Identifying resistance mutations in a drug target provides crucial information. Lentiviral transduction creates multiple types of mutations due to the error-prone nature of the HIV-1 reverse transcriptase (RT). Here we optimized and leveraged this property to identify drug resistance mutations, developing a technique we term LentiMutate. This technique was validated by identifying clinically relevant EGFR resistance mutations, then applied to two additional clinical anticancer drugs: imatinib, a BCR-ABL inhibitor, and AMG 510, a KRAS G12C inhibitor. Novel deletions in BCR-ABL1 conferred resistance to imatinib. In KRAS-G12C or wild-type KRAS, point mutations in the AMG 510 binding pocket or oncogenic non-G12C mutations conferred resistance to AMG 510. LentiMutate should prove highly valuable for clinical and preclinical cancer-drug development.

Significance:

LentiMutate can evaluate a drug's on-target activity and can nominate resistance mutations before they occur in patients, which could accelerate and refine drug development to increase the survival of patients with cancer.

Mutations in the target of a drug that confer resistance to this drug validates the drug's on-target activity and aids in the development of new drugs that circumvent clinical resistance, both by nominating potentially clinically relevant resistance mutations and by providing chemical–structural information about the target and the drug. Thus, the identification of resistance-conferring mutations in drug targets can inform both clinical and preclinical drug development programs. Multiple techniques, such as in vitro error-prone PCR (1), plasmid mutagenesis in bacteria with error-prone DNA replication (2), ethylnitrosourea (ENU) mutagenesis (3), use of mismatch repair-deficient cell lines (4), and Mutagenesis by Integrated TilEs followed by sequencing (MITE-seq; ref. 5) have been developed to identify resistance-conferring mutations. While each technique has unique strengths and weaknesses, all these techniques are limited by either solely identifying point mutations (2, 3), challenges in sequencing (3, 4), or restraints in the potential sequence space (5). Additionally, these techniques can be technically challenging, laborious, and/or expensive.

Lead contact and materials availability

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, R. Kittler. All unique/stable reagents generated in this study are available from R. Kittler without restriction.

Cell lines, cell culture, and patient samples

All cell lines were grown in RPMI-1640 (Millipore Sigma, catalog no. R8758) supplemented with 5% FBS (ThermoFisher Scientific, catalog no. 26140079). Cells were maintained in a 37°C incubator (NuAire, model NU-5810) with 5% CO2 and 95% relative humidity. Cell lines were routinely checked for Mycoplasma contamination (E-myco, catalog no. 25233, Bulldog Bio) and DNA fingerprinted (PowerPlex Fusion24, catalog no. DC2402, Promega). The following cell lines that were utilized for this paper were established by J.D. Minna and A. Gazdar at the NIH (referred to as NCI-Hxxxx) or UT Southwestern Medical Center (UTSW; referred to as HCCxxxx) and are available through ATCC or the Hamon Center for Therapeutic Oncology at UTSW: NCI-H1993 (RRID:CVCL_1512), NCI-H2009 (RRID:CVCL_1514), NCI-H358 (RRID:CVCL_1559), HCC827 (RRID:CVCL_2063). K562 (RRID:CVCL_0004) was obtained from ATCC (ATCC CCL-243). Lenti-X 293T cells were obtained from Takara (catalog no. 632180).

Samples of patients with chronic myeloid leukemia

After acquiring Institutional Review Board–approved informed written consent, bone marrow or peripheral blood specimens were collected from newly diagnosed patients with chronic myeloid leukemia (CML) or those with clinical resistance to one or more ABL1 tyrosine kinase inhibitor (TKI) treatments. Mononuclear cells were isolated by Ficoll-gradient centrifugation, and total RNA or genomic DNA was extracted using the Qiagen RNeasy or DNeasy kit, respectively. Samples were subjected to paired-end RNA sequencing (RNA-seq) or whole-exome sequencing (WES; details in the “RNA-seq and Exome-seq on samples of patients with CML” subsection below); ABL1 data is reported here and deposited with NCBI Gene Expression Omnibus (GEO; accession number GSE164664).

Immunoblots

Cells were washed with ice-cold PBS and then harvested via scraping, centrifugation, or trypsinization. Cells were lysed in RIPA buffer (50 mmol/L Tris, 150 mmol/L NaCl, 0.1% SDS, 1% IGEPAL CA-630, 1% sodium deoxycholate, 2 mmol/L MgCl2, pH 8) with 1 unit/μL benzonase (MilliporeSigma, catalog no. E1014), protease inhibitors (MilliporeSigma, catalog no. P8340), and phosphatase inhibitors (MilliporeSigma, catalog no. 4906845001) by rotating lysates at 4°C for 2 hours. Lysates were spun at max speed for 10 minutes, quantified using bicinchoninic acid (BCA; ThermoFisher Scientific, catalog no. 23225), mixed with 2X Laemmli buffer (BioRad, catalog no. 1610737EDU) and boiled for 5 minutes. 20 to 25 μg of protein was run on a 4% to 20% Mini-PROTEAN TGX gel (BioRad, catalog no. 4561096) at 200 V. Samples were transferred using the Trans-Blot Turbo RTA Mini Nitrocellulose transfer kit (Biorad, catalog no. 1704270) on the Trans-Blot Turbo Transfer System (Biorad, catalog no. 1704150) using manufacturers suggested protocols. After transfer, membranes were blocked for 1 hour using TBS Odyssey Blocking Buffer (LI-COR, catalog no. 927–50100), probed overnight at 4°C or at room temperature for 2 hours with primary antibody diluted in TBS Odyssey Blocking Buffer, washed three times for 5 minutes with TBST, incubated for 1 hour at room temperature with secondary antibodies diluted in TBS Odyssey Blocking Buffer, washed three times for 5 minutes with TBST, washed once with TBS for 5 minutes, then imaged using a LiCor Odyssey Fc. The following primary antibodies and dilutions were used: ABL1 1:500 (Santa Cruz Biotechnology, catalog no. sc-131) and GAPDH 1:5000 (GeneTex, catalog no. GTX627408). The following secondary antibodies were used at 1:10,000: goat polyclonal anti-Mouse IRDye 680RD conjugate (LI-COR, catalog no. 925–68070) and goat polyclonal anti-Rabbit IRDye 800CW conjugate (LI-COR, catalog no. 925–32211).

Lentivirus production

All cDNAs were cloned into the doxycycline-inducible lentiviral expression plasmid pLVX-TRE3G-IRES. To generate lentiviral particles, 1.5 million Lenti-X 293T cells (Clontech, catalog no. 632180) were forward transfected with 9 μg pCMV-dR8.91 (lentiviral packaging plasmid), 3 μg pMD2.G (lentiviral envelope plasmid) and 3 μg of pLVX-TRE3G-IRES, pLVX-EF1a-Tet3G, or pLentiCRISPRv2 using FuGene6 (Promega, catalog no. E2691) following the manufacturers protocol. After 12 hours, media was changed and viral supernatant was collected every day for 3 days and filtered through a 0.45 micron syringe filter (Corning, catalog no. 431220). When necessary, multiple lentiviral preps were pooled together and concentrated using Lenti-X Concentrator (Takara, catalog no. 631231). Cells were then transduced using the lentiviral supernatant and 6 μg/mL polybrene (Santa Cruz Biotechnology, catalog no. sc-134220). See the “Performing LentiMutate to identify drug resistance mutations” subsection of these Materials and Methods for details on how to perform LentiMutate.

Expression vectors

All cDNAs were cloned into the doxycycline-inducible lentiviral expression plasmid pLVX-TRE3G-IRES. RUVBL1 cDNA was reverse transcribed (SuperScript III, ThermoFisher catalog no. 18080051) from the H2009 cell line. EGFRdel746–750 cDNA was reverse transcribed from the HCC827 cell line. The HIV reverse transcriptase (RT)–M230I mutant (M-RT) was generated by site-directed mutagenesis of codon 230 (ATG→ATC) in the RT in the plasmid pCMV-dR8.91 with the Q5 Site-Directed Mutagenesis Kit (New England BioLabs, catalog no. E0554S). The BCR-ABL1 open reading frame was PCR amplified from the plasmid MSCV-(pBabemcs)-humanp210BCR-ABL-IRES-GFP that was a gift from Martine Roussel (Addgene plasmid # 79248; RRID:Addgene_79248), and the position of deleted nucleotides in this open reading frame for SV8 and SV11 variants can be found in Supplementary Table S6. For NTCD, the nucleotides 907–1122 were removed. KRAS–wild-type (WT) and KRAS-G12C cDNA was chemically synthesized using a codon-optimized reading frame (sequence shown below) because use of the endogenous cDNA sequence from H358 resulted in extremely low expression.

