Prototyping of a lateral flow assay based on monoclonal antibodies for detection of Bothrops venoms

. Bothrops spp. are responsible for roughly 70% of all snakebites in Brazil, and envenomings caused by their bites can be treated with three types of antivenom: bothropic antivenom, bothro-lachetic antivenom, and bothro-crotalic antivenom. The choice to administer antivenom depends on the severity of the envenoming, while the choice of antivenom depends on availability and on how certain the treating physician is that the patient was bitten by a bothropic snake. The diagnosis of a bothropic envenoming can be made based on expert identification of the dead snake or a photo thereof or based on a syndromic approach wherein the clinician examines the patient for characteristic manifestations of envenoming. This approach can be very effective but requires staff that has been trained in clinical snakebite management, which, unfortunately, far from all relevant staff has. Results: In this article, we describe a prototype of the first lateral flow assay (LFA) capable of detecting venoms from Brazilian Bothrops spp. The monoclonal antibodies for the assay were generated using hybridoma technology and screened in sandwich enzyme-linked immunosorbent assays (ELISAs) to identify Bothrops spp.- specific antibody sandwich pairs. The prototype LFA is able to detect venom from several Bothrops spp. The LFA has a limit of detection (LoD) of 9.5 ng/mL in urine, when read with a commercial reader, and a visual LoD of approximately 25 ng/mL.

Background: Brazil is home to a multitude of venomous snakes; perhaps the most medically relevant of which belong to the Bothrops genus.Bothrops spp.are responsible for roughly 70% of all snakebites in Brazil, and envenomings caused by their bites can be treated with three types of antivenom: bothropic antivenom, bothrolachetic antivenom, and bothro-crotalic antivenom.The choice to administer antivenom depends on the severity of the envenoming, while the choice of antivenom depends on availability and on how certain the treating physician is that the patient was bitten by a bothropic snake.The diagnosis of a bothropic envenoming can be made based on expert identification of the dead snake or a photo thereof or based on a syndromic approach wherein the clinician examines the patient for characteristic manifestations of envenoming.This approach can be very effective but requires staff that has been trained in clinical snakebite management, which, unfortunately, far from all relevant staff has.
Results: In this article, we describe a prototype of the first lateral flow assay (LFA) capable of detecting venoms from Brazilian Bothrops spp.The monoclonal antibodies for the assay were generated using hybridoma technology and screened in sandwich enzyme-linked immunosorbent assays (ELISAs) to identify Bothrops spp.specific antibody sandwich pairs.The prototype LFA is able to detect venom from several Bothrops spp.The LFA has a limit of detection (LoD) of 9.5 ng/mL in urine, when read with a commercial reader, and a visual LoD of approximately 25 ng/mL.

Introduction
Snakebite envenoming has plagued mankind since time immemorial, annually exacting a toll of 81,000-138,000 deaths and roughly six million disability-adjusted life-years (DALYs) [1,2].Brazil alone experiences approximately 26,000-30,000 snakebite envenomings per year, most of which are caused by Bothrops species, with the species B. atrox contributing with an especially high number of bites [3][4][5].In addition to the Bothrops genus, bites from Crotalus durissus, Lachesis muta, and Micrurus spp.also occur, although to lesser extents [3].The mainstay treatment of snakebite envenoming is antivenom, and, in Brazil, antivenoms are available at a genus and inter-genus level in the form of bothropic antivenom, crotalic antivenom, bothro-lachetic antivenom, bothro-crotalic antivenom, and elapidic antivenom.The corollary is that snakebite diagnosis in Brazil must be undertaken at the same level; healthcare providers must determine whether the bite was dry (i.e., a bite in which no venom was injected), or whether it warrants antivenom administration, and, if so, which antivenom is appropriate when factoring in the snake that caused the bite [6,7].This is usually accomplished via a syndromic approach, where the clinical manifestations of envenoming are compared to those associated with the different snake genera [6].Unfortunately, some of the syndromes overlap, as the venoms of most Central and South American pit vipers are known to cause similar local effects and coagulopathies [6,8].These overlapping syndromes can complicate matters for healthcare providers trying to identify the optimal treatment, especially as not all healthcare providers have received adequate training in clinical snakebite management [9].This is unfortunate, as early and correct treatment has been shown to correlate with better patient outcomes for snakebite victims both in Brazil and abroad [10][11][12][13][14].
Several snakebite diagnostics based on immunoassays have been described in the literature, with an interesting recent example of improvement being the use of a cation exchange tip to pre-concentrate antigen in samples for a multiplex lateral flow assay (LFA) [6,15].However, to the best of our knowledge, the only snakebite diagnostic commercially available and clinically used is Seqirus' Snake Venom Detection Kit, and nearly all previously reported snakebite diagnostics have relied on undefined mixtures of polyclonal antibodies.For further information on the state of the art of snakebite diagnostics, see the recent review by Knudsen et al. [6].
Brazil was chosen as a case for the diagnostic assay developed during this study due to the wide availability of high-quality antivenoms, which are subsidised by the state, and due to the potential of diagnostics to facilitate early and correct treatment of snakebite envenoming.In this regard, supportive diagnostic tools must be developed to distinguish Brazilian pit viper bites at the same level at which clinical decisions are made (e.g., the genus level).Such novel, diagnostic tools could support efforts to secure faster treatment in the Amazon region, e.g., through antivenom decentralisation.Empowering healthcare providers at remote clinics to diagnose snakebites more easily and with higher precision enables them to choose the appropriate antivenom to treat envenomings [16].Conversely, in more metropolitan areas of Brazil, snakebite victims do not have to travel as far to reach a healthcare facility, and thus might present to the hospital before clinical manifestations of envenoming develop.In such scenarios, diagnostic tools might be able to speed up the diagnosis by providing healthcare workers with an idea of the offending snake before the venom has exerted its full toxic effects on the patient.As such, diagnostic tools would allow healthcare workers to prepare the correct treatment regime early on.Finally, snakebite diagnostic tools might also help lower the incidence of misdiagnosis of snakebite patients.
Here, we describe an attempt to develop such a snakebite diagnostic tool in the form of an LFA.LFAs are well-suited for the diagnosis of snakebite envenoming for several reasons.Firstly, the widespread use of LFAs during the COVID-19 pandemic has demonstrated the feasibility of implementing rapid diagnostics in both larger clinics, more remote settings, and even for home use by consumers.Moreover, LFAs are affordable, rapid, and user-friendly, requiring no specialised equipment or training to operate them.Combined, this makes LFAs suitable for point of care (PoC) use, especially in primary care settings.Additionally, they can be mass-produced for as little as 0.10-3.00USD per test [17].The affordability of these tests is especially relevant, as snakebite envenoming is associated with poverty and often occurs in regions with low-resource healthcare systems, meaning that the cost of diagnosis and treatment could easily become prohibitively high [18][19][20].Finally, LFAs can be carried out in 5-20 min, making them appropriate diagnostics for an acute disease, such as snakebite envenoming.An additional advantage is that LFAs rely on paper-based, disposable materials, making them relatively sustainable compared to diagnostics requiring more plastic components and/or harmful chemicals (e.g., enzyme-linked immunosorbent assays, abbreviated ELISAs).In the work behind this article, we have developed a prototype sandwich LFA using monoclonal antibodies capable of detecting venom from multiple Brazilian Bothrops species without cross-reacting with venoms from other Brazilian vipers.If taken into clinical development, such a tool could support clinical diagnosis and decision-making regarding treatment, thereby contributing to more snakebite patients in Brazil receiving the appropriate antivenom as early as possible.

