BIOLOGICAL OVERVIEW

Sibling cell fate in the Drosophila adult external sense organ lineage is specified by prospero function, which is regulated by Numb and Notch.

Specification of cell fate in the adult sensory organs is known to be dependent on intrinsic and extrinsic signals. Prospero (Pros) acts as an intrinsic signal for the specification of cell fates within the mechanosensory lineage. The sensory organ precursors divide to give rise to two secondary progenitors: PIIa and PIIb. Pros is expressed in PIIb, which gives rise to the neuron and thecogen cells. Pros expression was first detected among the secondary progenitors in the nucleus of the more anteriorly located cell (PIIb). Interestingly, this was the first cell to divide in all cases examined. Prior to cell division, Pros becomes uniformly distributed on the cortical membrane and throughout the cytosol. Unlike the findings in the central nervous system, Pros is not asymmetrically localized in cells at any stage of the lineage. Asymmetric crescents of Pros immunoreactivity are not observed even after blocking mitosis with colcemid. These data suggest that Pros is expressed first in the nucleus and then generally in the cytosol of PIIb (Reddy, 1999).

Loss of Notch function generated using either a conditional mutant allele or by misexpressing Numb protein results in the ectopic expression of Pros in PIIa. This observation is consistent with a PIIa-to-PIIb conversion by Notch loss of function or nb gain of function. It is not clear whether the apparent negative regulation of Pros by Notch is a direct effect or merely reflects the altered fate of the cells that is caused by other molecular factors. Pros misexpression is sufficient for the transformation of PIIa to PIIb fate (Reddy, 1999).

Pros is not asymmetrically localized in PIIb; following division of this cell, the protein is detected in both the progeny. This is strikingly different from findings in the CNS where Pros, together with Numb and Miranda, is localized asymmetrically in the neuroblasts and inherited by the GMC. Following division of PIIb, Pros is inherited by nuclei of both progeny. Expression in one of the siblings decays rapidly and this cell differentiates as the neuron. Pros immunoreactivity is sustained in the thecogen cell possibly due to de novo synthesis. The requirement for pros function in PIIb precludes analysis of its later role after division. In experiments where Pros was misexpressed in all four cells of the sensory organ, a conversion of external to internal cells is observed, consistent with a PIIa-to-PIIb transformation. Neuronal cells form normally despite the fact that they express the thecogen cell marker. Similarly, Notch loss of function and Numb gain of function results in a conversion of all four cells of the lineage to neurons. Pros is expressed in all these cells under these conditions. These observations together demonstrate that pros activity is not sufficient for identity of the thecogen cell and that neuronal cell differentiation can occur normally despite Pros expression. The elucidation of pros function in the thecogen cell awaits the availability of a hypomorphic mutant allele that can allow loss of function in the thecogen cell without affecting the secondary progenitors. Pros has been shown to be expressed in several CNS- as well as PNS-associated glial cells and it is possible that it plays a role in the later differentiation and/or function of these cells (Reddy, 1999).

In the normal mechanosensory lineage, Notch is involved in the binary choice between thecogen and neuron. In this lineage Notch signaling is experienced by the cell that ultimately becomes the thecogen cell. This is distinct from the scenario at the secondary progenitor stage, where the cell that experiences Notch signaling does not express Pros. This means that, unlike in PIIb, Notch signaling does not downregulate Pros in the thecogen cell. There are several possible explanations for this finding. One possibility is that Pros protein is merely partitioned to the daughters of the PIIb after division. It therefore serves no role in the binary choice of thecogen versus neuron but is synthesized de novo after these cells have acquired their identity. At this later time point, the thecogen cell is no longer experiencing the Notch signal. Another possibility lies in the different effector mechanisms utilized for Notch activity. In three instances (during lateral signaling, PIIa-PIIb choice, as well as in the PIIa lineage) Notch activation results in release of Su(H) from its binding site on the cytoplasmic domain of Notch and its translocation to the nucleus. Su(H) protein can be detected in the nucleus of the socket cell; however, the role of Su(H) cannot be seen in the PIIb lineage. Extra copies of Su(H) do not produce thecogen-to-neuron transformations, suggesting that Notch signaling in the thecogen/neuron choice occurs by a Su(H)-independent mechanism. Thus the regulation of Pros expression could be mediated by Notch through a Su(H)-independent event (Reddy, 1999 and references).

Overlapping activators and repressors delimit transcriptional response to receptor tyrosine kinase signals in the Drosophila eye

Regulated transcription of the prospero gene in the Drosophila eye provides a model for how gene expression is specifically controlled by signals from receptor tyrosine kinases. prospero is controlled by signals from the Egfr receptor and the Sevenless receptor. A direct link is established between Egfr activation of a transcription enhancer in prospero and binding of two transcription factors that are targets of Egfr signaling. Binding of the cell-specific Lozenge protein is also required for activation, and overlapping Lozenge protein distribution and Egfr signaling establishes expression in a subset of equivalent cells competent to respond to Sevenless. Sevenless activates prospero independent of the enhancer and involves targeted degradation of Tramtrack, a transcription repressor (Xu, 2000).

Thus, Egfr signaling is required to activate pros expression in the R7 equivalence group but is restricted from activating pros expression in other cells by the distribution of the transcription factor Lz. The transcriptional effectors of the Egfr pathway combinatorially interact with Lz at an eye-specific pros enhancer to restrict enhancer activity to the R7 equivalence group. It is suggested that this mechanism is a primary means by which pros transcription is restricted to the R7 equivalence group. This combinatorial mechanism supposes that Egfr signaling inactivates Yan and activates Pnt, but modification of these transcription factors is not sufficient to activate the enhancer. Lz is also required to activate the enhancer. The only cells that contain Lz, activated Pnt, and inactivated Yan are R1, R6, R7, and cone cells. Thus, the enhancer is activated in a subset of Egfr-responsive cells. A similar combinatorial mechanism regulates shaven expression in cone cells. shaven expression requires both Lz and Egfr-induced regulation of Yan and Pnt. However, Notch signaling through Su(H) is also required for shaven expression in cone cells. This third input may explain why shaven has a more restricted expression pattern than pros, given that cone cells receive a robust Notch signal (Flores, 2000). In muscle and cardiac cells, RTK signaling is similarly integrated with other signal inputs and tissue-restricted transcription factors to regulate enhancer activity of the even skipped gene (Halfon, 2000). Thus, differential expression of genes in response to an RTK/Ras signal appears to be controlled by each gene's capacity to bind and be regulated by different combinations of transcription factors (Xu, 2000).

A model is presented for the regulatory inputs into prospero. (1) In eye progenitor cells, the presence of Yan represses pros transcription through its binding to the enhancer and competitively excluding Pnt from binding to the same sites. (2) Lz begins to be produced in progenitor cells after the first wave of photoreceptor differentiation. However, Lz alone cannot activate the enhancer in progenitor cells that have not received a Spitz signal. (3) When a progenitor cell receives a Spitz signal, Egfr is activated. This inactivates Yan, allowing activated Pnt to bind to the enhancer. At the morphogenetic furrow, the enhancer is inactive despite Egfr-stimulated cells containing inactive Yan and active Pnt since progenitor cells in this region do not contain Lz, which is also required for enhancer activity. Hence, photoreceptors R2, R3, R4, R5, and R8 do not express pros. It is only in cells that receive a Spitz signal and contain Lz that the combination of Lz and Pnt bound to the enhancer activate the enhancer. (4) Ttk88 reduces the level of pros transcription through a mechanism independent of the eye enhancer. This repression may not be strong enough to block the eye enhancer in the R7 equivalence group but acts to limit its level of transcription. (5) When a progenitor cell receives both a Spitz and Boss signal, stronger or longer signal transduction induces Ttk88 inactivation. This Egf represses pros transcription and leads to a specific increase of Pros in R7 cells (Xu, 2000).

