A programmable peptide interface for on-demand neural culturing platforms

Abstract

The precise spatial organization of neural cells into two-dimensional networks or three-dimensional spheroids is crucial for advancing neuroscience research and drug discoveries, yet remains challenging with conventional, single-function coatings. Here, we propose a programmable bifunctional peptide that integrates a silica-binding domain with a tunable cell-adhesive Arginine-Glycine-Aspartate (RGD) tripeptide. By systematically improving the RGD variant and linker rigidity, we introduced a single coating material that enables on-demand switching between two distinct functions: guiding the patterned growth of functional neural circuits on glass and facilitating the high-throughput formation of uniform neural spheroids. The latter exhibited high viability, extensive neurite outgrowth, and spontaneous electrophysiological activity, which validates their functional maturity. We establish by this work a versatile and reliable platform for advanced neural interface research, with significant potential for drug discovery and disease modeling.

Graphical abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12951-026-04032-x.

Introduction

The study of neural cells, encompassing neurons and glial cells, is fundamental to advancing our understanding of the brain and its neural circuits’ function, as well as the pathogenesis of neurological disorders [1, 2]. In vitro cell culture models serve as indispensable tools for such investigations, enabling controlled manipulation and high-resolution analysis that are often impossible to achieve in vivo [3–5]. A critical prerequisite for effective neural cell culture is the substrate interface, which must be prepared to mimic key aspects of the extracellular matrix (ECM) to support cell adhesion, survival, and function [6]. To this end, a variety of coating materials have been employed to functionalize substrates such as glass and plastic for neural cell culture. Traditional methods include the use of poly-D-lysine (PDL), poly-L-ornithine (PLO), laminin, and Matrigel, which promote general cell adhesion through charge interactions or integrin binding [7, 8]. While effective for creating homogeneous cell layers, these conventional coatings offer limited control over the precise spatial organization of cells. For advanced applications such as constructing in vitro models of neural networks, drug screening platforms, or bio-hybrid devices, the ability to pattern neural cells with high specificity and resolution is paramount [9, 10]. This spatial control allows for the directed formation of synapses and the reproduction of specific neural connectivity patterns, which is crucial for studying information processing in defined circuits [11].

Micropatterning techniques developed very fast in recent years [12], and some of them has been successfully applied to cell culture [13]. These include inkjet or microfluidic deposition for additive printing [14, 15], and microcontact printing (µCP) or photolithography for defining adhesive regions on pre-coated substrates [16–18], all of which allow controlled placement of adhesion cues or cells with micrometer-scale resolution for two-dimensional (2D) networks. However, each method carries inherent constraints: for instance, µCP suffers from stamp deformation and inconsistent protein transfer [19], while inkjet deposition may subject cells to shear stress and struggles with large-scale uniformity [20]. More fundamentally, these approaches are optimized for strong, fixed cell adhesion and are therefore not readily adaptable for applications requiring weak or tunable adhesion to guide subsequent three-dimensional (3D) morphogenesis.

Beyond 2D patterning, there is a growing demand for efficient and reproducible methods to generate 3D neural aggregates, such as neural spheroids or cerebral organoids [21, 22]. These 3D models better recapitulate the complexity and cell-cell interactions of native neural tissue, making them highly valuable for developmental studies, toxicity testing, and disease modeling [23]. Meanwhile, 3D neural cultures are increasingly recognized for their complexity and resemblance to in vivo neural microenvironments, while placing high demands on high throughput, controlled attrition rates, and high efficiency [24]. The reported techniques for neural spheroid culture can be categorized into scaffold-free and scaffolding methods based on whether an artificial ECM space is used [25]. In contrast to scaffold-based cultures (scaffolds made of Matrigels, synthetic polymers, etc.), neural spheroids formed by self-assembly generate their own ECM and therefore behave close to living cells [26]. Current scaffold-free techniques for generating neural spheroids, such as ultra-low-attachment (ULA) plates and hanging-drop cultures, often rely on non-adhesive surfaces or suspension culture, which can lead to heterogeneous aggregate sizes and shapes, complicating downstream analysis and reproducibility [27–29]. Although the widely used ULA approach forces neural cells and astrocytes to aggregate while defining culture dimensions using micro-patterned techniques, problems such as spheroid loss upon media change, low yield, and high time cost remain [27]. Thus, there is a clear need for alternative strategies to streamline workflows and offer improved reliability and control over the 3D assembly process.

