Kinetics of Nuclear Uptake and Site-Specific DNA Cleavage during CRISPR-Directed Gene Editing in Solid Tumor Cells

This article is featured in Highlights of This Issue, p. 799

Lung cancer is the leading cause of cancer-related death worldwide and accounts for an estimated 1.8 million newly diagnosed cases each year (1). Non–small cell lung carcinoma (NSCLC) makes up 85% of lung cancer cases and is further subtyped as adenocarcinoma, squamous cell carcinoma, and large-cell carcinoma. The 5-year survival rate is 18% despite advances in prevention, screening, and treatment (2). This high mortality rate is multifactorial; however, late stage diagnosis as well as resistance to treatment seems to have the most influence on poor patient outcomes. The first-line treatment for advanced or metastatic NSCLC includes platinum-based chemotherapy regimens and combinations with taxanes, as well as immunotherapy. Despite the various chemotherapy combinations available, a patient may become resistant to treatment and require two or three more lines of therapy (3, 4).

From a clinical perspective, the well-known toxicity of chemotherapy often remains a serious barrier to successful treatment and even enrollment in clinical trials (5–7). Even though CAR T-cell therapy is considered a breakthrough for many cancer types, the cost, the recently identified self-immunity, and even potential for biotherapeutic toxicity remain an impediment for widespread implementation (8–10). Currently, CAR T-cell therapy is not an option for NSCLC; thus, there is an unmet medical need for the development of innovative approaches for the treatment of lung cancer.

New developments in precision medicine show tremendous promise for the development of novel, personal therapeutics (11). More recently, Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR)–directed gene editing has already been approved for clinical trials by the FDA and similar organizations in Europe and China. Although the field of gene editing in mammalian cells has been active for almost two decades, the focus of the early work centered on the mechanism of action and circuitry that regulates single-base repair (12). Thus, the scientific foundation of gene editing has been well established (13–15). However, with the emergence of CRISPR/Cas, gene editing became democratized when the bacterial system was reengineered to work in human cells (16–20). The standard paradigm, “from bench to bedside,” has been accelerated because of the success and excitement that surrounds this technology. It is important to recognize that CRISPR-based therapeutic approaches have yet to be shown to be fully safe and lacking significant toxicity. Studies addressing these two key hurdles are now being carried out in both preclinical and phase I/II clinical trials. The Brilliance and the CTX001 clinical trial (21, 22) are among some of the current investigations that should yield important results.

We have begun to investigate the feasibility of using CRISPR-directed gene editing in combination with standard of care for the treatment of solid tumors, including NSCLC. We have taken a decidedly reductionist approach wherein we created a CRISPR/Cas9 gene editing tool to disable the Nuclear Factor Erythroid 2-Related Factor-Like (NRF2) gene. NRF2 is a master regulator of 100 to 200 target genes involved in cellular responses to oxidative/electrophilic stress (23, 24). NRF2 is also known to regulate the expression of genes involved in protein degradation and detoxification and is negatively regulated by Kelch-like ECH-associated protein 1 (KEAP1), a substrate adapter for the Cul3-dependent E3 ubiquitin ligase complex. Relevant to our overall objective, NRF2 expression increases significantly when a patient with NSCLC is treated with chemotherapy, through an activation of target genes that trigger the cyto-protective response (25–27). Several studies have demonstrated that increased NRF2 expression is one of the major contributing factors of chemoresistance in cancer cells, being more prevalent in NSCLC (24, 28–30).

Previously, we discovered that successful functional knockout (KO) of the NRF2 gene in chemo-resistant lung cancer cells significantly increased the anticancer activity of cisplatin, carboplatin, and vinorelbine in both cell culture and mouse models (31). Our results prompted further exploration of the feasibility of innovative therapeutic approaches aimed at reducing the clinical side effects of chemotherapy (11). This strategy centers on a combinatorial approach in which gene editing would disable the NRF2 gene removing its global regulation of chemoresistance.

Although CRISPR has already advanced into the clinical arena, fundamental molecular questions surrounding mode of action remain unanswered. First, is there a time frame correlation between the timing of nuclear penetration and the emergence of CRISPR-directed mutagenesis? And, second, is there enough sequence divergence between normal and tumor cell genomes so that CRISPR complexes can selectively cleave and disrupt the NRF2 gene in the tumor cell?

Regarding the first question, recent data suggest that a time lag exists between nuclear penetration and the appearance of gene editing activity in human cells (32). Therefore, we carried out a series of experiments to elucidate important reaction parameters that surround the coupling of nuclear penetration and gene editing activity. Our results indicate that the localization of the exogenously added CRISPR/Cas complex into the nucleus of the lung tumor cell is efficient and rapid; however, NRF2 gene disruption lags significantly, seen after about 8 hours as the gene editing components are visualized exiting the nucleus. With this data, we are able to create a snapshot of the time course and processing of CRISPR/Cas in a lung cancer cell.

