Development and deployment of a rapid recombinase polymerase amplification Ebola virus detection assay in Guinea in 2015

O Faye 1 2 , O Faye 1 2 , B Soropogui 3 , P Patel 4 , AA El Wahed 5 , C Loucoubar 6 , G Fall 1 , D Kiory 1 , N Magassouba 3 , S Keita 7 , MK Kondé 8 , AA Diallo 7 , L Koivogui 9 , H Karlberg 10 , A Mirazimi 10 11 , O Nentwich 12 , O Piepenburg 12 , M Niedrig 4 , M Weidmann 2 13 14, AA Sall 1 2 14 1. Arbovirus and viral hemorragic fever unit, Institut Pasteur de Dakar, Dakar, Senegal 2. These authors contributed equally to the paper 3. Laboratory for Hemorrhagic fevers of Guinea, Conakry, Guinea 4. Centre for Biological Threats and Special Pathogens, Robert Koch Institute, Berlin, Germany 5. Unit of Infection Models, German Primate Center, Goettingen, Germany, currently Division of Microbiology and Animal Hygiene, Georg-August-University, Goettingen, Germany 6. Group of Biostatistics, Bioinformatics and Modeling, Institut Pasteur de Dakar, Dakar, Senegal 7. Ministry of Health, Conakry, Guinea 8. Center for Research and training on malaria and prioritary diseases (CEFORPAG), Conakry, Guinea 9. National Public Health Institute, Conakry, Guinea 10. Public Health Agency of Sweden, Solna Sweden, 11. Karolinska Institute, Solna, Sweden 12. TwistDx, Babraham Research Campus, Babraham, Cambridge, United Kingdom 13. Institute of Aquaculture, University of Stirling, Stirling, United Kingdom 14. These authors jointly supervised the project


Introduction
As of 11 October 2015, the ongoing Ebola virus disease (EVD) epidemic in West Africa has resulted in more than 28,500 cases and over 11,300 deaths.The early symptoms of EVD (i.e.fever, fatigue, headache, vomiting and diarrhoea) are unspecific and present a challenge for clinical diagnosis [1].In humans, death occurs generally seven to 10 days after the onset of symptoms.Survivors can be ill for up to 22 days before recovering [2].Ebola virus (EBOV) infection is mainly diagnosed by various in-house and commercial real-time RT-PCR assays [3], used in up to 38 laboratories implemented at or close to Ebola treatment centres (ETC) in West Africa [4].Transmission of EVD occurs almost exclusively from human to human by direct contact with body fluids of symptomatic cases.Consequently, the control strategy for EVD epidemics relies on early identification of EBOV-infected patients and corpses for, respectively, isolation and safe burials.It is imperative to trace and follow up contacts and to implement infection control measures.Therefore, rapid EVD diagnostics impact on outcome of treatment, efficiency of contact tracing and subsequently community engagement, which is central to the successful control of the EVD epidemic.The World Health Organization (WHO) launched a call and consultation for an emergency procedure under its

Study design and samples
The Suspected EVD cases were defined as any person, alive or dead, suffering or having suffered from a sudden onset of high fever and having had contact with a suspected, probable or confirmed case of EVD, or any person with sudden onset of high fever and at least three of the following symptoms: headaches, anorexia/ loss of appetite, lethargy, aching muscles or joints,  B: Bespoke insert slots were cut out of the PVC and the foam layer to host the tubescanner, the box for the disinfection wipes, the waste container, the vortex, the minicentrifuge and two boxes of refill pipette tips.
C: Foam, PVC layer and instruments were assembled outside the suitcase.Electricity wires were stowed underneath the foam layer.The equipment was fixed to the PVC layer using hot glue.

D:
The setup was placed into the suitcase and the seams were sealed with hot glue.
breathing difficulties, and any person with inexplicable bleeding or any sudden inexplicable death.

