Transdifferentiation of human adult peripheral blood T cells into neurons

Direct Induction of iN Cells from Peripheral Blood Mononuclear Cells.

To investigate whether blood cells can be transdifferentiated to iN cells, we collected fresh blood from an adult healthy individual. We then isolated peripheral blood mononuclear cells (PBMCs) using gradient centrifugation and electroporated these cells with episomal vectors encoding the four transcription factors Brn2, Ascl1, Myt1l, and Ngn2, collectively termed the BAMN pool, which was previously found to generate iN cells from human fibroblasts (9), and enhanced green fluorescent protein (EGFP) into 3 million PBMCs. Transfected cells were then cultured in IL-2 and CD3/CD28 antibodies containing media supporting T cell growth (Fig. 1A). On day 3, we placed the nonadherent, transfected blood cells on different substrates, including primary mouse glia, mouse fibroblast SNL cell line, human primary fibroblasts, Matrigel, Polyornithine, or laminin-coated dishes. BAMN-transfected cells attached well on primary mouse glia cells and human fibroblasts, but on no other substrate. Cells transfected with EGFP alone did not attach to any substrate (SI Appendix, Fig. S1 B and C). On day 5, we changed the hematopoietic medium to the neuronal medium N3. Remarkably, after seeding of the human blood cells on murine glial cells, we noticed that glial cells deteriorated quickly. We reasoned that the nontransfected, activated human T cells presumably began to attack the mouse glia. Withdrawal of IL-2 and the T cell activators after day 3 not only mitigated this problem, but also improved reprogramming by 2.7- to 3.6-fold. Switching to neuronal media on day 3 also rescued glial viability but improved reprogramming only slightly (Fig. 1C and SI Appendix, Fig. S1E).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Generation of neuronal cells from peripheral blood cells. (A) Experimental outline of iN cell induction from PBMCs. (B) Morphological changes during iN cell induction from PBMCs. (Scale bars: 50 μm.) (C) The relative number of iN cells from T cells with or without T cell activator (anti-CD3/CD28), with or without IL-2, or a change to N3 media on day 3. n = 3 individuals. (D) Efficiency of iN cell induction of transduced cells from 35 individual donors without inhibitors at day 21. n = 1 for each donor. The number of iN cells on day 21 was divided by the number of total EGFP⁺ cells counted on day 1. (E and F) Relative iN cell induction (E) and efficiency of electroporation (F) from PBMCs of three individual donors that were kept at −80 °C or at 4 °C for 2 d relative to the fresh sample. *P < 0.05, paired t test. (G) Transdifferentiation efficiency of PBMCs from three individual donors kept at −80 °C or at 4 °C for 2 d relative to the fresh sample. Error bars represent SD.

To test the general applicability of our protocol, we obtained blood from a total of 35 healthy adult donors of various ages and ethnicities and both sexes. Surprisingly, the electroporation efficiencies varied significantly (SI Appendix, Fig. S1A), but nonetheless we were able to generate morphologically complex iN cells from all tested blood samples (Fig. 1D). The reprogramming efficiency varied as well, but the electroporation rate did not correlate with the reprogramming efficiency (SI Appendix, Fig. S1D). Unlike iPS cell reprogramming, reprogramming to iN cells was consistently lower in aged donors (Fig. 1D).

Because most biorepositories freeze patient samples, we next tried to induce iN cells from short- and long-term stored PBMCs at cold temperature. We isolated PBMCs from a fresh whole-blood sample, reprogrammed a fraction of the cells, and froze the remaining cells at −80 °C. Another fraction of the whole-blood sample was maintained at 4 °C for 2 d before subsequent PBMC isolation. We then reprogrammed the PBMCs isolated from the stored blood sample and the frozen PBMCs. There were no significant differences in the reprogramming efficiency and electroporation efficiency between fresh and frozen PBMCs, but the iN cell yield was substantially lower from PBMCs stored at 4 °C due to decreased transfection efficiency (Fig. 1 E–G). Thus, storage of freshly isolated cells at −80 °C did not affect reprogramming.

