AAV-DJ enables efficient, function-preserving neuronal transduction for circuit interrogation in human cortical and thalamic organoids

Summary

Neural organoids provide tractable models of human brain development, function, and disease, but adeno-associated viral vector (AAV)-based gene delivery is limited by inefficient transduction and insufficient comparative evaluation of the natural and engineered capsids. Here, we evaluated seven AAV capsids (AAV-2, -6, -9, -DJ, -2-retro, -PHP.eB, and -Cap-Mac) for transduction efficiency, cell type tropism, cytotoxicity, and functional integrity in human cortical and thalamic organoids. AAV-Cap-Mac demonstrated the highest transduction efficiency, with broad cellular tropism, but caused cytotoxicity and decreased neuronal activity. By contrast, AAV-DJ yielded efficient neuronal transduction, with preserved calcium dynamics and minimal cytotoxicity, which enabled rabies virus-based circuit tracing and all-optical physiology. Our findings reveal a capsid-dependent trade-off between transduction efficiency and functional integrity and identify AAV-DJ as an optimal serotype for robust gene delivery and physiological interrogation in organoid-based neuroscience. Combining AAV-DJ with emerging molecular and circuit-level technologies in organoids should facilitate mechanistic and translational studies of the human brain.

Graphical abstract

Highlights

• •AAV serotype comparisons reveal transduction efficiency versus toxicity trade-offs
• •AAV-DJ yields function-preserving neuronal transduction in human neural organoids
• •AAV-Cap-Mac exhibits efficient, broad transduction but elevated cytotoxicity
• •AAV-DJ enables all-optical physiology and monosynaptic circuit tracing

Motivation

Human neural organoids have emerged as powerful models of brain development and disease, but progress has been limited by the lack of validated viral tools for efficient and function-preserving gene delivery. To address this challenge, a comparative framework was developed to benchmark AAV capsids in human cortical and thalamic organoids by integrating measurements of transduction efficiency, cell type tropism, cytotoxicity, and physiological function, thereby identifying vectors suitable for functional imaging, optogenetics, and circuit tracing.

Adachi et al. systematically benchmark AAV capsids in human cortical and thalamic organoids by integrating gene delivery, cell type tropism, cytotoxicity, and neuronal function. By identifying AAV-DJ as a low-toxicity, function-preserving vector, this study offers a practical platform for imaging, manipulating, and tracing human neural circuits in vitro.

Introduction

The human brain emerges through the coordinated interplay of intrinsic genetic programs, cell type specification, and activity-dependent circuit assembly during development, yet this process remains largely inaccessible to direct experimental analysis in humans. Recent advances in stem cell biology have enabled the generation of three-dimensional (3D) neural organoids from pluripotent stem cells (PSCs), opening new avenues for the study of neurodevelopment and disease pathobiology.^1,2,3,4 These organoids partially recapitulate key hallmarks of the developing human brain, including the formation of the ventricular zone and cortical plate, radial organization, synaptogenesis, glutamatergic neurotransmission, and spontaneous neural activity.^3,5,4,6 Human neural organoids are, therefore, powerful experimental platforms for dissecting the molecular-, cellular-, and circuit-level mechanisms of human brain development, function, and disease that are inaccessible in vivo.

To fully exploit these organoid models, however, efficient and well-tolerated genetic and physiological access to defined cell populations is essential. Such access is required not only for monitoring gene expression and neural activity but also for manipulating molecular pathways and circuit function in a controlled manner. Among available gene delivery tools, viral vectors have emerged as powerful platforms for gene delivery in the mammalian brain.⁷ In particular, adeno-associated virus vectors (AAVs), members of the Parvoviridae family that consist of a 4.7-kb single-stranded DNA genome packaged within a non-enveloped protein capsid, are the most widely used because of their low immunogenicity, broad tropism, and efficient transduction of post-mitotic cells, such as neurons.^8,9 However, AAV capsid tropism is highly context dependent, and transduction efficiency can vary substantially across species, cell types, and developmental stages. As a result, AAV serotypes optimized in rodent models often do not translate effectively to the distinct cellular environment of non-human primates or human neural organoids.⁹ Gene delivery into human neural organoids has been attempted using bath application and direct microinjection of various AAV serotypes pseudotyped with different capsids, such as AAV2/2, AAV2/1, AAV2/5, AAV2/6, AAV2/8, and AAV2/9,^10,11,12 but efficient and function-preserving gene transfer remains a fundamental challenge. This limitation restricts both mechanistic studies, including gain-of-function and loss-of-function analyses, and the application of advanced molecular tools in complex organoid systems.

The transduction efficiency and cell type specificity of AAVs depend largely on the capsid, which determines receptor binding, cellular tropism, and intracellular trafficking.^8,9,13 Over the past decade, rational engineering of AAV capsids has yielded variants with properties tailored for specific applications.^14,15,16,17 For example, AAV-DJ is a hybrid capsid of AAV-2, AAV-8, and AAV-9 developed by gene shuffling that exhibits broad tropism, including efficient hepatic transduction.¹⁴ Similarly, AAV-2-retro was engineered for efficient retrograde infection via axon terminals,¹⁵ AAV-PHP.eB was created based on AAV-9 to cross the blood-brain barrier (BBB) and enhance central nervous system transduction in mice,¹⁷ and AAV-Cap-Mac was designed to improve brain-wide gene transfer in non-human primates.¹⁶ While pseudotyping of AAVs with capsids of different serotypes allows changes in transduction efficacy and cell type tropism,^8,9 there has been no comprehensive and systematic evaluation of AAV serotypes with these natural and engineered capsids in human neural organoids that considers not only transduction efficiency and tropism but also cytotoxicity and preservation of physiological function.

In this study, we systematically compared seven distinct AAV serotypes, three natural (AAV-2, AAV-6, and AAV-9) and four engineered capsids (AAV-DJ, AAV-2-retro, AAV-PHP.eB, and AAV-Cap-Mac), in cortical and thalamic organoids derived from human induced pluripotent stem cells (hiPSCs). Because organoid cellular composition changes substantially across developmental stages and can influence AAV performance, we focused primarily on later-stage organoids enriched in differentiated neurons and glia, which are more relevant for functional and circuit-level interrogation. By integrating measurements of transduction efficiency, cell type tropism, cytotoxicity, and physiological readouts, we identified a capsid-dependent trade-off between transduction efficiency and preservation of neural function. Among the capsids tested, AAV-DJ provided efficient neuronal transduction while preserving functional integrity in both human cortical organoids (hCOs) and human thalamic organoids (hThOs). Together, these findings offer a valuable methodological framework for organoid-based neuroscience and translational research of the human brain.

Results

Characterization of AAV serotypes for hCO labeling

We generated hCOs from hiPSCs, using a serum-free floating culture of embryoid body-like aggregates with quick reaggregation (SFEBq)-based method (Figures 1A and 1B).^4,6 Serial quantitative reverse-transcription PCR (RT-qPCR) measurements during culture revealed sharply reduced expression of the PSC marker octamer-binding transcription factor 4 (OCT4) and concomitant upregulation of the neural stem cell marker paired box 6 (PAX6) and the telencephalic markers Forkhead box protein G1 (FOXG1) and empty spiracles homeobox 1 (EMX1) (Figures 1C and 1D). In the early stage of differentiation (day 42), immunostaining revealed that PAX6⁺ neural progenitors formed ventricular zone-like structures outside the N-cadherin⁺ apical membrane. By contrast, in the late stage (day 89), cells positive for the neuronal markers microtubule-associated protein 2 (MAP2) and neuronal nuclei (NeuN; also known as RNA-binding fox-1 homolog 3, RBFOX3), the deep-layer cortical neuron markers T-box brain protein 1 (TBR1) and COUP-TF-interacting protein 2 (CTIP2), and the astrocyte markers glial fibrillary acidic protein (GFAP) and S100 calcium-binding protein beta (S100β) were distributed in cortical plate-like regions (Figures 1E and 1F). In day 89 organoids, more than 90% of NeuN⁺ cells expressed TBR1, indicating substantial differentiation toward deep-layer cortical neurons. These results indicated that day 42 hCOs correspond to an early neurogenic stage, while day 89 hCOs correspond to a later stage containing differentiated neurons and glia. Thus, these hCOs recapitulate key cytoarchitectonic features and genetic programs of the developing cortex.

Figure 1.

AAV capsid serotype determines transduction efficiency and cell type tropism in hCOs

(A) Methods for generating hCOs from hiPSCs.

(B) Stepwise culture treatments for generating hCOs from hiPSCs.

(C) RT-qPCR analysis for developmental markers in hCOs. n = 3 organoids from a set of experiments. Marker gene expression levels were normalized to GAPDH expression.

(D) Schematic representation of the marker expression patterns from hiPSCs to hCOs.

(E) Immunostaining for FOXG1, PAX6, N-CADHERIN, NeuN, MAP2, TBR1, and CTIP2 in hCOs on day 42. Scale bars: 500 μm (top) and 100 μm (bottom).

(F) Immunostaining for MAP2, NeuN, CTIP2, TBR1, and S100β in hCOs on day 89. Scale bars: 100 μm.

(G) Experimental scheme for the quantification of AAV-based transduction efficiency in hCOs shown in (H). Organoids were transduced with AAV-X-CAG-EYFP on day 72 of differentiation and analyzed on day 89, corresponding to 17 days post-transduction (dpt).

(H) Transduction efficiency of seven AAV serotypes in hCOs. Quantification shows the proportion (%) of EYFP⁺ cells relative to the total number of cells in each organoid. n = 8 organoids from 3 sets of differentiation experiments (AAV-2), n = 6 organoids from 3 sets of differentiation experiments (AAV-6, -9, -PHP.eB, and -Cap-Mac), n = 14 organoids from 4 sets of experiments (AAV-DJ), and n = 18 organoids from 4 sets of differentiation experiments (AAV-2-retro). ∗∗∗p < 0.001 vs. AAV-9, by Dunnett’s test. Red dots indicate results from individual organoids.

(I) Methods for evaluating cell-type-specific tropism of AAV-X-CAG-EYFP by immunostaining and confocal imaging for (J)–(Q).

(J) Immunostaining for EYFP and the neuronal marker MAP2 in hCOs on day 89. Images were acquired 17 dpt with AAV-DJ, 2-retro, or Cap-Mac (equal titers). Scale bars: 50 μm.

(K) Immunostaining for EYFP and the astrocyte marker GFAP in hCOs on day 89. Images were acquired 17 dpt with AAV-DJ, 2-retro, or Cap-Mac. Scale bars: 50 μm.

