IFITM proteins assist cellular uptake of diverse linked chemotypes

The search for cell permeable drugs has conventionally focused on low molecular weight, non-polar, and rigid chemical structures. However, emerging therapeutic strategies break traditional drug design rules by employing flexibly linked chemical entities composed of more than one ligand. Using complementary genome-scale chemical-genetic approaches we identified an endogenous chemical uptake pathway involving interferon induced transmembrane proteins (IFITMs) that modulates the cell permeability of a prototypical biopic inhibitor of MTOR (RapaLink-1, MW: 1784 g/mol). We devised additional linked inhibitors targeting BCR-ABL1 (DasatiLink-1, MW: 1518 g/mol) and EIF4A1 (BisRoc-1, MW: 1466 g/mol) whose uptake was facilitated by IFITMs. We also found that IFITMs moderately assisted some proteolysis targeting chimeras (PROTACs) and examined the physicochemical requirements for involvement of this uptake pathway.

To a mixture of compound 5 (100 mg, 0.0874 mmol) in dichloromethane (0.874 mL) was added trifluoroacetic acid (0.874 mL). The solution was stirred at room temperature for 1 h before concentrating in vacuo to afford a pale yellow semisolid that was used directly in the next step. To a mixture of the crude amine, trifluoroacetic acid salt and compound 4 (55 mg, 0.0961 mmol) in N,N-dimethylformamide (0.874 mL) was added N,N-diisopropylethylamine (46 μL, 0.262 mmol). The solution was cooled in an ice-water bath before the addition of 1-[bis(dimethylamino)methylene]-1H-1,2,3-triazolo[4,5-b]pyridinium 3-oxide hexafluorophosphate (37 mg, 0.0961 mmol) and stirred at room temperature overnight. The mixture was partitioned between ethyl acetate and water and the organic layer was washed with water (4×) and brine (2×), dried over sodium sulfate, filtered, and concentrated in vacuo. The crude was purified by flash chromatography over silica gel as follows: the crude was dry loaded into silica gel and an initial elution with ethyl acetate was made to remove an impurity. Upon switching to a methanol-dichloromethane solvent system, the desired product immediately eluted with additional impurities from the column. Fractions containing the desired product were then re-subjected to flash chromatography over silica gel eluting with a gradient from 0% methanoldichloromethane to 20% methanol-dichloromethane to afford compound 6 (99 mg, 0.0618 mmol, 71% yield over two steps) as a pale yellow solid.

DasatiLink-1.
To a mixture of compound 9 (47 mg, 0.0293 mmol) in dichloromethane (0.293 mL) was added trifluoroacetic acid (0.293 mL). The solution was stirred at room temperature for 6 h before concentrating in vacuo. The residue was partitioned between ethyl acetate and saturated sodium bicarbonate. The aqueous layer was extracted with ethyl acetate (3×) and the combined organics were washed with brine (2×), dried over sodium sulfate, filtered, and concentrated in vacuo. The crude was purified by flash chromatography over silica gel eluting with a gradient from 0% methanol-dichloromethane to 20% methanol-dichloromethane. Fractions containing the desired product were combined, concentrated in vacuo, and further purified by HPLC to afford DasatiLink-1 (32 mg, 0.0211 mmol, 72% yield) as a white solid. After allowing the reaction to cool to room temperature, the mixture was partitioned between dichloromethane and 5% citric acid in water. The aqueous layer was extracted with dichloromethane (4×) and the combined organics were washed with brine (2×), dried over sodium sulfate, filtered, and concentrated in vacuo. The crude was purified by flash chromatography over silica gel eluting with a gradient from 0% methanol-dichloromethane to 20% methanol-dichloromethane to afford rocagloic acid (66 mg, 0.138 mmol, 93% yield) as a white solid. and stirred at room temperature overnight. The mixture was partitioned between ethyl acetate and water and the organic layer was washed with water (4×) and brine (2×), dried over sodium sulfate, filtered, and concentrated in vacuo. The crude was purified by flash chromatography over silica gel eluting with a gradient from 0% methanol-dichloromethane to 10% methanol-dichloromethane to afford BisRoc-1 (33 mg, 0.0225 mmol, 72% yield) as a white solid. BisRoc-2. The same procedure as for BisRoc-1, using amino-PEG4-amine as starting material with scaled reagents, afforded BisRoc-2 (23 mg, 0.0199 mmol, 63% yield) as a white solid.

