Non-Ceruloplasmin Copper Identifies a Subtype of Alzheimer’s Disease (CuAD): Characterization of the Cognitive Profile and Case of a CuAD Patient Carrying an RGS7 Stop-Loss Variant

Alzheimer’s disease (AD) is a type of dementia whose cause is incompletely defined. Copper (Cu) involvement in AD etiology was confirmed by a meta-analysis on about 6000 participants, showing that Cu levels were decreased in AD brain specimens, while Cu and non-bound ceruloplasmin Cu (non-Cp Cu) levels were increased in serum/plasma samples. Non-Cp Cu was advocated as a stratification add-on biomarker of a Cu subtype of AD (CuAD subtype). To further circumstantiate this concept, we evaluated non-Cp Cu reliability in classifying subtypes of AD based on the characterization of the cognitive profile. The stratification of the AD patients into normal AD (non-Cp Cu ≤ 1.6 µmol/L) and CuAD (non-Cp Cu > 1.6 µmol/L) showed a significant difference in executive function outcomes, even though patients did not differ in disease duration and severity. Among the Cu-AD patients, a 76-year-old woman showed significantly abnormal levels in the Cu panel and underwent whole exome sequencing. The CuAD patient was detected with possessing the homozygous (c.1486T > C; p.(Ter496Argext*19) stop-loss variant in the RGS7 gene (MIM*602517), which encodes for Regulator of G Protein Signaling 7. Non-Cp Cu as an add-on test in the AD diagnostic pathway can provide relevant information about the underlying pathological processes in subtypes of AD and suggest specific therapeutic options.


Introduction
Alzheimer's disease (AD) is the most common neurodegenerative condition associated with dementia in the elderly. Hallmarks of AD are extracellular aggregates of amyloid-β (Aβ), and intracellular neurofibrillary tangles made of hyperphosphorylated tau (p-Tau) [1,2]). Recent programs aiming at the development of non-invasive AD biomarkers reached important achievements by employing plasma Aβ 40-42 (Neurology 3-Plex A-performed with SIMOA technology [3]), plasma phosphorylated tau 217 and tau181, biomarkers of the Cu panel underwent whole exome sequencing (WES) analysis to screen for potential causative gene variants. During the trial of this pilot clinical diagnostic path, we identified a patient with an RGS7 stop-loss gene variant associated with CuAD. Results also support the use of non-Cp Cu assessment as a stratification add-on test for the Cu-AD subtype, based on the characterization of the cognitive profile.

Results
During the trial period of our pilot diagnostic pathway, 30 patients were referred to our site. Twenty-eight received the diagnosis of AD (IGW2-criteria [28]). We first compared AD patients and healthy controls (CTRL) for the biomarkers of the Cu panel. We then used the non-Cp Cu biomarker as an add-on test to stratify the AD patients into two subgroups on the basis of a 1.6 µmol/L cut-off: we then compared the two AD subgroups for the outcomes of the extensive neuropsychological battery. One of the patients of the CuAD subgroup, a 76-year-old woman, showed significantly abnormal levels in more than two Cu biomarkers; she then underwent WES and was found to have an RGS7 stop-loss gene variant.

Alzheimer's Disease and Healthy Controls Comparison for Cu Panel
The AD and CTRL groups differed in sex and age, with the healthy group being younger than the AD group and 76% of the CTRL group being women (Table 1). For this reason, the analyses were adjusted to account for the potential confounding effect of age and sex. Serum Cu concentration was different between the two groups (Welch's t-test), being higher in AD patients than in CTRL. The two groups differed significantly in non-Cp Cu and Cu:Cp ratio values, which were higher in AD than in CTRL; ceruloplasmin biomarkers did not differ between the two groups ( Figure 1 and Table 2).

AD Subgroups Stratification on the Basis of the Add-On Non-Cp Cu Test
The AD population was stratified on the basis of the non-Cp Cu cut-off of 1.6 µmol into two subtypes (Table 3). Among the CuAD patients, one had severe Cu abnormalities, thus was studied apart as a case report (see Section 2.3).  Figure 1. Comparison of the Cu panel between CTRL and AD subjects. Data are mean and standard deviation of Cu, non-Cp Cu, ceruloplasmin, ceruloplasmin, and ceruloplasmin-specific activity (* p < 0.05). Ceruloplasmin activity is additional quality control of ceruloplasmin measurement as extensively discussed elsewhere [11]. Ceruloplasmin-specific activity (eCp:iCp) is the enzymatic activity per mg of Cp concentration. Raw data are shown, not adjusted for subjects' differences in sex and age.

