Peripheral blood mononuclear cells were isolated using density gradient centrifugation with Ficoll-Paque PREMIUM 1.073 (Cytiva) and SepMate-50 (Veritas) according to the manufacturer’s instructions. Freshly isolated peripheral blood mononuclear cells were seeded on plastic coverslips (Cell Desk LF, Sumitomo Bakelite) in an appropriate amount of monocyte attachment medium (PromoCell) in a density of 1 million/cm² and incubated for 2 h at 5% CO2, 37°C. Subsequently, the non-adherent cells were removed by three washing steps with warm monocyte attachment medium. Adherent monocytes were cultured in macrophage medium, which consisted of X-VIVO 15 (Ronza) supplemented with 10% heat inactivated fetal bovine serum (FBS) (Biological Industries), and stimulated with 50 ng/ml macrophage colony-stimulating factor (Wako) for 7 days to differentiate them into monocyte-derived macrophages (Lescoat et al., 2018). We evaluated 15 different experiments (using five donors, three-time experiments).

The conditioned media for neurodifferentiation consisted of Neurobasal (Gibco) supplemented with GlutaMAX, N2, and B27 supplements (0.5% each, Gibco), penicillin- streptomycin (Gibco), and brain-derived neurotrophic factor (BDNF) (20 ng/mL, Wako), glia-derived neurotrophic factor (GDNF) (20 ng/mL, Wako), neurotrophin-3 (NT3) (10 ng/mL, Wako), insulin-like growth factor 1 (IGF-1) (20 ng/mL, PEPROTECH), and basic fibroblast growth factor (bFGF) (20 ng/mL, Wako). Monocyte-derived macrophages were treated with the following chemical compounds: CHIR99021, 3 μM; Forskolin, 10 μM; Y-27632, 5 μM; Dorsomorphin, 2 μM; ISX-9, 3 μM; DB2313, 0.5 μM.

Cells on plastic coverslips (Cell Desk LF, Sumitomo Bakelite) were fixed in 4% paraformaldehyde (PFA) (Wako) at room temperature for 10 min; subsequently PFA was washed off with phosphate-buffered saline (PBS) three times. After washing, the coverslips were transferred to blocking buffer and incubated for 1 h. The blocking buffer consisted of 1% bovine serum albumin and 0.1% Triton X-100 in PBS. After blocking, cells were incubated with primary antibodies (all 1:500) overnight at 4°C. On the second day, primary antibodies were washed off thrice with PBS. The cells were incubated with the corresponding secondary antibodies for 30 min at room temperature. The secondary antibodies were tagged with the fluorophores: Alexa Fluor 488 and Alexa 594 (1:1,000, Invitrogen). After incubation, the secondary antibodies were washed off thrice with PBS. Cells were mounted on glass slides with mounting solution containing 4′,6′-diamidino-2-phenylindole (DAPI; Vector Laboratories). Immunostaining results were analyzed using a fluorescence microscope (Olympus DSU). Confocal microscopy (ZEISS LSM710) was used to capture images. For each sample, the cells were counted at designated times in 10 randomly selected fields of view. The antibodies used are listed in Supplementary Table 1.

Gene expression analysis was performed on the cultured cells. RNA was purified with the NucleoSpin^® RNA Plus kit (Macherey-Nagel) with initial quantitation conducted using a NanoDrop OneC spectrophotometer (Thermo Fisher Scientific). RNA quality was determined using a 2,200 Bioanalyzer (Agilent Technologies). Total RNA of the human cerebral cortex was obtained from Clontech (Cat. No. 636561).

The purified RNA samples were subjected to whole transcriptomic analysis using the GeneChip Human Clariom S Array (Thermo Fisher Scientific; the analysis was outsourced to Life Technologies Japan Ltd.). The transcriptomic profiles of the cells were analyzed using Transcript Analysis Console 4.0.2 (TAC) software from Applied Biosystems. Gene ontology enrichment analysis was performed using the PANTHER program.¹

The purified RNA samples were reverse-transcribed with SuperScript IV VILO Master Mix (Invitrogen) according to the manufacturer’s instructions. To select normalization factors, we used the free analysis program RefFinder² to calculate the relative stability of the expression of 16 housekeeping genes (human housekeeping gene primer set; Takara) in each cultured cell. Two genes (RPLP2 and RPS18) were selected as controls for stable endogenous expression. Quantitative real-time polymerase chain reaction was conducted with 400 nM primers, 25 ng cDNA template, and TB Green Premix Ex Taq II (Takara) in a Thermal Cycler Dice Real-Time System (software version 5.11B for TP800, Takara) (n = 3). Data were log2 transformed to fit the normal distribution assumption of one-way analysis of variance (ANOVA). Primers used are listed in Supplementary Table 1.

