Within the past dozen years, outbreaks of filoviral hemorrhagic fever within the human population have been occurring with increasing frequency, with an average of 1 epidemic now occurring every 1–2 years. Many of the outbreaks have been large (involving >150 cases), necessitating rapid responses from the international community to help implement infection control and surveillance. This increased activity, combined with today's climate of bioterrorism threats, has heightened the need for high-throughput methodologies for specific detection of these high-hazard viruses in sophisticated laboratory setups and mobile field laboratory situations. Using Zaire Ebola virus as an example, we describe here the development of a high-throughput protocol for RNA extraction and quantitative reverse-transcription polymerase chain reaction analysis that is safe, fast, and reliable. Furthermore, the applicability of this method to an outbreak setting was demonstrated by correct analysis of >500 specimens at a field laboratory established during a recent outbreak of Marburg hemorrhagic fever in Angola.
Outbreaks of hemorrhagic fever caused by the Ebola and Marburg viruses can spread easily within a human population by means of contact transmission and can cause high rates of morbidity and mortality. For hemorrhagic fever due to these filoviruses, the mortality rate among infected individuals can reach 30%–90%, resulting in the deaths of hundreds of people in a matter of months . The levels of public and government concern are often extreme, demanding a comprehensive public health response to identify all possible cases of hemorrhagic fever so that chains of transmission can be broken. An essential component of this response is efficient and reliable diagnostic testing that can rapidly identify the causative agent, followed by additional testing of suspected cases as they arise. The sheer volume of persons monitored and samples tested over the course of an outbreak can be daunting. During an outbreak of Ebola hemorrhagic fever in 2000 in Gulu, Uganda , thousands of potential cases were followed, resulting in laboratory testing of >1700 serum or blood samples by means of antigencapture ELISA; of these 1700 samples, 1100 underwent additional testing by polymerase chain reaction (PCR) [3, 4]. By the end of the outbreak, there was a total of 425 cases that resulted in 225 deaths . In this setting, nearly all laboratory testing for identification of acute cases was performed at an on-site, mobile field laboratory established by the Centers for Disease Control and Prevention (CDC) at a local hospital in Gulu in the center of the outbreak. This was the first time that such a field laboratory had been set up for the purpose of identifying acute cases of Ebola hemorrhagic fever.
The high-throughput capacity of the antigen-capture assay, combined with the sensitivity of a nested reverse-transcription (RT)-PCR assay, allowed for accurate, next-day results to be provided to field teams and isolation wards. In fact, the overall success of this field laboratory established an effective new approach to outbreak management by the international response team—namely, to establish on-site filovirus diagnostic capability. The greatest benefit of an on-site field laboratory is that it can provide early identification of acute cases, thereby allowing more rapid isolation of infected individuals.
The most sensitive specific assay for early case identification is RT-PCR [4, 6]. However, the nested RT-PCR assay used in Gulu showed limited throughput capacity and greater potential for false-positive results (as a result of cross-contamination), compared with the ELISAs performed in parallel. The challenge was to develop a molecular detection method that retained high sensitivity yet increased the throughput while decreasing the number of false-positive results. A robust assay system was developed, and the effectiveness of this strategy was demonstrated in the field during an outbreak of Marburg hemorrhagic fever (MHF) in Angola in 2005. The key developments in this methodology lay not only in the use of real-time RT-PCR for detection of viruses causing hemorrhagic fever, which has also been developed by other investigators [7–10] but, perhaps, even more so in the establishment of a new, robust, high-throughput RNA extraction protocol that yields consistent quantities of high-purity RNA over a minimum range of 5 logs. Although the process was primarily developed for high-hazard hemorrhagic fever viruses handled in biosafety level 4 (BSL4) environments, the method(s) can be easily applied to viruses handled in non-BSL4 conditions.
