Multiplexed in vivo homology-directed repair and tumor barcoding enables parallel quantification of Kras variant oncogenicity

Abstract

Large-scale genomic analyses of human cancers have cataloged somatic point mutations thought to initiate tumor development and sustain cancer growth. However, determining the functional significance of specific alterations remains a major bottleneck in our understanding of the genetic determinants of cancer. Here, we present a platform that integrates multiplexed AAV/Cas9-mediated homology-directed repair (HDR) with DNA barcoding and high-throughput sequencing to simultaneously investigate multiple genomic alterations in de novo cancers in mice. Using this approach, we introduce a barcoded library of non-synonymous mutations into hotspot codons 12 and 13 of Kras in adult somatic cells to initiate tumors in the lung, pancreas, and muscle. High-throughput sequencing of barcoded Kras ^HDR alleles from bulk lung and pancreas reveals surprising diversity in Kras variant oncogenicity. Rapid, cost-effective, and quantitative approaches to simultaneously investigate the function of precise genomic alterations in vivo will help uncover novel biological and clinically actionable insights into carcinogenesis.

Fig. 1

A cancer modeling platform that integrates AAV/Cas9-mediated somatic HDR with tumor barcoding and sequencing to enable the rapid introduction and functional investigation of putative oncogenic point mutations in vivo. a–d Schematic of the pipeline to quantitatively measure the in vivo oncogenicity of a panel of defined point mutations. a A library of AAV vectors is generated such that each AAV contains: (1) a template for homology-directed repair (HDR) containing a putatively oncogenic point mutation and a random DNA barcode encoded in the adjacent wobble bases; (2) an sgRNA targeting the desired endogenous locus to enhance HDR; and (3) Cre-recombinase to activate a conditional Cas9 allele (H11 ^LSL-Cas9) and other Cre-dependent alleles in genetically engineered mice. b The AAV library is delivered to a tissue of interest. c Following transduction, a subset of cells undergo AAV/Cas9-mediated HDR in which the locus of interest is cleaved by Cas9 at the sgRNA target site and repaired using the AAV HDR template. This results in the introduction of the desired point mutation and a unique DNA barcode at the targeted locus. d Somatic cells engineered with a point mutation may develop into de novo tumors if the introduced mutation is sufficient to initiate tumorigenesis and drive tumor growth. Two independent approaches can be used to analyze tumors: (1) the targeted region in individual tumors can be sequenced to characterize both alleles of the targeted gene, or (2) next-generation sequencing of the targeted region can be used to determine the number, size, and genotype of each tumor directly from bulk tissue in a quantitative and multiplexed manner

Fig. 2

Design and validation of an AAV targeting vector library to introduce all Kras codon 12 and 13 single-nucleotide non-synonymous point mutations into somatic mouse cells in a multiplexed manner. a AAV vector pool for Cas9-mediated HDR into the endogenous Kras locus (AAV-Kras ^HDR/sgKras/Cre). Each vector contained an HDR template with 1 of 12 non-synonymous Kras mutations at codons 12 and 13 (or wild-type Kras), silent mutations within the PAM (boxed sequence) and sgRNA homology region (PAM*), and a random 8-nucleotide barcode within the wobble positions of the adjacent codons for stable DNA barcoding of individual tumors. b Representation of each Kras ^HDR allele in the AAV-Kras ^HDR/sgKras/Cre plasmid library. c Diversity of the barcode region in the AAV-Kras ^HDR/sgKras/Cre plasmid library. d Schematic of the experiment to test for HDR bias. A Cas9-expressing cell line was transduced with AAV-Kras ^HDR/sgKras/Cre and Kras ^HDR alleles were sequenced to quantify HDR events. e Schematic of the PCR strategy to specifically amplify Kras ^HDR alleles introduced into the genome via HDR. Forward primer 1 (F1) binds to the sequence containing the 3 PAM* mutations, while reverse primer 1 (R1) binds the endogenous Kras locus, outside the sequence present in the homology arm of the Kras ^HDR template. F2 binds to the Illumina adaptor added by F1, R2 binds to a region near exon 2, and R3 binds to the Illumina adapter added in the same reaction by R2. f Frequency of HDR events for each Kras ^HDR allele plotted against the initial frequency of each Kras mutant allele in the AAV-Kras ^HDR/sgKras/Cre plasmid library. High correlation between the initial plasmid library and the representation of mutant Kras alleles following HDR suggests little to no HDR bias

