A TRIM21-based bioPROTAC highlights the therapeutic benefit of HuR degradation – Nature Communications

Ethical statement

These studies were conducted within the remit of a project licence approved by local Animal Welfare and Ethical Review Board (AWERB) committee and under a U.K. Home Office Project Licence in accordance with the U.K. Animals (Scientific Procedures) Act 1986 and EU Directive EU 2010/63/EU. Studies were performed according to the Home Office guidelines for the Care and Use of Laboratory Animals, and were also compliant with AstraZeneca policies on Bioethics and Good Statistical Practice in animal work.

Expression and purification of recombinant HuR

pET24a vectors encoding HuR (full-length—AA 1–326; RNA-recognition motif (RRM) 1 and 2 (RRM1 + 2)—AA 11–186 and RRM1—AA 11–98) with an Avi-tag and TEV-cleavable His-tag were used for protein expression in Escherichia coli BL21*λDE3 (New England Biolabs) following induction with a final concentration of 1 mM IPTG. Protein was purified by immobilised metal affinity chromatography (IMAC) and size-exclusion chromatography (SEC) with a HiLoad 16/600 Superdex 75 pg column (Cytiva), according to standard methods. Proteins were validated by SDS-PAGE and electrospray ionisation mass spectrometry (ESI-MS).

Identification and confirmation of HuR-VHH-binding domains

HuR-specific-phage VHH were selected by performing three rounds of panning on biotinylated HuR protein (HuR-RNA-recognition motif 1 (RRM1) and RNA-recognition motif 2 (RRM2) (RRM1 + RRM2) or HuR RRM1 only) using a Llamda® phage display VHH library (composed of four sub-libraries)75,76 built using Colibra® technology75, and licenced from Isogenica – strategy outlined in Supplementary Data 8. Separately, a control Cas9 specific-phage-VHH was also isolated from the aforementioned Llamda® phage display VHH library by panning the full-length Cas9 (without bound RNA or DNA).

In brief, for each round of selection, bound phage-antigen complexes were captured on Dynabeads™ M-280 Streptavidin (Invitrogen), with specific-phage-VHH eluted in 10 µg/ml trypsin. E. coli TG1 were infected with eluted phage-VHH and plated on 2TY + ampicillin + glucose (2TYAG) bioassay plates. The following day, colonies were collected in 2TYA media and used to initiate the next round of panning by inoculating 2TYAG media. Cultures were infected with M13 helper phage prior to centrifugation and re-suspension of the pellet in 2TY + ampicillin + kanamycin (2TYAK) for overnight incubation. For the third round of selection, individual colonies were cultured in 2TYA and the following day deep-well blocks containing 2TYAG media were inoculated. IPTG was added at a final concentration of 1 mM for phage-VHH expression.

Cell pellets were collected via centrifugation and resuspended in BugBuster (Merck) prior to the transfer of soluble protein lysate supernatant (expressed phage VHH) to a fresh plate. MaxiSorp™ ELISA plates were prepared with 50 µl biotinylated HuR (RRM1 only or RRM1 + 2) (1–2 µg/ml) following standard protocol. 50 µl soluble protein lysate (expressed VHH) was added, followed by 50 µl rabbit anti-myc-HRP (Abcam) in 3% skimmed milk. Fifty microlitres 3,3′−5′5-tetramethylbenzidine (TMB) substrate (Merck) was then incubated for 2–5 min before the addition of 50 µl 0.5 M sulphuric acid. Optical density was measured at 450 nm using an EnVision plate reader (PerkinElmer). Positive HuR-VHH (VHHHuR) binders were selected for sequencing analysis. Unique clones were prepared for larger scale expression and ELISA titrations against HuR RRM1.

Plasmids for VHH characterisation and bioPROTAC evaluation

Twenty-four lead phage-VHH were cloned into a pFC14K FLAG-HaloTag® mammalian expression vector via PCR amplification and subsequent sub-cloning. VHH were located N-terminally to tags. For co-immunoprecipitations, the full-length HuR antigen was cloned upstream of emGFP in pTuner vector, as a regular mammalian expression vector by removing the cassette responsible for controlling expression. For use in NanoBRET, full-length HuR was cloned into the pFN31K Nluc CMV-neo Flexi® vector (Promega).

