Genomic discovery of an evolutionarily programmed modality for small-molecule targeting of an intractable protein surface

Significance This manuscript reports on a member of the FK506/rapamycin family, WDB002, and the realization that FKBP-mediated recognition is a genetically programmable modality that enables engagement of topologically flat targets. FKBP-mediated recognition is thus nature’s strategy for drugging the “undruggable.” The surface of FKBP engages three completely unrelated targets—calcineurin, MTOR, and CEP250—with high-target affinity and specificity, using different constellations of amino acid residues. Target specificity is determined solely by the “variable domain” of the bound small molecule alone, suggesting the modality might be generalizable to other undruggable targets through variable domain engineering. Finally, since WDB002 targets CEP250, it may be a promising starting point for developing a treatment for COVID-19.

Sequences positive for KCDA were used to design PCR primers in the TaqMan format, which was used to screen 1:10 dilutions of the individual input DNAs. Samples with TaqMan Ct values of 25 or less were considered positives and sequenced individually on MiSeq using 2x125, 2x250 or 2x300 chemistry and assembled with Ray as before or with SPAdes 44 . Confirmation that a strain contained a large PKS cluster next to the KCDA, as well as other homologues of rapamycin cluster genes, resulted in acquisition of the strain, growth and isolation of high molecular weight DNA by phenolchloroform extraction. 50-100X coverage was generated using Pacific Biosciences SMRT technology and assembled with HGAP (Chin).

Strain Engineering and Fermentation
See Table S4 and S5 for the lists of strains and plasmids used in this study, and for product titer information. Strain-level engineering (rpsL mutant) and cluster-level engineering (+LAL and ΔcypB) yielded an 8.25-fold increase in WDB002 titer. Spontaneous streptomycin resistant colonies were isolated by plating spores of S22 on ISP2 with streptomycin at 20 ug/ml and incubated for 5-7 days at 30°C.
Resistant colonies were restreaked to ISP2 with streptomycin 20 ug/ml and amplified by PCR using primers S22 rpsL Forward and S22 rpsL Reverse. Point mutations were confirmed by Sanger sequencing (Genewiz, South Plainfield, NJ) using the same primers. S49 was sporulated and used for generation of the knock out strains.
Fermentation was carried out by picking colonies and inoculating to 20 ml GYM plus apramycin 50 ug/ml in a 125-ml baffled flask and cultured at 30°C for 48 hours.
Cultures were homogenized with a glass tissue homogenizer and 10% (v/v) was transferred to a secondary seed culture of GYM and incubated for 24 hours at 30°C. A 5% (v/v) inoculum was added to 500 ml of 8430 production media [Per Liter: 10 g Proflo cottonseed meal (ADM, Chicago, IL), 20 g D-Mannitol, 1 g Yeast Extract (Difco) 100 mg KH2PO4, 50 mg MgSO4, 20 mg CaCl2-2H2O, 4 ml ProFlo oil (ADM), 2.0 ml R2 trace elements 45 , pH adjusted to 6.5 with NaOH] in 2.8L Fernbachs and grown at 30°C for 6 days. Samples were harvested and processed for chemical detection.
Gene deletion constructs were created utilizing a rpsL dominance based counterselection strategy 46 . Marker-less deletion cassettes were created by fusion of the upstream and downstream homology regions of the cytochrome P450 genes (cypA, cypB or cypA-cyp B) by the method of Blodgett et al. 47,48 . Deletion cassettes were cloned into pJVD52.1 to generate the knockout vectors pWDB057, pWDB058, and pWDB059. These constructs were introduced to S49 by intergenic conjugation from E. coli JV36. Isolation of knock outs was done by the method of Blodgett et al. 48 Confirmed knock out strains were sporulated and pWDB041 was introduced to generate the final strains for fermentation.
A TTA-less version of the LAL gene from S18 was synthesized by Genscript (Piscataway, NJ) with the ermE* promoter and fd terminator and cloned into pSET152 to generate pWDB041.
Conjugations were done using the method of Blodgett et al. 48 Briefly Streptomyces spores were heat treated at 50°C for 10 minutes prior to use. Overnight cultures of the donor were washed in an equal volume of Lennox broth and 2 ul added to 100 ul heat treated spores and spotted to R2 medium 44 without yeast extract or sucrose.
Plates were incubated at 30°C for 20 hours for the knock outs or 37°C for 20 hours for integration, and then overlaid with 1 ml water with 2.4 mg nalidixic acid and 1.5 mg apramycin and incubated at 30°C for 3-5 days. Exconjugants were picked and restreaked to ISP2 plates with 25 ug/ml nalidixic acid and 50 ug/ml apramycin. Single colonies were picked and grown in GYM with apramycin and saved in 15% glycerol at -80°C.
Initial analysis was performed by plating 200 ul of mycelium grown in GYM for 48 hours spread to confluence on 8430 agar plates and grown for 7 days at 30°C.
Extraction was done by methanol soak overnight and purified over a Strata C-18U SPE column (Phenomenex, Torrance, CA). Samples were dried under vacuum and resuspended in DMSO. Determination of production was via LC/MS.
The combined filtrates are dried down to crude extract. The samples were first fractioned by MPLC silica gel column chromatography (25~40 uM with hexane-EtOAc as mobile phase). WDB001, WDB002, and WDB003 were then purified from enriched fractions by reversed phase HPLC chromatography (XBridge C18 column, MeOH/H2O, 0.1 % formic acid). For this study, a total of ~50 g WDB002 was isolated, 40.3 g of which was obtained via a single 500 L fermenter run. The structures of compounds were determined by the analysis of their mass spectrometry and 1D, 2D NMR data (Supplementary Information).

