Metabolic impairment of non-small cell lung cancers by mitochondrial HSPD1 targeting

All cell lines were cultured in media supplemented with 10% FBS, 1% Pen/Strep and 1% L-Glutamine (all from Sigma) at 37 °C and 5% CO₂ in a humidified incubator. H460, H1299, H520, SK-MES-1 and Calu-1 (from ATCC) were cultured in RMPI-1640 (Sigma). H23 and H838 (both from ATCC) cells were cultured in RMPI-1640, supplemented with 1 mM sodium pyruvate (Gibco). A549, 293T, BEAS-2B (ATCC), BEN (DSMZ), murine Lewis lung carcinoma cell line LL2, Ladi2.1 and Ladi3.1 cells (derived from p53fl/fl-LSL KRASG12D/ + mouse NSCLC model) were cultured in DMEM (Sigma). Human cells were STR-profiled, used between passages 3 and 15, examined for mycoplasma regularly (detection kit from Invivogen). The A549-Cas9 cell line was generated by lentiviral transduction with lentiCas9-Blast (Addgene #52962) and selection with 100 μg/mL of blasticidin (Sigma) for 3 days. KHS101 hydrochloride (4888) was purchased from Tocris; cisplatin (CAS 15663–27-1), Z-VAD-FKM (CAS 187389–52-2) and 2-Deoxy-D-Glucose (2-DG, sc-202010) were purchased from Santa Cruz Biotechnology; Necrostatin-1 (N9037-10NMG), Ferrostatin-1 (SML0583-5MG) and 1-Fluoro-2,4-dinitrobenzene (DNFB, D1529) were purchased from Sigma. HB072 was obtained as previously shown [23].

For in vitro drug treatment, cells were plated in a 96-well plate in low density (5–10% initial confluence) and incubated overnight. On the next day, cells were treated either with vehicle or with the drugs. Plates were loaded into the IncuCyte Zoom (Essen Bioscience) and scanned every 2–4 h. For each scan, phase contrast images were acquired from every well and analyzed by IncuCyte Zoom software. IC₅₀ values were calculated after 72 h of treatment as expression of confluency percentage of treated cells normalized to vehicle control (100% confluency) using Graphpad (nonlinear regression curve fitting model) to obtain the values. Proliferation assay was performed as described in [24]. Growth curves were analyzed with IncuCyte Zoom software.

For cytotoxicity death assay, cells were seeded in 96-well plates in low density (5,000 cells/well) and incubated overnight. 1000X Cytotox Green Reagent or Caspase 3/7 Green Apoptosis Assay Reagent (Essen Bioscience) was diluted in medium and working dilutions of the drug were prepared in Cytotox Green or Caspase3/7 Green supplemented media. After treatment, plates were loaded in IncuCyte Zoom and images were acquired in real-time for phase to quantify growth. Activity of green reagent was simultaneously acquired at the green channel to quantify death. IncuCyte Zoom software was used for the analysis and data export.

Oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) measurements were determined using the XFe96 Extracellular Flux Analyzer (Seahorse Bioscience/Agilent Technologies). Cells were seeded in specialized culture microplates at a density of 10,000–20,000 cells/well and prepared as describe in [25]. For Mito Stress Test (Agilent kit 103015–100) 1 μM oligomycin, 1 μM FCCP, 1 μM rotenone and 1 μM antimycin were sequentially injected at regular intervals. OCR, indicator for mitochondrial respiration, and ECAR, indicative of glycolysis, were measured.

Plasmids for HSPD1 knock down (TRCN0000029444, TRCN0000029445, TRCN0000029446, TRCN0000029447 and TRCN0000029448 for human cell lines) were purchased from Sigma. Scrambled pLKO.1 (referred to as pLKO) was used as non-targeting control. NLRC5 expression vector (EX-A3335-Lv105) and control vector (Ex-Neg-LV105b) were from GeneCopoeia. Plasmids containing RNAs for COX5B (gRNA sequence #1: ACTTCGCGGAGCTGGAACGC; gRNA sequence #2: CAGCCAAAACCAGACGACGC) and SLC6A8 (gRNA sequence #2: ACGGGGCCGTCGCCCTTGGC; gRNA sequence #3: GCCCTTACCATGCAGACCAG) knock-out (pLENTI-CRISPR-V2) were from Genscript. For production of lentiviral particles, 293 T cells were transfected with 8 μg knock-down/expression vectors and 2 μg of packaging vectors (pMDL, pVsVg and pRevRes) in complex with 24 μg PEI (Polysciences) in 0.9% NaCl. After 48 h, supernatant was collected, centrifuged and filtered. For transduction, 150,000 cells were seeded in a 6-well plate and infected in presence of 8 μg/mL polybrene (Sigma). Selection was done with 3 μg/mL puromycin (Sigma) and cells were maintained in 1 μg/mL puromycin.