Regarding the second question, Kerins and Ooi (33) cataloged somatic NRF2 mutations in various cancer cases reported in The Cancer Genome Atlas. They reported 214 cases of NRF2 mutations, predominately found within the Neh2 domain or known as the KEAP1-binding domain of the protein. These mutations disrupt KEAP1 binding and lead to constitutive expression of NRF2 in cancer cells. The most common NRF2 mutation noted, R34G, is of special interest because this mutation creates a new protospacer adjacent motif (PAM) for Cas9 recognition. The PAM consists of 2 to 6 nucleotides juxtaposed to the CRISPR seed binding sequence facilitating alignment and cleavage of the DNA. After cleavage, Cas9 creates blunt-end, double-stranded breaks which are then repaired by one of two mechanisms: nonhomologous end joining (NHEJ) or homology-directed repair (HDR). NHEJ is responsible for insertions or deletions (indel) at the cut site, generating genetic KO of the gene and potentially causing loss of protein expression. HDR is the mechanism by which genes can be repaired using gene editing if an appropriate donor DNA template is provided. We provide evidence that a unique PAM site regulates selective cleavage of a tumor cell genome, establishing a foundational basis for CRISPR-directed gene editing in lung cancer.

Human lung adenocarcinoma A549 cells (#CCL-185, passage numbers 1–12) and human lung squamous cell carcinoma NCI-H1703 (H1703) cells (#CRL-5889, passage numbers 1–12) were purchased from the ATCC. Cells were thawed, according to the manufacturer's protocol. A549 cells were grown in F-12K medium (ATCC) supplemented with 10% FBS (ATCC) and 1% Penicillin–Streptomycin Solution (ATCC). NCI-H1703 cells were grown in RPMI 1640 medium (ATCC) supplemented with 10% FBS (ATCC) and grown at 37°C in 5% CO₂. Both cell lines were tested for Mycoplasma upon thawing and before use in experiments using the MycoScope PCR Mycoplasma detection Kit (Genlantis, Cat. MY01100). The A549 cell line was previously characterized by the ATCC using short tandem repeat (STR) profiling, cellular morphology, and karyotyping. The H1703 cell line was previously characterized by the ATCC using STR profiling and cellular morphology. No further authentication was performed.

The NRF2 gene-coding sequence was entered into Benchling (https://benchling.com) and the gRNA 5′-TGGATTTGATTGACATACTT-3′ within exon 2 was selected. The GFP-labeled SpCas9 protein was a kind gift from Integrated DNA Technologies. Unlabeled Alt-R SpCas9, Alt-R tracrRNA ATTO 550, and designed Alt-R crRNA were purchased from Integrated DNA Technologies. RNA oligos (crRNA and tracrRNA) were mixed in equimolar concentrations to a final duplex concentration of 20 μmol/L, then annealed by heating the mix at 95°C for 5 minutes, and cooled to room temperature (15°C–25°C). Prior to mixing with cells, crRNA:tracrRNA duplex (20 μmol/L working solution) and spCas9 protein (62 μmol/L stock solution) were mixed together for a final complexed concentration of 20 pmol and set to incubate at room temperature for 15 minutes.

The NRF2 gene-coding sequence was entered into Benchling (https://benchling.com), and the R34G mutation–specific gRNA 5′-GATATAGATCTTGGAGTAAG-3′ and the nonspecific gRNA 5′-GTAACGGCAGACTTCTCCTC-3′ (HBB RNP) were selected. Alt-R SpCas9, Alt-R tracrRNA and designed Alt-R crRNA were purchased from Integrated DNA Technologies, and designed synthetic sgRNA was purchased from Synthego. RNA oligos (crRNA and tracrRNA) were mixed in equimolar concentrations to a final duplex concentration of 100 or 250 μmol/L, then annealed by heating the mix at 95°C for 5 minutes, and cooled to room temperature (15°C–25°C). Prior to mixing with cells, crRNA:tracrRNA duplex (100 μmol/L or 250 μmol/L) and spCas9 protein (100 μmol/L or 50 μmol/L) were mixed together and set to incubate at room temperature for 15 minutes. Final complexed concentrations were equimolar at 100 pmol or 5:1 ratio at 250:50 pmol of RNP.

The Lonza SF Cell Line 4D-Nucleofector X Kit (Lonza Inc.) was used for transfection of the RNP in the A549 and NCI-H1703 cell lines.

