General Base Swap Preserves Activity and Expands Substrate Tolerance in Hedgehog Autoprocessing

Abstract

Hedgehog (Hh) autoprocessing converts Hh precursor protein to cholesterylated Hh ligand for downstream signaling. A conserved active-site aspartate residue, D46, plays a key catalytic role in Hh autoprocessing by serving as a general base to activate substrate cholesterol. Here we report that a charge-altering Asp-to-His mutant (D46H) expands native cholesterylation activity and retains active-site conformation. Native activity toward cholesterol was established for D46H in vitro using a continuous FRET-based autoprocessing assay and in cellulo with stable expression in human 293T cells. The catalytic efficiency of cholesterylation with D46H is similar to that with wild type (WT), with k_max/K_M = 2.1 × 10³ and 3.7 × 10³ M⁻¹ s⁻¹, respectively, and an identical pK_a = 5.8 is obtained for both residues by NMR. To our knowledge this is the first example where a general base substitution of an Asp for His preserves both the structure and activity as a general base. Surprisingly, D46H exhibits increased catalytic efficiency toward non-native substrates, especially coprostanol (>200-fold) and epicoprostanol (>300-fold). Expanded substrate tolerance is likely due to stabilization by H46 of the negatively charged tetrahedral intermediate using electrostatic interactions, which are less constrained by geometry than H-bond stabilization by D46. In addition to providing fundamental insights into Hh autoprocessing, our findings have important implications for protein engineering and enzyme design.

Hedgehog (Hh) signaling is an important pathway for embryonic development, stem cell maintenance, and tissue repair in metazoans.^1–4 Abnormal activation of the Hh signaling pathway has been associated with many common types of cancers, including lung, breast, prostate, and ovarian cancer.^5–8 Hh autoprocessing (Figure 1A) is a required step preceding Hh signaling, in which the Hh precursor protein self-cleaves to release a cholesterylated Hh ligand.^1,9 The cleavage of the two-domain Hh precursor (HhN + HhC) is catalyzed by both the Hint (Hedgehog/intein) subdomain within HhC using intein-like chemistry and the sterol recognition region (SRR) which binds cholesterol.¹⁰ There are two steps in Hh autoprocessing: N–S acyl shift and then transesterification (Figure 1A). In the N–S acyl shift, the thiol group of a conserved cysteine 1 (C1) of HhC carries out a nucleophilic attack on the carbonyl of the preceding glycine, the last residue of HhN, and forms a thioester intermediate. The hydroxyl group of a cholesterol molecule then attacks the thioester, displacing HhC and cholesterylating HhN at its C-terminal glycine residue. Cholesterylation is a crucial step for Hh ligand generation. Abolishing cholesterylation prevents Hh ligand secretion, which allows the premature degradation of Hh ligand in the endoplasmic reticulum and suppression of downstream signaling events.^11,12 The lipidation of Hh ligand also restricts extracellular diffusion to facilitate the formation of signaling gradients during embryogenesis.¹³

Figure 1.

Hedgehog (Hh) autoprocessing and multiple sequence alignment of HhC domain. (A) Two catalytic steps in the autoprocessing of Hh precursor. (B) Multiple sequence alignment of HhC domain (only N-terminal region of HhC is shown) from different organisms using Clustal Omega.¹⁴ All the identical, conserved, and semiconserved residues are indicated using an asterisk (*), colon (:), or period (.), respectively. Sequences listed from top to bottom, except C. elegans, are from the following entries in UniProt (www.uniprot.org): Q02936, Q15465, G1TCI1, G3UC88, G5AKL0, Q98938, Q9U5Z6, A7RW45, H2LTU8, Q91610, and O61676. The sequence for C. elegans is from Wormbase (www.wormbase.org): W06B11.4b. C1, D/H 46, and ‘TXXH’ motifs are at the active site of the hint domain and are highlighted. Percent identity is based on HhC sequence alignment. (C) Conserved residues at the active site of the WT Hh Hint domain, including C1, D46, T69, H72, and C143.

