Skip to main content

Thank you for visiting You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Safety and feasibility of CRISPR-edited T cells in patients with refractory non-small-cell lung cancer

A Publisher Correction to this article was published on 18 June 2020

This article has been updated


Clustered regularly interspaced short palindromic repeats (CRISPR)–Cas9 editing of immune checkpoint genes could improve the efficacy of T cell therapy, but the first necessary undertaking is to understand the safety and feasibility. Here, we report results from a first-in-human phase I clinical trial of CRISPR–Cas9 PD-1-edited T cells in patients with advanced non-small-cell lung cancer ( NCT02793856). Primary endpoints were safety and feasibility, and the secondary endpoint was efficacy. The exploratory objectives included tracking of edited T cells. All prespecified endpoints were met. PD-1-edited T cells were manufactured ex vivo by cotransfection using electroporation of Cas9 and single guide RNA plasmids. A total of 22 patients were enrolled; 17 had sufficient edited T cells for infusion, and 12 were able to receive treatment. All treatment-related adverse events were grade 1/2. Edited T cells were detectable in peripheral blood after infusion. The median progression-free survival was 7.7 weeks (95% confidence interval, 6.9 to 8.5 weeks) and median overall survival was 42.6 weeks (95% confidence interval, 10.3–74.9 weeks). The median mutation frequency of off-target events was 0.05% (range, 0–0.25%) at 18 candidate sites by next generation sequencing. We conclude that clinical application of CRISPR–Cas9 gene-edited T cells is generally safe and feasible. Future trials should use superior gene editing approaches to improve therapeutic efficacy.


CRISPR–Cas9 technology offers a convenient, flexible and precise method for genome editing1,2. Cas9, typically Streptococcus pyogenes Cas9 (SpCas9)3, recognizes a protospacer-adjacent motif (PAM) upstream or downstream of the target sequence by using a guide RNA and induces double-stranded breaks in the target DNA. The clinical application of CRISPR editing has been unclear, as CRISPR–Cas9 could cause genotoxicity including off-target genome cleavage sites, chromosomal translocations or other complex changes, which may lead to unpredictable consequences4. Recent pilot clinical trials using CRISPR-edited stem cells and multigene-edited T cells in the treatment of HIV/AIDS and cancer, including multiple myeloma and sarcoma, support the safety and feasibility of CRISPR–Cas9 technology in the clinic5,6. Anti-PD-1 checkpoint inhibitors such as pembrolizumab have become the standard first-line therapy for advanced non-small-cell lung cancer (NSCLC) with PD-L1 expression7,8. Some patients may attain 5-yr survival rates of 15.5–23% with single-agent pembrolizumab9. We hypothesized that disruption of the PD-1 gene in T cells followed by ex vivo reinfusion of the gene-edited cells could be therapeutic. While specific data on lung cancer are lacking, preclinical data have shown that editing of PD-1 in tumor-infiltrating lymphocytes by zinc finger nuclease improves T cell cytotoxicity in melanoma, and that T cells edited via CRISPR–Cas9 show enhanced anti-tumor responses against gastric cancer10,11. As lung cancer is the most prevalent and fatal malignancy in China and worldwide12,13, we chose to focus on treatment of this disease.

Being conscious of the potential ethical implications and risk of off-target mutations, we enrolled patients with NSCLC who failed multiple lines of therapy in a dose-escalating phase I clinical trial. The primary objectives were to demonstrate the feasibility of gene-edited T cell therapy and to evaluate safety. The gene-edited cells were expanded ex vivo and re-infused as therapeutic T cells. To monitor the risk of off-target mutations, we performed whole-genome sequencing (WGS) and next generation sequencing (NGS) on genomic DNA isolated from edited T cells before infusion. In addition, we evaluated in vivo tracking of edited T cells, peripheral T cell receptor (TCR) clone diversity and unique TCR clones in peripheral blood mononuclear cells (PBMCs) during and following therapy.


Patients and treatment

A total of 22 patients were enrolled from 26 August 2016 to 21 March 2018 (Fig. 1). Five patients did not ultimately receive infusions because of insufficient expansion of T cells. We succeeded in manufacturing sufficient and high-viability (>90%) edited T cells (success rate, 77.3%, 17 of 22) for the remaining 17 patients. Five of 17 patients were excluded before T cell infusion for disease progression (two), consent withdrawal (one), pulmonary infection (one) and exacerbation of rheumatoid arthritis (one; Fig. 1a). The median time to prepare the T cells for reinfusion based on 17 patients was 25 d (range, 17–40 d). The median number of T cells for each infusion per patient was 1.33 × 109 (range, 0.48 × 109 to 2.38 × 109).

Fig. 1: Schematic showing patient flow and study design, including patient enrollment, planned dose cohorts, interventions, follow-up and outcome assessments.

a, A Consolidated Standards of Reporting Trials (CONSORT) diagram showing patient flow through the main stages in the phase I dose-escalation trial. b, Study design. The PD-1 gene-edited T cells were infused every 28 d (cycle) as three infusions, the first containing 20% of the indicated total cell number; the second, 30%; and the third, 50%. PD, progressive disease; TPS, tumor proportion score; ECOG PS, Eastern Cooperative Oncology Group performance status; CTX, cyclophosphamide (20 mg kg–1). Asterisks show that the patient could not be evaluated for treatment efficacy owing to early withdrawal from the trial due to a bacterial infection that caused exacerbation of pulmonary disease after the first treatment cycle.

Preparation and evaluation of edited T cells

A pair of single guide RNAs (sgRNAs; sgRNA1 and sgRNA2) was selected to target exon 2 of the PD-1 gene (Fig. 2a). The Cas9 and sgRNA plasmids were cotransfected into T cells by electroporation14, which was established as the standard procedure for this trial in 2015. Since the launch of our trial, the superior CRISPR–Cas9 ribonucleoprotein (RNP) delivery platform has been developed15. However, to maintain consistency with our preliminary data on quality and stability of cellular therapy, we chose to continue with the plasmid system. Editing efficiency (median, 20.1%; range, 8.7–31.2%) was detectable by T7E1 cleavage surveyor assay (Extended Data Fig. 1a). Flow cytometric analysis revealed markedly decreased PD-1 expression, with median disruption of 46.3% (range, 33.7–67.0%) in edited T cells (Fig. 2b). TA-clone sequencing showed that the median PD-1 gene-editing efficiency was 16% (range, 8–34%) in the first cycle (Extended Data Fig. 1b). NGS targeted sequencing of the PD-1 gene using two rounds of PCR amplification was performed for further validation. The median editing efficiency of all 12 patients was 5.81% (range, 0.42–24.85%) (Fig. 2c). In edited T cells, most cells were CD3 positive (median, 99.1%; range, 95.9–99.6%), with CD3CD8 double-positive T cells accounting for 73.5% (range, 38.5–93.0%) (Supplementary Table 1). A significant increase in the proportion of CD8+IFN-γ+ cells was found in edited T cells compared with unedited cells (P < 0.0001) (Fig. 2d). Long-term culture experiments showed a dramatically higher capacity for stable viability (>90%) and expansion number in edited T cells (versus unedited cells) over 30–40 d from patients B-01 and C-02 (Extended Data Fig. 2).

Fig. 2: CRISPR–Cas9-mediated PD-1 gene editing in T cells.

a, The pair of sgRNAs (sgRNA1 and sgRNA2) that was designed to target exon 2 of PD-1. The PAM sequences are marked with an underscore. The Cas9 and sgRNA plasmids were cotransfected into T cells by electroporation. b, The levels of PD-1 expression on edited T cells or unedited T cells in the first cycle were measured by flow cytometry. c, The frequency of PD-1 editing efficiency on all infused T cells was measured by NGS. Box-and-whisker plot shows the frequency of PD-1 editing efficiency; the box represents the values from lower quartile to upper quartile, the center line in the box represents the median data and the whiskers (vertical) lines outside the box represent the minimum and maximum values. Each dot represents the editing efficiency of infused cells in an individual cycle. n represents the number of tested samples (patient ID); n = 1 (pre-A-02, A-02, A-03, B-03 and C-01), n = 2 (A-01, A-04, B-02 and C-03), n = 3 (pre-A-01), n = 4 (C-02), n = 9 (B-01). d, The percentages of CD8+ IFN-γ+ cells in edited or unedited T cells on the first cycle were measured by flow cytometry (left panel), and analyzed by two-tailed paired t-test (right panel, n = 12 per group, normally distributed data, each dot represents individual data from each patient, center line and error bar represent mean ± s.d., mean of differences = 5.896, 95% CI of difference, 4.602–7.189, degrees of freedom = 11, P < 0.0001). Data are representative of three independent experiments. Chr., chromosome.

