Skinny Hedgehog, an Acyltransferase Required for Palmitoylation and Activity of the Hedgehog Signal

See allHide authors and affiliations

Science  14 Sep 2001:
Vol. 293, Issue 5537, pp. 2080-2084
DOI: 10.1126/science.1064437


One of the most dominant influences in the patterning of multicellular embryos is exerted by the Hedgehog (Hh) family of secreted signaling proteins. Here, we identify a segment polarity gene in Drosophila melanogaster, skinny hedgehog(ski), and show that its product is required in Hh-expressing cells for production of appropriate signaling activity in embryos and in the imaginal precursors of adult tissues. Theski gene encodes an apparent acyltransferase, and we provide genetic and biochemical evidence that Hh proteins from skimutant cells retain carboxyl-terminal cholesterol modification but lack amino-terminal palmitate modification. Our results suggest thatski encodes an enzyme that acts within the secretory pathway to catalyze amino-terminal palmitoylation of Hh, and further demonstrate that this lipid modification is required for the embryonic and larval patterning activities of the Hh signal.

In a genetic screen for components of the Drosophila Hh signaling pathway (1) we found that mutations in theskinny hedgehog (ski) gene produced phenotypes typical of those resulting from loss of hh orwingless (wg) function (2). Mutant individuals lacking zygotic function of ski survive until early pupal stages. However, lethality during the embryonic period and a strong segment polarity phenotype (Fig. 1, A to C) result from additional loss of the maternal component of ski function. To determine whetherski is required specifically for either the Hh or Wg pathway, we analyzed larval tissues in which the two pathways function independently of each other in distinct subpopulations of cells (3). Mutant imaginal discs show wild-type levels of Wg target gene expression but are abnormally small (Fig. 1, D and E). The expression of the Hh target genes decapentaplegic(dpp) and patched (ptc), respectively, is strongly reduced or absent in these discs (Fig. 1, F to I) (4–6). Thus, we conclude thatski acts in the Hh signaling pathway and that skimutant discs are undersized because dpp is expressed at abnormally low levels.

Figure 1

The ski product is a component of the Hh signaling system and is required in Hh producing cells for target gene expression in Hh transducing cells. (A toC) Cuticle preparations of a wild-type embryo (A), ahh AC mutant embryo (B), and aski –/– embryo (C) derived from aski –/– germline clone. All further micrographs (with the exception of Fig. 2A) are confocal images of third instar wing imaginal disc preparations stained with antibodies against the proteins indicated on the lower right of each panel. Genotypes are indicated on the lower left of each panel. (Dto I) Expression of Wg target gene Distalless (Dll) and the Hh targets, ptc-lacZ or dpp-lacZ in wild-type (D, F, and H) and ski mutant (E, G, and I) discs. (Jto M) Large clones of ski mutant cells marked by the absence of πMyc expression [green, (J) and (L)] and double-stained for the expression of dpp-lacZ [red, (K) and (M)]. The approximate position of the anteroposterior compartment boundary is indicated by a dotted line. Clones were induced in aMinute background.

To determine whether ski is required for the production of active Hh signal or for transduction of this signal, we assayed Hh target gene expression in genetic mosaics. Clones lackingski function in the anterior compartment of wing imaginal discs show no effect on dpp or ptc expression (Fig. 1, J and K). In contrast, discs with large clones of posterior compartment cells lacking ski display reduced expression of Hh target genes in adjacent anterior cells (Fig. 1, L and M). As this requirement for ski function is similar to that ofhh itself (7), we interpret these results as evidence that Ski is essential for effective production of the Hh signal. Ski appears not to be required for expression of thehh gene, as wild-type levels of hh transcription are observed in mutant discs (Fig. 2, A and B). Neither does Ski appear to be required for secretion of the Hh protein, as abnormal Hh protein accumulation in ski mutant cells is not observed (Fig. 2, C and D). One remaining possibility is that the ski gene product controls a maturation event critical for activity of the Hh signal.