For all experiments, A549 cells were seeded 48 hours prior to transfection and allowed to reach 60% to 80% confluency. On the day of transfection, A549 cells were harvested by trypsinization and washed twice with 1x PBS (−/−). Cells were resuspended at a concentration of 3 × 10⁵ cells/20 μL in SF/supplement solution, and 5 μL of 20 pmol labeled RNP complex was added for each sample. Lonza program CM-130 was used, and after 15 minutes of rest, cells were transferred to a 6-well plate or a 1.5 mL tube for further analysis.

A549 RNP fluorescence was measured by BD FACSAria II (BD Biosciences). Cells were harvested by trypsinization, washed 3 times, and resuspended in 1x PBS (−/−) for FACS analysis. Several controls were used to establish gates based on each fluorescent component. The controls consisted of mock transfection of A549 cells, transfection of RNP with Cas9 (GFP), and transfection of RNP with tracrRNA (ATTO 550). Gates were established based on these parameters. After FACS analysis, cellular genomic DNA was isolated from each sample using the DNeasy Blood and Tissue Kit (Qiagen, Cat. 69506). The region surrounding the CRISPR target site was PCR amplified using Q5 High-Fidelity 2X Master Mix (New England BioLabs, Cat. M0492; 762 bp, forward primer 5′-CACCATCAACAGTGGCATAATGTGAA-3′, reverse primer 5′-AACTCAGGTTAGGTACTGAACTCATCA-3′). The PCR reaction was purified using the QIAquick PCR Purification Kit (Qiagen, Cat. 28106) and sent out to GENEWIZ, LLC for Sanger sequencing.

A549 and NCI-H1703 cells were seeded 48 hours prior to transfection and allowed to reach 60% to 80% confluency. On the day of the transfection, cells were harvested by trypsinization and washed twice with 1x PBS (−/−). Cells were resuspended at a concentration of 3 × 10⁵ cells/20 μL in SF/supplement solution, and 5 μL of RNP complex was added to each sample. For transfection efficiency by FACS analysis, the pmaxGFP vector (0.2 μg) was added to the respective samples before transfection. Lonza program CM-130 was used, and after 15 minutes of rest, cells were transferred to a 6-well plate for 48 hours.

Transfection efficiency of A549 and NCI-H1703 cells was measured by BD FACSAria II 48 hours after transfection. Cells were harvested by trypsinization, washed 3 times, and resuspended in 1x PBS (−/−) for FACS analysis. After FACS analysis, cellular genomic DNA was isolated from each sample using the DNeasy Blood and Tissue Kit. The region surrounding the CRISPR target site was PCR amplified using Q5 High-Fidelity 2X Master Mix (530 bp, forward primer 5′-CACCATCAACAGTGGCATAATGTGAA-3′, reverse primer 5′-AACTCAGGTTAGGTACTGAACTCATCA-3′). The PCR reaction was purified using the QIAquick PCR Purification Kit, Sanger sequenced using a SeqStudio Genetic Analyzer (Applied Biosystems), and analyzed by Tracking of Indel DEcomposition (TIDE; ref. 34).

A549 cells were transfected with unlabeled Cas9 RNP as described above. Cells were harvested at the indicated time points, and the Western blot was carried out as previously described (31). Primary antibody incubation was performed overnight on a shaker at 4°C for Cas9 (1:5,000, Abcam ab210752), and GAPDH (1:10,000, Abcam ab8245) and secondary antibody (Jackson ImmunoResearch Laboratories) incubations were all done 1 hour at room temperature at a 1:10,000 dilution. The protein bands were visualized via chemiluminescence using a SuperSignal West Dura Extended Duration ECL (Pierce) and detected on the LI-COR Odyssey FC.

Transfected A549 cells were seeded in 4-well chambers (LabTek II). Before imaging at each indicated time point, cells were washed twice with PBS (−/−) and incubated with Hoechst 33258 (1:10,000, Invitrogen, Cat. H3569) for a minimum of 15 minutes at room temperature in the dark. Chambers were imaged at 20x magnification using the DAPI (358 nm), GFP (488 nm), and DsRed (560 nm) channels on the Zeiss Axio fluorescent observer.Z1 microscope. Random fields were imaged, and images were processed on the AxioVision software. The brightness and contrast of the images have been enhanced for better visualization.

A NRF2 gene expression plasmid (pcDNA3-EGFP-C4-NRF2) was purchased from Addgene. The plasmid was mutated using CRISPR-directed mutagenesis (CDM) as described by Sansbury and colleagues (35). CDM was achieved using two Cas12a cleavage sites (crRNA: 5′-ATTGACATACTTTGGAGGCAA-3′, 5′-GAGTTGTTCTTGTCTTTCCTT-3′), and two oligonucleotides, containing the intended R34G mutation (ssDNA: 5′-CAAGATATAGATCTTGGAGTAAGTGGAGAAGTATTTGACTTCAGTCAGCGACGGAAAGAGTATGAGCTGGAAAAACAGAAAAAACTTG-3′, 5′-CTTTTCAAGTTTTTTCTGTTTTTCCAGCTCATACTCTTTCCGTCGCTGACTGAAGTCAAATACTTCTCCACTTACTCCAAGATCTATA-3′), were annealed to create the template. Plasmids were amplified using Q5 High-Fidelity 2X Master Mix (901 bp, forward primer 5′-CATGGTCCTGCTGGAGTTCGTG-3′, reverse primer 5′-GAGGATGCTGCTGAAGGAATCCTC-3′).