(L) Immunostaining for EYFP and the oligodendrocyte marker OLIG2 in hCOs on day 89. Images were acquired 17 dpt with AAV-DJ, 2-retro, or Cap-Mac. Scale bars: 50 μm.

(M) Proportion (%) of MAP2⁺ neurons among AAV-transduced, EYFP-expressing cells. n = 5 organoids, 3 sets of experiments (AAV-DJ), n = 6 organoids, 2 sets of experiments (AAV-2-retro), and n = 7 organoids, 3 sets of experiments (AAV-Cap-Mac). ∗p < 0.05, by Tukey’s test.

(N) Proportion (%) of NeuN⁺ neurons among AAV-transduced, EYFP-expressing cells. n = 9 organoids, 3 sets of experiments (AAV-DJ, AAV-2-retro, and AAV-Cap-Mac). ∗∗p < 0.01, by Tukey’s test.

(O) Proportion (%) of GFAP⁺ astrocytes among AAV-transduced, EYFP-expressing cells. n = 5 organoids, 3 sets of experiments (AAV-DJ), n = 6 organoids from 2 sets of experiments (AAV-2-retro), and n = 7 organoids, 3 sets of experiments (AAV-Cap-Mac). ∗∗p < 0.01, by Tukey’s test.

(P) Proportion (%) of S100β⁺ astrocytes among AAV-transduced, EYFP-expressing cells. n = 9 organoids, 3 sets of experiments (AAV-DJ, AAV-2-retro, and AAV-Cap-Mac). ∗p < 0.05 and ∗∗p < 0.01, by Tukey’s test.

(Q) Proportion (%) of OLIG2⁺ oligodendrocytes among AAV-transduced, EYFP-expressing cells. n = 9 organoids, 3 sets of experiments (AAV-DJ, AAV-2-retro, and AAV-Cap-Mac).

We then examined the labeling efficiency of seven distinct AAV serotypes: AAV-2, AAV-6, AAV-9, AAV-DJ, AAV-2-retro, AAV-PHP.eB, and AAV-Cap-Mac.^{14,15,16,17,18} Each AAV encoded enhanced yellow fluorescent protein (EYFP) under the control of the ubiquitous CAG promoter. For the initial serotype comparison, we used the ubiquitous CMV early enhancer/chicken beta-actin (CAG) promoter to minimize cell type bias and to enable a more uniform (unbiased) evaluation of capsid-dependent transduction across the diverse cell populations present in organoids. Organoids were transduced at matched genome titers on day 72 and analyzed on day 89 (17 days post-transduction [dpt]) (Figure S1A). Whole-mount immunostaining using the clear, unobstructed brain/body imaging cocktails and computational analysis (CUBIC) method¹⁹ revealed a large number of EYFP⁺ cells in hCOs transduced with AAV-DJ, AAV-2-retro, or AAV-Cap-Mac (Figure S1B). To quantify the transduction efficiency, we dissociated AAV-transduced hCOs and analyzed them by flow cytometry on day 89 (Figure 1G). Among all capsids tested, AAV-Cap-Mac demonstrated the highest transduction efficiency, as evidenced by a significantly greater proportion of EYFP⁺ cells compared with AAV-9, which served as a reference serotype (Figure 1H). Moreover, transduction of AAV-DJ, AAV-2-retro, and AAV-2 also yielded larger numbers of EYFP⁺ cells than AAV-9 transduction, whereas AAV-6 and AAV-PHP.eB were as inefficient as AAV-9 (Figure 1H). These results indicated that three serotypes, AAV-DJ, AAV-2-retro, and AAV-Cap-Mac, have high transduction efficiency in hCOs.

We next evaluated which cell types were transduced by the three most efficient serotypes, AAV-DJ, AAV-2-retro, and AAV-Cap-Mac. Because hCOs on day 89 contained both neurons and glia, the samples were co-immunostained for the neuronal markers MAP2 and NeuN, the astrocyte markers GFAP and S100β, and the oligodendrocyte marker oligodendrocyte transcription factor 2 (OLIG2) (Figures 1I–1L and S2A–S2C). In the first analysis, we quantified the proportions of marker-positive cells within the EYFP⁺ population, thereby assessing the tropism of AAV-transduced cells, that is, which cell types were preferentially represented among the transduced cells. This approach evaluates the relative selectivity of each capsid independent of its overall transduction efficiency, thereby enabling a direct comparison across serotypes with different absolute transduction rates. All three serotypes transduced neuronal populations identified by MAP2 or NeuN (Figures 1M and 1N). In contrast, astrocytic populations immunopositive for GFAP or S100β were more frequently detected within the EYFP⁺ populations in the AAV-2-retro and AAV-Cap-Mac groups than in the AAV-DJ group (Figures 1O and 1P), while OLIG2⁺ cells were rarely detected within the EYFP⁺ populations for any of the tested serotypes (Figure 1Q). These results suggest that among AAV-transduced cells, AAV-DJ showed a higher tropism for neurons, whereas AAV-2-retro and AAV-Cap-Mac exhibited a broader tropism, including neurons and astrocytes.

To further account for possible differences in the overall cell type composition across organoids, we performed a reciprocal analysis by quantifying the fraction of marker-positive cells that were also EYFP⁺ to assess cell-type-specific transduction efficiency, that is, how efficiently each serotype transduced a given cell population. This approach revealed that NeuN⁺ neurons were most efficiently transduced by AAV-DJ, whereas S100β⁺ astrocytes were preferentially transduced by AAV-2-retro, and OLIG2⁺ cells were only rarely EYFP⁺ for any of the serotypes tested (Figures S2D–S2F). Thus, although the two quantification strategies emphasize different aspects of AAV performance, both supported the same overall conclusion that AAV-DJ shows a stronger neuronal bias, whereas AAV-2-retro and AAV-Cap-Mac exhibit a broader tropism for both neurons and certain glial cells.

AAV serotype-dependent cytotoxicity in hCOs

AAVs can differ not only in transduction efficiency but also in cytotoxicity.²⁰ We, therefore, evaluated serotype-specific cytotoxicity in hCOs by focusing on the three serotypes with the highest transduction efficiency: AAV-DJ, AAV-2-retro, and AAV-Cap-Mac. After transducing hCOs with these AAVs at a matched titer on day 72, we performed co-immunostaining for EYFP as an indicator of AAV transduction and cleaved (activated) caspase-3 as a marker of apoptotic cell death, on day 89 (17 dpt) (Figure 2A). To assess cytotoxicity specifically within the transduced population, we first quantified the fraction of EYFP⁺ cells that were also immunopositive for cleaved caspase-3. Approximately 10% of EYFP⁺ cells in the AAV-Cap-Mac group were also immunopositive for cleaved caspase-3, a proportion significantly higher than that in the AAV-DJ and AAV-2-retro groups, suggesting that AAV-Cap-Mac transduction is more cytotoxic (Figures 2B and 2C). To further examine the temporal profile of AAV-associated cytotoxicity, we compared organoids at 10 and 17 dpt, while earlier time points, such as day 3 or day 7, were not examined because sufficient EYFP expression required additional time after transduction. At 10 dpt, hCOs transduced with AAV-DJ, -2-retro, or -Cap-Mac exhibited little detectable cytotoxicity. By 17 dpt, however, the AAV-Cap-Mac group showed a marked increase in cleaved caspase-3⁺ cells, while such cells were still only barely detectable in the AAV-DJ and AAV-2-retro groups (Figure 2D). Thus, while AAV-Cap-Mac provides high transduction efficiency in hCOs, it also induces substantial delayed cell death. As a complementary analysis, we also quantified the fraction of cleaved caspase-3⁺ cells that were also EYFP⁺ at 17 dpt. This reciprocal analysis showed that the proportion of cleaved caspase-3 and EYFP double-positive cells was the highest in the AAV-Cap-Mac group, followed by the AAV-2-retro group, whereas only a small fraction of such double-positive cells was detected in the AAV-DJ group (Figure S2G). Together, these two analyses supported the same overall conclusion that AAV-Cap-Mac is associated with increased cytotoxicity, whereas AAV-DJ and AAV-2-retro maintain more favorable safety profiles.

Figure 2.

AAV serotype-dependent cytotoxicity in hCOs

(A) Schematic for viral labeling and 2D imaging. For immunohistochemical analysis, organoids were transduced by AAV-X-CAG-EYFP on day 72 of differentiation and fixed on day 89 (17 dpt).

(B) Immunostaining for EYFP and cleaved caspase-3 in hCOs on day 89. Scale bars: 50 μm.

(C) Proportion (%) of cleaved caspase-3⁺ cells among AAV-transduced, EYFP-expressing cells. n = 5 organoids from 3 sets of experiments (AAV-DJ and Cap-Mac) and n = 7 organoids from 3 sets of experiments (AAV-2-retro). ∗p < 0.05 and ∗∗p < 0.01, by Tukey’s test.

(D) Temporal profile of cell death induced by AAV transduction. ∗p < 0.05, by Welch’s t test.

Taken together, these results show that AAV serotypes differ markedly in their effects on cell viability in hCOs. AAV-Cap-Mac provides high transduction efficiency but induces delayed cytotoxicity, whereas AAV-DJ and AAV-2-retro combine efficient gene delivery with comparatively low toxicity.

Analysis of neural activity in AAV-transduced hCOs

In later stages of neural development, differentiated neurons begin to generate neural activity, which contributes to cortical organization and circuit formation.²¹ Because the evaluation of neural function requires relatively mature neurons, we focused on later-stage hCOs for these experiments. Compared with early-stage organoids (day 42), which were enriched in PAX6⁺ neural progenitors, later-stage organoids (days 83–89) contained abundant NeuN⁺ and MAP2⁺ neurons. Moreover, more than 90% of NeuN⁺ neurons were also positive for TBR1 or CTIP2, indicating advanced neuronal differentiation at this stage. Furthermore, rabies virus tracing indicated the presence of synaptically connected networks in these later-stage organoids. Together, these observations supported the use of late-stage organoids for calcium (Ca²⁺) imaging analyses.

Because our initial experiments showed that the transduction efficiency and cytotoxicity in hCOs varied by AAV serotype (Figures 1 and 2), we hypothesized that certain AAV serotypes would be better suited for functional assays of differentiated neurons in human neural organoids. To restrict transgene expression to neurons, we first generated AAVs expressing EYFP under the control of the neuron-specific human synapsin (hSyn) promoter and examined cellular specificity in hCOs. Organoids transduced on day 72 of differentiation contained numerous EYFP-expressing cells also immunopositive for NeuN on day 89 (Figures S3A and S3B), whereas we did not detect EYFP-expressing cells positive for S100β or OLIG2 (Figure S3C), indicating that the hSyn promoter drove transgene expression predominantly in neurons in hCOs.

