If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
State Key Laboratory of Medicinal Chemical Biology, College of Pharmacy and Key Laboratory of Medical Data Analysis and Statistical Research of Tianjin, Nankai University, Tianjin, ChinaCollege of Chemistry, Tianjin Normal University, Tianjin, China
Fungal infection has long been a chronic and even life-threatening problem for humans. The demand for new antifungal drugs has increased dramatically as fungal infections have continued to increase, yet no new classes of drugs have been approved for nearly 15 years due to either high toxicity or development of drug resistance. Thus, validating new drug targets, especially fungus-specific targets, may facilitate future drug design. Here, we report the crystal structure of yeast Hos3 (ScHos3), a fungus-specific histone deacetylase (HDAC) that plays an important role in the life span of fungi. As acetylation modifications are important to many aspects of fungal infection, the species specificity of Hos3 makes it an ideal target for the development of new antifungal drugs. In this study, we show that ScHos3 forms a functional homodimer in solution, and key residues for dimerization crucial for its deacetylation activity were identified. We used molecular dynamics simulation and structural comparison with mammalian hHDAC6 to determine unique features of the ScHos3 catalytic core. In addition, a small-molecule inhibitor with a preference for ScHos3 was identified through structure-based virtual screening and in vitro enzymatic assays. The structural information and regulatory interferences of ScHos3 reported here provide new insights for the design of selective inhibitors that target fungal HDAC with high efficiency and low toxicity or that have the potential to overcome the prevailing problem of drug resistance in combination therapy with other drugs.
). In recent years, the increase in hematopoietic stem cell transplantation, the widespread use of immunotherapeutic drugs, and the use of indwelling medical devices have led to a whole new group of patients being at risk for fungal infection (
). As the number of fungi caused infections is still increasing, the demand for antifungal agents has increased dramatically.
Most antifungal drugs developed before the 1990s act by perturbing the formation of the cell membrane in fungi. These drugs can be classified into several major groups, including polyenes, second-generation azoles, allylamines, flucytosines, and echinocandins (
). They are effective and save lives. However, many patients with deep fungal infection, like that caused by invasive aspergillosis, still die. Furthermore, since the last echinocandin, anidulafungin, was introduced to the market in 2006, no new classes of antifungal drugs were approved, as most candidate drugs either are highly toxic (cannot be taken orally), have problems with drug resistance, or have drug interactions (
). Studies have suggested that inhibiting fungal HDACs may have beneficial and synergistic effects, like reducing the virulence and growth of Candida spp. as well as decreasing their resistance to antifungal drugs (
). Inhibition of the activity of Aspergillus fumigatus HDAC (RpdA) significantly delays its germination, growth, and conidiation. These findings point to HDAC inhibitors (HDACi) as a promising class of drugs for treating fungal infections (
). Moreover, developing HDACi targeting fungal-specific HDACs will lower the toxicity of these antifungal drugs.
Like mammalian HDACs, fungal HDACs are grouped into three classes: class I includes Rpd3, Hos1, Hos2, and Hos3; class II has just one member, Hda1; and the members of class III, including Sir2, Hst1, Hst2, Hst3, and Hst4, are NAD dependent (
). Of these HDACs, Hos3 exists only in fungus. It was first reported in yeast, and it plays an important role in regulating transcription silencing during the G1/S transition of the mitotic cell cycle (
) found that although Hos3 can catalyze deacetylation on all four core histones (H3, H4, H2A, and H2B), it preferentially deacetylates histone H4, especially on sites lysine 5 and 8 (H4K5 and H4K8). Ahn et al. (
) later reported that Hos3-mediated deacetylation of H2BK11ac is related to the phosphorylation of H2BS10 and induces activation of the yeast apoptosis pathway. In addition to histone substrates, Hos3 also has non-histone substrates. It functions as a spindle position checkpoint by latching onto spindle pole bodies to deacetylate potential substrates and inhibit mitotic exit (
). In addition, Hos3 functions in the form of a homodimer instead of forming complexes with other proteins like mammalian HDACs. These features make Hos3 a good target for the development of fungus-specific HDACi. However, a lack of structural information impedes the depiction of the molecular mechanism governing the specificity of Hos3 catalyzation and organization, which are important for rational drug design.