>KRAS-WT

ATGACAGAATATAAACTGGTAGTAGTAGGTGCAGGTGGAGTTGGAAAATCCGCTCTTACAATTCAACTCATCCAAAACCACTTCGTCGATGAGTACGACCCCACTATCGAAGACAGTTATAGAAAACAGGTCGTTATCGACGGCGAGACGTGCCTTCTCGACATCCTGGATACCGCCGGCCAGGAAGAATATAGCGCCATGCGGGATCAATATATGCGTACAGGCGAAGGGTTCCTGTGCGTGTTCGCAATTAACAACACGAAGTCCTTCGAGGACATCCATCACTACCGGGAGCAGATCAAGCGTGTGAAAGATAGCGAGGACGTCCCGATGGTGCTTGTGGGCAACAAGTGCGATCTCCCAAGCCGAACTGTGGATACCAAGCAAGCACAAGATCTGGCTCGGTCCTACGGCATCCCCTTCATCGAGACGTCTGCCAAAACGAGGCAAGGCGTAGACGACGCATTTTACACGCTGGTGAGAGAGATCCGCAAGCACAAGGAGAAAATGTCTAAGGACGGAAAGAAGAAGAAGAAGAAGAGCAAAACGAAATGCGTGATCATGTAG.

>KRAS-G12C

ATGACAGAATATAAATTGGTCGTTGTTGGAGCTTGTGGTGTCGGGAAATCTGCACTTACCATTCAATTGATACAAAACCACTTCGTAGATGAGTACGACCCTACCATTGAAGACTCTTATAGAAAACAGGTTGTGATAGACGGGGAGACATGCCTGCTTGACATCCTGGATACTGCGGGACAGGAAGAATATTCAGCCATGCGCGATCAATATATGCGGACAGGCGAAGGGTTCCTGTGCGTGTTCGCTATTAACAACACAAAGAGCTTCGAGGACATACATCACTACCGTGAGCAGATCAAGCGGGTGAAAGATTCAGAGGACGTGCCCATGGTTCTCGTGGGCAACAAGTGCGACCTTCCCAGTAGGACTGTCGATACGAAGCAAGCCCAAGATCTGGCCCGGAGCTACGGTATCCCATTCATCGAGACTAGCGCTAAAACTCGGCAAGGCGTGGACGACGCATTTTACACTCTCGTCCGCGAGATCAGAAAGCACAAGGAGAAAATGTCTAAGGACGGCAAGAAGAAGAAGAAGAAATCCAAAACCAAATGCGTGATCATGTGA.

Performing LentiMutate to identify drug resistance mutations

To perform LentiMutate millions of cells are transduced with approximately one lentiviral integration per cell [i.e., transduction performed at a multiplicity of infection (MOI) of approximately 0.3], with greater numbers of cells being transduced for larger DNA/cDNA cargos to saturate the possible mutagenic space. For each batch of virus used on each target cell line we determined MOI in our experiments empirically by titrating the amount of virus such that approximately 30% of cells were resistant to puromycin, as the lentivirus we use contains a puromycin-resistance cassette. We estimate that the M-RT makes, on average, roughly 1 error per 3,500 nucleotides during transduction (Fig. 1D); thus, on average for a 1,000 bp cargo, we would expect: 1 mutation per 3,500 bases during transduction * 1,000 base pair cargo = 0.2857 mutations per transduced cells. Because each nucleotide in the cargo could be mutated to 3 different nucleotides, to saturate a 1,000 bp cargo, 3,000 (1,000 * 3) mutations would need to be created. To determine the minimal number of cells that would need infected: 3,000 mutations/0.2857 mutations per transduced cells = 10,500 transduced cells.

Figure 1.

Lentiviral mutagenesis can identify drug resistance mutations and can be improved by use of a more error-prone RT. A, WT- or PCR- mutated doxycycline-inducible cDNAs for RUVBL1 were transduced into H1993 cells and treated with a lethal (500 nmol/L) dose of the RUVBL1/2 inhibitor compound B in the presence (+doxycycline, 1 μg/mL) or absence (−doxycycline) of these cDNAs. Top, representative crystal violet-stained plates from one replicate. Bottom, RUVBL1 cDNAs were sequenced in resistant cells, and parental cells and mutation frequency between resistant and parental cells were compared. Positions in red are >5-fold enriched and overlap between experiments. Values are the average of two biological replicates. B, Outline of LentiMutate. C, Similar to Fig. 1A but using the cell line HCC827, the EGFRdel746–750 cDNA, and 1 μmol/L gefitinib. Top, representative crystal violet-stained plates from one replicate. Bottom, comparison of mutation frequencies in resistant cells with parental cells. Values are the average of two biological replicates. One μg/mL doxycycline was used. D, Sequencing of unique molecular identifier-tagged lentiviral integrants from cells transduced with EGFRdel746–750 cDNA using either WT-RT or M-RT to determine error rate. E, LentiMutate using the cell line HCC827, the EGFRdel746–750 cDNA, 100 nmol/L osimertinib, and the M-RT produced osimertinib-resistant cells with the EGFR-C792S mutation. Top, representative crystal violet-stained plates from one replicate. Bottom, comparison of mutation frequencies in resistant cells with parental cells. Values are the average of two biological replicates. One μg/mL doxycycline was used.

Figure 1.

Lentiviral mutagenesis can identify drug resistance mutations and can be improved by use of a more error-prone RT. A, WT- or PCR- mutated doxycycline-inducible cDNAs for RUVBL1 were transduced into H1993 cells and treated with a lethal (500 nmol/L) dose of the RUVBL1/2 inhibitor compound B in the presence (+doxycycline, 1 μg/mL) or absence (−doxycycline) of these cDNAs. Top, representative crystal violet-stained plates from one replicate. Bottom, RUVBL1 cDNAs were sequenced in resistant cells, and parental cells and mutation frequency between resistant and parental cells were compared. Positions in red are >5-fold enriched and overlap between experiments. Values are the average of two biological replicates. B, Outline of LentiMutate. C, Similar to Fig. 1A but using the cell line HCC827, the EGFRdel746–750 cDNA, and 1 μmol/L gefitinib. Top, representative crystal violet-stained plates from one replicate. Bottom, comparison of mutation frequencies in resistant cells with parental cells. Values are the average of two biological replicates. One μg/mL doxycycline was used. D, Sequencing of unique molecular identifier-tagged lentiviral integrants from cells transduced with EGFRdel746–750 cDNA using either WT-RT or M-RT to determine error rate. E, LentiMutate using the cell line HCC827, the EGFRdel746–750 cDNA, 100 nmol/L osimertinib, and the M-RT produced osimertinib-resistant cells with the EGFR-C792S mutation. Top, representative crystal violet-stained plates from one replicate. Bottom, comparison of mutation frequencies in resistant cells with parental cells. Values are the average of two biological replicates. One μg/mL doxycycline was used.

