Partial Reprogramming of Pluripotent Stem Cell-Derived Cardiomyocytes into Neurons

Direct reprogramming of somatic cells has been demonstrated, however, it is unknown whether electrophysiologically-active somatic cells derived from separate germ layers can be interconverted. We demonstrate that partial direct reprogramming of mesoderm-derived cardiomyocytes into neurons is feasible, generating cells exhibiting structural and electrophysiological properties of both cardiomyocytes and neurons. Human and mouse pluripotent stem cell-derived CMs (PSC-CMs) were transduced with the neurogenic transcription factors Brn2, Ascl1, Myt1l and NeuroD. We found that CMs adopted neuronal morphologies as early as day 3 post-transduction while still retaining a CM gene expression profile. At week 1 post-transduction, we found that reprogrammed CMs expressed neuronal markers such as Tuj1, Map2, and NCAM. At week 3 post-transduction, mature neuronal markers such as vGlut and synapsin were observed. With single-cell qPCR, we temporally examined CM gene expression and observed increased expression of neuronal markers Dcx, Map2, and Tubb3. Patch-clamp analysis confirmed the neuron-like electrophysiological profile of reprogrammed CMs. This study demonstrates that PSC-CMs are amenable to partial neuronal conversion, yielding a population of cells exhibiting features of both neurons and CMs.

transcription factors, namely Brn2, Ascl1, Myt1l, and NeuroD (BAMN), are required for this process, which takes only a few days 8 . These neural cell types could be utilized to model neurological disorders such as Parkinson's disease and Alzheimer's disease, to screen for potential neurotoxicities associated with pharmacological compounds in active drug development, or to potentially treat neurodevelopmental diseases or acquired neurological disorders such as spinal cord injury-induced paralysis 9 . Neural cell types are not the only electrophysiologically-active somatic cell type that has been produced via direct reprogramming. Indeed, direct reprogramming of fibroblasts by overexpression of Gata4, Mef2c, Tbx5, Hand2, Mesp1, Ets2, Myocardin, or microRNAs has been reported to generate cells that exhibit some features of cardiomyocytes such as sarcomeric protein gene expression, striation patterns, and spontaneous calcium transients [10][11][12][13][14] . While it remains to be clarified whether any of these in vitro directly reprogrammed cardiac cells exhibit the full repertoire of gene expression and structural and biochemical function as their target cell (i.e. fully functional cardiomyocytes), this approach represents a major departure from the developmental paradigm of stem/progenitor cells giving rise to differentiated daughter cells. It raises the possibility that somatic cells may be converted to cardiovascular cells by transcription factor overexpression. As a testament to the rapid pace of this field, direct reprogramming has also been able to generate pancreatic beta cells from exocrine cells and, more recently, functional hepatocytes from fibroblasts 15,16 . A number of these directly-reprogrammed somatic cell types are currently being considered for clinical translation 17 .
The direct reprogramming protocols for the aforementioned somatic cell types will continue to improve over time. However, in the case of electrophysiologically active cell types such as cardiomyocytes and neurons, both cell types have currently been produced by reprogramming either dermal fibroblasts or cardiac fibroblasts, which are structurally simple and electrophysiologically inert. To further evaluate the strength and efficacy of the direct reprogramming process, specialized, electrophysiologically-active cell types derived from different germ layers should also be tested for their propensity to interconvert. As a proof-of-principle, we examined the ability of recently described neurogenic reprogramming factors Brn2, Ascl1, Myt1l (BAM) (for mouse), plus NeuroD (BAMN) (for human) to convert mouse and human pluripotent stem cell-derived cardiomyocytes (PSC-CMs) into induced neurons 2 . Although the mesoderm-derived cardiac cell types and ectoderm-derived neurons arise from separate developmental origins, specialized cardiomyocytes of the cardiac electrical conduction network, such as Purkinje fibers, overlap with neurons in terms of gene expression for calcium and potassium channels needed for action potential propagation, intermediate filaments for the maintenance of spiny structure, and neural crest-associated markers [18][19][20] . These similarities may facilitate the reprogramming process between the two electrophysiologically active cell types.
