Crystal structures of 17-beta-hydroxysteroid dehydrogenase 13

doi:10.21203/rs.3.rs-2688507/v1

Download PDF

Article

Crystal structures of 17-beta-hydroxysteroid dehydrogenase 13

https://doi.org/10.21203/rs.3.rs-2688507/v1

This work is licensed under a CC BY 4.0 License

Journal Publication

published 24 Aug, 2023

Read the published version in Nature Communications →

Version 1

posted

You are reading this latest preprint version

Hydroxysteroid 17-beta-dehydrogenase 13 (HSD17B13) is a hepatic lipid droplet-associated enzyme that is upregulated in patients with non-alcoholic fatty liver disease. Recently, there have been several reports that predicted loss of function variants in HSD17B13 protect against the progression of steatosis to non-alcoholic steatohepatitis with fibrosis and hepatocellular carcinoma. Here we report the first known crystal structures of full length HSD17B13 in complex with its NAD⁺ cofactor and with small molecule inhibitors from two distinct series. These structures provide insights into a mechanism for lipid droplet-associated proteins anchoring to membranes as well as a basis for HSD17B13 variants disrupting function. Two series of inhibitors interact with the active site residues and the bound cofactor similarly, yet they occupy different paths leading to the active site. These structures provide ideas for structure-based design of inhibitors that may be used in the treatment of liver disease.

Biological sciences/Biophysics/Membrane structure and assembly

Biological sciences/Drug discovery

17 β-hydroxysteroid dehydrogenase 13

lipid droplet associated protein

inhibitor

crystal structure

enzyme activity

structure-activity relationship

thermal shift assay

Hydroxysteroid dehydrogenases (HSD) comprise a large family of enzymes that are involved in the biogenesis and metabolism of many steroid and non-steroid substrates in animals and human ¹. They are nicotinamide adenine dinucleotide (phosphate) (NAD⁺(P)/NAD(P)H)-dependent oxidoreductases, which interconvert ketones and the corresponding secondary alcohols at different positions of steroidal substrates (3α-, 3β-, 11β-, 17β-, 20α- and 20β-position)². Given their roles in steroid metabolism, HSD family members play important biological roles in human health and have been associated with a variety of diseases^1,3−7. For example, members of the 17b-HSD subfamily have been associated with breast and prostate cancer, polycystic kidney cancer, and Alzheimer’s disease, to name a few ⁸.

Recently, one of those members, 17β-HSD13 (HSD17B13), has garnered significant interest after a potential role in the pathogenesis of non-alcoholic fatty liver disease (NAFLD) was identified through genome-wide association studies ^9–13. NAFLD is a serious disease that affects around one quarter of the global population ¹⁴. The first breakthrough towards identifying the role of HSD17B13 in NAFLD pathogenesis in human occurred when an exome-wide association study revealed that a splice variant, rs72613567 or isoform D (isoD), conferred a protective effect against NAFLD disease severity. This variant was strongly associated with slowed progression of NAFLD toward nonalcoholic steatohepatitis (NASH) and advanced fibrosis ¹⁰. The isoD variant is the product of an insertion of adenine adjacent to the donor splice site of exon 6 (T > TA), which disrupts mRNA splicing. The isoD variant was shown to be translated into a truncated protein (P274del) that was unstable, with low protein level in human liver biopsy samples and decreased enzymatic activity against b-estradiol in vitro and in cell-based overexpression assays ¹⁰.

Subsequent genome-wide association studies have identified several additional human HSD17B13 variants that are protective against progression to advanced NASH and liver injury. These variants either strongly shifted the splicing equilibrium toward isoD ^11–13, resulted in a truncated protein product distinct from IsoD (Ala192LeuFsTERM or A192Lfs) ¹², or resulted in non-synonymous mutations (P260S) ¹¹. A192Lfs and P260S proteins are predicted to be unstable, with decreased protein expression levels, and indeed overexpressed P260S variant protein lacked enzyme activities. This together with the fact that HSD17B13 is significantly upregulated in the liver of patients with NAFLD ^{11 15}, suggests it may be a novel therapeutic target for NAFLD. Whether HSD17B13 variants identified to date therefore protect against NAFLD in human through loss of the scaffolding function or the enzymatic activity or both is unclear.

In fact, very little is known about the function of HSD17B13 beyond it being a liver-specific, lipid droplet (LD)-associated protein^16–18. LDs are storage organelles at the center of lipid and energy homeostasis. They have a unique architecture consisting of a hydrophobic core of neutral lipids, which is enclosed by a phospholipid monolayer that is decorated by a specific set of proteins^19–22. Although several domains of HSD17B13 were reported to be important for LD localization by mutagenesis studies ^17,23−25, currently a structural understanding of HSD17B13 LD association is lacking. In fact, although up to 100–150 major proteins were identified as LD-associated proteins in mammalian cells ^21,26, the mechanism of LD targeting of these proteins is still unclear ²⁷.

Because loss of function HSD17B13 variants are protective against NAFLD, small molecule HSD17B13 inhibitors may serve as oral medications for liver diseases. A key gap for designing potent small molecule inhibitors to HSD17B13, however, is the lack of any published HSD17B13 crystal structures. The available structure of the closest member of 17b-HSD subfamily is that of human 17b-HSD11 (HSD17B11) (PDB ID 1YB1). Like HSD17B13, HSD17B11 is a LD-associated protein, and HSD17B13 and HSD17B11 share 67% and 79% identity and similarity in sequences, respectively, in their core catalytic domains. The available structure for HSD17B11, however, is truncated which was likely done to facilitate crystallization. Specifically, it is missing the initial highly hydrophobic, LD-targeting N-terminal 22 residues ²⁸ and the final amphipathic C-terminal 25 residues, resulting in a soluble, non-membrane associated form of the protein. Additionally, the structure lacks the NAD⁺/NADH cofactor, and based on homology modeling, the active site is collapsed, and the protein is non-functional.

Here we report high resolution crystal structures of cofactor and inhibitor bound dog and human HSD17B13 proteins. These structures provide significant insight into the LD association of HSD17B13 (and HSD17B11) as well as the structural hypothesis behind the reported disease-associated variants. Combined with structure-based mutagenesis they further provide insight into substrate selectivity and inhibitor binding. Finally, comparing the structures of HSD17B13 with two small molecule inhibitors with distinct scaffolds highlights different potential vectors for structure-based design.

Small molecule inhibitors of HSD17B13

A series of HSD17B13 small molecule inhibitors were identified through a high throughput biochemical screening using recombinant human HSD17B13 and b-estradiol as substrate. Two compounds with distinct scaffolds emerged as validated hits (Fig. 1). Compound 1 is a fluorophenol containing compound that is a disclosed antagonist of N-methyl-D-aspartate (NMDA) NR2B receptor ²⁹. Compound 2 is a benzoic acid compound containing a sulfonamide linker, and it was reported to be an AKT1 kinase inhibitor ³⁰. Both were reasonably potent against HSD17B13 in biochemical assays using both b-estradiol or Leukotriene B4 (LTB4) as substrate and NAD⁺ as cofactor.

Structure determination of HSD17B13

HSD17B13 is a LD associated protein ¹⁶, and at the N-terminus of the protein there is a stretch of highly hydrophobic Leu and Ile residues (A1-15: MNIILEILLLLITII). Extensive protein engineering was pursued to remove these hydrophobic residues to favor protein purification and crystallization, but N-terminal truncation efforts generally resulted in low protein yield and/or aggregation (Supplementary Table 1). Two protective variants, IsoD and P260, were attempted and demonstrated reduced protein expression levels. In parallel, we also screened different expression systems (bacteria, Sf9, Expi293), various detergents, together with various engineered mutants, a few of which are listed in Supplementary Table 1. We eventually succeeded in generating and crystallizing the full-length human HSD17B13 protein with a C-terminal GSG linker and His tag to facilitate purification. The protein was extracted and solubilized using detergent micelles and octaethylene glycol monododecyl ether (C12E8), a non-ionic detergent, was identified as the suitable detergent for crystallization. Inclusion of cofactor NAD⁺ and inhibitors were also key for crystallization.

Initial crystallization of the human HSD17B13/compound 1 complex resulted in low resolution (5–6 Å) diffracting crystals. Screening of HSD17B13 from different species for crystallization led to the much-improved crystals of dog wild type HSD17B13 in complex with compound 1 and the cofactor, NAD⁺ (Table 1, compound 1-Dog WT). Subsequent mutation of four residues (G177, V178, T205, and I293) identified at the compound 1 binding site of dog HSD17B13 to the human ones (E177, G178, A205 and V293, dog mutant) further improved the resolution of these crystals (Table 1, compound 1-Dog mutant). We could remove compound 1 from these crystals by soaking with buffer containing NAD⁺ only and obtained NAD⁺ bound dog HSD17B13 crystals (Table 1, apo Dog WT or apo hereafter). Based on findings from dog HSD17B13 crystals, we changed four surface residues away from the active site of human HSD17B13 to those of dog protein at the dog HSD17B13 crystal interface (Q60K, I62R, R71H and E161K, human mutant) to improve resolution of the human HSD17B13 crystals, and obtained useful crystals of human HSD17B13 in complex with NAD⁺ and compound 2 (Table 1, compound 2-human). The structures were solved using molecular replacement method using published HSD17B11 crystal structure (before the highly accurate protein structure program AlphaFold 2 was available ³¹). There was one HSD17B13 dimer in the apo, compound 1-Dog WT and compound 1-Dog mutant crystals, and four HSD17B13 dimers in compound 2-human crystals.