Combined BMP and TGF-β Pathway Inhibition and PKA Activation Improved Reprogramming Efficiency.

We next sought to further increase the induction efficiency of iN cells using small molecules. Blockade of BMP and TGF-β pathways have been shown to promote neural induction during normal development, during ES cell differentiation, and from fibroblasts (23⇓–25). Moreover, cAMP has been reported to facilitate neuronal survival and maturation (26). Therefore, we treated reprogramming blood cells with compounds regulating these three pathways from day 5 to day 21 (Fig. 2A). Indeed, the number of iN cells was significantly increased by the adenylyl cyclase activator forskolin (1.9-fold), the BMP pathway blocker dorsomorphin (3.7-fold), and the TGF-β pathway inhibitor SB431542 (4.6-fold) when analyzed on day 21 (Fig. 2B). The combination of the three inhibitors (forskolin, dorsomorphin, and SB431542; 3sm) showed an additive effect on the number of iN cells relative to the cells seeded (8.7-fold) (Fig. 2 A and B). These combined improvements yielded a reprogramming efficiency of up to 6.2%, which translates into 54,265 iN cells from 1 mL of blood (Fig. 2B).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Small molecule treatment improves iN cell conversion efficiency and maturation. (A) Immunofluorescence analysis of iN cells with and without small molecules (3sm, three small molecules: forskolin, dorsomorphin, and SB431542; Cont, DMSO). (Scale bars: 50 µm.) (B) Fold change of improved iN cell formation on day 21 following various small molecule treatments as indicated. Do, dorsomorphin; Fo, forskolin; SB, SB431542. Data shown are average fold changes of three independent experiments using PBMCs from three different donors. The fold change over the control condition was plotted because the absolute reprogramming efficiency was variable among the three donors, but the fold change was consistent. *P < 0.05, paired t test. The error bars indicate SDs. Similar results were obtained with PBMCs from another set of three different donors. (C) Example traces of action potential firing recorded from PBMC-derived iN cells with or without 3sm at days 21 and 42 (n represents the number of cells patched that shows action potentials over the total number of cells patched). The experiment was performed with cells from three different donors, yielding similar results. (D) Sample traces (Left) and average values (mean ± SEM; Right) demonstrating the presence of voltage-gated Na⁺ and K⁺ channels in blood iN cells cocultured with glia with 3sm for 42 d. (Inset) Expanded view of the dotted boxed area. (E) Intrinsic properties of membrane potential (V_rest), capacitance (C_m), and input resistance (R_m) approach more mature values over time. (F) Maturation of blood iN cells on extended coculture with glia in 3sm from day 21 to day 42 as determined by increased action potential (AP) height and threshold. Bar graphs represent mean ± SEM. *P < 0.05; **P < 0.01. ns, not significant.

Blood iN Cells Exhibit Passive and Active Neuronal Membrane Properties.

We then tested the functional properties of blood iN cells that were generated with and without addition of the three inhibitors. Patch-clamp recordings of blood iN cells showed their ability to fire action potentials on step-current injection at 21 and 42 d after infection and expressed functional voltage-gated Na⁺ and K⁺ channels (Fig. 2 C and D). As expected, 3sm treatment and longer culture periods (day 42) yielded cells with parameters of more mature action potentials (height, threshold, faster depolarization and repolarization kinetics) and more mature intrinsic properties, such as increased capacitance and decreased input resistance, compared with control-treated cells and day 21 cells, respectively (Fig. 2 C, E, and F).

ROCK Inhibition Substantially Improves Formation of Neuronal Morphologies but Does Not Improve Functional Properties.