To generate the VHH-Fc fusion constructs, the VHHGFP_4 sequence (Addgene plasmid #35579)77 – as a non-HuR-targeting control—or VHHHuR_17 were cloned alongside the hIgG1-Fc coding sequence from pFuse-hIgG1-Fc1 (InvivoGen) into the pGEM-HE vector giving rise to a pGEMHE-VHH-hIgG1-Fc1. A Fc mutant (H433A) that cannot bind TRIM21 was generated by site-directed mutagenesis (Agilent). For the bioPROTAC constructs, the sequence encoding the TRIM21 Ring-B-box-coiled-coil (T21RBCC) (AA 1-255) and VHHHuR_17 or VHHGFP 77 were integrated in a cassette with a T7 promoter containing AG initiator sequences, suitable UTR for use in in vitro mRNA transcription78, Kozak consensus sequence and a C-terminal HA-tag. Constructs were generated in both an N- and C-orientation, or for the VHHHuR or VHHGFP domains alone. All cassettes were contained within a pcDNA 3.1(+) backbone (Invitrogen). For the T21RBCCΔRING, sub-cloning was completed.

For stable inducible expression of bioPROTACs, the ObLiGaRe doxycycline-inducible (ODIn) system was utilised48. This system requires two vectors; a pZFN1-T2A-ZFN2-AAVS construct encoding two zinc finger nucleases (ZFN) targeting the AAVS locus and a pBSK construct housing the AAVS1 locus, Tet-On 3G inducible expression system, neomycin-resistance gene and cassette encoding the transgene (T21RBCC-VHHHuR, T21RBCC-VHHGFP or VHHHuR) followed by a T2A peptide sequence and mCherry reporter.

All constructs were confirmed via sequencing analysis.

In vitro transcription

Prior to in vitro mRNA transcription, pcGEMHE or pcDNA 3.1 constructs were linearised via PCR 5’-capped modified RNA was synthesised according to the manufacturer’s protocol using HiScribe™ T7 ARCA mRNA Kit (New England Biolabs) or HiScribe™ T7 High Yield RNA Synthesis Kit (New England Biolabs), CleanCap® (Trilink) and 5-Methoxyuridine-5′-Triphosphate (5-moUTP) (40% final concentration) (Trilink) respectively. mRNA was purified using the MEGAclear kit (ThermoFisher Scientific) prior to quality control on the 2100 Bioanalyzer (Agilent).

Cell lines

The human HCT116 colorectal carcinoma cell line was obtained from the ECACC and the human A549 lung carcinoma, human U2OS osteosarcoma and human retinal pigment epithelial-1 (RPE-1) cell lines were obtained from the ATCC. All cell lines underwent short tandem repeat (STR) profiling prior to use and regular Mycoplasma screening. Cell lines were routinely passaged in DMEM medium (ThermoFIsher Scientific) supplemented with 10% foetal bovine serum (FBS), penicillin (105 U/L), and streptomycin (100 mg/L), and maintained at 37 °C and 5% CO2 in a humidified environment.

Transient transfection

For transient DNA expression, cells were transfected with FuGene HD transfection reagent (Promega) following the manufacturer’s protocol.

For transient mRNA expression, cells underwent electroporation or chemical transfection. Electroporation was used delivery of mRNA encoding Fc-fusions and was performed using the Neon® Transfection System (ThermoFisher). Cells were washed with PBS and resuspended in Buffer R (ThermoFisher) at a concentration of 8 × 107 cells/ml. For each electroporation reaction 8 × 105 cells (10.5 µl) were mixed with 2 µl of antibody or mRNA or protein to be delivered (0.5 µM). This mixture was taken up into a 10 µl Neon® Pipette Tip (ThermoFisher) and electroporated using the following settings: 1400 V, 20 ms, 2 pulses. Electroporated cells were transferred to medium supplemented with 10% FCS without antibiotics. Transfection of T21RBCC constructs was undertaken via reverse transfection of mRNA (0.5 µg/ml) using Lipofectamine RNAiMAX transfection reagent (ThermoFisher Scientific) according to the manufacturer’s protocol.

Stable cell line generation

For cell line generation using the ODIn system48, HCT116 cells were co-transfected with the ODIn vector (for either T21RBCC-VHHHuR, T21RBCC-VHHGFP or VHHHuR) and ZFN-AAVS vector and at a 2:1 ratio using FuGENE HD (Promega). For transgene selection integration, confluent cells were treated with G418 (500 µg/ml) (Sigma) for 10 days. For the induction of transgene expression, cells were treated with a final concentration of 100 ng/ml doxycycline for 24 h. Clonal selection was completed by fluorescence-activated cell sorting (FACS) of mCherry-expressing cells using the BD FACSAria™ II (BD Bioscience). Monoclonal populations were expanded and validated via immunoblotting and immunofluorescence.