Table 1

Data collection and refinement statistics
Crystal^a	Apo, Dog WT	Compound 1-Dog WT	Compound 1-Dog mutant	Compound 2-Human
Data collection
Space group	P2₁2₁2	P2₁2₁2	P2₁2₁2	P2₁
Cells
a, b, c (Å)	76.38 187.21 65.37	77.03 186.46 65.47	76.73 186.32 65.32	96.15 161.56 100.77
α, β, γ (°)	90, 90, 90	90, 90, 90	90, 90, 90	90 95.048 90
Resolution (Å)^b	93.60–2.47 (2.64–2.47)	71.19–2.23 (2.47–2.23)	93.16–1.93 (2.21–1.926)	100.38–2.65 (2.90–2.65)
Uni. reflections	29,109 (1,456)	32,599	36,866 (1,835)	67,940 (3,398)
R_pim^c	0.077 (0.579)	0.039 (0.57)	0.065 (0.404)	0.083 (0.589)
I / σI	8.3 (1.4)	14.2(1.4)	8.8 (1.8)	7.9 (1.5)
Completeness (%)	93.7 (50.6)	92.9(58.7)	93.6 (66.7)	93.4 (56.3)
Redundancy	6.6 (6.8)	6.6 (6.6)	6.2 (3.8)	3.4 (3.4)
Refinement
Resolution (Å)	30.86–2.47	23.16–2.22	21.1–1.91	37.9–2.65
No. reflections	29,084 (1,449)	32,551 (1,607)	36,628 (1,441)	67,909 (3,399)
R_work^d	0.200 (0.265)	0.210 (0.286)	0.200 (0.223)	0.212 (0.291)
R_free^e	0.227 (0.37)	0.228 (0.251)	0.224 (0.242)	0.244 (0.338)
No. atoms
Protein	4,678	4,553	4,142	17,331
Heterogen	88	116	144	777
Water	136	81	73	23
B-factors (Å²)
Protein	57.9	67.0	43.3	63.2
Heterogen	43.5	72.8	35.2	62.5
Water	48.4	46.6	24.3	35.6
R.m.s. deviations
Bond lengths (Å)	0.01	0.01	0.01	0.008
Bond angles (°)	1.1	1.1	1.1	0.98
Ramachandran Favored, outlier (%)^f	97.8, 0.0	96.5,0.3	96.9,0.0	96.3,0.5
^aData sets were collected on one crystal for each complex. Values in parentheses are for highest-resolution shell.
^bStatistics in the highest resolution bins are shown in parenthesis. Based on anisotropic scaling and merge.
^cR_pim=∑_hkl\(\sqrt[2]{1/(n-1)}\)∑_i\|I_i(hkl)-<I(hkl)>\|/∑_hkl∑_iI_i(hkl), where I_i (hkl) is the ith intensity measurement of reflection (hkl), <I(hkl) > is the mean intensity from multiple observations of that reflection, and n is the number of observations of that reflection.
^dR_work=∑_hkl│\|F_obs\|-\|F_calc\|│/∑_hkl∑│F_obs│, where F_obs and F_calc are observed and calculated structure-factor amplitudes, respectively.
^eR_free is calculated using 5% of reflections randomly excluded from refinement. For all crystals used in this study, the same set of reflections were chosen for exclusion.
^fPercentages of residues in the most favorable and disallowed regions of the Ramachandran plot.

Structural basis for the LD localization of HSD17B13:

In crystals both the human and the dog HSD17B13 formed dimers (Fig. 2). The overall structure of each HSD17B13 subunit could be divided into two parts: (i) the catalytic core domain (residues 29–259) that contained the cofactor binding site and the catalytic center and (ii) the membrane anchoring domain that consisted of the N-terminal helix (N2-P28) which was highly hydrophobic and a C-terminal proximal helix-turn-helix motif (P260-N286) which formed an amphipathic patch on the protein surface. Three proline residues at key positions in the protein marked the boundaries of these peptides: P28, P260 and P274. These prolines terminated the prior secondary structures and introduced sharp turns in the following peptides. Prolines are known to constrain the conformational space available to the protein backbone ³². Interesting to note that both the P274del and P260S human variants were reported to be likely loss of function with lower protein stability and expression levels than the wild type HSD17B13^10,11. And consistently, modifications around P28 resulted in minimal protein expressions (Supplementary table 1). The hydrophobic residues in the N-terminal helix of HSD17B13 were identified as critical for LD association in previous mutagenesis studies ²³. A similar N-terminal sequence in HSD17B11 was also shown to be essential for its LD-localization ²⁸.

The existence of the N-terminal hydrophobic helices and the amphipathic patches in the HSD17B13 dimer prompted us to model the interactions of the apo HSD17B13 dimer with lipid membranes (Fig. 1). In this model, the shape of HSD17B13 dimer resembles a sledge, with the N-terminal helices and the amphipathic patches running parallel to the surface of the lipid membrane. The N-terminal helices are at the lowest position of the HSD17B13 dimer facing down towards the lipid membrane, the helix-turn-helix motif at the second layer, and above them the catalytic cores of the HSD17B13 dimer. Based on this model, at the lowest position the N-terminal hydrophobic helices will be immersed completely in the non-polar interior of a single leaflet of the lipid membrane, much like runners on a sledge. At the second level, the two “faces” of the amphipathic helix enable the hydrophobic face to enter the interior of the membrane, while the positively charged face interacts favorably with the negatively charged head groups of the membrane. The rigidity of helix-breaking P28 maintains a sharp angle between the N-terminal membrane anchoring helix and the following peptides leading to the catalytic core, ensuring that they cannot form transmembrane helices that cross a lipid bilayer. Consistent with HSD17B13’s association with lipid droplets which are known to be formed by lipid monolayers surrounding cores of neutral lipids^16,19,20, the N-terminal helix lacks any positively charged residues which might otherwise mediate interactions with the lipid head groups of the second leaflet of a membrane bilayer. Based on our model, during the LD biogenesis, which was thought to be initiated by the accumulation of neutral lipids between the two leaflets of the endoplasmic reticulum (ER) ²⁰, HSD17B13 would preferentially partition onto the monolayer of the budding LDs.

While the HSD17B13 dimer has structurally conserved catalytic cores, its N-terminal LD-targeting helices are unique when compared to four representative HSD proteins targeting different cellular organelles: HSD17B4, a soluble, cytosolic enzyme ³³; HSD17B1, a protein at equilibrium between the cytosol and membrane ³⁴; HSD11B1, an HSD with an N-terminal transmembrane helix targeting the lumen of ER ³⁵; and HSD17B11 (Supplementary Fig. 1). When superimposed, the root-mean-square deviation (RMSD) was 1.2 Å for 258 Ca atoms of the core domains of dimers of HSD17B13 and HSD17B4 (PDB ID 1ZBQ). However, HSD17B4 lacks the N-terminal helices and the helix-turn-helix motifs which mediate HSD17B13’s LD association. The structure of HSD17B1 dimer (PDB ID 1QYW) could be superimposed to that of HSD17B13 (RMSD 1.7 Å for 291 Ca). But while HSD17B1 has a similar amphipathic helix-turn-helix motif, it lacks the N-terminal membrane anchoring helices. The HSD11B1 dimer (PDB ID 1XSE) has similar catalytic cores (RMSD 2.00 Å for 350 Ca) and helix-turn-helix motifs. It also has the N-terminal membrane anchoring helices, but at their N-termini there are basic residues (K4/K5), critical for interacting the ER bilayer ³⁶, which are absent in HSD17B13. Although the construct used in the published HSD17B11 structure (PDB ID 1YB1) lacked both the N-terminal helices and the helix-turn-helix motifs, the catalytic cores of HSD17B13 and HSD17B11 are similar (RMSD 0.6Å for 348 Ca).

Another key feature from the structure, is that the dimer interface of HSD17B13 is extensive (Supplementary Fig. 2). Analysis of the dimer interface using program PISA³⁷ showed that the HSD17B13 dimer interface buried 2,122 Å² surface area of each monomer. The main dimer interface involved residues 97 and 101, and peptides 128–157 and 175–207 in the catalytic core, burying 1,736 Å² surface area. Peptides 134–157 and 183–207 formed two a-helices. Y185 and Y189, two residues of the catalytic triad, located in the dimer interface, and third one of the triad, S172, was nearby. This suggests that the formation of the HSD17B13 dimer is important to the proper formation of the catalytic center of the protein. The C-terminal helix-turn-helix motif contributed to the rest of the dimer interactions.

