Colorimetric Test for Fast Detection of SARS-CoV-2 in Nasal and Throat Swabs

1Since its identification in China in late 2019, the SARS-CoV-2 epidemic has spread rapidly worldwide affecting millions of people, thus pushing the World Health Organization (WHO) to declare a COVID-19 outbreak a global health emergency. Mass testing is fundamental to identify and isolate clusters in order to limit and eventually eradicate SARS-CoV-2. The gold standard for diagnosing COVID-19 infection is a reverse transcription real-time polymerase chain reaction (real-time PCR) that is able to detect the virus genetic material (RNA) in samples collected via nasopharyngeal swab. Currently, only qualitative real-time PCR assays are available that yield positive/negative results without providing information about the viral load. Due to its complexity, real-time PCR tests are performed in certified laboratories, are time-consuming, require experienced personnel, and can hardly lend themselves to mass screening.− Huge efforts are put into overcoming such a bottleneck, thereby making nucleic acid amplification suitable for point-of-care tests, but the variety of methods and the lack of any commercial solution demonstrates that the gap between research and real applications is still to be filled. The main reason for that has to be found in the detection principle (RNA-extraction, reverse transcription, and amplification) whose complexity, though greatly reduced by several approaches (e.g., loop-mediated isothermal amplification (LAMP) and recombinase polymerase amplification (RPA) or CRISPR-based detection), is far from being used for quick POC tests.

2Lateral flow assays (LFA)s are among the actual biosensing platforms for home tests and potentially for mass screening,− but the relatively poor sensitivity inherent to this technology makes the quest for a different approach urgent. Biosensors based on metal nanoparticles are often proposed because of their unique optical properties, which makes them potentially suitable to develop easy-to-use and rapid colorimetric diagnostic tests for point-of-care applications or even for home use. Due to its surface chemistry and given its biocompatibility, gold is generally preferred to other metals. The physical process underlying this class of biosensors is the localized surface plasmon resonance (LSPR) that consists of coherent and nonpropagating oscillations of free electrons in metal nanoparticles arising when they interact with an electromagnetic wave whose frequency resonates with the plasmonic one., Colorimetric detection based on gold nanoparticles (AuNPs) takes advantage of the color change occurring in a colloidal suspension from red to blue as a result of LSPR coupling among the nanoparticles. AuNP aggregation can be regulated using biological mechanisms such as antigen–antibody (Ab) interaction, in which case different strategies can be used to immobilize Abs correctly oriented on the surface of the AuNPs, although the complexity of the standard procedures makes them unsuitable for industrial applications on a large scale.

3The photochemical immobilization technique (PIT) is a surface functionalization procedure that only requires UV activation of the Abs and leads to a high-density functionalized surface within minutes., PIT has proven itself to be effective in tethering Abs upright not only on flat surfaces,− but also on AuNPs, which were used either to ballast small antigen or to realize a colorimetric biosensor for detecting IgGs and estradiol. In the latter cases, the presence of the antigen was detected as a change in the absorbance that can be easily measured by a spectrophotometer or even by naked eye. An approach relying on nanoparticle aggregation induced by the presence of the antigen was also used to detect the influenza A virus, but no clinical application was reported to demonstrate the effectiveness of the whole procedure in clinical cases.

4Here, we report on the realization of a colorimetric biosensor that can be used for COVID-19 mass testing with sensitivity and specificity higher that 95% as demonstrated by a comparative analysis carried out on a total of 94 samples (45 positive and 49 negative), tested by standard real-time PCR in the Virology Unit of A.O.U. Federico II/Department of Translational Medicine of the University of Naples “Federico II”. The detection scheme is shown in Figure 1a,b and consists of a colloidal solution of PIT-functionalized AuNPs (f-AuNPs) against three surface proteins of SARS-CoV-2: spike, envelope, and membrane (S, E, and M, respectively, in Figure 1a). AuNP fabrication (20 nm diameter), characterization, and functionalization are described in the Supporting Information (see sections S1–S5 and Figures S1–S3 of SI), that contains a scalable procedure to realize the colloidal solution for COVID-19 test. In this approach, the sample was a solution of Universal Transport Medium (UTM, Copan Brescia, Italy), in which the specimen was dipped after its collection from the patient and without any additional treatment (see section S6 of Supporting Information). Our test was carried out by mixing 50 μL of the f-AuNP colloidal solution with 100 μL of sample and 100 μL of ultrapure water. The presence of the viral particles (virions) induced the formation of a nanoparticle layer on its surface (Figure 1b) that led to a redshift of the optical density (OD) in the extinction spectrum of the solution (Figure 1c). When the viral load was relatively high, i.e., Ct < 15 (see section S7 of Supporting Information), the color change from red to purple was visible even by the naked eye (Figure 1a,b).

