Research ArticleDiagnostics

A point-of-care diagnostic for differentiating Ebola from endemic febrile diseases

See allHide authors and affiliations

Science Translational Medicine  12 Dec 2018:
Vol. 10, Issue 471, eaat0944
DOI: 10.1126/scitranslmed.aat0944

Discerning febrile diseases

Many infectious diseases present with common clinical symptoms, such as fever, which complicates diagnosis at the point of need. Sebba and colleagues developed an assay using surface-enhanced Raman scattering (SERS) nanotags to distinguish Ebola virus infections from Lassa and malaria. The no-wash triplex assay workflow adds a small volume of blood and buffer to dried SERS reagents and delivers a readout within 30 min. The assay detected parasite- and virus-specific antigens spiked into the blood, Ebola infections in nonhuman primates, and Ebola and malaria infections in human blood samples collected from endemic regions during field testing. Although further testing is required, this assay could be useful during febrile disease outbreaks.

Abstract

Hemorrhagic fever outbreaks such as Ebola are difficult to detect and control because of the lack of low-cost, easily deployable diagnostics and because initial clinical symptoms mimic other endemic diseases such as malaria. Current molecular diagnostic methods such as polymerase chain reaction require trained personnel and laboratory infrastructure, hindering diagnostics at the point of need. Although rapid tests such as lateral flow can be broadly deployed, they are typically not well-suited for differentiating among multiple diseases presenting with similar symptoms. Early detection and control of Ebola outbreaks require simple, easy-to-use assays that can detect and differentiate infection with Ebola virus from other more common febrile diseases. Here, we developed and tested an immunoassay technology that uses surface-enhanced Raman scattering (SERS) tags to simultaneously detect antigens from Ebola, Lassa, and malaria within a single blood sample. Results are provided in <30 min for individual or batched samples. Using 190 clinical samples collected from the 2014 West African Ebola outbreak, along with 163 malaria positives and 233 negative controls, we demonstrated Ebola detection with 90.0% sensitivity and 97.9% specificity and malaria detection with 100.0% sensitivity and 99.6% specificity. These results, along with corresponding live virus and nonhuman primate testing of an Ebola, Lassa, and malaria 3-plex assay, indicate the potential of the SERS technology as an important tool for outbreak detection and clinical triage in low-resource settings.

INTRODUCTION

The accurate diagnosis of patients with an acute infectious disease is critical to effective treatment, but it is often not straightforward. The early symptoms of many acute diseases are similar, making diagnosis based on clinical symptoms alone extremely challenging (1, 2). This difficulty is especially notable in areas of the world that have a high burden of disease and in which a fever could indicate infection with one of many different diseases (3, 4). In these areas, a diagnostic test with the ability to differentiate among many potential causative agents is critical to ensuring the delivery of effective and efficient patient care.

From 2014 to 2016, the Ebola virus (EBOV) outbreak in West Africa highlighted the need for an expanded arsenal of diagnostic testing approaches. Empirical diagnosis based on clinical symptoms (for example, fever) is the most frequently used diagnostic tool in low-resource settings but is not very discriminatory in locations with a high incidence of malaria and regular occurrence of Lassa fever virus (LASV), and therefore, diagnosis of EBOV during the outbreak was primarily performed using a reverse transcription polymerase chain reaction (RT-PCR) assay (5, 6). Although RT-PCR was used successfully to diagnose Ebola infections, its deployment, and therefore clinical impact, was limited because of the infrastructure and training required to accurately run the assay. These limitations highlight the need for portable diagnostics with ambient temperature–stable reagents that can be deployed in low-infrastructure settings and that are capable of differentially diagnosing Ebola from other endemic diseases that present with similar clinical symptoms, such as malaria.

To bridge this gap, we developed a multiplexed diagnostic platform compatible with austere environments. The platform is based on a protein detection technology that uses surface-enhanced Raman spectroscopy nanoparticle tags (SERS nanotags). A number of practical advantages make this platform a strong candidate for a multiplexed point-of-need diagnostic assay. The SERS nanotag spectra are excited in the near infrared (NIR), minimizing optical background noise from the biological matrix. Unlike broad, featureless fluorescence emission spectra, SERS spectra have multiple sharp emission peaks that correspond to the vibrational mode of the SERS emitter, enabling multiple nanotags to be differentiated with high accuracy (7, 8). The SERS nanotags are engineered to be highly stable, avoiding a requirement for cold-chain transport and storage, and can be detected in complex biological fluids without sample preparation or wash steps, which simplifies user workflow. SERS nanotags with different optical signatures can be combined in a single assay tube with negligible interference, allowing easy multiplexing and enabling differential diagnosis. In addition, SERS nanotags can be incorporated into a sandwich immunoassay by using a pair of capture/detector antibodies, allowing us to leverage the specificity advantages afforded by this dual-binding approach.

With the goal of creating a diagnostic tool that would be useful in an outbreak of EBOV in an area with multiple “confuser” diseases, we first developed singleplex assays to detect proteins from three different infectious diseases: malaria (Plasmodium falciparum), EBOV, and LASV. We then developed assay readers compatible with operation over a range of environmental conditions, including lack of electricity. Using multiplexed, temperature-stable assay reagents that were prepackaged in an individual assay tube format, we tested samples collected in the field (for malaria) and repository samples from patients with EBOV disease. These results add an important new approach for the development of multiplexed point-of-care (POC) diagnostics.

RESULTS

SERS assay design and operation

SERS is an inelastic optical scattering phenomenon in which incident light is scattered at multiple distinct wavelengths based on the vibrational modes of the molecule undergoing scattering (the “Raman reporter”). Placement of the Raman reporter on the surface of a gold nanoparticle enhances the strength of the scattering by 4 to 8 orders of magnitude (9-11), enabling the Raman scattering signals to be detected by lower-cost optics suitable for point-of-need clinical diagnostics. Rather than attempting to detect proteins directly by SERS signal, we instead created a stable optical signal by encapsulating the SERS-active Raman reporter between a 60-nm gold nanoparticle, which acts as the nanotag core, and a 20- to 30-nm silica shell (Fig. 1A). Different Raman reporters provide different SERS spectra (Fig. 1B), with each spectrum having a unique optical “fingerprint”; this enables each nanotag to be used to detect a different target. Sulfhydryl groups are incorporated into the outer surface of the silica to provide convenient functional groups for bioconjugation to affinity reagents. The silica shell stabilizes the optical signal in complex biological fluids such as blood, preventing desorption of the Raman reporter from the gold surface and blocking spurious SERS signals from the blood sample itself.

