An Antigen Capture Assay for the Detection of Mycolactone, the Polyketide Toxin of Mycobacterium ulcerans

Buruli ulcer (BU), or Mycobacterium ulcerans disease, is a chronic necrotizing disease of skin and soft tissue (1) whose pathology is mediated by a plasmid-borne cytotoxin known as mycolactone (2, 3). Acquisition of the pMUM plasmid, which encodes the machinery necessary for mycolactone biosynthesis, led to the divergence of M. ulcerans from a common ancestor with the closely related M. marinum and is a defining characteristic of the species (4, 5). As such, all pMUM-containing mycobacteria are referred to as the mycolactone-producing mycobacteria (MPMs) (6). Mycolactone is a polyketide cytotoxin with immunosuppressive and analgesic properties, and at least six different variants of mycolactones have been described, all of which comprise an invariant lactone core with a C-linked short upper side chain and a longer lower acyl side chain (2, 3). As mycolactone is unique to the MPMs, it is an ideal target for the specific diagnosis of BU. mAbs targeting the core and/or the short upper side chain could allow detection of all known mycolactone variants because of the structural invariance in these parts of the molecules.

BU control is contingent upon early diagnosis and prompt treatment initiation. BU diagnosis currently hinges on the PCR detection of the M. ulcerans–specific insertion sequence (IS) 2404 in clinical samples. Clinical diagnosis is complicated by a multitude of differential diagnosis characteristic of the clinical picture (7, 8). The only decentralized laboratory diagnosis for the confirmation of clinical diagnosis currently available is the microscopic detection of acid-fast bacilli in wound swab samples or fine needle aspirates (9). Although this is a low-cost and relatively easy to perform method, it has low sensitivity. The highly specific and sensitive IS2404 PCR requires sophisticated laboratory equipment and well-trained laboratory personnel and is, in low-resource settings, typically only available in few reference centers. This in turn leads to delays in diagnosis and treatment, thus thwarting BU control. Moreover, PCR analysis is unreliable unless performed under strict quality control (10). An easier-to-use rapid diagnostic test (RDT) is, therefore, in urgent need for use in peripheral and field settings in low-resource BU-endemic regions.

We recently reported the generation of mAbs capable of specific binding to mycolactone and developed an Ag-competitive immunoassay based on these mAbs (11, 12). Although this assay is highly specific and sensitive, competitive assays do have a few shortcomings compared with Ag capture assays. For one, by using only one mAb, competitive assays have a higher probability of cross-reactivity (i.e., lower specificity) compared with Ag capture assays, which usually use two mAbs of different fine specificities (13, 14). Also, the need to have a suitable reporter molecule for competitive assays, like a biotinylated mycolactone variant (12), increases the complexity of scale-up for such an assay. Mycolactone is a notoriously difficult molecule to synthesize, even though several total syntheses strategies have been reported (15–18). Finally, conversion of immunoassays into point-of-care lateral flow formats is more facile for Ag capture than for competitive assays. For these reasons, we set out to develop a capture assay for mycolactone detection, which could, in the future, be converted into an RDT.

Approval for the collection of swab samples for BU diagnosis was obtained from the Cameroonian Comité National D’Ethique de la Recherche pour la Santé Humaine. Immunogenicity studies in mice and the generation of mAbs were performed under approval by the animal welfare committee of the Canton of Basel-Stadt (authorization number BE95/17). All animal experimentation was conducted in compliance with the Swiss Animal Welfare Act, Animal Welfare Ordinance, and the Animal Experimentation Ordinance.

The chemical synthesis of mycolactone and mycolactone derivatives (Fig. 1) has been described elsewhere (15, 16, 19). All synthetic products were HPLC purified and dissolved in DMSO to give 1 mg/ml stock solutions.

Mouse immunization and hybridoma generation were essentially done as described previously (11). Mouse immunization was done either with PG-180 or PG-203 (Fig. 1) coupled to BSA with the carbodiimide cross-linker EDC. Hybridoma selection was done by ELISA using a panel of biotinylated mycolactone derivatives comprising MG-158, MG-160, MG-161, PG-183, and PG-204 as target Ags (Fig. 1). For the selection tests, NeutrAvidin-coated plates (Thermo Fisher Scientific) were coated with each mycolactone derivative at a concentration of 1 µg/ml (100 µl/well) overnight at 4°C. Plates were washed three times with washing buffer (ddH₂O with 0.3% Tween-20) and blocked with SuperBlock T20 (Thermo Fisher Scientific) for 1 h at 37°C. Hybridoma culture supernatants were added and incubated for 2 h at 37°C, and after washing as described above, bound Abs were detected by incubating with goat anti-mouse IgG Abs coupled to HRP (SouthernBiotech) for 1 h at 37°C. Plates were again washed, and signal development was done with 3,3′,5,5′-tetramethylbenzidine (TMB; KPL SeraCare, catalog no. 5120-0047) for 7 min, after which the reaction was stopped with 0.5 M sulfuric acid. Selected hybridoma lines were cloned twice by limiting dilution, and mAbs were purified from culture supernatants by affinity chromatography using HiTrap Protein A HP columns (GE Healthcare) and a low-pressure liquid chromatography system (Model EP-1 Econo Pump; Bio-Rad Laboratories). All the mAbs were of IgG₁ isotype, except for LW1.1b, which was IgG_2b. Purified mAbs were dialyzed against PBS in Slide-A-Lyzer dialysis cassettes (Thermo Fisher Scientific) for 24 h and then sterile filtered and stored at 4°C until needed or at −80°C for long-term storage.

Binding of purified mAbs to the panel of biotinylated mycolactone derivatives used for hybridoma selection (Fig. 1) was assessed by ELISA. NeutrAvidin-coated plates (Thermo Fisher Scientific) were coated with each mycolactone derivative at a concentration of 1 µg/ml (100 µl/well) overnight at 4°C. Plates were washed three times with washing buffer (ddH₂O with 0.3% Tween-20) and blocked with SuperBlock T20 (Thermo Fisher Scientific) for 1 h at 37°C. Meanwhile, 5-fold serial dilutions of each purified mAb from a starting concentration of 10 µg/ml were prepared in PBS/Tween-20 (PBST). The ELISA plate was washed as described above, and the mAb dilutions were then added in. After incubating for 2 h at 37°C, plates were washed, and bound Abs were detected by incubating with goat anti-mouse IgG Abs coupled to HRP (SouthernBiotech) for 1 h at 37°C. Plates were again washed, and signal development was done using TMB (KPL SeraCare) for 7 min, after which the reaction was stopped with 0.5 M sulfuric acid.

mAbs at a concentration of 0.5 mg/ml were biotinylated using EZ-Link Sulfo-NHS-LC-Biotin (Thermo Fisher Scientific). Briefly, a Sulfo-NHS-LC-Biotin solution at 5.7 mg/ml was prepared by dissolving the appropriate amount of Sulfo-NHS-LC-Biotin powder in DMSO. For every 100 µl of mAb, 3 µl of the Sulfo-NHS-LC-Biotin solution was added, and the mixture was incubated at an ambient temperature for 30 min on a shaker and then dialyzed against PBS for 24 h. The biotinylated mAb was recovered and stored at 4°C until needed.

Ethanolic extracts were prepared from 8-wk-old cultures of an African M. ulcerans strain (S1013) cultivated on Middlebrook 7H9 agar plates supplemented with 0.2% glycerol and 10% oleic acid, albumin, dextrose, catalase supplement (20). Colonies were scraped from four well-grown plates with a sterile inoculating loop, transferred into 15 ml of absolute ethanol (Sigma), and incubated at an ambient temperature for 3 d protected from light. Afterwards, the suspension was vortexed for 1 min and centrifuged at 4000 × g for 10 min to pellet the bacterial debris. The supernatant was filtered through sterile 0.22-µm syringe filters and dried by vacuum centrifugation (SpeedVac, Thermo Fisher Scientific). The resulting mycolactone preparations were resuspended in DMSO and stored at −20°C until needed.

Swab samples were taken from ulcerative BU lesions as described (21) and were extracted, after long-term storage and transport, in 500 µl of PBS by vigorous bead vortexing. Quantitative PCR (qPCR) analysis was done with DNA extracted from 50 µl of the sample, as described (22). Lipid extracts were prepared from 400 µl of the sample as described previously (11). Briefly, each sample was divided into 50-µl aliquots, and 950 µl of a chloroform–methanol (2:1, v/v) solution was added to each aliquot. Samples were incubated for 2 h at 25°C with shaking, after which 200 µl of ddH₂O was added to induce a phase separation, and the samples were vigorously vortexed. After centrifuging for 10 min at 13,300 × g, the lower organic phase was transferred to a fresh tube and dried by vacuum centrifugation (SpeedVac, Thermo Fisher Scientific). The dried pellets were resuspended in 200 µl of ice-cold acetone, vigorously vortexed, and centrifuged at 13,300 × g. The supernatants for all aliquots of each sample were pooled into a fresh tube and again dried by vacuum centrifugation (SpeedVac, Thermo Fisher Scientific). The resulting lipid extract was stored at −20°C until needed.

MaxiSorp plates (Thermo Fisher Scientific) were coated with a capturing mAb at a concentration of 4 µg/ml (100 µl/well) in PBS and incubated overnight at 4°C. Plates were then washed three times with washing buffer (ddH₂O with 0.3% Tween-20) and blocked with SuperBlock T20 (Thermo Fisher Scientific) for 1 h at 37°C. Serial dilutions of (synthetic or extracted) mycolactone were prepared in a triethanolamine (TEA) buffer (0.2 M TEA [pH 7.5] with 20% DMSO). After washing the plates as described above, the mycolactone dilution series was added to the blocked plate and left to incubate for 2 h at 37°C. Plates were washed again, and biotinylated mAb (2 µg/ml) in the TEA buffer was added and incubated for 1.5 h at 37°C. Bound biotinylated mAb was detected using HRP-conjugated streptavidin (SouthernBiotech) diluted 1:5000 in PBST and incubated for 1 h at 37°C. After a final washing step, plates were developed with TMB incubated at an ambient temperature for 7 min, after which the reaction was stopped with 0.5 M sulfuric acid. Absorbance was measured at 450 nm with an ELISA microplate reader (Tecan Sunrise), and results were illustrated using GraphPad Prism version 8 (GraphPad Software, San Diego, CA) or R (version 3.6.1, package tidyverse).

Culture filtrates of two M. ulcerans strains isolated from BU lesions of patients from Cameroon (strain S1013) and Australia (strain S1251) were tested (23). M. ulcerans strains were cultured in BacT/Alert MP liquid medium (BioMerieux) for at least 8 wk. Cultures were harvested and spun at 13,300 × g to pellet the bacteria, and the supernatants were filtered through sterile 0.22-µm syringe filters. Culture filtrates were used directly in the assay without any mycolactone extraction. Briefly, MaxiSorp plates (Thermo Fisher Scientific) were coated with a capturing mAb at a concentration of 4 µg/ml (100 µl/well) in PBS incubated overnight at 4°C. Plates were then washed three times with washing buffer (ddH₂O with 0.3% Tween-20) and blocked with SuperBlock T20 (Thermo Fisher Scientific) for 1 h at 37°C. Serial dilutions of culture filtrates were prepared in a TEA buffer (0.2 M TEA [pH 7.5], with or without 20% DMSO), starting from the undiluted filtrate and proceeding in 2-fold dilutions down the series. The rest of the assay was done as described above.

The assay was optimized using a modified TEA buffer containing varying concentrations of the chaotropic salts magnesium chloride (MgCl₂), magnesium sulfate (MgSO₄), or ammonium thiocyanate, in the presence of up to 50% FBS or human serum. Where necessary, tyramide amplification was done to improve signals. Tyramide amplification is a signal enhancement method used for assays involving HRP, including ELISA and Western blots. In principle, HRP in the presence of hydrogen peroxide catalyzes the activation of labeled tyramide substrates (in this case, biotinylated tyramide), which then rapidly bind to nearby tyrosine residues (e.g., in HRP). This, therefore, increases the number of biotin molecules present in the reaction, thus increasing the overall signal gotten at the end of the assay.

Assays involving tyramide amplification were done as outlined below. MaxiSorp plates (Thermo Fisher Scientific) were coated with a capturing mAb at a concentration of 4 µg/ml (100 µl/well) in PBS incubated overnight at 4°C. Plates were then washed three times with washing buffer (ddH₂O with 0.3% Tween-20) and blocked with PBST containing 3% BSA for 1 h at 37°C. Serial dilutions of extracted or synthetic mycolactone were prepared in a TEA buffer without DMSO (0.2 M TEA [pH 7.5]) containing the different chaotropic salts and up to 50% serum from healthy donors. After washing the plates as described above, the dilutions were added to the blocked plate and left to incubate for 2 h at 37°C. Plates were washed again, and biotinylated mAb (2 µg/ml) in the test assay buffer was added and incubated for 1.5 h at 37°C. The plate was washed, and HRP-conjugated streptavidin (SouthernBiotech) diluted 1:50,000 in PBST containing 1% BSA was added and incubated for 1 h at 37°C. After washing, biotinyl tyramide (Sigma) prepared at a concentration of 1 µg/ml (70 µl/well) in a citrate buffer containing 0.02% hydrogen peroxide (KPL SeraCare, catalog no. 5120-0047) was added to the plate and incubated at an ambient temperature for 15 min. The plate was again washed, and HRP-conjugated streptavidin (SouthernBiotech) diluted 1:50,000 in PBST containing 1% BSA was added and incubated at an ambient temperature for 15 min. After a final washing step, plates were developed with TMB (KPL SeraCare, catalog no. 5120-0047) incubated at an ambient temperature for 5 min, after which the reaction was stopped with 0.5 M sulfuric acid. Absorbance was measured at 450 nm with an ELISA microplate reader (Tecan Sunrise), and results were illustrated using GraphPad Prism version 8 (GraphPad Software, San Diego, CA).

For extracts from swab samples, MaxiSorp plates (Thermo Fisher Scientific) were coated with a capturing mAb at a concentration of 4 µg/ml (100 µl/well) in PBS incubated overnight at 4°C. Plates were then washed three times with washing buffer (ddH₂O with 0.3% Tween-20) and blocked with PBST containing 3% BSA for 1 h at 37°C. Lipid extracts from swabs were resuspended in 150 µl of assay buffer (0.2 M TEA with 0.2 M MgCl₂ [pH 7.5]) containing a mixture of mouse Abs (1 mg/ml) and added to blocked plates. After incubating for 2 h at 37°C, plates were washed, and biotinylated mAb (2 µg/ml) in assay buffer was added and incubated for 1.5 h at 37°C. Plates were washed, and HRP-conjugated streptavidin (SouthernBiotech) diluted 1:5000 in PBST containing 1% BSA was added and incubated for 1 h at 37°C. After washing, biotinyl tyramide (Sigma) prepared at a concentration of 1 µg/ml (70 µl/well) in a citrate buffer containing 0.02% hydrogen peroxide (KPL SeraCare, catalog no. 5120-0047) was added to the plate and incubated at an ambient temperature for 15 min. The plate was again washed, and HRP-conjugated streptavidin (SouthernBiotech) diluted 1:25,000 in PBST containing 1% BSA was added and incubated at an ambient temperature for 15 min. After a final washing step, plates were developed with TMB (KPL SeraCare, catalog no. 5120-0047) and incubated at an ambient temperature for 7 min, after which the reaction was stopped with 0.5 M sulfuric acid. Absorbance was measured at 450 nm with an ELISA microplate reader (Tecan Sunrise), and results were illustrated using GraphPad Prism version 8 (GraphPad Software, San Diego, CA).

For the design of an Ag capture assay, usually a capture and a detecting mAb that recognize two nonoverlapping epitopes on the analyte are required. We therefore aimed at generating mAbs with different fine specificity for the relatively small mycolactone A/B molecule, which has a molecular mass of 743 Da. Two different immunogens, PG-180 with an intact lower side chain and an upper side chain that is extended by a linker and PG-203 with an unmodified upper side chain and the lower chain replaced by a linker (Fig. 1), were used to generate sets of mAbs with varying fine specificities (Table I). One of these sets of mAbs (JD5.1–JD 5.12) has been previously described (11, 12).

Immunizing mice with PG-180 led to the generation of two subsets of mAbs. Subset 1 mAbs recognized PG-183, the biotinylated variant of PG-180, but none of the derivatives which had the lower side chain replaced by a linker moiety, and subset 2a mAbs recognized all mycolactone variants with an intact core (MG-158, MG-160, MG-161, PG-183, and PG-204). Binding of subset 2a mAbs was not influenced by modifications of the upper chain, with the exception of PG-183 (with the upper chain being extended by the linker), which was only recognized with low affinity. Immunization with PG-180 thus seems to have generated mAbs that recognized primarily the hydrophobic lower part of mycolactone.

Three subsets of mAbs were obtained from PG-203–immunized mice. As expected, all of them recognized PG-204, the biotinylated variant of PG-203, and also MG-161, a PG-204 derivative with only a minor modification of the upper side chain (Fig. 1). Two mAbs (LW7.1 and 7.2) showed the same fine specificity as the subset 2a mAbs obtained after PG-180 immunization. Fine specificity of subset 2b mAbs resembled subset 2a mAbs in that they recognized all mycolactone variants with an intact core, but in contrast to the latter, their binding to PG-183 was also strong. Subset 3 mAbs were obtained from two different PG-203–immunized mice and include mAbs JD5.1–JD5.12, which have been described before (11, 12). This subset recognized PG-204 and MG-161 and, to varying degrees, the mycolactone derivatives MG-158 and MG-160 with more pronounced modifications at the upper side chain (Fig. 1). Seven mAbs within this subset (JD5.2, JD5.3, JD5.6, JD5.7, JD5.12, LW7.10, and LW7.11) were likely recognizing parts of the upper side chain distal from the core, as evidenced by their recognition of MG-161 and PG-204 (with little or no modification of this side chain) and lack of or only slight recognition of the remaining derivatives (which have more extensive modifications of this side chain). Immunization with PG-203 thus seems to have generated mAbs that recognized either primarily the core or the upper side chain of mycolactone.

Further analyses revealed that subset 1 mAbs, i.e., all those which primarily recognize the hydrophobic side chain of mycolactone, were only able to bind to PG-183 (which has the hydrophobic side chain intact) when it was affixed to a solid support but not when it was in aqueous solution. In contrast, all other mAbs, i.e., those which primarily recognized the lactone core and/or the upper side chain of mycolactone, were able to recognize mycolactone derivatives both on solid support and in solution. This is consistent with findings that indicate that because of its amphiphilic structure, mycolactone forms aggregates in aqueous solutions, with the hydrophobic side chain sequestered within the interior of these aggregates (24). Consequently, mAbs recognizing the hydrophobic side chain are unable to recognize aggregated mycolactone. Only by affixing mycolactones with an unmodified lower side chain to a solid support, thereby preventing aggregation, can they be recognized by subset 1 mAbs.

Mycolactone extracts were used to screen mAb pairs to select appropriate matched pairs. Typically, a capture assay is designed using two different mAbs, each recognizing nonoverlapping epitopes on the analyte in question. Thus, we were initially aiming for mAb pairs, in which one mAb would bind to the lactone core, whereas the other would recognize the upper side chain of mycolactone. However, mycolactone aggregates could potentially present epitope repeats, making it possible for the same mAb to function both as capturing and as detecting reagent. Therefore, we performed a screen with all feasible mAb pairs to identify both “mixed pairs” (with each of the two mAbs recognizing a different part of the mycolactone structure) and “same pairs” (with both mAbs recognizing parts of the same global region of the mycolactone structure). Indeed, when screening 1,360 mAb pairs to identify matched pairs suitable for an Ag capture assay, we confirmed that both mixed pairs and same pairs could function as matched pairs (Fig. 2). Of the 1360 mAb pairs, 213 potentially suitable matched pairs were identified.

Generally, these 213 matched pairs featured mAbs from subsets 2b and 3, and no pairings involving mAbs from subsets 1 and 2a functioned as matched pairs. There were 14 mAbs, comprising 9 subset 2b and 5 subset 3 mAbs, which were able to function both as capturing and detecting mAb in the same reaction. Except for mAb LW7.15, which gave high background signals, and mAbs JD5.11 and LW7.16, all subset 2b and 3 mAbs were present in at least one matched pair. Altogether, 9 of the 12 subclass 2b and 6 of the 16 subset 3 mAbs were involved in at least 10 matched pairs. Some mAbs showed marked differences in their suitability as detecting versus coating mAbs, as for example the subset 3 mAb JD5.10, which was present in 20 matched pairs as detecting mAb (more than any other mAb) but only in 8 pairs as coating mAb.

Of the 213 matched mAb pairs, 144 gave strong signals (OD₄₅₀ > 1.5) and did not show a prozone effect at relatively low concentrations of mycolactone and were thus selected for the next screening step. In this step, matched pairs that could detect the potentially nonaggregating mycolactone molecule PG-120 with a truncated lower side chain (Fig. 3A) were selected. Both intact mycolactone and mycolactone fragments that have lost the light-sensitive lower side chain may be present in biological samples, and matched pairs that recognize PG-120 should be able to detect both.

Following this screen (Fig. 3), matched pairs comprising the subset 3 mAb LW7.11 as a capturing mAb and any of the (core-specific) subset 2b mAbs LW7.5–LW7.9 as detecting mAb were prioritized for further tests. We have previously shown a TEA-based buffer to be beneficial for a mycolactone-competitive ELISA, yielding improved assay sensitivity compared with a PBST-based buffer (12). Interestingly, in the current mycolactone capture ELISA, the TEA-based buffer was crucial for mycolactone detection as little or no signals were obtained with a PBST-based buffer. Including DMSO into the buffer (i.e., 0.2 M TEA buffer with 20% DMSO) worked well for most mAb pairs and, in general, mAb pairs recognizing the mycolactone core were more likely to require added DMSO than pairs in which one mAb was binding to the short upper side chain of mycolactone. However, some pairs (notably those comprising mAbs LW7.10 and LW7.11) performed better without the addition of DMSO (Supplemental Fig. 1), leading to the selection of the TEA buffer without DMSO for further analyses involving these mAbs.

The sensitivity of capture assays with selected pairs of mAbs was estimated using synthetic mycolactone A/B. Whereas most tested pairs could detect as little as ∼2 ng of mycolactone in the assay volume of 75 µl, other matched pairs (such as LW7.12–LW7.5) showed a lower sensitivity (Fig. 4). Based on these results, the matched pair comprising LW7.11 as capturing mAb and LW7.5 as detecting mAb was selected for further assays.

In a next step, the LW7.11–LW7.5–based assay was used to detect mycolactone in biological samples without prior lipid extraction. Mycolactone could be detected directly in culture filtrates of African (S1013) and Australian (S1251) M. ulcerans strains, which produce primarily mycolactone A/B or mycolactone C, respectively (Fig. 5).

Given that diagnostic samples collected from BU lesions are serum-rich wound exudates or aspirates and because mycolactone is known to be bound by serum proteins (24), we assessed whether the capture assay is able to recognize mycolactone in the presence of serum proteins. Matrix interference by serum proteins is a common finding during capture ELISA development and may be ascribed to a variety of reasons. Potential causes include the cross-linking of the capturing and detecting mAbs or the binding of the target Ag to serum components, thus preventing its interaction with one or both mAbs (13, 25).

Typical ways of removing matrix interference include purifying the Ag or diluting the sample, both of which help to remove the interferents from the assay. However, diluting the sample may drive the mycolactone concentration below the limit of detection of the assay, and performing lipid extraction for every sample is an added complication. A more straightforward way of dealing with matrix interference is to use suitable chaotropic agents that are capable of breaking the (typically) low-affinity interactions of the interferents (25). We modified our assay buffer by adding different concentrations of MgCl₂, MgSO₄, or ammonium thiocyanate, all of which are commonly used chaotropic agents. Although the addition of MgSO₄ gave better signals than MgCl₂, the former resulted in an unfavorable signal-to-noise ratio when tyramide amplification was done. Therefore, following extensive testing (Fig. 6), we defined a modified version of the assay buffer, comprising 0.2 M TEA and 0.2 M MgCl₂ (pH 7.5), as best suited for the detection of mycolactone in the presence of serum.

Human serum presents an extra challenge to Ag capture assays owing to the presence of anti-animal Abs in the serum. In this study, human anti-mouse Abs (HAMA) led to cross-linking of the mycolactone-specific mAbs used in the assay, and this cross-linking could not be prevented by any of the chaotropic salts at any concentration tested. To this end, a mixture of mouse Abs were instead added to the reaction mix to prevent the cross-linking of the mycolactone-specific mAbs used in the assay. Both the addition of the chaotropic salts and the addition of extraneous mouse Abs to the reaction mix were crucial for preserving the mycolactone recognition in the assay in the presence of human serum. A final concentration of 1 mg/ml of mouse Abs was sufficient to reduce the cross-linking of the mycolactone-specific mAbs caused by HAMA. In addition, tyramide amplification was used to improve the signals obtained when the assay was performed in the presence of serum (Fig. 7).

In a pilot experiment, the utility of the assay in detecting mycolactones present in swabs obtained from IS2404 qPCR-positive BU lesions was shown (Fig. 8). Five of seven qPCR-positive swab samples tested, with cycle threshold values ranging from 15.5–26.5, yielded an absorbance above the 0.2 threshold of the ELISA. qPCR-negative controls tested negative.

The current gold standard assay for BU diagnosis is the highly specific and sensitive IS2404-detecting PCR. Although exquisitely sensitive owing to the high copy number of this IS in the M. ulcerans genome, routine application of this test is hampered by the necessity for sophisticated instrumentation with experienced personnel and rigorous quality control. As such, immunodiagnostic assays have been increasingly considered as practicable surrogates for molecular tests as they have the advantage of being comparatively easier to perform and potentially low cost while still being able to give timely and reliable results (9). Mycolactone makes an ideal target for the specific diagnosis of BU as it is unique to the MPMs, and an immunoassay that can reliably detect mycolactone could enable point-of-care laboratory diagnosis of BU.

We have generated panels of mAbs able to recognize mycolactone via a novel approach using modified synthetic nontoxic mycolactone variants with amine groups permitting the coupling of the polyketide to a carrier protein (11, 12). In this study, we have explored the use of mAbs with different fine specificities in the generation of an Ag capture assay for mycolactone detection.

That a macrolide can be detected in a sandwich ELISA with two full-sized mAbs is not unheard of. Indeed, such an assay was developed for the macrolide drug tacrolimus, which has a similar molecular mass (804 Da) as mycolactone (743 Da). For the generation of antitacrolimus mAbs, the use of truncated structures to constrain recognition to the desired epitopes was applied (26), a similar strategy which we have used for the generation of the antimycolactone mAbs used in this report. Selection of mAbs during hybridoma generation based on their reactivity with different mycolactone derivatives reinforced the establishment of sets of mAbs with diverse fine specificity.

We used two different derivatives of mycolactone to generate the mAbs described in this study, each with modifications to either the upper side chain (PG-180) or the lower hydrophobic side chain (PG-203). In both mycolactone derivatives, the core is present in its native form; therefore, it was unsurprising to see that immunizing mice with either derivative gave rise to mAbs recognizing the lactone core of mycolactone. As expected, mAbs recognizing only PG-183, the biotinylated form of PG-180, were generated only when mice were immunized with PG-180. In contrast, immunizing mice with PG-203 gave rise to mAbs that recognized structures with modified upper chains to varying degrees. Taken together, we could cover a broad range of epitopes on the mycolactone structure by using PG-180 and PG-203 as immunogens.

