The HCT116 cell line (CCL-247) was purchased from the American Type Culture Collection (ATCC) and cultured in McCoy's 5A medium (cat. no. 26600023, Gibco; Thermo Fisher Scientific, Inc.), supplemented with 10% fetal bovine serum (cat. no. 10270106, Gibco; Thermo Fisher Scientific, Inc.), 1% glutamine (cat. no. 25030024, Gibco; Thermo Fisher Scientific, Inc.) and 1% penicillin and streptomycin (cat. N 15140122, Gibco; Thermo Fisher Scientific, Inc.). The 293T cell line (CRL-3216, ATCC) was cultured in Iscove's modified Dulbecco's medium (IMDM cat. no. 12440053, Gibco; Thermo Fisher Scientific, Inc.) supplemented with 10% heat-inactivated fetal calf serum (cat. no. S11450H, R&D Systems, Inc.), 1% glutamine (cat. no. 25030024, Gibco; Thermo Fisher Scientific, Inc.) and 1% penicillin and streptomycin (cat. no. 15140122, Gibco; Thermo Fisher Scientific, Inc.). The 293T cell line, characterized by high transfectability and transgenic expression, was used as a packaging cell line for generating viral particles. The authenticity of the cell lines was verified at the beginning of the study by STR profiling, in accordance with ATCC product description. Cell lines were routinely tested for mycoplasma contamination using the LookOut^® Mycoplasma PCR Detection kit (cat. no. MP0035, MilliporeSigma). All cell lines were grown and maintained in a humidified incubator with 5% CO₂ and 95% (vol/vol) O₂. Hypoxic experiments were performed using the Galaxy 48-R incubator (New Brunswick Scientific Co., Inc.; Effendorf AG) at 0.5% O₂. HCT116 cells were also treated with hypoxia-mimic compound deferoxamine (cat. no. D9533, MilliporeSigma) at 250 µM for 6 h as a positive control of hypoxic experimental conditions.

TRAP1 transient silencing was performed using 80 nM siRNAs purchased from Qiagen (target sequence, 5′-CCC GGT CCC TGT ACT CAG AAA-3′; sense strand, 5′-CGG UCC CUG UAC UCAG AAA TT-3′; antisense strand, 5′-UUU CUG AGU ACA GGG ACC GGG-3′; cat. no. SI00115150, Qiagen GmbH). As a control, cells were transfected with the same amount of control siRNA (target sequence, N/A; sense and antisense sequence, proprietary; cat. no. SI03650318, Qiagen GmbH). Transient transfections of siRNAs were performed 48 h before the experiment using the HiPerFect Transfection Reagent (cat. no. 301705, Qiagen GmbH), according to the manufacturer's protocol.

mTOR inhibition in HCT116 cells was performed using everolimus (Afinitor, Novartis) at 1 µM for 24 h.

Glucose uptake and lactate production were assessed as previously described (23). All experiments were performed at least in triplicate.

Total RNA was extracted using the RNeasy Plus Mini kit (cat. no. 74034, Qiagen GmbH), according to the manufacturer's instructions and reverse transcribed into cDNA using reverse transcription reagents (Superscript IV Vilo, cat. no. 11756050, Invitrogen; Thermo Fisher Scientific, Inc.), according to the manufacturer's instructions. qPCR analyses were performed in the CFX96 Touch™ Real-Time PCR Detection System (Bio-Rad Laboratories, Inc.) using SsoAdvanced™ Universal SYBR^®-Green Supermix (cat. no. 1725271, Bio-Rad Laboratories, Inc.), according to the manufacturer's instructions. The reaction was conducted according to the following amplification protocol: 95°C for 3 min, 39 cycles at 95°C for 10 sec, 60°C for 30 sec and 65-95°C for 5 sec. Primers were purchased from MilliporeSigma or Invitrogen (Thermo Fisher Scientific, Inc.) and are listed in Table SI. Gene expression was analyzed according to 2^−ΔΔCq relative quantification method (26) using CFX Maestro software version 1.0 (Bio-Rad Laboratories, Inc.).

