Rapid in vivo multiplexed editing (RIME) of the adult mouse liver

Abstract

Background and Aims:

Assessing mammalian gene function in vivo has traditionally relied on manipulation of the mouse genome in embryonic stem cells or perizygotic embryos. These approaches are time-consuming and require extensive breeding when simultaneous mutations in multiple genes is desired. The aim of this study is to introduce a rapid in vivo multiplexed editing (RIME) method and provide proof of concept of this system.

Approach and Results:

RIME, a system wherein CRISPR/caspase 9 technology, paired with adeno-associated viruses (AAVs), permits the inactivation of one or more genes in the adult mouse liver. The method is quick, requiring as little as 1 month from conceptualization to knockout, and highly efficient, enabling editing in >95% of target cells. To highlight its use, we used this system to inactivate, alone or in combination, genes with functions spanning metabolism, mitosis, mitochondrial maintenance, and cell proliferation.

Conclusions:

RIME enables the rapid, efficient, and inexpensive analysis of multiple genes in the mouse liver in vivo.

INTRODUCTION

In the past decade, CRISPR/Cas9 (CRISPR-associated protein 9) technology has evolved as a powerful approach for generating genetically engineered mice, enabling researchers to circumvent the time-intensive processes associated with conventional approaches (e.g., embryonic stem cells). In addition, CRISPR/Cas9 has proven to be an efficient tool for reverse genetic screening both in vitro^[1,2] and in vivo.^[3,4] To overcome challenges associated with the delivery of the Cas9 endonuclease and gene-specific single-guide RNAs (sgRNAs), a Cre recombinase-dependent Rosa26-Cas9 knock-in mouse was developed,^[5] which has facilitated in vivo genetic screening in several mouse tissues.^[6–8] Here, we describe modified protocols that make feasible the editing of single or multiple genes in the adult mouse liver for rapid (< 1 month) assessment of gene activity. As a proof of concept for this rapid in vivo multiplexed editing (RIME) system, we investigated genes involved in mitosis, maintenance of mitochondrial DNA, and cell proliferation, including several whose function in the adult liver has not been widely researched.

MATERIALS AND METHODS

Mice

Rosa-LSL-Cas9-EGFP mice,^[5] maintained on a C57BL/6J background, were purchased from The Jackson Laboratory (strain no. 026175) and maintained as homozygotes. Six- to 8-week-old mice were retro-orbitally injected with adeno-associated virus (AAV) serotype 8 (AAV8)-EV, AAV8-U6-sgRNA-TBG-Cre, or AAV8-polycistronic tRNA-gRNA (PTG)-TBG-Cre viruses at 5 × 10¹¹ genome copies per mouse. All mouse procedures were performed in accordance with National Institutes of Health guidelines. All procedures used in this study were in accordance with, and had the approval of, the Institutional Animal Care and Use Committee of the University of Pennsylvania.

Plasmids and cloning

AAV8-Rep/Cap and Ad5-Helper plasmids were obtained from the Grompe lab. The backbone vector (empty vector; EV), AAV-U6-sgRNA-TBG-Cre, was cloned by replacing the pCBh promoter in the AAV: ITR-U6-sgRNA(backbone)-pCBh-Cre-WPRE-hGHpA-ITR (Addgene #60229). For promoter replacement, the thyroid hormone-binding globulin (TBG) promoter was PCR-cloned from pAAV.TBG.PI.eGFP.WPRE.bGH (Addgene #105535), using primers with an XbaI cloning site (5′-GGTTCTAGATGCATGTATAATTTCTACAG) and an AgeI cloning site (5′- GTCCATGCACTTGTCGAGGTC). For sgRNA cloning, sense and antisense oligo DNAs (Integrated DNA Technologies [IDT]; Table S1) were phosphorylated using T4 polynucleotide kinase (New England Biolabs; NEB) at 37°C for 30 min, annealed by ramp-down reaction from 95°C to 25°C at 5°C/min, and cloned by Golden Gate Assembly using SapI restriction enzyme (NEB) and T4 DNA ligase (NEB). Transformation was performed using Stbl3 bacteria (Thermo). After Sanger sequencing (Penn DNA Sequencing Core), validated clones were amplified at scale (150 ml), and plasmid DNA was isolated using the ZymoPURE Plasmid Maxiprep kit (Zymo Research). Endotoxin was eliminated by treating plasmid with Endozero columns (Zymo Research).

For the polycistronic tRNA-gRNA (PTG) system, we constructed AAV-hU6-sgEGFP-tRNA-Hnf4a-sg2-TBG-Cre (“PTG-EGFP/Hnf4a”) and AAV-hU6-Lats1-sg1-tRNA-Lats2-sg2-tRNA-Lats2-sg1-tRNA-Lats2-sg2-TBG-Cre (“PTG-Lats1/2”) by Gibson Assembly. For PTG-EGFP/Hnf4a cloning, a DNA cassette containing sgEGFP-tRNA-Hnf4a-sg2 was PCR-amplified using pGTR (Addgene #63143) as a template and primers listed in Table S2 (see also Figure S9C). For PTG-Lats1/2 cloning, three DNA cassettes (Lats1-sg1-tRNA-Lats1-sg2, Lat1-sg2-tRNA-Lats2-sg1, and Lats2-sg1-tRNA-Lats2-sg2) were PCR-amplified using pGTR and primers listed in Table S2 (see also Figure S10B). Then, DNA cassette(s) were assembled into the SapI-linearized AAV vector using NEBuilder (NEB).

AAV preparation

Cells (293T) grown to 90%–100% confluence in 15-cm dishes were replenished with 15 ml of fresh DMEM (Thermo), supplemented with 2% fetal bovine serum (FBS; EMSCO/FISHER) without antibiotics. For a 15-cm plate, 16 μg of AAV8-Rep/Cap plasmid, 16 μg of Ad5-Helper plasmid, 16 μg of AAV-U6-sgRNA-TBG-Cre plasmid, and 144 μl of 1 mg/ml of polyethylenimine (linear MW 25000) (Polysciences) were mixed in 9 ml of OptiMEM (Thermo). To target a gene, we typically cotransfected three sgRNAs to maximize knockout (KO) efficiency. For this end, we combined 5.3 μg of each sgRNA plasmid so that 16 μg in total was cotransfected to 293T cells. To generate AAV-EGFP-sg1, EV, AAV-Hnf4a-sg2, and PTG viruses, we used 16 μg of single transfer vector. After incubation at room temperature (RT) for 15 min, the plasmid/PEI complex was added to 293T cells in a drop-wise manner, and plates were gently shaken back and forth to mix the medium evenly. After incubation in a CO₂ incubator for 6 days, cells and culture supernatant were harvested into 50-ml tubes and centrifuged at 1900g for 15 min. The supernatant was transferred to new tubes, and a 1:40,000 volume of Benzonase (Sigma-Aldrich) was added and mixed thoroughly by inversion. After the digestion of nonviral DNA by incubating at 37°C for 30 min, virus-containing medium was centrifuged at 1900g for 15 min, and the supernatant was filtered with a 0.22-μM filter unit with polyethersulfone membrane (Thermo). Then, a 1:4 volume of 40% polyethylene glycol 8000 in 2.5 M of NaCl was added and mixed thoroughly by inversion. Following incubation at 4°C for at least overnight, precipitated AAV was collected by centrifugation at 3000g for 15 min. After removal of the supernatant, precipitate was homogenized in 100 μl of PBS per 15-cm dish by through pipetting. Non-AAV precipitate was eliminated by centrifugation at 2200g for 5 min. Smaller debris were further removed by filtrating the eluted AAV with a 0.45-μM filter columns. This crude AAV was titrated by qPCR using the AAV8-TBG-Cre (Penn Vector Core) as a standard template and primers listed in Table S3. Virus was then used directly for in vivo experiments without further purification.

