Evolutionary and structural bioinformatics identifies GPR89 as a conserved member of the LIMR protein superfamily

Abstract

Also called the Golgi pH regulator (GPHR), GPR89 is an orphan membrane protein found in nearly all eukaryotic lineages. Despite its broad phylogenetic distribution, the evolutionary history, structural diversity, and function of GPR89 remain poorly understood. In this study, we present a comprehensive bioinformatic analysis of GPR89 in Eukarya by integrating phylogenetic reconstruction, genomic synteny, sequence conservation, and structural modeling. While GPR89 is typically encoded as a single-copy gene, we identified lineage-specific duplications in both vertebrates and vascular plants. In contrast to the large sequence conservation, differences can be observed in whether plants and animals preserve the gene structure flanking GPR89. Structural clustering places GPR89 within the solute carrier (SLC) group, together with LIMR protein family members. Predicted structures reveal a unique intracellular helix hairpin and a conserved transmembrane core compatible with putative transport activity. This work provides a unifying framework for interpreting the existing evidence on the GPR89 function and proposes that GPR89 should be classified within the LIMR superfamily.

Graphical Abstract

Highlights

• •GPR89 is conserved in nearly all major eukaryotic lineages.
• •Structural analysis places GPR89 within the LIMR superfamily.
• •The predicted structure features a conserved core and a unique intracellular helical hairpin.

1. Introduction

GPR89 is a highly conserved orphan membrane protein found across unicellular eukaryotes, land plants, and animals [13], [16], [43], [53], [9]. Deficient GPR89 function in mammals has been associated with neurodegeneration [56], disrupted brain cholesterol synthesis [57], and compromised epidermal barrier function [60]. Originally annotated as a G protein-coupled receptor (GPCR), GPR89 displays nine predicted transmembrane domains, differing from the canonical seven-transmembrane GPCR structure. Although interactions with the G protein alpha subunit have been reported in plants such as rice, maize, and Arabidopsis thaliana [16], [42], [49], [70], GPR89’s classification as a functional GPCR remains unconfirmed due to the absence of a defined activation mechanism or a preferred signaling pathway [30].

Interestingly, GPR89 has been linked to a variety of molecular functions that might seem incompatible at first. In vascular plants, its localization and proposed roles vary widely; it has been described as a cold-activated calcium-permeable channel [42], an abscisic acid receptor [49], [70], and an ion channel present in the ER, Golgi, and the plasma membrane [30].

In animals, GPR89 has been referred to as the Golgi pH regulator (GPHR) and is predominantly located in the Golgi apparatus and endoplasmic reticulum [43], [56], [9]. A particularly intriguing hypothesis suggests that mammalian GPR89 forms a large-conductance, non-rectifying chloride channel (∼300 pS; P_Cl-/P_K+ ≈ 2.5) in the Golgi membrane, potentially corresponding to the GOLAC-2 current [43], [61]. This chloride conductance may provide the counterion flux necessary to support proton accumulation and maintain luminal pH gradient in the Golgi. Importantly, Golgi pH regulation is associated with several pathologies, including congenital glycosylation disorders, cancer, increased viral susceptibility, and intellectual disabilities [36], [52].

Lastly, in the unicellular eukaryote Trypanosoma brucei, GPR89 is found at the plasma membrane, where it likely functions as an oligopeptide transporter [53]. These divergent functions and localizations suggest that the different orthologs of GPR89 may have evolved distinct physiological roles across taxa, while maintaining a conserved core structure.

In this study, we investigated the evolutionary and structural landscape of GPR89 to better understand its roles across eukaryotic lineages. We show that it is a well conserved protein, retained as a single-copy gene in most taxa, with independent duplication events observed in both metazoa and vascular plants. Our clustering analysis strategy revealed that GPR89 displays strong structural similarity with members of the LIMR protein family, forming a group that is nested between solute carrier transporter (SLC) superfamilies. Moreover, sequence conservation and evolutionary covariance in AlphaFold2-predicted structures highlights a conserved hydrophobic core centered on transmembrane segment 5 (TM5). The latter forms part of a lateral groove that could be potentially involved in substrate binding and transport. Surrounding this core are polymorphic intracellular and extracellular loops, including a prominent cytosolic alpha-alpha hairpin (HP) super-secondary structure. The loop at the tip of the hairpin (HP_loop) shows marked sequence variability, especially among unicellular taxa, and may represent a modular domain contributing to the observed functional divergence of GPR89 across eukaryotic lineages. Overall, GPR89 represents a structurally conserved membrane protein that might have acquired lineage-specific subtle functional specializations, while retaining a core architecture associated with membrane permeability and organellar homeostasis.