Figure 2

Ski is required for a Hh modification that is essential for the activities of Hh, HhN, and HhCD2. (Aand B) Early third instar wild-type (A) and late third instar ski mutant (B) discs stained for the expression ofhh-lacZ. (C) Cells mutant fordispatched (disp, marked by the absence ofarmadillo-lacZ expression shown in red) fail to secrete Hh protein (green) properly (7). In contrast, cells mutant forski (D) do not show abnormally high levels of Hh protein. Discs in (C) and (D) were prepared and stained simultaneously under identical conditions. The approximate position of the anteroposterior compartment boundary is indicated with a dotted line. (E and F) Expression of aUAS-hhNHA transgene under control ofengrailed-Gal4 (en::hhNHA) in ahh ts background under nonpermissive conditions causes overgrowth in the anterior compartment and a stripe ofptc-lacZ expression (red). In a ski mutant background (F), neither of these activities of HhN can be detected. HhN protein is visualized with an antibody against the HA epitope. (G) Expression of a HhCD2 fusion protein (12) in a hh ts background under nonpermissive conditions results in a narrow stripe ofptc-lacZ expressing cells (red). HhCD2 (green) has no detectable activity when produced by ski mutant cells (H).

A critical step in maturation of the Hh protein is autoproteolytic cleavage at an internal site, during which the bioactive NH2-terminal product of cleavage acquires a COOH-terminally attached cholesterol (8, 9). This modification can be bypassed by the expression of truncated or chimeric Hh proteins, such as HhN (10, 11) or HhCD2 (12), which are produced in bioactive form without cleavage or cholesterol addition. Neither of these proteins is active in the absence of ski function (Fig. 2, E to H), indicating that Ski must control a different property of the Hh signal, one that is shared by HhNp, HhN, and HhCD2.

To investigate the role of Ski, we localized the ski gene to cytological position 63B6 (13) and used double-stranded RNA injections and genomic transgenes to identify a single transcript encoding ski function (14). This transcript contains an open reading frame encoding a protein of 500 amino acids (15). A single base pair in codon 145 is deleted in the ski allele ski Bg resulting in a frameshift mutation and a truncation of the presumptive protein product (Fig. 3A). Definitive identification of the ski gene was obtained from the complete rescue of all phenotypes associated with skimutations by a transgenic wild-type open reading frame expressed under control of the ubiquitously active promoter of thetubulinα1 gene (Fig. 4C). The Ski protein sequence exhibits an extraordinarily high content of hydrophobic amino acids, which fall into 10 to 12 putative membrane-spanning domains (Fig. 3A). There is a short but significant sequence homology between Ski and members of a diverse superfamily of membrane-associated acyltransferases (Fig. 3B) (16). The biochemically characterized proteins of this superfamily transfer fatty acids onto hydroxyl groups of membrane-embedded nonproteinaceous targets (17–19).

Figure 3

Properties of theDrosophila Ski and Hh proteins. (A) Hydrophobicity plot of the 500–amino acid Ski protein sequence. Positive values represent hydrophobic regions. Schematic representation of the proteins encoded by the wild-type andski Bg mutant allele, indicating the truncation caused by a frame shift (orange). Position and extent of the conserved stretch of residues shown in (B) is highlighted in green. (B) Sequence alignment of conserved residues in Ski and several other putative membrane-boundO-acyltransferases (16). The residues in red correspond to highly conserved polar residues, D341 and H381, of Ski that have been mutated to enzymatically inactivate Ski. Highly conserved hydrophobic residues and conservative substitutions are shown in green and blue, respectively. The human homolog (hs Ski1) can be aligned throughout most of its extent to Drosophila Ski (Z. Chamoun et al., data not shown). Abbreviations: hs Ski1, human Skinny hedgehog protein 1 (31); ce CAB16518, closest homolog of Ski in the Caenorhabditis elegans genome. dm Porc, Drosophila porcupine gene product (25); hs ACAT, human acyl-CoA:cholesterol acyltransferase-1 (19); mm DGAT, mouse acyl CoA:diacylglycerol acyltransferase (18); sc Are1, yeast alpha 2 repression gene 1 (32); si WaxSyn, jojoba embryo wax synthase (17); (C through E) Purification and mass determination of HhNp and HhNpC85S. Following initial enrichment by preparation of detergent-resistant membranes (DRM), HhNp and HhNpC85Sexpressed in Drosophila S2 cultured cells were further purified by binding to and elution from an antibody affinity column [HhNp elution shown in (C), arrow] (33). Masses of these proteins were determined by MALDI-TOF mass spectrometry (35) with horse skeletal myoglobin and bovine trypsinogen as internal standards (D and E). The average of two mass determinations for HhNp was 20,235 daltons; the average of three determinations for HhNpC85S was 19,982 daltons. (F) Signaling activity of HhN and HhNC85S. Media conditioned by S2 cells with stably integrated constructs for expression of HhN and HhNC85S were tested for activity in a cl-8 cell reporter assay (36). HhN induced reporter expression by more than 100-fold whereas HhNC85S had no inducing activity, despite similar protein levels in the medium (see Western blot in panel below; S2 denotes medium conditioned by the untransfected parental cell line).