RNA extraction and inactivation
Two extraction methods were used.In the first method, viral RNA was extracted from 100 µl serum or swab transport medium using the QIAamp Viral Mini Kit (QC; Qiagen, Hilden, Germany).RNA was eluted in 50 μl Tris-EDTA buffer.The second extraction protocol (SpeedXtract Nucleic Acid Kit (SE), Qiagen, Hilden, Germany) was a reverse extraction method extracting protein debris by way of magnetic beads after an initial 10 min heating step at 95 °C.It yielded 200 µl supernatant from 20 µl of serum or oral swab transport fluid diluted 1:2 with molecular grade water.
We added 5 µl of either eluate to the W-PCR (EBOV one-step real-time RT-PCR described in [6]) and the optimised RT-RPA, and 10 µl to the A-PCR (The RealStar Filovirus RT-PCR Kit, Altona-Diagnostics, Hamburg, Germany).
To test for inactivation of EBOV by the new SE kit, SE extract dilutions from 10 −1 to 10 −5 were added in triplicate onto 2 × 10 5 VeroE6 cells in a 96-well plate and incubated for five days.The supernatant was passaged three times by transfer to a new well, followed by a 3 h incubation, a wash, and another 48 h incubation step.Finally, cells were washed three times and RNA was extracted in 200 µl Trizol and submitted to an EBOV inhouse PCR.For each dilution, three more wells to which the supernatants had been added in the same manner were subjected to an immunofluorescence assay after passage 1 [8].A not extracted patient serum sample was used as positive control and showed virus growth on VeroE6 cells.

Real-time RT-PCR
The W-PCR was performed on the SmartCycler (Cepheid, Sunnyvale, US) using the RNA Master Hybridisation Probes kit (Roche, Manheim, Germany).A dried 10-fold primer and probe mix containing 100 pmol EBOZ FP and EBOZ RP and 50 pmol EBOZ P (TIB Molbiol, Germany) was used.The A-PCR was used on the SmartCycler according to the manufacturer`s instructions.Positive results above cycle threshold (Ct) 35 were regarded as equivocal and repeated for confirmation [9].

RT-RPA assay
The primers and the exo probe of an existing EBOV RT-RPA assay [7] were redesigned to adopt mismatches of the current EBOV outbreak strain (Figure 1, Table 1) following RPA design guidelines [10].
The RT-RPA was performed using a custom-made EBOVspecific exo RT kit with pellets containing optimised enzyme concentrations similar to the commercial TwistAmp RT exo kit [10,11] was analysed by threshold validation (mV/min).Slope validation was used to verify that the increase of fluorescence occurred at a sufficiently high rate, and was verified by first derivative analysis.

The mobile laboratory
The mobile laboratory consisted of a glovebox (Bodo Koennecke, Berlin, Germany), a Diagnostics-in-a-Suitcase (DiaS), and a solar panel and power pack set (Yeti 400 set, GOALZERO, South Bluffdale, US).
The disassembled glovebox was kept in a metal box (80 × 60 × 41 cm) with other necessary materials (disinfectant solution, extraction kits, filter tips, racks, vortex, heat block, autoclavable plastic bags and personal protective equipment (PPE).The total weight was 28 kg for the box and 16 kg for the DiaS.Sample inactivation and RNA extraction using the SE kit were done in the glovebox (Figure 2A,B).This allowed handling of hazard group 4 samples.The RT-RPA assay was performed in the DiaS (Figure 2A,C) containing the ESEQuant TS2 device with integrated touchscreen to operate the device and display the results (Qiagen, Lake Constance GmbH, Stockach, Germany).The DiaS was assembled using a trolley case (63 × 50 × 30.2 cm, Peli, Düsseldorf, Germany).The bottom layer of the DiaS contains foam to adsorb shocks during transportation which is covered by a PVC top layer fixed around inserted devices to provide water and chemical resistance (Figure 3, [12]).
Performance parameters of the test (sensitivity (Se), specificity (Sp), positive (PPV) and negative predictive  values (NPV) were estimated for each of the assays using real-time RT-PCR assays as reference test.The 95% confidence interval (CI) of performance parameters was calculated based on the exact binomial test.P values are derived from the exact binomial test.The calculated Se and Sp were considered statistically significant for p values < 0.05.We used Fisher's exact test to compare RT-RPA performance parameters in comparison with W-PCR and A-PCR as the reference method at different Ct ranges.