Cofactor binding site and a blocked substrate tunnel in apo HSD17B13:

The cofactor of HSD17B13, NAD⁺, was observed in all the crystals and in fact key for crystallization. The nicotinamide nucleotide of NAD⁺ interacts with the putative catalytic triad of HSD17B13 (S172, Y185 and K189)¹¹ and a long loop in HSD17B13, loop P218-T239. The adenine nucleotide of NAD⁺ forms hydrogen bonds with side chain of D67 and D93, and main chain of C94. Residues D67, D93-C94 are near or in peptide 71–106, which interestingly is deleted in an alternative splicing isoform variant of HSD17B13, isoB. Given these missing residues, isoB variant is unlikely to have a properly formed cofactor binding site and an intact dimer interface. The isoB variant was localized in the ER but not at LD, probably due to protein misfolding and aggregation ¹¹.

The interactions of NAD⁺ with HSD17B13 provide a structural explanation for the cofactor specificity of HSD17B13 (Fig. 3A). The side chain of a negative charged residue, D67, interacts with the 2’ and 3’ hydroxyl of the adenine mononucleotide of the cofactor. The negative charge of D67 would be repulsive with the phosphate group if a NADP cofactor bound there instead. As a comparison, HSD17B1 is a NADP specific enzyme, and it has been crystallized with NADP present enabling a comparison between cofactor interactions. In HSD17B1, L36 is present in place of D67 and faces away from the hydroxyls³⁸. Next to L36 is R37, and the positive side chain of R37 forms ion pair interactions with the 2’-phosphate of NADP bound in HSD17B1. The equivalent of R37 in HSD17B13 is I68, a hydrophobic side chain which further disfavors NADP binding. The 2’-phosphate and 3’-hydroxyl groups of NADP in HSD17B1 were also recognized by S11, equivalent of which is G45 in HSD17B13.

A prominent feature of the apo HSD17B13 structure is that the putative substrate binding site is completely blocked by the C-terminal peptide beyond the helix-turn-helix motif (Fig. 3B). This peptide, N286-K300 (C terminus), interacted extensively with the long loop P218-T239. Loop P218-T239 is equivalent to the substrate-binding loops of other HSD family proteins, and they were mostly disordered in the apo structures of those HSDs ³⁹. The C-terminal peptide of HSD17B13 additionally interacted with other parts of the protein. These interactions included hydrophobic contacts and hydrogen bonds involving main chain or side chain atoms. When a steroid molecule was docked in the active site of the apo HSD17B13 by homology modeling (Fig. 3B), the molecule would clash with the C-terminal peptide and loop P218-239. In fact, repeated soaking with substrate b-estradiol failed to result in ligand binding. Both sequence- and conformation-wise, the substrate-binding loops and the C-terminal peptides are the most variable in HSD proteins (Supplementary Fig. 1, Supplementary Fig. 3). As comparison, in the apo structure of HSD11B1, there was an empty cavity at the substrate site for access ³⁵. Substrate binding in HSD17B1 only caused minor changes at its substrate site ³⁴. The flexibility of the substrate-binding loops of other HSDs allowed substrates to readily access to their active sites ³⁹. In many instances the substrates could be soaked into crystals of those protein.

The blockage of the substrate binding site by the C-terminal peptide has two implications to the function of HSD17B13. First, filling the pocket with the C-terminal peptide stabilizes HSD17B13. As the isoD (P274del) variant lacks the C-terminal peptide as well as having an incompletely formed membrane interacting amphipathic helix, both factors likely contribute to isoD instability. Second, access to the substrate site needs to compete with the C-terminal peptide, which may reduce the catalytic efficiency of HSD17B13 with any substrate. Indeed, after screening recombinant HSD17B13 against 265 unique putative substrates including bioactive lipids and steroids, very few substrates were identified as has been previously reported ¹⁰. When compared to HSD17B2 using b-estradiol, one of the most active substrates identified in the screen, the enzymatic activity of HSD17B13 was much lower than that of HSD17B2 (Supplementary Fig. 4). In human HSD17B2 is widely expressed in many tissues, including liver ⁴⁰. Human HSD17B13 was also reported to have retinol dehydrogenase activity ¹¹, but multiple enzymes also efficiently oxidize retinol in liver ^41,42. Leukotriene B4, another one of the most active substrates identified, is also reported to be a substrate of other dehydrogenases ⁴³. The low efficiency of HSD17b13 towards these substrates when other dehydrogenases are present in liver, and are more efficient at catalyzing these reactions, begs the question of the biological relevance of HSD17B13 enzymatic activities of the substrates noted above.

Complex structures of compound 1 with dog HSD17B13

The overall structure of the compound 1 complex with dog mutant HSD17B13 did not have large changes in the catalytic cores of the HSD17B13 dimer compared to the apo HSD17B13 structure (0.31 Å for 437 Ca atoms of the dimer) (Fig. 4A). Compound 1 bound to both subunits of the dog mutant HSD17B13, displacing the C-terminal peptides from the substrate sites of both subunits. Missing the anchoring to the substrate sites from the C-terminal peptides, the adjacent P274-N286 helices in the helix-turn-helix motifs became disordered. Compound 1 binding also rearranged the P218-T239 loops, which interacted with the C-terminal peptides in the apo structure. Within the loop P229-L237 peptides became disordered, and the side chain of F220 adopted a different rotamer (Fig. 4B). The N-terminal helices packed against part of the helix-turn-helix motifs in the apo structure, and they became more flexible in the complex structure. Interestingly in the compound 1-dog WT HSD17B13 crystal, 1 only bound one subunit of the HSD17B13 dimer, and the unoccupied subunit adopted the apo conformation. Because dog WT HSD binds 1 weakly (shown below), subtle differences in the crystals could affect ligand binding, even though it crystallized in the same crystal form with the same crystal contacts as the dog mutant HSD, which binds 1 strongly.

At the ligand binding site compound 1 interacted with the catalytic center of the protein. The hydroxyl of the phenol made hydrogen bonds with the active residues S172 and Y185 (Fig. 4B). It contacted the positively charged nicotinamide ring of the bound cofactor, NAD⁺, with the closest distance at 3.1 Å to the C4N atom. The fluorophenol of 1 bound to HSD17B13 was likely deprotonated, with the oxygen atom carrying negative charge. The pKa of the phenol could be lowered by fluorine and amide substituents. The amide and the hydroxyl off the cyclohexyl of 1 donated hydrogen bonds to the side chain of T226 and main chain of K227, respectively. The non-polar atoms of 1 made extensive hydrophobic contacts with 12 residues of HSD17B13, including a p-stacking interaction between the terminal F-phenyl group and the side chain of Y181. All of the residues interacting with 1 were from one subunit, except A205 (human sequence), which was from the second subunit of the HSD17B13 dimer. The terminal fluorophenyl group of 1 extended toward a solvent facing opening left by the displaced C-terminal peptide (Fig. 4A).

We compared the residues contacting 1 directly between the dog and human HSD17B13 sequences. They are all identical between dog and human HSD17B13, except 205, which is a Thr in dog and an Ala in human. However, there are several residues adjacent to 1 interacting residues that are different between human and dog HSD17B13, among them are 117/118, which are Gly/Glu in human and Val/Gly in dog. In the displaced C-terminal peptide 293 is a Val in human and an Ile in dog. The dog HSD17B13 mutant (V117G/G118E/T205/I293V) had these residues changed from dog residues to human ones. These mutations did not change the overall structure of dog HSD17B13. The RMSD was 0.25 Å when 445 Ca atoms were superimposed between compound 1 dog WT and dog mutant HSD17B13 complex structures. However, local changes brought by these mutations were obvious (Fig. 4C). T205A changed a larger polar residue to a smaller hydrophobic one, and the terminal F-phenyl group that contacted this residue adjusted position accordingly. The side chain of E177 (human) pointed toward the inhibitor binding site (but did not have polar interactions with inhibitor or the rest of the protein), and V178 (dog) pointed outward and pushed against the second subunit of the HSD17B13 dimer. The combined effects of G177E and V178G mutations made an outward shift of the 177–180 peptide away from the ligand, leaving more space at the active site for inhibitors and substrates. I293V was in the C-terminal peptide and disordered in the complex. As a result, 1 had full occupancies in both HSD17B13 subunits in the crystal of the mutant protein (Fig. 4D).

