We generated a Cre-dependent Cas9 mouse by inserting a Cas9 transgene expression cassette into the Rosa26 locus (Figure 1A). The transgene consists of a 33 FLAG-tagged Streptococcus pyogenes Cas9 linked via a self-cleaving P2A peptide to an enhanced green fluorescent protein (EGFP) to facilitate visualization of Cas9-expressing cells. The transgene is driven by the ubiquitous CAG promoter and is interrupted by a loxP-stop (33 polyA signal)-loxP (LSL) cassette to render Cas9 expression inducible by the Cre recombinase (Figure 1A).

To evaluate the effect of prolonged Cas9 expression, we generated a constitutive Cas9-expressing mouse line by crossing the Cre-dependent Cas9 mouse to a β-actin Cre driver (Lewandoski et al., 1997). Resulting progenies of this cross were viable, and Cas9-P2A-EGFP expression was observed throughout the body (Figure 1B). The constitutive Cas9-expressing mice were fertile, had normal litter sizes, presented no morphological abnormalities, and were able to breed to homozygosity. At the cellular level, we also found no morphological abnormalities or upregulation in DNA damage and apoptosis markers (Figure S1 available online).

To further gauge whether constitutive Cas9 expression had adverse effects in cellular physiology, we used a panel of electrophysiological measurements to evaluate the health of Cas9-expressing neurons, a cell type particularly sensitive to perturbations. Therefore, we performed whole-cell patch clamp recordings in CA1 pyramidal neurons from acute hippo-campal slices to examine firing threshold (Figures 1C and 1D) and membrane properties (membrane excitability, input resistance, membrane capacitance, resting potential; Figures 1E– 1H and Table S1) and found no significant differences between wild-type and Cas9-expressing neurons.

With the conditional Cas9 mouse, tissue- and cell-type-specific promoters (Lewandoski, 2001) can facilitate defined spatio-temporal expression of Cas9. To demonstrate this, we crossed the Cre-dependent Cas9 mouse with two Cre driver strains, namely the tyrosine hydroxylase (TH-IRES-Cre) driver for dopaminergic neurons and the parvalbumin (PV-Cre) driver for a subtype of inhibitory interneurons (Hippenmeyer et al., 2005; Lindeberg et al., 2004). As predicted, Cas9 expression was restricted to TH- or PV-positive cells in the F1 progenies of these two crosses (Figure 1I–1J).

To determine whether the Rosa26 knockin construct provided functional levels of Cas9 expression, we set out to test whether a previously described U6-sgRNA lentiviral vector (Sanjana et al., 2014) could mediate indel formation ex vivo in primary immune cells. Several types of immune cells, such as innate immune dendritic cells (DCs), are often not accessible for genetic manipulation due to delivery challenges, short viability terms in culture, or both. Moreover, because existing cell lines do not mimic DC biology well, many studies are performed with primary cells derived ex vivo from precursors isolated from the bone marrow (BMDCs) (Figure 2A), which retain many critical characteristics of DCs in vivo (Amit et al., 2009; Chevrier et al., 2011; Garber et al., 2012; Shalek et al., 2013). We thus reasoned that Cas9-expressing cells derived from the constitutive Cas9-expressing mice may facilitate such applications, as genome editing would only require introduction of sgRNAs, which can be efficiently delivered using lentiviral vectors.

We first verified the expression of Cas9 in bone marrow from constitutive Cas9-expressing mice (Figure 2B). Similarly, we validated Cas9 expression in many other immune cell types (Figure S2). Two days after culturing bone marrow cells from the constitutive Cas9-expressing mice, we infected BMDCs with lentivirus encoding two different sgRNAs targeting early exons of either MyD88 (Figure 2C) or A20 (Figure 2D), two well-characterized positive and negative regulators of Toll-like receptor 4 (TLR4) signaling, respectively. At 7 days posttransduction, we activated cells with lipopolysaccharide (LPS) and performed functional analysis (Figure 2A). We found indels in 67%–78% of sequencing reads (Figures 2E and 2F), leading to reduction in mRNA (Figure 2G), and protein (Figure 2H).

