shh^−/− mice²⁰, Shh^Ala/Ala animals displayed normal separation of cerebral ventricles and two well-spaced (but small) eyes (Fig. 2a). Together these findings indicate that Shh-proteoglycan interactions mediated by the Cardin-Weintraub motif are specifically required for growth regulation but not for patterning activities.

Figure 2. Shh^Ala/Ala and Shh^Ala/− mice exhibit defects in growth, with normal patterning.

(a) Skeletal morphology, body and brain patterning are normal in Shh^Ala/Ala mice, but size of adult Shh^Ala/Ala mice is 11% less than wild type, and sizes of olfactory bulbs and cerebella are 30 and 31% reduced (arrows) respectively. Sagittal view of cerebellum shows normal patterning with reduced size in Shh^Ala/Ala. Scale bar, 1 mm. Shh^Ala/Ala animals also display well-spaced eyes.

(b) Spinal cord patterning is normal in Shh^Ala/Ala mice. In situ hybridization for Shh (expressed in notochord and floor plate), Nkx2.2 (motor neuron precursors), Nkx6.1 (ventral spinal cord), Isl1 (dorsal root ganglia and motor neurons), and Dbx1 (V0 interneurons) in E10.5 littermates. Scale bar, 100 μm.

(c) Growth defects are seen in Shh^Ala/−, without defects in patterning. Olfactory bulb and cerebellum are reduced in size (red arrows). Scale bar, 1 mm. Eyes of Shh^Ala/− mice are well spaced.

In mice, one copy of Shh is sufficient for normal development as Shh^+/− mice exhibit normal patterning and growth²⁰. One copy of Shh^Ala is also sufficient for patterning, as Shh^Ala/− mice were viable with normal limbs and digits. However, even more dramatic changes in growth were evident in Shh^Ala/− than in Shh^Ala/Ala animals (Fig. 2c). The overall size of Shh^Ala/− mice, and in particular, cerebellar and olfactory bulb size were greatly reduced compared to Shh^+/− animals. However, even when one copy of Shh^Ala is present, no defects were seen in tissue patterning (Fig. 2c). The characterization of Shh^Ala/Ala and Shh^Ala/− mice underscores the selectivity of the phenotype, and indicates Shh-proteoglycan interactions particularly regulate tissue growth.

Shh-PG interactions promote stem/precursor proliferation

To determine why proteoglycan interactions are critical for tissue growth, we focused on the cerebellum, as cerebella of Shh^Ala/Ala mice are one third smaller than their wild type littermates. As Shh regulates cell division and cell death in the nervous system^23,24, we asked whether the decrease in cerebellar size in Shh^Ala/Ala and Shh^Ala/− mice reflects abnormal proliferation or apoptosis. In early postnatal life, granule cell precursors (GCPs) undergo extensive proliferation in the external granule cell layer (EGL) then exit the cell cycle and migrate inwards past the Purkinje cell layer to reside in the internal granule cell layer (IGL). Previous studies have shown Purkinje cell-synthesized Shh promotes GCP proliferation^23,25. In Shh^Ala/Ala mice, GCP proliferation in the EGL is approximately 30% less than in wild type, as detected by percentage of EGL cells positive for M-phase marker phospho-Histone H3 and by S-phase labeling using bromodeoxyuridine (BrdU) (Fig. 3a–c). This decreased proliferation, which was seen throughout the cerebellum (Fig. 3d), indicates Shh-proteoglycan interactions promote mitogenesis of cerebellar precursors during early postnatal life across the rostral-caudal axis of the developing hindbrain.

Figure 3. Reduced proliferation of Shh^Ala/Ala cerebellar granule precursors is seen in the EGL of developing mice.

(a) p-Histone H3 (p-H3)-positive or BrdU (BrdU)-positive (both in red) cells are fewer in the EGL of P3 Shh^Ala/Ala mice as compared to Shh^+/+ littermates. DAPI is in grey or blue. Scale bar, 100 μm (left), 50 μm (right).

(b) p-Histone H3 mitotic indices are reduced in Shh^Ala/Ala mice (white bars) at multiple postnatal ages (Shh^+/+, black bars) (*p<0.001). Reduced proliferation in Shh^Ala/Ala mice is also seen in one wild type and mutant littermate pair at P9 (1.37% versus 1.07% for Shh^+/+ and Shh^Ala/Ala, respectively). Error bars are +/− s.e.m.