To determine the AAV serotypes most suitable for functional imaging analysis in neurons within hCOs, we generated AAV-hSyn-jGCaMP7f vectors for two-photon Ca²⁺ imaging. Based on the serotype comparison using AAV-X-CAG-EYFP, we selected the three best-performing capsids (AAV-Cap-Mac, AAV-2-retro, and AAV-DJ), together with AAV-9 as a reference control. Two-photon Ca²⁺ imaging was conducted 11 days after transduction with matched genome titers on day 72 (Figure 3A). All four serotypes yielded jGCaMP7f-expressing neurons distributed throughout the hCOs (Figure 3B). Among them, AAV-Cap-Mac-hSyn-jGCaMP7f yielded the largest number compared with AAV-2-retro and AAV-9 (Figures 3B and 3C), consistent with our initial transduction efficiency analysis (Figure 1). Spontaneous fluctuations in the jGCaMP7f fluorescence intensity were observed in subsets of labeled neurons, indicating intrinsic neural activity within hCOs (Figure 3D). Notably, however, both the proportion of these active neurons and the average signal magnitude differed markedly by AAV serotype. To evaluate the suitability of each AAV serotype for functional imaging, we first quantified the proportions of neurons exhibiting spontaneous Ca²⁺ transients (active neurons) among jGCaMP7f-labeled neurons. Neurons transduced with AAV-9 were largely silent, while neurons transduced with AAV-DJ- and AAV-2-retro-based vectors showed significantly larger fractions of active neurons than those transduced with AAV-9 and AAV-Cap-Mac (Figure 3E). Importantly, while AAV-Cap-Mac possessed the highest transduction efficiency, a large proportion of labeled neurons were functionally inactive. We also quantified the amplitude of jGCaMP7f signals in active neurons. AAV-DJ-transduced neurons showed the highest signal amplitude among all AAV serotypes tested (Figure 3F). In contrast, AAV-2-retro- and AAV-Cap-Mac-transduced neurons showed smaller neural response amplitudes.

Figure 3.

AAV-DJ transduction preserves Ca²⁺ dynamics in hCOs

(A) Schematic for viral labeling and two-photon Ca²⁺ imaging of hCOs. Organoids were transduced by AAV-X-hSyn-jGCaMP7f on day 72 of differentiation, and two-photon Ca²⁺ imaging was conducted on day 83 (11 dpt).

(B) Representative images of hCOs on day 83 after transduction with AAV-9, -DJ, -2-retro, or -Cap-Mac. Scale bars: 100 μm.

(C) Numbers of jGCaMP7f-expressing neurons in hCOs transduced by different AAV serotypes. n = 9 organoids from 3 sets of experiments (27 imaging fields: AAV-9, -DJ, and -2-retro), n = 8 organoids from 3 sets of experiments (24 imaging fields: AAV-Cap-Mac). ∗p < 0.05, ∗∗p < 0.01, and ∗∗∗p < 0.001, by Dunnett’s test. Red dots represent results from individual organoids.

(D) Time-series traces of jGCaMP7f signal changes in hCOs transduced with AAV-9, -DJ, -2-retro, or -Cap-Mac vectors.

(E) Proportions of active cells (exhibiting spontaneous Ca²⁺ transients) among jGCaMP7f-expressing neurons transduced by different AAV serotypes in hCOs. n = 9 organoids from 3 sets of experiments (27 imaging fields: AAV-9, -DJ, and -2-retro), n = 8 organoids from 3 sets of experiments (24 imaging fields: AAV-Cap-Mac). ∗p < 0.05 and ∗∗p < 0.01, by Dunnett’s test. Red dots represent the results from individual organoids.

(F) Average signal amplitudes of jGCaMP7f-expressing active neurons transduced by different AAV serotypes in hCOs. Distributions are shown as violin plots. Results are presented as medians. n = 9 organoids from 3 sets of experiments (27 imaging fields: AAV-9, -DJ, and -2-retro), n = 8 organoids from 3 sets of experiments (24 imaging fields: AAV-Cap-Mac). ∗∗p < 0.01, by Tukey’s test.

(G) Methods for evaluating the cytotoxicity of AAV serotypes by co-immunostaining for jGCaMP7f and cleaved caspase-3.

(H) Cleaved caspase-3⁺ neurons among jGCaMP7f-expressing neurons.

To examine whether impaired activity reflects the inherent cytotoxicity of the vector, we again conducted co-immunostaining for jGCaMP7f and cleaved caspase-3 (Figure 3G). Cleaved caspase-3-positive cells were more frequently observed in the AAV-Cap-Mac group and were also detected in the AAV-2-retro group, whereas few were detected in the AAV-DJ group (Figure 3H), consistent with our initial assessment using the CAG promoter (Figure 2C). Thus, while AAV-Cap-Mac was the most efficient vector for neuronal transduction, AAV-DJ was the most appropriate for Ca²⁺ imaging applications in hCOs.

Taken together, these results reveal a dissociation between the transduction efficiency and functional suitability in hCOs. AAV-Cap-Mac showed the highest transduction efficiency, but many transduced neurons were functionally silent and showed increased cytotoxicity. AAV-2-retro showed higher transduction efficiency but smaller neural response amplitudes. In contrast, AAV-DJ combined efficient neuronal transduction with a high proportion of active neurons, strong Ca²⁺ signals, and minimal cytotoxicity, indicating that it provided the best overall performance for functional imaging in hCOs.

Characterization of AAV serotypes suitable for hThOs

Because the suitability of AAV serotypes may vary depending on the type of neural organoid, we next evaluated AAV serotypes for hThOs, which differ from cortical organoids in terms of developmental program and cellular composition.⁹ We generated hThOs from hiPSCs following an SFEBq-based protocol (Figure 4A).²² To determine whether the induced organoids had thalamic identity, we performed immunostaining and quantified the expression of TCF7L2, a transcription factor enriched in the developing thalamus. TCF7L2⁺ cells were detected within the organoids, which indicated successful differentiation of hiPSCs into hThOs (Figures 4B and 4C).

Figure 4.

AAV-DJ demonstrates efficient transduction with preserved Ca²⁺ dynamics and minimal toxicity in hThOs

(A) Method for generating hThOs from hiPSCs.

(B) Immunostaining for the thalamic marker TCF7L2 in hThOs on day 54. Scale bars: 1,000 μm.

(C) Quantification of TCF7L2 expression in hCOs and hThOs. n = 4 hCOs from a set of experiments; n = 3 hThOs from a set of experiments. ∗p < 0.05, Welch’s t test.

(D) Experimental scheme for the quantification of transduced cells in hThOs, using a cell sorter.

(E) Transduction efficiency of seven AAV serotypes in hThOs. Quantification shows the proportion (%) of EYFP⁺ cells relative to the total number of cells in each organoid. n = 6 organoids from 3 sets of experiments. ∗∗p < 0.01 vs. AAV-9, Dunnett’s test.

(F) Method for evaluating cell-type-specific tropism of AAV vectors by immunostaining and confocal imaging (for G–J).

(G) Immunostaining for EYFP and the neuron marker MAP2 in AAV-transduced hThOs on day 70. Scale bars: 50 μm.

(H) Proportion (%) of MAP2⁺ neurons among AAV-transduced, EYFP-expressing cells. n = 10 organoids from 3 sets of experiments (AAV-DJ), n = 9 organoids from 3 sets of experiments (AAV-2-retro), and n = 5 organoids from 3 sets of experiments (AAV-Cap-Mac). Tukey’s test.

(I) Immunostaining for EYFP and the astrocyte marker GFAP in AAV-transduced hThOs. Scale bars: 50 μm.

(J) Proportion of GFAP⁺ astrocytes among AAV-transduced, EYFP⁺ cells. n = 10 organoids from 3 sets of experiments (AAV-DJ), n = 9 organoids from 3 sets of experiments (AAV-2-retro), and n = 5 organoids from 3 sets of experiments (AAV-Cap-Mac). ∗p < 0.05, by Tukey’s test.

(K) Schematic for viral labeling by AAV-X-hSyn-jGCaMP7f and two-photon Ca²⁺ imaging in hThOs (for L–P).

(L) Representative images of hThOs after transduction with AAV-9, DJ, 2-retro, or Cap-Mac vector. Scale bars: 100 μm.

(M) Numbers of jGCaMP7f-expressing neurons transduced by different AAV serotypes in hThOs. n = 9 organoids from 3 sets of experiments (26 imaging fields: AAV-9 and Cap-Mac), n = 8 organoids from 3 sets of experiments (24 imaging fields: AAV-DJ), and n = 9 organoids from 3 sets of experiments (27 imaging fields: AAV-2-retro). ∗∗∗p < 0.001, by Dunnett’s test.

(N) Time-series traces of jGCaMP7f signal changes in hThOs transduced with AAV-9, DJ, 2-retro, or Cap-Mac vectors.

(O) Proportions (%) of active cells (with spontaneous Ca²⁺ signals) among jGCaMP7f-expressing neurons transduced by different AAV serotypes in hThOs. n = 9 organoids from 3 sets of experiments (26 imaging fields: AAV-9 and Cap-Mac), n = 8 organoids from 3 sets of experiments (24 imaging fields: AAV-DJ), and n = 9 organoids from 3 sets of experiments (27 imaging fields: AAV-2-retro). Tukey’s test.

(P) Average signal amplitudes of jGCaMP7f-expressing active neurons transduced by different serotypes in hThOs on day 83. Distributions are shown as violin plots. Results are presented as medians. n = 9 organoids from 3 sets of experiments (26 imaging fields: AAV-9 and Cap-Mac), n = 8 organoids from 3 sets of experiments (24 imaging fields: AAV-DJ), and n = 9 organoids from 3 sets of experiments (27 imaging fields: AAV-2-retro). ∗p < 0.05, by Tukey’s test.

(Q) Schematic for viral labeling and 2D imaging. For immunohistochemical analysis, organoids were transduced by AAV-X-CAG-EYFP on day 55 of differentiation and analyzed on day 70.

(R) Immunostaining for EYFP and cleaved caspase-3 in hThOs. Scale bars: 50 μm.

(S) Proportions (%) of active cleaved caspase-3⁺ cells among AAV-transduced, EYFP-expressing cells. n = 8 organoids from 3 sets of experiments (AAV-DJ and 2-retro), and n = 7 organoids from 3 sets of experiments (AAV-Cap-Mac). ∗p < 0.05 and ∗∗p < 0.01, by Tukey’s test.

