Volume 38, Issue 12 p. 2241-2248
RESEARCH ARTICLE
Open Access

Harmonizing Genetic Testing for Parkinson's Disease: Results of the PARKNET Multicentric Study

Alessio Di Fonzo MD, PhD

Corresponding Author

Alessio Di Fonzo MD, PhD

Neuroscience Section, Department of Pathophysiology and Transplantation, Dino Ferrari Center, University of Milan, Milan, Italy

Foundation IRCCS Ca' Granda Ospedale Maggiore Policlinico, Neurology Unit, Milan, Italy

Correspondence to: Dr. Alessio Di Fonzo, Foundation IRCCS Ca' Granda Ospedale Maggiore Policlinico, Neurology Unit, Milan 20122, Italy; E-mail: [email protected]

Search for more papers by this author
Marco Percetti MD

Marco Percetti MD

Neuroscience Section, Department of Pathophysiology and Transplantation, Dino Ferrari Center, University of Milan, Milan, Italy

School of Medicine and Surgery, Milan Center for Neuroscience, University of Milan-Bicocca, Milan, Italy

Foundation IRCCS San Gerardo dei Tintori, Monza, Italy

Search for more papers by this author
Edoardo Monfrini MD, PhD

Edoardo Monfrini MD, PhD

Neuroscience Section, Department of Pathophysiology and Transplantation, Dino Ferrari Center, University of Milan, Milan, Italy

Foundation IRCCS Ca' Granda Ospedale Maggiore Policlinico, Neurology Unit, Milan, Italy

Search for more papers by this author
Ilaria Palmieri PhD

Ilaria Palmieri PhD

IRCCS Mondino Foundation, Pavia, Italy

Search for more papers by this author
Alberto Albanese MD

Alberto Albanese MD

IRCCS Humanitas Research Hospital, Milan, Italy

Search for more papers by this author
Micol Avenali MD, PhD

Micol Avenali MD, PhD

IRCCS Mondino Foundation, Pavia, Italy

Department of Brain and Behavior Sciences, University of Pavia, Pavia, Italy

Search for more papers by this author
Anna Bartoletti-Stella BSc, PhD

Anna Bartoletti-Stella BSc, PhD

IRCCS Istituto delle Scienze Neurologiche di Bologna, Bologna, Italy

DIMEC, University of Bologna, Bologna, Italy

Search for more papers by this author
Fabio Blandini MD, PhD

Fabio Blandini MD, PhD

Foundation IRCCS Ca' Granda Ospedale Maggiore Policlinico, Milan, Italy

Search for more papers by this author
Gloria Brescia PhD

Gloria Brescia PhD

Foundation IRCCS Ca' Granda Ospedale Maggiore Policlinico, Milan, Italy

Search for more papers by this author
Giovanna Calandra-Buonaura MD, PhD

Giovanna Calandra-Buonaura MD, PhD

DIMEC, University of Bologna, Bologna, Italy

DIBINEM, University of Bologna, Bologna, Italy

Search for more papers by this author
Rosa Campopiano BSc

Rosa Campopiano BSc

IRCCS Neuromed, Pozzilli, Italy

Search for more papers by this author
Sabina Capellari MD, PhD

Sabina Capellari MD, PhD

DIMEC, University of Bologna, Bologna, Italy

DIBINEM, University of Bologna, Bologna, Italy

Search for more papers by this author
Isabel Colangelo MSc

Isabel Colangelo MSc

Medical Genetics and Neurogenetics Unit, Fondazione IRCCS Istituto Neurologico C. Besta, Milan, Italy

Search for more papers by this author
Giacomo Pietro Comi MD

Giacomo Pietro Comi MD

Neuroscience Section, Department of Pathophysiology and Transplantation, Dino Ferrari Center, University of Milan, Milan, Italy

Foundation IRCCS Ca' Granda Ospedale Maggiore Policlinico, Neurology Unit, Milan, Italy

Search for more papers by this author
Giada Cuconato BSc

Giada Cuconato BSc

Department of Molecular Medicine, University of Pavia, Pavia, Italy

Search for more papers by this author
Rosangela Ferese BSc

Rosangela Ferese BSc

IRCCS Neuromed, Pozzilli, Italy

Search for more papers by this author
Caterina Galandra BSc

Caterina Galandra BSc

IRCCS Mondino Foundation, Pavia, Italy

Department of Molecular Medicine, University of Pavia, Pavia, Italy

Search for more papers by this author
Stefano Gambardella BSc

Stefano Gambardella BSc

Department of Biomolecular Sciences, University of Urbino “Carlo Bo”, Urbino, Italy

Search for more papers by this author
Barbara Garavaglia PhD

Barbara Garavaglia PhD

Medical Genetics and Neurogenetics Unit, Fondazione IRCCS Istituto Neurologico C. Besta, Milan, Italy

