|Home | About | Journals | Submit | Contact Us | Français|
The ability of transcription factors to directly reprogram the identity of cell types is usually restricted and is defined by cellular context. We show here that through ectopic expression of single C. elegans transcription factors, the identity of mitotic germ cells can be directly converted into that of specific neuron types (glutamatergic, cholinergic or GABAergic). This reprogramming event requires the removal of the histone chaperone LIN-53/RbAp48, a component of several histone remodeling and modifying complexes, and this removal can be mimicked by chemical inhibition of histone deacetylases. Our findings illustrate the ability of germ cells to be directly converted into individual, terminally differentiated neuron types and demonstrate that a specific chromatin factor provides a barrier for cellular reprogramming.
There is much interest in developing methods to direct cell differentiation to a specific cell fate. However, transcription factors required to induce the identity of specific cell types in a multicellular organism are remarkably ineffective in imposing their fate-inducing activity when ectopically expressed, resulting in the generally accepted paradigm that transcription factors can exert their activities only in specific cellular contexts (1, 2). Classic examples for such context dependency are the cell-type restricted ability of MyoD to induce muscle cell features (3), the region-restricted ability of ectopically expressed Pax-6/eyeless to induce ectopic eyes (4) or, as a more recent example, the restricted ability of a cocktail of transcription factors to directly reprogram the identity of pancreatic cell types (5). Our understanding of the mechanistic basis of the context dependency of transcription factor activity is limited. Overcoming such context dependency would have major implications for a variety of different applications. For example, the generation of specific cell type through transcription factor-mediated reprogramming strategies may allow the establishment of in vitro disease models for basic research or the provision of source material for cellular replacement therapies.
We sought to establish a system in which we can study the mechanistic basis of the context-dependency of transcription factor activity. To this end, we employed a genetic approach in the nematode C.elegans, using the Zn finger transcription factor CHE-1, which is required to induce the identity of a specific class of gustatory neurons, called ASE neurons (6, 7). CHE-1 exerts this activity through binding directly to a cis-regulatory motif (termed the "ASE motif") present in many ASE-specific terminal differentiation genes, such as chemoreceptors, signaling proteins, neurotransmitter transporters, receptors and others (6). Like CHE-1, several other invertebrate and vertebrate transcription factors are known to also co-regulate many terminal features of differentiated neurons in such a manner and have been termed "terminal selectors" (8). To test whether CHE-1 is not only required but also sufficient to induce ASE fate, we ectopically expressed CHE-1 throughout the entire animal in either larval or adult stages, using an inducible heat-shock promoter. Such misexpression results in broad ectopic expression of an artificial reporter of CHE-1 transcription factor activity, which is composed of a multimerized ASE motif ("8× ASE motif reporter")(6)(Fig.1). This indicates that, in principle, CHE-1 can exert its biochemical activity of DNA binding and transcriptional activation without spatial or temporal constraints. In contrast, postembryonic ectopic expression of che-1 during larval or in adult stages is able to induce markers for terminal ASE fate (the gcy-5 chemoreceptor and the ceh-36 homeobox gene; see Table S1 for list of markers) only in a small number of head sensory neurons but nowhere else in the animal (Fig.1; Table S2). This result illustrates the context-dependency of the ASE-fate inducing activity of CHE-1.
To test the hypothesis that the mechanistic basis for such context-dependency lies in inhibitory, perhaps chromatin-based mechanisms that may prevent CHE-1 from reprogramming the identity of other cells, we screened through an RNAi library that targets all genes in the C.elegans genome with predicted roles in chromatin regulation, based on the presence of characteristic protein domains (9)(Table S3). We found that RNAi-mediated knock down of lin-53, the C.elegans ortholog of the phylogenetically conserved, WD40 domain-containing Retinoblastoma (Rb)-binding protein RbAp46/48 (10), permits ectopically expressed CHE-1 to induce the ASE neuronal fate markers gcy-5 and ceh-36 in a large number of normally non-neuronal cells in the midbody region of larval and adult animals (Fig.2). Up to 52% of animals (n=227) show this effect (Table S4) and this effect can be observed using distinct, non-overlapping dsRNA clones that target lin-53 and six out of six tested transgenic che-1 lines (see Supplementary Methods).
