Lysoplex: An efficient toolkit to detect DNA sequence variations in the autophagy-lysosomal pathway.

The autophagy-lysosomal pathway (ALP) regulates cell homeostasis and plays a crucial role in human diseases, such as lysosomal storage disorders (LSDs) and common neurodegenerative diseases. Therefore, the identification of DNA sequence variations in genes involved in this pathway and their association with human diseases would have a significant impact on health. To this aim, we developed Lysoplex, a targeted next-generation sequencing (NGS) approach, which allowed us to obtain a uniform and accurate coding sequence coverage of a comprehensive set of 891 genes involved in lysosomal, endocytic, and autophagic pathways. Lysoplex was successfully validated on 14 different types of LSDs and then used to analyze 48 mutation-unknown patients with a clinical phenotype of neuronal ceroid lipofuscinosis (NCL), a genetically heterogeneous subtype of LSD. Lysoplex allowed us to identify pathogenic mutations in 67% of patients, most of whom had been unsuccessfully analyzed by several sequencing approaches. In addition, in 3 patients, we found potential disease-causing variants in novel NCL candidate genes. We then compared the variant detection power of Lysoplex with data derived from public whole exome sequencing (WES) efforts. On average, a 50% higher number of validated amino acid changes and truncating variations per gene were identified. Overall, we identified 61 truncating sequence variations and 488 missense variations with a high probability to cause loss of function in a total of 316 genes. Interestingly, some loss-of-function variations of genes involved in the ALP pathway were found in homozygosity in the normal population, suggesting that their role is not essential. Thus, Lysoplex provided a comprehensive catalog of sequence variants in ALP genes and allows the assessment of their relevance in cell biology as well as their contribution to human disease.

Introduction

The lysosome plays a central role in cellular clearance and energy metabolism by ensuring degradation and recycling of a variety of substrates. While extracellular substrates reach the lysosome via the endocytic and phagocytic pathways, intracellular material follows the autophagic pathway, mainly through the fusion of autophagosomes with lysosomes. In addition, recent studies have revealed that the lysosome is also involved in signaling pathways that play a crucial role in cell homeostasis and growth.

The physiological importance of lysosome-mediated pathways is underlined by their involvement in human diseases. Lysosomal storage diseases are due to mutations in more than 50 genes encoding lysosomal soluble acidic hydrolases, integral membrane proteins, activators, transporters, or nonlysosomal proteins that are necessary for lysosomal function. The precise molecular diagnosis is very important as new therapies are becoming available for specific types of LSDs. Unfortunately, for the majority of these disorders, the diagnosis is difficult due to considerable clinical overlap and clinical variability. This is particularly true for neuronal ceroid lipofuscinosis, a genetically heterogeneous subtype of LSDs with a devastating phenotype that can be caused by mutations of at least 13 different genes. Mutations in the autophagy-lysosomal (ALP) pathway can also cause common neurodegenerative disorders, such as Parkinson and Alzheimer diseases. It is likely that mutations in the ALP pathway will be identified in an increasing number of diseases.

Next-generation sequencing is revolutionizing our view of life sciences by allowing the investigation of variations in DNA/RNA sequences at a genome scale. In the last few years, whole exome sequencing or whole genome sequencing (WGS) approaches have received unlimited consideration as universal tests for most Mendelian conditions, with the exceptions of those caused by complex structural variations. As a cheaper and more focused alternative to WGS, WES has been demonstrated to be an excellent and cost-effective solution with a consistent number of advantages and, above all, a higher throughput on coding sequences that favors the identification of novel disease genes. However, accumulating data indicate that WES may not be a totally reliable tool to assess sequence variations in a given gene. Efficiency of WES is undermined by nonuniform targeting and coverage that, for a significant fraction of the exome, is too low to allow thorough variant detection. In addition, false positive calls are frequently observed in segmental duplications and processed pseudogenes. This reduces WES sensitivity as well as specificity, leading to labor-intensive validation steps usually based on conventional PCR reactions and Sanger sequencing. For practical reasons, the validation steps are usually restricted to specific genes, missing the remaining variability of the exome. Finally, there is an increasing need to properly manage incidental findings detected by WES/WGS.