Interestingly, the natural tendency of neural cells to detach and aggregate on glass surfaces, often viewed as a problem in conventional culture, may offer an unexpected opportunity. On glass surfaces, even with ECM coatings, neural cells often fail to maintain a monolayer and spontaneously form aggregates within a few days [30]. Inspired by this, we hypothesized that engineering a coating with uniform and moderate cell-binding capacity on glass could harness this propensity to facilitate the formation of size-controlled neural spheroids.

ECM-derived peptides have been widely applied to fabricate biocompatible materials for long-term neural cell culture and directed differentiation [31]. Although their integrin-binding activity is weaker than that of full-length ECM proteins, they provide advantages such as lower lot-to-lot variation and reduced risk of exogenous immunogens or pathogens [32] making them ideal for fusion constructs. Despite these advancements, the field faces a significant bottleneck: the lack of a versatile, tunable coating material that can seamlessly address both needs. Typically, materials are optimized for either strong, specific adhesion (suited for 2D patterning) or for promoting detachment and 3D aggregation, but not both. Researchers are often forced to use different, proprietary coatings for different experiments, increasing cost and experimental variability. Therefore, a single-platform solution whose interfacial properties can be precisely modulated would represent a substantial technological leap.

To bridge this gap, we have designed and synthesized a novel multifunctional peptide-based material. This peptide is achieved with two distinct functional domains: a glass-binding domain at one terminal, ensuring stable anchorage to the substrate surface, and a tunable cell-interactive domain at the opposite terminal. The key innovation lies in the programmability of the neural cell-binding motif. By adjusting the sequence and properties of this domain, we can precisely control the strength of cell-material interaction. In this study, we developed a Programmable Neural Anchor (PNA), which employed a validated silica-binding tag (Si-tag) [33] to facilitate ECM-derived peptide binding to the glass surface in a more efficient and uniform manner and maintained exposure of the C-terminal end of the ECM-derived peptide to promote neural cell adhesion [34]. We present the rational design, synthesis, and comprehensive characterization of this bifunctional peptide material. We demonstrate its efficacy in (i) creating high-fidelity patterns of primary neurons that form functional synapses, and (ii) serving as a standardized substrate for the reliable generation of uniform neural spheroids. This tunable platform technology not only addresses current limitations in neural cell culture but also opens new avenues for fundamental neurobiology research and therapeutic development.

Materials and methods

Synthesis and characterization of the peptide

The built coding sequence of PNAs were synthesized and cloned into pET28a plasmid between the NdeI sites and XhoI site (GenScript, Nanjing, China). Then, the PNA-contained plasmids were transformed into BL21(DE3) cells (Vazyme, Nanjing, China), respectively. For each type of plasmid, a single colony was seeded in 4 mL LB liquid medium (contains 10 µg/mL kanamycin) at 37 °C, 200 rpm (~ 0.6 OD600), and 1 mL of the culture was transformed into 100 mL fresh kanamycin-contained LB liquid medium for subculturing (~ 0.6 OD600), following an IPTG induction (0.05 mM) for 16 h at 15 °C, 180 rpm. Then, cells were harvested at 4,000 rpm and 4 °C for 20 min and washed twice with PBS buffer. Finally, the PNA proteins were purified through His-tag Protein Purification Kit (Beyotime Biotech, Shanghai, China) and analyzed using 15% SDS-PAGE gel.

Fabrication and functionalization of patterned substrates

Patterned substrates were fabricated on quartz glass coverslips using standard photolithography. The substrates were cleaned and spin-coated with AZ 1518 photoresist at 4000 rpm for 60 s, followed by soft-baking at 95 °C for 1 min, obtaining 1 μm thickness of photoresist. The photoresist was then exposed to UV light through a patterned photomask with an exposure dose of 90 mJ/cm², and processed in AZ 300 MIF developer for 40 s to reveal the desired patterns.

The patterned substrates were subsequently functionalized by incubation in PNA solutions with 10 to 100 µg/mL in PBS buffer for 20 min at room temperature. This step allowed for the specific adsorption of the peptide onto the exposed glass regions, while the photoresist-covered areas remained uncoated. The substrates were then gently rinsed with PBS buffer and F12 medium prior to cell culture.