Previously, we demonstrated that D46 coordinates two catalytic steps of Hh autoprocessing.¹⁵ Mutagenesis and crystal structure of Drosophila melanogaster (Dme) Hh Hint domain have shown several conserved residues at the active site (Figure 1C), including C1, D46, T69, H72, and C143.¹⁶ Mutagenesis of D46 to alanine (A), asparagine (N), arginine (R) and glutamate (E) abolishes Hh activity toward cholesterol.¹⁵ We were intrigued by multiple sequence alignment of Hh Hint domain (Figure 1B), which showed a charge-altering D-to-H substitution (D46H) in the sequences of Hh and Hh-related proteins in many organisms including C. elegans.^17,18 In solution chemistry, carboxylic acid and imidazole buffers can function almost interchangeably as acid/base catalysts; however, in enzymatic systems, changing a key catalytic functional group from carboxylic acid (D) to imidazole (H) usually leads to impaired or a total loss of function.^19,20 To investigate whether these natural “histidine mutants” in Hh and Hh-related proteins could carry out autoprocessing and cholesterylation, we characterized enzyme activity of D46H toward native cholesterol and a library of 17 non-native sterol and sterol-like molecules, and also made structural comparisons through NMR and X-ray crystallography.

To test if D46H preserves enzyme activity, we utilized our FRET-based in vitro assay²¹ (Figure 2A). The FRET reporter, C-HhC-Y, is a single polypeptide composed of cyan fluorescent protein (CFP) and yellow fluorescent protein (YFP) fused to the N- and C-termini of the Drosophila HhC domain, respectively. Before cleavage, C-HhC-Y exhibits FRET due to the spatial proximity of CFP and YFP, which is subsequently lost when CFP is cleaved off by cholesterol during autoprocessing. In the absence of cholesterol (Figure 2B), D46H exhibited stable FRET signal over the assay interval. In the presence of cholesterol, FRET signal of the D46H decreased with a rate similar to that of the WT, indicating a comparable autoprocessing activity as WT (Figure 2B, compare red and black traces). This is in contrast to D46A, D46N, D46E, and D46R mutations, which abolished cholesterylation.¹⁵ The cholesterylation activity of D46H and WT was further validated using a fluorescently tagged cholesterol derivative as substrate (Figure 2C). Precursor protein was first reacted with a molecule of cholesterol derivative where the side chain contains an azide (Figure S1A). Product mixtures were then subjected to strain promoted click chemistry with alkyne modified rhodamine (Figure S1B), followed by gel electrophoresis and fluorescence in-gel imaging. After the autoprocessing reaction, the product polypeptide, HhN, exhibits fluorescence under UV, confirming covalent sterylation brought about by both WT and D46H (Figure 2C, right panel). In addition, we also confirmed that D46H mutant can mediate Hh autoprocessing in human 293T cells (Figure S2), although less efficiently than the WT. The lower efficiency of D46H in cell culture may be due to altered interaction with protein disulfide isomerase (PDI), which is required for Hh autoprocessing in endoplasmic reticulum.²²

Figure 2.

Activity and structural characterization of D46H. (A) Schematics of FRET-based assays. C-HhC-Y represents CFP-HhC-YFP, in which C-chol = CFP-cholesterol and HhC-Y = HhC-YFP. (B) D46H has a similar autoprocessing rate as the WT in the presence of cholesterol. (C) Validation of HhC cholesterylation using a fluorescently tagged cholesterol derivative. (D) The WT Hh Hint domain (in blue, PDB: 1AT0) and D46H (in green, PDB: 6TYY) have almost identical 3D structure (RMSD = 0.4 Å). Conserved residues in the active site of WT and D46H were magnified on the right. (E) pK_a measurement of H46 by NMR pH titration, showing identical pK_a between D46 and H46.

The WT and D46H Hint domains have almost identical 3D structures, with an RMSD of 0.4 Å between two crystal structures (Figure 2D). Conformation of the active site is also preserved (Figure 2D). The X-ray data agrees with the minimal chemical shift perturbation (CSP) caused by D46H in solution NMR (Figure S3A,B). A key finding for D46 is an elevated pK_a of 5.8, which enables the proton shuttling through its side chain during WT Hh autoprocessing.¹⁵ To understand how D46H mediates Hh autoprocessing, the pK_a of H46 was measured by pH titration using long-range ¹H–¹⁵N HMQC NMR spectra (Figure S4). The titration curve for H46 was fit to a modified Henderson–Hasselbalch equation,²³ giving a pK_a value of 5.8 (Figure 2E), identical to the pK_a of D46. This remarkable correlation suggests similar catalytic roles of H46 in Hh autoprocessing as D46.

Recruitment of cholesterol by SRR subdomain of HhC is a crucial step in Hh autoprocessing. However, the specific interactions involved in sterol recognition remain largely unresolved. Studies show that HhC can also recognize diverse substrates in addition to cholesterol.^24,25 While HhC has a relatively broad tolerance toward sterols with alterations distant from the site of bond scission, its autoprocessing efficiency is sensitive to changes in sterol A-ring.²⁶ To investigate if D46H changes in substrate selectivity, the autoprocessing activities toward a library of 18 different sterols and sterol analogs (Figure S5) were characterized by Michaelis–Menten enzyme kinetic analysis. (Figures S6 and S7). Surprisingly, D46H exhibits increased promiscuity toward non-native substrates (Tables 1 and S1), especially for coprostanol and epicoprostanol (Table 1).

