key: cord-0804795-spar16ax
authors: Ye, H.; Nowak, C.; Liu, Y.; Li, Y.; Zhang, T.; Bleris, L.; Qin, Z.
title: Single-Molecule Detection of SARS-CoV-2 by Plasmonic Sensing of Isothermally Amplified Nucleic Acids
date: 2021-10-07
journal: medRxiv : the preprint server for health sciences
DOI: 10.1101/2021.10.05.21264561
sha: 9afa914cc5c75cfa62569de1a1d50fbf5c5345d5
doc_id: 804795
cord_uid: spar16ax

Single-molecule detection of pathogens such as SARS-CoV-2 is key to combat infectious diseases outbreak and pandemic. Currently colorimetric sensing with loop-mediated isothermal amplification (LAMP) provides simple readouts but suffers from intrinsic non-template amplification. Herein, we report that plasmonic sensing of LAMP amplicons via DNA hybridization allows highly specific and single-molecule detection of SARS-CoV-2 RNA. Our work has two important advances. First, we develop gold and silver alloy (Au-Ag) nanoshells as plasmonic sensors that have 4-times stronger extinction in the visible wavelengths and give 20-times lower detection limit for oligonucleotides than Au nanoparticles. Second, we demonstrate that the diagnostic method allows cutting the complex LAMP amplicons into short repeats that are amendable for hybridization with oligonucleotide-functionalized nanoshells. This additional sequence identification eliminates the contamination from non-template amplification. The detection method is a simple and single-molecule diagnostic platform for virus testing at its early representation.

One of the ultimate goals in biological sensing is to detect analytes at single-molecule level. [1] [2] [3] For nucleic acid tests, polymerase chain reaction (PCR) has been the gold standard for single-molecule diagnosis. [3, 4] However, its time-consuming protocols and the needs for laboratory infrastructure largely preclude it from point-of-care (POC) testing. [5, 6] Isothermal amplification methods, such as loop-mediated isothermal amplification (LAMP), have emerged as an alternative to PCR and allow POC testing without the need for thermal cycling. [7] Although simple, LAMP is susceptible to non-template amplification and its simple readout (e.g., based on pH change) cannot distinguish template versus non-templated amplification, thus susceptible to false-positive results. [8] [9] [10] [11] Previous efforts in implementing gold nanoparticles (AuNPs)-based plasmonic sensing on LAMP products have similar features. [12] [13] [14] Specifically, AuNPs have weak optical response and lead to little sensitivity enhancement, while their non-specific sensing of ionic strength change or conjugated primers limits the detection specificity. It is therefore important to develop techniques that can directly identify the amplified sequences via simplified detection scheme, in order to provide diagnostic platforms with simple readouts, high specificity, and single-molecule detection sensitivity.

In this study, we report a robust method for nucleic acid detection, based on plasmonic sensing of LAMP amplicons via DNA hybridization, termed as plasmonic LAMP. Using SARS-CoV-2 RNA as a model, we demonstrate that the plasmonic LAMP method achieved singlemolecule detection via colorimetric analysis, which is a desirable diagnostic toolkit to reduce the severity of COVID-19 pandemic. [9, 15, 16] Plasmonic LAMP has two distinctive features. First, we developed Au-silver (Ag) nanoshells as sensitive plasmonic labels. The Au-Ag shells have 4times stronger plasmonic extinction in the visible wavelengths and provide 20-times sensitivity All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted October 7, 2021. ; https://doi.org/10.1101/2021.10.05.21264561 doi: medRxiv preprint enhancement in plasmonic sensing of oligonucleotides over AuNPs. Second, we introduced restrict enzyme digestion and heat denaturation in order to cut the concatemer-like LAMP amplicons into short repeats that are amendable for subsequent hybridization with plasmonic sensors (Scheme 1). In contrast, previous studies relying on direct sensing of complex LAMP amplicons have met with limited success, due to the lack of hybridization mechanism. [7, 17, 18] The dual sequence identification of plasmonic LAMP, enabled by the primers and plasmonic sensors, eliminates the contamination from non-template amplification and thus improves the detection specificity and sensitivity. Our study is an important step toward advancing simple and ultrasensitive diagnosis.