MG132 treatment

Twenty-four hours post seeding, the HCT116 ODIn- T21RBCC-VHHHuR inducible cell line was co-treated with 5 µM MG132 and doxycycline (100 ng/ml) for 24 h.


To confirm VHH-HuR interactions, the A549 cell line was seeded in T25 flasks and co-transfected with 150 ng VHH-FLAG-HaloTag® (lead VHHHuR clones or a VHHCas9 control) and emGFP-HuR. Forty-eight hours later, cells were lysed in NP40 lysis buffer, and supernatants were collected for subsequent quantification via a BCA assay (ThermoFisher Scientific). Two hundred and fifty micrograms cell lysate was prepared in 250 µl lysis buffer then incubated with 5 µl Protein G (for anti-FLAG IP) or Protein A (for anti-HuR IP) Dynabeads (ThermoFisher) for 1 h at room temperature. Lysates were collected and incubated with anti-FLAG M2 Magnetic Beads (Sigma) or pre-prepared beads with anti-HuR antibody (Cell Signaling Technology) overnight at 4 °C. The following day, lysates were removed and beads washed with high salt wash buffer (1% NP40, 50 mM Tris HCl pH 7.5, 300 mM NaCl, 1 mM EGTA, 1 mM EDTA, 10 mM glycerophosphate, 50 mM sodium fluoride, 0.27 M sucrose, 5 mM sodium pyrophosphate, 1 mM sodium orthoVanadate). Bead-bound proteins were eluted in Laemmli buffer containing 10% β-mercaptoethanol. Lysate inputs and eluates were analysed via immunoblotting, as described below.


Further validation of the VHHHuR_8 and VHHHuR_17 interaction with HuR was completed using NanoBRET. HCT116 cells were seeded in 12-well plates and co-transfected with 0.5 ng HuR-NanoLuc (donor) and VHH-FLAG-HaloTag® (for VHHHuR clones or a VHHCas9 control) (acceptor), at a twofold serial dilution series, starting at 50 ng and a donor:acceptor ratio of 1:100. To maintain total DNA transfected, an empty pTuner vector was transfected for a total of 55 ng/condition. Twenty-four hours post transfection, a final concentration of 0.1 µM HaloTag® NanoBRET™ 618 Ligand (Promega) was added to cells. The following day, luciferase substrate was diluted in media (×166) and added for 10 min, then donor (450 mm) and acceptor (610 mm) emissions were measured using a luminometer.

Determination of VHHHuR potency

Following induction with a final concentration of 1 mM IPTG, VHHHuR_8, VHHHuR_17 and VHHCas9 proteins were purified from E. coli TG1 by IMAC and SEC with a HiLoad 16/600 Superdex 75 pg column (Cytiva), according to standard methods. Proteins were validated by SDS-PAGE and peptide mass finger printing.

To determine the VHHHuR binding constants (KD) to HuR, SPR was completed. A series S streptavidin Biacore chip (Cytiva) was docked into a T200 Biacore instrument (Cytiva), and priming was completed with running buffer (10 mM HEPES, 150 mM NaCl, 3 mM EDTA and 0.05% P-20, Cytiva). Biotinylated HuR RRM1, HuR RRM1 + 2 or full-length HuR protein (20 mg/ml) was injected for 420 s to give immobilisation signals of 1100, 1700 and 4800 respectively. Samples were prepared in assay buffer in a 384-well polypropylene microplate. Seven, threefold dilutions of VHHHuR_8/17/18, with a top concentration of 900 nM, were injected for 60 s with a dissociation time of 4000 s. All data were double referenced and globally fitted to a 1:1 binding model using the Biacore T200 Evaluation software.

To calculate VHHHuR inhibition constants (Ki), fluorescence polarisation assays were completed using a Musashi RNA-binding protein 1 (Msi1)-FITC probe (5′-rGrCrU rUrUrU rArUrU rUrArU rUrUrU rG/3FluorT/− 3′). For confirmation of the probe KD, full-length HuR protein (0–2000 nM) in the assay buffer (0.01 M HEPES pH 7.4, 0.15 M NaCl, 3 mM EDTA, 0.05% v/v Tween®20) and 10 nM Msi1-FITC probe were incubated in a 384-well plate and read immediately on a PHERAStar plate reader. Probe KD was determined to be 0.0037 µM. To evaluate RNA competition, unlabelled RNA (5′-rGrCrU rUrUrU rArUrU rUrArU rUrUrU rG-3′), VHHHuR (VHHHuR_8 or VHHHuR_17) or VHHCas9 (0–30 µM) were added to a plate followed by 8 nM Msi1-FITC then 8 nM HuR protein. Samples were incubated at room temperature for 5 min prior to reading on a PHERAStar plate reader. Ki was calculated by forcing the fit to baseline and using a modified version of the Munson-Rodbard equation with probe KD, as calculated above.