Fig. 1 Schematic and representative spectra of Ebola, Lassa, and malaria detection using SERS nanotag technology.

(A) Schematic of BD’s SERS nanotag technology. (B) Representative spectra from three SERS nanotags used in a sandwich immunoassay to detect Ebola, Lassa, and malaria histidine-rich protein 2 (HRP2) antigen. (C) Schematic of the HNW sandwich immunoassay using the SERS nanotags.

The stable SERS signal from the SERS nanoparticle has several unique properties that make it well suited for multiplexed POC diagnostics. First, a single 785-nm excitation source can be used, regardless of the choice of Raman reporter, and the scattered SERS signal is read in the NIR (820 to 914 nm), where clinical samples such as blood have minimal absorbance. This enables the SERS signal to be read directly in a whole blood sample without wash steps or other sample manipulation such as centrifugation. Second, because each Raman reporter provides a unique optical fingerprint, SERS tags with different Raman reporters can be readily differentiated from each other in complex mixtures, enabling multiplexed assays (where each SERS tag detects a different target) within a single assay tube.

Immunoassay development

Leveraging the unique optical properties of the SERS nanotags, we used the homogeneous no-wash (HNW) assay format, which is a particle-based sandwich immunoassay that does not require sample preparation or wash steps. This allows for a simple workflow that is well suited for outbreak detection in low-resource settings. Antibodies to the target (antigens specific to Ebola, Lassa, or malaria) are conjugated both to a SERS nanotag and to a magnetic microparticle. The antibodies may be either monoclonal or polyclonal, provided that the target antigen can be “sandwiched” (bound simultaneously to the antibodies on both the SERS nanotags and the magnetic particles). In this assay, the blood sample of interest is mixed in a tube containing liquid or dried SERS nanotags and magnetic microparticle reagents (Fig. 1C). Although dried reagents are preferred for field testing, liquid reagents are often used for early assay development and feasibility work. After reagents and sample are mixed, a magnet pulls magnetic microparticles and any magnetic microparticle–antigen–SERS nanotag complexes to the side wall of the reaction vessel (Fig. 1C). A laser outside the sample illuminates the side wall of the reaction vessel containing the concentrated magnetic microparticles; any measured SERS signal is proportional to the concentration of the target.

This technology enabled the development of a simple POC assay workflow for use in field testing (fig. S1). SERS tags and magnetic microparticle reagents for each pathogen were dried in a microcentrifuge tube to provide a single-use, temperature-stable disposable that minimized user steps by obviating the need for users to add liquid assay reagents to the tube. The clinical sample (45 μl of whole blood) and associated lysis buffer were added to the tube to rehydrate the dried assay reagents, and the tube was subsequently mixed for 25 min on an end-over-end rotary mixer. At the end of the mixing period, the mixing unit applied a bar magnet to the side wall of all the tubes for 5 min, pelleting the magnetic microparticle–SERS tag sandwich complexes on the side of the tube. The user then removed each tube in turn from the mixer and placed it on an optical reader for 30 s. The optical reader measured the SERS spectrum, which is a superposition of all the SERS spectra found in the magnetic SERS nanotag sandwich complexes. The reader provides a tag intensity that corresponds to the amount of each antigen in the clinical sample, removing the need for the end user to interpret the data.

Assay development

We first established the feasibility of developing HNW assays for malaria and two viral pathogens, EBOV and LASV. Malaria was included in the assay because it is a common cause of disease in areas where EBOV and LASV infections occur (3, 12) and presents with symptoms that are similar to the symptoms of early stage Lassa and Ebola infections. For malaria, we chose to develop an assay against HRP2, a common biomarker for detecting P. falciparum, which has been shown to be more sensitive than other malaria biomarkers and detects very low densities of P. falciparum (13). For EBOV, we chose to develop assays against the viral matrix protein VP40 (14), an antigen in Ebola lateral flow assays used successfully during the 2014–2016 outbreak (15). VP40 is an abundant antigen that is present in infected cells, infected virions, and virus-like particles that are shed from infected cells and that is present in the blood of infected individuals (16). The LASV assay targeted the nucleoprotein (NP), which is a well-known target for the development of antigen-based assays (17, 18).

All three assays were tested using recombinant antigen spiked into buffer and were confirmed to have good analytical performance (fig. S2). These results encouraged us to test the performance of these assays in antigen-spiked blood and using clinical specimens (malaria) or live virus under biosafety level 4 (BSL-4) conditions (Ebola and Lassa). The malaria assay demonstrated an excellent analytical sensitivity for recombinant antigen spiked into whole blood (Fig. 2A). For the Ebola assay, our primary aim was to detect Ebola Zaire, the species that has been responsible for most of the outbreaks. Figure 2B shows results of the singleplex Ebola assay at a range of Ebola Kikwit (Zaire) concentrations. The EBOV assay recognized Ebola Kikwit at concentrations between 1 × 105 and 1 × 106 plaque-forming units (PFU)/ml when spiked into healthy whole blood (Fig. 2B). Tests using our SERS LASV assay with different concentrations of Lassa Josiah showed reliable detection down to about 5 × 105 PFU/ml (Fig. 2C).

Fig. 2 Titration curves of singleplex assays and inclusivity testing in blood.

Titration curve of (A) malaria (P. falciparum) HRP2 recombinant antigen, (B) live Ebola Kikwit virus, and (C) live Lassa Josiah virus. (D) Malaria inclusivity testing using the singleplex HRP2 assay with malaria-negative whole blood samples, P. vivax serum samples, and P. falciparum whole blood samples. (E) Ebola inclusivity testing using the singleplex assay with high concentrations of various strains and species of EBOV and (F) Lassa inclusivity testing using the singleplex assay with high concentrations of various LASVs. Figure results of assays run in triplicate; error bars reflect SD.

