NEUROD1 efficiently converts peripheral blood cells into neurons with partial reprogramming by pluripotency factors

Significance

Here, we report a procedure for converting peripheral blood cells into functional neurons. Our method utilizes partial reprogramming strategies with Yamanaka reprogramming factors and the neuronal bHLH NEUROD1 to overcome epigenetic barriers. Cells produced through this process exhibit characteristics and genetic signatures similar to those of glutamatergic neurons. This method offers a minimally invasive and rapid culture protocol that can significantly advance the generation of patient-specific neurons for disease modeling and therapeutic development. While the conversion of blood cells into functional neurons has already been demonstrated earlier, the method reported in the current study introduces a different set of genes using viral vectors, significantly improving cell survival.

Abstract

The direct reprogramming of cells has tremendous potential in in vitro neurological studies. Previous attempts to convert blood cells into induced neurons have presented several challenges, necessitating a less invasive, efficient, rapid, and convenient approach. The current study introduces an optimized method for converting somatic cells into neurons using a nonsurgical approach that employs peripheral blood cells as an alternative source to fibroblasts. We have demonstrated the efficacy of a unique combination of transcription factors, including NEUROD1, and four Yamanaka reprogramming factors (OCT3/4, SOX2, KLF4, and c-MYC), in generating glutamatergic neurons within 3 wk. This approach, which requires only five pivotal factors (NEUROD1, OCT3/4, SOX2, KLF4, and c-MYC), has the potential to create functional neurons and circumvents the need for induced pluripotent stem cell (iPSC) intermediates, as evidenced by single-cell RNA sequencing and whole-genome bisulfite sequencing, along with lineage-tracing experiments using Cre-LoxP system. While fibroblasts have been widely used for neuronal reprogramming, our findings suggest that peripheral blood cells offer a potential alternative, particularly in contexts where minimally invasive sampling and procedures convenient for patients are emphasized. This method provides a rapid strategy for modeling neuronal diseases and contributes to advancements in drug discovery and personalized medicine.

The direct reprogramming of cells, a process that converts somatic cells into a variety of cell types by introducing specific genes, has immense potential for in vitro neurological studies (1). To overcome the obstacle of limited access to human neuronal tissue, the efficient generation of neurons from somatic cells has been achieved over the past decade (2). In particular, the introduction of neuronal bHLH-type transcription factors (TFs) such as ASCL1 and NEUROG/NEUROD family members, along with miRNA-9 and 124, has enabled the efficient conversion of fibroblasts into neuron-like cells in approximately 20 d (3–5).

However, surgical collection of dermal fibroblasts, which are typically used as source cells for direct-neuronal conversion, has slowed research progress. Hence, we aimed to explore less invasive alternatives by focusing on peripheral blood, which is an easily accessible and commonly used biopsy material. Prior attempts to convert blood cells into neurons (6–9) presented several challenges, such as the incorporation of multiple cellular processing steps, prolonged durations extending over several months, and low differentiation conversion efficiency (often below 10%); we sought to address these obstacles in the current study.

In this study, we developed a method for converting T cells into neurons by cotransducing NEUROD1 with Yamanaka reprogramming factors (OCT3/4, SOX2, KLF4, and c-MYC: OSKM) (10) into T cells using Sendai virus (SeV), which is known to cause minimal chromosomal damage as a single-stranded RNA virus (11, 12). Twenty days posttransduction, neuron-like cells were formed. Comprehensive gene expression and epigenomic analyses, and functional assessments, along with cell-type specific lineage-tracing experiments using Cre-LoxP system, indicated that a significant subset of cells transitioned into neuronal cells, with minimal indication of reverting to a pluripotent state. Additionally, DNA methylation analysis suggested that while these neurons retained certain epigenomic signatures from T cells, they also exhibited partial rejuvenation effects, potentially influenced by transient OSKM expression. Herein, we provide a detailed description of an efficient and convenient protocol for generating neuronal cells from T cells.

Results

Generation of Neuronal Cells from Human Peripheral Blood Cells.