Close modal

However due to biases in the substitution rate at each nucleotide, we multiply the minimal number of cells needed by 100, roughly the difference between the frequency of the most likely and least likely substitutions (10,500 cells × 100 = 1,050,000 cells). Then, to ensure that the rarest mutations are made multiple times, we multiple this number by 5 (1,050,000 × 5 = 5,250,000). Thus, to perform LentiMutate on a 1,000 bp cargo, we recommend transducing at least 5,250,000 cells and this number of cells must be maintained throughout the entire experiment. For smaller or larger cargos the numbers can be scaled accordingly. Greater numbers of infected cells will yield more reliable and reproducible results. For our experiments, cells containing TET3G (pLVX-EF1a-Tet3G) were transduced at an MOI of 0.3, such that for pLVX-TRE3G-EGFRdel746–750-IRES 2 million cells were infected, for pLVX-TRE3G-RUVBL1-IRES 500,000 cells were infected, for pLVX-TRE3G-BCR-ABL-IRES 15 million cells were infected, and for pLVX-TRE3G-KRAS-(G12C and WT) 3 million cells were infected. After infection and puromycin selection, for the RUVBL1/2 inhibitor compound-B experiments 500,000 H1993 cells/dish were plated into 4 15-cm dishes. Two plates received 1 μg/mL doxycycline for 2 days, and after 2 days 500 nmol/L compound B+ doxycycline (to respective plates) was added. Media was changed every 3 days until large resistant colonies emerged in the plates with doxycycline. For gefitinib/osimertinib/AMG 510 experiments, after infection and puromycin selection 2 million cells/dish were plated into 4 15-cm dishes. Two plates received 1 μg/mL doxycycline for 2 days, and after 2 days 1 μmol/L gefitinib, 100 nmol/L osimertinib or 1 μmol/L AMG 510, and + doxycycline (to respective plates) was added. Media was changed every 3 days until resistant colonies emerged in the plates with doxycycline. These cells were then trypsinized and split 1:2 with continued drug treatment to aid in the outgrowth of true resistant cells from persistors. For imatinib experiments, after infection and selection 15 million cells/dish were plated into 2 T175 tissue-culture treated flasks. One flask received 1 μg/mL doxycycline for 2 days, and after 2 days 1 μmol/L imatinib + doxycycline was added to respective plates. Cells treated with doxycycline + imatinib were pelleted after approximately 21 days of treatment. For all experiments, to identify resistance mutations, cells that survived drug treatment in the drug + doxycycline groups and the parental cell line were harvested and genomic DNA was isolated (Qiagen DNeasy Blood & Tissue Kit, catalog no. 69504). For EGFR the entire open reading frame, for BCR-ABL1 2 large amplicons, for RUVBL1 3 short barcoded amplicons, and for KRAS-WT/KRAS-G12C 2 short barcoded amplicons were then PCR amplified, purified (Qiagen QIAquick PCR purification kit catalog no. 28104), ran on a D5000 HS ScreenTape on the Agilent Tapestation 4200 and DNA concentrations determined by Qubit. For EGFR and BCR-ABL1 the PCR products were sheared on the Covaris S2 for 40 seconds and libraries were prepared with the KAPA Hyper Prep Kit. Samples were end repaired, 3′ ends were adenylated and then barcoded with multiplex adapters. Samples were size selected prior to PCR to remove small fragments. The short RUVBL1 and KRAS-WT/KRAS-G12C PCR products were directly (i.e., without sonication) ligated with the adapters. Four cycles of PCR were used to amplify the libraries, which were then purified with AmpureXP beads and validated on the Agilent Tapestation 4200. Before being normalized and pooled, samples were quantified by Qubit then ran using the Nextseq500 V2.5 150PE or MiSeq V2.5 250/300PE chemistry. Reads obtained from pooled samples were separated into sample-specific readIDs based on the barcode information. For each read in the pooled sample fastq file, the barcode and primer sequences were trimmed from the 5′ end. Using the trimmed fastq file and readIDs obtained in the first step, sample-specific fastq files were generated using BBMap (v 38.11). These reads were mapped to the reference cDNA sequence using BWA (v. 0.7.15). For identification of point mutations, read pileup information was obtained from mapped reads using the samtools (v.0.1.19) mpileup command. From the read pileup information, per base mismatch/indel information was obtained using pileup2base tool. Per base error ratios were computed using an in-house Perl script, and the frequency of mutations at positions was compared between the parental (no doxycycline, no drug) and +doxycycline +drug surviving cells. For identification of deletions, the trimmed reads were mapped to the reference cDNA sequences using speedseq, which in turn uses BWA-mem algorithm to map the reads. In addition to the alignment, speedseq tool marks duplicates and identifies split and discordant read pairs. The speedseq mapped files were used to call structural variations (SV) using Lumpy SV caller. Lumpy SV calls were further filtered using number of split and discordant supporting reads.

Inhibitors and antibiotics

Puromycin was purchased from InvivoGen (catalog no. ant-pr-1) and used at 1.5 μg/mL. Geneticin (G418) was purchased from ThermoFisher Scientific (catalog no. 11811031) and used at 800 μg/mL. Doxycycline was purchased from MilliporeSigma (catalog no. D9891) and used at 1 μg/mL unless otherwise stated. Compound B was synthesized by Daiichi-Sankyo and used at 500 nmol/L to identify drug-resistant cells. Gefitinib was purchased from Selleck Chemicals (catalog no. S1025) and used at 1 μmol/L to identify drug-resistant cells. Osimertinib was purchased from Selleck Chemicals (catalog no. S7297) and used at 100 nmol/L to identify drug-resistant cells. Imatinib was purchased from Selleck Chemicals (catalog no. S2475) and used at 1 μmol/L to identify drug-resistant cells. AMG 510 was purchased from MedChemExpress (catalog no. HY-114277) and used at 1 μmol/L to identify drug-resistant cells. MRTX849 was purchased from SelleckChemicals (catalog no. S8884).

Analysis of RT mutation rates

Genomic DNA was isolated (Qiagen DNeasy Blood & Tissue Kit, catalog no. 69504) from HCC827 cells transduced at an MOI of approximately 0.3 with EGFRdel746–750 cDNA that was packaged using either the WT-RT or M-RT and otherwise identical conditions. Two microgram of this DNA was used for two PCR cycles (using New England Biolabs Q5 Hot Start Polymerase, catalog no. M0494S) of a 243 bp amplicon of the EGFR open reading frame using the primers TTGCGGTGTCAATCTTACCANNNNNNNNCAGCGCTACCTTGTCATTCA and ATGAATCCAAACCGGATGAANNNNNNNNTTGCAGCCCATTTCTATCAA to label both ends of individual amplicon molecules with unique molecular identifiers (UMI) of 8 nucleotides. After removal of the free UMI-containing primers by Exonuclease I (NEB) digestion and column purification (Qiagen QIAquick PCR purification kit catalog no. 28104), the UMI-tagged amplicons were then further amplified using the primers TTGCGGTGTCAATCTTACCA and ATGAATCCAAACCGGATGAA and purified (Qiagen QIAquick PCR purification kit catalog no. 28104) to generate 299 bp products. For these, barcoded Illumina libraries were prepared with the KAPA HyperPrep Kit (Roche) for sequencing on the NextSeq500 (Illumina) with V2.5 150PE chemistry. PE reads were demultiplexed and from each read in the fastq file, 20bp primer was trimmed from the 5′ end and 8bp was extracted, which contained the UMI information. We used UMIs in which all base pairs had a sequence quality value of more than 30 for further analysis. The UMI combinations from plus and minus reads were quantified and assigned a UMI combination and its count to each readID. After preprocessing (including further trimming of the UMI and the sequences of the second primers), singletons from preprocessed fastq files were separated using fastq-pair tool. Resulting fastq files were mapped to the EGFR cDNA sequence using BOWTIE2 (v.2.3.5). To identify mismatches in each mapped sequence we used samtools calmd command and custom in-house python and perl scripts. First, for each read we identified mismatches in each position and combined this information with UMI counts. Then, for each base-pair mismatch, we calculated the ratio between the total number of reads that show a given mismatch and the total number of reads sequenced (given by total number of UMIs). Events with greater than 0.9 mismatch rate were used for further analysis.

Deleting endogenous BCR-ABL1 in the presence of exogenous BCR-ABL1

To make deletions in the endogenous BCR-ABL1, while sparing the exogenously expressed BCR-ABL1 deletion variants, we targeted a gRNA to the intron–exon junction of BCR-ABL1 in cells expressing either the SV11 or NTCD BCR-ABL1 deletion variants from doxycycline-inducible promoters. Cells stably transduced with pLVX-EF1a-Tet3G and pLVX-TRE3G-BCR-ABL1-SV11 or pLVX-TRE3G-BCR-ABL1-NTCD were cultured in the presence of 200 ng/mL doxycycline for 24 hours and then transduced with GFP-expressing LentiCrisprV2 with a gRNA targeting the intron–exon junction of endogenous BCR-ABL1 (gRNA sequence: GAAAAACTTCATCCACAGGT). After 24 hours, single GFP-positive cells were flow sorted into individual 96-well plates and clones were tested for dependency to doxycycline. The resulting clones that showed deletion of endogenous BCR-ABL1 while retaining expression of the exogenous vectors are called “gBCR-ABL1-SV11” or “gBCR-ABL1-NTCD” in the manuscript, depending on whether K562 cells expressing BCR-ABL1-SV11 or BCR-ABL1-NTCD were used first.

