Spatial transcriptomics reveals laminar and cell-type-specific A-to-I RNA editing signatures in the macaque cortex

Cheng Chen; Jing Yang; Zhiyong Zhu; Ziyu Wang; Yunqi Huang; Lingjun Ding; Lidan Luo; Tianming Lan; Yong Hou; Qiye Li; Shiping Liu; Ying Lei; Jinrong Huang

doi:10.1038/s41467-026-77434-y

. 2026 Sep 3;17:10504. doi: 10.1038/s41467-026-77434-y

Spatial transcriptomics reveals laminar and cell-type-specific A-to-I RNA editing signatures in the macaque cortex

Cheng Chen ^1,^2,^3,^#, Jing Yang ^3,^#, Zhiyong Zhu ³, Ziyu Wang ^1,³, Yunqi Huang ^1,³, Lingjun Ding ^2,³, Lidan Luo ⁴, Tianming Lan ⁵, Yong Hou ^1,⁶, Qiye Li ², Shiping Liu ^3,⁷, Ying Lei ^2,^3,^8,^✉, Jinrong Huang ^2,^3,^✉

PMCID: PMC13639015 PMID: 42834064

Abstract

ADAR-mediated A-to-I RNA editing is crucial for proper neuronal function. However, its spatial signatures in the cerebral cortex remain largely unexplored. To address this, we developed SRDRE, an analysis framework for identifying RNA editing sites in spatial transcriptomics data. By analyzing over 40 million cells across 142 spatially annotated cortical regions in the adult macaque cortex, we uncovered laminar and cell-type-specific RNA editing landscapes. Notably, neuronal RNA editing levels correlated with cortical hierarchy in the visual and somatosensory systems. Parvalbumin neurons exhibited the highest editing activity in most regions. Furthermore, we identified primate-enriched sites within ion channel genes, specifically in excitatory neuronal subtypes of the prefrontal cortex. Finally, in a preliminary exploratory analysis, we observed altered RNA editing in deeper layers of the dorsolateral prefrontal cortex in macaques exhibiting depressive-like behaviors, suggesting a potential link that warrants further investigation. These findings underscore the role of RNA editing and its potential contributions to cortical specialization, brain evolution, and neuropsychiatric disorders.

Subject terms: Computational neuroscience, Genetics of the nervous system

ADAR-mediated RNA editing is crucial for neuronal function, yet its spatial signatures in the cortex are unknown. Here, authors show that editing correlates with cortical hierarchy, and PV neurons exhibit the highest editing levels in most regions.

Introduction

Adenosine-to-inosine (A-to-I) RNA editing, catalyzed by the adenosine deaminase acting on RNA (ADAR) family of enzymes, is an evolutionarily conserved post-transcriptional mechanism that significantly contributes to transcriptome diversity^1–4. Unlike DNA mutations that are hard-wired, RNA editing is dynamically regulated and thus facilitates fine-tuning of gene expression. Inosine (I) is recognized as guanosine (G) by ribosomes, potentially leading to amino acid substitutions and altering protein-coding potential of transcripts³. A-to-I editing events are typically identified as A-to-G mismatches in high-throughput RNA sequencing data.

The importance of RNA editing is particularly evident in the brain. During brain development, RNA editing activity significantly increases, highlighting its critical role in neuronal maturation^5–7. This process is crucial for proper neuronal function, as many RNA editing sites act as key mediators of neurotransmission and contribute to synaptic plasticity^8–10. Aberrant RNA editing in the brain has been implicated in a range of neuropsychiatric disorders, including schizophrenia^11,12, autism spectrum disorder¹³, Alzheimer’s disease^14–16, and major depression¹⁷.

The cerebral cortex, or neocortex, exhibits substantial expansion in size and structural complexity in higher mammalian species. In the adult neocortex, which is organized into numerous distinct regions (approximately 180 in humans¹⁸), diverse cell types coexist and orchestrate sophisticated behaviors. A key organizational principle of the cerebral cortex is its hierarchy, which is particularly well-established in the visual and somatosensory systems¹⁹. Within this hierarchy, lower-level regions (e.g., the primary visual cortex) process basic features, while higher-level regions integrate these signals into progressively more complex representations. Given the critical role of RNA editing in brain function, it is particularly intriguing to examine its signatures in this highly specialized structure. Notably, the frequency of A-to-I RNA editing in primates exceeds that observed in non-primate mammalian species, such as mice, by at least an order of magnitude²⁰, suggesting a potential link between RNA editing and evolutionary complexity of the primate brain. However, it remains unclear whether RNA editing is associated with the hierarchical organization of the primate cortex.

RNA editing has been extensively studied in various species^21–29. However, most studies to date have provided tissue-averaged editing information, which obscures cellular heterogeneity and spatial context. Recent work has revealed that RNA editing can vary even within a single neuron, as demonstrated by differences between axons and neuronal bodies in squid³⁰. This spatial regulation of RNA editing highlights the need for techniques capable of capturing editing events within their native tissue context. Although methods such as fluorescence in situ hybridization^31,32, in situ sequencing³³, and endonuclease V immunostaining³⁴ have been developed to visualize A-to-I editing, these approaches are limited in their ability to provide transcriptome-wide insights. Comprehensive transcriptome-wide studies of RNA editing in complex tissues, such as the cerebral cortex, are still lacking. The emergence of sequencing-based spatial transcriptomics techniques now enables characterization of gene expression and transcriptome-wide RNA editing within their native spatial context^35,36. However, existing tools for identifying RNA editing events, such as GIREMI³⁷, REDItools³⁸, RES-Scanner³⁹, RNAEditor⁴⁰ and DEMINING⁴¹, are not applicable to spatial RNA sequencing data.

To address these limitations, we developed SRDRE (Spatially Resolved Detection of RNA Editing), an analysis framework for accurately identifying RNA editing events in spatial RNA sequencing data. We applied SRDRE to analyze publicly available spatial RNA sequencing data comprising over 40 million cortical cells from the crab-eating macaque (Macaca fascicularis; commonly referred to as the cynomolgus monkey). Through our analysis, we characterized the laminar and cell-type-specific landscape of RNA editing across the macaque cortex. Moreover, our analysis revealed primate-enriched editing patterns within ion channel genes, particularly at evolutionarily conserved sites such as GRIA2 (R/G) and KCNMA1 (S/G), in specific excitatory neuronal subtypes of the prefrontal cortex (PFC). Finally, in an exploratory analysis with a small sample size, we observed altered RNA editing patterns within the deeper layers of the dorsolateral prefrontal cortex (DLPFC) in macaques exhibiting depressive-like behaviors, suggesting a potential link that requires validation in future studies. SRDRE is freely available at https://github.com/JinRcn/SRDRE.

Results

Overview of SRDRE

We developed SRDRE, an analysis framework to identify RNA editing events directly from spatial RNA sequencing data (Fig. 1a). To remove potential genomic single nucleotide polymorphisms (SNPs), we employed a curated catalog of known A-to-I RNA editing events. For each spatial spot, RNA editing was detected using the mapped BAM (Binary Alignment Map) file, which contains aligned sequencing reads along with their spatial coordinates. PCR (Polymerase chain reaction) duplicates and reads aligned to multiple loci were excluded from the analysis to ensure accuracy. SRDRE then quantified the frequency of A-to-G mismatches at each RNA editing site across all spots, allowing for in situ visualization of RNA editing events within their spatial context. By integrating cell type and/or regional annotations established through marker gene expression, SRDRE enabled the identification of RNA editing patterns in specific transcriptome-defined cell types across distinct anatomical regions.

Comparisons of 3’-biased sequencing and full-transcript coverage methods

3’-biased sequencing methods, which primarily capture sequences near the 3’ end of transcripts, often exhibit sparse coverage in current spatial, single-cell (sc), and single-nucleus (sn) RNA-seq datasets. This sparse coverage poses a potential limitation for comprehensive RNA editing detection, as editing events can occur throughout the transcript. To evaluate the performance of 3’-biased sequencing relative to full-length transcript data, we incorporated Smart-seq, a plate-based method known for its reliability in single-cell analyses and its ability to sequence full-length mRNAs. We analyzed matched mouse cortical scRNA-seq datasets generated using 10X Genomics Chromium (a 3’-biased droplet-based method) and Smart-seq (full-length transcript profiling)⁴². This design allowed us to benchmark RNA editing detection accuracy between the methods. Evaluation of transcript coverage confirmed the expected 3’ bias in the 10X data (Supplementary Fig. 1a). RNA editing sites identified by both methods in glutamatergic neurons from the mouse primary motor cortex (MOp) showed substantial overlap (Supplementary Fig. 1b). Despite the coverage bias, RNA editing sites were largely evenly distributed across gene bodies in both datasets (Supplementary Fig. 1c). Detection sensitivity was evaluated through progressive sub-sampling of 10X MOp data and subsequent recall of editing events (Supplementary Fig. 1d). Approximately 15 million reads were sufficient to consistently recall over 50% of the sites annotated in the REDIportal database. Editing levels at common sites (n = 70,867) were highly correlated between methods (Supplementary Fig. 1e). A heatmap of key sites showed that most were reliably captured, with few dropouts (Supplementary Fig. 1f). The cell-type RNA editing index from both methods was highly correlated across cortical regions (Supplementary Fig. 1g). Overall, the results demonstrated a high degree of consistency, indicating that 3’-biased sequencing data are as effective as plate-based full-length techniques for RNA editing detection at cell-type resolution.

Assessment of coding site coverage in spatial transcriptomics data

To quantitatively evaluate the performance of spatial transcriptomics in terms of its coverage of coding sequence (CDS) sites, we analyzed datasets from two representative platforms: Spatial Enhanced Resolution Omics-sequencing (Stereo-seq) data from the macaque visual cortex⁴³ and Visium data from the macaque prefrontal cortex⁴⁴. To ensure the identification of bona fide RNA editing sites, we constructed a comprehensive A-to-I RNA editing catalog for the crab-eating macaque by integrating bulk-tissue RNA sequencing data with whole-genome DNA sequencing (WGS) data from the same individuals. This catalog revealed that A-to-I editing was the most abundant type of RNA editing in macaque (Supplementary Fig. 2a), consistent with previous studies in other species^{23,24,27–29}. Most sites (> 70%) were located in introns (Supplementary Fig. 2b). Additionally, a large majority of sites (> 86%) were found in Alu elements (Supplementary Fig. 2c). Notably, the observed depletion of G in the −1 base and enrichment of G in the +1 base flanking the editing sites align with the known sequence signature of ADAR enzymes (Supplementary Fig. 2d). In total, we identified 2,782,079 A-to-I RNA editing sites, of which 3290 were located within CDS, and 1716 of these CDS sites resided in non-repetitive regions (Supplementary Fig. 2e). Most CDS sites were detected in publicly available macaque bulk RNA-seq datasets (Supplementary Fig. 2f). Nonetheless, these identified CDS sites should be viewed as hypothesis-generating until further validated. The detailed methodology for catalog construction was described in the Methods section. Using this catalog as a benchmark, our analysis demonstrated that Stereo-seq detected a high proportion of CDS sites. Specifically, in the V1 region, it detected over 95% of CDS sites, over 96% of recoding sites, and over 91% of non-repetitive CDS sites (Supplementary Fig. 3a–c). The per-site coverage distribution confirmed adequate coverage for most critical recoding sites (Supplementary Fig. 3d), with key examples such as GRIA2 (Q/R) and KCNA1 (I/V) being reliably detected. Visium data captured a lower proportion of these sites (Supplementary Fig. 3e–h). This difference is potentially attributable to the smaller spatial coverage of the Visium data, which may have limited the overall sampling of sites. Specifically, the Visium dataset covered a smaller tissue area, only a subsection of the prefrontal cortex, whereas the Stereo-seq dataset provided a much larger capture area, spanning the entire visual cortex representation in that section.