Studies were conducted in adult CB-17/Icr-+/+Jcl (CB-17) mice aged 8 weeks (Clea Japan). The experimental animals had ad libitum access to food. Every effort was made to minimize the number of animals used and their suffering. Quantitative analyses were conducted by investigators blinded to the experimental protocol and sample identity.

Permanent focal cerebral ischemia was induced in mice by ligation and interruption of the distal portion of the left middle cerebral artery (MCA) as previously described (Nishie et al., 2021). In brief, under three types of mixed anesthesia, consisting of medetomidine, midazolam, and butorphanol, mice were placed in a lateral position. Under an operating microscope, an incision was made in the skin and the temporalis muscle was pushed aside to obtain a visual field for craniotomy. A burr hole was created in the skull using a drill. After carefully opening the dura mater, MCA occlusion (MCAO) was performed using electrocoagulation, followed by disconnection of the distal portion of the left MCA. The sham-operated mice underwent the same surgical procedure until the dura mater was opened.

A small molecules cocktail including CHIR99021, 10.54 μg/10 g; Dorsomorphin, 6.0 μg/10 g; Forskolin, 28.7 μg/10 g; Y27632, 23.7 μg/10 g; isoxazole-9 [ISX-9], 4.9 μg/10 g (Hu et al., 2015; Li et al., 2015; Gao et al., 2017; Yin et al., 2019); and DB2313, 2.5 μg/ mice weighted 10 g (Antony-Debré et al., 2017) or vehicle solution (dimethyl sulfoxide, DMSO 11.9 μL/ mice weighed 10 g) as control was injected into the peritoneum of mice daily starting 6 days after the induction of ischemic stroke. Fourteen days after the initial drug treatment, mice were sacrificed and transcardially perfused with 4% PFA. Whole brains were removed and post-fixed with 4% PFA for 24 h. Subsequently, the samples were embedded in paraffin and cut into 4 μm sections. The brain sections were subjected to immunohistochemistry using primary antibodies. Bound primary antibodies were visualized using Alexa Fluor 488 or 596 conjugated secondary antibodies (1:1,000, Invitrogen). Nuclei were counterstained with DAPI (Vector Laboratories). Images were acquired using an Olympus DSU or a Zeiss LSM 710 confocal microscope. The antibodies used are listed in Supplementary Table 1.

Phenotypic behavioral differences were assessed in sham-operated and MCAO mice in accordance with previous reports (Park et al., 2014; Tanaka et al., 2020), with minor modifications. Corner, wire hang, and basket tests were performed 2, 6, 13, and 20 days post stroke.

A corner test was conducted using two wooden cardboards at an angle of 30° (Park et al., 2014; Hatakeyama et al., 2019). The mouse was made to enter the corner upon its placement midway to the corner. As the mouse reached deep into the corner, it turned backward toward the open end. The direction toward which the mouse turned was recorded. Twenty trials were recorded per animal on indicated days. The ischemic mice showed a marked preference for turning toward the non-impaired side (left turn). The percentage of left turns was analyzed as an indicator of deficit.

The wire hang test was conducted using a square wire mesh plate (20 × 20 cm) made of warp and weft wires (0.8 mm in diameter) loosely woven in a mesh shape (Tanaka et al., 2020). Each mouse was placed on the wire mesh plate and allowed to acclimate to the environment for 10 s. The testing plate was then gently inverted and secured to the top of a cubic, open-topped glass box (25 × 25 × 25 cm). The latency to fall was measured.

The basket test was conducted using a rectangular wire mesh basket (29 × 29 × 29.5 cm) (Tanaka et al., 2020). The basket was then gently inverted and placed on a flat plate where sawdust was spread. The latency to climb down the vertical walls of the basket to the flat plate was recorded, with a maximum trial time of 200 s. Three trials were completed for each mouse, with an interval of 5 min.

