Cell culture
Cells were cultured in Dulbecco’s modified Eagle’s medium (Thermo Fisher Scientific) supplemented with 10% fetal bovine serum (Hyclone) and 1% penicillin/streptomycin (Gibco). A549 cells and H1975 cells were obtained from ATCC. The lung squamous cell line (human) RH2 were established in the laboratory of S.M.D. (UCLA). The head and neck squamous cell line (human) Tu686 was a gift from the laboratory of M.S.J. (UCLA). The mouse lung squamous cell line derived from mouse 5 (LPP) and mouse LUAD cell line derived from mouse 4 (LPP) were established in our laboratory. After resecting the tumour tissues, they were minced with razors and digested with collagenase/dispase (Sigma). Lysate was filtered through a 70-μm cell strainer. Dissociated single cells were centrifuged and resuspended in Dulbecco’s modified Eagle’s medium (10% fetal bovine serum). The cells were plated in tissue culture dishes and the medium was changed until there were enough colonies to expand the culture. All cells were grown at 37 °C in 5% CO2 in a humidified incubator, and a test for mycoplasma was carried out using the LookOut Mycoplasma PCR Detection Kit (Sigma). Cell identities were confirmed by Laragen Inc. The endoribonuclease-prepared siRNA used in this study was: GLUT1 (SLC2A1) (EHU028011, Sigma); OGT (EHU082301, Sigma).
GEMMs of lung tumour
We used five GEMMs in this study: (1) Kras-Lox-Stop-Lox-G12D; Rosa26-Lox-Stop-Lox-Luc (Kras); (2) Kras-Lox-Stop-Lox-G12D; LKB1 Lox/Lox; Rosa26-Lox-Stop-Lox-Luc mice (KL); (3) Kras-Lox-Stop-Lox-G12D; P53 Lox/Lox; Rosa26-Lox-Stop-Lox-Luc mice (KP); (4) Kras-Lox-Stop-Lox-G12D; LKB1 Lox/Lox; P53 Lox/Lox; Rosa26-Lox-Stop-Lox-Luc (KPL); (5) LKB1 Lox/Lox; P53 Lox/Lox; PTEN Lox/Lox; Rosa26-Lox-Stop-Lox-Luc (LPP). Lung tumours were induced by Ad5-CMV-Cre (VVC-U of Iowa-1174) or LentiCre (Kerafast) delivered intranasally as described previously37. Tumour growth was routinely monitored by bioluminescence imaging using an IVIS imager (PerkinElmer). All animal experiments were approved by UCLA’s Animal Research Committee (ARC) and carried out following ARC protocols and requirements. The tumour burden endpoints (morbidity; weight loss no greater than 20%; laboured breathing; impingement of animal’s ability to ambulate, eat and drink) allowed by our Institutional Animal Care and Use Committee were not exceeded. Lung tumours from different GEMM mice were collected and snap frozen in liquid nitrogen. Snap-frozen samples were stored at −80 °C until respirometry assay and western blotting analysis.
Subcutaneous implantation in NSG mice
A549, H1975, A549 Rho, RH2 and Tu686 cells were cultured in vitro under the conditions described above. Cells were collected and suspended in PBS, then mixed with Matrigel Membrane Matrix (Corning) and implanted subcutaneously on the flanks of NSG mice (2–4 × 106 cells per flank). All animal experiments were approved by UCLA’s ARC and carried out following ARC protocols and requirements. The tumour burden endpoints (tumour volume no greater than 2,000 mm3; morbidity; weight loss no greater than 20%; laboured breathing; impingement of animal’s ability to ambulate, eat and drink) allowed by our Institutional Animal Care and Use Committee were not exceeded. For the treatment study of the GLUT1 and GLUT3 inhibitor KL-11743 (C6), mice were treated with C6 (100 mg kg−1) delivered by oral gavage for 8 days. Subcutaneous tumours were either snap frozen in liquid nitrogen or fixed in 10% formalin overnight. Snap-frozen samples were stored at −80 °C until respirometry assay. Formalin-fixed samples were sent to the Translational Pathology Core Laboratory at UCLA for embedding and sectioning.