Performance and inclusivity against the malaria parasite were tested using 50 healthy, malaria-negative blood samples, 20 P. falciparum–positive whole blood samples, and 20 P. vivax–positive serum samples procured from commercial vendors and collected under institutional review board (IRB) approval for general research use. These results show detection of HRP2 in all P. falciparum–positive samples and, as expected, no HRP2 detection in the P. vivax–positive samples (Fig. 2D). Inclusivity testing showed similar performance between EBOV Kikwit and EBOV Makona (Fig. 2E, also Zaire) and limited affinity for other Ebola species: Sudan virus and Bundibugyo virus (fig. S3). The limited reactivity with Ebola Sudan and Bundibugyo reflects the fact that the VP40 antigen is well conserved among the EBOV Zaire strains but is not as well conserved between Ebola species. The LASV assay recognized the New Jersey, Weller, and Pinneo LASVs (Fig. 2F). Thus, the SERS technology provides analytical sensitivities similar to those found in enzyme-linked immunosorbent assay (ELISA) but in a format that tests for multiple pathogens simultaneously with a very simple workflow.

It should be noted that breadth of coverage within the different viral diseases showed some variability in our assays. The LASV assay showed reactivity across all clades tested (I, III, and IV) but lower reactivity with Weller and Pinneo. The broad sequence diversity of LASV suggests that multiple probes for this virus may be advantageous for increasing breadth of detection in future assays.

Multiplexed assay development

The strong performance of the individual malaria and hemorrhagic fever virus assays led us to test their performance in a multiplexed assay panel. Multiplexed detection of Ebola, Lassa, and malaria mixes three pairs of SERS nanotag/magnetic microparticle reagents, with each pair containing antibodies specific to a single disease target and using a spectrally distinct Raman reporter (Fig. 1B) on the respective SERS nanotags. When in the presence of their target antigen, SERS tags will bind and form a sandwich complex with the corresponding magnetic particles and will be detected in the pellet. If no target is present, only magnetic microparticles are pulled to the side of the reaction vessel, and no SERS signal is detected. To demonstrate the performance of the multiplexed assay, the malaria HRP2 assay reagents were combined with the EBOV and LASV reagents to create a triplex assay, with all reagents mixed in a single tube. The assay was tested by spiking each tube with live EBOV, live LASV, or a recombinant malaria HRP2 antigen. In addition, the triplex assay was also tested against Zika and dengue viruses. The multiplexed assay responded specifically to each target antigen, with no substantial cross-reactivity between the multiplexed assays (Fig. 3). A small nonspecific Ebola signal was observed, but this was well below the positivity threshold for the Ebola assay. No cross-reactivity with Zika or dengue viruses was observed.

Fig. 3 Triplex assay tested with various antigens spiked.

Signal seen from the EBOV, Lassa, and malaria assays is shown in blue, orange, and gray, respectively. Main figure shows signal generated after addition of ZIKV (MR766), ZIKV (PRVABC59), or DENV2 at 1 × 106 PFU/ml or recombinant Lassa NP, EBOV VP40, or P. falciparum HRP2 antigen to blood samples. Each bar represents an average of three experimental replicates; error bars represent SD.

Nonhuman primate testing

Motivated by the ability of the multiplexed SERS assay to detect live EBOV and LASV in spiked healthy human blood samples, we sought to characterize assay performance in nonhuman primate (NHP) models. The NHP model is a critical component of hemorrhagic fever assay development because active outbreaks of diseases such as Ebola are fortunately rare. Although banked (frozen) human disease samples from prior outbreaks can be used to characterize new assays, these samples are rare and can obscure the impact of fresh versus frozen blood on assay performance. The multiplexed SERS assay was tested in two rounds of NHP testing. In the first round of testing, the assay was tested using frozen longitudinal samples from two Ebola-infected NHPs. Using the multiplexed SERS assay, the Ebola infection was detectable at 6 to 7 days, which corresponded to viral loads of about 106 PFU/ml (Fig. 4A). A second round of testing was performed with fresh (nonfrozen) blood from nine late-stage Ebola Makona–infected NHPs and a frozen blood control from a healthy NHP (Fig. 4B). Results showed a strong signal for the EBOV assay, with little signal for the LASV or malaria assays. It should be noted that a single replicate from the disease state samples gave no signal, which we hypothesize occurred because of a missing Ebola assay reagent in that assay vial. Results from both rounds of testing demonstrate the robustness of the SERS assay in frozen NHP samples and fresh samples with the same blood matrix abnormalities expected in samples from human patients suffering with EBOV.

Fig. 4 Assay testing on samples from EBOV-infected NHPs.

(A) Time course of Ebola signals from blood from two different NHPs infected with Ebola Makona. Dashed line illustrates results from assays done on blood from one EBOV-infected NHP. Solid line illustrates results from assays done on blood from a second EBOV-infected NHP. Each point represents an average of three experimental replicates; error bars represent SD. (B) Triplex assay testing of NHP blood. Individual NHPs are represented on the x axis, one uninfected control (Neg) and nine late-stage Ebola Makona infected (1 to 9). Y axis shows assay signal from each of three triplex assays—EBOV HNW assay (blue diamonds), LASV HNW assay (orange triangle), and malaria HNW assay (gray circles).

Field site setup and sample collection

To determine the performance of the multiplexed HNW assay in West African laboratories, the SERS technology was tested at three sites using dried reagents and the POC workflow described previously: (i) Institut de Recherche en Santé, de Surveillance Epidémiologique et de Formations (IRESSEF) in Dakar, Senegal, (ii) Centre de santé in Bounkiling in Senegal, and (iii) at the Hemorrhagic Fever Laboratory (HFL) in Conakry, Guinea. A total of 586 clinical specimens were tested, with the SERS assay results compared against gold standard reference methods (PCR for Ebola and microscopy or lateral flow for malaria). Lassa status was not determined by any reference method; however, there were no active outbreaks in the area of collection during the study, and thus, all samples were assumed to be true negatives for Lassa. Both Senegalese sites tested the 3-plex Ebola, Lassa, and malaria assays, whereas the Guinean site tested a 2-plex Ebola and malaria assay (a 3-plex assay was also sent to the Guinean site but was not tested because of logistical issues and staff availability).