(c) BrdU proliferation indices are less in Shh^Ala/Ala mice (white bars) as compared to wild type (black bars) at multiple postnatal ages (*p<0.001). While no difference is seen at intermediate time points (P6 and P9), this, in part, reflects the cerebellar size difference between Shh^Ala/Ala and Shh^+/+ mice. There is still a significant difference in the number of proliferating cells/EGL length at P6 (3.6% in Shh^+/+ versus 2.7% in Shh^Ala/Ala). Error bars are +/− s.e.m.

(d) p-Histone H3 mitotic indices in P3 wild type (black bars) and Shh^Ala/Ala mice (white bars) in the anterior, middle and posterior cerebellum are not statistically different from their respective totals in (b), demonstrating that lobes throughout the cerebellum are affected in the mutant. Error bars are +/− s.e.m.

Ext1 and Ext2 are required for HSPG chain elongation, and so are necessary for generating glycan chains that interact with Shh. While we previously identified developmental changes in expression of ext1 and ext2 in BALB/c mice during the early postnatal period¹⁴, the 129/C57BL6 strain used here exhibited consistent expression of ext1 and ext2 in the postnatal period when proliferation is reduced in Shh^Ala/Ala (Supplementary Fig. 4). Many other genes required for synthesizing and modifying glycans are also appropriately expressed at this time²⁶. Thus, Shh and proteoglycans are expressed appropriately to regulate GCP proliferation, which is reduced when Shh-proteoglycan interactions are impaired.

The subventricular zone (SVZ) adjacent to the lateral ventricles and the subgranular layer (SGL) of the hippocampus represent two additional locations where there is significant Shh-regulated postnatal proliferation of neural stem/precursor cells^3,4,27, while the embryonic spinal cord represents a site where Shh regulates prenatal expansion of neural stem/precursors²⁸. Neural precursor proliferation in the spinal cord was reduced in Shh^Ala/Ala embryos as compared to Shh^+/+ littermates (Fig. 4a,b). Proliferation of neural stem/precursors in the SVZ and SGL of adult mutants was also decreased compared to age- and sex-matched Shh^Ctl/Ctl (Fig. 4c,d). As stem/precursor cells in the SVZ generate olfactory bulb neurons, decreased SVZ proliferation may contribute to the olfactory bulb hypoplasia (Fig. 2, Supplementary Fig. 5). Together, these findings demonstrate that Shh-proteoglycan interactions promote neural stem/precursor cell proliferation in multiple locations in both the developing and adult CNS, and thus highlight the widespread importance of proteoglycans in mitogenesis.

Figure 4. Embryonic and adult neural stem/precursor proliferation is reduced in Shh^Ala/Ala mice.

(a) More proliferating neural stem/precursors are seen in the spinal cord of Shh^+/+ mice (left) than in their Shh^Ala/Ala (right) E10.5 littermates. p-Histone H3 is in red and DAPI in blue. Scale bar, 100 μm.

(b) A reduction in the total number of p-Histone H3 positive (p-H3+) cells in the spinal cord of Shh^Ala/Ala embryos is seen compared to the number in Shh^Ctl/Ctl littermates (*p<0.05). A significant reduction in the total p-H3+ cells per unit spinal cord area is also seen in Shh^Ala/Ala E10.5 embryos (data not shown). Error bars are +/− s.e.m.

(c) Fewer p-H3+ cells are present in the SVZ of adult Shh^Ala/Ala mice (*p<0.05). Error bars are +/− s.e.m.

(d) Fewer BrdU positive (BrdU+) cells are seen in the hippocampal SGL of adult mutant mice (*p≤0.05). Error bars are +/− s.e.m.

Surprisingly, apoptosis was decreased in the EGL of Shh^Ala/Ala mutants compared with wild type littermates (Supplementary Fig. 6). As the Shh pathway can stimulate or suppress apoptosis under different conditions^24,29, it is possible that pro-apoptotic effects of Shh require proteoglycan interactions, while anti-apoptotic effects do not. Alternatively, decreased apoptosis in Shh^Ala/Ala mice may be an indirect consequence of the mutation. In either case, the increased survival did not mitigate the proliferation decrease, and the net result was a dramatically smaller cerebellum (30% smaller in Shh^Ala/Ala and 50% smaller in Shh^Ala/−).

