The principle of the biosensor HC-FIA
In the novel HC-FIA detection system, probe DNA-functionalized FNPs were used to amplify the signal of virus RNA and DNA hybridizing. With the help of monoclonal antibody S9.6, the nucleic acid detection of virus RNA was transformed into an immunoassay through the sophisticated design. This type of antibody specifically bond with double helix structure of RNA-DNA hybrid and such recognition depended little on base sequence [3]. The design principle of HC-FIA biosensor was shown in Figure 1. The murine antibody S9.6 (Ab1) was pre-fixed on the test line (T) of the lateral flow dipstick, and the control line (C) was coated with goat anti-rabbit IgG polyclonal antibody. Pre-coated FNPs labeled by anti-DNA-RNA S9.6 (Ab2) and rabbit IgG antibody were firstly contained in the reaction tube.
In the detection process, firstly, the SARS-CoV-2 in the throat swab or sputum sample was lysed and released, and the released RNA hybridized with the specific SARS-CoV-2 DNA probe. The resulting RNA-DNA hybrid was captured by S9.6 antibody (Ab2) labeled FNPs, and the complex flowed along the sample pad and the nitrocellulose membrane toward the absorbent paper under capillary forces. When passing through T area, it would be captured by the coated mouse anti-DNA-RNA monoclonal antibody (Ab1) to form an "Ab1-RNA-DNA hybrid-fluorescence labeled Ab2" complex and a fluorescent signal gradually generated. In the case of C area, FNPs labeled rabbit IgG antibody in the reaction tube and the pre-fixed goat anti-rabbit IgG polyclonal antibody bound to form a "goat anti-rabbit IgG polyclonal antibody-FNPs labeled rabbit IgG antibody" (Figure 1A).
The fluorescence signal of the nanoparticles emitted by the excitation light was received by the fluorescence analysis device, and the measurement value indicated the presence or absence of the target SARS-CoV-2 RNA based on the Cutoff value (Figure 1B&C). Figure 1D showed the appearance of the portable "suitcase laboratory" as alternative to large and precision equipment, which was able to give qualitative results in less than an hour undergoing only two steps, namely hybridization and immunofluorescence analysis (Figure 1E).
In this study, we chose to employ the data mode of T/C for the fluorescence signal to eliminate the background of difference test cards. By measuring the T/C value of throat swab samples from 211 healthy population, the average background value was calculated to be 49.95, with the standard deviation (S.D.) of 17.16 (Figure S1A). If the receiver operating characteristic (ROC) curve had been drawn for the test results of 100 clinical diagnosis cases (confirmed or excluded), the area under the curve (AUC) to evaluate the predictive accuracy was determined to be 0.999 with 95% confidence interval (CI) of 0.997-1.000 (Figure S1B). For the purpose of obtaining the highest sensitivity and specificity, the cutoff value was determined to be 102.07 corresponding to the Youden index point of 0.980. For 100.00 was also much close to 2 times the average background 49.95, T/C value of 100.00 was taken as the cutoff value for convenience in the following study.
Optimizing of the SARS-CoV-2 specific DNA probes
Optimizing DNA probes for the target RNA sequence of SARS-CoV-2 was the key factor to improve the efficiency of hybrid capture. All of DNA probe sequences we designed were shown in Table 1. The genome of SARS-CoV-2 is approximately 30,000 bases, including a variable number (6 to 11) of open reading frames (ORFs). The first ORF accounts for about 67% of the entire genome and encodes 16 non-structural proteins (nsp), and the remaining encodes helper proteins and structural proteins. The four main structural proteins are spike glycoprotein (S), small envelope protein (E), matrix protein (M), and nucleocapsid protein (N) [12-13]. To date, the targeted genomic sites in SARS-CoV-2 nucleic acid detection mainly include three conserved regions in the viral genome, namely the open reading frame 1ab (ORF1ab), where the RNA-dependent polymerase gene (RdRp) is located [12], the Nucleocapsid protein (N) gene and the envelope protein (E) gene.
Table 1 Sequences of the DNA probes used in the HC-FIA detection.