This work provides novel insight into direct somatic cell reprogramming by testing the strength of the neurogenic BAMN factors in activating the neurodevelopmental program in a non-ectodermal, highly-specialized, electrophysiologically active cardiac cell type, namely cardiomyocytes. We utilized single-cell qRT-PCR, immunofluorescence, time-lapse microscopy, and patch-clamp electrophysiology to characterize the sequential process of human and mouse PSC-CM neuronal conversion. We also identified partially reprogrammed, neuron-cardiomyocyte cells that harbor both cardiomyocyte and neuronal gene expression.

Lentivirus-Transduced mESC-CM.
To document the step-wise transition of Nkx2-5-eGFP + mESC-CMs into neuron-like cells, we performed time-lapse microscopy of BAM lentivirus-transduced eGFP+ CMs for three days following Dox treatment ( Fig. 2A, Supplemental Movie 1). Several eGFP+ cells readily underwent morphological change after BAM overexpression for only 3 days and displayed dendrite-like projections with an elongated shape. These cells expressed neuronal markers such as Tuj1 (Fig. 2B). With a longer treatment period, additional neuronal markers such as Map2 can be visualized. Some cells exhibited simultaneous expression of cardiac marker Nkx2-5 and of neuronal markers. Aside from the presence of these neuronal genes in cells that were previously CMs, some of these cells also expressed synapsin, a mature neuronal marker, in the expected distribution at nerve terminals at 21 days after Dox treatment. We determined that approximately 25% of mouse PSC-cardiomyocytes initiated neuronal conversion by their Tuj1+ cTnT− expression, whereas approximately 18% of cells exhibited a partial neuronal phenotype by their dual Tuj1+ cTnT+ cardiomyocyte-neuron gene expression. Real-time quantitative PCR analysis on mouse CMs treated with BAM lentivirus transduction and Dox induction confirmed the increased expression of neuronal specific genes (Fig. 2C, left column). Expression of cardiac marker Nkx2-5 remained largely unchanged during the conversion process. Cardiac-specific marker cTnT decreased transiently in doxycycline-treated transduced cells at day 4 post-dox treatment. Expression of atrial specific marker MLC2a was reduced in doxycycline-treated transduced cells at days 4 and 7 post-Dox. Notably, expression of most, if not all, sarcomeric proteins were only transiently reduced. This is likely due to the attenuation of BAMN expression at day 7 resulting in failure of maintenance of neuronal reprogrammed phenotype in  We also examined cardiac-specific markers (right column). Notably, the expression of cardiac-specific protein Nkx2-5 is largely retained following transduction. We observed a drop in the expression of atrial specific marker MLC2a at both day 3 and day 7. We also observed a drop in ventricular-specific marker MLC2v and cardiac troponin T (cTnT) at day 4, but this returned the same levels as the dox-untreated cells by day 7. *Indicates P < 0.05, NS indicates not significant. N = 3 biological replicates. Data are expressed as means ± SEM.

Neuronal Conversion of Human iPSC-Derived CMs Using BAMN-expressing Lentiviruses.
