Autosomal Dominant Gyrate Atrophy-Like Choroidal Dystrophy Revisited: 45 Years Follow-Up and Association with a Novel C1QTNF5 Missense Variant

We present a long-term follow-up in autosomal dominant gyrate atrophy-like choroidal dystrophy (adGALCD) and propose a possible genotype/phenotype correlation. Ophthalmic examination of six patients from two families revealed confluent areas of choroidal atrophy resembling gyrate atrophy, starting in the second decade of life. Progression continued centrally, reaching the fovea at about 60 years of age. Subretinal deposits, retinal pigmentation or choroidal neovascularization as seen in late-onset retinal degeneration (LORD) were not observed. Whole genome sequencing revealed a novel missense variant in the C1QTNF5 gene (p.(Q180E)) which was found in heterozygous state in all affected subjects. Haplotype analysis showed that this variant found in both families is identical by descent. Three-dimensional modeling of the possible supramolecular assemblies of C1QTNF5 revealed that the p.(Q180E) variant led to the destabilization of protein tertiary and quaternary structures, affecting both the stability of the single protomer and the entire globular head, thus exerting detrimental effects on the formation of C1QTNF5 trimeric globular domains and their interaction. In conclusion, we propose that the p.(Q180E) variant causes a specific phenotype, adGALCD, that differs in multiple clinical aspects from LORD. Disruption of optimal cell-adhesion mechanisms is expected when analyzing the effects of the point mutation at the protein level.


Introduction
Inherited diffuse choroidal dystrophies with peripheral onset include two well defined disorders, gyrate atrophy (MIM 258870) and choroideremia (MIM 303100) [1,2]. Gyrate atrophy is an autosomal recessively inherited disorder associated with mutations in the OAT gene and hyperornithinemia [1]. Choroideremia is an x-linked inherited disorder associated with mutations in the CHM gene [2]. Several simplex cases and few families which are not sufficiently defined to form another entity have been reported. Of these, autosomal dominant disorders include retinitis pigmentosa with predominant choroidal atrophy associated with a dominant RPE65 mutation [3] as well as a late-onset night blindness with trickle like macular dystrophy [4].
Autosomal dominant choroidal dystrophy has been rarely reported as diffuse choroidal dystrophy initially involving the posterior pole [5] and in few families with peripheral onset similar to gyrate atrophy. Some of the latter families were reported prior to the knowledge of the association of hyperornithinemia and gyrate atrophy [6]. One family was reported with exclusion of hyperornithinemia and the observed phenotype has been termed autosomal dominant gyrate atrophy-like choroidal dystrophy (adGALCD) [7].
Late-onset retinal degeneration (MIM 605670; LORD) is an autosomal dominantly inherited disorder with several distinct clinical signs, which include initially mid-peripheral temporal subretinal deposits developing into scalloped choroidal atrophy and progression towards the macula with choroidal neovascularization in later stages [8][9][10]. Most cases with LORD have been associated with the p.(S163R) variant in the C1QTNF5 gene, indicating that this particular variant is a founder mutation [11]. Few additional LORD families with different variants in the C1QTNF5 gene have been reported, including the recurrent p.(S163R) variant as well as additional missense variants (p.(P188T), p.(G216C), p.(P186S) and p.(S190W)) located in the gC1q domain of the C1QTNF5 protein [12,13] (Figure 1A,B).
The human C1QTNF5 gene encodes for a 25 kDa secretory protein expressed in adipose tissue [14], ciliary body of the eye and in the retinal pigment epithelium (RPE) [15]. C1QTNF5 is composed by three domains, namely the signal peptide (residues 1-15), the collagen domain (residues 30-98) and the gC1q domain (residues 103-243) [11,16] ( Figure 1A). Specifically, the collagen domain allows the association with the plasma membrane via collagen receptors, while the gC1q domain is responsible for RPE cell adhesion with other RPE cells and with Bruch membrane [17,18] (Figure 1C). As a member of the C1q family [19], C1QTNF5 protomers assemble into trimers due to the intertwining of their collagen domains. The association of the collagen domains brings the gC1q domains close to one another, thus favoring the packing of the hydrophobic "zipper" box at the protomer-protomer interface, ultimately leading to the constitution of a globular domain (or globular head). Additionally, six globular trimers furtherly assemble into a bouquet-like octadecameric multimer [18] via the association of their collagen triple helices ( Figure 1C,D).