SRDRE reveals regional RNA editing differences in the mouse brain

We applied SRDRE to analyze spatial RNA sequencing data obtained from a section of the adult mouse brain, where each spot on the section represented a mixture of cells (Fig. 1b). We performed A-to-I RNA editing detection on each spot using sites recorded in the REDIportal database of mouse³⁸. To assess global RNA editing activity, the RNA editing index (REI) is defined as the ratio of the number of Gs (edited bases) to the total number of As (unedited bases) and Gs. By calculating the REI for each spot, we observed spatial heterogeneity in RNA editing activity across the brain section (Fig. 1c). Among the regions analyzed, the thalamus exhibited a notably higher editing level compared to other regions, such as the cortex and hippocampus (Fig. 1d), in line with previous studies^35,36. Furthermore, to validate the robustness of our framework, we extended its application to spatial RNA sequencing data generated by long-read sequencing technology³⁶. As illustrated in Supplementary Fig. 4, our approach successfully detected RNA editing events in long-read data, demonstrating its versatility and compatibility with diverse sequencing platforms. This capability highlights the potential of SRDRE for integrating multi-modal sequencing data to provide a more comprehensive understanding of RNA editing dynamics in spatially resolved transcriptomics.

Visualizing A-to-I editing in single cells

Next, we expanded our analysis to investigate A-to-I RNA editing using macaque Stereo-seq data, which combines a large field of view with cellular resolution. As an illustration, we selected a representative tissue section from an adult macaque, which encompassed the occipital and parietal lobes, to showcase the analysis (Fig. 1e). The aligned reads from spatial RNA sequencing of fresh frozen tissues were obtained from a previous study⁴³. We computed the editing level for each site in individual cells, which was calculated as the number of Gs divided by the total number of As and Gs, and mapped their spatial coordinates. This approach enabled us to visualize RNA editing patterns at single-cell resolution across the tissue section. For example, we examined the editing pattern of a well-covered site, FBXO16 (K68R), in single cells (Fig. 1f). FBXO16 (K68R) was weakly edited in most cells but strongly edited in a small subset. We then quantified the proportion of cells with high editing levels (> 0.5) at this site for each neuronal type across cortical regions. Our results showed that inhibitory neurons in V1 and V6Av regions exhibited a higher proportion of high editing compared to those in other regions (Fig. 1f).

Quantifying global RNA editing activity using the Alu editing index

In humans and macaques, A-to-I editing sites are predominantly located within primate-specific Alu elements^45,46, which are short interspersed nuclear elements (SINEs) that are highly abundant in primate genomes. To assess global RNA editing activity, we utilized the Alu editing index (AEI), a well-established metric that quantifies the ratio of Gs to the total number of As and Gs within Alu elements. Previous studies have demonstrated the effectiveness of AEI in quantifying RNA editing activity in bulk RNA sequencing samples⁴⁷, making it suitable for evaluating editing dynamics at a broader scale.

In this study, we employed two adjacent Stereo-seq sections (10 μm apart) from the PFC of an adult macaque to perform regional AEI analysis (Supplementary Fig. 5). To ensure data reliability, we included only spots with a sequencing depth of at least 20 unique molecular identifier (UMI) counts covering known editing sites. These spots were then pooled based on their regional annotation, allowing us to calculate AEI values for specific layers within the 10mr (medial prefrontal area), 46d (dorso-lateral prefrontal area in the dorsal bank and lip of the principal sulcus), and 46v (ventro-lateral prefrontal area in the ventral bank and lip of the principal sulcus) regions of the PFC. Our analysis revealed a strong correlation between regional AEI in the two adjacent tissue sections (R = 0.77, p = 0.00028) (Fig. 1g). To confirm that the AEI reflects genuine editing activity rather than sequencing noise, we analyzed its relationship with read coverage in 72 independent human PFC single-nucleus RNA-seq replicates⁴⁸ and found no significant correlation (Supplementary Fig. 6). This high reproducibility underscores the reliability of our approach and provides robust evidence that AEI is a reliable and sensitive indicator of RNA editing activity in spatial RNA sequencing data.

Linking cell-type-specific RNA editing dynamics to cortical hierarchy

To gain a deeper understanding of the dynamics of RNA editing activity in the cortex, we collected Stereo-seq data of 152 serial sections spanning the entire cortex from three adult male macaques (Supplementary Data 1), as described in our previous study⁴³. A total of 142 spatially annotated cortical regions (Fig. 2a and Supplementary Data 2), categorized into nine lobes (prefrontal, frontal, occipital, temporal, parietal, auditory, somatosensory, cingulate, and insulate), were included in our cortex-wide RNA editing analysis. To ensure data reliability, only regions with a minimum read depth of 0.1 million UMI counts covering RNA editing sites were retained for further analysis. Our findings showed a significant correlation between regional RNA editing levels and the proportion of neurons across the lobes (Fig. 2b), consistent with previous observations in humans⁴⁹. However, no significant correlation was observed between regional RNA editing levels and neuron density (R = 0.034, p = 0.094) (Supplementary Fig. 7), suggesting that RNA editing activity is more closely associated with neuronal composition than with neuronal density.

Fig. 2 — a Over 40 million cells across 142 spatially annotated regions in the adult macaque cortex were involved in this study. b Correlation analysis depicting the relationship between the regional Alu editing index (AEI) and the proportion of neurons within each cortical region. Solid lines represent linear regression fits, with gray shaded bands denoting 95% confidence intervals. c Box plots illustrating AEI across different cortical regions for major cell types. Each point in the analysis represents pseudo-bulk data derived from regions with at least 0.1 million UMI counts covering known editing sites. For each box plot, the center line denotes the median, the box bounds represent the 25th and 75th percentiles, and the whiskers extend to the minimum and maximum values within 1.5 × IQR from the quartiles. Exact biological replicate numbers (n) for each region/cell type combination are provided in Source Data. d, e Negative association between the residual AEI and cortical hierarchy in inhibitory and excitatory neurons. Scatter plots show the correlation between the residual AEI and cortical hierarchy levels (as defined by Felleman and Van Essen) for the d visual and e somatosensory systems, after controlling for ADAR expression levels (ADAR1, ADAR2, and ADAR3), sequencing depth, and the proportion of neuronal subtypes. Analyses were performed separately for inhibitory and excitatory neurons. Only regions with at least 0.1 million UMI counts were included in the analysis. f Correlation analysis between AEI and ADAR1 expression specifically in the case of excitatory neurons. Spearman’s correlation (two-sided) was performed, and the correlation coefficient and p value are indicated. IN inhibitory neurons, EX excitatory neurons, ASC astrocytes, OLG oligodendrocytes. Source data are provided as a Source Data file.

To elucidate cell-type-specific RNA editing patterns, we calculated the AEI for inhibitory neurons (IN), excitatory neurons (EX), and non-neuronal cells such as astrocytes (ASC) and oligodendrocytes (OLG), across various cortical regions. Notably, we observed remarkable variation in AEI of these cell types among different cortical regions (Fig. 2c and Supplementary Data 3), indicating that RNA editing is subject to spatial regulation in the macaque cortex.

Interestingly, after controlling for neuronal proportion, ADAR expression (ADAR1, ADAR2, and ADAR3) and sequencing depth, we discovered a negative correlation between the regional AEI and the hierarchical level of specific regions in the visual (Supplementary Fig. 8a) and somatosensory systems (Supplementary Fig. 8b). To extend these findings with a more refined analysis, we next focused on neuronal AEI. Consistently, we observed a negative correlation with the hierarchical level in both the visual (Fig. 2d) and somatosensory systems (Fig. 2e). This analysis controlled for potential confounders, including the proportions of neuronal subtypes, ADAR expression levels, and sequencing depth. To further assess the statistical robustness of this correlation, we performed permutation tests by randomly shuffling the hierarchy labels 5000 times and recalculating the correlation coefficient for each permutation. The observed negative correlation coefficient fell far outside the distribution of coefficients from the permuted data (empirical p value < 2.2e−16) (Supplementary Fig. 9a, b). The hierarchical organization of visual and somatosensory cortical regions was defined according to the work of Felleman and Van Essen¹⁹. These findings suggest that RNA editing levels are dynamically regulated along the cortical hierarchy.

Layer- and cell-type-specific ADAR gene expression

To investigate the expression patterns of ADAR enzymes (including ADAR1, ADAR2, and ADAR3) across various cell types, we analyzed snRNA-seq data from more than 1 million cells derived from 142 spatially annotated cortical regions (Supplementary Fig. 10a)⁴³. Among the three ADARs family members, ADAR3 lacks deaminase activity. A recent study reported that ADAR3 can modulate neuronal differentiation⁵⁰. Our analysis revealed that both ADAR1 and ADAR2 were broadly expressed across different cell types (Supplementary Fig. 10b). In contrast, ADAR3 showed relatively high expression in specific inhibitory neuron subtypes, including LAMP5 (lysosome-associated membrane protein 5), VIP (vasoactive intestinal peptide), and RELN (reelin) neurons, which are derived from the caudal ganglionic eminence (CGE) (Supplementary Fig. 10b).

Further analysis using spatial RNA sequencing data revealed distinct expression patterns of ADARs across cortical layers and regions. In excitatory neurons, ADAR1 was broadly expressed in layers 2-6, whereas ADAR3 showed higher expression in upper layers, particularly layers 2-3 (Supplementary Fig. 11a). Of note, ADAR1 is the primary editor of repetitive elements, including Alu²⁴. In the case of excitatory neurons, we observed a moderate correlation between the normalized expression of ADAR1 and AEI (Fig. 2f). Moreover, in both IN and EX, ADAR1 expression tended to be lower in regions of higher cortical hierarchy within the visual system (Supplementary Fig. 11b). Taken together, our findings linked neuronal RNA editing to cortical hierarchy, suggesting a potential functional relevance of RNA editing in the hierarchical organization of the macaque cortex. However, further experiments are needed to validate the hypothesis and elucidate the underlying mechanisms.

Parvalbumin neurons exhibit the highest RNA editing levels in most cortical regions

We observed that neurons exhibited higher global RNA editing levels compared to non-neuronal cells in the macaque cortex (Fig. 3a, b), consistent with previous findings in purified human cortical cells⁴⁹. Our analysis of independent snRNA-seq data (n = 7) from the macaque PFC⁴⁴ further confirmed that global RNA editing levels are significantly higher in neurons than in non-neuronal cells. These results validated the framework used in this study and supported the notion that elevated RNA editing activity in neurons represents a widespread phenomenon. In the human PFC, the AEI was elevated in IN relative to EX⁴⁹. In most macaque cortical regions examined, the AEI of IN and EX was generally similar, with a few exceptions. Specifically, within the V1 region, higher AEI was observed in IN compared to EX (Fig. 3b). This observation aligns with a previous study reporting higher RNA editing levels in IN compared to EX within the V1 region of humans⁵¹, suggesting a conserved role for RNA editing in visual processing across primates.