RNA-seq and Exome-seq on samples of patients with CML

For RNA-seq, primary RNA of patients with CML was subjected to 2×125 bp paired-end Illumina sequencing using the KAPA Stranded Total RNASeq kit at the New York Genome Sequencing Center. Alignments were performed using STAR v2.5.4b, parameterized for fusion detection (https://github.com/STAR-Fusion/STAR-Fusion/wiki). Illumina WES data was generated for primary genomic DNA samples of patients with CML using the Nimblegen v3 library kit (OHSU Massively Parallel Sequencing Resource) and an Illumina HiSeq 3000 sequencer (Oregon State University Center for Genome Research and Biocomputing). Alignments were performed using BWA with preprocessing and genotyping performed using GATK v3.3. Specifically, the GATK UnifiedGenotyper was used to call small indels and SNVs.

From the RNA-seq data, reads were mapped to the reference ABL1 cDNA sequence using STAR to generate bam files and converted into fastq files using bedtools bamtofastq (v2.29.2). We inspected Sashimi plots generated from these bam files with the Integrative Genomics Viewer (IGV) to identify alternatively spliced transcripts. We then mapped fastq files to the alternative exon–exon spanning sequences of these alternatively spliced transcripts using bowtie (v2.1.0) and quantified the reads using a custom Unix script.

Expected sequence of novel BCR-ABL1 splicing isoforms identified in patient samples

>“WT BCR ABL1”, canonical_transcript_Exon10-Exon11

AAGACGAGGACCTCCAGGAGAGCTGCAGAGCACAGAGACACCACTGACGTGCCTGAGATGCCTCACTCCAAGGGCCAGGGAGAGAGCGATCCTCTGGACCATGAGCCTGCCGTGTCTCCATTGCTCCCTCGAAAAGAGCGAGGTCCCCCGGAGGGCGGCCTGAATGAAGATGAGCGCC

>Exon10-Exonv1-Exon11

AAGACGAGGACCTCCAGGAGAGCTGCAGAGCACAGAGACACCACTGACGTGCCTGAGATGCCTCACTCCAAGGGCCAGGGAGAGAGCGGTATACCCAAGACTGGGTAATTTATAAAGGAAAGAGGTTTCACTGACTCACAGTTCCACATGGCTGGGGAGGCCTCACAATCATGGCTGAAGGTGAATGAGGAGCAAGGTCACATCTTACTTGGCGGCAGGCAAGAGAGCTTGTGCAGGGGAAGTCCGCTTTATAAAACCATCAGATCTCGTGAGACTTATTCACTACCACGAGAATGTGGGAGAAACCTCCCCATGATTCATTGATCTCCACCTGACCCCACCGTTGACACGTGGGGCTTATTACAATTCAAGATCCTCTGGACCATGAGCCTGCCGTGTCTCCATTGCTCCCTCGAAAAGAGCGAGGTCCCCCGGAGGGCGGCCTGAATGAAGATGAGCGCC

>Exon10-Exonv2-Exon11

AAGACGAGGACCTCCAGGAGAGCTGCAGAGCACAGAGACACCACTGACGTGCCTGAGATGCCTCACTCCAAGGGCCAGGGAGAGAGCGACTGGGTAATTTATAAAGGAAAGAGGTTTCACTGACTCACAGTTCCACATGGCTGGGGAGGCCTCACAATCATGGCTGAAGGTGAATGAGGAGCAAGGTCACATCTTACTTGGCGGCAGGCAAGAGAGCTTGTGCAGGGGAAGTCCGCTTTATAAAACCATCAGATCTCGTGAGACTTATTCACTACCACGAGAATGTGGGAGAAACCTCCCCATGATTCATTGATCTCCACCTGACCCCACCGTTGACACGTGGGGCTTATTACAATTCAAGATCCTCTGGACCATGAGCCTGCCGTGTCTCCATTGCTCCCTCGAAAAGAGCGAGGTCCCCCGGAGGGCGGCCTGAATGAAGATGAGCGCC

>Exon10-Exonv3-Exon11

AAGACGAGGACCTCCAGGAGAGCTGCAGAGCACAGAGACACCACTGACGTGCCTGAGATGCCTCACTCCAAGGGCCAGGGAGAGAGCGTTCCACATGGCTGGGGAGGCCTCACAATCATGGCTGAAGGTGAATGAGGAGCAAGGTCACATCTTACTTGGCGGCAGGCAAGAGAGCTTGTGCAGGGGAAGTCCGCTTTATAAAACCATCAGATCTCGTGAGACTTATTCACTACCACGAGAATGTGGGAGAAACCTCCCCATGATTCATTGATCTCCACCTGACCCCACCGTTGACACGTGGGGCTTATTACAATTCAAGATCCTCTGGACCATGAGCCTGCCGTGTCTCCATTGCTCCCTCGAAAAGAGCGAGGTCCCCCGGAGGGCGGCCTGAATGAAGATGAGCGCC

>Exon10-Exonv4(part of Exon11)

AAGACGAGGACCTCCAGGAGAGCTGCAGAGCACAGAGACACCACTGACGTGCCTGAGATGCCTCACTCCAAGGGCCAGGGAGAGAGCGAAATGGTTTCCTCTGGATCGTTTTATGCGGTTCTTACAGCACATCACCTCTTTGCCCCCGACGGCTGTGACGCAGCCGGAGGGAGGCACTAGTCACCGACAGCGGCCTTGAAGACAGAGCAAAGCGCCCACCCAGGTCCCCC

Dose-response curves

One thousand five hundred to 2,500 cells in 50 μL media were plated into a 96-well plate in the presence (200 ng/mL Fig. 3D and Fig. 3G; 1,000 ng/mL Fig. 2B; or 500 ng/mL Fig. 4G) or absence of doxycycline with at least 8 technical replicates per condition. After 24 hours 50 μL of drugs were serially diluted and added (in the presence or absence of doxycycline) to cells and allowed to incubate for 4 days. After 4 days 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium (MTS; Promega catalog no. G3582) was added, absorbance was measured using a Molecular Devices SpectraMax 190 microplate reader. Absorbance values were normalized to respective number of drug controls. Curve fitting was performed in GraphPad Prism using a nonlinear fit of inhibitor versus normalized response – variable slope.

Figure 2.

LentiMutate with BCR-ABL1 identifies deletions in BCR-ABL1 that confer imatinib resistance. A, Top, cartoon domain structure of the BCR-ABL1 fusion protein p210. Bottom, BCR-ABL1 SVs inferred by sequencing of BCR-ABL1 in imatinib-resistant K562 cells after LentiMutate. B, Viability-based dose-response curves following imatinib or asciminib treatment in K562 cells harboring doxycycline-inducible BCR-ABL1 variants (35INS, NTCD, SV8, and SV11) with (+ dox) or without (−dox) expression of the BCR-ABL1 variant. Values shown are averages ± SEM of two biological replicates. Relative viability refers to the viability of cells treated with inhibitor relative to the viability of cells without inhibitor. One μg/mL doxycycline was used.

Figure 2.

LentiMutate with BCR-ABL1 identifies deletions in BCR-ABL1 that confer imatinib resistance. A, Top, cartoon domain structure of the BCR-ABL1 fusion protein p210. Bottom, BCR-ABL1 SVs inferred by sequencing of BCR-ABL1 in imatinib-resistant K562 cells after LentiMutate. B, Viability-based dose-response curves following imatinib or asciminib treatment in K562 cells harboring doxycycline-inducible BCR-ABL1 variants (35INS, NTCD, SV8, and SV11) with (+ dox) or without (−dox) expression of the BCR-ABL1 variant. Values shown are averages ± SEM of two biological replicates. Relative viability refers to the viability of cells treated with inhibitor relative to the viability of cells without inhibitor. One μg/mL doxycycline was used.

Close modal
Figure 3.

Deletions in BCR-ABL1 confer resistance to imatinib and similar deletions may be generated in patients with CML via alternative splicing. A, Outline of experiment performed to create cells used in Fig. 3B–D. B, Immunoblot demonstrating loss of endogenous BCR-ABL1 and expression of exogenous BCR-ABL1. C, K562 cells lacking endogenous BCR-ABL1 require expression of the BCR-ABL1 mutants for their viability. Background subtracted absorbance from the MTS assay is shown. D, Expression of BCR-ABL1 deletion mutants without endogenous BCR-ABL1 confers resistance to imatinib. Two hundred ng/mL doxycycline was used. E, Heatmap illustrating the percent of split reads supporting each novel isoform (rows) in 120 CML patient samples (columns). “Exon10-v1” denotes reads between ABL1 exon 10 and the novel v1 exon; “v1+v2+v3+v4” denotes the summation of split reads between exon 10 and any novel exon in each patient; “v1/2/3/4-exon11” denotes split reads between the novel exon and ABL1 exon 11 (summed for each patient). F, Cartoon diagram of the various BCR-ABL1 isoforms identified by split reads. G, Expression of BCR-ABL1 lacking ABL1 exon 11 in K562 confers resistance to imatinib. All values shown are averages ± SEM of two biological replicates. Relative viability refers to the viability of cells treated with inhibitor relative to the viability of cells without inhibitor. Two hundred ng/mL doxycycline was used.

