Cases were defined as individuals suspected of having subclinical WD without other definitive differential diagnoses. These individuals were expected to be enriched for pathogenic VUS, as they represent potential WD cases with incomplete phenotypic expression. Suspicions were primarily based on clinical manifestations (ocular, hepatic, or neurological), biochemical abnormalities (low serum ceruloplasmin, reduced copper levels, or elevated urinary copper excretion), a positive hepatocyte copper stain, or adherence to the Leipzig criteria and international guidelines for WD^[12-16]. Additionally, unexplained or early-onset steatosis was considered a qualifying feature^[4]. Briefly, patients without an alternative diagnosis who met at least one of the Leipzig criteria or exhibited unexplained steatosis were classified as cases. ATP7B genotype information was not available before genetic testing. The differential diagnosis included common viral hepatitis, hepatocellular carcinoma, alcohol-related liver disease, autoimmune disorders, and other neurological conditions.

Controls comprised randomly selected patients with diagnoses other than WD, including viral hepatitis, alcohol-related liver disease, autoimmune diseases, drug-induced liver injury, solid tumors, hematologic malignancies, Gilbert syndrome, hemochromatosis, Coombs-positive anemia, thalassemia, and other non-WD conditions. Expanding the case cohort further reduces the likelihood of undiagnosed cases in the control group. Randomization was performed using the RANDBETWEEN() function in Excel, with each control patient assigned a blinded identification number.

The patient inclusion period spanned from January 2015 to October 2025. This study was conducted in accordance with the Declaration of Helsinki and approved by the Ethics Committee of Shanghai Ruijin Hospital (approval number: 201617/201848).

Genomic DNA was extracted from peripheral whole blood samples. The entire coding and adjacent non-coding regions of the ATP7B gene were sequenced using either Sanger sequencing or next-generation exome sequencing. The choice was guided by cost-effectiveness, as detection rates were comparable between the two methods. In exome sequencing, the targeted regions achieved > 99.9% coverage, with a mean depth exceeding 200×. The percentage of the targeted area with a mean depth greater than 30× exceeded 99%. Key variants were validated by Sanger sequencing. Primer sequences are available in our previous publication^[17].

Ceruloplasmin levels, liver function tests, complete blood cell counts, and hemoglobin levels were assessed and included in the analysis due to their relative objectivity. This approach minimizes inter-operator variability compared to subjective clinical descriptions or examinations. Serum ceruloplasmin was measured using commercial reagents on an automatic biochemical analyzer (AU5800, Beckman Coulter, USA), with a lower limit of normal set at 200 mg/L.

Variants were initially classified according to the American College of Medical Genetics and Genomics (ACMG) guidelines^[18]. Automated variant classification was performed using a platform that integrates more than 20 databases to comprehensively annotate variants, including Human Phenotype Ontology (HPO)^[19], ClinVar^[20], OMIM^[21,22], the 1000 Genomes Project^[23], gnomAD (ExAC data assimilated)^[24], and dbSNP^[25]. These databases provide scoring modification functionality for 10 of the 28 manually scored evidence items in the ACMG guidelines. Based on the automated interpretation results, we conducted additional manual annotation.

Given the potential limitations of applying universal criteria to specific disorders^[26], we adopted a more stringent classification approach: variants classified as unequivocally pathogenic by ACMG and as disease-causing mutations (DM) in the HGMD database were considered confirmed pathogenic variants. All other variants were classified as VUS. Variants not reported in public databases as of 2025 were deemed novel. In the absence of confirmed paternity and maternity, these novel variants were assigned the PM2 (Moderate) evidence criterion within the ACMG framework.

Variant-level selection was conducted in close alignment with ongoing developments in population databases, genomic curation, and bioinformatics, as well as our internal medical, allelic, and locus-specific knowledge. According to ACMG guidelines, variants with a significantly higher prevalence in affected individuals compared to controls provide strong evidence of pathogenicity (PS4 criterion). Similarly, the absence of a variant in controls (or, if recessive, at extremely low frequency) in major population datasets constitutes moderate evidence of pathogenicity. To enhance interpretive power, we incorporated in-house disease controls into our assessment.

To reevaluate WD-specific VUS, we established a threshold based on serum ceruloplasmin level. Among the various biochemical markers used in WD diagnosis, ceruloplasmin levels are relatively objective and well validated^[12]. Given that levels below 100 mg/L are strongly indicative of WD^[27], we included VUS associated with such low ceruloplasmin concentrations, as well as cases in which only VUS were identified (i.e., no pathogenic variants detected). Recurrent VUS were also considered. Each selected VUS was reviewed on a case-by-case basis. For in silico splicing prediction, SpliceAI (https://spliceailookup.broadinstitute.org/) was used.