Complex structure of compound 2 with human HSD17B13

The crystal of the compound 2-human HSD17B13 contained four HSD17B13 dimer complexes in the asymmetric unit, with 2 occupying all the protein subunits. In some subunits, long rod-shaped electron densities were observed together with those of 2 at the binding sites (Fig. 5A). We built detergent C12E8 molecules into these densities, but these densities could be from the alky chain of the solubilized endogenous phospholipids as well. Additional detergent molecules were also seen around N-terminal helices of some HSD1713 subunits.

The overall structures of the catalytic cores of human HSD17B13 dimers upon 2 binding were essentially the same as those of the apo and the 1 bound dog HSD1713 (RMSD values ranged from 0.22–0.46 for 413–428 Ca atoms). Like 1, 2 binding also displaced the C-terminal peptides in HSD17B13. In detergent co-existing binding sites, these peptides, as extensions of the P274-N286 helices, shifted positions relative to those of the apo dog HSD17B13 to interact with bound C12E8 molecules. In binding sites absence of detergents, the C-terminal peptides were disordered. Loops P218-T239 were rearranged relative to the apo structure and within them peptides P229-L233 were disordered. The N-terminal membrane anchoring helices of four HSD17B13 dimers also adopted conformations different from those of the apo structure (Supplementary Fig. 5). Most of these helices were still parallel to the lipid monolayer surface when modelled on LD surface, however, some of these helices tilted and reached deeper (~ 26 Å) into the interior of the membrane, likely deep enough into the neutral lipids core of the LDs. Nevertheless, none of these helices could cross an ER membrane bilayer, the typical thickness of which is 37.5 ± 0.4 Å ⁴⁴.

Similar to 1, the negatively charged carboxylate of 2 formed hydrogen bonds with S172 and Y185 and contacted the positively charged nicotinamide ring of NAD⁺ at closest distances of 3.1 Å to the C4N and C5N atoms (Fig. 5B). The conserved interaction with NAD⁺, termed charge-transfer interaction, was well studied in NAD⁺-phenol model system ⁴⁵. One of the sulfonamide oxygen atoms formed a hydrogen bond with the main chain of F220. The terminal tosyl group extended toward the lipid membrane, interacting with the helix-turn-helix motif. The non-polar atoms of 2 made extensive contacts with hydrophobic side chains of 12 residues of HSD17B13, all from one subunit, as well as with the bound C12E8 molecules. There are distinct subsets of residues that contact only either 2 or 1. The C12E8 molecules themselves extended beyond the helix-turn-helix motif, likely entered the interior of the lipid membrane (if on LDs) or detergent micelles (in solubilized states).

There are also differences in the binding poses of 1 and 2. The carboxylate oxygen atom of 2 which formed hydrogen bonds with S172 and Y185 of HSD17B13 overlapped with the phenol oxygen atom of 1, but the benzene rings shifted positions relative to each other, leading the rest of the molecules interacting with different parts of the binding site (Fig. 5C). While the terminal fluorophenyl of 1 pointed toward an opening facing solvent, this opening was sealed by the extended P274-A291 helix in the 2 complex with detergent. The terminal tosyl group of 2 pointed toward an opening facing the membrane, through which the C12E8 molecule also exited. Nerveless, like 1 in the dog mutant protein, compound 2 is well ordered at the active site of human HSD17B13 (Fig. 5D).

Structure-activity relationship studies

To better understand structure-activity relationships of the ligand binding site of HSD17B13, we compared the specific enzymatic activities of purified dog, human, and mutated HSD17B13 against β-estradiol (Fig. 6A). At all titrated enzyme concentrations human wild-type HSD17B13 was significantly more active than the dog wild-type protein. Mutating residues 177 and 178 in dog protein to match human protein (G177E, V178G) significantly enhanced enzymatic activity. Further mutating residues 205 and 293 to match human sequence (T205A, I293V), matching what was used for crystallization, lead to an additional activity increase.

We next tested the cofactor selectivity in biochemical assay using NAD⁺ and NADP⁺ as cofactors, and using b-estradiol and LTB4 as substrates, respectively. While NAD⁺ could promote the oxidization of both substrates, NADP⁺ did not achieve significant activity compared to inhibited reactions (Fig. 6B). Given that cytosolic NAD⁺ concentration far exceeds those of NADH and NADP in liver ^46,47, it is expected that HSD17B13 is mainly an oxidative enzyme using NAD⁺ as cofactor.

To understand ligand binding, we tested several approaches such as surface plasmon resonance and isothermal calorimetry. Like many membrane/detergent-associated proteins, however, these methods were not amenable for HSD17B13. We next turned to thermal shift assays (TSA) for comparing ligand binding affinities ⁴⁸. TSA measures the changes in the thermal denaturation temperature of a protein under different conditions such as the presence compound, changes in buffer conditions, or sequence mutations. There are various readouts including intrinsic fluorescence or light scattering signals. We chose the static light scattering methods to detect the temperatures at which our proteins start to aggregate (Tagg), which are more suitable for membrane proteins that require detergent micelles for stability ⁴⁹.

Because we observed direct interactions between compound 1 and 2 with bound NAD in crystals, for the TSA experiments we first tested if binding of compound 1, the more potent of the two, to HSD17B13 is NAD⁺ dependent. While compound 1 resulted in a significant increase in thermal stability for human HSD17B13 in the presence of NAD⁺, there was no significant effect in the absence of cofactor (Fig. 6C). This was consistent for three additional close analogs of compound 1 (Supplementary Fig. 6). There was also a strong linear relationship between DTagg of compound 1 and its analogs, and the logarithm of their biochemical IC₅₀ values when the compound concentrations were kept the same and the same batch of protein was used (Supplementary Fig. 6), consistent with thermodynamic analysis that strong binders have larger stabilization effects on proteins ⁵⁰.

Finally, we measured the binding of compound 1 and 2 to dog WT, dog mutant HSD17B13, human HSD17B13, and HSD17B11 (Fig. 6D). Compound 1 did not stabilize dog WT HSD17B13 significantly, while compound 2 slightly destabilized the protein, probably due to the disruption of detergent micelles the protein was in. This might also explain why compound 1 only occupied one subunit of the HSD17B13 dimer in the dog WT HSD17B13 complex crystals despite 1 mM of compound 1 was used in co-crystallization, and compound 2 failed repeatedly in soaking into these crystals. Compounds 1 and 2 bound human HSD17B13 strongly. Dog mutant HSD17B13 bound compound 1 strongly, and similarly as the human HSD17B13 (dog mutant HSD1713/compound 1 Tagg: 47.8 ± 1.8°C; human HSD17B13/compound 1 Tagg: 49.3 ± 1.5°C; unpaired t test p = 0.1), indicating that the mutations introduced in dog HSD17B13 based on compound 1 interactions increased its binding affinity. Dog mutant HSD17B13 bound compound 2 strongly but did not achieve the same level of binding as the human HSD17B13 (dog mutant HSD17B13/compound 2 Tagg: 40.3 ± 0.6°C, human HSD17B13/compound 2 Tagg: 45.0 ± 1.5°C, p < 0.0001,). This mutant did not completely address residue differences at compound 2 site of human and dog HSD17B13. For example, compound 2 contacts I263 and F266 of human HSD17B13, which are both Tyr in dog HSD17B13. Finally, when tested with human HSD17B11, neither compounds 1 and 2 showed an effect on protein stability. Indeed, there are several residues that contacted both compound 1 and 2 in HSD17B13 that are different in HSD17B11, for example, C174, I179 and V219 in HSD17B13, which are Ala, Val and Asn, respectively.

Overall, our HSD17B13 structures suggest a novel mechanism for LD association where the protein embeds on top of the phospholipid membranes and the membrane anchoring helices do not cross lipid bilayers. The stretch of hydrophobic residues at the N-terminus and the key P28 are all essential for HSD17B13’s association with LD. Residues 1–28 are also necessary and sufficient for targeting HSD17B11 to LD²⁸. Residues 22–28 of HSD17B13 share weak sequence conservation with the PAT domain¹¹, a N-terminal domain of many well-known LD-associated proteins, such as Perilipin, ADRP, and TIP47^24,25. Inspection of PAT domain structures predicted by AlphaFold2³¹ showed conserved prolines at positions of turns in a layered amphipathic helix-turn-helix-turn-helix-turn-helix motif. The PAT domain of these proteins should structurally function similarly to the membrane associating domain of HSD17B13, except in HSD17B13 the peptides in the membrane associating domain are discontinuous in primary sequence (1–28 and 260–286). To the best of our knowledge, the only previously disclosed structure of a LD anchoring peptide was the NMR structure of the membrane anchoring motif of CGI-58 (PDB ID: 5A4H), which was largely unstructured ⁵¹.