Source data

To study the off-target impact of CRISPR–Cas9 on edited T cell genomes, we used sgRNAcas9 (v.2.0.6)16 to predict the 18 potential off-target (OT) sites (labeled as OT1–18, representing sites 1–18), which would indicate the top candidates for off-target activity (based on sequence homology to the PD-1-targeting guide RNA) (Supplementary Tables 2 and 3). We performed NGS with cSMART (circulating single-molecule amplification and resequencing technology)17 to detect off-target mutation frequencies and types, and their relative distributions of edited T cells infused in the first cycle (Fig. 3a) and the second cycle (Extended Data Fig. 3). The median mutation frequency of all off-target sites was 0.05% (range, 0–0.25%), which was much lower than that of the on-target site (median, 1.69% with PD-1; range, 0.03–11.08%). The modification ratio of on-target/off-target was 48.7. Mutation types such as frameshift indels (71.2%) and nonframeshift indels (28.6%) were mainly detected in on-target sites rather than off-target sites. In the predicted off-target sites, intergenic (48.1%) and intronic (35.1%) mutations composed a large proportion (Fig. 3a). The off-target effect was further assessed in seven samples of edited T cells infused in the second cycle, showing similar results. The modification ratio of on-target/off-target was 105.2 (Extended Data Fig. 3). Considering the limitation on the width of coverage by NGS, we have also performed WGS at 100× coverage using Cas-OFFinder18 for identification of potential off-target sites. This method is an unbiased and direct method for extensive mutation searching. Contrarily, Guide-seq19 is affected by considerable false negatives despite its high sensitivity20. In the current study, a total of 2,086 potential off-target sites were predicted by Cas-OFFinder. Indel events were not detected within 15 base pairs (bp) up- and downstream of the sites. When each site was broadened to 200 bp up- and downstream, 84 indel events in 53 sites were detected (Supplementary Table 4). All indel events were of 1-bp length variance on nucleotide repeats and thus were not considered to be true off-target events (Fig. 3b). There was a discrepancy in detectable off-target events between NGS and WGS, which we hypothesize is due to the differences in coverage of sequencing. Compared with NGS, WGS has the ability to detect potential off-target sites with higher frequency20. However, due to the limitations of our methodology, absence of detectable off-target sites does not entirely rule out off-target events.

Fig. 3: Off-target analysis by NGS and WGS.

a, Characteristics of off-target mutation types, frequencies and numbers determined by NGS for the edited T cells of 12 enrolled patients before the first cycle of infusion. Bar graph and pie graph above represent the types, the numbers and the composition of off-target mutations, color-coded according to the legend to the right. Intergenic (48.1%) and intronic (35.1%) mutations composed the majority proportion. Heatmap shows the mutation number of predicted off-target sites (18 off-target sites, OT1–18) and on-target sites for individuals. Bar graph on the right represents mean mutation frequencies of each site among all of the patients (bars represent mean ± s.d., each dot represents individual data, n = 12). The modification ratio of on-target/off-target was 48.7. Pie graph on bottom right shows the composition of on-target mutation. The mutation types of on-target consisted of frameshift or nonframeshift (deletion/insertion mutation) and stop-gain mutations, while the vast majority were deletion mutations (90.6%). b, The genome DNA assessed by WGS (100×). From 19 samples for infusion from 12 patients, 2,086 potential off-target sites predicted by Cas-OFFinder were further analyzed by bioinformatics filtration. Indel events were not detected within 15 bp up- and downstream (±15 bp) of the sites. When each site was broadened to 200 bp up- and downstream (±200 bp), 84 indel events in 53 sites were detected. All of the indels were of 1 bp length variance on nucleotide repeats and thus were not considered to be true off-target events. Pie graph represents the types and the composition of 84 indel events. Intergenic (38.1%) and intronic (39.3%) mutations composed the majority proportion. NcRNA, Non-coding RNA; 3′ UTR, 3′ untranslated region.

Source data

Safety and clinical outcomes

The first patient in the pre-A cohort (pre-A-01) developed grade 1 arrhythmia, detected as a premature beat on electrocardiography at day 1 of cell infusion. Cardiac evaluation, including myocardial enzyme test (creatine kinase-MB, troponin T and myoglobin), and echocardiography were normal (Extended Data Fig. 4). This patient received the second and third infusions on days 3 and 5, and no dose-limiting toxicities (DLTs) were observed. The second patient of the pre-A cohort (pre-A-02) did not experience any DLT. Per protocol, a total of ten patients were enrolled in dose-escalation cohorts A–C (Fig. 1b). At the time of data cut-off (31 January 2020), a total of 37 infusions were given to the enrolled patients. Patients’ characteristics are summarized in Extended Data Fig. 5.

Grade 1/2 treatment-related adverse events (AEs) occurred in 11 of the 12 patients, and no AEs occurred in patient C-03 (Table 1). There were no grade ≥3 treatment-related AEs. Details of AEs are summarized in Supplementary Table 5. Common treatment-related AEs included lymphopenia (n = 3, 25%), fatigue (n = 3, 25%), leukopenia (n = 2, 17%), fever (n = 2, 17%), arthralgia (n = 2, 17%) and skin rash (n = 2, 17%). Other treatment-related AEs included neutropenia, infusion-related reaction, hyperhidrosis, premature beats, hypertension, increased alanine aminotransferase, increased aspartate aminotransferase, thrombocytopenia and anemia (n = 1, 8%, Table 1). Duration of treatment-related AEs is shown in Extended Data Fig. 6, and the more durable AEs included arthralgia (52.4 weeks), premature beats (19.1 weeks) and fatigue (15.3 weeks). Serum cytokine analysis was done for all patients, and none of the patients had evidence of cytokine release syndrome (Fig. 4a). No DLTs were observed. During the follow-up for safety (up to 2 yr after the last treatment), 3 of the 12 patients had grade 1 treatment-related AEs (Extended Data Fig. 7).

Table 1 Summary of treatment-related AEs
Fig. 4: Persistence, TCR clones, cytokines and clinical outcomes of patients who received therapy with the gene-edited T cells.

a, Changes in IL-6 (left), IL-10 (middle) and tumor necrosis factor alpha (TNF-α) (right) in all patients during and after treatment. b, Bar graph representation of best percentage change from baseline tumor size, as determined based on Response Evaluation Criteria in Solid Tumors (RECIST), v.1.1. The dashed line represents the cut-off for PD (at 20% change in tumor size from baseline). An asterisk denotes PD with new lesions. c, Swimmer plot represents the treatment duration for 11 patients. Arrow represents that the patient was alive at the time of date cut-off (31 January 2020). Color bar represents the PFS time; blank bar represents the survival time after PD. Hollow and solid dots represent that patient C-02 experienced only intracranial progression and then intrapulmonary lesion PD, respectively. d, Frequency of edited PD-1 gene in PBMCs of each patient based on NGS from the time of first-cycle cell infusion up to 1 yr. e, TCR diversity of patients’ PBMCs collected at different time points during therapy. f, Frequency of unique TCR clones among patients’ PBMCs during the course of the trial. The unique TCR clones were defined as the different clones of only edited T cells in PBMCs after the first treatment from PBMCs of the baseline.

Source data

Among the enrolled 12 patients, the median progression-free survival (PFS) was 7.7 weeks (95% confidence interval (95% CI), 6.9–8.5 weeks). The median overall survival (OS) was 42.6 weeks (95% CI, 10.3–74.9 weeks) (Extended Data Fig. 8). The median follow-up was 47.1 weeks (range, 13.4–116.0 weeks). One patient was unavailable for assessment of efficacy due to early withdrawal. None of the patients attained partial response and two patients had stable disease (Fig. 4b). The 8-week disease control rate was 16.7% (2 of 12; 95% Cl, 2.1–48.4%). As of 31 January 2020, all patients had disease progression. Of 12 patients, 11 (91.7%) died from tumor progression. The one remaining patient (C-02) was still receiving other therapy (Fig. 4c). No treatment-related death occurred in our study.