Data analysis was performed using R version 4.5.2 (released October 2025) to assess associations and differences. All statistical tests were two-sided, and P-values < 0.05 were considered statistically significant. Normality of continuous variables was evaluated using the Shapiro-Wilk test, supplemented by visual inspection of histograms to assess distributional assumptions. Continuous variables were reported as mean ± standard deviation. Variables that deviated from normality were presented as medians (interquartile ranges, IQRs) and analyzed using non-parametric tests. Categorical variables were summarized as frequencies or percentages.

Differences in demographic and clinical characteristics between cases and controls, as well as within subgroups, were assessed using the Welch two-sample t-test or Mann-Whitney U test for continuous variables (age, protein levels, and red blood cell counts), as appropriate, and Pearson’s chi-square test with Yates’ continuity correction for categorical variables (sex). To compare the prevalence of molecular alterations between WD suspects and controls, logistic regression models were employed to assess the association between specific genetic variants and suspect status, adjusting for age and sex as confounders. Odds ratios (ORs) and corresponding 95% confidence intervals (CIs) were calculated to assess the strength of association between VUS in the case and control groups; results are presented in Supplementary Table 1. The detection of mutation only in cases and its absence in all controls, including both general and in-house study controls, was considered strong evidence of a potential disease association.

To confirm the linkage of tag variants, linkage disequilibrium (LD) between two loci was calculated using the coefficient of linkage disequilibrium (D), which quantifies the non-random association between alleles at two loci. D was calculated by comparing the observed frequency of haplotypes (allele combinations at the two loci) to the expected frequency under the assumption of random association. Normalized D' and r² value were also computed, where D' adjusts for the maximum possible D given the allele frequencies, and r² measures the correlation between alleles at the two loci, yielding a value between 0 and 1 that reflects the strength of LD. A positive D value and an r² > than 0.8 indicate strong linkage disequilibrium.

A total of 930 patients were included in the study, comprising 679 suspected WD cases and 251 non-WD disease controls. Demographic and clinical characteristics, including age, sex, serum ceruloplasmin level, alanine aminotransferase (ALT), total bilirubin, red blood cell (RBC) count, and hemoglobin levels, were compared between suspected WD cases and non-WD controls, and further stratified by ATP7B genotype (wild-type vs. variant). Suspected WD cases and controls were comparable in terms of sex distribution and age, but ceruloplasmin levels were significantly lower in the suspected WD cohort. Among patients with suspected WD, those carrying ATP7B gene variants had significantly lower ceruloplasmin levels compared to those with the wild-type genotype. Significant differences were also observed in other biochemical and hematological markers between cases and controls, as well as between the ATP7B variant and wild-type groups. These findings highlight the distinctiveness of the suspected group from the controls and suggest that molecularly diagnosed cases differ from those with other underlying etiologies. Detailed demographic comparisons are presented in Table 1.

Serum ceruloplasmin levels were compared across individuals with different ATP7B gene variants. Individuals with the wild-type ATP7B genotype exhibited significantly higher serum ceruloplasmin levels than those carrying any variant, whether pathogenic or a VUS. Variants were further categorized based on the co-occurrence of pathogenic variants, VUS, and the p.Leu770= variant (the most frequently observed VUS, discussed later in the mutational spectrum and VUS selection results). This classification yielded five subgroups: pathogenic-only, pathogenic with p.Leu770=, pathogenic with VUS, pathogenic with VUS and p.Leu770=, and VUS-only. All variant groups showed significantly lower median ceruloplasmin levels than the wild-type group. Notably, patients carrying VUS exhibited markedly reduced ceruloplasmin. Within the VUS subgroups, those with VUS in combination with known pathogenic mutations had lower median levels compared to those with VUS-only. This gradient of severity is consistent with the variable clinical impact of VUS - some likely pathogenic, others potentially benign or of intermediate effect. Figure 1 presents the detailed results. Significant differences in ceruloplasmin levels were observed in specific female subgroups, suggesting that the effect of ATP7B variants on ceruloplasmin may be modulated by sex.