Our structure-activity studies indicated that the differences in amino acids at the ligand binding sites (dog vs human HSD17B13, human HSD17B13 vs HSD17B11) affect the activities of both the inhibitors and the substrate tested. Interestingly, in mice HSD17B13 deficiency did not protect against liver injury ⁵², instead, it triggered hepatic steatosis and inflammation ⁵³. Among residues that were shown in our studies to be important for substrate and 1 activities, three residues are different between human HSD17B13 (E177, G178 and A205) and mouse HSD17B13 (G177, V178 and T205). For 2, in addition to residues 177 and 178, residues 266 and 281 are different between human and mouse HSD17B13 (Leu and Phe vs Ile and Ser, respectively). These amino acid differences in the substrate binding site of HSD17B13 between human and mouse may result in differences in substrate specificity and explain the difference in pathogenicity of this protein in human and rodents ^52,53. In fact, species differences in substrate selectivity and inhibitor binding have been observed for other HSD family members ^54–56.

The observation of a completely blocked substrate site in the apo HSD17B13 structure may explain the challenges we and other groups faced in identifying active substrates ¹⁰. While potent inhibitors could displace the C-terminal peptides, endogenous bioactive molecules may be less efficient in accessing the active site of HSD17B13. This together with data that reported protective variants have both diminished enzymatic activities and lower protein expression levels, raises the possibility that the pathogenicity of human HSD17B13 comes from the scaffolding function (for example, through ATGL interactions⁵⁷), rather from the enzymatic activity alone, of the protein on the LD surface. If the scaffolding function of HSD17B13 is important, then approaches of lowering HSD17B13 protein levels are needed, such as antisense or short interfering RNA ^58,59.

Altogether, our crystal structures provide important information for developing novel HSD17B13 binders. The observation of compound 1 and 2 occupying different parts of the HSD17B13 ligand binding site in our crystal structures further indicate that diverse chemical motifs can bind to HSD17B13, and multiple series of compounds could be developed as HSD17B13 inhibitors.

Hydroxysteroid dehydrogenases, HSD; HSD17B13, Hydroxysteroid 17-beta-dehydrogenase 13; non-alcoholic fatty liver disease, NAFLD; nonalcoholic steatohepatitis. NASH; octaethylene glycol monododecyl ether, C12E8; lipid droplet, LD; endoplasmic reticulum, ER; root-mean-square deviation, RMSD; thermal shift assay, TSA; Leukotriene B4, LTB4.

Data Availability

The atomic coordinates and structure factors (code 8G84, 8G89, 8G93 and 8G9V) have been deposited in the Protein Data Bank, Research Collaboratory for Structural Bioinformatics, Rutgers University, New Brunswick, NJ (http://www.rcsb.org/).

Supporting information

This article contains supporting information.

Acknowledgements.

We thank Suman Shanker for technical assistance in protein purification.

