Generation of cerebral cortical GABAergic interneurons from pluripotent stem cells

Abstract

Cortical interneurons are inhibitory GABAergic neurons that originate in the caudal or medial ganglionic eminences of the embryonic subcortical telencephalon, then migrate into the cortex. Differentiation conditions mimicking developmental signals found in the telencephalon are used to generate interneurons from embryonic and induced pluripotent stem cells. Stem cell-derived interneurons should be validated by migratory capacity, morphology, gene and protein expression, and electrophysiology.

Open in new tabDownload slide

INTRODUCTION

The cerebral cortex functions by an intricate mix of activities in neurons, glia, and vasculature. There are two major subclasses of cortical neurons, excitatory neurons using the neurotransmitter glutamate and inhibitory neurons using the neurotransmitter GABA. In general, the excitatory neurons have axonal projections extending beyond the neuron's local region, while the inhibitory neurons tend to project more locally and have hence been termed “interneurons.” These cells can be grouped into subclasses based on morphological, neurochemical, electrophysiological, and axonal targeting characteristics, as well as by their influences on local circuitry.¹ In addition, they can also be grouped based on distinctions in the origins of interneuron subclasses within the subcortical telencephalon.² Cortical subtypes, of which a dozen or more are thought to exist, are characterized by combinations of these subgrouping features.³ There is considerable controversy as to which features best define a cell “type”.⁴ This controversy is rendered more complex by species differences, particularly in the neurochemical domain,⁵ and primate and even human-specific interneuron subclasses have been proposed.⁶

Cortical interneuron (CIn) dysfunction has been implicated in a variety of neuropsychiatric disorders, including epilepsy, autism, schizophrenia, anxiety disorders, Alzheimer's disease, and attention deficit disorder, among others.^7,8 Hence, there has been tremendous interest in generating cortical GABAergic interneurons from human pluripotent stem cells both for the study of pathological processes and for the development of cell based-therapies. In this concise review, we discuss recent progress in understanding how interneuron subtypes are generated in vivo, and how that progress is being applied to the generation of rodent and human CIns in vitro. In addition, we will discuss approaches for the rigorous designation of interneuron subgroups or subtypes in transplantation studies, and challenges to this field, including the protracted maturation of human interneurons.

GENERATION OF CIns

In the early embryo, consisting of ectoderm, mesoderm, and endoderm, neural induction of the ectoderm takes place upon inhibition of transforming growth factor beta (TGFβ) and BMP signaling, which would otherwise activate SMAD signaling.^9,10 This serves as the rationale for differentiation protocols using dual-SMAD inhibition to rapidly and efficiently produce neural progenitors from human embryonic stem cells (hESCs).¹¹ Prior to being posteriorized in response to retinoic acid and other signaling molecules,¹² neural progenitors initially adopt an anterior identity, the most anterior of which, the telencephalon, expresses the transcription factor Foxg1.^13,14

Patterning of telencephalic primordium into distinct dorsal (pallial, cortical) and ventral (pallidal, subcortical) subdivisions with regionally restricted transcription factors (including Dlx1/2, Nkx2.1, Gsh2, Lhx6, and Pax6) is primarily accomplished by dorsoventral gradients of bone morphogenetic protein (BMP), WNT, and Shh (Figure 1).^15,16 The dorsal telencephalon, which gives rise to neocortex, hippocampus, cortical hem, choroid plexus, and parts of the amygdala and olfactory bulb, is mostly patterned by gradients of BMPs and WNTs secreted by the roof plate. On the other hand, ventral patterning of the telencephalon primarily relies on gradients of Shh.¹⁷ Of the ganglionic eminences, the medial ganglionic eminence (MGE) and caudal ganglionic eminence (CGE) are the primary sites from which inhibitory interneurons are generated in both rodents and humans.^2,18,19