The hydrophobic plateau constituted by residues 181-186 of each protomer [16] gives rise to an extended hydrophobic surface responsible for the binding of an RPE cell [15] to a neighboring cell or to the underlying Bruch membrane [18].
The purpose of the present study was to present long-term follow-up of the original adGALCD family [7] and of a distantly related second family to better characterize the clinical course and the retinal morphology and function in this disorder. In addition, we report an association of adGALCD with a novel C1QTNF5 missense variant and discuss possible genotype/phenotype correlation based on in silico molecular analysis. The three-dimensional structure of a C1QTNF5 protomer is shown in cartoon, with β-strands colored according to panel A, residues S163, P186, P188, S190 and G216, whose mutations are late-onset retinal degeneration (LORD)-associated, are shown in grey sticks and labeled. Residue Q180, whose E variant is autosomal dominant gyrate atrophy-like choroidal dystrophy (adGALCD)-associated in this work, is represented in red sticks and labeled. N atoms are colored in blue, O atoms in red, H atoms in white. (C) Schematic representation of the C1QTNF5-mediated interactions of retinal pigment epithelium (RPE) cells with neighboring cells and with Bruch membrane. The collagen domain binds to collagen receptors on the membrane of an RPE cell, while the globular heads of the bouquet-like C1QTNF5 multimer can interact with the globular heads of another C1QTNF5 multimer bound to an adjacent RPE cell or with Bruch membrane. The structure of the two globular heads framed in green is represented in panel D. (D) Three-dimensional structure of the two trimers constituting the globular heads of C1QTNF5 belonging to different RPE cells and involved in cell adhesion based on available X-ray crystallographic data. Protein structure is shown in cartoon, chain A of is colored in red, chain B in orange, chain C in green, chains A, B and C of the second molecule are represented in cyan, blue and purple, respectively. Residues Q180 are shown as red sticks and labeled.

Clinical Evaluation of Family Members
Two families were included in this study. For family BD 35 with three affected males, a father and his two sons, initial findings have been reported previously [7]. Ornithine blood levels were normal in all three patients. The mother of both sons (III:1) had normal findings at 68 years of age. The parents of patient III:2 died prior to 45 years of age without reported eye problems. The maternal grandmother of patient III:2 was reported to be affected. For one of the sons (IV:2), regular follow-up visits were undertaken between 38 and 63 years of age. In family ADRP 386, three affected siblings (IV:1, IV:2 and IV:3) were examined between 50 to 58 years of age. Further family members in three previous generations were reported to be affected ( Figure 2) but were not available for clinical examination. Onset of disease with peripheral atrophic lesions was documented in patient IV:2 from family BD 35 at 18 years of age and in patient IV:2 from family ADRP 386 at 24 years of age (Table 1). Functional problems were noted later at about 40 years of age (patients III:2 and IV:1 from family BD35; patients IV:2 and IV:3 from family ADRP 386) with the onset of night blindness. In the same patients, subjective problems due to visual field defects were noted at about 50 years of age. Snellen visual acuity ranged between 0.5 to 1.0 in less progressed stages and was reduced to 0.1 to 0.4 when the residual visual field was smaller than 10 degrees (Table 1). Refractive errors were most mildly hyperopic with moderate astigmatism.
Cataract surgery was performed in one of the patients by one of the authors (U.K.) without complications. In this patient (IV:2 from family BD 35) as well as in all other patients without prior cataract surgery no long anterior lens zonules were noted (Table 2). ERG: full-field electroretinography, mfERG multifocal ERG, IOL: intraocular lens, OU: both eyes, OD: right eye, ND: not done. For the three patients examined more than once, numbers in brackets indicate age of patient at the time of the examination.