To determine whether the neuronal conversion of CMs by BAM factor overexpression is conserved across different species, we differentiated human induced pluripotent stem cells (hiPSCs) into cardiomyocytes (CMs) using a previously-published protocol 22 (Fig. 3A) and treated these purified hiPSC-CMs with lentiviruses expressing BAMN, which has been shown to induce neuronal reprogramming in human cells (Fig. 3B) 23 . Following lentiviral transduction and Dox treatment, a phenotypic conversion of CM towards neuronal morphologies was observed as early as 3 days and became progressively more prominent by 5 days post-Dox treatment (Fig. 3C). Similarly to mESC-CMs, transduced hiPSC-CMs expressed human neuronal marker Tuj1 in conjunction with the pan-cardiomyocyte marker cTnT after 7 days of Dox treatment (Fig. 3D). Quantification of transduced cell populations (as determined by their ectopic expression of Ascl1) indicates that approximately 35% of transduced human iPSC-CMs adopted a Tuj1+ cTnT+ phenotype, supporting the initiation of their cardiomyocyte-to-neuron cell transition process. However, only 0.29% of transduced human CMs adopted a more neuronal phenotype by their Tuj1+ cTnT− expression profile. Single-cell quantitative RT-PCR was also performed on BAMN-treated hiPSC-CMs to assess the changes in neuronal and cardiac gene expression in the same cell over the three-week timecourse (Supplemental Figure 1). At an early timepoint (1 week), purified single hiPSC-CMs adopted a gene expression profile similar to that of control primary human heart tissue, where the expression of cTnT is high and the expression of neuronal genes such as Map2 is absent. At a later timepoint (2 weeks), a change in gene expression pattern is observed where the neuronal gene expression is increased in many cells while their cardiomyocyte gene expression is either unchanged or decreased. In one cell (2 WK-17), the expression of cardiomyocyte genes

Electrophysiological Characterization of Neuronal-like Cells Derived from Human iPSC-Derived
CMs. At baseline, we determined that the day 30 hiPSC-CMs we employed to initiate the neuronal reprogramming process were of atrial-like, ventricular, like, and nodal-like subtypes without any neuronal features (Supplemental Figure 2, Supplemental Tables 1 and 2). To determine whether the cells that exhibit neuronal morphologies from reprogrammed hiPSC-CMs also display electrophysiological characteristics of functional neurons, we performed single-cell current-clamp recordings in reprogrammed hiPSC-CMs at week 1 (7-10 days) and week 3 (18-21 days) after transduction with BAMN lentivirus and Dox treatment (Fig. 4A). We injected a 30 pA depolarization current of 1 s duration to test the firing ability of these cells, since the ability to fire a succession of repetitive action-potentials (AP) upon membrane depolarization is a unique characteristic of mature neurons 24 . At week 1 after transduction, the neuronal-shaped cells derived from hiPSC-CMs fired only abortive APs (Fig. 4Bi). Interestingly, at week 3 (i.e. day 21) of transduction, the neuronal-shaped cells could fire multiple and repetitive APs (Fig. 4Ci). These week 3 APs were similar to those exhibited by primary rat neuronal cultures (Supplemental Figure 3). There appears to be a progressive increase in the Na v and K v currents from days 7 to 21 as well which correlates well with the increased ability of neuronal-shaped cells to fire repetitive APs. Since the amplitude of APs and membrane excitability is determined by the sodium (Na v ) and potassium (K v ) currents, we characterized the changes of Na v and K v currents at week 1 and week 3 post-transduction. In whole-cell voltage-clamp mode, we observed rapidly inactivating inward Na v currents and persistent outward K v currents in response to depolarizing voltage steps that could be blocked by tetrodotoxin (TTX) and a cocktail of 4-aminopyridine (4-AP) and tetraethylammonium (TEA), respectively (Fig. 4Cii). Taken together, these data show for the first time the feasibility of inducing neuronal electrophysiological phenotype in human iPSC-CMs by transduction with neurogenic transcription factors.

Discussion
Using lentiviruses that overexpress neurogenic transcription factors previously shown to induce neuronal conversion from fibroblasts, we found that mouse and human PSC-CMs are amenable to partial neuronal phenotype conversion. We observed a significant number of cells expressing markers of both cardiomyocytes and neurons, suggesting the presence of cells in an intermediate, partially-reprogrammed state. Electrophysiological studies of reprogrammed cells with neuronal morphologies demonstrated the presence of mature neuronal action potentials in some of the reprogrammed cells. Taken together, these results demonstrate the ability of neuronal transcription factors to directly reprogram mesoderm-derived PSC-CMs into electrophysiologically active cells that exhibit variable degrees of neuron phenotype.