We first tested whether T cell–derived neuronal cells could be established from blood cells using four factors, namely NEUROD1, ASCL1, BRN2/POU3F2, and ZIC1 (N/A/B/Z), which have been previously used for neuronal induction in fibroblasts (3). T cells were stimulated and expanded using an anti-CD3 antibody. Subsequently, an optimal number of T cells was transduced with N/A/B/Z via Sendai virus at a multiplicity of infection (MOI) of 10. One day posttransduction, the virus was removed by medium replacement, and the cells were maintained in a medium with an adherent substrate (Fig. 1A and SI Appendix, Table S1), with medium changes every 3 d. Under these culture conditions, after 20 d, adherence and neuron-like morphological developments were minimal. Significant cell loss was observed during medium changes. We introduced four Yamanaka TFs O/S/K/M into our protocol, aiming to minimize cell loss, as adherent cells are usually observed during induced pluripotent stem cells (iPSCs) establishment from T cells. Remarkably, combining these O/S/K/M factors with the N/A/B/Z factors resulted in adherent cells within a week posttransduction. The cells began to develop neuron-like processes in the second week and exhibited complex dendritic morphology by the third week (Fig. 1B). Immunostaining confirmed the presence of βIII-tubulin and MAP2, although the survival rate was low (less than 50%). To improve the survival rate, we added 10% fetal bovine serum (FBS) to the differentiation medium, which resulted in a 10-fold increase in neuron yield. Consequently, the proportion of βIII-Tubulin positive cells was 119.7% relative to the initial/seed cell number and 22.4% relative to the surviving cells in the dish (Fig. 1 C–E).

Fig. 1.

Generation of neuronal cells from human peripheral blood cells. (A) Experimental design and schematic culture time course. SB: SB431542; LDN: LDN193189; FSK: Forskolin; Y27632; CHIR: CHIR99021. (B) Representative bright-field images of cells obtained at different culture time points. O/S/K/M refers to OCT3/4, SOX2, KLF4, and c-MYC, respectively. N/A/B/Z refers to NEUROG1, ASCL1, BRN2, and ZIC1, respectively. The yellow arrowheads indicate neurite structures. (Scale bars, 100 μm.) (C) Immunofluorescence images of the dish contents on Day 21. βIII-tubulin, MAP2, and Hoechst stains are labeled in green, red, and blue, respectively. Panels show samples with 10% serum or no serum added. TFs refer to TFs. (Scale bars, 50 μm.) (D) Cell numbers of O/S/K/M+N/A/B/Z under 10% serum conditions on Day 0 and the total cell numbers and βIII-tubulin-positive cell numbers on Day 21. Data are represented as mean ± SE (n = 3 independent experiments). (E) Comparison of the percentage of βIII-tubulin-positive cells on Day 21 under induction by O/S/K/M+N/A/B/Z, with and without serum supplementation. The P values are indicated at the top of the graph (n = 3; mean ± SE) ***P < 0.001; **P < 0.01 vs. Serum (−/−) condition (Student’s t test). (F) Comparison of the percentage of βIII-tubulin-positive cells on Day 21 among different extracellular matrix types, under O/S/K/M+N/A/B/Z induction with 10% serum. The P values are indicated at the top of the graph (n = 3; mean ± SE) *P < 0.05 vs. Matrigel (Student’s t test). (G and H) Representative immunostaining images on Day 21, comparing differentiation efficiency between induction with all eight factors and eight factors excluding c-MYC (G). Statistical analysis of the results (H). The P value is shown at the top of the graph (n = 3; mean ± SE) **P < 0.01 vs. 8Fs (Student’s t test). (I and J) Representative immunofluorescence images of differentiation efficiency on Day 21 after transduction with eight TFs, excluding one factor each from ASCL1, BRN2, ZIC1, and NEUROD1 (I). Statistical analysis of the results (J). The P values are indicated at the top of the graph (n = 3; mean ± SE) **P < 0.01; *P < 0.05, vs. control (Student’s t test). (K and L) Representative immunostaining images on Day 21 comparing differentiation efficiency between induction with eight TFs and induction with five TFs (O/S/K/M+NEUROD1). βIII-tubulin, MAP2, and Hoechst are labeled in green, red, and blue, respectively (K). Statistical analysis of the results (L). The P values (Student’s t test) are presented at the top of the graph (n = 3; mean ± SE).

For further optimization, we compared other basal culture media, including the media hormone mix, MHM (13), KBM502, and StemFit, which were originally optimized for neurons, T cells, and embryonic stem cells (ESCs)/iPSCs, respectively. We also assessed combinations of MHM + StemFit and MHM + KBM502. Subsequently, βIII-Tubulin-positive cells were observed in notably higher numbers (approximately 18- and 10-fold higher when StemFit plus MHM and StemFit alone, respectively) than those observed with MHM alone (SI Appendix, Fig. S1 A and B). The evaluation of various cell adhesion substrates, including the 511-laminin E8 fragment/iMatrix-511, bulk laminin, fibronectin, and Matrigel, showed that Matrigel yielded the highest number of neurons (Fig. 1F and SI Appendix, Fig. S1C). Thus, culture conditions comprising MHM + StemFit, with FBS and Matrigel as cell adhesion substrates, were established for subsequent neural induction experiments (SI Appendix, Table S1).