In this study, we examined the crystal structure of Saccharomyces cerevisiae Hos3 (ScHos3) and revealed the key residues of ScHos3 dimerization that are crucial for its deacetylation activity. Using molecular dynamics (MD) simulation and structural comparison with mammalian hHDAC6, we depicted specific features of the ScHos3 catalytic core. Based on these findings, we performed structure-based virtual screening to screen ScHos3-specific HDACi. Small-molecule compounds with good in vitro inhibition activity on fungal HDACs were found, and their selectivity against mammalian hHDAC6 was evaluated.
The overall structure of ScHos3
We initially attempted to crystallize full-length ScHos3 (a.a. 1–697), but no crystals were observed. We used limited proteolysis to promote crystallization (
). Elastase (1:500 w/w) was added to the protein solution, and after 30 min of digestion, a stable fragment appeared on examination by SDS-PAGE (Fig. S1). Subsequent N-terminal sequencing coupled with mass spectrometry was used to determine the N-terminal and C-terminal residues of the fragment, which were A34 and A510 (Fig. S2). By cloning the region of the gene that encodes A34 to A510, we were able to express and purify good-quality recombinant ScHos3 (a.a. 34–510) and successfully crystallize it.
The crystal structure of ScHos3 (a.a. 34–510) was determined by single wavelength anomalous dispersion at 2.4 Å resolution (PDB ID: 7WJL). All residues in the structure were well defined, except the regions encompassing residues S378-R381, Y463-D470, and D497-A510 were invisible (Fig. 1A). ScHos3 had a low sequence similarity to other mammalian HDACs. It shared ∼30% sequence identity with class I HDACs and ∼25% with class II HDACs (Fig. S3). However, the overall structure of ScHos3 was quite similar to that of previously determined class I and II HDACs, which includes a typical Rossmann fold with a central eight-stranded parallel β-sheet (β2–β9) surrounded by ten α-helices (α1–α10). Structural alignment with DALI showed that ScHos3 most resembled the CD2 domain of hHADC6, with a root-mean-square deviation (RMSD) of only 1.93 Å (https://www.ebi.ac.uk/msd-srv/ssm/). Besides the conserved globular rigid part, an additional β1-strand was antiparallel to the β-sheet and connected to α11 through a short loop. β1 and α11, together with α12 and the C-terminal helices (αC), formed the flexible part of the ScHos3 structure, which was different from other HDACs (Fig. 1, B and C).
The catalytic center of ScHos3 lay on the loop regions of the central αβ fold (Fig. 1A). The active site was composed of one tyrosine residue Y418 and two histidine residues H195 and H196 as well as a Zn2+ ion tetracoordinated by D233, H235, and D370 (Fig. 2A). Moreover, small side-chain hydrophobic residues, including G416, G417, and L234 were close to the Zn2+ ion binding area, and all aforementioned residues were well conserved in other HDACs (Figs. 2A and S3). Unlike other HDACs, no water molecules or K+ ions were identified in the active center of the ScHos3 structure (
) because of either poor density mapping or the wider aperture of the outer catalytic funnel of ScHos3 compared to other HDACs that allowed free movement of small molecules in and out (Fig. 2B; see below).
The catalytic center of ScHos3 was deep into a catalytic funnel on the surface. Compared to other HDACs, the catalytic funnel of ScHos3 was wider and more negatively charged (Fig. 2B). Two α-helices, α11 and α12, as well as two loop regions, loop A and loop B, formed the outer layer of the catalytic funnel (Fig. 1B). However, these regions differed quite a bit between the structures of hHDAC6 and hHDAC1 and the ScHos3 structure, as α11 and α12 were substituted by loop regions and loop B was replaced with a helix in both the hHDAC6 and hHDAC1 structures (Fig. 1, B and C). It is worth mentioning that loop A was very flexible in the ScHos3 structure, as part of the density map of this region was invisible in the structure, whereas a similar loop at this position in hHDAC6 and hHDAC1 was shorter and better defined. These structural differences are crucial clues to designing ScHos3-specific HDACi.