Search for more papers by this author
Andrea Gaudio BSc

Andrea Gaudio BSc

IRCCS Ospedale Policlinico San Martino, Genoa, Italy

Search for more papers by this author
Emiliano Giardina PhD

Emiliano Giardina PhD

Genomic Medicine Laboratory-UILDM, Santa Lucia Foundation IRCCS, Rome, Italy

Department of Biomedicine and Prevention, Tor Vergata University, Rome, Italy

Search for more papers by this author
Federica Invernizzi MSc

Federica Invernizzi MSc

Medical Genetics and Neurogenetics Unit, Fondazione IRCCS Istituto Neurologico C. Besta, Milan, Italy

Search for more papers by this author
Paola Mandich MD, PhD

Paola Mandich MD, PhD

IRCCS Ospedale Policlinico San Martino, Genoa, Italy

DINOGMI, University of Genoa, Genoa, Italy

Search for more papers by this author
Rossana Mineri PhD

Rossana Mineri PhD

IRCCS Humanitas Research Hospital, Milan, Italy

Search for more papers by this author
Celeste Panteghini MSc

Celeste Panteghini MSc

Medical Genetics and Neurogenetics Unit, Fondazione IRCCS Istituto Neurologico C. Besta, Milan, Italy

Search for more papers by this author
Chiara Reale MSc

Chiara Reale MSc

Medical Genetics and Neurogenetics Unit, Fondazione IRCCS Istituto Neurologico C. Besta, Milan, Italy

Search for more papers by this author
Lucia Trevisan MD

Lucia Trevisan MD

IRCCS Ospedale Policlinico San Martino, Genoa, Italy

Search for more papers by this author
Stefania Zampatti MD

Stefania Zampatti MD

Genomic Medicine Laboratory-UILDM, Santa Lucia Foundation IRCCS, Rome, Italy

Search for more papers by this author
Pietro Cortelli MD, PhD

Pietro Cortelli MD, PhD

DIMEC, University of Bologna, Bologna, Italy

DIBINEM, University of Bologna, Bologna, Italy

Search for more papers by this author
Enza Maria Valente MD, PhD

Enza Maria Valente MD, PhD

IRCCS Mondino Foundation, Pavia, Italy

Department of Molecular Medicine, University of Pavia, Pavia, Italy

Search for more papers by this author
on behalf of the PARKNET study group
First published: 26 September 2023

Alessio Di Fonzo and Marco Percetti share first co-authorship.

[Correction added on 12 October 2023, after first online publication: The ninth author's degree is updated in this version.].

Relevant conflicts of interest/financial disclosures: A.D.F. reports advisory board fees from Sanofi and speaking honoraria from Sanofi and Zambon. A.A. is Specialty Chief Editor of Frontiers in Neurology and President-Elect of the International Association for Parkinsonism and Related Disorders. A.A. received speaker's honoraria from Merz Pharma and Ipsen Pharma. G.C.B. received speakers's honoraria from Bial Italia. G.P.C. reports participation to advisory boards of Roche and Biogen. S.Z. received financial support from National Research council CNR. P.C. received speakers's honoraria from Abbvie. E.M.V. is Associate Editor of Journal of Medical Genetics; is Genetics Section Editor of Pediatric Research, of The Cerebellum, and of Neurological Sciences; is member of the Editorial Board of Movement Disorders clinical Practice; is member of the Steering Committee of ASAP GP2 (Global Parkinson genetic Program). E.M.V. received research support from the Italian Ministry of Health, CARIPLO Foundation, Telethon Foundation Italy, Pierfranco and Luisa Mariani Foundation, and European Community (Eranet Neuron). None of the other authors report any conflicts of interest.

Funding agencies: This work was supported by grants from the Italian Ministry of Health (Ricerca Corrente Reti 2021–2022-PARKNET project).

ABSTRACT

Background and Objective

Early-onset Parkinson's disease (EOPD) commonly recognizes a genetic basis; thus, patients with EOPD are often addressed to diagnostic testing based on next-generation sequencing (NGS) of PD-associated multigene panels. However, NGS interpretation can be challenging in a diagnostic setting, and few studies have addressed this issue so far.

Methods

We retrospectively collected data from 648 patients with PD with age at onset younger than 55 years who underwent NGS of a minimal shared panel of 15 PD-related genes, as well as PD-multiplex ligation-dependent probe amplification in eight Italian diagnostic laboratories. Data included a minimal clinical dataset, the complete list of variants included in the diagnostic report, and final interpretation (positive/negative/inconclusive). Patients were further stratified based on age at onset ≤40 years (very EOPD, n = 157). All variants were reclassified according to the latest American College of Medical Genetics and Genomics criteria. For classification purposes, PD-associated GBA1 variants were considered diagnostic.