In vivo tracking of edited T cells

To monitor the in vivo persistence of edited T cells after infusion, we used NGS to detect the disrupted PD-1 gene in DNA extracted from blood samples of treated patients per protocol. Presence of the edited PD-1 gene in PBMCs, acting as a surrogate for gene-edited T cells, was detected in 11 of the 12 treated patients (patient A-03 was not available for testing due to early withdrawal) at multiple time points between 8.0 and 52.0 weeks from the first infusion, and the allele frequency of edited PD-1 gene varied among the patients. Interestingly, one patient (B-01) with durable disease control showed a persistent, moderate level of edited PD-1 genes in PBMCs (Fig. 4d).

Peripheral TCR clonal diversity

High TCR diversity in PBMCs is associated with better response to immune checkpoint inhibitor (ICI) treatment21. The Shannon diversity index was calculated using TCR clonality of PBMCs from healthy donors and patients with refractory NSCLC. The median index of TCR clone diversity was significantly lower in our 12 treated patients at baseline than in healthy donors, with 5.65 (range, 2.22–7.44) and 8.11 (range, 5.17–8.86) (P = 0.0005), respectively (Extended Data Fig. 9). The TCR clone diversity was detectable in all patients and we observed a trend toward increased diversity over time in one patient (Fig. 4e). The median index was 6.54 (range, 2.56–8.29) at week 8, 7.05 (range, 6.48–8.57) at week 20 and 8.33 at week 68 following cell therapy.

We also tracked unique TCR clones by performing NGS of the complementarity-determining region 3 (CDR3). Unique TCR clones were defined as TCR clones that were undetectable in baseline PBMCs, but became detectable after the first infusion of gene-edited T cells. We observed that the unique clones were detected in all patients on day 6 and at week 8 after the first infusion with median frequency of 1.01% (range, 0.12–3.88%) and 0.74% (range, 0.33–3.29%), respectively. These TCR clones were detectable until week 76 in patient B-01 (Fig. 4f).

Clinical outcome of patient B-01

Patient B-01 had stable disease lasting for 76 weeks. This 55-yr-old woman had metastatic lung adenocarcinoma with positive PD-L1 expression (tumor proportion scores of 5%). She had failed three lines of treatment before infusion. At 8 weeks after the first T cell infusion, she developed pleural effusion which spontaneously resolved in 4 weeks without intervention. Positron emission tomography–computed tomography at 52 weeks showed only one tumor site with increased metabolic uptake (Fig. 5a). Four re-biopsy specimens were collected (before treatment, post-treatment weeks 24 and 54, and at tumor progression at week 76). Biopsy analysis revealed that there was minimal residual tumor, and increased infiltrating T cells (CD3+/CD8+ T cells) and CD68+ macrophages, at week 54 compared with the baseline, while the infiltrating T cells were hardly detected at tumor progression (Fig. 5b; semiquantitative analysis of staining density is presented in Extended Data Fig. 10).

Fig. 5: Outcomes of B-01.

a, Computerized tomography (CT) and positron emission tomography (PET) images of patient B-01. The patient underwent radiological imaging before T cell infusion (baseline) and at weeks 8, 12, 20, 52 and 76 after the first cycle of T cell infusion. Left pleural effusion developed at week 8, and then resolved at week 12 without any intervention. b, Representative photo micrographs showing histological staining of tumor specimens from patient B-01. At 54 weeks after cell infusion, tumor biopsies were performed in the high-metabolism and low-metabolism areas. Tumor re-biopsy was performed at tumor progression at week 76. Sections were stained, as indicated, with hematoxylin-eosin (H&E) or with immunohistochemistry antibodies directed against CD3, CD8 and CD68. Data are representative of three separate slides. Magnification, ×400. c, PD-1 editing efficiencies in preinfusion edited T cells (left axis) and postinfusion PBMCs (right axis) at different time points in patient B-01 were measured by NGS. d, Frequency of unique TCR clones among PBMCs (left axis) and TCR diversity of PBMCs (right axis) during therapy were measured by NGS. e, Heatmap represents the frequency distribution of TCR clones in the tumors and PBMCs at indicated time points. The top 20 TCR clones detected in four tumor samples and the overlapped clones among tumor samples were included for this analysis. Each colored bar represents a unique CDR3, and each color denotes the CDR3 frequency in each sample. The T cell repertoire in the tumor changed substantially during T cell therapy. The TCR clones abundant in baseline tumor were also found in the PBMCs after cell treatment. f, Sum of high-frequency TCR clones in the tumors of patient B-01 at baseline and post-treatment (weeks 24, 54 and 76 after cell treatment), and PBMCs at indicated time points. High-frequency clones were defined as those with abundance ≥0.1% in tumor tissue. g, The number of shared TCR clones between the tumor tissue and edited T cells (Venn diagram). The proportions of shared TCR clones with high frequency (≥0.1%) and low frequency (<0.1%) in tumor tissue at baseline (below bar). The high-frequency clone (≥0.1% in tumor) composed the majority (82%) of shared clones at the baseline. h, Changes in shared TCR clones from patient B-01 at baseline and weeks 24, 54 and 76 after cell treatment. Patient B-01 experienced tumor progression at week 76. ORR, objective response rate; SD, stable disease.

The median PD-1 editing efficiency of infused T cells from patient B-01 was 12.50% (range, 1.65–24.85%). The frequency of edited PD-1 gene detected in PBMCs was similar to the editing efficiency of infused T cells (Fig. 5c). This patient had persistent unique TCR clones and TCR diversity in the peripheral blood during the 17.5 months of clinical follow-up (Fig. 5d). Notably, the TCR clones with high frequency (≥0.1%) in tumors changed greatly during treatment and were also correspondingly detectable in the PBMCs (Fig. 5e,f). We further explored the shared TCR clones between tumor tissue and the edited T cells in this patient. Interestingly, the high-frequency clones (defined as ≥0.1%) in the tumor composed the majority (82%) of the shared TCR clones between tumor and the edited T cells at baseline (Fig. 5g). The numbers of shared clones increased by week 24 and week 54 during stable disease, and then declined at week 76 with disease progression (Fig. 5h).


Our study further supports the safety and feasibility of gene-edited T cell therapy by CRISPR–Cas9 technology in the clinic, in line with data from another recently published clinical trial6. The final dose is 4 × 107 T cells per kg given in a 28-d cycle. All treatment-related AEs were grade 1 or 2, and the more frequent events included lymphopenia, fatigue, leukopenia, fever, arthralgia and skin rash. Severity of AEs appeared to be dose-independent (Table 1). Only one patient had grade 1 arrhythmia after the first infusion, lasting 19.1 weeks (Extended Data Fig. 6), but the causal relationship to treatment is uncertain. We speculate that lymphocyte infiltration to the myocardium occurred in this patient, which is a known but rare complication of immune checkpoint inhibitors such as ipilimumab and nivolumab22, but we do not have sufficient evidence to confirm the causal relationship.

Off-target editing by the CRISPR–Cas9 system is a major concern for clinical application of the technology. Off-target events were not detected in a patient who received CRISPR-edited stem cell therapy5. For three patients treated with multigene-edited T cell therapy, there were minimal off-target events at TRAC and TRBC6. Similarly, our median mutation frequency of off-target cleavage was low in all sites by NGS, with the majority of these mutations being located in intergenic or intronic regions (Fig. 3a), which were unlikely to have significant effects on coding genes. Frameshift or nonframeshift mutations (insertion/deletion mutations) caused by CRISPR–Cas9 technology may induce gene dysfunction23,24. The mean frequencies of frameshift and/or nonframeshift mutations in the predicted sites appeared to be low, by both 20,000× NGS coverage (Fig. 3a) and 100× WGS coverage (Fig. 3b). We adopted the plasmid electroporation strategy, while the recent publication from the United States used RNP editing6. It appears that both strategies are associated with low incidence of off-target events, but this should be verified in future studies.

The persistence of edited T cells in the periphery and intratumorally could support the potential efficacy of the gene-edited T cells25,26. The disrupted PD-1 gene and unique TCR clones were detectable at 4 weeks in all patients, while the finding was more persistent in patients with stable disease (pre-A-01, B-01) (Fig. 4d,f). Notably, the high-frequency clones (≥0.1%) had a decreasing trend both in tumor and in PBMCs (Fig. 5f). Given these data and that the more frequent TCR clones are likely reactive to tumor neoantigens27, we postulate that the edited T cells were capable of recognizing tumor neoantigens. However, limited by the small sample size of the study, this hypothesis can only be validated in a future study.