NGS-based targeted approaches, i.e. the sequencing of selected genes and genomic regions of interest, have been developed as an alternative to WES/WGS. Using targeted approaches the ethical issues related to incidental findings are very limited. Our aim was to develop a robust sequencing platform to identify the sequence variations in the ALP pathway. A similar approach was used to detect sequence variations in mitochondrial pathways.

A potential obstacle to building a panel of bona fide ALP genes is represented by the fact that the identity of many key players is still unknown. Therefore, we used multiple bioinformatic tools to identify a comprehensive collection of ALP genes. The final, nonredundant, ALP list contains 891 genes, including 106 autophagy genes, 194 lysosomal genes, and 627 genes with a role in the endocytic pathway ().

Figure 1.

Subsets of genes that are part of the ALP gene list. Schematic diagram showing the composition of the Lysoplex gene list (n = 891 genes). Please note that some of the selected genes belong to more than one category.

Approximately 140 out of the total 891 genes, are mutated in human Mendelian diseases. We used the Haloplex enrichment protocol, which permitted accurate sequencing and high sequence coverage. This workflow was named Lysoplex.

Results

Generation of the ALP gene list

We selected the following sets of genes to generate the Lysoplex list: 1) a subset of genes encoding proteins with either an established role in autophagy or predicted by at least 2 of the following bioinformatics tools: AmiGO, Netview, UniProt, and MSigDB (see also Materials and Methods and Table S1); 2) a set of genes encoding proteins with lysosomal localization selected using a similar procedure involving both literature search and bioinformatic predictions; 3) all genes that are mutated in LSDs, including those coding for nonlysosomal proteins and the transcription factor TFEB, a master regulator of lysosomal biogenesis and autophagy, as well as 3 closely related paralogs (Mit/TFE members); 4) genes with a known/predicted role in the endocytic pathway; 5) genes involved in familial forms of Parkinson and Alzheimer diseases. As a result, the ALP list includes 891 genes (Table S1), for which all predicted exons and at least 10 nucleotides of flanking intronic sequences were included. The target region collectively spans 2.054 megabases of human DNA coding sequence ().

Lysoplex design and validation with mutation-known samples

To obtain a homogeneous and accurate coverage of the 891 ALP genes selected as indicated above, we designed capture oligonucleotide probes using the HaloPlex™ Target Enrichment System, which represents an excellent technology for complex targeting with high sensitivity and specificity. This procedure is based on restriction enzyme digestion of genomic DNA with multiple endonucleases and hybridization capture with probes that also work in the subsequent amplification step. We used this platform to sequence ∼100 DNA samples. On average, sequence reads that mapped to the ALP gene target represented more than 92% of total reads (), showing efficient and specific probe capture. This minimal “off-targeting” reduces the sequencing demand in comparison with other enrichment protocols and increases on-target coverage. Using a high-throughput setting of 32 libraries/lane, the average coverage depth was 40x for 95% and 100x for 80% of the ALP-target, which is significantly higher than the average depth we reached on the same genes with WES (). This setting led to a high specificity in variant calls, as demonstrated by the validation of 128/128 well covered (n > 20) variants tested by Sanger sequencing (100%).

Figure 2.

Performance of the Lysoplex toolkit. (A) Average yield of the Lysoplex capture. More than 92% of sequence reads truly map on ALP targets, indicating a very limited off-targeting in comparison with other methods. (B) Comparison between Lysoplex and WES. Comparison of the coverage distribution on all the 891 ALP genes between whole exome sequencing (blue boxes) and Lysoplex (red boxes). (C, D) Comparison of sequence coverage between WES and Lysoplex in 2 disease genes, namely CLN3 (C) and MTMR2 (D). Red boxes, Lysoplex; blue boxes, WES. Bold lines represent the median of all samples tested, the boxes include 50% of the values and the whiskers represent minimum and maximum values, excluding the outliers that are plotted individually.

To test the sensitivity of Lysoplex, we utilized a training set of 16 genomic DNAs from patients affected by 14 different forms of LSDs, whose molecular diagnosis was already known except for 2 cases (Table S2). These patients were affected by the following types of LSDs: Mucopolysaccharidoses (MPS) types I, II, IIIB and VI, Fabry disease, Niemann-Pick disease type C1 (NPC1), mucolipidosis type III (MLIII), late infantile neuronal ceroid lipofuscinosis 5 (CLN5), Gaucher disease, multiple sulfatase deficiency (MSD), metachromatic leukodystrophy (MLD), Pompe disease, and Danon disease.