To simplify the protocol, the necessity of each factor was evaluated. The four Yamanaka factors are crucial for establishing iPSCs, and each factor plays a significant role. In previous studies, in the absence of c-Myc, although the efficiency of iPSC establishment decreased, the generation of iPSCs was still feasible (14, 15). However, in the current study, the removal of c-Myc using our protocol resulted in significantly lower cell adhesion (Fig. 1 G and H and SI Appendix, Fig. S2A). Further assessment by omitting the N/A/B/Z factors revealed the essential role of NEUROD1 in neuronal cell generation (Fig. 1 I and J and SI Appendix, Fig. S2B). In fact, a five-factor (5F) approach, involving the addition of only NEUROD1 to O/S/K/M, resulted in a similar survival rate of βIII-Tubulin-positive cells as the eight-factor (8F) approach (Fig. 1 K and L and SI Appendix, Figs. S3 and S4).

Gene Expression Patterns in T Cell–Derived Neurons (TC-Ns) Resemble Those in Glutamatergic Neurons Differentiated from iPSCs.

To characterize the T cell–derived neuron or neuronal cell (TC-N) features, we profiled their transcriptomes using bulk RNA sequencing (RNA-seq). As controls, T cells, iPSCs, and iPSC-derived neurons (iPSC-Ns) with four factors (4F) (N/A/B/Z) and one factor (1F) (N) were profiled (Fig. 2A). Pairwise correction and clustering analysis, shown in SI Appendix, Fig. S5 A and B, revealed minor differences between donors, whereas appreciable differences were found across conditions, suggesting the high reproducibility of our protocols. Significantly, TC-Ns showed key marker gene expression patterns similar to those of iPSC-Ns, distinct from those of both T cells and iPSCs, as illustrated in Fig. 2B and SI Appendix, Fig. S5 A and B. Notably, the expression of SOX1, a neural stem cell (NSC) marker, was not elevated in TC-Ns, mirroring the direct neuronal induction from iPSCs. This suggests that the TC-Ns may have matured during culture or transdifferentiated into neurons, with limited involvement of the NSC stage (Fig. 2B and SI Appendix, Fig. S6 A and B). Assessment of the three germ layer marker genes (SI Appendix, Fig. S6C) revealed a low expression of most endoderm and mesoderm markers in both TC-Ns and iPSC-Ns, with the exception of ACTA2, the expression of which was notably higher in TC-Ns than in iPSC-Ns (SI Appendix, Fig. S6C). In addition, the high expression of GATA3 in TC-Ns likely reflected their hematopoietic origins (SI Appendix, Fig. S6D).

Fig. 2.

Global gene expression analysis of T cell–derived neuronal cells (TC-Ns). (A) Overview of the RNA-seq samples used in this study. The table displays four distinct combinations of TFs (1F, 4F, 5F, and 8F) employed in the experiment. Samples are derived from four different cell states: T cells, induced pluripotent stem cell (iPSCs), neurons induced from iPSC (iPSC-Ns), and neurons directly induced from T cell (TC-Ns). (B and C) Comparative analysis of gene expression across sample groups using averaged FPKM values, visualized in a heatmap. Expression patterns of typical markers for Blood cell, iPSC neural stem/progenitor cell (NSC/NPC), and neuron (B). Expression patterns of the typical markers of distinct neuronal subtypes (C). (D) Immunofluorescence images showing the expression of glutamatergic (VGLUT1) and GABAergic (VGAT) neuronal markers on Day 21 in TC-Ns (5F), TC-Ns (8F), and iPSC-Ns induced by Tet-On-driven expression of NEUROG2 (TetO-NEUROG2). (Scale bars, 100 μm.)

Regarding subtype-specific neuronal markers, TC-Ns and iPSC-Ns showed comparable patterns with particularly high expression levels of glutamatergic neuronal markers, as confirmed by immunostaining (Fig. 2D). Although cholinergic neuron marker expression was detected in the transcriptome, it was not observed in the immunostaining study. Notably, neural crest cell (NCC) marker gene expression was observed in TC-Ns but was absent in iPSC-Ns. This suggests that neural crest-related genes may be driven by the downstream gene network related to GATA3, which is highly expressed in blood cells (SI Appendix, Fig. S6E).

Intracellular Calcium Dynamics and Electrophysiological Analysis Demonstrate Neurophysiological Functionality in TC-Ns.

To assess the functional properties of TC-Ns, we conducted calcium imaging and electrophysiological analyses. Neuron-specific calcium activity was monitored in cells expressing jGCaMP7s and nuclear-localized mCherry under the control of the human SYN1 promoter (Fig. 3A) (16). Bath application of a high concentration of KCl (5 mM) significantly increased calcium signals (Fig. 3 B–D and SI Appendix, Fig. S7A), indicating the presence of voltage-dependent Ca²⁺ influx. These signals were blocked by tetrodotoxin (TTX), a voltage-gated Na⁺ channel blocker (Fig. 3 E–G and SI Appendix, Fig. S7B). Patch-clamp recordings revealed that most SYN1-tdTOMATO-positive cells (18/19) generated action potentials (APs) (Fig. 3 H–J), which were abolished by TTX application (n = 9, Fig. 3 I and J). We also observed voltage-dependent, fast-activating, and fast-inactivating inward currents in all cells studied (16/16), which diminished after TTX application (n = 6, Fig. 3K). Multielectrode array (MEA) analysis showed that spontaneous firing frequency increased with time in culture (SI Appendix, Fig. S9) and was suppressed by TTX and glutamate receptor blockers on day 30 (Fig. 3L). These results suggest that the TC-Ns share electrical properties with glutamatergic neurons, forming excitatory synaptic connections.

