What if cactus lipid-transfer proteins keep the same sturdy molecular core, but evolve by changing the flexible “gates” around their lipid tunnel? This study finds exactly that pattern in Cereus—while carefully stopping short of claiming proven ecological adaptation.

Cereus is widely distributed across arid and semi-arid environments of South America, and its adaptive capacity is partly associated with proteins involved in lipid metabolism and stress responses, such as non-specific lipid transfer proteins. Although nsLTPs are well-characterized in model plants, their structural diversity and functional roles in cacti remain largely unexplored.

The study characterizes the Cereus nsLTP repertoire by integrating transcriptomic data, phylogenetic analyses, structural modeling, and ligand-binding site mapping. Cereus comprises approximately 31 recognized species and ranges from coastal regions to Andean habitats above 3,000 meters in altitude.

This broad distribution reflects adaptive capacity to contrasting environmental conditions, including drought and temperature fluctuations typical of xeric and semi-arid ecosystems. Recent phylogenetic analyses used nuclear orthologous genes and a coalescent-based approach, supporting Cereus as monophyletic and partially recovering three of four subgenera.

At the molecular level, exploring Cereus proteins may uncover mechanisms underlying the genus’s evolutionary history and adaptive radiation. nsLTPs are small, around nine kilodaltons, basic proteins broadly distributed across higher plants, typically composed of four to five alpha helices and stabilized by eight cysteine residues forming four disulfide bonds.

This compact fold creates a hydrophobic tunnel-like cavity that allows nsLTPs to bind and transport a wide range of lipid molecules. Previous studies associate nsLTPs with responses to drought, salinity, cold, oxidative stress, and pathogen attack. In drought-tolerant and xerophytic systems, nsLTPs have been associated with cuticular lipid deposition, wax accumulation, membrane remodeling, and maintenance of cellular integrity under water deficit conditions.

Despite these advances, nsLTPs remain poorly characterized in non-model cactus lineages, and no integrated structural and evolutionary analysis had been performed for Cereus. The study addresses a gap: nsLTP diversity, structural features, and evolutionary history in non-model taxa such as Cereus and Cipocereus remain largely unexplored.

The integrative framework combines comparative transcriptomics, phylogenetic reconstruction, structural prediction, binding-site characterization, and codon-based evolutionary analyses. The central question is whether sequence variation is concentrated around predicted ligand-binding cavities while the overall nsLTP structural scaffold remains conserved.

Transcriptomic datasets represented seven taxa of Cereus and Cipocereus, with RNA sequencing performed from both root and epidermis. Only predicted nsLTP coding sequences identified from these transcriptomes were used in the study.

Redundant contigs were clustered with CD-HIT at a 90 percent identity threshold, and protein-coding sequences were used for orthology assignment in OrthoFinder. The nsLTP orthogroup contained 55 protein sequences, all originating from transcriptomic assemblies and used in subsequent structural and evolutionary analyses.

Structural predictions of Cereus nsLTP homologs were generated by AlphaFold3 using default settings. The standard AlphaFold confidence output, pLDDT, or per-residue confidence, was recorded to guide downstream interpretation. For cross-target comparability, the full predicted polypeptide was used, with no manual trimming or domain excision before alignment.

Figure two evaluates the AlphaFold model using a Ramachandran plot, which maps backbone dihedral angles to assess stereochemical quality. The plot shows residues concentrated in the favored regions, while the accompanying summary reports ninety-two percent in the core, eight percent in allowed regions, and zero percent in disallowed regions.

This supports the authors’ assessment that the predicted nsLTP structure has good geometric reliability for subsequent binding-site analysis. FTSite and FTMap were used to identify ligand-binding regions on protein surfaces.

Both methods probe the surface with small organic fragments and define consensus binding sites where multiple probe clusters overlap; FTMap further characterizes these clusters as energetically favorable hot spots. FTMap hot spots were ranked by cluster strength and mapped onto predicted cavities.

Across species and paralogs, high-density probe clusters frequently appeared in large and highly ligandable pockets, providing qualitative agreement between independent approaches. Molecular docking analyses were performed with AutoDock Vina version one point two point five to evaluate predicted ligand-binding properties.

Representative AlphaFold3 models were used as receptors, while five naturally occurring fatty acids associated with plant lipid metabolism were selected as ligands. The species phylogeny used for gene-tree reconciliation was built from 719 single-copy orthogroups identified with OrthoFinder.

Sequences from each orthogroup were aligned with MAFFT and analyzed with ASTRAL-III under a multispecies coalescent framework that accounts for incomplete lineage sorting. The resulting species tree served as the reference topology for gene-tree reconciliation in Notung, allowing inference of gene duplication and loss events.

The Mixed Effects Model of Evolution, or MEME, was applied to detect signatures of adaptive evolution in nsLTP coding sequences. MEME estimates the ratio of nonsynonymous to synonymous substitution rates and allows that ratio to vary across sites and across branches at a given site.

The analysis used a significance threshold of p less than zero point zero five, and sites below that threshold were considered candidates for episodic diversifying selection. Because recombination was not explicitly modeled, the inferred sites were treated as candidates for future investigation rather than definitive evidence of adaptive evolution.