Advances in lineage-specific reprogramming by transcription factor overexpression have raised significant promise that this approach may be useful to generate various types of cells for cell-based therapy. The most successful example of this approach is the generation of PSCs from fibroblasts by overexpression of highly potent transcription factors such as Oct4 and Sox2 that has been shown to be able to erase fibroblast epigenetic markings and induce epigenetic markings of pluripotent stem cells 25,26 . Further progress in this realm of research has now led to reprogramming of fibroblasts into various cell lineages such as neurons, cardiomyocytes, blood cells, hepatocytes, and oligodendrocytes with variable efficiency. Our results demonstrate that neuronal transcription factors can partially reprogram PSC-CMs into neuron-like cells. These results support the presence of pioneer transcription factor activity in BAM/BAMN for the neuronal lineage, given their potency and efficiency to induce neuronal lineage conversion regardless of the differences in germ layer origin or structural specialization of the starting cell population. Indeed, a very recent study employing chromatin immunoprecipitation of BAM factors followed by genome-wide high-throughput sequencing has confirmed that Ascl1 most likely plays a pioneer factor role during neuronal reprogramming 8 . This factor may open closed chromatin and allow for the entry of other neuronal transcription factors, further driving the neuronal conversion process by silencing non-neuronal genes. Regarding the silencing of the cardiogenic program, a recent study addresses the intermediate reprogrammed cell phenotype using single cell RNA sequencing and found that a competing myogenic transcriptional program must be silenced in order to complete the neuronal reprogramming process 27 .
Our data demonstrate that many PSC-CMs are partially reprogrammed as demonstrated by their expression of both CM and neuronal genes (Figs 2 and 3D, Supplemental Figure 1). For example, several reprogrammed cells are positive for the early cardiac marker Nkx2-5 as well as for neuronal markers such as Tuj1, Ascl1, and MAP2 (Fig. 2B). As mentioned by studies looking at cardiomyocyte or neuron production through direct reprogramming, there are a number of intermediate and partially-reprogrammed cells produced during the direct conversion process from fibroblast to cardiomyocyte and from fibroblast to neuron 1,10 . Interestingly, we observed a higher efficiency of neuronal conversion in mouse cardiomyocytes in comparison to human cardiomyocytes. The fact that the human PSC-CMs in our study achieved Tuj1+ cTnT− neuronal conversion at an extremely low rate (0.29%) reflects the difficulty of complete conversion from cardiomyocytes to neurons in this population.
The initial study that used BAM to produce induced neurons demonstrated that most of the converted cells were excitatory, glutamatergic neurons, largely of the cortical variety. That study also found some induced neurons expressed markers of GABAergic neurons 1 . However, despite the presence of some features of mature neurons and spontaneous action potentials in our reprogrammed cells that resemble the action potential of fully mature neurons, many cells continue to express cardiomyocyte genes. This is reminiscent of the immature cells produced in previous studies of in vitro direct reprogramming of fibroblasts into CMs 10,28,29 . Given the immaturity of these cardiomyocyte-converted neuron-like cells, would they be more amenable to conversion back to the cardiomyocyte state as well? Previous studies have indeed demonstrated that somatic tissues, namely fibroblasts, can be directly converted into cardiomyocytes, in both humans and mice, using the defined transcription factors Scientific RepoRts | 7:44840 | DOI: 10.1038/srep44840 Gata4, Mef2c, and Tbx5 (GMT) 10,30 . However, we have previously demonstrated that this conversion process is extremely inefficient, with less than 5% cardiomyocyte-like cells being produced from this fibroblast conversion Scientific RepoRts | 7:44840 | DOI: 10.1038/srep44840 process 28 . As such, we have not attempted to use the GMT protocol to convert cells such as the CM-neuronal cells back into cardiomyocytes. The electrophysiological, structural, and functional immaturity of directly reprogrammed cells remains a major issue in stem cell biology and regenerative medicine applications 31 . Further work will be needed to improve the fidelity of the reprogrammed phenotype and the efficiency of lineage conversion to be able to achieve a homogenous population of the desired cell type.