Figure 3.

Deletions in BCR-ABL1 confer resistance to imatinib and similar deletions may be generated in patients with CML via alternative splicing. A, Outline of experiment performed to create cells used in Fig. 3B–D. B, Immunoblot demonstrating loss of endogenous BCR-ABL1 and expression of exogenous BCR-ABL1. C, K562 cells lacking endogenous BCR-ABL1 require expression of the BCR-ABL1 mutants for their viability. Background subtracted absorbance from the MTS assay is shown. D, Expression of BCR-ABL1 deletion mutants without endogenous BCR-ABL1 confers resistance to imatinib. Two hundred ng/mL doxycycline was used. E, Heatmap illustrating the percent of split reads supporting each novel isoform (rows) in 120 CML patient samples (columns). “Exon10-v1” denotes reads between ABL1 exon 10 and the novel v1 exon; “v1+v2+v3+v4” denotes the summation of split reads between exon 10 and any novel exon in each patient; “v1/2/3/4-exon11” denotes split reads between the novel exon and ABL1 exon 11 (summed for each patient). F, Cartoon diagram of the various BCR-ABL1 isoforms identified by split reads. G, Expression of BCR-ABL1 lacking ABL1 exon 11 in K562 confers resistance to imatinib. All values shown are averages ± SEM of two biological replicates. Relative viability refers to the viability of cells treated with inhibitor relative to the viability of cells without inhibitor. Two hundred ng/mL doxycycline was used.

Close modal
Figure 4.

LentiMutate identifies mutations in WT KRAS or KRAS-G12C that can confer resistance to AMG 510. A, Results of LentiMutate on WT KRAS using 1 μmol/L AMG 510 in the H358 cell line. B, Results of LentiMutate on KRAS-G12C using 1 μmol/L AMG 510 in H358. Only substitutions enriched >3-fold are annotated, and values are the average of two biological replicates. S, synonymous mutation. C–F, The KRAS-G12C:AMG 510 cocrystal structure (PDB 6OIM) was used to model resistance mutations. Cyan, AMG 510; white, KRAS-G12C; red, oxygen atoms; blue, nitrogen atoms. C, V9 is normally buried in the hydrophobic core of the structure. D, The V9D substitution is predicted to expel the side chain out towards the ligand binding pocket, resulting in a steric clash. Red disks indicate significant van der Waals overlap between adjacent atoms. E, Y96 normally participates in a network of hydrogen bonds including waters (black spheres) and hydrophobic contacts with the pyrodopyrimidine scaffold. F, Y96H mutations result in loss of waters and hydrophobic contacts. G, Expression of indicated KRAS point mutants identified by LentiMutate in H358 cells confers resistance to AMG 510 and MRTX849. Values are averages ± SEM of two biological replicates. Relative viability refers to the viability of cells treated with inhibitor relative to the viability of cells without inhibitor. Five hundred ng/mL doxycycline was used.

Figure 4.

LentiMutate identifies mutations in WT KRAS or KRAS-G12C that can confer resistance to AMG 510. A, Results of LentiMutate on WT KRAS using 1 μmol/L AMG 510 in the H358 cell line. B, Results of LentiMutate on KRAS-G12C using 1 μmol/L AMG 510 in H358. Only substitutions enriched >3-fold are annotated, and values are the average of two biological replicates. S, synonymous mutation. C–F, The KRAS-G12C:AMG 510 cocrystal structure (PDB 6OIM) was used to model resistance mutations. Cyan, AMG 510; white, KRAS-G12C; red, oxygen atoms; blue, nitrogen atoms. C, V9 is normally buried in the hydrophobic core of the structure. D, The V9D substitution is predicted to expel the side chain out towards the ligand binding pocket, resulting in a steric clash. Red disks indicate significant van der Waals overlap between adjacent atoms. E, Y96 normally participates in a network of hydrogen bonds including waters (black spheres) and hydrophobic contacts with the pyrodopyrimidine scaffold. F, Y96H mutations result in loss of waters and hydrophobic contacts. G, Expression of indicated KRAS point mutants identified by LentiMutate in H358 cells confers resistance to AMG 510 and MRTX849. Values are averages ± SEM of two biological replicates. Relative viability refers to the viability of cells treated with inhibitor relative to the viability of cells without inhibitor. Five hundred ng/mL doxycycline was used.

Close modal

Crystal violet staining

Cells were fixed and stained with 4% formaldehyde (Fisher Scientific, catalog no. 50–980–487) in PBS containing 0.05% crystal violet (Millipore Sigma, catalog no. C6158).

Data and software availability

Sequencing data was deposited in GEO under accession number GSE164664.

Lentiviral transduction creates mutations in its cargo DNA and this property can be leveraged to identify cancer drug resistance mutations

We previously conducted a PCR-based mutagenesis screen to identify mutations in RUVBL1 that confer resistance to compound B, a RUVBL1/RUVBL2 inhibitor developed as an experimental cancer drug (6). In brief, for that study we performed PCR on RUVBL1 cDNA using the relatively error-prone Taq polymerase, these mutant cDNAs were cloned into a lentiviral doxycycline-inducible expression vector to generate a large library of cDNA variants, transduced as single integrants into non–small cell lung cancer (NSCLC) cells, the cDNA was expressed, and compound B added at a lethal concentration to identify drug-resistant clones, which were then pooled and sequenced to identify resistance-conferring mutations (6). After publishing these results, we repeated this experiment in an additional cell line and included a sequence-validated WT control of the RUVBL1 cDNA, i.e., the cDNA was not mutagenized by PCR. Surprisingly, this WT control produced similar results (Fig. 1A; Supplementary Tables S1 and S2). Independent cloning and expression of these mutations confirmed their ability to promote resistance to the inhibitor (Supplementary Fig. S1).

This unexpected result in the control prompted us to consider whether the lentivirus we used for transduction had created these resistance-conferring mutations. The lentivirus we and others use is derived from HIV-1, which, primarily due to an error-prone RT, has an error rate between 2 × 10–5 to 5 × 10–4 mutations per nucleotide per replication cycle (7–9). This is surprisingly similar to the error rate reported for Taq DNA polymerase, which is between 3 × 10–5 to 2 × 10–4 mutations per nucleotide per doubling (10, 11). Thus, we reasoned that lentiviral transduction alone may sufficiently mutagenize a transgene to perform facile mutagenesis screens. This property of lentiviral transduction is particularly advantageous as mutations and delivery of the cDNA are performed in a single reaction, obviating the need to monitor and maintain variant representation in a vector library through multiple steps such as PCR/cloning, transformation, and plasmid preparation like during PCR- or plasmid-based mutagenesis. Additionally, one can imagine designing RT enzymes better suited for this approach, similar to the development of polymerases with higher fidelity for PCR. We refer to this process/technique, where lentiviral transduction creates mutations in the cDNA of a purported anticancer drug target and resistance mutations are identified by drug treatment and sequencing, as LentiMutate (Fig. 1B).

To determine if LentiMutate would be applicable to other genes/drugs, we performed LentiMutate using mutant EGFR (del746–750) and gefitinib, an EGFR inhibitor, in an EGFR mutated and addicted NSCLC cell line, HCC827 (12). LentiMutate readily produced gefitinib-resistant cells and these cells harbored the well-known and clinically relevant gefitinib/erlotinib resistance mutation T785M (equivalent to T790M in full length-EGFR; ref. 13) (Fig. 1C; Supplementary Table S3). Because the EGFR inhibitor osimertinib is now the first line treatment for NSCLC patients with activating EGFR mutations (14), we performed this same experiment with osimertinib, expecting the well characterized C792S (equivalent to C797S in full-length EGFR) osimertinib resistance mutation (15) to emerge. However, osimertinib-resistant cells did not emerge.