A549 (parental and clonal) and NCI-H1703 cell lines were harvested, and genomic DNA was isolated using the DNeasy Blood and Tissue Kit. NRF2 exon 2 was amplified using Q5 High-Fidelity 2X Master Mix (530 bp, FWD 5′-CACCATCAACAGTGGCATAATGTGAA-3′, REV 5′-AACTCAGGTTAGGTACTGAACTCATCA-3′), and amplicons were purified using the QIAquick PCR Purification Kit. The gRNAs used for the in vitro cleavage assays were R34G gRNA (5′-GATATAGATCTTGGAGTAAG-3′) and HBB gRNA (5′-GTAACGGCAGACTTCTCCTC-3′).

RNA oligos (crRNA and tracrRNA) were mixed in equimolar concentrations to a final duplex concentration of 10 μmol/L, then annealed by heating the mixture at 95°C for 5 minutes, and cooled to room temperature (15°C–25°C). The crRNA:tracrRNA duplex (10 μmol/L) and spCas9 protein (100 μmol/L) were complexed together in PBS (−/−) to a final concentration of 1 μmol/L and incubated at room temperature for 10 minutes.

The reaction was assembled using 10X NEBuffer 3.1 (New England BioLabs, cat. B7203S; 1X final concentration), 1 μmol/L Cas9 RNP (molar concentrations were adjusted for each reaction), 100 ng DNA substrate, and nuclease-Free water. The reaction was incubated at 37°C for 60 minutes. The reaction was stopped by adding 1 μL Proteinase K (20 mg/mL) to the mixture at 56°C for 10 minutes and visualized by gel electrophoresis.

We sought to examine the temporal relationship among cellular penetration, nuclear transit, and gene editing activity vis-à-vis indel formation on the clinically relevant gene NRF2. To do so, we developed a CRISPR-directed gene editing strategy to target within the Neh2 domain of the NRF2 protein, encoded by exon 2 (Fig. 1A), using a dual fluorescently tagged CRISPR/Cas9 ribonucleoprotein (RNP) complex; the tracrRNA contained an ATTO 550 label and purified Cas9 protein was tagged with GFP (Fig. 1B). After cell preparation and RNP assembly, A549 cells were transfected by Nucleofection with the dual labeled particle. After a designated period of time, we examined the targeted cell population by: (1) FACS for single and dual transfection efficiency of the ribonucleoprotein components (Fig. 1C), (2) fluorescent microscopy to visualize cellular and nuclear uptake, (3) Western blot analysis for the presence and sustainability of Cas9 protein, and (4) gene editing activity on the NRF2 gene using Sanger sequencing and the TIDE program (34).

Fluorescence-activated cell sorting (FACS) enables analyses of the transfection efficiency and cellular uptake through visualization of all four quadrants for colocalization of the ATTO 550 and GFP fluorescent tags respectively (Fig. 2). The control data addressing compensation and gating position for FACS are shown in Supplementary Fig. S1. The red area (P1) of each left-hand graph represents the collected cell population analyzed for fluorescence intensity, shown in the right-hand graph. The Y axis (FITC channel) represents GFP fluorescence intensity, and the X axis (PE channel) represents ATTO 550 fluorescence intensity. Quadrant 1 (Q1) contains cells positive for only GFP (Cas9), whereas quadrant 2 (Q2) reflects cells containing both GFP and ATTO 550 (tracrRNA) fluorescence (outlined in green). Population 2 (P2), highlighted in green, was gated to capture GFP-positive cells that may have been omitted with standard gating (data not shown). Quadrant 3 (Q3) contains cells that do not exceed background fluorescence, and quadrant 4 (Q4) contains cells positive for only ATTO 550 (tracrRNA). The table under each plot contains raw data from each quadrant. The fluorescence (%) found in quadrant 2 is of interest because this population represents cells that contain both fluorescent tags, presumably a colocalization of the two components of the RNP complex.