FIGURE 1

Schematic of murine telencephalon development and example timeline of human stem-cell derived CIn differentiation. A, An embryonic day 15 mouse brain is illustrated on the left, showing the primary regions of cortical interneuron origination within the medial, lateral, and caudal ganglionic eminences (MGE, LGE, CGE) of the subcortical telencephalon. Similar subcortical origins have been identified in humans. The illustration on the right shows a coronal view through the telencephalon. The genes and signaling molecules shown are some of those frequently used to drive stem cell differentiation into telencephalic neuronal subgroups. The dorsally enriched genes associated with glutamatergic fate determination are highlighted in gray. The red region highlights the LGE which gives rise mainly to GABAergic striatal medium spiny neurons as well as interneurons of the olfactory bulb. The green region highlights the cortical interneuron-producing region of the MGE and preoptic areas. Mash1 is also strongly expressed in the MGE. B, An example of a human CIn differentiation protocol and associated timeline. Wnt inhibitor (XAV939) and dual SMAD inhibition factors (LDN193189, SB43154) initiate differentiation of ESCs or human induced pluripotent stem cell (hiPSCs; blue phase). Cells are then weaned onto media containing N2 supplement B (STEMCELL Technologies), and then are cultured from differentiation days (dd) 11 to 18 in the presence of ventralizing factors SHH and purmorphamine. From dd18 to 30, cells continue to divide and mature in neurobasal media containing B27 and retinoic acid (vitamin A). At roughly dd30, CIn precursors become postmitotic and can be selected (ie, via fluorescence activated cell sorting, FACS) to isolate a purer CIn population for downstream experiments. The further maturation of CIns is then dependent on trophic support and/or excitatory input from other cortical cells—either rodent or human cortical cultures in vitro, or by transplantation. CGE, caudal ganglionic eminence; CIn, cortical interneuron; LGE, lateral ganglionic eminence; MGE, medial ganglionic eminence

Open in new tabDownload slide

Nearly all CIns in mice can be grouped into three nonoverlapping subpopulations that express either the serotonin receptor 5HT3aR, the calcium binding protein parvalbumin (PV), or the neuropeptide somatostatin (SST).²⁰ The PV and SST-expressing subgroups, also generally distinguishable by their spiking characteristics and connectivity,²¹ originate primarily in the MGE where they depend on expression of Nkx2.1 for their fate determination.^22,23 After initial establishment of ventral telencephalic patterning, Shh signaling is required to maintain Nkx2.1 expression, and interneuron fates, by cycling MGE progenitors.^24,25 Interestingly, within the MGE, there is an inverted gradient in the level of Shh signaling, with higher levels of Shh signaling being present in the dorsal-most domain of the MGE that also expresses Nkx6.2.^25,26 Genetic fate mapping for Nkx6.2 reveals that these progenitors primarily give rise to SST-expressing CIns.²⁷ This result is consistent with transplantation studies,^28,29 and with both transgenic mouse and mouse embryonic stem cell (mESC) studies showing that higher levels of Shh signaling favor the SST-interneuron subgroup over those that express PV.^25,30

Upon cell cycle exit, Nkx2.1 expression becomes independent of Shh signaling and is maintained in PV or SST expressing striatal interneurons.^24,31 In contrast, Nkx2.1 must be downregulated in order for CIns to react appropriately to guidance cues directing their migration toward the overlying cerebral cortex.³² Around the time of cell cycle exit Nkx2.1 activates Lhx6,³³ which is required for normal migration as well as for aspects of terminal MGE-interneuron fates,^34,35 including in human stem cell-derived CIns (hCIns).³⁶

Although most knowledge regarding CIn development comes from rodent studies, this work has been complimented by a number of studies on the developing human and primate brain. Consistent with observations in the rodent and ferret brain,³⁷ CIns also originate in the GEs in primates including humans.^18,19,38 While some evidence suggests that there could be a cortical source of CIns in humans,^38-40 and cultured human cortical progenitors exposed to Shh signaling can generate GABA-expressing cells,⁴¹ the bulk of evidence from human fetal slice cultures or histological studies does not support the presence of substantial interneuron genesis in human cortex^18,19,42 (see also data on fetal Nkx2.1 expression from the Allen Brain Atlas).

DEFINING CIns

Prior to considering the literature on generating CIns from pluripotent stem cells, it is critical to consider criteria for establishing interneuron identification (Table 1). In terms of a general definition of CIns, we refer the reader to a consortium of researchers who have attempted to standardize terminology regarding these cells.⁴³ Within this framework, interneuron subtypes are classified broadly by their diverse molecular, morphological, and physiological characteristics. For instance, PV interneurons are defined by their molecular composition (the calcium binding protein PV), their electrophysiology (fast-spiking [FS]), and their morphology (ie, the chandelier subtype of the PV subclass). Other CIn subclasses contain different calcium-binding proteins (calbindin and calretinin), neurotransmitters and/or receptors, distinct firing patterns, and unique morphologies.