Probe
Number
|
Binding region
|
Sequence
(5’-3’)
|
Sequence length (bp)
|
cov01
|
ORF1ab segment
|
AACGATTGTGCATCAGCTGACTGAAGCATGGGTTCGCGGAGTTGATCACAACTACAGCCATAACCTTTCCACATACCGCAGACGGTACAGACTGTGTTTTTAAGTGTAAAACCCACAG
|
118
|
cov02
|
ORF1ab segment
|
TCCTCACTGCCGTCTTGTTGACCAACAGTTTGTTGACTATCATCATCTAACCAATCTT
|
58
|
cov03
|
ORF1ab segment
|
CTTATCATCTTGTTTTCTCTGTTCAACTGAAGGTTTACTTTCAGTTATAAATGGC
|
55
|
cov04
|
N segment
|
CAGACATTTTGCTCTCAAGCTGGTTCAATCTGTCAAGCAGCAGCAAAGCAAGAGCAGCATCACCGCCATTGCCAGCCATTCTAGCAGGAGAAGTTCCCC
|
99
|
cov05
|
N segment
|
GCCTTTACCAGACATTTTGCTCTCAAGCTGGTTCAATCTGTCAAG
|
45
|
cov06
|
N segment
|
CAGCAAAGCAAGAGCAGCATCACCGCCATTGCCAGCCATTCTAGCAGGAGAAGT
|
54
|
cov07
|
E segment
|
AAGCGCAGTAAGGATGGCTAGTGTAACTAGCAAGAATACCACGAAAG
|
47
|
cov08
|
E segment
|
CGAAGCGCAGTAAGGATGGCTAGTGTAACTAGCAAGAATACCACGAAAGCAAGAAAAAGAAGTACGCTATTAACTATTAACGTACCTGT
|
89
|
cov09
|
E segment
|
AGCAAGAAAAAGAAGTACGCTATTAACTATTAACGTACCTGT
|
42
|
We firstly found the RNA genome sequence of SARS-CoV-2 on the NCBI (National Center for Biotechnology Information) genebank, with the accession numbers: MN908947, MN908947.3, MN908947.2, NC_045512.1. Detailed analysis of other published sequences of SARS-CoV-2 genome revealed that no significant variation in these regions. With the aid of the design software Primer Premier 5.0, three probes were designed for each of the three regions. Among them, Cov01 and Cov04 located in the recommended region for the detection of the ORF1ab and N gene, respectively, announced by the Chinese Centers for Disease Control and Prevention (China CDC), while Cov08 accorded with the region of the E gene detection proposed by the literature [14] (Figure 1F).
Then the sequence alignment was conducted between the designed DNA probes and the human genome sequences, viruses, bacteria, mycoplasma, chlamydia, and other common pathogens. It only 100% matched the SARS-CoV-2 genomic sequence with no homology to human genomic DNA. Pseudoviruses carrying different segments of the target gene constructed using lentivirus as vectors were used as positive controls. The target gene sequences of pseudoviruses used was shown in Table S1. P1 was positive for the SARS-CoV-2 N segment, P2 was positive for the SARS-CoV-2 E segment, and P3 was SARS-CoV-2 ORF1ab segment was positive at a concentration of 3000 TU/mL. Physiological saline, purified water and other common pathogens positive pharyngeal swab samples were used as negative controls N1-N17. The information of the positive and negative references used in the study was listed in Table S2. And the test results were shown as Table S3, S4 and S5. For ORF1ab segment, probes CoV01 and CoV03 were selected, and for E segment, probe CoV08 was selected. In the case of N segment, probe CoV04 and probe CoV06 were the preferred ones to bind target RNA. Then the selected DNA probes were further optimized by combination in the next step (Table S6), and each group of probes should simultaneously target all the three segments. HC-FIA test results in Table S7 showed that the probe combination of Cov01, Cov04, Cov08 (Group 2) not only successfully discriminated positive control samples from negative controls, but was capable of detecting low viral titer positive samples (1000 TU/mL) with a positive rate higher than 0.95. Therefore, Group 2 was selected out as the final combination.
The specificity of HC-FIA
After optimizing the probe sequences, we next examined the specificity of the HC-FIA method in detecting SARS-CoV-2. The positive controls included pseudoviruses (P1-P3) with target genes, and five clinical throat swab samples (P4-P8), which was determined by the nucleic acid detection kit (Shanghai ZJ Bio-Tech Co.,Ltd.). The negative controls were the confirmed influenza A, influenza B, respiratory syncytial virus, chlamydia pneumoniae, adenovirus or other pathogens positive and SARS-CoV-2 negative throat swab samples (N5-N17), and pseudovirus positive for the N segment of MERS (N18) or SARS coronavirus (N19), as listed in Table S2. The detection results of 8 positive reference samples and 15 negative reference samples were in line with expectations, as shown in Table S8.
We proceeded to investigate whether there were cross-reactions between 55 common pathogens related to respiratory diseases and SARS-CoV-2. The source and quantitative information of each pathogenic microorganism was provided in Table S9. The virus titer was determined to be 106 pfu/mL by plaque assay, while for interference samples of bacteria, mycoplasma, and chlamydia, the concentration level was 107 cfu/mL. In addition, human genomic DNA was extracted and quantified to be 90-105 µg/mL from three whole blood samples. The results indicated that no significant cross reaction between all other pathogen samples (Table S10) or human genomic DNA (Table S11) and specific DNA probe sequences, and the HC-FIA method exhibited excellent specificity for SARS-CoV-2 detection.