We did not observe objective responses in this cohort of heavily treated patients. Several factors, including insufficient T cell expansion, lack of antigen specificity and/or tumor cells not being PD-L1 dependent, may account for the low response rate. While we did confirm that all enrolled patients had PD-L1+ tumors, PD-L1 expression by immunohistochemistry does not necessarily correlate with PD-L1 dependency. Five of the enrolled patients could not participate as a result of insufficient T cell expansion ex vivo. During primary cell culture and editing, some of the well-transformed cells failed to proliferate. This could be related either to poor quality of T cells from heavily treated patients or to the manufacturing procedure. Comparatively, in four of six patients whose T cells were manufactured by RNP editing, sufficient cell product was successfully obtained6. It is possible that the lack of tumor response was related to T cell anergy or exhaustion resulting from chronic exposure to tumor-associated antigens or anti-cancer therapy28. Unlike other cancer types, including melanoma and sarcoma and virus-related malignancies, there are currently no promising target antigens for lung cancer. Our study used T cells from peripheral blood for gene editing, which may include tumor-reactive T cells29. However, efficacy may be limited in patients with low frequencies of tumor-reactive T cells. Future clinical trials should use improved quantities of tumor-reactive or tumor antigen-specific T cells.

We used nucleofection to deliver plasmids encoding Cas9 and sgRNA into primary T cells to disrupt genomic PD-1 expression. A potential benefit of this approach is shorter exposure of the editing genomic components to the target cell, which may reduce the risk of off-target mutations30. However, this method may be associated with low editing efficiency. It is important to note that the editing protocol for the cells used in the current trial was based on the reagents available in 2015. Other technologies developed since the launch of our trial, such as RNA encoding the Cas9 and sgRNA31, Cas9 RNPs32, Cas9 messenger RNA or protein with chemically modified sgRNA33, may result in improved editing efficiencies. Using nucleofection, the median editing efficiency for PD-L1 was 5.81% (range, 0.42–24.85%). In contrast, using RNP editing for TRAC, TRBC and PDCD1, the editing efficiency was 45%, 15% and 20%, respectively6. The new approach appears to have advantage in editing efficiency over plasmid nucleofection. Our study has confirmed the safety and feasibility of gene-edited T cell therapy, while we anticipate that the higher editing efficiency with RNP editing would further improve the treatment efficacy.

We should emphasize that the long-term safety of clinical use of CRISPR–Cas9 technology remains important. After a median follow-up of 47.1 weeks (range, 13.4–116.0 weeks) as of 31 January 2020, none of the 12 patients had treatment-related severe AEs (Extended Data Fig. 7). The same was observed in the three patients who received RNP gene-edited T cells, but their duration of follow-up was much shorter6. We performed the gene-editing process ex vivo and demonstrated that the lifespan of the gene-edited T cells was short (Supplementary Fig. 1), suggesting that the risk of permanent genomic alteration is limited.

In summary, this study has demonstrated the safety and feasibility of CRISPR–Cas9 gene-edited T cell therapy targeting the PD-1 gene in a cohort of patients with advanced lung cancer. There were limited off-target effects observed with this approach, and we have demonstrated CRISPR-based technology to be clinically feasible. However, considering the limitations of the technology used in the study, future trials should use newer and more effective gene-editing systems.


Study design and patients

This is a dose-escalating phase I clinical trial with four cohorts. In the pre-A cohort, we aimed to assure the safety of CRISPR-edited T cells to humans; a total of two patients received 2 × 107 edited T cells per kg body weight and were carefully monitored for 28 d (pre-A cohort). After assurance of the absence of significant toxicity, we started enrollment in the three-cohort dose-escalation study (cohorts A, B and C; Fig. 1b). With three subjects to each cohort, each patient received 1 × 107, 2 × 107 or 4 × 107 T cells per kg in a series of three infusions on days 1, 3 and 5, respectively. The first infusion contained 20% of the total number of cells; the second infusion, 30%; and the third infusion, 50%. At 3 d before the first infusion, cyclophosphamide (20 mg kg−1) was given for lymphocyte depletion34. At day 28 from the first T cell infusion, and in the absence of major toxicity or disease progression, cycle 2 of the gene-edited T cell infusion would be administered, but without cyclophosphamide infusion. If the patients obtained clinical benefit, they would receive continuous treatment until unacceptable toxicity or withdrawal of consent, or disease progression. The unacceptable toxicity was defined as grade ≥3 cytokine release syndrome as well as grade ≥2 AE requiring systemic glucocorticoid therapy; in addition, patients would be withdrawn in the event of any AE, laboratory abnormality or intercurrent illness that the investigators judged would lead to great clinical risk in the patients.

Eligible patients were aged 18–70 yr and had histologically or cytologically confirmed stage IIIB or IV NSCLC35. Other major inclusion criteria included disease progression after three or more systemic therapies (including at least one molecular targeted therapy for patients with activating EGFR mutation or ALK fusion gene), positivity for PD-L1 expression as defined by tumor proportion score ≥ 1% using Ventana PD-L1 (SP142), adequate hematologic and organ functions, Eastern Cooperative Oncology Group score of 0–2 and a life expectancy of over 3 months. The re-biopsy samples were collected for exploratory analysis at the times of before treatment, half a year and 1 yr post-treatment, and at tumor progression if clinically feasible and agreed by patients. Key exclusion criteria were brain metastasis and/or spinal cord compression that were uncontrolled, active autoimmune disease, treatment with other immunomodulators within 28 d before the study and immunodeficiency. Complete eligibility criteria are provided in the study protocol (Supplementary clinical trial protocol). The patients were enrolled from 26 August 2016 to 21 March 2018.

The study protocol and subsequent amendments were approved by the Institutional Review Board of the West China Hospital, Sichuan University, Chengdu, China (2016 Review, No. 123). Written informed consent was obtained from all patients. Data were collected by the investigators. MedGenCell (Chengdu, China) provided the gene editing of T cells.

Study endpoints and assessments

The primary endpoints of this phase I dose-escalation study included safety and feasibility. Feasibility was defined by sufficient and viable (>90%) edited T cells being able to be manufactured from the majority (>50%) of enrolled patients. Secondary endpoints included objective response rate, 8-week disease control rate, PFS and overall survival. Other exploratory objectives included in vivo tracking of edited T cells, diversity and dynamics of TCR clones in PBMCs and re-biopsy tumor specimens (if available) and their potential correlations among efficacy outcomes.

All patients underwent baseline tumor assessment, including contrast-enhanced computed tomography and magnetic resonance imaging (MRI) of the brain, within 28 d before the first treatment. Tumor responses were assessed using Response Evaluation Criteria in Solid Tumors (RECIST 1.1) at weeks 8 and 12. Afterward, tumor responses were assessed every 8 weeks until disease progression or death.

AEs were monitored and graded according to the National Cancer Institute Common Terminology Criteria for Adverse Events (v.4.03) for 24 months after the last cell infusion or until the patient was lost to follow-up. AEs were identified using a predefined list of Medical Dictionary for Regulatory Activities (MedDRA, v.21.0) terms. DLTs were defined as events occurring within the first 28 d after the first cell infusion that required systemic glucocorticoids to be administered to patients, grade ≥2 autoimmune toxicities or any other grade ≥3 events, which were probably attributable to the T cell therapy. The association of AEs with treatment was determined by investigators.

CRISPR plasmid construction

To disrupt PD-1 in human T cells we designed a pair of sgRNAs (sgRNA1 and sgRNA2), showing the highest editing efficiency in the 293 T cell line, to target exon 2 of PD-1. The Cas9 expression vector pST1374-Cas9 (Addgene) was constructed as previously described14. The S. pyogenes Cas9 (SpCas9) was used in the study. For the construction of the sgRNA vector, the single vector pGL3-hPD1 containing two sgRNAs was constructed as follows. Oligos for generation of sgRNAs were synthesized and annealed to form sgDNAs. Subsequently, the pGL3-dual U6 sgRNA scaffold-PGK-Puro vector (YouBio) was cleaved using restriction endonucleases. Finally, the sgDNAs were cloned one by one into the BsaI or BbsI sites of pGL3-dual U6 sgRNA scaffold-PGK-Puro vector, forming tandem dual sgRNA expression vectors. pGL3-dual U6 sgRNA scaffold-PGK-Puro vector was modified from pGL3-U6 sgRNA scaffold-PGK-Puro vector (Addgene).

