Kras oncogene ablation prevents resistance in advanced lung adenocarcinomas

KRASG12C inhibitors have revolutionized the clinical management of patients with KRASG12C-mutant lung adenocarcinoma. However, patient exposure to these inhibitors leads to the rapid onset of resistance. In this study, we have used genetically engineered mice to compare the therapeutic efficacy and the emergence of tumor resistance between genetic ablation of mutant Kras expression and pharmacological inhibition of oncogenic KRAS activity. Whereas Kras ablation induces massive tumor regression and prevents the appearance of resistant cells in vivo, treatment of KrasG12C/Trp53-driven lung adenocarcinomas with sotorasib, a selective KRASG12C inhibitor, caused a limited antitumor response similar to that observed in the clinic, including the rapid onset of resistance. Unlike in human tumors, we did not observe mutations in components of the RAS-signaling pathways. Instead, sotorasib-resistant tumors displayed amplification of the mutant Kras allele and activation of xenobiotic metabolism pathways, suggesting that reduction of the on-target activity of KRASG12C inhibitors is the main mechanism responsible for the onset of resistance. In sum, our results suggest that resistance to KRAS inhibitors could be prevented by achieving a more robust inhibition of KRAS signaling mimicking the results obtained upon Kras ablation.

In vivo pharmacological treatments. Drug treatments were carried out in tumor-bearing K G12C P mice. Sotorasib (MedChemExpress) was administrated on 5 consecutive days per week by oral gavage (100 mg/kg), dissolved in 2% hydroxypropyl methylcellulose (HPMC) and 1% Tween 80. The tumor response was monitored by CT analysis.

Orthotopic and subcutaneous implantation of lung tumor cell lines. Implantations
were carried out in immunodeficient Foxn1 nu mice. Subcutaneous tumor formation was induced by injecting 10 6 cells resuspended in 100 µl of sterile PBS:Matrigel (1:1) (Corning, 354234) in the flank of anesthetized (4% isoflurane in 100% oxygen at a rate of 0.5 l/min) mice. The tumor size was measured periodically using a caliper and calculated as (short axis × short axis × long axis)/2. For transpleural orthotopic cell injections, 10 5 cells in 10 µl of sterile PBS:Matrigel (1:1) were injected through the intercostal space into the lung of anesthetized (4% isoflurane in 100% oxygen at a rate of 0.5 l/min) mice.