PGs delineate the Shh mitogenic niche in the EGL

There are two potential mechanisms to explain the effects of proteoglycans on Shh-induced proliferation. One is that proteoglycan interactions localize Shh to a mitogenic niche. Previous studies have suggested that proteoglycans regulate Hh dispersal^5,9,12,13 and localize Ihh in developing bone³⁰. Thus, without proteoglycan binding, Shh might not collect in the EGL, resulting in decreased proliferation. A second is that binding of Shh to proteoglycans on the surface of GCPs modulates Shh-dependent intracellular signal cascades^7,13. As these models are not mutually exclusive, we investigated both.

We previously developed an assay to evaluate the nature and location of mitogenic niches in neural tissues³¹. In this assay, GFP-labeled precursors are introduced onto organotypic slices and incorporate into the slices. We then incubate the slice culture with BrdU to label proliferating cells, and measure the proliferative index of GFP+ precursors in distinct microenvironments. Precursors exposed to the cerebellar EGL proliferate extensively due to mitogenic effects of Shh³¹. To determine whether Shh-proteoglycan interactions are important in delineating this niche, we introduced wild type GFP-labeled precursors or DiO-labeled Shh^Ala/Ala precursors onto slices cultured from wild type or Shh^Ala/Ala littermates. While wild type or Shh^Ala/Ala precursor proliferation is enhanced when cells are introduced onto a wild type EGL, the EGL of Shh^Ala/Ala mice does not provide a mitogenic niche for wild type precursors (Fig. 5a). To determine whether decreased proliferation is due to impaired proteoglycan interactions, we added heparan sulfates to cultures of wild type precursors on wild type slices. Excess glycans can compete with, and so diminish the actions of endogenous proteoglycans³⁰. Excess heparan sulfates phenocopied the Shh^Ala mutation, as there was decreased proliferation of cells introduced onto the EGL (Fig. 5b). Together these data indicate that interactions between Shh’s Cardin-Weintraub motif and endogenous proteoglycans promote neural precursor proliferation and establish the EGL mitogenic niche.

Figure 5. Shh^Ala cannot specify a mitogenic niche.

(a) Granule cell precursors (GCPs) from GFP+ mice were cultured on P6 Shh^+/+ or Shh^Ala/Ala cerebellar slices, or DiO-labeled Shh^Ala/Ala GCPs were cultured on wild type slices. Proliferation of introduced GCPs was analyzed by BrdU incorporation. The proliferative index is the percent (%) of GFP- or DiO-positive cells in each layer that are BrdU+ (*p<0.05 versus EGL of WT slice; **p<0.05 vs IGL of WT slice; no significant difference between EGL and IGL proliferation on Shh^Ala/Ala slices is seen). Error bars are +/− s.e.m.

(b) EGL mitogenic niche in Shh^Ala/Ala slices (light gray bar) is phenocopied by added exogenous heparan sulfates (HS) (dark gray bar) (*p<0.05). Error bars are +/− s.e.m.

(c) Shh immunostaining of wild type and Shh^Ala/Ala P6 cerebella shows a reduction in Shh staining in the EGLa of mutant mice (arrowhead) compared to wild type cerebella (arrows) (52%±11% of wild type (p=0.01)). EGLa: external granule cell layer, outer proliferative zone, EGLb: external granule cell layer, inner post-mitotic zone; ML: molecular layer; PCL: Purkinje cell layer; IGL: internal granule cell layer. Scale bar, 50 μm

In slice overlay cultures, proliferation of introduced precursor cells was decreased in the EGL but increased in the IGL of Shh^Ala/Ala slices (Fig. 5a). These data suggest that proteoglycan binding enables Shh to accumulate in the EGL rather than in other layers. To test this, we carried out immunohistochemistry for Shh. Consistent with previous studies^{31, 32} in wild type animals, Shh immunostaining was detected in Purkinje cells, and in the IGL and EGL (Fig. 5c, Supplementary Fig. 7). In Shh^Ala/Ala animals, Shh immunostaining was decreased in the outer EGL (EGLa) where proliferation occurs (Fig. 5c). These findings indicate proteoglycans localize Shh to a mitogenic niche within the outer EGL. Electron microscopy studies have identified a proteoglycan matrix in EGLa just under the pia and adjacent to proliferating GCPs³³. Thus one mechanism by which Shh-HSPG interactions stimulate precursor proliferation is by localizing ligand to the EGL mitogenic niche.