For proteins isolation, cells were lysed in RIPA buffer and quantified using Pierce BCA kit (Thermo-Fisher). For cytosolic/mitochondrial fractionation Cytochrome c Release Assay Kit (abcam, ab65311) was used. Proteins lysates (10–20 μg) were resolved on 5%-12% SDS–PAGE gels and transferred to PVDF membrane (Thermo-Fisher). Membranes were blocked in 5% Milk (BioRad) or 5% BSA (Sigma) in 1X TBST and incubated overnight at 4 °C in primary antibodies. Membranes were then washed with 1X TBST and incubated with secondary antibodies (Southern Biotech) for 1 h. The membranes were developed with ECL reagent (Thermo Fisher) on to X-ray films (Thermo-Fisher) using the chemiluminescence imager, AGFA CP100. Rabbit anti-HSPD1 (ab46798, 1:20,000) and rabbit anti-NLRC5 (ab117624, 1:1000) antibodies were purchased from Abcam; mouse anti-TOMM20 (H00009804-M01, 1:1000) was purchased from Abnova; anti-β-Actin (8H10D10) HRP conjugate (1:10,000) was purchased from Cell Signaling. Protein band density was quantified using ImageJ.

To assess clonogenic ability of knockdown cells or KHS101 treated cells, a colony formation assay was performed. Control (pLKO) and knockdown cells were plated in triplicates at a low density (1,500 cells/well) in a 6-well plate and they were allowed to grow until they formed visible colonies. To assess the clonogenic ability of cells upon KHS101 treatment, 200,000 cells were seeded in triplicates in 6-well plates, subsequently treated with 10 μM KHS101 for 5 days, after which media was replaced with normal media, whereas the DMSO cells were collected, counted and seeded in triplicates at low density (1,500 cells/well). Cells were allowed to grow until they formed visible colonies in the controls. Cells were washed with 1X PBS and fixed in 10% (v/v) formalin (Sigma) for 5 min. After washing, colonies were stained with 0.05% crystal violet solution (Biomatik, CAS 548629) for 30 min and washed twice with deionized water. Colonies were photographed and counted.

Cell cycle analysis was performed using the Propidium iodide (PI; Sigma) staining. 50,000 cells were seeded in triplicates in 12-well plates and incubated overnight. Samples were prepared as described in [26]. For ROS staining, 50,000 cells were seeded in 12-well plates and the day after they were treated with KHS101 in combination with cisplatin. After 24 h, samples were washed and stained for 30 min at 37 °C with CM-H2DCFDA 5 μM (Thermo Fisher, C6827). Then they were collected in FACS tubes, washed and re-suspended in FACS buffer (2% FBS, 5 mM EDTA in 1X PBS). Samples were run on Cytoflex FACS machine (Beckman) and data were analyzed using FlowJo software v10.6.

For mitochondrial staining, 100,000 cells were seeded in 12-well plates. The day after, MitoTracker Green FM (Cell Signaling Technologies, #9074S) was diluted to 400 nM directly into fresh media and added to the cells, which were incubated for 30 min at 37 °C. After incubation, media was changed, and live cells were imaged using IncuCyte-Zoom.

Total RNA was extracted using miRNeasy Mini Kit (Qiagen) according to the manufacturer's instruction. 500 ng of RNA were used to perform reverse transcription with the Tetro cDNA synthesis Kit (Bioline). cDNA was amplified with gene-specific (HK2, FBP1 and PKM) TaqMan probes (Applied Biosystem) using TaqMan Universal Master Mix II (Thermo Fisher Scientific). GAPDH was used as internal control. Quantifications were done by Applied Biosystems 7300 Real Time PCR system and relative mRNA expression level was calculated using the ΔΔCt method.

