Development and Interlaboratory Validation of a Simple Screening

Mar 24, 2016 - Copyright © 2016 American Chemical Society. *Phone: +81-3-3700-9397. Fax: +81-3-3700-7438. E-mail: [email protected]. Cite this:Anal...
0 downloads 0 Views 794KB Size
Download PDF
Article pubs.acs.org/ac

Development and Interlaboratory Validation of a Simple Screening Method for Genetically Modified Maize Using a ΔΔCq‑Based Multiplex Real-Time PCR Assay Akio Noguchi,† Kosuke Nakamura,† Kozue Sakata,† Nozomi Sato-Fukuda,† Takumi Ishigaki,† Junichi Mano,‡ Reona Takabatake,‡ Kazumi Kitta,‡ Reiko Teshima,† Kazunari Kondo,*,† and Tomoko Nishimaki-Mogami† †

National Institute of Health Sciences, 1-18-1 Kamiyoga, Setagaya-ku, Tokyo 158-8501, Japan National Food Research Institute, National Agriculture and Food Research Organization, 2-1-12 Kannondai, Tsukuba 305-8642, Japan

‡

S Supporting Information *

ABSTRACT: A number of genetically modified (GM) maize events have been developed and approved worldwide for commercial cultivation. A screening method is needed to monitor GM maize approved for commercialization in countries that mandate the labeling of foods containing a specified threshold level of GM crops. In Japan, a screening method has been implemented to monitor approved GM maize since 2001. However, the screening method currently used in Japan is time-consuming and requires generation of a calibration curve and experimental conversion factor (Cf) value. We developed a simple screening method that avoids the need for a calibration curve and Cf value. In this method, ΔCq values between the target sequences and the endogenous gene are calculated using multiplex real-time PCR, and the ΔΔCq value between the analytical and control samples is used as the criterion for determining analytical samples in which the GM organism content is below the threshold level for labeling of GM crops. An interlaboratory study indicated that the method is applicable independently with at least two models of PCR instruments used in this study.

A

since 2001.4 In several countries, including Japan, a quantitative screening method for GM foods requires construction of a calibration curve with standard plasmid DNA. The experimental conversion factor (Cf) value, which is the ratio of the copy number of a target sequence to that of an endogenous gene in GM seeds, must be calculated for each PCR instrument to convert the ratio to the weight/weight percentage of GM material. Moreover, recent GM maize events that lack the P35S sequence in the genome, such as MIR162, MIR604, 3272 and 5307, are not detected by the quantitative screening method currently used in Japan. MIR604 kernels were detected in nonidentity preserved maize samples produced in 2009 in the USA using an individual kernel detection system that involves event-specific analysis using extracted DNA from individual ground maize kernels but not using the qualitative screening method targeting P35S and GA21.5 A group testing method statistically evaluates GM organism (GMO) content based on qualitative PCR analysis for multiple small samples, consisting

n increasing number of genetically modified (GM) crops have been developed using recombinant DNA technology. These crops are widely cultivated as sources of food and feed in many countries.1 GM crops are generally approved for commercialization following safety evaluations. However, the use of GM crops for food remains controversial among consumers in many countries. The labeling of GM foods allows consumers to make informed food choices. Therefore, many countries mandate the labeling of foods containing a specified threshold level of GM crops (0.9% in the European Union and Russia, 1% in Australia and New Zealand, 3% in Korea, and 5% in Japan).2 In recent years, a number of GM maize events have been developed and approved for cultivation as compared with other major GM crops such as soybean, cotton, and canola. In Japan, 24 GM maize events and 168 varieties of stacked GM maize, combining two or more GM events, have been approved as of 2015.3 To monitor approved GM maize in Japan, a quantitative screening method targeting the Caulif lower mosaic virus 35S promoter (P35S) and GA21 has been implemented as an official method before the specific quantitative method for five GM maize events (Bt11, event176, GA21, MON810, and T25) © XXXX American Chemical Society

Received: November 16, 2015 Accepted: March 24, 2016

A

DOI: 10.1021/acs.analchem.5b04335 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

