NGLY1 encodes N-glycanase 1, a cytosolic deglycosylase that strips N-linked glycans from misfolded glycoproteins after they are retrotranslocated from the endoplasmic reticulum to the cytosol, a critical step in endoplasmic reticulum-associated degradation, or ERAD. Loss of NGLY1 also cripples the activation of NFE2L1, a transcription factor neurons and other cells rely on to rebuild the proteasome after proteotoxic stress. Clinically, the disorder presents as a multisystem neurodevelopmental syndrome featuring developmental delay, movement abnormalities, alacrima, seizures, peripheral neuropathy, and variable liver involvement. The allelic spectrum is broad, spanning nonsense, missense, frameshift, and splice-altering variants, and genotype-phenotype relationships remain incompletely resolved. Because prior human neuronal models had characterized NGLY1 loss largely within single genetic backgrounds, the central question of whether distinct patient genotypes converge on a shared molecular state or follow divergent routes had never been tested within a single system.

To resolve it, the team generated induced pluripotent stem cell-derived glutamatergic neurons from three patients carrying distinct biallelic genotypes: R401X/R401X, a homozygous nonsense variant in the PNGase domain predicted to yield no functional protein; Q208X/G310G, a nonsense plus splice-affecting combination expected to cause partial loss through truncation or aberrant splicing; and L318P/R390P, two missense variants in the TG domain predicted to destabilize the protein or reduce its activity. Two healthy control lines served as references. The neurons were differentiated and cultured for 28 days, then profiled using matched transcriptomics and proteomics, RNA-to-protein concordance analysis, differential exon usage testing, integrated pathway and phenotype network mapping, and longitudinal multielectrode array recordings during early network maturation. Immunofluorescence at day 28 confirmed that all lines produced extensive MAP2-positive dendritic processes and interconnected networks, though the NGLY1-deficient cultures showed greater cellular clustering and a less even distribution of processes, hinting that mutation-dependent alterations were already apparent.

The molecular results were unambiguous. Principal component analysis of both RNA-seq and proteomics data cleanly separated the three disease lines from the pooled controls, with each patient line occupying its own distinct region of the space. Differential expression revealed a graded response across genotypes: the R401X/R401X line yielded 9,040 differentially expressed genes, the L318P/R390P line 5,573, and the Q208X/G310G line 2,509, while at the protein level the counts were 1,081, 236, and 113, respectively. Crucially, these changes were predominantly line-specific. The largest transcriptomic intersection was unique to the R401X/R401X line, with only 404 genes shared across all three, and just 7 proteins were altered in common. Consistent with the underlying genotype, NGLY1 itself was significantly downregulated at both the transcript and protein level across all three lines, defining a small conserved cross-omics feature.

One of the most technically striking findings concerns what happens between the transcriptome and the proteome. When the team restricted both assays to the 4,128 gene products detected in common, most showed no significant change in either layer, but among those that did change, the largest classes reflected regulation detected at the RNA level alone. Spearman correlations between RNA and protein effect sizes were positive but modest, rising from 0.30 in the Q208X/G310G line to 0.40 and 0.57 in the other two, and Deming regression slopes below unity in every line, ranging from 0.21 to 0.44, indicated that protein fold changes were consistently compressed relative to their transcript counterparts. Yet the direction of change agreed strongly among features significant in both assays, exceeding 97 percent in two lines. The team ruled out statistical power and coverage artifacts as sole explanations: a simulation in which protein effect sizes were set equal to the observed RNA effects predicted far more differentially abundant proteins than were actually observed, indicating genuine post-transcriptional attenuation. This is mechanistically coherent with NGLY1 biology, since NGLY1 loss impairs ERAD and NFE2L1-mediated proteasome regulation and thereby alters protein turnover independently of transcription. A plausible interpretation is that transcriptional dysregulation in NGLY1-deficient neurons is partially buffered at the protein level by homeostatic control of protein abundance, a notable observation given the central role of NGLY1 in protein quality control.

Differential exon usage analysis added a second, non-mutually-exclusive contributor to this discordance: regulation at the level of transcript structure. Exon-usage changes were widespread in all three lines and, critically, were enriched among RNA-to-protein discordant genes in a mutation-dependent manner, with an odds ratio of 2.82 for protein-level discordance in the Q208X/G310G line but absent or depleted effects in the others. Because gene-level RNA-seq quantification aggregates reads across isoforms, a shift toward nonproductive or nonsense-mediated decay-targeted isoforms can reduce protein output without a proportional change in total transcript signal, offering a direct mechanism for a subset of the observed divergence. NGLY1 itself showed significant gene-level differential exon usage in two lines, and coordinated exon-usage changes were evident across canonical unfolded protein response and ER-proteostasis effectors and across the core nonsense-mediated decay machinery, indicating that NGLY1 deficiency remodels the very pathways responsible for ER quality control and transcript surveillance. The team also detected a focal, significant usage change at XBP1 in the R401X/R401X line, consistent with line-specific engagement of the unfolded protein response, though they note this signal is suggestive rather than definitive because the IRE1-mediated XBP1 splicing event involves a 26-nucleotide unconventional intron below the resolution of exon-bin counting.