Closer examination of the animals reveals that gcy-5 and ceh-36 induction is observed in cells located within the germ line (Fig.2). Differential interference contrast (DIC) microscopy reveals that germ cells of che-1heat-shock; lin-53(RNAi) animals loose their characteristic, fried egg-shaped nuclear and nucleolar morphology and rather adopt a larger, speckled neuronal nuclear morphology (Fig.2). Moreover, these cells grow cellular extensions resembling axo/dendritic projections (Fig.2; Suppl.Movies 1–3). These neuron-like cells indeed originate from the germline as they are not observed following ectopic che-1 expression and lin-53 removal in a glp-4 mutant background (Fig.S1), in which no germline is formed (11). Genetic removal of sperm (fem-3lf mutants; (12)) or oocyte (fem-3gf mutants (13)) or prevention of entry into meiosis (glp-1gf mutants; (14)) does not affect neuron induction, whereas severe reduction of the mitotic pool (glp-1lf mutants (14)) does, suggesting that it is the mitotic germ cells that become neuron-like (Fig.S1). This is further supported by the position of the converted cells relative to mitotic and meiotic markers (Fig.S2) and the observation that second and third larval stage animals which contain mostly mitotic, but no meiotic cells show germ cell to neuron conversion upon ectopic che-1 expression and lin-53 knockdown (Table S4). The conversion to neuron-like cells is efficient and fast; up to around 60 germ cells (out of a total of around 200 mitotic germ cells) undergo neuronal induction (Fig.S3) and morphological changes and marker induction first occur 6 hours after che-1 induction (Table S5). For comparison, the induction of the 8× ASE motif, an indicator of CHE-1 transcriptional activity occurs 4 hours after che-1 induction in the gonad. The induction of the 8 × ASE motif in the gonad of wild-type animals, as well as antibody staining conducted in both wild-type and lin-53(RNAi) animals, rules out the possibility that lin-53(RNAi) merely results in germline derepression of the che-1 transgene (we also note that lin-53(RNAi) does not result in germline derepression of a previously described transgenic array, let-858::gfp, known to be derepressed after loss of several different chromatin factors (15)).
We assessed the nature of these che-1-induced, neuron-like cells with a number of fate markers. Through antibody staining of a marker that labels specific germ cell structures, the P-granules, we confirmed that this germ cell feature is indeed lost upon ectopic che-1 expression and lin-53 knockdown (Fig.S4). Moreover, the reprogrammed cells express six out of six tested pan-neuronal reporter genes: the rab-3/Rab3 and snb-1/synaptobrevin genes, which encode presynaptic proteins normally exclusively expressed in all cell of the nervous system, the pan-neuronal axonal regulators unc-33/CRMP and unc-119 and the pan-neuronal signaling factor rgef-1 (Fig.3A; see Table S1 for transgenic reporters). Antibody staining against endogenous proteins also shows ectopic expression of presynaptic synaptobrevin/SNB-1 and RIM/UNC-10 proteins; those proteins appear to cluster in presynaptic specialization in the axonal extensions of the reprogrammed germ cells (Fig.3B; Fig.S5). The reprogrammed cells also express a ciliated marker gene, osm-6, a member of the intraflagellar transport particle, normally expressed exclusively in all ciliated sensory neurons, including ASE (Fig.4A). GFP-tagged OSM-6 protein also appears to cluster in particles in the induced neurons (Fig.4A) suggesting that these cells express and cluster intraflagellar transport particles. All tested components of the gene battery that combinatorially define ASE identity are also expressed in the induced neurons. That is, aside from the above mentioned chemoreceptor gcy-5 and the transcription factor ceh-36 (shown again in Fig.4A), the putative chemoreceptor gcy-7, which is normally exclusively expressed in ASE, and the vesicular glutamate transporter eat-4, normally expressed in ASE and a restricted number of additional head ganglia neurons, are also expressed in the reprogrammed germ cells (Fig.4A). Induction of eat-4 demonstrates that the reprogrammed cells are glutamatergic. The reprogrammed ASE-like cells do not express a battery of markers that are normally expressed in other neuron types, such as dopaminergic, serotonergic, cholinergic and GABAergic markers, among others (Fig.4B). This argues that the reprogrammed cells are not merely generic and/or mis-specified neurons but closely resemble normally differentiated ASE neurons. Taken together, animals ectopically expressing che-1 and lacking lin-53 not only contain their normal set of 2 ASE gustatory neurons in the head, but contain a gonad filled with dozens of ASE-like neurons.
Removal of RNA-binding gene regulatory factors has been shown to result in the formation of teratomas, i.e. cells of various origin and types, in the germline (16). The effects that we observe upon loss of lin-53 are not a reflection of teratoma formation, since we observe no expression of various neuronal fate markers other than those of ASE fate in the gonad of che-1heat-shock; lin-53(RNAi) animals. However, the very same neuronal markers (GABA neuron marker, cholinergic neuron marker) are expressed in the gonad of animals in which the translation regulator gld-1, a previously described repressor of teratoma formation (16), is knocked down (Fig.S6). Therefore, removal of lin-53 does not by itself trigger alternative developmental programs but primes germ cells to be responsive to a neuronal fate inducer like che-1.