Retinal findings started with peripheral and peripapillary chorioretinal atrophic lesions. The lesions presented as large atrophic areas which were sharply demarcated by small areas of preserved retinal pigment epithelium (RPE) between lesions (Figures 3 and 4). Thus, the findings resembled more the appearance of gyrate atrophy than choroideremia. During progression, the peripheral lesions circumferentially encroached towards the fovea, reaching first the inferior vascular arcades latest at 50 years of age (Figures 3 and 4). In contrast, the progression of the peripapillary atrophy from the disc towards the fovea was slow ( Figure 4). In late stages (patients III:2 from family BD 35 and IV:3 from family ADRP 386) only the foveal area was preserved. Pigmented lesions were either absent or minimal.  Fundus autofluorescence (FAF), near-infrared autofluorescence (NIA) and optical coherence tomography (OCT) were normal in the non-affected regions with a sharp border towards the affected areas (Figures 4 and 5). Affected areas showed absence of FAF, NIA and a complete loss of outer retinal structures in the OCT indicating loss of photoreceptors and RPE. OCT-angiography showed absence of choriocapillaris and few remaining large choroidal vessels with reduced flow in the affected areas (patient IV:2 from family BD 35, Figure 6). Progression of disease could be documented over time (Figure 4). Only in patient IV:3 from family ADRP 386 cystoid macular edema was observed ( Figure 5).    In patient IV:2 from family BD 35, full-field ERG was reported as slightly abnormal with an unspecified method at 18 years of age and the EOG showed a severely reduced light rise. At 38 and 42 years of age, ERG amplitudes according to ISCEV standards were markedly reduced. At 50 years of age, only residual responses were recordable in the ERG. Similarly, no measurable or small residual responses could be recorded in the father (patient III:2) as well as in all patients of family ADRP 386 between 50 to 58 years of age. The mfERG, recorded in three patients, showed centrally preserved responses with reduced amplitude but normal implicit time corresponding to areas with preserved retina and absent responses in atrophic areas.

Identification of a Novel Missense Variant in C1QTNF5
The self-reported family history of family BD 35 was indicative of an autosomal dominant pattern of inheritance for family BD 35 while it was consistent with an autosomal dominant inheritance mode for family ADRP 386 with affected individuals in four consecutive generations.
Patients III:2 and IV:2 from family BD 35 as well as patients IV:1 and IV:3 from family ADRP 386 underwent diagnostic genetic testing by whole genome sequencing. Putative pathogenic variants were only identified in the C1QTNF5 gene, namely a novel missense variant (c.538C > G/p.(Q180E)), which was found in heterozygous state in all family members that were available for genetic testing ( Figure 2). The variant is absent from the population database gnomAD (https://gnomad.broadinstitute.org/), indicating that it is very rare. In silico predictions of the p.(Q180E) variant using various online tools were discordant. It was predicted to be possibly damaging (PolyPhen2), disease causing (MutationTaster), neutral (PROVEAN) or tolerated (SIFT) (Appendix A Table A1). Sequence alignment of C1QTNF5 and orthologous proteins in other species showed that p.(Q180E) is fully conserved in 176 vertebrate species (Appendix A Table A2). Accordingly, the high level of evolutionary conservation is reflected by high scores obtained with PhyloP, CADD and FATHMM (Appendix A Table A1).
As far as could be established from the self-reported family histories the two families are either not or very distantly related (i.e., not being aware of their relatedness). In order to distinguish between the alternative hypotheses of recurrent mutation and identity by descent, we established haplotypes using SNP data obtained by whole genome sequencing. In addition, two annotated microsatellites in the vicinity of C1QTNF5 were analyzed. Appendix A Figure A1 shows that p.(Q180E) is associated with the same haplotype on all disease alleles analyzed. This is consistent with the mutant allele being identical by descent.
Patients of both families presented with similar clinical findings and segregated the same variant. This suggested an association with the novel p.(Q180E) variant with adGALCD.

Molecular Modeling of C1QTNF5 and Analysis of the Structural and Functional Effects of the Missense Mutation p.(Q180E)
The autosomal dominant nature of adGALCD and LORD implies that, in the pool of expressed C1QTNF5 proteins, half of the molecules will carry the Q180E mutation, whereas the other half will be unaffected (wild-type, WT). This fact, together with the supramolecular organization of C1QTNF5 ( Figure 1) opens up a very complex scenario, in which both WT and mutated protomers can form assemblies, whose protomers can either be mutated or not. We considered all these possibilities in our molecular modeling analyses.