Preparation of PD-1-edited T cells

PD-1-edited T cells were manufactured at MedGenCell (Chengdu, China). Briefly, PBMCs were isolated from 60–80 ml of peripheral blood from patients with NSCLC before every treatment cycle by centrifugation on a Ficoll density gradient. Approximately 5–10 × 106 cells were transfected by electroporation technology with the intended plasmids using a Nucleofector 2b device (Lonza) and the VPA-1002 Human T cells Nucleofector Kit (Lonza). After electroporation, cells were resuspended in X-VIVO15 medium (Lonza) containing 10% autologous serum with IFN-γ (2,000 IU ml−1) and DNase (3 μg ml−1) on day 1, and cultured with anti-CD3 antibody (0.4 μg ml−1), anti-CD28 antibody (0.4 μg ml−1) and interleukin-2 (IL-2) (1,000 IU ml−1) from day 2 at 37 °C in a 5% CO2 incubator. Cell culture medium was half-replaced every 2–3 d. Cell density was adjusted to 1–1.5 × 106 cells per ml. The final yield of cells was approximately 0.5–1.5 × 109 cells per bag with >95% cell viability at 20–28 d. The first and second cycles of cell infusion were conducted in the phase I clinical trial ward at West China Hospital, and the subsequent infusions were administrated in the inpatient ward at West China Hospital.

T7E1 cleavage assay and sequencing

To test the efficiency of PD-1 gene editing, the T7E1 cleavage assay was performed as follows. Briefly, cells were collected and digested in lysis buffer (10 μM Tris-HCl, 0.4 M NaCl, 2 μM EDTA, 1% SDS and 100 μg ml−1 Proteinase K). The genomic DNA was extracted by phenol-chloroform and precipitated using alcohol. Subsequently, targeted fragments of PD-1 were PCR-amplified from genomic DNA using rTaq (Takara), and the PCR products were purified with a PCR cleanup kit (Axygen). The primer sequences for human PD-1 are GTGGTGACCGAAGGGGACA (F) and GGATGACGTTACCTCGTGCG (R) from 5′ to 3′. Purified PCR products were denatured and re-annealed in NEBuffer 2 using the Veriti thermocycler (Applied Biosystems). Hybridized PCR products were digested with T7E1 (NEB) for 30 min and separated on 2% agarose gels. The purified PCR products were ligated with pEASY-T1 vector (TransGen Biotech). Ligation products were used for transformation and 50 colonies per transformation were sequenced using the universal primer M13F.

Flow cytometry

For detection of surface PD-1 expression, edited and unedited T cells were assessed by FACSAria II flow cytometry (BD Biosciences) after stimulation with anti-CD3/anti-CD28 Dynabeads (ThermoFisher Scientific). Cells were incubated with the following antibodies at room temperature in the dark for 30 min: CD3-BV510 (HIT3a, BD Biosciences), CD4-BV421 (RPA-T4, BD Biosciences), CD8-PE (HIT8a, BD Biosciences) and PD-1-APC (MIH4, BD Biosciences). For IFN-γ staining, cells were incubated with Leukocyte Activation Cocktail (BD Biosciences) and human PD-L1 protein (1 μg ml−1, ACRO Biosystems) at 37 °C for 4 h. Then, cells were collected and washed, followed by staining with CD3-BV510 and CD8-APC-Cy7 at room temperature for 30 min. After fixation and permeabilization, cells were incubated with IFN-γ-PE (B27, BD Biosciences). Isotype staining was used as the negative control to determine the positive gate. Data were analyzed using FlowJo software (v.7.6.1). Editing efficiency (%) was calculated with the following equation: editing efficiency (%) = (PD-1-positive unedited cells (%) − PD-1-positive edited cells (%))/PD-1-positive unedited cells (%).

Long-term cell culture

To test whether PD-1 disruption alters T cell viability or proliferation, the total cell number, viability and doubling time were assessed in a long-term culture. Cell culture medium was half-replaced every 2–3 d with fresh complete medium containing IL-2 (1,000 IU ml−1), and cell density was stabilized at 1–1.5 × 106 cells per ml. The viability of PD-1-edited and -unedited T cells, cultured for 40 d, was evaluated with trypan blue staining.

Detection of off-target sites by WGS analysis

Genomic DNA was isolated from PBMCs and PD-1-edited T cells using QIAamp Circulating Nucleic Acid Kit (Qiagen). The whole-genome DNA library constructed using edited T cells and unedited T cells was sequenced on the NovaSeq 6000 platform (Illumina) with 100× coverage. We used samtools (v.1.1), sambamba (v.0.6.8), strelka2 (v.2.8.3) and Fusionmap (v. to process sequence alignment data (BAM file) and variant call format (VCF file), and then aligned each 150-bp sequenced read to the human reference genome (hg19) using Burrows–Wheeler Aligner (BWA, v.0.7.10) in the ‘mem’ mode (‘BWA-MEM’) with default parameters36. Reads caused by PCR duplication were labeled by sambamba37, followed by strelka2 used for variants calling38. Lastly, all structure variant mutations were annotated with ANNOVAR (v.2019Oct24).

To predict potential off-target sites, we screened the whole human genome with the webtool Cas-OFFinder (v.2.4, Any sequences within five mismatches (or fewer; a bulge penalty equals two base mismatches) of the two 20-nucleotide guide RNA sequences, followed by an NRG PAM, were called, and we obtained 2,086 sites39. Sequence alignment data (BAM files) were integrated by sambamba to call structure variants using strelka2 with default parameters40 and to identify all translocations by Fusionmap. To get the candidate structure variants caused by off-target sites, we focused on insertion or deletion events (records contain ‘FILTER = PASS’ and ‘INFO does not contain SNVHPOL information’ in the VCF file) and filtered the indel events detected in unedited samples. To further confirm whether off-target sites occurred among these sites, each event was broadened to 200 bp up- and downstream to intersect with 2,086 potential off-target sites. A total of 84 indel events in 53 sites were detected.

Detection of off-target sites by NGS

Potential off-target sites were predicted using sgRNAcas9 (v.2.0.6)16. The top 18 sites (that is, highest similarity to the targeted PD-1 sequence) were considered potential off-target sites. Briefly, genomic DNA was isolated from PD-1-edited T cells with a QIAamp Circulating Nucleic Acid Kit (Qiagen). DNA was amplified by PCR for 14 cycles and further constructed DNA libraries. After denaturation of DNA to single strand, a bridging oligonucleotide synthesized with end sequences complementary to the PCR adaptors was used to facilitate ligation-mediated circularization with Taq ligase. The bridging oligonucleotide was also designed with a 7-bp NNNNNNN code and was synthesized as a mixture of degenerate molecules. The targeted regions within the circularized single-stranded molecules were amplified by inverse PCR using primer pairs located 16–29 bp apart.

In total, multiplexing of 36 inverse primer pairs was used to target mutation sites within the 18 locations. The pool of allelic molecules was subjected to pair-end sequencing on the NextSeq500 platform with 20,000× coverage of bases in the off-target sites. Finally, cSMART17 was used for analysis and identification of point mutations and indels through the fastq join program (v.1.3.1, The sequence reads were mapped to the human reference genome (hg19) using BWA-MEM. A minimum of one unique allelic read was used to define a read group with or without a mutation. Duplicate or higher-order reads with the same start and stop positions were distinguished by their unique barcodes and counted only once to correct PCR bias. The abundance of each mutation was expressed as the number of mutant molecules. For each region of potential off-target sequences, the number of unique allelic reads with mutations, the sum of fractions of each mutation and the fraction of indels in a specific region were calculated as the mutation rate of the given region. The number of unique allelic reads with deletions (Del) and insertions (Ins) of 1 bp or more are indicated. The indel frequency is the total number of unique allelic reads containing insertions and deletions for each off-target site expressed as a percentage of the total number of mapped unique allelic reads42. The primer pairs are listed in Supplementary Table 6.

