Co-targeting HSP90 alpha and CDK7 overcomes resistance against HSP90 inhibitors in BCR-ABL1+ leukemia cells

HSP90 has emerged as an appealing anti-cancer target. However, HSP90 inhibitors (HSP90i) are characterized by limited clinical utility, primarily due to the resistance acquisition via heat shock response (HSR) induction. Understanding the roles of abundantly expressed cytosolic HSP90 isoforms (α and β) in sustaining malignant cells’ growth and the mechanisms of resistance to HSP90i is crucial for exploiting their clinical potential. Utilizing multi-omics approaches, we identified that ablation of the HSP90β isoform induces the overexpression of HSP90α and extracellular-secreted HSP90α (eHSP90α). Notably, we found that the absence of HSP90α causes downregulation of PTPRC (or CD45) expression and restricts in vivo growth of BCR-ABL1+ leukemia cells. Subsequently, chronic long-term exposure to the clinically advanced HSP90i PU-H71 (Zelavespib) led to copy number gain and mutation (p.S164F) of the HSP90AA1 gene, and HSP90α overexpression. In contrast, acquired resistance toward other tested HSP90i (Tanespimycin and Coumermycin A1) was attained by MDR1 efflux pump overexpression. Remarkably, combined CDK7 and HSP90 inhibition display synergistic activity against therapy-resistant BCR-ABL1+ patient leukemia cells via blocking pro-survival HSR and HSP90α overexpression, providing a novel strategy to avoid the emergence of resistance against treatment with HSP90i alone.

Representative images of Sanger sequencing results of CRISPR-Cas9 mediated HSP90α-(A) and HSP90β-KO (B) clones.The vertical line indicates the mutation start site.Clone (C), empty vector (EV), wild type (WT) and non-targeting (NT) control.RT-PCR data confirming a notable increase in HSP90α mRNA transcripts upon HSP90β-KO in the K562 (C) or KCL22 (D) cell line models.(E) SNP-array comparing EV control and HSP90β-KO K562 cells.(F) Western blotting (WB) analysis of other noncytosolic HSP90 paralogues (GRP94 and TRAP1), HSR-related proteins (HSP70, HSP40 and HSP27) and HSP90 co-chaperones (AHA1 and CDC37) revealed no apparent change in HSP90α-KO cells.βactin served as a loading control.(G) WB results of the conditional short hairpin RNA (sh) mediated KD of HSP90α (α-KD) and HSP90β (β-KD) isoforms using doxycycline inducible Tet-Off system in K562 cells.Non-targeting control (NT), with (+) or without (-) Doxycycline.β-actin served as a loading control.(H) RT-PCR data validating the conditional HSP90β-KD (left panel) at mRNA level, and showing a notable increase in the HSP90α (middle panel) and HSP70 (right panel) mRNA transcripts upon conditional HSP90β-KD.WB analysis of HSP90α-KO (I) and HSP90β-KO (J) models to show effect on the binding preferences of specific client proteins on distinct HSP90 isoforms (with SURVIVIN as an HSP90α-dependent client and CDK4 and CDK6 as HSP90β-dependent client proteins).FKBP5 expression was found upregulated in both HSP90α/β-KO cells.GAPDH served as a loading control.(K) Expression of BCR-ABL and p-BCR-ABL Y412 and their related downstream pro-survival effectors (p-CRKL Y207 and p-STAT5a Y694 ) in HSP90α/β-KO cells, analysed by automated JESS WB. β-actin served as a loading control.(L) Bars show average protein quantification measurements of p-BCR-ABL Y412 , p-CRKL Y207 and p-STAT5a Y694 levels in HSP90α/β-KO cells compared to EV control (K562) cells.Error bars = SD of three independent replicates; p-values were calculated by unpaired two-tailed student's ttest.(M) Immunofluorescence imaging demonstrated a higher abundance of BCR-ABL foci (cytoplasmic/nucleocytoplasmic region) in HSP90α-KO cells than in HSP90β-KO or EV control cells.