Chemicals
We purchased rapamycin from LC Laboratories (cat# R-5000). When this synthesis was performed using the 7-membered oxepane form (WDB002b), the corresponding 7-membered form of the product will be isolated (WDB011b). While this material is stable in the solid state and some aprotic organic solvents, in aqueous buffer an isomerization reaction occurs to provide a 6-membered pyran isomer (WDB011). Spectral data for the 7-membered ring form can be found in Supplement Data.

Protein expression and purification
We expressed human FKBP12, CEP25011.4 and CEP25029.2 containing an N-terminal His-tags in E. coli BL21(DE3)pLysS cells using standard protocols. We harvested cells by centrifugation and resuspended the pellets with 500 mM NaCl and 3 mM TCEP in 1X PBS buffer (Lysis buffer). We lysed the cells by sonication, and clarified the lysate by centrifugation at 23,000g for 20 min. We added the clarified lysate to Ni-NTA beads (QIAGEN), equilibrated them with the lysis buffer, and eluted the protein in 1X PBS; 250 mM imidazole; 3 mM TCEP. We removed the N-terminal His-tags with TEV protease (Sigma), followed by size exclusion chromatography using Superdex 75 (GE Healthcare) in 12.5 mM HEPES pH 7.4; 75 mM NaCl. We also expressed a fragment of human CEP250, consisting of residues 2134-2231, with an N-terminal GST-His-tag; we harvested, resuspended and lysed this protein as described above. We first purified containing an N-terminal GST tag was expressed in E. coli BL21(DE3)pLysS cells using standard protocols. Lysed cells were purified by Glutathione affinity chromatography in PBS buffer using Glutathione Sepharose 4B beads (GE Healthcare). We removed the Nterminal GST-tag with TEV protease (Sigma), followed by purification on a Resource Q column (GE Healthcare) in HEPES-NaCl pH 7.5 buffer.

Identification of FKBP12-interacting proteins
We performed AP-MS/MS experiments using HEK293T cell lysates in a buffer containing 50 mM HEPES, pH 7.5; 120 mM NaCl; 0.5% octyl-beta-glucoside; 2 mM MgCl2, 2 mM CaCl2 and a protease inhibitor cocktail (Roche). We resuspended the cell pellets in lysis buffer, gently sonicated and centrifuged them; we used the supernatant from this for all subsequent experiments. We added N-terminal AVI-tagged and biotinylated FKBP12 to the cleared lysate to a final concentration of 4 µM and supplemented it either with DMSO or 10 µM compound and 60 µL of lysis bufferequilibrated, 50% slurry of Streptavidin agarose (Pierce). We allowed the reaction to proceed at 4 o C with gentle rocking for 60 min, after which time we washed the samples 3X with lysis buffer and 3X with lysis buffer without detergent. We eluted the remaining proteins 50 mM HEPES, pH 8.5; 7 M urea at room temperature for 2 x 15 min. Due to the strong nature of the biotin-streptavidin interaction, nearly all biotin-FKBP12 was retained on the resin. We digested the eluted samples in 4 M urea with Lys-C for 2 hrs at 37 o C, and then diluted them in 50 mM HEPES, pH 8.5; 0.9 M urea and added trypsin for digestion overnight at 37 o C. After digestion, we desalted the peptides using C18 spin columns and analyzed them using LC-MS/MS on a Velos-Pro OrbiTrap mass spectrometer operated in data-dependent Top20 mode with CID-based fragmentation.
We identified the peptides using Sequest with a target-decoy database 49 .

Identification of CEP250 binding domain
We initially homed in on the minimal CEP250 domain responsible for engagement with we generated successive truncations from both the N and C termini at 30 amino acid intervals. We cloned these deletion constructs into an in vitro transcription/translation vector (pT7CFE1-NHis) and then expressed them using HeLa cell extracts (Pierce #88892).
We screened each deletion construct for the ability to engage HA-tagged FKBP12 in the presence of WDB001. We incubated HA-FKBP12 with each His-tagged CEP250 fragments along with either DMSO or WDB001 at 4 o C for 1 hr. We enriched for the HA-FKBP12-associated proteins using anti-HA agarose and used western blots with anti-His antibodies to identify the interacting CEP250 fragments. We selected the protein spanning amino acids 1981-2281 as a minimized binding region since it was the smallest fragment that displayed WDB001-dependent FKBP12 binding. We considered any CEP250 fragments that we identified in both DMSO and WDB001 samples misfolded so we eliminated them.