The application of the Lysoplex procedure to the training set of 16 LSD patients allowed us to detect all of the previously known mutations and to identify the disease-causing mutations in the 2 patients in whom molecular diagnosis had not been previously made (Table S2).

Lysoplex application to molecular diagnosis of NCLs

To test the efficacy of Lysoplex in the analysis of patients with unknown/elusive molecular defects, we applied it to the search for disease-causing mutations in patients with clinically diagnosed NCLs, which are caused by mutations in at least 13 different genes. We carried out Lysoplex on 48 NCL patients, whose gene defect had not been previously detected (Table S3).

describes in “exploded pies” the success rate of Lysoplex to identify causative variants in the NCL patients. We detected both pathogenic variants (homozygous or compound heterozygous) in 29/48 cases. In 3 additional cases (; L33, L57 and L58), we only found a single heterozygous mutation in known NCL genes (CLN3, CLN6), which were compatible with the observed phenotype. We postulate that these patients may harbor a second mutation in the same gene that is not detectable, e.g., a deep intronic mutation or a large copy number variation.

Figure 3.

Success rate of Lysoplex in NCL patients. Schematic diagram summarizing the results of the Lysoplex procedure in the NCL patients analyzed. An overall detection rate of 67% was observed. In particular, mutations were identified in 8 NCL known genes. In one case, we found pathogenic mutations in the GLB1 gene (see text for details).

Table 1.