Methods and Results
The high-throughput RNA extraction/quantitative RT-PCR protocol described in the present study is centered around a 96-well nucleic acid preparatory station (Applied Biosystems) that easily lends itself to similarly formatted downstream RT and PCR setups. The formatting also complements the throughput and setup of 96-well antigen-capture IgG and IgM ELISA-based diagnostic assays performed in parallel [9, 11]. This RNA extraction/quantitative RT-PCR protocol is specifically designed for extracting nucleic acid from viruses normally handled in BSL4 containment. Here, we first outline the basic protocol as a point of reference and then describe solutions to specific hurdles arising during protocol development that are associated with high-hazard viruses. Finally, we present data to validate the overall extraction platform for use in an established laboratory environment as well as an outbreak setting. The validation of this method, using Ebola virus (EBOV) as the prototype example, describes safety testing (evaluation of virucidal activity) of RNA extraction buffers, as well as assessments of sensitivity, reproducibility, cross-contamination, and utility with multiple sample types.
All work with infectious EBOV or Marburg virus (MARV) was performed under BSL4 containment at the Centers for Disease Control and Prevention (CDC; Atlanta) or within a limited-access field laboratory in which personnel wore protective clothing and positive-pressure air-purifying respirators.
RNA extraction protocol. The extraction platform is designed to isolate RNA from either cellular (C) or noncellular (NC) liquid environments, by use of either C or NC lysis buffers, respectively. In general, we prefer to use the NC lysis buffer because it works well with both sample types, and, therefore, it is the buffer used throughout this protocol. In brief, 50 µL of sample are added to each well of a 96-well 1-mL chimneystyle masterblock (Greiner) containing 300 µL of 2× NC lysis buffer (Applied Biosystems). The 2× NC lysis buffer has a strong tendency to bubble; therefore, we added antifoam Y-30 (Sigma) diluted 1/2000 in lysis buffer. After addition of the sample, the masterblock is heat sealed for 4–5 s by use of a manual heat sealer (ABGene) and 20-µm nonpeelable seals (ABGene). The sealed masterblock is then dunked out of the high-containment laboratory in >3% disinfectant solution (Amphyl; Reckitt Benckiser).
To ensure complete virus inactivation in 2× lysis buffer, the sealed masterblock is shaken to mix the sample with lysis buffer and then is incubated at room temperature for at least 20 min before it is opened outside of biosafety containment. Just before the foil seal is punctured, the top surface is wiped with 10% bleach to minimize any spurious nucleic acid contamination, followed by an additional rinse with RNAse-free water. The foil seal is then punctured with dry 200-µL tips, followed by dilution to 1× with cold PBS (Ca++ and Mg++ free) containing poly-A (Sigma) carrier RNA (20 µg/well). An exogenous RNA target can be added at this time to serve as a control for RNA extraction efficiency. The RNA is incubated for 1 h on wet ice to facilitate RNA precipitation. Each sample (∼550 µL) is then added to the 96-well RNA purification tray and is extracted using a 6100 Nucleic Acid Preparation Station (Applied Biosystems) according to the manufacturer's instructions. Minor modifications to the protocol are to (1) perform 2 extra washes with 1× NC lysis buffer at the beginning of the protocol to remove a mild RT-PCR inhibitor associated with serum samples, and (2) use the Absolute RNA wash (Applied Biosystems; optional in the manufacturer's protocol) to remove any residual DNA contamination and/or aid with blood collection additives, such as heparin. After the final wash, RNA is eluted using 150 µL of elution buffer.
