PACS1 syndrome variant alters the proteomic landscape of developing cortical organoids

Summary

Patient-derived brain organoids have been used to model NDDs and identify pathogenic mechanisms. While most studies have evaluated the mechanisms by which NDD variants affect the transcriptome, studies examining the proteome remain limited. Here, we examined the effect of the PACS1 p.R203W variant, which causes PACS1 syndrome, on the proteomic landscape of brain organoids and compared the results with a transcriptomic dataset. Although we observed limited overlap between transcripts and proteins, most dysregulated processes converged. The proteomic data not only reinforced the importance of these developmental processes in p.R203W pathogenesis but also revealed dysregulated pathways not identified in the transcriptomic analysis. Key dysregulated transcripts and proteins in PACS1(+/R203W) organoids were classified as NDD-causative and ASD-risk genes. Our results highlight the potential of proteomics to complement transcriptomic studies by providing deeper insight into the mechanisms underlying NDDs while revealing additional therapeutic targets that may not be identifiable through transcriptomics alone.

Graphical abstract

Highlights

• •Protein dysregulation emerges across maturation in PACS1 p.R203W organoids
• •Dysregulated processes suggest accelerated maturation and increased synaptogenesis
• •Low proteome-transcriptome overlap, yet dysregulated processes converge
• •Proteomic analysis uncovered dysregulation missed by transcriptomics

Neuroscience; Proteomics; Stem cells research

Introduction

Neurodevelopmental disorders (NDDs) are heterogenous and usually present with complex etiology. Several large-scale genetic studies investigating the cause of NDDs such as autism spectrum disorder (ASD), intellectual disability (ID), and epilepsy have been conducted, and have led to the identification of hundreds of risk genes. However, our understanding of the mechanisms by which the observed genetic variants act has lagged and consequently has limited translation of genetic findings into clinical treatments.^1,2

In recent years, patient-derived brain organoids have been increasingly used to identify pathogenic mechanisms and possible therapeutic targets.³ While most of these studies investigate the mechanisms by which risk genes and pathogenic variants affect the transcriptome,^{4,5,6,7,8,9,10} studies considering the proteome are limited.^11,12,13 However, mRNA levels do not always predict protein levels, as protein expression and degradation are regulated by mechanisms beyond transcription.¹⁴ Here, we uncover that investigating the proteomic landscape of brain organoids during maturation provides further resolution of NDD pathogenesis, adding to the mechanisms observed with transcriptomic analysis. To do so, we examine the effect of the pathogenic variant PACS1 p.R203W, causative of PACS1 syndrome, on the proteomic landscape of brain organoids using tandem mass tag (TMT) mass spectrometry.

PACS1 syndrome (MIM615009) is characterized by ASD, ID, craniofacial abnormalities, and epilepsy^15,16,17 and results from a recurrent de novo c.607C>T variant in phosphofurin acidic cluster sorting 1 (PACS1) that leads to an arginine to tryptophan (p.R203W) substitution. This missense variant is present in all reported patients except for one known arginine to glutamine p.R203Q substitution.¹⁷ PACS1 encodes a multifunctional sorting protein required for furin localization to the trans-Golgi network (tGN).^18,19 In addition to tGN retrieval, PACS1 mediates the recycling of endocytosed furin from early endosomes to the cell surface and the delivery of proteins to the primary cilium.^20,21,22 PACS1 has been shown to interact with over 50 other proteins^{20,23,24,25,26} via acidic cluster sorting motifs.^24,27,28,29 PACS1 is also present in the nucleus^30,31 where it mediates DNA replication repair,^32,33 contributes to the nuclear export of viral transcripts.³¹

Recent reports explore the pathogenesis of the PACS1 p.R203W variant. Mice overexpressing PACS1 p.R203W exclusively in cortical neurons present with increased dendrite arborization and decreased postsynaptic currents, possibly due to a pathogenic interaction between PACS1 p.R203W and HDAC6. Antisense oligonucleotides (ASOs) directed toward both PACS1 and HDAC6 ameliorate these defects.³⁴ Transcriptomic analysis in patient-derived brain organoids revealed a PACS1 p.R203W dependent dysregulation in cell projection organization in immature neurons, while mature neurons exhibited an upregulation in synaptic signaling processes. Subsequent characterization of spontaneous electrical activity revealed that induced pluripotent stem cell (iPSC)-derived neurons harboring the p.R203W variant display asynchrony and an increased network burst duration.⁹ Finally, the PACS1 syndrome-associated variant altered PACS1 localization but had no effect on neuron morphology and synaptic patterns in C. elegans.³⁵ Although these studies have started to investigate the impact of the PACS1 p.R203W variant, the mechanisms by which the variant affects neurodevelopment are still poorly understood. This in turn results in limited therapeutic options for patients, which at present only encompass non-invasive behavioral approaches and sensory integration therapies to manage symptoms.^36,37 Here we assess the developmental proteomic landscape in PACS1^(+/R203W) organoids to further understand PACS1 syndrome pathogenesis. We then compared the dysregulated process uncovered both by proteomic and transcriptomic approaches. Our findings underscore the importance of integrating data across multiple regulatory levels to capture the complex pathogenic mechanisms underlying NDDs.

Results

Brain organoids capture proteomic developmental trajectories and processes observed in fetal brain

Proteins are the ultimate effectors of cellular processes. Thus, here we characterized the proteome of iPSC-derived brain organoids to better understand how p.R203W PACS1 impacts neurodevelopment. Organoids were collected at distinct differentiation time points: day 21, when proliferating neural precursors predominate; day 40, when some progenitors have transitioned into intermediate progenitors and newly born neurons; and day 90, when post-mitotic neurons are the most abundant population (Figure 1A).

Figure 1.

Cortical organoids recapitulate proteome dynamics during brain development

(A) Expression of neural precursor protein SOX2 (red) and projection neuron marker CTIP2 (magenta) on day 21, 40, and 90 organoids derived from control (up) and PACS1 p.R203W organoids (bottom). Scale bars, 200 μm. DAPI (blue).

(B) Total proteins detected in organoids were clustered by abundance changes over maturation time into three major trends: upregulated (red), downregulated (blue) or stable (gray) (graph on the left). Further classification detected four main abundance trends, based on the timing of change: steady over time (clusters steady -up and -down), or delayed to day 40 (clusters delayed -up and -down) (right). Total number of proteins assigned to a cluster is indicated in each graph; black lines in all figures represent mean trend of all proteins in each cluster.

(C and D) Top GO terms upregulated (red) and downregulated (blue) by cluster over maturation time. The inner pie plots represent processes common to the clusters in each figure. The nested pie charts show root and branch significant GO enrichments with the width of the pie correlating to –log(p) of the GO term. Processes in bold represent those mentioned in the text.

(E) Top GO terms upregulated (red) and downregulated (blue) shared between organoids (purple) and fetal brain (green) over time. Synapse and neurodevelopment-related processes (box). Organoid profiles correspond to clusters 1 to 4, while the fetal brain profiles correspond to clusters A to H. A full description of all enriched processes is included in Table S10.

Using TMT mass spectrometry, we identified 3,766 unique proteins across organoids at different developmental stages. Approximately 60% of these proteins overlapped with those detected in the fetal brain, whereas overlap decreased when compared with the postnatal or adult brain (Figure S1A and Table S1). Higher correlation on protein expression was primarily observed within organoids and brain samples at earlier developmental stages (post-conception week 16 through 26), with protein abundance correlation increasing with organoid maturation (Figure S1B). First, we aimed to understand the proteomic landscape of maturing control brain organoids. We identified groups of proteins with common temporal trajectories by performing time series and clustering analysis. We classified proteins into three temporal patterns: upregulation, downregulation, and stable abundance over time. The up and downregulated patterns were further clustered into two main profiles: (1) continuous protein change, with steady increase/decrease in abundance (clusters “steady-up” and “steady-down”); and (2) delayed trend, where protein abundance remained stable during early differentiation (from day 21 to day 40), and subsequently increased/decreased at day 90 (clusters “delayed-up” and “delayed-down”) (Figure 1B and Table S2).