To efficiently monitor and manipulate gene expression and neural function in hThOs, we evaluated which AAV serotypes most efficiently transduced hThOs. hThOs were transduced with AAV-X-CAG-EYFP (AAV-2, AAV-6, AAV-9, AAV-DJ, AAV-2-retro, AAV-PHP.eB, and AAV-Cap-Mac) on day 55, followed by whole-mount immunostaining and tissue clearing using the CUBIC method (Figures S4A and S4B). All AAV serotypes induced EYFP expression in cleared hThOs. Robust EYFP signals were detected in organoids transduced with AAV-Cap-Mac, AAV-DJ, and AAV-2-retro, whereas weaker labeling was observed with the other serotypes. For quantification and comparison of AAV transduction efficiency, we dissociated AAV-transduced hThOs into single cells and subjected them to flow cytometry (Figure 4D). AAV-Cap-Mac produced the highest proportion of EYFP-expressing cells and, thus, significantly outperformed AAV-9, whereas AAV-DJ and AAV-2-retro also showed higher transduction efficiency than AAV-9 but lower efficiency than AAV-Cap-Mac (Figure 4E). Thus, as in hCOs (Figure 1), AAV-Cap-Mac, AAV-DJ, and AAV-2-retro were identified as the best-performing serotypes for gene delivery in hThOs.

Next, we assessed the cell type tropism of AAV serotypes in hThOs by quantifying the proportions of neuronal and glial marker-positive cells within the EYFP⁺ population on day 70 (15 dpt) (Figure 4F). The serotypes AAV-DJ, AAV-2-retro, and AAV-Cap-Mac transduced MAP2⁺ neuronal populations (Figures 4G and 4H), while GFAP⁺ cells were more frequent among the EYFP⁺ population in the AAV-2-retro group than in the AAV-DJ group (Figures 4I and 4J). In addition, we examined the specificity of AAV transduction in the choroid plexus. During development, the choroid plexus arises from the most dorsal region of the neural tube under the influence of bone morphogenetic protein (BMP) and downstream Wnt signaling (Figure S5A),²³ and BMP4 has been reported to induce choroid plexus epithelial cells from PSC-derived neuroepithelial progenitors.^22,24,25 Consistent with these findings, hThOs generated by our BMP-based differentiation protocol contained small cell clusters expressing the choroid plexus marker prealbumin/transthyretin (TTR) outside the thalamic region on day 60 (Figure S5B). We, therefore, quantified the relative EYFP⁺ cell area within TTR⁺ domains as a measure of choroid plexus transduction for the three serotypes and found that AAV-Cap-Mac exhibited a higher tropism for choroid plexus-like structures than AAV-DJ and AAV-2-retro (Figure S5C). Thus, AAV-DJ exhibited a higher tropism for neurons in hThOs, whereas AAV-2-retro demonstrated tropism for both neurons and glia, and AAV-Cap-Mac exhibited a broader tropism in hThOs, including neurons, glia, and choroid plexus epithelial cells.

Subsequently, to characterize functionality in AAV-transduced hThOs, we added AAV-X-hSyn-jGCaMP7f at matched titers on day 72 and assessed neural activity of the transduced organoids using two-photon Ca²⁺ imaging on day 83 (Figure 4K). All serotypes successfully transduced hThO cells, as evidenced by broad jGCaMP7f expression (Figure 4L), with the AAV-Cap-Mac vector again labeling the largest number of neurons, followed by AAV-DJ and AAV-2-retro; all achieved higher values than AAV-9 (Figure 4M). Spontaneous Ca²⁺ transients were observed in subsets of labeled neurons (Figure 4N), with AAV-DJ and AAV-2-retro groups showing higher fractions of active neurons than the AAV-9 and AAV-Cap-Mac groups (Figure 4O). Furthermore, AAV-DJ-transduced neurons showed the largest signal amplitude among all AAV serotypes examined (Figure 4P). Taken together, AAV-DJ provided the most favorable balance of neuronal transduction and stable neural activity in hThOs. Although AAV-Cap-Mac yielded the broadest neuronal labeling, many labeled neurons were functionally inactive.

To evaluate serotype-specific toxicity in hThOs, we again focused on the three serotypes with the highest transduction efficiency (AAV-DJ, AAV-2-retro, or AAV-Cap-Mac). These AAVs were introduced into hThOs at an equivalent genome titer on day 55, and preparations were subsequently immunostained for EYFP as an indicator of AAV transduction and cleaved caspase-3 as a marker of apoptotic cell death, on day 70 (Figure 4Q). AAV-Cap-Mac induced greater cytotoxicity than AAV-DJ and AAV-2-retro, consistent with the reduced functional activity (Figures 4R and 4S). Taken together, these results indicate that AAV-DJ provides the best overall balance of efficient neuronal transduction, low cytotoxicity, and preservation of neural function in hThOs.

Monosynaptic rabies virus tracing combined with AAV-DJ in human neural organoids

Having demonstrated the transduction efficiency and low cytotoxicity of AAV-DJ in human neural organoids, we next examined whether AAV-DJ could also be used for neural circuit analysis by combining AAV-DJ-mediated gene delivery with circuit tracing using G-deleted rabies virus vectors (RVΔG), which label presynaptic neurons monosynaptically connected to the target neurons.^26,27 To analyze synaptic connectivity in hCOs, we first transduced organoids with AAV-DJ-hSyn-TVA-mCherry or AAV-DJ-hSyn-TVA-mCherry-P2A-oG (red), followed by the application of the envelope glycoprotein of subgroup A avian sarcoma and leukosis virus (EnvA)-pseudotyped RVΔG-EGFP (green). In this system, EGFP⁺mCherry⁺ cells are “starter cells” directly infected by EnvA-pseudotyped RVΔG through the receptor for subgroup A avian sarcoma and leukosis virus (TVA), whereas EGFP⁺mCherry⁻ cells are presynaptic input neurons labeled via monosynaptic viral spread in the presence of the optimized rabies virus glycoprotein (oG) (Figure 5A). To histologically validate this tracing strategy, we sectioned hCOs and performed immunohistochemistry 10 days after RVΔG application (Figure 5B). Both EGFP⁺ mCherry⁺ starter cells and EGFP⁺mCherry⁻ traced cells were observed within hCOs (Figure 5C). The EGFP⁺ cells were also positive for the neuron marker NeuN but negative for the astrocyte marker S100β (Figure 5C), indicating that rabies-mediated labeling occurred predominantly in the neurons. Next, we quantified the starter cells and traced presynaptic input cells, using flow cytometry (Figure 5D). In the control condition (TVA without oG), only a few EGFP⁺mCherry⁻ cells were observed, indicating no viral spread in the absence of oG. In contrast, the number of EGFP⁺mCherry⁻ cells was significantly enhanced in the presence of both TVA and oG (TVA + oG), indicating the transsynaptic spread of RVΔG from EGFP⁺mCherry⁺ starter cells to presynaptic cells following oG transcomplementation (Figure 5E). Moreover, in the TVA + oG condition, organoids containing more EGFP⁺mCherry⁺ starter cells also contained more EGFP⁺mCherry⁻-traced presynaptic neurons, whereas organoids with fewer starter cells showed fewer traced presynaptic neurons (Figure 5F). These results indicate that transsynaptic labeling scaled with the number of starter cells, as expected for monosynaptic rabies virus tracing. To normalize the tracing efficiency, we calculated the convergence index, defined as the ratio of EGFP⁺mCherry⁻ traced presynaptic cells to EGFP⁺mCherry⁺ starter cells. The convergence index was significantly higher in hCOs expressing both TVA and oG than in those expressing TVA alone (Figure 5G). Together, these results indicate that neurons within hCOs formed synaptically connected neural circuits that can be mapped by monosynaptic rabies virus tracing. Thus, AAV-DJ is a versatile vector not only for gene delivery and functional imaging but also for neural circuit tracing in human organoids.

Figure 5.

Monosynaptic rabies virus tracing in hCOs using AAV-DJ

(A) Experimental scheme for monosynaptic rabies virus tracing in hCOs. hCOs were transduced with AAV-DJ-hSyn-TVA-mCherry or AAV-DJ-hSyn-TVA-mCherry-P2A-oG, followed by infection with EnvA-RVΔG-EGFP. In this system, EGFP⁺mCherry⁺ cells are starter cells, whereas EGFP⁺mCherry⁻ cells are presynaptic input neurons labeled via RVΔG spread in the presence of oG.

(B) hCOs were transduced with AAV-DJ-hSyn-TVA-mCherry or AAV-DJ-hSyn-TVA-mCherry-P2A-oG, followed by infection with EnvA-RVΔG-EGFP.

(C) Representative immunofluorescence images of hCOs 10 days after EnvA-RVΔG-EGFP application, showing EGFP⁺mCherry⁺ starter cells and EGFP⁺ cells also positive for the neuron marker NeuN but not for the astrocyte marker S100β. Scale bars, 20 μm.

(D) Flow cytometric quantification of EGFP⁺mCherry⁺ starter cells and EGFP⁺mCherry⁻ traced presynaptic cells in hCOs under TVA alone and TVA + oG conditions.

(E) Plots of hCO cells under TVA alone and TVA + oG conditions.

(F) Quantification of EGFP⁺mCherry⁻-traced presynaptic cells. Under the TVA + oG condition, the number of traced presynaptic cells was significantly greater than that under the TVA alone condition.

(G) Convergence index (defined as the ratio of EGFP⁺mCherry⁻-traced presynaptic cells to EGFP⁺mCherry⁺ starter cells). n = 9 organoids from 3 sets of experiments. ∗∗∗p < 0.001, by Mann-Whitney U test.

