DDNS results compared with the standard algorithm
Stool samples were tested in parallel via DDNS and the standard culture-based detection algorithm. 2,339 prospective stool samples were processed over 26 nanopore sequencing runs across the 141-day period that sequencing was performed, averaging one sequencing run every 5.4 days. DDNS testing found 62 samples (2.7 % of total samples) positive for poliovirus; 36 for serotype 2 VDPV (1.58 %), 5 for Sabin serotype 1 (0.30 %), 19 for Sabin serotype 3 (0.90 %) and 2 for both serotypes 1 and 3 Sabin poliovirus (0.09 %) (Table 1). The standard cell culture algorithm identified polioviruses in 51 samples; 31 samples testing positive for serotype 2 VDPV (1.33 %), 4 for Sabin serotype 1 (0.17 %) and 16 for Sabin serotype 3 (0.68 %).
Table 1 – Poliovirus detection by DDNS and the standard cell-culture, ITD and Sanger sequencing algorithm
|
|
DDNS
|
|
|
Sabin 1
|
VDPV2
|
Sabin 3
|
Sabin 1 + Sabin 3
|
Negative
|
Cell-culture, ITD and Sanger sequencing
|
Sabin 1
|
3
|
0
|
0
|
0
|
1
|
VDPV2
|
0
|
27
|
0
|
0
|
4
|
Sabin 3
|
0
|
0
|
15
|
0
|
1
|
Negative
|
2
|
9
|
4
|
2
|
2271
|
The sensitivity and specificity of detection for each poliovirus type was determined for DDNS versus the cell-culture algorithm and vice versa (Table 2). Additional cVDPV2 detected by either method were not identified as contamination as sequences differed from those of other samples (as shown in Supplemental Figure 1). The sensitivity and specificity of the two methods did not differ significantly (Fisher’s exact test).
Table 2 – Sensitivity and specificity by sample for detection of Sabin 1 and Sabin 3 polioviruses and VDPV2 by the standard cell-culture algorithm versus DDNS and vice versa. P-values were generated using Fisher’s exact test.
|
Culture vs DDNS (95% CI, n/N)
|
DDNS vs culture (95% CI, n/N)
|
Test for difference between methods, P-value
|
Sabin 1
|
Sensitivity
|
43 (10-82, 3/7)
|
75 (19-99, 3/4)
|
0.55
|
Specificity
|
100 (100-100, 2331/2332)
|
100 (100-100, 2331/2335)
|
0.37
|
VDPV2
|
Sensitivity
|
75 (58-88, 27/36)
|
87 (70-96, 27/31)
|
0.24
|
Specificity
|
100 (100-100, 2299/2303)
|
100 (99-100, 2299/2308)
|
0.27
|
Sabin 3
|
Sensitivity
|
71 (48-89, 15/21)
|
94 (70-100, 15/16)
|
0.11
|
Specificity
|
100 (100-100, 2317/2318)
|
100 (99-100, 2317/2323)
|
0.12
|
Two stool samples were available for 1,118 AFP cases, with 37 cases positive for poliovirus by either method. 18 cases had full concordance between both methods and both samples testing positive (see Table 3). There were no cases where both samples tested positive by cell-culture and yielded no positive DDNS result, while in 9 cases with positive DDNS results no poliovirus was detected by the cell-culture algorithm.
Table 3 – Concordance of the cell-culture algorithm and DDNS for the testing of paired stool samples
|
|
DDNS
|
|
|
+ +
|
+ -
|
- -
|
Culture
|
+ +
|
18
|
0
|
0
|
+ -
|
3
|
4
|
3
|
- -
|
4
|
5
|
1081
|
A single sample or pair of samples were available for 1,159 AFP cases. The sensitivity and specificity of detection were calculated for each AFP case (see Table 4), and for only AFP cases where two stools were available (n =1,118 cases, see Supplemental Table 1).
Table 4 – Sensitivity and specificity by AFP case for detection of Sabin 1 and Sabin 3 polioviruses and VDPV2 by the standard cell-culture algorithm versus DDNS and vice versa.