Next, we used gene ontology (GO) analysis to determine the biological processes involved within these distinct patterns (Table S3). As expected, many developmental processes changed with time, including neuron differentiation, axon extension, dendrite morphogenesis, synaptic organization/function upregulation (Figure 1C), and cell cycle and stem cell population maintenance downregulation (Figure 1D). This is consistent with the differentiation of neural precursors to neurons during organoid maturation. Within the upregulated clusters, the steady-up contained mostly neurodevelopmental processes, while the delayed-up was mostly associated with signaling. Likewise, clusters steady-down and delayed-down shared a downregulation of cell cycle and DNA replication or repair (Figure 1D). Stable proteins were associated with basic functions like primary metabolism, regulation of gene expression, and ribosome/translation-related processes (Figure S1C).

Finally, we compared the observed organoid processes with the developing fetal brain dynamics obtained from a reference dataset³⁸ (Figure S1D and Table S4). We found significant overlap between clusters in organoids and the fetal brain (Table S5). Most enriched processes during fetal brain development were captured in the corresponding organoid clusters. These results confirm that our brain organoids recapitulate early stages of brain development, as previously reported,³⁸ supporting their use to model protein dysregulation in PACS1 syndrome.

PACS1 p.R203W pathogenic variant alters developmental trajectories

We then profiled the proteome of PACS1 syndrome patient-derived brain organoids to evaluate the effects of the PACS1 p.R203W pathogenic variant on protein regulation during development (Figure 1A). Organoids carrying the p.R203W variant developed similarly to control organoids by day 21, with only five proteins (HMGB3, HTATSF1, HM13, TCEAL3 and TCEAL5) showing upregulation (Figure S2A). However, upon maturation, 329 proteins in PACS1^(+/R203W) organoids showed dysregulation when compared to controls (Table S6). Oxidative phosphorylation (ATP6V0A1, CYC1, NDUFA7) and intracellular calcium ion homeostasis (ATP1B1, CALB2, LETM1, NPTN) were among the observed enriched processes resulting from 167 upregulated proteins in organoids carrying the p.R203W variant (Figure S2B and Table S7). By day 40, most dysregulated proteins were related to established PACS1 functions relating to endoplasmic reticulum (ER)-to-Golgi transport (ARF4, SEC23A, SNAPIN). Upregulated proteins at day 40 were involved in neuron projection development, such as the ligand of Eph-related receptor tyrosine kinases EFNB1 and the transcription factor FOXG1. This was confirmed to be upregulated by the western blot (Figure S2C). In contrast, neuron projection development and synapse signaling related proteins like EFNB3, NF1, VDAC3, and SV2A were enriched by day 90. Other upregulated processes include the metabolism of amino acids (ASS1, GLO1, CTH, HSD17B10, PLOD2, SMS), and glycerolipids (PITPNM1, AGPAT4, TMEM68). Interestingly, the PACS1 interactors PTBP1,³⁹ which is involved in mRNA processing, and WDR37,⁴⁰ were also upregulated (Figures S2A and S2B). On the other hand, 173 down-regulated proteins were enriched in processes including cell cycle division, DNA replication, and mRNA splicing. This further included the RNA-binding protein PCBP4, which is involved in mRNA stability⁴¹ and RNA-binding/splicing regulator PTBP2, which is also involved in brain development. The intermediate filament protein NES, which plays a role in proliferation and differentiation,⁴² was also found to be among the downregulated proteins. Decreased abundance of NES was confirmed by the western blot (Figure S2C).

After the data normalization and exploration phase (Figure S3), principal component analysis (PCA) indicated maturation time to be the strongest driver of variance when comparing PACS1^(+/+) and PACS1^(+/R203W) organoids at days 21, 40, and 90 (Figures S3C–S3E). Consequently, we investigated whether the presence of the PACS1 variant impacts groups of proteins that share common temporal trajectories. We used the short time series (STEM) Java analytical pipeline to compare maturation trajectories.⁴³ PACS1^(+/R203W) organoids also presented with the three major patterns (upregulation, downregulation, and stable) and profiles (steady vs. delayed) observed in controls (Figure 2A and Table S8). However, several proteins switched abundance patterns (Figure 2B). The profile rearrangement resulted in four additional clusters, for a total of eight clusters found in PACS1^(+/R203W) organoids. Additional clusters included proteins that were up or down-regulated when compared to the stable pattern in controls (“stable-up” and “stable-down” clusters) and those that lost the regulation observed in controls and remained stable over maturation (“up-stable” and “down-stable” clusters; Table S9).

Figure 2.

Organoids carrying the PACS1 p.R203W pathogenic variant present with altered protein dynamics during maturation

(A) Alluvial plot indicating the change of profile clustering in the PACS1 p.R203W organoids (right) in contrast to the clustering in controls (left). Breaks in the y axis were introduced to improve visualization of the dysregulated clusters by removing 2,000 proteins that remained stable over time.

(B) Four clusters of dysregulated proteins were identified in PACS1 p.R203W organoids; black lines in all figures represent mean trend of all proteins in each cluster. The plots are organized into a gain or loss of regulation categories: cluster stable-up (top), and stable-down (bottom) gained regulation while clusters up- and down-stable lost regulation (middle), (C) Enriched biological functions per dysregulated cluster Stable-up (top, red), Down-stable (middle, greyish blue) and Stable-down (bottom, blue). The Up-stable cluster had no significant enrichment due to cluster size. The nested pie charts show root and branch significant GO enrichments and the width of the pie correlating to –log(p) of the GO term. Neurodevelopment-related processes in (bold).

The stable-up cluster from PACS1^(+/R203W) organoids had a strong enrichment in oxidative phosphorylation (23 out of 33 proteins), brain development, and neuron differentiation. Proteins that gain upregulation include glucose transporter GLUT1, transcription factor FOXG1, and mitochondrial genes UQCRQ, VDAC1, and ATP5PF (Figure 2C and Table S10). The stable-down cluster was enriched in regulation of mRNA processing and translation. Proteins that gain downregulation include ribosomal subunit protein RPL38, alpha subunit of the elongation factor-1 complex EEF1A1, and developmental proteins Nestin,⁴² BRD2,⁴⁴ and Wnt family member 7B WNT7B.⁴⁵

The down-stable cluster was enriched in the metabolism of the nucleotides process. Proteins related to metabolism that lost their down regulation and remained stable over time have also been associated with brain development and function. These proteins include, ribokinase (RBKS), implicated in glucose metabolic disorders and recently associated to psychiatric disorders,⁴⁶ phosphoribosylpyrophosphate synthetase 1 (PRPS1), implicated in multiple syndromic and non-syndromic diseases causing hearing loss, hypotonia, and ataxia,⁴⁷ as well as catalase (CAT), implicated in oxidative stress-related diseases and ASD.⁴⁸

Finally, the up-stable cluster was comprised eight proteins which lost their temporal upregulation and remained stable in PACS1^(+/R203W) organoids, with no significant enrichment. In this group, however, several proteins have relevant functions in the brain, such as dendrite morphogenesis (SS18L1/CREST),⁴⁹ autophagy, and neuron projection regeneration after injury (VIM),⁵⁰ and regulation of intracellular pH homeostasis (CA2).⁵¹

These results suggest altered developmental trajectories in the PACS1^(+/R203W) organoids with premature or ectopic upregulation of selected neuronal/metabolic programs and processes related to known PACS1 functions (i.e., Golgi to ER vesicle transport). Although the observed protein clustering and enrichment of neuronal and metabolic processes could be driven by altered cell-type proportions, scRNA sequencing, and immunostaining confirmed that the PACS1 p.R203W variant does not alter the cytoarchitectural composition or organization of organoids.⁹ To better understand the possibility of altered developmental trajectories in PACS1^(+/R203W) organoids, we performed pseudotime analysis from our previously generated scRNA-seq data.⁹ Despite similar organoid composition, we observed a significant shift in the developmental trajectories of PACS1^(+/R203W) organoids when compared to controls (KS test, all lineages: D = 0.014613, p = 0.01803; Figure S4). The main developmental trajectory bifurcated into two branches, representing immature neurons and mature neurons (Figures S4A and S4B). Pseudotime ordering showed PACS1^(+/R203W) organoids contained a significantly greater density of cells at later pseudotime values than controls within both the immature (Figures S4C and S4D, middle) and mature (Figures S4C and S4D, right) neuronal lineages. A one-sided Kolmogorov-Smirnov test confirms the cumulative distribution of pseudoage values in the PACS1^(+/R203W) organoids in contrast to controls in all tested subgroups (Figure S4). Further investigation is required to determine whether these transcriptomic and proteomic shifts reflect accelerated maturation due to the PACS1 p.R203W variant. However, although PACS1^(+/R203W) organoids exhibited signatures of early neuronal maturation, they also showed dysregulation of numerous mRNA-related processes, including transport and processing, which may contribute to defective or incomplete differentiation (Table S10). Finally, our results further uncovered the dysregulation of known PACS1 interactors at the protein level that were not resolved on the transcriptome level.⁹