The M230I substitution in the lentiviral reverse transcriptase enzyme enhances the efficiency of LentiMutate

HIV-1 RT has been shown to have mutational bias (16) like all DNA polymerases and are more likely to make transitions than transversions. The C792S osimertinib-resistance mutation is the result of either a T → A or a G → C transversion. We reasoned that we could improve LentiMutate by modifying the lentiviral RT to become more error-prone and/or alter its mutational bias. A M230I substitution in the RT of HIV-1 has been shown to decrease fidelity without significantly impacting enzymatic activity (17). We introduced this mutation into the RT of our lentiviral packaging system and performed sequencing of unique molecular identifier-tagged lentiviral integrants from cells transduced with lentivirus packaged using either this mutated RT (“M-RT”) or the wild-type RT (“WT-RT”). Our results show that, on average, the M-RT increased the nucleotide substitution rate approximately three-fold, and importantly, reduced the M-RT enzyme's overall mutational bias (Fig. 1D), increasing the substitution rate at many nucleotides that had a relatively low substitution frequency with the WT-RT.

We next compared the efficiency of the M-RT with the WT-RT in creating osimertinib-resistance mutations by performing LentiMutate on EGFRdel746–750 with osimertinib in HCC827 using either the WT-RT or M-RT and otherwise identical conditions. Similar to prior experiments, the WT-RT did not produce osimertinib-resistant cells, however, cells transduced using the M-RT produced osimertinib-resistant cells, which harbored the C792S mutation (Fig. 1E; Supplementary Table S4).

LentiMutate identifies large deletions that confer drug resistance

To further validate LentiMutate on a different cancer and drug we chose CML, which harbors the BCR-ABL1 fusion gene/protein and is sensitive to imatinib (18). Mutations in the kinase domain of BCR-ABL1 that promote resistance to imatinib have been thoroughly studied, with the first report published nearly 20 years ago (19). We chose the K562 CML cell line as it harbors the BCR-ABL1 fusion gene/protein and is highly sensitive to BCR-ABL1 inhibitors (20) and performed LentiMutate using BCR-ABL1 (p210) and imatinib. Imatinib resistant cells readily emerged and harbored some of the previously identified imatinib-resistance mutations, such as M472I, E494K, and E255K, although we failed to identify many of the well-known resistance mutations such as T315I (Supplementary Table S5). However, we also identified numerous deletions (SVs) in BCR-ABL1 outside the ABL1 kinase domain that were strongly enriched and dominated the population of these imatinib-resistant cells (Fig. 2A; Supplementary Table S6). It is likely that many of the synonymous mutations highly enriched in the imatinib-resistant cells are passenger mutations with the BCR-ABL1 deletion mutations. LentiMutate may create large and small indels because the HIV-1 RT is known to switch templates and have low processivity (21). In comparison with other techniques used to identify resistance-conferring mutations, the ability of LentiMutate to create indels and recover them efficiently using deep sequencing is a unique advantage. It is interesting to note that for all other genes and inhibitors tested via LentiMutate in this manuscript, indels were only enriched in the BCR-ABL1 and imatinib experiment.

Two of these deletion mutations directly identified by LentiMutate, SV8 and SV11, as well as a 216-nucleotide region in BCR shared by the ten N-terminal in-frame deletion mutations, referred to as the N-Terminus Consensus Deletion (NTCD), were tested for drug resistance in K562 cells. We specifically chose to test SV8 and SV11 as these deletion mutations were highly enriched in the resistant population. As a comparison, we also tested the 35INS variant, a BCR-ABL1 variant detected in patient samples that truncates BCR-ABL1 at the end of the kinase domain due to the insertion and usage of a 35bp sequence in an intron that acts as a novel splicing site, producing a kinase dead protein. 35INS was originally believed to confer resistance to imatinib (22) but subsequently shown not to promote resistance (23). All the deletions directly identified by LentiMutate, as well as the NTCD, but not 35INS, conferred resistance to imatinib, an ATP-site directed ABL1 inhibitor (Fig. 2B). By contrast, while deletions in BCR (-SV8 and -NTCD) conferred resistance to the allosteric ABL1 inhibitor asciminib, deletions in ABL1 (-SV11) did not (Fig. 2B).

BCR-ABL1 is known to oligomerize when signaling. To rule out artifacts that might occur because we expressed these deletion mutants in the presence of WT BCR-ABL1, we sought to eliminate the endogenous BCR-ABL1 so that only the BCR-ABL1 deletion mutants were expressed. To do this, we deleted the endogenous BCR-ABL1 while expressing the exogenous, doxycycline-inducible BCR-ABL1 deletion mutants (Fig. 3A). We identified single-cell clones from this population that lacked WT BCR-ABL1 and lose expression of exogenous BCR-ABL1 (Fig. 3B) and viability (Fig. 3C) after withdrawal of doxycycline. These cells are resistant to imatinib (Fig. 3D), strongly suggesting that these deletion mutants alone can promote resistance to imatinib.

Numerous studies have examined samples of patients with CML for mutations in BCR-ABL1 that may confer resistance to BCR-ABL1 inhibitors; however, these studies tend to sequence only the kinase domain in ABL1. To search for events in patients with CML that may produce proteins similar to the BCR-ABL1 deletion mutants identified by LentiMutate we analyzed both Exome-Seq and RNA-seq data from 120 patients with CML that were either newly diagnosed or had relapsed on BCR-ABL1 inhibitors. Using the Exome-Seq dataset we did not identify any mutations that would create deletions similar to those identified by LentiMutate; however, this could be due to the limitations of indel calling using Exome-Seq. By contrast, in the RNA-seq dataset we identified split reads in 27.5% of patients that demonstrate either the inclusion of cryptic exons in the intron between the canonical exons 10 and 11 (v1, v2, v3) or alternative splicing between exons 10 and 11 (v4) in ABL1 (Fig. 3E). All 4 of these splice variants would be expected to truncate BCR-ABL1 (Fig. 3F) in a manner highly similar to the C-terminal BCR-ABL1 deletions identified by LentiMutate. To determine if these splice variants of BCR-ABL1 that we identified in patient samples could confer resistance to imatinib, we generated a BCR-ABL1 construct (BCR-ABL1-ex11del) that lacks exon 11, effectively mimicking all 4 of these novel BCR-ABL1 isoforms identified in patient samples, which differ only by a few C-terminal amino acids. Expression of BCR-ABL1-ex11del was able to confer resistance to imatinib (Fig. 3G). We note that alternative splicing has been documented to generate a truncated androgen receptor (AR) isoform, known as AR-V7, which is believed to promote clinical resistance to enzalutamide and abiraterone (24), analogous to the BCR-ABL1 isoforms identified in this study. Intriguingly, recent studies have shown that samples of patients with CML may have widespread, aberrant splicing (25), suggesting that altered splicing may be an unexpected source of variation and potentially resistance in CML.

Application of LentiMutate to identify resistance mutations for a promising clinical candidate, AMG 510

To demonstrate the potential ability of LentiMutate to prospectively identify resistance mutations before they occur in patient tumors, we turned to the KRAS-G12C inhibitor AMG 510 (sotorasib), which has recently shown promising activity in KRAS-G12C mutant NSCLC (26). KRAS mutations in NSCLC patients can be heterozygous (27); we therefore performed LentiMutate using either cDNAs for WT KRAS or KRAS-G12C and AMG 510 in the KRAS-G12C mutant and addicted NSCLC cell line H358. LentiMutate in H358 cells selected with AMG 510 using WT KRAS cDNA identified numerous KRAS mutations that have been shown to be oncogenic, such as KRAS-G12D and KRAS-Q61R (Fig. 4A; Supplementary Table S7). By contrast, LentiMutate using KRAS-G12C cDNA primarily identified point mutations that are predicted to alter drug binding, such as various mutations in cysteine 12, as well as mutations in the AMG 510 binding pocket, such as V9D and Y96N/H/D (Fig. 4B; Supplementary Table S8). Codon 9 is structurally conserved as a hydrophobic residue within the RAS superfamily and the V9D mutation alters the electrostatic environment, resulting in clashes between the aspartate, Y96, and AMG 510 (Fig. 4C and D). Y96 participates in hydrophobic interactions and a network of polar contacts that support ligand binding, including interactions between the hydroxyl and backbone of G10 and interactions with a network of waters. The Y96H mutation removes these polar contacts and electron density and is predicted to impair ligand binding (Fig. 4E and F). Expression of these mutants conferred resistance to AMG 510 (Fig. 4G), as well as a different KRAS-G12C inhibitor, MRTX849 (Fig. 4G). During review of our article, two articles were published that identified numerous mutations in samples of patients with cancer and in vitro that may confer resistance to MRTX849 (28, 29). Many of these mutations were also nominated by LentiMutate as AMG 510 resistance-conferring mutations, such as mutations at codons V9, G12, G13, Y96, Q61, and Y96. These results highlight the ability of LentiMutate to nominate clinically relevant resistance-conferring mutations prior to their identification in patient samples.

