Structural and evolutionary characteristics of dynamin-related GTPase OPA1

OPA1 information acquisition

The information of human OPA1 gene, exons and isoforms were obtained from the NCBI and Ensemble databases (Zerbino et al., 2018). The other vertebrate OPA1 protein sequences were acquired from the NCBI reference sequence (RefSeq) database (O’Leary et al., 2016). All partial sequences and low quality proteins were excluded. The crystal structures of proteins from dynamin family were obtained from the PDB database (Rose et al., 2017).

Sequence analysis

The theoretical isoelectric points and the grand average hydropathicity were calculated by ExPASy server (Bjellqvist et al., 1993). The eight human OPA1 isoforms and all vertebrates OPA1 proteins were aligned with MAFFT (Katoh & Standley, 2013), respectively. Using the longest splice variant OPA1 in vertebrates, the evolutionary conservations of amino acid residue positions in the OPA1 sequences was measured by using ConSurf algorithm (Ashkenazy et al., 2016). The best evolutionary substitution model was used and calculation was based on the Empirical Bayesian paradigm. The sequence and modeled structure of human OPA1 were used to show the nine-color conservation grades. Sequence logos were generated as graphical representations of the multiple sequence alignments of the amino acids (Crooks et al., 2004). Phylogenetic analyses were conducted using MEGA version 7 by the bootstrap neighbor joining method (Kumar, Stecher & Tamura, 2016).

Structural analysis

I-TASSER was used to predict the structures of the human OPA1 protein (Yang et al., 2015). The structural model was refined by fragment-guided molecular dynamics (FG-MD) simulations at the atomic-level (Zhang, Liang & Zhang, 2011). The quality assessment of Ramachandran plot had been used to quantitatively assess the accuracy of protein structure predictions. The statistical data of the Ramachandran plot was calculated by PROCHECK (Laskowski et al., 1993). The protein structures were displayed by PyMol (Schrödinger, LLC). Pairwise structural differences between human OPA1 and other proteins from dynamin family were measured by TM-align (Zhang & Skolnick, 2005). The structural similarity of two protein structures was measured with the TM-score that had the value (0, 1) (Xu & Zhang, 2010). The higher the value was, the more similar the two aligned structures were. The two modeled structures of human OPA1 protein were taken as target 1 and target 2, respectively, to submit to PRISM protein–protein docking server for predicting possible interactions and how the interaction partners structurally connect (Baspinar et al., 2014).

Protein purification and blue native page

A human OPA1 construct (exons 6–28) was expressed in Escherichia coli BL21 (DE3) cells as N-terminal His₆-SUMO-tagged fusion proteins using the pET28a plasmid (His₆-SUMO-OPA1). Cells were grown in LB medium at 37 °C to an OD₆₀₀ of ∼0.6. Protein expression was induced by addition of 100 µM IPTG and cultures were incubated overnight at 16 °C. The cell pellet was resuspended in 10 mM Tris-HCl pH 8.0, 20 mM imidazole and 500 mM NaCl buffer. Cells were lysed using a microfluidizer (Microfluidics). After centrifugation at 12,000 rpm for 1 h at 4 °C, the soluble extract was filtered and combined with Ni-NTA agarose (GE Healthcare) in batch. After beads were packed into a column, the combination was washed with resuspension buffer supplemented with 40 mM imidazole, and protein was eluted with an imidazole gradient to a final concentration of 500 mM. Peak fractions were pooled, and the His₆-SUMO-tag was cleaved by ULP1 peptidase in the 4 °C overnight to generate the OPA1 (exons 6–28) protein. Then diluted the NaCl concentration to 50 mM and applied to Resource Q column (GE Healthcare). Peak fractions were concentrated and loaded to a Superdex 200 10/300 gel filtration column (GE Healthcare) using the buffer with 10 mM Tris-HCl, pH 8.0 and 150 mM NaCl. The protein was concentrated to ∼10 mg/ml prior to freezing. All the samples were analyzed by SDS-PAGE electrophoresis. The procedure for SDS-PAGE was based on a general protocol as previously described (Simpson, 2006). All the samples heated at 95 °C for 10 min and loaded onto a polyacrylamide gel that was consist of 5% stacking and 10% resolving gel, which was run at 80 V for 0.5 h, and then at 100 V for 1 h.

Blue native page was performed as described previously with slight modifications (Wittig, Braun & Schagger, 2006). 10 mg of the OPA1 (exons 6–28) protein was loaded into a blue native page gel that was consist of 3.5% stacking and 4–10% gradient separating gel. The cathode buffer (7.5 mM imidazole with pH 7.0, 50 mM tricine, supplemented with 0.02% w/v coomassie brilliant blue G250) and the anode (25 mM imidazole with pH 7.0) were chilled at 4 °C before used. Then, electrophoresis was started at 80 V for 30 min and adjusted to 120 V for 10 h.