Figure 1. Flowchart of the procedure for the screening method. The ΔCq values of the analytical sample (ΔCq(AN)) and the control sample (ΔCq(CN)), in which the GMO content is unequivocally judged to be less than the threshold level for labeling of GM crops (5% in Japan), are determined by multiplex real-time PCR targeting the endogenous gene (SSIIb) and the target sequences (P35S and TNOS). The ΔΔCq value is then calculated. In the case that the GMO contents of the analytical samples A and B are distinctly less than that of the control sample, the GMO contents of the analytical samples F and G exceed 5%, and the GMO contents of the analytical samples C, D, and E are almost equal to that of the control sample, the assessment is as follows. The GMO contents of the analytical samples A and B are judged to be less than 5% because the ΔΔCq values of these samples are positive. The GMO contents of the analytical samples F and G are judged to have the potential to exceed 5% because the ΔΔCq values of these samples are negative. The GMO contents of the analytical samples C, D, and E are judged to be less than 5% (C and D) and have the potential to exceed 5% (E), respectively, because the ΔΔCq values of samples C and D are positive and zero, respectively, whereas that of sample E is negative. Given that the ΔΔCq values of the analytical samples C, D, and E are within or near the measurement uncertainty of that of the control sample, the judgments for these samples will vary among independent measurements but will not underestimate the GMO content.

of 20 maize kernels.6 To avoid nondetection of GM maize events that do not contain the P35S sequence, qualitative PCR is performed targeting P35S and the Agrobacterium nopaline synthase terminator (TNOS) contained in a number of GM maize events. However, these methods are time-consuming and/or require additional equipment. The comparative Cq (quantification cycle) method (ΔΔCq method), a convenient method with which to determine the differences in concentrations of the target genes between samples, does not require a calibration curve and Cf value.7 The difference in Cq values between the target and endogenous genes (ΔCq value) is calculated, and then the ΔCq value of the analytical sample (ΔCq(AN)) is compared with that of the control sample (ΔCq(CN)), of which the GMO content is unequivocally judged to be below the threshold level for labeling of GM crops (5% in Japan) (Figure 1), and the two genes must be amplified with the same PCR efficiency. If the difference in ΔCq value between the analytical sample and the control sample (ΔΔCq) value is positive or zero, the GMO content of the analytical sample is judged to be below 5%. However, if the ΔΔCq value is negative, the GMO content of the analytical sample has the potential to exceed 5%. The multiplex PCR method enables simultaneous amplification of two or more targets and is suitable to examine a number of analytical samples. Screening for GM maize incorporating multiplex PCR methods is expected to be simple and time-

saving. Several screening methods using multiplex PCR have been reported.8−11 However, these are qualitative methods or require a calibration curve to quantify GMO content. In this study, we developed a simple screening method targeting P35S and TNOS using ΔΔCq and multiplex PCR and then validated the method in an interlaboratory study. In this method, the ΔCq value between the target sequences (P35S and TNOS) and the endogenous gene (maize starch synthase IIb, SSIIb) is calculated using multiplex real-time PCR, and the ΔΔCq value between the analytical sample and the control sample is used as the criterion for determining analytical samples in which GMO content is below 5% (Figure 1). The GMO content of the control sample analyzed together with the analytical sample is unequivocally judged to be below the threshold level for labeling of GM crops (5% in Japan).

■

EXPERIMENTAL SECTION Maize Materials. Non-GM maize flour, 10% MON810 flour, 10% TC1507 flour, 10% 3272 flour, and 10% DAS59122 flour were purchased from Sigma-Aldrich (St. Louis, MO). MIR162 flour was purchased from the American Oil Chemists’ Society (Urbana, IL). DNA Extraction. Genomic DNA was extracted and purified from 1 g of maize flour using the DNeasy Plant Maxi Kit (QIAGEN, Hilden, Germany) as described previously.12 DNA concentrations were determined by measuring ultraviolet (UV) B

DOI: 10.1021/acs.analchem.5b04335 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

Optimization of Multiplex Real-Time PCR Assay. The concentration of primers and probes in the multiplex real-time PCR assay was optimized using the respective plasmid DNAs, pUC-P35S, or pUC-TNOS, diluted to 15, 75, 150, 375, and 750 copies/μL with 7 500 copies/μL of pUC-SSIIb. The concentration of primers and probes used in the optimization trial was 0.08−0.50 μM of SSIIb 3-5′ and SSIIb3-3′, 0.25−0.50 μM of P35S 1-5′ and P35S 1-3′, 0.30−0.50 μM of NOS ter 3-5′ and NOS ter 2-3′, 0.08−0.20 μM of SSIIb-TaqV, 0.10−0.20 μM of P35S-Taq, and 0.12−0.20 μM of NOS-Taq. Evaluation of the Screening Method. The PCR efficiency (E; %) for the target sequences was evaluated using the respective plasmid DNAs diluted to 50, 200, 1 000, 2 000, and 10 000 copies/μL according to the following formula:

absorption at 260 nm with a ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE). Samples were diluted to 20 ng/μL with sterile distilled water and stored at −30 °C until use. Multiplex Real-Time PCR Assay. For the screening of GM maize, multiplex real-time PCR assays were performed for detection of the endogenous gene encoding SSIIb and P35S and TNOS, which are widely introduced into GM maize events, using the ABI PRISM 7900HT Sequence Detection System (7900HT) (Thermo Fisher Scientific, Waltham, MA) and LightCycler 96 System (LC96) (Roche Diagnostics, Basel, Switzerland). The SSIIb-3 system (SSIIb 3-5′ and SSIIb 3-3′ with SSIIb-TaqV)13 and P35S-1 system (P35S 1-5′ and P35S 13′ with P35S-Taq)14 were used as the primer pairs and probes for detection of SSIIb and P35S, respectively, as previously described. For detection of TNOS, NOS ter 2-3′ and NOS-Taq were used as the antisense primer and probe, respectively, as previously described14 and NOS ter 3-5′ (5′-GCATGTAATAATTAACATGTAATGCATGAC-3′) was designed in this study as the sense primer. The oligonucleotide DNAs for primers and probes, except for SSIIb-TaqV, were purchased from Eurofins Genomics (Ebersberg, Germany). SSIIb-TaqV was purchased from Thermo Fisher Scientific. P35S-Taq and NOS-Taq were labeled with 6-carboxyfluorescein (FAM) and 6-carboxytetramethyl-rhodamine (TAMRA) at the 5′ and 3′ ends, respectively. For SSIIb-TaqV, 2′-chloro-7′-phenyl-1,4dichloro-6-carboxyfluorescein (VIC) was used instead of FAM. The 10 μL reaction mixture contained 1 μL of template DNA, 5 μL of FastStart Universal Probe Master (ROX) (Roche Diagnostics), 0.08 μM SSIIb 3-5′ and SSIIb 3-3′, 0.25 μM P35S 1-5′ and P35S 1-3′, 0.3 μM NOS ter 3-5′ and NOS ter 2-3′, 0.08 μM SSIIb-TaqV, 0.1 μM P35S-Taq, and 0.12 μM NOSTaq. The PCR protocol was as follows: 2 min at 50 °C, 10 min at 95 °C followed by 40 cycles of 30 s at 95 °C and 1.5 min at 59 °C. Each sample reaction was conducted in triplicate. Data Analysis. The data obtained with the 7900HT were analyzed using SDS2.1 software (Thermo Fisher Scientific). The baseline was set to cycles 3 through 15 and the ΔRn threshold for plotting Cq values was set to 0.2 during exponential amplification. Data obtained with the LC96 were analyzed using LC96 software (Roche Diagnostics). The difference in Cq value between P35S+TNOS, corresponding to the sum of P35S and TNOS detection, and SSIIb (ΔCq value) was calculated using the following formula: ΔCq = Cq(P35S + TNOS) − Cq(SSIIb)

E(%) = [10(−1/slope) − 1] × 100

(2)

where slope is calculated by linear regression analysis and theoretical PCR amplification is performed at 100% PCR efficiency. To evaluate the equivalency between the PCR assays for P35S and TNOS detection, we used the respective plasmid DNAs diluted to 15, 75, 150, 375, and 750 copies/μL with 7 500 copies/μL of pUC-SSIIb, and the genomic DNAs (20 ng/μL) diluted to 0.15, 0.5, 2, 5, and 10% of MON810, TC1507, DAS59122, MIR162, or 3272 with 100% of non-GM maize DNA (n = 3), and ΔCq values were calculated. To evaluate the sensitivity of the PCR assays for P35S and TNOS detection, we analyzed the respective plasmid DNAs diluted to 0.1, 2.5, 5, and 10 copies/μL with 7 500 copies/μL of pUCSSIIb (n = 21) and counted the number of positive results. Preparation of the Test Samples. MON810 or MIR162 flour was mixed with non-GM maize flour at 1, 2, 3, 4 and 5% (w/w) on a 1-g scale. Genomic DNA was extracted from the mixed flours or non-GM maize flour as described previously.12 The extracted DNA samples were diluted to 20 ng/μL with sterile distilled water. Ten microliters of each test sample was placed in a microtube and stored at −30 °C until use. Before dispatching to the laboratories, the homogeneity of the test samples was verified according to the procedure described in the International Harmonized Protocol for Proficiency Testing of Analytical Laboratories,15 except that the number of test samples was 10. As duplicate reactions for each test sample except the non-GM maize sample, 20 reactions in total were analyzed using the 7900HT and LC96. The copy number ratio of P35S+TNOS to SSIIb (CNR(P35S+TNOS)/SSIIb) was calculated using the following formula:

(1)

CNR (P35S + TNOS)/SSIIb = 2−ΔCq

where Cq(P35S+TNOS) is the Cq value of P35S+TNOS, and Cq(SSIIb) is the Cq value of SSIIb. Preparation of Plasmid DNAs. To evaluate the performance of the multiplex real-time PCR method for screening of GM maize, we prepared plasmid DNAs. Each DNA fragment (SSIIb, P35S, or TNOS) was amplified by PCR using each of the primer pairs and ligated into the pUC19 plasmid. The respective plasmids were designated pUC-SSIIb, pUC-P35S, or pUC-TNOS. Escherichia coli DH5α cells were transformed using these plasmids. The plasmids were extracted with the GeneElute Plasmid Miniprep Kit (Sigma-Aldrich). The purified plasmids were excised with NdeI. The resultant linearized plasmids were purified by phenol/chloroform extraction and ethanol precipitation. The purified plasmid was quantified by singleplex real-time PCR (Supporting Information, Experimental Section) and then diluted to the required concentration with tris/ethylenediaminetetraacetic acid buffer.

(3)

where PCR efficiency of each reaction was hypothesized to be 100% and subjected to one-way analysis of variance (ANOVA). Interlaboratory Validation. Interlaboratory validation of the developed screening method, conducted with the participation of 11 and 4 laboratories for the 7900HT and LC96, respectively, was organized by the National Institute of Health Sciences. Each laboratory received one tube with sterile distilled water as the nontemplate control, 22 tubes with the blind samples including two tubes with individual GM maize DNA samples, and two tubes with the non-GM maize DNA sample, the primers, and probes, and FastStart Universal Probe Master (ROX) used for multiplex real-time PCR as described above, as well as the experimental protocol from the National Institute of Health Sciences. The participants carried out multiplex real-time PCR for each sample and submitted the C

DOI: 10.1021/acs.analchem.5b04335 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

Table 1. Slope Values, 95% Confidence Interval (CI95%) of the Slope Value, Intercept Values, CI95% of the Intercept Value, and Coefficient of Determination (R2) Values of the Regression Lines Generated Using the Plasmid DNAs (P35S and TNOS) Diluted with pUC-SSIIb under Four PCR Conditions slope value PCR condition 7900HT 1 2 3 4 LC96 1 2 3 4

CI95% of slope value P35S

TNOS

intercept value

CI95% of intercept value

P35S

TNOS

P35S

P35S

TNOS

−3.37 −3.71 −3.59 −3.41

−3.92 −3.42 −3.27 −3.31

(−3.88, (−4.09, (−3.92, (−3.97,

−2.86) −3.34) −3.26) −2.85)

(−4.36, (−3.91, (−3.71, (−3.81,

−3.48) −2.93) −2.82) −2.82)

10.92 11.44 11.45 11.73

12.96 11.50 11.31 11.60

(9.79, 12.05) (10.61, 12.27) (10.72, 12.18) (10.49, 12.97)

(11.99, (10.41, (10.32, (10.49,

−3.50 −3.11 −3.69 −3.45

−4.24 −4.04 −3.50 −3.43

(−3.70, (−3.18, (−4.15, (−3.61,

−3.31) −3.03) −3.22) −3.29)

(−5.04, (−4.40, (−4.04, (−3.61,

−3.44) −3.68) −2.96) −3.26)

12.03 10.97 13.02 13.52

14.61 13.92 13.05 13.58

(11.61, (10.81, (11.99, (13.17,

(12.84, (13.12, (11.85, (13.19,

12.46) 11.13) 14.04) 13.87)

TNOS

R2 P35S

TNOS

13.93) 12.59) 12.30) 12.70)

0.998 0.997 0.993 0.992

0.995 0.994 0.996 0.993

16.38) 14.71) 14.25) 13.97)