Figure 4

Genetic and biochemical evidence that Ski functions as an acyltransferase for Hh proteins. (A) A transgene expressing wild-type Hh protein in P compartment cells (marked by GFP expression in green) rescues disc growth andptc-lacZ expression of hh ts mutant discs. Neither activity is observed with transgenes expressing HhC85S (B). (C and D) One copy of a transgene encoding wild-type Ski fully restores viability and Hh target gene expression in ski mutant animals (C). Ski protein with two amino acid substitutions at the presumed catalytic sites (D341 and H381) shows no detectable restoration of Hh target gene expression (D). (E to H) Wild-type or C25S mutant forms of mouse Shh were produced in cells of actin5c>Gal4expressing clones (marked by the absence of green CD2 staining). Shh induces high levels of dpp-lacZ expression in a wild-type background (E), and reduced, intermediate levels of dpp-lacZ(and only low levels of ptc-lacZ, not shown) inski mutant discs (F). Equal intermediate levels ofdpp-lacZ (or low levels of ptc-lacZ, not shown) are induced by ShhC25S in a wild-type (G) or skimutant background (H). dpp-lacZ expression is shown in red. Twofold magnifications of selected clones (small white squares) are shown in the upper right corner of panels (E) to (H) as single-labeled images to indicate that wild-type Shh protein produced fromski mutant cells and mutant Shh protein produced from wild-type cells retain the ability to induce dpp-lacZnonautonomously. (I through L) Reversed-phase HPLC analysis of HhNp proteins. (I) Partially purified HhNp and HhNpC85S from S2 cells (Fig. 3, C to E) were analyzed by RP-HPLC in a C4 column developed with a gradient of increasing acetonitrile:n-propanol (1:1) (37). The HhNpC85S protein elutes primarily in fractions 3 through 6, whereas the HhNp protein elutes primarily in fractions 8 through 11. (J) HhNp proteins from cells (33) incubated with dsRNA sequences (22) corresponding to yellow fluorescent protein (YFP) as a control and NH2- and COOH-terminal regions of the Ski coding sequence were fractionated by RP-HPLC in a C4 column. HhNp protein from YFP-treated cells showed a distribution indistinguishable from untreated S2 cells (R. K. Mannet al., data not shown), whereas HhNp from cells incubated with dsRNA corresponding to NH2- or COOH-terminal regions of Ski, or both, displayed a significant shift toward earlier elution. (K) HhNp protein from wild-type larval tissue (33) behaved similarly to that expressed in S2 cells, whereas homozygous ski larvae exclusively produced HhNp protein with an early pattern of elution. (L) On the basis of densitometric scanning, the ratio of late-eluting (fractions 8 through 11) to early-eluting (fractions 3 through 6) HhNp is plotted on a log scale. A ratio significantly greater than 1 is observed for HhNp in purified form, from YFP dsRNA-treated cells, and from wild-type third instar larvae, and less than 1 for samples from cells with impaired ski function.

It has recently been demonstrated that in addition to COOH-terminal addition of cholesterol, the human Sonic hedgehog (Shh) protein is further modified by NH2-terminal attachment of a palmitoyl adduct in a manner dependent upon the NH2-terminal cysteine residue (20). Our finding that ski encodes a putative acyl transferase raises the possibility that it could function as an enzyme for the palmitoylation of Hh protein. To determine whetherDrosophila Hh is NH2-terminally acylated, we purified the protein from Drosophila tissue culture cells and subjected it to mass spectrometric analysis. The experimentally determined mass of HhNp (20,235 daltons; Fig. 3C) exceeds by 238 daltons the average mass expected for the amino-terminal signaling domain linked to cholesterol alone (19,996.8 daltons). The difference between these values corresponds closely to the expected mass of a palmitoyl adduct in ester or amide linkage (238.4 daltons). Furthermore, substitution of a serine in place of the NH2-terminal cysteine results in a protein modified by cholesterol alone (Fig. 3D; average experimental mass, 19,982; 19,980.7 daltons expected). These results support the conclusion thatDrosophila HhNp undergoes NH2-terminal palmitoylation.