Nina, A. & Yvetta, K. in Dehydrogenases (ed Canuto Rosa Angela) Ch. 6 (IntechOpen, 2012).
Strauss, J. F. in Yen & Jaffe's Reproductive Endocrinology (Seventh Edition) (eds Jerome F. Strauss & Robert L. Barbieri) 66-92.e63 (W.B. Saunders, 2014).
Yang, S.-Y., He, X.-Y. & Miller, D. Hydroxysteroid (17β) dehydrogenase X in human health and disease. Molecular and Cellular Endocrinology 343, 1-6, doi:https://doi.org/10.1016/j.mce.2011.06.011 (2011).
McNamara, K. M. & Sasano, H. The role of 17βHSDs in breast tissue and breast cancers. Molecular and Cellular Endocrinology 489, 32-44, doi:https://doi.org/10.1016/j.mce.2018.10.019 (2019).
Mindnich, R., Möller, G. & Adamski, J. The role of 17 beta-hydroxysteroid dehydrogenases. Molecular and Cellular Endocrinology 218, 7-20, doi:https://doi.org/10.1016/j.mce.2003.12.006 (2004).
Penning, T. M. Molecular Endocrinology of Hydroxysteroid Dehydrogenases*. Endocrine Reviews 18, 281-305, doi:10.1210/edrv.18.3.0302 (1997).
Seckl, J. R. & Holmes, M. C. Mechanisms of Disease: glucocorticoids, their placental metabolism and fetal 'programming' of adult pathophysiology. Nature Clinical Practice Endocrinology & Metabolism 3, 479-488, doi:10.1038/ncpendmet0515 (2007).
Marchais-Oberwinkler, S. et al. 17β-Hydroxysteroid dehydrogenases (17β-HSDs) as therapeutic targets: protein structures, functions, and recent progress in inhibitor development. J Steroid Biochem Mol Biol 125, 66-82, doi:10.1016/j.jsbmb.2010.12.013 (2011).
Amangurbanova, M., Huang, D. Q. & Loomba, R. Review article: the role of HSD17B13 on global epidemiology, natural history, pathogenesis and treatment of NAFLD. Aliment Pharmacol Ther, doi:10.1111/apt.17292 (2022).
Abul-Husn, N. S. et al. A Protein-Truncating HSD17B13 Variant and Protection from Chronic Liver Disease. N Engl J Med 378, 1096-1106, doi:10.1056/NEJMoa1712191 (2018).
Ma, Y. et al. 17-Beta Hydroxysteroid Dehydrogenase 13 Is a Hepatic Retinol Dehydrogenase Associated With Histological Features of Nonalcoholic Fatty Liver Disease. Hepatology 69, 1504-1519, doi:10.1002/hep.30350 (2019).
Kozlitina, J., Stender, S., Hobbs, H. H. & Cohen, J. C. HSD17B13 and Chronic Liver Disease in Blacks and Hispanics. N Engl J Med 379, 1876-1877, doi:10.1056/NEJMc1804027 (2018).
Anstee, Q. M. et al. Genome-wide association study of non-alcoholic fatty liver and steatohepatitis in a histologically characterised cohort(☆). J Hepatol 73, 505-515, doi:10.1016/j.jhep.2020.04.003 (2020).
Loomba, R. & Sanyal, A. J. The global NAFLD epidemic. Nat Rev Gastroenterol Hepatol 10, 686-690, doi:10.1038/nrgastro.2013.171 (2013).
Su, W. et al. Comparative proteomic study reveals 17β-HSD13 as a pathogenic protein in nonalcoholic fatty liver disease. Proc Natl Acad Sci U S A 111, 11437-11442, doi:10.1073/pnas.1410741111 (2014).
Sunami, Y., Rebelo, A. & Kleeff, J. Lipid Droplet-Associated Factors, PNPLA3, TM6SF2, and HSD17B Proteins in Hepatopancreatobiliary Cancer. Cancers (Basel) 13, doi:10.3390/cancers13174391 (2021).
Horiguchi, Y., Araki, M. & Motojima, K. 17beta-Hydroxysteroid dehydrogenase type 13 is a liver-specific lipid droplet-associated protein. Biochem Biophys Res Commun 370, 235-238, doi:10.1016/j.bbrc.2008.03.063 (2008).
Liu, S. et al. Molecular cloning and expression analysis of a new gene for short-chain dehydrogenase/reductase 9. Acta Biochim Pol 54, 213-218 (2007).
Olzmann, J. A. & Carvalho, P. Dynamics and functions of lipid droplets. Nature Reviews Molecular Cell Biology 20, 137-155, doi:10.1038/s41580-018-0085-z (2019).
Wilfling, F., Haas, J. T., Walther, T. C. & Farese, R. V., Jr. Lipid droplet biogenesis. Curr Opin Cell Biol 29, 39-45, doi:10.1016/j.ceb.2014.03.008 (2014).
Bersuker, K. et al. A Proximity Labeling Strategy Provides Insights into the Composition and Dynamics of Lipid Droplet Proteomes. Dev Cell 44, 97-112.e117, doi:10.1016/j.devcel.2017.11.020 (2018).
Tauchi-Sato, K., Ozeki, S., Houjou, T., Taguchi, R. & Fujimoto, T. The Surface of Lipid Droplets Is a Phospholipid Monolayer with a Unique Fatty Acid Composition*. Journal of Biological Chemistry 277, 44507-44512, doi:https://doi.org/10.1074/jbc.M207712200 (2002).
Ma, Y. et al. Characterization of essential domains in HSD17B13 for cellular localization and enzymatic activity. J Lipid Res 61, 1400-1409, doi:10.1194/jlr.RA120000907 (2020).
Miura, S. et al. Functional Conservation for Lipid Storage Droplet Association among Perilipin, ADRP, and TIP47 (PAT)-related Proteins in Mammals, Drosophila, and Dictyostelium*. Journal of Biological Chemistry 277, 32253-32257, doi:https://doi.org/10.1074/jbc.M204410200 (2002).
Yamaguchi, T., Matsushita, S., Motojima, K., Hirose, F. & Osumi, T. MLDP, a Novel PAT Family Protein Localized to Lipid Droplets and Enriched in the Heart, Is Regulated by Peroxisome Proliferator-activated Receptor α*. Journal of Biological Chemistry 281, 14232-14240, doi:https://doi.org/10.1074/jbc.M601682200 (2006).
Fujimoto, Y. et al. Identification of major proteins in the lipid droplet-enriched fraction isolated from the human hepatocyte cell line HuH7. Biochim Biophys Acta 1644, 47-59, doi:10.1016/j.bbamcr.2003.10.018 (2004).
Walther, T. C. & Farese, R. V., Jr. Lipid droplets and cellular lipid metabolism. Annu Rev Biochem 81, 687-714, doi:10.1146/annurev-biochem-061009-102430 (2012).
Horiguchi, Y., Araki, M. & Motojima, K. Identification and characterization of the ER/lipid droplet-targeting sequence in 17beta-hydroxysteroid dehydrogenase type 11. Arch Biochem Biophys 479, 121-130, doi:10.1016/j.abb.2008.08.020 (2008).
KAWAI M., K. M., SAKURADA I., MORITA A. . THERAPEUTIC AMIDE DERIVATIVES. (2005).
KIRKPATRICK LYNN D, J. J. L. H. Pharmaceutical compositions and formulations including inhibitors of the pleckstrin homology domain and methods for using same. WIPO (PCT) patent (2009).
Jumper, J. et al. Highly accurate protein structure prediction with AlphaFold. Nature 596, 583-589, doi:10.1038/s41586-021-03819-2 (2021).
Breunig, S. L. & Tirrell, D. A. in Methods in Enzymology Vol. 656 (ed E. James Petersson) 545-571 (Academic Press, 2021).
Lee, S. A. et al. The Peroxisomal Localization of Hsd17b4 Is Regulated by Its Interaction with Phosphatidylserine. Mol Cells 44, 214-222, doi:10.14348/molcells.2021.2217 (2021).
Shi, R. & Lin, S.-X. Cofactor Hydrogen Bonding onto the Protein Main Chain Is Conserved in the Short Chain Dehydrogenase/Reductase Family and Contributes to Nicotinamide Orientation*. Journal of Biological Chemistry 279, 16778-16785, doi:https://doi.org/10.1074/jbc.M313156200 (2004).
Ogg, D. et al. The Crystal Structure of Guinea Pig 11β-Hydroxysteroid Dehydrogenase Type 1 Provides a Model for Enzyme-Lipid Bilayer Interactions*. Journal of Biological Chemistry 280, 3789-3794, doi:https://doi.org/10.1074/jbc.M412463200 (2005).
Mziaut, H., Korza, G., Hand, A. R., Gerard, C. & Ozols, J. Targeting Proteins to the Lumen of Endoplasmic Reticulum Using N-terminal Domains of 11b2-Hydroxysteroid Dehydrogenase and the 50-kDa Esterase *. Journal of Biological Chemistry 274, 14122-14129, doi:10.1074/jbc.274.20.14122 (1999).
Krissinel, E. & Henrick, K. Inference of macromolecular assemblies from crystalline state. J Mol Biol 372, 774-797, doi:10.1016/j.jmb.2007.05.022 (2007).
Huang, Y.-W. et al. Critical Residues for the Specificity of Cofactors and Substrates in Human Estrogenic 17β-Hydroxysteroid Dehydrogenase 1: Variants Designed from the Three-Dimensional Structure of the Enzyme. Molecular Endocrinology 15, 2010-2020, doi:10.1210/mend.15.11.0730 (2001).
Hwang, C. C. et al. The catalytic roles of P185 and T188 and substrate-binding loop flexibility in 3α-hydroxysteroid dehydrogenase/carbonyl reductase from Comamonas testosteroni. PLoS One 8, e63594, doi:10.1371/journal.pone.0063594 (2013).
MIETTINEN, M. M., MUSTONEN, M. V. J., POUTANEN, M. H., ISOMAA, V. V. & VIHKO, R. K. Human 17 β-hydroxysteroid dehydrogenase type 1 and type 2 isoenzymes have opposite activities in cultured cells and characteristic cell- and tissue-specific expression. Biochemical Journal 314, 839-845, doi:10.1042/bj3140839 (1996).
Belyaeva, O. V., Wu, L., Shmarakov, I., Nelson, P. S. & Kedishvili, N. Y. Retinol dehydrogenase 11 is essential for the maintenance of retinol homeostasis in liver and testis in mice. J Biol Chem 293, 6996-7007, doi:10.1074/jbc.RA117.001646 (2018).
Shirakami, Y., Lee, S.-A., Clugston, R. D. & Blaner, W. S. Hepatic metabolism of retinoids and disease associations. Biochimica et Biophysica Acta (BBA) - Molecular and Cell Biology of Lipids 1821, 124-136, doi:https://doi.org/10.1016/j.bbalip.2011.06.023 (2012).
Mesa, J. et al. Human prostaglandin reductase 1 (PGR1): Substrate specificity, inhibitor analysis and site-directed mutagenesis. Chem Biol Interact 234, 105-113, doi:10.1016/j.cbi.2015.01.021 (2015).
Mitra, K., Ubarretxena-Belandia, I., Taguchi, T., Warren, G. & Engelman, D. M. Modulation of the bilayer thickness of exocytic pathway membranes by membrane proteins rather than cholesterol. Proc Natl Acad Sci U S A 101, 4083-4088, doi:10.1073/pnas.0307332101 (2004).
Henneke, C. M. & Wedding, R. T. Charge transfer complexing of NAD with phenol and influence of phenol substituents on complex stability. Archives of Biochemistry and Biophysics 168, 436-442, doi:https://doi.org/10.1016/0003-9861(75)90272-6 (1975).
Masia, R. et al. Live cell imaging of cytosolic NADH/NAD+ ratio in hepatocytes and liver slices. American Journal of Physiology-Gastrointestinal and Liver Physiology 314, G97-G108, doi:10.1152/ajpgi.00093.2017 (2018).
Veech, R. L., Eggleston, L. V. & Krebs, H. A. The redox state of free nicotinamide-adenine dinucleotide phosphate in the cytoplasm of rat liver. Biochem J 115, 609-619, doi:10.1042/bj1150609a (1969).
Zhang, R. & Monsma, F. Fluorescence-based thermal shift assays. Curr Opin Drug Discov Devel 13, 389-402 (2010).
Senisterra, G. A. et al. Assessing the stability of membrane proteins to detect ligand binding using differential static light scattering. J Biomol Screen 15, 314-320, doi:10.1177/1087057109357117 (2010).
Layton, C. J. & Hellinga, H. W. Thermodynamic Analysis of Ligand-Induced Changes in Protein Thermal Unfolding Applied to High-Throughput Determination of Ligand Affinities with Extrinsic Fluorescent Dyes. Biochemistry 49, 10831-10841, doi:10.1021/bi101414z (2010).
Boeszoermenyi, A. et al. Structure of a CGI-58 Motif Provides the Molecular Basis of Lipid Droplet Anchoring*. Journal of Biological Chemistry 290, 26361-26372, doi:https://doi.org/10.1074/jbc.M115.682203 (2015).
Ma, Y. et al. 17-Beta Hydroxysteroid Dehydrogenase 13 Deficiency Does Not Protect Mice From Obesogenic Diet Injury. Hepatology 73, 1701-1716, doi:10.1002/hep.31517 (2021).
Adam, M. et al. Hydroxysteroid (17β) dehydrogenase 13 deficiency triggers hepatic steatosis and inflammation in mice. Faseb j 32, 3434-3447, doi:10.1096/fj.201700914R (2018).
Arampatzis, S. et al. Comparative enzymology of 11β-hydroxysteroid dehydrogenase type 1 from six species. Journal of Molecular Endocrinology 35, 89-101, doi:10.1677/jme.1.01736 (2005).
Inderbinen, S. G., Zogg, M., Kley, M., Smieško, M. & Odermatt, A. Species-specific differences in the inhibition of 11β-hydroxysteroid dehydrogenase 2 by itraconazole and posaconazole. Toxicol Appl Pharmacol 412, 115387, doi:10.1016/j.taap.2020.115387 (2021).
Blanchard, P.-G. & Luu-The, V. Differential androgen and estrogen substrates specificity in the mouse and primates type 12 17β-hydroxysteroid dehydrogenase. Journal of Endocrinology 194, 449-455, doi:10.1677/joe-07-0144 (2007).
Su, W. et al. Phosphorylation of 17β-hydroxysteroid dehydrogenase 13 at serine 33 attenuates nonalcoholic fatty liver disease in mice. Nature Communications 13, 6577, doi:10.1038/s41467-022-34299-1 (2022).
Pelechano, V. & Steinmetz, L. M. Gene regulation by antisense transcription. Nature Reviews Genetics 14, 880-893, doi:10.1038/nrg3594 (2013).
Friedman, S. L., Neuschwander-Tetri, B. A., Rinella, M. & Sanyal, A. J. Mechanisms of NAFLD development and therapeutic strategies. Nature Medicine 24, 908-922, doi:10.1038/s41591-018-0104-9 (2018).
Vonrhein, C. et al. Data processing and analysis with the autoPROC toolbox. Acta Crystallogr D Biol Crystallogr 67, 293-302, doi:10.1107/S0907444911007773 (2011).
McCoy, A. J. et al. Phaser crystallographic software. J Appl Crystallogr 40, 658-674 (2007).
Blanc, E. et al. Refinement of severely incomplete structures with maximum likelihood in BUSTER-TNT. Acta Crystallogr D Biol Crystallogr 60, 2210-2221 (2004).
Emsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. Features and development of Coot. Acta Crystallogr D Biol Crystallogr 66, 486-501 (2010).
The PyMOL Molecular Graphics System. v. 2.3 (2010).
One-way ANOVA followed by Dunnett’s multiple comparisons test was performed using GraphPad Prism version 90.0 for Windows v. 9.0 (GraphPad Sofware, www.graphpad.com, Sandiego, California USA).