Proteoglycan interactions modulate responses to Shh

While the data above indicate Shh-proteoglycan interactions are important for localizing Shh to proliferative zones and so establishing a mitogenic niche, it is also possible proteoglycans on neural stem/precursors bind Shh, thereby specifying proliferative responses. To address this possibility, we asked whether proteoglycans on GCPs affect Shh-induced responses at the level of an individual responding cell. Consistent with previous studies^{14,23,25,34,35}, wild type Shh induces robust GCP proliferation. In dissociated cell cultures, the proliferative response evoked by wild type Shh was much greater than that evoked by equivalent concentrations of Shh^Ala (Fig. 6a). In contrast, no consistent effects on survival were seen when dissociated GCP cultures were stimulated with either wild type Shh or Shh^Ala (Fig. 6b). Thus, the decreased proliferation observed in Shh^Ala/Ala mice is a direct consequence of the Cardin-Weintraub mutation, while decreased apoptosis may be an indirect result of the mutation. These data demonstrate Shh interacts with proteoglycans on individual GCPs, thereby inducing cells to proliferate, and suggest that the decreased proliferation observed in Shh^Ala/Ala mice results both from altered localization of ligand and from changes in intracellular signaling cascades.

Figure 6. Shh-proteoglycan interactions promote proliferation in dissociated cell cultures of GCPs, but are not needed for survival.

(a) The proliferative response to Shh (gray) (as assessed by BrdU incorporation) is greater than that of GCPs to Shh^Ala (black) (*p≤0.05, **p<0.01). Error bars are +/− s.e.m.

(b) No consistent effects on survival (as assessed by activated caspase 3 immunostaining) are seen in GCPs stimulated with Shh (gray) or Shh^Ala (black). Error bars are +/− s.e.m.

Shh responses are mediated by the transcription factors Gli1, Gli2 and Gli3³⁶. Therefore, we analyzed expression of these factors in Shh^Ala/Ala mice. While Gli transcription factors can function either as transcriptional activators or repressors³⁷, Gli1 only functions as an activator, and only one isoform of Gli1 has been reported³⁸. Gli1 protein levels were equivalent in P3 Shh^Ala/Ala and Shh^+/+ animals (Fig. 7a). Gli2 can function as activator and repressor, and distinct isoforms have been identified that subserve these different functions^38,39. In cerebella of mutant animals, the ratio of Gli2 activator (Gli^Act) to the shorter Gli2 repressor (Gli2^Rep) was reduced by 50.9% (Fig. 7b). These data suggest overall Gli2-dependent transcription is likely to be altered in mutants. Previous studies have indicated Gli3 functions primarily as a repressor⁴⁰. We did not detect any differences in Gli3^Rep levels, and we did not detect Gli3^Act in cerebellar tissue of either wild type or mutant animals (Fig. 7c). Thus Gli2 isoforms represent the Shh-dependent transcription factor(s) that are clearly altered in Shh^Ala/Ala cerebellum. As Gli2 activity is required to mediate expansion of GCPs⁴¹, changes in Gli2 may potentially explain the observed Shh^Ala/Ala phenotype.

Figure 7. Shh-proteoglycan interactions modulate transcriptional activity through the regulation of Gli2 isoforms and signaling kinetics.

(a) Gli1 protein levels in P3 Shh^Ala/Ala and Shh^+/+ cerebella are equivalent. At left, anti-Gli1 western blot of Shh^Ala/Ala and Shh^+/+ cerebellar lysates. At right, quantification of blotting results, Shh^Ala/Ala (white bar), Shh^+/+ (black bar). Error bars are +/− s.e.m.