Pull-down assay to assess binding of CEP25029.2 to WDB compounds
We incubated 20 µL of streptavidin magnetic beads (Pierce) with 0.5 µM biotinylated-FKBP12, 2.5 µM CEP25029.2 and 5 µM of compound or volume-matched DMSO in binding buffer (10 mM HEPES, pH 7.4; 150 mM NaCl; 1 mM 2-mercaptoethanol; 0.005% (v/v) Surfactant-P20; 1% BSA) and rotated the reaction overnight at 4 o C. We harvested the beads with a magnet and washed them four times with wash buffer (10mM HEPES, pH 7.4; 150 mM NaCl, 1 mM 2-mercaptoethanol; 0.005% (v/v) Surfactant-P20). We eluted the proteins from the beads in SDS loading buffer by heating for 5 min. We collected and analyzed the eluate by SDS-PAGE and visualized with Coomassie blue staining.

Ternary complex formation and crystallization
We incubated FKBP12 with three-molar excess of X1-encoded compounds at 4 o C overnight. We then added three-molar excess of this binary complex to CEP25029.2 or CEP25011.4 and incubated it at 4 o C overnight. We isolated ternary complexes from unbound proteins and compounds by size exclusion chromatography using Superdex

Data collection and structure determination
We collected the WDB002 ternary complex dataset at the 21ID-G (LS-CAT) beamline of the Advanced Photon Source and used HKL program suites to process the datasets. We obtained the initial molecular-replacement solution for the WDB002-ternary complex by PHASER in the CCP4 suite, using the coordinates of previously determined FKBP12 structure (PDB 1FKD) but omitting CEP250 as search models. The model was built through iterative cycles of manual model building of CEP250 in COOT 50 and structure refinement using REFMAC5 51 . The Ramachandran plots, calculated by MolProbity 52 , contain no residues in disallowed regions. We prepared all structure model figures with PyMol (The PyMOL Molecular Graphics System, Version 1.3, Schrödinger). For details on the data collection and structure refinement, see Table S3.

Surface Plasmon Resonance (SPR)
We analyzed small-molecule binding kinetics at 25 o C on a Biacore S200 SPR instrument (GE Healthcare) in running buffer (10 mM HEPES, pH 7.4; 150 mM NaCl; 0.05% (v/v) Surfactant-P20; 2% DMSO) that we prepared and filtered before use. To assess binding kinetics of FKBP12 binding ligands, we immobilized biotinlayted-FKBP12 on a CAP sensor chip to afford a highly stable and active surface suitable for kinetic measurements. We primed and charged the sensor chip with SA first. We passed 10 ug/ml of biotinlayted-FKBP12 over the sensor surface at a flow rate 2 µl/min for 60 s to stably immobilize ~600 RU of biotinlayted-FKBP12. We injected the compounds over a concentration range over target and control flow cells with multi-cycle runs at a flow rate of 50 µl/min for 100 seconds. After 300 seconds of dissociation, we regenerated the sensor surface with regeneration solution provided in the CAP chip Kit (GE Healthcare).
We performed ternary binding kinetic evaluations of HSBP1, CEP25011.4, WT CEP25029.2, and mutants of CEP25029.2 for each compound using a similar capture protocol except that a final of 100 RU of biotinlayted-FKBP12 was captured on the CAP sensor chip surface. After the surface was prepared, we used an ABA injection mode to inject compounds at a flow rate of 30 µl/min for 120 seconds to saturated FKBP12 on the chip surface, which allows us to isolate the binding of target proteins. To measure association, we injected the CEP250 or HSBP1 proteins over a concentration range in the presence of compound at a flow rate of 30 µl/min for 200 seconds. Dissociation was measured in the presence of compounds for 290 second. We obtained binding curves and kinetic data after subtracting the blank values, and we analyzed the data by fitting it to a 1:1 Langmuir-binding model provided by the Biacore S200 evaluation software.

Cell lines and reagents
The full length human FKBP12 ORF was cloned into the retroviral vector pMSCV-N-

Western blotting and immunoprecipitation
For Western blots, U2OS cells were grown as above with the addition of WDB002 (10 µM) for the times indicated. Cells lysates prepared using RIPA buffer were separated by SDS-PAGE and transferred to nitrocellulose before incubation with CEP250 and atubulin primary antibodies and detection using HRP-conjugated secondary antibodies and enhanced chemiluminescence. For immunoprecipitation, HeLa and U2OS stable cell lines expressing FLAG-tagged FKBP12 were grown as above and transfected with myc-CEP250-CTD constructs using HD Fugene. After 24 hours, cells were lysed using RIPA buffer and lysates incubated with FLAG-resin (M2 antibody; Sigma A2220).
Immunoprecipitates were analysed by Western blotting with FLAG (Sigma F3165) and myc (Cell Signalling 2276) antibodies.