Cell pellets were lysed in ice-cold RIPA buffer [20 mmol/l Tris (pH 7.5) containing 300 mmol/l sucrose, 60 mmol/l KC1, 15 mmol/l NaC1, 5% (v/v) glycerol, 2 mmol/l EDTA, 1% (v/v) Triton X100, 1 mmol/l phenylmethylsulfonylfluoride, 2 mg/ml aprotinin, 2 mg/ml leupeptin and 0.2% (w/v) deoxycholate] as previously described (27). The protein concentration was measured using the Bio-Rad protein assay kit (cat. no. 5000006, Bio-Rad Laboratories, Inc.), according to the manufacturer's instructions. Samples were resolved by SDS-PAGE using polyacrylamide 4-20% precast gels (Mini-PROTEAN TGX Stain-Free Gels, cat. no. 4568094, Bio-Rad Laboratories, Inc.) and transferred onto a nitrocellulose membrane (Trans-Blot Turbo Transfer Pack, cat. no. 1704158, Bio-Rad Laboratories, Inc.). The membrane was incubated for 60 min at room temperature with Western Blocker Solution (cat. no. W0138, MilliporeSigma) and immunoblotted with the following antibodies: Anti-TRAP1 (1:1,000 overnight 4°C; cat. no. sc-73604, Santa Cruz Biotechnology, Inc.), anti-HIF-1α (1:2,000 overnight 4°C; cat. no. ab16066, Abcam), anti-monocarboxylate transporter 4 (MCT4) (1:1,000 overnight 4°C; cat. no. sc-376140, Santa Cruz Biotechnology, Inc.), anti-GLUT1 (1:1,000 overnight 4°C; cat. no. ab32551, Abcam), anti-LDHA (1:1,000 overnight 4°C; cat. no. sc-137243, Santa Cruz Biotechnology, Inc.), anti-p70S6 kinase α (H-9) (1:1,000 overnight 4°C; cat. no. sc-8418, Santa Cruz Biotechnology, Inc.), anti-phospho-p70 S6 kinase (Thr389) (1:1,000 overnight 4°C; cat. no. 9205, Cell Signaling Technology, Inc.) and β-actin (1:2,000 1 h room temperature; cat. no. sc-47778, Santa Cruz Biotechnology, Inc.). The expression of specific proteins was detected using a secondary antibody labeled with peroxidase 1:2,000 for 1 h room temperature [goat anti-mouse (H + L)-HRP conjugate, cat. no. 1706516, Bio-Rad Laboratories, Inc.; goat anti-rabbit (H + L)-HRP conjugate, cat. no. 1706515, Bio-Rad Laboratories, Inc.) and the Clarity Western ECL Substrate (cat. no. 1705061, Bio-Rad Laboratories, Inc.). Protein expression levels were quantified using densitometric analysis, using ImageJ software v1.53e (National Institutes of Health) and normalized according to the expression of housekeeping genes.

Human p9T CRC organoids were generated by the Jacco van Rheenen Research Group at the Department of Molecular Pathology of Netherlands Cancer Institute of Amsterdam (NKI-AVL) and kindly provided to the authors' laboratory. Organoids were cultured in a humidified atmosphere at 37°C and 5% CO₂ with basal culture medium [Advanced DMEM/F12 (cat. no. 12634028, Gibco; Thermo Fisher Scientific, Inc.), 10 mM HEPES (cat. no. 15630056, Gibco; Thermo Fisher Scientific, Inc.), 2 mM Glutamax (cat. no. 35050061, Gibco; Thermo Fisher Scientific, Inc.), 100 U/ml penicillin and streptomycin (cat. no. 15140122, Gibco; Thermo Fisher Scientific, Inc.)] supplemented with 10% R-Spondin conditioned medium, 10% Noggin conditioned medium, 1X B27 (cat. no. 17504001, Gibco; Thermo Fisher Scientific, Inc.), 1.25 mM n-acetyl cysteine (cat. no. A9165, MilliporeSigma), 10 mM nicotinamide (cat. no. N0636, MilliporeSigma), 500 nM A83-01 (cat. no. SML0788, MilliporeSigma), 3 µM SB202190 (cat. no. CAY-10010399, Vinci-Biochem), 50 ng/ml human recombinant EGF (cat. no. 354052, BD Biosciences), 10 nM Leu15-Gastrin (cat. no. G9145, MilliporeSigma) and 10 nM prostaglandin E2 (cat. no. 14010, Vinci-Biochem). To confirm correct sample identity, the organoids were regularly tested by SNP analysis. Four different validated lentiviral reporter plasmids (pLKO.1_TRC cloning vector, plasmid#10878, Addgene, Inc.) for shTRAP1 were obtained from the NKI's Robotics and Screening Center Facility and tested to verify the silencing efficiency induced in CRC HCT116 cell line and organoids. Only two of these (shTRAP1#1 and shTRAP1#3) were selected for their high silencing efficiency and reduced mortality. The full hairpin sequences were as follows: shTRAP1#1, CCG GCC GCT ACA CCC TGC ACT ATA ACT CGA GTT ATA GTG CAG GGT GTA GCG GTT TTT G; and shTRAP1#3, CCG GCA GAG CAC TCA CCC TAC TAT GCT CGA GCA TAG TAG GGT GAG TGC TCT GTT TTT G.

