Background Retrospective clinical trials reported a reduced local relapse rate, as well as improved overall survival after injection of local anesthetics during cancer surgery. Here, we investigated the anticancer effects of six local anesthetics used in clinical practice.
Results In vitro, local anesthetics induced signs of cancer cell stress including inhibition of oxidative phosphorylation, and induction of autophagy as well as endoplasmic reticulum (ER) stress characterized by the splicing of X-box binding protein 1 (XBP1s) mRNA, cleavage of activating transcription factor 6 (ATF6), phosphorylation of eIF2α and subsequent upregulation of activating transcription factor 4 (ATF4). Both eIF2α phosphorylation and autophagy required the ER stress-relevant eukaryotic translation initiation factor 2 alpha kinase 3 (EIF2AK3, best known as PERK). Local anesthetics also activated two hallmarks of immunogenic cell death, namely, the release of ATP and high-mobility group box 1 protein (HMGB1), yet failed to cause the translocation of calreticulin (CALR) from the ER to the plasma membrane. In vivo, locally injected anesthetics decreased tumor growth and improved survival in several models of tumors established in immunocompetent mice. Systemic immunotherapy with PD-1 blockade or intratumoral injection of recombinant CALR protein, increased the antitumor effects of local anesthetics. Local anesthetics failed to induce antitumor effects in immunodeficient mice or against cancers unable to activate ER stress or autophagy due to the knockout of EIF2AK3/PERK or ATG5, respectively. Uncoupling agents that inhibit oxidative phosphorylation and induce autophagy and ER stress mimicked the immune-dependent antitumor effects of local anesthetics.
Conclusion Altogether, these results indicate that local anesthetics induce a therapeutically relevant pattern of immunogenic stress responses in cancer cells.
- Adjuvants, Immunologic
- Drug Evaluation, Preclinical
- Drug Therapy, Combination
- Cytotoxicity, Immunologic
Data availability statement
Data are available on reasonable request.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
- Adjuvants, Immunologic
- Drug Evaluation, Preclinical
- Drug Therapy, Combination
- Cytotoxicity, Immunologic
What is already known on this topic
Retrospective trials observed an improved survival after the use of local anesthetics during oncological surgery. The mechanisms by which these agents may exert an antitumor activity remain unclear.
What this study adds
Local anesthetics elicit direct and immune antitumor effects potentiated by immunotherapy.
How this study might affect research, practice or policy
The immune property of local anesthetics could be used during surgical procedures to control migrating residual cancer cells and opens up a novel field of research called onco-anesthesia.
Cancer results from two interlinked and age-dependent processes: (1) the genetic and epigenetic instability of cells activating oncogenes and inactivating tumor suppressor genes and (2) the loss of immunosurveillance mechanisms that usually eliminate or control neoplastic cells.1 2 As a result, modern anticancer agents have been designed to suppress oncogenic pathways or to (re)activate anticancer immune responses.3 4 Moreover, chemotherapeutic agents have turned out to be particularly efficient when they succeed in not only reducing tumor mass but also inducing antitumor immunity.5–7 Often such immunostimulatory effects are obtained secondary to chemotherapy-mediated activation of stress pathways such as autophagy and elements of the endoplasmic reticulum (ER) stress response before cell death occurs.8–11 Such stress-related and lethal pathways cause the emission of immunostimulatory danger-associated molecular patterns (DAMPs), in particular adenosine triphosphate (ATP, the release of which is favored by autophagy),12 13 calreticulin (CALR, the exposure of which is favored by partial induction of ER stress)8 14 and HMGB1 (the release of which is associated with cell death).15 Cancer cells in which these stress responses and DAMPs release or exposure pathways are interrupted become resistant to chemotherapy because of their failure to activate a specific immune response.5 8 In sum, it appears that therapy-elicited immunosurveillance plays a major role in determining the outcome of the clinical management of oncological disease.
Hitherto unsuspected external factors affect the immune-cancer dialogue. For example, induction of stress hormones including catecholamines and glucocorticoids,16 perturbation of the gut microbiota by broad-spectrum antibiotics17–19 and other co-medications including beta-blockers modulate the clinical efficacy of anticancer treatments.20–22 In this context, it would not be surprising that surgical interventions with their components of mental, nutritional, mechanic and inflammatory distress as well as the associated polypharmacy (including the administration of antibiotics, benzodiazepines, non-steroidal anti-inflammatory drugs, general and local anesthetics, vasoactive amines, glucocorticoids…) would affect anticancer immunosurveillance, as suggested by statistically significant epidemiological associations.23–34
Here, we investigated the possible impact of several local anesthetics that are currently used in the clinics on cancer immunosurveillance in preclinical models. For this, we determined the effects of six local anesthetics on immunogenic cancer cell stress pathways and determined their possible antineoplastic effects by injecting them into established tumors. We found that local anesthetics induce autophagy and facets of ER stress that are immunostimulatory, hence favoring an immune-dependent reduction of tumor progression.
Materials and methods
Human osteosarcoma U2OS, human cervical cancer HeLa, murine colon adenocarcinoma MC38 and murine breast cancer E0771 cells were purchased from the ATCC. Murine fibrosarcoma MCA205 cells were purchased from Merck. U2OS cells stably expressing HMGB1-GFP; MITOdsRED; GFP-LC3; GFP-LC3 Atg5-/-; RFP-LC3; mCherry-GFP-LC3; pSMALB-ATF4.5rep, XBP1ΔDBD-venus-RFP-FYVE; GFP-LC3 EIF2AK1-/-; EIF2AK2-/-; EIF2AK3-/-; EIF2AK4-/-; RFP-LC3 eIF2αS51A (mutant with non-phosphorylable version of eIF2α) and murine fibrosarcoma MCA205 EIF2AK3-/-; MCA205 Atg5KD, MCA205 CD39 were generated by our group in the past (Michaud et al, 2011; Shen et al, 2012; Zhou et al, 2016; Bezu et al, 2018; Humeau et al, 2020). U2OS GFP-ATF6 cells were a kind gift from Prof. Peter Walter (University of California, San Francisco, USA). Cell lines were frequently tested for mycoplasma contamination.