Fig. 3.

Intracellular calcium dynamics and electrophysiological analysis in TC-Ns. (A) Experimental time course and fluorescent-labeling vector structures. (B–G) Temporal imaging of GCaMP fluorescence signal alterations following 5 mM KCl treatment (B–D) and those in the presence of 1 µM tetrodotoxin (TTX) (E–G). Images display phase contrast (PH), mCherry-labeled neuronal nuclei, and green-labeled GCaMP fluorescence (C and F). Signal time courses for specific regions of interest (D and G). Cells #5 and #6 represent mCherry-negative nonneuronal cell nuclei. Consistent results were obtained from three or more replicates. (Scale bar, 50 µm.) (H–K) Whole-cell patch-clamp recordings were conducted on SYN1-tdTOMATO⁺ cells (red in H). Representative traces of membrane potentials in response to step current injections (Top in I). APs could be eliminated by 1 µM TTX (Bottom in I). Total number of APs during 500 ms stimulation period vs. injected current amplitude with (red, n = 8) or without (black, n = 16) TTX (J). Representative traces of membrane currents recorded following a ramp protocol (K). The fast inward currents was blocked by TTX (red in K). (L) Measurement of spike frequency activity by MEA in TC-Ns (5F and 8F). Variation in the number of spikes on Day 30 with 50 µM D-AP5 + 50 µM CNQX and 1 µM TTX treatments for iNs (n = 3; mean ± SE).

Characterization of TC-Ns Using Single-Cell RNA-Seq.

To characterize the heterogeneity of the cell populations, we analyzed the transcriptomes of TC-Ns (5F) on Day 21 using single-cell RNA sequencing (scRNA-seq). We obtained 4,883 cells and identified seven distinct cell clusters annotated as neurons, NSC-like cells (:NSC), iPSC-like cells (:iPSC), NCC-like cells (:NCC), T cells (:TC), unknown 1 (:NA1), and unknown 2 (:NA2). Assessing the expression of the canonical marker genes for neurons, NSCs, NCCs, iPSCs, and T cells were mostly specifically expressed in their respective clusters (Fig. 4 A–C and SI Appendix, Fig. S10A), except for typical NSC marker genes such as SOX1, PAX6, and MSI1 (SI Appendix, Fig. S10B). The top differentially expressed genes in the NSC cluster did not represent the typical set of NSC marker genes (SI Appendix, Fig. S10C), suggesting that the NSC clusters may represent a population of atypical NSCs or immature neurons. Focusing on the neuron cluster (Fig. 4D), the prevalent expression of the glutamatergic neuron marker SLC17A7 and the low expression of the GABAergic neuron marker SLC32A1 were consistent with the bulk RNA-seq observations (Figs. 2C and 4E and SI Appendix, Fig. S6B). As expected, these markers were mutually exclusive (Fig. 4F). Further clustering of the neurons yielded two clusters, with cluster 0 expressing peripheral nervous system markers and cluster 1 expressing central nervous system (CNS) markers (Figs. 4 D and G and SI Appendix, Fig. S11A). Among the CNS markers, CAMK2B and OTX2 tended to be coexpressed, whereas the hindbrain marker HOXB1 was highly expressed in distinct cells (Fig. 4G and SI Appendix, Fig. S11B).

Fig. 4.

Elucidating properties of cells generated by induction of peripheral blood T cells with five TFs using scRNA-seq analysis. (A) UMAP plot of 4,883 single-cell transcriptomes. Identification of seven distinct clusters: neural stem cells (:NSCs), Neurons (:Neuron), iPSC/iPSC-like cells (:iPSC), neural crest cells (:NCC), T cells (:TC), and nonannotated cells (:NA1, NA2). The Right Bottom panel depicts the number of constituent cells in each cluster. (B) Dot plot of T cell, iPSC, and neuron marker expression. (C) Feature plots and violin plots of representative marker genes for T cell, neuron, iPSC, and neural crest cell. (D) UMAP representation of the extracted neuron cluster cells from A. (E and F) Description of feature plots of neuronal subtype marker genes for UMAP (D) and scatter plots comparing a glutamatergic neuronal marker gene (SLC17A7/VGLUT1) and a glutamatergic synapse marker gene (DLG4/PSD95) with a GABAergic neuronal marker gene (SLC32A1/VGAT) (E). (G) Feature plots of central nervous system (CNS) and peripheral nervous system (PNS) marker genes (CAMK2B, OTX2, and HOXB1; PRPH/peripherin, PHOX2B, and RET, respectively).

scRNA-Seq Analysis Supports TC-N Induction with Limited Pluripotent State Reversion.