Genome-scale CRISPRi/a screening
Genome-scale CRISPRi/a screens were modeled after previous examples (28,29). Over the course of the screens, cells were grown in 500 mL Optimum Growth Flasks (Thomson) in 37 °C, 5% CO2 shaking culture [1300 revolutions per minute in a Multitron Incubator (Infors HT)]. K562 CRISPRi or CRISPRa cells were transduced with the five-sgRNA/gene human CRISPRi v2 (hCRISPRi-v2) or five-sgRNA/gene human CRISPRa v2 (hCRISPRa-v2) library respectively in the presence of polybrene (8 μg/mL) (29). Viral transduction was tittered to maximize singly transduced cells, targeting a multiplicity of infection (MOI) ≤ 1 (percentage of transduced cells 2 days after transduction = 20-40%). Transduced (sgRNA+) cells were selected with 2 doses of puromycin (1 μg/mL) up to 80-95% sgRNA+ in the population over the course of 5 days. Before the initiation of compound treatment, T0 samples were harvested with a minimum 1000-fold library coverage (approximately 100 million cells). The remaining cells were then divided into 5 treatment arms (DMSO, 1 nM sapanisertib, 1 nM rapamycin, 1 nM sapanisertib + rapamycin, and 1 nM RapaLink-1) with 2 biological replicates each. Cells were monitored for population doublings daily, and dilutions were made using complete media supplemented with the indicated compounds to maintain continuous selective pressure. Cells were cultured at a minimum 500fold library coverage (approximately 50 million cells) over 10 days, after which T10 samples were harvested with a minimum 1000-fold library coverage (approximately 100 million cells). Genomic DNA (gDNA) was extracted from T0 and T10 samples using NucleoSpin Blood XL (Macherey-Nagel). sgRNA protospacers were amplified directly from gDNA and processed for sequencing on an Illumina HiSeq 4000 as described previously (69).

Screen processing
Sequencing data from CRISPRi and CRISPRa screens were aligned to the hCRISPRi-v2 or hCRISPRa-v2 library respectively, counted, and quantified using the Python 2.7-based ScreenProcessing pipeline [https://github.com/mhorlbeck/ScreenProcessing (29)]. Phenotypes and Mann-Whitney P values were determined as described previously (28,29), although data detailed herein are not normalized to total population doublings. Additional analysis and plotting were performed in Prism 9 (GraphPad Software).

Large-scale chemogenomic profiling
High-throughput cell viability determination High-throughput drug screening and sensitivity modeling (curve fitting and IC50 estimation) was performed essentially as described previously (37). Cells were grown in RPMI or DMEM/F12 medium supplemented with 5% FBS and penicillin/streptomycin, and maintained at 37°C in a humidified atmosphere at 5% CO2. Cell lines were propagated in these two media in order to minimize the potential effect of varying the media on sensitivity to therapeutic compounds in our assay, and to facilitate high-throughput screening. To exclude cross-contaminated or synonymous lines, a panel of 92 SNPs was profiled for each cell line (Sequenom, San Diego, CA) and a pair-wise comparison score calculated. In addition, short tandem repeat (STR) analysis (AmpFlSTR Identifiler, Applied Biosystems, Carlsbad, CA) was performed and matched to an existing STR profile generated by the providing repository. Briefly, cells were seeded in 384 well plates at variable density to insure optimal proliferation during the assay. Drugs were added to the cells the day after seeding for adherent cell lines and the day of seeding for suspension cell lines. For tumor subtypes containing both adherent and suspension cells, all lines where drugged the same day (small cell lung cancer cell lines for example were all drugged the day after seeding). A series of nine doses was used with a 2-fold dilution factor for a total concentration range of 256 fold. Viability was determined using resazurin after 5 days of drug exposure, and data from treated wells were normalized to that of untreated wells.

Correlation analysis between drug sensitivity and basal gene expression
Dose-dependent growth inhibition of 935 cancer cell lines by RapaLink-1 and sapanisertib was determined as described above. Growth inhibition of 745 cell lines by rapamycin was obtained from the Genomics of Drug Sensitivity in Cancer database (GDSC2 release 8.3, accessed Oct. 4, 2020) (37,39). Gene expression data, reported as log2 transformed transcripts per million with a pseudocount of 1, was obtained from the DepMap (21Q4 release) (70). Spearman correlation coefficient between transcript level and area under the dose-response curve was calculated for each transcript using all cell lines present in both datasets (659 for RapaLink-1 and sapanisertib, 555 for rapamycin). Analysis and calculations (https://github.com/dwassarman/cellpanelr version 0.0.0.9001) were performed in R using tidyverse (71) and DepMap (70) packages and plotted in Prism 9 (GraphPad Software).