Temporally regulated phosphorylated proteins related to axonal and synaptic processes are also impacted in PACS1^(+/R203W) brain organoids

Post-translational phosphorylation is a main regulator of protein function. Thus, we analyzed whether the PACS1^(+/R203W) organoids also presented with alterations in phosphorylation during maturation. Quantification of phospho-peptides in organoids resolved a total of 4,238 high-confidence phosphorylation sites mapped to 1,754 unique proteins (Tables S6 and S11). Of these mapped phosphorylated peptides, 1,112 (63.4%) had a corresponding protein that allowed to correct phosphorylation occupancy by protein abundance (Figure S5). The time series and clustering analysis considering phosphorylation abundance uncovered the same eight clusters and stable patterns that resulted from the protein abundance analysis (Figure 3A and Tables S12 and S13). However, an additional distinct cluster composed of phosphorylated peptides between days 21 and 40 (“early-up” cluster) was observed in PACS1^(+/R203W) organoids (Figure 3B).

Figure 3.

Dysregulation of temporal phosphorylation dynamics in PACS1 p.R203W organoids

(A) Alluvial plot indicating the change of profile clustering in the PACS1 p.R203W organoids in contrast to controls. Breaks in the y axis were introduced to improve visualization of the dysregulated clusters by removing 3,000 modifications that remained stable over time.

(B) Five clusters of dysregulated phosphorylation trends were observed in PACS1 p.R203W organoids that either gained a temporal regulation or lost the original pattern observed in the controls; black lines in all figures represent mean trend of all proteins in each cluster. The plots are organized into a gain or loss of regulation categories: clusters stable-up and -down gained temporal regulation in contrast to the controls (orange), while clusters up- and down-stable lost temporal regulation (gray). Early-up is a new cluster exclusive to the phosphorylation dataset that also gained temporal regulation (green).

(C) Enriched biological functions per dysregulated cluster. The nested pie charts show root and branch significant GO enrichments and the width of the pie correlating to -log(p) of the GO term. Neurodevelopment-related processes (bold).

A total of 530 phosphorylation sites on 325 unique proteins were dysregulated in PACS1^(+/R203W) organoids (Table S14). Examples include two PACS1-known interactors, PACS2 (PACS1 paralog) and HDAC1, both significant at abundance and occupancy levels. PACS1^(+/R203W) organoids presented with increased PACS2-S437 phosphorylation, which is required for PACS2-mediated trafficking and repression of apoptosis.⁵² HDAC1 had decreased phosphorylation at residues S421 and S423, which is essential for its repressive function.⁵³ Interestingly, both serine residues are also phosphorylated by CK2 (found to have lost its temporal regulation and remained stable in PACS1^(+/R203W) organoids). Due to lack of specific antibodies for the aforementioned phosphosites, future validation experiments are required to confirm these phosphorylation events and their potential involvement in PACS1 syndrome pathogenesis.

Given that phosphorylation may result in either an induction or inhibition of protein function, and the activation status of a protein itself could have diverse effects on pathway activity, we used the phosphosite database to understand the functional impact of the observed dysregulated phosphorylation sites.⁵⁴ Phosphosite contained annotations on 51 out of 530 phosphorylation sites across 41 unique proteins, with few proteins having multiple phosphorylated sites (Figure S6 and Table S15). To perform the functional enrichment analysis, we grouped phosphorylated proteins that changed in abundance (clusters: stable -up and -down) or lost (clusters up- and down-stable) phosphorylation regulation in contrast to the controls. Although the early-up cluster also gained upregulation, due to its uniqueness to this dataset, we performed the functional enrichment analysis separately (Figure 3C and Table S16).

The phosphorylated proteins that gained up/downregulation in the PACS1^(+/R203W) organoids over time while remaining stable in controls are involved in axonogenesis, as well as synaptic function and organization processes (Figure 3C, orange pie plot). Examples of these proteins include synapsin1 (SYN1, significant at abundance and occupancy), implicated in both synaptogenesis and neurotransmitter release, whose phosphorylation at residues S568/S570 causes conformational changes that prevents its binding to actin or synaptic vesicles⁵⁵; synaptojanin1 (SYNJ1, significant at abundance and occupancy), which has a prominent role in synaptic vesicle recycling⁵⁶; the beta subunit of the AP-3 protein complex (AP3B2, significant at abundance level), a known PACS1 interactor that helps form clathrin-coated synaptic vesicles⁵⁶; and DNM1-S774 (significant at abundance level), which regulates clathrin-mediated endocytosis through its activation by dephosphorylation.⁵⁷ Further validation experiments are required to confirm the phosphorylation status of these proteins in PACS1^(+/R203W) brain organoids. These experiments will be prioritized upon the availability of the required phospho-specific antibodies.

Similarly, as observed in the other gain of regulation clusters, the early-up cluster had enrichment for axon-related function and organization, but showed novel terms related to signaling through semaphorin 3 (SEMA3E) (Figure 3C, green pie plot). This result is in consensus with single-cell RNA sequencing analysis, which identified SEMA3E to be upregulated in PACS1^(+/R203W) brain organoids.⁹ Our proteomic dataset shows that while there are no changes in SEMA3E, several proteins involved in its signaling pathway are phosphorylated, including CRMP1, DPYSL2, and DPYSL3 (all significant at abundance level). In particular, DPYSL2-S522 phosphorylation by CDK5 is involved in axonal guidance and spine development.⁵⁸ Future studies aimed at investigating SEMA3E signaling in PACS1^(+/R203W) models will confirm these observations.

Meanwhile, intracellular protein transport was the most enriched process in the clusters that lost regulation (Figure 3C, brown pie plot). Proteins that remained stable in PACS1^(+/R203W) organoids include GLI3 (significant at abundance and occupancy), which has a dual function as a transcriptional activator and a repressor of the sonic hedgehog (Shh) pathway dependent on phosphorylation⁵⁹ and LIN28A (significant at abundance level), an RNA-binding protein that is localized to the peri-ER and has been recently described as a global suppressor of genes in the secretory pathway.⁶⁰

Dysregulated protein abundance and phosphorylation status in PACS1^(+/R203W) organoids result in convergent processes and overlap with known NDD-related genes

We next analyzed the overlap of protein identities with dysregulated abundance and phosphorylation status in PACS1^(+/R203W) organoids, and with previously reported transcriptomic data.⁹ We re-analyzed the transcriptomic profiles of p.R203W organoids using a pseudobulk approach to estimate gene expression. This yielded a total of 454 differentially expressed genes between day 40 and day 90 (Figure S8 and Table S17). Surprisingly, we observed a low overlap of the total and dysregulated proteins, phosphopeptides, and the transcriptome, with just JAM2, a cell-adhesion protein from the immunoglobulin subfamily and key regulator of axon myelination⁶¹ (Figure 4A). Nevertheless, several affected processes converged (Figure S8). For example, neurogenesis, synapse organization, and axon-related processes are enriched due to dysregulated RNA expression, protein abundance, and phosphorylation status over time in PACS1^(+/R203W) organoids when compared to controls (Figure S8). Converging processes were observed also only at the protein and phosphorylation level including mRNA transport, actin and microtubule organization, and cell cycle dynamics. Finally, some processes were exclusive, including oxidative phosphorylation and energy production that resulted only from the dysregulation of protein abundance (purple); and RNA splicing showing a gain of upregulation in phosphorylation status over time (yellow) (Figure S7). These results suggest that PACS1^(+/R203W) organoid development is disrupted at different regulation levels.