0.999 0.999 0.995 0.999

0.990 0.998 0.993 0.999

Figure 2. Regression lines of the dilution series of the plasmid DNAs or the genomic DNAs. The ΔCq values were plotted against log copy number of the plasmid DNAs diluted with pUC-SSIIb using 7900HT (A) and LC96 (B) PCR instruments. The Cq values were plotted against log copy number of the plasmid DNAs using 7900HT (C) and LC96 (D). The ΔCq values were plotted against log GMO content of five GM maize events using 7900HT (E) and LC96 (F). The analysis was performed in triplicate per sample.

data to the National Institute of Health Sciences. Using the 1, 2, 3 or 4% sample as the control sample, the difference in ΔCq value between the 1, 2, 3, or 4% and 5% samples (ΔΔCq(1,2,3or4%)) was calculated from the submitted data using the following formula: ΔΔCq(1,2,3or4%) = ΔCq(5%) − ΔCq(1,2,3or4%)

The GMO content ratio (GCR) between the 1, 2, 3, or 4% and 5% samples was calculated from the submitted data using the following formula: GCR = 2−ΔΔCq(1,2,3or4%)

(4)

(5)

where the PCR efficiency of each reaction was hypothesized to be 100%. All submitted data were analyzed with Cochran’s test16 and Grubbs’s test17−19 to exclude outlier laboratories as

where ΔCq(1,2,3or4%) is ΔCq value of the 1, 2, 3, or 4% sample, respectively, and ΔCq(5%) is the ΔCq value of the 5% sample. D

DOI: 10.1021/acs.analchem.5b04335 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry described in the guidelines,20,21 and then the submitted data were statistically analyzed.

concentrations of the primers and probes for P35S were decreased to 50% (PCR condition 4). Under this condition, no difference between P35S and TNOS in ΔCq values (Figure 2A,B) and in the CI95% of the slope value (7900HT, −3.97 to −2.86 for P35S, −3.81 to −2.82 for TNOS; LC96, −3.61 to −3.29 for P35S, −3.61 to −3.26 for TNOS) and in the CI95% of the intercept value (7900HT, 10.49−12.97 for P35S, 10.49− 12.70 for TNOS; LC96, 13.17−13.84 for P35S, 13.19−13.97 for TNOS) was observed for both PCR instruments (Table 1). Furthermore, the regression lines showed high linearity (R2 = 0.990−0.999) (Table 1). Consequently, we decided to perform the multiplex real-time PCR assay using PCR condition 4 (0.08 μM of SSIIb 3-5′ and SSIIb 3-3′, 0.25 μM of P35S 1-5′ and P35S 1-3′, 0.3 μM of NOS ter 3-5′ and NOS ter 2-3′, 0.08 μM of SSIIb-TaqV, 0.1 μM of P35S-Taq, and 0.12 μM of NOSTaq). Evaluation of Multiplex Real-Time PCR Assay. The analytical performance of the optimized multiplex real-time PCR assay, using PCR condition 4, was evaluated using the plasmid DNAs containing a single target sequence for SSIIb, P35S, and TNOS and using the genomic DNAs extracted from MON810, TC1507, and DAS59122 (each containing a single genomic copy of P35S) and MIR162 and 3272 (each containing a single genomic copy of TNOS). In the ΔΔCq method, the target and endogenous genes must be amplified with the same PCR efficiency. The PCR efficiency for SSIIb, P35S, and TNOS calculated from the regression lines of the dilution series of each plasmid DNA were 97.6%, 95.1%, and 98.4%, respectively, for the 7900HT (Figure 2C and Table 2)