Western blotting

Liver tissue was mechanically homogenized on ice with a pellet pestle (Fisher) before lysis. Isolated hepatocytes or homogenized liver tissue was suspended at ~1:20 (v/v) in radioimmunoprecipitation assay buffer supplemented with Halt Protease and Phosphatase Inhibitor Cocktail (Thermo) and pipetted thoroughly. After incubation on ice for 30–60 min, lysate was centrifuged at 14,000g at 4°C for 15 min, and the supernatant was transferred to new tubes. Protein was then quantified using the Pierce BCA Protein Assay Kit (Thermo). Equal amounts of protein, ranging from 5 to 20 μg depending on target abundance, were loaded and separated in 4%–20% Mini-PROTEAN gels (Bio-Rad) and then transferred to polyvinylidene difluoride membrane (Millipore) at 100 V for 90 min. After blocking in 5% milk in PBS with Tween-20 at RT for 30 min, the membrane was probed with primary antibodies (Table S4) at RT for 1 h or 4°C overnight. Secondary antibody reaction was performed with donkey anti-rabbit, -mouse, or -goat antibodies conjugated with horseradish peroxidase (Jackson Immuno) at RT for 1 h. Blottings were developed with electrochemiluminescence (Thermo), ECL Plus (Thermo), or West Femto SuperSignal (Thermo) substrate and imaged using a ChemiDoc Imaging System (Bio-Rad).

Immunofluorescence

Tissue samples were fixed with Zinc Formalin Fixative (pH 6.25; Polysciences) and embedded in paraffin. Following dewaxing and rehydration, heat-induced epitope retrieval was performed by boiling specimens in R-buffer (Electron Microscopy Sciences) at 121°C for 15 min. Then, specimens were permeabilized with 0.1% Triton X-100 (Fisher). After treatment with Blocking One Histo at RT for 10 min, specimens were incubated with primary antibodies (Table S5) diluted in 1/20× Blocking One Histo at RT for 1 h or at 4°C overnight. Sections were then stained using donkey anti-rabbit, -rat, or -goat antibodies conjugated with Alexa Fluor 488 or Alexa Fluor 594 (Thermo) at a 1:300 dilution as well as DAPI (Thermo) at a 1:1000 dilution. After incubation at RT for 1 h, specimens were mounted in Aqua-Poly/Mount (Polysciences) and imaged on an Olympus IX71 inverted fluorescent microscope.

Hematoxylin-eosin, Oil Red O, and F-actin staining

Hematoxylin-eosin (H&E) staining was performed by the Penn Molecular Pathology and Imaging Core (MPIC), or by standard protocols. Oil Red O staining (frozen sections only) was performed by the MPIC. F-Actin was stained by incubating specimens with Flash Phalloidin Red 594 (BioLegend) at 1:500 dilution.

Hepatocyte isolation

Livers were perfused with 40 ml of Hanks’ balanced salt solution (HBSS; Thermo), followed by 40 ml of HBSS with 1 mM of EGTA (Sigma), and 40 ml of HBSS with 5 mM of CaCl₂ (Sigma-Aldrich) and 40 μg/ml of liberase (Sigma-Aldrich). Following perfusion, livers were mechanically dispersed with tweezers, resuspended in 10 ml of wash medium (DMEM supplemented with 5% FBS), and filtrated with a 70-μM cell strainer. Cells were centrifuged at 50g at 4°C for 5 min. Then, cells were resuspended in complete Percoll solution (10.8 ml of Percoll [Cytiva], 12.5 ml of wash medium, and 1.2 ml of 10× HBSS per mouse) and centrifuged at 50g at 4°C for 10 min. After a single wash with 10 ml of medium, cells were spun at 50g at 4°C for 5 min and then used for downstream experiments.

Enhanced green fluorescent protein flow cytometry

Cells were resuspended in flow buffer, consisting of HBSS (pH 7.4) supplemented with 25 mM HEPES (Thermo), 5 mM MgCl₂ (MedSupply Partners), 1× Pen/Strep (Thermo), 1× Fungizone (Thermo), 1× nonessential amino acids (Thermo), 1× Glutamax (Thermo), 0.3% glucose (Sigma-Aldrich), and 1× sodium pyruvate (Thermo). After adding 1 μg/ml of DAPI (Thermo), enhanced green fluorescent protein (EGFP) fluorescence was analyzed on an LSR II flow cytometer (BD Biosciences).

Ploidy analysis

For cellular ploidy analysis, freshly isolated hepatocytes were suspended in 10% FBS-DMEM supplemented with 15 μg/ml of Hoechst 33342 (Thermo) and incubated at 37°C for 30 min. Following the addition of TO-PRO-3 iodide (Thermo) at 1 mM, cells were analyzed on an LSR II flow cytometer. For nuclear ploidy analysis, nuclei were isolated by suspending cells in 10:1 (v/v) in nuclear isolation buffer (NIB; 15 mM of Tris–HCl [pH 7.5], 60 mM of KCl, 15 mM of NaCl, 5 mM of MgCl₂, 1 mM of CaCl₂, and 250 mM of sucrose) supplemented with 0.1% Nonidet P-40 (NP-40; Millipore) on ice for 5 min. After two washes in NP-40-free NIB followed by centrifugation at 600g for 5 min at 4°C, nuclei were resuspended in 500 μl of NIB containing 1 μg/ml of DAPI and analyzed on an LSR II.

Primary culture of hepatocytes

Culture medium consisted of DMEM/F12 (Corning) containing 2.4 g/l of NaHCO₃ and L-glutamine, supplemented with 5 mM of HEPES (Corning), 30 mg/l of L-proline (Sigma-Aldrich), 0.05% bovine serum albumin (Sigma-Aldrich), 10 ng/ml of epidermal growth factor (Sigma-Aldrich), insulin-transferrin-serine (ITS)-X (Thermo), 10⁻⁷ M of dexamethasone (Sigma-Aldrich), 10 mM of nicotinamide (Sigma-Aldrich), 1 mM of ascorbic acid-2 phosphate (Wako), gentamycin (Thermo), 10% FBS, 10 μM of Y^−27,632 (LC Laboratories), 0.5 μM of A-83–01 (AdooQ BioScience), and 3 μM of CHIR99021 (LC Laboratories). Isolated cells were plated on 12-well collagen-coated plates (IWAKI) at 2 × 10⁴ cells per well. Culture medium was replaced every 2–3 days.

Total RNA isolation and reverse transcription

Total RNA was extracted using a NucleoSpin RNA Kit (Takara), following the manufacturer’s instructions. Approximately 500 ng of RNA was reverse transcribed in 20 μl using a High Capacity cDNA Reverse Transcription Kit (Thermo). Complementary DNA (cDNA) was diluted 1:20 in water and used for qPCR.

Total DNA isolation

Total DNA was isolated from hepatocytes using the DNeasy Blood & Tissue Kit (Qiagen). DNA was diluted to 5–10 ng/μl and used for qPCR.

qPCR

qPCR was performed at 10 μl per well using the Bio-Rad CFX 384 qPCR machine (Bio-Rad): 3 μl of cDNA or genomic/mitochondrial DNA diluted in water as described above, 0.25 μl each of 10-μM forward and reverse primers (Table S3), 1.5 μl of H₂O, and 5 μl of SsoAdvanced SYBR reagent (Bio-Rad).