In each FACS plot of Fig. 2, quadrant 1 reveals almost no GFP fluorescence indicating that Cas9 is likely complexed with tracrRNA after introduction into the cell. Supplemental Data (Supplementary Fig. S1) confirm that the Cas9 (GFP) molecule is detectable in cells in quadrant 1 when transfected in the absence of other fluorescent components. GFP-fluorescing cells slowly disappear from all quadrants over 8 hours, indicating degradation through loss of signal. Similarly, ATTO 550-fluorescing cells are observed at high levels in quadrant 4 after 1 hour and gradually diminish over the next 24 to 48 hours (see Supplementary Fig. S1). After 12 hours, total fluorescence (the sum of quadrants 1, 2, and 4) declines precipitously (Fig. 2) correlating with the images provided by fluorescent microscopy. Taken together, we are observing the decomposition of the CRISPR/Cas9 complex over time, with Cas9 disappearing first followed by the tracrRNA.

In correlation with the FACS time course, we carried out gene editing analysis at the same time points. Exon 2 was amplified, Sanger sequenced, and analyzed by TIDE to quantify indel formation within the total, unsorted population (Fig. 3). Indel formation appears to rise slowly, but steadily, throughout the course of the reaction with significant levels visible only after 8 hours, coincident with the increase in fluorescent signal of quadrant 4 of the FACS plots (Fig. 2).

Next, we imaged nuclear entry, colocalization, and exit of the RNP components by fluorescent microscopy (Fig. 4A). Cells were stained with Hoechst 33258 to visualize the nucleus using the DAPI channel, whereas fluorescent CRISPR complex components were visualized using GFP and DsRed channels respectively. Brightness and contrast of the images are enhanced for better visualization of the fluorescent RNP components (see also Supplementary Fig. S2). The tracrRNA (ATTO 550) localizes rapidly into the nucleus, whereas Cas9 (GFP) appears punctate at the nuclear membrane, merging or coalescing into the nucleus after four hours. Between 1 and 8 hours, tracrRNA (ATTO 550) fluorescence appears throughout the nucleus and cytoplasm, and after 8 hours, both Cas9 and tracrRNA appear to exit the cell. At 24 hours, Cas9 (GFP) is barely visible apart from the punctate remnants surrounding the nucleus, and only small traces of tracrRNA (ATTO 550) remain visible in the nucleus. After 48 hours, both fluorescent components are barely detectable even with enhancement.

In Fig. 4B, we present a graphical representation of the individual and dual fluorescence obtained from FACS analysis. Combining FACS analyses and indel formation data, the reaction could be generally broken down into two phases. Phase I encompasses RNP delivery and nuclear uptake along with the initial signals of CRISPR-directed DNA cleavage and indel formation through NHEJ of a single target site; phase II reveals elevated levels of indel formation likely through resection and repair of the target site after cleavage, and the exit of the two fluorescent RNP components from the cell. Figure 4C represents Cas9 protein at each time point through Western blot analysis using an antibody directed against SpCas9. In our hands, the majority of Cas9 protein appears in the first 4 hours after introduction into the cell and becomes undetectable somewhere between 4 and 8 hours after transfection.

The total number of cells with both fluorescent signals in quadrant 2 is seen to decrease over time with the maximal level observed at 1 hour after nucleofection; a significant drop is seen between 4 and 8 hours. The totality of fluorescent cells in quadrant 2 and quadrant 4 remains consistent throughout most of the reaction time (until 12 hours), indicating that the cells receiving the RNP gradually lose GFP fluorescence first, while retaining ATTO 550 fluorescence for a longer period. These results suggest both labels are inside the targeted cell population for approximately 1 to 4 hours. The presence of intact Cas9 protein in the cell, as viewed by Western blot, confirms this time frame (see Fig. 4C).

A second question to support the concept of using CRISPR-directed gene editing for lung cancer centers on the selective activity of programmable nucleases on tumor cell genomes, avoiding similar activity on normal cell genomes. Due to the mechanism of CRISPR/Cas cleavage, a PAM sequence is required and can be selected through copious design of the crRNA seed sequence. As described above, tumor cell–specific sequence variance has already been identified in several lung cancer genomes. The R34G mutation was initially found because it inhibits KEAP1-mediated degradation of NRF2 which in turn causes increased NRF2 stability and nuclear accumulation (33, 36). This mutation occurs as a result of a base change from cytosine to guanine in the first base of codon 34 in NRF2, thereby creating a unique Cas9 PAM site (Fig. 5A).

As proof of concept, the following experiments were conducted to assess the level of discrimination and selectivity of the R34G mutation–specific CRISPR/Cas9 complex. An in vitro cleavage assay was employed to assess the discriminatory cleavage activities of this CRISPR/Cas complex. The RNP was incubated with varying DNA templates, and the cleavage products were visualized by gel electrophoresis. For this experiment, wild-type and R34G-mutated NRF2 PCR products (901 bp) from an NRF2 expression plasmid (Fig. 5B) were incubated with R34G-specific RNP (1 μmol/L). Cleavage products were visualized by gel electrophoresis, and the results are shown in Fig. 5C. A nonspecific RNP (HBB RNP) was also used to confirm specificity and selectivity of the new PAM site. In lane 6 (R34G NRF2–R34G RNP), two new bands (222 and 679 bp) appear and are the result of tumor genome–specific RNP cleavage; the R34G-targeting RNP only recognizes and cleaves the mutated DNA sequence.