Despite the heterogeneity of characteristics in different CIn subtypes, they also have a number of traits in common. First, CIns have the remarkable capacity to migrate nonradially,^44,45 including after their transplantation into adult forebrain.⁴⁶ This property has also been demonstrated with hCIns which migrate from the MGE to cortex when placed on the MGE of mouse embryonic slice cultures,^47,48 and from MGE-like spheroids into cerebral spheroids.^49,50 In addition, hCIns disperse widely after transplantation into mouse neocortex.^47,51,52 Stem cell-derived CIns should therefore have the capacity to migrate extensively and nonradially in vitro or in vivo.

Second, although within the cerebral cortex, there are a few interneuron-selective markers, for example, the neurotransmitter GABA (although also expressed by dentate granule cells), and the transcription factors Lhx6 (for MGE-derived CIns) or Prox1 (for CGE-derived CIns, but again expressed by dentate granulate cells), it is important to note that there are no interneuron-selective markers for cells generated in vitro. GABA and Dlx1, 2, and 5 are expressed at some point by virtually all GABAergic neurons of the forebrain, and Lhx6 is expressed by PV-containing projection neurons of the globus pallidus, as well as projection neurons of the amygdala.⁶⁷ SST (like PV) is expressed by subpopulations of striatal interneurons, and it is not clear how one would identify striatal vs CIns in vitro, apart from their persistent expression of Nkx2.1, which itself may be influenced by exogenous factors and hence unreliable in vitro. In addition, genes expressed in migratory neurons are not necessarily stably expressed in particular fates. For example, SST mRNA is far more widespread in the mouse embryo than expected for the relatively small population that expresses it in cortex,⁶⁸ and the SST-Cre mouse labels a fairly large number of deep-layer PV interneurons that do not have detectable SST by immunofluorescence. Similarly, calbindin is widely expressed by migrating interneurons, in addition to developing cortical pyramidal neurons,⁴⁴ then ends up being expressed by a mixed subset of PV+ and SST+ interneurons in adult cortex.^69-71 Finally, particularly regarding PV, it is important to note that this marker of an FS electrophysiological subclass of interneurons is at least partially activity and excitatory input-dependent.^72-74 Consistent with this notion, PV only becomes detectable relatively late in postnatal development in both rodents and primates, and is then detected in cells with complex dendritic arborizations, suggestive of fairly mature neurons.^75-77 It is thus highly questionable whether a stem cell-derived PV-immunolabeled cell, whether or not it originates from a Nkx2.1+ progenitor and co-expresses GABA, is actually indicative of the PV-subclass of CIn as defined by PV expression, MGE origin, and FS intrinsic firing properties in slices from mature cortex. “Real” PV-subclass CIns should be multipolar, aspiny, and FS (as defined in Reference 43).

An excellent example of this problem is evident in a recent publication that, while shedding innovative light on transcriptional regulation of Nkx2.1-lineage interneuron fates from human stem cells, specifically claims to have generated far larger percentages of PV interneurons than previous studies.⁶⁵ This designation of CIn PV fate is made solely on immunofluorescence for PV in differentiation day 35 cells that appear highly morphologically immature and have signal for “PV” mostly in their cell soma. In addition, they used an ultra-high concentration of a well-established polyclonal antibody that is generally used at least 10x more diluted. The point is not to criticize this particular paper (a similar finding is presented in another excellent paper, this time shortly after interneuron migration into organoids⁴⁹), but to highlight the importance of not designating PV fate, if one is referring to the FS CIn subgroup, based only on immunodetection in an obviously highly immature cell.

GENERATING INTERNEURONS FROM mESCs

In an effort to better understand and treat interneuronopathies, many researchers have turned toward pluripotent stem cells, which can be used to model development and to potentially generate cell-based therapies.⁷⁸ Initial efforts to generate CIns from embryonic stem cells largely used mouse over hESCs given their greater accessibility, relative lack of ethical considerations, ease of genetic modification, and perhaps most importantly, short maturation time in comparison to the protracted maturation of hESC-derived CIns. To date, the generation of CIns from mESCs has largely relied upon two strategies: (a) the addition of growth factors and other small molecules during cell culture that recapitulate the developmental signaling pathways that direct CIn development in vivo and (b) direct induction of CIns using lineage-specific transcription factors (discussed below).