A 3rd generation lentiviral system was used, that includes pVSV-G (pMD2.G; plasmid #12259, Addgene, Inc.), pMDLg/pRRE (plasmid #12251, Addgene, Inc.) and pREV AmpR (pRSV-Rev; plasmid #12253, Addgene, Inc.) plasmids. The propagation of plasmids was performed in the bacterial strain DH5a into LB medium. DNA was isolated using the PureLink™ HiPure Plasmid Midiprep kit (cat. no. K210004, Invitrogen; Thermo Fisher Scientific, Inc.).

The 293T cells were transfected with the pLKO.1 vector coding for shTRAP1 using the calcium phosphate-DNA co-precipitation technique. The cell supernatant containing viral particles was collected, filtered and directly used for cell transduction, or ultracentrifuged (49,100 × g) for 2 h at 7°C for organoid transduction. Lentiviral titers were determined using the qPCR Lentivirus Titration kit (LV900, Applied Biological Materials, Inc.), following the manufacturer's instructions. For the experiments, the amount of lentiviral supernatant used was calculated to achieve the multiplicity of infection (MOI) of 50. To ensure efficient transduction, the 293T and HCT116 cells were incubated with lentiviral supernatants at 37°C for 24 h in the presence of polybrene (8 µg/ml, cat. no. TR-1003, MilliporeSigma). Antibiotic selection was initiated at 24 h post-transduction and was carried out for 5 consecutive days.

Tumor-derived p9T organoids were trypsinized using TrypLE (cat. no. 12605010, Gibco; Thermo Fisher Scientific, Inc.), mixed with 100 µl concentrated virus, 8 ng/ml polybrene and 10 µM Rho-associated, coiled-coil containing protein kinase inhibitor (Y-27632; cat. no. Y0503, MilliporeSigma) in 15 ml tubes and centrifuged at 600 × g for 1 h at 32°C. Subsequently, the organoids were incubated at 37°C and 5% CO₂ for 6 h. Following incubation, the organoids were washed with Advance DMEM/F12 medium and spun down for 5 min at 160 × g 4°C to eliminate the virus. Following the removal of the virus-containing medium, organoids were plated in ~50 µl BME (cat. no. 353300502, Bio-Techne Corporation). Transduced organoids were grown in Advance DMEM/12 full of growth factors and fresh Y-27632 for 3 days after which selection was applied by the addition of puromycin (2 µg/ml, cat. no. A1113802, Thermo Fisher Scientific, Inc.) for 4 days.

Total RNA (300 ng) was transcribed for the synthesis of cDNA and biotinylated cRNA, according to the protocol of the Illumina TotalPrep RNA (cat. no. AMIL1791, Ambion; Thermo Fisher Scientific, Inc.) amplification kit. The hybridization, marking and scanning of 750 ng cRNA was performed on the Illumina Human HT12 v4.0 Expression BeadChip array (Illumina Inc.), following the standard protocol. All analyses were performed in triplicate for each sample. The BeadChip was then dried and scanned using the Illumina HiScanHQ system (Illumina Inc.). Data analysis was carried out through the free/open source environment R/Bioconductor (28). Probes with a low quality of fluorescence intensity signal were excluded from the subsequent analysis steps. Normalization was performed using the neqc procedure, a background correction through internal controls followed by Quantile Normalization. The differential analysis between the experimental conditions and the normalization were performed using the limma package (29).