2. Results

2.1. Repertoire of GPR89 across species

To understand the process of diversification of GPR89, we first performed a Maximum likelihood (ML) phylogenetic analysis using CDS from representative members of diverse taxonomic groups. Following calculation of hydrophobicity profiles, we recovered a total of 130 sequences, annotated as bona fide GPR89 orthologs in model species (Supplementary Table 1). The collected orthologs had an average length of 1368 nucleotides and the final alignment spanned 3913 nucleotides, displaying 248 conserved sites.

The Maximum Likelihood tree obtained shows the presence of the GPR89 gene among Metazoa, Archaeplastida, and Heterotrophic protists corresponding to Amorphea (Holozoa, Ascomycota, and Amoebozoa), Excavata, and SAR groups (Stramenopiles, Alveolata, and Rhizaria lineages) (Fig. 1). The gene tree is in agreement with previous phylogenetic hypotheses obtained by unrooted Neighbor-Joining trees [13], [30], [45]. Our phylogenetic reconstruction, recovers the monophyly among Metazoa, within Vertebrata, and its natural relations as monophyletic clades (Mammalia, Sauropsida, Anura, and Actinopterygii). Furthermore, the tree recovers Chlorophyta species as a sister group of Embryophyta (Fig. 1). Other relationships—such as those established among heterotrophic protists—appear as paraphyletic clades, including organisms from SAR and Opisthokonta (Fig. 1). Therefore, our analysis is consistent with both previous observations and the updated hypotheses for the speciation of Eukaryotic organisms [7].

Fig. 1.

Phylogenetic tree of GPR89 CDS across Eukarya. Best Maximum Likelihood phylogenetic tree displaying large groups in color code, Archaeplastida in green, heterotrophic protist in blue and Metazoa in orange. Numbers under the nodes represent bootstrap support values and red circles denote GPR89 intraspecific functional gene duplication. In parenthesis it indicates the number of sequences for each collapsed clade. Bar indicates nucleotide substitutions per site. The tree was rooted at the midpoint.

The branch lengths within major clades are relatively uniform, which supports a high degree of sequence conservation among closely related taxa (Fig. 1). An exception to this pattern is observed in heterotrophic protists, where significantly longer branch lengths indicate a diversification process, contrasting to the higher conservation observed within Metazoa and Archaeplastida. Among protists, Trypanosoma brucei emerges as the most divergent species.

A paralog of GPR89A, named GPR89B, was identified in both Metazoans and Archaeplastida, but is absent in heterotrophic protists (Fig. 1). Among Metazoa, the presence of more than one copy is relatively rare and, when observed—as in some representatives of Mammalia, Actinopterygii, Nematoda, Cnidaria, and Brachiopoda—it occurs in distant lineages, suggesting it results from lineage-specific, independent gene duplication events rather than a shared ancestral duplication (Fig. 1).

In contrast, multiple copies of the GPR89 gene are more frequently found in plants, particularly in green algae (e.g., Chlorella variabilis) and in a wide range of vascular plants, including Lycophytes, Gymnosperms, and Angiosperms (both Eudicots and Monocots) (Fig. 1). This broader retention of duplicated gene copies in plants aligns with the well-documented prevalence of whole-genome duplication (polyploidy) events in plant evolution and the generally higher tolerance for genetic redundancy and plasticity in plant genomes [39], [41], [48], [68].

To better understand the evolutionary history of GPR89, especially in groups with known duplication events and similar evolutionary time frames, we inspected the genomic organization around GPR89 orthologs. For this purpose, we compared the synteny found from from H. sapiens to Danio rerio (divergence 429 MYA; [37]) (Supplementary Figure 1) and that of A. thaliana to Marchantia polymorpha (divergence 480 MYA; [37]) (Supplementary Figure 1 A & B).