A major advance in generating CIns from mESCs came in 2005 when Sasai and colleagues developed a method to generate telencephalic precursors.⁷⁹ This was accomplished by initially growing embryoid bodies in defined media, followed by dissociation and replating onto an adherent surface. By inhibiting mesoendoderm induction pathways through combining Wnt inhibition by Dkk1 and Nodal/Tgfβ signaling by LeftyA, and later adding the ventralizing factor Shh, 5% to 15% of all cells generated were Foxg1⁺Nkx2.1⁺ and hence could be regarded as MGE-like progenitors. However, while many cells eventually expressed GABA, it is unclear whether these represented MGE-like CIns or other MGE-derived populations, such as projection neurons of the basal ganglia.

Building upon this work, Maroof and colleagues used the bacterial artificial chromosome approach to generate an Lhx6-GFP mESC reporter line to visualize and isolate CIn precursors of an MGE-like lineage. Rather than using early Wnt inhibition, their method combined initial BMP inhibition using Noggin with subsequent ventralization by Shh to generate Lhx6-GFP⁺ CIns. When transplanted into neonatal neocortex, these cells migrated extensively, adopted morphologies consistent with MGE-derived CIns, and expressed the mature interneuron markers PV or SST together with the correspondingly appropriate electrophysiological characteristics. In addition, gene expression patterns of the mESC-derived Lhx6-GFP cells closely resemble those from interneurons present in MGE-derived Lhx6⁺ cells in vivo.⁶¹ However, the method developed by Maroof and colleagues was inefficient such that only 2% to 5% of cells generated expressed Lhx6-GFP.

Improvements in generating CIns from mouse ESCs were demonstrated by Danjo et al, who used a serum-free embryoid body method to show that subregional specification of mESCs into lateral ganglionic eminence (LGE)-, MGE-, and CGE-like cell types could be achieved by varying the timing and concentration of Shh treatment.⁸⁰ They further showed that fibroblast growth factor (FGF) signaling, a regulator of rostral-caudal patterning in the developing forebrain,⁸¹ differentially modulated MGE- and CGE-like fates. While FGF8 increased Nkx2.1 expression at the expense of CoupTFII, FGF15/19 had the opposite effect. This corresponded to a shift in the generation of MGE-derived subgroups defined by expression of SST or PV, to mostly CGE-derived CIns such as vertically oriented bipolar or bitufted calretinin-expressing interneurons. Interestingly, a separate study also showed that Activin, a Tgfβ agonist, acting to inhibit Shh signaling, directed neural precursors toward a CGE-like fate.⁸² This result is remarkably consistent with the fate-switching effects of Shh inhibition in the MGE in vivo.²⁵

Using the Lhx6-GFP mESC line developed by Maroof and colleagues, Tyson and colleagues added the capacity to isolate MGE-like CIn progenitors by inserting a BAC-based Nkx2.1-mCherry reporter.⁸³ They also greatly enhanced telencephalic induction using the small molecule Wnt inhibitor XAV-939 in lieu of Dkk1, such that by differentiation day 12 (DD12) ~87% of all cells were Foxg1⁺. Next, based on rodent studies showing that birth date and Shh exposure differentially bias MGE progenitors to produce PV vs SST fated CIns^25,84 they found that earlier harvest of cells exposed to higher levels of Shh were greatly biased to become SST⁺ CIns, whereas later harvest of cells cultured without exogenous Shh were biased to become PV⁺ CIns.³⁰ Of note, PV interneuron fate was not only established based on marker expression, but on electrophysiological characteristics including the demonstration of FS by the PV+ subgroup.

It should be clear from the above discussion that advances in our understanding of normal development form the foundation for improvements in stem cell differentiation. Following this theme, Petros and colleagues, building upon studies of transgenic mice,^85,86 used in vivo fate mapping and fate manipulations to demonstrate that within the MGE, radial progenitors dividing asymmetrically (giving rise to one neuron, one progenitor) at the ventricular surface are biased toward the generation of SST⁺ interneurons, whereas symmetrical divisions (cyclin D2⁺, giving rise to two neurons) in the subventricular zone (SVZ) are biased toward giving rise to PV⁺ CIns.⁸⁷ Using this finding and evidence that inhibition of atypical protein kinase C (aPKC) increases SVZ divisions in embryonic cortex,⁸⁸ it was recently found that enhancement of symmetric divisions in vitro with an aPKC inhibitor further increases the fraction of transplanted CIns expressing PV vs SST.⁸⁹ This study also showed that Notch inhibition enhanced the generation of SST CIns, consistent with the fate switching effects of Notch inhibition in the MGE in vivo.⁸⁷ However, it remains unclear whether Notch signaling influences interneuron fate per se, or simply enhances the derivation of SST⁺ CIns by driving Nkx2.1⁺ progenitors out of the cell cycle before they can undergo a PV-interneuron generating, SVZ-like symmetric division.