The ETS factors Yan and Pnt have been implicated as substrates for activated MAPK, whose activities are modified upon phosphorylation. Both Yan and Pnt bind to the same sites in the pros eye enhancer except for one site that is Yan-specific. Their effects on enhancer activity are antagonistic; Yan represses while Pnt activates. One model is that Yan represses transcription by outcompeting Pnt for their binding sites, thereby preventing Pnt from activating transcription. This model is attractive since it has been found that Yan has a 100-fold greater affinity than Pnt for ETS factor binding sites in vitro. If this difference between purified fusion proteins in vitro is extrapolated to the fly eye, it would explain how Yan can outcompete Pnt and repress transcription. Results from mutagenesis of the binding sites is also consistent with this model. Mutated binding sites cause the enhancer to be inactive, which is the result predicted if Yan merely prevents Pnt from interacting with those sites. If Yan were actively repressing transcription in a manner dependent upon binding, then mutated binding sites would cause derepression and ectopic expression. Although a model where the binding sites are obligatory for both active repression by Yan and activation by Pnt cannot be excluded, the competitive binding model is the simplest one consistent with these data (Xu, 2000).

From these data it is proposed that two RTKs, Egfr and Sev, regulate pros by activating the Ras1 intracellular pathway in R7 cells, but these RTKs regulate pros differentially. Egfr regulates pros by modifying Yan and Pnt, which act directly through the eye-specific enhancer. The Egfr signal in R7 cells appears to occur before Sev, and it sufficiently inactivates Yan and activates Pnt to switch on the enhancer before the Sev signal. This sufficiency is demonstrated in sev mutants where enhancer activity in R7 cells is no different from wild-type. In contrast, the Sev signal in R7 cells is not sufficient to switch on the enhancer in the absence of the Egfr signal since the enhancer is inactive in Egfr mutant R7 cells (Xu, 2000).

Sev regulates pros in R7 cells by inactivating Ttk88, which otherwise represses pros through sequence elements distinct from the eye-specific enhancer. This is demonstrated by finding that overproduced Ttk88 blocks Sev from activating pros, and Sev can regulate the eye-specific enhancer only if it is linked with Ttk88 binding sites. It is not clear if the Sev signal is sufficient to inactivate Ttk88 without an Egfr input since the assay for Ttk88 activity is a reporter gene that includes the eye-specific enhancer. It is quite possible that Ttk88 inactivation in R7 cells requires both Sev and Egfr signals, since Ebi, acting downstream of Egfr to promote Ttk88 degradation, and Phyl/Sina, acting downstream of Sev, are both required to inactivate Ttk88 in R7 cells (Xu, 2000).

How do these RTKs selectively regulate particular transcription factors and thereby regulate different aspects of pros transcription? The most attractive model is that RTK selection reflects the timing or intensity of each signal. If it is timing, then there must be a time period of competence during which a factor is sensitive to any RTK signal, and the time period is different for each factor. Alternatively, the intensity of a signal may dictate which transcription factor activities are sensitive. For example, Yan and Pnt activities may be insensitive to signal strength that is less than or equal to the level achieved by Sev but not Egfr within R7 cells. Ttk88 activity may be insensitive to signal strength that is less than or equal to the level achieved by Sev or Egfr alone but not the combination of the two within R7 cells. Signal 'strength' may be determined by the level of Ras pathway activity or the length of time that the Ras pathway is active. Sensitivity of transcription factors might be set either by the affinities of these factors for binding sites in a gene such as pros, or by the ability of factors to be substrates for RTK-stimulated modification. Given that Yan and Pnt are modified by a very different mechanism from Ttk88, substrate sensitivity is a possible determinant. In summary, RTK signals may provide specificity to gene regulation based on quantitative variation in which threshold transcription responses are set by transcription factors that have different sensitivities to RTK signal strength (Xu, 2000).

Prospero acts as a binary switch between self-renewal and differentiation in Drosophila neural stem cells

Stem cells have the remarkable ability to give rise to both self-renewing and differentiating daughter cells. Drosophila neural stem cells segregate cell-fate determinants from the self-renewing cell to the differentiating daughter at each division. This study shows that one such determinant, the homeodomain transcription factor Prospero, regulates the choice between stem cell self-renewal and differentiation. The in vivo targets of Prospero have been identified throughout the entire genome. Prospero represses genes required for self-renewal, such as stem cell fate genes and cell-cycle genes. Surprisingly, Prospero is also required to activate genes for terminal differentiation. In the absence of Prospero, differentiating daughters revert to a stem cell-like fate: they express markers of self-renewal, exhibit increased proliferation, and fail to differentiate. These results define a blueprint for the transition from stem cell self-renewal to terminal differentiation (Choksi, 2006).

To identify sites within the Drosophila genome to which Prospero binds, use was made of an in vivo binding-site profiling technique, DamID. DamID is an established method of determining the binding sites of DNA- or chromatin-associated proteins. Target sites identified by DamID have been shown to match targets identified by chromatin immunoprecipitation (ChIP). DamID enables binding sites to be tagged in vivo and later identified on DNA microarrays. In brief, the DNA or chromatin-binding protein of interest is fused to an Escherichia coli adenine methyltransferase (Dam), and the fusion protein is expressed in vivo. The DNA-binding protein targets the fusion protein to its native binding sites, and the Dam methylates local adenine residues in the sequence GATC. The sequences near the protein-DNA interaction site are thereby marked with a unique methylation tag, over approximately 2-5 kilobase pairs (kb) from the binding site. The tagged sequences can be isolated after digestion with a methylation-sensitive restriction enzyme, such as DpnI (Choksi, 2006).

Dam was fused to the N terminus of Prospero, and transgenic flies were generated. The fusion protein is expressed from the uninduced minimal Hsp70 promoter of the UAS vector, pUAST, as high levels of expression of Dam can result in extensive nonspecific methylation and cell death. As a control for nonspecific Dam activity, animals expressing Dam alone were generated. To assess the sites to which Prospero binds during neurogenesis, genomic DNA was extracted from stage 10-11 embryos, approximately 4-7 hr after egg laying (AEL), expressing either the Dam-Prospero fusion protein or the Dam protein alone. The DNA was digested with DpnI and amplified by PCR. DNA from Dam-Prospero embryos was labeled with Cy3, and control DNA with Cy5. The samples were then cohybridized to genomic microarrays. Microarrays were designed that tile the entire euchromatic Drosophila genome. A 60 base oligonucleotide was printed for approximately every 300 bp of genomic DNA, resulting in roughly 375,000 probes on a single array (Choksi, 2006).

Log-transformed ratios from four biological replicates (two standard dye configurations plus two swapped dye configurations) were normalized and averaged. Regions of the genome with a greater than 1.4-fold log ratio (corresponding to approximately a 2.6-fold enrichment) of Dam-Prospero to the control over a minimum of four adjacent genomic probes were identified as in vivo Prospero binding sites. Using these parameters, a total of 1,602 in vivo Prospero binding sites were identified in the Drosophila genome. This work demonstrates that it is possible to map in vivo binding sites across the whole genome of a multicellular organism (Choksi, 2006).

Prospero is known to regulate the differentiation of photoreceptors in the adult eye, and recently sites have been characterized to which Prospero can bind upstream of two Rhodopsin genes, Rh5 and Rh6. A variant of the Prospero consensus sequence is found four times upstream of Rh5 and four times upstream of Rh6. Prospero was shown to bind this sequence in vitro, by band shift assay, and also by a 1-hybrid interaction assay in yeast. In addition, deletion analysis demonstrated that the consensus sequence is required for the Pros-DNA interaction both in vivo and in vitro. It was found that 67% of in vivo binding sites contain at least one Prospero binding motif. Combining in vivo binding-site data with searches for the Prospero consensus sequence reveals 1,066 distinct sites within the Drosophila genome to which Prospero binds during embryogenesis (Choksi, 2006).

A total of 730 genes have one or more of the 1,066 Prospero binding sites located within 1 kb of their transcription unit. Statistical analyses to determine GO annotation enrichment on the members of the gene list that had some associated annotation (519) was performed by using a web-based set of tools, GOToolbox. Using Biological Process (GO: 0008150) as the broadest classification, a list was generated of overrepresented classes of genes (Choksi, 2006).