Venoms
The whole venoms used in this study were procured in lyophilised form from the sources listed in the Supplementary Information (SI) Table S1.A precision scale was calibrated before each use and used to weigh 2 mg venom, which were subsequently dissolved in 1 mL of 0.14 M PBS of pH 7.4.The whole venom concentration in the resulting solution was calculated based on the weighed mass of lyophilised venom and the measured volume of buffer.Please note, the venoms used in this study are whole venoms, rather than specific toxins.

Immunisation
Immunisations were undertaken under ethical approval with j number 2015-15-0201-00680, University of Southern Denmark.Lyophilised whole venom from Bothrops atrox (a specimen from Brazil, Latoxan, L1210A) was reconstituted in a sterile 0.9% saline solution to a final concentration of 1 mg/mL.Either 5 or 10 μg venom, depending on the protocol, were mixed with aluminium hydroxide at a ratio of 1 mg aluminium hydroxide per 25 μg venom, in a solution of 0.05% methiolate, 50% Adjuvant P (Gerbu, 3111.6001), and 0.9% saline water to a volume of 100 μL to create an immunisation mixture.Immunisation mixtures were injected subcutaneously into four Naval Medicinal Research Institute (NMRI) mice: Two mice were injected with 5 μg venom, and two were injected with 10 μg venom.The mice were injected on days 1, 14, and 28, and they were bled on days 25 and 38 by taking a sample of 100 μL blood from the jaw of each mouse and C. Knudsen et al. transferring it to tubes containing EDTA.Subsequently, 200 μL 0.9% saline solution was added to the blood sample and the solution was centrifuged at 2000 x g for 5 min at room temperature, enabling the extraction of 250 μL plasma, which was tested in ELISA as described below to monitor antibody development.On day 42, the animals received an immunisation boost consisting of 20 μg whole venom dissolved in 0.9% saline to a volume of 100 μL, injected intravenously.

Bleed screenings
Clear 96-well plates (Thermo Scientific, MaxiSorp, 439454) were coated with 100 μL/well of 1000 ng/mL whole venom from either B. atrox, L. m. muta, or C. d. terrificus dissolved in carbonate buffer (pH 9.6, made from tablets as per the manufacturer's instructions, Medicago, 09-8922).The plates were incubated overnight at 4 • C or at ambient temperature and shaken at 300 rpm for 2 h, before being washed thrice with washing buffer (Ampliqon Laboratory Reagents, AMPQ40825.5).
Afterwards, 110 μL of mouse plasma diluted 1:55 in dilution buffer (10 mM phosphate, 140 mM NaCl, 0.5% w/v BSA, 0.0016% w/v phenol red, 0.05% v/v Tween-20, 0.1% v/v ProClin 950, pH 7.4) were added to each well.The plates were left shaking at 300 rpm for 1 h, and then washed thrice with washing buffer.Next, 100 μL of a 1:1000 dilution of horseradish peroxidase (HRP) conjugated, polyclonal, rabbit anti-mouse antibody (Dako, P0260) in dilution buffer (final concentration 1.3 ng/ mL) was added to each well.The plates were left shaking at 300 rpm for 1 h and subsequently washed thrice with washing buffer.Then, 100 μL of 3,3 ′ , 5,5 ′ tetramethylbenzidine (TMB) One substrate (Eco-Tek, 4380-12-15) were added to each well, and the plates were left in complete darkness for 12 min, before 100 μL of 0.5 M sulphuric acid were added to all wells to stop the reaction.The absorbance was measured at 450 nm and 620 nm on a Thermo Multiscan Ex plate reader.