Functional enrichment analysis was carried out through the Gene Set Enrichment Analysis (GSEA) (30) computational method and the MsigDB database (31) for the collection of annotated gene sets, focusing on the following categories: Hallmark, Gene Ontology and Pathways. The gene expression data generated in the present study have been deposited in the ArrayExpress database at EMBL-EBI (www.ebi.ac.uk/array-express) under the accession no. E-MTAB-10563.

TRAP1 expression analysis in CRC samples, compared with normal tissue, was performed using the TNMplot database (https://www.tnmplot.com/). The platform directly compares tumor and normal samples and performs a Mann-Whitney U or Kruskal-Wallis tests or a paired Wilcoxon test (in case of availability of paired normal and adjacent tumor) for statistical significance (32).

The unpaired t-test was performed using the t-test Calculator of GraphPad (online version) for the analysis of metabolic tests in silenced cells and related controls (Fig. 1C and D). Data are reported as mean values of least three independent experiments (± SD). RT-qPCR data were analyzed using two-way ANOVA and Sidak's multiple comparisons test. Statistical analysis was performed using GraphPad 7.0 software (GraphPad Software, Inc.). A value of P<0.05 was considered to indicate a statistically significant difference.

In the preliminary analyses, TNMplot gene chip data were used to confirm the increased expression of TRAP1 in human CRC samples (n=160) compared to normal human colorectal mucosa (n=160). Notably, while the normal colorectal mucosa was characterized by a low TRAP1 mRNA expression, a significant upregulation of TRAP1 expression was observed in malignant tissues (Fig. S1A) (P<0.001 Kruskal-Wallis test). Considering that hypoxia is the main stimulus to boost glycolytic metabolism in cancer cells (12,33-36), in order to establish the role of TRAP1 in hypoxia-induced metabolic rewiring, human CRC cells were used in the present study and in preliminary experiments; the optimal time for HCT116 cell exposure to 0.5% hypoxia to induce HIF-1α stabilization was evaluated. Western blot analysis revealed that HIF-1α maximal stabilization was observed after 4 h of exposure of the cells to 0.5% O₂ compared to normoxic conditions (20% O₂; Fig. 1A). In these experiments, HCT116 cells exposed to 250 µM deferoxamine for 6 h were used as a positive control.

To examine the role of TRAP1 in controlling HIF-1α stabilization in CRC, HCT116 cells were transiently silenced for TRAP1 and exposed to 4 h of hypoxia or normoxia and evaluated for HIF-1α stabilization using western blot analysis. Of note, although there were no significant differences in HIF-1α gene expression under the current experimental conditions (Fig. 1B, left panel), TRAP1 silencing induced a partial inhibition of HIF-1α stabilization at the post-translational level under hypoxic conditions (Fig. 1B, right panel).

In subsequent experiments, the association between TRAP1-mediated HIF-1α stabilization and its role in metabolic adaptive responses to reduced oxygen availability was investigated. Thus, TRAP1-silenced HCT116 cells were exposed to 0.5 and 20% O₂ for 4 h and analyzed for glucose uptake and lactate production. As was expected, hypoxia markedly enhanced both 2-DG uptake (Fig. 1C) and lactate production (Fig. 1D) in a high-TRAP1 background (siNEG). Of note, TRAP1 silencing significantly decreased 2-DG uptake (Fig. 1C) and partially impaired lactate production (Fig. 1D) under hypoxic conditions, thus suggesting that TRAP1 is required for hypoxia-induced metabolic rewiring.