5The extinction spectrum of f-AuNPs reported in Figure 1c (black continuous line) is not distinguishable from the spectrum of a mixed solution with a sample having Ct = 32 (dashed light blue curve, Ct32). On the contrary, red shift is observed for a sample with Ct = 15 (blue continuous line, Ct15), which becomes much more noticeable when Ct = 7 (red continuous line, Ct7). The contribution of the virion (surrounded by nanoparticles) to the extinction spectrum can be deduced by subtracting the spectrum of f-AuNPs from Ct15. In fact, the curve Ct15-(f-AuNPs) shows a peak at a wavelength comparable to that exhibited by Ct7 (approximately 560 nm).

6To further confirm that the spectrum Ct7 arose from f-AuNPs surrounding SARS-CoV-2, we simulated the virion as a 100-nm-diameter, dielectric sphere of 1.45 refractive index. A changeable number of gold nanoparticles (20 nm diameter) were randomly positioned on the sphere (see section S8 of Supporting Information). We implemented the “FDTD solutions” tool in Lumerical software that provides numerical solutions to the Maxwell’s equations by the finite-difference time-domain (FDTD) method within a Mie problem-like workspace. A sketch of the simulation workspace is depicted in Figure S4a, whereas Figure S4b shows the LSPR wavelength as a function of the number of AuNPs on the virion surface. The dependence of the plasmon resonance on the number N of AuNPs on the virion surface is highly nonlinear (Figure S4b). The maximum number of AuNPs that can be placed on a perfectly spherical surface with 100 nm diameter is N = 80, but the steric hindrance offered by surface proteins like the spike one surely limits the filling capacity of the virion surface. In fact, from Figure S4b we see that the experimental value of λLSPR = 560 nm is achieved when N = 70, a value only slightly smaller than the maximum achievable. The extinction spectrum corresponding to N = 70 AuNPs (scaled to the experimental one) is reported in Figure 1c (golden continuous line) and shows a more than satisfactory agreement with the experimental one (red continuous line), thereby confirming that the simple model proposed here is able to capture the essential physical processes underlying the virion detection.

7In order to test the validity of the colorimetric biosensor, we analyzed real samples previously examined by real-time PCR. The samples were from 45 positive to SARS-CoV-2 patients for which Ct ≤ 35 and 49 negative patients (Ct > 35). For all of them, we measured the optical density at 560 nm (OD560) by a commercial microplate reader (Figure 2a). It is quite evident the correlation between the Ct value (reported on the top scale) of the positives (red circles) and OD560, whereas all the negatives (identified by a progressive number in the bottom scale) are randomly distributed providing a “control” value of OD560neg = 0.22 ± 0.02. Also shown in Figure 2a is a horizontal line at an OD560 value of 0.252, which makes clear the ability to discriminate positives from negatives offered by our colorimetric biosensor. In fact, with such a threshold we get 96% and 98% for sensitivity and specificity, respectively. This result is even more remarkable if we consider that Ct > 30 correspond to a very low viral load for which the infection aptitude is questioned. To this aim, it is important to notice that in the early phase of the infection, high viral loads are detected by real-time PCR in upper respiratory specimens with low Ct readouts (Ct < 20, with our assay). In the late phase of the infection, a dramatic drop in viral loads is observed with higher Ct readouts (Ct > 30 with our assay). Positive results with high Ct readouts pose a diagnostic challenge, since they do not necessarily indicate active infection by a replicating virus. It has been observed using viral culture that patients with high Ct real-time PCR and protracted positivity are not infectious, suggesting that the assay likely detects non-active viral particles such as genetic material present in remnants of inactive virus, thereby making our approach of high diagnostic value.

8The receiver operating characteristic (ROC) curve from the data shown in Figure 2a is reported in Figure 2b together with the indication of three threshold values, all of them leading to sensitivity and specificity significantly higher than 90%. In particular, the highest threshold value (0.263) leads to 100% specificity while keeping the sensitivity at the remarkable value 94%. Overall, the high performance of the test associated with the colorimetric biosensor is demonstrated by the area under the ROC curve whose value is 0.98. The qualitative difference in the color between positives and negatives can be observed in Figure 2c that shows a picture of the multiwells containing the samples whose analysis is summarized in Figure 2a,b.

