Patient characteristics
The characteristics of patients with EGFR-mutated NSCLC who had clinically progressed after EGFR-TKI treatment are described in Table 2. The median age was 62 years (range, 39–83 years), and thirty-six patients (70.6%) were females. Forty-three out of fifty-one patients had stage IV disease (84.3%). Thirty patients (58.8%) had exon 19 deletion, eighteen patients (35.3%) had L858R point mutation, two patients (3.9%) had S768I point mutation, one patient (2.0%) had L861Q point mutation, and one patient (2.0%) had G719S point mutation. Ten patients (19.6%) received erlotinib therapy, thirteen (25.5%) received afatinib, and twenty-seven (52.9%) received gefitinib therapy. One patient (2.0%) received gefitinib and erlotinib therapy at different time points. The median months of time period from the start of TKI to the sample collection for EGFR testing was 17.5 months (range, 2–72 months).
Table 2
Characteristics | | No. of Patients |
n = 51 (100%) |
Age, median (range), years | | 62 (39–83) |
Sex | | | |
Female | | | 36 (70.6%) |
Male | | | 15 (29.4%) |
Histologic type | | |
Adenocarcinoma | | 50 (98.0%) |
Squamous cell carcinoma | 1 (2.0%) |
Tumor stage | | |
IB | | | 2 (3.9%) |
IIIA | | | 3 (5.9%) |
IIIB | | | 3 (5.9%) |
IVA | | | 21 (41.2%) |
IVB | | | 22 (43.1%) |
M category* | | |
M1a | | | 13 (25.5%) |
M1b | | | 10 (19.6%) |
M1c | | | 27 (52.9%) |
M1a + M1c† | | 1 (2.0%) |
Tissue EGFR genotyping | |
Exon 19 deletion | | 29 (56.9%) |
L858R | | | 17 (33.3%) |
S768I | | | 2 (3.9%) |
L861Q | | | 1 (2.0%) |
G719S | | | 1 (2.0%) |
Exon 19 deletion + L858R‡ | 1 (2.0%) |
Previous EGFR-TKI therapy | |
Erlotinib | | | 10 (19.6%) |
Afatinib | | | 13 (25.5%) |
Gefitinib | | | 27 (52.9%) |
>1 EGFR-TKIs | | 1 (2.0%) |
* The information of M category was reclassified at the time of EGFR testing. M category was based on the 8th TMN edition. M1a: lung metastases or pleural/pericardial malignant effusion or nodules; M1b: a single metastatic lesion in a single distant organ; M1c: multiple lesions in a single organ or multiple lesions in multiple organs |
†The M category of patient G (Supplemental Table 7) was M1a at first EGFR testing. M stage was reclassified to M1c at second EGFR testing. |
‡1 patient had both exon 19 deletion and L858R |
Abbreviation: TKI, tyrosine kinase inhibitor |
Validation Of Crispr-cppc
The analytical sensitivity of CRISPR-CPPC was evaluated using the Multiplex I cfDNA Reference Standard with allele frequencies of 5%, 1%, and 0.1% (Horizon Discovery, Cambridge, United Kingdom). The expected copy number of mutant alleles (3–109 copies) and the actual copy number of mutant alleles observed in these samples are presented in Table 3. The positive detection of mutant DNA after CRISPR-CPPC was approximately 2–6 times higher than the expected copies of mutant DNA. After mutant enrichment, the allele frequency was approximately 1.6–3.7 times higher than the expected allele frequency.
Table 3
Analytical sensitivity of CRISPR-CPPC in detecting EGFR T790M mutation
Reference Materials (T790M) | Expected allele frequency (%)* | Expected copies of mutant DNA per sample* | Expected copies of wild-type DNA per sample* | ddPCR after CRISPR-CPPC Detection positive (≥ 6 events/assay) |
Observed Mutant allele frequency (%) | Copies of mutant DNA per sample | Copies of wild-type DNA per sample |
5% Multiplex I cfDNA Reference Standard (HD777), 20 ng/µL | 4.9 | 109 | 2120 | 8.8 | 231 | 2409 |
1% Multiplex I cfDNA Reference Standard (HD778), 20 ng/µL | 1.1 | 24 | 2256 | 1.7 | 60 | 3376 |
0.1% Multiplex I cfDNA Reference Standard (HD779), 20 ng/µL | 0.1 | 3 | 2228 | 0.5 | 19 | 3842 |
* Expected allele frequency and copy number of wild-type and mutant DNA measured using ddPCR were provided by the manufacturer. |
Abbreviations: cfDNA, cell-free DNA; CRISPR-CPPC, CRISPR system combined with post-PCR cfDNA; ddPCR, droplet digital PCR |
A comparison of qPCR, ddPCR, and ddPCR with CRISPR-CPPC assay
Sixty samples from fifty-one patients were analyzed. All samples were subjected to qPCR, ddPCR and CRISPR-CPPC assay for detecting T790M (Supplemental Table 2). Samples that tested positive for T790M through two or more of the experimental methods (qPCR from cfDNA or tissue, NGS, ddPCR, and CRISPR-CPPC) were considered to be true positives (Supplemental Table 3). According to Kim et al., ddPCR without CRISPR-CPPC was considered positive if the measured events were ≥ 2 events/assay and negative if the events within a gated region were < 2 events/assay [30]. Based on the results of multiple assays, the sensitivities of CRISPR-CPPC and ddPCR were 92.0% and 64.0%, respectively (Supplemental Table 3). The PPA (%), NPA (%), and OPA (%) of CRISPR-CPPC and ddPCR compared to