(b) The ratio of Gli2 activator (Gli2^Act) to repressor (Gli2^Rep) is reduced in P3 Shh^Ala/Ala cerebella compared to Shh^+/+. At left, anti-Gli2 western blot. At right, quantification of results, Shh^Ala/Ala (white bar), Shh^+/+ (black bar) (*p<0.01). Error bars are +/− s.e.m.

(c) Gli3 repressor protein levels are unchanged in P3 Shh^Ala/Ala cerebella, as compared to Shh^+/+. At left, anti-Gli3 western blot. At right, quantification of results, Shh^Ala/Ala (white bar), Shh^+/+ (black bar). Error bars are +/− s.e.m.

(d) The ratio of Gli2 activator (Gli2^Act) to repressor (Gli2^Rep) is greater in cells stimulated with Shh compared to Shh^Ala. At top, anti-Gli2 western blot. At bottom, quantification of results, Shh (left bar), Shh^Ala (right bar) (*p<0.01). Error bars are +/− s.e.m.

(e) Kinetics of signaling differ for GCPs stimulated with Shh^Ala as compared to GCPs stimulated with Shh. At top, anti-Gli2^Act western blot of Shh- or Shh^Ala-stimulated GCPs (or blot against tubulin as a loading control). Numbers above blots are hours in culture. At bottom, anti-Gli2^Rep western blot of Shh- or Shh^Ala-stimulated GCPs (or blot against tubulin as a loading control).

Since the Gli2 activator/repressor ratio is altered in vivo in Shh^Ala/Ala cerebella and Gli2 has been shown to play a critical role in GCP proliferation⁴¹, we examined in greater detail Gli2 protein levels in GCPs acutely stimulated with wild type or mutant Shh. We found that GCPs stimulated with either Shh or Shh^Ala exhibited an increase in Gli2^Act protein (9.7% increase and 34.7% increase for Shh- and Shh^Ala-stimulated GCPs respectively). However, consistent with in vivo results, the Gli2^Act/Gli2^Rep ratio in GCPs acutely stimulated with Shh^Ala was lower than that seen in GCPs stimulated with Shh (Fig. 7d). Furthermore, the temporal profiles of Gli2^Act and Gli2^Rep differed when GCPs were stimulated with Shh^Ala rather than wild type Shh (Fig. 7e, Supplementary Fig. 8) (i.e average Gli2^Act levels declined by 3.7% and 31.4% between 24 and 32 hours post-stimulation by Shh and Shh^Ala respectively). These data indicate Shh interacts with proteoglycans on GCPs to alter the nature and timing of Gli2-dependent transcription.

PGs alter Shh-dependent gene expression and Shh kinetics

Changes in the Gli2 activator/repressor ratio are likely to alter expression of Shh-responsive genes. A number of Shh-responsive genes have been identified, including genes involved in proliferation and tissue patterning^{17,34,35, 42–44}. We analyzed Shh-responsive gene expression when dissociated wild type GCPs were stimulated acutely with equal concentrations of wild type Shh or Shh^Ala. Shh-dependent genes can be separated into three clusters that are differentially modulated by proteoglycan interactions (Fig. 8a). One cluster, including gli1 and ptc1, are similarly induced by Shh^Ala and wild type Shh. A second cluster of target genes (gli2 and N-myc) is induced to a greater extent by Shh^Ala than by wild type Shh. Surprisingly, expression of N-myc and gli2, genes implicated in proliferation, are in the gene set that is better induced by Shh^Ala. We note that Shh regulates both Gli2 and N-myc via transcriptional and post-transcriptional mechanisms ^{39, 45}, and so Gli2 and N-myc activity need not correspond to RNA levels. Induction of a third cluster, including gli3, Cyclin D1, Cyclin D2, and bmi-1, was diminished when cells were stimulated with Shh^Ala rather than wild type ligand. Cyclins D1 and D2 are cell cycle regulators at the G1/S transition that are induced in response to Shh stimulation³⁵ and have been particularly implicated in Shh-induced stem/precursor cell proliferation³⁵. The polycomb protein, Bmi-1, is a Shh target that is critical for self-renewal of stem cells and for cancer cell proliferation^17,18,44. Together, these data indicate the gene expression program induced by Shh is modulated by proteoglycans of responding cells. Moreover, Shh target genes that are altered when cells are stimulated by Shh^Ala include many implicated in proliferation and oncogenesis.