AD Subgroups Stratification on the Basis of the Add-on Non-Cp Cu Test
The AD population was stratified on the basis of the non-Cp Cu cut-off of 1.6 µmol into two subtypes (Table 3). Among the CuAD patients, one had severe Cu abnormalities, thus was studied apart as a case report (see Section 2.3). Table 3. Comparison of AD (non-Cp Cu < 1.6 µmol/L) and CuAD (non-Cp Cu ≥ 1.6 µmol/L).  Comparison of the Cu panel between CTRL and AD subjects. Data are mean and standard deviation of Cu, non-Cp Cu, ceruloplasmin, ceruloplasmin, and ceruloplasmin-specific activity (* p < 0.05). Ceruloplasmin activity is additional quality control of ceruloplasmin measurement as extensively discussed elsewhere [11]. Ceruloplasmin-specific activity (eCp:iCp) is the enzymatic activity per mg of Cp concentration. Raw data are shown, not adjusted for subjects' differences in sex and age. Subjects in the CuAD subgroup had higher values of Cu, ceruloplasmin, ceruloplasmin activity, and specific ceruloplasmin activity than patients in the AD subgroup. In addition, non-Cp Cu was about three times higher in the CuAD group than in subjects in the AD subgroup (1.14 ± 0.49 in AD vs. 3.29 ± 0.94 CuAD, p < 0.001). The two AD subgroups had the same years of disease duration and the same degree of severity (as indicated by the MMSE score, which did not differ between the two groups), but differed in neuropsychology test scores (Table 4), with FAB and CDT scores being lower in the CuAD subgroup (Table 4 and Figure 2  vs. AD) and neuropsychological tests (within-subject variable) was significant (p = Specifically, post-comparisons showed that CuAD patients performed significantly on FAB (B-H adjusted p = 0.006) as well as on CDT (B-H adjusted p = 0.024), and bord on MMSE (B-H adjusted p = 0.055).   We performed an exploratory correlation study between neuropsychological test results and Cu panel biomarkers. Analyses were performed on all AD patients without considering subgroup stratification (Table 5). Some tests of the neuropsychological battery had trends of correlation with Cu biomarkers, namely non-Cp Cu and ceruloplasminspecific activity (Table 5). After applying Benjamini & Hochberg's [34] correction to control for the false discovery rate, the only worse scores on the FAB scale correlated with abnormal values of all Cu biomarkers (Table 5).

Case Report of One Patient in the CuAD Subgroup
The patient was referred to our specialty memory clinic at the Experimental Alzheimer Center, Fatebenefratelli Roman Province, Rome, Italy at the age of 76 with a diagnosis of AD according to IWG-2 criteria [28] and a 14-year history of the disease.
She was positive for AD biomarkers in the CSF, namely decreased levels of Aβ40/42 and increased levels of P-tau or T-tau and had a positive FDG-PET (Figure 3), compatible with the diagnosis.
Her past medical history revealed no relevant pathological conditions. She had no surgical history. She had no history of smoking, alcohol, or drug use. There was no firstdegree family history of presenile dementia. Laboratory screening for markers of liver functions, magnesium, and phosphorus was normal, while iron metabolism markers and lithium were slightly below the normal reference range (Table 6).  Her past medical history revealed no relevant pathological conditions. She had no surgical history. She had no history of smoking, alcohol, or drug use. There was no firstdegree family history of presenile dementia. Laboratory screening for markers of liver functions, magnesium, and phosphorus was normal, while iron metabolism markers and lithium were slightly below the normal reference range (Table 6).  After completing the AD diagnostic pathway, the patient underwent a Cu panel screening and an extensive neuropsychological battery as per the add-on test Cu diagnostic pathway. The Cu study revealed more than two Cu biomarker abnormalities, namely non-Cu Cp (higher than 1.6 µmol/L) [11,36], Cu (lower than 11 µmol/L [37]), and ceruloplasmin activity that was higher than 140 IU [38] (Table 6). On this basis, she underwent WES analysis to screen for potential causative mutations.