Transmission electron microscopy

Electron microscopy imaging was performed at Penn Electron Microscopy Resource Laboratory.

Blood chemistry

All the blood chemistries (serum) were determined by IDEXX Laboratories.

Flow cytometry of nonparenchymal cells

To assess the cellular specificity of the AAV8-TBG-Cre system, Empty-sgRNA-vector–packaged AAV8 (AAV8-TBG-Cre) was injected into Cas9-EGFP mice. Livers were digested by the two-step liberase perfusion as described above. Then, the undigested remaining tissue was transferred to a 1.5-ml tube, minced with surgical scissors, and further digested with 10× concentrated liberase (400 μg/ml with a volume of ~430 μl per tube) at 37°C for 30 min. The digested tissue was filtered with a 70-μM cell strainer and combined to the cell suspension digested before. Cells were then centrifuged at 300g at 4°C for 5 min. Then, cells were suspended in 10 ml of ammonium-chloride-potassium lysis buffer (Quality Biological) and incubated on ice for 10 min to remove red blood cells. After centrifugation at 300g at 4°C for 5 min, cells were resuspended in 3 ml of flow buffer and filtered with a 40-μM cell strainer. Cells were transferred to a round-bottomed 96-well plate at 100–150 μl per well and centrifuged at ~800g at 4°C for 1 min. Cells were then resuspended in 100 μl per well of flow buffer containing fluorophore-conjugated antibodies (Table S6) and incubated on ice for 20 min. After two washes in flow buffer (150–200 μl per well, ~800g at 4° C for 1 min), cells were resuspended in flow buffer containing 1:1000× TO-PRO-3 (Thermo) and analyzed using an LSR II flow cytometer.

Indel analysis by deep sequencing

Genomic regions flanking each cut site were amplified by touchdown PCR with a Phusion Flash High-Fidelity PCR Master Mix (Thermo), using primers listed in Table S7. Annealing temperature was set to the calculated lower Tm + 7°C of the primer pairs and decreased by 1°C each cycle to the Tm −3°C in the first 10 cycles; then, PCR was continued for another 20 or 25 cycles at Tm −3°C. Library prep was performed with NEBNext Ultra II DNA kits (NEB), and amplicons were sequenced on an Illumina MiSeq using the 2 × 250 bp MiSeq Reagent Kit v3. Illumina indices were demultiplexed using Illumina bcl2fastq2. Sequencing adapters were trimmed using BBDuk from bbtools (https://jgi.doe.gov/data-and-tools/software-tools/bbtools/) with parameters “ktrim=r k=23 mink=1 hdist=1 tpe tbo.” Combined samples were demultiplexed using a custom Python script. Briefly, the first 6 bp of the forward and reverse reads respectively were treated as inline barcodes and assigned to their corresponding samples, considering both possible orientations of the paired-end reads. Paired-end reads were merged using FLASH merger (https://ccb.jhu.edu/software/FLASH/) with parameters “-m 10 -r 250 -f 420 -s 40 -p 33” that ensures a minimum overlap of 10 bp. We obtained 12,894 ± 6238 reads for 67 samples (mean ± SD). Reads were analyzed with respect to their expected CRISPR cleavage site(s) with a custom Python script. Briefly, for a single cut site, each read was aligned independently to the sequence directly preceding the cut site and to the sequence directly following the cut site using Blastn^[9] with a word size of 28. Extent of net insertions and deletions was calculated accordingly, and extent of microhomology was inferred based on overlap of the alignments. For sequences with two cut sites, a perfect joining of the two cut ends was considered to have zero indels, and sequences containing the excised region (not cleaved) were scored separately. Downstream analysis of the tabulated data for each sequence was performed in Excel and R v4.2.1. Custom Python scripts are available upon request. Sequencing data and the corresponding analysis have been deposited to the Gene Expression Omnibus (accession number GSE209543).

Assay for transposase-accessible chromatin with sequencing

A total of 50,000 hepatocytes were isolated from three AAV8-EV–injected (1 × 10¹² genome copies per mouse) liver samples at 4 days postinjection (dpi) and used as input for assay for transposase-accessible chromatin with sequencing (ATAC-seq) library preparation. Libraries were prepared as described,^[10] with minor modifications. Briefly, nuclei were isolated from cells using a solution of 10 mM of Tris–HCl pH 7.4/10 mM of NaCl/3 mM of MgCl₂/0.1% IGEPAL CA-630. Immediately following isolation, the transposition reaction was conducted using Tn5 transposase and TD buffer (Illumina) for 30 min at 37°C. Transposed DNA fragments were purified using a Qiagen MinElute Kit, barcoded, and PCR-amplified for seven to nine cycles depending on the samples using NEBNext High Fidelity 2× PCR master mix (New England Biolabs). Cycle number was determined empirically each time by qPCR. Libraries were then purified with AMPure XP beads. Paired-end 150 × 2 sequencing was performed by Novogene (Sacramento, CA), using a NovaSeq (Illumina). Reads were aligned to the mouse genome (mm10) using Bowtie2 with options ‘--very-sensitive -X 1000 --dovetail -1’^[11] and duplicates were removed using Picard (http://broadinstitute.github.io/picard/). Sequencing data and the corresponding analysis have been deposited to the Gene Expression Omnibus (accession no.: GSE209556).

Statistical analysis

For comparison of two groups, a two-sided t test was performed. For comparison of multiple groups, first ANOVA was conducted. If the global p-value by ANOVA was less than 0.05, multiple t-tests were performed using the EV control as the reference group. p-values were calculated using stat_compare_means function in the GGPUBR R package.

RESULTS

Efficient hepatocyte KO is achieved within 2 weeks after AAV injection

The RIME system has two components (Figure S1): (1) an AAV8 carrying Cre recombinase under the control of a tissue-specific promoter and an sgRNA under the control of the human U6 promoter and (2) host LSL-Cas9-EGFP mice,^[5] in which CAG-driven expression of Cas9 and EGFP is activated by Cre-mediated removal of an upstream transcriptional stop sequence. To target hepatocytes, we modified the existing AAV construct^[5] by replacing the ubiquitous CBh promoter with the hepatocyte-specific TBG promoter to drive Cre expression. AAV preparation from HEK293T cells was performed using a streamlined protocol (see Materials and Methods), and crudely processed viral preps were introduced into LSL-Cas9-EGFP mice by retro-orbital injection. To test the specificity of the delivery system, we injected AAV8 lacking an sgRNA (empty vector, AAV-EV) to Cas9-EGFP mice, harvested various tissue, and confirmed EGFP expression only in the liver (Figure S2A). Flow cytometry confirmed that >99% of Percoll-enriched hepatocytes were EGFP⁺ at 14 dpi (Figure S3A). In contrast, we did not observe EGFP expression in nonparenchymal cells, including epithelial cell adhesion molecule–positive biliary epithelial cells (BECs), Cd31⁺ vascular endothelial cells, Cd45⁺ whole immune cells, Cd45⁻Cd90⁺ periportal fibroblasts, and Cd45⁻Cd90⁻Cd105⁺ nonperiportal fibroblasts (Figure S2B–H).