We started with the synthesis of Au-Ag nanoshells through titrating AgNPs with HAuCl 4 . In a standard synthesis, an aqueous solution of HAuCl 4 is injected at a speed of 6 mL/h into a matrix solution containing AgNPs with an average size of 32 nm as the sacrificial templates ( Figure S1 ) and 2 mM sodium citrate (Na 3 CA) as a reductant and a colloidal stabilizer (see Methods for details). In this case, AuCl 2¯ is first generated from AuCl 4¯v ia Na 3 CA reduction and directs the galvanic replacement reaction with favored high standard reduction potential. [19] Figure 1A illustrates the growth of Au-Ag shells, where a pitting process can be observed in the beginning of reaction, followed by the cavity evolution and simultaneous shell formation as more HAuCl 4 are injected. Figure 1B -D show the transmission electron microscopy (TEM) images of products taken from the reaction, confirming the morphological transformation. In contrast, the absence of Na 3 CA leads to the formation of thin and porous cages at a HAuCl 4 load of 3.3 mL All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. Figure S2 , in equilibrium to 10 mL AuCl 2¯) . The different products can be attributed to the reaction stoichiometry, where one Ag atom is replaced by one Au atom in the case of AuCl 2¯, instead of three Ag atoms for AuCl 4¯. Also, the dissolution of the Ag template requires a longer time in the case of AuCl 2¯ since less openings are preserved on the surface. [19] The slow pitting process increased shell thickness and thus limits the red shift of the LSPR peak as compared to that of cages (more in next section). The energy-dispersive X-ray (EDX) mapping of an individual Au-Ag shell ( Figure 1E ) shows that the Au and Ag elements are distributed across the entire particle, confirming the alloyed structure. Figure S3 shows an atomic-resolution highangle annular dark-field scanning TEM (HAADF-STEM) image of a single particle, where the lattice spacing (1.43 Å) can be attributed to the interplane distance of the (110) planes of the face-centered cubic Au and Ag. The alloy structure is expected to improve the corrosion resistance of Au-Ag nanoshells. [20] We then evaluated the role of Na 3 CA in the shell growth by increasing its concentration from 0 to 20 mM while keeping other conditions the same. Due to the reducing ability endowed by Na 3 CA, we assume that the different reaction species of Au precursor (i.e., AuCl 4¯ and AuCl 2¯) would influence the reaction stoichiometry ( Figure S4A) . Specifically, without Na 3 CA, the reaction proceeds between AuCl 4¯ and AgNPs and results in porous cages. At low concentrations, Na 3 CA turns a small portion of AuCl 4¯ into AuCl 2¯ and leads to the mixture of cages and shells. In contrast, above a critical concentration (1~2 mM), Na 3 CA converts the majority of AuCl 4¯ into AuCl 2¯ and yields shells as the final product. As expected, the synthesized particles ( Figure S4B -G) have a cage-like morphology below 1 mM Na 3 CA and shell-like structure above 2 mM, confirming the growth mechanism. We also found that the reaction kinetics affect the shell growth ( Figure S5A ). Replacing the Na 3 CA with stronger All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted October 7, 2021. ; https://doi.org/10.1101/2021.10.05.21264561 doi: medRxiv preprint reducing agent like L-ascorbic acid leads to a mixture of small NPs and shells ( Figure S5B ). This is because the L-ascorbic acid can also reduce AuCl 4¯ into atoms for the self-nucleation growth of NPs. On the other hand, boosting the loading rate of HAuCl 4 by increasing its concentration and injection speed led to the formation of branched Au-Ag shells ( Figure S5C ). When either the HAuCl 4 concentration or injection speed increases, Au-Ag shells with a few holes on the surface can be observed ( Figure S5D, E) , suggesting a small portion of AuCl 4¯ participated in the reaction. In all cases, the increased deposition rate of Au atoms is key to the preferential overgrowth of branched or porous shells, given that the diffusion rate remains constant during the growth. [21] LSPR Properties of Au-Ag Nanoshells Next, we investigated the LSPR properties of the Au-Ag shells since they are critical for plasmonic sensing. Figure S6 shows the photographs of aliquots taken from the reaction in the presence and absence of Na 3 CA. The suspension of Au-Ag shells remains red-purple, while the cages change from yellow to blue with lower amount of Au precursor injected. The corresponding extinction spectra clearly reveal a different peak shift between these two structures. The LSPR extinction peak (LSPR λ max ) of Au-Ag shells shifted from 392 nm to 530 nm (Figure 2A) , while that of cages shifted to longer near-infrared wavelength ( Figure 2B ). The limited and slow peak shifting of Au-Ag shells ( Figure 2C ) arises from the thick and solid shells, as compared to the cages. [19, 22] Furthermore, the extinction intensity change of the cages and shells differs from each other, with a monotonic decrease and falling-rising trend, respectively ( Figure 2D ). This suggests the Au-Ag shells have higher extinction than the cages.