U2OS cell lines were cultured in Dulbecco’s modified Eagle’s medium (#41 966-02, Thermo Fisher Scientific) supplemented with 10% fetal bovine serum (#F7524, Sigma Aldrich), 1% non-essential amino acids (#11 140-035, Thermo Fisher Scientific), 1% HEPES (#15630080, Thermo Fisher Scientific) and 1% penicillin/streptomycin (#15140122, Thermo Fisher Scientific). For U2OS HMGB1-GFP, XBP1ΔDBD-venus-RFP-FYVE, GFP-LC3 and GFP-ATF6 0.5 mg/mL G418 (#10 131-27, Thermo Fisher Scientific) was added to the medium. MCA205 cell lines were cultured in Roswell Park Memorial Institute (RPMI) 1640 medium (#61870044, Thermo Fisher Scientific) supplemented with identical components. Cells were maintained in a humidified incubator at 37°C with 5% CO2. Cell culture consumables were purchased from Corning (New York, USA).
Antimycin A (A8674), BAY87-2243 (SML2384), bupivacaine (B5274), carbonyl cyanide m-chlorophenylhydrazone (C2759), carbonyl cyanide 4-(trifluoromethoxy)phenylhydrazone (C2920), chloroprocaine (1117008), dactinomycin (A1410), 2-deoxyglucose (D3179), 2,4-dinitrophenol (D198501), erastin (E7781), ferrostatin (SML0583), glucose (G7021), levobupivacaine (SML1092), lidocaine (L5647), mitoxantrone (M6545), necrostatin (N9037), oligomycin A (75351), prilocaine (P9547), ropivacaine (R0283), rotenone (557368), staurosporine (S6942), TNF-alpha (T6674), thapsigargin (T9033), tunicamycin (T7765) have been purchased from Merck-Sigma Aldrich. Oxaliplatin was purchased from Accord Healthcare (Ahmedabad, India). Bafilomycin A1 (1334), rapamycin (1292), torin 1 (4247) were purchased from Tocris bioscience (Bristol, UK). Smac-BV6 was purchased from CliniSciences (B4653), z-VAD-fmk (N1510.0025) was purchased from Bachem. Recombinant calreticulin was produced as described.8 14
Rabbit polyclonal antibody CALR (ab2907), rabbit monoclonal phosphoneoepitope-specific antibody against phospho-eIF2α (Ser 51) (ab32157, clone E90), mouse monoclonal antibody against β-actin (ab49900, clone AC-15) were purchased from Abcam (Cambridge, UK). Rabbit polyclonal antibody against HRI (sc-30143), mouse monoclonal antibody against PKR (sc-6282, clone B-10) were purchased from Santa-Cruz biotechnology. Rabbit monoclonal antibody against PERK (#3192, clone C33E10), rabbit polyclonal antibody against GCN2 (#3302), rabbit polyclonal antibody against eIF2α (#99 722 S), rabbit monoclonal antibody against LC3B (#2775), rabbit polyclonal antibody against GFP (#2555) were purchased from Cell Signaling Technology (Danvers, MA, USA). Antibody anti-PD-1 (BE0273, clone 29F.1A12) and rat IgG1 anti-horseradish peroxidase isotype control (BE0088, clone HRPN) were purchased from BioXcell (West Lebanon, NH, USA). Mouse monoclonal antibody against Atg5 (A2859) was purchased from Merck-Sigma Aldrich. Anti-rabbit Alexa fluor®−488 coupled secondary antibody came from Thermo Fisher Scientific (#A11034). APC anti-mouse CD11c (Clone N418) was purchased from Biolegend (117310).
Determination of IC20 and IC50
U2OS cells were seeded at 1,500 cells per well in 384-well plates (Greiner Bio-one; Kremsmünster, Austria). The day after, cells were treated at different concentrations for 8 hours. Then, cells were washed once with PBS and fixed in 3.7% formaldehyde (F8775, Merck Sigma Aldrich) supplemented with 1 µg/mL Hoechst 33 342 (H3570, Thermo Fisher Scientific) for 20 min. After washing with PBS, plates were immediately analyzed by fluorescence microscopy. Micrographs were analyzed with R by means of the EBImage package. Notably, nuclei were detected using the Hoechst staining from which average nuclear size and intensity were computed. These parameters were subsequently used to cluster cells as healthy cells (normal-sized nucleus, Hoechstdim) and dead cells (condensed nucleus, Hoechstbright). The number of dead cells was used to determine a dose-response curve by fitting the data point with a 4-parameter log-logistic function. This curve was used to determine the IC20 (concentration at which 20% of cells are dead) and the IC50 (concentration at which 50% of cells are dead) for each drug.
Extraction of proteins was realized in RIPA buffer (#89900; Thermo Fisher Scientific) associated with phosphatase and protease inhibitors (#88669; Thermo Fisher Scientific). Protein content was measured by Bio-rad laboratory DCTM Protein Assay reagent A, B and S (#500-0113, #500-0114 and #500-0115). A minimum of 15 µg of protein was solved in Laemmli buffer (Thermo Fisher Scientific), denaturated at 100°C and separated by electrophoresis using 4%–12% Bis-Tris gels (Thermo Fisher Scientific) in MES buffer (Thermo Fisher Scientific). Then, proteins were transferred to nitrocellulose membrane (Merck Millipore IPVH00010) in transfer buffer (25 mM Tris; 190 mM glycine; 20% methanol in H2O) at 200 mA for 1.5 hour. Membranes were washed in Tris-buffered saline with Tween20 buffer (TBST; 20 mM Tris, pH 7.5 150 mM NaCl 0.1% Tween 20 in H2O) and treated with blocking buffer (5% bovine serum albumin (BSA) in TBST) for 1 hour. Membranes were exposed to primary antibody diluted in 5% BSA in TBST overnight at 4°C. The day after membranes were washed with TBST and incubated with appropriate horseradish peroxidase-coupled secondary antibody (Southern Biotech, Birmingham, Alabama, USA) for 1 hour at room temperature. Then proteins were revealed with ECL (GE Healthcare, Chicago, Illinois, USA). β-actin (1:10,000) was used to verify equal protein loading.