Neural cell culture and differentiation

Human induced pluripotent stem cells (hiPSCs) were employed to generate functional human neurons through a multi-stage differentiation protocol sourced from STEMCELL Technologies (Shanghai, China). The differentiation process commenced with neural lineage specification: hiPSCs were induced to neural progenitor cells (NPCs) utilizing the SMAD inhibitor-based STEMdiff™ Neural Induction Kit (Catalog #08581). Subsequent neurogenic differentiation was achieved by culturing NPCs in STEMdiff™ Forebrain Neuron Differentiation Basal Medium (Catalog #08601), which supports regional specification toward forebrain neuronal subtypes. To promote synaptic maturation and electrophysiological functionality, differentiated neural cells were maintained in STEMdiff™ Neuron Maturation Kit (Catalog #08510)-supplemented media. The hiPSC line utilized for these experiments was obtained from Westlake University Cell Bank.

Immunostaining of neurons

Neuronal cultures were subjected to a standardized immunocytochemistry protocol to visualize cytoskeletal and nuclear components. Cell fixation was performed using 4% paraformaldehyde (PFA, Yuanye Biotechnology, Shanghai) for 10 min at room temperature, followed by membrane permeabilization with 0.3% Triton X-100 in phosphate-buffered saline (PBS) for 30 min to enable antibody penetration. To minimize nonspecific antibody binding, samples were incubated overnight at 4 °C in a blocking solution composed of 3% (w/v) bovine serum albumin (BSA) and 0.1% Triton X-100 in PBS. Primary antibodies—including anti-βIII-tubulin (TUJ1, 1:1,000; Abcam, UK) targeting neuronal microtubules and anti-TAU (1:1,000; Cell Signaling Technology, USA) recognizing microtubule-associated proteins—were diluted in blocking solution and applied for 2 h at ambient temperature. Species-specific secondary antibodies conjugated to fluorophores were subsequently incubated for 1 h under identical conditions. Nuclear counterstaining was performed using 4′,6-diamidino-2-phenylindole (DAPI, 1:500; Invitrogen, USA). Coverslips were mounted with Fluor mount Aqueous Mounting Medium to preserve fluorescence integrity. Mature neuronal phenotypes were analyzed via fluorescence microscopy, with dual labeling confirming cytoskeletal maturation and nuclear morphology.

Scanning electron microscope (SEM) imaging

For SEM analysis, cultured neurons were first immobilized with a fixative solution comprising 2.5% glutaraldehyde and 2% PFA at 4 °C for 1 h. Following primary fixation, samples were rinsed thoroughly with 0.1 M phosphate buffer (PB, pH 7.3) to remove any residual fixative. A secondary fixation was then performed by treating the cells with 1% osmium tetroxide in 0.1 M PB on ice for an additional hour. After this step, samples were washed gently with deionized water three times. Dehydration was carried out using a graded ethanol series, starting from 30% and progressing to 95%, and concluding with three exchanges of absolute ethanol. The dehydrated specimens were then subjected to critical point drying to preserve ultrastructure by replacing the ethanol with a transitional fluid under controlled temperature and pressure. Finally, the dried samples were sputter-coated with a thin, approximately 5-nm-thick conductive layer of gold to mitigate charging effects and enhance secondary electron emission during SEM imaging.

High-throughput generation and characterization of neural spheroids

The PNA-based substrate was prepared by dissolving purified PNA proteins in serum-free culture medium at predetermined concentrations (10, 50, and 100 ug/mL). The solution was subsequently introduced into sterile glass-bottom culture dishes and subjected to a two-hour incubation period at room temperature (25 °C) under light-protected conditions to facilitate optimal surface conjugation. 1 mL of the PNA solution was applied to a 20-mm-diameter glass-bottom dish to ensure complete coverage of the glass surface. The volume of the solution could be adjusted according to the coated area to maintain appropriate culture conditions. Following this immobilization phase, three PBS washes were gently performed to eliminate non-specific bound peptides while maintaining sterility conditions. Primary neural cells induced from iPSC were enzymatically dissociated and resuspended in neural maturation medium. Cell suspensions were seeded at optimized densities (5 × 10⁵ cells/cm²) onto peptide-functionalized substrates. Cultures were maintained in a humidified incubator (37 °C, 5% CO₂) with medium replacement every 72 h. Spheroid formation was monitored daily using phase-contrast microscopy, with mature 3D aggregates typically emerging within 5–7 days post-seeding.