Similar trends were observed for the Au-Ag shells and cages synthesized with different Na 3 CA concentrations ( Figure S4 , S7).

All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. To gain further insights into the LSPR properties of the Au-Ag shells, we performed a numerical simulation using the boundary element method. [23, 24] Based on the TEM image of the Au-Ag shell ( Figure 1D ) and the Au: Ag molar ratio of 76%: 24% obtained from inductively coupled plasma-mass spectrometry, we set the geometric model with Au: Ag molar ratio of 80%: 20% (homogeneously distributed), 30 nm interior diameter, and varied thickness (t) of wall ranging from 2.5 nm to 10 nm (Figure 2E inset). The simulation results suggest that increasing the shell thickness leads to blue shift of LSPR λ max and higher extinction cross-section area ( Figure 2E ). We also found that both the electric field intensity and distribution continuously increase with the shell thickness ( Figure S8 ). The enhancement in local field effect of Au-Ag shells is mainly ascribed to the collective effect of surface plasmons of unique shell structure, which condenses the interacting light to the structural center. [25] It should be pointed out that in previous studies, it has been reported that the holes on the surface of cages have a negligible effect on the LSPR λ max . [26] This phenomenon is also observed in the shells, as evidenced by the extinction spectra, where the shells with and without holes show close peak positions ( Figure   S5F ). Therefore, we did not include this parameter in the simulation. We also compared the LSPR properties of Au-Ag shells and 50 nm AuNPs ( Figure S9 ) at the same particle concentration and observed a 4-times higher peak extinction while maintaining similar peak position at visible wavelengths (500-600 nm, Figure 2F ). Taken together, the Au-Ag nanoshells possess stronger plasmonic extinction at visible range compared with AuNPs and have the potential for be a class of highly sensitive plasmonic sensors.