Fluorescence microscopy, image acquisition and analysis
One day before treatment, 1,500 U2OS cells either wild-type or stably expressing ATF6-GFP, GFP-LC3; GFP-LC3 Atg5-/-; RFP-LC3, mCherry-GFP-LC3, pSMALB-ATF4.5rep, XBP1ΔDBD-venus-RFP-FYVE, GFP-LC3 EIF2AK1-/-, EIF2AK2-/-, EIF2AK3-/-, EIF2AK4-/- or RFP-LC3 eIF2αS51A were seeded in 384-well plates (Greiner Bio-one; Kremsmünster, Austria). Upon treatment cells were washed once with PBS and fixed in 3.7% formaldehyde (F8775, Merck Sigma Aldrich) supplemented with 1 µg/mL Hoechst 33 342 (H3570, Thermo Fisher Scientific) for 20 min. ATF6-, XBP1-, pSMALB-ATF4.5rep-, LC3-, FYVE-expressing cells were immediately analyzed by fluorescence microscopy. For the analysis of eIF2α phosphorylation, cells were washed once with PBS after fixation and primary antibody diluted in blocking buffer (1:250, 2% BSA in PBS) was added for overnight at 4°C.35 Following, cells were washed thrice with PBS and incubated with AlexaFluor®-coupled secondary antibody (Thermo Fisher scientific) diluted 1:1,000 in 2% BSA for 45 min. Then, cells were washed once with PBS and fluorescence microscopic analyses were performed. A 20X PlanAPO objective (Nikon, Tokyo, Japan) was used for acquisition of four view fields per well. An automated image processing with the Custom Module Editor from MetaXpress Software (Molecular Devices) was used for image segmentation. Primary nuclear region of interest (ROI) was designed by a mask around the nucleus for cell count and for Hoechst, ATF4 or ATF6 fluorescence quantification. A secondary cytoplasmic ROI was used for the fluorescence intensity of phospho-eIF2α, ATF6, XBP1s or LC3 dots count. After exclusion of debris and dead cells, data were normalized and statistical analyses were performed with R software. Scale bars represent 10 µm.
Seven hundred and fifty U2OS wild-type cells were seeded in a 96-well plate. The next day, cells were transfected with small-interfering RNAs targeting Atg5 gene (ON-TARGETplus SMART pool SiRNA ATG5, L-004374-00-0005, Dharmacon, Lafayette, CO, USA) for 24 hours at a final concentration of 10 µM by means of DharmaFECTTM transfection reagent according to the manufacturer’s instruction (SMARTpool of 4 individual siRNAs). At day 3, medium was discarded. Cells were used at day 5.
Extracellular ATP quantification
Eight thousand U2OS wild-type cells per well were seeded in a 96-well plate. The next day, cells were treated for 24 hours. Then the supernatant was collected, centrifuged and transferred to a white bottom 96-well plate. Enzyme and substrate from ENLITEN ATP Bioluminescence Detection Kit (FF2000; Promega, Madison, Michigan, USA) were added. ATP-dependent substrate conversion was measured by luminescence at 560 nm with a Spectramax I3 multimode plate reader (Molecular Devices).
Calreticulin exposure by flow cytometry
Eight thousand U2OS wild-type cells per well were seeded in a 96-well plate. After 6 hours of treatment, cells were collected and transferred into a V-shape 96-well plate and centrifuged at 12,000 rpm for 5 min. Supernatant was removed and cells were incubated for 30 min at 4°C with primary rabbit monoclonal antibody against CALR (1:100 diluted in 1% BSA). Cells were washed, centrifuged at 12,000 rpm for 5 min and supernatant was removed. Cells were incubated with secondary AlexaFluor®488 goat anti-rabbit IgGs (1:1,000 diluted in BSA 1%) for 30 min at 4°C. Cells were washed, centrifuged at 12,000 rpm for 5 min and supernatant was removed. Finally, diamidino-2-phenylindole (DAPI, #62248, Thermo Fisher Scientific) was added before the analysis (1:400). Samples were analyzed using a CyAn ADP cytofluorometer (Beckman Coulter, Brea, California, USA) coupled to a HyperCyt loader (Intellicyt, Albuquerque, New Mexico, USA).
HMGB1 release by videomicroscopy
One day before treatment, 2×103 U2OS stably expressing HMGB1-GFP and H2B-RFP per well were seeded in a 384-well plate. The next day, cells were treated and observed by live-cell microscopy with a frequency of 1 image by hour for 24 hours. Then images were segmented and analyzed with R using the EBImage package. H2B-RFP was used to segment nuclei. Then, the obtained mask was used to measure nuclear GFP intensity over time, thus allowing to calculate the average signal loss.
Cell death assessment by flow cytometry
Eight thousand U2OS cells per well were seeded in a 96-well plate. After 8 hours of treatment, supernatant and cells were collected and transferred into a V-shape 96-well plate and centrifuged at 12,000 rpm for 5 min. Then, cells were incubated with diamidino-2-phenylindole (DAPI 0.5 µg/mL, #62248, Thermo Fisher Scientific) and with 3,3’-dihexyloxacarbocyanine iodide (DIOC 20 nM, #D273, Thermo Fisher Scientific) for 20 min at 37°C. Samples were analyzed using a CyAn ADP cytofluorometer (Beckman Coulter, Brea, California, USA) coupled to a HyperCyt loader (Intellicyt, Albuquerque, New Mexico, USA).
Autophagic flux measurement by U2OS mCherry-GFP-LC3 tandem reporter
To assess autophagic flux, U2OS cells stably expressing a tandem reporter mCherry-GFP-LC3 were seeded in a 96-well plate. The next day, cells were treated for 24 hours and observed by live-cell microscopy with a frequency of 1 image by hour for 24 hours. Then images were segmented and analyzed with R using EBImage package. The mCherry-GFP-LC3 tandem reporter emits GFP and mCherry signals in autophagosomes yet changes its fluorescence properties in the acidic environment upon lysosomal fusion. In consequence, the tandem reporter emits and solely mCherry signal in autolysosomes thus indicating autophagic flux.
Mitochondrial fiber assessment by videomicroscopy
One day before treatment, 8×103 U2OS stably expressing MITOdsRED were seeded in each well of a 96-well plate. Cells were stained with 3,3′-dihexyloxacarbocyanine iodide (DIOC 20 nM, #D273, Thermo Fisher Scientific) for 30 min. Then, cells were washed, treated and observed by live-cell microscopy with a frequency of 2 images by hour for 36 hours. Finally, images were segmented and analyzed with R using EBImage package. Briefly, the mitochondrial mask was detected using dsRed signal, and then skeletonized to measure the total length of the network.