Figure 4.

Protein abundance and phosphorylation sites dysregulated by PACS1 p.R203W are enriched in neurological disease-related genes

(A) Overlapping identities between all dysregulated RNAs, proteins, and identified phosphorylation sites observed in the PACS1 p.R203W organoids when compared to controls.

(B) Selected top neuro-related categories of gene-disease association type ontologies enriched in the PACS1 p.R203W dysregulated transcriptome, proteome and phosphoproteome, at a corrected p value of 0.01.

(C) NDD and ASD risk genes dysregulated in organoids carrying the p.R203W pathogenic variant. Genes are grouped by RNA, proteins, and phosphorylations. The RNA and protein abundance groups are colorized by dysregulation cluster while the phosphorylation clusters are colorized by major dysregulation trend (gain or loss) as presented in Figure 3: gain of regulation (clusters stable-up and -down) or loss of regulation (clusters up- and down-stable).

(D) Volcano plots representing the dysregulated RNAs (top), proteins (middle), and phosphorylations (bottom) at days 40 and 90. The circles in green tones represent ASD genes while the triangles in pink tones represent NDD genes.

Next, we investigated whether there are convergent dysregulated genes in PACS1 syndrome compared to other NDDs. We performed gene set enrichment analysis on the dysregulated proteins and transcripts using the DISGENET, the Simons Foundation ASD (SFARI), and the Online Mendelian Inheritance in Man (OMIM) gene databases. A total of 328 genes (20.7%) dysregulated in the PACS1^(+/R203W) organoids compared to controls are additionally reported in these databases. In DISGENET, we found 249 genes that showed a significant enrichment in multiple nervous system disorders (Figure 4B and Table S18). These disorders overlapped with phenotypes observed in PACS1 syndrome patients like ASD, ID, seizures, and abnormal morphogenesis of the nervous system or facial features,^16,62 as well as phenotypes yet to be reported, such as neurodegeneration. Interestingly, the core neurodevelopmental features of PACS1 syndrome like ID, delayed speech, and epilepsy¹⁵ were enriched in all levels of regulation, while the facial and muscular abnormalities were enriched in the proteome and phosphoproteome (i.e., wide nasal bridge, orbital separation, limb hypertonia, and unsteady gait).

We also identified multiple genes dysregulated in PACS1^(+/R203W) organoids that are reported as causing NDD in OMIM or as ASD-risk genes in the SFARI database (Figures 4C and 4D and Table S19). Fifty-five NDD and 57 ASD genes showed dysregulation in PACS1^(+/R203W) organoids over maturation (Figure 4C), mostly at the phosphoproteomic level. We also identified additional genes that were dysregulated at specific developmental timepoints across regulation levels (194 NDD and 98 ASD genes) (Figure 4D). These results highlight the complementarity of gene regulation mechanisms during development, supporting the benefits of including proteomic analysis to expand our understanding of NDD pathogenesis.

Discussion

NDDs are highly heterogenous and can arise from variants across hundreds of genes.^63,64 Several limitations have hampered our understanding of the mechanisms by which these variants act, limiting translation of genetic findings into the clinic.⁶⁵ Here, we leveraged the potential of proteomics in brain organoids to gain insights into the pathogenic mechanisms of PACS1 syndrome. Proteomic approaches are usually bulk and therefore lack single-cell resolution. Thus, we reanalyzed our transcriptomic data,⁹ using a whole-organoid pseudobulk approach to better compare proteomic and transcriptomic profiles. The recaptured gene regulation signals showed high concordance of affected biological processes between both approaches. Despite the low overlap of dysregulated genes identified at the mRNA, protein, and phosphorylation levels in PACS1^(+/R203W) organoids, most dysregulated processes converged. However, PACS1^(+/R203W) organoids also presented with dysregulated abundance and phosphorylation dynamics of temporally regulated proteins that were not identified by transcriptomics, highlighting the importance of complementing transcriptomic studies with proteomic analysis. Experimental approaches that target both proteomics and transcriptomics from the sample samples could help understand discrepancies in identified dysregulated gene identities.

At the transcriptomic level, we identified many neurogenesis and synaptic function-associated processes affected primarily in maturing neurons.⁹ The proteomic data not only reinforced the importance of these processes for p.R203W pathogenesis, but also detected dysregulated processes linked to known PACS1 functions. For example, PACS1’s role in protein sorting explains the overrepresentation of vesicle-related processes. Moreover, several established PACS1 interactors were uncovered as dysregulated proteins, such as its paralog, PACS2. PACS1 and PACS2 evolutionarily emerged as a duplication in higher metazoans from an ancient PACS gene.¹⁹ PACS1 function primarily involves sorting from endosomes to the tGN, which showed temporal dysregulation of over 20 proteins in PACS1^(+/R203W) organoids. Meanwhile, PACS2 evolved cargo-sorting functions from the Golgi to ER and vesicular coat preference for COPII; both processes upregulated in PACS1^(+/R203W) organoids. The upregulation in these processes is consistent with the potential gain of PACS2 regulation through increased phosphorylation at S437, which favors PACS2 trafficking function.⁵² The increase in phosphorylated PACS2 could explain the vesicular transport and mitochondrial protein import defects suggested by the dysregulated proteomic signature in PACS1^(+/R203W) organoids. Conversely, it may also explain the downregulation of apoptotic signaling pathway genes. Experiments designed to functionally measure these processes are required to confirm these observations. Of interest, a recurrent de novo missense mutation in the regulatory middle region of PACS2 is associated with PACS2 Syndrome, another NDD.⁶⁶ Future research aimed at understanding the interconnection of PACS1 and PACS2 function, as well as their dysregulation and compensatory effects in disease, will help shed light on the pathogenesis of both PACS1 and PACS2 syndromes.

The processes that lost downregulation over time were related to recently described PACS1 functions, including proteins related to the maintenance of genome stability and DNA repair via interaction with HDAC2, HDAC3, and HDAC6 (20, 22). Although we did not observe protein abundance and phosphorylation status changes in HDAC2, 3, or 6, we found downregulation in HDAC5 at day 40, and HDAC1 had decreased phosphorylation over maturation. This suggests that PACS1 could also modulate DNA stability through these proteins. HDAC1 interacts with RB1 and SMARCA4 (both dysregulated at the phosphorylation level) in a repressive complex in resting neurons. This complex is abolished in a calcium-dependent manner to enable transcription,⁶⁷ suggesting that the PACS1 p.R203W variant affects transcription regulation by destabilizing the repressive complex formed by HDAC1, RB1, and SMARCA4. Transcription dysregulation has been reported previously by our group and others.^9,32,34 However, it remains unclear whether the HDAC1, RB1, and SMARCA4 complex plays a role in these processes.

Processes unrelated to known PACS1 functions also emerged, such as the enrichment associated with regulating nucleotide metabolism and protein translation in the down-stable cluster. There is growing evidence of the impact of protein translation on the pathogenesis of NDDs, including ASD and fragile X syndrome.⁶⁸ Functional studies aimed at confirming these defects in protein translation would be a logical next step. This is the first time that either process has been implicated in the PACS1 syndrome pathogenesis, as these processes were not identified by transcriptomic analysis.⁹ Consequently, this result highlights the importance of collecting data from multiple regulatory levels to get a more complete picture of the pathogenic mechanisms when modeling NDDs.

Processes that seem unrelated to PACS1 function, but which have displayed upregulation in the PACS1^(+/R203W) organoids, should be further explored. Considering current evidence suggests that the PACS1 p.R203W pathogenic variant may result in a gain of function,^9,34,69 there is increased interest in generating targeted therapeutic strategies aimed at suppressing its toxic effects.^34,36 Thus, processes that seem unrelated to PACS1 function but with observed upregulation in the PACS1^(+/R203W) organoids should be a point of interest. Understanding whether the specific proteomic dysregulation results from the toxicity associated with the variant or as a compensatory mechanism to maintain cellular homeostasis would be especially relevant. Of further interest, most genes that we found dysregulated in PACS1^(+/R203W) organoids have also been linked to NDDs and ASD.^59,60,62 However, the overrepresentation of NDD- and ASD-related genes could also reflect the interconnectedness of neurodevelopmental pathways. Additional mechanistic analyses are required to understand if these effects are direct or indirect.