The three most significant classes of genes enriched in the list of putative Prospero targets are Cell Fate Commitment, Nervous System Development, and Regulation of Transcription. Utilizing GO annotation, it was found that nearly 41% of all annotated neuroblast fate genes (11 of 27) are located near Prospero binding sites and that approximately 9% of known cell-cycle genes are near Prospero binding sites. These include the neuroblast genes achaete (ac), scute (sc), asense (ase), aPKC, and mira and the cell-cycle regulators stg and CycE. In addition, it was found that the Drosophila homolog of the mammalian B lymphoma Mo-MLV insertion region 1 (Bmi-1) gene, Posterior sex combs, is located near a Prospero binding site. Bmi-1 is a transcription factor that has been shown to regulate the self-renewal of vertebrate hematopoetic stem cells. It is concluded that Prospero is likely to regulate neuroblast identity and self-renewal genes as well as cell-cycle genes directly, repressing their expression in the GMC (Choksi, 2006).

Prospero enters the nucleus of GMCs, and its expression is maintained in glial cells but not in neurons . Therefore the list of targets was searched for genes annotated as glial development genes. Prospero binds near 45% of genes involved in gliogenesis. Among the glial genes, it was found that the master regulator of glial development, glial cells missing (gcm), and gilgamesh (gish), a gene involved in glial cell migration, are both near Prospero binding sites and are likely directly activated by Prospero in glia (Choksi, 2006).

In summary, Prospero binds near, and is likely to regulate directly, genes required for the self-renewing neural stem cell fate such as cell-cycle genes. It was also found that Prospero binds near most of the temporal cascade genes: hb, Kruppel (Kr), nubbin (nub/pdm1), and grainyhead (grh) and to genes required for glial cell fate. The in vivo binding-site mapping experiments are supportive of a role for Prospero in regulating the fate of Drosophila neural precursors by directly controlling their mitotic potential and capacity to self-renew (Choksi, 2006).

The Drosophila ventral nerve cord develops in layers, in a manner analogous to the mammalian cortex. The deepest (most dorsal) layer of the VNC comprises the mature neurons, while the superficial layer (most ventral) is made up of the mitotically active, self-renewing neuroblasts. Neuroblast cell-fate genes and cell-cycle genes are normally expressed only in the most ventral cells, while Prospero is found in the nucleus of the more dorsally lying GMCs. If in GMCs, Prospero normally acts to repress neuroblast cell-fate genes and cell-cycle genes, then in a prospero mutant, expression of those genes should expand dorsally. Conversely, ectopically expressed Prospero should repress gene expression in the neuroblast layer.

The neuroblast genes mira, ase, and insc and the cell cycle genes CycE and stg show little or no expression in differentiated cells of wild-type stage 14 nerve cords. Expression of these neuroblast-specific genes was examined in the differentiated cells layer of prospero embryos and it was found that they are derepressed throughout the nerve cord of mutant embryos. mira, ase, insc, CycE, and stg are all ectopically expressed deep into the normally differentiated cell layer of the VNC. To check whether Prospero is sufficient to repress these genes, Prospero was expressed with the sca-GAL4 driver, forcing Prospero into the nucleus of neuroblasts. Prospero expression is sufficient to repress mira, ase, insc, CycE, and stg in the undifferentiated cell layer of the VNC. These data, combined with the Prospero binding-site data, demonstrate that Prospero is both necessary and sufficient to directly repress neuroblast genes and cell-cycle genes in differentiated cells. This direct repression of gene expression is one mechanism by which Prospero initiates the differentiation of neural stem cells (Choksi, 2006).

Having shown that Prospero directly represses genes required for neural stem cell fate, it was asked whether Prospero also directly activates GMC-specific genes. Alternatively, Prospero might regulate a second tier of transcription factors, which are themselves responsible for the GMC fate. Of the few previously characterized GMC genes, it was found that Prospero binds to eve and fushi-tarazu (ftz). In the list of targets several more GMC genes were expected to be found, but not genes involved in neuronal differentiation, since Prospero is not expressed in neurons. Surprisingly, however, it was foudn 18.8% of neuronal differentiation genes are located near Prospero binding sites (Choksi, 2006).

To determine Prospero's role in regulating these neuronal differentiation genes, in situ hybridization was carried out on prospero mutant embryos. Prospero was found to be necessary for the expression of a subset of differentiation genes, such as the adhesion molecules FasciclinI (FasI) and FasciclinII (FasII), which have roles in axon guidance and/or fasciculation. Netrin-B, a secreted protein that guides axon outgrowth, and Encore, a negative regulator of mitosis, also both require Prospero for proper expression. Therefore, in addition to directly repressing genes required for neural stem cell self-renewal, Prospero binds and activates genes that direct differentiation. These data suggest that Prospero is a binary switch between the neural stem cell fate and the terminally differentiated neuronal fate (Choksi, 2006).

To test to what extent Prospero regulates the genes to which it binds, genome-wide expression profiling was carried out on wild-type and prospero mutant embryos. While the DamID approach identifies Prospero targets in all tissues of the embryo, in this instance genes regulated by Prospero were assayed in the developing central nervous system. Small groups of neural stem cells and their progeny (on the order of 100 cells) were isolated from the ventral nerve cords of living late stage 12 embryos with a glass capillary. The cells were expelled into lysis buffer, and cDNA libraries generated by reverse transcription and PCR amplification. cDNA libraries prepared from neural cells from six wild-type and six prospero null mutant embryos were hybridized to full genome oligonucleotide microarrays, together with a common reference sample. Wild-type and prospero mutant cells were compared indirectly through the common reference (Choksi, 2006).

In the group of Prospero target genes that contain a Prospero consensus sequence within 1 kb of the transcription unit, 91 show reproducible differences in gene expression in prospero mutants. Seventy-nine percent of these genes (72) exhibit at least a 2-fold change in levels of expression. Many of the known genes involved in neuroblast fate determination and cell-cycle regulation (e.g., asense, deadpan, miranda, inscuteable, CyclinE, and string) show increased levels in a prospero mutant background, consistent with their being repressed by Prospero. Genes to which Prospero binds, but which do not contain an obvious consensus sequence, are also regulated by Prospero: CyclinA and Bazooka show elevated mRNA levels in the absence of Prospero, as does Staufen, which encodes a dsRNA binding protein that binds to both Miranda and to prospero mRNA (Choksi, 2006).

Expression of genes required for neuronal differentiation is decreased in the prospero mutant cells, consistent with Prospero being required for their transcription. These include zfh1 and Lim1, which specify neuronal subtypes, and FasI and FasII, which regulate axon fasciculation and path finding (Choksi, 2006).

The stem cell-like division of neuroblasts generates two daughters: a self-renewing neuroblast and a differentiating GMC. Prospero represses stem cell self-renewal genes and activates differentiation genes in the newly born GMC. In the absence of prospero, therefore, neuroblasts should give rise to two self-renewing neuroblast-like cells (Choksi, 2006).

The division pattern of individual neuroblasts was studied in prospero mutant embryos by labeling with the lipophilic dye, DiI. Individual cells were labeled at stage 6, and the embryos allowed to develop until stage 17. S1 or S2 neuroblasts were examined, as determined by their time of delamination. Wild-type neuroblasts generate between 2 and 32 cells, producing an average of 16.2 cells. Most of the clones exhibit extensive axonal outgrowth. In contrast, prospero mutant neuroblasts generate between 8 and 51 cells, producing an average of 31.8 cells. Moreover, prospero mutant neural clones exhibit few if any projections, and the cells are smaller in size. Thus, prospero mutant neuroblasts produce much larger clones of cells with no axonal projections, suggesting that neural cells in prospero mutants undergo extra divisions and fail to differentiate (Choksi, 2006).

Recently it was shown, in the larval brain, that clones of cells lacking Prospero or Brat undergo extensive cell division to generate undifferentiated tumors. Given that Prospero is nuclear in the GMC but not in neuroblasts, the expanded neuroblast clones in prospero mutant embryos might arise from the overproliferation of GMCs: the GMCs lacking Prospero may divide like neuroblasts in a self-renewing manner. It is also possible, however, that neuroblasts divide more frequently in prospero mutant embryos, giving rise to supernumerary GMCs that each divide only once. To distinguish between these two possibilities, the division pattern of individual GMCs was followed in prospero mutant embryos (Choksi, 2006).