The Q180E amino acid substitution could in first instance perturb the stability of each C1QTNF5 protomer. To evaluate such a possibility, Gln180 was replaced with the negatively charged Glu. The apparent change in free energy of folding with respect to the WT (∆∆G f app ) was calculated for the single protomer, resulting in a 2.56 kcal/mol destabilization. Replacement of superficial protein charges may perturb long-range electrostatics in proteins, with significant effects on protein stability [20]. A first effect of the Q180E mutation could therefore be the destabilization of the tertiary structure of C1QTNF5.
Residue Q180 is the last residue of β-strand E ( Figure 1B) and it is located between the hydrophobic plateau and the hydrophobic "zipper" box. The analysis of the protomer surface highlighted that residue Q180 is fundamental for the reciprocal interaction among the three protomers constituting the globular head of C1QTNF5. Indeed, the aggregation score provided by AggScore (11052) for residue Q180 resulted to be the highest value reported for the entire protein sequence, suggesting that mutations involving Q180 may have a detrimental effect on the association of the trimeric globular head and its stability. Considering the whole globular head, the substitution of a polar uncharged residue with a negatively charged residue close to the highly hydrophobic "zipper" interface altered the partial charge distribution and therefore the electrostatic potential on the protomer surface ( Figure 9A,C). Such alteration resulted in an electrostatic repulsion between the negative charges of two interacting protomers ( Figure 9B,D), affecting both the stability of the complex (∆∆G f app = 30.5 ± 1.0 kcal/mol, Table 3) and the affinity of the protomers for one another ∆∆G b app = 12.2 ± 5.5 kcal/mol, Table 3). On the other hand, the presence of two or even three mutations within the same trimer ( Figure 9B,D) further increased the electrostatic repulsion, thus causing a destabilization of the complex (∆∆G f app = 64.8 ± 1.6 and 101.12 ± 0.04 kcal/mol, respectively, Table 3) and a decrease in affinity of the protomers (∆∆G b app = 27.7 ± 6.6 and 44.7 ± 1.4 kcal/mol, respectively, Table 3). It is worth noting that the contribution of coulombic repulsion to the apparent binding free energy among protomers in the trimer impacted both affinity and stability. On the contrary, the contribution of solvation of the mutated trimer decreased the apparent affinity of the protomers (∆∆G b app S = 9.4 ± 1.2 kcal/mol, Table 4) but it was found to increase the stability of the ternary complex (∆∆G f app S = −45.5 ± 1.2 kcal/mol, Table 4). In summary, the Q180E substitution was predicted to generate a significant electrostatic repulsion among the protomers forming up a globular head, whose effects could propagate at longer distance, considering the significantly changed electrostatic potential generated by a trimer (Figure 10), which could affect the recognition between two adjacent globular heads approaching each other. Indeed, on a larger scale, the Gln-to-Glu substitution in position 180 also affected the interaction between adjacent globular heads belonging to different cells, as shown by the positive ∆∆G b app and ∆∆G f app shown by the mutants (Table 3). Similar to the interaction among protomers of the same globular head (Figure 10), the top-to-top interaction between globular heads was increasingly destabilized by the Q180E mutation proportionally to the number of mutations present in the two trimers. Indeed, the remapping of electrostatic potential involving globular heads' interface resulted in an increase of ∆∆G b app from 9.0 kcal/mol in the presence of 1 protomer carrying the mutation to approximately 60 kcal/mol in the presence of the homozygous variant (Table 3), while ∆∆G f app increased from 26.7 to 187.8 kcal/mol under the same conditions (Table 3). Interestingly, the presence of mutations was favorable in terms of electrostatic interactions between the two globular heads from both the affinity and stability standpoints, as shown by the negative ∆∆G values for both binding and folding (∆∆G b app C and ∆∆G f app C columns, Table 3). On the other hand, such positive coulombic contribution was counterbalanced by the larger, unfavorable increase in ∆∆G of solvation (∆∆G b app S and ∆∆G f app S columns, Table 3) resulting from the addition of a negatively charged residue to a highly hydrophobic patch.