Cloning of single sgRNA expression vectors
sgRNA protospacers targeting FKBP12 (also known as FKBP1A), IFITM1, IFITM2, IFITM3, and a negative control (NegCtrl) sequence were individually cloned into pCRISPRia-v2 (Addgene 84832) as described previously (29). Protospacer sequences are listed in Cloning of triple sgRNA expression vectors sgRNA protospacers targeting IFITM1, IFITM2, and IFITM3 or three negative control protospacers were cloned into pCRISPRia-v2 (Addgene 84832) using a two-step procedure as described previously (40)   Individual evaluation of sgRNA phenotypes Cells were transduced as described herein. 5 days after transduction, cells were divided into 5 treatment conditions (DMSO, 1 nM sapanisertib, 1 nM rapamycin, 1 nM sapanisertib + rapamycin, and 1 nM RapaLink-1). Cells were monitored for the percentage of sgRNA+ (BFP+) populations daily by flow cytometry, and dilutions were made using complete media supplemented with the indicated compounds to maintain continuous selective pressure. Increased relative sgRNA+ percentage over time corresponded to a resistance chemical-genetic interaction while decreased relative sgRNA+ percentage corresponded to a sensitizing chemical-genetic interaction.

Immunoblotting
Cells (500,000 cells in 2 mL per well) were seeded into 6-well plates and incubated at 37 °C overnight. Following treatment with compounds at the concentrations and times indicated, cells were placed over ice, transferred to 2 mL microcentrifuge tubes, and pelleted at 500g, 4 °C. The Internally normalized cellular fluorescence uptake assay K562 CRISPRi or CRISPRa cells stably expressing sgRNAs marked with BFP mixed at a 1:1 ratio with non-transduced (sgRNA-) cells (20,000 cells in 180 μL per well) were seeded into 96well round bottom plates and incubated at 37 °C overnight. Cells were treated with fluorescent compounds at the concentrations indicated (200 μL final volume per well) and incubated at 37 °C for 24 h. Cells were pelleted at 500g, washed twice with ice-cold PBS supplemented with 1% bovine serum albumin (Millipore) and 0.1% NaN3, and resuspended in the same before assessment by flow cytometry on an Attune NxT (Thermo Fisher Scientific). TAMRA fluorescence (YL-H: 561 nm excitation laser, 585/16 emission filter) and BFP fluorescence (VL1-H: 405 nm excitation laser, 440/50 emission filter) was measured for cells within each well. Relative cellular uptake was determined by dividing the median TAMRA fluorescence intensity of BFP+ populations by that of BFP-populations (Fig. 2B). Relative cellular uptake < 1 indicates decreased uptake resulting from the genetic perturbation and > 1 indicates increased uptake. Images were acquired from randomly selected fields of view and then exported to Inkscape for display or ImageJ for quantification. For quantification, cells were segmented manually and thresholded in the LysoTracker channel to generate a lysosome mask. Pixel intensity was measured within the cell as a whole, within the lysosome mask (endolysosomal) or within the cell as whole excluding the mask (intracellular). Quantification was performed on > 30 cells per condition from 3 distinct biological replicates.

ATP-site kinase pulldown
ATP-site competition binding assay (KdELECT) was performed by Eurofins DiscoverX as described previously (59). Compounds were assessed in 11-point 3-fold dilution series and compound mixtures were analogously diluted from a DMSO stock containing the 2 compounds at the ratio indicated. Pulldown measurements of DNA-tagged kinase by quantitative polymerase chain reaction (qPCR) were normalized to DMSO-treated values to determine relative ATP-site pulldown. A 4-parameter nonlinear regression model was fit to the data using Prism 9 (GraphPad Software) with the top parameter constrained to 100%. An outlier point corresponding to 152% relative ATP-site pulldown at 15.2 pM Dasatinib + Asciminib (1:100) was excluded from analysis and plotting in fig. S8C due to high likelihood of technical error associated with the measurement.