■

RESULTS AND DISCUSSION Design of Multiplex Real-time PCR Assay. To detect GM maize events that lack the P35S sequence in the genome, we used the TNOS sequence in addition to the P35S sequence. To simplify the procedure, ΔΔCq and multiplex real-time PCR were used. This method enabled simultaneous calculation of the Cq value corresponding to the sum of P35S and TNOS detection (Cq(P35S+TNOS)) using FAM-labeled P35S and TNOS and the Cq value for SSIIb (Cq(SSIIb)) using VIC-labeled SSIIb. The ΔCq value was calculated from these Cq values, and then the ΔΔCq value between the analytical sample and the control sample, of which the GMO content is unequivocally judged to be less than 5%, was used as the criterion for determining analytical samples in which GMO content was less than 5% (Figure 1). The control sample was analyzed simultaneously with the analytical sample. To optimize the concentration of primer pairs and probes in the multiplex real-time PCR assay, the respective plasmid DNAs, pUC-P35S, or pUC-TNOS, diluted to 15, 75, 150, 375, and 750 copies/μL with 7 500 copies/μL of pUC-SSIIb were used as the template for the multiplex real-time PCR using the 7900HT and LC96, and then the calculated ΔCq values were plotted against log copy number of the plasmid DNAs to generate regression lines. First, multiplex real-time PCR was carried out using 0.5 μM of each primer pair and 0.2 μM of each probe in the general singleplex real-time PCR (PCR condition 1) to determine the differences between P35S and TNOS in the 95% confidence interval (CI95%) of the slope value (7900HT, − 3.88 to −2.86 for P35S, − 4.36 to −3.48 for TNOS; LC96, − 3.70 to −3.31 for P35S, − 5.04 to −3.44 for TNOS) and in the CI95% of the intercept value (7900HT, 9.79−12.05 for P35S, 11.99−13.93 for TNOS; LC96, 11.61− 12.46 for P35S, 12.84−16.38 for TNOS) (Table 1 and Supporting Information, Figure S1A,B). Although the concentrations of the respective primers and probes were decreased to 60% (PCR condition 2), differences between P35S and TNOS in the CI95% of the slope value (−3.18 to −3.03 for P35S, − 4.40 to −3.68 for TNOS) and in the CI95% of the intercept value (10.81−11.13 for P35S, 13.12−14.71 for TNOS) were observed with the LC96 (Table 1 and Supporting Information, Figure S1D). In contrast, with 7900HT no difference between P35S and TNOS in the CI95% of the slope value (−4.09 to −3.34 for P35S, −3.91 to −2.93 for TNOS) and in the CI95% of the intercept value (10.61−12.27 for P35S, 10.41−12.59 for TNOS) was observed (Table 1 and Supporting Information, Figure S1C). In many reports of multiplex PCR, the concentrations of the primer pairs and probes for the endogenous gene are lower than those for the target gene.8,9,22 Therefore, the concentrations of the primers and probes for SSIIb were decreased to 16% and 40%, respectively (PCR condition 3). Subsequently, no difference between P35S and TNOS in the CI95% of the slope value (7900HT, −3.92 to −3.26 for P35S, −3.71 to −2.82 for TNOS; LC96, −4.15 to −3.22 for P35S, −4.04 to −2.96 for TNOS) and in the CI95% of the intercept value (7900HT, 10.72−12.18 for P35S, 10.32− 12.30 for TNOS; LC96, 11.99−14.04 for P35S, 11.85−14.25 for TNOS) was observed for both PCR instruments (Table 1 and Supporting Information, Figure S1E,F). Given that ΔCq values for P35S were slightly lower than those of TNOS in both PCR instruments (Supporting Information, Figure S1A−F), the

Table 2. PCR Efficiency (E), 95% Confidence Interval (CI95%) of E, and Coefficient of Determination (R2) Values of the Regression Lines Generated Using the Plasmid DNAs target 7900HT SSIIb P35S TNOS LC96 SSIIb P35S TNOS

E (%)

CI95% of E (%)

R2

97.6 95.2 98.4

(82.2, 120.0) (79.3, 118.5) (91.6, 106.4)

0.995 0.994 0.999

95.0 95.3 98.5

(84.7, 108.1) (90.7, 100.3) (91.6, 106.6)

0.998 0.999 0.999

and 95.0%, 95.3%, and 98.5%, respectively, for the LC96 (Figure 2D and Table 2). The CI95% of PCR efficiency for SSIIb, P35S, and TNOS was 82.2−120.0%, 79.3−118.5%, and 91.6−106.4%, respectively, for the 7900HT and 84.7−108.1%, 90.7−100.3%, and 91.6−106.6%, respectively, for the LC96 (Table 2). Furthermore, the regression lines showed high linearity (R2 = 0.990−0.999) (Table 1). These results indicated that theoretical PCR amplifications were performed for these sequences with almost equal PCR efficiency in this multiplex real-time PCR assay. To evaluate the equivalency between the PCR assays for different GM events, ΔCq values were calculated using the genomic DNAs diluted to 0.15, 0.5, 2, 5, and 10% of each of five GM events with 100% of non-GM maize DNA, and then the calculated ΔCq values were plotted against log GMO content of the genomic DNAs to generate regression lines. The regression lines for the five GM events showed high equivalency in the CI95% of the slope (7900HT, −3.80 to −2.80 for MON810, −3.70 to −3.00 for TC1507, −3.45 to −3.11 for DAS59122, −3.58 to −3.04 for MIR162, −3.67 to E