Oxygen consumption rate (OCR) measurements were performed using the Seahorse XFe96 Flux Analyzer and the XF cell Mito stress kit (SeaHorse Agilent, Santa Clara, California, USA) as previously described by our group (Sica et al, 2017). Briefly, 8×103 U2OS wild-type cells per well were seeded in a 96-well plate and treated for 6 hours. Then, cells were incubated with 200 µL/well of SeahorseXF assay medium (102365-100) (pH 7.4) containing 2 mM glutaMAX, supplemented with 10 mM glucose and 1 mM sodium pyruvate for 1 hour at 37°C without CO2. Then the cartridge was loaded with oligomycin A (27 mM), carbonyl cyanidem-chlorophenyl hydrazine (CCCP, 3 mM) and rotenone (13.75 mM) which the Seahorse apparatus injected sequentially while measuring the OCR.
Extracellular acidification rate (ECAR) measurements were performed using the Seahorse XFe96 Flux Analyzer and the XF cell Glycolysis kit (SeaHorse Agilent) as previously described by our group (Sica et al, 2017). Briefly, 8×103 U2OS wild-type cells per well were seeded in a 96-well plate and treated for 6 hours. Then cells were incubated with 200 µL/well of Seahorse XF assay medium (102365-100) (pH 7.4) for 1 hour at 37°C without CO2. Then the sensor cartridge was loaded with glucose (90 mM), oligomycin A (30 mM) and 2DG (200 mM) which the Seahorse apparatus injected sequentially while measuring the ECAR.
Measurement of transcription inhibition
Evaluation of transcription was assessed using two distinct methods: the incorporation of Click-iT chemistry-detectable 5-ethynyl uridine (EU) (#C10327, Thermo Fisher Scientific) measured as described in (Cerrato et al, in press36), as well as transmitted-light based cell image classification using trained deep neural networks, as described in Sauvat et al.37 Two thousand human osteosarcoma U2OS cells per well were seeded in 384-well µClear imaging plates (Greiner Bio-One) and let adhere for 24 hours. The following day, cells were pre-treated for 2.5 hours and then the treatment pursued in the presence of 1 mM EU for an additional hour. Cells were then fixed with 3.7% formaldehyde containing 1 µg/mL Hoechst 33 342 for 1 hour at room temperature and permeabilized with 0.1% Triton X-100 for 10 min. Alexa-Fluor®488-coupled azide was then added for 2 hours at room temperature following the manufacturer’s instructions. The EU intensity corresponding to the GFP signal in the nucleus was measured by fluorescence microscopy. Nuclear GFP values of each condition were normalized to the untreated control, and the percentage of transcription inhibition was calculated. In parallel, transmitted light images were acquired and a dual deep-learning algorithm was applied to classify cells based on their nuclear phenotype. Briefly, cell nuclei were first segmented using a pre-trained Fully Convolutional Neural Network (FCNN), allowing the extraction of single-cell patches, and thereafter classified using a Deep Convolutional Neural Network (DCNN) in a binary fashion as ‘inhibited’ or ‘control’ phenotypes. The percentage of nuclei bearing a transcription inhibition phenotype was calculated.
Assessment of phagocytosis
Phagocytosis was assessed as previously described.38 Bone marrow-derived dendritic cells (BMDCs) were provided from femurs of C57Bl/6 mice. Bone marrow was collected by flushing the bones with PBS, and clusters were dissolved by pipetting. Then, red blood cells were lysed with red cell lysis buffer (0.01 M Tris, 0.83% NH4Cl in Milli-Qwater). After washing and filtration, 1.5×106 viable cells were seeded into 6-well plates in 2 mL of BMDC culture medium (RPMI 1640 medium supplemented with 10% fetal bovine serum, 1% non-essential amino acids, 1% HEPES and 1% penicillin/streptomycin and 50 µM β-mercaptoethanol) supplemented with 20 ng/mL recombinant mouse GMCSF and 5 ng/mL IL-4. At day 6, half of the supernatant was removed and replaced by the original culture. Non-adherent BMDCs were harvested on day 7, counted, and 8×105 mouse fibrosarcoma MCA205 cells were cultured in standard 25 cm2 flasks for cell culture. The day after, MCA205 cells were labeled with 0.5 µM CellTracker Orange CMTMR (5-(and-6)-(((4 chloromethyl)benzoyl)amino) tetramethylrhodamine) dye (#C2927, Thermo Fisher Scientific) for 30 min at 37°C. Then, MCA205 cells were treated for 24 hours. After treatment, MCA205 cells were co-cultured with BMDCs in 6-well plates at 37°C at a 1:4 ratio (BMDC: MCA205) for 4 hours. Finally, cells were detached and BMDCs were stained with conjugated 1:100 anti-CD11c antibody diluted in 1% BSA for 30 min at 4°C in the dark. Cells were washed three times and fixed in 3.7% formaldehyde. Samples were analyzed through a BD LSRFortessa flow cytometer and acquired with a BD FACSDiva software (BD Biosciences). Phagocytosis was determined by assessing the ratio of CMTMR+ CD11c+ cells over total amount of CD11c+ BMDCs (FlowJo software and R software).
In vivo experiments
Female wild-type C57Bl/6 and nu/nu mice aged 6–8 weeks old were purchased from Envigo (Huntington, UK) and were kept in a pathogen-free animal facility with temperature-controlled environment. Mice received water and food ad libitum. The number of mice needed in each group was determined by ‘BiostatTGV’ software. Mice were randomized in each group based on tumor size before treatment. Tumor area was calculated with the formula length × width x π/4. Mice were sacrificed at a tumor size between 200 and 250 mm2 or upon appearance of any signs of discomfort. Tumor growth and survival were analyzed with TumGrowth software package39 available at https://github.com/kroemerlab.