The homodimer surface of ScHos3
As mentioned, ScHos3 forms a homodimer with intrinsic HDAC activity (
). To verify the dimerization state of ScHos3 in solution, we performed size-exclusion chromatography coupled to multi-angle light scattering (SEC-MALS) experiments on both full-length and crystallized ScHos3 (a.a. 34–510). The SEC-MALS results showed that the full-length ScHos3 was not stable and exhibited several states of aggregation in solution, which is a possible reason why it is unable to be crystallized from the full-length protein. By contrast, the crystallized ScHos3 was very uniform in size, with a molecular weight of ∼110 kDa, a value twice the calculated mass of monomer ScHos3 of 54 kDa (Fig. S4A). Moreover, in vitro enzymatic activity assay showed that the crystallized ScHos3 exhibited comparable deacetylation activity to the full-length ScHos3 (Fig. S4B). Thus, the crystallized ScHos3 (a.a. 34–510) was used for the following experiments unless explicitly specified.
As the homodimer state of ScHos3 in solution was confirmed, we set out to investigate the dimerization surface by structural analysis and site-directed mutagenesis. Although there was only one monomer in the asymmetric unit of the ScHos3 structure, a homodimer could be observed by symmetric operation in the crystal lattice (Fig. 3A). Interactions between the two monomers occurred principally on the α11 helix and the subsequent β1 strand, which are unique to the ScHos3 structure. The dimerization interface buried a total surface area of ∼2330 Å2, which indicates a rather strong intermolecular interaction. Polar interactions were major forces at the dimerization surface. In particular, side-chain amino group of K298 interacted with side-chain carboxyl group of E295 of the other molecule, and these two pairs of residues on the α11 helix formed the central interface of dimerization. Intermolecular polar interactions also occurred between side-chain amino group of K294 on the α11 helix and main-chain carbonyl group of K251 as well as between side-chain amino group of H308 at the end of the β1 strand and main-chain carbonyl group of I282. Furthermore, S301 and C303 on the loop region between α11 and β1 interacted with each other by main-chain carbonyl and amide groups, respectively (Fig. 3B).
To test whether the aforementioned residues on the interaction interface are crucial for the dimerization state of ScHos3, we designed, expressed, and purified several mutants of ScHos3. Their apparent molecular weights were examined by gel-filtration chromatography. First, we mutated both central K298 and K294 residues to alanine residues (KKmut); this KKmut was eluted from a Superdex 200 Increase 10/300 column at 13.20 ml, which is similar to the WT (13.14 ml). A single mutation of H308A (Hmut) at the outermost interaction surface did not affect the elution position too much (Fig. S5A). However, when the H308A mutation was added to KKmut, this triple mutant (KKHmut) eluted from the same column with two peaks, an initial small peak at 13.14 ml and a later larger peak at 14.47 ml, which corresponded to the estimated molecular mass of a dimer and monomer, respectively (Fig. 3C). These results indicate that KKHmut primarily forms a monomer state in solution. Further SEC-MALS experimentation verified that the molecular mass of KKHmut was ∼58 kDa (Fig. S5B), which is half the size of the WT dimer of 110 kDa (Fig. S4A). Thus, the dimerization state of ScHos3 was broken by mutating interacting residues on the central and outermost positions of the dimerization surface at the same time, which indicates a rather intimate interaction of the homodimer. More interesting is that the monomer fraction of KKHmut showed no detectable deacetylase activity in the in vitro deacetylation assay compared to the WT ScHos3 dimer (Fig. 3D), which proves that the dimerization state is the functional state of ScHos3 in solution, and interrupting the formation of the homodimer will badly compromise the deacetylase activity of ScHos3.