Inactivation
The inactivation of EBOV by the SE extraction procedure was confirmed in VeroE6 cells inoculated with SE extracts which were all negative in IFA.PCR results at passage 4 ranged from Ct 32 to undetectable.Since the IFA was negative, the PCR results were assumed to be due to remnant input RNA but not to actively replicating virus.

Analytical sensitivity and specificity of the RT-RPA assay
W-PCR and RT-RPA detected RNA standards over a range of 5 to 5 × 10 5 genome copies (GC)/reaction and 50 to 5 × 10 5 genome copies/reaction, respectively.RT-RPA assays could detect as little as 5 GC/reaction of a molecular RNA standard (data not shown) and 15 GC/reaction in EBOV-spiked human plasma samples (Figure 4A).No cross-detection of important differential diagnostic pathogens or any other filoviruses was observed for the Ebola RT-RPA assay (Table 2).

Performance of RT-PCR and RT-RPA assay using sera
Using a total of 141 sera extracted with QC, RT-RPA and W-PCR performances were assessed using the WHO-approved A-PCR as reference.Against the  A-PCR, the RT-RPA yielded a lower PPV (82% vs 100%, p = 3.05 × 10 −5 ), a higher corresponding Se (97% vs 91%, p = 0.19), a higher NPV (97% vs 86%, p = 0.09) and a lower Sp (79% vs 100%, p = 3.45 × 10 −4 ) than against the W-PCR (Table 3 rows 1 and 3, Table 4).The tendency of the results was even more pronounced in the subset of 84 samples with low viraemia (Ct values > 30, Table 3 rows 2 and 4).The difference between the PCR assays was analysed and revealed a reduced Se (77%) for the A-PCR compared with the W-PCR (Table 3 rows 7-8).
Samples determined as positive by the W-PCR but negative by the RT-RPA were also negative in the A-PCR, which missed some additional samples.There was no case of a negative RT-RPA result being positive in the A-PCR (Table 5).

Performance of RT-PCR and RT-RPA assay using swabs
In a preliminary test of RT-RPA efficiency on SE extracts from 47 swabs from deceased patients, all 47 samples scored positive in the W-PCR and the RT-RPA.Therefore, combined SE extraction and RT-RPA were deployed in the mobile laboratory and altogether 928 post-mortem swab samples (including the 47 preliminary ones) were tested.All 928 samples were also extracted by QC and tested by W-PCR and A-PCR.Overall, 120 samples scored positive both in W-PCR and RT-RPA, and only 67 of a subset of 83 samples scored positive in A-PCR.In reference to QC extraction and W-PCR, SE extraction and RT-RPA yielded a Se and Sp of 100% each (PPV: 100%; NPV: 100%).Since the results of W-PCR and RT-RPA were concordant, the significance of the results was not calculated (Table 5).
The prevalence of positives as tested by W-PCR and RT-RPA in the 928 swabs was 12.9%.Of the 928 postmortem samples tested, 790 were from suspected cases for whom no signs of disease were recorded and 138 from suspected cases for whom information on symptoms and onset of disease ranging from 1 to 35 days before death were available.Of the 120 positive cases, 53 belonged to the group without recorded symptoms and 67 belonged to the group with symptoms.Positive results were most frequent around day 6 after disease onset and no positive results were obtained later than 14 days after onset of disease (Figure 4B).