The modification ratio of on-target/off-target was calculated as follows:

$${\mathrm{mR}} = \frac{1}{{18}}\mathop {\sum }\limits_{a = 1}^n I_a(\frac{1}{{\mathop {\sum }\nolimits_{a = 1}^n T_{1a}}} + \frac{1}{{\mathop {\sum }\nolimits_{a = 1}^n T_{2a}}} + \ldots + \frac{1}{{\mathop {\sum }\nolimits_{a = 1}^n T_{18a}}})$$

where mR represents the modification ratio of on-target/off target; Ia represents the on-target frequency of patient number a; n represents the number of patients; T1a represents the frequency of off-target site 1 (OT1) in patient number a; and T18a represents the frequency of off-target site 18 (OT18) in patient number a.

Efficiency validation and in vivo tracking of edited T cells

We performed assays of NGS targeting the PD-1 gene on genomic DNA isolated from edited T cells and PBMCs using the QIAamp DNA Blood Mini Kit (Qiagen), and the DNA concentration was measured by Qubit dsDNA HS Assay Kit (Invitrogen). Thereafter, genomic regions of PD-1 were amplified by PCR with primers homologous to the region of interest and the appropriate Illumina forward and reverse adapters (Supplementary Table 7). Primer pairs for the first round of PCR (PCR1) for all genomic sites can be found in Supplementary Table 8, and the optimal conditions of the PCR1 reaction were as follows: 0.5 μM forward primer, 0.5 μM reverse primer, 10 ng of genomic DNA and 10 μl of Phusion Hot Start Flex 2× Master Mix (NEB). Reactions were incubated for 2 min at 95 °C; 19 cycles of 10 s at 98 °C, 20 s at 65 °C and 15 s at 72 °C; followed by 2 min at 72 °C for extension. PCR products were verified by 2% agarose gel electrophoretic analysis and compared with DNA standards (Quick-Load 100-bp DNA ladder). Unique Illumina barcoding primer pairs were added to each sample for the secondary PCR reaction (PCR2), which was performed in a manner similar to PCR1. Finally, PCR products purified by AMPure XP beads (NEB) and quantified with an Agilent Bioanalyzer 2100 were sequenced on a HiSeq X Ten instrument (Illumina) following the manufacturer’s protocols.

In brief, several steps, including demultiplexing of sequencing reads by MiSeq Reporter (Illumina), alignment of sequences by a modified version of MATLAB script and the Smith–Waterman algorithm and converting of bases (<30 of quality score) to N, were completed. Then, indels were quantified by the MATLAB script. After filtering out reads, insertion or deletion events of at least 2 bp within 30 bp up- or downstream of the Cas9 cleavage site were counted. Finally, the ratio of mutation reads to sequencing reads was calculated as the editing frequency of PD-1 of edited T cells and PBMCs.

Assessment of TCR clonal diversity

We used NGS to analyze CDR3 in the TCR-β chain and thereby identify unique TCR clones. The unique TCR clones were defined as the TCR clones undetectable in baseline PBMCs but available in the edited T cells of the first cycle. Immunosequencing of the CDR3 of human TCR-β chains was performed using the IR-seq platform from the Geneplus-Beijing Institute43. DNA from PBMCs, edited T cells and biopsies was amplified in a bias-controlled multiplex PCR system, followed by high-throughput sequencing. The absolute abundance of each unique TCR-β CDR3 was identified and quantitated. Then, the diversity of the TCR repertoire was calculated based on the Shannon index, similar to that previously described43.

$${\mathrm{Shannon}}\,{\mathrm{index}} = - \mathop {\sum }\limits_{i = 1}^s \frac{{n_i}}{N}{\mathrm{ln}}\frac{{n_i}}{N}$$

In this expression, ni is the clonal size of the T cell clonotype (the number of copies of a specific clonotype), i is the serial number of different clonotypes, i = 1, ..., n, S is the number of different clonotypes and N is the total number of TCR sequences analyzed.

Immunohistochemistry staining

For immunohistochemistry staining, tumors were fixed in 10% phosphate-buffered formaldehyde, embedded in paraffin and sectioned. Formalin-fixed, paraffin-embedded tissues were cut into 4-μm-thick sections and placed on plus-charged slides. Slides were heated at 65 °C overnight, and then were deparaffinized in xylene, hydrated through graded alcohols and water, pretreated by microwaving in Tris-EDTA buffer and treated for endogenous peroxidases with 3% hydrogen peroxide in PBS. Subsequently, immunostaining was performed by incubating with primary antibodies directed against CD3, CD8 or CD68 (PGM-1) (1:100, Dako) overnight at 4 °C. Then, slides were incubated with enzyme-labeled secondary antibody reagent (Dako) for 45 min at room temperature, followed by the ultraView DAB Detection on a Ventana Ultra instrument. Slides were analyzed under an optical microscope (Olympus), and analyzed using ImageJ software (v.1.47).

Detection of cytokines

Peripheral blood was collected in red-top tubes (CDRICH, BD Biosciences) and was processed within 2 h in the Department of Laboratory Medicine in the West China Hospital, which is certified by the College of American Pathologists. The serum of peripheral blood was separated and purified for cytokine detection. IL-10 and TNF-α were measured using chemiluminescence (IMMULITE 1000, Siemens). IL-6 was measured using rate nephelometry (IMMAGE800, Beckman Coulter).

Statistical analysis

We used the Kaplan–Meier method to summarize time-­to­-event variables including PFS and overall survival. PFS is defined as time from the date of first edited T cell infusion to the date of disease progression or death due to any reason. Overall survival is the duration from date of first edited T cell infusion to the date of death due to any reason. A subject who had not experienced disease progression before the cut-off date was censored at the last disease assessment date. Subjects alive until the cut-off date were censored at the last time at which they were known to be alive for overall survival analysis. IFN-γ production was compared between pre- and postedited T cells using two-tailed paired t-test. TCR diversity was compared between patients and healthy donors with the two-tailed Wilcoxon rank-sum test. Immunohistochemistry staining density was semiquantified by ImageJ software (v.1.47), and then compared using one-way analysis of variance with Sidak’s multiple comparisons test. Statistical significance was set at P ≤ 0.05 (two-sided). All other analyses, including those of AEs, efficacy and immunologic data, were descriptive. All analyses were performed using Prism, v.7.0 (GraphPad) or SPSS, v.22.0.

Reporting Summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability

All requests for raw and analyzed data and materials are promptly reviewed by the West China Hospital to verify whether the request is subject to any intellectual property or confidentiality obligations. Patient-related data not included in the paper were generated as part of clinical trials and may be subject to patient confidentiality. Any data and materials that can be shared will be released via a material transfer agreement. All other data that support the findings of this study will be provided by the corresponding author upon reasonable request when possible. Raw data for Figs. 24 and Extended Data Figs. 13, 6 and 810 are in the Source Data. The raw sequencing data reported in the study have been deposited in the Genome Sequence Archive for Human ( at the BIG Data Center, Beijing Institute of Genomics, Chinese Academy of Sciences, under accession number PRJCA002488.

Change history

  • 18 June 2020

    An amendment to this paper has been published and can be accessed via a link at the top of the paper.


  1. 1.

    Cong, L. et al. Multiplex genome engineering using CRISPR/Cas systems. Science 339, 819–823 (2013).

    CAS  PubMed  PubMed Central  Google Scholar 

  2. 2.

    Jinek, M. et al. A programmable dual-RNA-guided DNA endonuclease in adaptive bacterial immunity. Science 337, 816–821 (2012).

    CAS  PubMed  PubMed Central  Google Scholar 

  3. 3.

    Hsu, P. D., Lander, E. S. & Zhang, F. Development and applications of CRISPR-Cas9 for genome engineering. Cell 157, 1262–1278 (2014).

    CAS  PubMed  PubMed Central  Google Scholar 

  4. 4.

    Kosicki, M., Tomberg, K. & Bradley, A. Repair of double-strand breaks induced by CRISPR–Cas9 leads to large deletions and complex rearrangements. Nat. Biotechnol. 36, 765–771 (2018).

    CAS  PubMed  PubMed Central  Google Scholar 

  5. 5.

    Xu, L. et al. CRISPR-edited stem cells in a patient with HIV and acute lymphocytic leukemia. N. Engl. J. Med. 381, 1240–1247 (2019).

    CAS  PubMed  Google Scholar 

  6. 6.

    Stadtmauer, E. A. et al. CRISPR-engineered T cells in patients with refractory cancer. Science 367, eaba7365 (2020).

    CAS  PubMed  Google Scholar 

  7. 7.

