Modeling Synaptic Maturation From Growth Cone to Synapse in Human Organoids

ABSTRACT

Human neural organoids (NOs) provide a powerful platform for investigating synaptic development and dysfunction during early neurodevelopment. However, methodologies for isolating functional synaptic structures from these models remain limited. Here, we present a differential centrifugation protocol enabling the enrichment of growth cone particles (GCPs) and immature synaptosomes from air‐liquid interface cerebral organoids (ALI‐COs) at distinct developmental stages (Day 90 and 150). Notably, the method avoids density gradients, requires minimal starting material while maintaining reproducibility across human and murine tissues. Quantitative proteomic profiling revealed significant enrichment of growth cone markers (e.g., GAP43) and classical synaptosomal proteins (e.g., PCLO, BSN, SYN1). Transmission electron microscopy (TEM) confirmed the presence of membrane‐enclosed GCPs with fibrous content and mitochondria in Day 90 isolates, and immature synaptosomes containing synaptic vesicles on day 150. Functional viability of both types of synaptic structures was demonstrated through KCl‐induced depolarization, which triggered phosphorylation changes in growth cone proteins (GAP43, MARCKS, MARCKSL1), cytoskeletal regulators (DCLK1, SHTN1, MARK4, MAP1B) and protein kinases (CAMK2G, PRKCE) in Day 90 GCPs, as well as classical synaptic vesicle cycle proteins (SYN1, DNM1, RPH3A) at Day 150. Overall, this study establishes a centrifugation‐based protocol for isolating growth cones and immature synapses from human organoids, capturing key stages of synaptic development and enabling scalable, patient‐compatible models to study synaptic function and dysfunction in neurodevelopmental and neurodegenerative disorders.

Synapses are implicated in several neurological disorders and psychiatric diseases. The emergence and wide use of neural organoids provide a new opportunity to study human synapses in healthy and disease settings. Therefore, we developed a simple method for the enrichment of synaptosomes and growth cone particles from forebrain organoids. The method is based on differential centrifugation, works with small tissue amounts, and is highly reproducible. We validated the functionality of the isolated structures using KCl stimulation and phosphoproteomics. The method enables detailed mapping of protein composition and function during growth cone pathfinding, synaptogenesis, and establishment of neural circuits in organoids.

Abbreviations

2D

two‐dimensional

3D

three‐dimensional

ACN

acetonitrile

AGC

automatic gain control

ALI

air‐liquid interface

ALI‐CO

air‐liquid interface cerebral organoid

AZ

active zone

BP

biological process

BSN

Bassoon

CAMK2G

calcium/calmodulin‐dependent protein kinase II gamma

CAMK4

calcium/calmodulin‐dependent protein kinase type IV

CASK

Calcium/calmodulin‐dependent serine protein kinase

CC

cellular component

CO

cerebral organoid

DCLK1

doublecortin‐like kinase 1

DCX

Doublecortin

DDA

data dependent acquisition

DIA

data independent acquisition

DLGAP4

disks large‐associated protein 4

DNM1

dynamin‐1

DTT

dithiothreitol

EBs

embryoid bodies

ESI

electrospray ionization

EtOH

ethanol

FA

formic acid

FBO

dorsal forebrain organoid

FC

fold change

FDR

false discovery rate

FWHM

full width half maximum

GAP43

neuromodulin, alternative name: Growth associated protein 43

GCP

growth cone particle

GO

Gene Ontology

GRIA1

glutamate receptor 1

GRM2

metabotropic glutamate receptor 2

HBK

HEPES‐buffered Krebs‐like

HCD

higher energy collision dissociation

HPLC

high performance liquid chromatography

IAA

iodoacetamide

IHC

immunohistochemical

iPSC

induced pluripotent stem cell

LMB

Laboratory of Molecular Biology

m/z

mass to charge ratio

MAP1B

microtubule‐associated protein 1B

MAP 2

microtubule‐associated protein 2

MAPT

microtubule‐associated protein tau

MARCKS

myristoylated alanine‐rich C‐kinase substrate

MARCKSL1

MARCKS‐related protein

MARK4

microtubule‐associated regulatory kinase 4

MF

molecular function

MRC

medical research council

MS

mass spectrometry

MS/MS

tandem mass spectrometry

NCE

normalized collision energy

nLC

Nano Liquid Chromatography

NO

neural organoid

ON

overnight

P1

pellet after initial 800×g centrifugation

P12,000

pellet after 20 min 12 000×g centrifugation

P2

postnatal day 2

P22,000

pellet after 20 min 22 000×g centrifugation

P5,000

pellet after 10 min 5000×g centrifugation

PASEF

parallel accumulation serial fragmentation

PBS

phosphate buffered saline

PCA

principal component analysis

PCLO

Piccolo

PD

proteome discoverer

PPI

protein–protein interaction

PRKCE

protein kinase C type epsilon

PSC

pluripotent stem cell

PSD

postsynaptic density

PSD95

postsynaptic density protein 95, alternative DLG4

PTM

post‐translational modification

RP

reversed phase

RPH3A

Rabphilin 3A

RRID

Research Resource Identifier (see scicrunch.org)

RT

room temperature

S1

supernatant after initial 800×g centrifugation

SFSCM

serum free slice culture medium

SHANK1

SH3 and multiple ankyrin repeat domains protein 1

SHTN1

Shootin1

SP

sodium phosphate

SPS‐MS3

Synchronous Precursor Selection Multi‐Stage Mass Spectrometry

SYN1

Synapsin‐1

SYP

Synaptophysin

TEAB

(triethylammonium bicarbonate)

TEM

transmission electron microscopy

TFA

(trifluoracetic acid)

TMBST

TBS‐Tween20

TMT

tandem mass tags

TOMM20

translocase of outer mitochondrial membrane 20

VGCC

voltage‐gated Ca²⁺ channel

VGLUT1

vesicular glutamate transporter 1

W8

8‐week‐old

1. Introduction

Synapses are essential for the transmission of signals between neurons. Mutations in synaptic genes have been implicated in over 130 neurological conditions, comprising both neurodevelopmental and neurodegenerative diseases (Bayés et al. 2011) and include autism spectrum disorders (Masini et al. 2020), schizophrenia (Legge et al. 2021), epilepsy (Fukata and Fukata 2017), Alzheimer's disease (Wang and Reddy 2017) and Parkinson's disease (Nguyen et al. 2019). Alterations in the development of synapses, the process of synaptogenesis, are implicated in several neurodevelopmental disorders, such as schizophrenia, autism spectrum disorders and Fragile‐X syndrome (Habela et al. 2016; Telias 2019). This underscores the importance of studying synaptic function and development. A commonly used method to study synapses is density‐based enrichment of isolated nerve terminals called synaptosomes. Synaptosomes are formed when brain tissue is homogenized in an iso‐osmotic buffer, resulting in shear‐induced detachment of the synaptic boutons from axons, followed by resealing of the synaptic membrane (Figure 1A) (Hebb and Whittaker 1958; Whittaker 1959). Synaptosomes contain mitochondria, synaptic vesicles, and electron dense active zone (AZ) areas with docked vesicles, which can be visualized with transmission electron microscopy (TEM). The postsynaptic density (PSD) often remains attached to the AZ areas and can be observed with TEM depending on the orientation of the section. Synaptosomes with mitochondria can be viable and metabolically active and undergo respiration in a suitable buffer (Bradford 1970; De Belleroche and Bradford 1972). They can regenerate their membrane potential and be depolarized by chemical stimulation with for example, a brief pulse of elevated KCl concentration, leading to neurotransmitter release by synaptic vesicle exocytosis (Bradford 1970; De Belleroche and Bradford 1972; Nicholls and Sihra 1986) followed by endocytosis to regenerate synaptic vesicles (Cousin and Robinson 2001).

FIGURE 1.

(A) Schematic illustration of the formation of a synaptosome, including a TEM image of a mouse W8 synaptosome. m, mitochondrion; sv, synaptic vesicle; psd, postsynaptic density. (B) List of examples of how synaptosomes can be used. PTM: Posttranslational modification. (C) Workflow of synaptosome enrichment using differential centrifugation. The differential centrifugation method was applied to forebrain organoids at Day 100, ALI‐COs at Day 90 and 150, mouse brain at two developmental stages (postnatal Day 2 (P2) and adult Week 8 (W8)) and healthy cortical tissue from adult human brain. (D) Proteomic data PCA plots showing technical replicates of the different fractions, homogenate (H), P5,000, P12,000, P22,000, and Cytosolic fraction (n = 3–4 technical replicates), from differential centrifugation of human forebrain organoid tissue at Day 100, ALI‐COs at Day 90 (n = 3), and 150, human cortical tissue and newborn (P2) or adult (W8) mouse brain tissue.

Traditional methods for enrichment of synaptosomes from homogenized brain tissue use different kinds of discontinuous density gradient centrifugation (Booth and Clark 1978; Dunkley et al. 1986; Dunkley et al. 2008; Whittaker 1959). In these protocols the brain tissue is first homogenized in an iso‐osmotic sucrose/EDTA buffer followed by an initial low‐force centrifugation (800–1000×g) to pellet nuclei and non‐lysed whole cells in the so‐called P1 fraction. The first supernatant (S1) can then be subjected to density gradient centrifugation, resulting in subcellular fractionation of the material yielding relatively purified fractions of synaptosomes. If a pure fraction is not required, the density gradient centrifugation can be replaced by a simple second centrifugation step of S1 at a higher speed, resulting in a so‐called crude synaptosome fraction in the second pellet (Evans 2015).

Synapses have a complex composition comprising several thousand different proteins involved in processes essential for synaptic function such as synaptic vesicle cycling, neurotransmitter release and synaptic plasticity (Koopmans et al. 2019). Studies of synaptosomes have played a central role in brain research with significant contributions to the present knowledge about different kinds of neurotransmitters and synaptic protein composition and function (Evans 2015) (Figure 1B). Synaptosome research has benefitted from the development of mass spectrometry (MS)‐based proteomic techniques that now enable reliable, large‐scale identification and quantification of thousands of proteins in a single study [23, 24], together with characterization of various post‐translational modifications (PTMs) of proteins, such as phosphorylation, glycosylation, cysteine modifications, acetylation and ubiquitination (Boll et al. 2020; Kang et al. 2018; Larsen et al. 2007; Leutert et al. 2021; Melo‐Braga et al. 2015, 2014; Palmisano et al. 2012). MS‐based proteomics of synaptosomes has been utilized in studies of numerous neurological disorders such as Alzheimer's disease (Chang et al. 2013; Shen et al. 2022), schizophrenia (Paternoster et al. 2019; Zeppillo et al. 2022) and Parkinson's disease (Betzer et al. 2015; Plum et al. 2020). The combination of techniques has been used to study stimulation‐related changes in PTMs of synaptic proteins, mainly phosphorylation, providing a platform to study PTM‐regulated signaling in synaptic transmission (Boll et al. 2020; Engholm‐Keller et al. 2019; Kohansal‐Nodehi et al. 2016; Silbern et al. 2021). Studies of depolarization‐induced protein phosphorylation in synaptosomes have shown changes in phosphorylation or ubiquitination of proteins or sialylation of glycoproteins that are active in synaptic vesicle cycling, including endo‐ and exocytosis, actin dynamics and AZ proteins involved in neurotransmitter release (Ainatzi et al. 2025; Anggono et al. 2006; Boll et al. 2020; Engholm‐Keller et al. 2019; Imoto et al. 2024; Kohansal‐Nodehi et al. 2016; Silbern et al. 2021).