Deployment of the mobile laboratory to the local hospital in Guinea
The mobile laboratory was easy to transport to the point of need (Figure 2D-F).The setup of the mobile laboratory including the assembly of the glovebox and donning the PPE took ca 30 min.The SE step was performed in the glovebox for up to 10 samples in 30 min, while the RT-RPA needed 20 min including pipetting steps and mixing.We were able to power the mobile laboratory (peak energy need: 173 W) with the solar battery for up to 16 hours.Before moving to another spot, the glovebox and DiaS were disinfected with 2% bleach or 0.5% incidine.Altogether, setup, operation and disassembly of the unit was easy to perform in a timely manner.
Four Guinean biologists were equipped with and trained in the use of the mobile laboratory at the IPD in January 2015 in a five-day course.After a pilot phase in Guinea, the mobile laboratories were deployed in the Matoto district of Conakry to support testing of swabs from dead suspected cases, which was introduced to  improve community engagement in the EBOV response as well as community surveillance.

Discussion
In this study, we evaluated the analytical and clinical performance of an updated EBOV RT-RPA compared with reference real-time RT-PCR assays.The isothermal RT-RPA assay, which allows real-time detection of amplification from RNA samples using primers and a fluorescent restriction probe within 3 to 15 min [10].
We improved this assay by adaptating the primers to the new sequences of the EBOV strain circulating in West Africa and incorporating them into dried RT-RPA pellets.
In sera extracted by QC, the RT-RPA scored a Se of 91% and Sp of 100% in reference to the W-PCR (Se: 97% and Sp: 79% in reference to the A-PCR), which means it would miss out some weak positives while identifying all true negatives correctly.Results from SE extracted sera were similar (data not shown).Taking swabs is less invasive than taking serum, which makes it more acceptable to populations, but is also safer and easier for sampling and testing.Since SE extraction does not require the use of a centrifuge, we tried to combine the RT-RPA with SE extraction of swabs to simplify our mobile laboratory procedure.
During the analysis of the results, we noted that the widely used A-PCR was less sensitive than the W-PCR.This lower Se was also described by other teams in Guinea and Sierra Leone [14,15].A rapid detection test (ReEBOAg, Corgenix, Denver, US) was recently scored against the A-PCR with a Se of 91.8% and Sp of 84.6% and approved by the WHO for emergency use.Another recently described rapid detection test also scored a Se of 100% and a Sp of 96.6% against the A-PCR and was rated as a rule-out screening test by the authors because it would include all positives but would miss out on excluding all true negatives, therefore requiring a confirmatory test [15].Our data confirm their interpretation that the performance of these tests was underestimated when using the A-PCR as reference test.
Our data show that the combination of SE and RT-RPA is superior to the above rule-out tests as all true positive and negative post-mortem oral swabs are detected.Our previous work has shown that magnetic bead extraction is preferable to centrifuge-based extraction under field conditions as it obviates the need for a high-speed centrifuge (unpublished data).We therefore tested the novel magnetic bead-based SE extraction with its 15 min protocol.The materials for both SE extraction and RT-RPA are stable at ambient temperature (30-35 °C) for up to three months and this cold chain-independent combination proved to be well suited for field diagnostics.It scored very satisfactory results in swab extracts (Table 3, rows [13][14], indicating that the RT-RPA does not need a confirmatory test and can be used on site to correctly include positives and exclude negatives. The prevalence of EBOV in the 928 swabs tested was 12.9%.The day of death after onset of disease peaked at day 6 (range: 2-14 days) in the group of 67 swabpositive deceased for whom disease symptoms were recorded.For the ongoing EBOV outbreak in West Africa, the mean day of symptom onset is 11 days after infection and sera should ideally be collected during the acute phase of illness, within the first 10 days of the disease [2].We show here that the same is true for swabs, which could simplify diagnostics tremendously.
In 53 positive cases, symptoms were not recorded, which was mainly due to a lack of information in the records of the Safe and Dignified Burial teams that did the sampling.
When new EVD foci erupted in previously not affected western parts of Conakry in April 2015, the mobile laboratory was deployed to Matoto to support teams in charge of safe and dignified burials.Since it had been decided that all deceased should be tested, these teams collected swab samples from deceased of five neighbourhoods of Conakry (Matoto, Ratoma, Dixinn, Matam and Kaloum) and up to 50 samples had to be tested per day.The emergency response results were provided every 30 to 60 min to the field investigators and physicians.The rapidity and mobility of the RT-RPA method in the DiaS, in comparison with the average 3 to 4 h turnover with regular real-time RT-PCR, was appreciated by burial teams, health authorities, response teams and communities, as it allowed rapid clearance for normal burials deceased persons who were confirmed negative.The results also encourage the use of swabs from patients at ETCs.In that context, it would still be necessary to determine if swab samples can replace sera samples.
The deployment demonstrated that the mobile laboratory using glovebox, DiaS, SE and RT-RPA is a very good solution for decentralised biosafe diagnosis of EBOV, resulting in direct impact on community engagement for disease control.Moreover, this small mobile laboratory run by local teams is a sustainable contribution to future outbreak control.