Analysis and Identification of a Candidate Variant in the CuAD Patient
The initial Next Generation Sequencing panel of 74 genes (see method section) did not identify any pathogenic, likely pathogenic, or unknown significance variants. No common functional polymorphisms were found in the ATP7B gene.
The DNM2 variant shows an allelic frequency of 2:113.728 alleles for the European population according to GnomAD exomes, which corresponds to around one carrier in 29,000, moderately compatible with AD disease with late-onset, and highly compatible with AR disease frequency carrier. All in silico tools predict a possible damaging effect. DMN2 is associated with myotubular myopathy type 1 (MIM #160150) with an autosomal dominant trait of inheritance, Charcot-Marie-Tooth disease, axonal type 2M (CMT2M), Charcot-Marie-Tooth disease, dominant intermediate B (CMTDI1) (MIM #606482), and with Lethal congenital contracture syndrome type 5 (MIM #615368) following an autosomal recessive fashion. The patient showed no neuromuscular or polyneuropathy clinical signs.
The RGS7 variant presents an allelic frequency of 4:68.002 alleles for the European population according to GnomAD exomes, with no homozygous cases, which corresponds to around one carrier in 9000, doubly compatible with AD disease with late-onset and highly compatible with AR disease frequency carrier. To date, no MIM phenotype has been linked to RGS7 pathogenic variants ( Figure 4). The combined Annotation Dependent Depletion (CADD) score was 22.1, predicting this variant to be in the top 1% for deleteriousness [39].

In Silico Analysis of RGS7 Stop-Loss Mutation
Change at the 496th position in a wild-type sequence results in the loss of a stop codon and changes it into arginine (Arg) and extends the stretch of the protein sequence by 19 amino acids. In order to find similarly related sequences, Protein Blast was utilized. These sequences were then used for multiple sequence alignment, showing that 14 amino acids in the extended protein sequence were conserved in most of the sequences shown in Figure 5.

In Silico Analysis of RGS7 Stop-Loss Mutation
Change at the 496th position in a wild-type sequence results in the loss of a stop codon and changes it into arginine (Arg) and extends the stretch of the protein sequence by 19 amino acids. In order to find similarly related sequences, Protein Blast was utilized. These sequences were then used for multiple sequence alignment, showing that 14 amino acids in the extended protein sequence were conserved in most of the sequences shown in Figure 5. Only four of these 19 amino acids were in a helical and others were in a coil, according to secondary structure prediction. The dynamics of the protein were considerably different between mutant and wildtype proteins according to the B-factor study shown in Figure 6, which may result in a protein loss of or gain of function. Figure 6. B-factor shows a significant difference between mutant and wild-type RGS7 protein. Only four of these 19 amino acids were in a helical and others were in a coil, according to secondary structure prediction. The dynamics of the protein were considerably different between mutant and wildtype proteins according to the B-factor study shown in Figure 6, which may result in a protein loss of or gain of function. Only four of these 19 amino acids were in a helical and others were in a coil, according to secondary structure prediction. The dynamics of the protein were considerably different between mutant and wildtype proteins according to the B-factor study shown in Figure 6, which may result in a protein loss of or gain of function. Figure 6. B-factor shows a significant difference between mutant and wild-type RGS7 protein. Figure 6. B-factor shows a significant difference between mutant and wild-type RGS7 protein.
The Extended protein structure demonstrated the difference between the wild-type and the altered RGS7 protein (Figure 7a-c). The Extended protein structure demonstrated the difference between the wild-type and the altered RGS7 protein (Figure 7a-c). (c) Wild-type RGS7 protein structure. (d) Secondary structure prediction of extended protein sequence demonstrated that most of the amino acids were in a coil (C) structure, and only 4 residues were present in the helix (H) structure. (e) Solvent accessibility ranged between 0-9. Values range from 0 (buried residue) to 9 (highly exposed residue).
Secondary structure prediction of extended protein sequence exhibited most of the amino acids were in the coil (C) structure, and only four residues were present in the helix (H) structure (Figure 7d). Solvent accessibility ranged between 0-9. Values range from 0 (buried residue) to 9 (highly exposed residue) (Figure 7e).