Using this platform, we first estimated the kinetics of gene editing by targeting EGFP. Injection of AAV-EV resulted in the expression of EGFP in ~90% hepatocytes at 7 dpi and > 99% at 14 dpi, as assessed by flow cytometry (Figure S3A). By contrast, injection of AAV8 packaged with an sgRNA targeting EGFP (AAV-sgEGFP) led to a substantial decrease of EGFP⁺ cell fraction as well as mean fluorescence intensity (MFI) of EGFP⁺ cells at 14–21 dpi (Figure S3A). This observation was further validated by immunohistochemistry (IHC; Figure S3B) and western blotting (WB; Figure S3C). Sanger sequencing of bulk genomic DNA derived from isolated hepatocytes confirmed perturbation in the chromatograph at the proximity of the expected sgEGFP cut site (Figure S3D). Moreover, deep sequencing of PCR-amplified genomic DNA demonstrated that 61.5% of amplicons had insertions or deletions (indels) at 7 dpi. Indel abundance was further elevated to ~94% at 14 and 21 dpi (Figure S3E). Indel distribution indicated that the most common indel was a 1-bp insertion, suggesting that canonical non-homologous end joining (c-NHEJ) was the dominant mode of editing (Figure S3F).^[12] We also noted that 2- to 30-bp deletion were the most frequent deletion events, suggesting the contribution of alternative NHEJ, also known as microhomology-mediated end joining (MMEJ; Figure 3G).^[12] These data demonstrate that this system permits rapid and highly efficient gene editing in vivo.

FIGURE 3.

Mice tolerate severe mtDNA loss. (A) KO validation for Tfam and Polg by Western blotting. (B) KO validation for Polg2 by qRT-PCR using Polg2-sg2 as a forward primer. The upper diagram indicates the amplified region targeting the Polg2-sg2 cut site as an example. The lower panels indicate mRNA expression (qRT-PCR, normalized to beta-actin) of Tfam, Polg, and Polg2 following introduction of each of the indicated Polg2 sgRNAs. (C) Changes in mtDNA abundance in hepatocytes over time as determined by qPCR (n = 2–4 for each data point). mtDNA content were normalized to the Hk2 gene (nuclear encoded). The values of EV-injected hepatocytes at 2 wpi are set to 100%. (D) Body weights of mice injected with EV, sgTfam, sgPolg, and sgPolg2 measured at the designated time points. (E) Serum biochemistries at 6 wpi (n = 2–3). The units on the y-axes are U/L for ALP, ALT and AST, μM for Bile acid, and mg/dl for Cholesterol and Total bilirubin. (F) Blood biochemistry at 10 wpi (n = 2–4). Units on the y-axes are indicated in (E). (G) HE staining of EV, sgTfam, sgPolg, and sgPolg2-injected livers at 10 wpi. (H) Transmission electron microscopic images of mitochondria in EV and sgTfam-injected hepatocytes at 10 wpi. (I) Time course change in the transcripts of mtDNA-encoded genes in hepatocytes as determined by qPCR (n = 2–4 at each point). Expression levels are normalized to Gapdh. The values of EV-injected hepatocytes at 2 wpi are set to 100%. Statistical differences were calculated using multiple t test with EV as the reference. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. ALP, albumin; ALP, alkaline phosphatase; AST, aspartate aminotransferase; Fw, forward; Gapdh, glyceraldehyde 3-phosphate dehydrogenase; PAM, proto-spacer adjacent motif.

As another validation of the system, we targeted Fah (fumarylacetoacetate hydrolase). Deletion of Fah is known to cause a toxic accumulation of tyrosine catabolites within hepatocytes, and Fah null mice do not survive unless the upstream tyrosine catabolic pathway is inhibited or the liver is repopulated with wild-type hepatocytes.^[13] We designed AAV8 vectors packaged with two sgRNAs targeting Fah and injected AAVs into Cas9-EGFP mice. WB demonstrated a substantial reduction of Fah protein at 2 weeks postinjection (wpi) and an almost complete loss at 3 wpi (Figure S4A). Deep sequencing demonstrated that ~95% of each predicted cut site had indels at 2 wpi (96.6% for sg1; 94.8% for sg2), which was comparable at 3 wpi (97.3% for sg1; 97.6% for sg2) with a very consistent indel distribution (Figure S4B,C). Accordingly, Fah KO mice started to manifest weight loss at 3–4 wpi (Figure S4D) and mortality at ~5 wpi (Figure S4E). Surviving mice (44%; 4 of 9) exhibited repopulation by EGFP⁻ hepatocytes, which escaped AAV infection or Cre recombination (Figure S4F): EGFP⁻ escaper hepatocytes increased by 500-fold (0.2% at 3 wpi to 90% at 9 wpi; Figure S4F), highlighting the extremely high regenerative capacity of hepatocytes when placed under selective pressure. Another possible explanation for the extensive repopulation by EGFP⁻ hepatocytes is transdifferentiation-mediated regeneration by cells of non-hepatocyte origin, including BECs or putative liver stem cells.^[14] However, doubling the dose of AAV-sgFah resulted in the death of all animals (n = 4; Figure S4G), arguing against a contribution by those cell types.

Targeting hepatocyte nuclear factor 4 alpha in the adult liver

To assess our system functionally, we designed AAV8 vectors with sgRNAs targeting hepatocyte nuclear factor 4 alpha (Hnf4a), a known master regulator of hepatocyte identity and function. In a previous study, in which Alb-CreER^T2 mice were crossed to Hnf4a^loxP/loxP mice, Hnf4a deletion was noted to result in marked hepatocyte proliferation and hypertrophy within 2–3 weeks of tamoxifen treatment, contrasting with the phenotypes observed following embryonic deletion of Hnf4a.^[15] We therefore reasoned that knocking out Hnf4a (Hnf4a KO) in adult hepatocytes would provide a good test of the system’s fidelity.

We generated a pooled AAV8 viral prep carrying three sgRNAs against Hnf4a by cotransfecting HEK293T cells with plasmids carrying sgRNAs targeting different parts of the gene. At 10 dpi, we observed a decrease in Hnf4a mRNA in vivo by qRT-PCR with primer pairs proximal to each cut site (Figure S5A) and confirmed a complete loss of Hnf4a protein by WB (Figure 1A). Deep sequencing of each cut site using a 14-dpi KO sample demonstrated that ~90% of amplicons were mutated (85.8% for sg1, 89.9% for sg2, and 93.8% for sg3; Figure S5B,C). Accordingly, livers of mice injected with sgHnf4a became pale, a characteristic of steatosis (Figure 1B). H&E staining revealed scant cytoplasm, and Oil Red O staining confirmed an abnormal accumulation of oil droplets in sgHnf4a-injected hepatocytes (Figure 1C). Blood chemistries indicated that liver damage peaked at 10–14 dpi, with recovery evident by 22 dpi (Figure 1D), observations consistent with the finding that bulk Hnf4a mRNA expression began to recover by 3 wpi (Figure S5A), in alignment with previous work.^[15] Notably, we found that expression of hepatocyte-specific genes decreased in Hnf4a KO livers (Figure 1E), whereas expression of BEC genes increased (Figure S5D). These results suggest that Hn4a KO hepatocytes engage in some degree of hepatocyte-to-biliary reprogramming.^[16,17]

FIGURE 1.