Summary of mutations identified in NCL patients analyzed

	Patient ID	Nationality	Age of onset	First symptoms	Pre-screened Genes/ Enzymes	Allele 1	Genomic coordinates allele 1 (hg19)	Allele 2	Genomic coordinates allele 2 (hg19)
Neuronal ceroid lipofuscinosis 1 (CLN1/PPT1) NM_000310	L137	Indian*	late infantile	Epilepsy		c.217G>T; p.Gly73Trp¥	chr1:40558087	c.217G>T; p.Gly73Trp¥	chr1:40558087
L144	Indian*	infantile	Epilepsy		c.547G>T; p.Ala183Ser¥	chr1:40546149	c.547G>T; p.Ala183Ser¥	chr1:40546149
Neuronal ceroid lipofuscinosis 2 (CLN2/TPP1) NM_000391	L42	German	juvenile	Visual loss	CLN3, CLN5, CLN6, CLN8	c.622C>T; p.Arg208X	chr11:6638271	c.622C>T; p.Arg208X	chr11:6638271
L150	Lybian*	late infantile	Ataxia, epilepsy		c.1094G>A; p.Cys365Tyr	chr11:6637287	c.1094G>A; p.Cys365Tyr	chr11:6637287
L133	Indian*	late infantile	Ataxia, epilepsy		c.379C>T; p.Arg127X	chr11:6638858	c.379C>T; p.Arg127X	chr11:6638858
Neuronal ceroid lipofuscinosis 3 (CLN3)	L33	German	juvenile	Dementia	CLN3	c.1151T>C; p.Leu384Pro¥	chr16:28489104	n.f.	n.f.
NM_000086	L128	Egyptian*	juvenile	Ataxia, epilepsy		c.1173_1174delAT; p.Tyr392X¥	chr13:77575052	c.1173_1174delAT; p.Tyr392X¥	chr13:77575052
Neuronal ceroid lipofuscinosis 6 (CLN6) NM_017882	L21	Swiss	late infantile	Ataxia	CLN1, CLN2, CLN3	c.150C>G;p.Tyr50X	chr15:68510922	c.272_273insAACG; p.Val91GlufsX42	chr15:68506653
L24	Turkish*	early juvenile	Dementia, epilepsy	CLN1, CLN2, CLN3	c.311C>T; p.Ser104Phe	chr15:68504188	c.311C>T; p.Ser104Phe	chr15:68504188
L30	German	juvenile	Dementia, epilepsy	CLN1, CLN2, CLN3	c.373A>G; p.Ser125Gly¥	chr15:68504126	c.407G>A; p.Arg136His¥	chr15:68504092
L31	German	early juvenile	Dementia, epilepsy	CLN1, CLN2, CLN3	c.198+5C>T spl¥	chr15:68510869	c.659A>T; p.Tyr220Phe¥	chr15:68501981
L39	Turkish*	late infantile	Ataxia, epilepsy	CLN1, CLN2, CLN3	c.896C>T; p.Pro299Leu	chr15:68500518	c.C896T; p.Pro299Leu	chr15:68500518
L50	German	late infantile	Ataxia, epilepsy	CLN1, CLN2, CLN3	c.5A>G;p.Glu2Gly¥	chr15:68521918	c.5A>G;p.Glu2Gly¥	chr15:68521918
L57	Italian	adult	Visual loss, dementia	CLN1, CLN2, CLN3	c.728C>T; p.Ala243Val¥	chr15:68500686	n.f.	n.f.
L58	Italian	adult	Visual loss, dementia	CLN1, CLN2, CLN3	c.728C>T; p.Ala243Val¥	chr15:68500686	n.f.	n.f.
L129	German	late infantile	Ataxia, epilepsy		c.502C>G; p.Leu168Val¥	chr15:68500588	c.826T>G; p.Trp276Gly¥	chr15:68500588
L130	Indian*	late infantile	Ataxia, epilepsy		c.297G>T; p.Lys99Asn¥	chr15:68506628	c.297G>T; p.Lys99Asn¥	chr15:68506628
Neuronal ceroid lipofuscinosis 7 (CLN7/MFSD8) NM_152778	L20	German	juvenile	Ataxia, epilepsy	CLN1, CLN2, CLN3	c.1439G>T; p.Gly480Val¥	chr4:128841903	c.1439G>T; p.Gly480Val¥	chr4:128841903
L22	Swiss	juvenile	Ataxia, epilepsy	CLN3	c.881C>A; p.Tyr294Lys	chr4:128851955	c.881C>A; p.Tyr294Lys	chr4:128851955
L23	Tamilian	late infantile	Ataxia, epilepsy	CLN1, CLN2, CLN3	c.1394G>A; p.Arg465Gln	chr4:128841948	c.1394G>A; p.Arg465Gln	chr4:128841948
L40	Turkish*	late infantile	Ataxia, epilepsy	CLN1, CLN2, CLN3, CLN6, CLN8	c.1235C>T; p.Pro412Leu	chr4:128842794	c.1235C>T; p.Pro412Leu	chr4:128842794
L49	German	late infantile	Ataxia, epilepsy	CLN3, CLN6, CLN8	c.1103-2delA spl	chr4:128842927	dup17nt¥	chr4:128870968
L51	Indian*	late infantile	Ataxia, epilepsy	CLN1, CLN2	c.1420C>T; p.Gln474X	chr4:128841922	c.1420C>T; p.Gln474X	chr4:128841922
L56	Turkey*	late infantile	Ataxia, epilepsy	CLN1, CLN2,	c.1394G>A; p.Arg465Gln	chr4:128841948	c.1394G>A; p.Arg465Gln	chr4:128841948
L140	Indian*	late infantile	Ataxia, epilepsy		c.499delA; p.Ser167ValfsX6¥	chr4:128863253	c.499delA; p.Ser167ValfsX6¥	chr4:128863253
Neuronal ceroid lipofuscinosis 8 (CLN8) NM_018941	L37	Turkish*	late infantile	Ataxia, epilepsy	WES	c.789G>C; p.Trp263Cys	chr8:1728661	c.789G>C; p.Trp263Cys	chr8:1728661
L38	Turkish*	late infantile	Ataxia, epilepsy	WES	c.789G>C; p.Trp263Cys	chr8:1728661	c.789G>C; p.Trp263Cys	chr8:1728661
L60	German	late infantile	Ataxia, epilepsy	CLN1, CLN2, CLN3	c.598_599delAT; p.Met200ValfsX46¥	chr8:1728469	c.620T>G; p.Leu207Arg¥	chr8:1728492
L139	Indian*	late infantile	Ataxia, epilepsy		c.473A>G; p.Tyr158Cys	chr8:1719693	c.473A>G; p.Tyr158Cys	chr8:1719693
L141	Dutch	juvenile	Dementia, epilepsy		c.595C>G; p.His199Asp¥	chr8:1728467	p.Glu167X¥	chr8:1719719
Neuronal ceroid lipofuscinosis 11 (CLN11/GRN) NM_002087	L36	Turkish*	late infantile	Ataxia, epilepsy	CLN3, CLN6, CLN8	c.359C>A; p.Ser120Tyr¥	chr17:42427605	c.359C>A; p.Ser120Tyr¥	chr17:42427605
GM1-gangliosidosis (GLB1)	L44		early juvenile		CLN3	c.602G>A; p.Arg201His	chr3:33099712	c.622C>T; p.Arg208Cys	chr3:33099692