Synaptosome enrichment methods have traditionally been applied to rodent or human (postmortem or surgical) brain tissue. However, animal models are often not adequate when it comes to clinical translation into humans for disease mechanisms or drug development (McGonigle and Ruggeri 2014). Synaptosome functionality from postmortem tissue is limited by the postmortem interval and often only provides information about disease endpoints, while surgical sources are rare. In extensions, synaptosomes have also been prepared from cultured neurons (Bate and Williams 2012; Kishi et al. 1991), however, with very low and crude yields. The emergence of the neural organoid (NO) field, where human pluripotent stem cells (PSCs) are differentiated in three‐dimensional (3D) cultures to form brain‐like tissue, offers a promising new source of synapses derived from living human neuronal tissue, with potentially greater enrichment efficiency compared to 2D cultures. NOs provide a strong parallel representation of the human brain in vitro: they recapitulate the cellular events in early human brain development (Fair et al. 2020; Mulder et al. 2023; Qian et al. 2016; Renner et al. 2017) and gene expression in NOs has been shown to resemble that of the human fetal brain (Amiri et al. 2018; Camp et al. 2015; Qian et al. 2016). The 3D culture of NOs promotes increased cellular interaction and cytoarchitectural organization of the neurons into cortical layers in a timed manner reminiscent of the developing brain (Renner et al. 2017). This leads to higher complexity and cellular diversity than 2D cultures. In cerebral organoids, initial development of the neuroepithelial subventricular zone is followed by radial migration of neurons and formation of the cortical plate from around Day 30 of differentiation, with deep layer neurons appearing around Day 50 and upper layer neurons around Day 60 of differentiation, while inhibitory neurons gradually increase over time from around Day 30 of differentiation (Fair et al. 2020; Mulder et al. 2023; Qian et al. 2016; Renner et al. 2017). Astrocytes start to appear from around Day 60 of differentiation, but distribution has been shown to be sparse until around Day 140 (Mulder et al. 2023; Renner et al. 2017). The presence of a diverse range of cell types in NOs (excitatory neurons, inhibitory neurons, intermediate progenitors, radial glia cells, and astrocytes (Lancaster et al. 2013; Mariani et al. 2015; Paşca et al. 2015; Qian et al. 2016)) allows for development of functional synapses, as detected by immunohistochemical (IHC) labeling of synaptic markers and studies of electrical activity (Fair et al. 2020; Passaro and Stice 2020; Quadrato et al. 2017; Yakoub 2019). Spontaneous electrical activity has been recorded from NOs as early as 52–85 days of differentiation (Mariani et al. 2015; Qian et al. 2016) with later development of more coordinated network activity (3–4 months) (Fair et al. 2020) and complex dendritic morphology (Day 80 of differentiation) (Giandomenico et al. 2021; Qian et al. 2016), indicative of progressive maturation of synapse physiology. Using NO tissue for synaptosome preparation could therefore provide a platform for studying synapses developed from patient‐derived, induced PSCs (iPSCs), enabling functional studies of synapses in relation to diseases and drug development in a human setting. Synaptic function and synapse development could be studied in a controllable in vitro system that can be manipulated, for example, by CRISPR/Cas9 genetic engineering, labeling, or various kinds of stimulation.

Our aim was to establish a method for isolating viable synaptosomes from human NOs to support on‐going investigations of the molecular mechanisms underlying synaptic transmission in human synapses in health and disease. The resulting method provides a straightforward, reproducible, and efficient approach for enriching functional synaptic structures from NOs, enabling the study of synaptic proteins during early brain development, in disease models, and in contexts such as exposure studies or drug testing.

2. Methods

2.1. Generation of Dorsal Forebrain Organoids

Previously established human iPSC line IMR90‐4 (female) was obtained from WiCell Research Institute (Madison, WI, USA, RRID: CVCL_C437) under a material transfer agreement. All procedures were approved by The Scientific Ethical Committee of the Region of Southern Denmark and adhered to institutional and national guidelines. IMR90‐4 cells were cultured on growth factor reduced Matrigel‐coated (Corning, cat. no. 356230) plates and daily supplemented with mTeSR1 medium (Stem Cell Technologies, cat. no. 85870). The cells were used at passages between 34 and 39 to generate dorsal forebrain organoids (FBOs) according to the STEMdiff Dorsal Forebrain Organoid Differentiation Kit (Stem Cell Technologies, cat. no. 08620) with some modifications. In brief, on Day 0, IMR90‐4 cells were dissociated to single cells with Gentle Cell Dissociation Reagent (Stem Cell Technologies, cat. no. 07174) and seeded in one well of an AggreWell 800 24‐well plate (Stem Cell Technologies, cat. no. 34815) at a final density of 10 000 cells per microwell (3 × 10⁶ cells/Aggrewell) in 2 mL FBO Formation medium supplemented with 10 μM ROCK inhibitor (Sigma‐Aldrich, cat. no. Y0503). On Day 6 embryoid bodies (EBs) were transferred to low‐adherent 24‐well plates (Corning, cat. no. 3473) with 400 μL FBO Expansion Medium until Day 11 where they were moved to low‐adherent 6‐well plates (Corning, cat. no. 3776) with 2 mL medium/well and placed on an orbital shaker at 57 rpm (INFORS HT Celltron) to avoid organoid fusions. From Day 43 FBOs were supplemented with 100 U/mL of Penicillin–Streptomycin (Gibco, cat. no. 15140) and 1 μg/mL of Amphotericin B (Thermo Fisher Scientific, cat. no. 15290026). FBOs were cultured for 100 days before applying the differential centrifugation workflow.

All cell and organoid cultures were maintained with 5% CO₂ at 37°C.

2.2. Generation of Cerebral Organoids and Air‐Liquid Interface Cerebral Organoids (ALI‐CO) Cultures

H9 (female) human embryonic stem cells were purchased from WiCell (WA09, RRID: CVCL_9773) and approved for use in this project by the UK Stem Cell Bank Steering Committee. Cells were maintained in StemFlex medium (Thermo Fisher Scientific, cat. no. A3349401) on plates coated with growth factor reduced Matrigel (Corning, cat. no. 356230) and passaged twice weekly with EDTA (Merck, cat. no. 1.08418.1000). Cells were used at max passage 40.

Cerebral organoids were generated according to the STEMdiff Cerebral Organoid Kit (Stem Cell Technologies, cat. no. 08570) seeding 2000 cells per EB. At 15 days in vitro, organoids were manually excised from the Matrigel droplets (Corning, cat. no. 356234) in which they were embedded during the kit protocol, using a needle and fine scalpel under a stereomicroscope, and returned to organoid media. From Day 30, organoid media was supplemented with Matrigel (Corning, cat. no. 356234) dissolved at 2% (v/v).

For culture at the air‐liquid interface (ALI), cerebral organoids (COs) of Day 55–60 were processed to 300 μm slices as previously described (Giandomenico et al. 2021) and collected directly onto cell culture inserts (Millipore, cat. no. PICMORG50) with serum free slice culture medium beneath (SFSCM: Neurobasal (Thermo Fisher Scientific, cat. no. 21103049), 1× B‐27 supplement (Thermo Fisher Scientific, cat. no. 17504044), 1× Glutamax (Thermo Fisher Scientific, cat. no. 35050061), 0.5% (w/v) glucose). SFSCM was supplemented with 1× antibiotic‐antimycotic (Thermo Fisher Scientific, cat. no. 15240062) and additional 1:1000 (v/v) Amphotericin B (Merck, A2942‐20ML). Cell culture inserts were raised vertically within their respective wells using custom‐manufactured stages (MRC LMB mechanical workshop) such that more SFSCM, approximately 4.5 mL, could fit beneath each insert, permitting twice‐weekly full media changes.

ALI‐COs were cultured until Day 90, when most neural organoid models have started to show spontaneous electrical activity (Mariani et al. 2015; Qian et al. 2016) and ALI‐COs are expected to show more mature spine morphology (Giandomenico et al. 2021). Another batch was cultured until Day 150 of differentiation, where neural organoids are expected to show more coordinated network activity (Fair et al. 2020), before applying the differential centrifugation workflow. All cell, organoid, and ALI‐CO cultures were maintained with 5% CO₂ at 37°C.

2.3. Mouse Brains

All animal work was performed at the University of Southern Denmark animal facility (https://www.sdu.dk/en/om‐sdu/institutter‐centre/biolab_biomedicinsk_laboratorium) in accordance with the Animal Welfare Body (IACUC) at the University of Southern Denmark (License number 2021‐15‐0201‐01054). Mouse brain tissue was obtained from 2‐day‐old C57BL/6J mice or 8 weeks‐old male C57BL/6J mice (Taconic or Janvier, RRID:MGI:5650797). Mice were housed in the Biomedical Laboratory at University of Southern Denmark (SDU) under a 12:12 h light:dark cycle with food and water available ad libitum until use. C57BL/6J mice from Taconic were used for breeding mouse pups at the Biomedical Laboratory, SDU. Mice were euthanized by cervical dislocation, and their brains were immediately dissected, removing the cerebellum from adult individuals. The brains were used for either the differential centrifugation enrichment workflow (2‐day‐old (P2) and 8‐week‐old adult tissue (W8)) or a standard Percoll enrichment procedure (only adult tissue) as described below.

2.4. Human Cortical Tissue

During surgery and after informed consent (approval ID S‐20130048 from the Danish Research Ethics Committees), fresh human, non‐tumorous cortical tissue samples were collected from a non‐eloquent brain area within the exposed surgical field during surgeries for high‐grade gliomas conducted at Odense University Hospital, SDU. The tissues were collected directly in ice‐cold phosphate‐buffered saline (PBS, Gibco, cat. no. 70011‐036) and immediately processed within 1 h.

2.5. Serial Differential Centrifugation Enrichment of Synaptosomes

Mouse brains (W8 or P2), FBOs, ALI‐COs or surgery‐derived fresh human cortical tissue were homogenized in ice cold sucrose/EDTA buffer (0.32 M sucrose (Sigma, cat. no. 84097), 1 mM EDTA (Merck, cat. no. 1.08418.1000), 5 mM Tris base (Sigma, cat. no. 93352), pH 7.4, using around 700–800 μL buffer/100 mg of tissue) applying 6 up‐and‐down strokes with a tissue grinder (15‐mL Potter‐Elvehjem type Teflon‐glass tissue grinder, Wheaton) at 700 rpm. A homogenate aliquot of 100–200 μL was obtained and stored at −70°C as a reference/control sample for the enrichment. The remaining homogenate was centrifuged at 800xg for 15 min at 4°C to pellet nuclei and cellular debris, and after collecting the supernatant (S1 fraction) the centrifugation was repeated after redissolving the pellet (P1 fraction) in the half amount of sucrose buffer. The protein concentration of the S1 fraction was measured using a nanophotometer (N60, Implen), and the S1 fraction was split into 3 or 4 replicates of 1.5–3 mg of protein each. Dithiothreitol (DTT, Sigma cat. no. D9163) was added to the samples to a final concentration of 0.25 mM before starting the differential centrifugations (Dunkley et al. 2008). Finally, samples went through the differential centrifugations, 5000×g for 10 min, 12 000×g for 20 min, 22 000×g for 20 min, all at 4°C, and the pellets (P5,000, P12,000, P22,000) and the supernatant from the last centrifugation (S22,000, named the “cytosolic” fraction) were collected (Figure 1C).