To further study the unique structural properties of the catalytic funnel of ScHos3, we used MD simulation to reveal the dynamic conformations and capture the stable state of ScHos3. The CD2 domain of hHDAC6, which ScHos3 most resembled in the aforementioned DALI results, was used as a contrast of mammalian HDAC. Time evolutions of RMSD values showed that the structures of both ScHos3 and hHDAC6 were generally stable during the entire 200 ns of simulation, although a higher average RMSD was found in the ScHos3 system, which indicates local fluctuations of the ScHos3 structure (Fig. 4A). Per residue root-mean-square fluctuation (RMSF) analyses showed that most residues of ScHos3 had fluctuated little during the MD simulation, but a few regions were mobile and flexible, including residues 60 to 68 and 379 to 382 close to the catalytic funnel (Fig. 4, B and C). It is interesting that these two regions are located on α12 and loop A, respectively, which are unique sequential and structural elements of ScHos3 and the major components forming the outer layer of the catalytic funnel (Figs. 1B and S3). Structural superposition of the representative conformation after MD simulation and the initial conformation in the crystal structure showed that these two regions underwent apparent conformational changes and influenced the outer aperture of the catalytic funnel, which indicates their important role in regulating the enzymatic activity of ScHos3 (Fig. 4C). Furthermore, two distances depicting the outer aperture of the catalytic funnel around these two regions, denoted D1 and D2, were monitored during MD simulation (Fig. 4D). Time evolutions of the D1 and D2 distances of hHDAC6 were rather stable at 12 and 8 Å, respectively. However, both D1 and D2 of ScHos3 fluctuated more and ultimately tended to converge at ∼9 and 14 Å after 150 ns of MD simulation (Fig. 4, E and F). These data prove that the outer catalytic funnels of hHDAC6 and ScHos3 differ in shape, with the latter being more flexible and regulatable. Based on these findings, we set out to perform structure-based virtual screening of HDACi specific for ScHos3.
HDACi specific for ScHos3
Structure-based virtual screening for small-molecule inhibitors of ScHos3 was performed as described in the Experimental Procedures section. Briefly, 13 compounds with the highest binding scores (lowest binding free energies) were selected (Table 1 and Fig. S6). In vitro deacetylation assay showed different inhibitory activity of these compounds against ScHos3, as their IC50 numbers had a wide distribution. T2383 had the lowest IC50 of ∼0.7 μM (Fig. 5A and Table 1). Of note, T2383 showed a selective inhibitory effect between ScHos3 and hHDAC6, as the IC50 against hHDAC6 was about 4.0 μM, which is ∼6 times more than the 0.7 μM of ScHos3 (Fig. 5A). A similar comparison was made on the reported mammalian HDACi (R)-TSA. The results showed that the IC50 of TSA against ScHos3 and hHADC6 were 7.5 μM and 0.2 μM, respectively (Fig. 5A), which proves that TSA may not be a good inhibitor of fungal HDAC (
). To check whether the inhibitory effects of these small compounds are correlated with their binding affinities to ScHos3, we performed surface plasmon resonance (SPR) experiments and found that the dissociation constant (KD) of T2383 bound to ScHos3 was 1.5 μM, which was much lower than that of (R)-TSA (42.0 μM; Fig. 5B) and lower than those of the other four compounds selected with inferior but tolerable inhibitory effects (Fig. S7 and Table 1). These results indicate that T2383 is a promising inhibitor specific for ScHos3.