Neural spheroid formation was quantitatively assessed from three independent biological replicates (n = 6). For each condition, at least 3 random fields of view (2.5 mm × 2.0 mm) were imaged per replicate. All spheroids within these fields were manually counted, and their diameters were measured using ImageJ. Data for total spheroid number and diameter distribution are presented as mean ± standard deviation (SD) across the three replicates.

Functional neural activity recording

Neural activity was monitored and quantified using a commercial recording system (Axion Biosystems). Neural spheroids were transferred onto custom-designed 24-well multielectrode array (MEA) plates, each well equipped with 16 embedded electrodes. The MEA plates were connected to the recording platform for continuous extracellular signal acquisition. Signals were recorded in neural spike mode with a gain of 1,000 times to optimize signal-to-noise ratio. Spike detection and burst analysis were performed through integrated AxIS software. Spike events were identified when signal amplitude exceeded a threshold set at 6 standard deviations above the mean background noise level. Bursts were defined as episodes containing at least five successive spikes with inter-spike intervals not exceeding 100 ms. This setup enabled stable and reproducible electrophysiological recording of functional activity from the neural spheroids.

Results

Synthesis and characterization of the peptide

To create a versatile interface for neural cell manipulation, we propose a recombinant protein by integrating two distinct functional domains. The first domain is a high-affinity Silica-binding tag (Si-tag) for robust attachment to silica-based substrates such as glass [33]. Another is a multimer containing ECM-derived peptide [35], RGD motif variant, responsible for anchoring cells. Three candidate RGD peptides (from RGD-1 to RGD-3) with relatively high cell affinity were chosen for testing and comparison based on previous studies (Table 1). To ensure the robust fusion of two bio-bricks and the optimal presentation of ECM-derived peptide, a linker was inserted between Si-tag and RGD [36]. Subsequently, the PNA proteins were produced through the Escherichia coli (E.Coli) expression system and analyzed by SDS-PAGE, then purified with the help of their 3’-end His₆-tag.

Table 1.

Sequence information of RGD peptides

Multimer	Sequence	Journal	Year	References
RGD-1	[GSGAGGRGDGGYGSGSSS]4	Polymers	2018	[35]
RGD-2	[NGEPRGDTYRAY]4	Biomaterials	2010	[37]
RGD-3	[KGGPQVTRGDVFTMP]4	Acta Biomaterials	2017	[31]

Firstly, we designed an PNA construct by fusing a Si-tag with an RGD multimer via a classical flexible linker (EGKSSGSGSESKST). To further validate the anchoring capability of both functional domains, a GFP coding sequence was inserted between the Si-tag and RGD domain, yielding the recombinant protein PNA-GFP with a molecular weight of approximately 51.2 kDa (Fig. 1A–B). Neurons were collected and resuspended by culture medium, adding 100 ug/mL PNA-GFP, followed by 1 h incubation in a humidified incubator. When mixed with suspended neural cells, PNA-GFP fluorescence was clearly detected on the cell surface under the fluorescence microscope (Fig. 1C), confirming its ability to bind directly to cells. In addition, when applied to a patterned glass substrate, PNA-GFP fluorescence was observed exclusively on the exposed glass regions, while the photoresist-covered areas remained nonfluorescent (Fig. 1D), demonstrating its selective immobilization on silica-based surfaces.

Fig. 1.

Fabrication and characterization of the peptides: (A) The layout of coding sequences for the Programmable Neural Anchor (PNA). (a) The coding sequence of PNA proteins; (b) The coding sequence of GFP-labeled PNA protein. (B) The expressed PNA-GFP verified by SDS-PAGE; (C-D) Validation of PNA’s ability to anchor neural cells and glass surface. MW: molecular weight marker; PNA-GFP; PNA protein featuring a GFP motif inside

With the independent functionality of each domain confirmed, we proceeded to optimize the overall structure of the final PNA construct. We hypothesized that the orientation of the RGD multimer relative to the surface-bound Si-tag could critically influence its accessibility for cell binding. To test this, we generated two constructs: PNA-N, with the RGD multimer fused to the N-terminus of the Si-tag, and PNA-C, with the fusion at the C-terminus, both via the flexible linker (Fig. 2A). Both constructs, with a molecular weight of approximately 19 kDa, were efficiently expressed and purified (Fig. 2B). A quantitative fiber-optic binding assay, as previously established [33], verified that both PNA-N and PNA-C retained comparable and high binding affinity for silica substrates (Fig. 2C).