Cloning and hybridoma generation
On day 45 of the immunisation schedule, the mice were sacrificed, and their spleens were extracted.The spleens were reduced to a singlecell-suspension with a mortar and pestle and were immediately mixed with SP2/0-AG14 myeloma cells and a PEG solution to fuse the B cells from the spleen with the myeloma cells.The resulting cells were spread into 96-well microtiter plates and grown in HAT medium for 7-10 days to select successfully fused cells.Culture supernatant from the different wells were tested with ELISA as described above, with the exceptions that an IgG-specific HRP-conjugated antibody (Merck, AP127P) was used for detection, and instead of mouse plasma, 1:50 and 1:100 dilutions of supernatant from the hybridoma growth media were used.The cell cultures corresponding to wells with positive ELISA signals for B. atrox venom and negligible signals for C. d. terrificus and L. m. muta venom were selected for cloning.These cells were transferred to HT medium and cultured further.The cells were then sequentially diluted in 96-well microtiter plates and tested with ELISA as described above, until wells with single cells were identified.These monoclonal cell lines were expanded and used in future experiments.

Antibody purification
NaCl was added to antibody-containing hybridoma culture supernatant to a final concentration of 2.5 M.After the salt was dissolved, 1.25 g of Celpure P65 (Honeywell, 525235) was added, and the culture supernatant was filtered through a 0.45 μm filter (Durapore® Membrane Filters 0.45 μm, HVLP04700).An ÄktaPrimePlus system was washed with ultrapure water to remove air in the tubing.The system was then primed with filtered culture supernatant, and the purification procedure was started (rProtein A sepharose Fast Flow (GE healthcare: 17-1279-03)), and the antibodies were purified.

Cross-reactivity screenings
The antibody binding profiles were investigated by screening the antibodies in ELISAs against whole venom from  S1.The ELISAs were carried out as described for the bleed screenings, with the exception that 100 μL/well purified antibodies in dilution buffer at 1000 ng/mL were used instead of mouse serum.Later, these cross-reactivity screenings were repeated with sandwich ELISAs and LFAs, using the protocols described below.Additionally, the following venoms were screened with LFA:

Biotinylation of antibodies
The antibodies were buffer exchanged into carbonate buffer using Nap5 columns (Illustra, GE Healthcare, 17085302) according to the manufacturer's protocol.The antibodies were eluted from the Nap5 columns with 1.5 mL of carbonate buffer, and the eluate was collected into a Vivaspin 6 50 kDa MWCO column (Sartorius, V10631), which was centrifugated at 2113 x g for 5 min to concentrate the antibodies.The antibody concentration was determined using an Eppendorf Bio-Photometer (model 6131).Biotin-N-hydroxysuccinimide (Sigma-Aldrich, H1759-5 mg) was dissolved in DMSO to a concentration of 0.4 μg/mL.Biotin-N-hydroxysuccinimide was added to the buffer exchanged antibodies at a ratio of 55 μg Biotin-N-hydroxysuccinimide per mg antibody, and the samples were vortexed immediately for 1 min, before being left with end-over-end rotation for 2 h.The reaction was stopped through addition of 50 μL 1 M Tris per 2.5 mL sample.The antibodies were buffer exchanged into 0.14 M PBS with 0.1% NaN 3 , concentrated, and their concentrations were measured again.The biotinylated antibodies were evaluated in ELISAs as described previously, with the exception that antibody dilution series in the range 0.5-500 ng/ mL were made of the biotinylated and unbiotinylated antibodies, respectively, using dilution buffer.The biotinylated and unbiotinylated antibodies in the dilution series were detected with different reagents: The biotinylated antibodies were detected with HRP-conjugated streptavidin (for two replicate dilution series) and with a 1:1000 dilution of HRP-conjugated anti-mouse antibody (Dako, P0260) (for another two replicate dilution series), while the unbiotinylated antibodies were only detected with the 1:1000 dilution of the HRP-conjugated anti-mouse antibody (also for two replicate dilution series).The HRP-conjugated streptavidin used for detection was prepared by reconstituting lyophilised HRP-conjugated streptavidin (KemEnTec, 14-30-00) in 50% glycerol and leaving it with end-over-end rotation for at least 90 min, before being diluted 1:5000 in HRP-StabilPlus buffer (KemEnTec, 4530A).

Sandwich pair screening
100 μL of 1000 ng/mL of unbiotinylated antibodies in carbonate buffer were coated in individual wells on 96-well plates and incubated at 4 • C overnight or for 2 h at ambient temperature and 300 rpm.The plates were washed thrice with washing buffer, 100 μL of 1000 ng/mL whole venom dissolved in dilution buffer were added to each well, and the plates were incubated for another hour at ambient temperature and 300 rpm.The plates were washed thrice, and 100 μL of 1000 ng/mL biotinylated antibody in dilution buffer were added to each well, before the plates were incubated and washed again thrice.From here on, the protocol is identical to those previously described with HRP-conjugated streptavidin used as a detection reagent.