To assess specificity and fidelity of the R34G-mutation–specific RNP in lung tumor cells, we use the well-established non–small cell lung adenocarcinoma cell line A549 because it is often used as a gold standard for the discovery of anticancer therapeutics. We genetically engineered the A549 cell line to contain the R34G mutation in exon 2 of the NRF2 gene using previously established methods (37). A secondary cell line derived from a lung squamous cell carcinoma, NCI-H1703, was also selected to ensure reproducibility of the cell-based experiments. The parental A549 cell line, R34G-mutated A549 cell line (A549 R34G-6), and NCI-H1703 cell line were sequenced across exon 2 to verify the sequence of the Neh2 domain. Both A549 and NCI-H1703 parental cell lines contain no mutations within exon 2 of the NRF2 gene, whereas the A549 R34G-6 cell line contains a heterozygous R34G mutation (Fig. 6A). The A549 cell line is known to be bi- and tri-allelic; therefore, the R34G mutation in the A549 R34G-6 clone is believed to be on two of three alleles as seen by the double peak of the “G” nucleotide (black peak) as compared with the peak of the “C” nucleotide (blue peak), indicated by the red arrow under “A549 R34G-6” (Fig. 6A).

Genomic DNA from each cell line was used to amplify exon 2 of the NRF2 gene. The PCR products (530 bp) were then used in in vitro cleavage reactions where expected cleavage products would appear at 385 and 145 bp and uncut product would appear at 530 bp, indicated by the red lines along the ladder (Fig. 6B). The R34G-targeting RNP was used at two concentrations (1 and 5 pmol) with each sample (lanes 2, 3, 5, 6, 8, and 9). Minimal cleavage is seen with wild-type NRF2 sequence in the A549 and NCI-H1703 parental cell lines (lanes 3 and 6), whereas cleavage products (145 and 385 bp) are abundantly clear in the A549 R34G-6 cell line with the R34G-mutated NRF2 sequence (lanes 8 and 9). These data further confirm the specificity and efficacy of the R34G-targeting RNP, only recognizing the target DNA in the presence of the R34G mutation in exon 2 of NRF2. Minor residual cleavage of the wild-type NRF2 sequence can be explained by the nature of the cleavage reaction and sequence homology of the guide RNA to the target DNA as the cleavage reaction includes incubating target DNA with RNP for an hour. Cleavage specificity is further confirmed through dose-dependent cleavage reactions with a minimum of half the standard concentration and a maximum of 5 times the standard concentration of RNP. The range of concentrations does not result in cleavage of wild-type NRF2 sequence whereas with a R34G-mutated NRF2 sequence, cleavage products can be seen even with half of the standard concentration of the R34G-targeting RNP (Supplementary Fig. S3).

Finally, we evaluated selective CRISPR-directed gene editing at the NRF2 locus and measured indel formation in three separate cell lines (A549 parental, A549 R34G-6, NCI-H1703 parental), each of which is predicted to produce a different outcome. Figure 7 depicts the genetic analysis of each cell line after transfection of the R34G-targeting RNP under three different conditions: (1) equimolar concentration of duplexed guide RNA to Cas9 (top panel); (2) five times the amount of duplexed guide RNA to Cas9 (middle panel); (3) five times the amount of single guide RNA to Cas9 (bottom panel). The pMAX GFP vector, supplied by Lonza, was used as a control for transfection efficiency. TIDE analysis was employed to determine total indel formation efficiency. Wild-type or uncut sequences appear at the 0 tick mark as a pink bar and would indicate alleles unaffected by the cleavage activity of the R34G-specific RNP.

Figure 7A depicts the genetic analysis after transfection of the R34G-targeting RNP in the A549 parental cells. The A549 parental cells contain a wild-type NRF2 gene; therefore, there is no native CRISPR/Cas9 recognition site for the R34G-targeting RNP. Genomic DNA from the transfected cells was analyzed by TIDE analyses (shown in Fig. 7A). The top panel displays indel efficiency and sequence alignment of an unsorted population. According to TIDE analysis, there is a total indel efficiency of 11.8% and statistically insignificant insertions and deletions (black bars—P value > 0.001). The middle panel shows a total indel efficiency of 4.3% from a GFP-positive sorted population with only statistically insignificant indels. The bottom panel shows a total indel efficiency of 11.4% from a GFP-positive sorted population with one statistically significant insertion (+1 bp).