Next, we analyzed the transcriptional network that generates neurons from T cells. Based on the SeV vector sequences, we identified exogenous factors (exOCT3/4, exSOX2, exKLF4, exMYC, and exNEUROD1) in single-cell RNA-seq data and compared their expression with that of their endogenous counterparts (Fig. 5 A–C). Notably, the number of transgenes in iPSCs/iPSC-like cells was low, but exMYC and exNEUROD1 were still detectable in neurons. As demonstrated in Fig. 1G, Myc has proven to be a crucial factor in this induction process. Beyond its role in cell proliferation and survival, Myc has the potential to induce dynamic changes in chromatin structure (17, 18), which may have consequently facilitated neuronal differentiation. This abundance of exogenous factors was further confirmed by immunostaining using SeV-specific antibodies, where positive immunoreactivity was evident in βIII-tubulin-positive cells (SI Appendix, Fig. S12). Trajectory analysis using Monocle3 on UMAP indicated a continuum between NSC cluster and neuron cluster, while showing a notable distinction from iPSC cluster (Fig. 5D). Additionally, RNA velocity analysis using scVelo showed a gap between iPSC area (green cluster) and NSC area (blue cluster), with few connections. The transition from iPSCs to neurons appeared to be limited with most arrows terminating at the outer edge of the neuron area (orange cluster) (Fig. 5E), suggesting that this neuronal conversion predominantly follows a direct pathway with minimal involvement of a pluripotent state. This was consistent with the generation of neurons within the short duration of 20 d of culture (Fig. 5J).

Fig. 5.

Transgene expression, lineage tracing, and cellular trajectory analysis of TC-Ns. (A and B) Feature plots illustrating cells containing the transgene (O/S/K/M+N)-specific sequence (B) and cells containing the respective transgenes and corresponding endogenous genes (A). O/S/K/M and N refer to OCT3/4, SOX2, KLF4, and c-MYC; and NEUROG1, respectively. (C) Dot plots corresponding to Fig. 4 A and B. (D) Description of cell trajectory on UMAP using Monocle 3. The lineage of transdifferentiation is inferred and depicted as a trajectory of temporal evolution by comparing the expression profiles of cell populations. (E) RNA velocity analysis using the scVelo. The arrow indicates the direction in which the cellular transcriptome changes, whereas the arrow size indicates the speed of the change. (F) Experimental design for assessing Cre-LoxP system activation and FLEX switch induction, illustrating the culture timeline and lentiviral constructs. (G) Representative fluorescent images of cells showing activation of the NANOG, SOX1, and NESTIN/NES gene promoters/enhancers. (H) Quantification of TurboRFP-positive cells among iRFP670+EGFP+ populations. One-way ANOVA followed by Tukey’s multiple comparison test was performed. Each group: n = 12. Error bars represent SD. Statistical significance: *P < 0.05, ****P < 0.001.

Cre-LoxP System-Based Lineage Tracing in TC-N Induction.

To further investigate whether T cell–derived neuronal cells transitioned through a pluripotent stem or neural progenitor state, we performed a lineage-tracing experiment using the Cre-LoxP (FLEX switch) system to monitor promoter or enhancer activity during neuronal induction. Specifically, we introduced Cre under the control of the human NANOG, SOX1, and NES/NESTIN promoters or enhancers to track the state of pluripotent stem cell, neuroepithelial cell, and later-stage neural progenitor populations, including radial glia, respectively (Fig. 5F). We then assessed whether TurboRFP was expressed in cells that successfully underwent neural conversion (EGFP⁺iRFP670⁺), indicating prior activation of these promoters. The validation of these FLEX switch vectors was confirmed by analyzing cells reprogrammed into iPSCs using OSKM alone and their subsequent differentiation into neurons (SI Appendix, Fig. S13). Our results showed that during the conversion of T cells into neurons, 11.3% and 29.3% of the converted cells exhibited NANOG-Cre and SOX1-Cre activation, respectively, while NESTIN-Cre activation was observed in 67.8% of the cells (Fig. 5 G and H and SI Appendix, Fig. S14). These findings align with our scRNA-seq results, suggesting that while a substantial proportion of TC-Ns passed through a radial glia-like or neural progenitor-like state, the activation of pluripotency- or neuroepithelial-associated promoters/Cre was relatively limited.

WGBS-Seq Analysis Indicates Some Differences in DNA Methylation States Between TC-Ns and iPSC-Ns.