For confirmation of VHHHuR_17 inhibition of RNA binding, a series S streptavidin Biacore chip (Cytiva) was docked into an 8K Biacore instrument (Cytiva), and priming was completed as above. Biotinylated HuR RRM1 + 2 protein (20 mg/ml) was injected for 420 s. The A-B-A function of the 8K was used to measure RNA binding in the presence or absence of VHHHuR_17 by injecting 50 nM RNA (Musashi RNA-binding protein 1 (Msi1)-FITC probe (5′-rGrCrU rUrUrU rArUrU rUrArU rUrUrU rG/3FluorT/-3′) (B) with 1 µM HuRVHH_17 as the flanking solution (A). 1 µM of VHHHuR_17 was injected for 500 s prior to RNA to ensure equilibrium was obtained.


Cells were seeded in 24-well plates then treated with mRNA or doxycycline (100 ng/ml). Following incubation, cells were washed in PBS and cell lysates were collected and prepared in 1x Laemmli sample buffer (BioRad) containing 10X Bolt™ Sample Reducing Agent containing 500 mM DTT (ThermoFisher Scientific). Samples were denatured at 95 °C prior to being resolved by SDS-PAGE and transferred onto PVDF membrane. Membranes were probed with antibodies outlined in the appendix, and proteins were detected by enhanced chemiluminescence (Amersham, GE Healthcare) and X-ray films or the GelDoc™ XR (BioRad), or visualised using an Odyssey DLx imaging system (LI-COR). Analyses were completed using ImageLab software (BioRad).


HCT116 cells were seeded in 384-well plates then transfected with mRNA encoding bioPROTACs. Following 18 h incubation, cells were washed in PBS, fixed for 15 min at room temperature in 4% methanol-free formaldehyde (ThermoFIsher Scientific) then blocked in 3% BSA and 0.1% Triton X-100 for 1 h. Cells were stained overnight with mouse anti-HuR (ThermoFisher Scientific) and rabbit anti-HA (Abcam) antibodies. Nuclei were stained with Hoechst (ThermoFisher Scientific) and the entire cell was stained with HCS CellMask™ Deep Red Stain (ThermoFisher Scientific). Cells were visualised using the CV7000 spinning disk confocal microscope (Yokogawa Inc.) using a 20× objective and 2 × 2 binning. Analyses were undertaken using Columbus software (PerkinElmer) to quantify HuR abundance. Due to some evidence of epitope competition between the VHHHuR and HuR antibody, immunofluorescence was only used as an orthogonal approach.

Cell viability analysis

Parental and ODIn-HCT116 cell lines were prepared in phenol-red free DMEM medium (ThermoFisher Scientific) in a 384-well plate. After 24 h, cell lines were treated with a final concentration of 100 ng/ml doxycycline, then incubated for a further 72 h. Cell viability was assessed using the MTT (Sigma) cell assay according to the manufacturers protocol, and absorbance was measured at 570 nm using an EnVision plate reader (PerkinElmer).

3D colony formation assay

Parental and ODIn-HCT116 cell lines were prepared in 0.3% UltraPure low melting point agarose (ThermoFisher Scientific) diluted 1:1 in DMEM medium (ThermoFisher Scientific), and placed in 96-well plates pre-coated with 0.7% UltraPure low melting point agarose (ThermoFisher Scientific). Once set, DMEM supplemented with 500 µg/ml geneticin and 100 ng/ml doxycycline was added. Cells were grown for eight days, with regular media changes, then stained with SigmaFast BCIP/NBT solution (Sigma). Cells were visualised and colonies (size: 70–400 µm) were counted using the GelCount imager (Oxford Optronix).