Fig. 3

AAV/Cas9-mediated somatic HDR initiates oncogenic Kras-driven lung tumors. a Schematic of the experiment to introduce point mutations and a DNA barcode into the endogenous Kras locus of lung epithelial cells in Rosa26 ^LSL-Tomato ;H11 ^LSL-Cas9 (T;H11 ^LSL-Cas9), p53 ^flox/flox ;T;H11 ^LSL-Cas9 (PT;H11 ^LSL-Cas9), and Lkb1 ^flox/flox ;T;H11 ^LSL-Cas9 (LT;H11 ^LSL-Cas9) mice by intratracheal administration of AAV-Kras ^HDR/sgKras/Cre (8.4 × 10¹⁰ vector genomes per mouse). b Representative fluorescence and histology images of Tomato^positive lung tumors in LT;H11 ^LSL-Cas9, PT;H11 ^LSL-Cas9, and T;H11 ^LSL-Cas9 mice transduced with AAV-Kras ^HDR/sgKras/Cre. Scale bars = 5 mm. c Quantification of lung tumors in the indicated genotypes of mice transduced with the indicated AAV vectors (with and without sgKras). Each dot represents one mouse. d Representative FACS plot showing Tomato^positive DTCs in the pleural cavity of an LT;H11 ^LSL-Cas9 mouse with AAV-Kras ^HDR/sgKras/Cre-initiated lung tumors. Plot shows forward scatter/side scatter (SSC)-gated viable cancer cells (DAPI/CD45/CD31/F4-80/Ter119^negative). e Histology of lymphatic metastases from AAV-Kras ^HDR/sgKras/Cre-initiated lung tumors in PT;H11 ^LSL-Cas9 mice. Scale bar = 50 µm. f Number of AAV-Kras ^HDR/sgKras/Cre-transduced mice of each genotype that had disseminated tumors cells in their pleural cavity (DTCs > 10) or metastases out of the total number of mice analyzed. g Kras ^LSL-G12D/+ ;LT (KLT) and Kras ^LSL-G12D/+ ;PT (KPT) mice transduced with a 1:10,000 dilution of AAV-Kras ^HDR/sgKras/Cre developed approximately half as many tumors as the PT;H11 ^LSL-Cas9 and LT;H11 ^LSL-Cas9 mice transduced with undiluted virus. If all Kras ^HDR alleles in the AAV-Kras ^HDR/sgKras/Cre library are oncogenic, this suggests that AAV/Cas9-mediated HDR occurs in ~0.02% of transduced cells. Alternatively, if only 20% of the mutant alleles in the AAV-Kras ^HDR/sgKras/Cre library are oncogenic, this suggests that HDR occurs in ~0.1% of transduced cells. h Diverse HDR-generated oncogenic Kras alleles in individual lung tumors dissected from LT;H11 ^LSL-Cas9 and PT;H11 ^LSL-Cas9 mice transduced with AAV-Kras ^HDR/sgKras/Cre. Number of tumors with each allele is indicated. Alleles that were not identified in any lung tumors are not shown. For the Kras ^HDR alleles identified, Bonferroni corrected p values for likelihood of enrichment relative to WT are shown (Fisher’s exact test generalized for structural zeros; see Methods section)

Fig. 4

Introduction of mutant Kras variants into pancreatic cells using AAV/Cas9-mediated HDR drives the formation of metastatic PDAC. a Schematic of retrograde pancreatic ductal injection of AAV-Kras ^HDR/sgKras/Cre (∼1.7 × 10¹¹ vector genomes per mouse) into PT;H11 ^LSL-Cas9 mice to induce pancreatic cancer. b Histology of pancreatic tumors initiated by retrograde pancreatic ductal injection of AAV-Kras ^HDR/sgKras/Cre into PT;H11 ^LSL-Cas9 mice. Scale bars = 75 µm. c Histology of metastases in the lymph node (upper panel) and diaphragm (lower panel) in PT;H11 ^LSL-Cas9 mice with PDAC. Scale bars = 50 µm. d Incidence of PDAC, DTCs in the peritoneal cavity, and metastases in the indicated genotypes of mice (shown as the number of mice with cancer, DTCs, or metastases out of the total number of mice analyzed), 3–13 months after transduction with either AAV-Kras ^HDR/sgKras/Cre or AAV-Kras ^HDR/Cre (∼2.9 × 10¹¹ vector genomes per mouse). AAV-Kras ^HDR/sgKras/Cre was administered at the stock concentration (“undil.”) or at a 1:10 dilution (“1:10”). e HDR-generated oncogenic Kras alleles in individually dissected pancreatic tumor masses. Number of tumors with each allele is indicated. Alleles that were not identified in any pancreatic tumor masses are not shown