26 different lung cancer cell lines (Additional file 1: Table S1) were screened to determine their viability upon treatment with KHS101. The drug was added to the cells one day after seeding and the treatment was performed by nanodrop-dispensing using a Tecan Dispenser. 0.1% DMSO (solvent) and Staurosporine (1,0E-05 M) served as high control (100% viability) and low control (0% viability), respectively. After 72 h of incubation with the compound, cell plates were equilibrated to room temperature for one hour and CellTiter-Glo™ Luminescence Cell Viability Reagent (Promega) was added to the cell suspension. The luminescence was measured 1 h later using a luminometer. IC₅₀ values were expressed as percentage of proliferation in presence of solvent alone (100% = high control) as compared to cells treated with 1E-5 M Staurosporine (0% = low control). IC₅₀ calculation was performed using GraphPad Prism software with a variable slope sigmoidal response fitting model.

The CRISPR/Cas9 lentival library was generated as described in [25]. A549 Cas9 cells were transduced with serial dilutions of a virus to find the MOI of ~ 0.3. A549 Cas9 cells were transduced with lentiviral Human GeCKO v2 knockout pooled library part A and part B (a gift from Feng Zhang [27], Addgene # 1000000049) at MOI of 0.3 in the presence of 10 μg/mL of polybrene for 24 h, then replaced the virus medium with fresh growth medium and continued to culture the cells for 48 h. The cells were selected with 50 μg/mL of puromycin for 3 days. Subsequently, 4 million cells were seeded in 10-cm dishes and treated the day after either with vehicle or with 15/10 μM KHS101. A549 Cas9 pLKO were used as positive control. After 2 weeks, the media containing the drug was replaced with normal media to allow the cells to grow again. 40 million cells were then collected for genomic DNA isolation. Next generation sequencing was performed on the Illumina HiSeq 2500 platform in Deep Sequencing Facility of TU Dresden. The raw FASTQ files were analyzed with MAGeCK-VISPR [28].

Genomic DNA was extracted with NucleoSpin® Blood XL (Machery Nagel # 740,950.50) according to the manufacturer’s protocol. The first round PCR of Next Generation Sequence (NGS) was performed with 26 separate 100 μL redundant reactions, each containing 5 μg of DNA, 50 μL Q5® Hot Start High-Fidelity 2X Master Mix (NEB # M0494L), and 3 μL of a 10 μM solution of each primer (P5 and P7). The PCR was performed as described in [25].

A549 cells were treated either with vehicle (DMSO) or KHS101 10 μM for 48 and 96 h. After treatment the cells were fixed with 2% (v/v) glutaraldehyde in 0.04 M phosphate buffer pH 7.4 for 60 min, scraped off, collected in a tube and post-fixed with the same fixative for 24 h. The cells were then washed in PBS and resuspended in 15% BSA for 10 min. The cells were centrifuged, and the cell pellet fixed by slowly adding the fixative for 48 h at 4 °C. The cell pellets were cut in smaller pieces and prepared for EM like following. Pellet fragments were washed in phosphate buffer, stained with 1% osmium tetraoxide in phosphate buffer for 90 min, dehydrated in ethanol series and acetone and embedded in TAAB 812 Embedding Resin (T030, TAAB). Semithin (1 µm) sections were cut on Leica Ultracut UCT microtome and stained with toluidine blue. Ultrathin (70 nm) sections were collected on uncoated nickel grids (M200-NI, Electron Microscope Sciences). The grids were stained with 3% uranyl acetate for 15 min at 60 °C and 3% lead citrate (Leica Ultrostain 2) for 6 min at room temperature. The sections were analyzed with JEM-1400 Plus electron microscope, equipped with Quemsa TEM CCD camera and images obtained using Radius software (TEM Imaging Platform software).

Normalized gene expression values for HSPD1 were obtained for lung cancer patient samples from GEO (GSE30219) and mRNA z-score values for the TCGA profile (TCGA LUAD, Cell 2018) from the cbioportal platform [29]. Survival curves were generated using Kaplan–Meier estimate with the samples categorized into ‘low’ and ‘high’ HSPD1 expression groups based on the median of the HSPD1 mRNA expression value. Log-rank test was conducted to obtain the significance between the two groups in R software.