the qPCR results are presented in Supplemental Table 4. Compared to qPCR, CRISPR-CPPC and ddPCR showed 100% and 75% PPA, respectively. CRISPR-CPPC detected T790M variants from 15 samples whose T790M mutations were not detected by qPCR. (Supplemental Table 4). When compared to ddPCR, the PPA (%), NPA (%), and OPA (%) was 88.2%, 62.8 and 70.0%, respectively. (Supplemental Table 5). Eighteen samples showed discordant results between CRISPR-CPPC and ddPCR (Supplemental Table 6). CRISPR-CPPC detected T790M mutant alleles in sixteen T790M-negative samples by ddPCR, and ddPCR detected T790M in two T790M-negative samples by CRISPR-CPPC (sample No. 12 & 47). (Supplemental Table 5 & Supplemental Table 6). These two samples underwent further testing by NGS, which showed that one sample was T790M-positive with an allele frequency of 0.2% and the other was T790M-negative. The final clinical diagnosis of clinical progression was made by oncologists based on integration of patients' medical history and radiological findings. The researchers retrospectively reviewed the participant's medical records, including the final clinical diagnosis. Table 4 presents the analytical performance of CRIPSR-CPPC and ddPCR based on the results of multiple assays and final clinical diagnoses. The sensitivity and specificity of CRISPR-CPPC were increased up to 93.9% and 100.0%, respectively.
Table 4
Analytical performance of assays for detecting T790M mutation depending on clinical diagnosis
Method | T790M mutation was confirmed with multiple studies and or/and clinical diagnosis* | Sensitivity (95% CI) | Specificity (95% CI) | Accuracy (95% CI) |
Results | Pos (n = 33) | Neg (n = 27) |
ddPCR | Pos | 16 | 1 | 48.5% (30.8–66.5%) | 96.3% (81.0–99.9%) | 70.0% (56.8–81.2%) |
Neg | 17 | 26 |
CRISPR-CPPC | Pos | 31 | 0 | 93.9% (79.8–99.3%) | 100.0% (87.2–100.0%) | 96.7% (88.5–99.6%) |
Neg | 2 | 27 |
*T790M detected by more than one method (qPCR from cfDNA or tissue, NGS, ddPCR, CRISPR-CPPC) simultaneously is considered “true positive.” “Clinical diagnosis-T790M-positive” was defined when clinical history and image interpretation supported that a positive T790M result would be close to a true positive. Image interpretation was performed only for CRISPR-CPPC-positive samples. |
Abbreviations: Pos, positive; Neg, negative; CI, confidence interval; CRISPR-CPPC, CRISPR system combined with post-PCR cfDNA; ddPCR, droplet digital PCR; qPCR, real-time PCR; NGS, next-generation sequencing |
Ultra-sensitive Detection Of Crispr-cppc
A comparison of the allele frequency and positive calls of sixty samples is shown in Supplemental Table 2. Most samples showed approximately 1.2–13-times higher allele frequencies with the use of CRISPR-CPPC. In addition, approximately 1.6–562-times more positive calls were detected with the use of CRISPR-CPPC. The copy number comparison between pairs was statistically significant, with a P-value of < 0.0001, using the Wilcoxon signed-rank test.
We evaluated the performance of CRISPR-CPPC using the samples containing low copies of T790M mutant alleles from patients with EGFR-mutated NSCLC who had clinically progressed after EGFR-TKI treatment. The distribution of T790M copies according to detecting assays was depicted in Supplemental Fig. 3. The overall T790M positive copy number differences between CRISPR-CPPC and ddPCR are shown in Fig. 2. Figure 2 shows the trend of CRISPR-CPPC increasing the T790M positive copy numbers compared to ddPCR without CRISPR-CPPC, except for sample number 47. Among 51 samples with ≤ 10 copies of T790M alleles based on ddPCR, the positive T790M rate of ddPCR and CRISPR-CPPC was 15.7% (n = 8 / 51) and 45.1% (n = 23 / 51), respectively (Supplemental Fig. 3 & Supplemental Table 2). CRISPR-CPPC showed improved sensitivity of detecting T790M, notably in samples with T790M-low copies or T790M-negative by ddPCR.
The monitoring EGFR T790M in patient samples using CRISPR-CPPC
Among the 51 patients, eight patients had one or two follow-up EGFR mutation tests using the Roche cobas® EGFR Mutation Test v2. As shown in Supplemental Table 7, patients E, G, and H had a follow-up test to detect T790M using CRISPR-CPPC, but qPCR was unable to detect T790M. Using ddPCR, 0, 3, and 0 positive calls with a respective allele frequency (%) of 0, 0.3, and 0 were detected. Using CRISPR-CPPC in the first sample from patient H, the T790M variant was detected with six positive calls with an allele frequency (%) of 0.1. In the second sample from patient E and patient G, T790M was detected with eight and nine positive calls and an allele frequency (%) of 0.2 and 0.3, respectively. These results indicate that only CRISPR-CPPC could detect exceptionally low copies of T790M in patient samples.