2.6. Standard Percoll Gradient Enrichment of Synaptosomes From Mice

Mouse brains were homogenized in sucrose/EDTA buffer following a standard Percoll density gradient synaptosome enrichment procedure (Dunkley et al. 2008). Briefly, brains were homogenized with 6 up‐and‐down strokes at 700 rpm using a Teflon‐glass tissue grinder (Wheaton) and kept on ice or at 4°C during the following procedure until stimulation. A homogenate aliquot (100–200 μL) was obtained (stored at −70°C) as a reference/control for the enrichment and the remaining homogenate was centrifuged at 800×g for 10 min; the supernatant (S1) was obtained and DTT was added to a final concentration of 0.25 mM. Carefully, 2 mL of S1 was layered on top of each Percoll density gradient (4 layers of 3%, 10%, 15%, and 23% Percoll (Cytiva, cat. no. 17089102) in sucrose/EDTA buffer with 0.25 mM DTT) and centrifuged at 20000 rpm for 5 min when reaching top speed in a Sorvall RC 5C plus Harvest centrifuge using a Sorvall SS34 rotor. The F3 and F4 layers were collected and washed with sucrose/EDTA buffer followed by another centrifugation at 16000 rpm for 15 min. The pellets were collected into low‐binding Eppendorf tubes (Sorenson BioScience Inc., cat. no. 39640T), centrifuged for 10 min at 950×g, and the sucrose buffer was removed. Fractions for label‐free proteomic analysis were stored at −70°C.

2.7. Depolarization of Synaptosomal Fractions With High KCl

Stimulation of the fractions was performed as previously described (Craft et al. 2008). Fractions from the differential centrifugation series enrichment were resuspended in 37°C HEPES‐buffered Krebs‐like (HBK) control buffer (4.7 mM KCl (Sigma, cat. no. P9541), 118 mM NaCl (VWR, cat. no. 27810.295), 20 mM HEPES (Sigma, cat. no. H3375), 1.18 mM MgSO₄ (Merck, # 7487‐88‐9), 1.2 mM CaCl₂ (Merck, cat. no. 2366663), 0.1 mM Na₂HPO₄ (Merck, cat. no. 1.06580.0500), 10 mM glucose (Sigma, cat. no. G7021), 10 mM pyruvate (Gibco, cat. no. 11360), 25 mM NaHCO₃ (Sigma, #71345), pH 7.4, which had been bubbled with carbogen (95% O₂, 5% CO₂) for approx. 1 h before use) and incubated at 37°C for 1 h to allow regeneration of the resting membrane potential. Subsequently, 15 s depolarization was performed by adding an equal volume of high KCl HBK buffer (147.7 mM KCl, 20 mM HEPES, 1.18 mM MgSO₄, 1.2 mM CaCl₂, 0.1 mM Na₂HPO₄, 10 mM glucose, pH 7.4) resulting in a final concentration of 76.2 mM KCl. In the depolarizing buffer NaCl content is reduced to maintain constant osmolarity. The stimulation was stopped by snap‐freezing in liquid nitrogen, and the samples were kept at −70°C until further use. Control samples were stimulated in the same way but with HBK control buffer.

2.8. Sample Lysis and Digestion of Proteins

The sucrose/EDTA buffer from the enrichment protocols was removed from the homogenate samples and cytosolic fractions (S22,000) using 10 kDa spin filters (Amicon Ultra‐0.5 Centrifugal Filter Unit, MilliporeSigma, cat. no. UFC501024) by centrifuging at 11000×g for 20 min at 4°C. The samples were washed ×2 with 300 μL of 50 mM triethylammonium bicarbonate (TEAB, Sigma cat. no. T7408), pH 8. All fraction and homogenate samples were dissolved in lysis buffer (1% sodium deoxycholate (SDC, Sigma cat. no. D6750) in either 50 mM TEAB or 100 mM HEPES, pH 8) in low‐binding Eppendorf tubes (Sorenson BioScience Inc.). Samples were probe sonicated 3 × 10 s at 40% amplitude on ice, centrifuged for 10 min at 14000×g and the supernatants were transferred to new tubes. Protein concentrations were measured using a Nanophotometer (Implen). An aliquot of protein (10–20 μg for label‐free DIA analysis and 50–70 μg for TMT‐labeling and phosphopeptide enrichment) was prepared from each sample and the proteins were reduced with 10 mM DTT at room temperature (RT) for 20 min followed by alkylation with 20 mM iodoacetamide (IAA, Sigma cat. no. I1149) for 30 min in the dark at RT. After alkylation, the IAA reaction was quenched by raising the DTT concentration to 15 mM. Samples were predigested with 0.04 active units/mg Lys‐C (Lysyl Endopeptidase, Wako, cat. no. 129–02541) for 1 h at RT followed by overnight (ON) digestion with 5% trypsin (in‐house, methylated trypsin (Heissel et al. 2018)) at 37°C, pH 8. The following day, samples were incubated 1 h at 37°C with 1% additional trypsin.

2.9. Data Independent Acquisition (DIA) Workflow Using Label‐Free Quantification of Proteins

After Lys‐C/tryptic digestion, the SDC was precipitated using 2% formic acid (FA, Merck, cat. no. 1.11670.0250) followed by centrifugation and samples were lyophilized and subsequently resuspended in 0.1% FA. Peptide concentrations were measured using Pierce Quantitative Fluorometric Peptide Assay (Thermo Fisher Scientific, cat. no. 23290) according to manufacturer instructions and samples were diluted accordingly in 0.1% FA to a final concentration of 0.1 μg/μL.

Differential centrifugation fractions from FBOs, mouse (P2 and W8), human cortex and Percoll density gradient F3 and F4 fractions from adult mouse brain (W8) were analyzed the following way: a total of 2 μL (200 ng of peptides) was loaded onto a 15 cm × 75 μm, C18 1.6 m Aurora Elite column (ESI Source Solutions) coupled directly to a timsTOF Pro (Bruker) instrument using buffers A (0.1% FA) and B (95% acetonitrile (ACN, VWR cat. no. 83640.320), 0.1% FA). Peptides were eluted during a 30 min gradient (flow rate: 0.4 μL/min, increasing proportion of buffer B: from 0 to 2 min: 5%, from 2 to 18 min: 5%–28%, from 18 to 22 min: 28%–45%, from 22 to 23 min: 45%–95%, from 23 to 26 min: 95%, from 26 to 26.5 min: 95%–5%, from 26.5 to 30 min: 5%) on a Dionex Ultimate 3000 high performance liquid chromatography (HPLC)‐system (Thermo Fisher Scientific). Samples were run scanning from 100 to 1700 m/z, operating in positive ion mode and using a standard DIA‐PASEF (parallel accumulation serial fragmentation) method (mass range: 395.6 Da to 1020.6 Da, cycle time estimate: 1.06 s).

The ALI‐CO differential centrifugation fractions were analyzed in the same way with minor modifications in the gradient (30 min gradient at flow rate: 0.4 μL/min, increasing proportion of buffer B: from 0 to 2 min: 1%, from 2 to 2.5 min: 1%–5%, from 2.5 to 18 min: 5%–28%, from 18 to 22 min: 28%–45%, from 22 to 23 min: 45%–95%, from 23 to 26 min: 95%, from 26 to 26.5 min: 95%–1%, from 26.5 to 30 min: 1%) and DIA‐PASEF method (mass range: 400–1201 Da, cycle time estimate: 1.80 s).

The mouse (P2 and W8) label‐free proteomic samples were analyzed on an Orbitrap Exploris 480 mass spectrometer using a DIA method. A total of 200 ng of peptides was loaded onto a Vanquish Neo UHPLC system (Thermo Fisher Scientific) using a 2‐column setup (Trap column of 0.5 cm, inner diameter: 300 μm and Separation column of 23 cm, inner diameter: 100 μm). Peptides were eluted with an increasing amount of buffer B at a flow rate of 0.3 μL/min during a 35 min gradient: from 0 to 27 min: 2%–29%, from 27 to 30 min: 29%–42%, from 30 to 31 min: 42%–70%, from 31 to 34 min: 70%, from 34 to 35: 70%–100%. Peptides were analyzed using a standard DIA Scan method (precursor mass range (m/z): 400–1000, cycle time of 3 s, orbitrap resolution: 30000 full width half maximum (FWHM)).

2.10. Peptide Identification and Label‐Free Quantification

The raw data from DIA‐MS analyses was searched with DIA‐NN (v1.8.1) against computationally generated spectral libraries of human and mouse proteomes performed with the following search parameters: mass accuracy was set to 10 ppm whereas for MS1 mass accuracy to 15 ppm, peptide length range of 7–30 amino acids, precursor m/z range of 100–1700, fragment ion m/z range of 100–1700, neural network classifier in Single‐pass mode, Robust LC (high precision) mode was used with retention time (RT)‐dependent normalization enabled, maximum of 2 missed cleavages, N‐terminal methionine (M) excision and methionine oxidation as variable modification was enabled, C carbamidomethylation was used as fixed modification.

2.11. TMT‐Labeling of Samples for Phosphopeptide Enrichment

After Lys‐C/tryptic digestion, samples for phosphopeptide enrichment (high KCl stimulated fractions and controls) were labeled with isobaric Tandem Mass Tags (TMT) using TMTpro 16‐plex or TMTpro 18‐plex (Thermo Fisher Scientific, cat. no. A44520 and A52046) according to manufacturer instructions as follows: TMT‐reagent (0.25 mg) for 50 μg of sample were thawed for 5 min with frequent vortexing in 100% ACN. Reagents were mixed with 50 or 70 μg of peptides per sample in lysis buffer (1:5, ACN:sample v/v). The pH was adjusted to 8 with 1 M TEAB and the labeling reaction was allowed for 1–1.5 h at RT. Labeling efficiency and TMT intensities were tested by mixing 1 μL of each sample in 2% FA followed by centrifugation at 20000×g for 10 min to precipitate and pellet the SDC. Approximately 1 μg of the mixed sample was run on LC–MS/MS using an orbitrap Exploris 480 mass spectrometer. The TMT labeled samples were mixed according to the obtained test. The SDC was precipitated using 2% FA and pelleted with centrifugation (20000×g for 10 min), and the combined sample was lyophilized until a remaining volume of 100 μL, which was then used for TiO₂ enrichment of phosphopeptides.

2.12. TiO₂ Enrichment of Phosphopeptides

The enrichment of phosphopeptides was performed on the TMT labeled peptide mixture using TiO₂ (Larsen et al. 2007; Thingholm et al. 2006). This highly efficient method utilizes TiO₂ enrichment in the presence of glycolic acid, trifluoracetic acid (TFA), and ACN, which dramatically prevent binding of non‐phosphorylated peptides to TiO₂ (Larsen et al. 2007). After binding of the phosphopeptides to the TiO₂ resin (GL Sciences Japan, cat. no. 5020‐75010) in 5% TFA (Merck, cat. no. 1.08178.0050), 80% ACN, and 1 M glycolic acid (Sigma, cat. no. 124737) for 10 min, the solution was centrifuged, and the supernatant was re‐incubated with half the amount of TiO₂ for another 10 min. The supernatant from the second incubation was collected, labeled as “non‐modified” proteins, and dried for further use. The TiO₂ resin from the two incubations was washed using 200 μL loading buffer, 200 μL 80% ACN, 1% TFA, and 100 μL 10% ACN, 0.5% TFA, respectively. After the final wash, the beads were lyophilized for 5 min, and subsequently, the phosphopeptides were eluted with 5% ammonia water (Merck, cat. no. 1.05428.0500), pH 11. After incubation, the supernatant was collected and lyophilized without TiO₂ beads for subsequent high pH reversed phase fractionation.