Addendum (28 January 2016)
The World Health Organisation (WHO)-approved RealStar Filovirus Screen RT-PCR Kit 1.0 (Altona Diagnostics, Hamburg, Germany) has been widely used on Smart Cyclers throughout the Ebola outbreak in West Africa, although the manufacturer does not list this device in their instructions for use.The company published a note on its website on 10 July 2015, commenting the reported lack of sensitivity of this kit when used on a Smart Cycler.According to our findings, we consider that the reported lack of sensitivity is related to the restriction of using the kit for specific PCR cyclers only, as also found by other groups [20] previously.

Figure 2
Figure 2Mobile laboratory for Ebola virus diagnostics
Results from 138 swab samples from deceased suspected EVD cases with symptoms among a cohort of 928 samples from deceased with and without symptoms.Positive and negative results are scored over days after onset of disease.
Constance GmbH, Stockach, Germany) for real-time monitoring of fluorescence at 42 °C for 15 min, with brief mixing and centrifugation of the reaction tubes after 4 min.The resulting curves were analysed by TS2 Studio Version 1.8.2.0 (QIAGEN Lake Constance GmbH, Stockach, Germany).Increase of fluorescence intensity over time above the mean background signal A: Plasma samples spiked with inactivated Ebola virus (source: ENIVD) extracted and re-quantified by real-time PCR.Extracts (1.5 × 105, 7.0 × 102, 1.0 × 104, 1.0 × 105 Ebola genome molecules /reaction) were tested in triplicate by RT-RPA.The detection limit was 15 Ebola genome molecules/reaction in a maximum of 8 min.B:

Table 1
Primers and probe designed for the updated Zaire ebolavirus RT-recombinase polymerase amplification assay

Table 2
Pathogen nucleic acids used for evaluation of the Ebola virus RT-recombinase polymerase amplification assay BNI: Bernhard Nocht Institute; ND: not determined; GIN: Guinea; RKI: Robert Koch Institute; RPA: recombinase polymerase amplification; RT: reverse transcription; TT: threshold time.Ebola RT-RPA assay identified only Zaire ebolavirus but not the nucleic acids of other pathogens.

Table 3
Evaluation of Ebola virus diagnostic assays, serum and swab samples, Guinea, December 2014-May 2015 (n=1,069) Extraction method for RT-RPA.In all cases the reference test was tested with extracts from QC.
CI: confidence interval; CT: cycle threshold; NPV: negative predictive value; QC: QIAamp Viral Mini Kit; RPA: recombinase polymerase amplification; RT: reverse transcription; PPV: positive predictive value; SE: SpeedXtract nucleic acid extraction kit. a b Estimated proportions are given in decimals.c This comparison was tested on a smaller subset.

Table 4
Significance of the performance analysis results for Ebola virus diagnostic assays, serum samples, Guinea, December 2014-May 2015 (n=141)

Table 5
Concordance of results from Ebola virus diagnostic assays, serum and swab samples, Guinea, December 2014-May 2015 (n=928)