reported consanguinity

novel mutations

n.f. = not found

In total, pathogenic mutations were identified in 32 NCL patients in the following disease genes: CLN1 (n = 2), CLN2 (n = 3), CLN3 (n = 1), CLN5 (n = 1), CLN6 (n = 10), CLN7 (n = 8), CLN8 (n = 5), CLN11 (n = 1) and GLB1 (n = 1). Patient L44 turned out to be compound heterozygous for 2 mutations in the GLB1 gene, encoding β-galactosidase. Interestingly, mutations in GLB1 are responsible for GM1-gangliosidosis, a neurodegenerative type of LSD whose phenotype partially overlaps with that of CLN1. Overall, in all patients analyzed, the molecular diagnosis was in agreement with the clinical diagnosis, taking into account the phenotypic variability observed in NCLs. This is further exemplified by the fact that mutations in the CLN6 gene can cause both late infantile as well as adult NCL phenotypes ().

Of note, patient L42 had been previously proposed to represent a novel CLN locus, namely CLN9, based on a distinctive fibroblast phenotype, gene expression pattern, ultrastructural analysis of storage material, enzymatic activities, and sequencing of NCL genes. However, we found that this patient carries a homozygous CLN2 mutation.

Taken as a whole, the use of Lysoplex allowed us to identify causative mutations in 67% of the NCL patients () Eighteen patients had never received any test (“naïve” group), whereas the remaining 30 individuals had been tested for some NCL genes (“pre-screened”) but without positive findings. Interestingly, one pre-screened family had previously been studied by both linkage analysis and WES (; patients L37 and L38). There was, however, no substantial difference in the success rate between the pre-tested and the naïve group.

Interestingly, in addition to the causative mutations in a given disease gene, we also identified in some patients additional mutations in genes involved in other diseases (Table S4). This may add further complexity to the genotype/phenotype correlation.

We then looked for recessive mutations in novel NCL candidate genes, which could exert a pathogenic role in the 16 patients in whom Lysoplex could not identify any mutations in known disease genes. We focused our attention on loss-of-function variants such as frame-shift, nonsense, splice defects, and potentially damaging nonsynonymous amino acid substitutions in nondisease genes. We considered as novel candidate NCL genes those that showed these variants on both alleles in at least one NCL patient or those mutated in 2 unrelated patients. Following these selection criteria, we identified 3 potential candidate genes: STAB2 (stabilin 2), AGAP1 (ArfGAP with GTPase domain, ankyrin repeat and PH domain 1) and PLCG2 (phospholipase C, gamma 2 [phosphatidylinositol-specific]) (Table S5). However, functional analyses and in vivo studies in animal models are needed before we can conclude that these are causative mutations. Overall, we identified 37 different pathogenic mutations, including 21 novel mutations. They include 25 amino acid changes, 5 indels (insertions or deletions) including an insertion of 17 nucleotides, 5 nonsense mutations, and 2 splice site variants ().

Sequence variation in ALP genes

To assess the efficiency of Lysoplex in detecting sequence variations in ALP genes, we compared its performance to pre-existing public resources, primarily the Exome Variant Server (EVS). To avoid any bias due to sample selection, we first counted the number of synonymous variations/allele/kb in our patient cohort as compared with EVS data. We plotted each gene as a dot in a logarithmic scale where the X-axis represents the number of sequence variations/allele/kb in EVS and the Y-axis represents the same value obtained in the Lysoplex panel (). This analysis revealed that the vast majority of genes show a much higher number of variations in the Lysoplex samples vs. EVS. These data were also confirmed for all other categories of sequence substitutions, including missense and loss-of-function mutations (). This demonstrates that Lysoplex is much more sensitive than EVS in detecting sequence variability in ALP genes. We did not consider small insertions or deletions that grossly deviate from Hardy-Weinberg equilibrium, since it is well known that a high number of false positive variants are erroneously annotated in EVS and in all other databases.

Figure 4.