Constructs:

For protein and membrane purification, the following constructs were cloned into pFastBac unless otherwise indicated. For human HSD17B13 and HSD17B11, from N- to C-terminus: 10x His, a biotin acceptor peptide (GLNDIFEAQKIEWHE), a TEV protease site, and either human HSD17B13 (residues 2-300, NP_835236.2) or human HSD17B11 (residues 2-300, NP_057329.3). For dog HSD17B13 constructs, from N- to C-terminus: HSD17B13 (AA1-300, XP_038300006.1), a Gly-Ser-Gly linker, and a 11x His. For human HSD17B13 a similar version was made in pcDNA3.1 for mammalian overexpression. For substrate selectivity studies, the following combination of point mutations were made to dog HSD17B13: G177E/V178G or G177E/V178G/T205A/I293V. For human HSD17B13 crystallization, the following construct was made from N- to C- terminus: Human HSD17B13 (residues 2-300, NP_835236.2) with Q60K/I62R/R71H/E161K point mutations, a Gly-Ser-Gly linker, and 11x His. Proteins were expressed in Spodoptera frugiperda (Sf9, ATCC) using the Bac-to-Bac Baculovirus Expression System (Invitrogen) with infection at a cell density of 2 × 10⁶ cells ml⁻¹ for 72 hrs. For human HSD17B2, the construct was cloned instead into pcDNA3.1 and contained from N- to C-terminus: 6xHis, a TEV protease site, and human HSD17B2 (residues 2-387, NP_002144.1). The protein was expressed in Expi293 Cells (Life Technologies) grown in Expi293 Expression media to densities of 2.5 × 106 cells ml⁻¹. Cells were transiently transfected for 72 hrs with 1 ug plasmid per mL cell ratio in Opti-MEM reduced serum media (Life Technologies) using ExpiFectamine Transfection Kit (Life Technologies). Control and human HSD17B13 Expi293 membranes were similarly made using cells transfected with empty pcDNA3.1 plasmid pcDNA3.1 plasmid encoding HSD17B13.

Protein and Membrane Purification for Biochemical evaluation

For HSD17B2 and control membranes, transfected Expi293 cells were resuspended in 20 mM sodium phosphate pH 7.2, 20 mM NaCl, 250 mM sucrose, 10% glycerol, 1mM TCEP, 0.5 mM NAD⁺, cOmplete EDTA-free protease inhibitors (Roche), and Benzonase (MilliporeSigma). Cells were lysed using a microfluidizer, spun at 500xg for 10 min to remove unlysed cells, and membranes isolated by ultracentrifugation.

For purified human HSD17B13 and HSD17B11 protein used for biochemical evaluation, Sf9 cells were resuspended in 50 mM sodium phosphate pH 7, 500 mM NaCl, 10% glycerol, 1 mM TCEP, cOmplete EDTA-free protease inhibitors (Roche), and Benzonase (MilliporeSigma). Cells were lysed using a microfluidizer and pelleted by ultracentrifugation. Membranes were solubilized for 4-6hrs at 4˚C in 50 mM NaPhos pH 7.2, 300 mM NaCl, 50 mM Imidazole, 10% glycerol, 1 mM TCEP, 0.5 mM NAD⁺, and 0.5% DDM (Anatrace). The DDM-solubilized fraction was clarified by ultracentrifugation and incubated overnight at 4˚C with NiNTA agarose (Qiagen). Resin was washed through a series of steps with 50 mM sodium phosphate pH 7, 150 mM NaCl, 10% glycerol, 1 mM TCEP, 50 µM NAD⁺ and (1) 50 mM Imidazole, 0.5% DDM, (2) 50 mM Imidazole, 0.1% DDM, (3) 50 mM Imidazole, 0.05% DDM, and (4) 250 mM Imidazole, 0.02% DDM, and eluted with (6) 150 mM NaCl, 700 mM Imidazole, 0.02% DDM. Finally, the protein was further purified by S200 10_300 gel filtration chromatography in 25 mM sodium phosphate pH 7, 150 mM NaCl, 10% glycerol, 1 mM TCEP, 50 µM NAD⁺, and 0.02% DDM and concentrated in a 100 kDa MWCO centrifugal concentrator (Amicon)

For biochemical evaluation of dog HSD17B13 constructs, the purification was similar to above with a few changes. HEPES pH 7.4 buffer was used in place of sodium phosphate pH 7.0. Membranes were solubilized with 0.5% octaethylene glycol monododecyl ether (C12E8, Anatrace) instead of 0.5% DDM, and nickel resin was washed with decreasing [C12E8] to a final concentration of 0.005% C12E8. The final NiNTA fraction was run by S200 10_300 gel filtration chromatography in 50 M HEPES pH 7.4, 300 mM NaCl, 10% glycerol, 1 mM TCEP, 50 µM NAD⁺, and 0.005% C12E8.

Protein Purification for Crystallization

Similar to above, SF9 cells were collected and re-suspended in buffer containing 50 mM HEPES pH 7.4, 500 mM NaCl, 10% glycerol, 1 mM TCEP, EDTA-free cOmplete Protease Inhibitor cocktail (Roche), and Benzonase (MilliporeSigma). Cells were lysed using a micro-fluidizer and membrane was isolated by ultracentrifugation. The membrane pellet was resuspended in 50 mM HEPES pH 7.4, 300 mM NaCl, 50 mM Imidazole, 10% glycerol, 1 mM TCEP, 0.5mM NAD⁺ and solubilized by addition of 0.5% C12E8 (Anatrace) for 4 hours at 4 °C. The sample was clarified by ultracentrifugation, and the supernatant was batch bound to Ni-NTA agarose (Qiagen) overnight at 4 °C. The resin was washed in solubilization buffer containing 50 µM NAD⁺ with stepwise decrease to 0.005% C12E8. The eluted protein was further purified on by SEC in 50 mM HEPES pH 7.4, 300mM NaCl, 50 µM NAD⁺, 10% glycerol, 1 mM TCEP, 0.005% C12E8. HSD13B17 containing fractions were pooled and concentrated in a 100 kDa MWCO centrifugal concentrator (Amicon).

Inhibitors preparation and characterization

The preparation and characterization of compound 1 and its three analogs studied were reported before ²⁹. Compound 2 was repurified from internal collection and characterized using NMR and high-resolution mass spectrum (Supplementary Methods).

Biochemical assays

HSD17B13 enzyme inhibition potency of test compounds was determined using a purified protein biochemical enzyme activity assay, using NAD(P)H-Glo luciferase readout (Promega). HSD17B13 enzyme uses the oxidized form of nicotinamide adenine dinucleotide (NAD⁺) as a cofactor during metabolism of b-estradiol (substrate) to estradiol (product), while converting NAD⁺ to the reduced form (NADH). The Promega NAD(P)H-Glo™ assay is a homogeneous bioluminescent assay that generates a light signal from biochemical reactions that contain NADH (or nicotinamide adenine dinucleotide phosphate, NADPH). In the presence of NADH (or NADPH), the enzyme Reductase reduces a proluciferin reductase substrate to form luciferin. Luciferin then is quantified using Ultra-Glo™ Recombinant Luciferase (rLuciferase), and the light signal produced is proportional to the amount of NAD(P)H in the sample. Substrate mix composed of 12 µM final assay concentration (FAC) b-estradiol (Sigma, E8875) and 500 µM FAC NAD⁺ (Sigma, N8285) in assay buffer (25 mM Tris-HCl (Sigma T2444)and 0.02% Triton X-100, pH 7.6) was added (2 µL/well) to 384-well assay plates (Corning 3824) containing 80 nL of 50x FAC compound, serial diluted 1 in 3.162 in 100% DMSO for an 11 point concentration response curve (80 µM top concentration), with duplicate points at each concentration. The reaction was initiated by the addition of 2 µL/well purified HSD17B13 protein (30 nM FAC in assay buffer). Compound, substrate mix and HSD17B13 protein was incubated in the dark at room temperature for 2 hours before the addition of 3 µL/well NAD(P)H-Glo detection reagent, prepared from luciferase detection reagent and reductase/reductase substrate, as per manufacturer’s instructions (Promega, G9061). Detection reagent was incubated in the dark for 1 hour at room temperature before plates were read on the Envision plate reader (Perkin Elmer) using a luminescent protocol. Data expressed as relative luminescent units (RLUs) were then normalized to control wells using Activity Base (IDBS). Zero percent effect (ZPE) was defined as RLUs generated from uninhibited HSD17B13 protein (vehicle control). One hundred percent effect (HPE) was defined RLUs generated from wells containing 40 μM FAC of a Pfizer proprietary compound known to cause 100% inhibition of HSD17B13 protein. The concentration and % effect values for each compound were plotted by Activity Base using a four-parameter logistic dose response equation, and the concentration required for 50% inhibition (IC₅₀) was determined.