We further examined the Au-Ag nanoshells as labels for oligonucleotide detection. First, a sequence associated with the SARS-CoV-2 N gene (nucleocapsid phosphoprotein gene) was All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted October 7, 2021. ; https://doi.org/10.1101/2021.10.05.21264561 doi: medRxiv preprint chosen as the target and two complementary sequences were subsequently designed as probes (Table S1 ). Serial dilutions of the target oligonucleotides were mixed with a pair of shells-based sensors in a hybridization buffer and subjected to naked-eye observation or UV-Vis measurements (Figure 3A and S10A) . Figure S10B shows the digital photo taken from the completed assay solutions, where a color change can be observed at a target concentration of 250 pM against the blank sample (0 pM) as a reference. For quantitative analysis, the corresponding LSPR extinction spectra were recorded and normalized at 537 nm ( Figure 3B) . A calibration curve was obtained by plotting the extinction intensity (I ext ) ratio between λ max = 675 nm and 537 nm against target concentration ( Figure 3C) . A good linear relationship (R 2 = 0.99) was observed in 1-100 pM (inset of Figure 3C ). The limit of detection (LOD) was determined to be 3.6 pM, by calculating the concentration corresponding to a signal that is 3 times standard deviation above the zero calibrator. [27] Notably, this LOD is approximately 20 times lower than that of using 50 nm AuNPs as probes under same conditions ( Figure S11) . Because the Au-Ag shells and AuNPs have similar size and shape, the improved LSPR property including the 4-fold higher extinction intensity of the shells is likely the key to the sensitivity enhancement. It is worth noting that a ratiometric quantification method is adopted in the present work, which also contributes to the improved analytical performance.

We further assessed the detection selectivity by testing the probes with a variety of oligonucleotide sequences, including a non-complementary sequence (NC), a perfectly matched target sequence (T), and a group of single-base-pair (SBP) mismatched sequences (M n, n=1-7, see details in Figure 3D ). The M n sequences are taken from the SARS-CoV-2 Browser and have been confirmed to be polymorphism variants isolated in the US. [28] Therefore, an assay with SBP-mismatched tolerance is of paramount importance since false-negative results can cause All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted October 7, 2021. ; https://doi.org/10.1101/2021.10.05.21264561 doi: medRxiv preprint delayed treatment and increased risk of viral transmission. We tested sequences with these 7 different mutations and found that all showed positive detection above the negative control ( Figure 3E) . This result suggests a good tolerance for SBP mismatch. The slight difference among M n sequences may result from the mismatch locations and the direct hybridization mechanism. [29] Careful examination shows that monitoring the hybridization kinetics at room temperature can distinguish the SBP mismatch from the positive control sequence without mutations ( Figure 3F ) and provides a potential way to further detect SBP mismatch.

Lastly, we developed a diagnostic protocol to integrate reverse transcription loop-mediated isothermal amplification (RT-LAMP) with plasmonic Au-Ag nanoshells for single-molecule detection of infectious pathogens such as SARS-CoV-2 (Scheme 1). We first designed primers targeting six sites flanking a conserved region in the N gene of SARS-CoV-2 ( Figure 4A , sequence information in Table S2 ). With RT-LAMP, the viral RNA was reverse transcribed into complementary DNA and subsequently amplified via self-priming loop structures and a high strand-displacing DNA polymerase that generates large cauliflower-like concatemers. [7] The RT-LAMP products were verified by agarose gel electrophoresis revealing multiple bands in a ladder-like pattern indicative of successful LAMP amplification, whereas a sample without RNA input lacked any such bands (red box of Figure 4B ). The concatemers were then subjected to restriction endonuclease digestion by HincII and EaeI, which recognize and cleave GTY/RAC and Y/GGCCR sites, respectively. These sites are conserved throughout the long concatemers such that their cleavage by digestion collapses the concatemer bands and produces significantly shorter sequences more suitable for detection. Figure 4B (blue box) reveals clean bands located below 200 base-pair (BP) length, confirming the uniform digestion. Next, the reaction product All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted October 7, 2021. ; https://doi.org/10.1101/2021.10.05.21264561 doi: medRxiv preprint was denatured into single-strand DNA (ssDNA) by heating to 95°C and cooling on ice, followed by mixing with a set of shells-based probes and incubated at 62 o C for 10 min or room temperature for 30 min. Significantly, a color change can be observed at RNA input ≥ 10 copies/µL (Figure 4C ). Figure 4D shows corresponding LSPR extinction spectra normalized at 537 nm. A linear relationship (R 2 = 0.99) was observed covering 3 logs from 1 to 1,000 copies/µL ( Figure 4E) . The LOD was determined to be 1 copy/µL and confirmed the capability of plasmonic LAMP approach for the single-molecule RNA detection. When used respiratory syncytial virus (RSV) RNA as initial target, it led to insignificant spectral change as compared to the reference sample, indicating the high specificity of this assay. The protocol developed here only requires easily accessible heat blocks and is suitable for POC testing.