All-optical physiology of human neural organoids with AAV-DJ

To determine whether AAV-DJ can be used not only to monitor but also to control neuronal activity in human neural organoids, we combined the red-shifted channelrhodopsin ChRmine with the Ca²⁺ indicator jGCaMP7f.^28,29 Because optogenetic stimulation and all-optical physiology require neurons with well-developed membrane excitability, synaptic function, and network integration, we performed these experiments in day 130 hCOs, which represent a more mature stage of development than the day 83 organoids used for Ca²⁺ imaging. hCOs were co-transduced with AAV-DJ-CaMKIIα-jGCaMP7f-WPRE and AAV-DJ-CAG-ChRmine-mScarlet-Kv2.1-WPRE on day 122, and one-photon stimulation was delivered during two-photon Ca²⁺ imaging on day 130 (Figure 6A). Within hCOs, two classes of excitatory neurons could be distinguished by the viral expression patterns (Figure 6B)—neurons expressing jGCaMP7f alone and neurons co-expressing jGCaMP7f and ChRmine-mScarlet. While jGCaMP7f-expressing neurons showed spontaneous Ca²⁺ activity, only neurons co-expressing ChRmine-mScarlet reliably exhibited Ca²⁺ transients in response to 561 nm light stimulation (Figure 6C). Furthermore, synchronous neural firing was induced in multiple neurons within the same field of view upon 561 nm light exposure (Figure 6D). To verify that these neuronal responses were evoked by optogenetic stimulation rather than light exposure itself, we performed a control experiment using AAV-DJ-CAG-mScarlet-Kv2.1-WPRE, which does not express ChRmine (Figure 6E). When hCOs were transduced with an AAV carrying only mScarlet without ChRmine (control) (Figure 6F), no Ca²⁺ transients were evoked by 561 nm light stimulation despite ongoing spontaneous transients (Figure 6G). This absence of photobleaching, phototoxicity, or loss of responsiveness after repeated photostimulation indicates that AAV-DJ-mediated delivery of ChRmine and jGCaMP7f is well tolerated for optogenetics and two-photon imaging. Collectively, these results show that AAV-DJ is a versatile vector for simultaneous optical manipulation and monitoring of neuronal activity in human neural organoids.

Figure 6.

AAV-DJ enables one-photon optogenetic stimulation during two-photon imaging of hCOs

(A) Schematic for viral labeling and all-optical interrogation in the AAV-DJ-transduced neural organoids. Cortical organoids were transduced with AAV-DJ-CAG-ChRmine-mScarlet and AAV-DJ-CaMKⅡα-jGCaMP7f on day 122, and two-photon imaging at 920 nm and simultaneous one-photon stimulation at 561 nm were conducted on day 130.

(B) Representative images of cells expressing jGCaMP7f and ChRmine-mScarlet in AAV-DJ-transduced hCOs on day 130. Scale bars: 50 μm.

(C) jGCaMP7f-expressing cells in hCOs responding to photostimulation at 561 nm. Scale bars: 20 μm.

(D) Representative traces of jGCaMP7f signals in response to 561 nm photostimulation in hCOs transduced with AAV-DJ-CaMKⅡα-jGCaMP7f and AAV-DJ-CAG-ChRmine-mScarlet-Kv2.1.

(E) Schematic for viral labeling and all-optical interrogation in AAV-DJ-transduced neural organoids. Cortical organoids were co-transduced with AAV-DJ-CaMKⅡα-jGCaMP7f and either AAV-DJ-CAG-mScarlet-Kv2.1 (Control) or AAV-DJ-CAG-ChRmine-mScarlet-Kv2.1 (ChRmine), followed 8 days later by two-photon imaging of jGCaMP7f at 920 nm and one-photon stimulation at 561 nm.

(F) Representative images of cells expressing jGCaMP7f and mScarlet in AAV-DJ-transduced hCOs on day 130. Scale bars: 50 μm.

(G) Representative traces of jGCaMP7f signals in response to 561 nm photostimulation in hCOs co-transduced with AAV-DJ-CaMKⅡα-jGCaMP7f and either AAV-DJ-CAG-mScarlet (Control) or AAV-DJ-CAG-ChRmine-mScarlet (ChRmine).

Discussion

We systematically compared the transduction efficiency, cytotoxicity, and experimental utility of seven AAV serotypes in hCOs and hThOs, which represent two regional brain models with different developmental trajectories and cellular compositions. We identified a capsid-dependent trade-off between transduction efficiency and preservation of neural functional integrity. Among the serotypes tested, AAV-DJ offered the most favorable profile, enabling efficient neuronal transduction with minimal cytotoxicity and reliable preservation of neuronal function in both preparations, while AAV-Cap-Mac provided the highest overall transduction efficiency and broad tropism, but at the cost of increased cytotoxicity and impaired neuronal activity. These findings demonstrate that the suitability of viral vectors for human neural organoid analyses cannot be judged on transduction efficiency alone but must also be evaluated for how faithfully they preserve the biological state of the tissue.

The trade-off between the transduction efficacy and preservation of cellular functions is best illustrated by the contrasting performance of AAV-Cap-Mac versus AAV-DJ. The AAV-Cap-Mac vector showed the highest transduction efficiency in both cortical and thalamic organoids, with broad tropism, including neurons, glia, and choroid plexus epithelial cells. However, this efficient transduction was accompanied by increased cytotoxicity, as evidenced by a higher proportion of cleaved caspase-3-positive cells among the transduced cells. Critically, AAV-Cap-Mac toxicity was associated with reduced neuronal signaling activity, as indicated by smaller Ca²⁺ transient amplitudes and more numerous inactive cells. These observations align with previous reports on the serotype- and dose-dependent toxicity of AAVs in neural tissues.²⁰ In contrast, AAV-DJ showed high neuronal transduction efficiency, with substantially lower cytotoxicity and superior preservation of neuronal signaling function, as evidenced by larger spontaneous and optogenetically evoked Ca²⁺ transients (Table S3). The AAV-2-retro displayed intermediate properties, with tropism to both neurons and glia and functional performance preservation between that of AAV-DJ and AAV-Cap-Mac. Collectively, these results demonstrate that the viral vector choice is a fundamental determinant of experimental interpretability in studies of neural organoids. Among the serotypes tested, AAV-DJ was identified as being optimal for experiments requiring a precise analysis of neural dynamics and circuit-level activity manipulation.

The versatility of AAV-DJ can be extended beyond basic gene delivery. In addition to supporting functional Ca²⁺ imaging and all-optical physiology in human neural organoids, AAV-DJ enabled monosynaptic rabies virus tracing. By combining AAV-DJ-mediated delivery of TVA and oG with EnvA-pseudotyped RVΔG, we identified starter cells and traced presynaptic input neurons and, thereby providing direct evidence that hCOs contain synaptically connected neuronal networks. The value of AAV-DJ lies in its compatibility with a rapidly expanding set of molecular and circuit-level tools.^7,9 Recent advances in synthetic promoters, cell-type-specific enhancers, CRISPR-based genome engineering, RNA editing, and related targeting strategies are making it increasingly feasible to gain precise access to defined cell types.^{30,31,32,33,34,35} In particular, the growing repertoire of cell-type-specific enhancers offers a powerful strategy for restricting transgene expression to genetically defined cell types by using AAV-based systems. In addition, newly developed genetically encoded sensors and actuators for neurotransmitters, neuromodulators, membrane potentials, intracellular signaling pathways, and metabolic states are expanding the range of physiological variables that can be measured and manipulated in living tissues.^36,37,38 Combining AAV-DJ with such regulatory elements, sensors, and actuators should enable precise, subtype-resolved interrogation of human organoids, facilitating the dissection of cell-type-specific contributions to network activity, connectivity, and disease-relevant phenotypes. These capabilities suggest that genetically accessible organoids can evolve into experimentally tractable systems in which cell type identity, activity, connectivity, and molecular signaling can be monitored and manipulated, enabling molecular-, cellular-, and circuit-level dissection of the human brain in vitro.

These advances would help address a major bottleneck in human systems neuroscience. Many lines of evidence from animal models have demonstrated that neuronal circuits develop through a combination of genetic programs and activity-dependent mechanisms.²¹ However, distinguishing these contributions in the human brain has remained difficult due to ethical constraints and the lack of experimentally accessible model systems that allow manipulation at the molecular, cellular, and circuit levels. Human neural organoids have the potential to bridge this gap, but exploiting their full potential depends on genetic tools that allow precisely targeted manipulations without disrupting non-target functions. The methodology described here, which allows for high-resolution functional imaging, physiological manipulation, and circuit tracing in human neural organoid models using optimized AAV-mediated delivery, expands the experimental toolkit for organoid-based neuroscience and advances organoids as platforms for mechanistic analysis of the human brain. The present study will facilitate future research to determine the mechanisms of neurodevelopment, dissect human-specific neural circuit functions, investigate disease-associated phenotypes, and test therapeutic interventions in a three-dimensional, patient-relevant human context.^1,2,28,29

In conclusion, these comparisons highlight AAV-DJ as a versatile and functionally robust vector for studying the structure, function, development, and pathology of the human brain and the underlying molecular, cellular, and circuit-level mechanisms in human organoid models. Our study establishes a generalizable benchmark linking transduction, tropism, cytotoxicity, and physiological readouts in human organoids, thereby providing a practical framework for AAV-mediated functional interrogation and disease modeling in organoid-based neuroscience. Future studies integrating higher-resolution approaches, such as cell-type-specific enhancers, genetically encoded sensors and actuators, genome editing, single-cell transcriptomics, spatial transcriptomics, and multiplexed functional assays, will further refine our understanding of the human brain in both health and disease.^{31,32,37,38,39,40} These advances will open new opportunities for mechanistic and translational research in humans.

Limitations of the study

Several limitations warrant consideration. First, although we compared distinct capsids at matched genome titers, the optimal dose and relative performance of each serotype may vary with the promoter chosen, transgene design, cell type, and developmental stage of the organoid (Figure S6). The use of a ubiquitous CAG promoter, in particular, may introduce variability in the expression levels across cell types and maturation states, although this strategy was chosen to minimize cell type bias when comparing serotypes. Second, organoid heterogeneity represents an inherent challenge. Early-stage organoids are enriched in neural progenitors and exhibit prominent rosette structures, whereas later-stage organoids contain a higher proportion of differentiated neurons and glial cells. Therefore, variability in the cellular composition could affect the apparent transduction efficiency. To reduce this source of variation, we focused our analyses primarily on more mature organoids, which are more relevant for functional studies of neural circuits. Third, cell type tropism was inferred from marker-based analysis of the EYFP⁺ transduced population, which does not fully resolve the cellular diversity present in human neural organoids and may be influenced by organoid-to-organoid variability. Higher-resolution approaches, including single-cell RNA sequencing and spatial transcriptomics, are important for more precisely defining cell-type-specific AAV tropism. Finally, our analyses were performed within discrete post-transduction windows. Although our additional time-course analysis at days 10 and 17 revealed delayed cytotoxicity specifically following AAV-Cap-Mac transduction, earlier or longer-term analyses are important to more fully define the temporal profile of AAV-associated toxicity, transgene stability, and maturation-dependent changes in tropism.

Resource availability

Lead contact

Requests for further information, resources, and reagents should be directed to the lead contact, Fumitaka Osakada (osakada.fumitaka.x3@f.mail.nagoya-u.ac.jp).

Materials availability

All plasmids generated in this study are available from the lead contact upon reasonable request, with a completed material(s) transfer agreement.