Differential gene expression (DGE) analysis was performed between the KHS101-sensitive and KHS101-resistant cell lines identified from the drug sensitivity analysis. Briefly, the top four KHS101-sensitive cell lines (NCI-H460, NCI-H1581, LOU-NH91 and A549) and top four KHS101-resistant cell line (SK-LU-1, NCI-H1563, BEN and NCI-H838) were chosen for the analysis. Differential expression between the sensitive and resistant groups was performed using cancer cell line encyclopedia (CCLE) gene expression profile (GSE36133) with RMA log₂ signal intensity using GEO2R package in GEO with a moderated t-test parameter [30]. 33 up-regulated and 26 down-regulated genes were filtered with a p-value of < 0.05 and log₂ fold change cutoff of 2. In parallel, Pearson’s correlation analysis was performed between IC₅₀ values of KHS101 from 21 cell lines and 18,569 genes expression from CCLE. Based on a correlation value of 0.5 and p-value < 0.05, 298 positively and 97 negatively correlated genes were selected. An overlap analysis was performed with the DGEs list and correlation list to identify a stringent gene list of 12 genes. Correlation analysis between HSPD1 gene expression value obtained from GEO (GSE47206) and KHS101 dose response IC₅₀ values was determined using GraphPad Prism 8.

Similar to the differential gene expression analysis, a differential metabolite analysis was performed for the top four KHS101-sensitive and KHS101-resistant group cell lines with the information of 225 log₁₀ normalized quantitative metabolites levels across 177 CCLE cell lines [31]. 7 differential metabolites were identified with a log₁₀ fold change of 2 and p-value < 0.05.

Gene ontology analysis was performed for the top 20 gene candidates identified from the CRISPR knockout screen of A549 cells treated with KHS101. Genes were assessed for the enrichment using GO cellular component 2017b in Enrichr [32].

HSPD1 fitness was analyzed as described in [33]. HSPD1 gene’s cancer dependency was analyzed in a panel of NSCLC cell lines (n = 52) from PROJECTDRIVE RNAi screen [34]. Using redundant siRNA activity (RSA) gene level metric of threshold -3, the essentiality of the gene in cell lines was determined.

NSG strain (JAX) was used as an experimental model to study the tumor growth of HSPD1 knockdown cells (A549 and H1299). For subcutaneous injections, 0.5 × 10⁶ cells re-suspended in 50 μL 0.9% NaCl were mixed with Matrigel (Corning, 734–0268) in a ratio 1:1 (v:v). Cells were injected in flanks of 9–15-weeks-old NSG with 8 mice per group. Caliper measurements were taken twice a week and tumor volume was calculated using the formula (Length × Width² × π)/6. After 4/5 weeks, mice were euthanized by cervical dislocation and tumors were isolated and weighted.

C57BL/6 strain was used as experimental model to evaluate KHS101 effect on lung metastasis tumor formation. For the tail vein metastasis assay, 0.5 × 10⁶ LL2 cells were re-suspended in 100 μl PBS and injected in the tail vein of female C57BL/6, with 10 mice per group. After 2 weeks either vehicle (5% (v/v) Ethanol-15% (w/v) (2-Hydroxypropyl)-β-cyclo-dextrin (Captisol Technology)) or 6 mg/kg KHS101 (Cellagen Technology) were injected twice a day subcutaneously for 10 days. Lung metastases were monitored by bioluminescence imaging (BLI). Anesthetized mice were intraperitoneally injected with 50 mg/mL D-luciferin (Kayman Chemicals). Bioluminescence images were acquired with Lumina III in vivo Imaging System (IVIS, Perkin Elmer). Time of surrogate survival with disease was calculated as difference between the experimental endpoint and the day of appearance of the bioluminescence signal.

Statistical tests were performed with the GraphPad software (Prism) v.8 comparing groups of different conditions. In all tests at least three biological replicates were used and the statistical significance was considered at p ≤ 0.05.