|
Culture vs DDNS (95% CI, n/N)
|
DDNS vs culture (95% CI, n/N)
|
Test for difference between methods, P-value
|
Sabin 1
|
Sensitivity
|
50 (7-93, 2/4)
|
100 (22-100, 2/2)
|
0.47
|
Specificity
|
100 (100-100, 1155/1155)
|
100 (99-100, 1155/1157)
|
0.50
|
VDPV2
|
Sensitivity
|
70 (46-88, 14/20)
|
88 (62-98, 14/16)
|
0.26
|
Specificity
|
100 (99-100, 1137/1139)
|
99 (99-100, 1137/1143)
|
0.29
|
Sabin 3
|
Sensitivity
|
75 (43-95, 9/12)
|
90 (55-100, 9/10)
|
0.59
|
Specificity
|
100 (100-100, 1146/1147)
|
100 (99-100, 1146/1149)
|
0.62
|
Time taken to confirm poliovirus by VP1 sequence
During this study period, 27 samples with VDPV2 had the VP1 region sequenced using both diagnostic methods. Only samples of programmatic importance are sequenced following cell-culture (all serotype 2 viruses and any suspected vaccine-derived and wild-type polioviruses) whilst DDNS testing always yields a sequence for positive samples. For these 27 samples a median of six days were required between case onset and sample collection (range: 2-21 days) and a further median six days were required between sample collection and arrival of samples at the laboratory (range: 2-27 days). The time from receipt in the laboratory to a VP1 sequence took a median of 30 days (range: 21-41 days) via the standard algorithm, including a median of 8 days (range: 4-22 days) required for shipment between the virus isolation and sequencing lab, whilst DDNS was significantly quicker (p < 0.001, Mann–Whitney U test) requiring a median of seven days (range: 4-23 days) (see figure 1).
cVDPV2 outbreaks detected during the study period
During the study period, four cVDPV2 outbreaks occurred in the province of Maniema in the DRC and confirmed through routine testing (culture-ITD-sequencing). For two of the linages (RDC-MAN-3 and RDC-MAN-4) the sample confirming circulation (2nd case) was collected during the study period, whilst for RDC-MAN-2 the confirming sample was collected 42 days prior to the study period and processed during training. By the standard detection algorithm, the samples confirming these outbreaks required 27, 35 and 64 days respectively between sample collection and Sanger sequencing, a mean of 42 days. These same samples were processed in 6, 20 and 30 days respectively by DDNS, despite the samples for RDC-MAN-2 being collected prior to the study period, a mean of 23 days quicker. The fourth outbreak lineage, RDC-MAN-5, only had the first positive sample collected during the study period, but this sample similarly yielded a VP1 sequence 29 days earlier by DDNS. The geographic spread of cases identified by DDNS for the four outbreaks and relatedness of RDC-MAN-3 outbreak lineage is shown in figure 2. Based on the poliovirus VP1 molecular clock and these DDNS-derived VP1 sequences we estimate that the RDC-MAN-3 lineage emerged from a OPV2 vaccination campaign performed in the first quarter of 2020 (mean date 26.01.2020, 95% Highest Posterior Density 04.04.2019-16.09.2020.
Sequence comparisons
Where consensus cVDPV2 VP1 sequences were available from both DDNS and Sanger sequencing of culture isolate for the same sample, the similarity of the sequences was determined. The mean VP1 sequence identity comparing DDNS and the standard algorithm (including Sanger sequencing) for the 27 cVDPV2 with results for both methods was 99.98% (range: 99.60-100%). The absolute number of differences between sequences is shown in table 5.
Table 5 – Nucleotide differences in the VP1 region (903 base pairs) between sequences generated by Sanger sequencing of culture isolate and by DDNS from the same stool sample.
Number of nucleotide differences
|
Count (%)
|
0
|
25 (92.6%)
|
1
|
1 (3.7%)
|
2
|
0 (0%)
|
3
|
0 (0%)
|
4
|
1 (3.7%)
|
5+
|
0 (0%)
|
Costs and staff requirements
The DDNS assay consumable costs are approximately $20 per sample when performing multiplexed sequencing runs of 96 samples or $25 per sample when performing runs of 45 samples in lower throughput laboratories (see Supplemental Table 2). These figures include chloroform treatment, RNA extraction, nested PCR and nanopore sequencing. For chloroform treatment, cell-culture and qPCR alone, the cell-culture-based detection algorithm costs approximately $32 (see Supplemental Table 2), in addition to the cost of Sanger sequencing and shipping to the National Institute For Communicable Diseases (NICD) in South Africa where the sequencing is performed. Whilst some large items of equipment are required by both methods, DDNS avoids the requirements of microscopes, incubators, cell counters and tissue culture cabinets, whilst only needing the addition of a MinION or GridION sequencer (with MinIONs typically costing $1,000, including a sequencing reagents kit and a flow cell). Staff costs have not been included in the figures, yet the performance of DDNS at INRB required only five staff members; three laboratory scientists for RNA extraction, nested RT-PCR and nanopore sequencing and two bioinformaticians/data managers to perform data quality control and match the sequences to case data. Comparable steps from the cell-culture based algorithm requires four laboratory scientists for cell-culture and qPCR, two support staff for maintaining the facilities and the support of additional sequencing staff and bioinformaticians at the NICD.