In vivo tumor treatment
Murine MCA205 fibrosarcoma tumors were generated by subcutaneous (s.c.) injection of 1×105 cells into C57Bl/6 mice and nu/nu mice. Murine MCA205 EIF2AK3-/- fibrosarcoma tumors were generated by s.c. injection of 1×106 cells into C57Bl/6 mice. MCA205 Atg5KD fibrosarcoma tumors were generated by s.c. injection of 1×106 cells into C57Bl/6 mice. Tumors from MCA205 cells expressing transgenic ectoATPase CD39 were generated by s.c. injection of 1×105 cells into C57Bl/6 mice. Murine E0771 breast tumors were generated by s.c. injection of 1×106 cells into C57Bl/6 mice. Murine MC38 colon tumors were generated by s.c. injection of 1×106 cells into C57Bl/6 mice. When tumors became palpable, mice were randomized in experimental groups and treated with 25 µL of local anesthetics or CCCP or FCCP (diluted in PBS) injected intratumorally (i.t.) once a day for 2 days. Tumor growth and weight were monitored frequently.
In order to boost the immune response, the same treatment was combined with immune checkpoint blockade. At days 8, 12 and 16 after the first injection of local anesthetics or CCCP or FCCP, 200 µg/mouse of anti-PD-1 or corresponding isotype (diluted in 100 µL PBS) were injected intraperitoneally.
Alternatively, mice with palpable MCA205 fibrosarcomas were treated by i.t. injection of ropivacaine (4 mg/kg) or CCCP (0.25 mg/kg) in a volume of 25 µL supplemented with recombinant calreticulin (15 µg/mouse) once a day for 2 days.
Profiling of the tumor-infiltrating T lymphocytes
MCA205 fibrosarcoma tumors were generated by subcutaneous (s.c.) injection of 1×105 cells in C57Bl/6 mice. When tumors became palpable, 25 µL of PBS or lidocaine (3 mg/kg) or ropivacaine (4 mg/kg) were injected intratumorally (i.t). At day 9, tumors were harvested, weighed, and then processed before phenotyping as described in40. In short, tumors were mechanically and enzymatically dissociated with the tumor dissociation kit (Miltenyi) and the gentle MACS Octo Dissociator (Miltenyi). Tumor cell homogenates were filtered and washed twice with ice-cold PBS. Then, homogenates corresponding to 50 mg of the initial tumor sample, were stained with LIVE/DEAD Fixable Yellow dead cell stain (Thermo Fisher Scientific) and Fc receptors were blocked with anti-mouse CD16/CD32 (clone 2.4G2, Mouse BD Fc Block, BD Pharmingen). T lymphocytes were phenotyped with a set of fluorochrome-conjugated antibodies as follow. First, surface receptors were stained with anti-CD45 APC-Fire750 (clone 30F-11, Biolegend), anti-CD3 APC (clone 17A2, BioLegend), anti-CD4 PerCP-Cy5.5 (clone RM4-5, Thermo Fisher Scientific), anti-CD8a PE (clone 53–6.7, BD Pharmingen), anti-CTLA4/CD152 APC (clone UC10-4F10-11, BD Pharmingen), anti-GITR/CD357 BV785 (clone DTA-1, BD Pharmingen), anti-ICOS/CD278 BV421 (clone 7E.17G9, BD Pharmingen), anti-LAG3/CD223 BV605 (clone C9B7W, BP Pharmingen), anti-PD-1/CD279 APC-Fire750 (clone 29F.1A12, Biolegend), anti-TIGIT BV711 (clone 1G9, BD Pharmingen), anti-TIM3/CD366 PeCy7 (clone RMR3-23, Invitrogen), anti-VISTA PercP-Cy5.5 (clone MH5A, Biolegend). Second, after permeabilization of the cells in eBioscience FoxP3/Transcription Factor Staining Buffer (Thermo Fisher Scientific), intranuclear staining of the transcription factor FoxP3 was performed with anti-FoxP3 FITC (clone FJK-16s, Thermo Fisher Scientific). Stained samples were run through a BD LSRFortessa flow cytometer. All samples were acquired using BD FACSDiva software (BD biosciences) and analyzed using GraphPad Prism software.
In vitro, data of each experiment are presented as the mean of quadruplicates (or triplicates after automatic detection of outliers). Data are presented as mean±SD if one experiment representative of three independent experiments is shown. Data are presented as mean±SEM when resulting from three independent experiments. Statistical analyses were performed with R software (https://www.r-project.org). Student’s t-test was used to compare parametric data to a control using the t.test function from the stats R package. A pairwise multiple comparison test (Benjamin-Hochberg correction) was applied to compare each condition to another in a dataset with the pairwise.t.test function from the stats R package. Heatmaps were generated after sigmoidal scaling of raw data. In vivo, statistical analyses were performed with the TumGrowth software package39 available at https://github.com/kroemerlab. This R package allows for longitudinal analyses over entire segments of the tumor growth curves that can be subjected to automatic analyses of breakpoints in the tumor growth characteristics and outlier detection (if required). A type II analysis of variance (ANOVA) test and pairwise Wilcoxon test (with multiple comparison) were performed for tumor growth and a log-rank test was used for survival analysis. Tumor immune infiltrates were statistically evaluated with GraphPad Prism software. Samples were compared using a one-way ANOVA (Tukey’s multiple comparisons test). Outliers were excluded based on the ROUT test (Q=10%). For all test, significance was assessed for */#/$p<0.05, **/##/$$p<0.01, ***/###/$$$p<0.001.