Because lentiviruses can mutate their cargo DNA during integration, we demonstrate that by transducing cDNAs of anticancer drug targets into relevant drug-sensitive cancer cell lines, treating them with the anticancer drug, and sequencing the cDNA in surviving cells, drug resistance mutations can be readily identified; a technique we refer to as LentiMutate. This system offers many advantages: it uses biologically and clinically relevant cancer cell line models, is facile, requires minimal reagents, and unlike other techniques, can identify indels. The enzymatic origin and spectrum of mutations that can be identified by LentiMutate provides distinct advantages over other techniques, such as MITE-seq, which can be expensive and typically uses a small segment of the protein, and use of mismatch repair-deficient cell lines, which tends to identify only a few point mutations. These advantages of LentiMutate are exemplified by our findings with full length BCR-ABL1 and imatinib where we unexpectedly find deletions outside BCR-ABL1's kinase that confer resistance to imatinib. Despite the ability of LentiMutate to use larger segments of a protein in comparison with other techniques such as MITE-seq, our inability to detect the full spectrum of clinically known point mutations in BCR-ABL1 that confer resistance to imatinib highlights that the current version of LentiMutate may underperform on extremely large proteins, such as BCR-ABL1, which has an open reading frame (ORF) of over 6 kb. It is possible that while LentiMutate may have generated the full spectrum of clinically known imatinib-resistance mutations in BCR-ABL1 during reverse transcription, the cocreation of deleterious point mutations and indels, or the differential impact of various resistance-conferring mutations on cell fitness may have obscured the detection of these resistance-conferring point mutations in BCR-ABL1. While LentiMutate may have underperformed in this regard on BCR-ABL1, we note that for EGFR, which has an ORF of 3.6 kb, LentiMutate was able to identify the full spectrum of clinically relevant resistance mutations. Given that the EGFR ORF is larger than 90% of all human ORFs, we suspect that LentiMutate will perform well on the vast majority of human ORFs.

Using AMG 510, we also show that LentiMutate can provide potentially valuable preclinical insights. We identified mutations outside of cysteine 12 in KRAS-G12C that conferred resistance to AMG 510 and these mutations had not been found in prior preclinical studies of AMG 510 resistance (30–33). In addition, we show that mutations in WT KRAS or KRAS-G12C may promote resistance; thus, different resistance mutations may emerge in patients depending on the zygosity of mutant KRAS in individual patient tumors. While the mutations identified here in BCR-ABL1 and KRAS provide immediate insight into the structure-activity relationships between these targets and their inhibitors, whether these mutations are clinically relevant awaits further validation.

Because lentiviral transduction is already routinely performed in many laboratories, adoption and use of LentiMutate is relatively straightforward. We envision that LentiMutate could be used early during cancer-drug discovery to test specificity, inform the synthesis of second-generation inhibitors, and understand interactions between the drug and protein. It is noteworthy that LentiMutate could be used for any type of RNA/DNA/protein evolutionary process with a positive selection step, not just the identification of drug resistance mutations. Thus, LentiMutate may find use beyond informing anticancer drug development programs.

P. Yenerall reports a patent for "Methods and Compositions for Mutagenesis Screening in Mammalian Cells" application # 63158723 pending. C.A. Eide reports grants from NIH/NCI during the conduct of the study. S.K. McWeeney reports grants from NIH during the conduct of the study. K.D. Westover reports grants from NIH, US Department of Defense; grants from CPRIT during the conduct of the study; grants from Revolution Medicines outside the submitted work; and is on the scientific advisory board of Vibliome Therapeutics. B.J. Druker reports grants from NIH NCI R01 during the conduct of the study; personal fees from Aileron Therapeutics, Therapy Architects, LLC (Parent company of ALLCRON Pharma Inc.), Amgen, Aptose Biosciences, Iterion Therapeutics, Blueprint Medicines, Burroughs Wellcome Fund, Cepheid, GRAIL, Nemucore Medical Innovations, Recludix Pharma, VB Therapeutics, Vincerx Pharma, Vivid Biosciences, Gilead Sciences, ICON (formerly Molecular MD), Monojul personal fees and other support from EnlIven Therapeutics, Novartis; other support from Beat AML LLC, CureOne, Bristol-Myers Squibb; and other support from Pfizer outside the submitted work; in addition, B.J. Druker has a patent for Treatment of Gastrointestinal Stromal Tumors issued, licensed, and with royalties paid from Novartis. J.D. Minna reports grants from National Institutes of Health, grants from Cancer Prevention and Research Institute of Texas, and grants from Margot Johnson Foundation during the conduct of the study; personal fees from National Institutes of Health and personal fees from University of Texas Southwestern Medical Center outside the submitted work; in addition, J.D. Minna has a patent for LentiMutate pending. R. Kittler reports grants from NIH-NCI and grants from CPRIT during the conduct of the study; in addition, R. Kittler has a patent for Provisional US Patent application pending. No disclosures were reported by the other authors.

P. Yenerall: Conceptualization, investigation, visualization, methodology, writing–original draft, writing–review and editing. R.K. Kollipara: Investigation, visualization, methodology. K. Avila: Investigation. M. Peyton: Investigation. C.A. Eide: Investigation, methodology. D. Bottomly: Investigation. S.K. McWeeney: Investigation. Y. Liu: Investigation, methodology. K.D. Westover: Funding acquisition, investigation, methodology. B.J. Druker: Funding acquisition, investigation, methodology. J.D. Minna: Conceptualization, supervision, funding acquisition, writing–review and editing. R. Kittler: Conceptualization, supervision, funding acquisition, investigation, methodology, writing–review and editing.

The authors thank the UT Southwestern (UTSW) McDermott Sequencing Core for their technical assistance and Chelsea Burroughs for generating graphics. This study was supported by funding from the Simmons Comprehensive Cancer Center at UTSW (developmental funds to R. Kittler from P30CA142543), CPRIT (RP120732-P3 to R. Kittler, RP160652 to J.D. Minna, RP170373 to K.D. Westover), the NIH (NCI SPORE in lung cancer 5P50CA070907 to J.D. Minna, R01CA200787 to R. Kittler, R01CA244341 to K.D. Westover, and R01CA065823 to B.J. Druker), the Margot Johnson Foundation to J.D. Minna, and the Howard Hughes Medical Institute to B.J. Druker. R. Kittler is a John L. Roach Scholar in Biomedical Research and a CPRIT Scholar in Cancer Research.

Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).