To functionally characterize TRAP1-dependent HIF-1α stabilization in CRC, the role of TRAP1 in the HIF-1α-mediated transcriptional reprogramming of metabolic genes under hypoxic conditions was evaluated. HCT116 cells transiently silenced for TRAP1 were incubated for 8 and 24 h at 20 and 0.5% O₂ and assayed for the expression of both TRAP1 (Fig. S1B) and HIF-1α-inducible genes commonly involved in metabolic and angiogenic pathways (i.e., GLUT1, GLUT3, VEGFA, LDHA, PDK1, PKM2) (Fig. 2). Of note, TRAP1-silenced HCT116 cells failed to significantly activate the expression of some of the evaluated genes after 8 h of hypoxia compared to the control cells; the differential expression of GLUT3 (P<0.01) and PDK1 (P<0.001) was statistically significant (Fig. 2A). Consistently, TRAP1 silencing resulted in the reduced expression of all tested genes under 24 h of hypoxia (P<0.001; Fig. 2B). Under normoxic conditions, no significant differences were observed between the TRAP1-silenced and control cells at all time points, with similar expression levels observed for all tested genes (Fig. 2A and B). In parallel experiments, western blot analysis was used to evaluate HIF-1α stabilization and the expression of certain HIF-1α-target genes, i.e., GLUT1, LDHA and MCT4 in a high vs. low TRAP1 background under hypoxic conditions (Fig. 2C). Consistent with the RT-qPCR data, TRAP1 silencing led to the lower stabilization of HIF-1α and to lower levels of GLUT1, LDHA and MCT4 under hypoxic conditions compared to the control cells.

These experiments were further extended to a 3D-model of human CRC organoids, based on the rationale that the growth of organoids within a 3D matrix favors the spontaneous generation of hypoxic areas, as observed in tumors (37). For these experiments, pMOCK and TRAP1-silenced p9T organoids were generated. Four different shTRAP1 lentiviruses (#1, #2, #3 and #4) were first tested for the efficiency of transfection in both 293T and HCT116 cells evaluating TRAP1 silencing using RT-qPCR and western blot analysis (Fig. S1C, left and right panels, respectively), and shTRAP1 #1 and shTRAP1 #3 lentiviruses were selected for their highest silencing effectiveness. In subsequent experiments, shTRAP1 #1 and shTRAP1 #3 were tested for the efficiency of transduction in p9T organoids (Fig. S1D) and shTRAP1 #1 p9T organoids were used for further experiments aimed at validating the role of TRAP1 in the regulation of HIF-1α transcriptional activity. pMOCK and TRAP1-silenced p9T organoids were incubated for 8 and 24 h at 0.5 and 20% O₂ and analyzed using RT-qPCR for TRAP1 (Fig. S1E) and HIF-1α inducible genes (Fig. 3) expression levels. The majority of genes exhibited a significant and progressive upregulation in a TRAP1 high expression background under hypoxic conditions starting from 8 h (Fig. 3A) and reaching the highest level at 24 h (Fig. 3B). Notably, TRAP1 silencing resulted in a decreased expression of all tested genes under hypoxic conditions (Fig. 3), with maximal statistical significance at 24 h, as observed in the HCT116 cells.

As data described above demonstrating a role of TRAP1 in hypoxia-induced gene expression reprogramming, a whole genome gene expression profiling was performed in TRAP1-silenced and control HCT116 cells exposed to normoxia and hypoxia for 24 h. As was expected, oxygen deprivation promoted significant gene expression reprogramming with 2,665 downregulated and 2,041 upregulated genes compared to the normoxic control (abs logFC ≥0.60, P<0.05). Furthermore, 1,071 genes were differentially expressed (347 downregulated and 724 upregulated) in TRAP1-silenced compared to control HCT116 cells under hypoxic conditions (abs logFC ≥0.60, P<0.05) (Table SII).

GSEA was performed to obtain functional enrichments of differentially expressed genes focusing on the following categories: Hallmark, Gene Ontology (GO) and Pathway (abs enrichment score ≥0.60, P<0.05) (Table SIII). Although very few Hallmarks were enriched, the two most upregulated Hallmarks under hypoxic conditions were HYPOXIA and ANGIOGENESIS, confirming that, under the current experimental conditions, hypoxic stress regulated gene expression to promote oncogenic and angiogenetic pathways (10-16). Furthermore, several GOs and Pathways were functionally regulated in our experimental conditions. In particular, 453 (227 downregulated and 226 upregulated) and 201 (62 downregulated and 139 upregulated) GOs were enriched comparing hypoxia (siNEG 0.5% O₂) vs. normoxia (siNEG 20% O₂) and a low vs. high TRAP1 background under hypoxic conditions (siTRAP1 0.5% O₂ vs. siNEG 0.5% O₂), respectively. Among these, 58 enriched GOs were present in both datasets and most of these exhibited an opposite regulation after TRAP1 silencing under hypoxic conditions, being up/downregulated in response to hypoxia in a high TRAP1 background and down/upregulated upon TRAP1 silencing under hypoxic conditions (Table SIV).