Development of an RNA extraction method that ensures virus inactivation in fluids and tissues. The C and NC lysis buffers used in this protocol contained guanidine hydrochloride as the primary chaotrope for denaturation of protein. These proprietary buffers were supplied by the manufacturer, Applied Biosystems, with no assurances regarding the virucidal activity toward hemorrhagic fever viruses—or any other viruses—when used at the prescribed working concentration (1×) for extraction and precipitation of RNA. The Material Safety Data Sheets list the concentration range of the guanidine hydrochloride as 30%–60%. Given the highly pathogenic potential of hemorrhagic fever viruses, we could not assume that the lysis buffers were virucidal, and, thus, we sought to determine whether EBOV was indeed rendered inactive by immersion in 1× lysis buffer at room temperature. An essential component of this safety determination was the ability to assay for virus viability on tissue culture cells. Because concentrated guanidinium is highly toxic to living cells, treated virus was first diluted 240-fold in physiological buffer and centrifuged to equilibrium (60,000 g for 2 h in a SW28 rotor) through a 20% sucrose cushion that was layered on top of 60% sucrose. With this procedure, the highly diluted virus was reconcentrated at the 20%–60% interface, generally resulting in recoveries of >50% of input untreated virus (data not shown). Because all recovered virus was contained in a relatively small volume (∼5 mL), the entire fraction could be easily assayed for infectivity, thereby allowing a more rigorous assessment of virus inactivation than would be performed if the samples were simply diluted 100-fold after chaotrope treatment and then assayed directly at the diluted concentrations . Surprisingly, our initial plaque assay of NC lysis buffer showed that it did not provide complete virus inactivation at the 1× concentration. Treatment with 1.5×105 plaque-forming units (pfu) of EBOV for ∼10 min at room temperature resulted in only a ∼5-fold reduction in virus titer, compared with that measured for the untreated control (table 1). In comparison, no infectious virus was recovered when Tripure (Roche), a solution that additionally contained phenol, was mixed 5:1 with virus.
To determine the dilution at which the stock 2× lysis buffer was virucidal, if at all, virus was treated and assayed as described above, except the lysis buffer was mixed with virus at increasing ratios of lysis buffer to virus, starting from 1:1 and increasing to 5:1. Also in this experiment, the entire fraction collected from the 20%–60% interface was assayed for infectivity on Vero E6 cells. As shown in table 2, a minimum ratio of 3:1 (stock 2× lysis buffer to virus) was necessary for complete virus inactivation. Ratios of both 1:1 (the manufacturer's recommendation) and 2:1 resulted in incomplete inactivation of 2×105 pfu of EBOV. Thus, so long as lysis buffer and virus were mixed at a ratio ⩾3:1, virus inactivation was complete. On the basis of these data, a ratio of 6:1 was adopted for the final protocol, to incorporate a safety margin in the event that a virus sample is accidentally added twice during the initial extraction setup.
Extraction of viral RNA (vRNA) from infected tissues is a common need for diagnostic and experimental purposes. Therefore, we sought to determine the chaotropic penetration and virus inactivation capabilities of C and NC 2× lysis buffers on kidney and liver tissue specimens obtained from mice infected with mouse-adapted Zaire EBOV (ZEBOV). To set up this experiment, conditions were chosen that would likely provide the greatest utility for extraction of RNA from diagnostic or experimental tissues.We chose a reasonable amount of tissue (∼30 mg), which, after homogenization, would be unlikely to clog the filtration well of the RNA purification plate yet should still contain a sufficiently high viral load to serve as a rigorous test of viral inactivation.
As shown in table 3, previously frozen tissue specimens were immersed for 2, 4, and 8 h (in duplicate) in either C or NC lysis buffers. To control for any virus inactivation due to incubation at room temperature, infected tissue specimens of similar size were incubated in PBS (for 8 h only, also in duplicate). After each of the indicated incubation times, tissues treated in lysis buffer were immersed and washed for 10 min in 25 mL of sterile PBS to remove residual chaotrope. Each tissue was then homogenized (see the table 3 note) and subsequently assayed in its entirety for virus infectivity on Vero E6 monolayers, as described above. Results show that chaotrope penetration and virus inactivation were complete in all tissues tested, even samples inactivated for only 2 h. The homogenized chunks of untreated tissues revealed virus titers of ∼3×104 pfu/g for kidney tissue specimens and ∼2×108 pfu/g for liver tissue specimens. After adjusting for the size of the tissues tested, it was found that each piece had virus titers of ∼1×103 pfu for kidney tissue specimens and ∼5×106 pfu for liver tissue specimens, before inactivation. Thus, for the liver sample, >106 total pfu were assayed for virus inactivation. Results of IFA performed on cell scrapings from virus isolation attempts from lysis buffer-treated virus were negative, whereas results of IFAs performed on PBS-treated samples were brightly positive (data not shown).