Figure 7B depicts the genetic analysis after transfection of the R34G-targeting RNP in the A549 R34G-mutated clonal cell line (A549 R34G-6). The A549 R34G-6 cells contain a heterozygous R34G mutation in exon 2 of the NRF2 gene, creating a new CRISPR/Cas9 recognition site for the R34G-targeting RNP. Genomic DNA from transfected cells was analyzed for gene editing activity by TIDE analyses (shown in Fig. 7B). The top panel shows indel efficiency and sequence alignment of an unsorted population with a total indel efficiency of 43.0% and statistically significant insertions (+1, 30 bp) and deletions (−9, 41 bp) along with several insignificant deletions. The middle panel shows a total indel efficiency of 55.7% from a GFP-positive sorted population with statistically significant insertions (+1, 30 bp) and deletions (−1, 9 bp) and insignificant deletions. The bottom panel shows a total indel efficiency of 32.7% with significant insertions (+1, 30 bp) and deletions (−2 bp) and insignificant indels from a GFP-positive sorted population. The lower R² value indicates some of the sequence analysis is unaccounted for likely because of sequence quality.

Figure 7C depicts the genetic analysis after transfection of the R34G-targeting RNP in the NCI-H1703 cell line, which contains a wild-type NRF2 gene and no native CRISPR recognition site. Genomic DNA from transfected cells was analyzed for gene editing activity by TIDE analyses (shown in Fig. 7C). The top panel shows total indel efficiency of 0.0% from an unsorted population. The bottom panel shows total indel efficiency of 4.6% with only insignificant indels from a GFP-sorted population.

Taken together, these data confirm the highly specific activity of the R34G RNP in the presence of the R34G mutation. We can infer that the R34G mutation does in fact create a new CRISPR/Cas9 recognition site that is actively cleaved when tested in vivo. The R34G-targeting RNP differentiates between the wild-type and R34G-mutated NRF2 alleles within the A549 R34G-6 clonal cell line. This is seen by the remaining wild-type sequence (displayed at the 0-tick mark) in each panel of Fig. 7B, which approximately equates to the percentage of wild-type NRF2 allele present in the cells.

Clinical translation of CRISPR-directed gene editing is occurring at an unprecedented pace, and there is great hope for therapeutic breakthroughs (38, 39). And, although the technology appears to be democratized, there are several important questions that need to be addressed to ensure safety and success. We have taken a decidedly reductionist approach centering on some foundational molecular interactions that likely regulate the gene editing process. First, we begin to define the time required for CRISPR/Cas complexes to enter the cell, localize in the nucleus, and execute gene disruption. Second, we evaluate a new experimental strategy wherein CRISPR/Cas is designed to catalyze cleavage selectively on tumor cell genomes. This parameter is at the heart of the conundrum that has faced gene therapy applications in general: can a specific genetic disruption be made solely in tumor cell genomes while leaving the normal cell genomes genetically unaltered? While we are establishing a molecular foundation for CRISPR-directed gene editing for future clinical application, safety, control of expression in nontumor tissue, and a thorough evaluation of preclinical response data must be clearly addressed.

We analyzed the efficiency of cellular transfection of an RNP particle bearing differential fluorescent labels and assessed the emergence of indels. We imaged the nuclear transit and positioning of the tracrRNA and Cas9 as a complex as they relate to the timing of indel formation. Our results indicate that gene editing of the NRF2 gene in A549 cells may take place in two experimentally segmented phases. Cas9 levels diminish significantly between 4 and 8 hours correlating with their exit from the nucleus. In contrast, the fluorescently labeled tracrRNA transits into the nucleus rapidly and remains visible for up to 8 hours. Both fluorescent labels evolve as punctate particles on the nuclear membrane at 12 hours prior to being diluted and rendered undetectable between 24 and 48 hours.

The initialization of genetic disruption through gene editing begins with a CRISPR/Cas complex aligning in homologous register with a target gene, facilitating site-specific double-stranded DNA breakage (40–43). Double stranded DNA breaks are most often repaired through a process known as NHEJ, an imperfect and unfaithful process in which nucleotides are lost or gained, generating functional KO. Contrasting data centered around the question of how long it takes for double-strand breaks to be created and repaired has been published with bulk measurements consistently showing that double-strand breaks have a half-life of 10 to 60 minutes (32, 44–46). However, these measurements lack the sensitivity that is required to follow double-strand breaks occurring and being repaired at a single locus, the reaction mechanics of CRISPR/Cas9. A crude estimate from Kim and colleagues (47) suggested that indel formation by CRISPR/Cas occurs early in the process but is detected only after about 15 hours. These data align with the belief that sequence-specific nucleases direct indel formation only after multiple cycles of breakage and perfect repair (32). Our data suggest that the time delay between nuclear localization and detectable gene editing activity centers around 4 to 8 hours (48).