Figure 8. Proteoglycan interactions modulate Shh perdurance and differentially affect Shh-dependent gene expression.

(a) The gene pattern induced by Shh^Ala stimulation of wild type P6 GCPs is different than that of GCPs stimulated by equivalent amounts of Shh (*p<0.05). Error bars are +/− s.e.m.

(b) GCP proteoglycans modulate Shh ligand perdurance. At left, (top) anti-Shh western blot of Shh- or Shh^Ala-stimulated C3H10T1/2s (or blot against actin as loading control). At right, (top) anti-Shh western blot of Shh- or Shh^Ala-stimulated GCPs (or blot against tubulin as loading control). Numbers above blots are hours in culture. At bottom, quantification of results from one representative experiment. Shh (solid), Shh^Ala (dashed).

(c) Expression of Shh target genes in P1-2 Shh^Ala/Ala cerebella (P1-3 Shh^Ala/Ala cerebella for gli2) relative to Shh^+/+ littermates demonstrates that Shh-proteoglycan interactions differentially affect gene subsets (*p<0.01, **p<0.05). Error bars are +/− s.e.m.

Previous studies have demonstrated distinct responses to Shh can be elicited depending on ligand concentration²². To determine whether a shift in the dose response curve explains why Shh^Ala induces an altered program of gene expression, we performed dose response experiments using Shh or Shh^Ala protein. C3H10T1/2 differentiation as assessed by alkaline phosphatase induction (Supplementary Fig. 9a) is a standard assay for Shh activity; we found no shift in the Shh^Ala dose response curve compared to Shh. Furthermore, maximal efficacy of Shh^Ala was not reduced compared to Shh. We next measured gli1 and CyclinD2 mRNA levels in GCPs stimulated across a wide range of Shh and Shh^Ala doses (Supplementary Fig. 9b). Whereas gli1 expression did not differ significantly in response to equivalent doses of Shh and Shh^Ala throughout the dose range, CyclinD2 induction by Shh^Ala was diminished regardless of dose. These data indicate the altered biological activity of Shh^Ala does not reflect a shift in the dose response curve. As recent studies have highlighted the importance of Shh signaling kinetics in determining Shh responses⁴⁶, we asked whether proteoglycan interactions alter the ability of Shh to signal over prolonged periods of time. Equivalent amounts of Shh or Shh^Ala were added to either C3H10T1/2 or GCP cultures. After twenty-four hours of stimulation, we removed ligand from the media and analyzed Shh perdurance in the stimulated cells over the ensuing two days. In both systems, Shh^Ala levels declined more precipitously than wild type (Fig. 8b). Taken together, these data indicate that proteoglycans alter the kinetics of signaling, promoting a gene expression signature important for Shh-dependent precursor proliferation.

To determine whether changes in Shh-dependent gene expression explain the in vivo phenotype of Shh^Ala/Ala mice, we used quantitative RT-PCR to compare Shh-responsive gene expression in wild type and mutant cerebella. gli1 and ptc1 levels were equivalent in wild type and Shh^Ala/Ala cerebella. However, expression of a second set of genes including Cyclin D1, Cyclin D2, and bmi-1 was significantly reduced in Shh^Ala/Ala animals compared to wild type littermates (Fig. 8c). Decreased expression of CyclinD1, D2 and bmi-1 do not reflect a generalized decrease in cell cycle-associated proteins, as CyclinE expression was unchanged. A third set of Shh-regulated genes (gli2 and N-myc) was expressed at higher levels in Shh^Ala/Ala than in wild type mice. The observed changes in vivo do not reflect alterations in the distribution of cells expressing Shh-responsive genes as indicated by in situ hybridization studies for gli1, gli2 and gli3 (Supplementary Fig. 10). A critical feature of these data is that proteoglycan interactions do not uniformly increase or decrease Shh-dependent gene expression in vivo. Furthermore, clusters that are unchanged, increased, or decreased in Shh^Ala/Ala animals exhibit a striking similarity to gene expression patterns in GCPs acutely stimulated with wild type versus mutant Shh.