S1 or S2 neuroblasts were labeled with DiI as before. After the first cell division of each neuroblast, the neuroblast was mechanically ablated, leaving its first-born GMC. All further labeled progeny derive, therefore, from the GMC. Embryos were allowed to develop until stage 17, at which time the number of cells generated by a single GMC was determined (Choksi, 2006).

To determine whether mutant GMCs are transformed to a stem cell-like state, stage 14 embryos were stained for the three neuroblast markers: Miranda (Mira), Worniu (Wor), and Deadpan (Dpn). In wild-type embryos at stage 14, the most dorsal layer of cells in the VNC consists mostly of differentiated neurons. As a result, few or none of the cells in this layer express markers of self-renewal. Mira-, Wor-, and Dpn- expressing cells are found on the midline only or in lateral neuroblasts of the differentiated cell layer of wild-type nerve cords. In contrast, a majority of cells in the differentiated cell layer of stage 14 prospero mutant embryos express all three markers: Mira is found cortically localized in most cells of the dorsal layer of prospero nerve cords; Wor is nuclear in most cells of mutant VNCs; Dpn is ectopically expressed throughout the nerve cord of prospero mutants (Choksi, 2006).

Expression of neuroblast markers in the ventral-most layer of the nerve cord (the neuroblast layer), to exclude the possibility that a general disorganization of cells within the VNC contributes to the increased number of Mira-, Wor-, and Dpn-positive cells in the dorsal layer. The number of neuroblasts in a prospero mutant embryo is normal in stage 14 embryos, as assayed by Wor, Dpn, and Mira expression. Thus, the increased expression of neuroblast markers in prospero mutants is the result of an increase in the total number of cells expressing these markers in the differentiated cell layer. It is concluded that prospero mutant neuroblasts divide to give two stem cell-like daughters. GMCs, which would normally terminate cell division and differentiate, are transformed into self-renewing neural stem cells that generate undifferentiated clones or tumors (Choksi, 2006).

Therefore, Prospero directly represses the transcription of many neuroblast genes and binds near most of the temporal cascade genes: hb, Kruppel (Kr), nubbin (nub/pdm1), and grainyhead (grh), which regulate the timing of cell-fate specification in neuroblast progeny. Prospero maintains hb expression in the GMC, and it has been suggested that this is through regulation of another gene, seven-up (svp). Prospero not only regulates svp expression directly but also maintains hb expression directly. In addition, Prospero maintains Kr expression and is likely to act in a similar fashion on other genes of the temporal cascade. Intriguingly, Prospero regulates several of the genes that direct asymmetric neuroblast division (baz, mira, insc, aPKC). aPKC has recently been shown to be involved in maintaining the self-renewing state of neuroblasts (Choksi, 2006).

Prospero initiates the expression of genes necessary for differentiation. This is particularly surprising since prospero is transcribed only in neuroblasts, not in GMCs or neurons. Prospero mRNA and protein are then segregated to the GMC. Prospero binds near eve and ftz, which have been shown to be downstream of Prospero, as well as to genes required for terminal neuronal differentiation, including the neural-cell-adhesion molecules FasI and FasII. Prospero protein is present in GMCs and not neurons, suggesting that Prospero initiates activation of neuronal genes in the GMC. The GMC may be a transition state between the neural stem cell and the differentiated neuron, providing a window during which Prospero functions to repress stem cell-specific genes and activate genes required for differentiation. There may be few, or no, genes exclusively expressed in GMCs (Choksi, 2006).

Prospero acts in a context-dependent manner, functioning alternately to repress or activate transcription. This implies that there are cofactors and/or chromatin remodeling factors that modulate Prospero's activity. In support of this, although Prospero is necessary and sufficient to repress neuroblast genes, forcing Prospero into the nuclei of neuroblasts is not sufficient to activate all of the differentiation genes to which it binds (Choksi, 2006).

Neuroblasts decrease in size with each division throughout embryogenesis. By the end of embryogenesis, they are similar in size to neurons. A subset of these embryonic neuroblasts becomes quiescent and is reactivated during larval life: they enlarge and resume stem cell divisions to generate the adult nervous system. Neuroblasts in prospero mutant embryos divide to produce two self-renewing daughters but still divide asymmetrically with respect to size, producing a large apical neuroblast and a smaller basal neuroblast-like cell. The daughter may be too small to undergo more than three additional rounds of division during embryogenesis. prospero mutant cells eventually stop dividing, and a small number occasionally differentiate. This suggests that there is an inherent size limitation on cell division. The segregation of Brat, or an additional cell fate determinant, to the daughter cell may also limit the potential of the prospero mutant cells to keep dividing (Choksi, 2006).

The Prox family of atypical homeodomain transcription factors has been implicated in initiating the differentiation of progenitor cells in contexts as varied as the vertebrate retina, forebrain, and lymphatic system. Prospero/Prox generally regulates the transition from a multipotent, mitotically active precursor to a differentiated, postmitotic cell. In most contexts, Prox1 acts in a similar fashion to Drosophila Prospero: to stop division and initiate differentiation (Choksi, 2006).

It is proposed that Prospero/Prox is a master regulator of the differentiation of progenitor cells. Many of the vertebrate homologs of the Drosophila Prospero targets identified in this study may also be targets of Prox1 in other developmental contexts. Prospero directly regulates several genes required for cell-cycle progression, and it is possible that Prox1 will regulate a similar set of cell-cycle genes during, for example, vertebrate retinal development. In addition, numerous Prospero target genes have been identified whose orthologs may be involved in the Prox-dependent differentiation of retina, lens, and forebrain precursors (Choksi, 2006).

Coopted temporal patterning governs cellular hierarchy, heterogeneity and metabolism in Drosophila neuroblast tumors

It is still unclear what drives progression of childhood tumors. During Drosophila larval development, asymmetrically-dividing neural stem cells, called neuroblasts, progress through an intrinsic temporal patterning program that ensures cessation of divisions before adulthood. Previous work has shown that temporal patterning also delineates an early developmental window during which neuroblasts are susceptible to tumor initiation. Using single-cell transcriptomics, clonal analysis and numerical modeling, this study now identifies a network of twenty larval temporal patterning genes that are redeployed within neuroblast tumors to trigger a robust hierarchical division scheme that perpetuates growth while inducing predictable cell heterogeneity. Along the hierarchy, temporal patterning genes define a differentiation trajectory that regulates glucose metabolism genes to determine the proliferative properties of tumor cells. Thus, partial redeployment of the temporal patterning program encoded in the cell of origin may govern the hierarchy, heterogeneity and growth properties of neural tumors with a developmental origin (Genovese, 2019).

Central nervous system (CNS) tumors are rare and constitute less than 2% of all cancers in adults. In contrast, they represent more than 25% of cancer cases in children (including medulloblastoma, retinoblastoma, rhabdoid tumors (AT/RT), gliomas etc), suggesting that the developing CNS is particularly sensitive to malignant transformation. Moreover, unlike most adult tumors, pediatric tumors are often genetically stable and their initiation and progression do not necessarily require the accumulation of mutations in multiple genes. For example, the biallelic inactivation of a single gene is sometimes sufficient to trigger malignant growth as illustrated by mutations in the RB1 and SMARCB1 genes in retinoblastoma and rhabdoid tumors respectively. Recent studies suggest that CNS pediatric tumors such as medulloblastomas recapitulate the fetal transcription program that was active in the cell of origin. However, it remains unclear how the invalidation of single genes during fetal stages can disrupt on-going developmental programs to trigger malignant growth, and whether these fetal/developmental programs influence the heterogeneity, composition, and proliferative properties of cells composing CNS tumors (Genovese, 2019).