For the case of vertebrates, a comparison of the genomic regions displayed by the mammalian, sauropsid, and amphibian orthologs indicates that the single copy gene is flanked by five conserved genes: GJA8, GJA5, APC6, and BLC9 upstream, and by PDZK1 downstream. In fish, only PDZK1 remains in the syntenic group (Supplementary Figure 1 A). For the case of humans, the two copies of GPR89 arose from a unique duplication event that happened approximately 4.7 MYA [15]. These gene copies are located in close proximity to each other on chromosome 1, suggesting a duplication event that originates from a single locus (Supplementary Figure 1 A). In contrast, Embryophyta displays a similar synteny only at the family taxonomic level (Supplementary Figure 1B). For instance, in Brassicales, A. thaliana GPR89A (known as AtGTG1) on chromosome 1 has similar gene composition than his ortholog in B. oleracea (that is located on chromosome C2). The consistency of the syntenic region is lost over larger evolutionary distances, which can be explained by the more extensive genomic rearrangement reported in plants [55]. Thus, the remarkably conserved synteny in chordates contrasts with the absence of conservation displayed by plants.

Taken together, our findings suggest that the occurrence of multiple functional GPR89 copies across taxa did not stem from a single ancestral duplication, but rather from multiple, independent duplication events—such as segmental duplications and polyploidizations—reflecting distinct evolutionary pressures in Metazoan and Archaeplastid lineages (Supplementary Figure 1).

2.2. Detailed evolutionary trajectory of GPR89 among mammals

In addition to the canonical duplication, we identified a reported human pseudogene annotated as GPR89P. This prompted a broader search for GPR89A-related pseudogenes. Among eukaryotes, we found GPR89P restricted to mammals, particularly in most Anthropoid primates, and sporadically in Lagomorpha (rabbit), Scandentia (treeshrews), Rodentia (naked mole rat, degu, jerboa), Cetartiodactyla (blue whale, dolphin), and Marsupialia (wombat, koala). To further explore the evolutionary trajectory of GPR89 in mammals, we conducted a phylogenetic analysis including sequences for GPR89A, GPR89B, and GPR89P (Supplementary Table 2). The resulting gene tree confirmed one-to-one orthology of GPR89A across mammals (Supplementary Figure 1 C), showing the expected relationships among major lineages. The only exception was the paraphyly of Rodentia, likely due to long-branch attraction involving Jaculus jaculus and Dipodomys ordii [18], [5]. Among mammals, two coding copies of the gene were found exclusively in Homo sapiens and Vombatus ursinus, with each pair exhibiting near perfect CDS similarity (1 bp synonymous substitution in humans, completely identical in wombat). Moreover, our topology recovered monophyletic clades containing both the coding gene and pseudogene for Primate, Scandentia, Lagomorpha, Cetartiodactyla, and Marsupialia. These observations suggest species-specific duplications rather than a shared ancestral event for both coding genes and for the retrospseudogene copies as well (Supplementary Figure 1 C).

2.3. Amino acidic conservation patterns highlight critical regions

In order to identify possible conserved motifs, and regions of high variability in GPR89 orthologs, we translated our set of sequences to their respective protein primary structure (Supplementary table 1). The final set included representatives of Opisthokonta (Metazoa, Fungi, and Choanozoa), Chlorophyta (land plants and single-celled algae), as well as heterotrophic protists, such as members of the groups SAR, Rhodophyta, and Amoebozoa. We aligned the set of translated sequences and constructed an amino acid sequence identity histogram to detect conserved and variable regions within the protein sequence of human GPR89 (Fig. 2). We observed that the regions encoding for TM segments previously suggested in the literature [16], [43], [49], [70], [9] are not only well defined but largely conserved among all taxa (Fig. 2A). A thorough analysis of the amino acid distribution revealed 4 residues with conservation over 95 %, which mapped to their positions in human GPR89 correspond to P91 at TM3 and G154, L161 & G163 at TM5 (Fig. 2A & B).

Fig. 2.

Multiple Sequence Alignment of the translated sequences of curated GPR89 orthologs found across Eukarya. (A) Per-residue frequency histogram for the different amino acids. Structural annotations from the human GPR89A protein, obtained from the AlphaFold DB and DeepTMHMM, are mapped to their corresponding positions within the histogram: transmembrane helices (TM1–9, gold/hatched), hairpin helices (HP1&2, blue/hatched), and their connecting hairpin loop (HP_loop, magenta/hatched). Black triangles above the bars mark the positions of highly conserved residues (>95 %). A red rectangle delimits a cluster of highly conserved residues at TM5. (B) Amino acid frequency logo of the TM5 conserved cluster. Black arrows indicate 95 % conserved residues P91, G154, L161 & G163 mapped to the human GPHRA sequence. (C) MSA of the hairpin region of selected GPR89 orthologs from animals, plants, and one choanoflagellate. Residue similarity is indicated at 100 % (black), > 80 % (dark gray) & > 60 % (light gray). A consensus sequence of residues over 90 % conserved is shown on top. A highly variable region is observed surrounding and encompassing the HP_loop.