Among the 24 known pathogenic variants identified, the most frequently observed were p.Arg778Leu, p.Pro992Leu, and p.Ala874Val. The variant frequencies (in counts) were as follows: p.Arg778Leu (53), p.Pro992Leu (20), p.Ala874Val (15), p.Ser975Tyr (5), p.Ile1148Thr and p.Thr935Met (both 4), and p.Val1216Met (3). Variants occurred twice included p.Arg483fs, p.Arg919Gly, p.Glu332*, p.Gly869Arg, and p.Thr1178Ala. The remaining variants were detected only once: c.1543+1G>T, p.Asn1270Ser, p.Asn728Ser, p.Cys490*, p.Gly943Ser, p.Met769fs, p.Ser105*, p.Ser1365fs, p.Thr766Met, p.Thr888Pro, and p.Trp779*.

In addition, we identified 79 VUS, with the most common being p.Leu770=, p.Val1297Ile, p.Val1106Ile, p.Ile929Val, p.Ala1003=, p.Ala476Thr, and p.Asp196Glu. Furthermore, 26 novel variants were discovered. Detailed information on the detected variants – including chromosome coordinates, variant counts, HGVS^[28] annotations, and functional categorizations - is provided in Figure 2.

Variants associated with very low ceruloplasmin levels were further evaluated for VUS reclassification. The distribution of these genetic variants was analyzed by functional categories (e.g., exonic, intronic, splice-site). Additionally, zygosity and novelty were assessed for each variant.

We selected nine VUS for detailed reevaluation. Among these, the splice site mutation c.1543+40G > A, along with the missense variants p.Val1106Ile, p.Ala1018Val, p.Leu1088Ser, p.Gly1335Arg, as well as the synonymous variant p.Ala1003=, and p.Ile1338=, were reclassified as likely pathogenic. The synonymous variant p.Leu770= was the most frequently observed VUS.

In the LD analysis between p.Leu770= and p.Arg778Leu, both variants were observed in 54 cases, including four homozygous individuals. All cases and controls were genotyped at both loci, and no individual carried p.Leu770= in the absence of p.Arg778Leu. Moreover, zygosity concordance was observed. The calculated LD parameters were as follows: D = 0.0291, D' = 1.0000, and r² = 1.0000, indicating a strong linkage between the two loci. Consequently, the synonymous variant p.Leu770= - regardless of zygosity - should be regarded as a tag variant, and thus should not be designated as an independent pathogenic hit. Further studies in diverse populations are warranted before any recommendations can be made regarding modifications to diagnostic scoring systems, such as the Leipzig criterion.

Detailed descriptions and interpretations are provided in Table 2. Selected VUS, including the splice site variant c.1543+40G > A, are highlighted due to their novelty or presence in homozygosity in the VUS-only analysis. However, in-silico splicing prediction suggests that both c.1543+40G > A and the synonymous VUS are likely polymorphisms. A detailed visualization of these findings is presented in Figure 3.

The primary limitation of this study stems from the rarity of WD. As with most rare disease research, this field faces inherent challenges, including limited awareness, diagnostic delays, and insufficient funding. Furthermore, the study relies predominantly on ceruloplasmin levels as an objective biomarker for ATP7B dysfunction. While ceruloplasmin is among the first-line and most widely used clinical parameters, it does not directly reflect the functional status of the ATP7B protein. Emerging biomarkers - such as the measurement of ATP7B peptides^[40] and serum non-ceruloplasmin-bound copper^[41] - are currently under development and evaluation, and hold promise for more accurate assessment of ATP7B function in the future.

Another concern relates to the use of a very low ceruloplasmin cutoff for reclassifying VUS. This approach led to the exclusion of cases with ceruloplasmin levels above 100 mg/L and, critically, extremely rare WD cases with normal ceruloplasmin levels. As such, the reclassification strategy may be most applicable to variants associated with severe ceruloplasmin deficiency and may fail to capture milder or atypical disease presentations. Future case-control studies may help address this limitation - either by excluding ceruloplasmin from the analysis or by incorporating novel biomarkers that better reflect ATP7B function.

Additionally, the study lacks sufficient functional validation (e.g., in vitro copper transport assays) for the reclassified variants, limiting confidence in their pathogenicity assessment. The use of same-ethnic controls in applying ACMG population criteria may introduce potential bias. The possible misclassification within the “suspect” and “control” groups raises additional concerns.

Our study underscores the critical need to establish a dedicated ATP7B Gene Curation Expert Panel to ensure ongoing review and refinement of variant interpretations, encompassing both coding and non-coding regions. The identification of recurrent variants will further support the development of ethnically targeted screening strategies. Constant communication is of pivotal importance. Clinical laboratories should make concerted efforts to prioritize the reporting of any changes made that may affect clinical management. Particularly, the reclassification of any VUS should be prioritized and promptly communicated to clinicians, as it carries greater clinical impact than reclassifications such as from likely pathogenic to pathogenic.