GENERATING CIn-LIKE CELLS FROM HUMAN PLURIPOTENT STEM CELLS

Success in generating CIns from hESCs initially lagged behind that achieved in mouse in part due to their propensity to die after cell dissociation, paucity of reporter lines enabling cell subtype isolation, and intrinsically slower maturation rates. Early work in guiding hESCs toward telencephalic precursors has been reviewed elsewhere. ⁹⁰ Briefly, neural induction of hESCs, as with mESCs, depends on inhibition of mesendoderm induction pathways using a combination of Wnt, BMP, and Nodal antagonists to induce anterior neural tube character. Once this is achieved, Shh-signaling is used to induce ventral (pallidal, subcortical) telencephalic identity.

In 2007, Sasai and colleagues made the seminal finding that application of a selective Rho-associated kinase (ROCK) inhibitor, Y-27632, to hESCs tremendously reduces dissociation-induced apoptosis. This finding greatly enhanced the efficiency and reduced the expense of expanding and passaging hESCs.⁹¹ Two years later, Li et al showed that coordinated Wnt and Shh signaling differentially direct hESCs to dorsal vs ventral telencephalic fates.⁹² That same year, another group shared the pivotal discovery that dual SMAD inhibition using Noggin (exchangeable with either BMPR1A-Fc or LDN-193189) and SB431542, induced rapid and highly efficient neural conversion (>80%) of hESCs under adherent culture conditions, thereby obviating the need for stromal feeders or embryoid bodies.¹¹

Another advance came in 2011 when Goulburn et al generated an Nkx2.1-GFP hESC reporter line using homologous recombination.⁹³ Although this line is not NIH-registered, and has only one functional copy of Nkx2.1, the line enabled the isolation of MGE-like progenitors and their progeny. Now equipped with a dual SMAD protocol to achieve rapid neural induction¹¹ and small molecule Wnt inhibitors (Dkk1, XAV-939, or IWP-2) to induce early forebrain patterning,⁷⁹ a number of groups rapidly succeeded in efficiently generating relatively homogenous (upward of ~90%) cultures of Nkx2.1⁺Foxg1⁺ progenitors.^47,48,51,54

However, this group of studies involving the Nkx2.1-GFP hESC line highlighted the observation that hESC-derived CIns undergo a protracted maturation similar in nature to their in vivo counterparts, which take months or even years to fully mature. Experiments did suggest that hCIn maturation, as measured by intrinsic electrophysiological parameters, progressed faster when the hCIns are cultured on dissociated mouse cortex (excitatory neurons and astrocytes) than when cultured on a bed composed mainly of immature human pyramidal neuron-like cells (and some astrocyte-like cells).⁴⁷ Interestingly, whether hCIns were cultured with human or mouse cortical cells, in both preparations they appeared to mature faster than when very similar hCIns were cultured on rodent astrocytes alone.⁵¹ When transplanted into immunosuppressed mice, hESC-derived SST⁺ hCIns still took up to 30 weeks to achieve mature firing properties.⁵¹

GENERATING hCIn-LIKE SUBCLASSES FROM HUMAN PLURIPOTENT STEM CELLS

With the capacity to generate hCIn-like cells, there has been tremendous interest in particular CIn subclasses, again with some emphasis on the FS, PV-expressing subclass that has been implicated in neuropsychiatric disease.⁹⁴ With regard to other hCIn populations, several studies have reported methods to direct Foxg1⁺ progenitors into CGE-derived cell types, such as the vertically oriented, bipolar, calretinin-expressing interneurons, by treating cultures with Activin,⁸² or FGF19.⁴⁸

With regard to the generation of the main MGE-derived subgroups, those expressing SST and those expressing PV, there has been far more success with the SST subclass. The challenge in generating PV hCIns comes down to three issues: (a) having the correct protocol for generating these cells, (b) (discussed above) having the correct environment so that activity-dependent aspects of their maturation, including the expression of PV itself, can occur, and (c) having the necessary time for maturation to occur. Taken in reverse order, PV expression continues to mature for years after birth.^75,77,95 Hence even with the correct initial differentiation conditions and later neural environment, intrinsic alterations that accelerate maturation may be needed to create a practical PV subclass hCIn protocol.