Induction of autophagy and cell death by local anesthetics in vitro
Human osteosarcoma U2OS cells are widely used in cell biology due to their large size, flat morphology, and strong adherence to microscopy-compatible plastic surfaces.41 These cells have been equipped with multiple biosensors to explore immunology-relevant cell stress pathways including autophagy and the three arms of the ER stress response9 42 43 . In a first step, we determined the dose at which six different local anesthetics would cause the death of 20% cells within 8 hours of culture (online supplemental figure S1A-F). The dose required for killing 20% of cells (IC20) was in the range between 1.04 and 4.58 mM, depending on the anesthetic, for U2OS cells, in the same order of magnitude as for human cervical cancer HeLa cells (online supplemental figure S1G-L), for which a similar order of cytotoxic effects was determined (ranking the IC20 for both cell lines: levobupivacaine <bupivacaine < ropivacaine <chloroprocaine < prilocaine <lidocaine) (online supplemental figure S1A–L). We then determined the effects of each local anesthetic on cell stress at the IC20, while focusing our analysis on the 80% of viable cells (by excluding dead cells with condensed nucleus, Hoechstbright). U2OS cells that were either autophagy competent (wild type) or autophagy deficient (due to the knock-out of the Atg5 gene) were equipped with autophagy biosensor microtubule-associated proteins 1A/1B light chain 3B (LC3) fused to green fluorescent protein (GFP) (figure 1A) and then cultured either in control condition or in the presence of the potent autophagy inducer rapamycin to observe the formation of GFP-positive autophagic puncta in the cytoplasm. Similar to the positive control, rapamycin, the six local anesthetics all induced the formation of discrete autophagic puncta in wild type, but not in Atg5-/- cells (figure 1B,C). The induction of GFP-LC3 puncta was observed at the IC20 as well as at slightly lower doses (online supplemental figure S2A, B). Moreover, the six anesthetics used at the IC20 caused the lipidation of endogenous LC3 protein that leads to an increase in the electrophoretic mobility of LC3 detectable by immunoblot (as LC3-II) and that is coupled to redistribution of LC3 from the cytosol to autophagosomes and autolysosomes.44 Again, the generation of LC3-II was found to be Atg5-dependent (figure 1D and E). In addition, local anesthetics gave rise to the autophagy-associated phosphorylation of phosphatidylinositol (yielding phosphatidylinositol 3-phosphate, PI3P) that is detectable by means of a biosensor composed by FYVE zinc finger domain fused to red fluorescent protein (RFP).45 Indeed, the six local anesthetics caused the redistribution of RFP-FYVE toward cytoplasmic dots (online supplemental figure S2C, D). Moreover, the local anesthetics induced autophagic flux measured by means of a tandem mCherry-GFP-LC3 reporter that distinguishes autophagosomes (puncta with a red +green fluorescence) from autolysosomes (puncta with a purely red fluorescence due to quenching of GFP-dependent fluorescent signal at a low pH).46 Indeed, bupivacaine, levobupivacaine, lidocaine and ropivacaine induced an increase in the ratio of autolysosomes over lysosomes that was reverted by simultaneous exposure to the lysosomal inhibitor bafilomycin A1, indicative of the induction of autophagic flux (online supplemental figure S2E, F). In summary, local anesthetics induce signs of autophagy in cancer cells.
Induction of ER stress by local anesthetics
Autophagy is tied to the induction of the integrated stress response, which is part of the ER stress response and consists in the phosphorylation of eukaryotic initiation factor 2α (eIF2α) on serine 51.47–49 Local anesthetics (used at the IC20, 8 hours) induced eIF2α phosphorylation detectable by means of a phospho-neoepitope specific antibody and fluorescence microscopy (figure 2A,B). Upon knock-in mutation of eIF2α to replace serine 51 by a non-phosphorylable alanine residue (eIF2αS51A) (figure 2C), the induction of local anesthetic-induced RFP-LC3 puncta was blocked (figure 2D,E). However, the phosphorylation of eIF2α by local anesthetics was not decreased in Atg5-/- cells (figure 2F), and cell loss was attenuated in Atg5-/- cells as well (online supplemental figure S3). Collectively, these data suggest that local anesthetics induce cell death by a pathway that involves intertwined autophagic and ER stress responses. eIF2α is phosphorylated by four different eIF2α kinases (EIF2AK1 to EIF2AK4).50 We determined the effects of the knock-out of each of these kinases (online supplemental figure S4A) on autophagy. The knock-out of EIF2AK3 had the strongest inhibitory effects on this parameter induced by local anesthetics (figure 2G), in line with the fact that this kinase (best known as PERK) is strongly tied to ER stress51 . Local anesthetics also induced other signs of ER stress detectable with suitable biosensor cell lines, such as a cell line in which GFP is placed under the control of the ATF4 promoter (and that measures the translational activation of ATF4 expression downstream of eIF2α phosphorylation) (online supplemental figure S5A, B), the expression of a XBP1ΔDBD-venus fusion protein that is only in-frame for venus (a variant of GFP) when XBP1 has been spliced by IRE1 (online supplemental figure S5C, D), and the redistribution of a GFP-ATF6 fusion protein to the Golgi apparatus and the nucleus (online supplemental figure S5E, F).
In conclusion, local anesthetics trigger the activation of all three arms of the ER stress response, namely, EIF2AK3/PERK-dependent eIF2α phosphorylation leading to ATF4 translation, IRE1-mediated XBP1 activation, as well as activation of the latent transcription factor ATF6. The PERK-dependent phosphorylation of eIF2α is required for the induction of autophagy.
Autophagy-dependent and immune-dependent anticancer effects of local anesthetics
In the next step, we evaluated the anticancer effects of local anesthetics at clinically relevant concentrations on three different cancer types developing in immunocompetent mice, namely, cutaneous MCA205 fibrosarcomas, E0771 breast cancers and MC38 colon cancers. Lidocaine or ropivacaine significantly (p<0.05) inhibited the growth of all these cancers or increased the overall survival of mice (figure 3A–F). When combined with PD-1 blockade, additive effects were observed in several cases, for instance for ropivacaine +anti-PD-1 against MCA205 (figure 3B), as well as for lidocaine or ropivacaine +anti-PD-1 against MC38 (figure 3E and F). Importantly, the tumor growth-reducing and survival-extending effects of ropivacaine (figure 4) and all other local anesthetics (online supplemental figure S6) against MCA205 fibrosarcoma were only observed in immunocompetent wild type mice (figure 4A, online supplemental figure S6A) but were lost in nu/nu mice that lack mature T lymphocytes due to inborn athymia (figure 4B, online supplemental figure S6B). Moreover, MCA205 cells engineered to lose EIF2AK3/PERK (by CRISPR/Cas9 technology)52 or ATG5 (by means of an SH-RNA-mediated knockdown)13 and inoculated into immunocompetent wild-type mice formed tumors that were resistant against the anticancer effect of ropivacaine (figure 4C and D, online supplemental figure S4B). In the same way, the lack of PERK or ATG5 confers a cytoprotective effect on tumor cells after treatment in vitro with local anesthetics suggesting that ER stress and autophagy are both mechanisms implicated in the antitumor response induced by local anesthetics (online supplemental figures S7, S8A-C). When added to cultured cancer cells, local anesthetics induced the immunostimulatory release of ATP (which is autophagy-dependent)13 (figure 4E) and HMGB1 (which is cell death-related)15 (figure 4F), but failed to stimulate CALR exposure (which is linked to the integrated stress response but suppressed by the IRE1/XBP1 pathway) (figure 4G)9 11 14 53 . Accordingly, local injection of recombinant CALR protein together with ropivacaine into orthotopic MCA205 fibrosarcomas enhanced the survival of mice more efficiently than the injection of CALR or ropivacaine alone (figure 4H). Moreover, tumors designed to overexpress the ecto-ATPase CD39, which destroys extracellular ATP, were unable to respond to local anesthetics (online supplemental figure S9).