Table 1Inhibitory effects and binding abilities of selected compounds to ScHos3
). Like TSA and some other HDACi on the market (e.g., SAHA and Belinostat), T2383 is a hydroximic acid derivative, which means it shares an identical terminal hydroximic acid group to coordinate the Zn2+ ion in the active center of HDACs. However, the remaining structure of T2383 was different, as it was bigger and contained more aromatic groups (Fig. S8). Further molecular docking was performed to study the structural mechanism governing the preference of T2383 bound to ScHos3. T2383 and TSA were docked into representative structures of ScHos3 and hHDAC6 averaged from the last 20 ns of the MD simulation, respectively. As shown in Fig. S8, A and B, T2383 was docked into the catalytic centers of both ScHos3 and hHDAC6 and was close enough to coordinate the crucial Zn2+ ion. However, T2383 had more interactions with the surrounding residues of ScHos3, including polar interactions with the side chains of D148 and the catalytic residue Y418; π-π interactions between its phenyl group and the aromatic side chains of F205 and F285; and hydrophobic interactions between its 2-methyl indole group and the surrounding residues T67, V69, F205, M379, and Y418 (Fig. S8A). By contrast, although the phenyl ring of T2383 had similar π-π interactions with F620 and F680 of hHADC6, the hydrophobic indole ring seemed to be isolated from the catalytic center and exposed in the solvent, which suggests a loose binding mode (Fig. S8B). These results were consistent with the binding free energies of T2383 to ScHos3 and hHADC6, which are −9.69 and −8.04 kcal/mol, respectively. Similar comparisons of the less preferential binding of (R)-TSA to ScHos3 are shown in Fig. S8C. Polar interactions only occurred with the catalytic residue Y418, and although, like in T2383, the terminal group of TSA bent and inserted into a hydrophobic pocket formed by T67, V69, and M379, the dimethylaniline group of TSA did not seem to be as suitable for a pocket as the indole group of T2383. By contrast, (R)-TSA bound to hHADC6 in a more comfortable way (the docked conformation resembled that in the reported crystal structure of 5EDU); its hydroximic acid group formed more polar interactions with the catalytic residues Y782, H610, and H611; and the methyl group retained hydrophobic interactions with F620 and F680 (Fig. S8D). The binding free energies of TSA to ScHos3 and hHACD6 are −7.01 and −8.99 kcal/mol, respectively, which are consistent with the structural analyses.
As discussed previously, the α12 (a.a. 60–68) and loop A (a.a. 379–382) regions exhibited distinct sequential and structural features of the catalytic funnel of ScHos3, and two distances, D1 and D2, were defined to describe the flexibility of the outer catalytic funnel (Fig. 4). From the above docking results, we can see that the relatively larger D1 and smaller D2 of ScHos3 may promote the bending of the inhibitors and allow the free-floating terminus of the inhibitors to be stabilized in a hydrophobic pocket formed by residues mainly on α12 (T67, V69) and loop A (M379; Fig. S8, A and C). This bending may be either helpful (in the case of T2383) or harmful (TSA) to their inhibitory effects. To verify whether the hypothesized sequence differences between ScHos3 and human HDAC are indeed responsible for the preferential inhibition of T2383 of ScHos3, two ScHos3 mutants with deletion of α12 (Δ62–67) or loop A (Δ379–382) were expressed, purified, and examined by in vitro deacetylase activity assays. Results showed the deletion of α12 (Δ62–67) badly compromised the expression level and deacetylase activity of ScHos3. By contrast, although the deletion of loop A (Δ379–382) retained the good properties of protein, T2383 had evident reduced inhibitory effect on the Δ379–382 mutant, with an IC50 of ∼80 μM, compared to the 0.7 μM for WT ScHos3 (Fig. S9). These sequential and structural features of ScHos3 should be considered in the future when designing more suitable inhibitors specific for fungal HDACs.
In this study, we elucidated the crystal structure of ScHos3, a fungus-specific HDAC that shows intrinsic HDAC activity in its dimerization form. Through MD simulation and structural comparison with its mammalian homolog hHADC6, we revealed the unique structural features of the catalytic funnel of ScHos3. The small-molecule inhibitor T2383 with a preference for ScHos3 was further identified through structure-based virtual screening and in vitro enzymatic assay. T2383, an HDACi with broad-spectrum inhibitory effects on mammalian HDACs, has been approved by the FDA to treat multiple myeloma. In our study, T2383 had a greater inhibitory effect on ScHos3 than it did on hHADC6 (Fig. 5A). Whether T2383 inhibits Hos3 from other fungal species, especially those with high pathogenic properties, like Candida and Aspergillus spp., is worthy of in-depth study. The results of sequence alignment showed that all representative fungal Hos3 (including Hos3 from yeast and other infectious species) shared common features of α12 and loop A, which are not conserved in human HDACs (Fig. S3). This indicates that T2383 can potentially inhibit other fungal HDACs that are more disease relevant. In addition, preliminary in vitro deacetylase activity assay showed that T2383 had a comparable inhibitory effect on C. spp. CaHos3, with an IC50 of ∼2.8 μM compared to 0.7 μM for ScHos3 (Fig. S10). As acetylation modifications play important roles in regulating the stress response, antifungal tolerance, and virulence of C. spp., CaHos3 may be a good target for the development of antifungal drugs to fight against the drug resistance and drug toxicity that are common when treating Candida infection. However, much work will be required to optimize T2383 from an in vitro inhibitory study to a finished antifungal drug. One problem is that it is usually difficult for drugs that act on cellular targets—like CaHos2 inhibitors—to penetrate the fungal membrane (
). Future reports on the co-crystal structure of CaHos3 in complex with T2383 or its analogs, followed by iterative structure–activity relationship analyses, will indeed facilitate progress.