Comparison between the number of sequence variations found by Lysoplex and WES. Plotted data indicate the log of the number of variations/kb found in each gene by Lysoplex (Y-axis) and by public database (X-axis). If the sensitivity was identical, dots should tend to align. Dots above the line indicate a superior sensitivity of Lysoplex. Red dots indicate NCL genes and blue dots other disease genes. (A) synonymous variations; (B) missense variations; (C) loss-of-function variations.

Overall, we found 61 clear loss-of-function mutations (i.e., nonsense, indels leading to frameshifts and splice variants) and 488 missense variations with a high probability to exert a damaging effect on the corresponding protein product. The above mutations were found in a total of 316 different genes (Table S6). Interestingly, some of the loss-of-function variations were also found in the normal population at a frequency higher than 1% (). Some of the genes (n = 6) carrying common loss-of-function variations in the normal population are involved in autophagy (ATG3), endocytosis (ATP6V1G3, CBLC, SEC14L4) and membrane trafficking (INPP5D). These findings suggest that total loss-of-function in these genes does not appear to have major effects.

Table 2.

List of common loss-of-function sequence variations (frequency > 1%) in ALP genes

Gene symbol and RefSeq	Gene name	Biological function and biological process	Genomic coordinates (hg19)	Sequence variation	N Het samples	N Hom samples	Freq in EVS
ATG3 (NM_001278712)	autophagy related 3	Autophagic vacuole assembly	chr3:112253058	c.921_922insT; p.Leu307fs	9	2	0.601022
ATP6V1G3 (NM_133326)	ATPase, H+ transporting, lysosomal 13kDa, V1 subunit G3	Transmembrane transport and endocytosis	chr1:198505831	c.106C>T; p.Arg36X	2	0	0.019689
CBLC (NM_001130852)	Cbl proto-oncogene C, E3 ubiquitin protein ligase	Cell surface receptor signaling pathway and endocytosis	chr19:45296846	c.1115_1116insC; p.Asp372fs	4	0	0.064297
INPP5D (NM_001017915)	inositol polyphosphate-5-phosphatase, 145kDa	Membrane trafficking	chr2:234066969	c.956C>G; p.Ser319X	28	20	0.569864
LPL (NM_000237)	lipoprotein lipase	Protein binding	chr8:19819724	c.1421C>G; p.Ser474X	10	1	0.086345
SEC14L4 (NM_001161368)	SEC14-like 4 (S. cerevisiae)	Golgi-to-plasma membrane protein transport	chr22:30891264	c.400G>T; p.Glu134X	1	0	0.032062

Discussion

In the past few years, remarkable progress has been made in the characterization of the autophagic lysosomal pathway, discovering wide-ranging pathogenic roles in addition to vital homeostatic functions. Autophagy is involved in various conditions such as genetic diseases, metabolic disorders, neurodegenerative diseases, infections, and cancer.

While great attention has been given to the study of the mechanisms underlying the regulation and execution of autophagy in health and disease, the intrinsic genetic variation of the system in the human population has so far been ignored. Lysoplex was developed to map such genetic variation. When compared with public genome variant databases, the Lysoplex panel provided a significantly more accurate list of genetic variants that may affect the ALP pathways, in normal individuals and in patients.

When applied to the molecular diagnosis of LSDs, Lysoplex revealed a number of significant advantages over WES or WGS approaches. For example, the data load was approximately 1/10 and 1/100 compared to WES and WGS, respectively. This reduces costs and efforts, making Lysoplex applicable to a large number of samples. In addition, Lysoplex proved to be much more robust in terms of target reproducibility, specificity, and sensitivity, and its depth was such that there was no need for Sanger validation steps (i.e. to confirm variations or distinguish between homozygous and heterozygous status). The extent of off-targeting was minimal (8%) although slightly higher than that obtained with Motorplex (2%). However, Motorplex covers a target region that is 4-fold smaller than Lysoplex. It is also important to underline that Lysoplex is a very versatile tool as the list of genes to be analyzed can be easily updated as new relevant genes are identified.