Table 1 shows the distribution of the 586 clinical specimens tested at the Senegalese (n = 276) and Guinean (n = 310) field sites. Because no recent cases of Ebola have originated in Senegal and Senegal is an Ebola-free country (19), field-collected samples in Senegal allowed us to test not only the performance of the malaria-portion of the multiplexed SERS assay but also the specificity of the Ebola reagents in the SERS assay panel, as no Ebola-positive results were expected in Senegal. In contrast, samples from the Guinean site provided an opportunity to test the sensitivity of the SERS assay for Ebola using banked samples from the 2014 outbreak.

Table 1 Ebola and malaria status for the 586 clinical specimens at the Senegalese (276 samples) and Guinean (310 samples) field sites.

View this table:

The Bounkiling, Senegal site also provided an opportunity to test the SERS assay workflow in a remote, resource-limited facility with local staff in a region where malaria is endemic. Bounkiling is located in southern Senegal (about 329 km from Dakar). It is part of a poorly urbanized region with an urbanization rate of 17.7% and covers an area of about 2830 km (20). A total of 276 whole blood samples were collected in Bounkiling by venipuncture during the November 2016 to February 2017 malaria season. Aliquots of each sample were tested for malaria at the time of collection by microscopy and by lateral flow rapid test (SD Bioline Malaria Ag P.F.); the remainder of the sample was frozen (−20°C) for later SERS assay testing. On the day of SERS assay testing, malaria lateral flow testing was repeated. No discrepancies between the original and repeat lateral flow results were noted. No confirmatory gold standard Ebola testing was conducted on samples acquired in Senegal, and all Senegalese samples were assumed Ebola negative.

An additional 310 clinical specimens were tested at the HFL in Conakry, Guinea. HFL is the largest regional repository of banked human Ebola samples from the 2014 outbreak. Because of limited availability of banked whole blood samples, the Guinea Ebola samples were all serum samples rather than whole blood. In total, HFL tested 190 serum samples of known Ebola status based on PCR results (RealStar Filovirus Screen RT-PCR), consisting of 100 Ebola-positive samples [cycle threshold (CT) values ranging from 16 to 28] and 90 Ebola-negative samples. In addition, the cohort of Ebola samples was supplemented by an additional 120 whole blood samples [20 collected at HFL and 100 acquired from a site in Uganda via a clinical specimen supplier under IRB approval for general research use (Discovery Life Sciences)]. These additional 120 samples were included to provide a population of malaria positives and malaria/Ebola negatives to test the specificity of the Ebola results and the multiplexing performance of the SERS technology at the Guinean site. All 120 whole blood samples were tested for malaria by rapid test (SD Bioline Malaria Ag P.F.) on the day of SERS assay testing. Site teams were trained during a multiday session conducted in Dakar, during which protocol reviews were documented and hands-on proficiency was evaluated for each team member.

Clinical test results

Figure 5 plots the SERS assay signals for the Ebola VP40 and P. falciparum HRP2 antigens in the multiplexed assay. Figure 5A includes Ebola VP40 assay signals from all 586 specimens, grouped by Ebola status (negative/positive). As described above, all 276 Senegalese samples were presumed to be Ebola negative. Figure 5B plots P. falciparum HRP2 SERS assay signals from the 396 samples of known malaria status. The 190 serum samples from the HFL Ebola repository are excluded from Fig. 5B because their malaria status was unknown. Receiver operator characteristic (ROC) curves (Fig. 5, C and D) for both Ebola and malaria and fitted ROC curves based on a proper binormal statistical model (21). ROC curves were plotted, combining data from all sites for which Ebola or Malaria status was known, and included both serum and blood data and 2-plex and 3-plex data. Testing (fig. S4) indicated that the SERS signal intensity in serum and whole blood is comparable, in part because optical absorbance in whole blood is minimized in the NIR region and also because the incident laser interrogates a dense pellet of magnetic microparticles which blocks optical signal from the solution behind it.

Fig. 5 Results from clinical testing of the EBOV and malaria virus HNW assays.

Boxplots of (A) assay signals from Ebola-positive (pos) and Ebola-negative (neg) samples. (B) Assay signals from malaria-positive and malaria-negative samples. (C and D) Corresponding ROC curves for Ebola and malaria, respectively. Area under the curve (AUC) for each ROC curve is shown in the figure, along with associated 95% confidence intervals (CIs).

The 3-plex SERS malaria assay signals are about 30% lower at the low antigen concentrations when compared to the 2-plex assay signals but slightly elevated at the high spike level (fig. S5). During retrospective data analysis of the clinical results, the positivity thresholds for the 2-plex and 3-plex versions of the assay were found to be identical, implying that the difference in signal intensities did not result in notable performance differences across sites (fig. S6) and enabling us to merge datasets for a combined analysis. We hypothesize that this may be because the malaria antigen concentrations in the samples were relatively high, resulting in assay signals that were well above the positivity threshold. This hypothesis is supported by the fact that the HRP2 protein is expected to be highly abundant in our study population (individuals presenting with fever) (22). The Ebola 2-plex assay gave comparable signal intensities to the 3-plex within the error of the assay (fig. S5).

Overall, the empirical sensitivity for Ebola detection (n = 586) was 90.0% at a specificity of 97.9% for a 96.6% total accuracy. For the 396 samples of known malaria status, we observed an empirical sensitivity of 100.0% at a specificity of 99.6%, corresponding to a 99.7% total accuracy. The positivity threshold for Ebola detection was 0.392 and that for malaria was 0.07 arbitrary units (a.u.). None of the 276 clinical samples tested with the 3-plex assay were independently tested for LASV, and incidence of Lassa was expected to be minimal. All SERS assay signals for the Lassa antigen were below 0.134 a.u., consistent with the expected signal for negatives.

DISCUSSION

Here, we described the development and successful translation of a multiplexed diagnostic test from the laboratory to field deployment, demonstrating differential detection of Ebola, Lassa, and malaria infections from whole blood or serum samples. Data from live virus testing, NHP models, and clinical samples demonstrated the potential of the SERS assay technology to provide multiplexed, differential diagnosis of Ebola, Lassa, and malaria in a rapid test suitable for austere environments. Performance and user workflow testing at various field sites in West Africa showed good usability based on clinical staff feedback. During field testing, the SERS technology demonstrated excellent sensitivity and specificity for malaria diagnosis (>99.5%) and strong numbers for Ebola diagnosis.