Regarding the environment to provide activity-dependent aspects of CIn subclass maturation, three-dimensional human cerebral organoids have garnered considerable interest as systems to more faithfully recapitulate the in vivo microenvironment.⁹⁶ Cerebral organoids are capable of producing excitatory and inhibitory cortex-like neurons. ⁹⁷ When organoids of dorsal or ventral forebrain character are cultured in apposition, similar to what has been shown previously in mouse organotypic slice cultures,^98,99 interneurons from the ventralized (Shh exposed) spheroids migrate into the dorsal spheroids composed mainly of glutamatergic neurons and astrocytes.^49,50 Moreover, these interneurons contribute to functional microcircuits.

While the dorsal-ventral organoid fusion system clearly could be valuable for studying interneuron migration, at this point their value in studying interneuron subgroup or subtype development is less clear. Ventralized organoid-derived cells expressing either PV or SST have been identified in the dorsalized portion of organoids, but whether these express firing properties or axon targeting properties of native PV and SST subgroups remains to be determined. Regarding interneuron subgroup fates in organoids, one group found that higher levels of Shh increased SST mRNA expression at day 46,⁵⁰ consistent with the influence of Shh on mESC-derived Nkx2.1+ progenitors,³⁰ but it is not clear whether SST in a highly immature, possibly migratory neuron is a stable marker of mature SST interneuron subgroup fate. Another group noted the presence of PV+ neurons in fused ventral-dorsal organoids, but these had immature morphologies and were not demonstrated to have FS electrophysiological properties, or selective innervation of the axon-initial segment or soma and proximal dendrites.⁴⁹

Even if adequate environments and maturational acceleration are provided, it remains unclear whether the conditions needed for the specification of PV-subclass hCIns from Nkx2.1-expressing progenitors in vitro have been achieved. Recent work showing that an aPKC inhibitor, applied during directed differentiation of mESCs into MGE-like progenitors, enhances SVZ-like divisions and the generation PV interneurons may be of particular value in this endeavor.⁸⁹ It will be interesting to see whether this approach, applied to directed differentiations of hESCs into CIns, enhances the production of PV+ subtypes. A better understanding of the transcriptional changes that occur during CIn development would be of considerable value in developing new protocols to generate specific subtypes in vitro. Such an approach was taken by Close et al, whereby single cell RNA-seq was used to profile hESC-derived interneurons across various stages of maturation.¹⁰⁰ Efforts to identify combinations of transcription factors selectively expressed in PV-fate committed, Nkx2.1+ MGE progenitors in mice^101,102 may lead to improved screens for conditions that generate human Nkx2.1+, PV-fate-committed progenitors.

GENERATING INTERNEURONS FROM DIRECT INDUCTION

In recent years, several groups have reported success in generating CIns through direct induction of stem cells (ESCs and induced pluripotent stem cell [iPSCs]; mouse and human) and fibroblasts through supraphysiological expression of lineage-determining genetic factors.^{57,59,103,104} Petros et al used the reverse tetracycline transactivator system to inducibly express Nkx2.1 in an Lhx6-GFP mESC line.¹⁰⁴ Not only did Nkx2.1 induction enhance the production of Lhx6⁺ CIn precursor-like cells, it was capable of doing so in the presence of Shh antagonists, thereby rendering Shh dispensable for the generation of MGE-derived interneurons.¹⁰⁴ A similar but more elegant approach to generate induced CIns from mESCs used a modular gain-of-function approach, whereby Nkx2.1 is expressed using the nestin promoter, together with a bicistronic tet-inducible construct to drive the expression of Dlx2,¹⁰⁵ and a Dlx5/6- promoter element-based GFP reporter of forebrain GABAergic fate.¹⁰⁶ Using this system, they found that Pou3f4 improved the general efficiency of mESC-derived interneuron differentiation while LMO3 augmented the production of PV subgroups.¹⁰³