The adult mammalian cerebral cortex is typically organized into six distinct layers, comprising upper layers (layer 1 to 4) and deep layers (layer 5 and 6). The upper layers primarily facilitate intracortical connections, whereas the deep layers are responsible for long-distance projections to other brain regions⁵². Previous studies have extensively explored layer-dependent gene expression in the primate cortex⁵³; however, our understanding of laminar differences in RNA editing remains limited. In this study, we calculated the AEI for each layer across various cortical regions. Interestingly, we observed dynamic patterns of laminar cell-type A-to-I editing, and representative cortical regions exemplifying these dynamics were shown (Fig. 3c). For instance, within the F1 region, the AEI of IN was relatively higher in layer 3 compared to layer 6. The AEI were largely similar between upper and deep layers across most brain regions. A significant layer-specific difference was exclusively observed in excitatory neurons of V1, but not in inhibitory neurons or other examined regions (Supplementary Fig. 12). ADAR3 is believed to suppress RNA editing and was more abundant in layers 2-3 (Supplementary Fig. 11a). However, no statistically significant differences in AEI were found between layers 2-3 and 4-6 in any of the regions studied (Supplementary Fig. 13).

Next, we calculated the AEI for different neuronal subtypes across cortical regions. Inhibitory neurons were classified into distinct subtypes based on marker genes, including LAMP5, VIP, RELN, PV (parvalbumin), and SST (somatostatin), while excitatory neurons were grouped based on their layer preferences (e.g., L2/3, L2/3/4, L3/4, L4, L4/5, L4/5/6, L5/6, and L6; L denotes layer). In most regions examined, the AEI was found to be higher in PV neurons compared to other neuronal subtypes (Fig. 3d and Supplementary Data 4). For instance, in the motor cortex (e.g. F1, F2, F3, and F4), PV emerged as the neuronal subtype with the highest level of editing, while the EX-L2/3/4 subtype displayed relatively lower editing levels compared to other subtypes.

To further validate these observations, we conducted a comprehensive analysis of sc/snRNA-seq data from three different sources: mouse cortex⁴², macaque PFC⁴⁴, and human PFC^54–56. Within the investigated regions of the mouse cortex, PV neurons exhibited the highest level of editing among all neuronal subtypes (Supplementary Fig. 14a). Similarly, PV neurons in macaque PFC and human PFC also demonstrated relatively higher levels of editing compared to other neuronal subtypes (Supplementary Fig. 14b). These results imply that RNA editing is not uniform throughout the various neuronal subtypes but instead varies in a cell-type dependent manner, potentially reflecting subtype-specific functional requirements or regulatory mechanisms.

More CDS sites are enriched in inhibitory neurons than excitatory neurons

The importance of A-to-I editing events within CDS in neuronal function, particularly in neurotransmission, has been well-documented^9,10. For instance, AMPA (α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid) glutamate receptors containing subunits with edited R/G site exhibit faster recovery rates from desensitization and respond more rapidly to synaptic impulses⁵⁷. To identify differences in editing levels at CDS sites between major neuronal populations, we performed pairwise comparisons between IN and EX across cortical regions. We found that more sites exhibited IN-enriched editing patterns in most of the regions investigated (Fig. 4a). A similar trend was found in non-repetitive CDS sites (Supplementary Fig. 15). For example, among the 1,199 sites analyzed in V1 region, 23 showed IN-enriched patterns, while only 9 exhibited EX-enriched patterns (Fig. 4b). We then examined CDS sites in excitatory neurons and identified differentially edited sites between the upper and deep layers (Supplementary Fig. 16a-c and Supplementary Data 5, 6).

Fig. 4 — a Number of IN- and EX-enriched CDS sites in various cortical regions. b Volcano plot showing IN- and EX-enriched non-repetitive CDS sites in V1 (primary visual cortex). Differential editing sites were identified by Fisher’s exact test (two-sided) with Benjamini-Hochberg (BH) correction. Cell-type-enriched sites were defined as those with an adjusted p value of less than 0.05 and an editing difference of at least 10%. c Representative recoding sites in excitatory neurons, showcasing their spatial editing patterns in both the visual and somatosensory systems. d Association between editing level and cortical hierarchy in a specific layer, focusing on the recoding site GRIA2 (R764G). e Similar analysis as in (d), but for the recoding site KCNA1 (I400V). Spearman’s correlation (two-sided) was performed, and the correlation coefficient and p value are indicated. The analyzed CDS sites had a minimum depth of five reads supporting both edited and non-edited status at the cell type level. Recoding sites are denoted by the original amino acid followed by the edited amino acid. IN inhibitory neurons, EX excitatory neurons. Source data are provided as a Source Data file.

We also observed that the expression of ADAR2, but not ADAR1, tended to be higher in IN than in EX across most cortical regions (Supplementary Fig. 17a, b), consistent with previous findings in the human PFC⁴⁹. Given that ADAR2 is the primary editor of coding editing sites²⁴, the greater number of sites enriched in IN can be partially attributed to the higher expression of ADAR2 in these cells. These observations pointed to the existence of cell-type-specific regulatory mechanisms in RNA editing. On the other hand, certain sites, such as GRIA2 (R764G) and GRIA3 (R775G), showed low specificity between IN and EX across most cortical regions.

To quantify the spatial dynamics of RNA editing, we calculated the standard deviation (SD) of editing levels for each recoding site across 142 brain regions. We focused our analysis on excitatory neurons as a representative neuronal subtype and included only previously reported sites that were detected in at least 48 (one-third) of the regions. The recoding sites related to neurotransmission and ion channels did not show significantly greater spatial variation than sites in other functional categories (Supplementary Fig. 18a, b). The recoding sites associated with neurotransmission and ion channels, such as GRIA2 (Q607R), GRIA2 (R764G), GRIA3 (R775G), and KCNA1 (I400V), are known to regulate neuronal excitability⁹. The editing level of GRIA2 (Q607R) increases rapidly in the early neuronal lineage and reaches nearly 100% in mature neurons¹⁰. This high editing level is critical since AMPA glutamate receptors containing the GluA2 subunit with a fully edited Q/R site are impermeable to Ca²⁺, a property that is crucial for neuronal survival and synaptic function¹⁰. As expected, the editing level of GRIA2 (Q607R) remained constant and approached nearly 100% across all examined cortical layers and regions (Fig. 4c). In contrast, the editing level of GRIA2 (R764G) exhibited a more dynamic pattern. Specifically, in layer 2 of the cortex, the editing level of GRIA2 (R764G) showed a negative correlation with the cortical hierarchy within the visual system (Fig. 4d). Similarly, the editing level of KCNA1 (I400V) in layers 3 and 4 was also negatively correlated with the cortical hierarchy in the visual system (Fig. 4e). The I400V site is located in K_V1.1, a potassium voltage-gated channel encoded by the KCNA1 gene. Edited K_V1.1 channels recover faster from inactivation⁹, indicating that RNA editing at this site may fine-tune neuronal excitability. These findings suggest that the regulation of editing at certain recoding sites is both laminar- and hierarchy-dependent, likely reflecting the functional demands of specific cortical regions and layers.

Primate-enriched RNA editing patterns occur at evolutionarily conserved sites

The evolutionary expansion of the neocortex represents a fundamental prerequisite for higher cognitive functions in primates⁵⁸. To better understand the role of RNA editing in this evolutionary process, we extended beyond previous tissue-level studies⁵⁹ by performing a comparative analysis of conserved editing sites at cell-type resolution. We focused on the PFC, a brain region critically associated with higher-order cognitive processes, across three evolutionarily representative species: mouse, macaque, and human. First, we compiled a comprehensive list of evolutionarily conserved editing sites (n = 160) across mouse, macaque, and human genomes (Supplementary Data 7). Genes harboring these conserved sites exhibited a notable enrichment in ion channel-related functions (Supplementary Fig. 19), consistent with previous mammalian studies⁶⁰. For the mouse, we utilized data (n = 7 independent biological replicates) from previous studies covering prefrontal subregions such as the prelimbic cortex (PL), infralimbic cortex (ILA), and orbital cortex (ORB)⁴². For macaque and human, we obtained publicly available snRNA-seq data of healthy individuals (macaque: n = 7; human: n = 37) from previous studies conducted on the PFC^44,54–56. We focused on specific neuronal subtypes, including EX-L2/3, EX-L6, IN-PV, and IN-SST, which showed a higher read coverage at conserved sites. Overall, our observations indicated a tendency for higher RNA editing levels in primates at these conserved sites (Fig. 5a). One notable example is the K/R recoding in NEIL1 (Nei like DNA glycosylase 1), which leads to a 30-fold decrease in thymine glycol cleavage activity on duplex DNA⁶¹. For the EX subtypes analyzed, we found that NEIL1 (K/R) was highly edited in macaques and humans but showed low editing levels in mice (Fig. 5b), consistent with previous tissue-level observations⁶². The vast majority of genes harboring conserved sites show similar expression levels across the three species, with no systematic bias toward higher expression in human or macaque relative to mouse (Supplementary Fig. 20a, b). Therefore, these results demonstrate that the observed editing patterns are unlikely to be confounded by interspecies differences in gene expression.

By performing differential RNA editing analysis of conserved sites across species, we identified species-enriched editing sites (Fig. 5c and Supplementary Data 8, 9). Primate-enriched editing patterns were observed at conserved sites within specific neuronal subtypes. For instance, GRIA2 (R/G) displayed a primate-enriched pattern in EX-L2/3 and EX-L6, but not in IN-SST (Fig. 5b). Similarly, KCNMA1 (S/G) exhibited primate-enriched editing in EX-L6 but not in EX-L2/3 (Fig. 5b). Notably, both sites are located within ion channel genes. On the other hand, certain sites, such as GRIA2 (Q/R) and GRIA3 (R/G), showed a consistent and highly edited pattern across neuronal subtypes in all three species (Fig. 5b). In addition to recoding sites, certain sites located in introns or 3′-untranslated regions (3′-UTRs) also exhibited primate-enriched editing in specific cell types (Supplementary Fig. 21). Among the conserved sites shared by all three species, a greater number exhibited human-enriched patterns compared to macaque (Fig. 5d and Supplementary Data 10).

To further expand our analysis, we examined conserved editing sites (n = 1,221,330) shared between macaque and human. We calculated the conserved editing index, which represents the weighted average of editing levels for these conserved sites in the two species. Consistent with previous findings⁷, we observed higher editing levels in humans compared to macaques across the investigated cell types, with more pronounced differences in EX-L2/3 than in EX-L6 (Fig. 5e). Of the 1,221,330 conserved editing sites, 1,116,746 (91.4%) were located within Alu repeats, while the remaining 104,584 (8.6%) were in non-Alu regions. In both Alu and non-Alu sites, the conserved editing index was higher in humans than in macaques across all investigated cell types (Supplementary Fig. 22).