Finally, the presence of the PACS1 p.R203W variant seems to affect every regulatory level as we see convergence in processes related to the clinical features that patients manifest. This includes ASD, ID, and seizures,^15,16 as well as phenotypes yet to be reported, such as neurodegeneration, all reflected in the dysregulation of mRNA and protein abundance as well as phosphorylation status. Interestingly, the facial and muscular abnormalities including wide nasal bridge, orbital separation, limb hypertonia, and unsteady gait, resulted mainly from enrichments in the proteome and phosphoproteome. This study provides a foundation for hypothesis generation and for designing experimental approaches aimed at validating the proposed pathogenic mechanisms. Our results highlight the potential of proteomic analyses to complement transcriptomic studies to enhance our understanding of NDD mechanisms and uncover additional therapeutic targets that may facilitate therapy development.

Limitations of the study

While we captured a broad range of proteins and phosphorylated peptides that recapitulate aspects of early brain development in brain organoids, this dataset does not fully represent the complexity or temporal dynamics of neurodevelopment. Proteomic data were generated from bulk tissue and therefore lack the single-cell resolution of transcriptomic approaches, limiting direct comparisons between these technologies. Likewise, the phosphopeptide findings require further validation to determine their mechanistic roles in PACS1 p.R203W pathogenesis. Pathway and biological function enrichment analyses are also limited by incomplete gene and pathway annotations. Although useful for interpreting large datasets, these analyses are descriptive and correlative and do not establish causal relationships. Similarly, pseudotime analysis provides a computational ordering of cellular states based on molecular profiles rather than a direct measure of developmental time and, therefore, should be interpreted with caution. Lastly, although brain organoids recapitulate key features of human brain development, they lack cell types that emerge later in development or originate from other germ layers, including astrocytes and mesoderm-derived microglia. These cells may also contribute to PACS1 syndrome pathogenesis. Therefore, the mechanisms identified in this study should be viewed as a foundation for future investigation rather than a comprehensive account of disease pathogenesis.

Resource availability

Lead contact

Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Alicia Guemez-Gamboa (alicia.guemez@northwestern.edu).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• •Raw TMT mass spectrometry data have been deposited in the Proteomics Identifications Database (PRIDE) public repository under project accession PRIDE: PXD080130, and project https://doi.org/10.6019/PXD080130.
• •This study does not report original code.
• •Any additional information required to reanalyze the data reported in this study is available from the lead contact upon request.

Acknowledgments

This research was supported in part through the computational resources and staff contributions provided for the Quest high-performance computing facility at Northwestern University which is jointly supported by the Office of the Provost, the Office for Research, and Northwestern University Information Technology.

This work was funded by the PACS1 Syndrome Research Foundation (A.G.-G.) and by National Institutes of Health NINDS R01NS123163 (A.G.-G.) as well as NIMH R01MH125235 (M.L.M and B.D.), R01MH118497 (M.L.M.). Confocal microscopy was done at the Northwestern University Center for Advanced Microscopy funded by NCI CCSG P30 CA060553.

Author contributions

This work was conceptualized by A.G.-G. L.E.R. cultured the organoids, collected samples, and prepared them for proteomic analysis. S.R., B.D., J.G., and M.L.M. performed the TMT mass spectrometry processing of the samples. X.G.-M. performed all bioinformatic steps for transcriptomics and proteomics analysis with help from A.W. to process figures, and help from S.R., B.D., and L.K. to refine the proteomic experimental design. A.W., A.L.S., J.R., C.L.B., and T.A.S. performed wet lab experiments (immunohistochemistry, western blotting, and imaging) and data analysis derived from those experiments. Writing was done by X.G.-M. and edited by A.G.-G., M.L.M., L.E.R., J.R., A.L.S., T.A.S., and C.L.B. All authors read and approved the final manuscript.

Declaration of interests

The authors declare no competing interests.

Declaration of generative AI and AI-assisted technologies in the writing process

The authors did not use AI or AI-assisted technologies in the writing process.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-Ctip2 monoclonal antibody (25B6) (recommended dilution 1:500)	Abcam	Cat #: ab18465; RRID: AB_2064130
Anti-SOX2 polyclonal antibody (recommended dilution 1:200)	MilliporeSigma	Cat #: AB5603; RRID: AB_2286686
Anti-Nestin Antibody, clone 10C2 (recommended dilution 1:1000)	MilliporeSigma	Cat #: MAB5326; RRID:AB_2251134
Anti-FOXG1 antibody [EPR18987] (recommended dilution 1:100)	Abcam	Cat #: ab196868; RRID: AB_2892604
IRDye® 800CW Goat anti-Rabbit IgG Secondary Antibody	LICORbio	Cat #: 926–32211; RRID AB_621843
IRDye® 680RD Goat anti-Mouse IgG Secondary Antibody	LICORbio	Cat #: 926–68070; RRID AB_10956588
IRDye® 800CW Goat anti-Mouse IgG Secondary Antibody	LICORbio	Cat #: 926–32210; RRID AB_621842
Donkey anti-mouse IgG(H + L) highly cross-adsorbed secondary antibody, Alexa Fluor 594 (recommended dilution 1:200)	Thermo Fisher Scientific	Cat#:A-21203; RRID:AB_141633
Donkey anti-Mouse IgG (H + L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor™ Plus 488	Thermo Fisher Scientific	Cat #: A32766TR; RRID: AB_2866493
Chemicals, peptides, and recombinant proteins
Matrigel hESC-qualified matrix, LDEV-free	Corning	Cat #: 354277
DMEM/F-12 (Dulbecco’s Modified Eagle Medium/Nutrient Mixture F-12)	Gibco	Cat #: 11320033
Y-276322HCl	Selleck Chemicals	Cat #: S1049
Accutase	MilliporeSigma	Cat #: SCR005
Paraformaldehyde	MilliporeSigma	Cat #: 158127
Normal donkey serum	SouthernBiotech	Cat #: 0030–01
Bovine serum albumin	MilliporeSigma	Cat#: A2153
DAPI	Abcam	Cat#: ab285390
Fluoromount-G™ Mounting Medium	Thermo Fisher Scientific	Cat#: 00–4958–02
PVDF Transfer membrane	Thermo Fisher Scientific	Cat#: 88518
Intercept blocking buffer	LICORbio	Cat#: 927–60001
Halt Protease & Phosphatase Inhibitor Cocktail (100×)	Thermo Fisher Scientific	Cat#: 1861284
RIPA Lysis Buffer (10×)	EMD Millipore Corp	Cat#: 20–188
4× Laemmli Sample Buffer	Bio-Rad	Cat#: 1610747
Pierce BCA Protein Assay Kit	Thermo Fisher Scientific	Cat#: 23225
Critical commercial assays
STEMdiff™ Dorsal Forebrain Organoid Differentiation Kit	STEM CELL Technologies	Cat#:08620
STEMdiff™ Trilineage Differentiation Kit	STEM CELL Technologies	Cat#: 05230
STEMdiff™ Neural Organoid Maintenance Kit	STEM CELL Technologies	Cat#: 100–0120
mTeSR™ medium	STEM CELL Technologies	Cat#: 05825
AggreWell™800 24-well Plate	STEM CELL Technologies	Cat#: 34811
STEMdiff Cerebral Organoid Maturation Kit	STEM CELL Technologies	Cat#:08571
Costar ultra-low attachment 6-well plate	Corning	Cat#:3471
Axygen 200 mL wide bore pipette tips	Corning	Cat#:TF-205-WB-R-S
37 μm Reversible Strainer, Large (50 mL)	STEM CELL Technologies	Cat#: 27250
Micro BCA™ Protein Assay Kit	Thermo Fisher Scientific	Cat #:23235
TMTpro™ 16plex Label Reagent Set	Thermo Fisher Scientific	Cat #:A44520
Pierce™ High pH Reversed-Phase Peptide Fractionation Kit	Thermo Fisher Scientific	Cat #: 84868
HALO 120 Å ELV C18	HALO	Cat #: 92272–402
AssayMAP 5 μL Fe(III)-NTA cartridges	Agilent	Cat #: G5496-60085
Acclaim PepMap RSLC C18, 2 μm, 100 Å, 75 μm i.d. × 50 cm, nanoViper	Thermo Fisher Scientific	Cat #: 164540
S-TrapTM micro columns	Protifi	Cat #: C02-MICRO-10
Deposited data
Raw TMT mass spectrometry data	This paper	PRIDE: PXD080130
Single cell sequencing data	Rylaarsdam et al.9	NCBI Gene Expression Omnibus (GEO): GSE250386
Human fetal brain proteomics	Djuric et al.38	ProteomeXchange: PXD004076
Experimental models: Cell lines
Control cell line 1 (+/+), female (WT1)	Coriell	GM03651
CRISPR-corrected cell line 2 (+/+), female (WT2)	Rylaarsdam et al.9	PACS1002i-GM27159
CRISPR-corrected cell line 3 (WT3), male (WT3)	Rylaarsdam et al.9	PACS1003i-GM27161
Control cell line 4 (WT4), male (WT4)	Coriell	GM03652
Control cell line 5 (WT5), female (WT5)	Wicell	GM27160
CRISPR p.R203W insertion (+/R203W), female (P1)	Rylaarsdam et al.9	GM03651(+/R203W)
Affected individual (+/R203W), female (P2)	Wicell	PACS1002i-GM27159
Affected individual (+/R203W), male (P3)	Wicell	PACS1003i-GM27161
Software and algorithms
Proteome Discoverer 2.4	Thermo Fisher Scientific	https://www.thermofisher.com/
R version 4.1.1	R Core Team70	https://www.r-project.org
STEM v1.3.13	Ernst and Bar-Joseph43	https://doi.org/10.1186/1471-2105-7-191
limma v3.58.0	Ritchie et al.71	https://doi.org/10.1093/nar/gkv007
Seurat v5.3	Hao et al.72	https://doi.org/10.1038/s41587-023-01767-y
GeneOverlap v.1.40.0	Shen et al.73	http://shenlab-sinai.github.io/shenlab-sinai/
UpSetR (v.1.4.0)	Gehlenborg74	https://doi.org/10.1093/bioinformatics/btx364
DESeq2 v1.46	Love et al.75	https://doi.org/10.1186/s13059-014-0550-8
Metascape v3.5.20240901	Zhou et al.76	https://doi.org/10.1038/s41467-019-09234-6
Cytoscape v3.10.2	Shannon et al.77	https://doi.org/10.1101/gr.1239303
webr v.0.1.5	Moon et al.78	https://github.com/cardiomoon/webr
ggplot2 v3.5.1	Wickham79	https://github.com/tidyverse/ggplot2
Monocle3	Cao et al.80	https://doi.org/10.1038/s41586-019-0969-x