Taken together, these studies indicate Shh-proteoglycan interactions selectively promote neural stem/precursor proliferation by two mechanisms. First proteoglycans localize the ligand within developing tissue and so establish the mitogenic niche, and second they alter ligand perdurance, preferentially activating intracellular cascades that culminate in mitogenesis and precursor renewal.

DISCUSSION

Using a genetic approach, we find that Shh-proteoglycan interactions are required for proliferative, rather than patterning responses to Shh. We generated mutant mice that express Shh that cannot bind proteoglycans but can bind to Ptc. Mutants exhibit a selective deficit in neural stem/precursor cell proliferation. We identified two distinct activities of proteoglycans in Shh-dependent proliferation: these molecules localize Shh to mitogenic niches and they also trigger a gene expression program important for cell division and stem cell renewal.

Proteoglycans are tremendously diverse; each proteoglycan represents a complex, yet poorly understood readout of the glycogenes, the genes encoding core proteins, sugar transporters, glycosyltransferases, sulfatases, and acetylating enzymes. As proteoglycans interact with many growth factors and other molecules, analyses of mutants that interfere with individual steps of proteoglycan synthesis are difficult to interpret. To identify the functions of proteoglycans for Shh responses in vivo, we took an alternative approach. Mutations in the Cardin-Weintraub motif of Hh proteins interfere with proteoglycan binding, but do not alter Ptc affinity, lipid modifications, or expression level. Therefore Shh^Ala/Ala and Shh^Ala/− mutants provide a unique genetic approach for identifying functions of Shh-proteoglycan interactions.

We find proteoglycans are specifically needed for proliferative responses to Shh, but are dispensable for most of its patterning activities. One intriguing finding is that proteoglycans localize Shh to proliferative zones, and also function at the single cell level to determine the nature of the response. Previous studies using mutations in the HSPG synthesizing enzymes of the ext gene family, mutations in the glypican Dally-like or deletion of the Cardin-Weintraub motif have indicated that HSPGs are critical for appropriate localization of Hh proteins ^{11–13, 47}. Here we show that Shh that cannot bind to HSPGs do not accumulate in the EGL mitogenic niche. Thus, direct interactions of Shh with proteoglycans are important for Shh dispersal from the Purkinje cell layer and/or its sequestration in the EGL. Altered Shh localization in mutant animals impairs the establishment of mitogenic niches in developing brain.

In addition to their function in localizing Shh, we find that proteoglycans on receiving cells modulate the Shh response. Previous studies have indicated that proteoglycans function in Hh-responding cells^7,13. Surprisingly, some groups have found that glypicans on receiving cells are needed for full-strength signaling while others have found that responding cell glypicans compete with Ptc1 for Hh interaction, thereby inhibiting Hh activity^11,13. Our studies suggest proteoglycan interactions differentially affect intracellular signaling cascades downstream of Smoothened, and so selectively regulate Shh target gene subsets, thereby modifying the response.

As Shh target gene expression is primarily regulated by the Gli transcription factors³⁶, we examined levels and isoforms of Gli1, Gli2 and Gli3 proteins in wild type and mutant animals. Whereas there were no discernable differences between Gli1 and Gli3 protein levels of wild type and mutant cerebella, clear differences were seen in Gli2 proteins. Gli2 is the major transcription factor driving Shh-induced proliferation^41,48 and proliferation is reduced in multiple locations pre- and post-natally in Shh^Ala/Ala animals. Gli2 proteins can function as either transcriptional activators or repressors; unprocessed Gli2 functions as a weak transcriptional activator^37,39, while Gli2 cleavage and removal of the C-terminal activator domain generates a transcriptional repressor³⁹. Therefore both Gli2 protein levels and the ratio between unprocessed and processed forms are likely to be critical for the net Gli2-dependent transcriptional response. In vivo, gli2 mRNA is lower in wild type than in Shh^Ala/Ala cerebella, but the Gli2 activator/repressor ratio is two-fold greater in wild type. Similarly, gli2 mRNA levels are lower in GCPs acutely stimulated with Shh than Shh^Ala, but GCPs acutely stimulated with Shh had a larger Gli2 activator/repressor ratio than those stimulated with Shh^Ala.