Discussion
The main result of this study is that AD patients displayed neuropsychological heterogeneity after stratification into two subgroups based on the non-Cp Cu add-on test, demonstrating our working hypothesis. Another result is the identification of an RGS7 stop-loss variant in a CuAD patient with severely abnormal Cu, supporting a potential role of the RGS7 protein in regulating Cu trafficking. Furthermore, current data suggest an association of the RGS7 stop-loss variant with CuAD. However, the data presented are merely descriptive and further experimental evidence is needed to claim a causative connection between the RGS7 stop-loss variant and AD onset or progression, as well as the role of RGS7 as a protein involved in regulating Cu trafficking. protein structure. (c) Wild-type RGS7 protein structure. (d) Secondary structure prediction of extended protein sequence demonstrated that most of the amino acids were in a coil (C) structure, and only 4 residues were present in the helix (H) structure. (e) Solvent accessibility ranged between 0-9. Values range from 0 (buried residue) to 9 (highly exposed residue).
Secondary structure prediction of extended protein sequence exhibited most of the amino acids were in the coil (C) structure, and only four residues were present in the helix (H) structure (Figure 7d). Solvent accessibility ranged between 0-9. Values range from 0 (buried residue) to 9 (highly exposed residue) (Figure 7e).

Discussion
The main result of this study is that AD patients displayed neuropsychological heterogeneity after stratification into two subgroups based on the non-Cp Cu add-on test, demonstrating our working hypothesis. Another result is the identification of an RGS7 stop-loss variant in a CuAD patient with severely abnormal Cu, supporting a potential role of the RGS7 protein in regulating Cu trafficking. Furthermore, current data suggest an association of the RGS7 stop-loss variant with CuAD. However, the data presented are merely descriptive and further experimental evidence is needed to claim a causative connection between the RGS7 stop-loss variant and AD onset or progression, as well as the role of RGS7 as a protein involved in regulating Cu trafficking.

Comparison of the AD Subtypes
CuAD subtype patients scored lower on the FAB test than AD subjects with normal non-Cp Cu, despite the fact that the two subgroups did not differ in disease duration and severity of cognitive decline, as evidenced by the MMSE score. CDT was also lower in CuAD than in typical AD, but it did not reach the statistical threshold after the B-H adjustment, likely because of the small size of the sample. Furthermore, after applying Benjamini & Hochberg's [33] correction, we noted that worse scores on the FAB scale correlated with abnormal values of all Cu biomarkers. The FAB test and the CDT specifically assess executive functions, i.e., the set of mental processes aimed at developing adaptive cognitive-behavioral patterns in response to novel and challenging environmental conditions. The prefrontal cortex, located in the frontal lobe, plays a key role in executive functions. Dopamine and noradrenalin are catecholamines that play a crucial role in the cortical-subcortical circuitry involved in executive functions. Cu is a component of metallo-proteins and a cofactor of enzymes (dopamine β-hydroxylase, monoamine oxidase) involved in catecholamine balance [41]. Thus, it is reasonable that the Cu imbalance observed in the CuAD subtype could result in an alteration of the catecholamine balance and explain the executive function deficits exhibited by these patients. As a matter of fact, catecholamine balance is altered in other disorders of Cu metabolism, primally in WD [41].

Case Report: DNM2 and RGS7 Mutation and Alzheimer's Disease
Extensive analysis of WES data allowed the identification of a small number of candidate genes.
DNM2 pathogenic variants cause a spectrum of neuromuscular diseases but no involvement in dementia or altered Cu metabolism has been documented.
To the best of our knowledge, RGS7 is a gene with no phenotype associated to date. Expression of RGS7 is especially abundant in the frontal cortex and overall, in the brain, where other RGS proteins are also temporally and spatially expressed [42]. It has been linked to a number of functions, such as synaptic plasticity, memory, learning, and maintenance of dopaminergic neurons, and is thus involved in addiction and psychiatric diseases [43][44][45].
This protein can also form a complex with 14-3-3 proteins, which has been linked to different neurodegenerative disorders including Parkinson's and Lewy body [46][47][48].