Faced with the complexity of brain development and neural tumors in mammals, simple animal models can represent a powerful alternative to investigate basic and evolutionary conserved principles. The development of the CNS is undoubtedly best understood in Drosophila. The Drosophila CNS arises from a small pool of asymmetrically-dividing neural stem cells (NSCs), called neuroblasts (NBs). NBs possess a limited self-renewing potential. They divide all along development (embryonic and larval stages) to self-renew while generating daughter cells named Ganglion Mother Cells (GMCs). GMCs then usually divide once to produce two post-mitotic neurons or glia. NBs are the fastest cycling cells during development, able to divide every hour during larval stages when most of the neurons are produced. However, all NBs terminate during metamorphosis and are absent in adults. Two antagonistic RNA-binding proteins, IGF-II mRNA-binding protein (Imp) and Syncrip (Syp) are essential to first promote and then conclude this formidable period of activity. During early larval development (L1/L2), NBs express Imp that promotes NB self-renewal. Around late L2/early L3, NBs silence Imp to express Syp that remains expressed until NB decommissioning during metamorphosis. This Imp-to-Syp transition is essential to render NBs competent to respond to subsequent pupal pulses of the steroid hormone ecdysone and initiate a last differentiative division. Failure to trigger the transition results in NBs permanently dividing in adults. The Imp-to-Syp transition appears to be mainly regulated by a NB intrinsic timing mechanism driven by the sequential expression of transcription factors. This series of factors, also known as temporal transcription factors, has been first identified for its ability to specify different neuronal fates produced by NBs as they divide. In addition, temporal transcription factors also schedule the Imp-to-Syp transition to ensure that NBs will not continue cycling in adults. Recent transcriptomic analyses indicate that other genes are dynamically transcribed in NBs throughout larval stages, although their function and epistatic relationship with temporal transcription factors and the Imp/Syp module are unclear. All together, these studies highlight a complex, but still relatively unexplored, temporal patterning system in larval NBs (Genovese, 2019).

Perturbation of the asymmetric division process during early development can lead to NB exponential amplification. In such conditions, the NB-intrinsic temporal program limiting self-renewal appears to become inoperant, and uncontrolled NB amplification is observed. Serial transplantations of asymmetric division-defective NBs have revealed an ability to proliferate for months, if not years, demonstrating tumorigenic characteristics. Perturbation of asymmetric divisions can be induced by the inactivation of the transcription factor Prospero (Pros) in type I NB lineages (most lineages in the ventral nerve cord (VNC) and central brain (CB)). During development, Pros is strongly expressed in GMCs where it accumulates to induce cell cycle-exit and neuronal or glial differentiation. GMCs that lack pros fail to differentiate and revert to a NB-like state. This triggers rapid NB amplification at the expense of neuron production. Previous work has shown that inactivation of pros in NBs, and their subsequent GMCs, before mid-L3 (L3 being the last larval stage) leads to aggressive NB tumors that persist growing in adults. In contrast, inactivation of pros after mid-L3 leads to transient NB amplification and most supernumerary NB properly differentiate during metamorphosis, leading to an absence of growing tumors in adults. Interestingly, propagation of NB tumor growth beyond normal developmental stages is caused by the aberrant maintenance of Imp and the transcription factor Chinmo from early-born GMCs, the latter representing the cells of origin of such aggressive tumors (Narbonne-Reveau, 2016). Chinmo and Imp positively cross-regulate and inactivation of either in NB tumors stops tumorigenic growth. Because pros-/- NB tumors can only be induced during an early window of development, and are caused by the biallelic inactivation of a single gene, they represent an exciting and simple model to investigate the basic mechanisms driving the growth of tumors with an early developmental origin, such as in the case of pediatric CNS cancers (Genovese, 2019).

NB tumors can also be induced from type II NBs (a small subset of NBs in the central brain) or from neurons upon inactivation of the NHL-domain family protein Brat or Nerfin-1 respectively. In both cases, tumor growth appears to rely on the aberrant expression of the Chinmo/Imp module arguing for a general tumor-driving mechanism in the developing Drosophila CNS (Narbonne-Reveau, 2016). Interestingly, in the different types of NB tumors, Chinmo and Imp are only expressed in a subpopulation of cells, demonstrating heterogeneity in the population of tumor NBs (tNBs). However, the full repertoire of cells composing the tumor, the rules governing the cellular heterogeneity and the mechanisms determining the proliferative potential of each cell type remain to be investigated (Genovese, 2019).

This study used single-cell RNA-seq, clonal analysis and numerical modeling to investigate these questions. A subset of genes involved in the temporal patterning of larval NBs were identified that are redeployed in tumors to generate a differentiation trajectory responsible for creating tumor cell heterogeneity. This cellular heterogeneity results in NBs with different types of metabolism and different proliferative properties. This study also deciphered a robust hierarchical scheme that drives reproducible heterogeneity through the dysregulated but fine-tuned transition between the two RNA-binding proteins Imp and Syp. This work thus identifies a core larval NB temporal patterning program, the disruption of which not only causes unlimited growth but has an overarching role in governing the cellular hierarchy, heterogeneity and metabolism of NB tumors (Genovese, 2019).

This study demonstrates that temporal patterning, not only determines which cells are susceptible to cancer transformation during development (Narbonne-Reveau, 2016), but also has an overarching role in governing different aspects of CNS tumor organization such as hierarchy, heterogeneity and the proliferative properties of the different types of cells via the regulation of their metabolism (Genovese, 2019).

Given the recent discovery that temporal patterning is conserved in the developing mammalian brain (Telley, 2019), this study could shed light on an ancestral mechanism that governs the progression of CNS tumors with developmental origins (Genovese, 2019).

The rules governing the initiation and progression of CNS pediatric tumors that often exhibit stable genomes are still unclear. Previousl work has demonstrated that temporal patterning in Drosophila larval NBs delineates a window of time during which the Chinmo/Imp oncogenic module is expressed and makes early larval NBs prone to malignant transformation (Narbonne-Reveau, 2016). This study finds that after tumor initiation, temporal patterning is partly recapitulated in tNBs where it generates differentiation trajectories to constrain tumor composition and growth. This is illustrated by the presence of about 20 genes (Imp, chinmo, Lin-28, E23, Oatp74D, Gapdh2, Sip1/CG10939, plum/CG6490, SP1173, Chd64, CG10512, CG44325, CG5953, Syp, E93, lncRNA:noe, CG15646 and stg), previously identified to be temporally regulated in some larval NBs, that are differentially regulated along the pseudotime/differentiation trajectory reconstructed from single-cell RNA-seq analysis of tNBs, and/or differentially expressed in Imp+ vs Syp+ tNBs. Thus, this study identified what appears to be a subset of a core temporal patterning program encoded in central brain and ventral nerve cord NBs that becomes deregulated upon asymmetric-division defects during early development (Genovese, 2019).

Notably, the larval temporal transcription factor Cas and Svp, known to schedule the Imp-to-Syp transition during development are not enriched in Imp+ tNBs suggesting that they do not play a role in regulating the Imp-to-Syp transition along the trajectory in tumors. Interestingly, while Syp is transcriptionally regulated in larval NBs, it seems rather post-transcriptionally regulated in tNBs as Syp RNAs are present throughout all clusters. This suggests that different mechanisms may be operating in tumors than during development to regulate the Imp-to-Syp transition (Genovese, 2019).

This study observed that the proportions of Imp+ and Syp+ tNBs systematically reach an equilibrium over a few days with a 20/80 (+/-10) ratio in poxn > prosRNAi tumors. This suggests that the regulation of the Imp-to-Syp transition in tumors is not random and the predictability of the final proportions possibly implies robust underlying constraints. By investigating the population dynamics of Imp+ and Syp+ tNBs in prosRNAi tumors, this study has deciphered a finely tuned hierarchical division scheme that appears to constrain the growth and cellular heterogeneity of the tumor. Imp+ tNBs is shown in the tumorigenic context favor a symmetric self-renewing mode of divisions (in more than 60% of divisions) while unlikely to exit the cell-cycle. This allows the perpetuation of a small subset of Imp+ tNBs that are endowed with a seemingly unlimited self-renewing potential by the Imp/Chinmo module. Imp+ tNBs can also make symmetric and asymmetric divisions that generate Syp+ tNBs, leading to the production of a population of Syp+E93+ tNBs that accumulates through limited self-renewal, and have a high propensity for exiting the cell-cycle. Moreover, this study could shows that Syp+E93+ tNBs are unable to generate Imp+ tNBs, demonstrating a rigid cellular hierarchy reminiscent of development. In addition, in this context, Syp acts as a tumor suppressor by limiting tNB proliferation while Imp acts as an oncogene by promoting tNB proliferation and propagation of tumor growth. Together, these data argue for a scenario where cooption of the Imp-to-Syp transition is responsible for installing a hierarchical mode of tumor growth with Imp+ tNBs propagating unlimited growth in a CSC-like manner, while Syp+E93+ tNBs acts as transient amplifying progenitors with limited self-renewing abilities. Although the Imp/Syp RNA-binding proteins have an essential and antagonistic role in governing the proliferative properties of tumor cells, the function of the other redeployed temporal patterning genes is unknown (except for chinmo, downstream to Imp and Syp, that is essential for tumor growth). As many are linked with the Imp+ tNB state, it will be important in the future to decipher how they contribute to establish or maintain the CSC-like identity (Genovese, 2019).