An important degree of polymorphism, including a variety of insertions/deletions and natural mutations, was found in the portion of the intracellular linker located between TM5 and TM6 (Fig. 2A & B). This region is predicted to fold in an alpha-alpha helix hairpin (HP) with a short loop connecting the helices in most collected sequences (HP_loop) (Fig. 2A & C). The most dramatic changes in the polymorphic region corresponding to large insertions in the HP_loop (Fig. 2A). These are present in orthologs from unicellular eukaryotes such as Amoeba, green algae, Ascomycota, and Excavata. Interestingly, in flowering plants, such as Oryza sativa and A. thaliana, this linker has been implicated in binding the alpha subunit of the G protein heterotrimer [16], [42], [49], [70].

To inspect whether certain polymorphisms are associated with major taxonomic groups, representative sequences of chordates and vascular plants were inspected without the large gaps introduced by most unicellular organisms (Fig. 2C). Notably, both mammals and flowering plants have clear consensus sequences respective to their own groups. For example, a member of an ancient Tracheophyte lineage (Selaginella moellendorffii) shows a variable region closely resembling that of flowering plants, while an Actinopterygii (Danio rerio) displays a sequence that is closer to the one in mammals. On the other hand, the choanoflagellate Monosiga brevicollis, which lacks some of the distinct insertions of unicellular organisms found in this region, shows substitutions that do not clearly align with the distinguishing characteristics of plants or animals (Fig. 2C). Overall, these analyses suggest that putative functional properties arising from this sequence segment, such as G_alpha binding reported in plants, might have evolved within their respective taxonomic group.

2.4. Structural alignments identify GPR89 as a member of the LIMR superfamily

Protein structure is closely related to function, and is more sensitive to distant homologies than the primary sequence alone [10], [29]. For this reason, we decided to search for structural similarities instead of primary sequence homology. A preliminary search was performed using the predicted structure of human GPR89 (Uniprot ID: B7ZAQ6) against the full human AlphaFold Protein Structure Database via the Dali server [28], [64]. Surprisingly, across the entire database, only the four members of the poorly studied LIMR family (Pfam PF04791) emerged as the closest structural matches, despite the fact that their primary sequences share less than 15 % identity (Table 1). Further structural examination shows that they all share 9 transmembrane helices, with a characteristic linker between TM5 & 6 (Supplementary Figure 2).

Table 1.

Structural comparison of human GPR89A against the AlphaFold Database using the Dali Server. Top hits (Z-scores > 10) are shown.

Protein	Uniprot accession	%Id	RMSD	Z-score
LMBD1	Q9NUN5	15	5.6	19.6
LMBD2	Q68DH5	10	4.3	18.7
LMBRL	Q6UX01	13	8.7	17.7
LMBR1	Q8WVP7	13	5.7	17.5

The namesake member of this family, LIMR (i.e., LMBRL), has been associated with the endocytic transport of hydrophobic molecule-binding proteins, such as lipocalin-1 and β-lactoglobulin ([66]; Fluckinger et al., 2007). Another member, LMBRD1, participates in the lysosomal cobalamin transport, acting as an escort protein of the ABCD4 transporter [14], [35], [54].

It has been suggested that GPR89 orthologs act either as a transporter or an ion channel (Maeda et al., 2009; [42], [53]). One of the most thorough functional studies was done in Trypanosoma brucei (Rojas et al., 2015), where authors suggest a role as an oligopeptide transporter, similar to SLC15, a member of the Major Facilitator Superfamily (MFS, CATH 1.20.1250.20) of solute carrier transporters (SLC). Members of the SLC superfamily are found throughout the eukaryotic cell in both the plasma membrane and organellar membranes. The SLC group comprises over 400 different transporters. According to their sequence homology and characteristic function, SLC are organized in multiple distinct families [19], [26], [62]. Thus, we focus our efforts on whether the human GPR89 ortholog is part of the large group of transporter proteins including SLCs and ABCs. The multiple structure alignment prediction was performed over 480 human proteins by using TM-Vec, a deep learning approach that efficiently approximates TM-scores from primary sequences without explicit structural superposition [25]. We included most human SLC and ABC transporters, plus representative members of other families, such as chloride and calcium channels, and GPCRs. Explicit structural outgroups were used to root the tree, i.e. beta-barrels (VDAC1 & TOM40) and globins (myo & hemoglobin). From the predicted TM-scores rendered by TM-Vec, we constructed a distance tree (distance = 1 - TM-score) using unweighted pair group method with arithmetic mean (UPGMA) analysis (Fig. 3). The resulting tree can be roughly categorized into two main branches: the MFS/DMT superfamily-containing branch (CATH: 1.20.1250.20 & 1.10.3730.20 respectively) and the APC superfamily-containing branch (CATH: 1.20.1740.10/1.20.1730.10). Consistent with the Dali results, GPR89 groups well with all LIMR family members, with the closest alignment being to LMBD2 (Fig. 3A; tree file in Supplementary Materials). This group lies in the APC-containing branch, between the SLC39 (Zinc-Iron permeases) and SLC42 (Ammonium transporters) families (Fig. 3A).