Gold-conjugation of antibodies
The antibodies were buffer exchanged into ultra-pure water, concentrated, and the concentration was measured as described above for biotinylations.The antibodies were gold-conjugated using 40 nm gold particles at 15 OD/mL from a Naked Gold Conjugation Kit (Bio-Porto Diagnostics, NGIB18) according to the manufacturer's protocol.In addition to the salt tests described in the manufacturer's protocol, the suitability of the gold-conjugated antibodies was also evaluated in terms of false positives by comparing the results of positive and negative samples on LFA strips (see below).

Lateral flow assays
75 μL running buffer (0.14 M PBS with 5% BSA (w/v), 0.5% Tween-20 (v/v), 0.1% ProClin 950 (v/v)) were added to a tube and mixed with 5 μL 0.1 mg/mL biotinylated capture antibody (dissolved in running buffer), 5 μL gold-conjugated detection antibody, and 15 μL sample.The mixture was allowed to incubate for 5 min at room temperature.After 5 min, a commercially available LFA strip (BioPorto Diagnostics, gRAD1-120) was inserted into the tube.The strip was then either read every 10 s for 15 min (kinetic measurements) or read once after 15 min (point measurement) using a Cube Reader (ChemBio Diagnostics).Samples consisted of either running buffer, pooled human serum (Sigma-Aldrich, H4522-100 mL), or pooled human urine (Lee Biosolutions, 991-03-P) spiked with whole venom.For the interference study, each interferant was diluted in the matrix (pooled human serum or pooled human urine) to the final concentration seen in SI Table S2.The measurements were analysed in GraphPad Prism 9 (version 9.4.0)using a one-way ANOVA analysis followed by a post-hoc Dunnett analysis comparing each interferant mean to the mean of the control.

Sample preparation for native mass spectrometry
B. atrox venom and antibody samples were fractionated and exchanged into 200 mM ammonium acetate by size exclusion chromatography (SEC) as previously described [21,22].These experiments were performed on a Superdex Increase 200 10/300 GL column (Cytiva, Massachusetts, United States) pre-equilibrated with 200 mM ammonium acetate at the rate of 0.5 mL/min.Samples were collected and stored at 4 • C until used.The concentration of toxins in the SEC fractions was not adjusted prior to mixing with the antibody.

Native mass spectrometry
All mass spectrometry (MS) experiments were performed on a SELECT SERIES Cyclic IMS mass spectrometer (Waters, Manchester, U. K.), which was fitted with a 32,000 m/z quadrupole, equipped with an electron capture dissociation (ECD) cell (MSvision, Almere, Netherlands) in the transfer region of this mass spectrometer.Approximately 4 μL of sample were nano-sprayed from borosilicate capillaries (prepared in-house) fitted with a platinum wire.Spectra were acquired in positive ion mode, with the m/z range set to 50-8000.Acquisitions were performed for 5 min at a rate of 1 scan per second.The operating parameters for the MS experiments were as follows: capillary voltage, 1.2-1.5 kV; sampling cone, 20 V; source offset, 30 V; source temperature, 28 • C; trap collision energy, 5 V; transfer collision energy, 5 V; and Ion guide RF, 700 V.This instrument was calibrated with a 50:50 acetonitrile:water solution containing 150 μM caesium iodide (99.999%, analytical standard for HR-MS, Fluka, Buchs, Switzerland) each day prior to measurements.

Biolayer interferometry
All biolayer interferometry (BLI) experiments were performed on a Sartorius Octet K2 with Data Acquisition Software.High Precision Streptavidin 2.0 biosensors (Sartorius, SAX2) were used.Biotinylated IgGs were diluted to 1 μg/mL in kinetics buffer (Sartorius, 18-1105), and B. atrox whole venom was injected at two concentrations, namely 1 μg/ mL, and 10 μg/mL.A baseline signal was measured for 120 s in kinetics buffer.This was followed by the loading of the biotinylated IgGs onto the biosensor tip for up to 600 s.After this, a new baseline signal was measured for 120 s.The venom was allowed to associate for 600 s, followed by a dissociation step of 600 s in kinetics buffer.The biosensor tip was regenerated in NaOH pH 10, followed by a neutralisation step in kinetics buffer.A reference biosensor devoid of biotinylated IgG was used as the negative control.The results were analysed in Sartorius Octet K2 Data Acquisition Software and plotted using Graphpad Prism software.