(N) Cell cycle analysis of HSP90α/β-KO cells revealed no significant changes in sub-G1, S, G0/G1 and G2/M phases when treated with vehicle (DMSO) or the positive control vorinostat (SAHA).As anticipated, the administration of the positive control, vorinostat, induced significant changes (determined by two way ANOVA test, n=3) in all cell cycle phases, when compared to the vehicle (DMSO) treatment group.Displayed are representative images of colonies taken from the colony forming unit (CFU) assay after plating K562-(O, upper panel) or KCL22-(P, upper panel) HSP90α/β-KO cells.A reduced colony sizes compared to the respective controls were observed by HSP90β-KO cells, whereas colonies of the HSP90α-KO cells were loosely packed with atypical morphology.Bar graphs illustrating the results of the CFU assay indicate a growth disadvantage to HSP90β-KO K562 (O, lower panel) and to KCL22 (P, lower panel) cells in comparison to HSP90α-KO counterparts or respective controls, evidenced by the significant decrease in total colony numbers (n=3, unpaired twotailed student's t-test).as the specificity cutoff criteria.Black dots represent secreted proteins that are not significantly regulated, while grey dots represent significantly regulated secreted proteins, but below log2(FC) threshold.Blue and red dots represent significantly downregulated and upregulated secreted proteins, respectively.The third replicate of the EV control in the secretome data was omitted from the statistical analysis due to its significant deviation from the other four replicates.(B) WB analysis of LCP2 and TOP2A expression in K562 and KCL22 cells (left panel).WB analysis of HSP90α-KO in BCR-ABL1+ BCP-ALL cell line SUPB15 (right top panel).As monoclonal section of HSP90α-KO SUPB15 cells was not possible in several attempts, therefore HSP90α-KO pool cells were used to determine the LCP2 and TOP2A expression (right bottom panel).Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) served as a loading control.(C) Synergy maps of PU-H71 and Mitoxantrone (TOP2i) combination matrix for K562 (upper panel) and KCL22 (lower panel) cells, using Zero Interaction Potency (ZIP) method (1).Visualization was done using SynergyFinder package.(D) Bars show average protein quantification measurements of LCK and p-LCK Y505 levels in HSP90α-KO cells compared to non-targeting control (KCL22) cells.Error bars = SD of three independent replicates; significance was calculated by unpaired two-tailed student's t-test.(E) WB analysis showing rescue experiment.KCL22 HSP90α-KO cells were transiently transfected with HSP90α overexpression (OE) construct.NT stands for non-targeting control.β-actin served as a loading control.Significance was calculated using unpaired t-test.(C) WB (JESS) analysis of the BCR-ABL1 expression in three BCR-ABL1+ BCP-ALL PDX samples.Synergy maps (generated using ZIP method (1) of PU-H71 and THZ1 (CDK7i) combination matrix using K562 and KCL22 cells (D) and multi tyrosine kinase inhibitor (TKI)-resistant counterparts of K562 and KCL22 cell lines, referred to as K562r and KCL22r (E).Synergy maps of the PU-H71 and THZ1 combination using multi tyrosine kinase inhibitor resistant SUPB15 cells, referred to as SUPB15r (F) and murine pro B cell line model BA/F3, expressing multi TKI-resistant BCR-ABL1 T315I mutant made resistant to the third generation TKI Ponatinib, referred to as BA/F3 BCR-ABL1 T315I-PNr (G).(H) Synergy maps of the PU-H71 and THZ1 combination using two relapsed (TKI-resitant) BCR-ABL1+ BCP-ALL patient derived xenograft (PDX) cells.The visualization of the synergy data was performed using SynergyFinder package.(I) Caspase 3/7 glo assay (which assesses apoptosis induction) was conducted following the treatment of peripheral blood-derived mononuclear cells (PBMCs) obtained from three healthy individuals.These cells were treated with PU-H71, THZ1, either alone or in combination, and their response was compared to that of leukemia K562 cells.Error bars = SD of three independent replicates; p-values were calculated by unpaired two-tailed student's t-test.(J) Quantification of the immunofluorescence imaging data (shown in main Figure 6E) to assess the levels of HSP90 and HSP70 proteins, after treatment of BCR-ABL1+ BCP-ALL PDX cells with PU-H71, THZ1, either alone or in combination.Error bars = SD (n=4); p-values were calculated by unpaired two-tailed student's t-test.

Generation of KCL-22 and SUP-B15 KO cells:
The transfection was carried out using the Amaxa Nucleofection system (SF Cell Line Kit, #V4XC-2032).Sanger sequencing was performed to validate the herein reported variants using the following primers: basis set was used for all atoms except iodine for which aug-cc-pVDZ-PP was used.