In an attempt to increase the reprogramming efficiencies even further, we screened six additional compounds—IWP2, DAPT, retinoic acid, SU5402, Y26732, and SP600125—targeting pathways previously implicated in neural differentiation in combination with forskolin, dorsomorphin, and SB431542. Of the single compounds tested, we found that ROCK inhibition increased both the relative number of iN cells and the morphological complexity of reprogramming cells compared with the DMSO control (Fig. 3 A–D). Other combinations of small molecules did not have additional effects (Fig. 3E). Moreover, the addition of neurotrophic factors did not substantially increase the formation of iN cells, but long-term drug treatment yielded more iN cells than transient small molecule treatment (Fig. 3F).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

ROCK inhibition improves morphological maturation, but not functional maturation. (A) Relative number of TUJ1-positive (yellow bars) or MAP2-positive (blue bars) cells with 3sm (forskolin, dorsomorphin, and SB431542) and additional small molecules from PBMCs from three individual donors on day 21, normalized to the no treatment control condition (Cont). n = 3 individuals. *P < 0.05 relative to DMSO (only with 3sm). (B) Relative average length of neurites with small molecules from PBMCs from three individual donors. The length was normalized to the no treatment control (Cont). n = 3 individuals. *P < 0.05 relative to the DMSO. (C) Example pictures of iN cells with indicated inhibitors. Green, EGFP fluorescence (Scale bars: 50 µm.). (D) iN cells generated with 3sm plus ROCK inhibitor express neuronal markers, including MAP2. (Scale bars: 50 µm.) (E) The effect of indicated small molecule combinations on reprogramming efficiency. *P <0.05, paired t test. n = 3 individuals. (F) The effect of the duration of 3sm plus ROCK inhibitor with (yellow bars) or without (blue bars) GDNF and BDNF on reprogramming efficiency. n = 3 individuals. (G) Representative traces of action potential responses of PBMC-derived iN cells under control condition (N3 only; Top) and in the presence of 3sm plus ROCK inhibitor when cocultured with glia for 21 d (Left) or 42 d (Right). (H) Graph showing the total number of neurons in four conditions: DMSO control (black line), 3sm treatment for 21 d (blue line), 3sm plus ROCK inhibitor for 21 d (red line), and 3sm plus ROCK for 14 d and 3sm for 7 d. n = 3 individuals. *P < 0.05.

Importantly, however, when we tested their functional properties, cells generated in the presence of ROCK inhibition were not able to generate mature action potentials, unlike cells grown without the ROCK inhibitor (Fig. 3G). The lack of mature action potentials suggests that ROCK inhibition simply affects cytoskeletal rearrangements but perturbs functional neuronal maturation. To assess whether transient ROCK inhibition is sufficient to increase conversion efficiency and still allow for functional maturation, we removed the ROCK inhibitor and 2 wk after 3sm plus ROCK inhibitor treatment and found no beneficial effect (Fig. 3H).

Molecular Characterization of Blood-Derived iN Cells.

To assess the transcriptional changes induced by reprogramming, we performed RNA-sequencing (RNA-seq) on the donor PBMCs as well as on EGFP⁺ and PSA-NCAM⁺ blood iN cells purified by magnetic cell sorting. A total of 6,941 genes were differentially expressed between the PBMCs and blood iN cells (Fig. 4A). The up-regulated genes were enriched for Gene Ontology terms such as nervous system development and synaptic transmission, while the down-regulated genes were enriched for cellular defense responses (Fig. 4A). Pan-neuronal markers (TAU, TUBB3, MAP2, and NCAM) were up-regulated while blood surface markers (CD8, CD45, and CD3) were silenced (Fig. 4 B and C). As expected, proproliferative genes were down-regulated and negative cell cycle regulators were induced (Fig. 4 D and E). These results suggest that blood iN cells have silenced hematopoietic transcriptional programs and have adopted a pure neuronal identity.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Blood iN cells express genes characteristic of excitatory, postmitotic neurons. (A) Heatmap showing 6,941 genes differentially expressed between PBMC and blood iN cells. Two biological replicates per population, greater than twofold change and P < 0.05. Shown are the seven most significant (P < 0.05, Bonferroni-corrected) Gene Ontology terms among up- and down-regulated genes using PANTHER. (B) Induction of pan-neuronal markers. (C) Suppression of blood cell genes. (D) Down-regulation of cell cycle activators. (E) Induction of antiproliferative cyclin-dependent kinase inhibitors. (F) Immunofluorescence of blood iN cells showing expression of the excitatory marker vGLUT and subtype markers SATB2 and CTIP2 at 21 d after infection and cultured with 3sm on glia (Scale bars: 50 µm.). (G) Quantification of day 21 blood iN cells grown on glia in 3sm conditions by immunofluorescence for indicated markers. (H) Expression of region-specific markers by the PBMC-derived iN cells by RNA-seq. (I) Validation of the neurotransmitter-specific markers by RNA-seq. (J) Validation of the cortical subtype-specific markers by RNA-seq.