Development of a real-time RT-PCR assay for detection of ZEBOV. To quantitatively measure the virus-specific RNA yields by use of the high-throughput RNA extraction platform, we developed a real-time RT-PCR assay designed to detect all known strains of ZEBOV. To do this, we aligned all publicly available nucleoprotein sequences and designed the primers and probe in areas of high conservation (figure 1A) while adhering to commonly accepted strategies for quantitative RT-PCR probe/primer design . The resulting quantitative RT-PCR assay was then tested to determine its ability to quantitatively detect ZEBOV RNA, by use of ∼2.5 kb of negative-sense in vitro RNA transcripts corresponding to the entire nucleoprotein open reading frame. Linear regression analysis shows quantitative detection over an 8-log range, with a predicted sensitivity of 3–38 RNA copies (figure 1B). This assay was specific for ZEBOV and did not react with close relatives, such as Sudan EBOV or MARV (data not shown).
Development of a protocol for safely removing samples from the BSL4 environment. A necessary step for removal of samples from our BSL4 laboratory is decontamination of all exposed surfaces by means of immersion in ⩾3% Amphyl (or 10% bleach) for ⩾3 min, to ensure complete virus inactivation. With this Amphyl immersion step in mind, a protocol was needed for safe export of samples out of the BSL4 laboratory in lysis buffer, without sample-to-sample cross-contamination or Amphyl penetration into the sample, yet with efficient and consistent vRNA recovery maintained. We found that heatsealing, 96-well, 1-mL masterblock plates with nonpeelable (20- ·m) foil seals worked best for this purpose. With the use of chimney-style masterblocks with a slightly raised lip at the top of each well, the Amphyl disinfectant has complete accessibility to the outside of each well, whereas the interface between the nonpeelable foil seal and the plastic at the top of each well is never exposed. In addition, the interface is flash-heated to well over 100°C, and the inside of each well is decontaminated by the lysis buffer.
This exit protocol for samples in chaotrope was tested for efficient vRNA recovery without cross-contamination by aliquoting 5000 pfu of EBOV, diluted in cell culture media (containing 10% fetal calf serum [FCS]), into the 96-well masterblock in a checkerboard configuration (figure 2A). All 96 wells contained NC lysis buffer. Total RNA was extracted as described in the “RNA extraction protocol” subsection of the Methods and Results section and assayed for EBOV RNA by use of the quantitative RT-PCR assay described above. In this way, any leakage of a spiked sample into a neighboring well would be easily detected, and, conversely, leakage of Amphyl into one of the spiked wells would likely be detected as a false-negative result. As shown in figure 2B, 4 complete runs totaling 96 EBOV-spiked wells were examined this way without a single cross-contamination or Amphyl penetration event noted. In addition, the RNA recovery was consistent and of expected quantity, as is demonstrated by quantitative RT-PCR cycle threshold (Ct) values that are well within the range predicted by the standard curves (described in the following subsection) in which virus was diluted in either serum or whole blood.
Demonstration that the high-throughput method works well for both serum and whole blood. After establishing that we could safely extract RNA from media spiked with high titers of EBOV without cross-contamination, we sought to determine the ultimate sensitivity and reproducibility of the high-throughput RNA extraction protocol in different sample types. To do so, ZEBOV was diluted in either whole blood or undiluted FCS over a 5-log range, with the total RNA of each dilution extracted in replicates of 8. Each of these extractions was then assayed for vRNA by use of the quantitative RT-PCR assay described above. As shown in figure 3A, similar regression curves, whose slopes are within 0.2 U, are obtained regardless of the virus diluent used. Furthermore, as shown in figure 3B, the end point of detection was consistently ∼.05 pfu per extraction (2 pfu/ mL starting material). This limit of detection (LOD), compared with the LOD in which we measured RNA copies (figure 1B), predicts a ratio of RNA copies to plaque-forming units of ∼1000, which is similar to the ratio of RNA copies to plaqueforming units previously calculated by strand-specific quantitative RT-PCR performed for the detection of Sudan EBOV . Together, these findings indicate that the RNA extraction protocol routinely extracts to levels of <50 RNA copies per assay. As a further indication of the reproducibility of this method, we set up separate experiments in which 96 replicates of 5000 pfu were extracted from either whole blood or serum (figure 3C). These experiments produced average Ct values within ∼1 Ct of that predicted by their respective standard curves, with SDs well below 1 Ct.