Experimental model and study participant details

This study was performed in compliance with relevant ethical regulations and approved by the Northwestern University Institutional Review Board (IRB STU00215054). Human-derived induced pluripotent stem cell (iPSC) lines are deidentified, and no information about the donors is linked and/or available. All donors consented to tissue being used for research purposes. Female and male cell lines were used but sex was not evaluated as an independent variable. PACS1 syndrome affects both sexes and there is no evidence that disease severity is influenced by sex. Two control lines derived from a 25-year-old female (WT1, GM03651) and a 24-year-old male (WT4, GM03652) and the 39-year-old unaffected mother (WT5, GM27160) of a 3-year-old daughter with PACS1 syndrome (P2, PACS1002i-GM27159) and an additional 6-year-old male patient (P3, PACS1003i-GM27161) were used. Three isogenic pairs were used in total with the p.R203W variant introduced in the WT1 background (P1) and corrected (WT2 and WT3) in both patient lines (P2 and P3) as reported in Rylaarsdam et al.⁹ The karyotyping of all lines was determined normal, and short tandem repeat (STR) profiling was used for authentication.⁹ All iPSC lines expressed markers of pluripotency and demonstrated the capacity to differentiate the three germ layers using the STEMdiffTM Trilineage Differentiation Kit (Stem Cell Technologies, 05230).⁹ All cell lines were regularly tested for mycoplasma contamination and genotyped with Sanger sequencing. All cell lines were differentiated in 3 independent trials with at least 3 replicates each, from which individual organoids were randomly collected at maturation day 40 and 90. For samples collected at day 21, multiple organoids from the same differentiation were pooled into one sample due to their smaller size. The total number of samples collected per cell line are as follows: WT1, 12 (day 21: 5, day 40: 4, day 90: 3); P1, 10 (day 21: 4, day 40: 4, day 90: 2); WT2, 8 (day 21: 3, day 40: 3, day 90: 1); P2, 7 (day 21: 3, day 40: 3, day 90: 1); WT3, 7 (day 21: 3, day 40: 2, day 90: 2); P3, 10 (day 21: 4, day 40: 3, day 90: 3); WT4, 9 (day 21: 4, day 40: 3, day 90: 1); WT5, 12 (day 21: 2, day 40: 1, day 90: 1). See Table S20 for additional sample information.

Method details

Generation of forebrain organoids from iPSCs

Three isogenic pairs of PACS1^(+/+) and PACS1^(+/R203W) iPSC lines were previously generated⁹ and differentiated to brain organoids using the kit based on the protocol outlines by Sloan et al.⁸¹ (Stem Cell Technologies, 08620). iPSCs were maintained in mTeSR+ medium (Stem Cell Technologies, 05825) on plates pre-coated with Matrigel (Corning, 354277) dissolved in DMEM/F12 at a concentration of 10 μg/mL. Stem cells were passaged as needed using ReLeSR medium (Stem Cell Technologies, 05872). AggreWell800 24-well plate (Stem Cell Technologies, 34811) were prepared and 3 million cells from each line were seeded per well. STEMdiff Forebrain Organoid Formation Medium was exchanged for 5 days and on day 6, embryoid bodies from each Aggrewell were filtered through a 37 μM strainer (Stem Cell Technologies, 27250) and transferred to two wells of a 6-Well Ultra-Low Adherent Plate (Corning, 3471) containing STEMdiff Forebrain Organoid Expansion Medium. STEMdiff Forebrain Organoid Expansion Medium was changed every other day until day 26 when it was replaced with STEMdiff Forebrain Organoid Differentiation Medium. Organoids were cultured in differentiation media until day 43. Following day 43, organoids were cultured in STEMdiff Forebrain Organoid Maintenance Medium until collection.

Organoid cryosectioning and staining

Organoids were fixed with 4% paraformaldehyde at 4 °C overnight and transferred to 30% sucrose for at least 16h at 4 °C. Organoids were then embedded in OCT, snap frozen in dry ice and ethanol, and stored at −80 °C. Frozen organoids were cryosectioned 15–20 μm thick, dried for at least 1 h and washed with PBS for OCT removal. Antigen retrieval was performed by steaming slices at 99 °C for 10 min in a 10 mM citrate solution at pH 6.0. Slides were washed and blocked in a 10% NDS, 0.3% Triton X-100 PBS solution and then incubated with primary antibodies at 4 °C overnight. Primary antibodies used in this study include: CTIP2 (Abcam, ab18465, Dilution 1:200); SOX2 (EMD Millipore, Dilution 1:200). The overnight incubation was followed by PBS washes and secondary antibody incubation at 1:200 in a 5% NDS, 0.15% Triton X-100 solution for 90 min at room temperature. Secondary antibodies used in this study include donkey anti-mouse Alexa Fluor 488 (Invitrogen) and donkey anti-rat Alexa Fluor 594 (Invitrogen, A21209, Lot 2078918). DAPI (2 μg/mL) was incubated for 20 min. Slides were mounted with Fluoromount-G (Thermo Scientific, 00–4958–02) and imaged on a Nikon W1 spinning disk confocal microscope or BZX710 Keyence epifluorescent microscope.