Focusing on the top five more significantly enriched GOs in siTRAP1 0.5% O₂ vs. siNEG 0.5% O₂, a positive regulation of ribosomal small subunit biogenesis emerged, including the co-transcriptional assembly of tricistronic pre-rRNA into the small subunit (SSU) processome and the subsequent maturation of SSU rRNA 18S, contained in the 40S subunit of the mature ribosome (Fig. 4A). Noteworthy, these GOs were negatively regulated by hypoxia and their downregulation was prevented by TRAP1 silencing under hypoxic conditions. Instead, none of these GOs was enriched following TRAP1 silencing under normoxic conditions (siTRAP1 20% O₂ vs. siNEG 20% O₂).

To validate the possible role of TRAP1 in sustaining the ribosome biogenesis inhibition under hypoxic conditions, six genes were selected (i.e., IMP3, LTV1, PWP2, RCL1, RRS1 and UTP3) commonly involved in different steps of ribosome biogenesis and maturation and differentially expressed in the current experimental conditions, thus contributing to the enrichment of previous GO categories (abs logFC ≥0.60, P<0.05). The expression levels of these genes were evaluated using RT-qPCR in both HCT116 cells and p9T organoids silenced or not for TRAP1 and exposed to hypoxia for 24 and 8 h, respectively (Fig. 4B and C). These time points were selected based on data reported in Figs. 2 and 3, demonstrating an earlier transcriptional reprogramming in response to hypoxia in CRC organoids. No difference in expression was observed for all tested genes in TRAP1-silenced cells (Fig. 4B) and shTRAP1#1 p9T organoids (Fig. 4C) compared to controls under normoxic conditions; this was consistent with the GSEA data demonstrating no enrichment of ribosome biogenesis/assembly GO categories under normoxic conditions. Conversely, as was expected, the hypoxic stimulus significantly suppressed the expression of these genes in both the CRC experimental models and TRAP1 silencing attenuated the inhibitory effects of hypoxia by promoting a higher expression of these genes under hypoxic conditions (Fig. 4B and C). Two genes (i.e., PWP2 and RCL1) were undetectable in p9T organoids and, thus, are not illustrated in Fig. 4.

Based on the hypothesis that the TRAP1 modulation of ribosome biogenesis under hypoxic conditions may involve the mTORC1/p70S6K pathway, western blot analysis was performed to evaluate the phosphorylation status of p70S6K, a master regulator of protein synthesis and ribosome biogenesis, in TRAP1-silenced HCT116 cells exposed to hypoxia and treated with the mTOR inhibitor everolimus at 1 µM for 24 h (Fig. 5A). Unlike oxygen deprivation that suppressed p70S6K phosphorylation, TRAP1 silencing enhanced p70S6K activation/phosphorylation compared to the respective controls and this occurred both under normoxic and hypoxic conditions. Moreover, the mTOR inhibitor, everolimus, hampered p70S6K phosphorylation under normoxic and hypoxic conditions, and in high and low TRAP1 expression backgrounds (Fig. 5A).

To corroborate the link between the TRAP1 regulation of the mTOR pathway and its control of ribosome biogenesis under hypoxic conditions, four ribosome biogenesis-related genes (i.e., IMP3, LTV1, RRS1 and UTP3) were further evaluated using RT-qPCR in TRAP1-silenced cells exposed to normoxia and hypoxia for 24 h in the presence or absence of everolimus (Fig. 5B). Data are reported as the fold change (FC) increase in gene expression in TRAP1-silenced cells compared to the respective control under normoxic and hypoxic conditions, and in the presence or absence of everolimus. As observed in Fig. 4A, TRAP1 silencing resulted in the upregulation of the expression of these genes under hypoxic conditions, and this occurred in parallel with mTOR signaling activation (Fig. 5A). Of note, mTOR signaling inhibition by everolimus in TRAP1-silenced cells attenuated the upregulation of these genes (Fig. 5B). According to previously described data, no significant differences (FC <1.5) were observed after TRAP1 silencing under normoxic conditions (siTRAP1 20% O₂ vs. siNEG 20% O₂).