It should be emphasized that the enzyme digestion and heat denaturation (steps ii and iii, Figure 4A ) involved in the plasmonic LAMP are necessary. Without step iii, directly detecting the RT-LAMP products via oligonucleotide sensors results in no color change since RT-LAMP amplicons are double-strand DNA (dsDNA) with no available binding sites (sample c, d, Figure   S12 ). Without step ii, the produced ssDNA after heating could easily re-anneal back as dsDNA in the hybridization buffer due to the extensive stem-loop and cauliflower-like structure of the RT-LAMP products. [30] Likewise, this would prevent the binding of the NP probes and does not lead to color change (sample e, Figure S12 ). In contrast, with step ii and iii, the dsDNA can be cut into uniform and short structures, which are amendable for subsequent hybridization detection with short oligonucleotides as probes, leading to highly specific and sensitive detection. [7] In contrast, previous reports rely on indirect sensing, where the ionic strength change in a salt solution or molecules labelled on primers essentially induce the aggregation of AuNPsbased sensors. [12] [13] [14] [31] [32] [33] The non-template amplification could still lead to false-positive results.

All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted October 7, 2021. ; https://doi.org/10.1101/2021.10.05.21264561 doi: medRxiv preprint

To demonstrate the advances of plasmonic LAMP over commercially available colorimetric LAMP kit, we compared their detection performance for SARS-CoV-2 RNA. Figure 5A shows the photographs of colorimetric LAMP incubating with different SARS-CoV-2 RNA inputs and RSV RNA as a negative control. The reaction was conducted for up to 45 min at 65°C. It can be seen that it only shows color change for SARS-CoV-2 RNA molecules of 50 copies/µL after 30 min incubation, while no color change can be observed for a short reaction time (e.g., 15 min).

In contrast, at time point of 45 min, all samples turn into yellow color, resulting in undistinguished positive and negative controls. This is indeed caused by the non-template amplification products and thus compromise the detection sensitivity. [9, 10] We then performed plasmonic LAMP for the samples processed with 15 min and 45 min amplification. Spectral analysis of the completed assay solutions suggests that plasmonic LAMP pushes down the detection limit to 50 copies/µL at 15 min and discriminates signals for 5 copies/µL against control sample at 45 min ( Figure 5B) . Evidently, plasmonic sensing enhances the detection sensitivity and specificity of LAMP and in turn makes the plasmonic LAMP a reliable singlemolecule diagnostic approach for nucleic acid detection.

To demonstrate the potential clinical use, we applied the integrated plasmonic LAMP approach to detect nasal swab samples that were spiked with SARS-COV-2 RNA. RNA of six different concentrations in the range of 0-100 copies/µL were spiked in healthy nasal swab samples and tested. The analytical performance shows sensitive detection ( Table 1 ) and suggests that the Au-Ag-shells-based sensors work well in a complex sample matrix and have potential uses in clinical setting. Taken together, the established diagnostic approach should be readily adopted as a nucleic-acid detection platform simply by changing the set of primers and probing sequences.

All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. 