Altered RNA editing in the DLPFC of macaques with depressive-like behaviors

RNA editing alterations have been observed in the DLPFC of individuals with depressive disorder¹⁷. However, it is still unclear whether RNA editing changes in depressive disorder are associated with the layer structure within the DLPFC. To address this, we analyzed spatial RNA sequencing data from macaques exhibiting depressive-like behaviors, as described in a previous study⁴⁴. Three highly belligerent populations were screened, and depressive-like behaviors (including huddling, sitting alone, locomotion, amicable interactions, and ingestion) were quantitatively analyzed⁴⁴. Based on social rank assignment, the top two or three macaques in each group were identified as controls (no depressive-like behaviors), while the lowest-ranking four were classified by behavioral severity into those with significant or mild depressive-like behaviors⁴⁴. Using unsupervised clustering based on gene expression features, we arranged the spots in each slice sample according to their cortical structure (layers 2-6) (Fig. 6a). Comparisons between macaques with significant depressive-like behaviors (n = 2 independent biological replicates) and healthy controls (n = 3) revealed significant differences in AEI of spots located in the deeper cortical layers (layers 5 and 6), but not in the upper layers (layers 2-4) (Fig. 6b). No significant differences in AEI were observed between macaques with mild depressive-like behaviors (n = 3) and healthy controls (n = 3) (Fig. 6b). However, the study’s small sample size limits its statistical power, and potential confounders (e.g., stress, handling, age) were not controlled for. These limitations warrant further investigation with larger samples in future research.

Fig. 6 — a Representative section of macaque DLPFC showing spatial segmentation (L for Layer). b Significant differences in Alu editing index (AEI) of spatial spots were observed between macaques with significant depressive-like behaviors (DL) and the control group in deeper layers (including layers 5 and 6), while no significant differences were found in upper layers (layers 2-4). The RES group represents macaques with mild depressive-like behaviors. Exact biological replicate numbers (n) for each group are provided in Source Data. c Significant differences in AEI between individuals with MDD (n = 37) and healthy controls (n = 35) were observed in multiple cell types within the human DLPFC. The significance of differences in AEI between groups was assessed using a two‑sided Wilcoxon rank-sum test. The resulting p values were adjusted using the Benjamini-Hochberg (BH) method to control for multiple testing. MDD major depressive disorder, DLPFC dorsolateral prefrontal cortex, IN inhibitory neurons, EX excitatory neurons, ASC astrocytes, OLG oligodendrocytes, OPC oligodendrocyte precursor cells. For each box plot, the center line denotes the median, the box bounds represent the 25th and 75th percentiles, and the whiskers extend to the minimum and maximum values within 1.5 × IQR from the quartiles. Source data are provided as a Source Data file.

We further extended our analysis to human samples to explore potential cell-type-specific dysregulation of RNA editing. Utilizing public snRNA-seq data, we examined the DLPFC of individuals with major depressive disorder (MDD) who died by suicide (17 males, 20 females), as well as matched healthy controls (17 males, 18 females)^48,55. Our analysis revealed significant differences in AEI between individuals with MDD and healthy controls in multiple cell types within the human DLPFC (Fig. 6c). We then leveraged snRNA-seq data to focus on conserved editing sites—those present in the same cell types in both macaque and human—and identified a subset that was dysregulated in both species (Supplementary Data 11). These findings underscored the significance of investigating the specific contributions of laminar and cell-type RNA editing to the development of depressive-like behaviors and related disorders.

Discussion

We developed an analysis framework, SRDRE, to detect RNA editing events in spatial RNA sequencing data and validated its performance on two widely adopted spatial transcriptomics platforms: Stereo-seq and Visium. Using SRDRE, we uncovered patterns of RNA editing across cortical layers and neuronal subtypes. Cortical regions with a higher proportion of neurons exhibited elevated global RNA editing levels. Notably, this study provided the evidence that neuronal RNA editing activity is closely associated with the hierarchical level of regions within the visual and somatosensory systems.

Neurons exhibited significantly higher global editing levels than non-neuronal cells, consistent with their requirement for transcriptomic diversity in supporting neural plasticity⁹. Among neuronal subtypes, PV neurons play a critical role in regulating local circuitry, brain networks, and memory processing^63,64. In our study, we observed that PV neurons exhibited higher global editing levels in most regions of the macaque cortex compared to other neuronal subtypes. This finding was consistent with observations in both mouse and human cortical regions. These discoveries raise the intriguing possibility that the heightened RNA editing activity in PV neurons is closely linked to their unique electrophysiological properties, particularly their ability to generate high-frequency firing patterns. Such firing patterns are essential for the rapid and precise regulation of cortical circuits, which underlies their critical role in information processing and network synchronization^64,65. The conservation of this phenomenon across species highlights the potential evolutionary importance of RNA editing in maintaining the functional integrity of PV neurons.

The extraordinarily high frequency of A-to-I RNA editing in primates, compared to non-primate mammalian species, may be attributed to the abundance of primate-specific Alu elements²⁰. These repetitive elements, which are highly prevalent in primate genomes, serve as substrates for RNA editing and likely enhance the complexity of transcriptomic regulation in primates. We expanded upon previous tissue-level studies by performing a cross-species comparative analysis of evolutionarily conserved RNA editing sites at cell-type resolution, thereby providing higher precision in identifying species-enriched editing patterns. Using sc/snRNA-seq data from the PFC of humans, macaques, and mice, we mapped RNA editing events in excitatory and inhibitory neuronal subtypes, such as EX-L2/3, EX-L6, IN-PV, and IN-SST. This high-resolution approach revealed primate-enriched editing patterns that were previously obscured in bulk tissue analyses. Notably, these patterns were particularly pronounced at evolutionarily conserved sites within ion channel genes, including GRIA2 (R/G) and KCNMA1 (S/G), suggesting their functional relevance in fine-tuning synaptic plasticity. Moreover, a greater number of conserved RNA editing sites showed human-enriched patterns compared to those in macaques. This observation raises intriguing questions about the role of RNA editing in the evolutionary expansion of the neocortex and the development of higher cognitive functions in humans. Future studies should focus on elucidating the functional consequences of these editing events, particularly their contribution to primate-specific cognitive abilities and their potential role in shaping the complexity of the human brain.

We uncovered significant alterations in RNA editing within the deeper layers of the DLPFC in macaques exhibiting depressive-like behaviors. To extend these findings, we analyzed snRNA-seq data from humans with MDD and identified significant RNA editing alterations across multiple cell types compared to healthy controls. Furthermore, we identified a subset of conserved editing sites—present in the same cell types of both species—that were dysregulated in both macaques and humans. Collectively, our results implicate layer- and cell-type-specific RNA editing alterations in the pathophysiology of neuropsychiatric disorders, warranting future studies to elucidate the underlying molecular pathways and their therapeutic potential.

This study has several limitations. First, the analysis included a limited number of biological replicates, which may constrain the generalizability of the findings. Second, potential biases arising from differences in sex and age were not accounted for during cross-species comparisons. Additionally, sparse data and limited detection efficiency of current spatial RNA sequencing assays constrain this study. Future studies incorporating spatial full-length RNA sequencing, deeper sequencing, and larger sample sizes will enable a more comprehensive understanding of RNA editing at single-cell or even subcellular resolution.

This study validated the utility of spatial RNA sequencing data for studying RNA editing and provided one of the most detailed investigations of laminar and cell-type-specific RNA editing in the primate cortex to date. We believe that SRDRE will serve as a valuable tool for the scientific community, enabling further exploration of RNA editing in spatial transcriptomics data and facilitating new discoveries.

Methods

Ethics statement

To create an A-to-I RNA editing catalog of the crab-eating macaque, three adult macaques (one female and two males, 6 years of age) were included in this study. Animal sample collection and handling procedures were conducted in accordance with the national guidelines for the large experimental animals and were approved by the Institutional Ethics Committee of BGI (Ethical Approval No. BGI-IRB19125-T4). The macaques were deeply anesthetized with a combination of ketamine hydrochloride (10 mg/kg) and sodium pentobarbital (40 mg/kg), and subsequently euthanized by exsanguination.

A-to-I RNA editing catalog of mice and humans

A-to-I RNA editing sites (Version 2.0) in mice and humans were downloaded from the REDIportal database (http://srv00.recas.ba.infn.it/atlas/index.html)³⁸.

Characterization of macaque A-to-I RNA editing via detection of DNA-RNA differences

By combining bulk RNA-Seq with whole-genome DNA sequencing (WGS) from the same animals, we were able to de novo identify RNA editing sites, excluding potential genomic single nucleotide polymorphisms (SNPs). The genomic DNA was isolated from the liver tissue of each macaque. DNA libraries were generated utilizing the MGI Easy Universal DNA Library Prep Kit and sequenced on the BGISEQ-500 platform with 100 bases paired-end reads. Clean DNA sequencing reads were mapped to the Macaca_fascicularis_5.0 genome (ensembl release-102), using BWA⁶⁶ (version 0.7.15-r1140) with the parameters ‘mem -t 4 -M -Y -R’. RNA sequencing of 81 samples from over 30 anatomically defined regions within the macaque brain was conducted after ribosomal RNA depletion. The clean RNA-seq reads were mapped to the Macaca_fascicularis_5.0 genome (ensembl release-102) by HISAT2⁶⁷ (version 2.1.0) using the parameters: ‘-q --no-mixed --no-discordant’. High consistency with results from different RNA editing detection methods, such as REDItools³⁸ and RES-Scanner³⁹, was demonstrated in our previous study²⁹. In this study, RES-Scanner³⁹ was employed to accurately identify RNA editing events based on the detection of DNA-RNA differences.

Validation of macaque CDS sites using independent bulk RNA-seq data

To validate the identified CDS sites, we utilized independent bulk RNA-seq data from macaques (Macaca fascicularis), including DRP005990, SRP164572, and SRP410543. These datasets were downloaded from the NCBI Sequence Read Archive (SRA) and represent transcriptomes from a variety of tissues. The clean RNA-seq reads were mapped to the Macaca_fascicularis_5.0 genome (Ensembl release 102) using HISAT2⁶⁷ (version 2.1.0) with the parameters: ‘-q --no-mixed --no-discordant’. A site is considered detected if it meets two criteria: first, the total read count (combining both edited and non-edited reads) is at least 5; and second, at least one read supports the edited status.

Comparative analysis of RNA editing in scRNA-seq data using two distinct platforms

We acquired scRNA-seq data from multiple regions of the mouse cortex using two distinct platforms: the droplet-based 10X Genomics Chromium (10X) and the plate-based Smart-seq method⁴². The analyzed regions encompassed primary motor cortex (MOp), primary somatosensory cortex (SSp), auditory cortex (AUD), visual cortex (VIS), anterior cingulate cortex (ACA), prelimbic cortex (PL), infralimbic cortex (ILA), orbital cortex (ORB), agranular insular cortex (AI), and retrosplenial cortex (RSP). We calculated the RNA editing index for each major cell type across regions.