Gli2 expression and processing are both Shh-regulated^39,48. The ability of proteoglycans to affect the levels and proportions of the Gli2 isoforms in vivo and in vitro would alter Shh-dependent transcriptional activation and repression, and so could account for increases and decreases in Shh-target gene expression. One set of Shh-regulated genes, including classical targets ptc1 and gli1, are similarly induced by Shh that can and cannot bind proteoglycans. A second set, including gli2 and N-myc, are preferentially increased by Shh that cannot bind to proteoglycans. Induction of a third set of Shh-regulated genes, including bmi-1, and D-type Cyclins, genes implicated in stem cell maintenance and in tumor biology, requires proteoglycan interactions. Differences in Shh target genes are observed in vivo and in vitro and are not due to shifts in dose response curves, but may be explained by changes in ligand signaling kinetics. Taken together, analysis of the mutant phenotype indicates that Shh-HSPG interactions particularly affect overall Gli2 transcriptional activity, thereby promoting proliferative responses.

Changes in ligand intracellular localization or ligand perdurance provide two possible mechanisms whereby Shh-proteoglycan interactions could selectively modulate signaling. Shh signaling kinetics are important in determining morphogen responses⁴⁶. Here we find that Shh-proteoglycan interactions affect Shh perdurance in responding cells and alter the temporal profile of Shh signaling. It has been proposed that signal duration is proportional to Shh concentration⁴⁶, our data indicate that proteoglycan interactions alter that relationship. In this way, proteoglycan interactions can modulate the pattern of gene induction and the biological response to Shh. Proliferation without appropriate patterning is a cardinal feature of tumor biology. The Hedgehog pathway may promote oncogenesis when it stimulates proliferation without patterning. It is intriguing that glypican overexpression can augment growth in cancers that depend on Shh activity, including rhabdomyosarcoma, prostate cancer, and pancreatic carcinoma^49,50. Similarly, D-type Cyclins and bmi-1, gene targets that require Shh-HSPG interactions, play central roles in tumor stem cell biology^17,18.

In summary, we have demonstrated that Shh-proteoglycan interactions are selectively critical for Shh-dependent mitogenesis, functioning in two ways. First, proteoglycans localize Shh to establish mitogenic niches. Second, proteoglycans on responsive cells selectively promote intracellular pathways that lead to precursor proliferation. Both mechanisms contribute to the phenotype of the Shh^Ala/Ala mice. Thus, Shh-proteoglycan interactions promote proliferation within mitogenic niches by localizing Shh to these proliferative zones and by modulating intracellular signaling cascades and transcriptional programs.

METHODS

Section binding assay

To evaluate proteoglycan binding in situ¹⁴, cryosections were treated with PBS, vehicle control, 500 mU/ml of heparinases (Sigma), or 500 mU/ml of PI-PLC for 1 h (37°C), then O/N (4°C). Shh:AP or Shh^Ala:AP was added (1 h, RT), sections were washed and immunostained with anti-alkaline phosphatase.

Size measurements

6 month old animals were weighed, then sacrificed. Total brain and cerebella were weighed. Olfactory bulb dimensions (OB) were measured from 10 μm thick coronal cryosections and overall volume was calculated. At least 3 pairs of Shh^Ala/Ala and age-matched Shh^Ctl/Ctl were assessed.

Shh staining quantification

Quantification was done on mid-sagittal cerebellar sections from three wild type-Shh^Ala/Ala littermate pairs, stained in parallel for Shh. Staining was analyzed in primary and secondary fissures using NIH Image J software. In each fissure, three lines from the pia through the IGL were drawn perpendicular to the pia, and an intensity plot was determined for each line (plot profile function).

Proliferation and apoptosis quantification in vivo

Analyses of p-Histone H3 staining, BrdU labeling, and TUNEL were performed on cerebella from 3–5 matched wild-type and mutant littermate pairs. Proliferation or apoptotic index was calculated by dividing the number of positive cells by the number of DAPI-stained cells in the EGL. Comparable locations (midpoint of primary, secondary and tertiary fissures within vermis) were assessed. In coronal sections of adults, every 9th section of the SGL or SVZ was stained with p-Histone H3, BrdU, and activated caspase 3. At least 3 pairs of adult Shh^Ala/Ala animals and age-matched controls were analyzed. SGL was the inner half of the thickness of the dentate gyrus, extending two cell widths into the hilus.