More recently, groups have used similar approaches to derive hCIns from stem cells and fibroblasts.^57,59 Colasante et al showed that a combination of five transcription factors (Foxg1, Sox2, Ascl1, Dlx5, and Lhx6, with Dlx5 being dispensable) was capable of converting mouse and human fibroblasts, as well as human iPSCs, directly into CIns (henceforth termed iCIns).⁵⁷ This was achieved using lentiviral transfection of mouse embryonic fibroblasts from GAD67-GFP embryos, which were used as a reporter of GABAergic conversion.⁵⁷ Interestingly, ~90% of murine and human iCIns expressed PV, while only rare cells stained positive for SST.⁵⁷ Applied to human cells, they estimated the rate of conversion to a GABAergic identity to be approximately 30%. After 4 weeks in coculture, murine iCIns produced sustained, high-frequency (72 Hz) firing patterns and integrated into synaptic circuits.⁵⁷ However, after transplantation into adult mouse hippocampi, PV staining could no longer be detected.⁵⁷ It is thus not clear whether this phenotype is due to the Dlx5 transgene directly driving the PV promoter rather than a general PV-FS interneuron fate, and whether it is a stable, survivable system. Less than a year later, Sun et al reported that lentiviral transfection of a phospho mutant form of ASCL1 (previously shown to convert human fibroblasts into neurons more efficiently than wild-type ASCL1), together with LHX6, DLX2, and miR-9/9*-124 (previously shown to increase the efficiency of neuronal conversion from non-neuronal cells) was capable of converting ~85% of human stem cells into a GABAergic identity.⁵⁹ The majority of these cells expressed SST, together with a smaller proportion of calretinin (CR), calbindin (CB), and neuropeptide Y (NPY) subtypes. Very few or no cells expressed vasoactive intestinal protein (VIP), parvalbumin (PV), or reelin (RELN). Although no FS iCIns were identified, a variety of other AP patterns were seen, many of which were consistent with reports of mature SST firing patterns in vivo.^59,107 In similar studies, Wernig and colleagues demonstrated the efficacy of transient expression of Ascl1 and Dlx2 to generate induced hCIns.⁶⁴ Differentiation into neuronal morphologies could be accelerated by addition of MYT1L, without grossly affecting fate. Finally, Yuan et al combined dual-SMAD inhibition with induced expression of Lhx6 to generate hCIns that contained SST or PV.¹⁰⁸ Electrophysiological analyses showed impressive levels of subclass-related characteristics, but the extent to which both markers were expressed in the same cells was not reported, and 3 to 5 months after transplantation only 60 of over 4500 cells were PV+.

Taken together, these studies show that hCIns can be directly induced from a variety of mouse and human cell types using lineage-specific genetic factors. However, it bears mentioning that stem cell differentiation by direct induction with cell subclass-specifying transcription factors may have a general problem of leaving key areas of heterochromatin inaccessible, and hence leaving unexpressed some genes encoding proteins necessary for their fully natural function. To this end, partial directed differentiation by dual-SMAD inhibition, together with induced transcription factor expression, may be the more useful approach for generating the most natural-like neuronal subtypes.^108,109

FUTURE DIRECTIONS

Recent studies using single cell RNA-sequencing have provided evidence that transcription codes for specifying CIn subtypes already exist when the progenitors are around the time of cell cycle exit in the MGE.^101,102 Although those studies did not characterize such a code, progress in both mouse¹¹⁰ and in human stem cell systems⁶⁵ suggests that the detailed characterization of the genetic events leading to interneuron subtype fate determination will be established. We expect that future studies will capitalize on these advances to create protocols for the efficient generation of distinct CIn subgroups or subtypes that, together with approaches to intrinsically accelerate neuronal maturation, will be scalable for therapeutic use.

ACKNOWLEDGMENT

This work was supported by the following NIH Grants F31NS108622 (N.S.), 5F30MH105045 (D.J.T.), and R01MH066912 (S.A.A.).

CONFLICT OF INTEREST

The authors declared no potential conflicts of interest.

AUTHOR CONTRIBUTIONS

All authors contributed to the conceptualization and writing of this article.

DATA AVAILABILITY STATEMENT

Data sharing is not applicable to this article as no new data were created or analyzed in this study.

REFERENCES

Author notes