Odorant binding proteins from Hermetia illucens: potential sensing elements for detecting volatile aldehydes involved in early stages of organic decomposition

Because conventional analytical instruments could not be readily used for malodour monitoring in the field, many researchers have resorted to using 'electronic nose' techniques (recently reviewed by Covington et al [16]). Key publications include that of Schiffman and Osuna who reported the use of such a system for evaluating the odours from swine operations [1], Stuetz et al for sewage odour monitoring [17], Tanaka et al for measurement of oral malodour [18], Wilson [19] for hazardous gases, Schnürer et al [20] for detection of fungal volatiles indicative of food spoilage, together with networks of sensing devices for malodour monitoring [21].

Despite many research publications, all of these approaches suffered from lack of sensitivity and selectivity required for the application due to the types of gas sensors utilised (usually metal oxide gas sensors or electrochemical gas sensors). Hence, researchers have turned to utilising and adapting components of the biological olfactory system as nanolayers to create hybrid biomimetic sensing systems, where the recognition element of a gas or odour sensor is an odorant binding protein or a receptor coupled to a transducer (reviewed by Persaud [22]). These have promise in improving conventional 'electronic nose' systems to the level where they can be utilised for routine odour detection and measurements. The frontier of nanotechnology is heading towards the development of increasingly miniaturized devices and the use of biological components as nanolayers opens new possibilities in this field.

1.1.1. Odorant binding proteins

Hermetia illucens (Linnaeus, 1758) (Diptera: Stratiomyidae) is a bioconverter insect commonly known as the Black Soldier Fly (BSF). It is one of the most studied insect species for the bioconversion of organic waste and it is a promising and sustainable source of proteins, lipids, and bioactive compounds (i.e. chitin and antimicrobial peptides) [43–48]. Adult gravid females are attracted by VOCs coming from decomposing organic substrates and some species of bacteria normally present on decaying matter [49]. Once laid nearby organic matter, eggs hatch after 3 days and the newborn larvae, also thanks to their strong mouthparts and specific digestive enzymes, voraciously feed for 15–20 days on a variety of decomposing substrates of both plant and animal origin, including fruits, vegetables, human and animal faeces, manure and carrion, previously selected by the ovipositing females [50–55]. The work presented here develops previous research by Scieuzo et al [56] who investigated the larval and adult transcriptome and reported an in-depth study of potential OBPs involved in odour recognition in H. illucens. This organism has evolved an arsenal of OBPs, with different grades of gene expression, that allow the colonization of the most disparate environments with characteristic smells of decomposition. These OBPs may have specific functions in the behaviour of the insect and the choice of appropriate substrates for deposition of eggs.

The opportunity to incorporate these natural proteins, able to bind volatile products of decomposition, for the development of innovative biosensors mimicking the biological system has been exploited in the present work. Available data on the three-dimensional structures and diameter of the main binding pockets [56] confirmed their use for the development of biosensors at nanoscale, offering significant advantages in terms of enhanced detection sensitivity and specificity. Here, we report the expression, characterization and performance of immobilised H. illucens OBPs (HillOBPs) on quartz crystal microbalance transducers, capable of sensitively detecting chemical species associated with organic decomposition.

Analytical grade reagents for RNA purification, bacterial growth and chemicals were of analytical grade purchased from Sigma-Aldrich, Fisher Scientific or VWR (UK). T4 DNA Ligase and EcoRI/BamHI restriction enzymes were supplied by New England Biolabs (USA). Primers were synthesized by Macrogen Europe (Netherlands), SuperScript^® for RT-PCR was ordered at Invitrogen (USA) and the KOD DNA Polymerase at Merck Millipore (USA). DH5-α and BL21 (DE3) Escherichia coli chemically competent cells were purchased respectively from Invitrogen (USA) and Agilent (UK). For the molecular cloning pCR^™II-TOPO^® vector was ordered at Thermo Fisher, while pET-22b(+) vector was kindly provided by Prof. Paolo Pelosi (Italy). Plasmid mini/midi kit, Gel extraction kit and HiPrep^™ and HisPrep^™ columns were obtained respectively from Invitrogen (USA), Bio-Rad (USA) and GE Healthcare (UK), dialysis membrane tubing was ordered at the Spectra/Por^® (USA). The JLMQ USB interface for 4 QCM sensors (JLM Innovation GmbH, Tuebingen, Germany) was used as a readout system for 20 MHz quartz crystal microbalances with gold coated electrodes fabricated by IMM-CNR (Italy). Odorants, purchased from Sigma-Aldrich (UK), were diluted in recommended solvents (https://pubchem.ncbi.nlm.nih.gov/) for in vitro binding assays by using a quartz cuvette with a light path of 1 cm (Quartz Suprasil 10 mm, Hellma) and freshly used without any dilutions on the day of the experiments in glass vials (Supelco, 40 ml) for biosensor acquisition. The fluorescent probe used to measure the binding activity in solution was the N-phenyl-1-naphthylamine (1-NPN), provided by Sigma-Aldrich (UK).

RNA was isolated from 500 mg of Hermetia illucens larvae using TRI-Reagent, following the manufacturer's protocol and the purified RNA sample was stored at −80 °C until use. In order to remove all the genomic DNA contaminations, total RNA (5 μg) was treated with DNAse I and the full-length cDNA synthesis was performed in a reaction mixture containing dNTP mix (10 mM), oligo (dT)₂₀ primers (50 μM) and the SuperScript III RT (200 U μl⁻¹). The full length of H. illucens OBP (HillOBP) genes was synthesized following conditions established for the reverse transcriptase and amplified by conventional polymerase chain reaction (PCR) (Thermocycler, Applied Biosystems) using specific primers (table S1). Primers related to specific contigs were named wild-type (WT) and six histidine tagged (HisTag) (WT-OBP_C57; HisTag-OBP_C11107; HisTag-OBP_C21691; HisTag-OBP_C1173); the EcoRI and BamHI sites were added to the forward and reverse primers, respectively. Amplicons were generated starting from 1 μl of cDNA template in a 50 μl of solution containing saline buffer (10X), dNTP mixture (2 mM), MgCl₂ (25 mM), forward and reverse primers (10 μM each), KOD^® High Fidelity DNA polymerase (2.5 U μl⁻¹). PCR conditions were used according to the polymerase protocol and each amplicon was electrophoresed on 1.2% (w/v) agarose gel, stained with UltraPure^™ Ethidium Bromide (0.5 μg ml⁻¹) (Thermo Fisher). The PCR product was first cloned in pCR^™II-TOPO^® following the manufacturer's protocol. Then DNA from positive colonies was digested with EcoRI and BamHI and then ligated with T4 DNA ligase into the EcoRI/BamHI-treated pET-22b(+) vector (30 ng), following the manufacturer's protocol. The entire volume of ligation reaction was used to transform E. coli DH5-α cells with heat shock method (30 min in ice—45 s at 42 °C–5 min in ice) and clones were propagated growing in a Luria–Bertani (LB) broth at 37 °C. The clone selection was carried out in plates with LB agar to which was added 50 μg ml⁻¹ ampicillin (Amp^R ). Screening of E. coli clones was carried out using the Colony-PCR technique and the final recombinant plasmids were isolated using the FastPlasmid^™ Mini kit (5 PRIME). DNA fragments were sequenced and analysed with Clustal Omega (https://www.ebi.ac.uk/Tools/msa/clustalo/) setting the parameters by default, in order to verify the sequence fidelity.

The steps involved in the production of OBPs are outlined in figure 1.

Zoom In Zoom Out Reset image size

The heterologous expression of recombinant HillOBPs was carried out with specific plasmids transformed in E. coli BL21 (DE3) pLys and inoculated in 1 l of LB medium supplemented with 50 μg ml⁻¹ ampicillin (Amp^R ). Isopropyl-ß-D-thiogalactoside (IPTG) (0.4 mM) was added when the optical density (O.D._{600 nm}) value reached 0.6 at 37 °C, shaking at 225 rpm for 2 h. After expression, cells were harvested by centrifugation at 4000 rpm (4 °C) for 20 min and resuspended in equilibration buffer (50 mM Tris-HCl, pH 7.4), with addition of phenylmethylsulfonyl fluoride (PMSF) (1 mM). The cell suspension was passed through a sonicator (Bandelin Sonopuls) to give a crude cell-free supernatant enriched with soluble proteins, which was clarified by centrifugation at 4000 rpm (4 °C) for 2 h. The HillOBPs extraction was optimized treating the pellet with lysis buffer (50 mM Tris-HCl, 0.5 M NaCl, 8.0 M Urea, pH 7.4) added with dithiothreitol (DTT) (1 mM), and stirring for 1 h at room temperature, in order to release insoluble proteins from the inclusion bodies. A centrifugation at 4000 rpm (4 °C) for 20 min to remove insoluble cellular debris was followed by four cycles of dialysis at 4 °C for 16 h in Spectra/Por^® dialysis membrane tubing (MWCO, 12–14 kDa) against equilibration buffer to dilute salts and favour the protein refolding. Ion exchange (IEX) and affinity chromatography were, respectively, the two techniques performed for the purification of the wild-type HillOBP_C57 and 6-His-tagged HillOBPs (C-ter: HillOBP_C11107, HillOBP_C21691; N-ter: HillOBP_C1173), according to the manufacturer's instructions. For the purification of HillOBP_C57, the crude extract was loaded to a HiPrep^™ Q Sepharose Fast Flow 16/10 column, pre-equilibrated with equilibration buffer (50 mM Tris-HCl, pH 7.4) and the bound HillOBP_C57 was eluted with elution buffer (50 mM Tris-HCl, 0.5 M NaCl, pH 7.4). The HisPrep^™ Fast Flow 16/10 column, pre-equilibrated with equilibration buffer (50 mM Tris-HCl, 500 mM NaCl, 35 mM Imidazole, pH 7.4) was used to purify the remaining 6His-tagged HillOBPs. The crude extracts, to which imidazole (35 mM) was previously added, were loaded into the column and, thereafter, the bound HillOBPs were separately eluted with the same buffer, with addition of imidazole (500 mM). The eluates were analysed in 12.5% (w/v) SDS-PAGE gels and stained with 0.1% (w/v) Coomassie brilliant blue (R-250) to confirm the presence of HillOBPs free of contaminants. The molecular weights, predicted with ProtParam—ExPASy tool, were confirmed using Novex^™ Sharp pre-stained protein marker and the concentration was estimated using bovine serum albumin (BSA) as a standard [57], and the Beer–Lambert's law [58].

A delipidation treatment was used to decrease non-specific binding from any pockets of the recombinant HillOBPs produced in E. coli cells, as trapped endogenous lipids could affect the active sites of proteins. The fatty acid residues were easily removed, without denaturing or permanent loss of HillOBPs binding activity, by using 1.0 g of pre-swollen Sephadex LH-20 in 50 mM sodium acetate buffer pH 4.5, was shaken for 3 h at room temperature. The slurry was spin washed twice with the same buffer and then equilibrated with 25% (v/w) of the same buffer. The protein samples pH was adjusted to 4.5 with acetic acid, mixed with the slurry (5:1) and stirred at 4 °C for 1 h. The supernatant, cleaned from fatty acids previously trapped in the slurry, was collected following two centrifugation steps at 15 000 g (4 °C) for 5 min. Finally, the supernatant containing protein was dialysed overnight against 50 mM Tris-HCl pH 7.4, centrifuged and filtered with a 0.8 μm filter. The diluted protein fractions were transferred to Amicon^® Ultra-15 Centrifugal Filter Devices with a 10 kDa molecular weight cut-off regenerated cellulose membrane (Merck Millipore) and centrifuged at 4000 rpm (4 °C), for a protocol time adapted to obtain a volume equal to 1/10 compared to the initial volume and, as consequence, higher concentration values per milliliter.

The AT-cut quartz crystal was sandwiched between two gold electrodes, with a diameter of 7.95 mm for the crystal and 4.9 mm for the geometric surface of the gold layer, to build a quartz crystal microbalance (QCM) (figure 2). The golden electrodes were sputtered on the quartz crystal on an adhesion layer of titanium. The QCM was driven by a Pierce oscillator circuit that created a oscillating electric field of 2.4 V amplitude which induced an acoustic wave. The thickness of QCM determined a base resonant frequency of approximately 20 MHz. In the AT-cut crystal, the temperature dependence of the resonant frequency is known to be negligible in the range of 0 °C–40 °C. The QCM gold surface was cleaned by dipping the crystal into a strong oxidizer Piranha solution (1:3, 30% H₂O₂: H₂SO₄) for 4 min, to remove any organic contaminants from the gold surface and allow the deposition of the self-assembled monolayer. QCMs were then thoroughly rinsed with distilled water, absolute ethanol and dried at air, before starting the protein immobilisation process.

Zoom In Zoom Out Reset image size

Prior to the deposition of HillOBPs, the gold electrode surfaces were chemically modified with thioctic acid (TA), an alkanethiol used as self-assembled monolayer (SAM). The assembly of the lipid layer was favoured by immersion of QCMs into a solution of 10 mM TA in ethanol for 20 h under nitrogen flow (20%), to allow a higher efficiency of TA binding on the gold surface. To remove any unbound molecules of TA, the electrodes were then rinsed twice with absolute ethanol and left to dry at room temperature. TA free carboxylic groups were activated through a reactive intermediate, pipetting alternately on the both sides of gold surface 20 μl of a solution, composed of 180 mM ethyl-(dimethylaminopropyl)-carbodiimide (EDC) and 180 mM of N-hydroxy-succinimide (NHS) in 10 mM sodium phosphate buffer (pH 7.0). The SAM activation process lasted for 2 h at room temperature and was then interrupted with two washes in distilled water. QCMs were dried in air, 10 μl of protein solution (0.3–1.5 mg ml⁻¹) was pipetted over the gold surface of each electrode and left for 1 h at room temperature, to create the peptide bonds between the amino groups in R side chains of HillOBPs and the activated carboxylic groups of SAM. Each deposition process was followed by rinsing with distilled water and drying at air, three times for each QCM side. The QCM gold area was coated with the peptide monolayer, fixed to the SAM by a covalent bond; the thick film of the unbound fluid was removed manually after the deposition, avoiding contact with the outer edge of each crystal.

The acquisitions were carried out under strictly controlled conditions in terms of resonance frequency of the microbalances, in order to obtain a stable baseline. The surface coverage of QCM resonator with HillOBPs as sensing layer and the adsorption of analyte molecules to characterize the binding interactions with this biochemical recognition film were monitored by variations in the resonance frequency expressed by the Sauerbrey equation [59], where, ρ_q and μ_q are the density (2.648 g cm⁻³) and shear modulus (2.947 × 10¹¹ g cm⁻¹ s²) of the quartz, f₀ is the fundamental frequency of the piezoelectric quartz crystal (19453 ± 50 KHz) and A is the crystal piezoelectrically active geometrical surface (0.496 cm²), that is the area of the deposited film on the gold electrode. According to this equation, there is a decrease in the system frequency proportional to the mass changes (Δm), as a consequence of the material deposition on the crystal surface. The frequency shift (Δf) was used to calculate the number of HillOBP molecules immobilised on each QCM and monitor the VOC interactions. To provide evidence that the response signals obtained from the biosensor are specific, a series of control measurements have been included. Firstly, QCM devices fabricated and functionalized with the SAM, but without immobilised HillOBPs, were tested. The response of the biosensor was also tested with 2-phenylethanol, a compound recognized as a ligand with poor binding affinity to insect OBPs, on the basis of experiments performed in Cali and Persaud [42]. The resulting HillOBP-based sensors were then exposed for 20 s to pulses of vapours from target VOCs indicative of organic decomposition. QCM measurements with JLMQ USB interface with MultiSens software (JLM Innovations, Germany) were performed at room temperature with a relative humidity (RH) of 22% and, between each acquisition, the baseline was established with clean air flowing at the rate of 20 ml min⁻¹.

Fluorescent competitive binding assays were employed for a functional characterization of HillOBP_C57, previously identified as the OBP expressed at the highest level in H. illucens [56] and the most powerful target in the nanobiosensor assembled in the present work. All the in vitro interactions between HillOBP_C57 and VOCs of interest were monitored using the fluorescent probe N-phenyl-1-naphthylamine (1-NPN) and measuring the fluorescence variations produced following the competition between 1-NPN and the ligand of interest for the same binding pocket. To establish the probe dissociation constant (K_D), 1-NPN was added in different concentrations (0–16 μM) to the Tris-HCl buffer (50 mM, pH 7.4) containing HillOBP_C57 (2 μM) and excited at the specific wavelength of 337 nm, with an emission spectrum recorded between 380 and 450 nm. Fluorescence signals were measured with each addition of 1-NPN and, as result, the intensity increased several fold with a shift of the emission wavelength to a maximum of 407 nm. The K_D of ligands was measured using Tris-HCl buffer (50 mM, pH 7.4) containing HillOBP_C57 (2 μM) and 1-NPN (5 μM), titrated with solutions of each ligand to final concentrations of 0–16 μM. The fluorescence measurements were performed in three replicates on a Perkin Elmer LS-55 Luminescence Spectrometer at 25 °C in a right-angle configuration, with 5 nm of slit widths on the excitation and emission monochromators. Emission fluorescence was monitored with a scan speed of 300 nm min⁻¹ and spectra were recorded by FL WinLab software. The K_D for the HillOBP_C57/1-NPN complex was calculated with SigmaPlot^™ 12 (Systat Software, Inc.) (www.sigmaplot.com) and the equation K_D = IC₅₀/1+([1-NPN]/K_D,1-NPN) was used to calculate the dissociation constant of HillOBP_C57 for each tested ligand, where IC₅₀ is the probe concentration required to occupy 50% of pockets, halving the maximum fluorescence intensity.

Nine independent replicates were obtained starting from three data acquisitions for each VOC, tested on three different QCMs, assembled in series with the same HillOBP stock. Raw sensing data relating to the interactions with VOCs were normalised on the basis of the number of calculated immobilised HillOBPs on the sensor surface. The related frequency curves and the different VOC concentrations allowed to study the kinetics of adsorption and desorption, in order to extrapolate useful information related to dissociation constants, affinity, sensitivity and limit of detection (LOD) of the HillOBP-VOC complex. Inspection of multidimensional data sets from an array of different OBPs responding to individual VOCs was performed using Principal Component Analysis (PCA) [60–62]. PCA is a general mathematical method of extracting significant information from a large complex data set, by projecting data in a global model in this case of protein-ligand space. As an orthogonal regression in the space, PCA defines the structure of variance–covariance of a data set through a system of coordinates whose number of dimensions is less than the number of original variables.

A neural network based on radial basis functions (RBF) was used to test discrimination between malodour types. A confusion matrix shows the overlap and discrimination power of datasets measuring the separability of classes and predict the binding propensities of each VOC towards a given HillOBP. Values were in the range of 0 and 1, values lower than 0.5 showed a rather bad separability, while high values, near 1, indicated a good separability of classes.

Based on the de novo transcriptome analysis [56], four different HillOBPs (HillOBP_C57, HillOBP_C11107, HillOBP_C21691 and HillOBP_C1173) were expressed as recombinant proteins in a heterologous system. To determine the influence of 6His-tag on the protein structure, HillOBP_C57 was cloned without tag, HillOBP_C11107 and HillOBP_C21691 with tag at C-terminal and HillOBP_C1173 at N-terminal. Each nucleotide sequence was subcloned in pCR^™II-TOPO^® and, after an EcoRI/BamHI restriction digestion, it was fused to the 3'-end of the expression vector pET-22b(+) and the resulting recombinant plasmid was propagated in DH5-α and then expressed in BL21(DE3) E. coli competent cells (figure 1). Recombinant HillOBPs were detected in purified fractions and all the observed molecular weights were compatible with the predicted size of the mature full-length recombinant proteins (table S2). In the case of 6His-tagged HillOBPs, the best purification yields were obtained by coupling affinity chromatography with anion exchange chromatography, because the presence of tag was not enough to improve the purification yields, as the protein was folded into a hydrophobic conformation. Once delipidated and concentrated, HillOBPs were immobilized and used as biorecognition elements on QCMs, to test the ability to detect volatile organic compounds (VOCs) of interest associated with organic decomposition.

The HillOBPs immobilization, through a covalent binding between the alkanethiols of the self-assembled monolayer (SAM) and the free amino groups of the protein side chains, resulted in an easy, fast and inexpensive technique for the QCM-based biosensors technology (figure 2). A Pierce oscillator circuit allowed the quartz crystal to oscillate with its own resonance frequency (∼20 MHz), according to the phenomenon of piezoelectricity and, considering the molecular weight, the HillOBP surface densities on the QCM devices were calculated. The experimental measurements of biomolecular interactions between OBPs and VOCs are based on the one-to-one binding mechanism, arised from properties of the molecules involved. On that basis, it was possible to estimate the VOCs loading per unit area (cm²) (table S3).

Competitive binding assays were used to investigate in vitro the HillOBP_C57 binding affinities, exploiting a competition between free ligands and the 1-NPN fluorescent reporter for the same main binding pockets. When excited at 337 nm in presence of HillOBP_C57, 1-NPN showed a strong emission spectrum shift from 450 nm to approximately 380 nm, as well as a drastic increase in fluorescence intensity, suggesting the 1-NPN suitability for investigating in general OBP binding properties in fluorescent competitive assays. Titration with the best HillOBP_C57 concentration (2 μM), chosen on the basis of background noise-free signal, and increasing concentrations of 1-NPN (0–16 μM), provided a dissociation constant (K_D) equal to 3.27 ± 0.54 μM (figure 3(a)). Subsequently, the fluorescence quenching by a competitive ligand was measured by adding aliquots (0–16 μM) of VOCs to the protein solution (HillOBP_C57, 2 μM) pre-equilibrated with 5 μM of 1-NPN probe (figure 3(b)). All titrations showed a great reduction in fluorescence intensity and, as consequence the highest binding affinities, in presence of three specific compounds (isobutyraldehyde, isovaleraldehyde and 2-methylbutyraldehyde), belonging to the same chemical class (aldehydes) (figure 3(c)). The low values of the binding affinities recorded for the remaining compounds are not reliable since the nature of the VOCs, strongly hydrophobic, could alter the result following the micelles formation, particularly in presence of fatty acids [63, 64].

Zoom In Zoom Out Reset image size

When a QCM is loaded with a thin rigid film, which is purely elastic, the resonance frequency decreases with increased mass. If the film is much thinner than the thickness of the resonator, the decrease in resonance frequency (f) is linearly correlated to the increase in mass (m) (or thickness) according to the Sauerbrey relation:

$$\begin{eqnarray}&&{\rm{\Delta }}{m}=-{C}\displaystyle \frac{{\rm{\Delta }}{f}}{{n}},\end{eqnarray} \tag{ 1 }$$

where C is a constant dependent on the material and n is the overtone number which is 1 in this case indicating the fundamental frequency. However, if a viscous component is introduced as well, such as a protein, this gives a viscoelastic or soft film and there will be energy losses in the system, quantified by an energy dissipation factor, D. Depending on the conformation of the protein and how it may change in shape when it binds a ligand, the resonance frequency, f, and energy dissipation factor of the QCM can enter a nonlinear regime where propagation of the shear wave, and the softness of the film, will have a dampening effect on the oscillating motion quantified by D.

The biosensor response was investigated with 42 VOCs indicative of organic decomposition [56], by monitoring the frequency changes when the HillOBP-coated QCMs were exposed to VOC vapours at a fixed concentration of 4 ppm. During the exposure phase, the HillOBPs on QCMs act as recognition layers for VOC molecules and the complex formation causes a mass increase in real time that leads to a proportional reduction in the resonance frequency of the microbalances. The system showed a fast and reversible response within 20 s starting from the exposure, following the introduction of a specific analyte in vapour-phase by pumping. The baseline value was restored during the regeneration phase, when the adsorbed VOC molecules were released by switching to clean air and the frequency of the sensors rose as the mass decreased. The complete desorption of VOC vapours, achieved by injecting fresh air in the sensing chamber until the frequency returned to baseline, was indicative of good reversibility and recovery of the device. In presence of different VOCs, during a constant time of acquisition, HillOBP_C57, HillOBP_C11107 and HillOBP_C1173 sensors had a much stronger response with a negative frequency shift only in presence of isobutyraldehyde, isovaleraldehyde, 2-methylbutyraldehyde and butyric acid. The same VOCs evoked an opposite response from the HillOBP_C21691 sensor with an increase in frequency, interpretable as change in viscoelastic properties of the OBP layer. Computational analysis of this OBP indicated it was devoid of pockets involved in ligand-binding for these ligands, and the interaction with the aldehydes and fatty acid may be entirely due to surface adsorption. The frequency responses of HillOBP_C57, HillOBP_C11107 and HillOBP_C1173 sensors decreased immediately after analyte injection and reached an equilibrium state approximately after 5–10 s, indicating a faster response time only for these specific ligands (figure 4). Data relating to VOC interactions have been normalized by multiplying the frequency variation of each analyte by the number of immobilized HillOBP molecules (table S3). Sensors respond to analytes irrespective of different ambient humidity levels. Figure S1 (available online at stacks.iop.org/NANO/33/205501/mmedia) shows the QCM resonance frequencies at 22%RH, and the changes in frequencies for two immobilized proteins (HillOBP_C57, HillOBP_C11107) without VOCs (blue line, Humidity) and with analytes at 22%RH. After normalizing the data vectors across the array of QCM sensors, principal component analysis (PCA) was used to reduce the multidimensional space (i.e. responses from four independent sensors) towards the 42 VOCs, without losing information of the technical replicates. Of practical interest is not the discrimination between individual chemical species but between compounds related to spoilage origin (in this case either microbial or chemical decomposition).

Zoom In Zoom Out Reset image size

The HillOBP-sensor responses were presented on a PCA score plot, dividing the analysed VOCs in two main categories, related to the spoilage origin (microbial and chemical), based on the most common causes of organic degradation, such as the presence of contaminants, physical/chemical agents and animal activity (table S4). The dataset was projected onto the three-dimensional space spanned by the vectors (principal components) that correspond to the maximum variance of the dataset. The PCA score plot indicated that the device is generally able to discriminate between microbial and chemical spoilage, however with some overlap of signals between microbial and chemical spoilage, due to the fact that all the produced VOCs following chemical reactions are themselves generated by metabolic reactions triggered by microorganisms (figure 5).

Zoom In Zoom Out Reset image size

To test discrimination, a radial basis function neural network was trained against approx 70% of the data set shown in figure 5 and tested against the remaining data to determine whether the microbial and chemical classes could be robustly discriminated. This gave a true positive rate of 0.88 and a true negative rate of 1.0 in discriminating these classes with a 94% average system accuracy due to the overlap between the two classes.

Reaction kinetic studies performed with this QCM-based biosensor prototype allowed calculation of theoretical affinity values based on the kinetics of the on/off response to a pulse of vapour. Figure 6 shows the modelled affinities to a range of compounds for the best and worst performing sensors; the HillOBP_C57 sensor gave the highest affinity values towards isobutyraldehyde, 2-methylbutyraldehyde, isovaleraldehyde, butyric acid, while the lowest affinity values were observed from the HillOBPs_C21691 sensor towards many compounds belonging to different chemical classes. On the other hand, all the HillOBP-coated QCM exhibited a very slow frequency change in the detection of VOC vapours belonging to other chemical classes (terpenoids, ketones, esters, aromatics) (figure 6). These data are in agreement with the heat maps related to the OBPs gene expression level in H. illucens shown in Scieuzo et al [56], where the OBP_C57 was found to be the most expressed protein, with 100% identity between larvae and adults. Experimentally, another finding was obtained in vitro, through the binding assays in solution (figure 3): HillOBP_C57 showed the lowest levels of dissociation constants and, consequently, the highest levels of affinity in presence of the same VOC categories.

Zoom In Zoom Out Reset image size

At the molecular level, biomolecular interactions were calculated starting from the equilibrium binding constant (K_eq) and the dynamic interactions between immobilized HillOBPs and VOCs assuming the following equilibrium:

$$\begin{eqnarray}&&{A}+B\,\iff {AB},\end{eqnarray} \tag{ 2 }$$

$$\begin{eqnarray}&&{{K}}_{{\rm{eq}}}=\displaystyle \frac{\left[{AB}\right]}{\left[{A}\right]\left[{B}\right]}=\displaystyle \frac{{{k}}_{{\rm{on}}}}{{{k}}_{{\rm{off}}}},\end{eqnarray} \tag{ 3 }$$

where [AB], [A] and [B] are equilibrium concentrations of the complex HillOBP/VOC, free HillOBP and free VOC respectively; k_on and k_off are time constants relative to the association/dissociation of the complex. The affinity and the strength of binding between HillOBPs and VOCs were calculated as the ratio k_off/k_on (k = Δf / time_90%), before the saturation was reached, as shown in figure 7. The sensor responses (Δf) were calculated as the differences between the corresponding HillOBPs-coated QCM frequencies at the beginning and the end of the exposure phase of 4.0 ppm VOC vapour at room temperature. When the HillOBPs were employed in the VOC ligand binding, the QCM response was a sudden decrease in frequency to the steady state level. The frequency returned to the original value, as a consequence of the fast and reversible binding between VOCs and HillOBPs, within a time related to the binding affinity and difference in sensing responsivity was based on the different affinities of HillOBPs with different chemical structures of VOCs belonging to the same chemical classes as shown in figure 6.

Zoom In Zoom Out Reset image size

The changes in frequency of the QCMs were proportional to the concentration of VOCs to which they were exposed.

Isobutyraldehyde, 2-methylbutyraldehyde and isovaleraldehyde concentrations were initially set at 6.4 ppm and subsequently increased to 204.8 ppm, considering the saturation limit (table S5). As expected, the dynamic responses of HillOBP_C57 sensor under exposure to increasing concentrations showed a good correlation with loading (frequency shifts) and a clear dependency between concentration and response. A Freundlich isotherm could be fitted to the concentration-response curves reflecting the adsorption/desorption of analytes on the QCMs (figure 8). Consequently, it was possible to calculate the HillOBP_C57 sensor sensitivity (S, Hz ppm⁻¹) defined as the initial slope of the curves shown in figure 8. HillOBP_C57 exhibited the highest sensitivity in presence of Isobutyraldehyde (5.45 Hz ppm⁻¹) with a correlation coefficient (R²) of 0.98. Several reports indicated that the sensitivity of the QCM method depends largely on the molecular weight of the analyte [65, 66]. In this case, as 2-methylbutyraldehyde and isovaleraldehyde have the same molecular weight, the sensitivity is approximately equal. The correlation between frequency shift (Δf) plotted against VOC concentrations made it possible to determine, furthermore, the limit of detection (LOD) of the most interesting VOCs belonging to the aldehydes chemical class, for the HillOBP_C57-coated sensor (table 1).

Zoom In Zoom Out Reset image size