Nucleotide sequences of the epidemic strain of Zaire Ebola virus from West Africa were obtained from GenBank (KJ660346, KJ660347, and KJ660348) [3]. Synthesis of the target segments of the EBOV genome was outsourced to Thermo Fisher Scientific Inc. (Beijing, China), including the full sequence of the 3’-untranslated region (UTR) and the NP fragment, and portions of the sequence of the GP, L, and 5’-UTR regions. Synthesis of the target sequences was divided into two parts: the first encompassed nearly 2400 bp (the 3′-UTR and a large part of the NP region), and the other encompassed approximately 2250 bp (the remaining part of the NP region, and selected parts of the GP, L, and 5′-UTR regions). The primers used are listed in S1 Table. The two synthesized fragments were subcloned into the pACYC-MS2 vector at the KpnI/PacI and BglII/KpnI restriction sites, respectively, to form the recombinant plasmids pACYC-MS2-NP and pACYC-MS2-GP/L. The expression and purification of the two types of MS2 virus-like particles (VLPs) was performed according to previously published protocols [9, 10].

Two different types of EBOV MS2 VLPs were used as positive samples. The sample panel was comprised of three negative and seven positive samples, ranging from 8 × 10² to 5 × 10⁷ copies/mL, in accordance with previously reported clinical outcomes [11]. Three negative samples and dilution buffer for positive ones were composed of 30% normal human serum and 70% normal saline.

Of the seven positive samples, No. 1402 contained the highest concentration of EBOV VLPs (Table 1). No. 1408 also contained hepatitis C virus (HCV) and measles virus (MV) VLPs, to verify whether EBOV detection is impacted by the presence of other viruses. Repeatability was evaluated by the use of two replicate specimens (No. 1403 and No. 1407). No.1410 was prepared to evaluate the ability to detect slightly positive samples. Additionally, one specimen (No. 1405) contained only a very weak concentration of the NP fragment and a moderate concentration of the GP/L EBOV fragment. Prior to distribution, each sample panel was confirmed using three commercial kits manufactured in China (Daan Gene Co., Ltd., Guangzhou; ZJ Bio-Tech Co., Ltd., Shanghai; and Puruikang Bio-Tech Co., Ltd., Shenzhen).

The stability of two types of EBOV MS2 VLPs with the dilution buffer of 30% normal human serum and 70% normal saline was evaluated at different times and temperatures. Two different concentrations of VLPs were used: 10⁵ copies/mL and 10³ copies/mL. For each of the concentrations, 15 samples were included. Three samples were stored at -80°C as a control. The remaining 12 samples were divided into three groups and incubated at different temperatures: room temperature, 4°C, and -20°C. In each group, two samples were removed and stored at -80°C for 10 days or 20 days. After 20 days, all 15 samples for each concentration were simultaneously assessed using EBOV commercial assays (real-time RT-PCR kit, Daan Gene Co., Ltd., Guangzhou, China).

The stability of the samples during distribution and transport was monitored. A sample panel was sent unmarked by mail to a deliberately non-existent destination and returned to NCCL. The returned sample panel was assessed simultaneously with another that had been stored at -80°C, using an EBOV RT-PCR assay manufactured by Daan Gene Co., Ltd. In all assays, stability was designated as a reduction of < 0.5 log₁₀ copies/mL, compared with the results for samples stored at -80°C.

All qualified laboratories, institutions for inspection and quarantine, and manufacturers of kits for EBOV detection in mainland China were invited to participate in this investigation in November 2014. All participating organizations were identified by early December (Fig 1). All samples were distributed by NCCL on December 10, 2014. Qualitative results, threshold cycle (Ct) values, RNA extraction and detection methods, and primer sources were all required to be reported to NCCL by December 30, 2014. The full EQA process and related time points are shown in Fig 1.

The results were classified into three categories: competent (100% correct responses), acceptable (≤ 2 incorrect results), or improvable (> 2 incorrect results). All data were analyzed using SPSS, version 19.0; the variances of different groups were compared using Fisher’s exact test or Pearson’s chi-square test, when appropriate.

Construction of the recombinant plasmids pACYC-MS2-NP and pACYC-MS2-GP/L was successful. A series of experiments was conducted to verify the effects of VLP expression and packaging capability. Target protein purification in approximately 60 minutes (S1A Fig) and observation of a 14-kD band following sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) (S1B Fig) were preliminary evidence for successful VLP expression. After digestion with RNase A and DNase I at 37°C, observation of a single band over 1 kb in length in 1% agarose gel electrophoresis (S1C Fig) indicated the stability of the EBOV VLPs. RT-PCR results showing a target band between 2–2.5 kb (S1D Fig) served as further evidence of successful encapsulation of the target EBOV segments in VLPs after extraction of nucleotides. The concentrations of the two types of EBOV VLPs were each approximately 5 × 10¹¹ copies/mL.