The aforementioned studies of Hos3 inhibitors focus on the interference of its catalytic center. Intervention of the dimerization surface may also effectively hinder its activity, as the homodimer state is the functional quaternary structure of ScHos3 in solution (Fig. 3), and CaHos3 is mostly in the homodimer state during the purification as well. However, the dimerization surface of ScHos3 is large and cannot be easily broken by a single mutation of the interacting residue. Thus, peptide mimics like in the protein–protein interaction library may be a good choice for initial screening for interrupters of dimerization.
In summary, Hos3 is an ideal target for the development of antifungal drugs with low toxicity, as it exists only in fungi and plays an important role in their life span. The structural information about ScHos3 reported here provides new insight into the design of selective inhibitors that target fungal HDAC. It may facilitate the development of new antifungal drugs with high efficiency and low toxicity or even overcome the prevailing problem of drug resistance in combination therapy with other drugs.
Protein expression and purification
The full-length ScHos3 (a.a. 1–697) gene was amplified by PCR from the yeast genomic DNA library. Genes encoding ScHos3 and its truncation (a.a. 34–510) or mutants (H308A, K294AK298A, K294AK298AH308A) were subcloned into an engineered pET28a-smt3 vector with an N-terminal His6-SUMO tag. The fusion protein was expressed in the BL21(DE3) codon plus RIL strain of Escherichia coli. Cells were induced by 0.2 mM isopropyl β-D-1-thiogalactopyranoside when the cell density reached A600 ∼0.8, and cell cultures were grown for an additional 20 h at 16 °C. Cells were harvested by centrifugation and resuspended in lysis buffer (20 mM Tris 8.0, 500 mM NaCl, 10 mM imidazole, 1 mM β-Me, 5 mM PMSF), then lysed by sonication. The cell lysate was clarified by centrifugation at 18,000 rpm for 50 min at 4 °C, and the supernatant was loaded onto a Ni-excel affinity column (GE healthcare). After being washed with ten column volumes of washing buffer (20 mM Tris 8.0, 500 mM NaCl, 20 mM imidazole, 1 mM β-Me), the fusion protein was eluted from the affinity column with elution buffer (20 mM Tris 8.0, 500 mM NaCl, 250 mM imidazole, 1 mM β-Me) and digested by sumo protease for 2 h at 4 °C. The sample was dialyzed to buffer A (20 mM Tris 8.0, 200 mM NaCl, 1 mM DTT) and loaded onto a HiTrap SP HP column (GE Healthcare). Untagged proteins were eluted with buffer B (20 mM Tris 8.0, 1 M NaCl, 1 mM DTT) and further purified through a HiLoad 10/300 Superdex 200 column (GE Healthcare) in buffer A. The Δ62–67 and Δ379–382 mutants of ScHos3 (a.a. 34–510) were expressed and purified as above except that the SP column purification was saved as the expression levels were low. High purified proteins were collected and concentrated for crystallization and enzymatic assays.