If Ski functions as the Hh palmitoyl-transferase, then the absence of Ski should result in Hh protein that is not NH2-terminally modified, with a consequent loss or reduction of Hh signaling activity. Genetic evidence for this possibility derives from the functional equivalence between mutations that abolish the enzymatic activity of Ski and alterations in Hh proteins that prevent their NH2-terminal acylation. If the wild-type skialleles are replaced by a ski transgene in which two presumptive active-site residues (16) of Ski are mutated, Hh signaling activity is greatly reduced (Fig. 4, C and D). The same phenotypes are observed when posterior wing cells express the NH2-terminally mutant form of Hh (HhC85S) instead of wild-type Hh protein (Fig. 4, A and B). Interestingly, although mouse Shh protein also depends on ski function for maximal activity, it retains some signaling activity if secreted fromski mutant cells (Fig. 4, E and F). The same partial reduction in Shh activity is observed with a mutant form of Shh in which the NH2-terminal cysteine residue is mutated (Fig. 4G). The activity of ShhC25S (21) is not further reduced when secreted from ski mutant cells (Fig. 4, F and H). These results suggest that Ski is specifically required for the NH2-terminal addition of palmitate to Hh.

Biochemical evidence for Ski's role in Hh palmitoylation was obtained by examination of Hh protein produced from cells with reducedski function. Hh proteins were analyzed by reversed-phase high pressure liquid chromatography (RP-HPLC), which resolves doubly-modified HhNp from proteins lacking the NH2-terminal palmitate adduct (compare HhNp to HhNpC85S; Fig. 4I). Note that a small proportion of HhNp in normal cells and tissues displays the earlier elution profile characteristic of the HhNpC85Sprotein, suggesting that autoprocessing and cholesterol modification precede palmitoylation during Hh biogenesis. Drosophilacultured cells expressing HhNp were treated with double-stranded RNA corresponding to two regions of the ski coding sequence (22). A significantly increased proportion of the protein from these cells eluted at a position indicative of absence of the NH2-terminal palmitate (Fig. 4, J and L). A similar shift in elution was observed for HhNp from tissues of mutant third instar larvae, except that all of the HhNp eluted earlier, as would be expected for a complete loss of palmitate transfer (Fig. 4, K and L). Reduction or loss of Ski function thus reduces the hydrophobicity of HhNp to that of HhNpC85S, consistent with a loss of NH2-terminal palmitoylation and with a role for Ski function in palmitate transfer.

It has been proposed that the linkage of palmitate to the NH2-terminus of ShhNp is via an amide bond (20), but the mechanism of amide formation is not clear. The membership of Ski in a family of enzymes catalyzing O-linked acyl transfers strongly argues for a mechanism in which a thioester intermediate is formed with the side chain of the NH2-terminal cysteine, followed by a rearrangement through a five-membered cyclic intermediate to form the amide (20). It is curious that addition of cholesterol, the other lipid modification of the mature Hh signal, also involves a cyclic intermediate and thioester chemistry but proceeds in reverse, from amide to ester (9, 11).

In summary, ski encodes an apparent acyltransferase critically required for Hh palmitoylation, a modification vital for signaling activity in vivo. Our analysis indicates that nonacylated Hh protein is secreted efficiently and reaches target cells (23). The severe hh-like phenotypes ofski mutant animals demonstrate that the NH2-terminal acyl-group plays a critical role in enabling Hh to gain access to and inactivate Ptc. Finally, we note that Porcupine, required for production of active Wnt signal (24,25), is a member of the same family of acyl transferases as Ski (16) (Fig. 3B), raising the possibility of a more widespread role for acylation of secreted protein signals in development.

  • * These authors contributed equally to this work.

  • To whom correspondence should be addressed. E-mail: pbeachy{at} (P.A.B.), basler{at} (K.B.)


View Abstract

Stay Connected to Science

Navigate This Article