Pathway analysis moved from individual genes to coordinated biological programs and revealed a stratified architecture. A single program, extracellular matrix, collagen and adhesion, was over-represented in every line and every assay layer and reached concordant significance even in the least affected line, marking it as the principal convergent axis of NGLY1 deficiency across genetic backgrounds. Beyond this shared core, most programs were engaged to markedly different extents and frequently in opposing directions across lines: muscle and contractile differentiation, ion transport and membrane potential, GPCR and receptor signaling, development and morphogenesis, and synaptic transmission were broadly affected but direction-divergent, so the same program could be activated in one genotype and repressed in another. The R401X/R401X line was by far the most transcriptionally and translationally engaged, uniquely reaching significance for proteasome, ERAD and ubiquitin quality control and for mitochondrial energy metabolism, alongside strongly upregulated translation machinery and broadly downregulated synaptic programs. The L318P/R390P line was intermediate and predominantly RNA-driven, with downregulation of trans-synaptic signaling and neuronal system pathways, while the Q208X/G310G line showed the most restricted response, concentrated on extracellular matrix organization and collagen formation.

To connect these molecular findings to the clinical presentation, the team constructed an integrative network linking concordant RNA and protein markers to their enriched pathways and, in turn, to the documented symptom domains of NGLY1 deficiency derived from Human Phenotype Ontology annotations. A core set of markers was altered concordantly across all three lines, including secreted and matrix-associated factors such as CXCL14, PTN, S100A6, GPNMB, SORCS2, SFRP2, and MTTP, several with upregulated direction across backgrounds. Around this shared core, each line contributed its own specific markers, for example RELN, ISG15, CAMK2B, MAPT, and CSPG5 in the R401X/R401X line and TFAP2B, NFIA, PLCB1, and NRGN in the L318P/R390P line. Crucially, the convergent matrix and adhesion signature and the line-specific synaptic and neuronal programs together plausibly account for the multisystem symptom spectrum of the disorder, mapping onto the most penetrant clinical features including growth impairment, feeding and swallowing dysfunction, liver injury, neurodevelopmental impairment, hypotonia, sensorimotor neuropathy, hyperkinetic movement disorder, speech impairment, seizures, and cerebral and cerebellar atrophy. The authors emphasize that these marker-to-symptom edges are ontology-based annotations used for hypothesis generation rather than causal or diagnostic inference.

The functional validation layer came from longitudinal multielectrode array recordings, which are non-invasive and permit repeated measurements of network maturation throughout the culture. All lines exhibited spontaneous activity from day 11 onward, and control cultures showed progressive increases in active electrodes, weighted mean firing rate, and both individual and network bursting, reflecting normal maturation. In contrast, NGLY1-deficient neurons exhibited reduced mean firing rate, a slower progressive increase in active electrodes, delayed onset of bursting, and altered network bursting parameters. Differences were genotype-specific and could indicate either hypoexcitable or hyperexcitable phenotypes: the L318P/R390P line displayed an increased number of network bursts relative to controls, suggesting hyperexcitability, yet a smaller number of spikes per network burst suggested weaker network strength and hypoexcitability. The R401X/R401X line showed a distinctive pattern of few network bursts but very high spikes per burst. Principal component analysis of 49 multielectrode array parameters at day 29 showed distinct, non-overlapping clustering of each patient line, mirroring the separation seen in the transcriptomic and proteomic data and suggesting a functional connection with the molecular consequences of each genotype. The team also found distinct differences in the profiles of ionotropic glutamate receptor and GABA receptor subunits across lines, which could explain the mix of excitability changes, potentially compounded by deficits in GABA production previously reported in NGLY1-deficient midbrain organoids.

The authors are candid about the central constraint of their comparative design: with a single patient line per genotype, genotype and genetic background are not separable, and the study does not include isogenic gene-corrected controls. They therefore treat the findings as discovery-stage and hypothesis-generating rather than a definitive attribution of specific features to NGLY1 loss, a constraint intrinsic to comparative work in an ultra-rare disorder where patient material is fundamentally limited. Their immediate next direction is to incorporate isogenic gene-corrected controls to determine which of the prioritized features are directly attributable to NGLY1 variation rather than to background-dependent variation, building on their own prior work that applied CRISPR-corrected isogenic controls to R401X iPSC-derived neural stem cells. A second boundary is the restriction to a single neuronal lineage, and brain organoid models constitute a complementary extension that would allow these signatures to be evaluated within a more physiologically diverse environment. A final consideration concerns enzymatic activity: although NGLY1 transcript and protein abundance were reduced, residual deglycosylation activity may vary across genotypes, and direct assessment of genotype-resolved catalytic activity in differentiated neurons remains an important component of future validation. More broadly, the work underscores the importance of accounting for genetic heterogeneity when investigating disease mechanisms in rare neurodevelopmental disorders, where overlapping clinical phenotypes may mask substantially divergent underlying biology. For biomarker development, the message is clear from the outset: heterogeneous downstream biology can coexist with partially overlapping clinical presentations, and candidate markers identified in one model background may not generalize across the full disease spectrum, so a panel-based approach capturing multiple endpoints is likely more appropriate than a one-metric-fits-all model.

Subject of Research: NGLY1 deficiency

Article Title: Genotype-specific molecular and functional trajectories in NGLY1 deficiency: a comparative multi-omics study of patient-derived iPSC neurons

Article References: Amniouel, S., Shyr, Z. A., Borschel, W. F., Tambe, M., Farkhondeh, A., & Zheng, W. (2026). Genotype-specific molecular and functional trajectories in NGLY1 deficiency: a comparative multi-omics study of patient-derived iPSC neurons. Frontiers in Cell and Developmental Biology, 14, Article 1966226. https://doi.org/10.3389/fcell.2026.1966226

Image Credits: AI Generated

DOI: 10.3389/fcell.2026.1966226

Keywords: NGLY1 deficiency, rare diseases, iPSC neurons, multi-omics, genotype heterogeneity, MEA, proteomics, transcriptomics, Genotype-specific, molecular, functional, trajectories