Fig. 5

Introduction of mutant Kras variants into muscle cells using AAV/Cas9-mediated HDR induces invasive sarcoma. a Schematic of intramuscular administration of AAV-Kras ^HDR/sgKras/Cre (1.6 × 10¹¹ vector genomes per mouse) into the gastrocnemii of PT;H11 ^LSL-Cas9 mice to induce sarcomas. b, c Histology of stereotypical sarcoma (b) and invasive sarcoma (c) initiated by intramuscular administration of AAV-Kras ^HDR/sgKras/Cre into the gastrocnemii of PT;H11 ^LSL-Cas9 mice. Scale bars = 75 µm. d Sarcoma incidence in PT;H11 ^LSL-Cas9 mice 3–7 months after intramuscular administration of AAV-Kras ^HDR/sgKras/Cre. Incidence represents the number of mice that developed sarcomas out of the total number of mice injected. e HDR-generated oncogenic Kras alleles in sarcomas. Number of tumors with each allele is indicated. Alleles not identified in any sarcomas are not shown

Fig. 6

Multiplexed, quantitative analysis of Kras mutant oncogenicity using AAV/Cas9-mediated somatic HDR and high-throughput sequencing of barcoded lung tumors. a Pipeline to quantitatively determine the number, size, and genotype of individual tumors directly from bulk lung samples by high-throughput sequencing of tumor barcodes. b–e Number of lung tumors harboring each mutant Kras allele normalized to its initial representation (mutant representation in the AAV plasmid library divided by WT representation in the AAV plasmid library) and relative to WT (mutant tumor # divided by WT tumor #). Variants present in significantly more tumors than WT (two-sided Fisher’s exact test; p < 0.05) are colored blue; darker blue indicates no significant difference from G12D (p > 0.05), lighter blue indicates significantly less tumors with the indicated variant than G12D (p < 0.05). Error bars are bootstrapped 95% confidence intervals. Bar plots were generated from pooled data from all mouse genotypes (N = 15) (b), or individually from LT;H11 ^LSL-Cas9 (N = 6) (c), PT;H11 ^LSL-Cas9 (N = 6) (d), or T;H11 ^LSL-Cas9 (N = 3) (e) mice. Note that different y axis scales are used in each plot

Fig. 7

High-throughput sequencing of pancreatic tumor masses enables spatial mapping of tumor clones and phylogenetic tracking of metastases. a Analysis pipeline to identify Kras ^HDR alleles in AAV-Kras ^HDR/sgKras/Cre-initiated tumor masses within the pancreata of PT;H11 ^LSL-Cas9 mice. b Diverse HDR-generated Kras alleles identified by tumor barcode sequencing of pancreatic tumor masses from three PT;H11 ^LSL-Cas9 mice. Numbers of uniquely barcoded primary tumors with each allele (including those identified by individual tumor analyses, as shown in Fig. 4e) are indicated. Alleles not identified in any pancreas tumor masses are not shown. For the Kras ^HDR alleles identified, Bonferroni-corrected p values for likelihood of enrichment relative to WT are shown (Fisher’s exact test generalized for structural zeros; see Methods section) (p values <0.05 are bold). c Multi-region sequencing of a large pancreatic tumor mass in a PT;H11 ^LSL-Cas9 mouse transduced with AAV-Kras ^HDR/sgKras/Cre revealed a diverse spectrum of mutant Kras alleles and uncovered relationships between primary tumors and their metastatic descendants. Each dot represents a tumor with the indicated Kras variant and a unique barcode within each sample (labeled 1–4). Dots connected across different primary tumor samples (labeled 1–3) shared the same Kras variant-barcode pair, and are thus presumably regions of the same primary tumor that were present in multiple samples. A colored line links primary tumors and lymph node metastases harboring the same Kras variant-barcode pair, indicating a clonal relationship. The size of each dot is scaled according to the size of the tumor that it represents (diameter of the dot = relative size^1/2). Since the size of pancreatic tumors was not normalized to a control, tumor sizes can only be compared to other tumors within the same sample. Thus, the largest tumors within each sample have been scaled to the same standard size. g gallbladder, sto stomach, duo duodenum, pan pancreas, sp spleen, ln mesenteric lymph nodes