The multiplexed SERS assay showed strong sensitivity and specificity for EBOV in human clinical samples with values of 90.0 and 97.9%, respectively (empirical). These results compare favorably to other rapid diagnostic tests that were studied during and after the EBOV outbreak, including the eZYSCREEN lateral flow assay that has received CE-IVD clearance (65 and 99%) (23), a Naval Research Laboratory–developed Rapid Diagnostic Test (RDT) (87.8 and 97.5%), a Defense Science and Technology Laboratory–developed Rapid Diagnostic Test RDT (100 and 92%) (24), and a Viral Hemorrhagic Fever Consortium–developed RDT (25), with a World Health Organization–evaluated specificity of 95% and a specificity of 86.6% (26). The SERS-based assays also compared well to a PCR-based multiplexed diagnostic (Biofire FilmArray), which showed 95.7% sensitivity at 100% specificity (27). Although it is difficult to definitively compare the performance of these assays against one another, because many were evaluated under outbreak conditions and rigorous comparison was not possible (28), these results suggest that our multiplexed SERS–based diagnostics maintains similar sensitivity and specificity for EBOV detection and adds simultaneous diagnostic assays for malaria and Lassa. A direct comparison of all assays using the same set of samples would provide additional important information about the relative merits of the different assays.

There are a number of assay advantages that the SERS technology provides that might encourage its use. Considering that diagnosis of infections such as Ebola come with heightened biosafety concerns (29), it is beneficial that the SERS assay requires only the addition of blood to a tube that is then closed and tested directly, with no need to reopen the test vial. The assay requires fewer handling steps and offers a more rapid time to result from blood sampling compared to RT-PCR approaches (~30 min versus 1 to 4 hours). Its portability can enable near-patient testing, simplifying the logistics and biosecurity concerns associated with transporting samples for analysis. With the ability to detect multiple diseases from the same sample while maintaining a POC workflow that uses ambient temperature–stable reagents and instruments that operate over a wide range of temperatures and humidity (and in the absence of electricity), SERS-based assays could be a valuable tool to help ensure appropriate clinical treatment and triage of febrile patients during pandemic situations.

One additional strength of this multiplexed assay is that it offers the potential to simplify the identification of co-infections that could complicate clinical treatment decisions. A retrospective study of samples collected during the 2014 Ebola outbreak in Sierra Leone found that malaria and Ebola co-infection was common and resulted in an increased mortality risk (30). Currently, there are no reports of other POC field-tested diagnostic devices that can simultaneously detect these co-infections, leaving health care workers with limited tools and information to make challenging clinical decisions. During field testing, the SERS platform identified several clinical samples at the HFL as co-infected with Ebola and malaria. It is important to note that malaria infection status was unknown for these samples, and because of logistical issues, we were unable to test these banked Ebola-infected serum samples with a malaria RDT approved for serum; however, many of these EBOV-positive, “potentially co-infected” samples tested positive for malaria during limited “spot” testing with a malaria RDT approved for whole blood. These results suggest that these samples were true co-infections and demonstrate the potential of the SERS technology to simultaneously detect Ebola and malaria co-infections from a single sample.

Outbreak diseases present a unique challenge in translational medicine because the availability of human specimens for assay development and testing is limited. Multiplexing of rare diseases only compounds the challenges. Here, we have taken a multipronged approach to addressing the field performance risks of a multiplexed POC differential Ebola diagnostic and have aggregated data from live virus testing, NHP models, and clinical specimens. Healthy human blood samples spiked with live virus provide information about the sensitivity and specificity of the technology and may provide the best means of broad cross-reactivity testing because potential assay interferents and competitors can be spiked at controlled loads. The analytical limit of detection for the SERS HNW technology (105 to 106 PFU/ml) is consistent with other antibody-based approaches (31, 32), suggesting that the technology is capable of diagnosing diseases at time points similar to other immunoassays; however, unlike other sandwich immunoassays such as ELISAs, the SERS technology requires no sample preparation or wash steps. Specificity for each of the pathogens was good, and no cross-reactivity was observed with dengue or Zika, other potentially endemic diseases.

Live virus spiked into healthy whole blood provides key information about analytical performance but does not always reflect the actual performance seen in true clinical specimens. Virus titers in infected individuals may vary wildly among individuals or as disease progresses. Moreover, the potential interferents or cross-reactors in blood samples can be difficult to predict or may vary on the basis of ethnicity or demographics, making it imperative to test the specificity of the assay in the field among the intended use population. In our work, we found that nonspecific or background signal, presumably due to nonspecific antibody-mediated SERS–magnetic particle complexes, was elevated in the West African population relative to blood taken from healthy U.S.-based donors or from Ugandan donors. We attempted to determine the source of the higher background in the West African patient population but were unable to isolate a specific cause. Workflow and environmental studies were conducted to understand the impact of variations in assay timing, assay temperature and humidity, and errors in sample volume; however, none of these were found to affect the nonspecific signal. No correlation with patient demographics (gender and malaria status) was observed. We consistently observed that the average nonspecific signal from the Ebola portion of the multiplexed assay was higher in the West African population than in U.S. or Ugandan populations, whereas the malaria nonspecific signal was independent of patient population. We speculate that the Ebola antibodies are more prone to nonspecific interactions than the malaria antibodies and interact with autoimmune components in the West African patient population. Further work would be required to better understand this phenomenon. However, because good separation was observed between the SERS signal for Ebola-positive samples and Ebola-negative samples, the higher nonspecific background in the West African population does not appear to adversely affect the multiplexed SERS HNW assay performance.

In addition to live virus testing and field performance characterization of banked human specimens, we used NHP models to complete our characterization of the assay’s expected performance in an outbreak situation. The NHP data illustrate the performance of the assay in fresh (nonfrozen) clinical specimens. It also challenges the assay format in that blood from a patient with hemorrhagic fever may have considerable hemolysis and clotting abnormities that may not be apparent in banked samples that have undergone freeze/thaw before testing (33). Because the SERS multiplexed assay is conducted directly in a mixture of the blood sample and assay buffer without any wash steps, it was important to demonstrate that disease effects such as hemolysis do not interfere with the signaling mechanism. By leveraging previously developed NHP models, we were able to probe the potential impact of diseased blood on assay performance. In this case, blood from Ebola-infected NHPs, although visually distinct from healthy blood, showed no interference in the SERS assay, demonstrating the potential of the SERS technology to perform in the presence of the complex matrix etiologies observed in patients with hemorrhagic fever.