To test whether lin-53 removal also permits the conversion of germ cells into other neurons types, we tested two other terminal selector genes (8), the phylogenetically conserved Pitx-type homeobox gene unc-30, a terminal selector required for the generation of GABAergic motor neurons in the ventral nerve cord (17, 18) and the EBF-like transcription factor unc-3, required for the generation of two types (A- and B-type) of cholinergic motor neurons in the ventral nerve cord (19, 20). When ectopically mis-expressed, neither unc-30 nor unc-3 is able to induce GABAergic or cholinergic neuron fate in the germ line, respectively. However, upon removal of lin-53, heat-shock induction of either unc-30 or unc-3 results, like che-1 induction, in germ cells losing their characteristic morphology and instead adopting neuron-like nuclear morphology and growing axonal projections (Fig.5A,B). In the case of unc-3, the ectopic neurons now express a marker characteristic of cholinergic A/B-type ventral cord motor neurons (acr-2; Fig.5B), whereas ectopic expression of unc-30 results in the expression of the GABAergic marker unc-47 (Fig.5A). In neither case do we observe any ASE marker expression to be induced (>100 animals scored); also, neither cholinergic nor GABAergic markers are induced by che-1 (Fig.4B). We conclude that upon loss of lin-53, germ cells acquire the ability to be reprogrammed into distinct neuron types through the activity of neuron-type specific terminal selector transcription factors.
Does lin-53 removal prime germ cells to only respond to factors that induce neuronal fates or can they now respond to other factors as well? To address this question, we ectopically expressed a selector gene, the C. elegans MyoD homolog hlh-1 that was previously shown to be able to ectopically induce muscle fate in early embryos (21). We find that hlh-1 is unable to convert the germ cells of lin-53(RNAi) animals into muscle cells (Fig.S7). These negative results need to be cautiously interpreted, but may represent a first hint toward a target selectivity of lin-53. That is, lin-53 may only restrict the developmental potential specifically toward a neuronal developmental program while other factors may serve to prevent the induction of other, non-neuronal differentiation programs.
LIN-53, which is ubiquitously expressed (10), is one of several, phylogenetically conserved histone chaperones which are thought to assist in the recruitment of various distinct types of histone modifiers or remodelers, including histone methyltransferases, histone acetylases and histone deacetylases to histone H3 and histone H4 (22, 23). LIN-53 orthologs in various species have been found to be integral components of at least 6 different protein complexes, each displaying diverse biochemical and biological roles - the NURD and NURF nucleosome remodeling complexes, the CAF-1 chromatin assembly factor complex, the Sin3A transcriptional repressor complex, the PRC2 histone methyltransferase complex and the HAT1 histone acetyltransferase complex (22, 23) (Table S6). RNAi of representative members of each of these complexes either did not phenocopy the lin-53(RNAi)-induced reprogramming ability of ectopic CHE-1 expression or could not be interpreted due to early lethality induced by RNAi (Table S6). However, since two of the LIN-53-associated complexes (NURD and Sin3a) each contain at least two histone deacetylase (HDAC) components and since the reprogramming role of lin-53 may involve several complexes, we sought to broadly inhibit HDAC function by using two distinct chemical inhibitors, valproic acid and trichostatin A (24). At sub-lethal doses, animals treated with either drug survived and permitted heat-shock induced che-1 to induce ASE fate in the germ line (and no other cell type) even in the presence of functional lin-53 (Fig.6). Even though these drug effects may be unrelated to normal lin-53 function, these results nevertheless provide a strong indication that histone modifications are key players in restricting the ability of a transcription factor to reprogram cellular identity.
Our finding that the removal of a single chromatin factor, together with the induction of single transcription factors, can produce distinct and specific neuron types in a heterologous cellular context is a testament to the simplicity of programs that control neuronal differentiation. The main role of complex, multi-stage neuronal developmental programs may be little more than to activate a terminal regulatory routine, that is, a single or a combination of terminal selectors which then induce a specific differentiated state. This notion may apply to more complex systems as well, since recent work has shown that it takes as little as two transcription factors to drive a differentiated fibroblast toward a specific neuronal fate (25). Our findings are also a testament to the pluripotency of germ cells (26). This totipotency is normally kept in check by a variety of transcriptional and post-transcriptional mechanisms (26), but is unleashed either spontaneously in pathological situations (germ cell tumors)(27) or upon culturing germ stem cells under specific conditions, which transform these cells into cells indistinguishable from pluripotent embryonic stem cells (28). We have shown here that the ability of germ cells to be directly reprogrammed into neurons can be unleashed, even in the adult animal, through removal of a single gene that we speculate to be involved in rendering neuronal differentiation genes inaccessible to transcriptional induction by, for example, contributing to the formation of facultative, i.e. conditional and developmentally regulated heterochromatin.
Seen in a broader context, our results indicate that the reprogramming of cellular identity may critically depend not just on providing the correct transcription factor(s) that induce a specific fate, but also on the removal of inhibitory mechanisms that restrict transcription factor activity. We anticipate that the disablement of such inhibitory mechanisms may provide an efficient strategy to ectopically generate neuron types in other organisms or perhaps even in cell culture using isolated germ cells.
This study was funded by the NIH (R01NS039996-05; R01NS050266-03). O.H. is an Investigator of the HHMI and B. T. is a Francis Goelet Research Scientist. We thank T. Schedl, M. Krause, J. McGhee, P. Sengupta, Y. B. Qi, Y. Jin, M. Nonet, S. Strome and members of the Hobert lab for reagents, I. Greenwald and members of the Hobert lab for comments on the manuscript and Q. Chen for expert assistance in strain generation.
Material and Methods are in the Supplementary Online Material