DCs specialize in pathogen detection and initiation of appropriate immune responses (Mellman and Steinman, 2001). Therefore, we measured the expression of 276 representative genes of the LPS response, using the Nanostring nCounter, in cells targeted for Myd88 or A20 as compared to controls (Figure 2I). As predicted, depletion of MyD88 resulted in a reduction of inflammatory response genes, whereas depletion of A20 resulted in an increase of inflammatory response genes. These effects were comparable to those observed with shRNA-mediated knockdown in independent experiments (Figure 2I). Taken together, our results demonstrate the potential of constitutive Cas9-expressing mice for efficient perturbation in primary cells.

To demonstrate direct genome editing in vivo in the Cre-dependent Cas9 mouse, we applied AAV-mediated expression of Cre and sgRNA in the brain (Figure 3A). We constructed an AAV-U6-sgRNA-Cre vector (Figure 3B) containing an sgRNA targeting a highly expressed neuronal-specific RNA-splicing factor, NeuN (RbFox3) (sgNeuN) (Figure 3C). A chimeric AAV1/2 vector was packaged using an equal mixture of AAV1 and two packaging plasmids, which was previously shown to efficiently infect neurons (Konermann et al., 2013). The packaged virus was delivered via stereotactic injection into the prefrontal cortex of Cre-dependent Cas9 mice (Figure 3A). Three weeks postinjection, we dissected the injected brain region. Deep sequencing of the NeuN locus showed indel formation near the predicted cleavage site (Figure 3C), suggesting that Cas9 was functional and facilitated on-target indel formation. Furthermore, we observed NeuN protein depletion only in the injected region (Figures 3D– 3F), but not in controls (noninjected and LacZ-targeted sgRNA, termed sgLacZ) (Figures 3D–3F). To quantify this effect, we performed immunoblot analysis on tissue samples from three mice and found that NeuN protein expression was reduced by 80% (Figure 3G).

To study the mutagenic effect of Cas9-mediated genome editing at the single-cell level, we set out to isolate individual neuronal nuclei according to a previously established method (Okada et al., 2011). To enable this, we generated a single vector (AAV-sgRNA-hSyn-Cre-P2A-EGFP-KASH) that expresses sgRNA, Cre, and EGFP fused to the KASH (Klarsicht, ANC-1, Syne Homology) nuclear transmembrane domain in neurons (L.S. and M. Heidenreich, unpublished data). After harvesting injected brain tissues, we used fluorescence-activated cell sorting (FACS) to isolate single EGFP-positive neuronal nuclei. Targeted amplicon sequencing of the target site from populations of EGFP-positive nuclei revealed an average indel rate of 85% (Figure 3H). Sequencing of 167 single nuclei showed that 84% (n = 141) of transduced cells had biallelic modifications, leaving 9% (n = 15) with monoallelic modification and 7% (n = 11) unmodified (Figure 3I). The 141 biallelic mutant nuclei contained both frame-shift (phase of 3n+1 or 3n+2) and non-frame-shift mutations (phase of 3n), and the 15 monoallelic mutant nuclei all contained frame-shift mutations in the modified alleles (Figure 3J).

The multiplexability of Cas9-mediated genome editing enables modeling of multigenic diseases processes such as tumorigenesis. The genomes of cancer cells possess complex combinations of genetic lesions (Weinberg, 2007), which usually include gain-of-function mutations in proto-oncogenes as well as loss-of-function mutations in tumor suppressor genes (Garraway and Lander, 2013; Lawrence et al., 2013). It is challenging to elucidate the roles of these complex combinations of genetic alterations in tumor evolution largely due to the difficulty of creating transgenic animals with multiple perturbed alleles. We therefore set out to model multi-lesion lung cancer in the Cre-dependent Cas9 mouse.

We chose to model lung cancer mutations involving the onco-gene KRAS and the tumor suppressor genes p53 and LKB1 (Stk11), which are the top three most frequently mutated genes in lung adenocarcinoma in human (46% for p53, 33% for KRAS and 17% for LKB1) (TCGA-Network, 2014). To model the dynamics of mutations in these three genes, we built a single vector capable of generating KRAS^G12D mutations while simultaneously knocking out p53 and LKB1 in the Cre-dependent Cas9 mouse.