To understand whether HSPD1 could be an attractive target for NSCLC, we performed a fitness analysis from a previously published dataset of a pan-cancer and genome-wide CRISPR/Cas9 screening of 18,009 genes, which included n = 21 adenocarcinoma (ADC) and n = 11 squamous cell carcinoma (SqCC) cell lines [35]. As a result, the HSPD1 gene scored high, ranking 25^th in ADC and 71^st in SqCC cell lines, out of 805 and 559 total ‘fitness’ genes, respectively (Fig. 1A). This initial observation was supported by the results from an independent dataset based on a large-scale RNAi screen of 7,837 genes performed in 52 NSCLC cell lines [34]. HSPD1 was a significant cancer-dependent gene in over 70% of the cells (Fig. 1B, Additional file 1: Table S2), confirming the importance of this gene for NSCLC survival. In addition, survival analysis on HSPD1 mRNA expression showed that the patients with ‘high’ levels of HSPD1 carry a worse prognosis compared to those with ‘low’ HSPD1, in terms of overall (Fig. 1C), relapse-free (Fig. 1D) and disease-specific survival (Fig. 1E). We then investigated the prevalence of HSPD1 expression in tissue samples from a small cohort of NSCLC patients (n = 30) by immunohistochemistry (IHC), and observed positive HSPD1 protein expression in 100% of the cases (Fig. 1F-I). The expression pattern was cytoplasmic, granular or diffuse, with some degree of heterogeneity (semi quantitatively scored in a three-level scale), with very high protein levels identified in 64% of specimens (Fig. 1J). No difference was observed between ADC and SqCC samples in terms of HSPD1 expression levels. To confirm HSPD1 expression in NSCLC cell lines grown in vitro, a panel of 10 different cell lines, including three mouse-derived ones (Ladi3.1, LL2 and Ladi2.1) [33], was subjected to western blot quantification. The results indicated detectable HSPD1 protein in all samples (Fig. 1K). Separated protein isolation from mitochondria and cytosolic fractions confirmed that HSPD1 is predominantly localized in the mitochondria in NSCLC cells (Fig. 1L). Therefore, HSPD1 is a mitochondrial protein essential for NSCLC with a strong prognostic value and high prevalence of expression in tissue samples, making it a very valuable candidate for targeted inhibition.

To study the function of HSPD1 in NSCLC cells in vitro, we analyzed the effects of its knockdown by transducing 5 different NSCLC cell lines with up to 5 independent shRNA sequences (called #44, #45, #46, #47 and #48) (Fig. 2A, Additional file 1: Fig. S1A). Proliferation was evaluated through a real-time cell confluency assay (using the Incucyte system). Shortly after shRNA-based inhibition of HSPD1 expression, the cells showed a decrease in proliferation compared to the controls transduced with a scrambled/non-targeting shRNA, referred to as pLKO (Fig. 2B, Additional file 1: Fig. S1B). In particular, the phenotype was stronger for shRNA sequences providing higher levels of HSPD1 suppression (i.e., #48). HSPD1 knockdown cells were not able to form colonies compared to the control cells when plated at lower density (1,500 cells/well) as observed over two weeks (Fig. 2C-D, Additional file 1: Fig. S1C-D). In line with a strong and durable effect on cell proliferation, a cell cycle analysis (by FACS) showed a significant increase in the percentage of cells in sub-G1 and G1 phases in HSPD1 knockdown cells, in contrast to a reduction in cells residing in S or G2-M phases (Fig. 2E-H, Additional file 1: Fig. S1E-F). To evaluate whether HSPD1 loss-of-function could also affect the growth of normal lung cells, we induced the knockdown in an immortalized non-tumorigenic bronchial epithelial cell line (BEAS-2B). As a result, BEAS-2B cells showed a dramatically reduced growth in vitro (Additional file 1: Fig. S1G-H); however, they were able to completely recover their proliferative capacity as early as one week post infection in contrast to the cancer cells showing a sustained reduction in proliferation capacity (Additional file 1: Fig. S1I).

HSPD1 has a critical function in mitochondrial metabolism [23]. Therefore, to investigate whether the growth suppression observed with HSPD1 knockdown was linked to metabolic alterations, an extracellular metabolic flux analysis was performed using two different cell lines. Throughout the experiments, knockdown cells showed a significant reduction in their basal respiration and in their ATP-linked respiration, suggesting that the alterations in the mitochondrial metabolism may be responsible for the reduction in cell proliferation (Fig. 3A-B). To investigate whether this reduction could be due to a decrease in mitochondrial number, a specific-mitochondrial staining was performed in HSPD1 knockdown cells, which, did not show any difference in the fluorescent intensity of the mitochondrial dye (Additional file 1: Fig. S2A). Finally, we tested whether HSPD1 knockdown inhibited cell growth in vivo. To this end, we injected shHSPD1-expressing A549 and H1299 cells subcutaneously in NSG mice to evaluate the tumor growth rate over time. The results markedly indicated that HSPD1 knockdown cells formed very small and slower-growing tumors compared to control pLKO cells, in terms of both tumor volume and tumor weight (Fig. 3C-H), an effect which was maintained for the entire duration of the experiment, hence confirming that HSPD1 is required for NSCLC growth.