In our study we detected all disease-causing mutations in a set of LSD patients. Lysoplex was successful in identifying the molecular defect in 67% of the NCL patients analyzed, which is a significant result when considering both the locus and allele heterogeneity of this class of diseases. Remarkably, patient L42 belongs to one of the 2 pairs of NCL siblings of German and Serbian descent, respectively, who were previously proposed to be affected by mutations in an undefined CLN9 locus. Lysoplex analysis allowed us to reclassify L42 as CLN2. Since the other sib pair (of Serbian origin) was reclassified earlier after homozygosity mapping as CLN5, no further evidence for an elusive CLN9 gene exists. Two others of the pre-tested patients, L37 and L38 (), had been previously analyzed by WES, but the pathogenic mutations were originally missed due to unclear family history (presence of one additional sibling with an uncertain diagnosis). Furthermore, in one of the patients originally diagnosed as affected by NCL, we identified missense mutations in both alleles of the GLB1 gene, resulting in the molecular diagnosis of GM1 gangliosidosis. This disease is associated with accumulation of saposin and subunit c of the V-ATPase, and is characterized by common manifestation of ER and oxidative stress with a resulting phenotype that overlaps that observed in NCL. This suggests that some cases with a clinical diagnosis of NCL may indeed have GM1-gangliosidosis. There is another report in which a patient diagnosed with adult NCL was found to carry heterozygous mutations in SGSH defective in MPS IIIA (Sanfilippo syndrome A) and associated with lysosomal accumulation of heparin sulfate. Thus, Lysoplex appears to be instrumental for the differential diagnosis and correct classification between different types of LSDs. This result emphasizes the value of testing multiple genes for the same broad class of genetic disease (e.g., LSDs) with partially overlapping phenotypes.

Considering that ALP genes may be candidates for other genetic conditions, and that approximately 85% of them are not yet assigned to any disease, Lysoplex can effectively be used for the discovery of novel disease genes. Lysoplex application to a relatively small set of patients resulted in the identification of several homozygous or compound heterozygous mutations in putative novel disease genes for NCLs. These are: STAB2, that encodes a cell receptor that enables the scavenging and clearance of multiple ligands from the circulation salvaging their building block monomers by lysosomal degradation, as well as MAPK1/ERK2-MAPK3/ERK1 activated signaling, AGAP1 that encodes an endosome-associated, phosphoinositide-dependent ADP-ribosylation factor GTPase-activating protein, and PLCG2 that encodes a transmembrane signaling enzyme whose gain-of-function mutations have been associated with autoimmune and inflammatory diseases mediated by B cells. More data, such as the identification of additional families and/or functional assays, are needed to definitely prove the involvement of the above candidate genes in NCL pathogenesis.

Another application of Lysoplex may be in the field of modifier genes, complex disorders, and polygenic inheritance. It is well known that patients that share the same mutations may have different phenotypic spectrum. Thus, the effect of the primary molecular defects may be modified by the presence of additional and variable elements located in other genes that encode proteins involved in the same pathways. Remarkably, we have identified additional mutations in other disease genes in patients harboring pathogenic mutations in NCL genes (Table S4). This may represent disease modifiers that explain intra-familial phenotypic variability. However, to test their role as modifier genes, genotype-phenotype correlation studies are needed on a much larger number of cases.

There is increasing evidence that mutations in genes involved in the ALP pathway may predispose to common neurodegenerative diseases. Cohorts of patients affected by common neurodegenerative diseases, such as Parkinson and Alzheimer diseases, may be screened using Lysoplex to identify additional genes predisposing to such diseases. It may also be interesting to use Lysoplex for the screening of somatic mutations in ALP genes in tumor samples due to the known involvement of autophagy in cancer.

Finally, Lysoplex is an exceptionally efficient tool to map amino acid sequence variations in proteins encoded by ALP genes. Such amino acid sequence variations may have functional relevance. Indeed, we have observed a high frequency of rare, nonconserved, amino acid changes as well as truncating variants. The observation of such protein variation may not be a total surprise, considering the predicted human genome variability, but it is remarkable considering the crucial roles of ALP genes in cell homeostasis. Much of this variability is overlooked when performing global WES or WGS approaches. In practice, WES or WGS data are usually filtered and the discordant variability is not dissected gene by gene according to function. Furthermore, when the coverage is low, the variability is always submerged by numerous sequencing errors and gaps that, in contrast, do not affect Lysoplex.

This catalog of functional variants may serve 2 main purposes. It provides a useful tool for basic researchers to study the biological relevance of protein variants in the ALP pathway. At the same time, clinical researchers can use this catalog to identify candidate causative or modifier genes for human diseases.