Antibody discovery and characterisation
Mice were immunised with whole venom from Brazilian B. atrox specimens.ELISAs on the plasma from the mice were used to confirm that venom-specific antibodies had been raised.Once the immunisation schedule had been completed, the mice were euthanised, and B cells were harvested from their spleens.Hybridoma cell lines were generated, screened for expression of antibodies specific to venom, and cloned to monoclonality.This resulted in 38   ).The fourth antibody, conversely, was shown to also react with C. horridus venom (SI Figure S1).The antibodies that had been selected based on their binding profiles were biotinylated and screened against all 38 monoclonal antibodies in sandwich ELISAs to find the sandwich pairs necessary for the LFA.Our hypothesis was that only one sandwich pair component would need to have the desired genus specificity for the pair to be genus-specific.With the three selected detection antibodies, we found ten possible sandwich pairs that could detect B. atrox venom.All these pairs utilised the same detection antibody (antibody 86-14), while no sandwich partners were found for the other two detection antibodies.We subsequently investigated the binding profiles of these ten sandwich pairs against the panel of 21 venoms and confirmed that the binding profiles of the sandwich pairs reflected the binding profile of the genus-specific detection antibody (SI Figure S2).Out of the ten sandwich pairs, four seemed especially interesting due to the combination of their recognition of venoms from multiple Bothrops species and the comparatively high signals they yielded in the sandwich ELISAs (SI Figure S2).These four pairs were selected for further evaluations in LFAs.

LFA prototyping
The four selected non-genus-specific capture antibodies were biotinylated as preparation for use in LFAs, and the genus-specific, monoclonal detection antibody was conjugated to gold nanoparticles (AuNP- mAb).The gold-conjugation was carried out at ten different pHs, and the absorption spectra of each version of the conjugated antibody were investigated.Additionally, the blank signal (i.e., the test line signals of LFAs in which no antigen was added) of the different AuNP-mAbs were measured in LFAs to find the optimal conjugation conditions.Following successful conjugation, LFA prototypes were established for each of the four sandwich pairs using Generic Rapid Assay Device (gRAD) LFA strips (Fig. 1) [23].The gRAD is a commercially available, universal sandwich LFA that is not antigen specific.The gRAD's single test line is composed of biotin-binding proteins and its control line is immobilised anti-mouse antibodies.This generic configuration makes it possible to rapidly prototype LFAs, as long as the capture antibody is biotinylated and the detection antibody is murine.
The gRAD-based prototypes were used to detect a 1000 ng/mL solution of whole venom from B. atrox (this concentration is higher than the expected levels detected after a bite [24,25], and was intended as a positive control), and the test line and control line intensities were quantified using a low-cost, commercial LFA reader.Out of the four prototypes, the one relying on antibodies 86-14 and 86-11 provided the highest test line signal, so it was decided to keep working with this prototype.To assess the prototype's ability to detect the target antigen across a wide dynamic working range, the LFA was tested in decreasing concentrations (0 ng/mL -1,000,000 ng/mL) of B.atrox venom spiked into running buffer, pooled human urine, and pooled human serum.The resulting test line signal intensity was divided by the resulting control line signal intensity to give the T/C ratio, which was plotted against venom concentration to give a calibration curve.The results show an increase in signal intensity when the antigen concentration increases until a point between 10,000 and 100,000 ng/mL after which the signal intensity starts decreasing (see Fig. 2A, Fig. 2B, and SI Figure S3).This reduction of test line intensity at high antigen concentrations is characteristic of the hook effect [26].To determine the visual (by naked eye) and digital (by low-cost reader) limit of detection (LoD) of the prototype, LFA running buffer, pooled human serum, and pooled human urine were spiked with 0-20 ng/mL of B. atrox whole venom, and the test and control line signal intensities were measured (Fig. 2C and D).Linear regression was used to find the linear functions that best described the data.The LoDs were calculated using the formula LoD = 3.3⋅(σ/S), where σ is the standard deviation of the test line signal, S is the slope of the curve, and 3.3 is a constant [27].The parameters used in the calculations and the resulting LoDs can be seen in Table 1.Spiking serum with venom resulted in false positive signals (i.e., test line signal on tests in which no antigen was added) (Fig. 2C and D).However, the curve based on the serum samples had a steeper slope than the curves based on the LFA buffer and urine samples.The LoD as calculated based on data from the digital reader was therefore determined to be lower in serum samples (8.0 ng/mL) than in urine samples (9.5 ng/mL) and LFA buffer samples (10.3 ng/mL).However, when reading by eye, it was not possible to distinguish the signal of 8.0 ng/mL venom in serum from a blank sample.The stronger false positive signal observed in the serum samples indicates unspecific interactions between serum components and the assay.Due to the degree of false positives, the current version of the assay was not deemed suitable for detection of venom in clinical serum samples.However, a future iteration of the assay, in which the false positive rate in serum is reduced, might find utility for this application.False positives when measuring in serum samples have also been observed in other LFAs [28,29].Interestingly, a recent study of a venom-detecting LFA reports an improvement in the LoD due to a decrease in false positive signal in blood-derived samples, when using a cation tip for pre-concentration of the sample [15].A visual LoD of 25 ng/mL was comfortably achievable with the assay (SI Figure S3).While the visual LoD is slightly higher than the digital LoDs obtained with the commercial reader, it still indicates that the test is useable even without any specialised equipment.Additionally, the interactions of the IgGs with B. atrox whole venom were assessed using BLI.This technique enables the determination of apparent k off values for bivalent IgGs, as these are independent of antigen (ligand) concentration but cannot be used to determine k on values in this scenario, as the ligand concentration is not known for the employed batch of snake venom.The apparent k off values for the different IgGs were determined to be 7.1⋅10 − 3 /s and 4.4⋅10 − 3 /s (SI Table S3), indicating that the antibodies are slow to release bound ligand.Moreover, visual inspection of the sensorgrams indicates that the antibodies could have apparent affinities in the lower nM range (SI Figure S4), although determination of the antigen concentration would be needed to confirm this.