Detection of RGS7 Variants Highlights the Relevance of Alternative Pathways of Intracellular Cu Trafficking in Alzheimer's Disease
We have described in detail elsewhere [8] the main biological processes that connect Cu imbalance with major degenerative processes related to the Aβ and tau proteins in AD. The finding of RGS7 involvement in CuAD adds new insight into a less explored pathway that could help elucidate the role of non-Cp Cu in this disease.
The current results support growing evidence on the role of the endocytosis mechanisms of GPCRs in cellular Cu uptake and trafficking [16,19]. The transport machinery of GPCRs has been proposed to control a clathrin-dependent pathway in which NK A and B, and the neuropeptide Y, pair with Cu and enter the cell by pinocytosis. This Cu/peptide uptake has been described primarily in nerve cells and in the hippocampus [21]. NK A and B and GNrH are low molecular weight compounds that bind Cu and travel into the blood circulation as part of the non-Cp Cu serum component [20,21]. Furthermore, it has been shown that NK B can limit Cu uptake into astrocytes [21]. A new hypothesis poses that specialized astrocytes producing fatty acid-binding protein 7 (FABP7) might be related to AD pathogenesis [49]. FABP7+ astrocytes have been proposed to support adult neurogenesis in the dentate gyrus of the hippocampus. It is known that astrocytes are the main controller of Cu homeostasis in the brain (review in [50]). Abnormal non-Cp Cu levels may threaten Cu brain homeostasis. We do not have causal evidence demonstrating a direct link between the RGS7 loss-stop variant and the onset of AD. Nevertheless, we cannot rule out that the high non-Cp Cu serum concentration (4 µmol/L, Table 6) of the patient carrying the RGS7 loss-stop variant, along with other RGS7 altered mechanisms, could have had effects in accelerating AD pathogenesis. In line with previous evidence [20,21], FABP7+ astrocytes might control the Cu/ NK A and B and other neuropeptides up-take through endocytosis, which is dependent on GPCRs and controlled by the RGS7 protein.
Theoretically, it could be biologically plausible that FABP7+ astrocytes may be involved in restraining excess labile Cu in the brain. We are still far from defining RGS7 as a new gene causally linked to Cu metabolism and AD. Nonetheless, our results partially corroborate the new FABP7+ astrocytes' hypothesis [49]: non-Cp Cu serum expansion associated with the RGS7 stop-loss variant might decrease the ability of FABP7+ astrocytes to clear labile brain Cu, and FABP7+ astrocytes might no longer protect neurogenesis in the hippocampus, accelerating the disease cascade [49,50].

The Presence of the RGS7 Stop-Loss Variant Highlights the Relevance of Alternative Pathways of Intracellular Fe Trafficking in Alzheimer's Disease
Another finding of the current study is that the AD patient carrying the RGS7 stop-loss variant has lower levels of circulating markers of iron metabolism. Iron, transferrin, TIBC, and transferrin saturation are low, while ceruloplasmin activity measured by the orthodianisidine assay (ceruloplasmin-referred oxidative activity) is strongly increased. The serum markers of Cu and Fe in this patient are very peculiar and both could be explained by the dysregulation of metal transport controlled by the clathrin pathway, which is dependent on GPCRs and controlled by the RGS7 protein. Further research is needed to explore RGS7 s involvement in metal serum concentration.
A limitation of our study regards the size of the sample and possible sampling bias related to mild cases, which are underrepresented. The results of our pilot study need to be confirmed in other populations and especially in early cases. In addition, a study on the frequency of the RGS7 stop-loss variant in a larger population of patients with CuAD is in progress.

Materials and Methods
Non-Cp Cu has been advocated as a stratification add-on biomarker of the CuAD subtype (review in [13]). To further circumstantiate this concept, we performed a pilot clinical study focused on improving diagnostic accuracy for use in precision medicine applied to AD. The primary goal of precision medicine is to treat patients with the appropriate drugs based on their individual biological and genetic makeup. To this end, we have developed a pilot diagnostic pathway for future therapeutic programs based on anti-Cu drugs. In brief, our pilot diagnostic pathway can be summarized, as follows:

1.
Application of the standard diagnostic pathway for AD according to the IGW2 criteria [28]; 2.
Apply the stratification add-on biomarker to classify the AD population into two subgroups on the basis of the established non-Cp Cu cut-off (1.6 µmol/L) [13]; 3.
Search for genetic variants that might explain the disease-associated metabolic subpathways, i.e., searching for mutations that may explain the altered Cu metabolism.
Furthermore, in order to collect additional evidence on the CuAD subtype, we applied one of the four strategies devoted to this scope and detailed elsewhere [13,51]. In brief, after classifying AD patients into subgroups based on the non-Cp Cu cut-off, we compared the subgroups for the neuropsychological testing outcomes to find heterogeneity, as heterogeneity speaks in favor of disease subtypes [51].