Proteomic sample preparation

For non-kinetic analyses, HCT116 cells were seeded in six-well plates and reverse transfected with mRNA encoding VHHHuR, T21RBCC-VHHHuR, VHHHuR-T21RBCC, VHHGFP, T21RBCC-VHHGFP and VHHGFP-T21RBCC. For kinetic analyses, the HCT116 ODIn- T21RBCC-VHHHuR cell line was seeded in a 6-well plate 24 h prior to doxycycline induction at 24 h intervals over a 72-h period. At the endpoint, 1.5 × 106 cells/condition were collected following trypsinisation, and washing in cold PBS. Cell pellets were resuspended in S-Trap lysis buffer (5% SDS, 50 mM triethylammonium bicarbonate (TEAB) buffer, pH 7.55) and solubilised using a Retsch mill (MM400) bead beater for 2 min at frequency 30 Hz. Protein concentration was measured using a BCA assay kit (Thermo Fisher). Fifty micrograms of protein lysates were digested using micro S-Trap method (Protifi.com) according to the manufacturer’s protocol79. Proteins were reduced using 20 mM tris(2-carboxyethyl)phosphine for 15 min at 60 °C, alkylated using 80 mM iodoacetamide for 1 h at room temperature, and digested on a micro S-Trap cartridge using mass spectrometry grade trypsin/lys-C (Promega) for 2 h at 47 °C. Trypsin/Lys-C digested peptides were eluted with 50 mM TEAB buffer, followed by 0.2% formic acid (FA) in water, and 50/50 acetonitrile/water with 0.2% FA. Eluted peptides were dried then reconstituted in 0.15% FA in water.

LC–MS/MS analysis

LC–MS/MS analysis was conducted on a timsTOF Pro mass spectrometer (Bruker) coupled with a nanoElute LC-system and nano-electrospray ion source (CaptiveSpray Source, Bruker). Samples were loaded onto a 15 cm × 75 µm, 1.9 ReproSil, C18 column (PepSep.com) maintained at 50 °C. The peptides were separated using a gradient generated using solvent A (composed of 0.15% FA in water), and solvent B (composed of 0.15% FA in acetonitrile). Peptides were eluted at a flow rate of 500 nl/min over a 51 min gradient, from 4–24% solvent B (36 min), 24–36% solvent B (7 min), 36–64% solvent B (5 min), and 64– 98% solvent B (3 min). Data-dependent acquisition (DDA) was performed in PASEF mode with six PASEF scans at a duty cycle close to 100%. MS acquisition recorded spectra from 100-1600 m/z and ion mobility was scanned from 0.85–1.30 Vs/cm2 over a ramp time of 100 ms. The total duty cycle time was 1.15 s. The collision energy was linearly increased from 27 to 45 eV as a function of ion mobility. An active exclusion of 0.4 min was applied to precursors that reach a target intensity of 20,000 units. Data-independent acquisition (DIA)-PASEF mode was performed with a scheme that consists of two rows of 32 windows (eight PASEF scans per row and four steps per PASEF scan) with a 25 m/z isolation width80. The mass scan range was from 100 to 1700 m/z and ion mobility was scanned from 0.57–1.47 Vs/cm2 over a ramp time of 100 ms. The collision energy was ramped linearly from 20 to 52 eV as a function of mobility.

tims-TOF MS data analysis

To generate a comprehensive spectral library for the DIA analysis, we created a hybrid library that contained MS data of samples analysed in DDA mode and followed DIA analysis with technical replicates. The combined DDA and DIA acquisition raw files were analysed via Spectronaut (Biognosys AG) software with Pulsar search engine (SN14.10.201222) to build the library using UniProt human proteome database (UP000005640, 96,797 entries). The search parameters were set as default but included an additional deamidation (NQ) in variable modifications. DIA files (36 files for analysis of non-kinetic data and 20 files for analysis of kinetic data) were processed via Spectronaut using the default settings with precursor and protein FDR cut-off set to 0.01, quantification data filtering set to Q-value 0.5 percentile with global imputing, and cross run normalisation strategy set to local normalisation.

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE81 partner repository with the dataset identifiers PXD033221 (non-kinetic data) and PXD033222 (kinetic data).