For comparison between two groups, statistical analysis was performed using the Student’s t-test. For comparisons among three or more groups, we used ANOVA followed by a post-hoc test (Tukey–Kramer test). All data are presented as the mean ± standard error of the mean (SEM).

We first examined small molecule candidates that chemically convert macrophages into neuronal cells (Figure 1A). Blood samples were obtained from healthy adults who provided informed consent, and peripheral blood mononuclear cells were isolated by centrifugation. Subsequently, the cells were cultured in monocyte attachment medium, and non-adherent cells were discarded. The remaining adherent monocytes were cultured in 10% FBS and 50 ng/ml recombinant human macrophage colony-stimulating factor. After 7 days, most cells were positive for macrophage markers Iba1 and CD206, indicating that monocyte-derived macrophages were induced (Figures 1B, C; Lescoat et al., 2018). We defined this time point as Day 0.

Day 0 macrophages were cultured with small molecules to investigate whether cells with a neuron-like morphology could be induced. CHIR99021, Dorsomorphin, Forskolin, ISX-9, and Y27632 were used as the small molecules. These compounds have been reported to be able to convert astrocytes and fibroblasts into neurons (Hu et al., 2015; Li et al., 2015; Yin et al., 2019). However, on Day 7, no cells presented neuron-like structures; rather, most of them transformed into cells with foamy structures (Supplementary Figure 1A1). Since the appearance of foam cells could morphologically indicate the activation of macrophages (Russell et al., 2009; Grajchen et al., 2018; Enos et al., 2019), we hypothesized that a chemical compound that silences macrophage specific genes is necessary for successful differentiation. PU.1 is one of the genes that play an important role in the differentiation of monocytes into macrophages (Klemsz et al., 1990; Zakrzewska et al., 2010). High levels of PU.1 expression in PU.1+/- fetal liver hematopoietic progenitors promoted macrophage differentiation (DeKoter and Singh, 2000). DB2313, a derivative of the heterocyclic diamidine family (Supplementary Figure 1B), efficiently inhibits PU.1 gene expression by an allosteric mechanism (Antony-Debré et al., 2017). Therefore, we cultured macrophages with a combination of six compounds: CDFIYB (CHIR99021, Dorsomorphin, Forskolin, ISX-9, Y27632, and DB2313). Some macrophages underwent drastic morphological changes and complex neurite-like structures were observed on Day 7 (Figure 1D).

Immunostaining indicated that these cells presented the neuronal markers doublecortin (DCX), beta-III tubulin (TUJ1), microtubule-associated protein 2 (MAP2), and neuronal nuclear protein (NeuN) (Figures 1E–I). However, these cells did not present the NueroD1 (Figure 1J). In contrast, some cells did not express any neuronal markers but remained positive for Iba1 and CD206 (Figures 1E, F). These results suggested that some macrophages that successfully differentiated into neurons and others that failed to differentiate existed simultaneously (Figure 1K). Among the TUJ1-positive cells, approximately 5% were MAP2-positive and 30% were DCX-positive, indicating that most of the induced neuronal cells were immature (Figure 1L). Conversion efficiency decreased when cells were cultured with all but one of the six compounds (Supplementary Figures 1 A1–A12, C). We defined cells treated with CHIR99021, Dorsomorphin, Forskolin, ISX-9, and Y27632 as Chemical Cocktail (CC) 5 cells, and those treated with these five compounds plus DB2313 as CC6 cells.