KHS101 is a synthetic small molecule reported to impair the growth of glioblastoma cells in vitro and in vivo via inhibition of HSPD1 chaperone activity [23]. As KHS101 leads to metabolic exhaustion and apoptosis in glioblastoma cells, we first tested as to whether the compound elicits cytotoxicity in NSCLC cell lines. After treatment with KHS101, 6 human NSCLC cell lines showed a significant reduction in cell proliferation that was dependent on the compound concentration (Fig. 4A, Additional file 1: Fig. S2B-C). To verify the specificity of HSPD1 targeting, we exposed the cells to HB072, an HSPD1 non-targeting chemical analogue of KHS101 [23] (Fig. 4B), which did not significantly affect cell growth (Fig. 4C, Additional file 1: Fig. S3A). After 5 days of continuous drug treatment, NSCLC cells lost their clonogenic growth ability, as shown in a colony formation assay (Fig. 4D-E, Additional file 1: Fig. S3B). In line with HSPD1 knockdown experiments, cell cycle analysis showed a significant reduction in the percentage of cells in S and G2-M phases concomitant with an increase of cells in G1 phase (Fig. 4F-I), as assessed 24 h after KHS101 treatment. Finally, since many conventional anti-cancer therapeutic strategies fail to target non-proliferating cancer cells, we sought to test whether the KHS101-induced growth inhibitory effect was maintained altering the proliferative state of the cells in a controlled fashion. Therefore, after being synchronized overnight using a low FBS concentration (0.5%), the cells were treated with KHS101 in presence of media with increasing FBS, ranging from 0 to 5%. As a result, no differences were measured in KHS101 efficacy in the different conditions by a real-time proliferation assay (Additional file 1: Fig. S3C), indicating that KHS101 is effective independently of the cells’ proliferative status. In conclusion, these data indicate that chemical targeting of HSPD1 markedly reduces the growth of NSCLC cell lines.

To test whether KHS101 affects energy metabolism in NSCLC cell lines, we performed an extracellular flux analysis 24 h after KHS101 treatment. The results indicated a strong reduction in cellular basal respiration and in ATP-coupled respiration capacity. Concomitantly, a significant increase was observed for basal glycolysis (Fig. 5A-B). We next asked as to how fast KHS101 would induce such alterations and treated the cells for shorter periods (ranging from 0.5 h to 6 h), observing that glycolytic activity increases simultaneously with oxidative phosphorylation (OXPHOS) reduction from less than one hour after treatment (Additional file 1: Fig. S4A-B), and suggesting that the metabolic rewiring induced by KHS101 is an early event, as previously described [23]. We further investigated metabolic alterations evaluating the expression of some key metabolic enzymes in A549 HSPD1 knockdown cells and A549 cells treated with KHS101 for 72 h. We observed a significant increase in hexokinase 2 (HK2) expression (Additional file 1: Fig. S4C), which was consistent with the detected increase in glycolysis. In addition to metabolic flux phenograms, cell morphology was found altered in cells exposed to 10 µM KHS101. Treated cells presented an altered ultrastructure and, after 96 h, severe swelling of mitochondria and sparse presence of cristae was observed. Moreover, an increased number of secondary lysosomes was found indicating a marked cellular degeneration associated with a loss of microvilli (Fig. 5C). These alterations were already detectable as early as 48 h after treatment (Additional file 1: Fig. S4D). As a result of these alterations, and differently to HPSD1 knockdown, KHS101 treatment induced cell death, as indicated by the incorporation of fluorescent Cytotox green dye (Fig. 5D, Additional file 1: Fig. S4E). To test whether the affected cells were undergoing apoptosis, we performed a caspase 3/7 activity assay, which showed specific activation of apoptosis 2 days after the treatment (Fig. 5E, Additional file 1: Fig. S4F-G). Co-administration of the pan-caspase inhibitor Z-VAD-FMK could efficiently suppress this caspases activation (Additional file 1: Fig. S4I) but did not rescue the cells from KHS101 cytotoxicity (Fig. 5F, Additional file 1: Fig. S4H), thus indicating that cell death is not occurring solely as a result of apoptotic induction. Interestingly, no protection was observed also in presence of necroptosis or ferroptosis inhibitors (Additional file 1: Fig. S4J-K), suggesting that the metabolic alterations caused by KHS101 could not be rescued by blocking cell death pathways. Furthermore, to investigate the impact of KHS101 in vivo, we tested the sensitivity of the tumor-forming mouse-derived NSCLC cell line LL2 (Additional file 1: Fig. S5A-B), and subsequently injected them in the tail vein of BL6 mice to evaluate the growth of lung metastasis upon KHS101 treatment. Lung tumors-bearing mice that were treated subcutaneously for 2 weeks with 6 mg/kg KHS101 (twice a day) did not show a significant difference in tumor area, as evaluated by IVIS measurement. However, KHS101-treated mice showed a prolonged disease-specific survival time (Additional file 1: Fig. S5C), in line with a non-significant trend for a reduced number of lung lesions, as macroscopically detected (Additional file 1: Fig. S5D).