Analysis of spatial RNA sequencing data from the mouse brain section

The genome-aligned BAM file for the mouse brain section (Slide Serial Number: V19L01-041-C1) was obtained from https://www.10xgenomics.com/datasets/adult-mouse-brain-coronal-section-fresh-frozen-1-standard. We used Scanpy⁶⁸ (version 1.8.2) for unsupervised clustering, grouping spots based on gene expression patterns. The identified clusters were mapped to anatomical regions through alignment with the Allen Mouse Brain Atlas reference. Additionally, the genome-aligned BAM file of the mouse brain (GSM4656182), generated using 10x Genomics Visium and Nanopore long-read sequencing, was obtained from a previous study (GSE153859)³⁶. A-to-I RNA editing was detected for each spot using the mouse REDIportal database (V2.0), and the RNA editing index was subsequently computed.

Aligning spatial RNA sequencing data to the macaque genome

Stereo-seq data were collected from 152 serial sections spanning the entire cortex of three adult male macaques (Macaca fascicularis), aged 6 years (Macaque #1), 4 years (Macaque #2), and 7 years (Macaque #3), respectively. To analyze these data, we employed the Stereo-seq Analysis Workflow (SAW) pipeline to align the reads to the reference genome (Macaca fascicularis 5.0) and generate mapped BAM files. This approach has been previously described in detail⁶⁹. Each section of the macaque brain yielded a single BAM file, which contained sequencing reads as well as the spatial coordinates of all spots. This standardized process ensured the generation of reliable and consistent data for further analysis. For Visium spatial transcriptomics sequencing data, Space Ranger was employed.

A-to-I RNA editing detection using spatial RNA sequencing data

To avoid overcounting PCR duplicates, we implemented a deduplication step by removing reads with identical unique molecular identifiers, spatial barcodes, and alignment coordinates. Additionally, we excluded reads that mapped to multiple genomic coordinates to ensure accurate counting. Subsequently, we focused on identifying A-to-G mismatches recorded in the species-specific catalog, and these were quantified across all spots. To maintain data integrity, we set a minimum sequencing base quality score threshold of 20, ensuring reliable and high-quality results.

Quantification of RNA editing index

The RNA editing index is defined as the ratio of the number of Gs to the total number of As and Gs in each sample, cell type, or region. For primate species, the Alu editing index (AEI) specifically focuses on the ratio of Gs to the total number of As and Gs within Alu elements, which are a specific type of repetitive DNA sequence. To ensure reliable and meaningful results, spots with a minimum of 20 UMI counts covering known RNA editing sites were retained for calculation. This filtering step helps to increase the signal-to-noise ratio and ensures that only spots with sufficient coverage and reliable data are included in the analysis.

Quantification of RNA editing at cell-type resolution in macaque Stereo-seq data

To assign sequencing reads to specific cell types, cell-type annotations of the macaque cortex were obtained from the original study⁴³. By combining reads belonging to the same cell type, we could accurately quantify RNA editing events specific to each cell type. This approach allowed for a more comprehensive understanding of the cell-type-specific RNA editing landscape within the macaque cortex.

Comparisons between two adjacent sections from macaque PFC

To evaluate the reliability and consistency of our method, we computed the correlation of the AEI for the same region between two sections. This analysis allowed us to assess the robustness of the AEI measurements and determine the degree of concordance. To address the issue of data sparsity inherent to spatial RNA-seq data, we employed a strategy to aggregate individual spots into pseudo-bulk pools based on their regional annotations. To ensure sufficient sequencing coverage and mitigate the potential effects of low coverage, we focused on analyzing pseudo-bulk data organized by region with a minimum of 0.1 million UMI counts supporting both edited and non-edited status. By applying this criterion, we could enhance the accuracy and robustness of our RNA editing activity measurements within the specific regions of interest.

Neuron density

Neuron density across brain regions is defined as the number of neurons in a region divided by the area of that region. The area of brain regions is calculated using a grid of sampling spots, where each 2 × 2 block (4 adjacent spots) corresponds to 1 square micrometer (µm²).

Correlation between regional Alu editing index and cortical hierarchy levels

The hierarchy levels of visual and somatosensory cortical regions in the macaque were determined based on the work of Felleman and Van Essen¹⁹. To rigorously assess the association between regional AEI and cortical hierarchy while controlling for potential confounders, we performed a linear mixed model (LMM) analysis using the lme4 package (v1.1-37). The model specified the regional AEI as the dependent variable. Cortical hierarchy was included as the primary fixed-effect predictor, with neuronal proportion, expression levels of ADAR1, ADAR2, and ADAR3, and total sequencing depth (coverage) as additional fixed-effect covariates. To account for the nested structure of the data (samples within individual donors), random intercepts were incorporated using the formula: AEI ~ Hierarchy + Neuronal proportion + ADAR1 + ADAR2 + ADAR3 + Coverage + (1 | Individual / Sample). The significance of fixed effects was assessed using Type III ANOVA. To specifically evaluate the contribution of cortical hierarchy, we compared the full model against a reduced model without the hierarchy term. The residual AEI, derived from the difference between the observed values and those predicted by the reduced model, was then used to confirm the unique association with hierarchy, independent of all other covariates. This relationship was evaluated using Spearman’s correlation.

Correlation between cell-type Alu editing index and cortical hierarchy levels

To rigorously assess the association between cell-type AEI and cortical hierarchy while controlling for potential confounders, we conducted a LMM analysis using the lme4 package (v1.1-37). The model specified neuronal subtype AEI as the dependent variable, with cortical hierarchy as the primary fixed-effect predictor. Additional fixed effects included neuronal subtype proportion, expression levels of ADAR1, ADAR2, and ADAR3, and total sequencing depth (coverage). The nested data structure (samples within individual donors) was modeled as random intercepts using the formula: AEI ~ Hierarchy + Neuronal subtype proportion + ADAR1 + ADAR2 + ADAR3 + Coverage + (1 | Individual / Sample). To specifically evaluate the contribution of cortical hierarchy, we compared the full model against a reduced model without the hierarchy term. The residual AEI, derived from the difference between the observed values and those predicted by the reduced model, was then used to confirm the unique association with hierarchy, independent of all other covariates. This relationship was evaluated using Spearman’s correlation.

Cell-type identification in the macaque DLPFC using snRNA-seq data

The raw expression data were obtained from the GEO dataset (GSE201687). To ensure data quality, low-quality nuclei were filtered out based on the criteria specified in the original study⁴⁴. Specifically, nuclei expressing fewer than 200 genes or with a mitochondrial UMI rate exceeding 20% were excluded. To normalize the data, the LogNormalize function in the Seurat package⁷⁰ (v4.4.0) was utilized. Principal component analysis (PCA) was then conducted, and the top ten principal components were selected for unsupervised clustering of the nuclei. To account for any batch effects across different samples, the Harmony⁷¹ (v1.2.0) was employed. The nuclei were visualized in a two-dimensional space using Uniform Manifold Approximation and Projection (UMAP) and subsequently categorized into distinct clusters. Additionally, major cell types were identified based on the expression of canonical marker genes. These included GAD1 for inhibitory neurons, SLC17A6 for excitatory neurons, CSF1R for microglia, AQP4 for astrocytes, MBP and PDGFRA for oligodendrocyte progenitor cells/oligodendrocytes.

Spatial domain identification of macaque DLPFC using Visium spatial transcriptomics

We obtained raw expression data from the GEO repository (accession: GSE201687). For spatial domain identification, we implemented a two-stage analytical approach. Initial clustering was performed using Scanpy⁶⁸ (version 1.8.2) with low resolution parameters. The analysis pipeline included data loading through the sc.read_visium function, followed by total count normalization (sc.pp.normalize_total) and log1p transformation (sc.pp.log1p), maintaining default parameters throughout the process. To enhance spatial domain resolution, we subsequently applied GraphST⁷² with optimized parameters. The spatial neighborhood radius was set to 50, and cluster identification was conducted using the mclust method. The optimal number of clusters (n_clusters) was determined based on preliminary Scanpy clustering results. Cortical layer annotation was performed using well-established layer-specific marker genes^43,73, including: CUX2 and GPR83 (layer 2), GPC5 and PDZD2 (layer 3), RORB (layer 4), ETV1 (layer 5), and TLE4 with SEMA3E (layer 6).

Differential RNA editing analysis between cell types or conditions

We aggregated the UMI counts supporting both edited and non-edited status for each editing site from single cells or spots corresponding to the same cell type within the same region. To ensure statistical reliability, editing sites were required to have a minimum of five reads supporting both edited and non-edited status. To identify significant differences in RNA editing levels between cell types, or between different conditions (e.g., disease vs. control) within the same cell type, we employed Fisher’s exact test. The resulting pvalues were adjusted using the Benjamini-Hochberg (BH) method to control for multiple testing. RNA editing sites that displayed a difference in editing level of at least 10% and an adjusted p value below 0.05 were classified as cell-type-enriched or condition-specific sites.

Identification of conserved editing sites

To identify evolutionarily conserved A-to-I RNA editing sites across mouse, macaque, and human, we integrated editing site datasets from all three species. We started with 15,683,855 human sites and 107,094 mouse sites from REDIportal (version 2.0), and supplemented these with 2,782,079 macaque editing sites from our own study. To enable a direct genomic comparison, we converted the genomic coordinates of the mouse (mm10) and macaque (macFas5) sites to the human reference genome (hg19) using the UCSC liftOver tool with the corresponding chain files (mm10ToHg19.over.chain.gz and macFas5ToHg19.over.chain.gz). We then intersected the lifted-over datasets to compile a comprehensive list of conserved editing sites, which yielded 160 sites that are edited in all three species.

Differential RNA editing analysis between species

To identify species-enriched RNA editing sites, we aggregated the UMI counts supporting edited and non-edited status for each conserved editing site from single cells corresponding to the same cell type across different species. We normalized the UMI counts from the combined set of selected cell types (EX-L2/3, EX-L6, IN-PV, and IN-SST) to a total depth of 100,000 UMI counts per species. The relative proportions of edited and unedited reads were preserved for each site. This normalization ensured that cross-species comparisons were not biased by differences in overall sequencing depth. However, despite normalization, species-specific differences in mapping efficiency or repeat content may still influence absolute editing levels. To ensure statistical reliability, only editing sites with a minimum of five reads supporting both edited and non-edited status were included in the analysis. To identify significant differences in RNA editing levels between species, we employed Fisher’s exact test. The resulting p values were adjusted using the Benjamini-Hochberg (BH) method to control for multiple testing. RNA editing sites exhibiting a difference in editing level of at least 10% and an adjusted p value below 0.05 were classified as species-enriched sites. Sites showing enriched editing patterns in both macaque and human compared to mouse were classified as primate-enriched sites.