In summary, we have, for the first time, converted mouse and human CMs into neuronal-like cells. Many of the converted cells exhibit partial features of both CMs and neurons. The fact that only three or four neuronal transcriptional factors are needed to convert mesoderm-derived, electrophysiologically-active CMs into ectoderm-derived, electrophysiologically-active neuronal-like cells suggests that distinct cell fates are regulated by very few key transcription factors.

Methods
All experiments were performed in accordance with relevant guidelines and regulations of the Stanford University School of Medicine.

In vitro differentiation of murine Nkx2-5-eGFP ESCs and flow sorting for eGFP + ESC-CMs.
Mouse ESCs were cultured in accordance with previous studies and adapted to gelatin-coated dishes in the presence of leukemia inhibitory factor (LIF) for 2 days prior to differentiation 32 . CM differentiation of ESCs was performed per a previously-published protocol 21 . On the day of flow sorting, embryonic bodies (EBs) were digested with trypsin/EDTA for 3 minutes followed by collagenase A + B solution (10 mg/ml each) for 30 minutes. These cells were then re-suspended in HBSS (Life Technologies) with 20% FCS (Sigma). The eGFP-expressing cells were analyzed and sorted using a FACS Aria II system, and flow cytometry data were analyzed using FACS Diva Software (Becton Dickinson).

Maintenance of hiPSCs and differentiation of hiPSCs to hiPSC-CMs.
Human iPSCs were maintained on Matrigel (BD Biosciences) coated plates in E8 medium (Life Technologies), which was changed daily 33 . Prior to cardiac differentiation, hiPSCs were passaged once every four days with 0.5 mM EDTA at 37 °C for 5 minutes. After hiPSCs had achieved 95% confluence, cardiac differentiation was initiated by treating cells with 6 μ M CHIR99021 (Selleck Chemicals) in RPMI + B27 without insulin for 48 hours (day 0 to day 2) according to a previously-published protocol 23 . The medium was then changed to RPMI + B27 without insulin for 1 day, and 5 μ M IWR1 (Selleck Chemicals) was added between days 3 to 5 of differentiation. From day 5 to 7, cells were returned to RPMI + B27 without insulin and without small molecules. Cells were then maintained on RPMI + B27 with insulin from day 7 onwards. Once spontaneous beating could be observed (usually between days 10 to 12 of differentiation), the medium was changed to RPMI without glucose + B27 with insulin for glucose starvation for 3 days to eliminate non-CMs in culture. This step typically resulted in greater than 90% CM purity. The surviving CMs were then passaged with trypLE TM (Life Technologies) and replated for further studies. Viral transduction. Replication-incompetent, VSVg-coated lentiviral particles carrying doxycycline (Dox)-inducible cDNA for BAM (mouse studies) or BAMN (human studies) expression were packaged in 293 T cells with a Fugene HD kit, where the DNA to Fugene volume ratio of 1 to 2.5 was given per the manufacturer's suggested protocol. Viruses produced were further concentrated by PEG-it virus concentration solution (System Biosciences) and stored at − 80 °C until needed. To induce neuronal phenotype conversion in mouse and human PSC-CMs, we transduced FACS-purified eGFP+ CMs from in vitro differentiated Nkx2-5-eGFP ES cells and human iPSC-CMs at 2 passages after glucose-starvation with prepared lentiviruses according to previously published protocols 1,2 . After PSC-CMs had been incubated for 16-20 hours in lentivirus-containing media, the cells were switched into fresh N3 media containing doxycycline (2 μ g/ml) to activate BAM or BAMN expression. The media was changed every 2-3 days for the remainder of the culture period.