The gene encoding the CD2 domain of human HDAC6 (a.a. 495–850) was cloned into a pMAT9s vector in frame with an N-terminal maltose binding protein (MBP) tag. The fusion protein was expressed in the BL21(DE3) strain of E. coli, and cells were harvested and lysed as described above except that the lysis buffer used for the MBP affinity column was 20 mM Tris 8.0, 500 mM NaCl, and 5 mM PMSF. After being washed with ten column volumes of washing buffer (20 mM Tris 8.0, 500 mM NaCl), the fusion protein was eluted from the affinity column with elution buffer (20 mM Tris 8.0, 500 mM NaCl, 10 mM maltose) and concentrated for enzymatic assay.
The full-length CaHos3 (a.a. 1–713) gene was cloned, expressed, and purified like the full-length ScHos3 except that a HiTrap Heparin HP column (GE Healthcare) was used instead of the SP column for ion-exchange purification. There were three peaks on the last gel-filtration chromatography. The main peak, representing a homodimer state, was collected and concentrated for enzymatic assay.
Crystallization and diffraction
All crystallization experiments were conducted at 20 °C with the sitting drop vapor diffusion method. Briefly, 1 μl ScHos3 (a.a. 34–510, ∼8 mg/ml) was mixed with 1 μl reservoir solution in 48-well Cryschem plates and equilibrated against 100 μl reservoir solution. The optimized crystallization condition was 0.05 M Hepes 7.0, 0.1 M ammonium acetate, 0.02 M magnesium chloride, and 8% PEG 8000. ScHos3 crystals reached sizes suitable for X-ray diffraction after 2 to 3 days. For X-ray diffraction, the crystals were rapidly frozen in liquid nitrogen with a cryoprotectant prepared by adding 20% (v/v) glycerol to the reservoir solution. For phasing, heavy atom iodide was introduced into the crystals by soaking the crystals for 0.5 to 24 h in cryoprotectant with different concentrations of sodium iodide. All diffraction data were collected at Shanghai Synchrotron Radiation Facility Beamline BL17U using a Quantum 315r CCD detector (ADSC) at a wavelength of 0.9789 Å for native crystals and 0.9792 Å for heavy atom–soaked crystals. The diffraction images were processed with HKL2000 (
), and subsequent computational refinements with local noncrystallographic symmetry restraints were performed with PHENIX. Detailed statistics on data collection and refinement are summarized in Table S1. All figures of protein structures were prepared in PyMOL (http://pymol.sourceforge.net/).
Enzymatic activity assay
Deacetylation assay was performed in triplicate at 37 °C in a flat-bottom 96-well black plate (Costa). Briefly, enzymes and different concentrations of inhibitors were mixed in HDAC assay buffer (50 mM Tris 8.0, 137 mM NaCl, 2.7 mM KCl, 1.0 mM MgCl2). To initiate the reaction, we added 80 μM fluorogenic substrate H4K5(Ac) peptide (GRGK[Ac]-AMC) to the aforementioned solution in a 100 μl reaction system. After incubating for 30 min, the reaction was stopped with the addition of 10 μl developer solution (1 μM trypsin and 10 μM TSA in HDAC assay buffer). The fluorescence intensity was measured with a Tecan Spark reader (λex = 360 nm, λem = 450 nm). To calculate IC50 values, we analyzed the data using logistic regression. The fluorescence intensity of the negative control (which contained 2% DMSO instead of inhibitors) was set as 100% enzymatic activity.
Size-exclusion chromatography coupled to multi-angle light scattering
SEC-MALS was used to determine the molecular weight of WT and mutant ScHos3 (a.a. 34–510) in solution. Experiments were performed at 4 °C in a buffer containing 20 mM Tris 8.0 and 200 mM NaCl. Purified protein samples (1 mg/ml, 500 μl) were injected into a Superdex 200 increase 10/300 column (GE Healthcare) and flowed through the column at a rate of 0.4 ml/min. Light scattering at 663 nm, ultraviolet absorbance at 280 nm, and the refractive index were monitored during the procedure. BSA was used to standardize the system. ASTRA 6.1 was used to record and process data.