In summary, we have developed the plasmonic LAMP approach for the single-molecule detection of SARS-CoV-2 RNA. With enzyme digestion and heat denaturation, plasmonic LAMP allows identifying amplicons via DNA hybridization utilizing Au-Ag nanoshells as sensors, yielding detection limit at 1 RNA copy per microliter. The Au-Ag shells can be prepared from AgNP templates through titration with Au + ions and have thick and hole-free surface. They show LSPR peak around 530 nm and high peak extinction, offering substantially enhanced detection sensitivity (20-times) compared to that of conventional AuNPs-based assay. Plasmonic LAMP has improved detection performance and only requires easily accessible heat blocks. Our work provides a diagnostic toolkit with simple readouts and ultralow detection limit that has potential for clinical applications. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted October 7, 2021. ; https://doi.org/10.1101/2021.10.05.21264561 doi: medRxiv preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted October 7, 2021. ; https://doi.org/10.1101/2021.10.05.21264561 doi: medRxiv preprint

Single-molecule nucleic acid interactions monitored on a label-free microcavity biosensor platform

Label-free single-molecule detection of DNA-hybridization kinetics with a carbon nanotube field-effect transistor

Ultra-sensitive and rapid detection of nucleic acids and microorganisms in body fluids using single-molecule tethering

Diagnostics based on nucleic acid sequence variant profiling: PCR, hybridization, and NGS approaches

Inhibition mechanisms of hemoglobin, immunoglobulin G, and whole blood in digital and real-time PCR

Microbial diagnosis of bloodstream infection: towards molecular diagnosis directly from blood

Loop-mediated isothermal amplification of DNA

Visual detection of isothermal nucleic acid amplification using pH-sensitive dyes

Point-of-care bulk testing for SARS-CoV-2 by combining hybridization capture with improved colorimetric LAMP

A All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted

Reduced false positives and improved reporting of loopmediated isothermal amplification using quenched fluorescent primers

Multiplexed recombinase polymerase amplification assay to detect intestinal protozoa

High sensitivity, loop-mediated isothermal amplification combined with colorimetric gold-nanoparticle probes for visual detection of high-risk human papillomavirus genotypes 16 and 18

Ultrasensitive and closed tube colorimetric loop mediated isothermal amplification assay using carboxyl modified gold nanoparticles

Comparison of the Accula SARS-CoV-2 test with a laboratory-developed assay for detection of SARS-CoV-2 RNA in clinical nasopharyngeal specimens

Evaluation of the Cue Health point-of-care COVID-19 (SARS-CoV-2 nucleic acid amplification) test at a community drive through collection center

Rapid and sensitive colorimetric detection of microsporidian Enterocytozoon hepatopenaei (EHP) based on spore wall protein (SWP) gene using loop-mediated isothermal amplification combined with DNA functionalized gold nanoparticles as probes

Reverse transcription loop mediated isothermal amplification (RT LAMP) combined with colorimetric gold nanoparticle (Au All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted

A comparative study of galvanic replacement reactions involving Ag nanocubes and AuCl 2¯ or AuCl 4¯

Silverbased plasmonic nanoparticles for and their use in biosensing

Ru nanoframes with an fcc structure and enhanced catalytic properties

Galvanic replacement: a simple and versatile route to hollow nanostructures with tunable and well controlled properties

MNPBEM-A Matlab toolbox for the simulation of plasmonic nanoparticles

An analytic model for the dielectric function of Au, Ag, and their alloys

Nonlinear optical behavior of Au@Ag coreshell nanostructures

Facile synthesis of gold-silver nanocages with controllable pores on the surface

Limit of blank, limit of detection and limit of quantitation

SARS-COV-2 mutation browser

Sensitive and selective detection of DNA fragments associated with Ganoderma boninense by DNA-nanoparticle conjugate hybridisation

Characterization of denaturation and renaturation of DNA for DNA hybridization

All rights reserved. No reuse allowed without permission.(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.The copyright holder for this preprint this version posted October 7, 2021. All rights reserved. No reuse allowed without permission.(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.The copyright holder for this preprint this version posted October 7, 2021. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.The copyright holder for this preprint this version posted October 7, 2021. ; https://doi.org/10.1101/2021.10.05.21264561 doi: medRxiv preprint All rights reserved. No reuse allowed without permission.(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.The copyright holder for this preprint this version posted October 7, 2021. ; https://doi.org/10.1101/2021.10.05.21264561 doi: medRxiv preprint