Subjects
A total of 41 healthy control subjects (CTRL) and 30 consecutive patients with dementia were recruited at the Experimental Alzheimer Center, Fatebenefratelli Roman Province, Rome, Italy. Twenty-eight of them received a final diagnosis of AD according to International Working Group (IWG) 2 Criteria for AD (IWG-2 criteria) [28], with gradual progressive change in memory function for >6 months, objective evidence of hippocampal amnestic syndrome, a marked reduction in cerebrospinal (CSF) Aβ42 [52,53] and in the Aβ42/Aβ40 ratio [28], increased levels of P-tau or T-tau in the CSF, or a positive amyloid positron emission tomography (PET) or fluorodeoxyglucose (FDG)-PET, and MMSE ≤ 25, and were enrolled. Two patients did not meet the IGW2 AD criteria and were excluded. Healthy controls were individuals with no sign of neurological disorders and with normal cognitive function, selected mainly among spouses. Exclusion criteria were conditions affecting Cu metabolism, evaluated on the basis of past medical history. Participants underwent blood sampling after an overnight fast, and analyses of an extensive panel of Cu metabolism biomarkers. Patients underwent an extensive neuropsychological battery MMSE [29], MOCA [30], FAB [31], CDT [32], and NPI [33]. The study protocol was approved by the local ethics committee and subjects provided written informed consent.

Biochemical Analyses
The panel of Cu metabolism biomarkers is made of: Ceruloplasmin concentration (Cp): Immunoturbidimetric assay (Futura System S.r.l.); Cu concentration in serum measured by Abe [54] Colorimetric assay (Randox) and confirmed in 30% of the samples by Atomic absorption Spectrometry (A Analyst 600 Perkin Elmer); non-Cp Cu Calculated: Cu-(0.472*iCp) as proposed by Walshe [11]; Copper:ceruloplasmin ratio (Cu:Cp) is an additional Cu index that serves as an useful internal quality control to verify ceruloplasmin calibration, and offers information about the actual stoichiometry between Cu and ceruloplasmin in the specimens: a Cu:Cp value of 6.6 is the theoretical optimal ratio for healthy subjects [55]. Cu:Cp calculation: Cu/Cp [MW of Cp (132KDa)] as proposed by Twomey [56]. Non-Cp Cu (exchangeable Cu) was measured by Ultrafiltration +AAS600 Perkin Elmer measure of Cu (adapted from [36]). The percentage of transferrin saturation was calculated as follows: the value of serum iron divided by the total iron binding capacity [TIBC = transferrin (mg/dL) × 1.25] × 100.

Statistical Analysis
The normality of the distribution of variables was assessed by the Kolmogorov-Smirnov test (p > 0.05). Association between the two groups and dichotomous variables were assessed by Chi-square or, when appropriate, by Fisher's exact test. Differences between the groups in continuous variables were assessed by parametric t-test or, when the two samples had different variances, by Welch's t-test.
To account for multiple comparisons in the analysis of cognitive abilities between AD and CuAD, the Benjamini & Hochberg's [34] correction was applied to the p values. A p-value < 0.05 was considered statistically significant.
To compare the difference found among the two groups (AD, CuAD), Cohen's d and the corresponding 95 Confidence Interval (95% CI) were calculated. The interpretation was based on benchmarks suggested by Cohen [35]: d = 0.2 indicates a small effect size, d = 0.5 medium, and d = 0.8 large. In addition, the repeated measures ANOVA model was performed, considering the type of neuropsychological tests as a within-subject factor and the groups as a between-subject factor. Since the neuropsychological tests had different ranges of variations, for each test a normalization was applied, as follows: (observed value-Scale Minimum value)/(Scale Maximum value-Scale Minimum value).
A Pearson's correlation coefficient was calculated to evaluate the association between the neuropsychological test battery vs. each biomarker of the Cu panel (Cu, non-Cp Cu, ceruloplasmin, ceruloplasmin activity, and ceruloplasmin-specific activity). To control for the false discovery rate (FDR), Benjamini & Hochberg's correction of the p-value was applied.
Statistical analyses were performed with IBM SPSS statistic version 16.0 and version 9.3.1 of the GraphPad Prism program for graphs.
The main purpose of our pilot study was to identify two subtypes of AD patients using a non-Cp Cu biomarker as a stratification add-on test (cut-off 1.6 µmol/L) after the application of the mainstream diagnostic pathways according to the IGW-2 criteria [28]. Our working hypothesis was that AD patients display neuropsychological heterogeneity after their classification into two subgroups based on the non-Cp Cu add-on test. At the end of recruitment, 30 consecutive patients have been enrolled: 2 patients were excluded since they did not meet the IGW2 criteria, an additional one (a 76-year-old CuAD patient) was not included in the statistical analysis since she was studied separately as a case report, thus 27 patients entered the classification study. The pathogenicity of variants was predicted using the following in silico prediction algorithms: VarSome and Franklin variant interpretation (Genoox).
For WES analysis, we included all exons and surrounding 10 intronic bases, with a minimum vertical coverage of 10×, discharging all calls without PASS quality and with an allele fraction (AF) < 0.3. For a dominant mode of inheritance, we selected all variants with MAF < 0.00001 assuming possible presence in GnomAD due to late onset of disease, while for an autosomal recessive mode of inheritance, we selected all variants with MAF < 0.005 with no more than 1 homozygous present in GnomAD database. Variants were classified according to ACMG criteria, and we kept only P, PP, or VUS variants. Gene prioritization was made using Gene Ontology, GTEx, and REACTOME according to their expression, localization and function, and also manually curated according to the published literature. In silico prediction was realized using the UMD predictor, Human Splicing Finder, SIFT, Polyphen 2 HumDiv, Polyphen 2 HumVar, Mutation Taster, Mutation Assessor, Provean, M-CAP, LRT, CADD, and DANN. For genes with unknown modes of inheritance, in the case of heterozygous mutations, we used RVIS score and LoFTool to predict possible intolerance to non-synonymous variants and loss-of-function.