Multiple sgRNAs were designed targeting the first exon near the G12D site in KRAS and early exons in p53 and LKB1 using a previously developed informatics tool (Hsu et al., 2013). To select an effective sgRNA for each gene, we individually expressed each sgRNA as well as Cas9 in a cancer cell line (Neuro-2a) and quantified the editing efficiency using the SURVEYOR nuclease assay. Most of these sgRNAs were capable of inducing indels in specific targeted loci (Figure S3A). One sgRNA was chosen for each of the three genes and was then used to generate a single vector containing U6 expression cassettes for all three sgRNAs. Transient transfection of this single vector resulted in significant indels in all three genes (35% KRAS, 44% p53, and 30% LKB1) (Figure S3B).

Whereas p53 and LKB1 mutations in patient tumors are often loss-of-function in nature, mutations in KRAS are often missense mutations that result in a gain of function (TCGA-Network, 2014). To model a missense gain-of-function KRAS mutation, we designed an HDR donor template, which consists of an 800 base-pair (bp) genomic sequence homologous to a region encompassing the first exon of the mouse KRAS gene. This HDR donor encoded: (1) a glycine (G) to aspartate (D) mutation in the 12th amino acid position (G12D), resulting in the oncogenic KRAS^G12D mutation; (2) 11 synonymous single-nucleotide changes to aid the distinction between the donor and wild-type sequences; and (3) protospacer-adjacent motif (PAM) mutations to prevent donor DNA cleavage by Cas9. To test the efficiency of HDR-mediated missense KRAS mutation, we used deep sequencing to assess the rate of G12D incorporation in Neuro-2a cells in vitro. Four percent of the sequencing reads identically matched the donor sequence, which included the G12D mutation and synonymous mutations (Figure S3C). Collectively, these data indicate that a single vector can stimulate HDR events to induce KRAS^G12D mutations as well as facilitate efficient editing of the p53 and LKB1 genes in vitro.

After confirming that Cas9 was expressed in the lung after induction by Cre recombinase (Figures S4A and S4B), we evaluated the efficiency of AAV-mediated gene delivery (Figure 4A). A Firefly luciferase (Fluc) expression vector was packaged using two different serotypes of AAV (AAV6-Fluc and AAV9-Fluc), both of which are capable of transducing the lung efficiently (Bell et al., 2011; Halbert et al., 2002; Limberis and Wilson, 2006). The viruses were delivered using both intranasal (i.n.) and intratracheal (i.t.) methods (Figures 4A, 4B, and S4C). Using in vivo luminescence imaging, we detected luciferase activity in the lung 1 week after i.t. delivery of AAV9-Fluc, but not AAV6-Fluc. Both AAV6-Fluc and AAV9-Fluc showed positive signal in the lung by 3 weeks, but AAV9 generated a higher intensity (Figure S4C). Based on these data, we chose i.t. delivery of AAV9 vectors for all subsequent in vivo studies.

We generated a single AAV vector integrating the KRAS^G12D HDR donor, the U6-sgRNA cassettes for KRAS, p53, and LKB1, as well as an expression cassette containing Cre recombinase and Renilla luciferase (AAV-KPL, Figure 4C). The AAV-KPL vector was packaged using AAV9 (AAV9-KPL) and delivered i.t. into Cre-dependent Cas9 mice, with a relatively low dose of virus for sparse targeting. Four weeks later, lungs were harvested from AAV9-KPL-transduced mice and AAV9-sgLacZ-transduced controls and were characterized by Illumina sequencing. We identified indels in p53 and LKB1 at the predicted cutting site, with 0.1% p53 indels and 0.4% LKB1 indels in the whole lung (Figures 4D, 4E, and 4G). A large fraction of these indels potentially disrupted the endogenous gene function because they were mostly out of frame (i.e., 3n+1bp or 3n+2bp in length) (Figures 4D and 4E). Furthermore, we detected 0.1% KRAS^G12D HDR events in the genomes of the lung cells, as indicated by the synonymous-SNP-barcoded G12D reads (Figures 4F and 4H). The frequency of indels in p53, and LKB1 and targeted KRAS^G12D mutations increased over time, as we detected 1.3% p53 indels, 1.3% LKB1 indels and 1.8% KRAS^G12D mutations in the whole lung at 9 weeks postdelivery (Figures 4G–4H). This was potentially due to the selective growth advantage of mutant cells. These data suggest that delivery of AAV9-KPL into the lungs of Cre-dependent Cas9 mice generates the KRAS^G12D mutation and putative loss-of-function mutations in p53 and LKB1.