2.13. Desalting of “Non‐Modified” Peptides

The flow‐through from the TiO₂ enrichment was dried and subsequently resolubilized in 2 mL 0.1% TFA. The peptides were desalted on a Waters Oasis HLB Cartridge (Waters cat. no. 186000132). The HLB cartridge was washed with 2 mL 100% Methanol (VWR cat. no. 20864.320) followed by 2 mL 100% ACN using a 5 mL syringe. The HLB resin was equilibrated with 2 mL 0.1% TFA and the sample was loaded slowly using the 5 mL syringe. The HLB cartridge was washed with 2 mL 0.1% TFA and the peptides were eluted with 1 mL 50% ACN in water and subsequently the HLB cartridge was washed with 1 mL 70% ACN eluting remaining peptides. Both elutions were lyophilized and stored until further use.

2.14. High pH Reversed Phase (RP) Fractionation

The “non‐modified” peptide and phosphopeptide samples were dissolved in high pH RP fractionation solvent A (20 mM ammonium formate (Riedel‐de Haën cat. no. 25204), pH 9.3) and pH was adjusted with 1 M TEAB to around 9. Samples were fractionated on a Dionex Ultimate 3000 HPLC system (Thermo Fisher Scientific) using an Acquity UPLC‐Class CSHTM C18 column (Waters) into 15 concatenated fractions with the following gradient of solvent B (80% ACN and 20% buffer A): sample loading with 2% solvent B, peptide elution (and fraction collection every 122 s) from 2% to 50% in 59 min, 50%–70% in 10 min, 70%–95% in 5 min and for 10 min at 95% solvent B (during which collection of fractions was terminated). Flow rate was 5 μL/min. Fractions were lyophilized and resuspended in 0.1% FA for HPLC‐ Tandem MS (MS/MS) analysis.

2.15. Data Dependent Acquisition (DDA) Nano Liquid Chromatography‐Electrospray‐Tandem Mass Spectrometry (nLC‐ESI‐MS/MS)

All samples were dissolved in 0.1% FA and analyzed by nLC‐ESI‐MS/MS using an EASY‐nLC (Thermo Fisher Scientific) with buffer A (0.1% FA) and buffer B (95% ACN, 0.1% FA) connected online to an Orbitrap Eclipse Tribrid Mass Spectrometer (Thermo Fisher Scientific, USA). The separation was performed on an in‐house‐made fused silica capillary two‐column setup, a 3 cm pre‐column (100 μm inner diameter packed with Reprosil‐Pur 120 C18‐AQ, 5 μm (Dr. Maisch GmbH)) and an 18 cm pulled emitter analytical column (75 μm inner diameter packed with Reprosil‐Pur 120 C18‐AQ, 3 μm (Dr. Maisch GmbH)). The peptides were eluted with increasing amount of the buffer B (95% ACN, 0.1% FA) from 2% to 40% in 90 and 140 min for phosphopeptide fractions and “non‐modified” peptides, respectively. All spectra were generated using DDA acquisition with positive ion mode mass spectrometry. The full MS was performed in the mass range of 450–1500, in the Orbitrap with a resolution of 120 000 FWHM, a maximum injection time of 50 ms and an Automatic Gain Control (AGC) target value of 1 × 10⁶. Hereafter, the peptides were selected for MS/MS using higher energy collision dissociation (HCD) with normalized collision energy (NCE) setting as 35, resolution of 50 000 FWHM, AGC target value of 1 × 10⁵ ions and maximum injection time of 150 ms.

Additionally, the “non‐modified” samples were run on the Eclipse Orbitrap in TMT Synchronous Precursor Selection Multi‐Stage Mass Spectrometry (SPS‐MS3) mode with real‐time search (Ting et al. 2011) enabled against a Uniprot human/mouse protein database (static modifications: carbamidomethyl (C), TMTpro 16‐ or 18‐plex (Kn), maximum missed cleavages: 1) with a maximum search time of 35 ms. The MS/MS isolation window was set to 2 m/z and peptides with confident identification using the built‐in search engine were selected for MS3 fragmentation using HCD with NCE setting at 55, Orbitrap resolution of 30 000 FWHM, AGC target value of 500% with a scan range of 100–500 m/z.

2.16. Peptide Identification and Quantification of “Non‐Modified” Peptide and Phosphopeptide Samples

All raw files were searched against the Uniprot database of Homo sapiens or Mus Musculus in Proteome Discoverer (PD) 2.5.0.305 (Thermo Fisher Scientific, USA) using an in‐house Mascot server (v2.6) and SEQUEST HT. For HCD fragmentation, search parameters were as follows: cleavage specificity Trypsin/P; precursor mass tolerance of 10 ppm and fragment mass tolerance of 0.05 Da. The phosphopeptides search included the parameters: Maximum missed cleavages, 2; Static modifications, TMTpro (K), TMTpro (N‐term) and Carbamidomethyl (C); Dynamic modifications, Phosphorylation (S, T, Y). For SPS‐MS3 of “non‐modified” peptides, the data were searched in PD 2.5.0.305 using SEQUEST with the same Uniprot Human/Mouse database as the real time search on the Eclipse. Search parameters were as follows: cleavage specificity Trypsin/P; precursor mass tolerance of 10 ppm and fragment mass tolerance of 0.8 Da. The search included the parameters: Maximum missed cleavages, 2; Static modifications, TMTpro (K), TMTpro (N‐term) and Carbamidomethyl (C). For the phosphopeptides the search was performed first in Mascot and then in SEQUEST using the same Uniprot databases and same search parameters, except for 0.05 mass tolerance in MS2 data and variable phosphorylation on S/T/Y. The percolator software in PD 2.5 was used for filtering for false discovery rate (FDR) of < 1% for proteins and peptides. All datasets from PD 2.5 were exported to Excel (Microsoft) for further processing. For the Mascot searches peptides were accepted if the Mascot score was ≥ 15.

2.17. Experimental Design, Statistical Analysis and Data Processing

No statistical methods were used to predetermine sample size. This study was designed as a method development and proof‐of‐principle investigation aimed at adapting a synaptosome enrichment protocol based on differential centrifugation for neural organoid tissue and benchmarking its performance against established brain tissue preparations.

Experiments were performed once per tissue type using material derived from a single preparation for each source, including dorsal forebrain organoids, ALI‐COs, mouse brain tissue (P2 and W8), and adult human cortical tissue. These tissue types represent independent biological systems used to evaluate the robustness and generalizability of the enrichment strategy rather than biological replicates intended for statistical inference.

For each tissue type, 3–4 technical replicates were included to assess the technical reproducibility of the enrichment and downstream processing workflow, including stimulation and phosphoproteomic analysis. Technical replicates were used to evaluate the consistency of measurements within each experiment and do not represent independent biological samples.

Sample numbers were determined by experimental feasibility, limited availability of starting material—particularly for organoid and adult human brain samples—and the minimum material requirements of phosphoproteomic analyses. Functional viability of the enriched structures was assessed through stimulation‐induced phosphorylation responses, providing an orthogonal validation of the method beyond enrichment efficiency alone.

No formal tests for outliers were performed. All data points obtained from the experiments were included in the analyses, and no data were excluded.

Statistics of the label‐free, proteomic datasets was performed with the online available PolyStest (Schwämmle et al. 2020), which does not assume normality, using the final PolyStest FDR values with Benjamini‐Hochberg‐based correction for multiple testing. Before statistical analysis, all label‐free DIA datasets were sorted to obtain a maximum of 25% missing values for each protein across the whole dataset. Regulations in label‐free datasets were considered significant when FDR ≤ 0.01 and |fold change (FC)| ≥ 1.5. For TMT‐labeled phosphopeptide datasets, statistics was derived from Proteome Discoverer applying the background‐based t‐test with Benjamini‐Hochberg‐based correction for multiple testing. Regulations in phosphopeptides were considered significant when adjusted p‐value ≤ 0.05, fold change ≥ 1.3 up or down and grouped abundance CV < 30%. “Non‐modified” TMT‐labeled data was used only for identification of background protein lists for Gene Ontology (GO) term enrichment analyses.

For the proteomic DIA datasets, GO term enrichment analyses (Cellular component [CC]) were performed in Cytoscape (v3.10.0) applying the StringApp (Doncheva et al. 2019) using confidence score cutoff 0.7 and filtering for redundant terms in GO term enrichment analysis (cutoff: 0.4). Full lists of identified protein groups per dataset were used as background for each enrichment analysis, for example, the full list of protein groups identified from the FBO fractions dataset was used as background for the GO term enrichment analysis of FBO fractions etc.

For the phosphopeptide datasets, proteins with regulated phosphopeptides from P5,000 and P12,000 fractions were analyzed together. GO enrichment analysis (CC and Molecular Function [MF]) was performed with Cytoscape (v3.10.0) and regulated phosphopeptides of two selected protein–protein interaction (PPI) clusters (confidence score cutoff 0.4) were visualized using the OmicsVisualizer app (Legeay et al. 2020). All identified master proteins in “non‐modified” samples (based on ≥ 2 unique peptides) and phosphopeptide datasets combined were used as background for the enrichment analysis.

Dot plots, principal component analysis (PCA) plots, and bar graphs were made in R (v4.3.0) (R Core Team 2013) using the package ggplot2 (Wilkinson 2011), and ChatGPT (vAugust 3, 2023) was used for coding assistance. Heatmaps were made using the pHeatmap package (Kolde 2019) in R.

The statistical test used for changes in synaptic and growth cone marker proteins was a Wilcoxon Rank Sum test, suited for data where normal distribution and equal variance cannot be assumed. Tests were conducted as one‐sided, testing for up‐regulation in P5,000, P12,000, and P22,000 fractions and down‐regulation in Cytosolic fractions.

2.18. Immunofluorescent Labeling

Organoids were fixed with 4% paraformaldehyde (Thermo Fisher Scientific, cat. no. 28908) in PBS (Gibco, cat. no. 70011‐036) for 1 h at RT and soaked in 30% sucrose in PBS ON or until further processed. Organoids were embedded in OCT mounting media (VWR, 361603E), snap frozen in < −50°C ethanol and sectioned in 30 μm sections on a Leica CM1860 cryostat at −17 to −20°C. Sections were placed on microscope slides (Thermo Fisher Scientific, cat. no. J1800AMNZ) and stored at −20°C. Slices were hydrated with PBS and washed 2 × 5 min with wash buffer (0.1% TritonX100 (Plusone, 17‐1315‐01) in PBS). Permeabilization and blocking buffer (5% donkey serum (BioWest, S2170‐100)) was added followed by 30 min incubation at RT. Primary antibodies diluted in wash buffer incl. 1% donkey serum (BioWest) were added (see dilutions in Table 1) and slices were incubated ON at 4°C to allow binding. Slices were washed in wash buffer for 3 × 10 min followed by incubation for 1 h in the dark with secondary antibodies diluted in wash buffer and 1% donkey serum (see dilutions in Table 2). Secondary antibodies were removed, and slices were washed with PBS for 3 × 10 min followed by mounting with DAPI‐containing ProLong Diamond Antifade mountant (Invitrogen, P36965). The slices were stored in the dark at 4°C until image acquisition. The images were obtained with confocal microscopy on a Nikon A1R confocal unit coupled to a Ti‐2 LFOV microscope. Images were analyzed with ImageJ (v1.53t).

TABLE 1.

Primary antibodies for labeling of synapses in FBO and ALI‐CO slices.

Antigen	Species	Company	Cat. no.	RRID	Dilution
MAP2	Mouse	Sigma	M1406	AB_477171	1:2000
PSD95	Goat	Abcam	ab12093	AB_298846	1:100
SYP	Mouse	Merck Millipore	S5768	AB_477523	1:200
SYN1	Rabbit	Abcam	ab254349	AB_2920663	1:500
VGLUT1	Rabbit	Synaptic Systems	135 303	AB_887875	1:1000

TABLE 2.