The division parameters defined by clonal analysis and modeling approach could capture both the hierarchical aspect of tumor growth as well as the global population dynamics: from an initial homogenous pool of larval Imp+ tNBs to the stable heterogeneity observed during adulthood. It could also resolve the paradoxical observation that Chinmo+Imp+ tNBs end up in minority despite exhibiting a higher average mitotic rate. Although, like all models, it is not expected that this model would perfectly recapitulate all the parameters regulating tumor growth and heterogeneity (for example, this study has neglected apoptosis and neuronal differentiation that occur at low levels), it is thought that this model provides a reasonable and useful ground on which further studies can be performed for a more detailed understanding. On these lines, while the division pattern this study has described with a numerical model provides estimates of division probabilities in poxn > prosRNAi tumors, it says nothing as to how these probabilities are biologically set within the tumor. A possible scenario is that cell fate determination upon division relies on signals received by immediate neighboring tumor cells, resulting in effective probabilities at the scale of the whole tumor. Such a micro-environment dependent regulation of the Imp-to-Syp transition in tumors would strongly contrast with the cell-intrinsic regulation of the Imp-to-Syp transition that systematically occurs in NBs around early L3. Future studies will aim at deciphering the mechanisms that interfere with the developmental progression of the temporal patterning, upon asymmetric-division defects, to favor the self-renewing mode of divisions undergone by the Chinmo+Imp+ tNBs, allowing perpetuation of a population of CSC-like cells (Genovese, 2019).

Noteworthy, prosRNAi and snr1/dSmarcb1RNAi tumors exhibit different but reproducible ratios of Imp+ and Syp+ tNBs. This suggests the existence of tumor-specific mechanisms that fine-tune the Imp-to-Syp transition. Such mechanisms may be related to the tumor cell of origin, or to the genetic insult that initiated NB amplification. Further analysis will help identifying tumor-intrinsic signals regulating the balance between Chinmo+Imp+ tNBs and Syp+E93+ tNBs in various types of NB tumors (Genovese, 2019).

Until recently, the existence of temporal patterning in mammalian neural progenitors remained uncertain. Elegant single-cell transcriptomic studies of embryonic cortical and retinal progenitors in mice have now revealed that they transit through different transcriptional states that are transmitted to their progeny to generate neuronal diversity, similar to temporal patterning in Drosophila (Clark, 2019; Telley, 2019). However, it remains unknown whether temporal patterning determines the cell of origin and governs the growth of CNS tumors in children. Along these lines, the finding that the transcriptional programs operating in cerebellar progenitors during fetal development are recapitulated in medulloblastomas is promising (Vladoiu, 2019). By uncovering the overarching role of temporal patterning in governing tumor susceptibility during CNS development and in constraining tumor properties during cancer progression in Drosophila, this work thus possibly provides a new conceptual framework to better understand CNS tumors in children (Genovese, 2019).

Because of the difficulty to investigate metabolism at the single-cell level, it has been difficult to determine how heterogeneous is the metabolic activity of cells in tumors, and how it is controlled. Using a combination of single-cell and bulk RNA-seq approaches, this study has found that progression of temporal patterning provides a tumor-intrinsic mechanism that generates heterogeneity in the proliferative abilities of tumor cells through the progressive silencing of glucose and glutamine metabolism genes (Genovese, 2019).

Consequently, Chinmo+Imp+ tNBs, that lie at the top of the hierarchy, highly express glycolytic and respiratory/OXPHOS genes, as well as Gdh, that are down-regulated by the Imp-to-Syp transition. This default high expression of both glutamine and glucose metabolism genes in CSC-like Chinmo+Imp+ tNBs likely favors sustained self-renewal, but could also confer plasticity and a way to adapt cellular metabolism to different environmental conditions as frequently observed in CSCs (e.g., glutamine can compensate for glucose shortage) (Sancho, 2016) (Genovese, 2019).

This study showed that Syp+E93+ tNBs exhibit a reduced size, and that knock-down of glycolytic (Gapdh1 or Pglym78) or respiratory/OXPHOS genes (Cyt-c-p or Cyt-C1) prevented propagation of tumor growth in adults. Thus, reduction of biosynthesis and energy production through down-regulation of glucose and glutamine metabolism genes after the Imp-to-Syp transition could progressively exhaust Syp+E93+ tNB growth and self-renewing ability, ultimately leading to cell-cycle exit (Genovese, 2019).

With the demonstration that temporal patterning regulates glycolytic, TCA cycle and OXPHOS genes in NB tumors, this work provides a tumor-intrinsic mechanism that creates metabolic heterogeneity to control the proliferative potential of the various tumor cells. It was also observed that Syp+E93+ tNBs associated with lowest levels of metabolic and cell-cycle genes also upregulate genes of the E(spl) genes. Interestingly, expression of Hes genes (orthologs of Enhancer of split genes) in vertebrate neural stem cells is associated with the maintenance of a quiescent state in adults. Thus, E(spl) genes may promote the quiescent tNB state identified with the clonal and numerical analysis while preventing their differentiation in neurons (Genovese, 2019).

Down-regulation of the mRNA levels of metabolic genes after the Imp-to-Syp transition could be due to the silencing of a transcriptional activator or to an increased mRNA degradation. On one hand, Chinmo is a likely candidate for the first scenario, as its inactivation reduces growth in NBs (Narbonne-Reveau, 2016) and this study showed that it is a direct target of both Imp and Syp. On the other hand, the second scenario is consistent with Imp orthologs in human being able to promote OXPHOS and proliferation in glioma cells, through the post-transcriptional regulation of mitochondrial respiratory chain complex subunits (Genovese, 2019).

This study has also identified a small population of tNBs expressing various stress or growth arrest factors. One of these factors, Xrp1, is a transcriptional target of p53 in the response to irradiation. Xrp1 expression has also recently been linked to defects in translation rates, together with the expression of Irbp18 and GstE6. Thus, these factors may label a subset of tNBs undergoing DNA or translational stress. The reason and consequences of such cellular stresses in tumor progression need to be further investigated (Genovese, 2019).

Transcriptomic analyses have revealed strong similarities in the differentiation trajectories of tNBs in tumors and of NBs in larvae. Yet, it is surprising that the down-regulation of glutamine and glucose metabolism genes has not been detected in NBs during larval development, after the Imp-to-Syp transition (Ren, 2017). It is possible that the glial niche surrounding NBs, that is known to influence NB growth properties during larval stages, somehow sustains high levels of glucose metabolism genes in late Syp+E93+ NBs. Given that this glial niche is absent in tumors, Syp+E93+ tNBs may not be able to sustain the high expression of metabolic genes imposed by the Imp/Chinmo module, leading to progressive cell-cycle exit (Genovese, 2019).

Chinmo and Imp are reminiscent to oncofetal genes in mammals, in that their expression decrease rapidly as development progresses while they are mis-expressed in tumors. Along these lines, the three IMP orthologs in humans (also called IGF2BP1-3) are also known as oncofetal genes. They emerge as important regulators of cell proliferation and metabolism in many types of cancers including pediatric neural cancers. Along evolution, the ancestral Syncrip gene has been subjected to several rounds of duplication and has diverged into five paralogs in mammals, some of them emerging as tumor suppressors with an important role in tumor progression (Genovese, 2019).