Fig. 2.

Characterization of neural cells PNA constructs. (A) Schematic layout of coding sequences for PNA-C and PNA-N. (B) Expression of PNA-C and PNA-N confirmed by SDS-PAGE. (C) Binding affinities of PNA-C and PNA-N to silica substrates validated by a previously reported fiber-optic method [33]. Data are presented as mean ± SD from three independent biological replicates (n = 3). (D) Immunostaining of iPSC-derived neural cells. (E) Effects of N- and C-terminal RGD fusion on neural cell adhesion and growth. (F) Example of software auto label for cell-body and neurites. (G) Average neurite length per cell-body on different coatings changes with time. Laminin: laminin-coated glass; PNA-C: PNA construct with C-terminal RGD multimer; PNA-N: PNA construct with N-terminal RGD multimer

The biological efficacy of the two configurations was then assessed using mature neurons differentiated from induced pluripotent stem cells (iPSCs). The neuronal identity and maturity of these cells were confirmed by immunocytochemistry, showing positive staining for canonical neural markers (Fig. 2D). When seeded onto glass substrates coated with either PNA-N or PNA-C, the neurons exhibited markedly different adhesion and growth behaviors over a seven-day culture period. The PNA-C coating demonstrated good ability in promoting robust cell adhesion and extensive neurite outgrowth (Fig. 2E). Further, to quantitatively compare their performance, iPSC-derived neurons were cultured on PNA-C, PNA-N, and laminin coatings and monitored in using an Incucyte Live-Cell Analysis System. To ensure clear morphological identification, cells were seeded at a low density and continuously imaged for 2 days. The integrated NeuroTrack analysis module was employed to quantify key parameters, including average neurite length and cell confluence (Fig. 2F and Fig. S1A). The results demonstrated that the PNA-C coating supported neuronal adhesion and neurite outgrowth to a level comparable to the laminin control, both of which significantly outperformed the PNA-N coating, as shown in Fig. 2G and Fig. S1B (n = 6). This finding aligns with previous literature indicating that a solvent-exposed, C-terminal presentation of the RGD motif enhances its recognition by cell surface integrins, thereby facilitating superior cell attachment [34].

Patterning of neural networks through peptide interface

Leveraging the specific binding affinity of our PNA protein for silica, we introduced precise, customizable cellular microenvironments for neural culture. The fundamental premise is that by spatially controlling the presentation of the PNA protein on a glass substrate, we can direct and confine neural cell adhesion exclusively to the predefined patterns.

To achieve this, we employed a standard photolithography process to create micron-scale patterns on quartz glass, as schematically illustrated in Fig. 3A (with detailed procedures as mentioned above). We first sought to visually confirm the successful and selective adsorption of our protein onto the patterned substrate. Incubation with the PNA-GFP fusion protein resulted in clear and intense fluorescence that precisely delineated the exposed glass regions, while the photoresist-protected areas remained dark, confirming the high-fidelity patterning of the bioactive coating (Fig. 3B). Having established the spatial control over protein immobilization, we next investigated the resulting cellular organization. When iPSC-derived neural cells were seeded onto these PNA-patterned substrates, they adhered selectively and proliferated exclusively within the protein-functionalized areas. After seven days in culture, the neural cells formed well-defined networks that reproduced the underlying microscopic patterns. The cells confined to these patterns exhibited robust health and extensive neurite outgrowth, forming intricate intercellular connections.

Fig. 3.

Patterning of neural networks through peptide interface. (A) Fabrication process of patterned glass substrate; (B) Customizable coating of peptide and growing of neurons under optical microscope; (C) Customizable coating of peptide and growing of neurons under SEM

To obtain a higher-resolution view of the cell-material interface and neuronal morphology, we performed SEM. The SEM images (Fig. 3C) revealed that the neurons attached firmly and spread extensively on the PNA-coated patterns. Notably, the cells displayed a flattened morphology with numerous, elaborate neurites that extended along the pattern guidance, forming a dense meshwork. This superior cellular architecture underscores the effectiveness of the PNA interface in not only capturing cells but also in supporting their maturation and complex networking. Collectively, these results demonstrate that our bifunctional peptide interface provides a robust and versatile platform for the precise implementation of structured neural networks.