    Mok, T. S. K. et al. Pembrolizumab versus chemotherapy for previously untreated, PD-L1-expressing, locally advanced or metastatic non-small-cell lung cancer (KEYNOTE-042): a randomised, open-label, controlled, phase 3 trial. Lancet 393, 1819–1830 (2019).

    CAS  PubMed  Google Scholar 

  8. 8.

    Reck, M. et al. Pembrolizumab versus chemotherapy for PD-L1-positive non-small-cell lung cancer. N. Engl. J. Med. 375, 1823–1833 (2016).

    CAS  PubMed  Google Scholar 

  9. 9.

    Garon, E. B. et al. Five-year overall survival for patients with advanced nonsmall-cell lung cancer treated with pembrolizumab: results from the phase I KEYNOTE-001 Study. J. Clin. Oncol. 37, 2518–2527 (2019).

    CAS  PubMed  PubMed Central  Google Scholar 

  10. 10.

    Su, S. et al. CRISPR-Cas9-mediated disruption of PD-1 on human T cells for adoptive cellular therapies of EBV positive gastric cancer. Oncoimmunology 6, e1249558 (2016).

    PubMed  PubMed Central  Google Scholar 

  11. 11.

    Beane, J. D. et al. Clinical scale zinc finger nuclease-mediated gene editing of PD-1 in tumor infiltrating lymphocytes for the treatment of metastatic melanoma. Mol. Ther. 23, 1380–1390 (2015).

    CAS  PubMed  PubMed Central  Google Scholar 

  12. 12.

    Chen, W. et al. Cancer statistics in China, 2015. CA Cancer J. Clin. 66, 115–132 (2016).

    PubMed  Google Scholar 

  13. 13.

    Bray, F. et al. Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J. Clin. 68, 394–424 (2018).

    PubMed  PubMed Central  Google Scholar 

  14. 14.

    Su, S. et al. CRISPR-Cas9 mediated efficient PD-1 disruption on human primary T cells from cancer patients. Sci. Rep. 6, 20070 (2016).

    CAS  PubMed  PubMed Central  Google Scholar 

  15. 15.

    Schumann, K. et al. Generation of knock-in primary human T cells using Cas9 ribonucleoproteins. Proc. Natl Acad. Sci. USA 112, 10437–10442 (2015).

    CAS  PubMed  Google Scholar 

  16. 16.

    Xie, S., Shen, B., Zhang, C., Huang, X. & Zhang, Y. sgRNAcas9: a software package for designing CRISPR sgRNA and evaluating potential off-target cleavage sites. PLoS ONE 9, e100448 (2014).

    PubMed  PubMed Central  Google Scholar 

  17. 17.

    Lv, W. et al. Noninvasive prenatal testing for Wilson disease by use of circulating single-molecule amplification and resequencing technology (cSMART). Clin. Chem. 61, 172–181 (2015).

    CAS  PubMed  Google Scholar 

  18. 18.

    Bae, S., Park, J. & Kim, J. S. Cas-OFFinder: a fast and versatile algorithm that searches for potential off-target sites of Cas9 RNA-guided endonucleases. Bioinformatics 30, 1473–1475 (2014).

    CAS  PubMed  PubMed Central  Google Scholar 

  19. 19.

    Tsai, S. Q. et al. GUIDE-seq enables genome-wide profiling of off-target cleavage by CRISPR-Cas nucleases. Nat. Biotechnol. 33, 187–197 (2015).

    CAS  PubMed  Google Scholar 

  20. 20.

    Manghwar, H. et al. CRISPR/Cas systems in genome editing: methodologies and tools for sgRNA design, off-target evaluation, and strategies to mitigate off-target effects. Adv. Sci. 7, 1902312 (2020).

    CAS  Google Scholar 

  21. 21.

    Robert, L. et al. CTLA4 blockade broadens the peripheral T-cell receptor repertoire. Clin. Cancer Res. 20, 2424–2432 (2014).

    CAS  PubMed  PubMed Central  Google Scholar 

  22. 22.

    Johnson, D. B. et al. Fulminant myocarditis with combination immune checkpoint blockade. N. Engl. J. Med. 375, 1749–1755 (2016).

    PubMed  PubMed Central  Google Scholar 

  23. 23.

    Sui, T. et al. CRISPR-induced exon skipping is dependent on premature termination codon mutations. Genome Biol. 19, 164 (2018).

    PubMed  PubMed Central  Google Scholar 

  24. 24.

    Cho, S. W. et al. Analysis of off-target effects of CRISPR/Cas-derived RNA-guided endonucleases and nickases. Genome Res. 24, 132–141 (2014).

    CAS  PubMed  PubMed Central  Google Scholar 

  25. 25.

    Kalos, M. & June, C. H. Adoptive T cell transfer for cancer immunotherapy in the era of synthetic biology. Immunity 39, 49–60 (2013).

    CAS  PubMed  Google Scholar 

  26. 26.

    Rapoport, A. P. et al. NY-ESO-1-specific TCR-engineered T cells mediate sustained antigen-specific antitumor effects in myeloma. Nat. Med. 21, 914–921 (2015).

    CAS  PubMed  PubMed Central  Google Scholar 

  27. 27.

    Pasetto, A. et al. Tumor- and neoantigen-reactive T-cell receptors can be identified based on their frequency in fresh tumor. Cancer Imunol. Res. 4, 734–743 (2016).

    CAS  Google Scholar 

  28. 28.

    Zarour, H. M. Reversing T-cell dysfunction and exhaustion in cancer. Clin. Cancer Res. 22, 1856–1864 (2016).

    CAS  PubMed  PubMed Central  Google Scholar 

  29. 29.

    Dijkstra, K. K. et al. Generation of tumor-reactive T cells by co-culture of peripheral blood lymphocytes and tumor organoids. Cell 174, 1586–1598.e1512 (2018).

    CAS  PubMed  PubMed Central  Google Scholar 

  30. 30.

    Mandal, P. K. et al. Efficient ablation of genes in human hematopoietic stem and effector cells using CRISPR/Cas9. Cell Stem Cell 15, 643–652 (2014).

    CAS  PubMed  PubMed Central  Google Scholar 

  31. 31.

    Ren, J. et al. Multiplex genome editing to generate universal CAR T cells resistant to PD1 inhibition. Clin. Cancer Res. 23, 2255–2266 (2017).

    CAS  PubMed  Google Scholar 

  32. 32.

    Kim, S., Kim, D., Cho, S. W., Kim, J. & Kim, J. S. Highly efficient RNA-guided genome editing in human cells via delivery of purified Cas9 ribonucleoproteins. Genome Res. 24, 1012–1019 (2014).

    CAS  PubMed  PubMed Central  Google Scholar 

  33. 33.

    Hendel, A. et al. Chemically modified guide RNAs enhance CRISPR-Cas genome editing in human primary cells. Nat. Biotechnol. 33, 985–989 (2015).

    CAS  PubMed  PubMed Central  Google Scholar 

  34. 34.

    Curti, B. D. et al. Phase I trial of anti-CD3-stimulated CD4+ T cells, infusional interleukin-2, and cyclophosphamide in patients with advanced cancer. J. Clin. Oncol. 16, 2752–2760 (1998).

    CAS  PubMed  Google Scholar 

  35. 35.

    Edge, S. B. et al. AJCC Cancer Staging Manual 7th edn (Springer, 2010).

  36. 36.

    Li, H. & Durbin, R. Fast and accurate short read alignment with Burrows–Wheeler transform. Bioinformatics 25, 1754–1760 (2009).

    CAS  PubMed  PubMed Central  Google Scholar 

  37. 37.

    Tarasov, A., Vilella, A. J., Cuppen, E., Nijman, I. J. & Prins, P. Sambamba: fast processing of NGS alignment formats. Bioinformatics 31, 2032–2034 (2015).

    CAS  PubMed  PubMed Central  Google Scholar 

  38. 38.

    Kim, S. et al. Strelka2: fast and accurate calling of germline and somatic variants. Nat. Methods 15, 591–594 (2018).

    CAS  PubMed  Google Scholar 

  39. 39.

    Zhou, H. et al. Whole genome analysis of CRISPR Cas9 sgRNA off-target homologies via an efficient computational algorithm. BMC Genomics 18, 826 (2017).

    PubMed  PubMed Central  Google Scholar 

  40. 40.