Hnf4a-LKO causes liver injury and repopulation by escaper cells. (A) WBs of Hnf4a at the designated time points. (B) Representative macroscopic images of EV- or sgHnf4a-injected mouse livers at 10 dpi. (C) Representative images of H&E (top) and Oil Red O staining (bottom) at 14 dpi. (D) Serum biochemistries of EV- or sgHnf4a-injected mice (n = 4 at each time point). Units on the y-axes are U/l for ALP, ALT, and AST and mg/dl for cholesterol, HDL, total bilirubin, and triglyceride. (E) Gene expression analysis of hepatic marker genes by qRT-PCR (n = 4 for each time point). Data are normalized to Gapdh expression. (F) Quantification of EGFP⁺ cells by flow cytometry (n = 4–6 for each time point). (G) Representative immunofluorescent images for EV- or sgHnf4a-injected livers at the designated time points. White arrows indicate EGFP⁻Hnf4a⁺ cells. Yellow arrows indicate EGFP⁺Hnf4a⁺ cells. Arrowheads indicate EGFP⁺Hnf4a⁻ cells. Statistical differences were calculated using multiple t test. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Alb, albumin; ALP, alkaline phosphatase; AST, aspartate aminotransferase; Cebpa, CCAAT enhancer binding protein alpha; CV, central vein; Gapdh, glyceraldehyde 3-phosphate dehydrogenase; PV, portal vein; Trf, transferrin; Ttr, transthyretin.

To obtain insight into the origin of cells contributing to recovery of Hnf4a expression, we measured the abundance of EGFP-labeled cells by flow cytometry. Whereas the vast majority of hepatocytes were EGFP⁺ at 14 dpi (96.9 ± 0.71%), the peak of labeling, Hnf4a KO mice exhibited a decrease in EGFP⁺ fraction as early as 22 dpi (88.9 ± 2.0%) and a further decrease at 84 dpi (37.5 ± 7.4%; Figure 1F). As was the case for the Fah KO experiment, this observation raised the possibility that hepatocytes that had escaped Hnf4a deletion were responsible for repopulating the liver at the expense of Hnf4a-deficient cells. IHC analysis at 84 dpi confirmed that most hepatocytes in Hnf4a KO livers expressed Hnf4a, whether they were EGFP⁺ or EGFP⁻ (Figure 1G). Thus, cells that escaped either Cre-mediated recombination or Cas9-mediated editing of the Hnf4a gene were the dominant source of hepatocytes >2 months after infection. Given that these results recapitulated and expanded findings previously obtained through traditional gene-targeting approaches (see Discussion), we proceeded to target genes whose functions have not been widely investigated in the liver.

Kinesin family member 11 is necessary for the maintenance of hepatocyte ploidy

In rodents, ~70%–90% of adult hepatocytes are polyploid.^[18,19] Polyploidization of hepatocytes occurs mainly through cytokinesis failure and is developmentally programmed to begin at weaning and increase with age.^[19] Genetic studies have identified multiple cell-cycle–related genes as suppressors or promoters of polyploidization in hepatocytes, including members of the E2F family (reviewed in an earlier work^[18]). Deletion of E2F family members can lead to either an increased (E2f1, E2f2, and E2f3) or a decreased (E2f7 and E2f8) number of polyploid hepatocytes through suppression or promotion of endoreplication, respectively. However, these genes have versatile functions in cell-cycle regulation, leaving the detailed molecular mechanism of hepatocyte polyploidization unclear.

To evaluate whether disrupting mitotic spindle formation would affect the ploidy status of quiescent hepatocytes, we targeted the Kif11 (kinesin family member 11) gene. Kif11, also known as kinesin-5, is essential for mitosis, where it participates in self-assembly of the microtubule-based mitotic spindle.^[20] Disruption of Kif11 in dividing cells results in mitotic catastrophe and cell death, and thus germline KO of Kif11 causes early embryonic lethality.^[21] Kif11 also contributes to axon extension in adult neurons,^[20] but its role in other tissues—particularly in quiescent cells—remains open for study.

We injected AAV-EV or AAV-sgKif11 viruses into 6-week-old LSL-Cas9-EGFP mice to generate Kif11 KO mice (or EV controls) and harvested hepatocytes after 2 weeks. Neither EV nor Kif11 KO hepatocytes had detectable levels of Kif11 protein by WB (Figure 2A), consistent with the quiescent state of most hepatocytes in the normal adult liver.^[22] However, upon plating in optimized hepatocyte culture medium,^[23] EV hepatocytes readily expressed Kif11, whereas Kif11 KO hepatocytes failed to express the protein, confirming efficient deletion (Figure 2A). Deep sequencing of the genomic regions flanking each predicted cut site showed that editing efficiency of Kif11 was 60%–90% (88.7% for sg1; 59.8% for sg2; and 79.3% for sg3) at 2 wpi, which increased to 86%–91% (87.1% for sg1; 90.8% for sg2; and 86.4% for sg3) at 3 wpi (Figure 2B, Figure S6A). Although the kinetics of Kif11 editing were slower than that of EGFP, Fah, and Hnf4a, the gene was still likely to have been efficiently edited as early as 2 wpi. Specifically, the probability that a given allele escaped editing at all the three cut sites was estimated to be only 0.94% at 2 wpi, assuming that editing at each cut site is independent (i.e., rate of editing escape = 0.113 × 0.402 × 0.207 × 100).

FIGURE 2.

LKO of Kif11, a gene involved in mitotic spindle formation, causes elevation of hepatocyte ploidy. (A) WBs of freshly isolated hepatocytes or hepatocytes cultured in the presence of three signaling inhibitors—Y27632 (ROCK inhibitor), A83–01 (TGFβ inhibitor), and CHIR99021 (GSK3 inhibitor)^[18]—for 5 days. Hepatocytes were isolated from Cas9-EGFP mice injected with EV and sgKif11 at 2 wpi. (B) Quantification of editing efficiency at each cut site of the three Kif11-sgRNAs, as determined by deep sequencing using EV control and KO samples at 2 and 8 wpi. (C) Representative flow cytometry plots of Hoechst33342-loaded hepatocytes isolated from EV- or sgKif11-injected mice at the designated time points. (D) Quantification of data in panel C (n = 3–5 for each data point). (E) Representative H&E images of EV or sgKif11-injected mouse livers at the designated time points. (F) Summary of serum chemistries of mice with Kif11 LKO at 13 wpi. Samples from age-matched WT (no AAV injection) mice are shown as controls. Values are mean ± SEM. ALP, albumin; ALP, alkaline phosphatase; AST, aspartate aminotransferase; FSC-A, forward scatter area; Gapdh, glyceraldehyde 3-phosphate dehydrogenase; GSK3, glycogen synthase kinase-3; ROCK, Rho-associated coiled-coil containing protein kinase; WT, wild type.

To assess the consequences of Kif11 loss, we measured ploidy in EV and Kif11 KO hepatocytes by performing Hoeschst staining on cells obtained from mice between 2 and 12 wpi. Whereas EV hepatocytes showed minimal changes in their ploidy profile over the 12-week period, Kif11 KO hepatocytes exhibited a dramatic increase in 8c and 16c hepatocytes over the same time period, particularly between 4 and 8 wpi (Figure 2C,D). H&E staining confirmed that both nuclear and cell size were increased in hepatocytes from Kif11 KO livers (Figure 2E). Intriguingly, most hepatocytes in Kif11 KO livers were mononucleated (Figure 2E), suggesting that the increase in ploidy was the result of endoreplication rather than failed cytokinesis (following successful karyokinesis). Flow cytometry and quantification of ploidy in isolated nuclei support this conclusion (Figure S6B,C).