Tunable and high-throughput formation of neural spheroids

Having established the utility of our PNA interface for guiding 2D neural patterning, we next explored its capacity to direct the self-assembly of neural cells into 3D spheroids. We hypothesized that by systematically tuning the molecular structure of the PNA, specifically the cell-adhesive RGD variant and the structural rigidity of the linker, we could precisely control the cell-cell versus cell-substrate interactions that govern spheroid formation. To this end, we prepared a series of PNA variants with modified linkers (Fig. 4A). All new constructs, PNA-1, PNA-2, and PNA-3, were successfully expressed and purified, with molecular weights confirmed by SDS-PAGE (approximately 24.3 kDa, 24.0 kDa, and 24.7 kDa, respectively) consistent with theoretical values (Fig. S2). Meanwhile, PNA-GFP was set as PNA-4 for comparison.

Fig. 4.

Implementing PNA variants for tunable neural spheroid formation. (A) Schematic of the PNA variants and representative micrographs of neural spheroids formed on glass surfaces coated with each variant. PNA-1, −2, and − 3 were designed with RGD-1, −2, and − 3 motifs, respectively, connected via a rigid α-helical linker. PNA-4 was constructed by replacing the rigid linker in PNA-3 with GFP. (B) Quantification of the total number of spheroids per field of view (2.5 μm × 2 μm) formed on each PNA coating. Data are presented as mean ± SD from three independent biological replicates (n = 6). (C) Distribution of spheroid diameters per field of view (2.5 μm × 2 μm) for each PNA variant. Data are presented as mean ± SD from three independent biological replicates (n = 6)

We first constructed PNA-1 by replacing the original flexible linker with a well-defined rigid α-helical motif (17-Helix: AEAAAKEAAAKEAAAKA) [38]. We expected that this rigid linker would restrict the spatial freedom of the RGD motif, thereby stabilizing its orientation and promoting full exposure of the C-terminal cell-binding site to enhance integrin engagement. Following the determination of the optimal neural cell seeding density for neural spheroid formation (Fig. S3), cells were seeded at 5 × 10⁵ cells/cm² on PNA-1-coated glass. Notably, PNA-1 successfully induced the formation of neural spheroids with homogeneous size, confirming that linker rigidity is a critical factor in controlling spheroid formation. The growing process of neural spheroids was showing in Fig. S4.

Subsequently, we introduced two additional RGD motifs (RGD-2 and RGD-3) into the PNA framework, generating PNA-2 and PNA-3 for comparative evaluation. As summarized in Fig. 4A, PNA-3 proved to be the most effective variant, producing a high density of numerous, well-defined, and size-uniform neural spheroids. While PNA-1 and PNA-2 also supported spheroid formation, they produced comparatively fewer aggregates under the same conditions. The superior performance of PNA-3 highlights the importance of RGD motif affinity in determining the efficiency and uniformity of spheroid generation. Furthermore, evaluation of different PNA coating concentrations revealed that optimal spheroid formation occurred at 50 µg/mL, with no significant improvement at higher concentrations, thus establishing this as the ideal coating parameter.

Building on the success of PNA-3, we next asked whether further increasing the distance of the RGD motif from the glass substrate, by replacing the rigid linker with GFP, could reduce steric hindrance to promote even greater cell binding and aggregation [39]. The resulting construct, PNA-4, confirmed this possibility: its coated surfaces adsorbed more neural cells initially and ultimately yielded a larger number of neural spheroids with a greater average size than those formed on PNA-3. However, contrary to the highly uniform spheroids generated by PNA-3, the size distribution of aggregates on PNA-4 was broad and heterogeneous, indicating a loss of spheroid uniformity. These findings demonstrate that while extending the spacer arm with GFP enhances the bioactivity of the RGD motif and facilitates large-scale aggregation, it compromises the control over spheroid size. This underscores the critical need to optimize spacer length to balance sufficient cell adhesion with the cell reorganization dynamics necessary for consistent size control, a key objective for future research.

To corroborate these distinct morphological outcomes, we performed a quantitative analysis of the total number of spheroids formed (Fig. 4B) and their diameter distribution (Fig. 4C) across all PNA variants. The data confirmed that PNA-3 coatings consistently produce a high yield of spheroids with a narrow size distribution. In contrast, other variants offer a spectrum of outcomes in terms of spheroid count and dimensions. This tunability is a key advantage of our platform, as it allows researchers to select a specific PNA formulation to generate spheroids tailored to distinct experimental requirements, whether for high-throughput screening or for studies requiring spheroids of a precise size.