Secondary antibodies for labeling of synapses in FBO and ALI‐CO slices.

Secondary antibody	Company	Cat. no.	RRID	Dilution
Alexa Fluor 488 donkey‐anti‐mouse IgG (H + L)	Life Technologies	A21202	AB_141607	1:5000
Alexa Fluor 647 donkey‐anti‐rabbit IgG (H + L)	Life Technologies	A31573	AB_2536183	1:500
Alexa Fluor 568 donkey‐anti‐goat	Life Technologies	A11057	AB_142581	1:1000

2.19. Transmission Electron Microscopic (TEM) Analysis

Fractions were fixed in 3% glutaraldehyde (Merck, cat. no. 1042390250) in 0.1 M sodium phosphate (SP) buffer (78 mM Na₂HPO₄, 22 mM NaH₂PO₄ (Sigma‐Aldrich, cat. no. S3139), pH 7.3) for 0.5–1 h on ice. The fixing solution was replaced with 0.1 M SP buffer and samples were stored at 4°C until further processing. The fractions were embedded in 4% Batco agar at 45°C (BD diagnostics, 214 010) and washed ×2 in 0.1 M SP buffer. The samples were postfixed in 1% OsO4 (Sigma‐Aldrich, cat. no. 201030, in 0.1 M SP buffer) for 1 h at RT and covered in folio followed by washing ×2 in Milli‐Q water. The samples were subsequently dehydrated in a series of increasing concentrations of ethanol (EtOH): 50% EtOH (10 min), 70% EtOH (10 min), 96% EtOH (10 min) and 99% EtOH (3 × 20 min). The EtOH series was followed by an intermediate solution of propylene oxide (Merck, cat. no. 8070271000) (2 × 10 min). Samples were then gradually infiltrated in Epon (812 Resin, TAAB Laboratories Equipment Ltd., T031) and incubated in a rotor ON in pure Epon. Next, the samples were embedded in Epon and polymerization was allowed for 48 h at 60°C. The specimens were cut into semi‐thin sections of 2 μm with a glass knife (LKB Bromma 7800, Leica Microsystems) on an ultramicrotome (Reichert Ultracut S, Leica, Microsystems) followed by staining with 1% toluidine blue (VWR, cat. no. 34187.185) in 0.1% Borex (VWR, cat. no. 27727.231), to identify a section of interest. Ultra‐thin sections of 50–60 nm were cut with a diamond knife (MD1035) on an ultramicrotome (Reichert Ultracut S, Leica, Microsystems) and collected on 150 mesh copper grids (Gilder, cat. no. G150‐C3) and coated with a parlodion‐amyl acetate film (EMS). Finally, the sections were contrasted using 2% uranyl acetate (Polyscience, cat. no. 21447) and 1% lead citrate (Reynold, cat. no. 1963). The micrographs were examined with a Talos L120C transmission electron microscope operating at 80 kV and images were acquired using a Ceta camera and Velux software (Olympus).

2.20. Western Blotting

Samples were lysed in 1% SDC and 50 mM TEAB, pH 8, and sonicated 3 × 10 s on ice at 40% amplitude. Following sonication, samples were centrifuged at 14000×g for 10 min at RT and supernatant was transferred to new Eppendorf tubes (Sorenson BioScience Inc.). Protein concentration was determined by Nanophotometer (Implen). For each sample, 3.2 μg protein was separated using Bolt 4%–12% Bis‐Tris Plus Gels (Thermo Fisher Scientific, cat. co. NW04125BOX) with samples loaded in NuPage LDS Sample buffer (Thermo Fisher Scientific, cat. no. NP0007) with NuPage Sample Reducing agent (Thermo Fisher Scientific, cat. no. NP0004) and PageRuler Plus pre‐stained protein standard (Thermo Fisher Scientific, cat. no. 26620) for size reference. Proteins were transferred to activated PVDF membranes (Millipore, cat. no. IPVH08100) using the Trans‐Blot Turbo Transfer System (BioRad, cat. no. 1704150) and membranes were blocked with 5% skimmed milk (VWR, cat. no. 999999‐99‐4) in TBS‐Tween20 (TBST, Sigma, cat. no. P1379) buffer for 1 h and incubated overnight at 4°C in 5% milk/TBST with the following primary antibodies: rabbit anti‐Translocase of outer mitochondrial membrane 20 (TOMM20) (Abcam, RRID: AB_2889972) 1:1000 or mouse anti‐Synaptophysin (SYP, Millipore, RRID: AB_477523) 1:200. Mouse anti‐β‐actin HRP‐linked (Abcam, RRID: AB_867494) 1:50000 was used as loading control. Following 3 × washes in TBST, the membranes were incubated for 2 h at RT in TBST with the following secondary antibodies: anti‐rabbit IgG, HRP‐linked (Cell Signaling, RRID: AB_2099233) 1:10000 or anti‐mouse IgG, HRP‐linked (Abcam, RRID: AB_955440) 1:10000. Following 3 × washes in TBST, the membranes were visualized with Immobilon ECL Ultra Western HRP Substrate (Millipore, car. No. WBULS0100) using an Amersham 680 Imager (GE Healthcare).

3. Results

3.1. The Differential Centrifugation Protocol Is Highly Reproducible Across Brain Tissue Types

After several attempts to enrich synaptosomes from NOs using a standard Percoll density gradient method yielding no clear F3/F4 (synaptosome enriched) layers, we decided to develop a different approach. Differential centrifugation strategies have previously been used in combination with TMT‐based proteomics to characterize organelle‐specific proteins in various cell types. A study from 2019 by Geladaki et al. observed mitochondria pelleting at 3000×g, Peroxisomes at 5000xg, lysosomes at 12000×g and Golgi and ER from 15000 to 30000×g (Geladaki et al. 2019). Based on an estimated density of synaptosomes, we expected these structures to be pelleted around 5000–12 000xg. The workflow we developed for differential centrifugation enrichment of synaptic structures from human NOs is shown in Figure 1C. This protocol was applied to multiple neural tissue types including two types of NOs (FBOs at Day 100 and ALI‐COs at Day 90 and 150, cell culture images are shown in Figure S1A) together with newborn mouse brain (postnatal Day 2, P2), adult (Week 8, W8) mouse brain and adult human cortical tissue (surgery derived, healthy tissue). Prior to the enrichment from NOs, the presence of synapses in the NO tissues was confirmed with IHC labeling of pre‐ and postsynaptic markers (Synapsin‐1 [SYN1], Vesicular glutamate transporter 1 [VGLUT1], synaptophysin [SYP], Postsynaptic density protein 95 [PSD95, also called DLG4], and Microtubule‐associated protein 2 [MAP2]), showing co‐localization in both organoid types (Figure S1B–D). A standard procedure for enrichment of synaptosomes, based on discontinuous Percoll density gradient centrifugation (Dunkley et al. 2008), was performed in parallel on adult mouse brain (W8) for comparison. The Percoll‐based synaptosome enrichment generated the synaptosome‐containing fractions F3 and F4 and the pellet from the first 800×g centrifugation (P1) was used as a negative reference for synapse‐specific proteins for the Percoll enrichment procedure. Four technical replicates of the new differential centrifugation samples were examined with LC–MS/MS‐based DIA proteomics and principal component analysis (PCA) (Tables [Link], [Link], [Link], [Link], [Link], [Link]) showed similar fraction patterns between each of the tissue types, indicating high reproducibility between brain species and organoid types (Figure 1D).

3.2. Evaluation of Protein Content From the Differential Centrifugation Fractions

To assess enrichment across fractions, we identified fraction‐specific proteins (defined as fold change (FC) ≥ 1.5 relative to the corresponding total homogenate sample, FDR < 0.01, see lists in Tables [Link], [Link], [Link], [Link], [Link], [Link]) and performed Gene Ontology Cellular Component (GO CC) enrichment using the full dataset as background. Figure 2 summarizes enriched GO CC terms for each fraction derived from FBOs, ALI‐COs (Day 90 and 150), newborn mouse (P2) and adult human cortex; top 20 results from each tissue are in Figure S2A–X and full results can be found in Table S11.

FIGURE 2.

Summary of Gene Ontology Cellular Component (GO CC) enrichment in the different fractions (Homogenate (H), P5,000, P12,000, P22,000, and Cytosolic) from the differential centrifugation protocol. The enrichment was based on fraction‐specific proteins (FC ≥ 1.5 compared to homogenate, FDR ≤ 0.01) for each of the five tissue types: forebrain organoids at Day 100 (FBO), ALI‐COs at Day 90 and 150, mouse brain at postnatal Day 2 (P2) and healthy cortical tissue from adult human brain (Human cortex). For all tissue types n = 4 technical replicates, except ALI‐COs Day 90, which was based on n = 3 technical replicates. Mint: Mitochondrial terms; green: neuronal and synaptic terms; blue: Endomembrane system terms; pink: Cytosolic or nucleus terms. The terms were selected manually based on designated key‐words (e.g., “synaptic”, “mitochondrial”, “lysosomal,” “golgi,” “cytosolic”) together with the average FDR across tissue types, total number of proteins and number of times the term appeared across the datasets. All enriched term FDRs are below 0.05.

For all tissues, P5,000 fractions were enriched in mitochondrial proteins. In addition, FBOs and Day 150 ALI‐CO also showed enrichment of synaptic terms, suggesting pre‐synaptic enrichment. In contrast, Day 90 ALI‐COs and human cortex P5,000 fractions primarily showed mitochondrial terms. All P12,000 fractions showed enrichment in several synaptic terms, together with mitochondrial, lysosomal, and a few ER‐related terms. Synaptic terms in the FBO and in both Day 90 and 150 ALI‐COs included both pre‐ and postsynaptic terms (Figure 2, Figure S2F,J). Human cortex P12,000 fraction revealed synaptic and ion channel‐related enrichment (Figure 2, Figure S2V). P22,000 fractions also showed substantial enrichment in synaptic terms; however, this was accompanied by significant enrichment in endomembrane‐related terms (e.g., endosome, ER, and Golgi). Mitochondrial terms were absent. Cytosolic fractions were enriched in cytosolic and nuclear proteins, consistent with partial nuclear disruption during homogenization.

Overall, the GO enrichment patterns were highly consistent across the tissue types investigated, suggesting similar organelle distributions between organoids and human/mouse brain tissue. This was supported by Western blot analysis of the mitochondrial marker TOMM20, which confirmed declining mitochondrial distribution across fractions from FBO and human cortex (Figure S3A).

3.3. Enrichment of Synaptic Proteins by Differential Centrifugation

To delve into the GO enrichment patterns, we compared synaptic protein enrichment in organoid fractions to adult mouse brain tissue (W8) processed with traditional Percoll density gradients. We focused on a panel of 30 synaptic proteins (Table S1), quantified across all datasets (25 detected in ALI‐CO fractions, see detailed lists in Table S12). Their abundance in the pellet fractions (P5,000, P12,000, and P22,000) was compared to the corresponding total homogenate sample levels, using the corresponding Percoll gradient method fractions (F3 and F4) as positive controls (Figure 3A).

FIGURE 3.