Despite hepatocyte enlargement, the overall size of livers in Kif11 KO mice was unaltered at 13 wpi (Figure S6D), suggesting that total hepatocyte number was reduced. Cell death followed by hepatocyte hypertrophy may be one mechanism, given that Miyaoka et al. reported that hepatocytes compensate for mild tissue loss (30% partial hepatectomy) by increasing hepatocyte volume rather than through cell division.^[24] To determine whether cells that had escaped Kif11 loss were more fit than Kif11 KO cells, we again measured the fraction of EGFP⁺ over time. Whereas most hepatocytes remained EGFP⁺ through weeks 8 and 12 wpi (EGFP⁺ rate = 96.6 ± 2.2% and 95.9 ± 1.6%, respectively), we observed substantial repopulation by escaper (EGFP⁻) cells at 24 wpi (EGFP⁺ rate = 60.4 ± 16.9%; Figure S6E). These results suggest that loss of Kif11 is detrimental to hepatocyte viability, even under homeostatic conditions where rates of hepatocyte proliferation are low. Indeed, we found that there was a slight, but significant, increase of blood alanine aminotransferase (ALT) level as well as a decrease in triglyceride level in Kif11 KO mice preceding substantial repopulation (13 wpi; Figure 2F). This observation suggests that Kif11 loss reduces hepatocyte viability, which, in turn, triggers wild-type hepatocytes to enter the cell cycle.

Mice tolerate transient loss of mitochondrial DNA in hepatocytes

Mitochondrial diseases can present at any age and manifest in any organ,^[25] and germline mutations that disrupt mitochondrial function are the most common cause of mitochondrial diseases.^[26] Although mitochondria have their own ~16.5-kb genome, encoding 37 genes, the genes responsible for maintaining mitochondrial DNA (mtDNA) are encoded in the nucleus. Mutations in mitochondrial maintenance genes result in mtDNA depletion syndrome (MDS), but the mechanisms of mtDNA loss, and the tissue specificity of the resulting disease phenotypes, remain unclear.^[26] To date, there has been limited investigation of the major mitochondrial maintenance genes in the liver, where disease phenotypes associated with MDS are common.

To address this knowledge gap, we used RIME to identify liver phenotypes associated with loss of three mtDNA maintenance genes—transcription factor A, mitochondrial (Tfam), polymerase gamma (Polg), and Polg2—the germline deletion of which cause embryonic lethality at E8.0–10.5 in the mouse.^[27–29] Tfam was originally identified as a transcription factor necessary for mtDNA transcription and was subsequently found to be important for mtDNA replication and maintenance in vitro.^[27,30] Heterotrimeric polymerase γ, the mitochondria-specific polymerase, is composed of a catalytic subunit encoded by Polg and a homodimeric accessory subunit encoded by Polg2, which is required for the efficient binding of Polg to mtDNA. Defects in all three genes are associated with MDS, but expressivity and tissue tropism vary; specifically, defects in TFAM and POLG typically involve liver diseases, whereas liver involvement with POLG2 mutations is rare.^[25,26,31]

To inactivate these genes in the liver, we injected LSL-Cas9-EGFP mice with AAV viruses carrying three targeting sgRNAs as described above. We first performed WB to confirm protein loss. Whereas Polg protein was absent by 2 wpi, Tfam persisted longer, but was largely absent by 6–8 wpi (Figure 3A). This extended persistence of Tfam is likely attributable to greater protein or mRNA stability, because we confirmed efficient editing at 2 wpi as assessed by deep sequencing (25.6% for sg1; 98.8% for sg2; 92.3% for sg3; probability that a given allele remains intact estimated at 0.069%; Figure S7A). The predicted cut sites of Tfam-sg2 and Tfam-sg3 were located back to back with a 53-bp gap, thereby providing the opportunity to interrogate whether simultaneous editing of the adjacent loci results in the removal of the intervening sequences. Indel distribution of Tfam cut sites for sg2 and sg3 clearly demonstrated that >30-bp deletions dominated the indels (Figure S7B). Indeed, genome browser views of the Tfam KO sample demonstrated that most of the amplicons flanking these cut sites had a ~53- to 54-bp gap (Figure S7C), highlighting the reliable destruction of the target gene by simultaneous editing of the two cut sites. We were unable to detect Polg2 by WB using two commercially available antibodies (data not shown). As an alternative, we confirmed by deep sequencing that AAV-sgPolg2–infected hepatocytes had efficient editing at each cut site (65.3% for sg1; 94.6% for sg2; and 95.2% for sg3; Figure S7A,B). In addition, qRT-PCR using the Polg2-sgRNAs as a forward primer for each of the three cut sites confirmed the reduction of Polg2 transcripts near the cut sites (Figure 3B). Based on these results, we conclude that all the three genes are susceptible to in vivo knockout using this technique.

Unexpectedly, we observed that liver-specific deletion of either Tfam or Polg2 led to reduction of Polg protein as early as 2 wpi, whereas knocking out Polg or Polg2 led to mild reductions of Tfam protein (Figure S8A). qRT-PCR did not detect down-regulation of Polg mRNA by Tfam KO or Polg2 KO (Figure S8B). These results suggest that Tfam and Polg2 are required for Polg stability at the protein level, consistent with a recent report that Polg2 KO or knockdown in vitro leads to reductions in Polg and Tfam.^[32] We also observed that Tfam KO hepatocytes down-regulated Polg2 mRNA expression (Figure S8B), suggestive of additional regulation at the RNA level. Collectively, these data underscore a highly interdependent regulatory system among these three genes.

Given that Tfam, Polg, and Polg2 are involved in mtDNA maintenance, we performed qPCR to measure the abundance (DNA) of several mtDNA-encoded genes, including mt-Nd1 (NADH–ubiquinone oxidoreductase chain 1) and mt-Rnr2 (16S ribosomal RNA, mitochondrial), in KO livers. Analysis of hepatocytes isolated from Tfam KO and Polg KO mice between 2 and 10 wpi revealed steep declines in mtDNA levels (Figure 3C). Specifically, mtDNA levels decreased by ~60%–80% at 4 wpi and by >95% at 8 wpi (Figure 3C). Despite this dramatic loss of mtDNA, there were few gross manifestations until 8 wpi, when some animals began to exhibit weight loss (Figure 3D). Likewise, there was little evidence of laboratory abnormalities at 6 wpi, when KO animals had lost anywhere from 60% to 90% of mtDNA, aside from elevated bile acids in 1 Polg KO animal (Figure 3E). By 10 wpi, however, Polg KO and Tfam KO animals exhibited signs of hepatocellular death and cholestasis (Figure 3F). Liver injury was confirmed by H&E staining (Figure 3G), and transmission electron microscopy of 10 wpi Tfam KO livers revealed giant mitochondria with poorly developed cristae (Figure 3H). As expected, EGFP-negative escaper cells repopulated livers of Polg KO and Tfam KO animals, although this did not occur to a significant degree until 13 wpi, well after the manifestation of these phenotypes (Figure S8C). Surprisingly, Polg2 KO animals exhibited only slight/undetectable changes in body weight (Figure 3D), liver chemistries (Figure 3F), or histology (Figure 3G) at 10 wpi despite the nearly complete loss of mtDNA at that time point (Figure 3C). Additional follow-up revealed evidence of liver damage in Polg2 KO mice a week later, at 11 wpi (Figure S8D).

Given that reductions in Polg levels was readily and robustly observed in Tfam KO and Polg2 KO hepatocytes (Figure S8A), we sought to determine whether Polg loss is the common cause of liver injury in all liver knockout (LKO) animals. To this end, we exploited our system’s ability to target multiple genes simultaneously through the coinjection of multiple AAVs. Using this approach, we generated Polg/Polg2 double KO (DKO) and Tfam/Polg/Polg2 triple KO (TKO) mice. Polg/Polg2 DKO slightly exacerbated the weight loss at 10 wpi, whereas Tfam/Polg/Polg2 TKO mice (3 of 3) died at 9–10 wpi (Figure S8E). These results indicate an additive effect of simultaneous KO of these three genes, suggesting that both distinctive and redundant mechanisms underlie the liver injuries caused by their loss.