Fig. 3.

Pairwise structural alignment prediction of GPR89 with SLC transporters and other transmembrane proteins. (A) Expanded unrooted tree depicting the UPGMA groups from the TM-Vec prediction. Branch colors and numbers at the tips indicate the CATH fold classification of the transmembrane domains for each protein; black signifies no annotated CATH classification. Non-transporters are indicated: CAC (*), CLCN (**), CFTR(***) and GPCR (*****). The sky blue triangle indicates GPHRA, and the gray shaded sections correspond to the main structural groups. The black dotted section was selected for the reanalysis shown in panel B. (B) Pairwise structural alignment of the GPR89 neighborhood. The numbers at each node show their UPGMA distance (average 1 - TM-Score). A distance cutoff of 0.4 (dotted line) reflects an 80 % probability of shared CATH classification between random alignments [67]. (C) Structural Alignments showing the citosolic view between GPR89 and representative members of each family are presented at the bottom. Blue: GPR89, Yellow: SLC42 (PDB ID: 8CSX), Orange: SLC4 (PDB ID: 6CAA), Magenta: SLC39 (PDB ID: 7Z6N), Red: LMBRL.

Since TM-Vec analysis considered the entire protein sequence, unwanted noise in the grouping procedure might have been introduced by soluble or intrinsically disordered domains. To address this possibility, we performed a pairwise structural alignment considering the transmembrane domains exclusively, using TM-align [69]. The advantages of this approach include the utilization of explicit three-dimensional coordinates and the computation of inter-query Euclidean distance. Despite some grouping rearrangements, GPR89 still clusters with the LIMR family, being the earliest diverging member of the group, with an average pairwise distance to other members of 0.351 (Fig. 3B). This value falls below the established TM-score threshold of 0.4 for pairwise analysis, which indicates a > 80 % probability of sharing a CATH structural classification [67]. Above this threshold, we may only infer a shared topology at best.

In short, our evidence suggests that GPR89 is a novel member of the LIMR superfamily, sharing their characteristic 9TM and HP domain. Structurally, the transmembrane region of the group resembles SLC transporters more than ABC transporters, Calcium and Chloride Channels, or GPCRs. Nevertheless, taking into account the limitations of performing in silico structural analyses over predicted targets, further experimental data would be necessary to confirm our predictions.

2.5. GPR89 shows a conserved transmembrane groove

To better understand the implications of similarities and differences displayed by GPR89 orthologs, we inspected the predicted structures available in the AlphaFold database [64], accounting for representative organisms of the various species and groups included in the present work. The predicted structure for human GPR89A (Uniprot ID B7ZAQ6), calculated by AlphaFold 2 (AF2) [31], [64], shows the expected 9 transmembrane helices and a long helix hairpin (HP) that shapes the TM5-TM6 cytosolic linker (Fig. 4). These structural features are consistent with other predicted models in the database, such as those from A. thaliana (UniProt: Q9XIP7), D. discoideum (Q54QM5), and T. brucei (Q580H0). In these models, the prediction of a 9TM architecture with a long helical hairpin constituting the majority of the TM5-TM6 linker is a systematic feature, likely constituting a structural signature for GPR89. In general, the large insertions displayed in various organisms occur predominantly in and around the cytosolic HP_loop.

Fig. 4.