The Ag capture assay this study described was able to recognize different natural variants of mycolactone, could recognize mycolactones secreted into M. ulcerans culture filtrates without prior lipid extraction, and could recognize mycolactone in the presence of serum. Although we only assessed the recognition of secreted mycolactones A/B and C, we would expect all natural variants of mycolactone to also be recognized in the assay because they all share the same core and upper side chain structures. The prior development of a more efficient ELISA running buffer was crucial to this assay because little or no signals could be obtained in the typically used PBS-based buffers (12). We found that some mAb pairs performed better in the presence of 20% DMSO in the assay buffer, whereas others were indifferent to or impaired by its presence.

With its propensity to form aggregates (which present epitope repeats) in aqueous solutions, we could show that mycolactone could not only be detected by pairs of capturing and detecting mAbs with different fine specificity but also by “same pairs” with the same specificity. In fact, 14 mAbs yielded mycolactone detection signals when they were used both as capturing and as detecting reagents in the same reaction. However, careful selection of optimal capturing and detecting mAbs led to the identification of a small number of combinations, in which capturing and detecting mAbs were not identical.

Following systematic selection, we found that the best-performing pairs contained the subset 3 mAbs LW7.10 or LW7.11 as capturing mAbs and the subset 2b mAbs LW7.5–7.9 as detecting mAbs. Based on their binding patterns to mycolactone derivatives, these capturing mAbs most likely recognize parts of the mycolactone short upper side chain distal from the core. This is evidenced by the abrogation of their binding with increasing modification of this part of the mycolactone molecule. In contrast, the detecting mAbs LW7.5–LW7.9 most likely recognize the core of the mycolactone structure as they could recognize all derivatives containing this substructure to almost identical levels, independent of modifications of the upper side chain. Not surprisingly, therefore, such matched pairs were also able to recognize the derivative PG-120 with truncated lower side chain, potentially a nonaggregated mycolactone. This is an important characteristic, given that the lower side chain is known to be light sensitive because of its extended conjugated double bonds (24, 27). Changes in this hydrophobic side chain could yield structures similar to PG-120; therefore, it was expedient to identify matched pairs that could still recognize mycolactone molecules that have lost the hydrophobic side chain.

Mycolactone has been reported to be bound by serum proteins (24), and this can hamper its recognition by Abs. Human serum, in addition, is rich in anti-animal Abs causing matrix interferences well known in the development of Ag capture assays (28). We could get around these challenges by 1) incorporating a chaotropic salt into the assay system and 2) adding extraneous mouse Abs to suppress cross-linking of the mycolactone-specific mAbs by HAMA. The reduction of signal experienced in the presence of serum could be restored at least partially by including a tyramide amplification step into the protocol. Further optimization could possibly be done by chimerizing the mAbs used as reagents in this assay to remove the chances of mAb cross-linking by HAMA.

The mycolactone capture assay could recognize as little as ∼2 ng of mycolactone. This is comparable to what has been described for other mycolactone-detecting assays, such as 2 ng for fluorescent thin-layer chromatography (29) and 1 ng for the competitive ELISA we recently described (12). The amounts of mycolactone present in fine needle aspirates or tissue biopsies collected from different types of BU lesions have been estimated by liquid chromatography with tandem mass spectrometry. Detected amounts ranged from 0–1970 ng/ml (30). Samples from nodules and plaques on average had slightly higher detectable mycolactone than those from ulcers and edema forms of the disease. Thus, the Ag capture assay described in this study appears suitable for detection of mycolactone in clinical samples. This was confirmed in a pilot experiment by testing a small number of swab samples from BU patients. The swabs tested had been stored for an extended period of time, potentially leading to partial degradation of mycolactone. Nevertheless, the sensitivity of the present assay was sufficient to detect mycolactone in the majority of the qPCR-positive samples. However, for the use as a diagnostic test with serum–protein–containing swab samples and fine needle aspirates from BU lesions, the ELISA conditions will have to be adapted further and validated in comparison with the current diagnostic gold standard for BU, IS2404 PCR testing. Subsequent conversion of the ELISA into a lateral flow format that could be used as a RDT in field settings is envisioned.

The authors have no financial conflicts of interest.