Human authenticated leukaemia NALM-6 cells, RRID:CVCL_0092 were cultured in T175 tissue culture flasks (VWR, Leicestershire, UK). They were maintained at 37°C in humidified air balanced with 5% CO ₂ in RPMI 1640 medium (GIBCO, Invitrogen, Paisley, UK) supplemented with 10% heat-inactivated fetal calf serum (Thermo Scientific, Epsom, UK), 2 mM L-glutamine and 100 U ml ^-1 penicillin-streptomycin (Invitrogen). Culture medium was replaced every day.

Before gas exposure, NALM-6 cells were cultured at 1 × 10 ⁶ per ml density on 30 mm ² Petri dishes (VWR, Leicestershire, UK), or 24-well plates with a seeding density of 5 × 10 ⁴ per ml. Cells were used 12 hours after seeding. Cells were placed in 1.5 L purpose-built airtight, temperature-controlled chambers equipped with inlet and outlet valves and an internal electric fan, which was used to provide a continuous delivery and mixture of gases. The chamber was connected to calibrated flow meters and an in-line vaporizer was used to deliver the desired composition (Datex gas monitor, Helsinki, Finland) of sevoflurane (3.6%, MAC 2.0) (Abbott Laboratories, Maidenhead, UK) in 21% oxygen and 5% CO ₂ balanced with nitrogen (BOC, Guildford, UK). The chamber was pre-flushed with the aforementioned gas mixture to ensure that a stable gas composition was achieved, and a closed system was established to prevent leakage. Gas treatment was given at the desired sevoflurane concentration for two, four, and six hours at 37°C. At the end of treatment, cells were harvested for further analysis. Cells used as the naïve control group were placed in an identical gas chamber containing 21% oxygen and 5% CO ₂ balanced with nitrogen at 37°C.

A clinical formulation of propofol was used. It was dissolved in 10% intralipid (Astra-Zeneca, London, UK). On the day of the experiment, the stock solution of propofol was diluted with medium to the desired concentrations (5 and 10 μg/ml, equivalent to 28 and 56 μM, respectively). For the intralipid control (vehicle control), 10% intralipid was added to the cell medium to recreate the amount of intralipid in the highest (10 μg/ml) dose of propofol being used. Propofol-supplemented medium was added to the cell cultures for six hours. At the end of treatment, cells were harvested for further analysis. Cells used as the naïve control group were treated with identical medium with no propofol or intralipid.

For cell surface receptor experiments, leukaemia cells were incubated with CXCR4 (eBiosciecne, 17-9999-42, APC) and CD49d (eBioscience, 12-0499-42, PE) antibodies for 30 minutes at 4°C. After incubation, cells were washed with PBS and samples were run on a flow cytometer (CyAn ADP; Beckman Coulter, US). In addition, isotype controls were used to establish background fluorescence signals including: 1) mouse IgG2a kappa isotype control eBM2a (eBioscience, 17-4724-81, APC) and 2) mouse IgG1 kappa isotype control P3.6.2.8.1 (eBioscience, 12-4714-82, PE). In order to gate cells, the scatter plot of forward scatter versus side scatter was initially gated to have viable cells. Then, duplicates were excluded by plotting forward scatter height versus area. Finally, protein expression represented by mean fluoresce intensity (MFI) was recorded using a histogram. Each assay included 50,000 gated events. All samples were analysed via FlowJo software. For propidium iodide experiments, cells were stained with propidium iodide (e-Bioscience, 00-6990-50, PE-Texas red) and incubated in the dark at room temperature for 5 minutes. A count of 50,000 cells per sample were analysed with a flow cytometer (CyAn ADP; Beckman Coulter, US). In order to determine the cell death via propidium iodide, the scatter plot of forward scatter versus the PE-Texas was initially analysed in naïve control samples to determine the population of healthy cells. We determined any cells with an intensity larger than 10 ² in the PE-Texas red channel to be dead cells.