To characterize the regional and neurotransmitter identity of PBMC-derived iN cells, we performed immunofluorescence and considered the results in conjunction with the RNA-seq data. The two methods yielded very consistent results. We found that blood iN cells expressed the vesicular glutamate transporter (VGLUT) but not the vesicular GABA transporter, GAD65, or tyrosine hydroxylase, suggesting that blood iN cells are excitatory neurons similar to fibroblast iN cells and Ngn2-ES iN cells (8, 9, 27) (Fig. 4 F, G, and I). Based on the RNA-seq results, we found expression of forebrain markers and cortical layers II–V but not layers I, IV, and VI (Fig. 4 H and J). Immunofluorescence analysis demonstrated that in fact almost all blood iN cells were positive for SATB2 and CTIP2 but negative for REELIN, as suggested by the RNA results (Fig. 4 F and G). SATB2/CTIP2 double-positive cells are found in layer V of the mouse cortex (28).

The Blood-to-Neuron Conversion Does Not Involve a Proliferative Neural Progenitor State.

Several groups have reported the conversion of blood cells into proliferative neural cells using subsets of iPS cell reprogramming factors (20, 21). In contrast, our reprogramming factors are not proproliferative and our approach yields postmitotic neurons directly. Nevertheless, it may be possible that even our reprogramming process involves a proliferative intermediate progenitor. To address this question, we decided to stain the cells at different time points between days 3 and 21 during reprogramming with the neural progenitor marker Sox1 and the proliferation marker Ki67 (SI Appendix, Fig. S2). To our surprise, Ki67-positive cells decreased only after the first week of reprogramming. Nonetheless, all Ki67-positive cells were Sox1-negative, and the number of proliferative cells declined rapidly after day 7. Thus, no proper neural progenitor cells are formed as transient intermediates. The initial persistence of proliferative cells is likely due to the cytokines that we used to activate lymphocytes.

Blood iN Cells Are Stable Without Persistent Transgene Expression.

To examine whether the neuronal identity is dependent on continued transgene expression, we first examined whether the transgenes were perhaps already silenced in our original transfection protocol (Fig. 1A). We had used a bicistronic delivery of Ngn2 and Ascl1 linked by the T2A self-cleaving peptide. After cleavage, the T2A peptide sequence is predicted to be fused in frame with Ascl1 and thus can serve as a tag of the exogenous Ascl1. Immunostaining with T2A antibodies showed that as many as 40% of blood iN cells were T2A-Ascl1 negative, and FACS analysis demonstrated that approximately the same fraction of cells had silenced the EGFP transgene, which was co-electroporated together with the reprogramming factors (SI Appendix, Fig. S3 A and B). Because we used the mouse cDNAs for all four reprogramming factors to reprogram human cells, we could accurately distinguish the exogenous (mouse) factors from the endogenous (human) ASCL1, NGN2, BRN2, and MYT1L genes in our RNA-seq dataset. This analysis clearly demonstrates that the exogenous factors were effectively silenced in the EGFP⁻/PSA-NCAM⁺ iN cell population (SI Appendix, Fig. S3C). Thus, blood iN cells can adopt a stable neuronal identity without the need for continued expression of exogenous reprogramming factors.