To investigate the phenotypic effect of delivery of AAV9-KPL, injected animals were imaged by micro-computed tomography (μCT). Two months postinjection, all (5 out of 5 = 100%) AAV9-KPL-treated animals developed nodules in the lung, whereas none (0/5 = 0%, Fisher’s exact test, one-tail, p = 0.008) of the AAV9-sgLacZ-treated animals showed detectable nodules (Figures 5A and 5B and Movie S1). Quantification showed that the average total tumor burden at 2 months was ~33 mm³ (Figure 5C), which was close to 10% of the total lung volume (Mitzner et al., 2001).

In detail, we observed multiple tumor nodules in the lung of AAV9-KPL-treated, but not AAV9-sgLacZ-treated, mice (Figure 5D). Moreover, certain lobes were dominated by one large tumor, whereas others had many tumors of various sizes (Figures 5D–5G). The sizes of tumors significantly increased over time (Figure 5E). The total tumor area per lobe also significantly increased over time (Figure 5H). The variation in tumor size suggests dynamic tumor initiation and growth across lobes and animals, which is reminiscent of the complexity observed in human tumors (Herbst et al., 2008).

To understand the pathology of tumors formed by injection of the AAV9-KPL vector, we performed hematoxylin and eosin (H&E) staining and immunohistochemistry (IHC). Pathology showed that lungs of AAV9-KPL-treated mice developed multiple grade I and grade II bronchial alveolar adenomas at 1 month postinjection (Figure S4D). In patients, this tumor type is thought to arise from type II pneumocytes (also called alveolar type II cells) at the junction between bronchial and terminal bronchial (Park et al., 2012). These tumors progressed to grade III lung adenocarcinomas in 2 months and occasionally became invasive grade IV adenocarcinomas (Figures 6A–6C and Table S2). None of the AAV9-sgLacZ mice showed any tumors detectable by histology (t test, one-tailed, p = 0.004) (Figures 5D, 6A, and 6B). Clara cell secretory protein (CCSP) staining showed that almost all tumors were adjacent to the clara cells of bronchials (Figure 6D). In addition, most tumors (178/182 = 95.7%) stained positive for pro-surfactant C (pSPC), a marker for type II pneumocytes (Figure 6E), implying that most of these tumors originated from this cell type. Many tumor cells also stained positive for Ki67, an indicator of active cell cycle (Figure 6E), signifying more rapid proliferation in these tumor cells as compared to adjacent normal tissue. All tumors grades II to IV contained CD31-positive endothelial cells (Figure 6E), suggesting that these tumors induced angiogenesis. These data suggest that AAV9-KPL delivery generates a spectrum of tumors from bronchial alveolar adenomas to invasive adenocarcinomas in the lungs of Cre-dependent Cas9 mice in less than 2 months.

We dissected the largest EGFP-positive tumors and adjacent tissues without visible tumors from the lungs of AAV9-KPL-treated mice (Figure 7A) and performed captured Illumina sequencing of KRAS, p53, and LKB1 loci. We detected KRAS^G12D mutations and p53 and LKB1 indels in dissected tumors with low background in adjacent tissues (Figure 7B). The frequencies of p53 or LKB1 indels in dissected tumors from AAV9-KPL-treated animals were highly enriched over the whole lobe, adjacent tissues, and tissues from AAV9-sgLacZ controls, suggesting that p53 and/or LKB1 loss of function were driving the formation of these particular tumors. Interestingly, the frequency of KRAS^G12D mutations was lower in these dissected tumors compared to whole-lobe tissue (Figure 7B). Moreover, KRAS^G12D mutations were much lower than the frequencies of p53 and LKB1 indels. At the whole-lobe level, we observed 0.1% KRAS^G12D HDR at 4 weeks and 1.8% at 9 weeks (Figure 4H), suggesting that KRAS^G12D mutation frequency increases over time in whole tissue but is not enriched in the large tumors (Figure 7B). We did, however, observe a KRAS^G12D-dominated tumor in one out of three small tumors laser captured from lung sections, with 65% KRAS^G12D HDR reads (Figures S5A and S5B). These data suggest that the largest and fastest-growing dissected tumors do not depend on oncogenic KRAS for growth, either because p53;LKB1 double-mutant tumors outcompete KRAS^G12D-containing tumors or because activating KRAS^G12D by HDR in this model has longer latency. This heterogeneity of mutations is reminiscent of mosaicism of multiple mutations in the clinic, with different patients having different combinations (Figures S5C and S5D).