Cross-species gene expression analysis

Single-nucleus/single-cell RNA sequencing (sn/scRNA-seq) data from mouse, macaque, and human were processed and analyzed. Four neuronal subtypes (EX-L2/3, EX-L6, IN-PV, and IN-SST) were isolated based on pre-annotated cell type labels using the Seurat package. Gene expression matrices were normalized to a scale of 10,000 transcripts per cell and natural log-transformed using the LogNormalize method. For cross-species comparison, we computed the mean expression of genes harboring evolutionarily conserved editing sites with Seurat’s AverageExpression function, grouping the normalized data by neuronal subtype and species.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Supplementary information

Supplementary Information^{(12.4MB, pdf)}

Peer Review file^{(2.2MB, pdf)}

41467_2026_77434_MOESM3_ESM.pdf^{(181.3KB, pdf)}

Description of Additional Supplementary Files

Supplementary Data 1-11^{(536.3KB, xlsx)}

Reporting Summary^{(144.6KB, pdf)}

Source data

Source Data^{(5.5MB, xlsx)}

Acknowledgements

We are grateful to Professor Cai Song for her insightful comments. We sincerely thank the China National GeneBank for providing technical support.

Author contributions

J.H. designed the study and developed the framework. J.H., C.C., J.Y., Z.Z., Z.W., Y.H., L.L., and T.L. conducted the study. Q.L. and Y.H. provided valuable advice. J.H., C.C., J.Y., Z.Z., and L.D. prepared the figures. J.H. led data interpretation and wrote the manuscript with inputs from all other authors. J.H., S.L., and Y.L. co-supervised the study.

Peer review

Peer review information

Nature Communications thanks Meng How Tan and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

The project was supported by National Key R&D Program of China (2024YFC3505100), National Natural Science Foundation of China (No. 32571341), National Key R&D Program of China (2022YEF0203200, 2022YFC3400400, 2021YFA0805100, and 2020YFE0205900), National Science and Technology Innovation 2030 Major Program (STI2030-2022ZD0211700, STI2030-2022ZD0205000 and STI2030-2021ZD0200100), Zhejiang Province ‘Vanguard’ R&D Program (2023C03SA103409), and Hangzhou Leading Innovation Team Project.

Data availability

The raw WGS and bulk RNA-seq data of the macaque are available in the China National GeneBank (CNGB) database (https://db.cngb.org) under accession number CNP0002819. The raw stereo-seq and snRNA-seq data of the macaque are also available in CNGB (CNP0002035). The macaque bulk RNA-seq data used for CDS site validation can be accessed in the NCBI Sequence Read Archive (SRA) under accession numbers DRP005990, SRP164572, and SRP410543. The Visium and snRNA-seq data of the macaque are available in the Gene Expression Omnibus (GEO) under accession number GSE201687. The Visium data of the mouse are available at 10x Genomics (https://www.10xgenomics.com/datasets/adult-mouse-brain-coronal-section-fresh-frozen-1-standard) and in GEO (GSE153859). The scRNA-seq data of the mouse cortex are available in GEO under accession number GSE185862. The snRNA-seq data of human cortex are available in GEO (GSE144136, GSE213982, and GSE207334) and SRA (SRP132816). Source data are provided with this paper.

Code availability

The original code for SRDRE is available in GitHub (https://github.com/JinRcn/SRDRE) and Zenodo (https://doi.org/10.5281/zenodo.21759254)⁷⁴.

Competing interests

The authors declare no competing interests.

Footnotes

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

These authors contributed equally: Cheng Chen, Jing Yang.

Contributor Information

Ying Lei, Email: leiying1@genomics.cn.

Jinrong Huang, Email: jinrcn@foxmail.com.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-77434-y.

References

1.Yablonovitch, A. L., Deng, P., Jacobson, D. & Li, J. B. The evolution and adaptation of A-to-I RNA editing. PLoS Genet.13, e1007064 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
2.Porath, H. T., Knisbacher, B. A., Eisenberg, E. & Levanon, E. Y. Massive A-to-I RNA editing is common across the Metazoa and correlates with dsRNA abundance. Genome Biol.18, 185 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
3.Eisenberg, E. & Levanon, E. Y. A-to-I RNA editing—immune protector and transcriptome diversifier. Nat. Rev. Genet.19, 473–490 (2018). [DOI] [PubMed] [Google Scholar]
4.Zhang, P. et al. On the origin and evolution of RNA editing in metazoans. Cell Rep.42, 112112 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
5.Hwang, T. et al. Dynamic regulation of RNA editing in human brain development and disease. Nat. Neurosci.19, 1093–1099 (2016). [DOI] [PubMed] [Google Scholar]
6.Rajendren, S. et al. Profiling neural editomes reveals a molecular mechanism to regulate RNA editing during development. Genome Res.31, 27–39 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
7.Cuddleston, W. H. et al. Spatiotemporal and genetic regulation of A-to-I editing throughout human brain development. Cell Rep.41, 111585 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
8.Nishikura, K. A-to-I editing of coding and non-coding RNAs by ADARs. Nat. Rev. Mol. Cell Biol.17, 83–96 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
9.Rosenthal, J. J. & Seeburg, P. H. A-to-I RNA editing: effects on proteins key to neural excitability. Neuron74, 432–439 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
10.Behm, M. & Ohman, M. RNA editing: a contributor to neuronal dynamics in the mammalian brain. Trends Genet.32, 165–175 (2016). [DOI] [PubMed] [Google Scholar]
11.Choudhury, M. et al. Widespread RNA hypoediting in schizophrenia and its relevance to mitochondrial function. Sci. Adv.9, eade9997 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
12.Breen, M. S. et al. Global landscape and genetic regulation of RNA editing in cortical samples from individuals with schizophrenia. Nat. Neurosci.22, 1402–1412 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
13.Tran, S. S. et al. Widespread RNA editing dysregulation in brains from autistic individuals. Nat. Neurosci.22, 25–36 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
14.Khermesh, K. et al. Reduced levels of protein recoding by A-to-I RNA editing in Alzheimer’s disease. RNA22, 290–302 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
15.Ma, Y. et al. Atlas of RNA editing events affecting protein expression in aged and Alzheimer’s disease human brain tissue. Nat. Commun.12, 7035 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
16.Chen, C. et al. Comprehensive characterization of the transcriptional landscape in Alzheimer’s disease (AD) brains. Sci. Adv.11, eadn1927 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
17.Weissmann, D. et al. Region-specific alterations of A-to-I RNA editing of serotonin 2c receptor in the cortex of suicides with major depression. Transl. Psychiatry6, e878 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
18.Van Essen, D. C. & Glasser, M. F. Parcellating cerebral cortex: how invasive animal studies inform noninvasive mapmaking in humans. Neuron99, 640–663 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
19.Felleman, D. J. & Van Essen, D. C. Distributed hierarchical processing in the primate cerebral cortex. Cereb. Cortex1, 1–47 (1991). [DOI] [PubMed] [Google Scholar]
20.Eisenberg, E. et al. Is abundant A-to-I RNA editing primate-specific? Trends Genet.21, 77–81 (2005). [DOI] [PubMed] [Google Scholar]
21.Dillman, A. A. et al. mRNA expression, splicing and editing in the embryonic and adult mouse cerebral cortex. Nat. Neurosci.16, 499–506 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
22.Chen, J. Y. et al. RNA editome in rhesus macaque shaped by purifying selection. PLoS Genet.10, e1004274 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
23.Li, Q. et al. Caste-specific RNA editomes in the leaf-cutting ant Acromyrmex echinatior. Nat. Commun.5, 4943 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
24.Tan, M. H. et al. Dynamic landscape and regulation of RNA editing in mammals. Nature550, 249–254 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
25.Ye, L. Q. et al. The RNA editome of Macaca mulatta and functional characterization of RNA editing in mitochondria. Sci. Bull.62, 820–830 (2017). [DOI] [PubMed] [Google Scholar]
26.Liscovitch-Brauer, N. et al. Trade-off between transcriptome plasticity and genome evolution in cephalopods. Cell169, 191–202.e11 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
27.Buchumenski, I. et al. Systematic identification of A-to-I RNA editing in zebrafish development and adult organs. Nucleic Acids Res.49, 4325–4337 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
28.Duan, Y. et al. A-to-I RNA editing in honeybees shows signals of adaptation and convergent evolution. iScience24, 101983 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
29.Huang, J. et al. A porcine brain-wide RNA editing landscape. Commun. Biol.4, 717 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
30.Vallecillo-Viejo, I. C. et al. Spatially regulated editing of genetic information within a neuron. Nucleic Acids Res.48, 3999–4012 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
31.Jepson, J. E., Savva, Y. A., Jay, K. A. & Reenan, R. A. Visualizing adenosine-to-inosine RNA editing in the Drosophila nervous system. Nat. Methods9, 189–194 (2011). [DOI] [PubMed] [Google Scholar]
32.Mellis, I. A., Gupte, R., Raj, A. & Rouhanifard, S. H. Visualizing adenosine-to-inosine RNA editing in single mammalian cells. Nat. Methods14, 801–804 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
33.Lundin, E. et al. Spatiotemporal mapping of RNA editing in the developing mouse brain using in situ sequencing reveals regional and cell-type-specific regulation. BMC Biol.18, 6 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
34.Quillin, A. L., Arnould, B., Knutson, S. D. & Heemstra, J. M. Spatial visualization of A-to-I editing in cells using endonuclease V immunostaining assay (EndoVIA). ACS Cent. Sci.10, 1396–1405 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
35.Bai, Z. et al. Spatially exploring RNA biology in archival formalin-fixed paraffin-embedded tissues. Cell187, 6760–6779.e24 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
36.Lebrigand, K. et al. The spatial landscape of gene expression isoforms in tissue sections. Nucleic Acids Res.51, e47 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
37.Zhang, Q. & Xiao, X. Genome sequence-independent identification of RNA editing sites. Nat. Methods12, 347–350 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
38.Lo Giudice, C., Tangaro, M. A., Pesole, G. & Picardi, E. Investigating RNA editing in deep transcriptome datasets with REDItools and REDIportal. Nat. Protoc.15, 1098–1131 (2020). [DOI] [PubMed] [Google Scholar]
39.Wang, Z. et al. RES-Scanner: a software package for genome-wide identification of RNA-editing sites. Gigascience5, 37 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
40.John, D., Weirick, T., Dimmeler, S. & Uchida, S. RNAEditor: easy detection of RNA editing events and the introduction of editing islands. Brief. Bioinform.18, 993–1001 (2017). [DOI] [PubMed] [Google Scholar]
41.Fu, Z. C., Gao, B. Q., Nan, F., Ma, X. K. & Yang, L. DEMINING: a deep learning model embedded framework to distinguish RNA editing from DNA mutations in RNA sequencing data. Genome Biol.25, 258 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
42.Yao, Z. et al. A taxonomy of transcriptomic cell types across the isocortex and hippocampal formation. Cell184, 3222–3241.e26 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
43.Chen, A. et al. Single-cell spatial transcriptome reveals cell-type organization in the macaque cortex. Cell186, 3726–3743.e24 (2023). [DOI] [PubMed] [Google Scholar]
44.Wu, J. et al. Integrating spatial and single-nucleus transcriptomic data elucidates microglial-specific responses in female cynomolgus macaques with depressive-like behaviors. Nat. Neurosci.26, 1352–1364 (2023). [DOI] [PubMed] [Google Scholar]
45.Barak, M. et al. Evidence for large diversity in the human transcriptome created by Alu RNA editing. Nucleic Acids Res.37, 6905–6915 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]
46.An, N. A. et al. Evolutionarily significant A-to-I RNA editing events originated through G-to-A mutations in primates. Genome Biol.20, 24 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
47.Roth, S. H., Levanon, E. Y. & Eisenberg, E. Genome-wide quantification of ADAR adenosine-to-inosine RNA editing activity. Nat. Methods16, 1131–1138 (2019). [DOI] [PubMed] [Google Scholar]
48.Maitra, M. et al. Cell type specific transcriptomic differences in depression show similar patterns between males and females but implicate distinct cell types and genes. Nat. Commun.14, 2912 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
49.Cuddleston, W. H. et al. Cellular and genetic drivers of RNA editing variation in the human brain. Nat. Commun.13, 2997 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
50.Karlstrom, V. et al. ADAR3 modulates neuronal differentiation and regulates mRNA stability and translation. Nucleic Acids Res.52, 12021–12038 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
51.Ansell, B. R. E. et al. A survey of RNA editing at single-cell resolution links interneurons to schizophrenia and autism. RNA27, 1482–1496 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
52.Belgard, T. G. et al. A transcriptomic atlas of mouse neocortical layers. Neuron71, 605–616 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
53.He, Z. et al. Comprehensive transcriptome analysis of neocortical layers in humans, chimpanzees and macaques. Nat. Neurosci.20, 886–895 (2017). [DOI] [PubMed] [Google Scholar]
54.Velmeshev, D. et al. Single-cell genomics identifies cell type-specific molecular changes in autism. Science364, 685–689 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
55.Nagy, C. et al. Single-nucleus transcriptomics of the prefrontal cortex in major depressive disorder implicates oligodendrocyte precursor cells and excitatory neurons. Nat. Neurosci.23, 771–781 (2020). [DOI] [PubMed] [Google Scholar]
56.Ma, S. et al. Molecular and cellular evolution of the primate dorsolateral prefrontal cortex. Science377, eabo7257 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
57.Lomeli, H. et al. Control of kinetic properties of AMPA receptor channels by nuclear RNA editing. Science266, 1709–1713 (1994). [DOI] [PubMed] [Google Scholar]
58.Rakic, P. Evolution of the neocortex: a perspective from developmental biology. Nat. Rev. Neurosci.10, 724–735 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]
59.Paz-Yaacov, N. et al. Adenosine-to-inosine RNA editing shapes transcriptome diversity in primates. Proc. Natl. Acad. Sci. USA107, 12174–12179 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
60.Pinto, Y., Cohen, H. Y. & Levanon, E. Y. Mammalian conserved ADAR targets comprise only a small fragment of the human editosome. Genome Biol.15, R5 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
61.Yeo, J., Goodman, R. A., Schirle, N. T., David, S. S. & Beal, P. A. RNA editing changes the lesion specificity for the DNA repair enzyme NEIL1. Proc. Natl. Acad. Sci. USA107, 20715–20719 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
62.Gabay, O. et al. Landscape of adenosine-to-inosine RNA recoding across human tissues. Nat. Commun.13, 1184 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
63.Hijazi, S., Smit, A. B. & van Kesteren, R. E. Fast-spiking parvalbumin-positive interneurons in brain physiology and Alzheimer’s disease. Mol. Psychiatry28, 4954–4967 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
64.Hu, H., Gan, J. & Jonas, P. Interneurons. Fast-spiking, parvalbumin(+) GABAergic interneurons: from cellular design to microcircuit function. Science345, 1255263 (2014). [DOI] [PubMed] [Google Scholar]
65.Sohal, V. S., Zhang, F., Yizhar, O. & Deisseroth, K. Parvalbumin neurons and gamma rhythms enhance cortical circuit performance. Nature459, 698–702 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]
66.Li, H. & Durbin, R. Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics25, 1754–1760 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]
67.Kim, D., Paggi, J. M., Park, C., Bennett, C. & Salzberg, S. L. Graph-based genome alignment and genotyping with HISAT2 and HISAT-genotype. Nat. Biotechnol.37, 907–915 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
68.Wolf, F. A., Angerer, P. & Theis, F. J. SCANPY: large-scale single-cell gene expression data analysis. Genome Biol.19, 15 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
69.Gong, C. et al. SAW: an efficient and accurate data analysis workflow for Stereo-seq spatial transcriptomics. GigaByte2024, gigabyte111 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
70.Hao, Y. et al. Integrated analysis of multimodal single-cell data. Cell184, 3573–3587.e29 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
71.Korsunsky, I. et al. Fast, sensitive and accurate integration of single-cell data with Harmony. Nat. Methods16, 1289–1296 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
72.Long, Y. et al. Spatially informed clustering, integration, and deconvolution of spatial transcriptomics with GraphST. Nat. Commun.14, 1155 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
73.Herring, C. A. et al. Human prefrontal cortex gene regulatory dynamics from gestation to adulthood at single-cell resolution. Cell185, 4428–4447.e28 (2022). [DOI] [PubMed] [Google Scholar]
74. Huang, J. JinRcn/SRDRE: v1.0. Zenodo 10.5281/zenodo.21759254 (2026). [DOI]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Information^{(12.4MB, pdf)}