Finally, we examined whether the neurons produced in this study retained any somatic epigenetic memory, comparing the DNA methylation profiles of TC-Ns with iPSC-Ns. We homogenized the neuronal states by isolating PSA-NCAM positive, SSEA3 negative cells on day 21 of induction (Fig. 6A and SI Appendix, Fig. S13). The overall methylation level distribution was largely consistent across the four groups (SI Appendix, Figs. S14–S18). However, a comparative analysis of differentially methylated regions (DMRs) indicated that TC-Ns retained more of the original T cell methylation patterns (R² = 0.646) compared to iPSC-Ns (R² = 0.419), as expected (Fig. 6 B and C). Further examination of the distribution of hyper- and hypomethylated DMRs revealed that TC-Ns exhibited higher methylation levels across various DMR target regions relative to iPSC-Ns (Fig. 6D). When assessing overall similarity to iPSCs, iPSC-Ns showed the highest correlation (R² = 0.837). Moreover, gene ontology (GO) enrichment analysis of DMRs highlighted an intriguing finding: Immune-related DNA methylation was more pronounced in iPSC-Ns than in TC-Ns (Fig. 6E). Meanwhile, TC-Ns still exhibited a moderate level of similarity to iPSCs (R² = 0.747), suggesting that the OSKM reprogramming factors may have induced partial epigenetic resetting or rejuvenation at the epigenetic level (Fig. 6B). However, despite this partial reprogramming effect, TC-Ns retained distinct epigenetic features compared to iPSC-Ns, particularly in immune- and neural-related regions. This finding is consistent with scRNA-seq predictions and Cre-LoxP study, which suggest that TC-Ns were primarily generated through direct conversion rather than transitioning through a fully reprogrammed iPSC state (Fig. 5J).

Fig. 6.

Epigenetic profiling and partial reprogramming pathways of TC-Ns. (A) Preparation outline for the four groups (T cell, TC-N, iPSC, and iPSC-N) used in WGBS-seq. (B and C) Comparative analysis of DMRs among the four groups. The whole genome is split by 2 kb subbin, then methylation level for each bin and Pearson correlation analysis were performed. Heatmap represents the correlation between samples. R²: the square of Pearson correlation (B). The average methylation level of the DMR to draw the methylation level distribution violin boxplot (C). (D) DMR distribution plot in CG context across different functional regions. For various DMR target regions [e.g., promoter, exon, intron, CpG island (CGI), CGI shore, repeat, transcription start site (TSS), transcription end site (TES)], the distribution of hypermethylated (high methylation DMR) and hypomethylated (low methylation DMR) regions was analyzed in TC-Ns compared to iPSC-Ns. (E) DMR feature extraction between TC-Ns and iPSC-Ns. GO enrichment analysis of the top 100 genes with the highest and lowest DNA promoter methylation variation. Only terms with –log10(P) ≥ 2 are shown. (F) Overview of partial reprogramming methods used in our study: This study demonstrates a rapid and efficient method for inducing neurogenesis in peripheral blood T cells in vitro using O/S/K/M+N factors. Unlike methods that involve reprogramming through an intermediate iPSC stage, this approach allows neuronal differentiation within approximately 20 d, with most cells following a direct or neural progenitor-mediated pathway. While the possibility of rare transitions through an undifferentiated state cannot be entirely excluded, the resulting neurons exhibit distinct DNA methylation patterns compared to those derived via the iPSC route. By simplifying the process and reducing the time required for neuronal induction, this method provides a more efficient alternative for generating neurons.

Discussion

Over the last few decades, there has been an increasing focus on direct cellular reprogramming from fibroblasts to neurons (19); however, direct neuronal conversion from peripheral blood cells, a nonsurgical and simpler method, remains less explored, possibly owing to the substantial lineage difference between these cell types. This gap typically necessitates complex, multistage protocols involving long-term culture, and various support conditions, such as glial feeder cells, induction of neural differentiation post proliferation of neural stem cells via partial reprogramming, and cell sorting are necessary (6, 7, 9). A previous study by Tanabe et al. (8) also demonstrated the direct induction of functional neurons from blood cells but encountered challenges due to insufficient gene transduction efficiency, hindering the production of an adequate number of cells with a conversion efficiency of less than 10%.

In our study, the number of neurons obtained was nearly equal to the number of T cells initially seeded. We overcame these challenges by successfully generating neurons from T cells in approximately 20 d, under a straightforward, glial cell-free condition. As a result, we achieved high-effeciency neuronal conversion by the present method. This was achieved by overexpressing NEUROD1 along with the four Yamanaka factors (O/S/K/M) without fully inducing an iPSC state.