Results

In 186 of 648 (29%) patients, the diagnostic report listed at least one variant, and the outcome was considered diagnostic (positive) in 105 (16%). After reanalysis, diagnosis changed in 18 of 186 (10%) patients, with 5 shifting from inconclusive to positive and 13 former positive being reclassified as inconclusive. A definite diagnosis was eventually reached in 97 (15%) patients, of whom the majority carried GBA1 variants or, less frequently, biallelic PRKN variants. In 89 (14%) cases, the genetic report was inconclusive.

Conclusions

This study attempts to harmonize reporting of PD genetic testing across several diagnostic labs and highlights current difficulties in interpreting genetic variants emerging from NGS-multigene panels, with relevant implications for counseling. © 2023 The Authors. Movement Disorders published by Wiley Periodicals LLC on behalf of International Parkinson and Movement Disorder Society.

Introduction

Patients with early-onset Parkinson's disease (EOPD) are mostly defined as subjects with disease onset before 50 years.1 It has been clearly established that EOPD often recognizes a genetic basis.2 Although PRKN is the most commonly mutated gene in recessively inherited EOPD with onset before 40 years, several other genes have also been implicated.3 In particular, heterozygous pathogenic variants in LRRK2 and GBA1, both dominantly inherited with low to very low penetrance, have been frequently detected in patients with EOPD, with ethnic-specific variability.4 In the past, these genes used to be Sanger sequenced in selected patients (eg, based on positive family history, very early onset, atypical presentations). However, most PD genes have numerous exons and lack mutational hot spots, making Sanger sequencing impractical in the diagnostic setting.

The advent of next-generation sequencing (NGS) has revolutionized the diagnosis of genetically heterogeneous diseases such as EOPD, allowing to sequence all disease-related genes simultaneously and rapidly in large cohorts of patients, and has now fully substituted Sanger sequencing in most diagnostic labs. In contrast, NGS unraveled a much higher genetic variability than expected, resulting in the identification of several variants of difficult interpretation, which often represent a challenge in the diagnostic setting.

Although labs are bound to adopt specific criteria to classify variants according to their presumed pathogenicity, guidelines on which variants should be included in the diagnostic report and how certain variants should be interpreted are still missing, leading to wide variability and inconsistencies in genetic testing reports.

Within the PARKNET project, this study aimed to harmonize the interpretation of PD genetic testing across the network, as well as to provide a comprehensive assessment of the genetic basis of EOPD in Italy.

Patients and Methods

Patient Recruitment

The National Virtual Parkinson Institute, a consortium of Italian Scientific Institutes for Medical Research and Care (Istituti di Ricovero e Cura a Carattere Scientifico [IRCCS]) focused on PD research and management funded by the Ministry of Health, has undertaken PARKNET, an ambitious project to share knowledge and expertise, as well as clinical and genetic data of patients with PD among participating centers. Data-sharing agreements were signed by each institute.

In the frame of PARKNET, we retrospectively collected data from 648 patients with PD with age at onset (AO) younger than 55 years referred for PD genetic testing in the years 2017–2022 at one of the following institutions: Foundation IRCCS Ca' Granda Ospedale Maggiore Policlinico of Milan, Foundation IRCCS Istituto Neurologico Carlo Besta of Milan, IRCCS Humanitas Research Hospital of Rozzano, IRCCS Ospedale Policlinico San Martino of Genoa, IRCCS Istituto delle Scienze Neurologiche of Bologna, Foundation Mondino IRCCS of Pavia, IRCCS Istituto Neurologico Mediterraneo of Isernia, and Foundation Santa Lucia IRCCS of Rome (Fig. S1). Diagnosis of PD was always made by specialists in movement disorders using the Movement Disorder Society clinical diagnostic criteria.5 The decision to set the cutoff of AO to 55 years was stimulated by the observation that several patients with onset between 50 and 55 years had also been referred to diagnostic genetic testing, even if not properly defined as EOPD. Shared data included a minimal clinical dataset (AO, family history for PD, consanguinity), as well as the complete list of variants included in the diagnostic report and the final interpretation of results: either positive (one or more variants listed in the report, allowing to reach a definite genetic diagnosis), inconclusive (one or more variants listed in the report, but not sufficient to reach a genetic diagnosis), or negative (no variants listed in the report). Written informed consent for genetic analyses for diagnostic and research purposes was obtained from all patients. Stratification of collected data was performed according to AO, differentiating a subgroup of patients with very EOPD (vEOPD; AO ≤ 40 years), and to family history for PD (Fig. S4).

Genetic Analysis

The study design is schematically shown in Figure S2.