Diagnosis of MHF in an outbreak setting. During the spring of 2005, a large outbreak of MHF erupted in the Uige province of northern Angola [9, 15, 16]. As part of the international response to the outbreak, the CDC established a mobile field laboratory in the capital city of Luanda. This was operated in parallel to a field laboratory set up in Uige by the National Microbiology Laboratory of the Public Health Agency of Canada (PHAC). The workhorse diagnostic assay used by the CDC laboratory was a quantitative RT-PCR assay designed to detect all known strains of MARV . This assay was used in conjunction with the high-throughput RNA extraction protocol described here. The field laboratory was operated for ∼2 months, during which time a total of 505 samples (175 blood, 326 oral/nasal swab, and 4 breast milk samples) underwent PCR analysis. Of these samples, 180 tested positive for the presence of MARV. The concordance with the PHAC laboratory, which also used PCR-based assays, was >99%, ultimately resulting in only a single case for which the findings of the 2 laboratories were in discordance. These results, combined with previous data obtained at the CDC, validated our high-throughput protocol in both a sophisticated laboratory setting and a field situation.
In the present article, we describe a methodology for highthroughput detection of high-hazard viruses nominally handled in a BSL4 environment. In our study, we combined a newly developed quantitative RT-PCR assay for ZEBOV with a 96-well RNA extraction platform, to demonstrate consistent and quantitative RNA extractions over a 5-log range. Furthermore, we demonstrated that the extraction method was safe, reliable, and applicable to a field setting, as demonstrated by the successful analysis of >500 clinical specimens from a recent large outbreak of MARV infection in 2005 in Angola.
One of the most surprising observations during this study was the inability of the standard guanidine-based chaotropic lysis buffer to completely inactivate infectious virus. The mechanism of the RNA extraction chemistry used here involves selectively precipitating total RNA into micellar structures that are trapped on a filtration membrane and subsequently washed. The micelles are then dissolved, thus releasing the purified RNA. Filoviruses are typical enveloped, single-strand negativesense RNA viruses and are not known to be particularly resistant to inactivation, compared with some viruses (such as poliovirus) that are acid resistant and are routinely purified in 1% SDS. We speculate that the failure to inactivate virus at the 1× concentration of lysis buffer resulted from (1) the exact concentration of the guanidine (which is proprietary) not being sufficiently high enough for complete denaturation, and/or (2) some intact virus being precipitated and encapsulated into micellar structures and sequestered from the denaturing action of the guanidine. At higher concentrations of the lysis buffer, virus denaturation was increasingly more complete.
After we performed the RNA extraction protocol, we used a high-capacity random hexamer cDNA synthesis step (figure 1B), which introduced an additional 3-fold dilution of the extracted RNA. One tenth of the cDNA synthesis was then used to program the PCR amplification. Collectively, this procedure dictates that only 1/30 of the extracted RNA was actually assayed in each reaction. Compared with a 1-step protocol, this 2-step protocol used half of the RNA, which, at first glance, might not seem to be beneficial. However, there are 3 major advantages afforded by the 2-step method: (1) separation of the RT step from PCR amplification allows for each of these steps to proceed under optimal buffer conditions, resulting in greater RT and cDNA amplification efficiency; (2) generation of separate total RNA conversion to cDNA allows for up to 9 additional nonmultiplexed assays before having to go back to use of the original RNA; and (3) cDNA tends to be more stable at higher temperatures, which is a valuable consideration for downstream analyses of samples prepared in a field setting, given that the cold chain from a field laboratory is often inconsistent. In fact, side-by-side experiments with a 1-step protocol showed no increased sensitivity, compared with the experiments with the 2-step protocol (data not shown).