Electrophysiology. For electrophysiological assays, cultured hiPSC-CMs or reprogrammed cells were dissociated using TrypLE for 10 min at 37 °C, centrifuged at 200xg, suspended in RPMI media supplemented with B27 + insulin (RPMI + B27 with insulin), filtered through a 70 or 100 μ M cell strainer (BD Biosciences), and plated as single cells (1 × 10 5 cells per well of a 24-well plate) on No. 1 (8 mm) glass cover slips (Warner Instruments, Hamden, CT, USA) coated with Matrigel (1:50 ratio). Immediately after plating, cells were grown in RPMI + B27 with insulin media supplemented with 2 μ M thiazovivin and allowed to attach for 48-72 hours, changing the media every other day. Cells were analyzed at indicated times after transduction. Using an EPC-10 patch-clamp amplifier (HEKA, Lambrecht, Germany), whole-cell action potentials and currents were recorded in standard current and voltage clamp mode, respectively, as shown previously [34][35][36] . Glass micropipettes (2-3 MΩ tip resistance) were fabricated from standard wall borosilicate glass capillary tubes (Sutter BF 100-50 −10 , Sutter Instruments, Novato, CA) using a programmable puller (P-97; Sutter Instruments) and filled with the following intracellular solution: 123 mM K-gluconate, 10 mM KCl, 1.0 mM MgCl 2 , 10 mM HEPES, 1 mM EGTA, 0.1 mM CaCl 2 , 1 mM K 2 ATP, 0.2 mM Na 4 GTP and 4 mM glucose; pH was adjusted to 7.2 with KOH. Membrane potentials were kept around − 65 to − 70 mV, and step currents were injected to elicit action potentials. For voltage-dependent current recordings, the same internal solution as was used. Whole-cell sodium and potassium currents were recorded at a holding potential of − 90 mV, and 200 ms voltage steps ranging from − 80 mV to +90 mV were delivered at 10 mV increments. Series resistance and cell capacitance were compensated to between 50-80% in all voltage clamp recordings as shown previously 35,36 . All recordings were made at room temperature except for the cardiac action potentials, which were recorded at 35-37 °C. The extracellular solution contained 140 mM NaCl, 5 mM KCl, 2 mM CaCl 2 , 2 mM MgCl 2 , 10 mM HEPES, and 10 mM glucose; pH was adjusted to 7.4 with NaOH. A single beating CM or cell exhibiting neuronal morphology was selected for action potentials (APs) or current recording. Data were acquired using PatchMaster software (HEKA, Germany), digitized at 1.0-10 kHz and analyzed using IGOR Pro V:6.2.2.2, a custom-built MatLab script, and MS Excel software. The following are the criteria used for classifying observed APs into ventricular-, atrial-and nodal-like hiPSC-CMs: for ventricular-like cells, the criteria were a negative maximum diastolic membrane potential (less than or equal to 50 mV), a rapid AP upstroke, a long plateau phase, AP amplitude > 90 mV and AP duration at 90% repolarization/AP duration at 50% repolarization (APD 90 /APD 50 ) < 1.4. For atrial-like cells, the criteria were an absence of a prominent plateau phase, a negative diastolic membrane potential (less than or equal to 50 mV) and APD 90 / APD 50 > 1.7. For nodal-like cells, the criteria were a more positive MDP, a slower AP upstroke, a prominent phase 4 depolarization and APD 90 /APD 50 between 1.4 and 1.7 34 .
Gene expression analysis. For gene expression analysis, cultured cells were rinsed with PBS, immediately harvested using Trizol ® reagent (Invitrogen, Carlsbad CA), and stored at − 80 °C until needed. Total RNA from each sample was purified from cell lysate using QIAGEN RNAeasy kit per the manufacturer's instructions. Quantitative PCR was performed on cDNA made from reverse-transcribed RNA using the iScript TM cDNA synthesis kit (BioRad, Hercules, CA). For single cell analysis, cells were rinsed with PBS and then digested with a combination of trypsin and collagenase A and B (10 mg/mL each). Single cells were then manually selected under bright-field microscopy and subjected to cell lysis and target gene-specific reverse transcription and amplification according to a previously-published protocol 37 . Quantitative PCR was then performed using the BioRad CFX96TM qPCR system with SYBR ® Green substrate (BioRad, Hercules, CA) for 40 cycles. A gene expression heat map for single cell analysis was generated in R software.
Statistical analysis. Data presented as mean ± standard error of the mean (SEM) unless otherwise specified.
Comparisons were conducted via Student's t-test with significant differences (*) defined by P < 0.05.