Functional validation of cultured neural spheroids

To comprehensively evaluate the functional integrity and physiological relevance of the neural spheroids generated by our PNA platform, we conducted a series of assessments focusing on viability, morphological maturation, and electrophysiological activity. First, we dissociated the spheroids into suspensions and performed a live/dead assay using Calcein-AM and Propidium Iodide staining kit. Fluorescence imaging and subsequent quantitative analysis revealed an exceptionally high cell viability exceeding 90% (Fig. 5A), indicating that the spheroid formation process maintained excellent cellular health.

Fig. 5.

Functional validation of achieved (or built) neural spheroids. (A) Live/dead staining of neural spheroid; (B) Growing of neural spheroids on culture dish and live/dead staining; (C) Neural spheroids on MEA chip; (D) Spike rates recorded from neural spheroids across different days; (E) Typical overlaid spike waveform from neural spheroids on Day 20; (F) Immunostaining of iPSC-derived neural spheroid

We next investigated the developmental potential and structural plasticity of the spheroids. Upon re-seeding onto laminin-coated dishes, the spheroids attached and began to extend neurites radially into the surrounding space within seven days. Remarkably, when adjacent spheroids were in proximity, their outgrowing neurites interconnected, forming prominent nerve bundle-like structures between separate aggregates (Fig. 5B). This observed network formation is a hallmark of mature and active neural tissue. A follow-up live/dead assay conducted after this week of extended culture confirmed that the spheroids remained robustly healthy, with viability sustained above 90%.

To definitively confirm functional maturity, we monitored the electrophysiological activity of the spheroids using a commercial MEA system. Multiple spheroids were transferred onto an MEA chip for non-invasive, long-term recording (Fig. 5C). The raster plot of spikes from neural spheroids was showing in Fig. S5. Analysis of spike rates over time revealed that spontaneous electrical activity emerged around day 8 in culture. This activity progressively increased in frequency and amplitude, stabilizing by day 20, which suggests the establishment of a mature and functional neuronal network within the spheroids (Fig. 5D). The representative superimposed spike waveforms recorded on day 20 exhibited characteristic shapes and amplitudes (Fig. 5E), consistent with typical neuronal action potentials reported in prior literature.

Collectively, these data provide compelling evidence that the neural spheroids fabricated using our tunable PNA interface are not only viable but also functionally competent. They exhibit key attributes of functional neural networks, including structural plasticity, spontaneous electrophysiological activity, and the capacity for inter-spheroid communication. This functional validation underscores the suitability of our platform for producing high-quality neural spheroids amenable to a wide range of downstream applications, including drug screening, disease modeling, and regenerative medicine studies.

To validate the structural integrity of our neural spheroids, immunostaining analysis was performed on neural spheroids to assess their internal architecture and maturity. Spheroids were fixed and immunostained for key neuronal markers. TUJ1 was used to visualize the dense cytoskeletal network and neurites, while NeuN served as a specific marker for mature neuronal nuclei. High-resolution confocal microscopy imaging of central cross-sections (now presented in Fig. 5F) revealed a highly interconnected and intricate internal structure. The images show robust and continuous TUJ1 signal throughout the spheroid core, indicating an extensive, dense meshwork of neuronal processes. Concurrently, clear and abundant NeuN-positive nuclei are evident, confirming the predominant and mature neuronal identity of cells within the spheroid interior. The intricate co-localization and interweaving of neurites between cell bodies demonstrate substantial cell-cell interactions and network integration deep within the spheroid. It confirms that our PNA-derived spheroids maintain structural integrity, high cellular viability, and a mature neuronal phenotype throughout their 3D volume, substantiating their suitability for functional neural studies.

Conclusion

In this study, we have successfully designed, synthesized, and validated a novel programmable neural interface, which effectively addresses the long-standing challenge of creating a single, tunable platform for both structured 2D neural network patterning and high-throughput 3D neural spheroid formation. The core innovation of our work lies in the rational fusion of a high-affinity silica-binding tag with a tunable cell-adhesive RGD multimer, connected by an optimized linker. This molecular design enables precise spatial control over neural cell adhesion on patterned glass substrates by simply modulating the RGD motif and linker rigidity, enabling the directed self-assembly of uniform, healthy neural spheroids.