Data and code availability

• •All data reported in this study will be shared by the lead contact upon reasonable request.
• •This study does not report original code.
• •Any additional information required to reanalyze the data reported in this study is available from the lead contact upon request.

Acknowledgments

We thank the members of the Osakada laboratory for their valuable discussions. This work was supported by Grants-in-Aid from the Japan Society for the Promotion of Science (20K21476, 23K24034, and 25K02553 to F.O.); CREST from the Japan Science and Technology Agency (JPMJCR1851 and JPMJCR25B3 to F.O.); Brain/MINDS 1.0 and 2.0; CREST from the Japan Agency for Medical Research and Development (JP20dm0207058, JP23gm1510011, and JP24wm0625110 to F.O.), SRF Foundation (to F.O.), and Hoansha Foundation (to F.O.). S.A. was supported by the CIBoG WISE Program and THERS Make New Standards Program for the Next Generation Researchers from MEXT.

Author contributions

Conceptualization, S.A. and F.O.; methodology, S.A. and M.N.; investigation, S.A., M.N., and K.Y.; visualization, S.A. and R.F.T.; supervision, F.O.; validation, S.A. and M.N.; writing – original draft, S.A.; and writing – review & editing, S.A., M.N., A.Y.S., R.F.T., and F.O.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
rabbit anti-FOXG1	Takara	Cat# M227; RRID: AB_2827749
mouse anti-N-CADHERIN	Cell Signaling Technology	Cat# 14215; RRID: AB_2798427
rabbit anti-PAX6	Abcam	Cat# ab195045; RRID: AB_2750924
chicken anti-MAP2	Abcam	Cat# ab5392; RRID: AB_2138153
rabbit anti-NeuN	Cell Signaling Technology	Cat# 24307; RRID: AB_2651140
rat anti-CTIP2	Abcam	Cat# ab18465; RRID: AB_2064130
rabbit anti-TBR1	Cell Signaling Technology	Cat# 49661; RRID: AB_2799364
mouse anti-GFAP	Cell Signaling Technology	Cat# 12389; RRID: AB_2631098
rabbit anti-S100β	Sigma-Aldrich	Cat# S2532; RRID: AB_477499
rabbit anti-OLIG2	Abcam	Cat# ab109186; RRID: AB_10861310
rabbit anti-GFP	Abcam	Cat# ab6556; RRID: AB_305564
rat anti-GFP	Nacalai Tesque	Cat# 04404-84; RRID: AB_10013361
chicken anti-GFP	Abcam	Cat# ab13970; RRID: AB_300798
rat anti-mCherry	Thermo Fisher Scientific	Cat# M11217; RRID: AB_2536611
rabbit anti-Cleaved Caspase-3	BD Biosciences	Cat# 559565; RRID: AB_397274
rabbit anti-TCF7L2	Cell Signaling Technology	Cat# 2569; RRID: AB_2199816
mouse anti-TTR	Santa Cruz Biotechnology	Cat# sc-377517; RRID: AB_11150473
DAPI	Wako	Cat# 049-18801
Alexa Fluor 488 donkey anti-chicken IgY	Jackson ImmunoResearch Labs	Cat# 703-545-155; RRID: AB_2340375
Alexa Fluor 488 donkey anti-mouse IgG	Jackson ImmunoResearch Labs	Cat# 715-545-151; RRID: AB_2341099
Alexa Fluor 488 donkey anti-rabbit IgG	Jackson ImmunoResearch Labs	Cat# 711-545-152; RRID: AB_2313584
Alexa Fluor 488 donkey anti-rat IgG	Jackson ImmunoResearch Labs	Cat# 712-545-153; RRID: AB_2340684
Alexa Fluor 594 donkey anti-mouse IgG	Jackson ImmunoResearch Labs	Cat# 715-585-151; RRID: AB_2340855
Alexa Fluor 594 donkey anti-rabbit IgG	Jackson ImmunoResearch Labs	Cat# 711-585-152; RRID: AB_2340621
Alexa Fluor 594 donkey anti-rat IgG	Jackson ImmunoResearch Labs	Cat# 712-585-153; RRID: AB_2340689
Alexa Fluor 647 donkey anti-chicken IgY	Jackson ImmunoResearch Labs	Cat# 703-605-155; RRID: AB_2340379
Alexa Fluor 647 donkey anti-mouse IgG	Jackson ImmunoResearch Labs	Cat# 715-605-151; RRID: AB_2340863
Alexa Fluor 647 donkey anti-rabbit IgG	Jackson ImmunoResearch Labs	Cat# 711-605-152; RRID: AB_2492288
Bacterial and virus strains
AAV-2-CAG-EYFP-WPRE3	This paper	N/A
AAV-6-CAG-EYFP-WPRE3	This paper	N/A
AAV-9-CAG-EYFP-WPRE3	This paper	N/A
AAV-DJ-CAG-EYFP-WPRE3	This paper	N/A
AAV-PHP.eB-CAG-EYFP-WPRE3	This paper	N/A
AAV-2-retro-CAG-EYFP-WPRE3	This paper	N/A
AAV-Cap-Mac-CAG-EYFP-WPRE3	This paper	N/A
AAV-DJ-hSyn-EYFP-WPRE	This paper	N/A
AAV-9-hSyn-jGCaMP7f-WPRE	This paper	N/A
AAV-DJ-hSyn-jGCaMP7f-WPRE	This paper	N/A
AAV-2-retro-hSyn-jGCaMP7f-WPRE	This paper	N/A
AAV-Cap-Mac-hSyn-jGCaMP7f-WPRE	This paper	N/A
AAV-DJ-hSyn-TVA950-mCherry-WPRE	This paper	N/A
AAV-DJ-hSyn-TVA950-mCherry-P2A-oG-WPRE	This paper	N/A
AAV-DJ-CAG-mScarlet-Kv2.1-WPRE	This paper	N/A
AAV-DJ-CAG-ChRmine-mScarlet-Kv2.1-WPRE	This paper	N/A
AAV-DJ-CaMKⅡα-jGCaMP7f-WPRE	This paper	N/A
EnvA-SAD-B19ΔG-EGFP	This paper	N/A
Chemicals, peptides, and recombinant proteins
PBS	Wako	Cat# 045-29795
Fetal bovine serum	Corning	Cat# 35-079-CF
Fetal bovine serum	Sigma-Aldrich	Cat# 173012
iMatrix-511	Nippi	Cat# 892021
StemFit AK02N medium	Ajinomoto	Cat# RCAK02N
Accutase	Nacalai Tesque	Cat# 12679-54
Y-27632	MedChemExpress	Cat# HY-10583
N2 supplement	Wako	Cat# 141-08941
Penicillin-Streptomycin	Wako	Cat# 168-23191
MEM-NEAA	Wako	Cat# 139-15651
L-Glutamine	Wako	Cat# 073-05391
Insulin	Wako	Cat# 099-06473
D-Glucose	Nacalai Tesque	Cat# 16806-25
PD0325901	Wako	Cat# 162-25291
BMP7	PeproTech	Cat# 120-03P
Neurobasal medium	Thermo Fisher Scientific	Cat# 21103049
B27 supplement minus vitamin A	Thermo Fisher Scientific	Cat# 12587001
BDNF	PeproTech	Cat# 452-02
Ascorbic Acid	Wako	Cat# 013-12061
IWR-1-endo	Wako	Cat# 037-25131
A-83-01	Wako	Cat# 035-24113
LDN-193189	Sigma-Aldrich	Cat# SML0559
2-Mercaptoethanol	Wako	Cat# 133-14571
bFGF	PEPRO TECH	Cat# 100-18B
KSR	Thermo Fisher Scientific	Cat# 10828028
Polyethylenimine (PEI)	Polyscience	Cat# 24765-2
Proteinase K	Wako	Cat# 161-28701
DNase I	Sigma-Aldrich	Cat# D4263
Triton X-100	Wako	Cat# 167-21101
Blocking One	Nacalai Tesque	Cat# 03953-95
Sucrose	Wako	Cat# 190-00013
Nitrilotriethanol	Wako	Cat# 145-05605
(Ethylenedinitrilo)tetrakis(2-propanol)	Wako	Cat# 201-13145
Urea	Wako	Cat# 219-00175
DMEM/F12	Sigma-Aldrich	Cat# D8062
DMEM	Wako	Cat# 043-30085
0.25% Trypsin-EDTA	Wako	Cat# 201-16945
PrimeSTAR Max DNA Polymerase	Takara	Cat# R045A
EmeraldAmp Max PCR Master Mix	Takara	Cat# RR320
Critical commercial assays
TB Green Fast qPCR Mix	Takara	Cat# RR430B
CytoFLEX Daily QC Fluorospheres	Beckman	Cat# B53230
Experimental models: Cell lines
HEK293T	RIKEN Bioresource Center Cell Bank	RRID: CVCL_0063
Human iPSC line	Kyoto University	Clone: 1383D6
B7GG	Salk Institute for Biological Studies	N/A
BHK-EnvA	Salk Institute for Biological Studies	N/A
HEK-TVA800	Salk Institute for Biological Studies	N/A
Experimental models: Organisms/strains
E. coli DH5α	Toyobo	Cat# DNA-903
E. coli Stbl3	Thermo Fisher Scientific	Cat# C737303
E. coli XL10-gold	Agilent	Cat# 200314
Oligonucleotides
Primer: human_GAPDH_F (5′- TCACCAGGGCTGCTTTTAAC -3′)	Sigma-Aldrich	N/A
Primer: human_GAPDH_R (5′- TGACGGTGCCATGGAATTTG -3′)	Sigma-Aldrich	N/A
Primer: human_OCT4_F (5′- GCAAAACCCGGAGGAGGAGTC -3′)	Sigma-Aldrich	N/A
Primer: human_OCT4_R (5′- CCACATCGGCCTGTGTATATC -3′)	Sigma-Aldrich	N/A
Primer: human_PAX6_F (5′- TCACCATGGCAAATAACCTG -3′)	Sigma-Aldrich	N/A
Primer: human_PAX6_R (5′- CAGCATGCAGGAGTATGAGG -3′)	Sigma-Aldrich	N/A
Primer: human_FOXG1_F (5′- ATGATCCCCAAGTCCTCGTT -3′)	Sigma-Aldrich	N/A
Primer: human_FOXG1_R (5′- GTGGTGGTTGTCGTTCTGG -3′)	Sigma-Aldrich	N/A
Primer: human_EMX1_F (5′- CCGGGACCCTTCTCCATTTCT -3′)	Sigma-Aldrich	N/A
Primer: human_EMX1_R (5′- ACGTCGCTGGCCTGGAA -3′)	Sigma-Aldrich	N/A
Primer: WPRE_F (5′- GCTGTTGGGCACTGACAATT -3′)	Sigma-Aldrich	N/A
Primer: WPRE_R (5′- CCACATAGCGTAAAAGGAGCAA -3′)	Sigma-Aldrich	N/A
Primer: WPRE3_F (5′- CTTCCCGTATGGCTTTCATTTT -3′)	Sigma-Aldrich	N/A
Primer: WPRE3_R (5′- GCCGTGGCAAGAACTAACCA -3′)	Sigma-Aldrich	N/A
Recombinant DNA
pAAV-CAG-EYFP-WPRE3	This paper	N/A
pGP-CaMKⅡα-jGCaMP7f-WPRE	This paper	N/A
pDP2	Grimm et al.18	N/A
pDP6	Grimm et al.18	N/A
pDP9	Grimm et al.18	N/A
pAAV2/2-retro-rep/cap	Addgene	Cat# 81070
pAAV2/DJ-rep/cap	Cell Bio Labs	Cat# VPK-420-DJ
pAAV2/PHP.eB-rep/cap (pUCmini-iCAP-PHP.eB)	Addgene	Cat# 103005
pAAV2/Cap-Mac-rep/cap (pUCmini-iCAP-AAV.CAP-Mac)	Addgene	Cat# 200658
pHelper	Cell Bio Labs	Cat# 340202
pAAV-hSyn-TVA950-mCherry-WPRE	Nishimura et al.6	N/A
pAAV-hSyn-TVA950-mCherry-P2A-oG-WPRE	Nishimura et al.6	N/A
pAAV-CAG-mScarlet-Kv2.1-WPRE	This paper	N/A
pAAV-CAG-ChRmine-mScarlet-Kv2.1-WPRE	Takeuchi et al.41	N/A
pSADB19ΔG-EGFP	Osakada et al.27	N/A
Software and algorithms
Fiji (ImageJ)	NIH	RRID: SCR_002285
MATLAB	MathWorks	RRID: SCR_001622