1.
Cadwell
RC
,
Joyce
GF
.
Randomization of genes by PCR mutagenesis
.
PCR Methods Appl
1992
;
2
:
28
33
.
2.
Azam
M
,
Latek
RR
,
Daley
GQ
.
Mechanisms of autoinhibition and STI-571/imatinib resistance revealed by mutagenesis of BCR-ABL
.
Cell
2003
;
112
:
831
43
.
3.
Bradeen
HA
,
Eide
CA
,
O'Hare
T
,
Johnson
KJ
,
Willis
SG
,
Lee
FY
, et al
.
Comparison of imatinib mesylate, dasatinib (BMS-354825), and nilotinib (AMN107) in an N-ethyl-N-nitrosourea (ENU)-based mutagenesis screen: high efficacy of drug combinations
.
Blood
2006
;
108
:
2332
8
.
4.
Wacker
SA
,
Houghtaling
BR
,
Elemento
O
,
Kapoor
TM
.
Using transcriptome sequencing to identify mechanisms of drug action and resistance
.
Nat Chem Biol
2012
;
8
:
235
7
.
5.
Melnikov
A
,
Rogov
P
,
Wang
L
,
Gnirke
A
,
Mikkelsen
TS
.
Comprehensive mutational scanning of a kinase in vivo reveals substrate-dependent fitness landscapes
.
Nucleic Acids Res
2014
;
42
:
e112
.
6.
Yenerall
P
,
Das
AK
,
Wang
S
,
Kollipara
RK
,
Li
LS
,
Villalobos
P
, et al
.
RUVBL1/RUVBL2 ATPase activity drives PAQosome maturation, DNA replication and radioresistance in lung cancer
.
Cell Chem Biol
2020
;
27
:
105
21
.
7.
Hu
WS
,
Hughes
SH
.
HIV-1 reverse transcription
.
Cold Spring Harb Perspect Med
2012
;
2
:
a006882
.
8.
Abram
ME
,
Ferris
AL
,
Das
K
,
Quinones
O
,
Shao
W
,
Tuske
S
, et al
.
Mutations in HIV-1 reverse transcriptase affect the errors made in a single cycle of viral replication
.
J Virol
2014
;
88
:
7589
601
.
9.
Roberts
JD
,
Bebenek
K
,
Kunkel
TA
.
The accuracy of reverse transcriptase from HIV-1
.
Science
1988
;
242
:
1171
3
.
10.
Potapov
V
,
Ong
JL
.
Examining sources of error in PCR by single-molecule sequencing
.
PLoS One
2017
;
12
:
e0169774
.
11.
McInerney
P
,
Adams
P
,
Hadi
MZ
.
Error rate comparison during polymerase chain reaction by DNA polymerase
.
Mol Biol Int
2014
;
2014
:
287430
.
12.
Amann
J
,
Kalyankrishna
S
,
Massion
PP
,
Ohm
JE
,
Girard
L
,
Shigematsu
H
, et al
.
Aberrant epidermal growth factor receptor signaling and enhanced sensitivity to EGFR inhibitors in lung cancer
.
Cancer Res
2005
;
65
:
226
35
.
13.
Thomas
RK
,
Greulich
H
,
Yuza
Y
,
Lee
JC
,
Tengs
T
,
Feng
W
, et al
.
Detection of oncogenic mutations in the EGFR gene in lung adenocarcinoma with differential sensitivity to EGFR tyrosine kinase inhibitors
.
Cold Spring Harb Symp Quant Biol
2005
;
70
:
73
81
.
14.
Soria
JC
,
Ohe
Y
,
Vansteenkiste
J
,
Reungwetwattana
T
,
Chewaskulyong
B
,
Lee
KH
, et al
.
Osimertinib in untreated EGFR-mutated advanced non-small-cell lung cancer
.
N Engl J Med
2018
;
378
:
113
25
.
15.
Thress
KS
,
Paweletz
CP
,
Felip
E
,
Cho
BC
,
Stetson
D
,
Dougherty
B
, et al
.
Acquired EGFR C797S mutation mediates resistance to AZD9291 in non-small cell lung cancer harboring EGFR T790M
.
Nat Med
2015
;
21
:
560
2
.
16.
McCallum
M
,
Oliveira
M
,
Ibanescu
RI
,
Kramer
VG
,
Moisi
D
,
Asahchop
EL
, et al
.
Basis for early and preferential selection of the E138K mutation in HIV-1 reverse transcriptase
.
Antimicrob Agents Chemother
2013
;
57
:
4681
8
.
17.
Svarovskaia
ES
,
Cheslock
SR
,
Zhang
WH
,
Hu
WS
,
Pathak
VK
.
Retroviral mutation rates and reverse transcriptase fidelity
.
Front Biosci
2003
;
8
:
d117
34
.
18.
Druker
BJ
,
Tamura
S
,
Buchdunger
E
,
Ohno
S
,
Segal
GM
,
Fanning
S
, et al
.
Effects of a selective inhibitor of the Abl tyrosine kinase on the growth of Bcr-Abl positive cells
.
Nat Med
1996
;
2
:
561
6
.
19.
Gorre
ME
,
Mohammed
M
,
Ellwood
K
,
Hsu
N
,
Paquette
R
,
Rao
PN
, et al
.
Clinical resistance to STI-571 cancer therapy caused by BCR-ABL gene mutation or amplification
.
Science
2001
;
293
:
876
80
.
20.
Thiesing
JT
,
Ohno-Jones
S
,
Kolibaba
KS
,
Druker
BJ
.
Efficacy of STI571, an abl tyrosine kinase inhibitor, in conjunction with other antileukemic agents against bcr-abl-positive cells
.
Blood
2000
;
96
:
3195
9
.
21.
Huber
HE
,
McCoy
JM
,
Seehra
JS
,
Richardson
CC
.
Human immunodeficiency virus 1 reverse transcriptase. Template binding, processivity, strand displacement synthesis, and template switching
.
J Biol Chem
1989
;
264
:
4669
78
.
22.
Laudadio
J
,
Deininger
MW
,
Mauro
MJ
,
Druker
BJ
,
Press
RD
.
An intron-derived insertion/truncation mutation in the BCR-ABL kinase domain in chronic myeloid leukemia patients undergoing kinase inhibitor therapy
.
J Mol Diagn
2008
;
10
:
177
80
.
23.
O'Hare
T
,
Zabriskie
MS
,
Eide
CA
,
Agarwal
A
,
Adrian
LT
,
You
H
, et al
.
The BCR-ABL35INS insertion/truncation mutant is kinase-inactive and does not contribute to tyrosine kinase inhibitor resistance in chronic myeloid leukemia
.
Blood
2011
;
118
:
5250
4
.
24.
Antonarakis
ES
,
Lu
C
,
Wang
H
,
Luber
B
,
Nakazawa
M
,
Roeser
JC
, et al
.
AR-V7 and resistance to enzalutamide and abiraterone in prostate cancer
.
N Engl J Med
2014
;
371
:
1028
38
.
25.
Schmitz
U
,
Shah
JS
,
Dhungel
BP
,
Monteuuis
G
,
Luu
PL
,
Petrova
V
, et al
.
Widespread aberrant alternative splicing despite molecular remission in chronic myeloid leukaemia patients
.
Cancers
2020
;
12
:
3738
.
26.
Hong
DS
,
Fakih
MG
,
Strickler
JH
,
Desai
J
,
Durm
GA
,
Shapiro
GI
, et al
.
KRAS(G12C) inhibition with sotorasib in advanced solid tumors
.
N Engl J Med
2020
;
383
:
1207
17
.
27.
Uchiyama
M
,
Usami
N
,
Kondo
M
,
Mori
S
,
Ito
M
,
Ito
G
, et al
.
Loss of heterozygosity of chromosome 12p does not correlate with KRAS mutation in non-small cell lung cancer
.
Int J Cancer
2003
;
107
:
962
9
.
28.
Awad
MM
,
Liu
S
,
Rybkin
II
,
Arbour
KC
,
Dilly
J
,
Zhu
VW
, et al
.
Acquired resistance to KRAS(G12C) inhibition in cancer
.
N Engl J Med
2021
;
384
:
2382
93
.
29.
Tanaka
N
,
Lin
JJ
,
Li
C
,
Ryan
MB
,
Zhang
J
,
Kiedrowski
LA
, et al
.
Clinical acquired resistance to KRASG12C inhibition through a novel KRAS switch-II pocket mutation and polyclonal alterations converging on RAS-MAPK reactivation
.
Cancer Discov
2021 Apr 6 [Epub ahead of print]
.
30.
Canon
J
,
Rex
K
,
Saiki
AY
,
Mohr
C
,
Cooke
K
,
Bagal
D
, et al
.
The clinical KRAS(G12C) inhibitor AMG 510 drives anti-tumour immunity
.
Nature
2019
;
575
:
217
23
.
31.
Ryan
MB
,
Fece de la Cruz
F
,
Phat
S
,
Myers
DT
,
Wong
E
,
Shahzade
HA
, et al
.
Vertical pathway inhibition overcomes adaptive feedback resistance to KRAS(G12C) inhibition
.
Clin Cancer Res
2020
;
26
:
1633
43
.
32.
Xue
JY
,
Zhao
Y
,
Aronowitz
J
,
Mai
TT
,
Vides
A
,
Qeriqi
B
, et al
.
Rapid non-uniform adaptation to conformation-specific KRAS(G12C) inhibition
.
Nature
2020
;
577
:
421
5
.
33.
Adachi
Y
,
Ito
K
,
Hayashi
Y
,
Kimura
R
,
Tan
TZ
,
Yamaguchi
R
, et al
.
Epithelial-to-mesenchymal transition is a cause of both intrinsic and acquired resistance to KRAS G12C inhibitor in KRAS G12C-mutant non-small cell lung cancer
.
Clin Cancer Res
2020
;
26
:
5962
73
.
This open access article is distributed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0) license.

Supplementary data