One of the best features of this system is the high-throughput surge capacity afforded by the 96-well extraction format. At full capacity, a few hundred samples could be analyzed in 1 day. Unlike our experience with PCR during the outbreak of Ebola hemorrhagic fever in Uganda in 2000, the ultimate bottleneck in throughput capacity is now in the initial phases of sample preparation—namely, cataloguing, labeling, and aliquoting—and not in the RNA extraction/quantitative RT-PCR analysis. In fact, during the recent outbreak of MHF in Angola, the throughput capacity allowed us to easily process an accumulation of >70 samples in 7 h on the first day that our laboratory was established. Our increased surge capacity also has obvious implications toward our increased ability to respond to threats of intentional release of hemorrhagic fever viruses, both domestically and internationally [17, 18].
In recent years, newly developed hemorrhagic fever assays [19, 20] and the emergence of mobile diagnostic capabilities [4, 8, 9], such as the one described in the present study, have become tools with enormous potential for helping manage outbreaks of filovirus infection in sub-Saharan Africa. Clearly, the greatest benefit is early case identification for rapid isolation of infected individuals. However, another important but less quantifiable benefit is the aid to response teams charged with monitoring contacts of individuals with known cases of viral hemorrhagic fever. The initial symptoms of filovirus hemorrhagic fever are general and not specific to EBOV or MARV infection. In these instances, a laboratory diagnosis, in the hands of mobile response teams, can provide concrete and persuasive arguments to justify to the patient why it is in the interest of the community for them to travel to the hospital isolation ward when they might not otherwise choose to do so. A final benefit of high-throughput diagnostic technology that is usable in the field will be to aid in the implementation of several promising vaccines and treatment regimens that are on the horizon [21–24]. The ability to precisely quantify viral loads in multiple and varied specimens in a field setting will be invaluable for monitoring efficacy of these new therapeutics or vaccine approaches in outbreaks.
We thank Heinz Feldmann, Allen Grolla, Lisa Fernando, Steven Jones, Ute Ströher, and Jim Strong from the National Microbiology Laboratory of the Public Health Agency of Canada for their assistance in processing the field specimens obtained in Uige, Angola. The management of the outbreak was greatly assisted by the Ministry of Health, Republic of Angola; the World Health Organization; and the International Response Team.We also thank Filomena Gomes da Silva of the Instituto Nacional de Saúde Publica and Fernando del Castillo and Amilcar Tanuri of the Centers for Disease Control and Prevention (CDC; Atlanta) Global Aids Program, for their excellent logistical assistance with the CDC field laboratory. The efforts of Daniel Kertesz of the World Health Organization were also critical for sending the initial sets of diagnostic specimens to the CDC. We would also like to thank the US Embassy to Angola and the US Office of Foreign Disaster Assistance for their essential financial and logistical support, as well as the Chevron Corporation for the loan of a crucial electrical generator. The CDC field laboratory was well supported by Pierre E. Rollin, Amy L. Hartman, Martin J. Vincent, and James A. Comer, and technical assistance was received from Darcy A. Bawiec, Bobbie R. Erickson, Jennifer B. Oliver, Deborah L. Cannon, Kimberly A. Slaughter, and Thomas L. Stevens. We would also like to thank Mike Frace and Brian Halloway for making available multiple components of the CDC Biotechnology Core Facility, which immensely facilitated this study.
Supplement sponsorship. This article was published as part of a supplement entitled “Filoviruses: Recent Advances and Future Challenges,” sponsored by the Public Health Agency of Canada, the National Institutes of Health, the Canadian Institutes of Health Research, Cangene, CUH2A, Smith Carter, Hemisphere Engineering, Crucell, and the International Centre for Infectious Diseases.