Dependency of co-factor (NAD⁺ and NADP⁺) was determined using a purified protein HSD17B13 biochemical enzyme activity assay, using NAD(P)H-Glo luciferase readout (Promega). Substrates b-estradiol (Sigma, E8875) and Leukotriene B4 (LTB4, Cayman), 15 mM FAC in assay buffer (25 mM Tris-HCl (Sigma T2444) and 0.02% Triton X-100, pH 7.6), and 500 µM FAC co-factor (NAD⁺or NADP, Sigma, N8285) in assay buffer were added (2.5 µL/well each) to 384-well assay plates (Corning 3824) containing 2.5 mL of either 30 mM FAC inhibitor in 1% DMSO/assay buffer or vehicle (1% DMSO/assay buffer), in triplicate. The reaction was initiated by the addition of 2.5 µL/well purified HSD17B13 protein (20 nM FAC in assay buffer). Compound, substrate, co-factor and HSD17B13 protein was incubated in the dark at room temperature for 2 hours before the addition of 10 µL/well NAD(P)H-Glo detection reagent, prepared from luciferase detection reagent and reductase/reductase substrate, as per manufacturer’s instructions (Promega, G9061). Detection reagent was incubated in the dark for 1 hour at room temperature before plates were read on the Envision plate reader (Perkin Elmer) using a luminescent protocol. Data expressed as relative luminescent units (RLUs) were exported from the plate reader and plotted in GraphPad Prism.

The activities of membranes containing overexpressed HSD17B13 and HSD17B2 were measured in the same assay format. Cofactor NAD⁺(Sigma, N8285), prepared in assay buffer (25 mM Tris-HCl (Sigma T2444) and 0.02% Triton X-100, pH 7.6) at 2 mM final assay concentration (FAC), was added (2.5 µL/well) to 384-well assay plates (Corning 3824) along with either 2.5 mL of 50 mM FAC b-estradiol (Sigma, E8875) in assay buffer/2.5% DMSO, or vehicle (2.5% DMSO/assay buffer), in duplicate. The reaction was initiated by the addition of 5 µL/well membrane titration (HSD17B13, HSD17B2, or pcDNA3.1 blank) in assay buffer. Substrate, cofactor and membrane was incubated in the dark at room temperature for 1 hour before the addition of 10 µL/well NAD(P)H-Glo detection reagent, prepared from luciferase detection reagent and reductase/reductase substrate, as per manufacturer’s instructions (Promega, G9061). Detection reagent was incubated in the dark for 1 hour at room temperature before plates were read on the Envision plate reader (Perkin Elmer) using a luminescent protocol. Data expressed as relative luminescent units (RLUs) were exported from the plate reader into excel, and background (non- b-estradiol wells) was subtracted from the wells containing b-estradiol, for a change in signal (delta, or Δ). Background subtracted data were plotted in GraphPad Prism, vs. membrane concentration (mg/well) on the x-axis.

The activities of dog WT, dog mutant, and dog G177E/V178G HSD17B13 proteins were tested in the same assay format as the human HSD17B13. Co-factor and substrate mix containing 12 µM final assay concentration (FAC) NAD⁺ (Sigma, N8285) and 12 µM FAC beta-estradiol (Sigma, E8875) in assay buffer (25 mM Tris-HCl (Sigma T2444) and 0.02% Triton X-100, pH 7.6) (2.5% DMSO final) was added (5 µL/well) to 384-well assay plates (Corning 3824). The reaction was initiated by the addition of 5 µL/well purified protein titration (wild-type dog, dog mutant and additional G177E/V178V dog mutant, and wild type human HSD17B13) in assay buffer. Substrate, co-factor and membrane was incubated in the dark at room temperature for 2 hours before the addition of 10 µL/well NAD(P)H-Glo detection reagent, prepared from luciferase detection reagent and reductase/reductase substrate, as per manufacturer’s instructions (Promega, G9061). Detection reagent was incubated in the dark for 1 hour at room temperature before plates were read on the Envision plate reader (Perkin Elmer) using a luminescent protocol. Data expressed as relative luminescent units (RLUs) were exported from the plate reader into excel, and plotted in GraphPad Prism, vs. enzyme concentration (nM).

Thermal Shift Assay

TSA studies were run on the UNit instrument (Unchained Labs, CA, USA). Experiments were run in either 25 mM sodium phosphate pH 7, 150 mM NaCl, 10% glycerol, 1 mM TCEP, 0.05% DDM, + different cofactors (NAD⁺, NADH, NADP, NADPH; MilliporeSigma) or 25 mM HEPES pH 7.2, 300 mM NaCl, 0.5 mM TCEP, 0.05% DDM, 10% glycerol, ± 500 µM NAD⁺, and ± 100 µM compound 1 (from 2 mM DMSO stock) or DMSO or 50 mM Hepes pH 7.5, 300 mM NaCl, 0.5 mM TCEP, 0.05% DDM, 500 µM NAD⁺, ± 50 µM compounds (from 2 mM stock) or DMSO. Protein aggregation was measured by static light scattering at 266 and 473 nm with a 1 °C/sec ramp rate from 20-80 ˚C. The data was processed using the UNcle software (Unchained Labs, CA, USA) and temperatures at which the protein started aggregation were recorded as Tagg. Reported Tagg and analysis was based on measurements at 266 nm unless otherwise indicated.

Protein crystallization

For the canine proteins, protein was incubated at 12 mg/ml with 1 mM NAD⁺ and 1 mM compound 1 with the addition of 0.125% β-octyl-glucoside. Sitting drop vapor diffusion crystallization was set up by mixing 300 nl protein complex with 300 nl of reservoir solution containing 30% PEG3350, 0.2 M ammonium chloride. Crystals grew at room temperature over 2 weeks.

Human protein was crystallized with 1 mM NAD⁺ and 1 mM compound 2 with the addition of 0.5% β-octyl-glucoside. Vapor diffusion was carried out in the same fashion as the dog HSD17B13 complex with reservoir containing 25% PEG3350, 0.1 M Bis-Tris pH 5.5, 0.2M Li₂SO₄. Crystals grew at room temperature over 2 weeks.

Crystallographic data collection, structure determinations and refinements

Crystal data sets were collected at beamline 17-ID at IMCA CAT, APS (Argonne National Laboratory, Chicago). The wavelength used was 1 Å for all crystals. Crystals were kept at 100 K during data collection using cryo stream. Data were processed using the program autoPROC⁶⁰. Data sets were scaled and merged using anisotropic resolution cut off, and ellipsoidal data completeness were reported (Table 1).

The initial dog WT HSD17B13/compound 1 complex structure was solved with molecular replacement method using the published HSD17B11 crystal structure as model (PDB ID 1YB1), using the program Phaser⁶¹. Subsequent dog HSD17B13 apo/complex structures were solved using rigid body refinements. The compound 2/human HSD17B13 complex structure was solved with molecular replacement method using refine dog HSD17B13 structure as the starting model. Structure refinements were carried out using the program Buster⁶² and manual model building using the program COOT⁶³. Final refinement statistics are listed in Table 1. For the preparation of structural figures in this manuscript the graphic program PyMol was used⁶⁴.

There is NO Competing Interest.

ManuscriptSupplementaryv2.docx
Crystal structures of 17-beta-hydroxysteroid dehydrogenase 13
D1000272406valreportfullP1.pdf
PDB structure validation report
D1000272331valreportfullP1.pdf
PDB structure validation report
D1000272375valreportfullP1.pdf
PDB structure validation report
D1000272341valreportfullP1.pdf
PDB structure validation report

Download PDF

Journal Publication

published 24 Aug, 2023

Read the published version in Nature Communications →

Version 1

posted

You are reading this latest preprint version

Crystal structures of 17-beta-hydroxysteroid dehydrogenase 13

Status:

Journal Publication

Version 1

Abstract

Figures

Introduction

Results

Discussion

Abbreviations

Declarations

References

Methods

Additional Declarations

Supplementary Files

Status:

Journal Publication

Version 1