Materials and Methods

Patients

The training set of 16 genomic DNA samples belonged to patients affected by 14 different LSDs (Table S2). In all cases, the enzyme activities were measured in cultured fibroblasts, using artificial fluorogenic substrates, according to standard procedures. NCL patients received a diagnosis based on clinical evaluation, biochemical assays, and electron microscopy analysis. Some of the patients had already undergone some molecular tests (). All procedures on patients were approved by the Ethics Boards of the participating institutions and adhered to the tenets of the Declaration of Helsinki.

Selection of Lysoplex genes

The complete gene list includes 891 members (Table S1): 98 previously described lysosomal genes; 9 genes colocalizing with the lysosomal marker LAMP2; the 4 MIT family members; all the known lysosomal and nonlysosomal protein-coding genes with a role in the different LSDs selected by using the Online Mendelian Inheritance in Man (OMIM).

Lysoplex also includes 10 genes involved in hereditary forms of Parkinson disease and 4 involved in hereditary forms of Alzheimer disease. Moreover, bioinformatics prediction tools, including AmiGO, Netview, and UniProt were used. For lysosomal genes, in each database the complete human gene list was downloaded restricting the output to the “lysosome” term: the AmiGO list included 306 genes, the Netview list included 484 genes and the UniProt list included 305 genes. The 3 lists were compared in pairs: 83 genes predicted to be lysosomal by at least 2 out of 3 tools were considered good candidate genes and thus added to the final list. In Table S1 this subset of genes is listed and associated with the specific bioinformatics tools' pair: 16 genes were shared by AmiGO and Netview lists, 63 genes were shared by AmiGO and UniProt lists, one gene was shared by Netview and UniProt lists and 3 genes were shared by all 3 lists.

As for the lysosomal genes, the selection of the 106 genes mainly playing a role in the autophagic pathway was performed by both literature analysis and bioinformatics tools. In detail, 73 genes were mapped to 12 autophagic subcategories on the GO hierarchy rooted in the “Autophagy”(GO:0006914) accession term; 2 more genes (SNX3 and CHPT1) were introduced because they were predicted as “autophagic vacuole” genes by Netview.

For the selection of genes with a known role in the endocytic pathway and trafficking, both literature analysis (n = 325) and the Molecular Signatures Database (MSigDB) were used. MSigDB, which is one of the most widely used repositories of well-annotated gene sets, was used to dissect lists of genes included in different “endocytosis” gene sets: we thus found 53 genes annotated by the AmiGO term (GO:0006897) “endocytosis”; 14 genes annotated by the GO term (GO:0030139) “endocytic vescicle”; 23 genes collected in the GO term (GO:0006898) “receptor mediated endocytosis”; 11 genes annotated by the GO term (GO:0030100) “regulation of endocytosis”; 178 genes included in the “endocytosis” term in KEGG and finally 23 genes included in the Reactome gene set term “transferrin endocytosis and recycling.”

Finally, the information about the specific cellular compartment, biological function and biological process associated with each Lysosplex gene was obtained by EntrezGene.

Targeted sequencing analysis

The libraries were sequenced using the HiSeq1000 system (Illumina inc., San Diego, CA, USA). The generated sequences were analyzed using an in-house pipeline designed to automate the analysis workflow and composed of modules performing every step using the appropriate tools available to the scientific community or developed in-house. Paired sequencing reads were aligned to the reference genome (UCSC, hg19 build) using BWA and sorted with SAMtools and Picard (http://picard.sourceforge.net). Genome Analysis Toolkit (GATK) with parameters adapted to the Haloplex-generated sequences was then used to perform local realignment around insertions-deletions (indel) and single nucleotide variants (SNV) and small indel calling. The called SNV and indel variants produced with both platforms were annotated using ANNOVAR with: the relative position in genes using the RefSeq gene model, amino acid change, presence in dbSNP v137, frequency in NHLBI Exome Variant Server (http://evs.gs.washington.edu/EVS), 1,000 genomes project, multiple cross-species conservation and prediction scores of damaging on protein activity. The annotated variants were then imported in the internal variation database, which stores all the variations found in internal sequencing projects. The database is then queried to generate the filtered list of variations and the internal database frequency in samples with unrelated phenotype is used as further annotation and filtering criterion. The alignments at candidate positions were visually inspected using the Integrative Genomics Viewer (IGV).