Antigen identification
To identify the antigen recognised by the sandwich pair, the antibodies 86-14 and 86-11 were screened against B. atrox venom proteins using electrospray ionisation mass spectrometry (ESI-MS) to look for binding partners.These mass spectra were acquired under 'soft' ionisation conditions, which allow non-covalent interactions to be preserved and transferred into the gas phase.Prior to screening, B. atrox venom was fractionated by SEC to separate the venom components by mass.This was done to generate venom protein mixtures, which were less complex than the whole venom, to make antigen identification easier, as well as to exchange the proteins into a spray solution that was appropriate for ESI-MS.Fig. 3A shows the SEC profile of the venom, which contained multiple protein peaks (labelled in the chromatogram), which corresponded to proteins of different masses.The size exclusion fractions for these peaks were mixed with the two antibodies in a 1:1 (volume:volume) ratio to screen for binding.Both 86-11 and 86-14 antibodies only bound to venom proteins from SEC fraction four (elution volume 15-16 mL) of B. atrox venom.To better understand the nature of these antibody antigen interactions, 86-14 and 86-11 were titrated against B. atrox SEC fraction four.Fig. 3B and C show the native mass spectra of both antibodies prior to mixing with the venom fraction, where the most abundant charge states were 24 + and 25 + for 86-11 and 86-14, respectively.Fig. 3D and E shows the mass spectra of the antibodies mixed in a 1:1 (volume:volume) ratio with a diluted (one in five) shown are the averages of duplicate measurements, and the error bars represent the standard deviations.Blue indicates that venom from Brazilian snake specimens were used.Purple indicates that the snake species is found in Brazil, but that the venom was extracted from a non-Brazilian specimen or that the origin of the specimen is unknown.Green indicates that the species is not found in Brazil.(For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)B. atrox SEC fraction four.Within these spectra, two prominent charge state distributions were detected within the m/z region 5400 to 7600 for 86-11 and 86-14.The difference in masses between the charge state series corresponded to the antibodies complexed with a 23.3 kDa protein (168.1 kDa vs. 144.8kDa for 86-11, and 169.1 kDa vs. 146.5 kDa for 86-14).When mixed with the undiluted venom fraction, the most prominent charge state series have masses corresponding to the antibodies bound to 46.6 kDa of antigen (Fig. 3F and G).Taken together, these titration experiments indicate that the antibodies are binding two 23.3 kDa venom proteins.To further test these observations, ions corresponding to antibody complexes with two toxins were isolated and fragmented by tandem MS (MS/MS) and subjected to collisional energy to dissociate the antigens.
For the MS/MS experiments, the 28 + charge states of the antibody toxin complexes were isolated and subjected to collisional energy.This was done to dissociate the antigens from the antibody to accurately calculate their masses.Fig. 4 shows the MS/MS spectra of each antibody: toxin complex after the application of collisional energy.For both antibodies (Fig. 4A and B), the ejected protein from the complexes had the mass of 23.3 kDa.A negative control was included to check the fragmentation patterns of only the 86-11 and 86-14 antibodies (SI Figure S5).The data from these spectra show that the only antigen ejected has a mass of 23.3 kDa, and that this mass did not correspond to the fragmentation of the antibody.Taken together, the ESI-MS data presented in Figs. 3 and 4 show that the antibodies bind specifically to the 23.3 kDa toxin twice.The masses of the ejected toxins are within the expected sequence mass range of type I snake venom metalloproteases, a major component of Bothrops venoms [30], and could fit with the mass of a toxin such as atroxlysin (uniprot P85420).However, further experiments are required to confidently identify the antigen.

Assay characterisation
To further characterise the LFA, it was screened against the panel of 21 venoms used in the ELISAs, as well as an additional 24 venoms to determine which venoms were recognised (see SI Table S1).The results indicate that the assay is specific towards the venoms of Bothrops species from both Brazil and some of its neighbouring countries, while it cannot detect venoms from viperids such as Crotalus and Lachesis spp.(Fig. 5).
Generally, LFAs are compatible with several different sample matrices.However, this requires careful optimisation to prevent matrix effects/false positives.To evaluate the effect of different sample matrices on venom detection by our antibodies, we spiked pooled human urine samples, pooled human serum samples, pooled human plasma (containing heparin) samples, and pooled human plasma (containing EDTA) samples with various amounts of B. atrox venom and detected the venom using both ELISAs and LFAs (Fig. 6 and SI Figure S6).The results in urine were comparable to the results in running buffer, both in terms of the low occurrence of false positives on negative tests and high signals on positive tests.In LFAs, the serum samples generally elicited lower signal intensities than the other types of samples.Plasma samples with heparin had the strongest false positive signals and only medium signal strength in true positive samples.
To assess whether compounds that might be found in human sample  matrices could cause false positives in the LFA, serum and urine samples were spiked with a series of compounds (the individual concentrations of which can be seen in SI Table S2) and tested with LFAs in the absence of antigen.Additionally, the influence of sample pH on the presence of false positives was assessed.The results of the experiment can be seen in Fig. 7. Urine samples generally yielded lower test line signals and higher control line signals than serum samples.The only tested compounds to cause false positives were glycine and heparin, and the only tested conditions to cause false positives were pHs 4.0, 5.0, and 5.5, and in all cases these false positives only occurred when tested in serum, indicating that serum causes matrix effects for the reported LFA.