Peer Review file^{(2.2MB, pdf)}

41467_2026_77434_MOESM3_ESM.pdf^{(181.3KB, pdf)}

Description of Additional Supplementary Files

Supplementary Data 1-11^{(536.3KB, xlsx)}

Reporting Summary^{(144.6KB, pdf)}

Source Data^{(5.5MB, xlsx)}

Data Availability Statement

The original code for SRDRE is available in GitHub (https://github.com/JinRcn/SRDRE) and Zenodo (https://doi.org/10.5281/zenodo.21759254)⁷⁴.

[CR1] 1.Yablonovitch, A. L., Deng, P., Jacobson, D. & Li, J. B. The evolution and adaptation of A-to-I RNA editing. PLoS Genet.13, e1007064 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR2] 2.Porath, H. T., Knisbacher, B. A., Eisenberg, E. & Levanon, E. Y. Massive A-to-I RNA editing is common across the Metazoa and correlates with dsRNA abundance. Genome Biol.18, 185 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR3] 3.Eisenberg, E. & Levanon, E. Y. A-to-I RNA editing—immune protector and transcriptome diversifier. Nat. Rev. Genet.19, 473–490 (2018). [DOI] [PubMed] [Google Scholar]

[CR4] 4.Zhang, P. et al. On the origin and evolution of RNA editing in metazoans. Cell Rep.42, 112112 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR5] 5.Hwang, T. et al. Dynamic regulation of RNA editing in human brain development and disease. Nat. Neurosci.19, 1093–1099 (2016). [DOI] [PubMed] [Google Scholar]

[CR6] 6.Rajendren, S. et al. Profiling neural editomes reveals a molecular mechanism to regulate RNA editing during development. Genome Res.31, 27–39 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR7] 7.Cuddleston, W. H. et al. Spatiotemporal and genetic regulation of A-to-I editing throughout human brain development. Cell Rep.41, 111585 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR8] 8.Nishikura, K. A-to-I editing of coding and non-coding RNAs by ADARs. Nat. Rev. Mol. Cell Biol.17, 83–96 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR9] 9.Rosenthal, J. J. & Seeburg, P. H. A-to-I RNA editing: effects on proteins key to neural excitability. Neuron74, 432–439 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR10] 10.Behm, M. & Ohman, M. RNA editing: a contributor to neuronal dynamics in the mammalian brain. Trends Genet.32, 165–175 (2016). [DOI] [PubMed] [Google Scholar]

[CR11] 11.Choudhury, M. et al. Widespread RNA hypoediting in schizophrenia and its relevance to mitochondrial function. Sci. Adv.9, eade9997 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR12] 12.Breen, M. S. et al. Global landscape and genetic regulation of RNA editing in cortical samples from individuals with schizophrenia. Nat. Neurosci.22, 1402–1412 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR13] 13.Tran, S. S. et al. Widespread RNA editing dysregulation in brains from autistic individuals. Nat. Neurosci.22, 25–36 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR14] 14.Khermesh, K. et al. Reduced levels of protein recoding by A-to-I RNA editing in Alzheimer’s disease. RNA22, 290–302 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR15] 15.Ma, Y. et al. Atlas of RNA editing events affecting protein expression in aged and Alzheimer’s disease human brain tissue. Nat. Commun.12, 7035 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR16] 16.Chen, C. et al. Comprehensive characterization of the transcriptional landscape in Alzheimer’s disease (AD) brains. Sci. Adv.11, eadn1927 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR17] 17.Weissmann, D. et al. Region-specific alterations of A-to-I RNA editing of serotonin 2c receptor in the cortex of suicides with major depression. Transl. Psychiatry6, e878 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR18] 18.Van Essen, D. C. & Glasser, M. F. Parcellating cerebral cortex: how invasive animal studies inform noninvasive mapmaking in humans. Neuron99, 640–663 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR19] 19.Felleman, D. J. & Van Essen, D. C. Distributed hierarchical processing in the primate cerebral cortex. Cereb. Cortex1, 1–47 (1991). [DOI] [PubMed] [Google Scholar]

[CR20] 20.Eisenberg, E. et al. Is abundant A-to-I RNA editing primate-specific? Trends Genet.21, 77–81 (2005). [DOI] [PubMed] [Google Scholar]

[CR21] 21.Dillman, A. A. et al. mRNA expression, splicing and editing in the embryonic and adult mouse cerebral cortex. Nat. Neurosci.16, 499–506 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR22] 22.Chen, J. Y. et al. RNA editome in rhesus macaque shaped by purifying selection. PLoS Genet.10, e1004274 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR23] 23.Li, Q. et al. Caste-specific RNA editomes in the leaf-cutting ant Acromyrmex echinatior. Nat. Commun.5, 4943 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR24] 24.Tan, M. H. et al. Dynamic landscape and regulation of RNA editing in mammals. Nature550, 249–254 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR25] 25.Ye, L. Q. et al. The RNA editome of Macaca mulatta and functional characterization of RNA editing in mitochondria. Sci. Bull.62, 820–830 (2017). [DOI] [PubMed] [Google Scholar]

[CR26] 26.Liscovitch-Brauer, N. et al. Trade-off between transcriptome plasticity and genome evolution in cephalopods. Cell169, 191–202.e11 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR27] 27.Buchumenski, I. et al. Systematic identification of A-to-I RNA editing in zebrafish development and adult organs. Nucleic Acids Res.49, 4325–4337 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR28] 28.Duan, Y. et al. A-to-I RNA editing in honeybees shows signals of adaptation and convergent evolution. iScience24, 101983 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR29] 29.Huang, J. et al. A porcine brain-wide RNA editing landscape. Commun. Biol.4, 717 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR30] 30.Vallecillo-Viejo, I. C. et al. Spatially regulated editing of genetic information within a neuron. Nucleic Acids Res.48, 3999–4012 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR31] 31.Jepson, J. E., Savva, Y. A., Jay, K. A. & Reenan, R. A. Visualizing adenosine-to-inosine RNA editing in the Drosophila nervous system. Nat. Methods9, 189–194 (2011). [DOI] [PubMed] [Google Scholar]