[18F]FBnTP synthesis
Synthesis of the radioactive [18F]FBnTP probe was carried out as previously described9,38.
PET–CT imaging
PET–CT imaging and analysis were carried out on GNEX as previously described37,39. PET signals were measured as percentage of injected dose per gram after 1 h uptake and normalized to heart signal. Tumours with a tumour-to-heart [18F]FBnTP uptake ratio of ≥0.5 are defined as [18F]FBnTPHI; those with a ratio of <0.5 are defined as [18F]FBnTPLO. Tumours with a tumour-to-heart [18F]FDG uptake ratio of ≥0.2 are defined as [18F]FDGHI; those with a ratio of <0.2 are defined as [18F]FDGLO (ref. 26).
Ex vivo respirometry analysis on frozen tissues
Tumour tissues were isolated from GEMMs of lung cancer and snap frozen using liquid nitrogen. Frozen samples were stored in −80 °C until use in the Seahorse experiments. Frozen tissues were thawed on ice and homogenized in MAS buffer (70 mM sucrose, 220 mM mannitol, 5 mM KH2PO4, 5 mM MgCl2, 1 mM EGTA, 2 mM HEPES pH 7.4) with protease inhibitor cocktail (Roche). Homogenates were centrifuged at 1,000g for 10 min at 4 °C and supernatant was collected. Protein concentrations were determined by BCA assay kit (Thermo Fisher). Homogenates (12 µg per well for human samples and 6 µg per well for mouse samples) were loaded into a Seahorse XF96 microplate in MAS buffer (20 µl each well) and centrifuged at 2,000g for 5 min at 4 °C. After centrifugation, the volume was increased to 150 µl by adding 130 µl MAS containing cytochrome c (10 µg ml−1). At port A, substrates of NADH (1 mM) were injected to determine the respiratory capacity of mitochondrial complex I; succinate (5 mM) + rotenone (2 µM) were injected to determine the respiratory capacity of mitochondrial complex II. The following compounds were injected so that final concentrations were as follows—port B: rotenone (2 µM) + antimycin (4 µM); port C: TMPD (0.5 mM) + ascorbic acid (1 mM); port D: azide (50 mM). OCR rates were measured using a Seahorse XF96 Extracellular Flux Analyzer (Agilent Technologies) and normalized to mitochondrial content quantified by MTDR. Homogenates were stained with 500 nM MTDR for 10 min followed by two wash steps to remove the dye (Thermo Fisher). MTDR fluorescence was read on a Tecan Spark plate reader (Ex: 633 nm; Em: 678 nm).
In vitro respirometry analysis on cultured cells
Cells were seeded into a Seahorse XF96 microplate before the assay and maintained in a tissue culture incubator (37 °C in 5% CO2) to reach 90–100% density before running the Seahorse assay. To measure complex I and complex II respiratory capacity, cells were permeabilized with 4 nM XF plasma membrane permeabilizer (Agilent). Permeabilized cells were started in state 3 with substrates of pyruvate (5 mM) + malate (0.5 mM) and 4 mM ADP for respiration-driven through complex I; succinate (5 mM) + rotenone (2 µM) and 4 mM ADP for complex II. Following state 3 measurements, injections included the following—port A: oligomycin (2 µM); ports B and C: FCCP (B: 0.75 µM and C: 1.35 µM); port D: rotenone (1 µM) + antimycin (2 µM). To measure OCR and extracellular acidification rate in intact cells, cells were washed twice and incubated with freshly prepared assay medium (Seahorse XF Base Medium + 2 mM l-glutamine + 1 mM pyruvate + 10 mM glucose) for 30 min. The following compounds were injected in the order of oligomycin (2 µM), FCCP (0.75/1.35 µM), rotenone (1 µM) + antimycin A (2 µM). Cell count (per well) was determined by the number of nuclei stained with Hoechst (10 µg ml−1) and quantified by an Operetta High-Content Imaging System (PerkinElmer). OCR and extracellular acidification rate were normalized to cell count per well. For nutrient-dependent respirometry analysis, conditions of UK5099 (5 µM), BPTES (3 µM) and etomoxir (3 µM) were applied.