Discussion
The LFA prototype presented here is the first report of the use of monoclonal antibodies in an LFA to detect Bothrops venom.The LoD of the prototype, when read with a reader, was 8.0 ng/mL in serum samples and 9.5 ng/mL in urine samples, with a visual LoD of approximately 25 ng/mL in buffer.These LoDs would possibly be sufficient to detect venom in most patient samples.However, the current version of the assay is likely not appropriate for venom detection in patient serum samples due to the propensity of serum samples to cause false positive signals in the assay.A literature study found that reported snake venom concentrations in patient samples across species varied from <1 ng/mL to >1000 ng/mL, and that venom concentrations were generally higher after bites caused by vipers than after bites caused by elapids [31].Two case studies of patients with bothropic envenomings revealed serum venom concentrations of 33.7 ng/mL and 62.6 ng/mL 5 h post-bite, respectively, well above the LoDs reported here [24,25].However, it is unknown whether these venom concentrations are representative of average bothropic envenomings.Generally, venom concentrations in patient samples are expected to vary depending on the sample matrix (e. g., urine or serum), time since the bite occurred, the body size of the patient, and other parameters.If the LoD of the prototype proves to be too high to detect venom in a certain sample type or at a certain time (e.g., in a urine sample immediately after the bite), it might be possible to improve the LoD by enriching the venom from the sample, or alternatively to measure another sample type (e.g., a wound swab or serum sample) [15,32].Conversely, if the antigen concentration in the sample is high enough to bring the assay into the hook effect range, it might be possible to dilute the sample.The LoDs reported in this study are lower than LoDs reported for other immunoassays intended for detection of snake venoms in Brazil, and on par (i.e., within the 5-10 ng/mL range) with LoDs reported for other venom-detecting LFAs [6,15].Additionally, the assay is also faster than other venom-detecting assays reported for Brazil [6].However, LFAs (e.g., Ref. [15]) and ELISAs (e.g., Ref. [33]) have been reported targeting different snake venoms from other parts of the world and which have lower LoDs than the current prototype LFA [6], e.g., LoDs ≤1.5 ng/mL in the case of the ELISA developed by Maduwage et al. [33].When comparing LoDs, it is worth noting that the LoDs reported in this study are defined by the concentration of whole venom in the mock patient samples, rather than by the concentration of whole venom after dilution of the sample in the LFA running buffer or by the concentration of the specific antigen recognised by the monoclonal antibodies.
One limitation of this study is that the LFA was not evaluated on real patient samples.Therefore, it remains to be seen to what extent the LFA is capable of detecting venom in real samples and whether any untested compounds interfere and cause false positives or false negatives.It is promising that the more extreme pHs tested in this study did not cause false positives in the urine samples, as the physiological pH of urine is known to vary considerably.Conversely, the physiological pH of blood is much less variable.It is possible, however, that anticoagulants commonly used to prepare blood-derived samples, e.g., citrate, could affect the pH of a sample.Two compounds that were not tested in this study, but which could potentially cause issues are antivenom and biotin.Antivenom is often added to snakebite patient blood samples to prevent clotting, and it is likely that the polyvalent antivenom antibodies can block the epitopes recognised by the monoclonal antibodies employed in the LFA.This potential issue might be alleviated by measuring serum or plasma samples derived from blood samples collected before antivenom administration or by using different sample types.As for biotin, it has recently been suggested that overconsumption of biotin by the general population might lead to elevated biotin levels in plasma, which could potentially interfere with immunoassays relying on biotin-binding, such as the gRAD [34].
A study showed that 4.2% of snakebite victims in Brazil received two or more kinds of antivenom, and that 10.5% of patients bitten by Bothrops spp. received polyvalent antivenom [35].Administration of polyvalent or multiple antivenoms might indicate a lack of confidence in the identification of the type of snakes involved in snakebite accidents.This notion could be supported by another study of 1063 snakebites in Brazil, in which it was found that only 44% of snakebites were identified at the genus level [36].Lack of confidence in the identification of the snake involved in an accident is potentially problematic, as it has been demonstrated that a delay in treatment, either due to insufficient or incorrect administration of antivenom, for Crotalus bites in Brazil led to an increased risk of acute renal failure for the patients [13].Additionally, it has been argued that there is a systematic lack of training of healthcare professionals in clinical snakebite management in certain states in Brazil [37,38].Thus, snakebite diagnostic tools could potentially help mitigate the risk of inappropriate antivenom administration and promote the use of monovalent antivenoms, especially in cases where the treating personnel is not trained in clinical snakebite management.Use of monovalent antivenoms can be advantageous when the snake is known, as they contain more antibodies specific to a given type of snake, thereby potentially enabling healthcare providers to administer a smaller volume of antivenom [39,40]; this would be preferable because many antivenoms are highly immunogenic, and higher doses are more likely to elicit adverse events.In the future, the LFA might also be used to confirm envenoming prior to administration of upcoming first-line-of-defence drugs, if drug candidates such as varespladib and marimastat are approved for treatment of snakebite envenoming [41].This could in turn improve patient outcome even further.
Taken together, the data presented here constitute a proof-ofconcept for a rapid diagnostic test for Bothrops envenomings.Potentially, a diagnostic tool, such as the LFA presented here, could be used to confirm or disprove suspected Bothrops envenomings, thereby guiding the choice of antivenom not only for the 10.5% of patients with bothropic bites who receive polyvalent antivenom but also for the 34.4% of patients with crotalic bites who receive either bothropic or bothrocrotalic antivenom [35].