The sensitivity and precision of HC-FIA
Furthermore, the pseudoviruses samples containing three sections of target genes were serially diluted to titers of 5000, 2500, 1000, 800, 500, 250 and 100 TU/mL (absolutely quantified by digital PCR). Average the T/C values of 20 parallel tests. Figure 2A showed that the limit of detection (LOD) of SARS-CoV-2 N segment, E segment or ORF1ab segment positive pseudovirus was as low as 1000 TU/mL with positive rate greater or equal to 95%, that is, 500 copies/mL as determined by digital PCR. Moreover, when the titers of pseudoviruses samples reached 108 TU/mL, no significant HOOK effect was observed (Figure S2). Simultaneously, it is indicated that the linear range of the assay located in the titer range from 103 to 106-7 TU/mL.
Throat swab samples from 3 critical patients, proved to be positive by RT-qPCR and viral load further quantified based on digital PCR, were mixed with negative throat swabs samples to prepare serial dilutions of 2000, 1000, 500, 400 and 250 copies/mL. Similarly, the results showed that the LOD for clinical samples was revealed to be 500 copies/mL with positive rate greater or equal to 95% (Figure 2B). Moreover, the clinical throat swab samples close to the critical value of positive ones (512, 489, and 497 copies/mL of the ORF1ab segment according to digital PCR) were employed to verify the LOD. Being performed 20 times in parallel, the positive rates of the critical samples were documented to be higher than or equal to 95%.
To evaluate the performance of HC-FIA kit in SARS-CoV-2 detection for precision, 20 times of parallel tests for each clinical throat swab sample were performed in 5 consecutive days. And the representative samples we chose included a positive one (1348 copies / mL), a critical one (512 copies / mL) and a negative one (0 copies / mL). The average T/C values of the three batches in detecting the positive sample were as follows: 199.92 ± 8.25 (CV: 4.13%), 200.68 ± 7.91 (CV: 3.94%) and 199.03 ± 7.43 (CV: 3.73%). For the critical sample, they were 109.17 ± 5.68 (CV: 4.65%), 110.80 ± 5.63 (CV: 5.08%), 111.48 ± 4.67 (CV: 4.19%), and for the negative one, 44.66 ± 3.36 (CV: 7.52%), 43.99 ± 2.72 (CV: 6.18%), and 44.72 ± 2.98 (CV: 6.66%). The batch-to-batch CV were 3.89%, 4.66%, and 6.74%, lower than 10%. In summary, the above results indicated that the reagents and related products of HC-FIA exhibited good precision and reproducibility in detecting SARS-CoV-2.
The robustness of HC-FIA
In order to examine the robustness of HC-FIA, the effects of the endogenous interference substances, such as hemoglobin, mucin, and exogenous interference substances, namely common clinical drugs utilized in the treatment of patients with respiratory infection, including antiviral drugs, antibiotic and hormone, were systematically evaluated.
There were 18 clinical throat swab samples used for this experiment, including six critical samples (500-530 copies/mL for ORF1ab segment based on digital PCR), six negative samples, six positive samples. The T/C value of the previous interference-free test was the basic value, and the ratio of the later T/C value with interference to the basic value was used to investigate whether the interference existed or not. In the prepared interference samples, the hemoglobin concentrations were 0.5, 1.0, and 2.0 g/L, and in the case of mucin, the concentrations were 5, 10, 20 g/L. The drug concentrations of exogenous interference samples were provided in Table S12-S14 according to the peak plasma concentration in vivo. As expected, all the ratios of interference samples to the corresponding basic ones located in the range of 0.9 to 1.1 (Table S15-S17 for hemoglobin, Table S18-S20 for mucin and Table S21-23 for exogenous interference substances), indicating the SARS-CoV-2 HC-FIA detection reagent had excellent anti-interference property.
The clinical evaluation of HC-FIA in SARS-CoV-2 detection
In order to further evaluate the performance of HC-FIA detection reagents, a randomized double-blind clinical trial was performed by comparing it with RT-qPCR or clinical diagnosis results in three independent medical institutions. The SARS-CoV-2 RT-qPCR detection kit produced by Shanghai ZJ Bio-Tech Co.,Ltd., which had been approved by NMPA and considered to be of better quality was chosen for contrast research. The clinical diagnosis results were provided by the designated hospitals of CIVID-19, determined from the CT images combined with clinical manifestations of the patients. A total of 734 samples provided by 670 subjects were tested in parallel, among which there were 593 throat swabs samples and 141 sputum samples. The case enrollment criteria referred to “Diagnosis and treatment of novel coronavirus infection pneumonia (Trail Version 6.0)” of China. The raw data of the clinical trial in Excel was provided as a separate supporting document.