Synaptically Competent iN Cells Can Be Derived from CD3⁺ T Cells.

PBMCs consist of fairly heterogenous hematopoietic cell populations. We wondered whether iN cells could be established from a more defined cell type. After characterizing freshly transfected cells, we found that our electroporation conditions greatly favor CD3⁺ T cells, suggesting that the vast majority of PBMC iN cells are in fact T cell-derived (SI Appendix, Fig. S4 A and B). To more specifically test whether T cells can be converted, we introduced the BAMN factors into four defined cell populations purified based on CD3 and CD4 expression. Indeed, morphologically complex iN cells were induced from CD3⁺/CD4⁻ and CD3⁺/CD4⁺ cells, but not from CD3⁻ cells (Fig. 5A). T cell iN cells showed passive and active neuronal membrane properties when cocultured with glia for 18 d and recorded on day 21 (Fig. 5 B and C and SI Appendix, Fig. S4E). Finally, we confirmed the T cell origin of PSA-NCAM FACS-purified iN cells, by demonstrating genomic VDJ rearrangements at TCR-β locus (SI Appendix, Fig. S4 C and D).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Generation of synaptically competent iN cells from T cells. (A) Relative reprogramming efficiency of iN cells from the four indicated PBMC populations. The plotted efficiency was normalized by the electroporation efficiency and the efficiency of the CD3⁺/CD4⁻ cell population was set to 1 (n = 3 individuals). (B) Recording configuration of EGFP-labeled blood iN cells cocultured with unlabeled iPS cell-derived neurons. The recording electrode (Rec) was placed onto an EGFP-positive blood iN cell (white arrowhead) surrounded by nonfluorescent human iPS cell-derived neurons (black arrowheads). (Scale bar: 50 µm.) (C) Example traces of action potential firing recorded from iN cells derived from CD3⁺/CD4⁻ (red) or CD3⁺/CD4⁺ (blue) T cells. N represents the number of cells patched that show action potentials over the total number of cells patched. (D) Representative traces of AMPA receptor-mediated spontaneous network activity (Top, black) recorded from a T cell-derived iN cell, indicating successful integration into the human synaptic network. The trace in red (Bottom) represents an expanded view of the boxed area. Spontaneous PSCs recorded from a T cell-derived iN cell (Top, black) and subsequently blocked by CNQX and picrotoxin (Bottom, black). The trace in red (Middle) represents and expanded view of the boxed area. This pattern was observed in 2 of the 27 cells patched. (E) Evoked PSCs (five trials) in response to extracellular field stimulation recorded from a blood iN cell (Top), which was subsequently blocked by CNQX and picrotoxin application (Bottom).

The defining property of a neuron is its ability to make synaptic connections. Therefore, we asked whether BAMN-induced T cell iN cells would be able to form functional synapses. Local administration of GABA and AMPA on the soma and proximal dendrites of day 21 T cell iN cells recorded in voltage-clamp mode yielded prominent GABA_A receptor-mediated inhibitory postsynaptic currents (PSCs) and AMPA receptor-mediated excitatory PSCs, respectively (SI Appendix, Fig. S4 G and H), demonstrating the presence of functional neurotransmitter receptors. To address whether the T cell iN cells have the capacity to form synapses and functionally integrate into existing neuronal networks, we plated EGFP-labeled iN cells at 14 d postinduction onto unlabeled human iPS cell-derived neurons. At 21 d after coculture, we performed voltage-clamp recordings from EGFP-positive cells and observed synaptic AMPA receptor-mediated spontaneous network activities, indicating their successful integration into the synaptic network (Fig. 5D). In addition, spontaneous PSCs could also be observed in these cells, as confirmed by application of the specific AMPA and GABA_A receptor antagonists CNQX and picrotoxin, respectively (Fig. 5D). Finally, evoked PSCs could also be elicited by activating surrounding axons via extracellular field stimulation of the vicinity (Fig. 5E), unambiguously demonstrating that blood iN cells can receive functional synaptic inputs from other neurons.