Sample preparation for proteomic analysis

Total protein was extracted from the organoids by sonication in lysis buffer (1% sodium dodecyl sulfate (SDS), 125 mM triethylammonium bicarbonate (TEAB), 75 mM NaCl, and HaltTM protease and phosphatase inhibitors). Lysate samples from two additional control cell lines were included to improve protein detection. See Table S20 for a detailed description of mass spectrometry samples.

Mass spectrometry sample preparation

Protein extraction and trypsin digestion. Total protein was extracted from the organoids by sonication in lysis buffer (1% sodium dodecyl sulfate (SDS), 125 mM triethylammonium bicarbonate (TEAB), 75 mM NaCl, and HaltTM protease and phosphatase inhibitors). Lysates were cleared by centrifugation at 12,000 g for 15 min at 4 °C. Lysate samples from two additional control cell lines were included to improve protein detection. A different randomized block was used for each stage of sample processing, TMT labeling, and mass spectrometry analysis, balanced for genotype and timepoint. Total protein was estimated using microBCA assay according to the manufacturer instructions (Cat. No. 23235, Thermo Fisher). 20 μg total protein from each sample was reduced with dithiothreitol (DTT), alkylated with iodoacetamide (IAA), and digested with trypsin on S-Traps (Protifi) per manufacture protocol.

TMT labeling. Samples were first assigned to one of five sets of TMTPro16 blocks (plex’s), such that each plex was balanced for genotype and timepoint, with a random order in each plex (Tables S21 and S22). Organoid samples from two additional, age-matched control cell lines with no isogenic pair (WT4 and WT5) were included within each plex to improve peptide diversity capture. However, these samples were later removed from downstream analyses since they lacked an isogenic p.R203W. Pooled reference control for internal reference scaling (IRS) normalization was included within each plex. These reference controls were prepared by pooling equal amounts of lysate from all samples, which were digested separately, and then split in aliquots for TMT labeling. Peptide digests were reconstituted in 20 μL 100 mM TEAB and sonicated for 15 min at RT for TMT labeling. TMTPro reagents were brought to room temperature and dissolved in 20 μL Optima LC/MS-grade acetonitrile (ACN). 6 μL of ∼59 mM TMTPro reagents was added to 20 μL reconstituted peptide accordingly. All TMT plex’s were labeled at the same time. For the reference control, peptides were reconstituted in 80 μL. A total of 40 μL was labeled with 12 μL TMTPro 134 N. Labeling was performed at RT for 90 min with shaking at 400 rpm. The reaction was quenched by adding 3 μL 5% hydroxylamine/100 mM TEAB at RT for and shaken 15 min (6 μL for reference channel). All samples with in a plex were mixed along with 20 μg aliquots of each labeled pooled control, giving 320 μg total TMT labeled peptides per plex, and dried in a vacuum concentrator. Of this, 80 μg were used for peptide/protein quantification, and 240 μg was subject to phosphopeptide enrichment. Prior to peptide/protein quantification, the 80 μg aliquot was separated into 8 fractions using the Pierce High pH Reversed Phase Fractionation Kit per manufacturer protocol. The resulting fractions were dried down in a speed vac, and resuspended in 3% ACN, 0.1% formic acid.

Phosphopeptides enrichment. For phosphopeptide enrichment, the 240 μg TMT-labeled peptide aliquot were desalted with C18 spin column. Samples were resuspended and sonicated for 10 min in 300 μL buffer A (LC MS water 0.1% TFA) with 2% ACN. Resin was packed by centrifugation at 5,000 g for 2 min, activated twice with 300 μL ACN, and conditioned twice with 300 μL buffer A. Peptides were loaded once by centrifugation at 3,000 g for 1 min. Column was washed three times with 300 μL buffer A, and twice with 300 μL 5% methanol buffer A. Peptides were eluted twice with 300 μL 50% ACN 0.1% TFA and dried in a vacuum concentrator. Phosphopeptides were enriched using the AssayMAP Agilent Bravo instrument with Fe(III)-NTA IMAC resin according to the manufacturer’s instructions (Agilent, Cat. No. G5496-60085). Peptides were resuspended in 210 μL 80% ACN 0.1% TFA, bath sonicated for 10 min, vortexed for 10 min, and quickly centrifuged. The resin was primed with 50% ACN 0.1% TFA and equilibrated with equilibration/wash buffer 80% ACN 0.1% TFA. Samples were loaded into the syringes then dispensed into the Flow Through Collection plate. Syringes were washed with the Cartridge Wash Buffer discarded into waste, then repeated with Syringe Wash Buffer. The elution buffer was aspirated into the syringes and dispensed into the Eluate Collection plate. Phosphopeptides were eluted in 18 μL 1% ammonium hydroxide solution and acidified with 1.8 μL 20% FA, and subject to high pH reverse phase fractionation as described above but condensed to 5 fractions.

MS data acquisition and raw data processing

For peptide/protein quantification, TMT labeled peptides (∼1 μg) were loaded onto a heated PepMap RSLC C18 2 μm, 100 Å, 75 μm × 50 cm column (Thermo Scientific) and eluted over 180 min gradients optimized for each high pH reverse-phase fraction.⁸² Sample eluates were electrosprayed (2000 V) into a Thermo Scientific Orbitrap Eclipse mass spectrometer for analysis. MS1 spectra were acquired at a resolving power of 120,000. MS2 spectra were acquired in the Ion Trap with CID (35%) in centroid mode. Real-time search (max search time = 34 s; max missed cleavages = 1; Xcorr = 1; dCn = 0.1; ppm = 5) were used to select ions for synchronous precursor selection for MS3. MS3 spectra were acquired in the Orbitrap with HCD (60%) with an isolation window = 0.7 m/z and a resolving power of 60,000, and a max injection time of 400 ms. For phosphopeptide quantification, 4 μL (out of 20) of the TMT labeled phosphopeptide enrichments were loaded onto a heated PepMap RSLC C18 2 μm, 100 Å, 75 μm × 50 cm column (Thermo Scientific) and eluted over a 180 min gradient: 1 min 2% B, 5 min 5% B, 160 min 25% B, 180 min 35% B. Sample eluate was electrosprayed (2000 V) into a Thermo Scientific Orbitrap Eclipse mass spectrometer for analysis. MS1 spectra were acquired at a resolving power of 120,000. MS2 spectra were acquired in the Orbitrap with HCD (38%) in centroid mode with an isolation window = 0.4 m/z, a resolving power of 60,000, and a max injection time of 350 ms.

The raw MS files were processed using Proteome Discoverer version 2.4 (Thermo Scientific, Waltham, MA). MS spectra were searched against the Homo sapiens Uniprot/SwissProt database. SEQUEST search engine was used (enzyme = trypsin, max. missed cleavage = 4, min. peptide length = 6, precursor tolerance = 10 ppm). Static modifications include carbamidomethyl (C, +57.021 Da) and TMT labeling (N-term and K, +304.207 Da for TMTpro16). Dynamic modifications include oxidation (M, +15.995 Da), Phosphorylation (S, T, Y, +79.966 Da, only for phosphopeptide dataset), acetylation (N-term, +42.011 Da), Met-loss (N-term, −131.040 Da), and Met-loss + Acetyl (N-term, −89.030 Da). PSMs were filtered by the Percolator node (max Delta Cn = 0.05, target FDR (strict) = 0.01, and target FDR (relaxed) = 0.05). Proteins were identified with a minimum of 1 unique peptide and protein-level combined q values <0.05. Reporter ion quantification was based on corrected S/N values with the following settings: integration tolerance = 20 ppm, method = most confident centroid, co-isolation threshold = 70, and SPS mass matches = 65.