Predicted structure and conservation of human GPR89A. (A) AlphaFold2-predicted structure of GPR89A, showing 9 transmembrane helices and two cytosolic helices forming a hairpin. Colored TM residues correspond to those predicted by the DeepTMHMM server. (B) Conservation scores from ConSurf, using the MSA from Fig. 3, were mapped onto the structure and colored as a gradient from magenta (most) to cyan (least) conserved.(C) A closer look at the highly conserved TM5 and its vicinity. (D-F) Lateral hydrophilic cavity observed from the cytosolic (D), membrane (E) and extracellular/luminal perspectives (F). The surface is colored by calculated electrostatic potential; red: negative charge, blue: positive charge, white: neutral. TM5 is depicted in yellow.

To better visualize the relative position of the conserved residues, we displayed the conservation score at each position of the MSA previously obtained (Fig. 2), in the context of the predicted structure for GPR89A (Fig. 4 B & C). By doing this structural mapping, we noted that conserved residues of different helices appear in close proximity (<4 Å atom to atom) to the well conserved TM5 helix, which appears tilted relative to the other transmembrane helices (Fig. 4C). A large, partially hydrophilic vestibule, formed by TM5, TM6 & TM7, is accessible from the cytosolic face of the predicted structure (Fig. 4 D & E). This opening displays polar residues (R300, R357, Y390, H423, D427). The cavity appears inaccessible from the extracellular side of the predicted structure (Fig. 4F).

3. Discussion

The present work confirms the broad conservation of GPR89 across eukaryotic lineages and reveals its predominant retention as a single-copy gene, which suggests its expression is under tight regulatory control. This would be consistent with a strong purifying selection and a function of general importance, underscored in recent literature [23], [56]. The strong congruence between the phylogeny of GPR89 and the species tree [7] further indicates evolutionary robustness preserved along speciation events. An exception is found in protists, which exhibit exceptional divergence, suggesting a distinct evolutionary path potentially involving lineage-specific innovations, restricted to specific protein domains.

Despite this conserved evolutionary signature, GPR89 exhibits diverse molecular roles and subcellular localizations across taxa. In vertebrates, it is found at the Golgi and endoplasmic reticulum membranes, where it likely supports the anion flux that is essential for luminal acidification [43]. It has been suggested that GPR89 orthologs in plants mediate calcium permeability and the detection of hormonal and environmental cues [16], [42], [49], [70]. In Trypanosoma brucei, it localizes primarily to the plasma membrane where it supports the transport of small peptides [53]. Collectively, this evidence suggests a general role in membrane transport.

Our data strongly suggest that GPR89 shares a structural topology with the LIMR superfamily. Members of the LIMR protein family have been described as endocytic receptors [20], [66], transporter escort proteins [14], [35], or mediators of GPCR signalling cascades [11], [40], [47], indicating functional plasticity.

A shared feature is the HP_loop, a distinct intracellular helix containing a Gα binding site in plant GPR89 [16], [42], [49], [70]. This Gα-binding capacity aligns well with a role in modulating membrane permeability and may represent a common feature within the LIMR family.

The apparent functional diversity of GPR89 across eukaryotes could be explained by a hypothetical role as an escort protein, which interacts with clade-specific transport proteins or membrane receptors. However, a significant challenge to this model arises from the functional recovery demonstrated by Rojas et al., where the Trypanosoma brucei gene complemented a transporter-deficient strain of E. coli. Bacteria lack the complex eukaryotic machinery (e.g., G proteins, clathrin) required for an escort function, therefore, this finding suggests that at least TbGPR89 may operate as a transporter by itself.

Our findings thus support the transporter hypothesis originally proposed by Rojas et al. [53]. The structural classification places GPR89 and the related LIMR group within the SLC transporter superfamily, with closer structural similarity to the APC family. Moreover, AlphaFold2 structural models revealed a conserved transmembrane scaffold that would be compatible with solute translocation. Consistent with known SLC architecture [4], [51], the predicted structure is organized in two sections, displaying a central hydrophobic core surrounding a specific transmembrane segment (TM5) [12], [17]. TM5 is flanked by conserved transmembrane helices, and polymorphic cytoplasmic and extracellular loops. The cytoplasmic loops display kingdom-specific variations that are particularly evident in the intracellular helix hairpin (HP) and the associated HP_loop, which may govern regulatory interactions or subcellular targeting. It is important to note that no structure of the putative LIMR superfamily or GPR89 ortholog has been resolved, and that AlphaFold predictions are less reliable for orphan proteins with no close experimental evidence [65].