To model sigma-hole interactions of the iodine, an additional pseudo atom was introduced and charge fitted as described by Rendine et al. (22).The starting structure was solvated in TIP3P water model (23) in a rhombic dodecahedral box; the protein was placed at least 1.4 nm from the box edges.Na + ions were added to neutralize the system.An initial relaxation of the system was performed over 50,000 steps of steepest descent algorithm, followed by 100 ps NVT simulationunder at 300 K using the Berendsen thermostat (24), and 100 ps NPT simulation at 300 K and 1 atm using the Parrinello-Rahman barostat (25).
The final production was performed for a total of 1 s for the wildtype and S164F HSP-90.Covalent bonds were constrained using the LINCS algorithm (26) and Particle Mesh Ewald (PME) algorithm ( 27) was used to calculate long range interaction.Leap frog integrator was used to calculate the equation of motion with a timestep of 2 fs (28).
Visual Molecular Dynamics (VMD) program ( 29) was used to perform the visual analysis of the final trajectories.The RMSD and RMSF were calculated using the GROMACS modules gmx rmsf and gmx rms.

6 :
Fig.S6 Generation of HSP90α/β-KO K562 cells: For generation of KO cells, the following guide RNA (gRNA) sequences were utilized: For 2 x 10 5 cells, 100 pmol of Cas9-GFP protein (IDT, Alt-R S.p. HiFi Cas9-GFP V3, #10008100) was mixed with 120 pmol of gRNA (crRNA:tracrRNA 1:1) and assembled for 20 min at RT. Afterwards, the labelled ssODN was added and the mixture was combined with the cell suspension (resuspended cells in Nucleofector solution SF) and the electroporation enhancer (IDT).The complete volume was transferred to the Nucleocuvette module, placed in the 4D-Nucleofector system (Lonza) and electroporated with the program CA-137 (KCL-22) or CV-104 (SUP-B15).After 72 h, the cells were sorted for GFP and monoclonal selected via semi-solid cloning.Quantitative real-time PCR: RNA extraction was performed with the Maxwell RSC viral total nucleic acid purification kit (Promega, Madison, WI, USA) with the Maxwell RSC 48, following manufacturer's instructions.2 μg of total RNA was used for cDNA synthesis (QuantiTect Reverse Transcription kit, Qiagen, Hilden, Germany).Quantitative real-time PCR was carried out using BioRad cycler (BioRad).Reactions were carried out in triplicates in three independent experiments.The mean Ct values of the housekeeping gene B2M and GAPDH was used to normalize the variability in expression levels.Primer sequences: K562 cells (EV, HSP90α-KO (C1), or HSP90β-KO (C1); five biological replicates) were washed three times with PBS and shock frozen in liquid nitrogen.Proteins were extracted from frozen cell pellets as described elsewhere(3).Briefly, cells were lysed and homogenized in urea buffer using a TissueLyser (Qiagen) and, after centrifugation (15 min, 16000 rcf, 4°C), supernatants were collected.After determination of protein concentration (Pierce 660 nm Protein Assay, Thermo Fischer Scientific), samples were adjusted to 0.5 mg/ml total protein concentration with SDS buffer (final 7.5% glycerol, 3% SDS, 37.5 mM Tris/HCl pH 7.0) and 10 µl were reduced (final 20 mM dithiothreitol, 20 min, 56°C), alkylated (final 80 mM iodoacetamide, 15 min, r.t., protected from light) and finally underwent tryptic digestion (200 ng trypsin in 50 mM triethylammonium bicarbonate) after applying a slightly modified sp3 protocol(4) using 50 µg 1:1 mix Sera-Mag SpeedBeads.Peptides were reconstituted in 0.1% trifluoracetic acid and subjected to LC-MS analysis.ii) Secretomes K562 cells (EV, HSP90α-KO (C1), or HSP90β-KO (C1); five biological replicates) were washed three times with PBS and three times with FCSfree medium and incubated for 24 h in FCS-free medium at a density of 0.8 million cells / mL.The conditioned medium was collected by centrifugation (5 min, 800 x g, 4°C) and filtering through a 0.2 µm membrane (Acrodisc 32 mm Syringe Filter with 0.2 µm Supor Membrane; Pall, #4652).Aliquots were shock frozen in liquid nitrogen