Sequence analysis of human OPA1 isoforms

The human OPA1 gene contains eight spliced variants (Fig. 1A). Proteolytically processed by OMA1 and YME1L proteases in the domains were corresponding to exons 5 and 5b, containing the cleavage sites S1 and S2, respectively. In principle, each mRNA variant form can produce a long isoform and one or more short isoforms. However, the four isoforms (3, 5, 6 ,8) including exon 4b completely processed into s-forms. We confirmed that exon 4b-encoded peptide was hydrophobic while the others around were hydrophilic (Fig. 1B), which was consistent with the previous study (Elachouri et al., 2011). The hydrophobicity of exon 4b-encoded peptide might be helpful to recruit the OMA1 protease, thus, promoting the cleavage at site S1. Although exon 5b-encoded peptide was hydrophilic as a whole, there was a hydrophobic region just before site S2, hypothesizing that the hydrophobic segment can associate with YME1L protease to promote the cleavage at site S2.

We also noticed that the N-terminus of OPA1 s-forms were positively charged (Fig. 1C). The isoelectric point (pI) value of exon 5-encoded peptide was 6.8, while the pI value of the remaining peptide segment increased to 8.6 after being cleaved at site S1. Similarly, the pI value of exon 5b-encoded peptide rose from 6.8 to 9.5 after the site S2 was cut. This result revealed that the N-terminal of OPA1 s-forms all had positive charge no matter it was cleaved at site S1 or S2.

Structure characteristic of human OPA1

Taking the longest splice variant 8 including 30 exons as an example, we used the threading approach to model the three-dimensional structure of human OPA1 protein excluding MTS and TM regions (Fig. 2). According to quality assessment, the accuracy of the OPA1 protein structural models were acceptable (Fig. S1). The structure showed that OPA1 protein could be clearly divided into two regions, the N-terminal and C-terminal region. The N-terminal region (encoded by exons 3–8) was rich in alpha helix, which did not include specific domains. The difference among the 8 spliced variants of OPA1 protein, l-forms and s-forms was the region of N-terminal. The N-terminal region of l-forms contained 138–228 amino acid residues, while s-forms included only 104–134 residues.

The C-terminal region of OPA1 protein was a dense structure containing multiple domains and was identical for all isoforms. In the human OPA1 protein, each domain of the C-terminal region had perfectly corresponded with exons. The GTPase domain corresponded with exons 9–16, middle domain corresponded with exons 19–22 and a GED corresponded with exons 26–28. The peptide encoded by exons 17–18 formed a long helix that connected GTPase domain and middle domain. The exons 23–25 encoded a peptide, a PH domain, located between middle domain and GED.

The PH domain of classical dynamins is responsible for their interaction with negatively charged lipid membranes (Praefcke & McMahon, 2004). The pI analysis indicated that the pI value of PH domain of human OPA1 protein was 7.7, especially the pI value of exon 25-encoded peptide was as high as 9.5 whereas the pI value of the middle domain and GED next to PH domain were 5.0 and 6.8, respectively. Thus, the PH domain of OPA1 carried positive charge, might be involved in interaction with negatively charged phospholipid molecules.

Comparing the structure of the C-terminal region of human OPA1 protein with the nine-known structures of the dynamin superfamily proteins (Fig. 3), the results showed that OPA1 and other human dynamins had two common domains: a GTPase domain and an α-helical bundle. Interestingly, human OPA1 protein was similar to dynamin 1, dynamin 3, Drp1, MxA and MxB proteins, and the TM-score of all was higher than 0.5, although these proteins did not mediate membrane fusion like OPA1. Comparing OPA1 with mitofusin 1, atlastin 1 and altastin 3, which were involved in membrane fusion, the TM-score of all was minor than 0.3. In addition, similar to dynamin 1, dynamin 3 and Drp1, OPA1 had a PH domain, while other proteins involved in fusion process did not have. Significant structural differences indicated that OPA1 might have a unique mechanism for controlling inner membrane fusion.

Dimerization of human OPA1 protein

To determine the oligomeric form of OPA1, we expressed the common part of the eight spliced variants of human OPA1 (exons 6–28) protein in E. coli. The protein was detected by SDS-PAGE electrophoresis (Fig. S2). From the peak position of the OPA1 (exons 6–28) elution profiles of a S200 size-exclusion column, the result indicated that the molecular weight of the protein was about 170 kD, thus, we speculated the protein may exist as dimers in solution. Additionally, in blue native electrophoresis, we found nucleotide-free OPA1 formed stable dimers and multiple oligomers, indicating that the OPA1 molecule itself could be interacted and formed a stable multimer.