The inherent modularity of SLCs allows subtle molecular adaptations to have major functional consequences. For instance, prestin (SLC26A5) has lost the capacity to transport and evolved into a molecular actuator for electromotility in mammalian cochlear hair cells, sustaining the hearing process in mammals [38]. Therefore, it would be reasonable to suggest that GRP89 might be an ancient protein that retained transport capacity while other members of the LIMR protein family lost their transport capacity and turned into modulatory proteins.

In summary, phylogenetic stability, conserved transmembrane architecture, and structural classification places GPR89 within a well-supported structural and evolutionary context as part of the LIMR superfamily.

4. Materials and methods

4.1. Phylogenetic analyses

We used the ortholog prediction function of Ensembl database [27] and BLAST on the NCBI platform [2] to search for ortholog and paralog genes of GPR89 on major groups of vertebrates, metazoan model species, major groups of Viridiplantae (including green algae), and unicellular eukaryotic organisms such as heterotrophic protists. After corroborating the number of transmembrane segments with DeepTMHMM and AlphaFold 2 Protein Database [24], [64], the accuracy of the CDS translation was confirmed rendering an initial data set of 130 CDS including GPR89A and GPR89B copies for Archaeplastida, Heterotrophic protist, and Metazoa (See details of the taxonomy sampling and the accession numbers on Supplementary Table 1). In instances where the nucleotide sequence was incomplete, we manually annotated the CDS by retrieving the corresponding genomic segments from Ensembl databases. Exon sequences were subsequently recovered using the exons of a canonical sequence from a closely related species for each case, in Geneious 2023.2.1 (https://www.geneious.com) (referred as “own annotation” in Supplementary Table 1).

For the phylogenetic analyses on Supplementary Figure 1, we looked for orthologs and paralogs of GPR89A and GPR89P among major groups of Angiospermae and Mammals (Supplementary Table 2). In this case, we retrieved GPR89P performing a BLASTN search on the Ensembl database with default settings against the whole genome of model mammals organisms, using the functional CDS of each species as a query sequence. When GRP89P was present, we manually annotated the CDS using the coding exons of the functional gene of each species in Geneious 2023.2.1 (https://www.geneious.com). Therefore, we obtained a new data set, containing 99 Mammalian sequences (including GPR89A, B and P). Sauropsid sequences were used as outgroups for mammals.

For both data sets, the Multiple Sequence Alignment (MSA) was performed using MAFFT [33], [34] on Geneious Prime 2023.2.1, allowing the software to choose the strategy to determine best fit. Then, we performed a visual inspection of the alignments. The conservative sites were obtained on MEGA11 [58], and Model Finder option was used by default on IQ-TREE 2 [32] to select the best fitting model for nucleotide substitutions (Fig. 1: SYM+I+G4, Supplementary Figure 1 C: GTR+F+I+G4). We further used IQ-TREE 2 [44] to conduct a maximum likelihood analysis in order to obtain the best phylogenetic tree. This analysis was run four times to corroborate the topology of this tree. Bootstrapping was performed to obtain the consensus from 1000 replicates. We report the best Maximum Likelihood phylogenetic tree, obtained with the following minimum Log-likelihood values: (Fig. 1: −105917.056, Supplementary Figure 1 C: −21678.235). The corresponding values for bootstrap support were obtained from the consensus tree.

4.2. Synteny assessment

From the Ensembl Compara database [27] we obtained the loci of the genes. From Genomicus 109.1 vertebrate database [46], and PLAZA5.0 Angiosperm database [63] we obtained the genomic context for predicted functional copies of GPR89. We examined upstream and downstream coding genes that are common to model organisms within these major groups. Specifically, for the case of mammals, we also look forward to retro-pseudogenes synteny but due to the fact that we were not able to find an evident pattern, synteny for pseudogenes is not reported.

4.3. Amino acidic frequency analysis

The 128 non-redundant sequences from the CDS were translated in silico, and then the amino acid sequences were aligned using MAFFT (L-INS-i strategy). In order to facilitate the visualization in Fig. 3C, we minimized the number and length of the gaps by removing protists and other sequences.

4.4. Multiple Structural Alignment

We selected a collection of human transporters, defined as such in the IUPHAR/BPS Guide to Pharmacology [1]. Sequences included members of the Solute Carriers (SLC), ABC transporters, human GPR89A, LIMR protein homologs and representatives of other transmembrane protein families, such as GPCRs, Calcium and Chloride Channels. Beta-barrel proteins (VDAC1 and TOM40) & globulins (myoglobin and alpha-hemoglobin) were chosen as structural outgroups. The amino acid sequence of these proteins was obtained directly from Uniprot, via an automatic Python script and manual accession. Transporter subunit OSTB was excluded from the analysis for its short length (128 a.a.). The final list comprises 482 sequences in total.