In Silico Analysis of RGS7 Stop-Loss Mutation
Data collection: The nucleotide sequence and protein sequence of RGS7 was downloaded from NCBI [58]. For the mutant sequence, T is replaced with C (T > C) at position 1432 (Ter496Argext*19) in the wildtype sequence and translated into the protein sequence, and Open Reading Frame (ORF) analysis was carried out using the expasy tool [59].
BLAST Analysis: The Basic Local Alignment Search Tool (BLAST) identifies areas where sequences are locally similar. It can be used to infer functional and evolutionary relationships between sequences. It is used to find closely similar sequences to RGS7 [58].
Multiple Sequence Alignment (MSA) analysis: The alignment of three or more biological sequences (protein or nucleic acid) of comparable length is known as multiple sequence alignment. The results allow for the study of the evolutionary links between the sequences and the inference of homology. Clustal was used for the analysis of MSA [60]. B-factor and secondary structure analysis: The B-factors of protein crystal structures reflect the fluctuation of atoms about their average positions and provide important information about protein dynamics [61]. Secondary server structure, solvent accessibility prediction, and B-factor analysis were done by using protocols as described in [62] Three-dimensional structure prediction: Robbeta was used to predict 3D structures. To predict the three-dimensional protein structure, this server uses homologous sequences or homology modeling [63]. Using the Chimera tool, wild-type and mutant structures were superimposed [64].

Conclusions
For the same disease duration and severity, the AD and CuAD subgroups differed in terms of cognitive profile, which was worse specifically in executive functions in subjects with higher non-Cp Cu. This heterogeneity argues in favor of the existence of a CuAD subtype, confirming previous evidence (review in [13]). The CuAD subtype had worse performance in FAB, and borderline changes in CDT, likely related to catecholamine circuit alteration associated with Cu imbalance. A patient of the CuAD subtype with severely abnormal Cu biomarkers turned out to be a carrier of an RGS7 stop-loss variant that might explain her abnormal Cu serum profile, likely associated with a disturbed Cu uptake and cell distribution regulated by RGS7. It is known that non-Cp Cu crosses the BBB [10][11][12][13], so non-Cp Cu serum expansion, with levels higher than 1.6 µmol/L, may threaten Cu homeostasis in the brain. A new hypothesis poses that FABP7+ astrocytes in the dentate gyrus may participate in adult neurogenesis, and that might be related to AD pathogenesis [49]. Since astrocytes are the main controller of Cu homeostasis in the brain, abnormal non-Cp Cu levels may be associated with FABP7+ astrocytes in a disease-associated metabolic sub-pathway, eventually leading to AD. Overall, this study supports potential AD prevention strategies, such as a low-Cu diet and Zinc therapy, to be offered specifically only to individuals with CuAD/mild cognitive impairment (MCI). We are currently carrying out a Zinc therapy phase II clinical trial in CuMCI (Alzheimer's Association Part the Cloud: Translational Research Funding for Alzheimer's disease PTC-19-602325).