Patients were included if both a diagnostic gene panel and a dosage analysis for SNCA, PRKN, PINK1, and PARK7 had been performed. Panels varied among institutions, but all shared a minimal set of 15 genes associated with autosomal dominant (SNCA, LRRK2, VPS35, GBA1), X-linked (RAB39B), and autosomal recessive PD (PRKN, PINK1, PARK7, ATP13A2, PLA2G6, DNAJC6, SYNJ1, FBXO7, VPS13C, PTRHD1) (Table S1). NGS enrichment was based on either HaloPlex (Agilent), SureSelect XT (Agilent), Nextera (Illumina), or TruSeq Custom Amplicon (Illumina). All labs guaranteed a mean 30-fold coverage of at least 90% nucleotides across all genes. Multiplex ligation-dependent probe amplification (MLPA) was performed to determine exon copy-number variants (CNVs) using P051-D2 and P052-D2 kits (MRC Holland). Sanger sequencing was used to confirm all reported variants. Each center was requested to indicate all single-nucleotide variants (SNVs) and CNVs included in the genetic report, and whether it was defined positive or inconclusive. Data related to each variant were collected from online databases (gnomAD v2.1.1,6 dbSNP, ClinVar).

Harmonization of the Interpretation Pipeline

The harmonization process consisted of applying the same internationally adopted criteria to classify each reported variant. Each SNV was (re)classified as benign (<−6 points), likely benign (from −6 to −1 points), variants of unknown significance (VUSs; from 0 to 5 points), likely pathogenic (from 6 to 9 points), or pathogenic (≥10 points), according to the latest American College of Medical Genetics and Genomics (ACMG) criteria, because not all laboratories had formerly used this classification.7, 8 ACMG classification was first obtained using Varsome Premium9 and then manually curated to take into account additional relevant factors, such as data from the recent literature or cosegregation with another deleterious variant (in case of recessive genes). GBA1 variants previously classified as “complex,” “severe,” “mild,” and “risk” were all classified as pathogenic.10 To calculate the diagnostic yield, we included only variants classified as pathogenic and likely pathogenic, hereby collectively referred to as “deleterious variants.” Notably, according to recent evidences, the PVS1 criterion was not flagged for loss-of-function LRRK2 variants.11 All deleterious GBA1 variants were considered diagnostic, although being aware of their very low penetrance. Regarding autosomal recessive genes, only biallelic deleterious variants were considered diagnostic.12, 13

Statistical Analysis

χ2 test was used to assess statistically significant differences within PD and vEOPD groups. A P value <0.05 was set to consider a difference between groups statistically significant.

Results

Harmonization and Diagnostic Yield

We reviewed the diagnostic reports of 648 patients with PD with AO ≤ 55 (mean AO, 42.6 ± 9.7 years; male: n = 369, 57%) (Table 1).

TABLE 1. Demographic data and genetic results
PD (AO ≤ 55 y) vEOPD (AO ≤ 40 y)
Demographic data
Total patients, n 648 157
Mean AO ± SD 42.6 ± 9.7 y 32.3 ± 9.7 y
Male sex, n (%) 369 (57) 105 (67)
Positive family history for PD, n (%) 172 (26.5) 59 (37.6)
No variant alleles listed in genetic report, n (%) 462 (71.3) 86 (54.8)
At least one variant listed in genetic report, n (%) 186 (28.7) 71 (45.2)
Genetic results (postharmonization)
Positive diagnoses (P/LP variants), n (%) 97 (15.0) 38 (24.2)
  • GBA1a
58 (9.0) 14 (8.9)
  • PRKN biallelica
18 (2.8) 12 (7.6)
  • LRRK2a
9 (1.4) 4 (2.5)
  • PINK1 biallelic
7 (1.1) 6 (3.8)
  • SNCA
3 (0.5) 1 (0.6)
  • PARK7 biallelic
1 (0.2) 1 (0.6)
  • RAB39B
1 (0.2) 1 (0.6)
  • VPS13C biallelic
1 (0.2)
  • PLA2G6 biallelicb
1 (0.2)
Inconclusive diagnoses, n (%) 89 (13.7) 33 (21.0)
  • Monoallelic P/LP variants in recessive genes
13 (2.0) 6 (2.5)
  • Other variants (VUS, LB, or B)
76 (11.7) 27 (17.2)
  • a One patient tested positive for both GBA1 and PRKN.
  • b One patient tested positive for both LRRK2 and PLA2G6.
  • Abbreviations: PD, Parkinson's disease; AO, age at onset; vEOPD, very early-onset Parkinson's disease; SD, standard deviation; VUS, variant of unknown significance; LB, likely benign; B, benign.

In 186 of 648 patients (29%), at least one SNV/CNV was listed in the report. The identified variants were considered diagnostic in 105 patients (16%), whereas in 81 cases (13%) the report listed one or more variants, but the outcome was considered inconclusive.

After harmonization, the interpretation of genetic testing changed in 18 of 186 patients (10%). In five, the outcome was modified from inconclusive to diagnostic after reclassifying the reported GBA1 variants as pathogenic. Conversely, 13 reports previously considered as diagnostic were declassified as inconclusive, including four cases with single heterozygous pathogenic variants in a recessive gene and nine cases where the reported variants were reclassified as VUS or likely benign (Table S3).