Cre-dependent Cas9 knockin mice were generated by homologous recombination in R1 embryonic stem cells and implanted in C57BL6/J blastocysts using standard procedures, as described previously (Heyer et al., 2010; Nagy et al., 1993). Briefly, a codon-optimized 3 × FLAG-NLS-SpCas9-NLSP2A-EGFP expression cassette was cloned into the Ai9 Roas26 targeting vector (Madisen et al., 2010) and further verified by sequencing. Linearized targeting vector was electroporated into R1 embryonic stem cells followed by G418 and DTA selection for a week. Targeted single-ES cell colonies were screened by PCR with primers amplifying both recombinant arms. PCR products were sequenced to further validate correct insertion. Correctly targeted colonies were injected into C57BL6/J blastocysts for generating chimeric mice. High-percentage male chimeric mice were mated with C57 BL6/J female mice (Jackson Laboratory) to establish germline-transmitted founders. Genotypes of Cas9 mice were determined by amplifying a 4.5 kb product from purified mouse tail DNA (forward: GCAGCCTCTGTTCCACATA CAC; reverse: ACCATTCTCAGTGGCTCAACAA).

SgRNAs were designed using the CRISPRtool (http://crispr.mit.edu) to minimize potential off-target effects. SgRNA sequences and genomic primers are listed in Table S3.

Genomic DNA was extracted from both cells and tissue using Quick Extract DNA extraction solution (Epicenter) following the recommended protocol. Treated cells and tissues were used as a template for PCR for both captured sequencing and SURVEYOR nuclease assay using high-fidelity polymerases (Thermo Scientific) as previously described (Hsu et al., 2013). Genomic PCR products were subjected to library preparation using the Nextera XT DNA Sample Prep Kit (Illumina) or using customized barcoding methods. Briefly, low-cycle, first-round PCR was performed to amplify the target site. Second-round PCR was performed to add generic adapters, which were then used for a third round of PCR for sample barcoding. Samples were pooled in equal amounts and purified using QiaQuick PCR Cleanup (QIAGEN), quantified using Qubit (Life Technologies). Mixed barcoded library was sequenced on an Illumina MiSeq System.

Illumina sequencing reads were mapped to specific amplicons of KRAS, p53, or LKB1 as reference sequences using Burrows-Wheeler Aligner (Li and Durbin, 2010) with custom scripts. Insertions and deletions were called against reference using VarScan2 (Koboldt et al., 2012). Indel length distribution, indel phase, and donor allele frequencies were processed using custom scripts as described previously (Chen et al., 2014; Hsu et al., 2013). The rate of KRAS^G12D HDR was calculated as donor allele frequency based on the G12D mutations as well as the synonymous barcoded SNPs.

The AAV vectors used for stereotaxic injection into the mouse brain were cloned between AAV serotype 2 ITRs and also included the human U6 promoter for noncoding sgRNA transcription, the ubiquitous Cbh promoter (a hybrid form of the CBA promoter) (Gray et al., 2011), HA-tagged Cre recombinase for recombination of loxP-SV40polyA-LoxP, WPRE, and human growth hormone polyA sequence. For nuclei sorting, a similar plasmid was cloned between AAV serotype 2 ITRs and also included the human U6 promoter for noncoding sgRNA transcription, the ubiquitous hSyn promoter, HA-tagged Cre recombinase for recombination of loxP-SV40polyA-LoxP, P2A, EGFP-KASH fusion for nuclei labeling, WPRE, and human growth hormone polyA sequence. Complete plasmid sequences and annotations are found in Data S1.