Among the 670 cases enrolled in the trial, 313 cases were males, accounting for 46.72%, and 357 cases were females, accounting for 53.28%. As shown in Figure 3, the enrolled population was in line with demographic characteristics of SARS-CoV-2 infection according to different age groups [15]. Moreover, the visiting rate and diagnosis rate probably basically balanced with clinical actual conditions.
A total of 621 cases of HC-FIA test results were consistent with clinical diagnosis (Table 2). Among them, 210 cases were confirmed, and 411 cases were excluded. 49 cases were inconsistent with the clinical diagnosis results, of which 27 cases were confirmed, with the HC-FIA test results negative, and 22 cases were excluded, with the HC-FIA test results positive.
Table 2 Qualitative analysis of the 670 cases*.
|
Clinical Diagnosis
|
Confirmed
case
|
Excluded
Case
|
Total
|
HC-FIA
|
Positive
|
210
|
22
|
232
|
Negative
|
27
|
411
|
438
|
Total
|
237
|
433
|
670
|
*Note:Each case was included only once. One case that simultaneously sampling sputum and throat swabs was only enrolled the sputum sample.
Table 3 Qualitative analysis of the 734 samples*.
|
RT-qPCR
|
Positive
|
Negative
|
Total
|
HC-FIA
|
Positive
|
249
|
4
|
253
|
Negative
|
0
|
481
|
481
|
Total
|
249
|
485
|
734
|
*Note:All tests of all cases are included in the table. If repeated measure was required, the retested results were enrolled in the group for analysis.
As shown in Table 3, a total of 730 samples of HC-FIA test was consistent with RT-qPCR results, of which 249 were positive and 481 were negative. There are 4 samples among them inconsistent with each other, namely, the HC-FIA test reagents made the judgement of positive while the RT-qPCR results were negative. Detailed analysis of the 4 samples are as follows: 3 of them were clinically diagnosed as COVID-19 excluded cases, and the RT-qPCR results agreed well with the clinical diagnosis, indicating the HC-FIA test gave false positive results. One sample was clinically diagnosed as a confirmed case, and herein the HC-FIA test was in accordance with the clinical diagnosis.
According to the data in Table 4, statistics analyses gave direct evidence that HC-FIA showed high agreement with clinical diagnosis, with the Kappa coefficient calculated to be 0.8393, higher than 0.75. For the contrast research of HC-FIA and RT-qPCR, similar conclusion was reached since the all Kappa coefficients obtained were higher than 0.98, regardless of the sample type (Table 5).
In summary, our HC-FIA test kit was proved to have high-consistency with the clinical diagnosis results. Meanwhile, complete statistical analyses indicated the HC-FIA test kits had no statistical difference with the RT-qPCR products approved for marketing, namely the performances of the two test kits were equivalent for clinical application.
Table 4 Statistical analysis of HC-FIA and clinical diagnosis results for all cases.
|
Sensitivity
|
Specificity
|
Total data
|
Kappa
coeffiency
|
|
Consistency rate
|
95% confidence interval
|
Consistency rate
|
95% confidence interval
|
Consistency rate
|
95% confidence interval
|
Throat swabs
|
87.69%
|
83.08%~92.30%
|
95.01%
|
92.82%~97.20%
|
92.53%
|
90.38%~94.68%
|
0.8323
|
Sputum
|
91.94%
|
85.16%~98.72%
|
92.41%
|
86.57%~98.25%
|
92.20%
|
87.77%~96.63%
|
0.8419
|
All samples
|
88.61%
|
84.57%~92.65%
|
94.92%
|
92.85%~96.99%
|
92.69%
|
90.72%~94.66%
|
0.8393
|
|
|
|
|
|
|
|
|
|
|
Table 5 Statistical analysis of HC-FIA and RT-qPCR results for all samples.
|
Sensitivity
|
Specificity
|
Total data
|
Kappa
coeffiency
|
|
Consistency rate
|
95% confidence interval
|
Consistency rate
|
95% confidence interval
|
Consistency rate
|
95% confidence interval
|
Throat swabs
|
100%
|
99.85%~100%
|
99.26%
|
98.43%~100%
|
99.49%
|
98.92%~100%
|
0.9883
|
Sputum
|
100%
|
99.74%~100%
|
98.73%
|
96.26%~100%
|
99.29%
|
97.90%~100%
|
0.9856
|
All samples
|
100%
|
99.87%~100%
|
99.18%
|
98.38%~99.98%
|
99.46%
|
98.93%~99.99%
|
0.9879
|