Differential diagnosis, or the ability to differentiate among diseases that present with similar clinical symptoms, remains a critical gap in low-resource settings. Initial symptoms of EBOV infection are nonspecific (fever and chills) and overlap with other common endemic diseases such as malaria, dengue, tuberculosis, or typhoid fever (34). In outbreak situations, limited resources tend to be necessarily focused on the identification, treatment, and management of the outbreak disease. An Ebola rapid test may identify Ebola-positive patients, but no good paradigm exists for handling febrile patients who test Ebola negative during an outbreak. A recent study suggests that the 2014 Ebola outbreak may have led to large-scale undertreatment of patients with malaria in Guinea, potentially leading to larger numbers of unnecessary malaria deaths than deaths from Ebola itself (35). The multiplexed SERS diagnostic platform reported herein is well suited for field deployment and POC use to differentially detect Ebola, Lassa, and malaria infections from whole blood or serum samples. Performance and user workflow results reported here are encouraging and suggest that the SERS technology could be a valuable tool to help ensure appropriate clinical treatment and triage of febrile patients during pandemic situations.

This study does have limitations. In particular, the performance of the SERS technology in human samples in an outbreak is undetermined, and challenges due to the sample matrix (fresh versus frozen blood or serum versus whole blood samples) present risks that are hard to anticipate and mitigate. This highlights a recurring issue: Developing and validating diagnostic methods for rare diseases such as Ebola present a unique challenge in translational medicine. We have attempted to address the field performance risks of a multiplexed POC differential Ebola diagnostic by aggregating data from live virus testing, NHP models, and clinical sample testing, but final performance of the assays will not be known until tested under outbreak conditions.

MATERIALS AND METHODS

Study design

The goal of this study was to develop a multiplexed POC diagnostic to differentially detect Ebola, Lassa, and malaria infections directly from a whole blood sample. To evaluate performance of the diagnostic compared to reference methods, readers and tests kits were deployed and tested at several field sites in West Africa using banked repository samples and samples collected in the field in Senegal and Guinea. All studies were conducted in accordance with the ethical principles that originate from the Declaration of Helsinki and the Belmont Report. The protocol for the Senegal arm of the study was approved by the Comité National et Ethique pour la Recherche en Santé in Senegal (reference number MSAS/DPRS/CNRS). The protocol for the Guinea arm of the study was approved by the Comité National et Ethique pour la Recherche en Santé in Guinea (reference number 128/CNERS/16) and the Columbia University IRB (approval number AAAQ9717). Sample collection was done under informed consent from adults (>18 years old) with febrile illness. Samples were excluded from anyone without consent, less than 18 years of age, or having chronic heart or kidney disease, cancer, diabetes, or other life-limiting or life-threatening chronic diseases. Field sites were selected on the basis of (i) the availability of malaria and Ebola clinical samples in repository, (ii) proximity of the collection site to disease endemic locations, and (iii) diversity of field site conditions. For sample size selection, we followed previously described general guidelines (36). We aimed to achieve assay sensitivities and specificities of about 95%, with a lower confidence interval (CI) of 85% for both sensitivity and specificity, requiring a sample size of 93 positives and 93 negatives for both malaria and Ebola. During the course of the study, 486 negative and 100 positive Ebola samples and 230 negative and 163 positive malaria samples were tested. On the basis of the proper binormal model (PBM) fit, the actual achieved 95% CIs of the sensitivity at 95% specificity are 0.844 to 0.951 for Ebola and 0.999 to 1.00 for malaria.

Senegalese samples were collected after informed consent by venipuncture into a 5-ml evacuated EDTA tube (BD) and given a sample number that is not patient identifiable. Malaria RDTs (SD Bioline Malaria Ag P.F.) and thick-smear microscopy were completed as reference tests for malaria status, whereas Ebola status was assumed negative because samples were collected from an Ebola-free region.

The Guinean specimens from healthy individuals were collected via venipuncture into a 5 ml evacuated EDTA tube and labeled with a patient deidentified number. These specimens were tested for malaria using a malaria RDT (SD Bioline Malaria Ag P.F) and then stored and refrigerated for up to 48 hours. The Guinean cohort was supplemented with frozen whole blood specimens obtained via a clinical specimen supplier (Discovery Life Sciences) from a site in Uganda under IRB approval for general research use. Malaria status of the supplemental specimens was also determined by RDT (SD Bioline Malaria Ag P.F.), and Ebola status was assumed negative because samples were collected from an Ebola-free region. The bulk of the specimens tested in Guinea was taken from the HFL repository and were frozen serum samples with known Ebola status by PCR (RealStar Filovirus Screen RT-PCR) and unknown and untested malaria status. All samples at both sites were tested as a single replicate for each method, and no outliers have been reported or eliminated from the data analysis.

Malaria or Ebola status determined by reference tests was not blinded to the investigator. For each clinical sample, site staff completed a case report form (CRF) to indicate malaria and Ebola status and SERS raw data. Disease status, as determined by SERS results, was not presented to the investigator because data were processed and positivity thresholds were calculated after study completion. During the field study, values from the SERS assay were displayed in real time on the reader screen. The data were also automatically uploaded daily to a cloud server for data storage and to ensure concordance with the CRFs. To evaluate performance of the diagnostic, reference test results were compared to the results obtained by the SERS technology.

Viruses

All work with infectious virus and potentially infectious materials derived from animals was conducted in a BSL-4 laboratory in the National Institute of Allergy and Infectious Diseases (NIAID) Integrated Research Facility (IRF)–Frederick. All stocks were propagated in VERO C1008 (E6) cells, kidney (African green monkey), Working Cell Bank, NR-596 obtained through BEI Resources [NIAID, National Institutes of Health (NIH), Manassas, VA] using the Minimum Essential Medium-alpha, GlutaMAX, no nucleosides (Gibco, Thermo Fisher Scientific) supplemented with 2% U.S.-origin, certified, heat-inactivated fetal bovine serum (HI-FBS; Gibco, Thermo Fisher Scientific). After harvest, HI-FBS was diluted to 10% final concentration before cryopreservation (see table S1 for reference and passage information).