Western blotting

Protein was extracted from organoids using a mixture of a 1× RIPA buffer (EMD Millipore Corp, #20–188) and protease and phosphatase inhibitor cocktail (Thermo scientific, #1861284) in a 1:100 ratio. Organoids were lysed in this mixture using an insulin syringe and then incubated on ice for 5 min. Afterward, samples were centrifuged for 10 min at 10,000 × g at 4°C, and the resulting protein lysate supernatant was isolated. To quantify the protein lysate, a Pierce BCA Protein Assay Kit (Thermo scientific, #23225) was used. The samples were then mixed with prepared 4× Laemmli sample buffer (Bio-Rad, #1610747) and boiled at 95°C for 10 min. Protein lysate was loaded and ran on SDS page gels and then transferred to 0.45 μm PVDF membranes (thermo scientific, #88518). Membranes were blocked for an hour using Protein Free Blocking Buffer (LICORbio, #927–60001) and then immunoblotted in a primary antibody solution composed of primary antibody at the indicated dilution, 10 μL Tween 20 and protein free blocking buffer for a total of 5 mLs. Membranes were incubated for 1–3 nights at 4°C with gentle shaking. Primary antibodies used included Nestin Polyclonal antibody (1:1000), Anti-FOXG1 antibody [EPR18987] (1:1000), and Anti-alpha Tublin antibody [DM1A] (1:10,000). After primary antibody incubation, membranes were washed three times with TBST and then incubated at room temperature for an hour with a secondary antibody solution composed of 10 μL Tween 20, 5 μL 10% SDS buffer, and protein free blocking buffer for a total of 5 mLs. Secondary antibodies were used at 1:1000 and included IRDye® 800CW Goat anti-Rabbit IgG Secondary Antibody, IRDye® 680RD Goat anti-Mouse IgG Secondary Antibody, and IRDye® 800CW Goat anti-Mouse IgG Secondary Antibody. After incubation, membranes were washed three times with TBST and then imaged for protein signal using the Odyssey® Fc Imaging System (Li-Cor) and quantified with Image Studio (Li-Cor). Two membranes with two replicates per genotype each were used. Signal values were normalized by calculating a lane normalization factor to alpha-Tubulin within each membrane and then dividing each lane signal to its corresponding normalization factor. Normalized values in the WT were then averaged per membrane to then calculate the fold change in expression in p.R203W samples using the same reference. Fold change values were then log2 transformed to match the scale of the proteomic data.

Bioinformatic processing

Protein intensity data values were first normalized to adjust each TMT experiment to an equal signal per channel. A scaling value for each TMT plex and the internal reference scaling (IRS) factor were then calculated using the pooled reference samples. After which, a median normalization was performed followed by log2-transformation (Figures S4A and S4B). Relative occupancy of phosphorylated sites was also calculated for those peptides where the corresponding protein was also quantified by subtracting the peptide abundance from the log2 normalized values.⁸³ Both occupancy and phosphopeptide abundance were used for downstream analyses. Further analysis through PCA indicated a batch effect associated with multiplex labeling in the second component of the graph (Figures S4C and S4D), which we ameliorated in the R package limma (v3.58.0) by duplicate correlation for each observation to model the plex as a batch, and included covariates to fit the limma model as ∼ genotype + day + cell line, while using the function removeBatchEffect for visualization of the model fit.⁷¹ Only the proteins and phosphopeptides present in all plexes and with high detection/localization confidence were used for downstream analysis. This stringent filter was implemented to better control batch effects, it filtered sex-dependent proteins and addressed the impossibility of differentiating the missing values from non-expressed proteins. Despite this stringent filter, we captured a diverse pool of proteins and phosphopeptides similar to other publicly available datasets.³

Time series clustering

We used the Short Time-series Expression Miner software (STEM, v1.3.13),⁴³ which compares and assigns each expression profile from the dataset to a set of distinct and representative models of temporal expression profiles based on the total timepoints. These model profiles are independent of the data and correspond to possible profiles of a protein change in expression over time. For its implementation, the batch-corrected normalized protein abundances in the organoids were averaged by isogenic line and timepoint. Samples that corresponded to the same genotype, isogenic line, timepoint, and differentiation were averaged to generate a representative replicate of that group. Then, a single input file containing all timepoints for each genotype, isogenic, and differentiation group were generated for STEM (Figure S9 and Tables S23 and S24). We also used a fetal brain proteome previously reported by Djuric et al.³⁸ and selected the prenatal samples from all reported brain regions (ventricular zone, intermediate zone, subplate zone, and cortex). Fetal brain samples were processed using each region as a replicate for clustering. The STEM algorithm then transforms all trends to start at 0, generating a vector (0, v₁—v_0, v₂—v₀, …, v_n—v₀) for each protein. Proteins are then assigned to the model profile to which their time series most closely matches based on correlation coefficient. The number of proteins assigned to each model profile is then tested for statistical significance against the number of proteins expected to be assigned to a profile. This number is estimated by randomly permuting the original time points, scaling the protein abundances, and then assigning them to the most closely matching model profiles. The average number of proteins assigned to a model profile over all permutations was used as the estimate of the expected number of genes assigned to the profile. This process was performed using a significance level of 0.05 of the Bonferroni-corrected p-value. The values used for each parameter are presented in Table S25. Several preliminary runs were performed to select the values that both offered clustering of most proteins and diversity of profiles, while also controlling the number of highly similar profiles. Protein overlap among analyzed datasets were performed using the packages GeneOverlap (v.1.40.0)⁷³ and UpSetR (v.1.4.0).⁷⁴

Transcriptome pseudobulk and cell lineage pseudotime analysis

Previously published PACS1 syndrome single cell data⁹ was reanalyzed to quantify pseudo-bulk gene expression profiles. Data was first normalized using standard Seurat commands v5.3⁷² and the count matrix was aggregated by organoid sample using DESeq2 v1.46 (number of organoids per group: WT D40 = 2, p.R203W D40 = 3, WT D90 = 2, p.R203W D90 = 2).⁷⁵ Differential gene expression was assessed using a likelihood ratio test with an interaction design model on organoid age and genotype as the full model to test effects over maturation time, and a reduced model without the interaction term. For comparison of cell type lineages across pseudotime, all genotypes were down sampled to 10,000 randomly selected cells per organoid and used monocle3 to calculate transcriptomic landscape transitions from progenitor cells toward neuronal lineages.⁸⁰ A one-sided Kolmogorov-Smirnov test was used to test whether the p.R203W trajectories were accelerated in contrast to controls by comparing the distribution of pseudotime values (Figure S4). Cell type proportions between genotypes were previously assessed, with no significant differences.⁹

Functional processes – gene enrichment analysis

Testing for enrichment was performed with the online tool Metascape (v3.5.20240901)⁷⁶ using a hypergeometric test followed by the Benjamini-Hochberg p-value correction, selecting statistically significant terms at a cutoff value below −1.3 Log(q-value),⁷⁶ and then simplifying redundancy in the resulting lists using rrvgo in R (v1.16).⁸⁴ Enrichment tests were performed using the Gene Ontology (GO) biological processes^85,86 and the gene-disease association (DisGeNET)⁸⁷ databases. Visualizations were performed using webr (v.0.1.5)⁷⁸ and ggplot2 functions (v3.5.1).⁷⁹ The GO-term network was performed in Metascape using default settings and the human genome as a reference background. In brief, the 20 top-score clusters (lowest p-values) were selected from each dysregulated expression trend and then used the Kappa similarity statistic above 0.3 to connect similar terms.⁷⁶ The resulting network was then edited in Cytoscape (v3.10.2).⁷⁷ Additionally, we consulted the SFARI Gene database (2024 Q3 release)⁸⁸ and OMIM⁸⁹ to detect dysregulated proteins that have been annotated as ASD-risk genes and/or pathogenic variants on such genes cause NDD.

Quantification and statistical analysis

The objective of this investigation is to provide further understanding of the pathogenic mechanisms of PACS1 syndrome by investigating proteomic landscape regulation during brain organoid maturation. We used tandem mass tag (TMT) mass spectrometry in brain organoids generated from control and iPSC lines carrying the unique PACS1 p.R203W variant. Organoids were collected on days 21, 40, and 90 to better understand the maturation process. Protein and phosphorylated peptides were processed through mass spectrometry. Outputs were normalized using IRS scaling⁹⁰ in R. Processed outputs were then fitted into a linear model and tested for differential expression over maturation time using the package limma, with genotype as fixed effect, accounting for day and isogenic line as covariates, and calculating duplicate correlation to model plex (batch).⁷¹ Normalized data were also aggregated by isogenic pairs (Table S20) and imported into the short time series (STEM)⁴³ pipeline for expression trend clustering (Figure S9).