The new diagnostic yield was 15% (97/648); in an additional 89 patients (14%), the variants listed in the report were considered inconclusive (Tables 1 and S4). Deleterious variants in the GBA1 gene (n = 58, 8.0%; mean AO: 43.6 ± 9.3 years) and biallelic deleterious SNVs/CNVs of PRKN (n = 18, 2.6%; mean AO: 33.3 ± 11.2 years) were the most common diagnoses, followed by heterozygous deleterious variants in LRRK2 (n = 9, including p.G2019S n = 6, 0.9%, mean AO: 42.3 ± 12.8 years; p.R1441C n = 3, 0.5%, mean AO: 40.7 ± 20.6 years), biallelic deleterious variants in PINK1 (n = 7, 1.1%; mean AO: 32.6 ± 12.5 years), and heterozygous deleterious variants in SNCA (n = 3, 0.5%, mean AO: 44.6 ± 12.6 years). Single patients carried deleterious variants in RAB39B, PARK7, VPS13C, and PLA2G6, respectively.

Notably, two patients received a double genetic diagnosis: one carried a GBA1 heterozygous variant and a PRKN homozygous variant, and another carried the LRRK2 p.G2019S variant in combination with a homozygous likely pathogenic missense variant in PLA2G6. In 16 patients who received a definite diagnosis, the diagnostic report also listed one or more additional variants in a distinct gene, which were classified as VUSs or likely benign/benign in all cases but one (a heterozygous likely pathogenic variant in DNAJC6).

Among the 89 patients with inconclusive reports, 13 carried a single heterozygous deleterious variant in a recessive gene (PRKN n = 7, PINK1 n = 2, PLA2G6 n = 3, PARK7 n = 1), whereas 5 additional cases were biallelic carriers of two variants in a recessive gene (PINK1 n = 2, PLA2G6 n = 1, FBXO7 n = 1, VPS13C n = 1), both classified either as VUS or benign/likely benign, thus not allowing to reach a definite diagnosis. The remaining patients carried single or multiple nondeleterious variants in one or more genes.

Focusing on the vEOPD subgroup (n = 157; male n = 105, 67%; mean AO: 32 ± 9.7 years), 71 of 157 (45%) had one or more variants reported, which were initially considered diagnostic in 43 (27%) and inconclusive in 28 (18%). Postharmonization, a definite diagnosis could be confirmed in 38 patients (25%). Deleterious GBA1 variants (n = 14, 9%; mean AO: 32.3 ± 8.7) and biallelic PRKN SNVs/CNVs (n = 12, 8%; mean AO: 30 ± 8.9) were the most common diagnoses, followed by biallelic PINK1 variants (n = 6, 4%; mean AO: 28.8 ± 8.4 years). Other genetic causes were rarer.

Variants

Overall, 263 variant alleles in 14 genes were listed in the examined genetic reports (Tables 2 and S5). Among these alleles, 148 were classified as deleterious, 50 as VUSs, and 65 as benign or likely benign. Of the deleterious variants, 134 were diagnostic, whereas 14 represented single heterozygous variants in recessive genes.

TABLE 2. Genetic findings: summary of the most involved genes
Positive Inconclusive
Total P/LP variants, biallelic in AR genes Monoallelic P/LP variants in AR genes VUSs LB/B variants
All genes 263 134 (51%) 14 (5.3%)a 50 (19%) 65 (24.7%)
GBA1 70 65 (92.9%) 5 (7.1%) 0
LRRK2 34 9 (26.5%) 10 (29.4%) 15 (44.1%)
SNCA 3 3 (100%) 0 0
VPS35 4 0 3 (75%) 1 (25%)
PRKN 55 36 (65.4%) 7 (12.7%) 7 (12.7%) 5 (9.1%)
PINK1 28 14 (50.0%) 2 (7.1%) 8 (28.6%) 4 (14.3%)
PARK7 12 2 (16.7%) 1 (8.3%) 1 (8.3%) 8 (66.7%)
ATP13A2 2 0 0 0 2 (100%)
PLA2G6 17 2 (11.8%) 3 (17.6%) 6 (35.3%) 6 (35.3%)
DNAJC6 7 0 1a (14.3%) 2 (28.6%) 4 (57.1%)
SYNJ1 6 0 0 3 (50%) 3 (50%)
FBXO7 6 0 0 2 (33.3%) 4 (66.7%)
VPS13C 18 2 (11.1%) 0 3 (16.7%) 13 (72.2%)
PTRHD1 0 0 0 0 0
RAB39B 1 1 (100%) 0 0
  • Note: Number and frequency of variant alleles divided by gene and American College of Medical Genetics and Genomics classification (out of a total of 263 alleles reported in 186 subjects with Parkinson's disease). For the purpose of this table, homozygous variants are counted twice. Some variants (within the same gene or distinct genes) co-occurred in the same individual. The two GBA1 complex alleles are counted as single variants. Percentages are calculated on the total number of variants for each gene. The most prevalent category for each gene is highlighted in bold.
  • a One heterozygous LP variant in DNAJC6 occurred in a patient carrying a GBA1 deleterious variant.
  • Abbreviations: AR, autosomal recessive; VUS, variant of unknown significance.