Discussion
Autosomal dominant gyrate atrophy-like choroidal dystrophy (adGALCD) appears to be a distinct entity that has been described so far in three families. Two of them are included in this manuscript, while images consistent with the ocular phenotype have already been reported for another family without detailed information [6]. As all three families were observed in Germany, it might well be that they are all related. This assumption is supported by haplotype analysis for the two families reported in the present study.
Compared to the clearly defined entities of gyrate atrophy and choroideremia, the onset is later and the rate of progression in adGALCD is slower, e.g., reading ability may be preserved up to 70 years of age. The chorioretinal lesions are more clearly demarcated compared to choroideremia. In contrast to gyrate atrophy, early onset of cataract formation and hyperornithinemia were not observed, and cystoid macular edema was only seen in one patient.
The finding of a putative pathogenic variant in the C1QTNF5 gene supports the inclusion of adGALCD into the spectrum of LORD. However, the clinical findings differ in several aspects between adGALCD and LORD ( Table 4). The onset of retinal alterations is earlier and more peripheral in adGALCD compared to LORD [21][22][23][24][25][26]. In contrast to adGALCD with peripheral retinal changes present at least at the end of the second decade of life, patients with stage 1 LORD are reported to show a normal retina until 40 years of age, while long anterior lens zonules and iris atrophy may already by present [26,27]. The peripheral lesions in adGALCD are associated with early peripheral visual field loss, while the peripheral visual fields remain intact in LORD even in progressed stages [25]. Characteristic findings of stage 2 LORD are subretinal deposits which present in a pseudodrusen-like pattern with sub-RPE material and a thickening between RPE and Bruch membrane in the OCT [9,11,[21][22][23][28][29][30]. These pseudodrusen-like lesions first develop in the temporal middle periphery progressing to scalloped chorioretinal atrophy as well as pigmentary changes, all of which were not observed in the adGALCD families. Although scalloped chorioretinal atrophic lesions have been observed in adGALCD and LORD [4,9,13,[21][22][23][24][25][31][32][33][34], scalloped lesions in LORD are located in the mid-periphery and are much smaller in size compared to those observed in the patients in the present and previous study [7]. Stage 3 LORD is characterized by atrophic macular alterations or choroidal neovascularizations [11,24,27,34,35], whereas in adGALCD neovascular lesions were not observed and the fovea was preserved until the final stage. As such, the phenotype in the families presented in this study much more resembles gyrate atrophy than LORD. Distinct characteristics of LORD in FAF and OCT were not observed in adGALCD: no sub-RPE deposits in the OCT and no structural alterations prior to RPE loss in the FAF were detected [25,26,29,33,36].
In addition, retinal function was earlier and more severely affected in adGALCD as indicated by earlier and markedly reduced ERG and multifocal ERG responses. Normal or well-preserved ERG has been reported up to 60-67 years of age in LORD [23,33,37].
A frequent additional finding in LORD are long anterior lens zonules [12,22,24,25,28,32,38,39], although they are not present in all LORD patients [13] and may also occur in the absence of LORD [40]. Long anterior lens zonules were not observed in the patients in this study. In addition, no problems during cataract surgery were observed or reported in contrast to LORD [41].
A similarity between adGALCD and LORD is the development of night blindness at about 40 years of age in most patients [26]. This appears unexplained for both disorders, while at that age still sufficient rod-rich areas along the vascular arcades appear to be structurally normal in the patients presented here as well as in LORD. As previously suggested [26], C1QTNF5 gene variants may affect rod function with an additional mechanism which still has to be defined.