Thus, the respective oncogenic and tumor suppressor roles of IMP and SYNCRIP gene families appear to have been conserved in humans and they may not be restricted to tumors of neural origin. This study therefore raises the exciting possibility that these two families of RNA-binding proteins form a master module at the top of the self-renewal/differentiation cascades, that regulates CSC populations and hierarchy in a spectrum of human cancers (Genovese, 2019).

Multiple lineages enable robust development of the neuropil-glia architecture in adult Drosophila

Neural remodeling is essential for the development of a functional nervous system and has been extensively studied in the metamorphosis of Drosophila. Despite the crucial roles of glial cells in brain functions, including learning and behavior, little is known of how adult glial cells develop in the context of neural remodeling. This study shows that the architecture of neuropil-glia in the adult Drosophila brain, which is composed of astrocyte-like glia (ALG) and ensheathing glia (EG), robustly develops from two different populations in the larva: the larval EG and glial cell missing-positive (gcm⁺) cells. Whereas gcm⁺ cells proliferate and generate adult ALG and EG, larval EG dedifferentiate, proliferate and redifferentiate into the same glial subtypes. Each glial lineage occupies a certain brain area complementary to the other, and together they form the adult neuropil-glia architecture. Both lineages require the FGF receptor Heartless to proliferate, and the homeoprotein Prospero to differentiate into ALG. Lineage-specific inhibition of gliogenesis revealed that each lineage compensates for deficiency in the proliferation of the other. Together, the lineages ensure the robust development of adult neuropil-glia, thereby ensuring a functional brain (Kato, 2020).

The neuropil-glial architectures in larval ventral nerve cords (VNC) and brains are composed of a small number of neuropil-glia, generated during the embryonic stage, whereas a large number of neuropil-glia form the glial architecture in the adult VNC (Enriquez, 2018) and brain (Awasaki, 2008; Kremer, 2017). The neuropil-glial architecture appears to change in concert with the remodeling of the brain. One group proposed a model for the generation of the adult neuropil-glia architecture, in which both larval ALG and EG undergo programmed cell death. Others reported that the cell bodies of larval ALG persist during pupal life, and they re-infiltrate their fine process into the neuropil at the late pupal stage. Neuropil-glia for an adult brain are generated from a small number of specific larval neuroblasts, termed as type II neuroblasts. However, they are not accountable for the entire architecture of neuropil-glia in an adult brain; the superiormost and the inferior regions of a brain remain uncovered by neuropil-glia generated by type II neuroblasts. Thus, a broad conceptual view of how the architecture of neuropil-glia undergoes remodeling remains to be elucidated (Kato, 2020).

This study has investigated the fate of larval ALG, EG and glial cell missing-positive (gcm⁺) cells, and found that the larval EG dedifferentiate, proliferate and redifferentiate into adult ALG and EG. Together with the gcm⁺ lineage, the larval EG lineage generates the adult neuropil-glia architecture. Finally, to investigate the interaction between the lineages in the development of the neuropil-glial architecture, this study evaluated whether one lineage compensates for the failure of gliogenesis in the other lineage (Kato, 2020).

The adult architecture of neuropil-glia is formed from two lineages: the differentiated larval EG lineage and the gcm⁺ lineage. Both lineages require htl for proliferation and Pros for differentiation of ALG. Each lineage compensates for the failure of the other to proliferate. Thus, the architecture of adult neuropil-glia develops robustly to ensure a functional adult brain (Kato, 2020).

Previous studies have suggested that the adult neuropil-glia are derived from larval neuroblasts, and larval neuropil-glia (both L-EG and L-ALG) undergo programmed cell death during metamorphosis. Given that neuroblasts give rise to gcm⁺ cells, which then generate mature glial cells, the fate of gcm⁺ cells was traced and were shown to generate adult neuropil-glia. Although this result is mostly consistent with previous reports, the area occupied by adult neuropil-glia derived from gcm⁺ cells was larger than that occupied by type II neuroblast-derived glia, as reported by Ren (2018). The results suggest that type II neuroblasts are not the sole origin of gcm⁺ cells that generate adult neuropil-glia. Furthermore, this study demonstrates that L-EG also participates in the genesis of adult neuropil-glia. Collectively, this study demonstrates that adult neuropil-glia are generated from gcm⁺ cells and L-EG (Kato, 2020).

The L-EG and gcm⁺ lineages undergo proliferation at the early pupal stage to generate the architecture of neuropil-glia in the adult, which is more complex and has 100-fold more glial cells than the larva. Adult flies process a vast amount of sensory information and exhibit complex behaviors, such as courtship, aggression, flight and walking. Accordingly, the structure of the adult brain is more elaborate, with more subdivided neuropils and 20-fold more neurons than the larval brain. Thus, the cell proliferation of both lineages leads to an increase in the number of glial cells, which likely occurs in coordination with the elaboration of adult neural circuits (Kato, 2020).

Neuron-glia interactions underlie the adjustment of glial cell numbers to neuronal structure through cell survival or cell proliferation in flies and vertebrates. This study shows that the FGF receptor htl is required for cell proliferation in both L-EG and gcm⁺ lineages in early pupal life. In flies, the htl ligand Pyramus, which is secreted from neurons, regulates the proliferation of htl-positive cortex glia during the larval stage (Avet-Rochex, 2012). Similarly, it is possible that htl ligands from neurons non-cell autonomously regulate the proliferation of lineage cells. Consistent with this notion, the data show that the total number of neuropil-glia in an area is limited, unless htl is constitutively activated. Thus, such non-cell autonomous regulation may adjust the numbers of neuropil-glia in adult neural circuits, thereby enabling the complex behavior of adult animals (Kato, 2020).

ALG and EG were present in both larval and adult brains, and each cell type shares morphological features and the expression of certain markers between stages. The data show the similarities in the developmental program of neuropil-glia for embryos/larvae and adults. It was ascertain that Pros is required for the differentiation of adult ALG, as it is for the development of embryonic/larval ALG. In embryos and larvae, Pros is also required to maintain the proliferative ability of ALG. The KD of pros in the lineages results in fewer GFP+ neuropil-glia at the 50% pupa stage in some areas of the brain. This implies that Pros is involved in the regulation of cell number in the development of adult neuropil-glia. The KD of pros in the cell lineages results in the appearance of Naz+ cells. The exact identity of these cells remains unknown. Rather than differentiating into adult ALG, the persisting weak Naz+ cells in adult brains may have acquired EG-like characteristics. However, it is difficult to assess this possibility because of the lack of markers that can clearly identify adult EG. Alternatively, they may be undifferentiated cells that have failed to differentiate into adult ALG. This notion is consistent with the fact that EG are Naz- and the undifferentiated cells at the 25% pupa stage are weak/faint Naz+ cells (Kato, 2020).

This study has established that htl is required for the proliferation of the L-EG and gcm⁺ lineages. In the development of embryonic/larval neuropil-glia, the number of ALG in htl^AB42 null mutants is similar to that in the control, suggesting that htl is not involved in cell proliferation in embryos. Instead, htl is required for the proper organization of IG/ALG during embryogenesis, and for extending the fine projections of larval ALG into the neuropil. In the current analysis, the cell-proliferation phenotype of htl KD emerged in the early pupal stage; thus, the role of htl in later stages was not specifically investigated. htl is also expressed in neuropil-glia at the 50% pupa stage, when their maturation is initiated. Thus, these results do not rule out the involvement of htl in the maturation of adult ALG in later pupal stages. Nevertheless, they show that the developmental programs for larva and adult neuropil-glia partially differ (Kato, 2020).

The plasticity of glial cells in terms of their differentiation is well established in vertebrates. Radial glial progenitors generate neurons and, subsequently, some of them generate oligodendrocytes and astrocytes during development. Radial glia progenitors persist in adults and transform into neural stem cells. Both astrocytes and NG2-glia [oligodendrocyte precursor cells (OPCs)] in adult brains proliferate after injury, and generate astrocytes and oligodendrocytes, respectively. In some cases, astrocytes may even transdifferentiate into neurons after injury. NG2-glia/OPCs also generate astrocytes in cell culture, in development and after injury. Some studies also reported that NG2 glia generate neurons in adult mice. Drosophila ALG also proliferate in response to injury in larval ventral nerve cords. However, whether they differentiate into different glial subtypes or neurons is currently unexplored. Foo (2017) reported the presence of adult neural progenitor cells in Drosophila that can generate glial cells and neurons in response to a defect in glial cells. In contrast, the changes of L-EG into the progenitor state, in which cells proliferate and then differentiate, are developmentally regulated. Thus, it may serve as an excellent model for the investigation of glial cell plasticity (Kato, 2020).