GBA1 was the most commonly mutated gene, with 70 variant alleles reported, of which 65 (93%) were classified as deleterious. The founder p.N409S (N370S) and p.L483P (L444P) variants were the commonest, representing nearly half of the reported variants (n = 30, 43%), either alone or in combination with other GBA1 variants [p.R87W (R48W), p.R159Q (R120Q), p.E365K (E326K)]. In the PRKN gene, 55 variant alleles were reported. The majority of PRKN variants were pathogenic or likely pathogenic (n = 43, 78%), with the majority being represented by CNVs, in particular deletions or duplications of exon 3. Of the 28 PINK1 variant alleles, more than half were deleterious (n = 16, 57%, of which 13 were truncating). Conversely, of 34 LRRK2 variant alleles, only 9 (26%) could be classified as deleterious, including 6 p.G2019S and 3 p.R1441C.

Discussion

In this study, we reviewed the genetic reports of 648 patients with PD with AO ≤55 years provided by eight Italian genetic laboratories in the frame of the PARKNET project, with the main goal to harmonize the interpretation of diagnostic PD genetic tests across all participating centers. In addition, we used this large wealth of data to assess the diagnostic yield and determine the contribution of each gene to the diagnosis of EOPD in Italy.

To harmonize interpretation of genetic results, we applied the latest ACMG guidelines to (re)classify each variant that had been listed in the diagnostic reports; in line with these widely accepted guidelines, only patients carrying pathogenic or likely pathogenic variants in a PD-related gene (biallelic in case of recessive genes) were considered “positive,” for example, received a definite diagnosis of a genetic disease. Notably, not all laboratories were CLIA accredited, and some had not adopted ACMG criteria to classify variants. Interestingly, the reanalysis led to changing the outcome of the diagnostic report in 18 patients (10%): 5 could eventually receive a definite diagnosis after upgrading of the detected GBA1 variants to pathogenic, whereas in 13 patients, the former genetic diagnosis could not be confirmed, either because the identified variant was (re)classified as VUSs or likely benign or due to the lack of a second deleterious variant in a recessive gene. This outcome change has major implications for genetic counseling of patients and families, considering at one end the diagnostic delay and, at the other end, the former delivery of a genetic diagnosis that could not be confirmed based on further assessments.

Despite the rising awareness toward PD genetic testing and the increasing requests of diagnostic tests for patients with EOPD, no specific guidelines exist, leading to inconsistencies across distinct laboratories in the choice of variants to be listed in the diagnostic report and, most importantly, in their final interpretation. The case of the patient carrying the homozygous variant p.A516V in PLA2G6 is emblematic of the still-open challenge in the interpretation of variant pathogenicity and highlights the limitations of current ACMG criteria. Indeed, according to such criteria, this variant is classified as a VUS, applying PM1 (mutational hot spot) and PM2 (absence in controls) at the supporting level. Upon discovering a positive family history and reexamining clinical and imaging data, we were able to manually flag two additional criteria at the supporting level, namely, PP1 (cosegregation into multiple family members) and PP4 (highly specific phenotype), yet, to strictly adhere to current guidelines,8 we did not flag PP5 (reliable source reports the variant as pathogenic), although the same variant had formerly been reported in three unrelated patients with features in the spectrum of PLA2G6-associated neurodegeneration.14-17 Thus, despite that this homozygous variant clearly appears as causative of the early-onset dystonia-parkinsonism phenotype shown by the patient, ACMG criteria still do not reach the sufficient strength to unequivocally classify the variant as pathogenic/likely pathogenic. This case well illustrates the current controversies and methodological challenges in reaching a genetic diagnosis that should be properly recognized toward continuous improvement of diagnostic criteria. In this light, ClinGen has recently established an international PD gene and variant curation expert panel to develop consensus over genes with low confidence that have been cited in the literature and/or included on commercially available genetic testing panels, as well as over the pathogenic relevance of specific gene variants (https://clinicalgenome.org/). Yet, this effort is still ongoing and, even upon completion, it will require some time to enter routine clinical practice. Our work highlights the importance of developing international consensus guidelines for PD genetic testing, providing guidance to the diagnostic laboratories as regard genes to be included in NGS panels, as well as consolidated criteria for variant classification, interpretation, and reporting.