Given that manifestations of liver damage became apparent only 2 weeks after mtDNA levels dropped below 10%, we hypothesized that persistence of mitochondrially encoded RNAs might have allowed cells to survive in the face of mtDNA depletion. To test this, we performed a time course of mt-Rnr2 and mt-Nd1 RNA expression in the three mutants. Levels of mt-Rnr2 and mt-Nd1 RNA were maintained at 50% of control levels at 6 wpi in Polg KO and Tfam KO mice and at 8 wpi in Polg2 KO mice, consistent with the delayed kinetics of liver damage in these mice (Figure 3I). These results suggest that hepatocyte function and viability are temporarily preserved in cells lacking mtDNA through the persistence of mitochondrially encoded RNAs. Past proteomic studies have estimated that the median half-life of mitochondrial proteins in the liver is 4–5 days.^[33] Hence, levels of mitochondrially encoded proteins would be expected to undergo steep reductions between 8 and 10 wpi (when mRNA levels nadir), a timing that is consistent with the expansion of escaper (EGFP-negative) hepatocytes at 10–13 wpi (Figure S8C). These findings provide insight into the concept of “mitochondrial threshold effects,” wherein failure of the mitochondrial electron transport chain occurs once the frequency of mtDNA mutations exceeds a certain threshold, thought to be 60%–80%.^[34] Our data indicate that hepatocytes can tolerate a more severe and complete loss of mtDNA for several weeks, raising the possibility that the liver has a higher threshold for mtDNA loss compared to other tissues.

PTG: A multiplexed polycistronic sgRNA system

The ability to target multiple genes simultaneously (Figure S8E) could vastly simplify the time-consuming breeding necessary to generate KO mice carrying multiple mutant alleles. To confirm that such an approach could work, we coinjected AAVs carrying sgRNAs targeting EGFP and Hnf4a (Figure S9A, left). As predicted, this resulted in livers that were both steatotic and EGFP negative (Figure S9B). Next, we asked whether it would be possible to modify the system to incorporate multiple sgRNAs into a single viral vector. To do so, we introduced a polycistronic sgRNA expression platform (PTG) that uses the cell’s tRNA-processing machinery (Figure S9A, right; Figure S9C; see Materials and Methods for details). Mice injected with AAV carrying the PTG-EGFP/Hnf4a construct recapitulated features of AAV coinjection, including steatosis and loss of EGFP expression (Figure S9B). Reductions in Hnf4α levels following AAV-PTG-EGFP/Hnf4a injection were confirmed by WB (Figure S9D). Reductions in EGFP levels were confirmed by flow cytometry, although loss of signal (MFI) appeared to be slower with the polycistronic vector as compared to coinjection (Figure S9D,E). These findings suggest that multiplexed gene editing in vivo can be streamlined with polycistronic viral vectors.

We next sought to apply this system to another biological context. To this end, we chose to knock out large tumor-suppressor homologs 1 and 2 (Lats1/2), which regulate liver development by antagonizing the transcriptional coactivators, Yes-associated protein (YAP) and transcriptional coactivator with PDZ-binding motif (TAZ).^[35–37] We first confirmed that coinjection of AAV-Lats1-sg1/2 and AAV-Lats2-sg1/2, but not individual injection of Lats1-sg1/2 or Lats2-sg1/2, resulted in liver overgrowth at 3 wpi (Figure 4A–C, Figure S10A). In contrast to one earlier study, in which animals survived for 2 months after injection of adenovirus-Cre into Lats1^−/−; Lats2^fl/fl mice,^[35] no animals survived beyond 4 wpi in our AAV-coinjected Lats1/Lats2 DKO cohort (n = 6).

FIGURE 4.

RIME-mediated KO of Hippo pathway components Lats1/2 induces liver overgrowth and hepatobiliary reprogramming. (A) Schematic representation of the strategy of DKO of Lats1/2 by coinjection of four AAVs (left) and by injection of single AAV-PTG-Lats1/2 (right). (B) Representative macroscopic images of Lats1/Lats2 DKO at the designated time points. Percentages shown in photos indicate liver/body weight ratios. (C) Liver/body weight ratios at the designated time points are shown for Lats1/Lats2 DKO mice generated by either coinjection or PTG strategies. Mice harvested 2–6 weeks after injection of EV served as controls. (D) WB for validation of KO of Lats1 and Lats2 as well as the subsequent effects on Yap and Taz phosphorylation. (E) Bulk Sanger sequencing results for each of the cut sites targeted by two Lats2 sgRNAs. (F) Immunofluorescence images for Yap/Taz, EGFP, and Ki67 using livers harvested from mice injected with EV, Lats1-sg1/2 + Lats2-sg1/2 (coinjection), or PTG-Lats1/2 at the designated time points. Nuclei were counterstained with DAPI. Arrows indicate cells costaining for Ki67 and nuclear-localized Yap/Taz. Gapdh, glyceraldehyde 3-phosphate dehydrogenase; PAM, proto-spacer adjacent motif; pTaz, phosphorylated Taz.

Next, we designed a PTG-Lats1/2 vector that incorporated the two Lats1 sgRNAs and two Lats2 sgRNAs used in the coinjection experiment (Figure 4A, right, Figure S10B). In contrast to the coinjection cohort, the PTG-Lats1/2 cohort exhibited only modest liver overgrowth at 3 wpi (Figure 4B,C). By 5–6 wpi, however, liver overgrowth in the PTG-Lats1/2 cohort was comparable to that of the coinjection cohort at 3 wpi (Figure 4B,C). WB confirmed substantial reductions in Lats1 levels (Figure 4D), but because we were unable to reliably monitor Lats2 levels by WB (data not shown), we performed bulk Sanger sequencing to confirm mutations in Lats2 in coinjected mice at 2 wpi and PTG mice at 6 wpi (Figure 4E). Deep sequencing demonstrated that editing efficiency at the Lats1-sg2, Lats2-sg1, and Lats2-sg2 cut sites were lower in PTG cohorts at 2 and 3 wpi than the coinjection cohort, but increased to the level comparable to the coinjection cohort at 3 wpi (Figure S10C,D). Accordingly, levels of phosphorylated Yap (pYap) were diminished in the coinjected cohort at 2 wpi and in the PTG cohort at 6 wpi, whereas total YAP levels did not change significantly (Figure 4D). Unexpectedly, we observed a robust increase in the level of total Taz, but not total Yap, at 3 wpi in coinjected mice and 6 wpi in PTG mice (Figure 4D), suggesting distinct mechanisms regulating Yap and Taz stabilization by Lats1 and Lats2.

Finally, we compared the histological consequences of Lats1/2 inactivation in the liver using these two approaches. H&E staining as well as immunofluorescence staining with biliary markers revealed the emergence of hepatocyte-derived BEC-like cells in both the coinjection and PTG cohorts (Figure S11), as expected.^[38] In the control (EV) liver, Yap and Taz staining was limited to hepatocyte cytoplasm, and hepatocyte proliferation (measured by Ki67) was low (Figure 4F, left). In mice that received a coinjection of AAV-Lats1-sg1/2 and AAV-Lats2-sg1/2, by contrast, Yap and Taz were detectable in both the nucleus and cytoplasm, and hepatocyte proliferation was elevated (Figure 4F, middle). Similar effects were observed in the PTG-Lats1/2 cohort, although more normal regions were observed as compared to the coinjection cohort (Figure 4F, right). In summary, the PTG strategy permits mosaic inactivation of multiple genes in the mouse liver.