In the stability evaluation, the variations of two concentrations of samples in different temperatures were presented in form of mean Ct±standard deviation (SD), which indicated in S2 Table. Samples stored at different conditions showed less than 0.5 log₁₀ variations in concentration, compared with samples incubated at -80°C. Meanwhile, the comparisons of sample panels saved in -80°C and from blind mail were showed in S2 Fig. No statistical difference was observed between test results for sample panels sent unmarked through the mail and those stored at -80°C.

The sample panel was verified before distribution using the three commercial kits mentioned above; the results were concordant with the results shown in Table 1. The results for No. 1403 and No. 1407 were similar. No obvious interference of HCV or MV with EBOV detection was observed for No. 1408. The slightly positive samples (No. 1405 and No. 1410) could be detected.

A total of 19 agencies took part in the assessment of the quality of molecular detection of EBOV, including three provincial Centers for Disease Control and Prevention (CDC) (Shanghai, Zhejiang, and Jiangsu), six Entry-Exit Inspection and Quarantine agencies in port cities (Fujian, Hainan, Hebei, Yunnan, Zhejiang, and Zhuhai), one public health platform in Shanghai, one fixed point hospital qualified for EBOV detection(Chongqing), and eight manufacturers of EBOV detection kits(Daan Gene Co., Ltd., ZJ Bio-Tech Co., Ltd., HuaDa BioTech Co., Ltd., KeHua Bio-Engineering Co., Ltd., TianLong Bio-Tech Co., Ltd., BioPerfectus Technologies, Sansure Bio-Tech Co., Ltd., and Puruikang Bio-Tech Co., Ltd.). All participating organizations had certification in China for EBOV screening and diagnosis.

At the 19 participating agencies, both magnetic bead and spin column methods were used for nucleic acid extraction. The paramagnetic particle method was preferred by nine organizations, the spin column method by the other ten (Table 2). All of the laboratories that used the spin column method used manual extraction, whereas 2/3 of the agencies using the paramagnetic particle method used instruments. For detection, only one agency used RT-PCR and agarose gel electrophoresis, whereas most chose real-time PCR. Among the 18 agencies that selected quantitative real-time PCR, 83.33% used commercial kits. The specified primers offered by the national CDC [12] (Table 3), one of which returned results for both the NP and GP sites, were preferred by two provincial CDCs. Additionally, one agency used a real-time PCR assay developed in-house for EBOV diagnosis.

The performance of the 19 participating agencies is shown in Table 2. The performance of 14 of the 19 agencies (73.68%) was found to be competent (100% correct responses). Three participants (15.79%) reported one false-negative and one (5.26%) reported two false-negatives; these were classified as acceptable (21.05%) according to the criteria stated above. One participant (5.26%) was unable to obtain any positive results in this assessment, and was thus included in the improvable group.

In all, among the 200 results returned, 12 (6%) were incorrect, including 8.57% false-negative results and no false-positives. Seven of these incorrect results were returned by one agency. No. 1405 was not detected by five different laboratories, which reported very weakly positive results for the NP (8 × 10² copies/mL) fragment. The pair of duplicate specimens (No. 1403 and No. 1407) with similar Ct values was reported correctly by 18 institutions. No. 1410 (NP: 8 × 10³ copies/mL, GP/L: 2 × 10³ copies/mL), with a Ct value close to the declared lower limit (1 × 10³ copies/mL) of several kits, was detected by 17 participants. In addition, target sequences of other viruses (No. 1408 with HCV and MV VLP) did not produce cross-reactivity with the EBOV fragments.

We next evaluated the performance of different assays and participants. The commercially available assay from Daan Gene was the most frequently used and produced 100% accurate results at five agencies. Several participants selected the kits manufactured by Shanghai ZJ Bio-Tech Co., Ltd., one of which failed to detect the weakly positive specimens (No. 1405 and No. 1410). The primers offered by the Chinese CDC were selected by two participants, which achieved “competent” results. One participant used the RT-PCR and agarose gel electrophoresis method, and returned qualitative results (100% accurate rate) without a Ct value. However, another used an assay developed in-house and achieved only 30% accuracy, failing to detect any positive specimens. Additionally, no significant differences (P = 0.179) were observed between results obtained using commercial kits and those obtained using assays developed in-house.