Whole-animal perfusion and tissue fixation
Fixative solution (2% paraformaldehyde, 2.5% glutaradehyde, 0.15 M cacodylate and 2 mM Ca2+) was freshly prepared every time. Animals were anaesthetized by ketamine (200 mg kg−1) and xylazine (10 mg kg−1) delivered by intraperitoneal injection. Anaesthesia took effect after several minutes. The depth of anaesthesia was tested using tail pinch and paw prick, and it was ensured that the breathing did not stop. The body cavity was cut with scissors, and cut up the midline to the sternum. The heart was exposed and a needle was placed into the left ventricle and then the right atrium was snipped with iridectomy scissors. The animal was perfused for 30 s at the flow speed of 8 ml min−1 with Ringer’s solution supplemented with 2% xylocaine and 1,000 U herapin. Then the valve on the pump was switched on to fixative solution and perfusion was carried out for 5–8 min. After continuous perfusion, mouse lung tumour was collected, placed in ice-cold fixative solution and fixed overnight in the fridge. Fixed tissues were washed three times with 0.15 M cacodylate buffer (2 mM Ca2+) and stored in the same buffer at 4 °C until further processing.
Sample preparation for SBEM imaging
Post-fixed tissues of mouse lung tumour were washed in 0.15 M cacodylate buffer (2 mM Ca2+). Tissues were stained for 1 h with 2% osmium and 1.5% potassium ferrocyanide in 0.15 M cacodylate buffer (2 mM Ca2+). Tissues were washed five times (5 min each time) with double-distilled (dd)H2O, and then placed in filtered TCH buffer (0.05 g thiocarbohydrazide in 10 ml ddH2O) for 20 min at room temperature. Tissues were washed five times (5 min each time) with ddH2O and then stained with 2% osmium in ddH2O for 30 min at room temperature. Next, tissues were stained with 2% uranyl acetate overnight at room temperature and lead aspartate solution (0.66% (w/v) lead in 0.03 M aspartic acid) for 30 min in a 60 °C oven. Tissues were dehydrated in serial ice-cold ethanol (70%, 90% and 100%) and ice-cold acetone after washing in ddH2O. Tissues were then embedded in serial Durcupan resin (50%, 75%, 100%) and solidified in a 60 °C oven for 2 days.
SBEM
SBEM volumes were collected on a Zeiss Gemini 300 microscope equipped with a Gatan 3View 2XP microtome system. The volumes were collected at 2.5 kV using a 30-μm aperture, the gun in analytic mode, and the beam in high-current mode. Focal charge compensation with nitrogen gas was used to mitigate charging artefacts. The dwell time was 1 μs. The pixel size was either 5 or 6 nm, and the Z step size was always 50 nm. Following data collection, the images were converted to .mrc format and cross-correlation was used for rigid image alignment of the slices using the IMOD image processing package40.
MicroCT
MicroCT imaging was carried out on a Zeiss Versa 510 microscope. The wet lung specimens were imaged with no contrast staining in buffer at 40 kV, using 360° of specimen rotation and 801 projection images were collected. The pixel size was 11.5 μm. Specimens stained for SBEM and embedded in epoxy were imaged at 80 kV, using 360° of specimen rotation and 1,601 projection images were collected. The pixel size was between 5 and 8 μm.
3D visualization and analysis of SBEM images
Individual cells were segmented using Amira software (Thermo Scientific). Sequential SBEM images were processed with median filtering and contrast adjustment to enhance the appearance of intracellular space. Interactive thresholding was applied to distinguish cellular and extracellular space and generate binary images. Binary images were processed with the separate object algorithm to fragment cells without clear extracellular boundaries. The connected components algorithm was applied to create individual labels for each identified cell. Segmented individual cells in each stack were visualized with a voxelization rendering.