The C1QTNF5 gene product is present in the lateral and basal membrane of retinal pigment epithelial cells and the ciliary body [42]. It interacts with HTRA1 in mice and affects extracellular matrix turnover, which might explain the development of subretinal deposits in LORD [43][44][45]. Following the resolution of X-ray structures of the globular domain of C1QTNF5 [16,18], molecular models were built to explain the mechanism of C1QTNF5-mediated cell adhesion among RPE cells and RPE-Bruch membrane. The typical bouquet-like octadecameric organization of C1QTNF5 ensuring the appropriate scaffold for cell-cell communication can be reached only by specific molecular recognition between C1QTNF5 subunits within the same and with surrounding globular head domains. Point mutations of key amino acids involved in the protein-protein recognition process can therefore lead to destabilization of these supramolecular assemblies. In case of autosomal dominant diseases, homo-and hetero-complexes can be formed by the expressed WT and mutant proteins, thus generating a complex molecular scenario that has emerged recently, with specific effects depending on the nature of the complex [46,47]. In the case of C1QTNF5, the scenario is extremely complex, with up to 64 possible assemblies between wild-type and individually mutated protomers. As a general conclusion, our modeling study predicts that, independently on the number of mutated protomers, the p.(Q180E) substitution will perturb both the stability of the globular heads and the recognition between adjacent trimers. This would result in a perturbation of the molecular scaffold linking to adjacent cells, ultimately disrupting the native cell-adhesion mechanisms.

Patient Enrollment and Retrieval of Blood Samples
Patients were recruited and clinically examined either at the Eye Hospital of the University of Tübingen (family ADRP 386) or at the AugenZentrum Siegburg (family BD 35). Genomic DNA of patients was extracted from peripheral blood using standard protocols. Samples from all patients were recruited in accordance with the principles of the Declaration of Helsinki and were obtained with written informed consent accompanying the patients' samples. The study was approved by the institutional review board of the Ethics Committee of the University Hospital of Tübingen under the study numbers 349/2003V and 116/2015BO2.

Clinical Examination
Family BD 35 included four members who have been examined, three of them affected males including the father and both of his sons (see pedigree depicted in Figure 2). Details of these examinations have been reported previously [7]. The father (III:2) was lost to follow-up due to stroke-associated death and one son (IV:2) due to a brain tumor. The second son (IV:2) was re-examined at different time intervals between 38 to 63 years of age. In addition to basic ophthalmologic examination, he underwent visual field testing, fullfield electroretinography (ERG) and multifocal ERG (mfERG) recording, wide-angle fundus autofluorescence (FAF), near-infrared autofluorescence (NIA), spectral domain OCT (OCT) examinations and OCT-angiography. Family ADRP 386 comprises three siblings. All three siblings underwent in addition to basic ophthalmologic examination visual field testing, ERG and mfERG recording, FAF and OCT examination. Please note that the phenotype in family ADRP 386 was initially described as autosomal dominant retinitis pigmentosa (ADRP) but that the original family ID was kept after refinement of the clinical diagnosis based on examination of more family members and longer follow-up.
Electrophysiologic examinations and non-invasive retinal imaging were obtained as described previously after medical dilatation of the pupil [48,49]. Full-field ERG and mfERG were measured according to the most recent ISCEV standards [50,51] at the time of recording. FAF and NIA were obtained with a confocal scanning laser ophthalmoscope (Heidelberg Retina Angiograph 2, Heidelberg Engineering, Heidelberg, Germany) using 30 • and 50 • lenses. Volume and single scan SD-OCT as well as OCT-angiography were performed with a Spectralis OCT (Heidelberg Engineering, Heidelberg, Germany).

Diagnostic Genetic Testing
Patients III:2 and IV:2 from family BD 35 and patients IV:1 and IV:3 from family ADRP 386 underwent diagnostic genetic testing by means of whole genome sequencing. Details of sequencing and variant classification have been described previously [52]. Segregation analysis in family members was performed using conventional Sanger sequencing.

Haplotype Analysis
Haplotype analysis was performed in all available family members by genotyping two microsatellites in the vicinity of C1QTNF5 (D11S614 and D11S4129). Primers for microsatellite amplification were as follows: D11S614-forward: 5 -ACAGACCCACCAGGACTAT-3 , D11S614-reverse: 5 -CCCGGATGTCTGCAAGGTGG-3 ; D11S4129-forward: 5 -ACAGCGA CCACATCTCCTGC-3 , D11S4129-reverse: 5 -GGCCACTGCCCTTACCATCA-3 . Genotyping of microsatellites was performed as described previously [53]. Extended haplotype analysis was performed in patients III:2 and IV:2 from family BD 35 and patients IV:1 and IV:3 from family ADRP 386 using informative genotype data of 41 SNPs spanning a physical region of 2 Mb interval surrounding C1QTNF5. Genotypes were considered informative if they could be unambiguously assigned to a haplotype. The genotype data were obtained from the whole genome sequencing datasets.