Multiple pathways contribute to the repair of a Cas9-induced double-strand break (49), and relative activity of each is dependent on the locus, the metabolic state of the cell, and the half-life of the bound CRISPR complex. Indel formation likely emerges after multiple rounds of cleavage and repair once the cell has extinguished its energy supply which then causes error-prone repair of the double-strand breaks. It is possible that NHEJ initially prevents access of the Microhomology-mediated End Joining (MMEJ) pathway proteins to the double-strand break. Only after several hours wherein NHEJ has failed to repair the break, does the MMEJ pathway engage. We detect gene editing activity several hours after the RNP particle has successfully been transduced to the nucleus, a time frame that aligns with the reaction mechanics of NHEJ and MMEJ. Our data suggest an inverse relationship among important steps in the gene editing reaction: cellular uptake, nuclear localization of multiple components, indel formation, and particle and/or component exit.

We also evaluated a strategy that exploits a DNA change that creates a unique PAM site in the NRF2 gene. A NRF2 gene expression plasmid was used as a template to generate PCR products that we showed were specifically and uniquely cleaved by the R34G RNP, in the presence of the R34G mutation. We followed up with a series of experiments that utilized a genetically modified cell line containing the R34G mutation as a new PAM site to create an in vivo PCR product that was subsequently cleaved again specifically by the R34G RNP. Finally, we targeted several cell lines directly using the R34G RNP and showed by TIDE analysis that efficient and highly specific indel formation is observed in a cell line that bears the R34G PAM site. Thus, by multiple iterations of the same basic experiment, we conclude that the R34G RNP demonstrates a high level of specificity of NRF2 gene disruption in an efficacious and efficient fashion. We believe that we have begun to establish a foundational basis for moving such a strategy forward to more complex targets in Xenograft mouse models.

NRF2 mutations occur in about 15% of lung squamous cell carcinomas, among other common cancer types (33). Disrupted NRF2 protein accumulates in the cytoplasm and nucleus, causing these cells to become resistant to toxic insults such as platinum-based drugs or radiation. Ultimately, this allows the cells to continue to grow and proliferate regardless of the environment. Despite the prevalence of mutations and the dismal prognosis of patients with elevated NRF2 expression, there is no targeted therapy. Clinically, we envision a unique approach to target NRF2 mutations using the gene editing tool, CRISPR/Cas9. This approach would also minimize the risk of off-target effects and improve delivery method. These mutations have been noted in other cancer types as well which makes this approach a platform therapy.

The CRISPR-directed gene editing strategy could evolve as an effective augmentative therapy to approach patients with locally advanced lung cancer, especially where the current standard of care has failed and there may be resistance to standard chemotherapy. We recognize that enhancing sensitivity to chemotherapy in such patients could only provide an improvement in the quality of life for a limited amount of time. However, in some cases, current and alternative therapies are employed often to achieve the same end point (50). As such, we remain realistic and only cautiously optimistic. Ultimately, if this proves to be an effective strategy, it could be incorporated into an earlier stage including those patients treated adjunctively following surgical resection, or concurrently in combination with radiotherapy. By destroying the NRF2 gene in a tumor cell population, the source of chemo-resistant cells is immaterial, because CRISPR-directed gene editing is active on tumor cells regardless of origin. A differential PAM site exists in tumors from esophageal, head and neck, uterine-endometrial, and bladder-urothelial cancers which might provide a path to exploit tumor cell sequence variance as a method of selection.

No potential conflicts of interest were disclosed.

Conception and design: K. Banas, B.-C. Yoo, E.B. Kmiec

Development of methodology: K. Banas, N. Rivera-Torres, P. Bialk

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. Banas

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Banas, N. Rivera-Torres, P. Bialk, B.-C. Yoo, E.B. Kmiec

Writing, review, and/or revision of the manuscript: K. Banas, P. Bialk, E.B. Kmiec

We thank the members of the Kmiec laboratory for input and advice. We thank Dr. Lynn Opdenaker at the CTCR Flow Cytometry Core Facility for running, analyzing, and sorting cells. This project was supported by grants from the National Institute of General Medical Sciences; P20 748GM103446 and P20GM109021, and from the Ammon Foundation. This content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

This article is dedicated to the Lung Cancer Working Group from the Helen F. Graham Cancer Center and Research Institute that was established in 2017 to help formulate the molecular basis of CRISPR-directed gene editing in solid tumors. This group includes Drs. Nicholas J. Petrelli, GregMasters, Michael Guarino, Christopher Koprowski, Jamil Katri, Jerry Castellano, and Pat Swanson.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.