DOI: 10.1021/acs.analchem.5b04335 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

Table 3. Slope Values, 95% Confidence Interval (CI95%) of the Slope Value, Intercept Values, CI95% of the Intercept Value and Coefficient of Determination (R2) Values of the Regression Lines Generated Using the Genomic DNAs of Five GM Maize Events Diluted with Genomic DNA of 100% Non-GM Maize template DNA 7900HT MON810 TC1507 DAS59122 MIR162 3272 LC96 MON810 TC1507 DAS59122 MIR162 3272

slope value

CI95% of slope value

intercept value

CI95% of intercept value

R2

−3.30 −3.35 −3.28 −3.31 −3.39

(−3.80, (−3.70, (−3.45, (−3.58, (−3.67,

−2.80) −3.00) −3.11) −3.04) −3.12)

6.48 6.71 6.46 6.62 6.70

(6.14, (6.47, (6.35, (6.44, (6.51,

6.83) 6.95) 6.57) 6.81) 6.89)

0.993 0.997 0.999 0.998 0.998

−3.32 −3.43 −3.33 −3.25 −3.33

(−3.56, (−3.54, (−3.79, (−3.51, (−3.42,

−3.08) −3.31) −2.88) −2.99) −3.24)

7.88 7.76 7.95 8.06 8.24

(7.72, (7.68, (7.64, (7.88, (8.18,

8.04) 7.84) 8.26) 8.24) 8.30)

0.998 0.999 0.995 0.998 0.999

−3.12 for 3272; LC96, −3.56 to −3.08 for MON810, − 3.54 to −3.31 for TC1507, −3.79 to −2.88 for DAS59122, −3.51 to −2.99 for MIR162, −3.42 to −3.24 for 3272) (Figure 2E, F and Table 3). In contrast, the equivalency in the CI95% of the intercept between the five GM events was less than that of the slope (7900HT, 6.14−6.83 for MON810, 6.47−6.95 for TC1507, 6.35−6.57 for DAS59122, 6.44−6.81 for MIR162, 6.51−6.89 for 3272; LC96, 7.72−8.04 for MON810, 7.68−7.84 for TC1507, 7.64−8.26 for DAS59122, 7.88−8.24 for MIR162, 8.18−8.30 for 3272) (Figure 2E, F and Table 3). The difference in the CI95% of the intercept between the five GM events was predicted to result from errors of DNA extraction and/or DNA quantification. Furthermore, the regression lines showed high linearity (R2 = 0.993−0.999) (Table 3). These results indicated that the quantitative performance for P35S and TNOS were equivalent in this multiplex real-time PCR assay, regardless of the GM event, for both PCR instruments. In the Minimum Information for Publication of Quantitative Real-Time PCR Experiments (MIQE) guidelines, the limit of detection (LOD) is defined as the lowest concentration at which 95% of the positive samples are detected.23 By analyzing a low copy number (2.5, 5, and 10 copies/well sample) of the plasmid DNAs pUC-P35S and pUC-TNOS, diluted with pUCSSIIb (n = 21 per sample), we confirmed that the present multiplex real-time PCR assay could detect at least five copies of target DNAs using both PCR instruments (Table 4), which represents increased sensitivity compared with that previously reported.6 Furthermore, positive rates for 0.1 copy/well sample were 4.8−9.5%, which suggested that it was not possible to overestimate the LOD (Table 4). Interlaboratory Study for Method Validation. For validation of the developed method, an interlaboratory study was performed using blind samples of DNA solutions consisting of five different concentrations of MON810 and MIR162 as test samples. The homogeneity of CNR(P35S+TNOS)/SSIIb variances between 10 of the randomly selected test samples was confirmed by Cochran’s test [Cochran’s test statistic value = 0.197−0.578 < 0.602 (= 5% critical Cochran’s test statistic value for duplicates)], and no significant differences between test samples were observed by one-way ANOVA [F-ratio = 0.30−2.65 < 3.02 (= 5% critical F value for duplicates)] using the two PCR instruments (Supporting Information, Table S1). These results indicated that the test samples were homogeneous, i.e., all of identical quality.