Experimental model and study participant details

All experiments were approved by Nagoya University and performed in accordance with institutional guidelines.

Human iPSC culture

Human induced pluripotent stem cells (hiPSCs) originally established at Kyoto University were a gift from Drs. Masato Nakagawa and Shinya Yamanaka, CiRA, Kyoto University, and are also available from the RIKEN Cell Bank (HPS1006). All procedures involving hiPSCs were conducted following institutional guidelines and approved ethical protocols using de-identified donor material. Cells were maintained on culture dishes coated with iMatrix511 (Nippi Inc., Tokyo, Japan) in StemFit AK02N medium (Ajinomoto, Tokyo, Japan) at 37°C under a humidified atmosphere containing 5% CO₂, as previously described.^6,42 At near-confluence, hiPSCs were dissociated into single cells using Accutase (Nacalai Tesque, Kyoto, Japan) and reseeded at 1.1 × 10³ cells/cm² on iMatrix511-coated dishes in StemFit AK02N supplemented with 10 μM Y-27632 (Wako, Osaka, Japan). The medium was replaced daily with Y-27632-free StemFit AK02 from the next day onward.

Generation of human cortical and thalamic organoids

Human cortical organoids (hCOs) were generated from hiPSCs using an SFEBq-based protocol.^4,6,43 Subconfluent hiPSCs were dissociated into single cells using Accutase and seeded into low-adhesion 96-well V-bottom plates at 1.0 × 10⁴ cells per well in DMEM/F12 + KSR medium (DMEM/F12, 15% (v/v) KSR, 1% (v/v) MEM-NEAA, 1% (v/v) L-glutamine, and 100 μM β-mercaptoethanol). From day 0, the medium also contained 50 μM Y-27632, 500 nM A-83-01, 100 nM LDN-193189, and 2 μM IWR-1-endo. Half of the medium was replaced every 2 days. From day 4 onward, Y-27632 was removed from the replacement medium. On day 10, the organoids were transferred to low-adhesion 6-cm dishes and cultured on an orbital shaker at 70 rpm. From days 10–18, organoids were maintained in a 1:1 mixture of DMEM/F12 and Neurobasal medium supplemented with 1% (v/v) B27 supplement (minus vitamin A), 0.5% (v/v) N2 supplement, 1% (v/v) L-glutamine, 0.5% (v/v) MEM-NEAA, 0.025% (v/v) insulin, 50 μM β-mercaptoethanol, 1% (v/v) penicillin/streptomycin, and 20 ng/mL basic FGF. From day 18 onward, organoids were cultured in DMEM/F12 + Neurobasal medium supplemented with 20 ng/mL BDNF and 200 μM ascorbic acid. After 18 days, the medium was completely replaced every 2 days.

Human thalamic organoids (hThOs) were also produced from hiPSCs using an SFEBq-based protocol.^22,43 Subconfluent hiPSCs were dissociated into single cells using Accutase and seeded into low-adhesion 96-well V-bottom plates at 1.0 × 10⁴ cells per well in DMEM/F12 + KSR medium (DMEM/F12, 15% (v/v) KSR, 1% (v/v) MEM-NEAA, 1% (v/v) L-glutamine, and 100 μM β-mercaptoethanol) supplemented with 50 μM Y-27632, 500 nM A-83-01, 100 nM LDN-193189, and 4 μg/mL insulin. Half of the medium was replaced every 2 days. From day 4 onward, Y-27632 was removed from the replacement medium. On day 8, the organoids were transferred to low-adhesion 6-cm dishes and cultured on an orbital shaker at 70 rpm. These organoids were then maintained in DMEM/F12 (2% (v/v) B27 supplement (minus vitamin A), 1% (v/v) N2 supplement, 24.8 mM D(+)-glucose, and 100 μM β-mercaptoethanol) containing 30 ng/mL BMP7 and 1 μM PD0325901. This medium was used until day 16, with complete replacement every 2 days. From day 16 onward, organoids were cultured in DMEM/F12 + Neurobasal medium (1:1) supplemented with 20 ng/mL BDNF and 200 μM ascorbic acid, with replacement every 2 days.

Method details

Quantitative reverse-transcription polymerase chain reaction

Gene expression levels were assessed by RT-qPCR as previously described.^6,42,43,44 Briefly, total RNA was extracted and purified using the FavorPrep Tissue Total RNA Purification Mini Kit. Subsequently, 500 ng of total RNA was reverse-transcribed into cDNA using oligo(dT) primers and PrimeScript RT Master Mix (Takara, Kyoto, Japan). Expression levels of target genes were then measured by RT-qPCR using gene-specific primers and TB Green Fast qPCR Mix (Takara) on the LightCycler system (Roche, Basel, Switzerland). Primer sequences used in this study are listed in Key Resources Table (KRT).

Organoid sectioning and immunostaining

Organoids were fixed with 4% paraformaldehyde (PFA) in PBS (PFA/PBS) and processed for cryosectioning as described previously.^6,42,43,44 Briefly, fixed organoids were transferred sequentially to 15% (v/v) sucrose/PBS overnight and 30% (v/v) sucrose/PBS overnight for cryoprotection, embedded in a 2:1 mixture of OCT compound (Sakura Finetek, Torrance, CA, USA) and 30% (v/v) sucrose/PBS, sectioned at 20–40 μm on a cryostat (CM3050S, Leica, Wetzlar, Germany), and mounted on glass slides.

Immunostaining was performed as previously described.^44,45 Sections were rinsed with PBS, incubated with a mixture of Blocking One (Nacalai Tesque) and 0.1% (v/v) Triton/PBS, and incubated with the indicated primary antibodies at 4°C overnight. Sections were then washed with PBS three times and incubated with secondary antibodies for 2 h at room temperature. The following primary antibodies and working dilutions were used: anti-FOXG1 (1:500, rabbit; Takara), anti-N-CADHERIN (1:200, mouse; Cell Signaling Technology, Danvers, MA, USA), anti-PAX6 (1:600, rabbit; Abcam), anti-CTIP2 (1:500, rat; Abcam, Cambridge, UK), anti-TBR1 (1:400, rabbit; Cell Signaling Technology), anti-GFP (1:1,000, rabbit; Abcam), anti-GFP (1:1,000, rat; Nacalai), anti-GFP (1:1,000, chicken; Abcam), anti-mCherry (1:1,000, rat; Thermo Fisher Scientific), anti-MAP2 (1:500, chicken; Abcam), anti-NeuN (1:100, rabbit; Cell Signaling Technology), anti-GFAP (1:200, rabbit; Cell Signaling Technology), anti-S100β (1:1,000, mouse; Sigma-Aldrich), anti-OLIG2 (1:500, rabbit; Abcam), anti-cleaved caspase-3 (1:500, rabbit; BD Pharmingen, Franklin Lakes, NJ, USA), anti-TCF7L2 (1:500, rabbit; Cell Signaling Technology), and anti-TTR (Prealbumin) (1:50, mouse; Santa Cruz Biotechnology, Dallas, TX, USA) (Table S1). The secondary antibodies used were donkey anti-mouse IgG, anti-rabbit IgG, anti-rat IgG, or anti-chicken IgY labeled with Alexa Fluor 488, Alexa Fluor 594, or Alexa Fluor 647 (1:1,000; Jackson Laboratories, West Grove, PA, USA) (Table S1). Cell nuclei were counterstained with DAPI (1.25 μg/mL). Specimens were imaged using a confocal microscope (LSM800, Zeiss, Oberkochen, Germany) equipped with 10× (NA 0.45, Zeiss) and 20× (NA 0.75, Zeiss) objective lenses.

Tissue clearing and whole-mount immunohistochemistry

Organoids were also immunostained as whole-mount preparations after clearing using the Clear, Unobstructed Brain/Body Imaging Cocktails and Computational Analysis (CUBIC) method.^19,43 Briefly, organoids were fixed with PFA/PBS overnight at 4°C, washed three times with PBS, incubated in 50% (v/v) CUBIC Reagent-1 in H₂O for 3 h at 37°C followed by incubation in CUBIC Reagent-1 overnight at 37°C, and washed in PBS three times for 2 h each. For immunostaining, the organoids were incubated with PBS containing 0.1% Triton X-100, 5% Blocking One reagent, and anti-GFP antibody (1:1,000, rabbit, Abcam) for 48 h at 37°C under gentle agitation, washed three times with PBS for 2 h each, and incubated with Alexa Fluor 488-conjugated anti-rabbit IgG and DAPI (1.25 μg/mL) for 48 h at 37°C. For refractive index matching, organoids were incubated sequentially in 50% (v/v) CUBIC Reagent-2 in H₂O for 3 h at room temperature and then in 100% CUBIC Reagent-2 overnight. Cleared specimens were imaged on a confocal microscope with a 10× objective. Z-stacks were collected from the surface to a depth of 250 μm.