Table 1.

Expanded Substrate Tolerance of D46H for Structurally Diverse Substrates

substrate	efficiency (kmax/KM) (×103 M−1 s−1)		D46H/WT
	WT	D46H
cholesterol	2.1 ± 6%	3.7 ± 12%	1.8 ± 13%
epicoprostanol	<0.0002	0.063 ± 16%	>315
coprostanol	0.011 ± 8%	2.4 ± 7%	218 ± 11%
abiraterone	0.38 ± 20%	2.0 ± 3%	5.3 ± 20%
A-nor-cholestanol	0.29 ± 18%	1.3 ± 15%	4.5 ± 23%
allocholesterol	2.3 ± 9%	7.6 ± 11%	3.3 ± 14%
desmosterol	1.9 ± 9%	5.7 ± 13%	3.0 ± 16%
galeterone	8.0 ± 16%	19 ± 11%	2.4 ± 19%
5-androsten-3β-ol	0.94 ± 8%	1.6 ± 7%	1.7 ± 11%
ergosterol	1.6 ± 15%	2.3 ± 11%	1.4 ± 19%

For cholesterol, D46H and WT have similar kinetic parameters, where D46H has a K_M of 1.0 μM and k_max of3.6 × 10⁻³ s⁻¹, while WT has a K_M of 1.9 μM and k_max of 3.8 × 10⁻³ s⁻¹ (Figure 3A). Coprostanol, a microbial metabolite of cholesterol²⁷ with a bent A-ring, shows much higher autoprocessing activity by D46H than WT. D46H has a K_M of 2.4 μM and k_max of 2.9 × 10⁻³ s⁻¹, while WT has a K_M of 25 μM and k_max of 0.3 × 10⁻³ s⁻¹ for coprostanol (Figure 3B). The efficiency of D46H over WT is >200-fold greater (Table 1). For epicoprostanol, D46H has a K_M of 16 μM and k_max of 1.0 × 10⁻³ s⁻¹, while WT shows no detectable reactivity (Figure 3C). The efficiency of D46H over WT is >300-fold greater for epicoprostanol (Table 1). These data demonstrate that while WT Hh is less tolerant toward cholesterol analogs as substrates, D46H exhibits comparable or enhanced activity toward non-native substrates, indicating an expanded substrate scope of the D46H.

Figure 3.

Michaelis–Menten enzyme kinetics of D46H and WT toward cholesterol (A), coprostanol (B), and epicoprostanol(C). (D) 3D structures of cholesterol (cyan), coprostanol (orange), and epicoprostanol (yellow) are aligned using PyMOL.

As a general base, D46 and H46 are expected to have equivalent efficiency due to their identical pK_a as stated above. However, the D46H may be more substrate promiscuous because H46 is able to promote formation of the negatively charged tetrahedral addition intermediate through favorable electrostatic interaction with its side chain imidazolium (Figure S8). Electrostatic stabilization interactions are longer range and less geometrically constrained than H-bonding (Figure 4), which is the only mode of transition state stabilization possible for D46. It is conceivable that a geometric analog of cholesterol, like coprostanol, could displace D46 out of its catalytically competent position, preventing H-bonding and thereby suppressing catalysis. With H46, electrostatic stabilization would be less sensitive to geometric alteration of the substrate, allowing reaction with more structurally diverse sterols.

Figure 4.

H46 side chain can provide favorable electrostatic stabilization not available in D46.

Changing a key enzymatic residue from aspartate to histidine is expected to compromise function.^19,20 There are few examples of active-site D-to-H mutation that preserve enzyme activity. The closest example is Δ⁵-3-ketosteroid isomerase (KSI), where the WT active-site residue D38 shuttles the sterol proton while the D38H mutation lowers the catalytic efficiency by ~20-fold.²⁸ It was suggested that in KSI a neutral H38 may be more favorable for binding the hydrophobic sterol than D38.²⁸ Here we have a surprising case where in vitro kinetics, pK_a values, 3D structure, and active-site conformation are nearly identical between the WT and D46H, indicating that D-to-H mutation, even at a critical active-site position, can be a synonymous substitution. The one change we observed is that the D46H exhibits increased activity toward non-native substrates. These observations have important implications in enzyme engineering for expanding or altering substrate specificity.