Our results demonstrate the dual functionality of the PNA system. First, it serves as a highly specific and effective coating for guiding the growth of iPSC-derived neurons into predefined, complex micropatterns, resulting in the formation of mature synaptic networks, as confirmed by fluorescence images and SEM. Second, by fine-tuning the cell-substrate interaction strength, we identified an optimal PNA variant (PNA-3) that promotes the efficient and consistent formation of 3D neural spheroids with high viability and a narrow size distribution. Functional analyses, including live/dead assay and MEA recordings, confirmed that 3D spheroids exhibit spontaneous and synchronized electrophysiological activity, a critical indicator of their physiological relevance and functional maturity. This confirms that our PNA-based fabrication process preserves the intrinsic electrophysiological properties of the neural cells.

It is worth noting that, similar to other bioactive coatings such as Matrigel and laminin, the PNA-coated substrates are intended for single-use to ensure optimal performance and reproducibility. Limited reusability (2–3 cycles) was observed for the spheroid-forming variants, albeit with declining efficiency, as the cell removal process between cycles can compromise coating integrity. The long-term stability of the PNA coating under culture conditions is another critical feature. The Si-tag provides robust anchorage to glass, maintaining stability for weeks. Furthermore, the cellular interaction stability is application-tuned: PNA-C is designed to be permanent for patterned network formation, supporting cultures for over four weeks, whereas for spheroid formation, stability of PNA 1 to PNA 4 is moderate, enabling long-term spheroid culture and analysis.

Looking forward, the PNA platform opens up numerous exciting avenues for both fundamental research and translational applications (Fig. 6). Firstly, the modular nature of the design invites the incorporation of other bioactive peptides (e.g., IKVAV for enhanced neurite outgrowth or specific ligands for different cell types) to create bespoke microenvironments for studying specific neurobiological processes. Secondly, this technology holds significant promise for advancing disease modeling. The ability to reliably generate uniform, functional human neural spheroids from patient-derived iPSCs provides an excellent tool for drug screening and toxicology studies for neurological disorders, and the compatibility of PNA-engineered tissues with multi-well formats further supports high-throughput screening workflows.

Fig. 6.

Schematic diagram of the programmed peptide-modified neural platform and its applications

In addition, given their inherent compatibility with planar microelectronic devices such as microelectrodes and microelectrode arrays (MEAs), PNA-coated interfaces offer a promising platform for constructing stable neuro-electrode hybrids. This feature could enable long-term, high-fidelity electrophysiological recordings from both patterned 2D neuronal networks and 3D organoid assemblies. Beyond modeling single-tissue systems, the tunable and interface-selective properties of PNA components may further support the construction of cross-dimensional disease models. For example, platforms that capture interactions along the gut-brain axis or immune-neural crosstalk offer opportunities for mechanistic studies requiring communication across spatial and biological scales. Finally, the ability of PNA modules to program cell-type-specific adhesion suggests that different organoid types could be selectively positioned and co-cultured on the same substrate, thereby enabling modular “body-on-a-chip” configurations. This capability supports advanced applications such as multi-organoid crosstalk models and higher-order organoid assemblies.

Altogether, the PNA technology presented here transcends the limitations of conventional, single-purpose coating materials. It establishes a versatile and powerful paradigm for neural interface implementation, offering unprecedented control over neural cell organization in both two and three dimensions. We anticipate that this platform will become a valuable tool for the neuroscience and bioengineering communities, accelerating progress in understanding neural development, modeling diseases, and developing new therapeutic strategies.

Author contributions

HZ: Wrote the original draft, did most of the experiments about cell culture and chip fabrication. XS: Wrote the original draft, did most of the experiments about peptide synthesis and characterization. NH: Neural cell culture and immunostaining. KX: Stem cell culture. NS: Neural signals recording. YS: Data analysis. SB: Paper review and editing. MS: Paper review and editing, Project administration, Funding acquisition.

Funding

This research was supported by: the National Natural Science Foundation of China (Grant No. W2431058); the “Pioneer” and “Leading Goose” Research and Development Program of Zhejiang under Grant 2024C03002; Project of Westlake Institute for Optoelectronics under Grant 2023GD004.

Data availability

No datasets were generated or analyzed during the current study.

Declarations

Consent for publication

All the authors agree to the publication of the article.

Competing interests

The authors declare no competing interests.