Notably, bulk RNA-seq analysis indicated that the quality of the neurons obtained using this method was comparable to those generated using all the N/A/B/Z factors (Fig. 2 B and C). Intriguingly, neuronal bHLH TF ASCL1, which has been implicated in the induction of fibroblast-to-neuron transdifferentiation by promoting local DNA demethylation and dynamic chromatin remodeling (20, 21), was not an essential factor in our conversion system. Nakashima et al. also showed that the differentiation method altered the chromatin in NEUROD1-expressing blood-related cells (22, 23). Thus, further investigation is needed to distinguish this method from that using conventional bHLH-type factors, such as ASCL1, NEUROG1, and NEUROG2 (19), in order to reveal its unique impact on chromatin states.

Our protocol predominantly generated glutamatergic neurons that displayed distinct gene expression patterns similar to those of iPSC-Ns. These neurons showed an exclusive relationship with VGAT-positive cells. Conversely, TC-Ns exhibited high expression of NCC marker genes. We hypothesized that GATA2/3 signaling, which is active during early NCC differentiation, facilitates neuronal cell production in T cells. Additionally, using CHIR9901 (a WNT/β-catenin pathway activator) and Y27631 (a ROCK inhibitor) have been suggested to support NSC and peripheral neuron-like cell production (24–26), although peripheral neuron marker-positive cells were not predominant.

Our scRNA-seq analysis revealed SeV-derived exogenous genes remained in some cells. Additionally, trajectory and RNA velocity analyses indirectly suggested a neuronal conversion occurred with limited involvement of an iPSC state. To further verify this, we performed lineage-tracing experiments using Cre-LoxP to track the activation history of specific gene enhancers/promoters. The results indicated substantial activity of neuronal progenitor or radial glial cell marker gene (NESTIN+) enhancer in many populations, while pluripotency (NANOG+) and neuroectoderm (SOX1+) promoter activity was limited.

Our method is not only beneficial in terms of conversion efficiency but also provides valuable insights into the nature of reprogrammed neurons. One key advantage is its potential to examine the epigenetic memory inherited from the original cells. Indeed, while TC-Ns exhibit physiological functions and RNA-seq profiles consistent with glutamatergic neurons, they retain distinct DNA methylation states in blood lineage-related genes. On the other hand, previous studies have shown that prolonged expression of OSKM factors leads to epigenetic rejuvenation (27–30), which could potentially erase donor-specific aging signatures. However, this also suggests that rejuvenated cells could be a valuable resource for regenerative medicine and gene therapy applications (31, 32). In the future, further single-cell-level analysis of DNA methylation states in TC-Ns could help determine whether this model can be leveraged in different contexts—either preserving donor-dependent epigenomic information for personalized medicine or utilizing rejuvenated cells for regenerative medicine.

A limitation of our study was the inability to produce highly purified or subtype-specific neurons. We prioritized cell survival by employing basic medium conditions in combining media for pluripotent stem cell and neuronal cultures with small molecular compounds, as previously reported (8), without time-based adjustments. In addition, residual SeVs may affect neuronal maturation and subtype specificity. Future optimization of culture conditions and alternatives to FBS or temperature-sensitive SeV systems may enable more efficient production of pure, subtype-specific neurons and xeno-free conditions (33), with applications not only in neurological disease research but also in gene therapy and regenerative medicine strategies incorporating epigenetic rejuvenation.

Materials and Methods

Preparation and Culture of Human Peripheral Blood Mononuclear Cells.

Human peripheral blood mononuclear cells (PBMCs) were purchased from Lonza Biosciences (Cat. No. CC-2703; Walkersville, MD) or sourced from fresh blood donations. Fresh peripheral blood was collected from healthy Japanese donors in their 30s, and PBMCs were subsequently isolated with Ficoll using the sucrose density gradient method. Cells were cultured in KBM502 medium (16025020; KOHJIN BIO, Saitama, Japan) supplemented with 281 IU/mL interleukin-2 (IL-2) and a plate-bound anti-CD3 antibody (sc-19590; Santa Cruz, Dallas, TX) in tissue culture dishes. Experiments using human blood samples were conducted with the approval of the Ethics Committee of Keio University School of Medicine (Approval No.: 20080016). Written informed consent was obtained from all study participants prior to sample collection.

Preparation of SeV.

All Sendai viruses (SeVs) were purchased from ID Pharma Co. Ltd. (Tsukuba, Japan). SeVs encoding the four Yamanaka factors (OCT4, SOX2, KLF4, and c-MYC) were utilized as a component of CytoTune-iPS 2.0 kits (ThermoFisher Scientific, Carlsbad, CA). SeVs encoding iN factors, including NEUROD1, ASCL1, BRN2/POU3F2, and ZIC1, were synthesized via a contractual agreement (ID Pharma, Tokyo, Japan). The Reference Sequence (RefSeq) transcript IDs for each iN factor were as follows: NEUROD1, NM_002500.5; ASCL1, NM_004316.4; BRN2/POU3F2, NM_005604.4; and ZIC1, NM_003412.4.