Next, we investigate the immunogenic properties of local anesthetics. After treatment in vitro, MCA205 cells were labeled with CellTracker Orange (CMTMR) and their engulfment by bone marrow-derived CD11c+ dendritic cells (BMDCs) was assessed. Local anesthetics induced phagocytosis (figure 5A, online supplemental figure S10). Accordingly, local anesthetics elicited signs of an immune response in vivo by increasing CD4+ T lymphocytes, as well as the ratio of CD8+ T lymphocytes over FoxP3+CD4+ T regulatory cells (Tregs) in the tumor immune infiltrate. Moreover, local anesthetics caused intratumoral CD8+ T cells to upregulate glucocorticoid-induced TNFR-related receptor (GITR) and inducible costimulatory (ICOS) (figure 5B–F).
Immunogenic cell death inducers have been reported to induce apoptosis, necroptosis or ferroptosis and to frequently inhibit DNA-to-RNA transcription.54 Further results showed z-VAD-fmk and ferrostatin, which are inhibitors of pro-apoptotic caspases and ferroptosis, respectively failed to inhibit tumor cell death induced by local anesthetics. In contrast, necrostatin-1 attenuated the cytotoxic effects of lidocaine on tumor cells suggesting local anesthetics induce necroptosis (online supplemental figure S8D). Moreover, local anesthetics failed to inhibit transcription (online supplemental figure S11).
Altogether, these results suggest that local anesthetics can exert anticancer effects through a unique pathway that depends on the induction of autophagy in tumor cells as well as the presence of an intact immune system, but can be improved by the co-injection of CALR protein.
Inhibition of oxidative phosphorylation by local anesthetics has autophagy-dependent anticancer effects
Several reports indicate that local anesthetics can cause an inhibition of oxidative phosphorylation in different cell types including neurons, chondrocytes and cardiomyocytes.55–58 Therefore, we determined the capacity of different local anesthetics to inhibit oxygen consumption by U2OS cells. When added to cells that are in a state of maximal respiration (due to the addition of the uncoupling agent carbonylcyanide-3-chlorophenylhydrazone, CCCP), the addition of the six different local anesthetics (used at the IC20) caused an immediate reduction of oxygen consumption that was as efficient as that caused by the positive control rotenone, an inhibitor of respiratory chain complex I (figure 6A). Moreover, preincubation of cells with each of the six local anesthetics for 6 hours reduced basal respiration, which could be partially increased by addition of CCCP, but then was further inhibited by rotenone (figure 6B). This regimen of preincubation with local anesthetics enhanced basal glycolysis (as determined by the extracellular acidification rate, ECAR, reflecting lactate accumulation) that was not further increased by inhibition of oxidative phosphorylation by oligomycin (figure 6C). Of note, when added to U2OS cells expressing a mitochondrion-targeted dsRED biosensor, local anesthetics and mitochondrial inhibitors including CCCP, carbonylcyanide-4-trifluorometh-oxyphenylhydrazone (FCCP) and rotenone indistinguishably induced progressive rounding up of mitochondria, as determined by video fluorescence microscopy (figure 6D and E). These results suggest that local anesthetics alter cellular metabolism by inhibiting oxidative phosphorylation. We therefore examined the possibility that mitochondrial inhibitors would induce a similar pattern of stress responses as local anesthetics. CCCP (but less for rotenone) stimulated the formation of GFP-LC3 puncta (figure 6F and G). Both CCCP and rotenone caused the immunofluorescence-detectable phosphorylation of eIF2α (figure 6H and I) and the redistribution of FYVE-RFP to dots (online supplemental figure S12A, B). CCCP induced the activation of the three arms of the ER stress response, while rotenone only induced ATF4 and XBP1 activation, not that of ATF6 (online supplemental figure S12C-H). Rotenone was also able to activate the three hallmarks of immunogenic cell death (ATP release, CALR exposure, HMGB1 exodus), while CCCP only translocated calreticulin at the plasma membrane and activated the release of HMGB1 (online supplemental figure S13A-C). Nonhierarchical clustering analysis of all these cellular effects indicated that local anesthetics induce a pattern of ER- and autophagy-related stress responses that most closely resembles those elicited by the uncoupler CCCP (figure 6J). In sum, local anesthetics perturb mitochondrial morphology and function as they induce cellular stress responses similar to those caused by CCCP.
Driven by these findings, we evaluated the possible anticancer effects of CCCP and its analog FCCP on fibrosarcomas, mammary carcinomas and colon cancers developing in immunocompetent mice. Local injection of CCCP and FCCP caused tumor growth reduction that in several instances, in particular in MCA205 and MC38 cancers, was additive with the effects of PD-1 blockade (online supplemental figure S14). Again, tumor growth reduction by CCCP was lost when relevant stress pathways were blocked in the tumor cells by knock-out of PERK or knockdown of ATG5 (figure 7A–C), but was enhanced by local injection of recombinant CALR protein (figure 7D). Thus, CCCP apparently compromises tumor growth through mechanisms that resemble those triggered by local anesthetics.
In this paper, we report that the six most widely used local anesthetics can trigger T lymphocyte-dependent tumor growth reduction that is variably improved by systemic administration of anti-PD-1 antibodies or local injection of CALR, but lost upon genetic inhibition of the integrated stress response (the EIF2AK3/PERK-dependent phosphorylation of eIF2α) and autophagy. Mechanistically, local anesthetics induce the integrated stress response (with activation of the ER stress-relevant kinase EIF2AK3/PERK, leading to the phosphorylation of eIF2α) which is required for the induction of autophagy.48 In addition, local anesthetics stimulate a full-blown ER stress response with activation of the transcription factors ATF4, XBP1 and ATF6. Although EIF2AK3/PERK-catalyzed eIF2α phosphorylation is required for CALR exposure,14 the activation of the IRE1/XBP1 arm of the ER stress response suppresses CALR exposure,9 53 explaining the net result that local anesthetics fail to stimulate the translocation of CALR from the ER lumen to the cell surface. However, local anesthetics were able to elicit other immunostimulatory stress signals including the release of ATP and HMGB1 from cancer cells, likely explaining their immune effects. Taken together, these data suggest that the administration of a local anesthetic into a (still immunologically ‘cold’ tumor) causes local stress, the stress response lasting longer than the presence of the drug, hence setting of a cascade that ignites the recruitment of immune effectors and renders the tumor immunologically ‘hot’.