Molecular Modeling of C1QTNF5 Protein Structure and Analysis of the p.(Q180E) Effects on Stability and Affinity
All molecular modeling analyses were performed within the environment of the chemical simulation software Maestro/Bioluminate (Schroedinger, New York, USA). Human C1QTNF5 trimer was modeled using as the starting structure the PDB file with entry 4F3J [16], which provided the highest resolution (1.34 Å) among the available structures, as well as the symmetric operators for reconstituting the trimeric biological unit. The hexameric assemble of two globular heads was built superimposing the trimers to the top-to-top interacting protomers of PDB file 4NN0 (resolution 1.42 Å) [18]. The trimeric globular heads were reconstituted using the symmetric operators previously included in PDB file 4F3J [16]. This allowed us to evaluate the effects of the pathogenetic point mutation within each individual protomer making up a globular head (trimer), as well as a couple of globular heads from two different cells, interacting with each other (interacting trimers, Figure 1C,D).
Protein structures were prepared according to the pipeline of the Protein Preparation tool, briefly consisting of the assignment of bond orders using Chemical Components Dictionary database (www.pdb.org, wwPDB Foundation, Piscataway NJ, USA), addition of H atoms, selection of the most probable rotamer and deletion of water molecules closer than 3.5 Å only to other water molecules. Structure refinement included sampling of water orientation, usage of crystal symmetry to optimize H-bonds with the other protomers and prediction of the protonation states of ionizable residues at pH 7.5 by PROPKA prior to H-bond assignment and optimization. Finally, protein structure was minimized using the OPLS3e forcefield (Schroedinger, New York, NY, USA) until the Root-Mean Square Displacement of the heavy-atom reached 0.3 Å.
The role of residue Gln180 (in the WT protein) in protomer association was investigated by the "Protein Surface Analysis" tool. A well-performing index based on the analysis of the distribution of hydrophobic and electrostatic patches on the surface of the protein, the AggScore index [54], was calculated to estimate aggregation propensity of the protomer.
The electrostatic potential surface was calculated by the "Poisson-Boltzmann Electrostatic Potential Surface" tool, by setting solute dielectric constant to 1, solvent dielectric constant to 80, solvent radius to 1.4 Å, temperature to 298 K, grid extension to 5 Å.
The "Residue Scanning" tool was employed to introduce the p.(Q180E) mutation in any combination of both the three protomers constituting a globular domain and the two interacting globular domains, thus resulting in 7 and 63 combinations, respectively. Each mutagenized structure was subjected to automatic selection of the most probable rotamer and energy minimization using the same parameters as those employed for the wild-type (WT) model.
The Molecular Mechanics/Generalized Born and Surface Area Continuum Solvation (MM/GBSA) method was used to predict the effects of residue mutation on protomer stability and affinity, in terms of relative changes compared to the WT, by using the specific thermodynamic cycle. It should be noticed that the free energy computed by the method is based on the MM force field without explicit contributions from conformational changes, therefore the obtained free energy values (∆∆G app in Table 2, expressed in kcal/mol) should be taken as "apparent" values and considered as approximate indexes, which are useful in comparisons, rather than rigorously defined thermodynamic quantities. Protomerprotomer affinity was calculated for each of the three protomers against the other two protomers. The resulting 21 combinations were pooled for the number of mutations present in the trimeric assembly and the variations in the apparent binding affinity (∆∆G b app ) and apparent stabilities (∆∆G f app ). Presented in Table 3 are the average ± standard deviation of the respective group, whose size is also reported. Trimer-trimer affinity was calculated between the two globular heads, the 63 combinations obtained by "Residue Scanning" were grouped by the number of variants in the hexameric assembly with the same approach as that employed for protomer-protomer affinity. The contribution of electrostatics and solvation to the relative ∆∆G app values for both binding and stability was specifically reported in Table 3 and discussed.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author. The data are not publicly available due to health system data regulations.

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.