    Kim, D. et al. Digenome-seq: genome-wide profiling of CRISPR-Cas9 off-target effects in human cells. Nat. Methods 12, 237–243 (2015).

    CAS  PubMed  Google Scholar 

  41. 41.

    Wang, Z. et al. Application of single-molecule amplification and resequencing technology for broad surveillance of plasma mutations in patients with advanced lung adenocarcinoma. J. Mol. Diagn. 19, 169–181 (2017).

    CAS  PubMed  Google Scholar 

  42. 42.

    Shen, B. et al. Efficient genome modification by CRISPR-Cas9 nickase with minimal off-target effects. Nat. Methods 11, 399–402 (2014).

    CAS  PubMed  Google Scholar 

  43. 43.

    Jia, Q. et al. Local mutational diversity drives intratumoral immune heterogeneity in non-small cell lung cancer. Nat. Commun. 9, 5361 (2018).

    CAS  PubMed  PubMed Central  Google Scholar 

Download references


This clinical trial was supported by the 1.3.5 Project for Disciplines of Excellence, West China Hospital, Sichuan University (grant no. ZYJC18001); the West China Hospital Foundation of New Technology (grant nos. XJS2016003 and 190160012); the Sichuan Cancer Society Foundation (grant no. SCS-KT001); the National Science and Technology Major Project (grant no. 2017ZX09304023); and the National Natural Science Foundation of China (grant no. 81672982). We thank all of the study participants, H. Wakelee and G. P. Gao for providing insightful advice on this study, J.S. Kim for advice on the off-target effects, J.Y. Li and the nursing team for clinical care, Q. Lu for data collection, Q. Zhang for clinical ECG diagnosis, L. Wang for supporting preclinical study, J. Jiang for data and safety monitoring, M. Zhao for data management and S. Wang for statistical support.

Author information




Y. Lu, J.X., L.D. and T.M. were involved in study design. Y. Lu and T.D. contributed to study concepts. T.D., K.Y. and Y. Zeng were responsible for manufacturing of therapeutic cells. X. Zhou, M.H., R.T., Z.D., Y.G., J.Z., Yongsheng Wang, L.L., Y. Zhang, Y. Liu, B.Z., M.Y., L.Z., Y. Li, Q. Z. and B.Y. were involved in data acquisition. Y. Lu, Yu Wang, H.S. and M.L were involved in quality control of data and algorithms. J.X., X. Zhou, X.Y., J.S., J.L., Yuqi Wang, X.S., W.W., X. Zhang, L.Y., X.X. and C.C. were involved in data analysis and interpretation. Yu Wang and H.S. contributed to statistical analysis. Y. Lu, J.X., R.T. and T.M. wrote the manuscript. Yuquan Wei and W.L. were involved in administrative support and supervision. All authors approved the article for submission and publication.

Corresponding author

Correspondence to You Lu.

Ethics declarations

Competing interests

The authors declare no competing interests.

Additional information

Peer review information Saheli Sadanand was the primary editor on this article and managed its editorial process and peer review in collaboration with the rest of the editorial team.

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Extended data

Extended Data Fig. 1 T7E1 cleavage assay, TA cloning sequencing of PD-1 disruption in cultured T cells.

a, DNA amplified from edited or unedited T cells were subjected to T7E1 cleavage assay. T cells from a healthy person served as a control. The blue arrow indicates the expected bands for uncut (no mismatch) PD-1; the red arrow, expected bands from the T7E1 assay. Marker, DL2000 Marker (Innova GENE Biosciences, Ontario, Canada). b, The efficiency of PD-1 editing was analyzed by TA cloning on day 21 after electroporation. Source data

Extended Data Fig. 2 Long-term effects of PD-1 disruption in cultured T cells in Patient B-01 and C-02.

Viability of PD-1 disruption in long-term cultured T cells. Compared to the rapid decrease in viability of unedited T cells after day 30, the viability of edited T cells was over 90% and remained high until day 40. Total cell numbers of PD-1 disruption in long-term cultured T cells. The numbers of edited T cells increased slowly until around day 30, reflecting delayed proliferation likely due to the electroporation procedure; after day 30, the numbers of edited T cells increased rapidly. By contrast, the numbers of unedited cells, decreased rapidly after day 30. Source data

Extended Data Fig. 3 Off-target analysis by next generation sequencing (NGS).

Characteristics of on-target and off-target mutation types, frequencies and numbers determined by next-generation sequencing (NGS) for the edited T cells of 7 patients prior to the second cycle of infusion. Bar graph and pie graph above represent the types, the numbers and the composition of off-target mutation, color-coded according to the legend in the top-right corner. Intergenic (44.4%) and intronic (39.1%) mutations composed the majority proportion. Heatmap shows the mutation number of predicted off-target sites (18 off-target sites, OT1-18) and on-target site for individuals. Bar graph on right represents mean mutation frequencies of each site among the 7 patients. The mutation frequencies at these off-target sites and the on-target site were 0.05% (range 0.00–0.22%) and 4.09%, respectively. The modification ratio of on-target/off-target was 105.2. Pie graph on bottom-right shows the composition of on-target mutation. The mutation types of on-target consisted of frameshift or nonframeshift (deletion/insertion mutation), and stopgain, while the vast majority was the deletion mutation (88.5%). Data in bar graph are shown as mean ± s.d. Source data

Extended Data Fig. 4 Electrocardiograph and echocardiography of patient Pre-A-01 during treatment.

a, Electrocardiograph images of patient Pre-A-01 during T-cell therapy. Patient Pre-A-01 had no history of heart disease. The baseline electrocardiograph (before infusion) showed normal results. However, the electrocardiograph on day 1 after the first infusion showed a premature beat; the electrocardiograph on the day 113 showed a premature beat similar to that on day 1. Each image is representative of 3 independent tests. b, Representative images from baseline echocardiography (before infusion, left) and echocardiography conducted on day 26 after the first infusion (right). No cardiac lesion or obvious change in functional parameters was found.

Extended Data Fig. 5

Baseline characteristics of all treated patients.

Extended Data Fig. 6 Duration of treatment-related adverse events.

Different colors are used to represent each patient. Bar length represents duration of the adverse effect. All related AEs were grade 1 or 2. Grade 2 AEs are outlined in black. Source data

Extended Data Fig. 7 Follow-up diagram and data.

a, Follow-up diagram after treatment. b, Summary of treatment-related AEs and severe adverse events (SAEs) during follow-up.

Extended Data Fig. 8 Kaplan-Meier estimates of survival in 12 patients.

a, Overall survival. b, Progression-free survival. Source data

Extended Data Fig. 9 TCR diversity in healthy donors and patients.

Comparison of TCR diversity (Shannon index) in PBMCs of 11 healthy donors (n = 11) and 12 patients with refractory NSCLC (n = 12). Data are shown as median ± 95% confidence interval [CI], each dot represents an individual data. P value was calculated using the two-tailed Wilcoxon rank-sum test. Median of differences was -2.457, 95% CI for difference was −4.096 to −1.067, P = 0.0005. Source data

Extended Data Fig. 10 The immunohistochemistry staining density was semi-quantified by ImageJ software.

The data was compared using one-way Anova with Sidak’s multiple comparisons test (n = 3 per group). Data are shown as mean ± s.d., * P < 0.05, ** P < 0.01, ***P < 0.001 and ****P < 0.0001. Source data

Supplementary information

Supplementary Information

Supplementary Fig. 1, Tables 1–8, and Notes

Reporting Summary

Source data

Source Data Fig. 2

Statistical Source Data

Source Data Fig. 3

Statistical Source Data

Source Data Fig. 4

Statistical Source Data

Source Data Fig. 5

Statistical Source Data

Source Data Extended Data Fig. 1

Unprocessed gels

Source Data Extended Data Fig. 1

Statistical Source Data

Source Data Extended Data Fig. 2

Statistical Source Data

Source Data Extended Data Fig. 3

Statistical Source Data

Source Data Extended Data Fig. 6

Statistical Source Data

Source Data Extended Data Fig. 8

Statistical Source Data

Source Data Extended Data Fig. 9

Statistical Source Data

Source Data Extended Data Fig. 10

Statistical Source Data

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Lu, Y., Xue, J., Deng, T. et al. Safety and feasibility of CRISPR-edited T cells in patients with refractory non-small-cell lung cancer. Nat Med 26, 732–740 (2020).

Download citation

Further reading


Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing