Structural basis of sensory receptor evolution in octopus

Abstract

Chemotactile receptors (CRs) are a cephalopod-specific innovation that allow octopuses to explore the seafloor via 'taste by touch'¹. CRs diverged from nicotinic acetylcholine receptors to mediate contact-dependent chemosensation of insoluble molecules that do not readily diffuse in marine environments. Here we exploit octopus CRs to probe the structural basis of sensory receptor evolution. We present the cryo-electron microscopy structure of an octopus CR and compare it with nicotinic receptors to determine features that enable environmental sensation versus neurotransmission. Evolutionary, structural and biophysical analyses show that the channel architecture involved in cation permeation and signal transduction is conserved. By contrast, the orthosteric ligand-binding site is subject to diversifying selection, thereby mediating the detection of new molecules. Serendipitous findings in the cryo-electron microscopy structure reveal that the octopus CR ligand-binding pocket is exceptionally hydrophobic, enabling sensation of greasy compounds versus the small polar molecules detected by canonical neurotransmitter receptors. These discoveries provide a structural framework for understanding connections between evolutionary adaptations at the atomic level and the emergence of new organismal behaviour.

Extended Data Fig. 1 |. Octopus chemotactile receptors exhibit high rates of molecular evolution.

a, Maximum likelihood treeof nicotinic receptor-like genes From Octopus bimaculoides showing the number of introns per gene. CRT1 is high lighted in the phylogeny.b, CRs are intronless genes arranged in tandem arrays in a single chromosome. The two main groups of CRs in octopus map to two distinct tight clusters in Chromosome 15. c, Alignment of the two arrays of intronless CRs in Chromosome 15 suggest diFFerences in overall architecture between dusters. Only genes with complete coding sequences (CDS) are mapped. d, Phytogeny of acetylcholine receptor-like genes of Octopus bimaculoides depicting variation in dN/dS values for tip branches inferred underthefree-ratio model and mapped to the phylogeny. e, Phylogeny from Fig. 1b depicted as ultrametric and with bootstrap values.

Extended Data Figure 2 |. Antibody control and biochemistry.

a, Anti-CRT1 is specific to expressed CRT1 as it did not label another CR (CR840) expressed in HEK293 cells. Representative of 3 independent transfections. Scale bar = 30 μm. b, Fluorescence-detection size-exclusion chromatography (FSEC) trace of CRT1 pentamer and SDS-page gel of final EM sample, representative of 4 independent purifications.

Extended Data Fig. 3 |. EM data processing.

a-d, TMD z-slices of 3D reconstructions from CRT1 in corresponding detergent conditions. Inset number indicates overall map resolution. A cartoon in panel a represents a schematic diagram of TMD helices, in initial purifications in DDM, both in the absence and presence of the terpenoid agonist nootkatone, the transmembrane domain (TMD) was poorly resolved, and the predicted fourth TMD helix, M4. was entirely absent. We thus collected EM datasets in two other detergents, lauryl maltose neopentyl glycol (L-MNG), and glyco-dlosgenin (GDN). L-MNG stabilized a n asymmetric TMD conformation, while GDN resulted in a well ordered and symmetric TMD, and the best overall resolution. Attempts to reconstitute the receptor in lipid nanodiscs resulted in profound aggregation. e, Representative cryo-electron micrograph of CRT1 in GDN detergent micelle from dataset of 4043 dose-fractionated micrographs. Scale bar indicates 100 nm. f, Projection images from the final selected 2D classes. g-i, 3D classification results; good classes selected for further processing are boxed In red. j, 3D reconstructed maps from the final 3D classification, which are shown in side-view and top-view from ECD. Selected 3D classes are boxed in red to generate a final 3D map. k, Unsharpened 3D map where M1-M3 TMD helices from an individual subunit are labeled in black text. l, Sharpened map colored by local resolution. m, Half map FSC plot for masked and unmasked maps with resolutions indicated at FSC = 0.143.

Extended Data Fig. 4 |. Cryo-EM density of the CRT1 receptor.

a, Cryo-EM density map of the CRT1 receptor tor representative adjacent subunits colored in blue and cyan. Density map of diosgenin colored in gray at a threshold level of 0.03. b, Orthosteric binding site of CRT1, where residues within 5 Å of diosgenin are shown as sticks. c, Orthosterlc binding site of $\text{α}7$ nicotinic receptor (PDB:7KOX), where residues within 5 Å of epibatidine are shown as sticks. d, Calculated interface areas and interaction energies ($\text{Δ}$¹G) for protein and diosgenin using PDBePISA. Calculated solvent accessible area and volume of the binding pocket for CRT1 and $\text{α}7$ nicotinic receptor using CASTp3.0⁴¹. e-g, Cryo-EM density segments of M1-M3 helices at a threshold level of 0.02. h-m, Cryo-EM density segments of Loop A-F in the orthosteric binding site at a threshold level of 0.02. j-l, Cryo-EM density segments of Cys-loop, M1M2 loop, and M2M3 loop at a threshold level of 0.02.

Extended Data Fig. 5 |. CRT1 Secondary structure prediction from AlphaFold2.

a, A subunit of CRT1 cryo-EM structure colored in dark blue. b, Predicted AlphaFold2 CRT1 monomer colored by per-residue confidence score (pLDDT). Regions of high expected accuracy are colored in blue: regions of low expected accuracy are colored in red. c, $\text{Cα}$ r.m.s.d between the CRT1 cryo-EM structure and the AlphaFold2 model (monomer comparisons) as a function of residue number. Helical regions are shown in green boxes. Blue boxes indicate $\text{β}$-strands next to the most divergent $\text{β}4$-$\text{β}5$ loop and $\text{β}8$-$\text{β}9$ loop regions. Unmodeled post-M3 helix is shaded in gray. d, Seque nee of CRT1 and secondary structure prediction by AlphaFold2. Helices are shown in green cylinders and strands are indicated by blue arrows. M4 helix amino acids are colored in red, blue, and black for negatively charged, positively charged, and hydrophobic residues, respectively. S-Sbond2 is in the Cys-loop. e, Sequence alignment of TMD helices (M1-M4) for CRT1 and other Cys-foop receptors. Same color scheme is followed as in Panel d.

Extended Data Fig. 6 |. Subunit and transmembrane pore conformation of CRT1 compared to the human α7 nicotinic acetylcholine receptor.

a, Comparison of single subunit structure of CRT1 and the human $\text{α}7$ nicotinic receptor (PDB:7KOX); MA and MX helices generally conserved in nicotinic receptors are absent in the CRT1 structure. b, Comparison of Loop C boxed in a; Disulfide bond on Loop C of $\text{α}7$ is shown as spheres. c, View at 90° rotation about pore axis from a. d, Comparison of ECD boxed in c; disulfide bonds are shown as spheres. e, M2 helices of the CRT1 receptor from opposing subunits (chains A and C) with pore-lining residues shown as sticks. Colored spheres indicate the pore diameter by displaying blue spheres (pore diameter > 5.6 Å), green spheres (2.8 Å < pore diameter < 5.6 Å), and red spheres (pore diameter < 2.8 Å), f, M2 helices of the $\text{α}7$ nicotinic receptor in an activated state (PDB :7KOX). g, M2 helices of $\text{α}7$ in a desensitized state (PDB:7KOQ). h, M2 helices of $\text{α}7$ in a resting state (PDB:7KOO). i, Pore diameters of CRT1 and $\text{α}7$ in panels e-h as a function of a distance along the pore axis. Structures were aligned using the M2 heiix Leu9’ at the midpoint of the pore, which we defined as y = 0.

Extended Data Fig. 7 |. CRT1 pharmacological properties.

a, CRT1 exhibited dose-dependent responses to nootkatone, zearalenone and GDN in patch-clamp experiments. Nootkatone EC50 = 16.1 μM, 95% CI = 13.1 – 18.7 μM, zearalenone EC50 = 4.8μM, 95% CI = 4.3 – 5.4 μM, GDN EC50 = 125.1 NM, 95% CI = 91.2 −184.6 nM. n = 8 cells per 1igand. b, Minimal desensitization was measured in response to low concentrations of agonist while higher concentrations produced inhibition with large wash off currents, consistent with moderate pore block. p < 0,0001 for concentration, two-tailed student’s t-test (n = 6 cells), c, Heat map of normalized axial nerve and arm responses to 50 μM of the indicated compound (except GDN at 5 μM to avoid micelle formation). Nootkatone and zearalenone elicited the most robust arm activity among CRT1 agonists or other molecules. Some differences in agonist efficacy in isolated receptors and arms are expected due to solubility issues at concentrations used for arm experiments, particularly with detergent molecules, unknown features of native CR signal transduction, neural integration. And differences in experimental preparations. n = 7–8 arms, d, WT CRT1 exhibited dose-dependent activity in response to nootkatone. Y78A mutant channels were insensitive to nootkatone except at higher concentrations which inhibited activity, n = 8 cells per condition, e, Current-voltage (I-V) relationships showing ligand-gated activity in WT CRT1 versus constitutive activity in agonist-binding site mutants. Ligands did not increase currents In mutant channels and all channels were sensitive to CR blocker mecamylamine (1mM). Data are represented as mean ±SEM.

Extended Data Fig. 8 |. CRT1 vs. α7 sequence alignment and residues under strongly positive selective pressure.

a, Side view of CRTI cryo-EM single subunit structure with residues under highly positive selection (LRT > 3) colored in red. b, Sequence alignment of CRT1 and human $\text{α}7$ nicotinic receptor (UniProt accession number: P36544). Unresolved post M3 region of CRT1 model is colored in gray. Red boxes indicate residues with LRT > 3 highlighted in panel a, and yellow boxes highlight the conserved Cys-loop disulfide bond.

Fig. 1 |. Octopus CRs are divergent and structurally distinct from related neurotransmitter receptors.

a, Anti-CRT1 (green) localized to the extracellular dendritic ending of putative receptor cells in the sucker epithelium of O. bimaculoides arms. Anti horseradish peroxidase is shown in purple. Representative of three octopuses. Scale bars, 1 mm (bottom left), 100 μm (middle), b. Octopus CRs form a clustered expansion of 26 intronless genes that diverged from octopus nicotinic AChRs. Scale bar, branch length, c. The non-synonymousor synonymous substitution rate indicates accelerated evolution of CRs versus nicotinic receptors, ***P < 0.001 for likelihood ratio test (LRT). d, Cryo-EM map and atomic model of octopus CRT1 with a single subunit highlighted from the homopentamer and diosgenin indicated in grey. Disulphide bonds in the agonist binding site and Cys-loop are shown as yellow spheres; N-linked glycans are shown as sticks. e. Top views per pendicular to the membrane viewed from the extraceilular space of the receptor map and model, f. Single subunit structure of octopus CRT1.

Fig. 2 |. Octopus chemotactile receptor permeation pathway.

a, CRT1 ion permeation pathway coloured by hydrophobicity with El04 indicated as spheres; the front subunit has been removed for clarity. b, Top view comparisons of CRT1 and human $\text{α}7$ nicotinic receptor show negatively charged El04 points towards the channel pore. c, Residues lining the permeation pathway from two M2 helices; the tan spheres illustrate pore shape. d, e, Mutating E104 to alanine or lysine decreased Ca²⁺ selectivity (d), leaving El04K with the least Ca²⁺ permeation (e). P < 0.001 for WT versus E104Aor E104K, P < 0.01 for E104A versus E104K. Two way analysis of variance (A NOVA) with post hoc Tukey’s test ($n=7$ cells). Data are represented as the mean± s.e.m.

Fig. 3 |. Octopus chemotactile receptor binds poorly soluble molecules.

a, CRT1 was sensitive to terpenes (35 μM nootkatone, 35 μM atrac tylon, 25 μM costunolide, 10 μM polygodial) but not ACh (100 μM ACh + modulator 10 μM PNU) in patch clamp recordings. The $\text{α}7$ nicotinic receptor was activated by ACh but not terpenes. P < 0.001, two-way ANOVA with post-hoc Bonferroni test ($n=6$ cells), b, Extracellular domains of octopus CRT1 (blue) and human $\text{α}7$ nicotinic receptor (grey). The orange box indicates partially overlapping agonist binding sites. Disulphide bonds are shown as blue and yellow spheres for CRT1. c, The surface of the agonist site is coloured by molecular lipophilicity potential from hydrophilic (KPI, green) to hydrophobic (HPO, brown). Agonists are shown as sticks. Part of Loop C from $\text{α}7$ was removed to visualize bound epibatidine. d, Structure-based sequence alignment of agonist binding site elements. Hydrophobic residues in 5 Å of bound diosgenin in theCRT1model are coloured orange. The yellow box Indicates a disulphide bond present only in CRT1. e, Screening of Xenopus oocytes expressing CRT1 identified active ligands with structural simiiarities to GDN and terpenes ($n=4$).Current-voltage relationship ofCRT1 activity in response to 15 μM nootkatone, 5 μM zearalenone or 125 nMGDN. f, The CR blocker mecamylamine(1mM) abolished ligand-evoked activity. P < 0.001, two-way ANOVA with post-hoc Bonferroni test (n = 4 cells), g, Arm axial nerve innervating suckers responded to 3 kDa filtered fish extract, 50 μM nootkatone or zearalenone. P < 0.001, one-way ANOVA with post-hoc Tukeys test ($n=7-8$). h, Amputated octopus arms exhibit autonomous responses to fish extract or 50 μM ligands but not control. P< 0.001, one-way ANOVA with post-hoc Tukey s test ($n=8$). Data represent the mean ± s.e.m.

Fig. 4 |. Evolution of the orthosteric binding site facilitates sensory adaptation.

a, Side view of the CRT1 orthosteric binding site showing residues involved in hydrophobic interaction with the diosgenin moiety of GDN. b, LRT to diversify selection indicated that ligand interaction sites (red) are under positive selective pressure. Threshold for significance, LRT = 3 (P < 0.05). c, Mutating hydrophobic residues coordinating GDN and under selective pressure resulted in constitutively active CRs with reduced sensitivity to 15 μM nootkatone, 5 μM zearalenone 125 nMGDN. Wild-type (WT) and mutant activities were blocked by 1 mM mecamylamine. Compared with WT, mutations increased average basal activity; P < 0.001, one way ANOVA with post-hoc Tukey’stest ($n=5-6$ cells), d, Mutations reduced ligand-dependent currents. P < 0.001, two-way ANOVA with post-hoc Tukey’s test ($n=\mathrm{5–6}$ cells). Data represent the mean ± s.e.m.