Exposure model

Rhesus macaques (Macaca mulatta) were exposed intramuscularly to a target dose of 1000 PFU of EBOV/Mak-C05, and animals were monitored at least daily by trained, experienced comparative medicine program staff. The animals were provided food and water ad libitum and checked at least daily according with the protocol. All efforts were made to minimize painful procedures; the attending veterinarian was consulted regarding painful procedures, and animals were anesthetized before phlebotomy and virus infection. After the development of clinical signs, animals were checked multiple times daily. Following approved Animal Care and Use Committee protocol and in accordance with current American Veterinary Medical Association Guidelines on Euthanasia and institute standard operating procedures, when clinical scores reached defined criteria, animals were euthanized under deep anesthesia to minimize pain and distress. Research was performed in accordance with animal study protocols approved by the Animal Care and Use Committee of the NIAID Division of Clinical Research. Protocols adhere to the recommendations stated in the Guide for the Care and Use of Laboratory Animals and were developed in compliance with the U.S. Department of Agriculture Animal Welfare Act regulations, U.S. Public Health Service Policy on Humane Care and Use of Laboratory Animals, NIH policies and guidelines, and other federal statutes and regulations relating to animals and experiments involving animals. The NIAID IRF–Frederick, where the research was conducted, is fully accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International.

Malaria assay testing

The singleplex malaria assay performance and inclusivity/exclusivity testing was performed using frozen P. falciparum whole blood specimens from a site in Uganda and frozen P. vivax serum specimens from a site in Vietnam obtained via a clinical specimen supplier (Discovery Life Sciences) under IRB approval for general research use. Malaria-negative whole blood samples were obtained via a clinical specimen supplier (Biological Specialty Corporation) under IRB approval for general research use.

Assay reagents and disposable design

SERS nanotags were synthesized by BD (Research Triangle Park, NC) according to methods described previously (10, 11, 37) and using the following Raman dyes: 2,5-bis(4-pyridyl)-1,3,4-thiadiazole (Tag 493; catalog no. 10650010, Acros Organics), 4,4′-dipyridyl (Tag 420; catalog no. 289426, Sigma-Aldrich), and 4,4′-dipyridyl-d8 (Tag 421; catalog no. 613606, Sigma-Aldrich). Antibodies were conjugated to thiolated SERS nanoparticles using an N-hydroxysuccinimide (NHS)–maleimide linker (catalog no. 22622, Thermo Scientific). Functionalized magnetic beads (carboxylic acid, catalog no. 65011, Dynabeads MyOne) were purchased from Life Technologies (Carlsbad, CA) and later conjugated to antibodies via standard NHS/1-ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride (NHS/EDC) chemistry. Lassa antibodies were provided by the U.S. Army Medical Research Institute of Infectious Diseases (USAMRIID). SERS nanotags and magnetic beads were conjugated to antibodies to P. falciparum HRP2, Ebola VP40, and Lassa NP using NHS-maleimide and NHS/EDC conjugation chemistries, respectively. For laboratory testing using liquid reagents, the assay reagents were used as is. Vials containing dried reagents were made by lyophilizing reagents into spatially separated spots on a polymer release liner. After lyophilization, this release liner was removed, and the reagent spots were placed in an assay tube to generate a “unitized assay tube” for testing. The tubes were all visually inspected to ensure that all reagents were present.

Preliminary testing was performed using commercial recombinant HRP2 antigen (catalog no. A3000, CTK Biotech), recombinant Ebola VP40 antigen (catalog no. 30-1837, Fitzgerald), and Lassa NP antigen (catalog no. CSB-YP356406LCPe1, CUSABIO Life Science).

The lysing and blocking buffer is formulated using sodium chloride (catalog no. 71394, Fluka), Tris Ultrapure (catalog no. 15568-025, Invitrogen), EDTA (catalog no. AM9260G, Ambion), Triton X-100 (catalog no. T8787, Sigma-Aldrich), bovine serum albumin (catalog no. 001-000-162, Jackson ImmunoResearch), and sodium azide (catalog no. S0209, TechNova).

Field site assay workflow

After thawing the specimens to be tested, the user adds 45 μl of sample (whole blood or serum) and 455 μl of lysis buffer to a tube containing reagents. Immediately after liquid addition, the user inserts the capped tubes into a mixing unit for automatic mixing and magnetic pelleting (up to eight tubes can be run simultaneously per available mixer). This mixing unit rotates tubes end over end for 25 min followed by automatic application of a magnet bar to the tubes for pelleting. After 5 min of pelleting, the mixer alarms to inform the user that the samples are ready for reading. The user transfers the tube to a pelleting fixture containing a focusing magnet for 15 s to improve the pellet form and align the pellet. The user then transfers the tube to the SERS reader where a barcode is read on each tube, and the user is requested to enter the sample number allowing association of sample to tube and keeping the sample deidentified for analysis.

SERS instrumentation

The SERS reader used during field testing is a stand-alone instrument consisting of an internal computer for processing, optical components to collect the Raman spectrum from the sample, and wireless communication devices to upload data to a cloud server via either Wifi or 3G connectivity. In the reader, the pellet is interrogated with a 785-nm laser (Innovative Photonic Solutions), and the resulting Raman spectrum is collected via a Raman spectrometer (Hamamatsu Photonics K.K.). The resulting SERS spectrum is analyzed for presence and quantity of the various SERS tags, as described in the next section. Data collected from the readers at all field sites were displayed in real time on the reader screen and automatically uploaded to an Azure blob for cloud storage and for remote analysis. It should be noted that the reader used for malaria assay development work was an earlier version of the POC reader. This early version used similar optical and mechanical components as the POC reader and was controlled via an external laptop. The newer field-deployed POC reader was developed specifically for ease of use and operation in austere environments where ambient conditions are not well controlled.