We performed RNA microarray analysis on human brain cortex total RNA, Day 0 macrophages, Day 7 CC5 cells, and Day 7 CC6 cells. First, we compared the expression levels of 21,448 genes in the human brain cortex and Day 0 macrophages. Based on these results, we defined two groups of genes: cortex enriched genes (CEGs), whose expression was upregulated more than four-fold in the brain cortex, and macrophage-enriched genes (MEGs), whose expression was upregulated more than four-fold in Day 0 macrophages (Figure 2A and Supplementary Figure 2A). For these two groups of genes, we examined changes in gene expression levels from Day 0 macrophages to Day 7 CC6 cells. Of the 2,418 MEGs, the expression levels of 1,271 genes were suppressed to < one-half in Day 7 CC6 cells, and gene ontology analysis indicated significant enrichment of biological processes related to the interleukin-3-mediated signaling pathway, detection of fungus, and detection of bacterial lipopeptides, indicating that macrophage properties were repressed (Figure 2B and Supplementary Figure 2B). Of the total 2,814 CEGs, the expression levels of 625 genes were increased > two-fold in Day 7 CC6 cells, and gene ontology analysis indicated significant enrichment of biological processes related to positive regulation of axonogenesis, protein localization to synapse, and postsynaptic intermediate filament cytoskeleton organization, indicating that neural differentiation was induced (Figure 2C and Supplementary Figure 2C). Gene ontology analysis enrichment analysis of the group B genes and group B genes classified in Figure 2A was no regulation of neurons (Supplementary Figures 2D, E). Comparison of Day 7 CC5 cells with Day 0 macrophages exhibited no difference in the expression of CEGs such as TUJ1 and INA and MEGs such as PU.1 and MAFB (Supplementary Figure 2F).

Based on the results of the RNA microarray, we performed reverse transcription-quantitative polymerase chain reaction validation by comparing Day 0 macrophages, CC5 cells, and CC6 cells over time (Figure 2D). In CC6 cells, the expression of proneural genes such as ASCL1 and NGN2 and neuron-related genes increased over time, while the expression of macrophage related genes such as CD11b, CD68, and PU.1 was suppressed. Interestingly, although CC5 cells demonstrated increased expression of proneural and neuron-related genes to the same extent as CC6 cells on Day 1, upregulation could not be maintained, and these genes were suppressed by Day 7. The reduction in the increase of CEGs on day 3 CC6 cells compared to Day 1 and Day 7 was observed. Macrophage-related genes were not sufficiently suppressed in CC5 cells throughout the study period. Finally, we performed RNA microarray analysis of Day 0 macrophages and Day 7 CC6 cells to determine whether the microglia were affected by the drug cocktail. The expression of all seven microglia-specific enriched genes was inhibited (all P < 0.001) (Supplementary Figure 2G).

To investigate the therapeutic effects of these six drugs against stroke, we prepared a CB-17 wild-type mouse model of acute ischemic stroke. The left middle cerebral artery was permanently occluded in mice using electrocoagulation (Supplementary Figure 3A). Two days after the onset of ischemia, MAP2 immunoreactivity was markedly decreased in the left cerebral cortex, whereas TUJ1 immunoreactivity remained positive. We defined the MAP2-lost area as the infarct area (Popp et al., 2009). Twenty days after ischemia, the infarct cortical tissue gradually shrank, and TUJ1 immunoreactivity decreased in the outer cortex, dividing the infarct area into TUJ1-positive and negative areas (Supplementary Figure 3B). Macrophages accumulated more at the infarct area on Day 6 than on Day 2 (Supplementary Figure 3C), consistent with a previous report’s findings (Breckwoldt et al., 2008). We examined whether systemic administration of these six drugs could induce transdifferentiation of macrophages into neurons in the infarct area. Intraperitoneal injection of the six drug-cocktail CDFIYB was performed daily from days 6 to 20 (Figure 3A). Immunostaining of the brain tissues on Day 20 exhibited a new MAP2 positive layer that emerged in the infarct area, suggesting cortical regeneration (Figures 3B, C).

This newly generated cortical layer was confined to the TUJ1+ infarct area (Figure 3D). The number of Iba1+ or CD206+ cells in the infarct area significantly reduced, indicating that macrophages were consumed (Figures 3E, F and Supplementary Figures 4A, B). Although we investigated the possibility of small molecule cocktails which stimulate the activation of endogenous neural stem/progenitor cells (NSPCs) in the subventricular zone or dentate gyrus, there was no significant change in the number of DCX+ cells in the dentate gyrus in this study (Supplementary Figures 4C, D). We did not observe any morphological differences between sham animals and contralateral brains. Finally, in several assessments of motor sensory function, mice treated with small molecule compounds showed significant neurological improvement (Figures 3G–I). Additionally, adverse effects such as weight loss, decreased appetite, and liver or kidney dysfunction were not observed in mice administered DMSO or the six drugs.