Finally, we investigated the in vitro effects of KHS101 in combination with cisplatin, one of the most commonly used chemotherapeutic drug used in relapsing NSCLC [36]. Strikingly, a low cisplatin dose (1 or 3 µM) that alone was not able to strongly impair cell growth or survival, showed to induce a marked growth suppression (Fig. 5G, Additional file 1: Fig. S6A) and increase in cell death (Fig. 5H, Additional file 1: Fig. S6B) when combined with KHS101 (also at sub-lethal doses), resulting in a 2 to 4-fold reduced relative growth compared to cisplatin alone in three different cell lines (Additional file 1: Fig. S6C). Cisplatin exerts its activity via upregulation of reactive oxygen species (ROS) [37] and by fluorescent ROS staining of treated cells we found that the combination of cisplatin and KHS101 synergistically increased ROS activation (Fig. 5I-J). These data suggest that HSPD1 targeting can also induce cell death and increase the efficacy of chemotherapeutic treatments.

We further profiled in vitro sensitivity to KHS101 by performing a screening on 26 different lung cancer cell lines, measuring cell viability 72 h after treatment. Interestingly, none of the cell lines in the panel appeared to be completely resistant (Fig. 6A, Additional file 1: Fig. S7A). KHS101-induced cytotoxicity was independent of the lung cancer type (NSCLC or SCLC) or NSCLC histological subtype (Fig. 6B). However, cells showed different grades of sensitivity towards the compound with IC₅₀ values ranging from 2 to 14 μM (Fig. 6C, Additional file 1: Fig. S7A). These results were independently validated comparing KHS101 sensitivity in the most sensitive and resistant cell lines based on percentage of inhibition of cell proliferation (Additional file 1: Fig. S8A). To find possible determinants of drug resistance, cell lines with different genetic alterations were compared, but no particular genetic background was found as significantly protective (Additional file 1: Table S3). To identify possible molecular determinants of drug sensitivity, we performed a transcriptomic analysis on cell lines based on their IC₅₀ values. HSPD1 mRNA expression correlated with IC₅₀ values obtained from the screening (Fig. 6D). In order to have a stringent and genome-wide comparison, we used a 2-sided statistical approach. First, the 4 most resistant (NCI-H838, BEN, NCI-H1563 and SK-LU-1) and the 4 most sensitive NSCLC cell lines (H460, NCI-H1581, LOU-NH91 and A549) were grouped and the differentially expressed genes between the two groups were identified and ranked according to their fold differences (Additional file 1: Table S4-5). Moreover, using available data from all the cells in the screening, we correlated the expression of each gene with the KHS101 IC₅₀ values and ranked them based on Pearson metrics (Additional file 1: Table S6-7). The results from these two analyses were overlapped (Additional file 1: Fig. S8B), obtaining a list of potential up- and down- regulated genes in cells resistant to KHS101 (Fig. 6E). The top up-regulated resistance-conferring gene was the transcription factor NLRC5, which acts as a transactivator for the MHC class I genes [38] (some of which were also up-regulated in the analysis, like HLA-J or HLA-F). To validate this finding, we lentivirally overexpressed NLRC5 in sensitive NSCLC cells (Fig. 6F), and detected a significantly decreased cytotoxic effect of KHS101, indicating cellular resistance (Fig. 6G, Additional file 1: Fig. S8C). Among the genes down-regulated in the resistant cell lines, the top hit was SLC6A8 (Fig. 6E), a sodium chloride-dependent creatine transporter 1 [39]. This evidence was validated by introducing a specific CRISPR/Cas9-based knock-out of SLC6A8, which increased cells viability of sensitive cell lines treated with KHS101 (Fig. 6H). Since SLC6A8 is a creatine transporter we speculated that it could alter KHS101 sensitivity by changing the metabolic features of the cells. Analysis of a previously published metabolome profiling of NSCLC cell lines [31] revealed that cells that we determined to be the most KHS101 sensitive had a higher creatine level (Fig. 6I), which was consistent with the results from the transcriptomic analysis (i.e. higher expression of the importer) suggesting a possible effect of creatine metabolism in the response to KHS101. Creatine can be phosphorylated by the cytosolic creatine kinase (C-CK) or by the mitochondrial isoenzymes (MtCK) to provide a rapid source of ATP [40] thereby fueling ATP-dependent cellular processes. To test whether differences in cellular creatine levels could be correlated with the cells’ metabolic state, we performed extracellular flux analysis of cells with higher (A549 and H460) and lower creatine (H838 and BEN) baseline levels [31] and found that cells with higher creatine had higher basal respiration (Fig. 6J). To prove this correlation further functionally, we inhibited creatine kinase activity with the specific inhibitor DNFB which decreased the basal respiration level of A549 cells (Fig. 6K), concomitantly increasing the cells’ resistance to KHS101 as assessed by cell viability assays (Fig. 6L). Taken together these findings suggest that a higher dependency on OXPHOS due to the alteration of creatine metabolism activity could determine the sensitivity to KHS101 of NSCLC cells.