Induction of Neurons from Human PBMCs.

Human PBMCs were directly seeded into culture dishes coated with anti-CD3 antibody in KBM502 medium. After 24 h, the cells were infected with SeVs, and after another 24 h, the cells were harvested and centrifuged at 200×g for 5 min at 25 °C. After removal of the supernatant, the cells were reseeded onto precoated 6-, 12-, or 24-well plates. Both the coating substrate and the culture medium were varied as part of the experimental design. Detailed information regarding the substrates is provided below.

Culture plates were precoated with various substrates, including Matrigel (354230, Corning, NY) at concentration of 1:100, fibronectin (F0895, Sigma, St. Louis, MO), laminin (L2020, Sigma), and lamin 511-E8/iMaterix-511 fragment (892021, Nippi Corp, Yokohama, Japan) at concentrations ranging from 10 to 50 µg/mL, 10 to 50 µg/mL, and 0.6 µg/cm² to 3.0 µg/cm², respectively. The induction medium was also varied as part of the study. We used MHM (13), StemFit (RCAK02N, Reprocell, Beltsville, MD), and KBM502, in addition to a 1:1 mixture of MHM and StemFit and a 1:1 mixture of MHM and KBM502. In the electrophysiological study, cells were seeded on Cell Disk LF1 (MS-92132, Sumitomo Bakelite, Tokyo, Japan) with the same precoating manner as described above. The composition of the cell maintenance solution used in the patch-clamp analyses was based on the protocol described in ref. 34, with details provided in SI Appendix.

All the induction media were supplemented with the following additives: 10 µM SB431542 (S4317, Sigma), 2 µM CHIR99021 (10-1279, Focus Biomolecules, Plymouth Meeting, MA), 500 nM LDN-193189 (Sigma, SML0559), 3 µM forskolin (F3917, Sigma), 10 µM Y-27632 (08945-42, Nacalai Tesque, Kyoto, Japan), 10 ng/mL BDNF (B-250, Alomone Labs, Jerusalem, Israel), 10 ng/mL GDNF (G-240, Alomone Labs), and 10 ng/mL NT-3 (N-260, Alomone Labs). Depending on the experimental conditions, FBS (Lot: 14K525; Sigma-Aldrich, St. Louis, MO) was added.

Single-Cell-Level Dissociation for scRNA-Seq and WGBS-Seq.

On day 21, the 5F-TC-N and iPSC-N samples were gently rinsed once with PBS and subsequently treated with Accutase (12679-54, Nacalai Tesque) for 60 min, at 37 °C. This process was followed by careful dissociation of the cells to obtain individual cells, using a combination of tapping and pipetting. The cells were then centrifuged at 200×g for 5 min at 4 °C 40. After centrifugation, the cells were resuspended in PBS, sieved through a 40-µm cell strainer, and cellular viability was evaluated.

Generation of iPS cells from human PBMCs, induction of neurons from human iPS cells, high-content image analysis, calcium imaging, electrophysiological study by patch-clamp analsys and MEA, RNA extraction and bulk-RNA-seq-based gene expression analysis, scRNA-seq library preparation and the data analysis, trajectory analysis, RNA velocity analysis and WGBS-seq library preparation, and the data analysis are described in detail in SI Appendix.

Institutional Review Board Statement.

The study was conducted in accordance with the Declaration of Helsinki and approved by the Clinical Research Committee of Keio University School of Medicine (No. 20080016).

Informed Consent Statement.

Written informed consent has been obtained from the participants in this study to publish this paper.

Supplementary Material

Appendix 01 (PDF)

Movie S1.

Live-cell imaging video of TC-Ns treated with KCl, using GCaMP. The video captures oscillation changes in 21-day cultured TC-Ns expressing Lenti-Syn1-jGCaMP7s following the application of PBS(-) at 30.00 seconds and subsequent treatment with 5 mM KCl at 1 minute and 30.00 seconds. The transient spike-like increase in fluorescence intensity induced by KCl treatment is shown. The video is compressed to 20 times the actual speed.

Movie S2.

Live-cell imaging video of TC-Ns treated with TTX, using GCaMP. The video captures oscillation changes in 21-day cultured TC-Ns expressing Lenti-Syn1-jGCaMP7s following the application of PBS(-) at 2 minutes and 30.00 seconds and subsequent treatment with 1 μM TTX at 5 minutes and 0.00 seconds. The cessation of spontaneous oscillations following TTX treatment is depicted. The video is compressed to 20 times the actual speed.

Data, Materials, and Software Availability

The gene expression data obtained in this study, including multiple NGS datasets from T cell-derived neuronal induction experiments, have been deposited in GEO (35). All other data are included in the article and/or supporting information.