Conclusions
Here, we have presented a prototype LFA capable of distinguishing the venoms of several Bothrops spp.from the venoms of non-Bothrops spp.The LFA had an LoD of 9.5 ng/mL in urine, when read with a commercial reader, and a visual LoD of approximately 25 ng/mL.With further improvements to the prototype's diagnostic sensitivity and specificity, and a thorough validation study, it is possible that the LFA could empower healthcare providers who have limited or no experience in clinical snakebite management to diagnose snakebite victims more confidently.This might be especially valuable in remote clinics and as a support to ongoing efforts to decentralise antivenom in the Brazilian Amazon, thereby bringing treatment closer to those who need it most and ameliorating the burden of the highly neglected disease of snakebite envenoming.

•
Development of monoclonal antibodies for genus-specific snake venom detection.• Development of sandwich lateral flow assay with a limit of detection in the single-digit ng/mL range in spiked urine samples.• Proof of concept for a rapid diagnostic test for Bothrops envenomings in Brazil.

Fig. 1 .
Fig. 1.Schematic representation of the generic rapid assay device (gRAD), which was used during this study.The gRAD is a commercially available, lateral flow strip intended for sandwich assays.It has a test line consisting of a biotin-binding protein, which can immobilise a biotin-conjugated capture antibody.The control line consists of immobilised anti-mouse antibodies.

Fig. 2 .
Fig. 2. Calibration curves of sandwich pair 86-14 and 86-11 as measured with LFAs.A&B) B. atrox whole venom was dissolved in either LFA running buffer, pooled human urine, or pooled human serum at various concentrations and measured in duplicates by LFA to assess the concentration-signal correlation of the assay.Both the test line intensities as quantified with a commercial reader A) and the test line intensity to control line intensity (T/C) ratios B) are shown here.C&D) Low concentrations of B. atrox venom were dissolved in LFA running buffer, pooled human serum, and pooled human urine, respectively, and measured in six replicates in LFAs to assess the LoD of the test.Error bars represent the standard deviation between two measurements (A&B) and six measurements (C&D), respectively.

C
. d. terrificus, and L. m. muta, to identify antibodies that bind only to Bothrops venom, without cross-reacting to venoms from either of these two other medically relevant pit vipers.Out of these 38 antibodies, four bound only B. atrox venom (B.atrox signal >3.0, other signals <0.5).These four antibodies were screened in further ELISAs against a panel of 21 Latin American snake venoms (including the original three venoms), and three out of the four antibodies were selected, as they retained a binding profile where only Bothrops venoms elicited strong signals (Bothrops signal >2.5, other signals <0.5

Fig. 3 .
Fig. 3. Size exclusion chromatography (SEC) and mass spectrometry data for the titration of B. atrox venom against the monoclonal antibodies 86-11 and 86-14.A) The size exclusion chromatogram of the B. atrox venom, with the peak corresponding to the antigen fraction highlighted in blue.Native mass spectra of B) antibody 86-11 (green) and antibody 86-14 (red), C) antibody 86-11 prior to mixing.The mass spectra for antibodies 86-14 and 86-11 mixed with the five times diluted (D & E, respectively) and undiluted (F & G, respectively) fraction from the SEC of B. atrox venom.(For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 4 .Fig. 5 .
Fig. 4. Tandem mass spectrometry experiments for the antibody:(2) toxin complexes.Collision induced dissociation (CID) of antibody 86-11 (green) and antibody 86-14 (red) complexed the antigens are shown in spectra A) and B), respectively.(For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 6 .
Fig. 6.Comparison of matrix effects in LFAs.B. atrox venom was diluted in LFA running buffer, pooled human serum, pooled human plasma (containing heparin), pooled human plasma (containing EDTA), and pooled human urine.The antigen in the dilution series was detected with the LFA, and the test line intensities A) and T/C ratios B) are displayed here.The results shown are the averages of triplicates, and the error bars represent the standard deviations.

Fig. 7 .
Fig. 7. Screening of potential interferents in LFAs.Serum A) and urine B) samples were spiked with different compounds and tested in LFAs in the absence of antigen to investigate whether the compounds could cause false positives.In addition, the solvents (water, DMF, DMSO, and methanol) used for dissolving the compounds were tested.Serum and urine at their natural pHs, and pH-adjusted serum and urine samples were also tested.The error bars represent the standard deviations between triplicate measurements.

Table 1
monoclonal cell lines expressing antibodies capable of recognising B. atrox venom.The 38 resulting antibodies were screened in ELISAs against whole venom from B. atrox, Parameters used for LoD calculations.Blank signal refers to the test line intensity when the LFA is used to test a sample not containing the target antigen.S is the slope of the curve, and σ is the standard deviation of the test line signal.Eight replicates were performed for the blank measurements, and six replicates were performed for all other measurements.