Machine-learning-based segmentation of nucleus and mitochondria in SBEM images
Our method segments nuclei (class 1) and mitochondria (class 2) from the background (class 0) in a 3D SBEM volume. The method takes individual slices from the volume (2D images) as input and predicts the segmentation for the slices independently. To carry out the segmentation, a deep CNN was used, which is a function that takes an image as input and outputs logits (or softmax confidence) corresponding to each class. The class with the highest response is chosen for each pixel to yield the segmentation map for each slice. Specifically, an encoder–decoder architecture based on U-Net41, with three modifications was chosen. First, ResNet42 blocks (with 32, 64, 128, 256 and 256 filters) was used instead of standard convolutional blocks. Second, as mitochondria are small objects, we chose to limit spatial downsampling to one-eighth of the original image. To do so, the max-pooling operation was removed after the third layer. Third, an atrous spatial pyramid pooling43 layer was used with strides 6, 12, 18 and 24 as the last layer of our encoder to increase the field of view of our filters. All weights use random uniform initialization, and we used leaky-ReLU activations. Our model is trained using the standard cross-entropy loss and is optimized using Adam44 with β1 = 0.9 and β2 = 0.999. We used an initial learning rate of 1 × 10−4 and decreased it to 1 × 10−5 after 150 epochs for a total of 200 epochs. We use a batch size of 8 and resize each image to 768 × 768. We carry out random horizontal and vertical flips, rotations between [−20°, 20°], zero-mean additive Gaussian noise with standard deviation of 0.08, and crop sizes up to 90% of the image height and width as data augmentation. Each augmentation was applied with a 50% probability. Our model was trained on ten densely labelled images, and was validated on three additional labeled images that are not included in the training set. Training takes about 8 h on an Nvidia GTX 1080 GPU, and inference takes about 11 ms per 2D image. Our segmentation model was evaluated using the Sorensen–Dice index (DICE), the Jaccard index (also referred to intersection-over-union, IoU), precision and recall. DICE and IoU are metrics to gauge the similarity of two sets, commonly used to evaluate segmentation results. Our segmentation model achieves an average of 0.92 DICE score, 0.86 IoU, 0.916 precision and 0.927 recall, as compared to that of manual registration.
Mitochondrial motility and mitochondrial crista analysis
A total of 2 × 105 cells were plated per well into CELLview 4-compartment glass-bottom tissue culture dishes (Greiner Bio-One) 48 h before the imaging section, and maintained in a tissue culture incubator (37 °C in 5% CO2). Cells were stained with TMRE (15 nM, Thermo Fisher) for 1 h to analyse mitochondrial motility, and with 10-N-nonyl acridine orange (100 nM, Thermo Fisher) for 1–3 h to analyse mitochondrial cristae. Live-cell imaging was carried out on the Zeiss LSM 880 with Airyscan using the alpha Plan-Apochromat 100×/1.46 Oil DIC M27 objectives. Images were deconvolved by Airyscan Processing in ZEN software. Image analysis was carried out using ImageJ (Fiji) in the order of: background subtraction, crop region of interest, adjust thresholds and measure parameters. Mitochondrial motility was analysed using a program developed in the laboratory of O.S.S. To measure mitochondrial motility, each field of view was imaged every 20 s, 10 times (yielding a total of 10 frames). Mitochondrial placement in frame 1 of each time series is identified as the reference point. The following 9 frames (n = 2, 3, 4…10) in the image set were then overlaid on the reference image, and the overlapping area between frame n and frame 1 was calculated and denoted as overlap_area(n). The non-overlapping mitochondria area between frame (n) and the reference image was denoted as travel_area(n) and calculated with the formula: travel_area(n) = mitochondria_area(n) − overlap_area(n). Mitochondrial motility was then calculated as the displacement ratio for each frame, ratio(n) = travel_area(n)/overlap_area(n). We then fitted a linear regression to the ratio for each image frame (n > 1) with the time intervals (20 s, 40 s…180 s), and the regression line is set in the form of ratio = a (slope) × time intervals + b (intercept). The value of “a (slope)” for each time series of image sets is our readout of mitochondrial motility. Macros were designed for automated segmentation and quantification of mitochondrial crista structure and density according to previously described methods19,21. The original images used for classification of mitochondrial cristae were 2D images, and the consensus for classification was obtained from two.