Further, we predicted the possible mode of dimerization of the C-terminal region by using the human OPA1 protein modeled structure. Protein–protein docking results showed that the C-terminal region of the human OPA1 protein could form two stable homodimers modes. In the first mode of homodimer, the GTPase domains of the two monomers were approximately 130 Å apart, and the dimeric interface involved in 23 pairs of interaction among 15 amino acid sites, 10 pairs of which were from the interaction between the linker domain and middle domain, 12 pairs of which were from the interaction between the middle domain and the GED, while one pair of sites was between the two same GED domains (Fig. 4A). In the second homodimer, the GTPase domains of the two monomers were approximately 100 Å apart, and the dimeric interaction interface involved in 33 pairs of interaction among 18 amino acid sites, 17 pairs of which were from the interaction between two middle domains, 12 pairs of which were from the interaction between the middle domain and the GED, and four pairs from the interaction between the two GED domains (Fig. 4B). These results indicated that, similar to dynamin 1 and MxA, OPA1 could assemble to dimers and higher-order oligomers via middle domain and GED, which was different from mitofusins that mediated membrane fusion by GTPase domains (Cao et al., 2017). Additionally, interactions between GTPase-GTPase domains had been reported to be essential for the function of dynamins, but our results showed neither of these two dimeric interfaces of nucleotide-free OPA1 proteins involved in the GTPase domain.

Conserved sites in OPA1 protein

Furthermore, we collected all the vertebrate OPA1 proteins by sequence similarity search. Totally, more than 900 of OPA1 protein sequences were discovered from 205 species including several spliced variants. The sequence alignments was shown in the form of sequence logo map (Fig. S3). Using the longest spliced variant of OPA1 protein sequence in each species of vertebrates, we analyzed conserved amino acid sites in the OPA1 proteins and displayed them through the modeled structure of the human OPA1 protein (Fig. 5A).

The results showed that the N-terminal region of OPA1 protein was significantly less conservative than the C-terminal region (Fig. 5B). For those sites with the highest conservation (grade 9), the N-terminal region only occupied 27% of all sites and 66% in the C-terminal region. For those sites with the lowest conservation (grade 1), 43% of all sites were in the N-terminal region and 10% in the C-terminal region. Specifically, in the N-terminal region, the exon 4 and exon 5-encoded peptides had the lowest conservation while the peptides encoded by exons 6–8 had a higher conservation. In the C-terminal region, the conservation of the GTPase domain and the linker encoded by exons 17–18 were the highest, while the PH domain was less conservative. These results indicated that the linkers encoded by exons 17–18 and exons 6–8, next to the GTPase domain, may be functionally important. A large swing in the linker during the GTPase cycle might cause the power-stroke that led to fusion. However, being the PH domain less conserved, it was speculated that its function might vary from species to species.

Disease-related sites in human OPA1 protein

Human OPA1 mutations have been associated to a large spectrum of neurodegenerations. 171 amino acid site variants in the human OPA1 had been analyzed by the locus-specific database dedicated to OPA1 (Data S1). The positions of the amino acid sites in human OPA1 protein were in accord with those of spliced variant 8. These variations include substitutions, duplications, deletions and insertions, but do not include synonymous mutations, nonsense mutations, and frameshift mutations. Of the 171 variants, 136 were associated with diseases, 10 were not related to diseases and 25 were variants of unknown significance. About 80% of disease-related variant sites were the highest conservative in the vertebrate OPA1 proteins (grade 9), while only 30% of the variant sites unrelated to disease had the corresponding values. Besides, 60% of variant sites that were not related to diseases had the lowest conservation (grade 1), while less than 10% of disease-related variant sites were the lowest conservative. Therefore, disease-related mutations in the OPA1 protein occurred mainly at sites with highly evolutionary conservation.

The N-terminal region of OPA1 protein occupied 34% of the total length, but the disease-related mutation sites located in this region accounted for only 15% of the total disease-related mutation sites in the OPA1 while the mutations unrelated to diseases in this region occupied 70% of the total mutation sites unrelated to diseases. The disease-related variant sites in the N-terminal region were concentrated in the exons 1–2 coding region (8 variants) and the exons 6-8 coding region (11 variants), while the variant sites unrelated to disease were concentrated on the exons 4, 4b and 5b (five variants). For the C-terminal region, more than 25% of the sites in the GTPase domain and the linker region encoded by exons 17–18 were found to have disease-related mutations, respectively. And about 20% of sites in the GED were also found to present disease-related variation. Therefore, most of the disease-related mutations in the OPA1 protein were located in the C-terminal region, while in the alternative splicing region, almost no disease-related mutation occurred.