Of note, reportedly local anesthetics induce various modalities of cell stress and death (autophagy, apoptosis, necrosis…) in a time-and dose-dependent fashion, as previously described to explain their neurotoxicity.59 60 In line with the literature,61 the use of specific caspase inhibitor such as z-VAD-fmk failed to interfere with tumor cell death mediated by local anesthetics. Toxicities of local anesthetics may involve damage of cell membranes and mitochondria. Thus, local anesthetics inhibit oxidative phosphorylation and perturb mitochondrial morphology in cultured cancer cells, in line with previous observation suggesting uncoupling effects on mitochondria from normal cells.55–58 The mitochondrial uncoupler CCCP fully mimicked the cell stress-inducing and immune-dependent anticancer effects of local anesthetics at all levels, including the stimulation of autophagy, all arms of the ER stress response and two hallmarks of ICD. Moreover, local injection of CCCP or that of its close structural and functional analog FCCP had anticancer effect that required EIF2AK3/PERK and ATG5 expression by the tumor cells. Altogether, these results can be interpreted to mean that the bioenergetic effects of local anesthetics are responsible for their immunostimulatory function. Although long-term uncoupling effects would probably compromise the function of tumor-infiltrating immune cells,62 the transient presence of locally injected anesthetics (and presumably CCCP and FCCP as well) appears to be compatible with immune effector functions as demonstrated by the induction of tumor cell phagocytosis by immature dendritic cells induced in vitro and the signs of a T cell-mediated anticancer immune response observable within tumors.63
Irrespective of the exact mechanisms that account for the pro-immunosurveillance effects of local anesthetics, there is ample clinical evidence in favor of anticancer effects of such compounds. Indeed, several retrospective clinical trials observed a decrease in recurrence and a better survival after local anesthetics injection alone or in supplement of general anesthesia during solid tumor resection.64–70 A meta-analysis conducted on 21 studies supported a positive association between local anesthetics and oncological outcomes compared with general anesthesia (overall survival p=0.026; recurrence free survival p=0.047).71 Thus, local anesthesia performed for tumor biopsies, for melanoma exeresis, regional block anesthesia (such as block of the axillary brachial plexus, serratus anterior plane, or pectoralis nerve) realized in the context of tumorectomy and lymphadenectomy, may directly induce immunogenic stress in cancer cells. In addition, local anesthetics injected close to the tumor site might reduce the migration of residual malignant cells released during surgery. Whether these effects are solely due to activation of immunogenic stress or rather involve additional effects such as the well-studied anti-inflammatory effects of local anesthetics or the direct activation of natural killer cells remains an open question.63 72 73
The most important limitation of this study resides in its preclinical nature, requiring validation at the clinical level. More than twenty prospective trials are currently evaluating the prognostic and immune effects of local anesthetics administered via distinct routes (intraperitoneal injection, spinal anesthesia, epidural anesthesia, infusion pumps…) to patients with multiple different cancer types (online supplemental table 1). The results of these studies may influence future guidelines in onco-anesthesia.
Data availability statement
Data are available on reasonable request.
Patient consent for publication
All mice experiments were conducted according to EU Directive 63/2010 with the following protocols approved by the Ethical committee of the Gustave Roussy Cancer Campus (CEEA IRCIV/IGR no.26, registered at the French Ministry of Research, (2018100208217977v3; 2020070716191430v2; 2020100809149728v2; and 2021061408119739v2).
We are grateful to the support of the Gustave Roussy Cancer Campus, Université Paris-Saclay.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Contributors LB performed most experimental procedures, analyses, preparation of the manuscript and is responsible for the overall content as the guarantor. AWC performed HMGB1 video, oxygen consumption and glycolytic flux experiments and participated in autophagic experiments. AS helped in statistical analyses and image acquisition. JH generated eIF2αS51A cells as well as the knockout cells lacking EIF2A kinases 1-4. WX generated the MCA205 EIF2AK3 knockout cell line. GC performed the experiments of transcription. PL, LZ, SZ participated in phagocytosis experiments. JLN and JP participated in the characterization of the tumor immune infiltrate. PvE generated recombinant calreticulin. OK helped in figure design. FB provided intellectual input. GK and LB conceived and directed the project, and wrote the manuscript.
Funding GK is supported by the Ligue contre le Cancer (équipe labellisée); Agence Nationale de la Recherche (ANR)—Projets blancs; ANR under the frame of E-Rare-2, the ERA-Net for Research on Rare Diseases; Association pour la recherche sur le cancer (ARC); Cancéropôle Ile-de-France; Chancellerie des universités de Paris (Legs Poix), Fondation pour la Recherche Médicale (FRM); a donation by Elior; European Research Area Network on Cardiovascular Diseases (ERA-CVD, MINOTAUR); Gustave Roussy Odyssea, the European Union Horizon 2020 Project Oncobiome; Fondation Carrefour; High-end Foreign Expert Program in China (GDW20171100085 and GDW20181100051), Institut National du Cancer (INCa); Inserm (HTE); Institut Universitaire de France; LeDucq Foundation; the LabEx Immuno-Oncology; the RHU Torino Lumière; the Seerave Foundation; the SIRIC Stratified Oncology Cell DNA Repair and Tumor Immune Elimination (SOCRATE); and the SIRIC Cancer Research and Personalized Medicine (CARPEM). LB is supported by Bristol-Myers-Squibb (BMS). Foundation for Research in Immuno-Oncology. AWC was supported by El Programa Nacional de Becas 'Don Carlos Antonio Lopez' (BECAL). JH was supported by the Philanthropia Foundation. WX and SZ received support by the China scholarship council. PvE acknowledges support by the Fondation pour la Recherche Médicale (EQU201903007853).
Competing interests GK and OK are cofounders of Samsara Therapeutics. JP is the inventor of patents for cancer vaccination licensed to Turnstone Biologics.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.