Quantification of distance between mitochondria and nucleus in fluorescent images
Given the centre of a mitochondria, we want to find its minimum distance to the surface of a nucleus, which is segmented as an ellipse. The result can be obtained by solving a constrained optimization problem using Lagrange multipliers. The original ellipse of the nucleus is tilted by a degree and is centred at an arbitrary coordinate. To ease the computation, we translate and rotate the original ellipse to a standard ellipse that is centred at (0,0) with the major axis aligned with the x axis and the minor axis aligned with the y axis. The centre of the mitochondria is mapped to the new coordinate system accordingly. Through substitution, the gradient of the objective function that involves only y is a fourth-degree polynomial, which has at most four possible roots that satisfy the criteria. We then choose the coordinate (x,y) on the ellipse that has the minimal distance to the target nucleus.
Immunohistochemistry
Immunohistochemistry staining and analysis were carried out as previously described11. Tissue slides were probed with the following antibodies: CD34 (1:800, ab8158 Abcam); TTF1 (1:1,000, M3575 Dako); CK5 (1:1,000, ab52635 Abcam).
Oil red O staining
The oil red O staining was carried out following the protocols of a commercial kit (StatLab, KTOROPT). Frozen tissues were fixed by O.C.T compound (Fisher HealthCare) and sectioned at the Translational Pathology Core Laboratory (UCLA). The frozen tissue sections were fixed in 10% formalin for 5 min, and then rinsed in distilled water. Slides were immersed in propylene glycol for 2 min, followed by immersion in preheated oil red O solution at 60 °C for 6 min. Slides were then washed in 85% propylene glycol for 1 min and running distilled water for 1 min. The nuclei were stained with modified Mayer’s haematoxylin for 1 min and rinsed in distilled water for 1 min. The slides were mounted with aqueous medium and covered with a coverslip.
Western blotting
Whole-cell lysates of lung tumours isolated from GEMMs were generated by homogenizing snap-frozen tumour tissues in lysis buffer (20 mM Tris pH 7.5, 150 mM NaCl, 1% (v/v) Triton X-100, 50 mM sodium fluoride, 1 mM EDTA, 1 mM EGTA, 2.5 mM pyrophosphate, 1 mM sodium orthovanadate, complete protease inhibitor cocktail). Whole-cell lysates were centrifuged at 2,000g for 5 min and supernatants were transferred to empty tubes. Supernatants were stored at −80 °C until use. Whole-cell lysates of in vitro-cultured cells were generated by homogenizing the cells SDS lysis buffer (100 mM Tris pH 7.5, 100 mM NaCl, 1% SDS, protease inhibitor cocktail) followed by heat inactivation at 90 °C for 10 min. Protein concentration was determined by BCA assay (Thermo Fisher). Lysates were run on 4–12% Bis-Tris gels (Thermo Fisher) to separate the proteins, and then transferred to PVDF membrane. Membranes were stained with Poceau S to confirm transfer efficiency. Membranes were then probed with the following antibodies: SP-C (1:5,000, AB3786 Millipore); GLUT1 (1:2,000, GT11-A, Alpha Diagnostic); NDUFS1 (1:1,000, ab169540, Abcam); O-linked N-acetylglucosamine (1:1,000, ab2739, Abcam); SDHA (1:1,000, 5839, Cell Signaling Technology); SDHC (1:1,000, ab155999, Abcam); actin (1:5,000, A3853, Sigma); tubulin (1:2,500, T9026, Sigma).
Statistics and reproducibility
Statistical analyses were carried out on GraphPad Prism 9 or R studio. Differences between groups were determined using unpaired two-tailed t-test or one-way ANOVA if more than two groups were compared. For treatment studies, Dunnett’s test was used to compare every mean to a control mean. For non-treatment studies, Tukey’s test was used to compare every mean to every other mean. Data are presented as mean ± s.e.m. unless specified otherwise. Numbers of biological replicates are indicated in the figure legends. All experiments were repeated in at least duplicate. No statistical methods were used to predetermine sample size.
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
More News
Daily briefing: The body odours that make mosquitoes want to bite you
Daily briefing: Why climate change is making flights rougher
How researchers in remote regions handle the isolation