HEK293FT cells were transfected with the plasmid of interest, pAAV1 plasmid, pAAV2 plasmid, helper plasmid pDF6, and PEI Max (Polysciences, Inc. 24765-2) in Dulbecco's modified Eagle medium (DMEM from Life Technologies, 10569-010). At 72 hr posttransfection, the cell culture media was discarded. Then the cells were rinsed and pelleted via low-speed centrifugation. Afterward, the viruses were applied to HiTrap heparin columns (GE Biosciences 17-0406-01) and washed with a series of salt solutions with increasing molarities. During the final stages, the eluates from the heparin columns were concentrated using Amicon ultra-15 centrifugal filter units (Millipore UFC910024). Titering of viral particles was executed by quantitative PCR using custom Cre-targeted Taqman probes (Life Technologies).

2–3 month old animals were anesthetized by intraperitoneally injection of ketamine (100 mg/kg) and xylazine (10 mg/kg). We also administered buprenorphine HCl (0.1 mg/kg) intraperitoneally as a pre-emptive analgesic. Once subject mice were in deep anesthesia, they were immobilized in a Kopf stereo-taxic apparatus using intra-aural positioning studs and a tooth bar to immobilize the skull. Heat is provided for warmth by a standard heating pad. We drilled a hole on the surface of the skull at 1.94 mm anterior to Bregma and 0.39 mm lateral for injection into the prefrontal cortex. Using a 33 G Nanofil needle and World Precision Instrument Nanofil syringe at a depth of [–2.95 mm, we injected 1 ul (1 × 10¹³ viral genome copies) AAV into the right hemisphere of the brain. Injection rates were monitored by the World Precision Instruments UltraMicroPump3. After injection, we closed the incision site with 6-0 Ethilon sutures (Ethicon by Johnson & Johnson). Animals were postoperatively hydrated with 1 ml lactated Ringer's solution (subcutaneous) and housed in a temperature controlled (37°C) environment until achieving ambulatory recovery. Meloxicam (1–2 mg/kg) was also administered subcutaneously directly after surgery. For downstream analysis, EGFP+ tissue was dissected under a stereotactic microscope.

The single vector design was cloned utilizing ITRs from AAV2, human U6 promoter for noncoding sgRNA transcription, the short EFs promoter derived from EF1a, Renilla luciferase for in vivo luciferase imaging, P2A for peptide cleavage, Cre recombinase for recombination of LSL, a short polyA sequence, and an 800 bp KRAS^G12D homologous recombination donor template. Complete plasmid sequences and annotations are found in Data S1.

AAV6 and AAV9 were packaged and produced in HEK293FT cells and chemically purified by chloroform. In brief, HEK293FT cells were transiently transfected with the vector of interest, AAV serotype plasmid (AAV6 or AAV9), and pDF6 using polyethyleneimine (PEI). At 72 hr posttransfection, cells were dislodged and transferred to a conical tube in sterile DPBS. 1/10 volume of pure chloroform was added and the mixture was incubated at 37^°C and vigorously shaken for 1 hr. NaCl was added to a final concentration of 1 M and the mixture was shaken until dissolved and then pelleted at 20,000 3 g at 4^°C for 15 min. The aqueous layer was discarded while the chloroform layer was transferred to another tube. PEG8000 was added to 10% (w/v) and shaken until dissolved. The mixture was incubated at 4^°C for 1 hr and then spun at 20,000 × g at 4^°C for 15 min. The supernatant was discarded and the pellet was resuspended in DPBS plus MgCl₂ and treated with Benzonase (Sigma) and incubated at 37^°C for 30 min. Chloroform (1:1 volume) was then added, shaken, and spun down at 12,000 × g at 4C for 15 min. The aqueous layer was isolated and passed through a 100 kDa MWCO (Millipore). The concentrated solution was washed with DPBS and the filtration process was repeated. The virus was titered by qPCR using custom Taqman assays (Life Technologies) targeted to Cre.

