The genome of the African elephant Loxodonta africana contains genes encoding several lipocalins, which were downloaded from the NCBI and the UniProt databases. In order to identify those sequences that could be reliably involved in chemical communication, we built a phylogenetic tree including all the lipocalins from the African elephant and selected species of mammals: the pig, the cow, the horse and the rabbit (Figure 1). The phylogenetic tree was built using the software FigTree v1.4.4 with the sequences obtained from a BLAST search of the NCBI database, using the elephant lipocalins as queries. We can observe, as already reported in a more general analysis involving 25 mammalian species (Pelosi and Knoll, 2022), that different lipocalins segregate into well separated clades. The large cluster containing the cow and pig OBPs, which have been the first to be isolated from biological tissues and characterised, also include sequences classified as MUPs/SALs and lipocalin 9 (LCN9). Another cluster also includes some sequences annotated as OBPs, together with members of the VEG protein group. Three sequences from these two functionally related clusters are from the African elephant (OBP1, OBP2 and VEG) and are aligned in Figure 2.

While the roles of OBPs and MUPs/SALs have been well established as semiochemical carriers, for VEGs and LCN9 there are still aspects to be clarified. In particular VEG proteins were reported to be endowed with different functions (Glasgow, 2021), such as carriers for taste compounds (Schmale et al., 1990), scavengers of toxic chemicals (Redl, 2000) and transporters of vitamins and nutrients (Gasymov et al., 2002). It is possible that different VEG proteins may be involved in different functions, including chemosensing, depending on the animal species. Regarding LCN9 members, no hypotheses can be proposed, based on the scanty information available for these proteins, although their clustering with OBPs makes their role worthy of investigation (Pelosi and Knoll, 2022).

As LafrOBP1 has been described in our previous work, here we turned our attention to LafrOBP2. Its sequence shares only 15% of its residues with LafrOBP1, while it appears to be more similar to LafrVEG with 31% identity. LafrOBP2 was not found in the trunk wash (Zaremska et al., 2022), unlike the two other proteins, but its presence in other body fluids cannot be excluded. Further insight into the physiological role of LafrOBP2 might come from the identification of the glands and organs where this protein is produced and likely performs its functions. As a first contribution to unveil the biological role of this protein we decided to apply a “reverse chemical ecology” approach. The strategy consists in evaluating the affinity of a protein towards a number of pure chemicals with the aim of identifying the best fitting compounds and therefore suggesting which could likely be its physiological ligands (Leal, 2017; Zhu et al., 2017; Choo et al., 2018; Zaremska et al., 2021).

The expression of LafrOBP2 in bacteria produced the protein in large amounts (about 50 mg/l of culture) but mostly present in the pellet, after sonication, as inclusion bodies. However, the protein was successfully solubilised by denaturation in urea/DTT and refolded by extensive dialysis. Purification was performed by anion-exchange chromatography on a Hi-Prep Q column. Supplementary Figure S1 reports the electrophoretic analysis of samples of the expression and purification steps.

Before using the purified protein in ligand-binding experiments, we wanted to verify whether LafrOBP2 was present in solution as a monomer or a dimer. In fact, this protein contains three cysteine residues at positions 62, 104 and 154 of the mature sequence (Figure 3A). While Cys62 and Cys154 are very likely connected by a disulphide bridge, as predicted by a three-dimensional model, Cys104 might be available for an intermolecular disulphide bridge connecting two units of the protein. To verify such a possibility, we analysed samples of the purified protein on SDS-PAGE in denaturing and non-denaturing conditions. The results clearly show that LafrOBP2 is present almost entirely as a monomer (Figure 3B). In fact, the bands migrating with apparent molecular size of about 40 kDa may indicate the presence of a dimeric form, which however accounts for only a small fraction of the protein. On the other hand, the migration of the monomeric protein as a doublet may be explained with the possibility that some mercaptoethanol could have diffused from the second well and reduced part of the protein. A closer look at the model also shows that dimerization of the protein involving Cys104 seems rather unlikely. In fact, this residue, although located on the surface of the protein, is not protruding enough to allow interaction with another molecule of the protein. Such interaction is further hindered by repulsion forces generated by the presence of negatively charged residues (Asp83, Asp107) in the proximity of Cys104 (Figure 3A).

The recombinant LafrOBP2 binds the fluorescent probe N-phenyl-1-naphthylamine (1-NPN) with a dissociation constant of 3.6 μM (Figure 4A), thus enabling us to test the affinity of other ligands in a competitive binding approach. Our first choice of potential ligands included the two known pheromones for the Asian elephant, (Z)-7-dodecenyl acetate and frontalin, and three other insect pheromones identified in elephants: exo-brevicomin, (E)-β-farnesene and (2E,6E)-farnesol. None of these chemicals showed any appreciable affinity to the protein (Supplementary Figure S2A). In particular, the peculiar behaviour of (Z)-7-dodecenyl acetate, whose curve increases with the concentration of the ligand, also observed with bombykol (Supplementary Figure S2D) has been explained with the ability of these compounds to form micelles which entrap molecules of the fluorescent probe, thus increasing the emission intensity (Sun et al., 2012; Leal and Leal, 2015; D'Onofrio et al., 2020).

Therefore, we adopted a more general approach and tested a selection of chemicals different in size, functional groups and structure. All the tested compounds are listed in Table 1 together with their binding parameters. Our list also included some classes of insect pheromones, such as homologous series of long-chain linear alcohols, aldehydes, esters and fatty acids, as well as γ-lactones and macrocyclic ketones. In fact, the adoption of similar or even identical chemical structures as pheromones by insects and mammals is well documented, in particular in elephants (Schulte and LaDue, 2021; Voznessenskaya et al., 2022).

LafrOBP2 proved to be tuned to linear unsaturated C16-aldehydes with dissociation constants between 2.4 and 3.4 μM. While several of the tested compounds did not exhibit any appreciable affinity for the protein, others were found to be weak ligands with dissociation constants in the order of 10–20 μM, as roughly evaluated after extrapolating the displacement binding curves. The only exceptions are linoleic and linolenic acids which proved to be good ligands with dissociation constants of 4.0 and 5.4 μM, respectively. Figure 4B reports the actual displacement curves recorded with the five best aldehydes, while their dissociation constants are graphically compared in Figure 4C. The displacement curves of representative linear alcohols, fatty acids and esters are reported in Supplementary Figures S2B–D. We summarise our binding data as follows:

All chemicals were purchased from Merck, Austria, unless indicated otherwise. Restriction enzymes and kits for DNA extraction and purification were from New England Biolabs, Germany. Oligonucleotides and the synthetic gene encoding LafrOBP2 were custom synthesised at Eurofins Genomics, Germany.

The African elephant’s lipocalin protein sequences were obtained from the NCBI and the UniProt databases. These sequences were used in a BLAST search against the NCBI protein database limited to the following species: pig (Sus scrofa), cow (Bos taurus), horse (Equus caballus) and rabbit (Oryctolagus cuniculus). A phylogenetic tree was built with the software FigTree v1.4.4, using default settings.

The synthetic gene encoding LafrOBP2 was digested with restriction enzymes NdeI and EcoRI at 5′- and 3′-ends, respectively, and ligated (using T4 DNA ligase, New England Biolabs, Germany) into pET30a (Novagene, Darmstadt, Germany) previously digested with the same enzymes. The only modification of the expressed protein was an additional starting methionine deriving from the starting codon ATG introduced by the Nde I restriction enzyme. Competent E. coli DH-5α cells were transformed with this construct. Positive colonies containing the plasmid with the gene were selected by PCR using T7 (5′-TAA TAC GAC TCA CTA TAG GG-3′) as forward primer and the specific reverse primer 5′-TTA CTC CGT TTC CGG AGT-3′, and used to transform competent E. coli BL-21 (DE3) cells. At OD₆₀₀ of 0.8, protein expression was induced with 0.4 mM IPTG and the culture was incubated for two more hours. Cells were lysed by sonication yielding the protein in the pellet, which was then dissolved in 8 M urea/5 mM DTT. Solubilised protein was refolded by extensive dialysis against 50 mM Tris–HCl, pH 7.4 and finally purified by anion-exchange chromatography on HiPrep-Q 16/10 20 ml (Bio-Rad, United States), using a linear gradient of 0–1 M sodium chloride in 50 mM Tris–HCl, pH 7.4 (Liu et al., 2014; Brulé et al., 2020).

Affinity of the fluorescent probe N-phenyl-1-naphthylamine (1-NPN) to LafrOBP2 was measured by titrating a 2 μM solution of the protein in 50 mM Tris–HCl, 1 M NaCl, pH 7.4 with 1 mM 1-NPN aliquots in methanol to final concentrations of 2 to16 μM. Dissociation constants of other ligands were determined by displacing the fluorescent reporter 1-NPN from the complex. A mixture of protein and 1-NPN, both at the concentration of 2 μM in 50 mM Tris–HCl, 1 M NaCl, pH 7.4, was titrated with 1 mM methanol solutions of each ligand to final concentrations of 2 to 16 μM. Spectra were recorded on a PerkinElmer FL 6500 spectrofluorometer in a right-angle configuration, at room temperature, with slits of 5 nm for both excitation and emission, using 1 cm path quartz cuvettes. The excitation wavelength was 337 nm and intensity values were recorded in correspondence with the peak maximum, around 420 nm. Data for 1-NPN were analysed with GraphPad Prism version 8.0.0, GraphPad Software, San Diego, California United States, www.graphpad.com. Dissociation constants of other ligands were calculated from the corresponding [IC]₅₀ values (the concentration of each ligand halving the initial value of fluorescence), using the equation: K_D = [IC]₅₀ /1 + [1-NPN]/K_NPN, where [1-NPN] is the concentration of free 1-NPN and K_NPN the dissociation constant of the complex OBP/1-NPN.

The three-dimensional model of LafrOBP2 was generated with the on-line software Swiss Model (Bertoni et al., 2017; Waterhouse et al., 2018; Studer et al., 2021) using the human OBP2a crystal structure (PDB: 4run.2.A; identity: 50%) as a template. Figures were created with Chimera software (Pettersen et al., 2004).