Proteomics data analyses

For the non-kinetic study, the lead bioPROTAC T21RBCC-VHHHuR and the controls VHHHuR, VHHGFP and T21RBCC-VHHGFP underwent the outlined analyses. For the kinetic study all conditions were analysed. Peptide intensities were aggregated at the protein level. The resulting protein intensities were filtered and normalised to the total intensities of each sample. In-house scripts were developed to correct the batch effects using the plate information (https://github.com/AstraZeneca/trim21-bioprotac). To find differentially expressed proteins, a linear model was developed, and comparisons between constructs were defined using the Bioconductor limma package. False discovery rate (FDR)-adjusted p-values were calculated using the Benjamini Hochberg procedure and discoveries assigned based on a 5% FDR threshold and log2FC < −1 (downregulation) or log2FC > 1 (upregulation). Volcano plots and the Venn diagram were generated using the R packages ggplot2, ggrepel. Venn diagrams were generated using the R package VennDiagram. HuR protein-protein interaction information was retrieved from the BioGRID database (https://thebiogrid.org/). The web interface of UniprotKB (https://www.uniprot.org/) was used to extract the functional annotation (keywords and gene ontologies) of proteins present i.e. different subsets. Python scripts were used to process the output obtained from UniprotKB and barplots were generated using R (https://github.com/AstraZeneca/trim21-bioprotac).

Xenograft model

Athymic nude female mice were obtained at 8 weeks of age from Envigo and housed in specific pathogen-free and standardised environmental conditions according to UK Home Office regulations. Mice received irradiated aspen chip bedding, nesting material, a cardboard tunnel, and wooden chew blocks. Mice were housed on a 12/12 light/dark cycle, with ad libitum UV-treated water and sterilised RM1 rodent diet. The maximum tumour burden was not allowed to exceed 10% of body weight, using the following formula based on calliper measurements of length(l) and width(w): volume = (pi/6)*l*w2. This maximum burden was not exceeded during this work. Sex was not considered in the study design, as we concluded for a human xenograft tumour model the sex of the recipient would not have a significant impact on the results seen.

For tumour engraftment, 8–12-week-old athymic nude mice were anaesthetised and injected subcutaneously in the flank with 5 × 106 HCT116-ODIn cell lines (20 mice for each of T21RBCC-VHHHuR or VHHHuR groups) in 100 µl sterile PBS. Russ Lenth’s power tool was used to inform group sizes. Once tumours reached an average size of 150 mm3, tumour-bearing mice were size-matched and randomly assigned into ten mice per experimental group prior to doxycycline treatment. For groups receiving doxycycline, mice were switched to a sterilised chow containing 625 ppm doxycycline hyclate (equivalent to 545 mg/kg doxycycline) (Ssniff). Throughout the duration of the study, tumour size was routinely measured with electronic calipers, enabling tumour volume to be calculated in mm3 (length × width x width/2). Tumours were monitored until study endpoint or until an average tumour diameter of 15 mm or maximum volume of 1500 mm3 was reached. At endpoint, mice were euthanised via cervical dislocation with secondary confirmation, and tumours were then resected for ex vivo analysis. The maximum tumour burden was not exceeded during this work (greater than 10% of body weight; determined using the following formula based on calliper measurements of length(l) and width(w): volume = (pi/6)*l*w2).

In vivo imaging

Tumour-bearing mice were evaluated for mCherry expression at 4 days post doxycycline treatment by placing under recoverable isoflurane anaesthesia and imaging using an IVIS Spectrum (PerkinElmer) connected to XGI-8 Gas Anaesthesia System (Caliper Life Sciences). Once anaesthetised, mice were positioned on their sides on the IVIS stage enabling images to be captured with a field of view of 21.5 cm (FOV ‘D’). Images were acquired by epiluminescence with excitation 587 nm and emission 610 nm for mCherry detection.

Ex vivo analysis

Tumours were homogenised using zirconium oxide beads (1.4 mm and 2.8 mm) (Bertin Corp.) in PBS, containing protease/phosphatase inhibitors (New England BioLabs) and benzonase nuclease (Merck), with the Precellys tissue homogeniser (6500 rpm, 3 × 30 s oscillations) (Bertin Technologies). Lysates were collected and prepared in 10× RIPA buffer (Merck), prior to protein quantification using the BCA protein assay kit (ThermoFisher Scientific). Tumour lysates were assessed by immunoblotting, as above.

Statistical analyses

Tumour growth rate analysis was completed for each group from day 15 (first measurement post doxycycline) until day 36 (study endpoint), by fitting each animal’s tumour volume data to an exponential model using equation ‘log10(tumour volume) = a + b * time + error’ where a and b are coefficients that correspond to the log initial volume and growth rate respectively, as previously described82. Growth rate summary metrics calculated for each animal were then used for statistical analysis to compare groups – treating each animal as the experimental unit.

Average (mean), standard deviation (s.d.) and statistical significance based on Student’s t-test (two-tailed) or multiple comparisons using a one-way or two-way ANOVA with post-hoc test were calculated in GraphPad Prism.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.