In this cohort of 648 PD probands, a genetic diagnosis could be reached in 15% of patients and increased to 24% in the subgroup of vEOPD. These figures are in line with previous NGS studies in French, Spanish, Chinese, and Taiwanese populations,18-21 suggesting that the diagnostic yield is similar across different populations worldwide and not strictly correlated with population-specific genetic variants. However, the frequency of founder mutations is different across populations: for instance, LRRK2 p.G2019S and GBA1 p.N409S (N370S) are more common in European patients, whereas LRRK2 p.G2385R and GBA1 p.L483P (L444P) occur more frequently in Asians.22, 23

Notably, GBA1 variants emerged as the commonest determinants in EOPD, being detected in 9% of the whole cohort and accounting for more than half of all genetic diagnoses. There is now consensus evidence that GBA1 heterozygous variants cause autosomal dominant PD with highly reduced penetrance (https://www.clinicalgenome.org/affiliation/40079/), and therefore such variants should be included in the diagnostic reports and included in the assessment of the diagnostic yield of PD genetic testing. Although the two founder variants, p.N409S (N370S) and p.L483P (L444P), were the most frequent ones, they collectively accounted for less than half of all GBA1 diagnoses, confirming our previous findings that sequencing of the entire GBA1 gene is necessary to avoid false negative results.10

The PRKN gene was the second most frequently mutated gene in this study: frequencies are consistent to those of different populations,18, 21 with the exception of Ashkenazi Jews, who seem to be more rarely mutated.13 Up to 40% of variant alleles were represented by CNVs, highlighting the importance of complementing the diagnostic testing with molecular strategies able to identify PRKN genomic rearrangements.

LRRK2 p.G2019S and p.R1441C were collectively found in ~1.4% of patients with PD with AO ≤55 years, in line with previous reports in the Italian population.24-26 This low frequency of LRRK2 diagnoses compared with other studies may reflect the fact that LRRK2-related PD may manifest at older ages.24, 26 Moreover, it is expected that several patients with EOPD (especially those with positive family history) have already been screened for LRRK2 p.G2019S and p.R1441C in the past years, and thus may have been missed by this study.

PINK1 biallelic and SNCA heterozygous pathogenic variants occurred in 1.1% and 0.5% of patients, respectively, whereas few other genes (PARK7, VPS13C, PLA2G6 or RAB39B) were mutated in single patients, usually presenting more complex phenotypes. It must be noted that the cohort screened in this study was characterized by a pure parkinsonian syndrome, which likely explains the underrepresentation of genes associated with more complex phenotypes (ie, SYNJ1, ATP13A2, FBXO7, PTRHD1). Subgroup stratification showed enrichment of pathogenic variants in patients with vEOPD, leading to a higher diagnostic yield (p = 0.00002). In particular, patients with AO ≤40 years presented a higher chance to have biallelic PRKN or PINK1 variants compared with the entire cohort (P = 0.00007).

Upon reclassification of variants, 89 genetic tests (14%) yielded inconclusive results, yet important differences should be highlighted within this group. Indeed, 13 patients carried a single heterozygous pathogenic or likely pathogenic variant in a recessive gene, suggesting the existence of a second cryptic variant escaping NGS and MLPA testing, and thus prompting further studies to explore this possibility. In two additional cases, two VUSs in trans were detected in a recessive gene (either PINK1 or PLA2G6). Although this variant's classification impeded to reach a definite diagnosis at present, functional experiments may be warranted to prove pathogenicity of these VUSs, which would potentially allow to reclassify them as deleterious.

Conversely, 74 reports listed heterozygous variants in one or more genes, which were classified as VUSs, likely benign, or benign. Similar variants were also listed in the reports of 16 patients who already had a definite diagnosis related to the presence of deleterious variants in another PD gene. It is questionable whether these variants should have been included at all in the diagnostic reports because, according to ACMG guidelines, VUSs should be reported only where further testing or investigations have the potential to change their classification to likely pathogenic, which is improbable for benign and likely benign variants, and also for some low-score VUSs. In fact, these variants are almost invariably unlikely to be disease causing and could be potentially confusing if included in a report. Notably, most variant alleles in the SNCA, GBA1, and PRKN genes were deleterious (many of them having already been reported in patients with PD); conversely, nearly half of PINK1 and the majority of LRRK2, PARK7, PLA2G6, and VPS13C variant alleles were classified as nondeleterious, suggesting caution when interpreting variants encountered in these genes. It is also worth noticing that variant classification is still a highly dynamic process, with frequent fine-tuning of the adopted criteria that might result in classification changes, especially for VUSs. In this light, laboratories should implement a regular reassessment of NGS data, especially regarding variants whose classification is not straightforward.

Finally, it is interesting to note that, despite NGS and MLPA screening, the majority of patients with PD remained without a genetic diagnosis. Many Whole Exome Sequencing projects have been recently conducted in large cohorts of patients but have resulted in very few novel PD-causative genes.27 These observations suggest either the possibility of an incomplete detection of variants in already known PD genes using conventional molecular approaches, or the existence of other still-unknown genetic factors, possibly acting in a multifactorial model to influence PD susceptibility even in patients with early onset of the disease.28, 29 It is expected that large ongoing projects on PD genetics, such as the worldwide GP2 effort (Global Parkinson's Genetic Project; https://gp2.org), will eventually address this major unsolved issue.