Indel patterns of hepatocytes in the homeostatic liver reflect NHEJ-dominated mutagenesis

Cas9-mediated double strand breaks are mainly repaired by c-NHEJ and MMEJ pathways, which can elicit a variety of indels. The overall spectrum of indels induced by Cas9 have been investigated intensively by in vitro studies.^[39] MMEJ repair accounts for most of the deletions with 2- to 30-bp sizes, amounting to 30%–70% of all indels,^[40–44] whereas NHEJ-derived 1-bp insertions account for 10%–25% of all events.^[39] Given that MMEJ requires end resection, which occurs only in G1 cells,^[45] the high MMEJ contribution rates observed in past studies are likely to have resulted from the proliferative characteristics of cells cultured in vitro. To gain insight into Cas9 editing modes in vivo, we summarized the indel profiles for the 20 sgRNAs of the nine genes assessed in the present study (Figure S12A), except for Tfam-sg2/sg3 cut sites, where editing was mostly caused by simultaneous deletion of the two cut sites. Although the editing pattern varied among sgRNAs, 1-bp insertions accounted for the majority of indels (50.8 ± 26.8%), indicating that NHEJ is the dominant mode of editing in hepatocytes in vivo. We observed that 3- to 30-bp deletion events accounted for 12.3 ± 8.3% of indels (Figure S12A), suggesting the contribution of MMEJ as well. Microhomology-mediated deletions (defined as ≥ 5-bp deletions with ≥ 2-bp-long homology) were overall low (1.89 ± 1.71%), but could be as high as >6% (Hnf4a-sg1, Polg2-sg2; Figure S12B,C), suggesting that the MMEJ pathway can be active even in quiescent hepatocytes.

Editing efficiency is dependent on chromatin accessibility

In our study, we noted variability in editing efficiency of the assessed cut sites across different genes. A recent in vitro study reported that the chromatin state of target sites affects the activity of sgRNA-guided Cas9 nuclease.^[44] We thus asked whether there is a correlation between editing efficiency and chromatin accessibility, using ATAC-seq data obtained from adult mouse hepatocytes. Specifically, we calculated the mean ATAC-seq read depth (a measure of chromatin accessibility) across the 500-bp flanking regions of each cut site and found a positive correlation between editing efficiency and chromatin accessibility (Figure S13A,B). These results suggest that chromatin accessibility influences the efficiency of genomic editing by CRISPR/Cas9 in vivo as it does in vitro.

DISCUSSION

This report describes RIME, a suite of tools that facilitate the inactivation of single or multiple genes in the adult mouse liver. To demonstrate the feasibility and use of this system, we knocked out eight genes with functions spanning metabolism, mitosis, mitochondrial maintenance, and cell proliferation. Earlier work has demonstrated the potential of CRISPR/Cas9-based LKO by adopting the hydrodynamic tail vein injection to deliver Cas9 and sgRNAs to hepatocytes.^[46–48] However, the low efficiency of gene transfer by hydrodynamic injection makes this approach impractical for whole-liver KO. To circumvent this obstacle, other investigators have taken the additional step of repopulating Fah KO host mice with gene-edited hepatocytes, in which whole-liver KOs were achieved after several months.^[46,48] The RIME approach described here uses off-the-shelf tools—an LSL-Cas9-EGFP mouse colony, tailored vectors, and a streamlined viral production protocol. This allowed most of our experiments to move from concept to realization within a month, with evidence of gene inactivation occurring 1–2 weeks after the injection of AAVs.

In addition to its speed, RIME provides other advantages. For example, the system contains a built-in lineage trace (the result of Cre-mediated expression of an EGFP reporter), a feature that offers a reinterpretation of Hnf4a’s role in the adult liver. In an earlier study,^[15] an increase in hepatocyte proliferation was observed following tamoxifen treatment of Alb-CreER^T2; Hnf4a^loxP/loxP mice, a result that was interpreted as indicating that Hnf4a has tumor-suppressive (cell-cycle–inhibitory) activity. However, this conclusion was based upon the assumption that tamoxifen treatment achieved Hnf4a KO in 100% of hepatocytes. With the benefit of lineage tracing, our experiments reveal reduced fitness of Hnf4-adeficient hepatocytes, resulting in their replacement by Hnf4a wild-type cells. Consequently, the increase in proliferation in Hnf4a-deficient livers is most likely attributable to the expansion of escaper cells rather than a cell-autonomous effect of Hnf4a.

A second advantage of the system is its ability to perform multigene KO knockout experiments. This can be achieved through one of two approaches: (1) simultaneous coinjection of multiple AAVs carrying individual gene-targeting sgRNAs or (2) use of a polycistronic (PTG) AAV viral construct containing up to four independent sgRNAs. Using the coinjection strategy, we were able to demonstrate gene interactions between Tfam, Polg, and Polg2 in mitochondrial maintenance and recapitulate features of Lats1/Lats2 DKO mice.^[38] Although the coinjection and PTG strategies gave similar findings, the efficiency of multigene editing by PTG was lower, likely attributable to the requirement for simultaneous processing of four sgRNAs from a single template. Thus, whereas there is room for optimization of the PTG strategy, the main advantage of the polycistronic system is the ability to knock out multiple genes within the same cell, which makes it useful for mosaic analysis.

In summary, RIME represents a fast, cost-effective, and straightforward means of assessing conditional gene function in vivo. Moreover, the ability to perform multiplexed gene targeting provides an opportunity to assess questions of genetic epistasis and gene redundancy. As the spectrum of AAV capsids with tissue-specific tropism continues to evolve, this approach can be extended to other organs and organisms to address a wide range of biological questions without a need to create conditionally targeted embryonic stem cells or perform multiple rounds of breeding.

Abbreviations:

AAVs

adeno-associated viruses

AAV8

adeno-associated virus serotype 8

ALT

alanine aminotransferase

ATAC-seq

assay for transposase-accessible chromatin with sequencing

BECs

biliary epithelial cells

Cas9

CRISPR-associated protein 9

cDNA

complementary DNA

DKO

double knockout

dpi

days postinjection

EGFP

enhanced green fluorescent protein

EV

empty vector

Fah

fumarylacetoacetate hydrolase

FBS

fetal bovine serum

HBSS

Hanks’ balanced salt solution

H&E

hematoxylin-eosin

Hnf4a

hepatocyte nuclear factor 4 alpha

Kif11

kinesin family member 11

KO

knockout

Lats1/2

large tumor-suppressor homologs 1 and 2

LKO

liver knockout

MDS

mtDNA depletion syndrome

MMEJ

microhomology-mediated end joining

mtDNA

mitochondrial DNA

mt-Nd1

NADH–ubiquinone oxidoreductase chain 1

mt-Rnr2

16S ribosomal RNA, mitochondrial

NIB

nuclear isolation buffer

Polg

polymerase gamma

PTG

polycistronic tRNA-gRNA

RIME

rapid in vivo multiplexed editing

RT

room temperature

sgRNA

single-guide RNA

TAZ

transcriptional coactivator with PDZ-binding motif

TBG

thyroid hormone-binding globulin

Tfam

transcription factor A, mitochondrial

WB

western blotting

wpi

weeks postinjection

YAP

Yes-associated protein