To further investigate KHS101 resistance at a broader level, we performed a genome-wide CRISPR/Cas9-based dropout screening. A genetically barcoded whole-genome library (GeCKO v2) was transduced in A549 cells overexpressing the Cas9 nuclease. These cells were treated with two different KHS101 concentrations (10 and 15 μM) or DMSO as control (Fig. 7A). None of the cells were able to survive the treatment with the 15 μM KHS101 dose, suggesting that is not possible to develop a complete resistance to this compound by the elimination of a single gene. By contrast, we could observe non-growing and morphologically altered cells surviving the 10 μM dose. After 2 weeks of survival (after all control pLKO-infected Cas9-A549 cells treated with the same KHS101 dose underwent complete cell death), the drug treatment was stopped to allow the selected cells to grow and be processed for the DNA preparation and the next-generation sequencing steps. The sequencing results on untreated cells indicated that it was possible to detect a total of 116,700 sgRNAs (all gRNAs excluding those with no counts) present in the library, with a 97% representation, a level similar to previous screenings [25], while in the treated group this percentage decreased to 56%, as a result of the drug selection. ‘Hits’ were ranked and gene ontology analysis on the 20 highest ranking ‘hits’ (Additional file 1: Table S8) indicated that the genes whose knockout increased cellular resistance to KHS101 were mainly involved in regulating mitochondrial structure and activity (Fig. 7B). The COX5B gene, a subunit of the cytochrome c oxidase, was the ‘top hit’. In order to independently validate this finding, we knocked out COX5B by two independent gRNAs in A549 cells, and in both cases, we found the emergence of a more KHS101-resistant phenotype (Fig. 7C). COX5B knock-out cells presented a reduction in their basal respiration and ATP-linked respiration (Fig. 7D), strengthening the notion that KHS101 sensitivity is linked with cellular oxidative activity, in line with the results obtained from creatine metabolism. Conversely, blocking glycolysis by 2-deoxy-D-glucose (2-DG) treatment significantly enhanced KHS101 sensitivity in different cell lines (Fig. 7E), further confirming that cancer cells respond to KHS101 proportionally to their OXPHOS-dependence. Therefore, HSPD1 targeting induces a mitochondrial metabolic breakdown in NSCLC, which results in loss of growth potential and cell death, and the small molecule KHS101 is particularly active in OXPHOS-dependent cells.