Surface plasmon resonance
SPR experiments were performed on a BIAcore T200 SPR system (BIAcore, Cytiva) at 25 °C in Hepes buffer (20 mM Hepes 7.4, 200 mM NaCl, 5% DMSO). Protein was immobilized on CM5 chips by amine coupling, and the diluted inhibitors (0–200 μM) were flowed through the chips at a rate of 30 μl/min with solvent correction performed at the same time. Background binding to blank immobilized flow cells was subtracted, and KD values were calculated with the 1:1 binding kinetics model or steady state affinity model built into BIAcore T200 Evaluation Software (version 3.2).
Two systems, hereafter designated ScHos3 and hHDAC6, were built to perform MD simulation. The initial structure of the two systems was derived from the crystal structure of ScHos3 determined in this work and the CD2 domain of the crystal structure of hHDAC6 (deduced from PDB ID: 5EDU by the removal of the MBP tag and TSA) (
). The distance between the edges of the box and the closest atoms of the complex was 12 Å. Na+ was added as counterions to neutralize each system.
For each solvated system, 5000-step energy minimization for the whole residues was performed, followed by a combined equilibration process with a 500 ps constant volume ensemble to heat the system from 0 to 300 K and a 500 ps constant pressure ensemble at a constant pressure of 1 bar. During equilibration, a force constant of 10 kcal mol−1 Å−2 as a harmonic constraint was applied. Then 200 ns MD simulation of each system was performed with the AMBER18 software package in constant pressure ensembles at 300 K with the constraint released (
). The cut-off value of the van der Waals interactions was set to 10 Å.
The cpptraj module in AmberTools18 was used to calculate the conformational descriptors along each MD simulation, including RMSDs, RMSFs, and the defined distances D1 and D2. Distance D1 was measured between the Cα atoms of residues T67 and D148 of ScHos3 and those of H500 and S568 of hHDAC6, respectively. Distance D2 was measured between the Cα atoms of V69 and M379 of ScHos3 and those of P501 and L749 of hHDAC6, respectively. The illustrated representative structure for each system was derived from the average structure during the last 20 ns of MD simulation.
Virtual screening and molecular docking
Receptor-based virtual screening was performed with the AutoDock4.2 program to screen the lead compounds against ScHos3 (
). The ligand for molecular docking was derived from an in-house virtual database that included approximately 2000 molecules and was prepared by retaining the molecules with hydroximic acid groups in the ZINC library (
). The receptor was the representative structure of ScHos3 obtained from the MD simulation. The energy grid was generated to cover the binding pocket of ScHos3 with a size of 18 × 18 × 18 Å3. The number of docking runs for each compound was set to 500. From both docking scoring and visual inspection, 13 compounds were selected and purchased from TopScience and examined by in vitro enzymatic activity assay.
In addition, molecular docking was also performed to investigate the selectivity of small-molecule inhibitors against ScHos3 and hHDAC6. Each receptor structure was derived from its representative structure during MD simulation. The parameters in the calculations for both systems were the same as those used in the virtual screening.
Atomic coordinates and associated structure factors have been deposited in the Protein Data Bank (PDB ID code: 7WJL).
The authors declare that they have no conflict of interest with the contents of this article.
We thank Shanghai Synchrotron Radiation Facility beamline scientists for their help with data collection, Dr Wensi Yang and Dr Lin Zhang of the Institute of Biophysics for their participation and discussions of the work, and Ms Weiqun Shen of Peking University for her help with MS analyses and N-terminal amino acid sequencing.
N. Y. conceptualization; N. Y. and J. S. methodology; N. Y. supervision; N. Y. funding acquisition; N. Y. writing-reviewing and editing; N. P. and S. C. data curation; N. P., S. C., and J. S. validation; N. P. and J. S. writing-original draft.
Funding and additional information
This work was supported by the National Natural Science Foundation of China ( 31870737 and 32170549 ), the Chinese Ministry of Science and Technology ( 2019YFA0508902 and 2018YFA0107004 ), Tianjin Funds for Distinguished Young Scientists ( 17JCJQJC45900 ), and Fundamental Research Funds for the Central Universities of Nankai University.