In vitro migration assay was performed by loading 2 × 10 ⁵ pre-treated cells into the upper chamber of the migration chamber. Cells were allowed to migrate for six hours before being collected in the lower chamber. The lower chamber was supplied with 20 ng/ml SDF-1 as a chemoattractant. Finally, samples were run on a flow cytometer (CyAn ADP; Beckman Coulter, US) with Accu check counting beads (Thermo Fisher, PCB 100, UK) to determine absolute cell counts in each sample.

NALM-6 cells were pre-treated with propofol and sevoflurane with respective concentrations. Recombinant human OPN (R&D systems, 1433-OP-050/CF) was thrombin-cleaved by incubating 24 μg recombinant full length human OPN in 20 mM Tris-HCL (pH7.6), 80 mM NaCl, 2 mM CaCl ₂, and 0.1 unit thrombin (Merck, 10602400001) for ten minutes at 37 degrees. Then, full length OPN and thrombin-cleaved OPN were coated at 20 μg/ml the night before experiment. Then, 5 × 10 ⁵ cells of NALM-6 cells were labelled with calcein-AM stock solution from the Vybrant cell adhesion assay kit (ThermoFisher, UK). Pre-labelled NALM-6 cells were allowed to adhere for two hours under 37°C. Finally, fluorescence readings were read at 494 nm. The percentage of adhesion was determined by dividing the corrected (background subtracted) fluorescence of adherent cells by the total corrected fluorescence of cells added to each microplate well and multiplying by 100%.

All animal work was in accordance with the animal ethics committee (AWERB) at Imperial College London, UK and the UK Home Office regulations (ASPA, 1986). C57BL/6 female mice were purchased from Charles rivers UK Ltd with minimum age >6 weeks with weight around 20-22 g. All animals were allowed for 7 days of acclimation before procedures.

Prior to procedures, animals were randomly allocated into three groups: 1) naïve control group, 2) propofol treated group and 3) sevoflurane treated group. In naïve control group, animals were given one million untreated GFP-NALM-6 cells through tail vein injection followed by intravital microscopy. In propofol and sevoflurane treated group, animals were given one million pre-treated GFP-NALM-6 cells (pre-treated with 10 μg/ml propofol and 3.6% sevoflurane for six hours in vitro, respectively) followed by intravital microscopy. On the day of experiments, we would normally do three intravital microscopy procedures in the order of 1) naïve control group, 2) propofol treated group and 3) sevoflurane treated group to minimise potential confounders. All members involved in the mouse study were aware of group allocations at all stages.

Initially, 5 × 10 ⁴ of NALM-6 cells were seeded in a 24-well plate. After 12 hours, TransDux and TransDUX MAX enhancer (System Biosciences, Palo Alto, USA) were mixed and added to culture at a concentration of 1X. Then, pre-packed GFP lentivirus (System Biosciences, Palo Alto, USA) was added to NALM-6 cells at a multiplicity of infection (MOI) of 20. Cells were incubated in 5% CO ₂ at 37°C for 72 hours. Finally, GFP-expressing NALM-6 cells were sorted using FACS Aria-II cell sorter (BD, USA), before seeding in a T175 flask with fresh medium.

Intravital microscopy was performed using a Leica SP5 and a Zeiss LSM 780 upright confocal microscopes, both with a motorized stage. ¹¹ The SP5 was fitted with the following lasers: Argon, 546, 633 and a tunable infrared multiphoton laser (Spectraphysics Mai Tai 690-1020). The Zeiss LSM 780 was fitted with the following lasers: Argon, 561, 633 and a tuneable infrared multiphoton laser (Spectraphysics Mai Tai DeepSee 690-1040). The signal was visualized with a Leica HCX IRAPO L ×25 water immersion lens (0.95 N.A) and a W Plan-Apochromat ×20 DIC water immersion lens (1.0 N.A), respectively. Collagen bone second harmonic generation signal and GFP and CFP signals were generated through excitation at 840 and 870 nm and detected with external detectors. Internal detectors were used to collect DsRed and Cy5 signals (and on some occasions, GFP). Anaesthesia was induced in mice with 4% isoflurane mixed with pure oxygen. This was gradually reduced to approximately 1% as anaesthesia stabilised. Animals were randomly assigned to each group (naïve control, propofol treated, and sevoflurane treated).