The structural alignment was carried out using TM-Vec [25]. This Deep-Learning tool predicts the TM-score (structural similarity) of proteins via their primary structure. We used the large CATH dataset-trained model [25] and the full amino acid sequences previously described. The TM-score prediction was turned to a symmetric distance matrix, where structural distance was defined as 1 - TM-score (0 = identical structures, 1 = complete dissimilar structures). Then we ran the UPGMA analysis on MAFFT online (https://mafft.cbrc.jp/alignment/server/phylogeny.html). We used the unrooted ultrametric tree. This tree file can be found in Newick format on our Supplementary Material.

For the pairwise structural alignment of transmembrane domains, experimental structures from representative neighbors of the previous analysis were recovered from the Protein Data Bank [6]. For close neighbors with no structure available, the AlphaFold Protein Structure Database structure was used. The soluble domains were removed from the pdb files, and the structures were aligned using TM-align [69]. Beta-barrel proteins (VDAC1 and TOM40) were selected as structural outgroups.

4.5. Structural predictions

All 3D models for the different GPR89 orthologs and paralogs we found present in Eukarya were obtained from the freely available AlphaFold Protein Structure Database [64]. We present human GPR89A as a representative structural model. All inspected models shared the general 9TM topology and the structural signatures discussed in this paper.

4.6. Figure rendering and accessibility

The trees were visualized using Figtree 1.4.4 (http://tree.bio.ed.ac.uk/software/figtree/) and vectorial silhouettes for phylogenetic figures were taken from https:/phylopic.org. These figures were colored and prepared on Inkscape 1.1.2.

MSA data was plotted with a custom Python script [8] and the sequence logo was made using LogoMaker [59]. For accessibility, Google’s Turbo color palette was chosen for labeling (http://research.google/blog/turbo-an-improved-rainbow-colormap-for-visualization/), and all figures were tested using a Color Blindness Simulator in Python (https://medium.com/@er_95882/colour-blindness-simulator-in-python-detecting-visualisation-problems-in-presentations-and-reports-71595526dc0e).

The degree of conservation of the different residues obtained from the MSA was mapped onto the structure of GPR89 using the web server ConSurf [21], [3]. The mapped structure was then exported to the molecular visualization software UCSF ChimeraX [22], [50], and the conservation score was manually annotated into the PDB file. The transmembrane helices were annotated according to the prediction of the DeepTMHMM server [24]. Coulombic surface potentials were calculated by ChimeraX.

CRediT authorship contribution statement

Luka Robeson: Writing – review & editing, Visualization, Methodology, Formal analysis. Brauchi Sebastian: Writing – review & editing, Writing – original draft, Project administration, Methodology, Formal analysis, Data curation, Conceptualization. Camila A. Quercia-Raty: Writing – original draft, Visualization, Investigation, Formal analysis, Data curation.

Declaration of Generative AI and AI-assisted technologies in the writing process

During the preparation of this work, the authors used ChatGPT by OpenAI to correct the language and improve the readability of the text. After using this tool, the authors reviewed and edited the content as needed and took full responsibility for the content of the published article.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Glossary

APC

Amino acid-polyamine-organocation superfamily

CDS

Coding DNA Sequence, i.e., the coding region of a gene

COLD1

Protein chilling tolerance divergence 1, aka GPR89

DMT

Drug/Metabolite Transporter superfamily

GPCR

G protein-coupled receptor

GPHR

Golgi pH regulator, aka GPR89

GTG

GPCR type G protein, aka GPR89

HP

alpha helix hairpin structure

HP_loop

Short turn connecting the HP helices.

LIMR

Lipocalin-1-interacting membrane receptor, aka LMBR1

LMBR

Limb region 1 protein homolog

MFS

Major Facilitator Superfamily

ML

Maximum Likelihood

MSA

Multiple sequences alignment

MYA

Million years ago

SLC

Solute carrier transporter

SAR

Stramenopiles, Alveolata and Rhizaria lineages. Major clade of eukaryotic organisms

UPGMA

Unweighted pair group method with arithmetic mean

Appendix A. Supplementary material

Supplementary material

Supplementary material

Supplementary material

Supplementary material

Supplementary material

Supplementary material