(A) Changes in the levels of 30 selected synaptic markers in the differential centrifugation fractions compared to the corresponding total homogenate samples from each tissue type (n = 25–30 synaptic marker proteins per tissue type based on the average Log2FC of 3–4 technical replicates). An enrichment of synaptosomes from adult mouse brain (W8) using the conventional Percoll density gradient centrifugation method was performed in parallel to compare the synaptic protein enrichment levels. The conventional Percoll F3 and F4 synaptosome enriched fractions are shown. **p < 0.001; ***p < 0.0001. Error bars represent the mean ± SD. (B) Heatmaps of the normalized abundances of the same 30 synaptic markers as used in A for each fraction and tissue or enrichment protocol (n = 25–30 markers detected per dataset). P1: Pellet from the first centrifugation, containing mainly whole cells, cell debris and nuclei to serve as a negative control. (C) Western blots of the synaptic marker SYP in differential centrifugation fractions from human forebrain organoids and human cortical tissue, at same scale to enable direct comparison between levels in fractions from FBOs and human cortex.

Consistent enrichment of synaptic proteins was observed in all pellets (P5,000, P12,000, and P22,000) from all tissue types. The highest enrichment was observed in P12,000 and P22,000 fractions from organoids and newborn mouse brain, with fold changes (35%–65% enrichment) comparable to Percoll‐based synaptosome fractions (40%–50% enrichment, see Table S2 for details). P22,000 fractions generally showed a higher variation. The adult brain samples (human and mouse W8) showed weaker or no enrichment in the P22,000 fractions, likely due to developmental differences in synaptic structure and composition affecting synaptosome density. Synaptic protein levels in the cytosolic fractions were depleted, confirming successful pelleting of synaptic compartments.

A heatmap of individual protein abundance (Figure 3B) further highlighted a trend of higher postsynaptic protein enrichment in P22,000 fractions from organoids and newborn mouse (P2), while presynaptic proteins were more prominent in P5,000 and P12,000 fractions. For example, the presynaptic proteins Protein piccolo (PCLO), Bassoon (BSN), and SYN1 showed significant enrichment of 2.3, 1.7, and 1.5 FC respectively in the ALI‐CO Day 150 P12,000 fraction, while postsynaptic proteins like Glutamate receptor 1 (GRIA1) and Metabotropic glutamate receptor 2 (GRM2) showed significant enrichment of 1.5 and 1.7 FC respectively in the P22,000 fraction (Table S12). This distribution pattern of pre‐ and postsynaptic proteins was not conserved in fractions from human cortex and adult mouse brain (Figure 3B, Figure S3B), possibly reflecting developmental differences between the young and adult stages, or the composition and density of the applied tissues (whole brain vs. organoids).

Western blot analysis of the presynaptic marker SYP supported the mass spectrometry data, showing increased levels in P5,000 and P12,000 fractions from both FBOs and human cortex (Figure 3C, Figure S3C,D). As expected, the more mature human cortex samples exhibited higher SYP levels than FBOs.

3.4. Enrichment of Growth Cone Markers

Given the early developmental stage of NOs, they are expected to contain immature synapses and actively growing axons tipped by growth cones that undergo structural and molecular changes that lead to the establishment of functional synaptic connections when they reach their final target. To assess whether the differential centrifugation protocol also enriches growth cone structures, we analyzed the distribution of 28 growth cone marker proteins (Table S3, detailed data in Table S13), curated from two proteomic studies (Estrada‐Bernal et al. 2012; Nozumi et al. 2009).

In young tissue types (mouse P2, FBOs, ALI‐COs Day 90 and 150), growth cone proteins were significantly enriched in all pellet fractions (P5,000, P12,000, and P22,000) and depleted in the cytosolic fractions (Figure 4A, see details in Table S4). The highest overall enrichment was observed in the P12,000 fractions of organoids (41%–47% increase), while in mouse P2, P22,000 showed higher average enrichment (42% increase) than P12,000 (26% increase); however, with greater variability.

FIGURE 4.

(A) Levels of selected growth cone markers in the fractions compared to homogenate (organoids and newborn mouse (P2), n = 27–29 detected markers per dataset, average Log2FCs of 3–4 technical replicates). The proteins were selected based on two proteomic studies determining markers of growth cones (Estrada‐Bernal et al. 2012; Nozumi et al. 2009). *p < 0.01, **p < 0.001, and ***p < 0.0001. Error bars represent the mean ± SD. (B) Abundance levels of the well‐known growth cone marker GAP43 in the homogenate and the four fractions derived from ALI‐CO Day 90 and 150, FBO and newborn mouse brain (P2). Stars in B indicate significant FC compared to the corresponding Homogenate level with *p < 0.05, **p < 0.01, and ***p < 0.0001. Error bars represent the mean ± standard error of the mean (SEM). Individual data points are shown.

Focusing on the canonical growth cone marker Neuromodulin (GAP43), we observed clear enrichment in the P12,000 fraction across all young tissues, with peak levels in the P22,000 fractions from ALI‐CO Day 90 and mouse P2 (Figure 4B). For example, GAP43 showed a significant FC of 2.8 in the P12,000 fraction from Day 90 ALI‐COs and 2.5 from Day 150 ALI‐COs. These findings indicate that the workflow not only enriches synaptic material but also captures developing growth cones, the precursor of synaptosomes.

3.5. TEM Evaluation of Synaptosome Structures

Apart from functional GO terms and protein enrichments, the morphology of the fractions was next investigated. Functional synaptosomes require intact, membrane‐enclosed structures capable of maintaining a membrane potential, depolarization, and vesicle release. To verify the structural integrity of the enriched particles, we examined the P5,000, P12,000, and P22,000 fractions from human cortex, adult mouse brain (W8), and ALI‐COs using transmission electron microscopy (TEM).

In adult brain samples (human and mouse), synaptosomes with clearly identifiable mitochondria and synaptic vesicles were readily observed in both the P5,000 and P12,000 fractions (human cortex in Figure 5A,B; mouse W8 and wider view of the same fields in Figure S4A,B,D,E). Some synaptosomes also showed electron‐dense active zones (AZs) with attached postsynaptic densities (PSDs) (Figure S4A,D).

FIGURE 5.

Transmission electron microscopy (TEM) images of differential centrifugation fractions from human cortex and ALI‐COs at Day 90 and 150. P5,000 and P12,000 fractions from human cortex (n = 1 biological replicate per tissue type distributed into the differential centrifugation fractions) (A, B) showing mature synaptosomes, and from ALI‐COs (C–F) showing the presence of growth cone particles (GCPs) at Day 90, and immature synaptosomes at Day 150. Scalebars: 500 nm. is, immature synaptosome; s, synaptosome; m, mitochondria; g, GCP, white arrows: synaptic vesicles.

ALI‐COs at Day 90 showed a few small synaptosome‐like structures in the P5,000 and P12,000 fractions together with larger membrane‐enclosed structures presenting with fibrous content, large vesicles, and mitochondria (Figure 5C,D, Figure S4F,G), consistent with previously described growth cone particles (GCPs) (Gordon‐Weeks and Lockerbie 1984; Grove et al. 1973; Pfenninger et al. 1983). This interpretation is supported by the observed enrichment of growth cone marker proteins, including GAP43 (Figure 4A,B).

By day 150, ALI‐CO P5,000 and P12,000 fractions contained intact synaptosomes with mitochondria and synaptic vesicles (Figure 5E,F). These vesicles were fewer in number compared to those in mature synapses, aligning with the known scarcity of synaptic vesicles during early synaptogenesis (Mozhayeva et al. 2002). Accordingly, we refer to these structures as immature synaptosomes.

Both the P5,000 and the P12,000 fractions contained non‐synaptic mitochondria (e.g., Figure S4A,D,F). Finally, the P22,000 fractions (Figure S4C,J) predominantly contained smaller membrane‐bound particles, in line with the proteomic GO Cellular Component enrichment analysis showing a large representation of endomembrane and general membrane‐associated terms in P22,000 fractions.

3.6. Phosphoproteomic Response to KCl Depolarization of Immature Synaptic Structures From ALI‐COs

If synaptosomes are sealed and maintain a membrane potential then they can be considered metabolically active, such that KCl stimulation will result in membrane depolarization and opening of voltage‐gated Ca²⁺ channels (VGCCs) allowing influx of Ca²⁺ ions. The resulting release of neurotransmitters by Ca²⁺ mediated exocytosis is then followed by ultra‐fast endocytosis of the synaptic vesicles (Silbern et al. 2021). These molecular functions are tightly controlled by protein phosphorylation and dephosphorylation via multiple kinases and phosphatases (Engholm‐Keller et al. 2019; Kohansal‐Nodehi et al. 2016; Silbern et al. 2021). To test whether the derived GCPs and immature synaptosomes were metabolically active and capable of stimulation‐induced exo‐/endocytosis, we stimulated the P5,000 and P12,000 fractions of the NOs, using a high concentration of KCl (76.2 mM). Stimulation of synaptic structures derived from the FBOs did not yield any significant changes in phosphorylation, likely due to too low an amount of material (data not shown). ALI‐COs generally yielded more material per tissue gram, though this tendency was not statistically significant (Figure S4K,L). Stimulation of enriched synaptic structures from the ALI‐COs at Day 90 and 150 resulted in significant regulation of phosphorylation in 219 proteins at Day 90 and 73 proteins at Day 150 (see full lists of detected and regulated phosphopeptides in Tables S14 and S15). Sixteen proteins showed stimulation‐dependent changes in phosphorylation at both timepoints. These comprised proteins important for exo−/endocytosis at the synapse, like Dynamin1 (DNM1), and PCLO, cytoskeletal proteins Doublecortin (DCX), Microtubule‐Associated Protein 1B (MAP1B), and the synaptic scaffold protein Disks large‐associated protein 4 (DLGAP4) (Figure 6A). A well‐described key indication of active endocytosis in synaptosomes is the dephosphorylation of S744/S778 phosphosites in DNM1, which allows scission and release of endocytosed vesicles from the membrane (Cousin and Robinson 2001; Graham et al. 2007). At both timepoints, DNM1 shows significant dephosphorylation of the doubly phosphorylated S774/S778 peptide, indicating active endocytosis (Figure 6B).

FIGURE 6.

(A) Venn diagram of proteins with significant KCl stimulation induced changes in phosphorylation (P5,000 and P12,000 fraction proteins combined) showing the overlapping proteins with regulation in both Day 90 and 150 ALI‐CO fractions. (B) DNM1 S774 or S774/S778 down‐regulation was detected at both timepoints (P5,000 fraction at Day 150; P12,000 fraction at Day 90). Error bars represent the mean ± standard error of the mean (SEM). Individual data points are shown. (C) GO molecular function (MF) or Cellular Compartment pathway enrichment among the proteins with stimulation‐dependent phospho‐regulation in ALI‐CO fractions from Day 90 and 150. (D) PPI STRING network (functional associations, incl. text mining) of growth cone proteins and their changes in phosphorylation at Day 90. (E) PPI STRING network (functional associations, incl. text mining) of proteins involved in synaptic vesicle cycling (extracted using the SynGO portal online tool) and their stimulation‐dependent changes in phosphorylation in Day 150 ALI‐CO fractions. (n = 3 technical replicates per condition (stim/ctl) and fraction).