Live cell kinase occupancy profiling
Compound treatment and preparation of cell lysates for proteomics analysis K562 CRISPRi cells (1 × 10 6 /mL) were maintained in RPMI medium (Gibco) supplemented with 10% (v/v) fetal bovine serum (FBS) (Axenia BioLogix), penicillin (100 U/mL, Gibco), and streptomycin (100 µg/mL, Gibco). Cells were pretreated with DMSO, dasatinib + asciminib (10 nM, 100 nM, or 1 μM), or DasatiLink-1 (10 nM, 100 nM, or 1 μM) at 37 ℃ for 4 h, followed by treatment with XO44 (2 mM) at 37 ℃ for another 30 min. Each sample was prepared in triplicate. Cell pellets were collected by centrifugation at 500g, 4 °C and lysed in 100 mM HEPES pH 7.5, 150 mM NaCl, 0.1% NP-40, 1 mM PMSF, and 1× cOmplete EDTA-free protease inhibitor cocktail (Sigma-Aldrich #11873580001). Lysates were cleared by centrifugation (16,000g, 4 °C, 30 min). Protein concentration was determined by protein BCA assay (Thermo Fisher #23225). Cell lysates were normalized to 5 mg/mL with lysis buffer for subsequent pulldown-MS analysis. TMT labeling of tryptic peptides TMT labeling was performed with the TMT10plex kit (Thermo Fisher Scientific #SK257743) according to manufacturer's recommendations with minor modifications. Briefly, peptides (25 μg) were reconstituted in 50 μL of 30% MeCN in 200 mM HEPES buffer pH 8.5. TMT reagents were reconstituted in 40 μL of MeCN per vial, and 6 μL of this solution was incubated with each sample for 1 h at RT. Reactions were quenched by adding 9 μL of 5% hydroxylamine and incubated at RT for 15 min, followed by adding 50 μL of 1% TFA to acidify the solution. TMTlabeled samples were pooled and concentrated by Speedvac to remove MeCN, and desalted using C18 OMIX Tips (Agilent #A57003100). Peptides were eluted with 50% MeCN, 0.1% TFA, and dried by Speedvac. scans were acquired at a resolution of 120,000 with an AGC of 4e5, m/z scan range of 400-1600, a maximum ion injection time of 50 ms, a charge state of 2-6, and a 60 s dynamic exclusion time. MS2 spectra were acquired via collision-induced dissociation (CID) at a collision energy of 35%, in the ion trap with an automatic gain control (AGC) of 1e4, isolation width of 0.7 m/z and an auto maximum ion injection time. For real time search, MS2 spectra were searched against human reviewed Swiss-Prot database (accessed Sept. 16,2020) with the digestion enzyme set to trypsin. Methionine oxidation was set as a variable modification, while carbamidomethylation of cysteine and TMT modification were set as constant modifications. For MS3 acquisition, a synchronous precursor selection (SPS) of 10 fragments was acquired in the orbitrap for a maximum ion injection time of 105 ms with an AGC of 2.5e5. MS3 spectra were collected at a resolution of 60,000 with higher-energy C-trap dissociation (HCD) collision energy of 55%.

Protein identification and TMT quantification.
Raw files were analyzed with Thermo Scientific Proteome Discoverer (2.4) software against the human reviewed Swiss-Prot database (accessed Sept. 16,2020). Trypsin was selected as the digestion enzyme with a maximum of 2 missed cleavages and a minimum peptide length of 6. Cysteine carbamidomethylation and TMT-6plex on K and peptide N-terminus were set as fixed modifications, while methionine oxidation and acetylation of protein N-terminus were set as variable modifications. Precursor tolerance was set to 10 ppm, and fragment tolerance was set to 0.6 Da. Peptide-spectrum match (PSM) and protein false discovery rate (FDR) were set to 1% and 5%, respectively. Reporter ion intensities were adjusted to correct for impurities during synthesis of different TMT reagents according to the manufacturer's recommendations. For quantification, PSMs with an average reporter signal-to-noise threshold (< 9) and synchronous precursor selection (SPS) mass matches threshold (< 75%) were removed from final dataset. Quantified PSMs were summarized to their matched proteins. Median protein intensities for each TMT channel were used to normalize protein intensities across all channels. Normalized treatment intensities for each replicate were divided by DMSO values, log2 transformed, and averaged for calculation of the mean log2 fold change for each condition (e.g. dasatinib + asciminib or DasatiLink-1). Contaminant keratin (KRT) proteins were excluded from plots.

Chemical-genetic interaction mapping
Cells treated with compounds were evaluated for viability as described herein. For

Physicochemical property determination
Unless otherwise specified, physicochemical properties of compounds were computed using the Mcule property calculator (77) and listed in table S1. Simplified molecular-input line-entry system (SMILES) strings were inputted to https://mcule.com/apps/property-calculator/.       (A) 1 H-15 N heteronuclear single quantum coherence (HSQC) spectra of ABL1 kinase domain in the presence of dasatinib (blue), asciminib (green), dasatinib + asciminib (red), and DasatiLink-1 (black). (B) Chemical shift differences for assigned residues in ABL1 kinase domain resulting from interactions with different inhibitors as in (A). δ (ppm) > 0.1 indicates a major chemical shift difference. (C) ATP-site pulldown of ABL1 kinase domain in the presence of inhibitor with or without addition of 100-fold molar excess asciminib. Data represent two biological replicates.