9To measure the dose–response curve of the biosensor, we assessed the optical density OD560 of samples obtained by serial dilutions (1:10) of an initial volume with very high viral load (Ct = 7). The results are shown in Figure 3, in which OD560 is reported as a function of the relative concentration of SARS-CoV-2. For convenience, the top scale reports the equivalent Ct obtained by considering that 1:10 dilution corresponds to log210 ≈ 3.32 change in Ct, whereas the vertical dashed blue line identifies the starting concentration. As it turns out, a detectable signal can be appreciated even after 7 serial dilutions (1:107) confirming the wide detection range resulting from the validation measurements reported in Figure 2.

10The behavior of the OD560 as a function of the virion concentration C reported in Figure 3 can be described by the three-parameter Hill equation

11In eq 1, OD560(0) = OD560neg = 0.22 is the OD of the “control” as deduced from the negative samples, Δ(OD560) is the maximum OD560 variation, K is the concentration (in this case it is a relative concentration) at which the OD560 variation reaches 50% of its maximum. and n is the so-called Hill’s coefficient. The best fit of the experimental data with eq 1 yielded the curve in Figure 3 and the following values for the parameters: Δ(OD560) = 0.45 ± 0.11, K = 0.03 ± 0.11, and n = 0.11 ± 0.02.

12The (relative) concentration K is essentially undetermined as a consequence of the high detection range of our method, which entails small response to the concentration changes (low sensitivity) and, hence, high uncertainty on the concentration measurements. A value for n significantly smaller than 1 indicates the occurrence of a negatively cooperative binding, a behavior which we expect by considering that the probability a f-AuNP binds the virion reduces as the surface covering grows. Eventually, Δ(OD560) is the range of optical densities spanned by the biosensor. The shaded area in Figure 3 contains the values of OD560 within 3 standard deviation (SD) from the mean value of the control (OD560neg + 3SD = 0.28). Thus, according to 3SD criterion, we obtain Ct = 36.5 as the limit of detection of the biosensor in terms of real-time PCR cycle threshold. Thus, although real-time PCR sensitivity is hard to surpass, in the context of COVID-19 pandemic in which the detection of SARS-CoV-2 is required in nasopharyngeal swabs, our method is an efficient alternative when quick and widespread response cannot be provided by standard laboratory methods.

13In conclusion, we realized a colorimetric biosensor based on a colloidal solution of AuNPs (20 nm, OD ≈ 1), each of them functionalized with Abs against one of the three surface proteins of SARS-CoV-2 (spike, envelope, and membrane). The ratio among the three kinds of functionalized AuNPs was 1:1:1. Although both the ratio and the size of AuNPs are still susceptible to optimization to allow one to push even further the limit of detection, the current performances of the biosensor would already permit its use as a test for mass screening, since the detection is based on the interaction among the virions and the pAb-functionalized AuNPs (single step detection) without any pretreatment (e.g., RNA extraction and amplification). The comparison of the readout of our biosensor at 560 nm with the threshold cycle (Ct) of a real-time PCR proved that viral loads corresponding to Ct = 36.5 are detected by the colorimetric biosensor. This threshold is of particular importance because it corresponds to a very low viral load for which the infecting capacity is likely negligible. Such a good performance has to be ascribed to a high filling ratio of the virion surface that results from the presence of multiple Abs (three proteins are targeted) and an effective AuNP surface functionalization procedure (PIT). In fact, through PIT not only is one Fab always exposed to make AuNP highly “reactive”, but also the Abs are attached to the surface (side-on position) without any linker (e.g., protein A), the latter being detrimental for the plasmonic interactions among AuNPs on which the colorimetric biosensor is based.

14Another remarkable feature of the biosensor described here relies on its sensitivity to the virion rather than to its content (RNA). The importance of this is twofold: (1) after the calibration of the optical response, the biosensor lends itself as a powerful tool to quantify the viral load, a nontrivial issue in diagnostic assays in virology; (2) being sensitive only to the virions, the biosensor detects the presence of active viral particles; thus, our method is apt to assess the actual degree of infectiveness of a specimen.

15As a final remark, we point out that the colorimetric solution described here can be easily modified to target other viruses. Thus, we expect that single-step colorimetric detection of viruses can become a general technique to be used for laboratory applications as well as point-of-care testing.