Mixing was accomplished with an end-over-end rotary mixer (HulaMixer, Thermo Fisher Scientific) except for the field studies where we integrated the magnetic pelleting with the rotary mixing unit to further simplify the workflow. The integrated mixer consists of a timing unit and a bar magnet on an arm added to the rotary mixing module so that the bar magnet is automatically contacted with the assay tubes at the end of the incubation period to form the magnetic pellet.

Both the reader and mixer used for field testing incorporated a rechargeable battery and were capable of operating for multiple hours in the absence of electricity. In addition, both the reader and mixer were capable of operating over a wide range of temperatures and humidity.

Data analysis

The measured SERS spectrum consists of a superposition of optical peaks from each of the SERS nanoparticles detected in the sample (one unique SERS spectrum for Ebola, Lassa, and malaria). To determine the signal (tag weight) contribution from each SERS particle to the measured spectrum, the spectrum is deconvolved by least-squares regression using archived pure reference spectra for each SERS tag, a background spectrum, and a combination of Legendre polynomials, similar to methods described in detail elsewhere (38). This fitting method uses the entire spectral fingerprint to quantitatively extract signals from the multiplexed SERS spectrum. The algorithm calculates values of the weights (the linear equations coefficients) for each reference spectra that lead to the smallest difference between the fitted model and the measured spectrum. The intensity of the tag weight is directly proportional to the amount of SERS tags captured in the magnetic pellet. Figure S7A shows the SERS spectrum collected from a malaria-positive clinical sample (blue) and the fitted spectrum determined from the least-squares deconvolution (red). Because the fitted curve closely matches the collected spectrum, the blue curve is hidden by the red curve. Figure S7B shows the residuals between the fitted and collected spectrum, and fig. S7C shows the library of pure reference spectra for Tags 420, 421, 440, and 493 and the background. Fitted Legendre polynomials and calculated tag weights are also shown for each spectrum. Figure S8 shows similar data using a malaria + Ebola–positive clinical sample.

Statistical analysis

ROC curves were generated, and parameters including AUC, sensitivity/specificity at various thresholds, and CIs for fitted curves were calculated using standard methods (39, 40). To generate empirical and fitted ROC curves, we used Metz ROCKit software (41), with PBM fitting (21). Optimal positive/negative cutoff (threshold) values for malaria and Ebola SERS HNW data were independently determined by calculating the Youden index point, which corresponds to the point on the ROC curve where the distance from the “coin toss” line, Se = 1 – Sp, is maximized. Sensitivity and specificity were determined for both empirical and fitted ROC curves based on the Youden index point. Raw data for experiments where N < 20 are reported in table S2.

SUPPLEMENTARY MATERIALS

www.sciencetranslationalmedicine.org/cgi/content/full/10/471/eaat0944/DC1

Fig. S1. Schematic demonstrating the HNW assay workflow.

Fig. S2. Determination of assay sensitivity for Ebola, Lassa, and malaria HNW assays using recombinant antigens in assay buffer.

Fig. S3. Ebola inclusivity testing using the triplex assay with high concentrations of Sudan virus and Bundibugyo virus and a negative control.

Fig. S4. Signals obtained from recombinant antigens spiked into matched human blood, plasma, or serum that was provided by a single donor.

Fig. S5. Signal intensities for malaria and Ebola antigens spiked into duplex and triplex assay formats using blood as a test medium.

Fig. S6. ROC curves for the HNW malaria assay.

Fig. S7. Example SERS spectra from a clinical sample assay that indicated a single infectious agent.

Fig. S8. Example SERS spectra from a clinical sample assay that indicated a co-infection with malaria and EBOV.

Table S1. Reference and passage information for viruses used in the NHP model.

Table S2. Raw data for experiments where N < 20.

REFERENCES AND NOTES

Acknowledgments: We thank the following BD associates for contributions to this body of work: M. Cash, J. Thomas, A. Sonnenberg, R. Payne, D. Dias, T. Jensen, S. Shanmugam, and J. Sollome. We thank BU/NEIDL research technicians A. Soucy, M. Lei, and A. J. Devaux for contributions to assay development. We thank R. Schoepp and M. Poli from USAMRIID for providing the LASV antibodies used in this work. We thank the following associates from district Sanitaire de Bounkiling: S. N. Kaly, T. I. Coly, E.-h. A. Diop, O. Niang, and S. D. Cissokho. We thank the following associates from IRESSEF Dakar: N. p. Mze and D. d. Marre. Funding: This work was supported by the Paul G. Allen Family Foundation. NHP samples were provided via an interagency agreement between NIAID Division of Microbiology and Infectious Diseases and the NIH Office of Research Services/Division of Veterinary Resources, NOR15003-001-00000. Author contributions: D.S. and A.G.L. jointly developed the assays and designed the performance testing experiments. J.J. and A.N.H. completed the live virus and NHP testing. M.K. and H.F. fabricated the assay reagents. E.F. and R.N. developed and validated the SERS reader and mixer. J.R. and K.P. developed the assays and completed the inclusivity and exclusivity testing. E.C. helped develop the malaria assay. A.A., C.D., and S.M. tested the samples at the Senegalese field site. D.A., B.S., N.M., and Y.B.A. tested the samples at the Guinean field site. J.H.C. and K.W. oversaw all technical work. D.S., A.G.L., E.F., H.F., M.B., Y.B.A., A.A., J.J., A.N.H., J.R., S.M., J.H.C., and K.W. discussed, analyzed, and interpreted the results and wrote and edited the manuscript. Competing interests: Patents relevant to this work include the following: US 6514767 B1, US 7192778 B2, EP 1226422 A4, US 7443489 B2, and EP 2295954 A1 (“Surface enhanced spectroscopy-active composite nanoparticles”); US 6861263 B2 (“Surface-enhanced spectroscopy-active sandwich nanoparticles”); and US 8497131 B2 (“Surface enhanced spectroscopy-active composite nanoparticles comprising Raman-active reporter molecules”). The authors declare that they have no competing interests. Data and materials availability: SERS reagents and instrumentation may be available through a material transfer agreement with Becton, Dickinson and Company. Lassa antibodies may be available through a material transfer agreement with USAMRIID.
View Abstract

Stay Connected to Science Translational Medicine


Editor's Blog

Navigate This Article