Intranasal and intratracheal delivery of adeno-associated virus were performed accordingly to previously established protocol (DuPage et al., 2009). In brief, 2- to 3-month-old animals were anesthetized using isoflurane and were set up in a biosafety cabinet. For intranasal delivery, previously titrated virus solution in 50 ml sterile saline was pipetted directly into one nostril of the mouse. For intratracheal delivery, a gauge-24 catheter was inserted to the trachea, and virus solution in 75 μl sterile saline was pipetted to the top of the catheter to allow animal to gradually breathe in the solution. A titer of 1 × 10¹¹ viral genome copies was administered to each mouse. Animals after the procedure were kept warm using a heat lamp for recovery.

Mice were given a lethal dose of Ketamine/Xylazine and transcardially perfused with 0.9% saline followed by 4% paraformaldehyde using a peristaltic pump (Gilson) and fixed overnight. Tissue sectioning was performed on a vibratome (Leica VT1000S) at a thickness of 40 μm. Sections were rinsed three times in phosphate-buffered saline (PBS) with 0.1% Triton X-100 (PBSTx) and blocked with 5% normal goat serum (NGS) (Cell Signaling Technology) in PBS-Tx for 1 hr. Sections were incubated with primary antibodies diluted in PBS-Tx with 5% NGS overnight at 4^°C on an orbital shaker. The following is a list of primary antibodies that were utilized: anti-cleaved caspase 3 (CC3) (1:1,000, Cell Signaling Technology, 9664), anti-yH2AX (1:1,000, Millipore, 05-636) anti-NeuN (1:800, Cell Signaling Technology, 12943), anti-GFP (1:1,600, Nacalai Tesque, GF090R), anti-parvalbumin (1:1,000, Sigma Aldrich, P3088), and anti-tyrosine hydroxylase (1:1,000, Immunostar, 22941). The following day, sections are washed three times in PBS-Tx and then incubated in PBS-Tx with the appropriate AlexaFluor 405, 488, 568 and/or 647 secondary antibody (1:400, Life Technologies) for 3 hours at room temperature on an orbital shaker. After the incubation, sections are rinsed with PBS-Tx three times and then mounted onto superfrost microscope slides (VWR). The sections are then coverslipped with VECTASHIELD HardSet Mounting Medium with DAPI (VECTOR Laboratories) and visualized under a confocal microscope (Zeiss LSM 710, Ax10 ImagerZ2, Zen 2012 Software).

For labeling neuronal nuclei, we generated an AAV-U6-sgRNA-hSyn-Cre-P2A-EGFP-KASH vector. At 3 weeks postinjection, we dissected the EGFP+ injection site in the prefrontal cortex under a stereotactic microscope. As previously described (L.S. and M. Heidenreich, unpublished data), tissue was gently homogenized in 2 ml ice-cold homogenization buffer (HB) (320 mM Sucrose, 5 mM CaCl, 3 mM Mg(Ac)₂, 10 mM Tris [pH 7.8], 0.1 mM EDTA, 0.1% NP40, 0.1 mM PMSF, 1 mM β-mercaptoethanol) using 2 ml Dounce homogenizer (Sigma Aldrich); 25 times with pestle A, followed by 25 times with pestle B. An additional 3 ml of HB was added and the mixture was put on ice for 5 min. Gradient centrifugation was performed using 5 ml 50% OptiPrep density gradient medium (Sigma Aldrich) containing 5 mM CaCl, 3 mM Mg(Ac)₂, 10 mM Tris pH 7.8, 0.1 mM PMSF, 1 mM β-mercaptoethanol. The nuclei-containing mixture was gently added on the top of a 29% iso-osmolar OptiPrep solution in a conical 30 ml centrifuge tube (Beckman Coulter, SW28 rotor). Samples were centrifuged at 10,100 × g (7,500 rpm) for 30 min at 4^°C. The supernatant was removed and the nuclei pellet was gently re-suspended in 65 mM β-glycerophosphate (pH 7), 2 mM MgCl₂, 25 mM KCl, 340 mM sucrose, and 5% glycerol. Intact nuclei were labeled with Vybrant DyeCycle Ruby Stain (Life Technologies) and sorted using a BD FACSAria III. EGFP⁺ nuclei were sorted into individual wells of a 96 well-plate containing 5 μl of QuickExtract DNA Extraction Solution (Epicenter).