GO enrichment analysis of proteins with stimulation‐dependent changes in phosphorylation showed a difference in stimulation profiles of Day 90 and 150. Stimulation at Day 90 primarily showed changes within cytoskeleton terms, while at Day 150, the proteins with phospho‐regulations were enriched in both cytoskeleton, neuron projection, and synapse/growth cone terms (Figure 6C). Characteristic for the regulation of phosphorylation in the Day 90 fractions, was a cluster of proteins specific for growth cones and important for growth cone motility (MARCKS (Myristoylated alanine‐rich C‐kinase substrate), MARCKSL1 (MARCKS‐related protein), PRKCE (Protein kinase C epsilon type) and GAP43, Figure 6D) (Gatlin et al. 2006; Nozumi et al. 2009). GAP43 also plays an important role in Ca²⁺ signaling at the synapse (Haruta et al. 1997). Furthermore, proteins involved in neurite outgrowth and axonal elongation showed changes in phosphorylation in Day 90 fractions including DCLK1 (Serine/threonine‐protein kinase DCLK1), SHTN1 (Shootin 1), MARK4 (MAP/microtubule affinity‐regulating kinase 4) and MAP1B. In synaptic structures from Day 150 ALI‐COs, consistent with the later developmental stage and presence of synaptic vesicles in the enriched structures, we observed stimulation‐dependent phosphorylation changes in proteins involved in the synaptic vesicle cycle including exo−/endocytosis proteins (e.g., PCLO, DNM1, SYN1, RPH3A (Rabphilin 3A) and CASK (Peripheral plasma membrane protein CASK), Figure 6E). Other proteins showing changes in phosphorylation were MAPT (Microtubule‐associated protein tau), CAMK4 (Calcium/calmodulin‐dependent protein kinase type IV) and DLGAP4 which all play a role in synaptic signaling and/or plasticity (Guo et al. 2017; Ho et al. 2000; Pavinato et al. 2023; Won et al. 2017) (Rasmussen et al. 2017).

Among the proteins with regulations in phosphorylation at both timepoints are also proteins associated with neurodevelopmental and psychiatric disorders such as autism spectrum disorders (SHANK1 (SH3 and multiple ankyrin repeat domains protein 1) (Gong and Wang 2015), DLGAP4 (Schob et al. 2019), and MAP1B (Liu et al. 2015)), intellectual disability (MAP1B (Liu et al. 2015; Walters et al. 2018), CAMK2G (Calcium/calmodulin‐dependent protein kinase type II subunit gamma) (Proietti Onori et al. 2018), DLGAP4 (Schob et al. 2019), and CASK (Moog and Kutsche 1993)), and schizophrenia (DCLK1 (Håvik et al. 2012), MARCKS (Pinner et al. 2014), PRKCE (Zhao et al. 2013), and SHANK1 (Fromer et al. 2014)).

To assess the capability of depolarization‐induced Ca²⁺ signaling and the level of maturity in the Day 90 and 150 ALI‐COs, we examined the presence of VGCCs in the stimulated fractions, estimated the ratio of excitatory to inhibitory synapses, and examined the presence of astrocytic markers in the ALI‐CO cultures. At both timepoints, VGCCs of types L, N, R, and T were detected, demonstrating the presence of the machinery for depolarization‐dependent Ca²⁺ influx. To estimate the distribution of excitatory and inhibitory synapses in the ALI‐COs used for enrichment and stimulation, we calculated the whole‐organoid proportions of glutamate to GABA receptors (abundance of all subtypes were summed). In the Day 90 ALI‐COs, the glutamate‐to‐GABA receptor ratio was 2:1 and in Day 150 ALI‐COs, 1.4:1, indicating an increase in the proportion of GABAergic synapses (see details in Tables S6 and S7). We detected the following astrocytic markers in Day 150 ALI‐COs: GFAP, S100B, ALDH1L1, AQP4, SLC1A2, SLC1A3, APOE, GJA1, and ALDOC. Of these, only GFAP, ALDOC, SLC1A2, SLC1A3, and APOE were detected in Day 90 ALI‐COs (Tables S6 and S7).

4. Discussion

Our study establishes a differential centrifugation protocol that enables isolation of functional GCPs and immature synaptosomes from human NOs, addressing critical limitations of traditional synaptosome enrichment methods for this kind of material. By integrating proteomic, ultrastructural, and functional analyses, we demonstrate that ALI‐COs recapitulate key stages of synaptic development: Day 90 fractions showing stimulation‐dependent changes in cytoskeletal regulators (e.g., MARCKS, MARCKSL1) and growth cone markers (GAP43), while Day 150 fractions transition toward immature synaptosomes with changes in phosphorylation in synaptic vesicle cycle proteins (e.g., SYN1, RPH3A) and the key ultra‐fast endocytosis protein DNM1. This temporal progression is consistent with the expected features of in vivo synaptogenesis, where axonal growth and pathfinding precede synaptic maturation, providing a human‐specific model to study synaptic development and dysfunction. The findings further reflect known cerebral organoid development where the cortical plate, including upper and deeper layer neurons, is formed and sparse spontaneous electrical activity is observed from Day 30 to 0, while further maturation and network formation, congruent with active synaptic transmission, develops from around Day 90 and onward (Fair et al. 2020; Mariani et al. 2015; Qian et al. 2016).

4.1. Methodological Advancements and Validation

The simplicity of the protocol, bypassing density gradients, and the need for only minimal tissue volume allows scalability across species and organoid models. Vesicle‐dense synaptosomes were successfully enriched from human cortical tissue in the P5,000 and P12,000 fractions as confirmed by TEM, demonstrating the effectiveness of the method even with limited tissue samples and relatively immature organoids.

Proteomic profiling revealed robust enrichment of synaptic proteins in P5,000 and P12,000 fractions (e.g., 2.3‐fold increase in PCLO in Day 150 ALI‐COs vs. homogenate), while non‐synaptic mitochondrial contamination was predominantly confined to P5,000 fractions (validated with TOMM20). The P22,000 fractions showed enrichment in postsynaptic proteins and endomembrane structures. Ultrastructural analysis by TEM confirmed membrane‐enclosed GCPs with fibrous content resembling actin networks, larger vesicles and mitochondria in Day 90 isolates, which is consistent with other studies on GCPs (Gordon‐Weeks and Lockerbie 1984; Grove et al. 1973; Pfenninger et al. 1983). As in previous GCP enrichments, the fractions likely also contain detached filopodia from growth cones (Pfenninger et al. 1983). There is also the possibility that a fraction of the isolated structures could be detached growth cone‐like structures from the leading process of migrating neurons (Miyata and Hayashi 2023; Nakajima et al. 2024). Day 150 fractions contained immature synaptosomes marked by sparse synaptic vesicles in the structures. The findings align with the expected developmental shifts in NOs over time.

4.2. Functional Maturation and Phosphoproteomic Signatures

Functional viability was validated through KCl‐induced depolarization of synaptic structures isolated from ALI‐COs. This triggered phosphorylation changes in growth cone markers (e.g., GAP43, MARCKS, MARCKSL1), cytoskeletal regulators (DCLK1, SHTN1, MARK4), and calcium‐dependent kinases (CAMK2G, PRKCE) in Day 90 GCPs. These phosphorylation dynamics align with established mechanisms of growth cone motility, in which Ca²⁺ influx modulates actin‐microtubule interactions (Henley and Poo 2004; Nelson et al. 2013; Schneider et al. 2023). In this process, VGCCs, which were detected in both D90 and D150 ALI‐CO fractions, play important roles for pathfinding and extension of the growth cone (Henley and Poo 2004). Rodent studies have demonstrated that growth cones and isolated GCPs can depolarize in response to stimulation, leading to Ca²⁺ transients and cytoskeleton remodeling, which is highly controlled by phosphorylation (Belardetti et al. 1986; Lipscombe et al. 1988; Lockerbie et al. 1991; Neely and Gesemann 1994). It is, for example, known that PRKCE‐dependent MARCKS phosphorylation is involved in growth cone adhesion and turning (Gatlin et al. 2006). This supports our interpretation that the Day 90 phosphorylation profile reflects a growth cone–dominated stimulation response. The presence of detached filopodia in the fractions is considered unlikely to contribute substantially to the stimulation profile, since they typically lack mitochondria (Smith and Gallo 2018). However, a possible fraction of growth cone‐like particles from the leading process of migrating neurons could potentially influence the results. The cytoskeletal machinery has been shown to differ between leading process growth cones and axonal growth cones (Miyata and Hayashi 2023), which makes the distinction important to consider in future studies.

In Day 150 fractions, depolarization induced regulation of phosphorylation in synaptic vesicle cycling‐related proteins (SYN1, DNM1, CASK, RPH3A, DMXL2). GO enrichment analysis showed enrichment in cytoskeleton, synaptic, and neuron projection related terms, indicating a more mature stimulation profile compared to Day 90 fractions. Importantly, DNM1 S774/S778 dephosphorylation, a strong indicator of active endocytosis and synaptic vesicle cycling (Cousin and Robinson 2001; Graham et al. 2007), was observed in both Day 90 and 150 fractions.

The estimated proportions of excitatory to inhibitory synapses based on receptor abundance in the ALI‐CO cultures (2:1 for Day 90 ALI‐COs and 1.4:1 at Day 150) are aligned with the expected increase in the proportion of interneurons over time (Fair et al. 2020). This change in proportion of glutamatergic to GABAergic synapses is unlikely to explain the shift in phosphorylation profile observed from Day 90 to 150, since the presynaptic transmission machinery is expected to be conserved between GABA and glutamatergic neurons. However, it is a clear indication of neuronal network maturation in the ALI‐COs, which aligns with the more mature phosphorylation profile at Day 150. A further indication of maturation in the organoids was the detection of mature astrocytic markers such as S100B, ALDH1L1, ALDOC, GJA1, and AQP4 only in Day 150 ALI‐COs (Jurga et al. 2021). Astrocytes are known to increase functional maturation of cortical neurons (Kuijlaars et al. 2016; Sloan et al. 2017), which further supports the interpretation of a change in maturity causing the shift in phosphorylation profile.

The functional changes in protein phosphorylation in KCl stimulated synaptic structures from ALI‐COs, which was not possible to show in FBO isolates, could indicate a higher level of maturity in the ALI cultures, which have previously been shown (Giandomenico et al. 2021). A higher number of synapses/growth cones and/or more mature, spine‐like morphology could result in a better quality of the enriched structures. The lack of significant changes in phosphorylation upon KCl stimulation of structures from FBOs could, however, also be due to too low an amount of material.

4.3. Implications for Disease Modeling and Functional Studies

The protocol offers value beyond structural enrichment. Phosphoproteomic profiling of Day 150 ALI‐COs identified phosphorylation changes in neurodevelopmental disorder‐associated proteins (e.g., SHANK1, DLGAP4 and CAMK2G), offering a platform to study synaptic dysfunction in conditions like autism spectrum disorder, intellectual disability and schizophrenia. For instance, CAMK2G is a protein kinase implicated in intellectual disability (Proietti Onori et al. 2018), which here showed depolarization‐responsive phosphorylation, suggesting its role in activity‐dependent synaptic maturation. Similarly, the stimulation‐dependent regulation of phosphorylation in MARCKS and MARCKSL1, cytoskeletal modulators linked to neuronal migration defects (El Amri et al. 2018), underscores the method's potential to probe early pathogenic mechanisms.