We acknowledge some limitations of this study. First, this is a retrospective multicentric study, and despite that we included only patients who underwent both NGS of a minimal shared gene set and MLPA analysis, testing methodologies, as well as criteria for reporting variants, varied among labs, possibly affecting sensitivity and specificity of testing. Moreover, all panels were designed to target only exonic regions and intron-exon boundaries, missing possible pathogenic variants and rearrangements within noncoding regions. Also, GBA1 complex rearrangements could have been missed or underestimated using this NGS approach because of the very high sequence homology of GBA1 with its pseudogene, leading to misalignment of reads. Indeed, only two complex alleles were reported in the entire cohort. Furthermore, as mentioned earlier, LRRK2 carrier frequency could also have been underestimated because of previous screenings of patients with a positive family history. Finally, the authors acknowledge the potential inclusion of genetic data formerly reported in other studies, because the occasional patients may have independently undergone sampling for diagnostic and research purposes tests in different centers, and with diverse pseudonymization.

In summary, in this article, we report the results of a project aimed at harmonizing the diagnostic testing and reporting strategies across different institutions in Italy. We confirm that GBA1 and PRKN are the commonest mutated genes in PD with early AO, and that a very small set of genes (including GBA1, LRRK2, PRKN, PINK1, and SNCA) accounts for nearly all mutated PD patients lacking atypical features. Assessment of variants according to the latest ACMG criteria led to change the diagnostic status in 18 of 186 patients, with relevant consequences for counseling of patients and their families. This highlights the importance of developing an international consensus strategy for PD genetic testing, variant interpretation, and reporting in the diagnostic setting.

Acknowledgments

We are grateful to the patients and families for taking part in this study. Open access funding provided by BIBLIOSAN.

    Author Roles

    (1) Research Project: A. Conception, B. Organization, C. Execution;

    (2) Statistical Analysis: A. Design, B. Execution, C. Review and Critique;

    (3) Manuscript: A. Writing of the First Draft, B. Review and Critique.

    A.D.F.: 1A, 1B, 2A, 2C, 3A, 3B

    M.P.: 1A, 1B, 1C, 2A, 2B, 3A

    E.M.: 1A, 1B, 1C, 2C, 3B

    I.P.: 1B, 1C

    A.A.: 1B, 1C, 3B

    M.A.: 1B, 1C

    A.B.-S.: 1B, 1C

    F.B.: 1B, 2C, 3B

    G.B.: 1B, 1C

    G.C.-B.: 1B, 1C

    R.C.: 1B, 1C

    S.C.: 1A, 1B, 1C, 2C, 3B

    I.C.: 1B, 1C

    G.P.C.: 1A, 1B, 1C, 2C, 3B

    G.C.: 1B, 1C

    R.F.: 1B, 1C

    C.G.: 1B, 1C

    S.G.: 1A, 1B, 1C, 2C, 3B

    B.G.: 1A, 1B, 1C, 2C, 3B

    A.G.: 1B, 1C

    E.G.: 1A, 1B, 1C, 2C, 3B

    F.I.: 1B, 1C

    P.M.: 1A, 1B, 1C, 2C, 3B

    R.M.: 1A, 1B, 1C, 2C, 3B

    C.P.: 1B, 1C

    C.R.: 1B, 1C

    L.T.: 1B, 1C

    S.Z.: 1B, 1C

    P.C.: 1A, 1B, 2C, 3B

    E.M.V.: 1A, 1B, 1C, 2A, 2B, 2C, 3A, 3B

    Financial Disclosures

    This work was supported by grants of the Italian Ministry of Health (Ricerca Corrente Reti 2021-2022-PARKNET project). A.D.F. reports advisory board fees from Sanofi and speaking honoraria from Sanofi and Zambon. A.A. is Specialty Chief Editor of Frontiers in Neurology and President-Elect of the International Association for Parkinsonism and Related Disorders. A.A. received speaker's honoraria from Merz Pharma and Ipsen Pharma. G.C.B. received speakers's honoraria from Bial Italia. G.P.C. reports participation to advisory boards of Roche and Biogen. S.Z. received financial support from National Research council CNR. P.C. received speakers's honoraria from Abbvie. E.M.V. is Associate Editor of Journal of Medical Genetics; is Genetics Section Editor of Pediatric Research, of The Cerebellum, and of Neurological Sciences; is member of the Editorial Board of Movement Disorders Clinical Practice; is member of the Steering Committee of ASAP GP2 (Global Parkinson genetic Program). E.M.V. received research support from the Italian Ministry of Health, CARIPLO Foundation, Telethon Foundation Italy, Pierfranco and Luisa Mariani Foundation, and European Community (Eranet Neuron). None of the other authors report any conflict of interest.

    Data Availability Statement

    The complete list of variants is available in Table S2. The data that supports the findings of this study are available in the supporting information of this article.