A total of 55 nsLTP proteins were recovered from transcriptomic datasets representing seven taxa of Cereus and Cipocereus. Multiple sequence alignment revealed strong conservation of the canonical eight-cysteine motif, a defining structural feature of plant nsLTPs.

All eight cysteine residues were conserved across the analyzed sequences despite substantial variation in the intervening regions. In panel B, the cysteine columns are marked across the enlarged alignment, and each forms a tall, narrow green peak reaching roughly four bits of information.

The neighboring columns are much more mixed in letter identity and height, making the contrast between the cysteine positions and intervening sequence especially clear. TM-scores consistently fell within the range interpreted as the same topology, around zero point five to zero point seven, with aligned cores of approximately 60 to 80 residues and backbone RMSDs around two to three angstroms.

Models with higher loop divergence also tended to show greater pocket-volume variability and altered rim polarity, suggesting an association between peripheral structural variation and tunnel geometry. The AlphaFold3 model revealed the canonical nsLTP alpha-helical scaffold with a conserved hydrophobic cavity.

The predicted structure of Cereus jamacaru was compared with experimentally determined plant nsLTPs in the Protein Data Bank. Structural superposition with Solanum melongena and Triticum aestivum confirmed that the overall nsLTP fold is highly conserved despite sequence divergence.

The AlphaFold3-predicted structure revealed three major binding regions characterized by multiple overlapping FTMap probe clusters. Site number one showed the highest probe diversity, with sixteen probe types distributed across three overlapping clusters, indicating a structurally accessible binding environment.

Looking across the probe column, Site number one has a three-level pattern: Cluster one lists sixteen probes, Cluster two lists eleven, and Cluster three lists three. Site number two has two clusters with thirteen and four probes, while Site number three presents a single cluster listing fifteen probes.

Representative nsLTP models underwent molecular docking against a panel of lipid-related ligands to test whether structural variation around predicted binding pockets could influence ligand recognition. Docking scores differed significantly among ligands and protein variants, indicating that the predicted binding cavities are not energetically equivalent.

All proteins retained the characteristic hydrophobic tunnel, but predicted binding energies varied according to ligand identity and local pocket geometry. Maximum-likelihood analysis recovered all Cereus nsLTPs within a single, well-supported lineage relative to representative plant nsLTPs used as external references, supporting a monophyletic origin of this gene family within the genus.

Orthologous nsLTP sequences from different Cereus species were extensively intermingled throughout the phylogeny rather than forming species-specific clusters. This pattern indicates that much of the family’s diversification predates the radiation of extant Cereus species, with many duplicated copies retained across descendant lineages.

Several terminal clades contained multiple copies from the same species, revealing lineage-specific paralogous duplications superimposed on the ancestral pattern. The evolutionary history therefore appears to involve both ancient duplication events shared among species and more recent duplications restricted to individual lineages.

The MEME analysis used 39 aligned sequences totaling 133 codon sites; 18 sites showed evidence of episodic positive selection, corresponding to about 13.5 percent. Nine sites displayed p less than zero point zero five, while the others did not reach that threshold.

On average, each codon under positive selection was detected in one to two branches, suggesting that selective pressure acts locally in specific clades or lineages rather than uniformly across the gene. These results identify candidate codons associated with episodic sequence diversification rather than definitive targets of adaptive evolution, and the functional consequences remain unknown.

In panel A, the plotted points fall at distinct codon positions, with several below the dashed p-value threshold and others above it, showing that the signal is uneven across the sequence. In panel C, the red positively selected residues cluster near the colored binding regions, including Site one, Site two, and Site three, rather than appearing uniformly across the structure.

The proposed gate-modulation framework combines a conserved four-helix scaffold that maintains the hydrophobic tunnel with loop-level diversification at rim and entrance regions that tunes tunnel accessibility and local physicochemical properties. Structural comparisons, pocket and hot-spot mapping, phylogenetic context, and episodic selection signals place functional tunability in peripheral, flexible regions rather than in the structural core.

Because AlphaFold3 generates static structural predictions, the dynamic behavior of these regions cannot be directly evaluated. Interpretations involving conformational flexibility, ligand-induced rearrangements, or alternative structural states require future molecular dynamics simulations or experimental structural approaches.

High confidence scores for the alpha-helical scaffold and consistent binding-region identification across complementary methods support the structural reliability of the predicted models for comparative evolutionary analyses. The results should be interpreted as hypotheses about nsLTP evolutionary diversification in Cereus, rather than direct evidence of ecological adaptation or physiological specialization.

Biochemical assays, ligand-binding experiments, molecular dynamics simulations, gene expression analyses, and comparative physiological approaches will be necessary to determine whether the molecular differences influence protein function. Episodic diversifying selection affected only one or a few branches for each candidate codon rather than acting consistently across the entire phylogeny.

Together with ancestral orthologs and lineage-specific paralogs, these observations are consistent with localized evolutionary diversification within a conserved protein family. Although the biological significance of individual substitutions cannot be determined from sequence analyses alone, the candidate sites provide a rational starting point for future biochemical and functional studies.

Cereus nsLTPs preserve a conserved four-helix scaffold and hydrophobic tunnel, while loops, pockets, duplications, and candidate selection sites diversify locally. The result is a structural hypothesis for functional tuning—not direct proof of stress adaptation.