[CR32] 32.Mellis, I. A., Gupte, R., Raj, A. & Rouhanifard, S. H. Visualizing adenosine-to-inosine RNA editing in single mammalian cells. Nat. Methods14, 801–804 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR33] 33.Lundin, E. et al. Spatiotemporal mapping of RNA editing in the developing mouse brain using in situ sequencing reveals regional and cell-type-specific regulation. BMC Biol.18, 6 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR34] 34.Quillin, A. L., Arnould, B., Knutson, S. D. & Heemstra, J. M. Spatial visualization of A-to-I editing in cells using endonuclease V immunostaining assay (EndoVIA). ACS Cent. Sci.10, 1396–1405 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR35] 35.Bai, Z. et al. Spatially exploring RNA biology in archival formalin-fixed paraffin-embedded tissues. Cell187, 6760–6779.e24 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR36] 36.Lebrigand, K. et al. The spatial landscape of gene expression isoforms in tissue sections. Nucleic Acids Res.51, e47 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR37] 37.Zhang, Q. & Xiao, X. Genome sequence-independent identification of RNA editing sites. Nat. Methods12, 347–350 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR38] 38.Lo Giudice, C., Tangaro, M. A., Pesole, G. & Picardi, E. Investigating RNA editing in deep transcriptome datasets with REDItools and REDIportal. Nat. Protoc.15, 1098–1131 (2020). [DOI] [PubMed] [Google Scholar]

[CR39] 39.Wang, Z. et al. RES-Scanner: a software package for genome-wide identification of RNA-editing sites. Gigascience5, 37 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR40] 40.John, D., Weirick, T., Dimmeler, S. & Uchida, S. RNAEditor: easy detection of RNA editing events and the introduction of editing islands. Brief. Bioinform.18, 993–1001 (2017). [DOI] [PubMed] [Google Scholar]

[CR41] 41.Fu, Z. C., Gao, B. Q., Nan, F., Ma, X. K. & Yang, L. DEMINING: a deep learning model embedded framework to distinguish RNA editing from DNA mutations in RNA sequencing data. Genome Biol.25, 258 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR42] 42.Yao, Z. et al. A taxonomy of transcriptomic cell types across the isocortex and hippocampal formation. Cell184, 3222–3241.e26 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR43] 43.Chen, A. et al. Single-cell spatial transcriptome reveals cell-type organization in the macaque cortex. Cell186, 3726–3743.e24 (2023). [DOI] [PubMed] [Google Scholar]

[CR44] 44.Wu, J. et al. Integrating spatial and single-nucleus transcriptomic data elucidates microglial-specific responses in female cynomolgus macaques with depressive-like behaviors. Nat. Neurosci.26, 1352–1364 (2023). [DOI] [PubMed] [Google Scholar]

[CR45] 45.Barak, M. et al. Evidence for large diversity in the human transcriptome created by Alu RNA editing. Nucleic Acids Res.37, 6905–6915 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR46] 46.An, N. A. et al. Evolutionarily significant A-to-I RNA editing events originated through G-to-A mutations in primates. Genome Biol.20, 24 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR47] 47.Roth, S. H., Levanon, E. Y. & Eisenberg, E. Genome-wide quantification of ADAR adenosine-to-inosine RNA editing activity. Nat. Methods16, 1131–1138 (2019). [DOI] [PubMed] [Google Scholar]

[CR48] 48.Maitra, M. et al. Cell type specific transcriptomic differences in depression show similar patterns between males and females but implicate distinct cell types and genes. Nat. Commun.14, 2912 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR49] 49.Cuddleston, W. H. et al. Cellular and genetic drivers of RNA editing variation in the human brain. Nat. Commun.13, 2997 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR50] 50.Karlstrom, V. et al. ADAR3 modulates neuronal differentiation and regulates mRNA stability and translation. Nucleic Acids Res.52, 12021–12038 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR51] 51.Ansell, B. R. E. et al. A survey of RNA editing at single-cell resolution links interneurons to schizophrenia and autism. RNA27, 1482–1496 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR52] 52.Belgard, T. G. et al. A transcriptomic atlas of mouse neocortical layers. Neuron71, 605–616 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR53] 53.He, Z. et al. Comprehensive transcriptome analysis of neocortical layers in humans, chimpanzees and macaques. Nat. Neurosci.20, 886–895 (2017). [DOI] [PubMed] [Google Scholar]

[CR54] 54.Velmeshev, D. et al. Single-cell genomics identifies cell type-specific molecular changes in autism. Science364, 685–689 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR55] 55.Nagy, C. et al. Single-nucleus transcriptomics of the prefrontal cortex in major depressive disorder implicates oligodendrocyte precursor cells and excitatory neurons. Nat. Neurosci.23, 771–781 (2020). [DOI] [PubMed] [Google Scholar]

[CR56] 56.Ma, S. et al. Molecular and cellular evolution of the primate dorsolateral prefrontal cortex. Science377, eabo7257 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR57] 57.Lomeli, H. et al. Control of kinetic properties of AMPA receptor channels by nuclear RNA editing. Science266, 1709–1713 (1994). [DOI] [PubMed] [Google Scholar]

[CR58] 58.Rakic, P. Evolution of the neocortex: a perspective from developmental biology. Nat. Rev. Neurosci.10, 724–735 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR59] 59.Paz-Yaacov, N. et al. Adenosine-to-inosine RNA editing shapes transcriptome diversity in primates. Proc. Natl. Acad. Sci. USA107, 12174–12179 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR60] 60.Pinto, Y., Cohen, H. Y. & Levanon, E. Y. Mammalian conserved ADAR targets comprise only a small fragment of the human editosome. Genome Biol.15, R5 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR61] 61.Yeo, J., Goodman, R. A., Schirle, N. T., David, S. S. & Beal, P. A. RNA editing changes the lesion specificity for the DNA repair enzyme NEIL1. Proc. Natl. Acad. Sci. USA107, 20715–20719 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR62] 62.Gabay, O. et al. Landscape of adenosine-to-inosine RNA recoding across human tissues. Nat. Commun.13, 1184 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR63] 63.Hijazi, S., Smit, A. B. & van Kesteren, R. E. Fast-spiking parvalbumin-positive interneurons in brain physiology and Alzheimer’s disease. Mol. Psychiatry28, 4954–4967 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR64] 64.Hu, H., Gan, J. & Jonas, P. Interneurons. Fast-spiking, parvalbumin(+) GABAergic interneurons: from cellular design to microcircuit function. Science345, 1255263 (2014). [DOI] [PubMed] [Google Scholar]

[CR65] 65.Sohal, V. S., Zhang, F., Yizhar, O. & Deisseroth, K. Parvalbumin neurons and gamma rhythms enhance cortical circuit performance. Nature459, 698–702 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR66] 66.Li, H. & Durbin, R. Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics25, 1754–1760 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR67] 67.Kim, D., Paggi, J. M., Park, C., Bennett, C. & Salzberg, S. L. Graph-based genome alignment and genotyping with HISAT2 and HISAT-genotype. Nat. Biotechnol.37, 907–915 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR68] 68.Wolf, F. A., Angerer, P. & Theis, F. J. SCANPY: large-scale single-cell gene expression data analysis. Genome Biol.19, 15 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR69] 69.Gong, C. et al. SAW: an efficient and accurate data analysis workflow for Stereo-seq spatial transcriptomics. GigaByte2024, gigabyte111 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR70] 70.Hao, Y. et al. Integrated analysis of multimodal single-cell data. Cell184, 3573–3587.e29 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR71] 71.Korsunsky, I. et al. Fast, sensitive and accurate integration of single-cell data with Harmony. Nat. Methods16, 1289–1296 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR72] 72.Long, Y. et al. Spatially informed clustering, integration, and deconvolution of spatial transcriptomics with GraphST. Nat. Commun.14, 1155 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR73] 73.Herring, C. A. et al. Human prefrontal cortex gene regulatory dynamics from gestation to adulthood at single-cell resolution. Cell185, 4428–4447.e28 (2022). [DOI] [PubMed] [Google Scholar]

[CR74] 74. Huang, J. JinRcn/SRDRE: v1.0. Zenodo 10.5281/zenodo.21759254 (2026). [DOI]

PERMALINK

Spatial transcriptomics reveals laminar and cell-type-specific A-to-I RNA editing signatures in the macaque cortex

Cheng Chen

Jing Yang

Zhiyong Zhu

Ziyu Wang

Yunqi Huang

Lingjun Ding

Lidan Luo

Tianming Lan

Yong Hou

Qiye Li

Shiping Liu

Ying Lei

Jinrong Huang

Abstract

Introduction

Results

Overview of SRDRE

Fig. 1. Identification of A-to-I RNA editing using spatial transcriptomics data.

Comparisons of 3’-biased sequencing and full-transcript coverage methods

Assessment of coding site coverage in spatial transcriptomics data

SRDRE reveals regional RNA editing differences in the mouse brain

Visualizing A-to-I editing in single cells

Quantifying global RNA editing activity using the Alu editing index

Linking cell-type-specific RNA editing dynamics to cortical hierarchy

Fig. 2. RNA editing patterns across cortical regions in the macaque brain.

Layer- and cell-type-specific ADAR gene expression

Parvalbumin neurons exhibit the highest RNA editing levels in most cortical regions

Fig. 3. Cell-type-specific RNA editing patterns.

More CDS sites are enriched in inhibitory neurons than excitatory neurons

Fig. 4. Coding sequence (CDS) sites in inhibitory and excitatory neurons.

Primate-enriched RNA editing patterns occur at evolutionarily conserved sites

Fig. 5. Cross-species comparisons of RNA editing in conserved sites among mouse, macaque, and human.

Altered RNA editing in the DLPFC of macaques with depressive-like behaviors

Fig. 6. Alterations of RNA editing in macaques exhibiting depressive-like behaviors and humans with major depressive disorder (MDD).

Discussion

Methods

Ethics statement

A-to-I RNA editing catalog of mice and humans

Characterization of macaque A-to-I RNA editing via detection of DNA-RNA differences

Validation of macaque CDS sites using independent bulk RNA-seq data

Comparative analysis of RNA editing in scRNA-seq data using two distinct platforms

Analysis of spatial RNA sequencing data from the mouse brain section

Aligning spatial RNA sequencing data to the macaque genome

A-to-I RNA editing detection using spatial RNA sequencing data

Quantification of RNA editing index

Quantification of RNA editing at cell-type resolution in macaque Stereo-seq data

Comparisons between two adjacent sections from macaque PFC

Neuron density

Correlation between regional Alu editing index and cortical hierarchy levels

Correlation between cell-type Alu editing index and cortical hierarchy levels

Cell-type identification in the macaque DLPFC using snRNA-seq data

Spatial domain identification of macaque DLPFC using Visium spatial transcriptomics

Differential RNA editing analysis between cell types or conditions

Identification of conserved editing sites

Differential RNA editing analysis between species

Cross-species gene expression analysis

Reporting summary

Supplementary information

Source data

Acknowledgements

Author contributions

Peer review

Peer review information

Funding

Data availability

Code availability

Competing interests

Footnotes

Contributor Information

Supplementary information

References

Associated Data

Supplementary Materials

Data Availability Statement

ACTIONS

PERMALINK

RESOURCES

Similar articles