Adeno-associated virus production and transduction

Adeno-associated virus vectors (AAVs) were produced in HEK293T cells by transient transfection as previously described.³⁹ For production of AAV-2, AAV-6, and AAV-9, HEK293T cells were co-transfected with the corresponding pDPX vector and either the genomic vector pAAV-CAG-EYFP-WPRE3 or pAAV-hSyn-jGCaMP7f-WPRE.¹⁸ For production of AAV-DJ, AAV-2-retro, AAV-PHP.eB, and AAV-Cap-Mac, HEK293T cells were co-transfected with pHelper, the serotype-specific rep/cap vector, and either the genomic vector pAAV-CAG-EYFP-WPRE3 or pAAV-hSyn-jGCaMP7f-WPRE.^14,15,16,17 Transfected cells were cultured for 6 h in a humidified atmosphere of 5% CO₂ at 37°C, washed in PBS, cultured for 1 day, transferred to a humidified atmosphere of 3% CO₂ at 35°C, and cultured for another 2 days before harvesting. Viral particles were released by three freeze-thaw cycles, and cell debris was removed by centrifugation. Viral particles were isolated from the supernatant by iodixanol density gradient centrifugation (15%, 25%, 40%, and 58% OptiPrep; Serumwerk Bernburg, Bernburg, Germany) at 16,000 × g for 4 h at 4°C. The 40% iodixanol fraction containing the purified viral particles was collected, and AAV genome titers were measured using quantitative PCR (KRT). The aliquoted AAVs were stored at −80°C until further use. Similar yields were obtained across lots of the different AAVs.

For organoid transduction, 10 μL of AAV vector solution was added to 3 mL of culture medium per well in 6-well plates, with each well containing three organoids. Vectors were used at titers of 5.0 × 10¹¹ to 5.0 × 10¹² vector genomes (vg)/mL, corresponding to approximately 1.7 × 10⁹ to 1.7 × 10¹⁰ vg per organoid. Within each experiment, all AAV serotypes were applied at equivalent genome titers to enable direct comparison of transduction efficiency. The titers of all AAVs used in this study are listed in the Supplementary Information (Table S2).

G-deleted rabies virus production

EnvA-pseudotyped G-deficient rabies virus vector (EnvA-RVΔG) was prepared as previously described.^27,43,46 Briefly, RVΔG was rescued by transfecting B7GG cells with pcDNA-SAD-B19N, pcDNA-SAD-B19P, pcDNA-SAD-B19L, pcDNA-SAD-B19G, and rabies virus genome vector pSAD-B19ΔG-EGFP using Lipofectamine 2000 (Thermo, Waltham, MA. USA). For pseudotyping RVΔG with EnvA, BHK-EnvA cells were infected with unpseudotyped RVΔG. Virus-containing supernatant was collected and concentrated by two rounds of ultracentrifugation (Beckman Coulter, Brea, CA, USA). HEK293T cells were used to check for unpseudotyped RVΔG contamination. Virus solutions were aliquoted and stored at −80°C until use.

Organoid dissociation and flow cytometry

Organoids were dissociated into single cells as previously described.⁶ Briefly, organoids were collected, washed once with PBS, incubated in a 1:1 mixture of TrypLE Select (Thermo) and 1 mM EDTA solution for 30 min at 37°C, triturated into small cell aggregates by pipetting, and incubated for an additional 15 min at 37°C. The enzymatic reaction was stopped by adding FACS buffer supplemented with 2% (v/v) fetal bovine serum (FBS) and 10 μM Y-27632. Dissociated cells were centrifuged at 200 × g for 5 min, followed by resuspension of the cell pellet at 1.0 × 10⁶ cells/mL in pure FACS buffer. The suspension was filtered through a 40-μm cell strainer and immediately analyzed using an acoustic focusing flow cytometer (Attune NxT, Thermo). Debris was excluded by forward- and side-scatter gating. Cells expressing enhanced yellow fluorescent protein (EYFP) were counted under 488-nm laser excitation.

Monosynaptic rabies virus tracing

For retrograde monosynaptic tracing, hCOs were transduced with AAV-DJ-hSyn-TVA950-mCherry or AAV-DJ-hSyn-TVA950-mCherry-P2A-oG, followed by infection with EnvA-RVΔG-EGFP. Ten days after rabies virus application, hCOs were fixed overnight in 4% PFA/PBS at 4°C for immunostaining. For flow cytometry, hCOs were dissociated into single cells 10 days after rabies virus application, resuspended in FACS buffer, filtered through a 40-μm cell strainer, and immediately analyzed on a cell sorter (CytoFLEX SRT, Beckman Coulter). Debris was excluded based on forward scatter (FSC) and side scatter (SSC) profiles, and singlets were selected and doublets excluded by sequential gating. Fluorescence gates for enhanced green fluorescent protein (EGFP) and mCherry were defined using nontransduced control cells and single-color compensation controls. EGFP⁺mCherry⁺ cells were classified as starter cells, whereas EGFP⁺mCherry⁻ cells were classified as traced presynaptic input cells. Cell numbers in each population were quantified using CytExpert software (Beckman Coulter).

Two-photon calcium imaging

Two-photon calcium imaging of neural organoids was performed as previously described.^6,44 The lower half of each organoid was embedded in 4% low-melting point agarose and submerged in oxygenated artificial cerebrospinal fluid (ACSF) consisting of 125 mM NaCl, 2.5 mM KCl, 1.25 mM NaH₂PO₄⋅H₂O, 25 mM NaHCO₃, 25 mM D-glucose, 2 mM CaCl₂, and 1 mM MgCl₂. Imaging was performed using a two-photon fluorescence microscope equipped with a GaAsP nondescanned detector and resonant scanner (A1R-MP+, Nikon, Shinagawa, Japan).⁴⁷ jGCaMP7f was excited with a 920-nm femtosecond pulse laser (InSight DeepSee+, Spectra-Physics, Santa Clara, CA, USA). Time-lapse images were acquired using a 25× water-immersion objective lens (NA: 1.1, Nikon) at 512 × 512-pixel resolution and 30.0 Hz.

All image processing and statistical analyses were performed using MATLAB (R2022b, Mathworks, Natick, MA, USA). For processing of time-lapse data, cross correlation-based rigid image registration was performed to correct for image displacement caused by motion.⁴⁸ Fluorescent intensity fluctuations from individual cells were then quantified. Regions of interest (ROIs) corresponding to individual cells were semimanually defined using the “Cell Magic Wand” plugin of ImageJ.⁴⁹ Raw jGCaMP7f traces were smoothed using a Savitzky–Golay filter (filter width = 301 frames) to reduce high-frequency electrical noise while preserving the shape of calcium transients. The baseline for each cell (F₀) was computed using a low-pass percentile filter (10th percentile, cut-off frequency = 1/15 Hz), and the normalized fluorescence signal change was computed as ΔF/F₀ to evaluate the calcium signaling activity of each neuron. Active cells were classified based on the following three criteria calculated from ΔF/F₀: skewness of the ΔF/F₀ distribution >1.0, ΔF/F₀ standard deviation >0.04, and at least one calcium transient during the recording period. Calcium transients in jGCaMP7f traces were detected using the MATLAB findpeaks function with settings “MinPeakHeight = 0.2,” “MinPeakDistance = 30,” and “MinPeakWidth = 15”. Mean calcium transient amplitude was defined as the average peak height of all detected Ca²⁺ events during the designated recording period.

One-photon optogenetic stimulation during two-photon calcium imaging

Simultaneous optogenetic stimulation was performed during two-photon calcium imaging on a two-photon fluorescence microscope (A1R-MP+, Nikon) equipped with a 920-nm pulsed laser (InSight DeepSee+, Spectra-Physics) and a 561-nm pulsed laser (Civil Laser, Hangzhou, China). One-photon optogenetic stimulation was delivered through a custom-built extension adapter (Ex2pO) coupled to an optical fiber.⁴¹ One-photon stimulation pulses were synchronized to the turnaround period of the two-photon imaging resonant scanner using a field-programmable gate array (FPGA)-based multifunction board (Analog Discovery 2, Digilent, Boston, MA, USA). One-photon stimulation of ChRmine was delivered with the 561-nm pulsed laser. Time-lapse images were acquired with a 25× water-immersion lens (NA: 1.1, Nikon) at 512 × 512-pixel resolution and 30.0 Hz. The lower half of each AAV-transduced organoid was embedded in 4% low-melting point agarose and submerged in oxygenated ACSF.

All analyses were performed with MATLAB (R2022b, Mathworks). Calcium imaging data were first processed using Suite2p for motion correction and automated ROI segmentation. ROIs positively classified as cells by Suite2p were selected for subsequent analysis. The fluorescence signal for each cell was processed in several steps. First, the raw fluorescence trace (F) was corrected for neuropil contamination by subtracting 70% of the signal from the corresponding neuropil ROI (F_neu), yielding a corrected trace (F_corr = F – 0.7×F_neu). Slow baseline drifts arising from photobleaching were computationally removed using a custom percentile-based filter. Baseline fluorescence (F₀) was defined as the 15^th percentile of the corrected signal distribution, and the normalized fluorescence change was computed as ΔF/F₀. To reduce high-frequency noise while preserving the shape of calcium transients, the ΔF/F₀ traces were smoothed using a Savitzky–Golay filter (7^th order polynomial, 15-frame window). The resulting processed traces (F_preprocessed) were used for all subsequent analyses.

Quantification and statistical analysis

All statistical analyses were performed using MATLAB (R2022b, Mathworks). Results are expressed as mean ± S.E.M. unless otherwise stated. jGCaMP7f signal amplitudes in Figures 3F and 4P are presented as medians, and distributions are shown as violin plots. The numbers of organoids and independent differentiation experiments are indicated in the corresponding figure legends. Multiple group means were compared by one-way analysis of variance, followed by the indicated post hoc test selected according to the experimental design and the specific comparison of interest. When AAV-9 served as a predefined reference serotype, Dunnett’s multiple comparison test was used to compare each capsid group against AAV-9. When the aim was to compare all serotype groups with one another, Tukey’s multiple comparison test was used. For comparisons between two groups, Welch’s t test was used, as in Figures 2D and 4C. Probability values < 5% were considered statistically significant for all tests.

Published: August 13, 2026