Six- to eight-week old constitutive Cas9-expressing female mice were used for all DC experiments. Bone marrow cells were collected from femora and tibiae and plated at concentra tion of 2 × 10⁵/ml on nontreated tissue culture dishes in RPMI medium (GIBCO, Carlsbad, CA, Invitrogen, Carlsbad, CA), supplemented with 10% FBS (Invitrogen), L-glutamine (Cellgro), penicillin/streptomycin (Cellgro), MEM nonessential amino acids (Cellgro), HEPES(Cellgro), sodium pyruvate (Cellgro), β- mercaptoethanol (GIBCO), and GM-CSF (20 ng/ml; Peprotech). SgRNAs targeting Myd88 and A20 were cloned into a guide-only lentiviral vector (Sanjana et al., 2014). At day 2, cells were infected with lentiviruses encoding guide RNAs. Cells were expanded in the presence of GM-CSF. At day 7, infected cells were selected by adding puromycin (Invitrogen) at 5 mg/ml. At day 9, 100 ng/ml LPS (Invivogen) was added 30 min prior to protein analysis or 3 hr prior to mRNA expression profiling. Flow cytometry for GFP detection was performed with BD Accuri C6. Western blot was done using anti-Myd88 (R&D Systems AF3109) and anti-actin (Abcam, ab6276) antibodies.

DCs were processed and analyzed as previously described (Amit et al., 2009). In brief, 5 × 10⁴ cells were lysed in TCL buffer (QIAGEN) supplemented with β-mercaptoethanol. 5% of the lysate was hybridized for 16 hr with a previously described Nanostring Gene Expression CodeSet (Geiss et al., 2008) and loaded into the nCounter Prep Station followed by quantification using the nCounter Digital Analyzer. Counts were normalized using control genes as previously described (Amit et al., 2009) and fold changes were calculated relative to cells transduced with sgRNA targeting GFP and nontargeted controls. Heat maps were created using GENE-E (http://www.broadinstitute.org/cancer/software/GENE-E/).

μCT imaging was performed using standard imaging protocol with a μCT machine (GE Healthcare). In brief, animals were anesthetized using isoflurane and were set up in the imaging bed with a nosecone providing constant isoflurane. A total of 720 views were acquired for each mouse using a soft-tissue-fast-scan setting. Raw image stacks were processed for lung reconstruction using the standard ROI tool (MicroView). Rendering and quantification were performed using render volume tool and measurement tool in MicroView. Tumor burden was calculated as the sum of the sizes in mm³ from all detectable tumors per mouse, with 3–4 mice per group.

Mice were sacrificed by carbon dioxide asphyxiation. Lungs were dissected under a fluorescent stereoscope, fixed in 4% formaldehyde or 10% formalin overnight, embedded in paraffin, sectioned at 6 μm, and stained with hematoxylin and eosin (HE) for pathology. For tumor size quantification, images were taken tiling the whole lobe, merged into a single lobe, and tumors were manually outlined as region of interest (ROI) and were subsequently quantified using ImageJ (Schneider et al., 2012). Sections were de-waxed, rehydrated, and stained using standard immunohistochemistry (IHC) protocols as previously described (Chen et al., 2014). The following antibodies were used for IHC: anti-Ki67 (abcam ab16667, 1:100), anti-CCSP (Millipore, 1:500), anti-pSPC (Millipore AB3786, 1:500) and anti-CD31 (abcam, ab28364, 1:50). IHC was quantified in 10 randomly chosen low-magnification fields per lobe with 3 mice per group using ImageJ (Schneider et al., 2012).

TCGA lung adenocarcinoma (LUAD) mutation data were downloaded from the TCGA data portal (https://tcga-data.nci.nih.gov/tcga/). Mutation annotation format file of LUAD was reprocessed to show KRAS coding missense mutation as well as Tp53 and STK11 loss-of-function mutations in 227 patients.

All animal work was performed under the guidelines of Division of Comparative Medicine (DCM), with protocols (0411-040-14, 0414-024-17 0911-098-11, 0911-098-14 and 0914-091-17) approved by Massachusetts Institute of Technology Committee for Animal Care (CAC), and were consistent with the Guide for Care and Use of Laboratory Animals, National Research Council, 1996 (institutional animal welfare assurance no. A-3125-01).