and stored at -80°C.An aliquot (400 µl) per cell type and replicate was thawed on ice in the presence of protease inhibitor cocktail (added 50 µl of a solution of 1 cOmplete ULTRA tablet, mini, EDTA-free in 2 mL water; Roche, #05892791001), supplemented with SDS buffer (added 50 µl of 30% glycerin, 12% SDS, 150 mM Tris base), reduced (added 40.5 µL of 100 mM dithiothreitol; 20 min at 56 °C under shaking), alkylated (added 54 µL of 300 mM iodacetamide; 15 min at r.t.protected from light), and quenched (added 40.5 µL of 100 mM dithiothreitol; 20 min at r.t.).Applying a slightly modified sp3 protocol (4), proteins were precipitated (added 10 µL of 20 mg/mL 1:1 bead-mix of pre-washed Sera-Mag SpeedBeads GE #45152105050250 and #65152105050250 in water; added 645 µL ethanol abs.p.a.; 15 min at 24 °C under shaking), washed (3x 80% ethanol, 1x acetonitrile) and digested (100 ng trypsin in 20 µL 50 mM triethylammonium bicarbonate).Peptides were reconstituted in 0.1% trifluoracetic acid and subjected to LC-MS analysis.b) LC-MS analysis: For the LC-MS analysis, a Q Exactive Plus Hybrid Quadrupole-Orbitrap (for proteomes) or a Orbitrap Fusion Lumos Tribrid (for secretomes) mass spectrometer (Thermo Fisher Scientific), operated in positive mode and coupled with a nano electrospray ionization source connected with an Ultimate 3000 Rapid Separation liquid chromatography system (Dionex / Thermo Fisher Scientific, Idstein, Germany) equipped with an Acclaim PepMap 100 C18 column (75 µm inner diameter, 25 cm length, 2 mm particle size from Thermo Fisher Scientific) was applied using a 120 min LC gradient.Capillary temperature was set to 250°C or 275°C and source voltage to 1.4 kV or 1.5 kV for the Q Exactive or Lumos mass spectrometer, respectively.MS survey scans had a mass range from 350 (Q Exactive) or 200 (Lumos) to 2000 m/z at a resolution of 140,000 (Q Exactive) or 120,000 (Lumos).The automatic gain control (AGC) was set to 3,000,000 (Q Exactive) or the normalized AGC target was set to 62.5% (Lumos) and the maximum fill time was 80 ms (Q Exactive) or 60 ms (Lumos).The ten most intensive peptide ions per survey scan were isolated and fragmented by high-energy collision dissociation (HCD) (Q Exactive) or a cycle time of 2 s was employed (Lumos).c) Data analysis: MaxQuant (version 2.0. The atomistic MD-simulations of wildtype-HSP90 and the S164F variant were performed in GROMACS v.2020.1(15-17)  using the AMBER99SB-ILDN forcefield(18) for the protein and GAFF(19) for PU-H71.The partial charges of the ligand were derived via the RESP method (20) from ab inito calculations to Gaussian 16 A.03(21) at the HF level of theory.

Western Blotting:
Cells were harvested by centrifugation at 400 x g for 5 min at 4°C and washed three times with ice-cold PBS and then snap-frozen in liquid nitrogen.50 µL RIPA lysis buffer (Thermo Fisher Scientific, #89900) supplemented with cOmplete tm (Roche, #11697498001) and PhosSTOP (Sigma-Aldrich, #4906845001) was used per 1 million cells.Cells were lysed on ice for 1h with periodically vortexing, centrifuged two times (10000 x g for 20 min at 4 °C) and protein quantification was performed by BCA-Assay (Thermo Fisher Scientific, #23227).10 -20 µg lysate were separated by SDS-PAGE at 50 V for 30 mins during stacking phase with subsequent 100 V for 2h during seperation phase and blotted onto 0.45 µm nitrocellulose membrane (Cytiva, #10600002) at 100 V for 1h or overnight (30 V, 16 h).Membranes were washed twice with TBS and analyzed for their protein content by ponceau staining (Sigma-Aldrich #P7170) staining.After three washes with TBS-T, membranes were blocked in 5% BSA (Sigma-Aldrich, #A3294) TBS-T solution, washed three times with TBS-T and incubated with primary antibody solution in 5% BSA solution overnight at 4°C.Membranes were washed three times with TBS-T, incubated for 1h with secondary HRP-conjugate (Cell Signaling Technologies, #7074 and #7076) at 1:2000 in TBS-T solution, washed again three times in TBS-T and lastly one time in TBS.For