Implantation of mouse and PDX tumors in immunodeficient mice. PDX lung tumors
(1) and lung tumors from K G12C P mice were cut into small fragments (approx. 5 mm 3 ), embedded in Matrigel and introduced through a transversal incision in the flank of anesthetized (4% isoflurane in 100% oxygen at a rate of 0.5 l/min) mice. The tumor size 5 was measured periodically using a caliper and calculated as (short axis × short axis × long axis)/2.
Micro-CT Imaging. Mice were anesthetized by inhalation of 4% isoflurane in 100% oxygen at a rate of 0.5 l/min. The parameters for CT acquisition were set as follows: 100 µA, 50 kV, projections 360, 10 shots, set binning 2x2 and set exposure 100 milliseconds.
Lung images were acquired using a SuperArgus COMPACT (Sedecal). Image processing, analysis and 3D rendering was performed using the 3D Slicer Viewer Software. Tumor volume was calculated according to the following formula: (Short axis × short axis × long axis)/2. APC anti-F4/80 (1:400, clone BM8 eBioscience) and PE-Cy7 anti-CD11b (1:800, clone M1/70 BD Pharmingen) were included in a second multicolor panel for macrophage identification. 0.1 μg/ml of DAPI (Molecular Probes) was used to identify dead cells. 6 Samples were collected on a FACS Canto II flow cytometer (BD Biosciences) equipped with 488nm, 405nm and 633nm laser lines. Cell aggregates were excluded using pulse processing and at least 10,000 events of interest were collected. Data were analyzed using FlowJo V10 (BD Biosciences).
Cell proliferation and colony formation assays. 300 cells per well were dispensed in 96-well plates in DMEM supplemented with 10% FBS in triplicate, and proliferation was assessed by the MTT cell viability assay (Sigma) every two days. Cell growth was calculated relative to day 0. For 2D colony formation, 5000 cells per plate were cultured in 10 cm diameter dishes in DMEM supplemented with 10% FBS for 7-10 days. Then, plates were fixed with 0.1% glutaraldehyde and stained with 0.2% Crystal Violet.
In vitro drug treatments. Cells (4000 cells per well) were dispensed in 96-well plates in DMEM supplemented with 10% FBS and grown for 24h. Alternatively, cell lines were seeded on ultra-low attachment 96-well plates, with clear round bottom and black opaque microplate body (Corning, 4515) at a density of 4000 cells per well and grown for 4 days until spheroids formation. Then cells were treated with threefold dilutions of sotorasib (MedChemExpress), or with BAY 11-7082 (10 µM) and Stattic (5 µM) as single or combined treatments for 72h, and viability was determined by MTT or Cell Titer Glo assay (Promega). Every condition was seeded in triplicate and normalized to DMSOtreated controls. To calculate the IC50, values were plotted and fit to a sigmoid doseresponse curve using GraphPad Prism (v8.4.0) software. To activate NF-kB, cells were exposed to 20 mg/ml tumor necrosis factor a (TNF-a, R&D Systems). RAS activation assay. RAS-GTP levels were determined using the Ras Activation Assay kit (Cell Biolabs) according to manufacturer's instructions. Cell lysates were incubated with Raf-1 RBD (Ras Binding Domain) agarose beads for 1h at 4ºC in gentle agitation.
After washing with lysis buffer, the beads were denatured in 1X loading buffer and 1X reducing agent (NuPAGE, Invitrogen), and boiled for 5 min at 95ºC. Pull-down samples and whole-cell lysates were detected by Western blot using anti pan-RAS antibodies.  (2). Briefly, cells exposed to colcemid to arrest mitosis at the 9 metaphase stage were treated with a hypotonic solution and fixed with glacial acetic acid and methanol. After dehydration, the samples were denatured in the presence of the specific probe at 73°C and incubated overnight for hybridization at 37°C. Finally, the slides were washed in 20X SSC buffer with Tween 20 and mounted on fluorescent mounting media (DAPI in antifade solution). A Leica DM 5500B fluorescence microscope with a 100x oil-immersion objective, Leica DM DAPI, green, and orange fluorescence filter cubes, and a CCD camera (Photometrics SenSys camera) connected to a PC running the Zytovision image analysis system (Applied Imaging Ltd., UK) with Z stack software was used to image 200 cells. The z-stack images were manually scored by two independent investigators by counting the number of co-localized signals.
GST activity assay. GST activity was detected with the colorimetric GST Assay kit (PromoCell) following manufacturer's recommendations. A total of 50 µg of protein lysate was loaded in 96-well plates in duplicate and the reaction was initiated by adding glutathione (GSH) and CDNB (1-Chloro-2,4-dinitrobenzene). The absorbance was measured every minute at 340 nm and the specific GST activity (U/mg) was calculated according to the manufacturer's guidelines. following manufacturer's protocols. Efficient nuclear and cytoplasmic fractionation was confirmed by western blot analysis using Lamin B antibodies for the nuclear fraction and GAPDH antibodies for the cytoplasmic fraction.
Whole exome sequencing. Genomic DNA was isolated from tumors using the saltingout method. Then, DNA samples were subjected to quality control and subsequent library construction for whole exome sequencing (WES) using the SureSelect XT Mouse All  (5) to discard contaminating reads from mouse stroma. Good quality reads were then aligned to the Mus musculus (GRCm38/mm10) or human (GRCh38) genome reference using the Burrows-Wheeler Aligner (BWA-MEM) (6). Following the alignment, we used Picard to mark duplicates and Genome Analysis Toolkit (GATK) for base quality score recalibration. Collection of alignment and coverage metrics was performed with samtools and Picard. Targeted bases were sequenced to a mean depth of 100x, with at least 80% of targeted bases sequenced to 30x coverage or higher.
Germline and medium-to-high frequency variants were detected using the GATK algorithm HaplotypeCaller, while MuTect2 was used for the calling of somatic lowfrequency variants. Variants were selected using the hard-filtering recommendations from GATK, and annotated with SnpEff and Ensembl Variant Effect Predictor (VEP) tools (7,8). Those variants reported as single nucleotide polymorphisms (SNPs) by the NCBI database dbSNP were removed. When population allelic frequencies were available (human), we also considered as polymorphisms those with allelic frequencies higher than 0.01. In mice, we generated a list of likely germline events by using a pooled 13 sample of tail DNA from two mice of mixed (129/Sv-C57BL/6) background and subtracted them from the list of variants detected in tumor samples.
Copy number variations (CNVs) were detected by using CNVkit software (9). A pooled copy-number reference was constructed from two control samples (DNA from mouse tails). Then, each sample was individually compared with the reference to calculate copy number ratios, and circular binary segmentation (CBS) algorithm (10) was used to infer copy number segments. Copy number variants were labelled as losses or gains relative to the overall sample wide estimated ploidy. Genes with copy number gains or losses above a threshold of 2 (by default), equivalent to single-copy gains and losses in a completely pure tumor sample, were reported. Also, CNVs covering <3 bins were not considered.