The method's compatibility with patient‐derived organoids enables direct investigation of disease‐relevant synaptic phenotypes in a human context. For example, the method could be used to study how FMR1 loss in Fragile‐X syndrome causes reduced synaptogenesis (Hanson and Madison 2007), or to explore mechanisms underlying altered synaptogenesis observed in schizophrenia models, where human iPSC‐derived neurons exhibit fewer dendritic spines and immature presynaptic terminals compared to controls (Brennand et al. 2011; Habela et al. 2016). In parallel, CRISPR/CAS approaches can be used to model specific gene perturbations or SNPs, for example, in synaptic adhesion molecules like SHANK3, NLGN3/4 and NRXN1 that disrupt presynaptic assembly in autism spectrum disorders (Südhof 2008). The depolarization‐dependent phosphoproteomic workflow further allows examination of signaling abnormalities linked to VGCCs such as CACNA1C, which is a major risk locus for psychiatric diseases like schizophrenia and bipolar disorder (Harrison et al. 2022; Ripke et al. 2013) and the causal gene in Timothy Syndrome (Splawski et al. 2004). The method could also provide important functional insights into synaptic vesicle cycle proteins involved in monogenic epileptic disorders such as SYN1, STXBP1, and DNM1 (Jiang et al. 2025). Furthermore, the ability to isolate metabolically active GCPs permits investigations into axon guidance responses (e.g., semaphorin and netrin signaling), which are important for proper network formation (Kolodkin and Tessier‐Lavigne 2011). Exploiting the in vitro setting of iPSC‐derived organoids, the method bypasses ethical and logistical constraints of postmortem or primary human tissue and allows systematic dissection of human synaptic gene function in health and disease settings.

4.4. Limitations and Future Directions

In addition to developmental stage differences, a key distinction between brain organoids and native brain tissue lies in their structural and cellular composition. Mature brain tissue is densely packed and includes abundant white matter, comprising myelinated axons and oligodendrocytes, as well as astrocytes, microglia, and vasculature. These all contribute to tissue density, biochemical complexity, and centrifugation behavior. Most neural organoids, in contrast, lack myelination and supporting glial populations, particularly oligodendrocytes and mature astrocytes, which emerge only under prolonged culture or directed differentiation protocols (Paşca et al. 2015; Madhavan et al. 2018). The absence of white matter tracts and compact axon bundles likely reduces physical resistance during homogenization and may shift the sedimentation profiles of synaptic components, particularly in differential centrifugation protocols. The reduced extracellular matrix and lower tissue stiffness of organoids compared to intact brain (Lancaster et al. 2013; Quadrato et al. 2017) may lead to more efficient extraction of membrane‐bound structures but may also influence organelle integrity. These structural differences should be considered when comparing subcellular enrichment yields and interpreting the spatial distribution of synaptic and cytoskeletal proteins.

The relatively limited number of depolarization‐responsive phosphorylation events correlates with the low presence of GCPs and immature synaptosomes detected by electron microscopy, suggesting that only a small fraction of physiologically sealed, metabolically active synaptosomes are present. This may reflect limited structural maturity, suboptimal purification, or biophysical differences in lipid or protein composition between NOs and native brain tissue. While extra‐synaptosomal mitochondrial contamination in P5,000 fractions remains a challenge, the P12,000 fraction provides the purest enrichment of functional synaptic structures for stimulation experiments. Further optimization of fractionation protocols or incorporation of immunoaffinity‐based purification strategies could help to further reduce non‐synaptic contaminants.

Additionally, extending organoid culture beyond Day 150 may yield more synaptosomes with mature synaptic vesicle pools, resulting in higher yield of functional active synaptosomes and thereby enabling studies of neurotransmitter release kinetics. Systematic evaluation of batch‐to‐batch organoid variability, regional patterning, and synapse subtype composition may be critical to increase reproducibility and translational relevance. Furthermore, variable yield of enriched synaptic structures between organoid models (e.g., spheric vs. ALI cultures) will affect the scalability of the method. These will be important aspects to address for effective scalability of the model, for example, drug screening. The protocol's reproducibility across human and murine tissues underscores its versatility. Future studies incorporating biological replication across independent organoid differentiations will enable quantitative assessment of inter‐sample variability and formal statistical testing of biological questions. Coupled with advances in spatial proteomics, this method could be deployed to map subcellular proteome dynamics during synaptic maturation or drug responses. For example, integration of synaptosome depolarization with single‐organoid real‐time calcium imaging could offer deeper insight into human‐specific signaling cascades underlying synaptic plasticity. Applying subsequent sorting of the isolated fractions based on neurotransmitter types, For example, Fluorescence Activated Synaptosome Sorting of GABAergic and glutamatergic synaptosomes (Biesemann et al. 2014), could allow studying the parallel development of inhibitory and excitatory synapses, highly relevant for neurodevelopmental disease modeling and neuronal network formation. Since this model enables access to early‐stage synaptic structures, it may also allow future investigation into upstream regulators of synaptogenesis, such as guidance cues, adhesion molecules, and cytoskeletal remodeling enzymes.

5. Conclusion

By integrating a differential centrifugation strategy with functional and proteomic profiling, we have established a practical framework for isolating active synaptic structures from human neural organoids. Th approach reveals temporally distinct molecular signatures that reflect developmental transitions in synapse formation and enables the identification of phosphorylation events linked to growth cone remodeling and synaptic vesicle cycling. Beyond enriching specific compartments, the protocol facilitates future mechanistic interrogation of synaptic signaling, with the potential to be adapted for high‐throughput studies and for patient‐specific disease models. As cerebral organoid technology matures, this platform is anticipated to provide a new opportunity to explore human synaptic development and dysfunction in an experimentally accessible, physiologically responsive system.

Author Contributions

Marie S. Øhlenschlæger: investigation, writing – original draft, methodology, validation, visualization, formal analysis, data curation. Lucrezia Criscuolo: investigation, validation, writing – original draft, methodology. Pia Jensen: project administration, writing – review and editing, validation, investigation, methodology, formal analysis, writing – original draft. Daniel J. Lloyd‐Davies Sánchez: investigation, writing – original draft. Magdalena Sutcliffe: investigation, writing – review and editing. Santosh Bhosale: investigation. Helle Bogetofte: investigation. Muhammad Tahir: investigation. Lene A. Jakobsen: investigation. Maria Pihl: investigation. Jonathan Brewer: resources. Veit Schwämmle: formal analysis. Frantz R. Poulsen: resources. Kristine Freude: resources. Madeline A. Lancaster: resources. Phillip J. Robinson: methodology, writing – review and editing, writing – original draft, supervision. Martin R. Larsen: supervision, conceptualization, funding acquisition, writing – review and editing, methodology, project administration, investigation, formal analysis.

Funding

This work was supported by Danmarks Frie Forskningsfond, 9040‐00381B. Lundbeck Foundation, R336+2020‐1113. Novo Nordisk Fonden, NNF18SA0032928. Danish Agency of Higher Education and Science, 5229‐00012B. National Health & Medical Research Council Australia, GNT1069493, GNT1052494, GNT1047070.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: (A) Light microscopic images of IMR90‐4 iPSCs and forebrain organoids (FBOs) or ALI‐COs at different differentiation stages. (B–D) Immunohistochemical (IHC) labelling of synapses in FBOs (B, C) at Day 100 and ALI‐COs (D) at Day 100. (B) FBO Day 100 labelling of presynaptic markers Synaptophysin (SYP) and Vesicular glutamate transporter 1 (VGLUT1) and postsynaptic marker Postsynaptic density protein 95 (PSD95) at Day 99. Scalebar: 5 μm. (C) FBO day 100 labelling of presynaptic marker Synapsin 1 (SYN1), postsynaptic marker PSD95 and dendritic marker Microtubule associated protein 2 (MAP2) at Day 100. Scalebar: 20 μm. (D) ALI‐COs labeled at Day 100 for same markers as in C (scale bar: 10 μm). White arrows pointing at areas with co‐localization or adjacent expression of the pre‐ and postsynaptic markers. For all IHC images, n = 3 organoids.

Figure S2: A–X Gene Ontology (GO) Cellular Component enriched terms (top 20, FDR ≤ 0.05) based on fraction specific proteins (FC ≥ 1.5 compared to homogenate, FDR ≤ 0.01). P5,000, P12,000, P22,000, and Cytosolic fractions from differential centrifugation of forebrain organoids (Day 100), ALI‐COs Day 90, ALI‐COs day 150, newborn mouse brain tissue (postnatal Day 2 [P2]), adult mouse brain tissue (Week 8 [W8]) and surgery derived adult human cortical tissue. Data were based on n = 3–4 technical replicates per tissue type.

Figure S3: (A) Full western blot of the mitochondrial marker, TOMM20, and β‐actin, in differential centrifugation fractions from forebrain organoids and human cortex in n = 3 technical replicates of the differential centrifugation. (B) Heatmap of the normalized abundances of n = 30 selected synaptic markers for each fraction from the differential centrifugation workflow applied to adult mouse (W8) brain tissue. (C, D) Full western blots from Figure 3C, of the synaptic marker, Synaptophysin, and β actin, in differential centrifugation fractions from (C) forebrain organoids (incl. standard protein ladder) and (D) human cortex in n = 2 technical replicates of the differential centrifugation, with protein ladders shown. All blots were cut before antibody incubation.

Figure S4: (A–J) TEM images of differential centrifugation fractions from adult mouse, adult human cortex and ALI‐COs day 90 and 150 (n = 1 biological replicate per tissue type distributed into the differential centrifugation fractions). (A–C) Adult mouse fractions P5,000, P12,000 and P22,000. Synaptosomes (s) are clearly present in the P5,000 and P12,000 fractions. The P22,000 fraction contains mainly smaller undefined membrane structures. (D, E) Human cortex fractions P5,000, P12,000. Synaptosomes are clearly visible in both fractions. (F, G) Day 90 ALI‐CO fractions P5,000, P12,000. A few synaptosomes or immature synaptosomes (is) could be seen, however, many structures resembled growth cone particles (g). (H–J) Day 150 ALI‐CO fractions P5,000, P12,000, and P22,000. Immature synaptosomes are seen in the P5,000 and P12,000 fractions with the presence of mitochondria and some synaptic vesicles. Scalebars (A–J): 500 nm. s, synaptosome; is, immature synaptosome; m, mitochondria; g, growth cone particle (GCP); white arrows, postsynaptic density attached to the synaptosome. (K, L) Comparison of yield from the differential centrifugation between ALI‐CO (slice‐culture) and FBO (spherical culture) organoids (n = 3–4 organoid batches). Error bars represent the mean ± SD. (K) Total protein amount (μg) in the first supernatant (S1) used for differential centrifugation, per total weight of the used organoid tissue and (L) per number of organoids used to begin with. The number of ALI‐COs is corrected assuming that each initial cerebral organoid generates at least 4 ALI‐CO slices. Error bars represent the mean ± SD.

Table S1: Selected synaptic proteins that are mainly expressed in the synapse.

Table S2: Results from Wilcoxon Rank Sum tests of synaptic marker protein levels for each fraction compared to homogenate samples (n = 25–30). FC, fold change; FBO, forebrain organoid; ALI‐CO, air‐liquid‐interface cerebral organoid; D, Day; P2, postnatal Day 2; W8, 8‐week‐old.

Table S3: Selected synaptic proteins that are specific for growth cones or growth cone particles (GCPs).

Table S4: Results from Wilcoxon Rank Sum tests of growth cone marker protein levels for each fraction compared to homogenate samples (n = 27–29). FC, fold change; FBO, forebrain organoid; ALI‐CO, air‐liquid‐interface cerebral organoid; D, day; P2, postnatal day 2.

Table S5:

Table S6:

Table S7:

Table S8:

Table S9:

Table S10:

Table S11:

Table S12:

Table S13:

Table S14:

Table S15:

Øhlenschlæger, M. S. , Criscuolo L., Jensen P., et al. 2026. “Modeling Synaptic Maturation From Growth Cone to Synapse in Human Organoids.” Journal of Neurochemistry 170, no. 5: e70458. 10.1111/jnc.70458.

Data Availability Statement

The data that supports the findings of this study are available in the Supporting Information of this article. The mass spectrometry proteomic and phosphoproteomic data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD072589.