Table 4. Sensitivity of the Multiplex Real-Time PCR Assay target sequence 7900HT P35S

TNOS

theoretical copy number

number of positive

positive rate (%)

10 5 2.5 0.1 10 5 2.5 0.1

21/21 21/21 19/21 2/21 21/21 20/21 15/21 1/21

100 100 91 9.5 100 95 71 4.8

10 5 2.5 0.1 10 5 2.5 0.1

21/21 20/21 18/21 1/21 21/21 21/21 16/21 2/21

100 95 86 4.8 100 100 76 9.5

LC96 P35S

TNOS

In two laboratories using 7900HT, weak detection of P35S +TNOS was observed in one of six reactions (three reactions × two samples) containing non-GM maize sample. Given that P35S+TNOS was not detected in the nontemplate control in all laboratories, weak detection of P35S+TNOS in the non-GM maize sample may not be attributed to contamination by GM maize during the experiment. Although the cause of this phenomenon is unknown, we speculate that contamination from natural Caulif lower mosaic virus or Agrobacterium adhering to non-GM maize kernels may be responsible. Given that the weak amplification of P35S+TNOS in the two laboratories (Cq = 37.8, 37.9) did not affect the Cq values of positive samples (Cq(1%) = 32.9−33.5), we used the data from these two laboratories in the statistical analysis. No P35S+TNOS detection was observed in non-GM maize samples in the other participating laboratories. The regulatory threshold level of GMO content for labeling is 5% in Japan. Therefore, a screening method in Japan must determine whether the GMO content of the analytical sample is less than 5%. Therefore, we defined the control sample as a sample in which the GMO content is unequivocally judged to be less than 5%, and the ΔΔCq value between the analytical sample and the control sample was selected as the criterion for F

DOI: 10.1021/acs.analchem.5b04335 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

Table 5. Variance Analysis of GMO Content Ratio (GCR) Values in the Interlaboratory Study Obtained Using the 7900HT PCR Instrument trueness

precision

confidence interval

mean

bias

repeatability

reproducibility

repeatability

reproducibility

retained laboratories

theoretical value

GCR

true value (%)

RSDr (%)

RSDR (%)

CIr, 95%

CIR, 95%

1 2 3 4

10 10 10 10

5.00 2.50 1.67 1.25

5.13 2.56 1.66 1.27

2.6 2.2 −0.4 2.0

17.6 18.5 19.2 16.8

20.0 20.8 20.6 21.8

(4.51, (2.23, (1.45, (1.13,

5.74) 2.88) 1.88) 1.42)

(4.43, (2.19, (1.43, (1.09,

5.83) 2.92) 1.90) 1.46)

1 2 3 4

11 11 11 11

5.00 2.50 1.67 1.25

5.50 2.74 1.75 1.25

10.0 9.5 4.9 −0.4

13.7 13.6 12.8 13.1

14.8 18.6 17.1 18.3

(5.00, (2.49, (1.60, (1.14,

6.01) 2.99) 1.90) 1.35)

(4.95, (2.39, (1.55, (1.09,

6.05) 3.08) 1.95) 1.40)

mixing level (%) of control sample MON810

MIR162

Table 6. Variance Analysis of GMO Content Ratio (GCR) Values in the Interlaboratory Study Obtained Using the LC96 PCR Instrument trueness

mixing level (%) of control sample

precision

confidence interval

mean

bias

repeatability

reproducibility

repeatability

reproducibility

CIr, 95%

CIR, 95%

retained laboratories

theoretical value

GCR

true value (%)

RSDr (%)

RSDR (%)

1 2 3 4

4 4 4 4

5.00 2.50 1.67 1.25

5.48 2.73 1.57 1.29

9.6 9.1 −6.0 3.2

14.7 15.0 13.7 13.7

14.9 15.6 13.2 11.4

(4.20, (2.08, (1.23, (1.01,

6.76) 3.38) 1.91) 1.57)

(4.18, (2.05, (1.24, (1.06,

6.77) 3.40) 1.90) 1.52)

1 2 3 4

4 4 4 4

5.00 2.50 1.67 1.25

4.84 2.73 1.65 1.28

−3.2 9.2 −1.0 2.5

6.3 5.9 5.3 5.7

7.1 5.8 12.3 7.0

(4.36, (2.47, (1.51, (1.16,

5.32) 2.99) 1.79) 1.40)

(4.29, (2.48, (1.33, (1.14,

5.39) 2.98) 1.98) 1.42)

MON810

MIR162

21.8%, and from 12.8% to 19.2%, respectively. These data fulfilled the method acceptance criterion (RSDr < 25%) and method performance requirements (RSDR < 35%; absolute value of bias