Surgery to attach the headpiece was then performed as described below. 1)

Sterile forceps and scissors were used to carefully remove the central portion of the scalp to expose the calvarium area to be imaged. Then, a small incision was made at the back of the head between the ears by lifting the skin up with the forceps. While holding the skin up, scissors were slid under the skin and gently cut along the outside of the desired imaging area.

Large three-dimensional ‘tile scans’ of the entire BM cavity space were acquired by stitching adjacent, high-resolution z-stack images using a surgically implanted imaging window that ensures steady positioning of mice on the microscope. Time-lapse datasets were acquired by focusing on one area of BM. One image was taken every three minutes for a total of 22 images. Then, all 22 images were combined together to produce a movie (time-lapse). The calvarium has been demonstrated to be equivalent to the long bones such as the femur with regards to haematopoietic stem cell frequency, function, and localization, and is the only BM compartment that allows longitudinal imaging through minimally invasive surgery. Blood vessels were highlighted by I.V. injection of 50 μl of 8 mg/ml 500 kDa Cy5-Dextran (Nanocs, MA). After the intravital microscopy, all mice were euthanised via cervical dislocation.

Microscopy data was processed using multiple platforms. Tile scans were stitched using ZEN black, RRID:SCR_018163 (Zeiss, Germany) software. Raw data were visualised and processed using Fiji/Image J, RRID:SCR_00285 (National Institutes of Health, Maryland, USA). Cell tracking was performed using FIJI plugin MTrackJ. For accuracy in cell tracking data, videos were registered when required before using four-dimensional data protocols implemented in Fiji. Three-dimensional data rendering and measurement of cell distances were performed in Volocity, RRID:SCR_002668 (Perkin Elmer, MA, USA).

The sample size required for the experiments was estimated based on the results of preliminary data. Statistical differences between the means of two data groups were determined using two-tailed unpaired Student’s t-test, and p values<0.05 were considered statistically significant. Multiple group comparisons were performed using ANOVA with a Kruskal-Wallis nonparametric or Bonferroni’s post hoc test for comparisons GraphPad Prism 7.0, RRID:SCR_002798. A p value<0.05 was considered statistically significant.

Before assessing their effects on cell migration, cell toxicity of propofol and sevoflurane were assessed via propidium iodide staining. No cell death was detected in NLAM-6 and Reh cells after exposure to propofol and sevoflurane ( Figure 1a- c).

We investigated whether the migration of leukaemia cells was affected by general anaesthetics. An in vitro migration chamber was set up where 2 × 10 ⁵ of NLAM-6 (pre-treated with propofol or sevoflurane for six hours) were loaded into the upper chamber and allowed to migrate for six hours to the lower chamber where 1 nM of SDF-1, a CXCR4 ligand, was applied. We observed that 10 μg/ml (56 μM) of propofol led to a significant reduction in the number of cells migrating to the lower chamber (almost a 50% decrease) ( Figure 2a). Similarly, sevoflurane, on average, reduced cell migration to the lower chamber by approximately 35%, compared with naïve control ( Figure 2a). Intralipid (vehicle control) had no effect on cell migration.

In order to validate our in vitro migration results, we set up a mouse model to study migration using intravital microscopy. ¹² In our model, we used immune-competent C57BL-6 mice instead of SCID mice for homing and migration experiments as it is a well-established model for non-maintenance cancer studies. ¹³ GFP-NALM-6 cells (pre-treated with 10 μg/ml propofol or 3.6% sevoflurane for six hours) were injected intravenously. Then, intravital microscopy was performed. Only cells present in all time lapse images were tracked and included in our displacement analysis. In addition, cells inside blood vessels were not included in analysis. Displacements were calculated by adding all displacements in each time interval. The mean speed was calculated by averaging all mean speed recorded in each time interval. Our results showed that general anaesthetic pre-treated cells appeared to achieve smaller displacements compared to naïve cells, and their mean migration speed was also significantly reduced ( Figure 2b). It was very obvious that after propofol and sevoflurane exposure, the number of fast-moving cells was significantly reduced ( Figure 2b).

After revealing that cell migration was reduced by general anaesthetics, we investigated whether the homing process was affected by anaesthetics specifically in terms of 1) homing location of cells (expressed by the distance between each cell to the nearest endosteal surface); 2) The number of cells entering bone marrow. Tile scan imaging ( Figure 3a) of large areas of mouse calvarium bone marrow was carried out and the distance between each cell and the nearest endosteal bone surface was calculated. We used this distance as an indication for homing location. Cells in the blood vessels were not included in our analysis. Results showed that the average distance between a leukaemia cell and the endosteal bone surface was increased by propofol and sevoflurane, indicating disruption of the homing process ( Figure 3b). Naïve control cells consistently move to locations approximately within 100 μm of the endosteal surface in bone marrow ( Figure 3b). Presumably, this is a discrete region near vessels demonstrated previously, ⁸ though the previous publication did not measure the distance between cells and the endosteal bone surface. After exposure to propofol or sevoflurane, a fair proportion of cells was found in locations further away from the endosteal bone surface ( Figure 3b).

We then studied whether treatment with general anaesthetics affected the number of cells entering bone marrow by counting the number of GFP+ cells from tile scans. Our results demonstrated that treatment with propofol and sevoflurane caused a significant reduction in the number of NALM-6-GFP cells entering bone marrow ( Figure 3c). After either treatment, fewer than 100 GFP-NLAM-6 cells entered the bone marrow compared to an average of nearly 200 cells entering the bone marrow for naïve control cells.

CXCR4 is crucial for migration ¹⁴ and homing ⁸ of leukaemia cells in the bone marrow. One study demonstrated that B-ALL cells (NALM-6 cells) home to discreet, discontinuous locations near vessels in bone marrow that express the adhesion molecules E-selectin and SDF-1. ⁸ Disruption of the interactions of the SDF-1-CXCR4 axis inhibited the homing of GFP-NALM-6 cells to these vessels. ⁸ We hypothesised that CXCR4 was one of the key factors in general anaesthetics-mediated reduction of the migration and homing of GFP-NALM-6 cells in vivo. In order to validate this, we treated NALM-6 cells with propofol (5 and 10 μg/ml) and sevoflurane (3.6% MAC 2) followed by flow cytometry analysis of surface CXCR4 expression. Our results indicated that both propofol and sevoflurane significantly reduced surface CXCR4 expression ( Figure 4a). More specifically, 5 and 10 μg/ml of propofol (28 and 56 μM) reduced CXCR4 expression in a dose-dependent manner ( Figure 4a). Only six hours of sevoflurane significantly reduced CXCR4 ( Figure 4a). These results suggest that CXCR4 might be the key behind general anaesthetics-mediated reduction of homing and migration in leukaemia cells.

In addition to CXCR4, we then speculated that the changes in the number and position of GFP-NALM-6 cells homed to the BM was due to the disruption of interactions between leukaemia cells and components of bone marrow matrix. We firstly investigated the expression of integrin α4 (CD49d), which is one of the dimers of α4β1 integrin or very late antigen 4 (VLA-4) (the other one being integrin β1), after exposure to general anaesthetics. VLA-4 is an essential mediator for engraftment of NALM-6 cells in bone marrow by anchoring NALM-6 cells to bone marrow matrix and supporting leukaemia growth. ^{10 , 15 , 16} The prime target for VLA-4 adhesion is osteopontin (OPN). Interestingly, no significant change was detected in the expression of the α4 subunit after anaesthetics exposure (Supplementary figure 1a and b). Next, we hypothesised that general anaesthetics might affect the adhesion of NALM-6 cells to OPN. Given that OPN has various isoforms in bone marrow, we tested the adhesion of leukaemia cells to thrombin-cleaved OPN, which is the most dominant form of OPN in bone marrow. ⁹ Similar to a previous study, ¹⁰ NALM-6 cells only weakly adhered to full-length OPN but they showed strong adhesion to thrombin-cleaved OPN (Supplementary figure 1c). Consistent with our hypothesis, when cells were pre-treated with propofol and sevoflurane, the percentage of cells adhered to thrombin-cleaved OPN was significantly reduced ( Figure 4b and c).