This study showed that adult neuropil-glia are derived from two lineages: L-EG and gcm⁺. What is the significance of having two lineages to establish the architecture of adult neuropil-glia? The peculiar distribution patterns of the lineages may relate to the evolution of insect brains. In numerous hemimetabolous insects (e.g. locusts and cockroaches), the mandibular, maxillary and labial ganglia, which are mostly occupied by L-EG lineage cells in flies, are detached from the protocerebrum, deutocerebrum and tritocerebrum, and located more inferiorly to (i.e., below) the esophagus. In these insects, L-EG may generate neuropil-glia for the mandibular, maxillary and labial ganglia, whereas gcm⁺ cells may generate neuropil-glia for the protocerebrum, deutocerebrum and tritocerebrum. In contrast, in flies, the two different populations appear to generate neuropil-glia for one structure (i.e. a brain) as all of the areas are fused together. This notion is consistent with the idea that the segmental distribution pattern of specific embryonic neuroblasts is evolutionarily conserved between Drosophila and hemimetabolous insects (Kato, 2020).

This study demonstrates that inhibition of glial proliferation in one lineage is rescued by the other lineage. This indicates that, regardless of evolutionary relevance, the multiple lineages (i.e., L-EG and gcm⁺ cells) ensure robust development of the adult neuropil-glia architecture. Such robust development of glial architecture has been reported in several contexts. In the thorax and brain, neuroblasts generate adult neuropil-glia and compensate for the failure of gliogenesis from other neuroblasts. This study reveals that the ability to compensate for deficiencies in a lineage is not restricted to neuroblast-derived glia and is greater in scope. Each lineage (L-EG or gcm⁺) rescues the entire loss of the other lineage. A similar mechanism is involved in the development of mouse oligodendrocytes. Two lineages of oligodendrocyte precursor cells (a ventral and a dorsal population) generate oligodendrocytes in embryos and in postnatal animals. One lineage completely takes over the brain when the other fails to develop, preventing any locomotor defect. Therefore, multiple lineages with glial ability to adjust to the surroundings guarantee the robust development of glial architecture. Thus, it is suggested that glial plasticity may be a widespread strategy for ensuring the robust development of functional brains (Kato, 2020).

Neuronal upregulation of Prospero protein is driven by alternative mRNA polyadenylation and Syncrip-mediated mRNA stabilisation

During Drosophila and vertebrate brain development, the conserved transcription factor Prospero/Prox1 is an important regulator of the transition between proliferation and differentiation. Prospero level is low in neural stem cells and their immediate progeny, but is upregulated in larval neurons and it is unknown how this process is controlled. This study used single molecule fluorescent in situ hybridisation to show that larval neurons selectively transcribe a long prospero mRNA isoform containing a 15 kb 3' untranslated region, which is bound in the brain by the conserved RNA-binding protein Syncrip/hnRNPQ. Syncrip binding increases the mRNA stability of the long prospero isoform, which allows an upregulation of Prospero protein production. Adult flies selectively lacking the long prospero isoform show abnormal behaviour that could result from impaired locomotor or neurological activity. These findings highlight a regulatory strategy involving alternative polyadenylation followed by differential post-transcriptional regulation (Samuels, 2020).

Many key regulators of NB proliferation and differentiation have been identified and characterised. Recently, an increasing number of RBPs, the key regulators of post-transcriptional processes, have been implicated in neurodevelopment, suggesting that the importance of post-transcriptional regulation in the brain has so far been underestimated. This study has applied smFISH to examine the regulation of the transcription factor (TF) Pros, a master switch promoting neuronal differentiation. Pros expression is regulated by the differential stability of its mRNA isoforms depending on their 3' UTR lengths. Unstable short mRNA isoforms produce sufficient Pros protein to prevent GMCs and their neuronal progeny from reverting back to NB identity, but a switch to the more stable proslong isoforms is required to upregulate Pros protein in larval neurons. These findings highlight the capacity of cell type-specific alternative 3' UTRs to mediate different modes of post-transcriptional regulation of mRNA isoforms (Samuels, 2020).

Surprisingly, it was found that neurons do not de-differentiate into NBs in syp mutants, despite the loss of neuronal Pros upregulation in the absence of proslong. Previous work has shown that Pros elimination in young or middle-aged larval neurons causes de-differentiation of neurons and their reversion to NBs. This work suggests that the low levels of Pros remaining in the syp mutant (provided by prosshort) are sufficient for neurons to maintain their identity. proslong is not required to drive differentiation of GMCs to neurons or to maintain neuronal identity. Although this study has not uncovered the function of proslong in larval/pupal brain development, the impaired locomotive activity of the proslong-REDr adult flies indicates a role of proslong in the adult brain, either because of an earlier neuronal specification event or due to a function of Pros in the adult brain (Samuels, 2020).

Exclusion of the 3' UTR extension from pros transcripts in the proslong-REDr brains reduces the number of pros transcripts and Pros protein in the neurons. However, the pros transcript levels are still much higher in proslong-REDr than in syp mutant brains. This residual pros signal suggests that Syp can stabilise pros mRNA through binding to additional regions of the transcript, perhaps the 5' UTR sequence that is also unique to the proslong transcripts, or some shared sequence included in all transcripts (Samuels, 2020).

The 3' UTR extension of proslong may mediate a second regulatory step, at the level of translation. The upregulation of Pros protein in the neurons is lost in the proslong-REDr brains, despite the relatively high levels of remaining pros mRNA transcripts. While the pros exon smFISH signal is much higher in the proslong-REDr brains, compared to the syp mutant, the Pros protein levels are similar between the two genotypes. This result suggests that the proslong 3' UTR extension includes additional regulatory sequences that promote increased pros translation, either via Syp or an unidentified second RBP. This hypothesis would explain why a moderate decrease in pros transcript levels in proslong-REDr brains leads to a loss of neural Pros protein upregulation (Samuels, 2020).

These experiments show that Pros expression is controlled at two levels: alternative polyadenylation and then differential mRNA stability, regulated through Syp binding. The molecular mechanism underlying the cell type-specific choice of pros isoform has not yet been identified. In Drosophila embryos, Elav is recruited at the promoter of extended genes and is required to extend the 3' UTR of brat. Future experiments will determine whether pros differential polyadenylation is regulated at the promotor region by a similar Elav-dependent mechanism (Samuels, 2020).

Many key regulators in the brain also have complex gene structures such as multiple isoforms and long 3' UTRs, hallmarks of post-transcriptional mechanisms. Such genes include the temporal regulator neuronal fate, chinmo, the driver of cell growth and division, myc and the mRNA-binding proteins, Brat and Imp. Quantitative smFISH approaches combined with genetics and biochemistry will allow the detailed disentanglement of the transcriptional and post-transcriptional mechanisms regulating these genes (Samuels, 2020).

Mammalian SYNCRIP/hnRNPQ is an important regulator of neural development and has a number of post-transcriptional roles including regulating mRNA stability through binding at the 3' end of transcripts. Prox1, the mammalian orthologue of Pros, is a tumour suppressor that regulates stem cell differentiation in the brain as well as many other organ systems. Like Drosophila pros, prox1 has several isoforms including alternative 3' UTRs, and a burst of Prox1 expression is required to drive the differentiation of immature granular neurons in the adult hippocampus. It is plausible that post-transcriptional regulation by RBPs helps determine the expression and translation of Prox1. Application of quantitative smFISH to mammalian systems will uncover whether Prox1 expression levels, like pros, are regulated through differential stabilisation of different 3' UTR isoforms (Samuels, 2020).

Identification of genomic enhancers through spatial integration of single-cell transcriptomics and epigenomics

Single-cell technologies allow measuring chromatin accessibility and gene expression in each cell, but jointly utilizing both layers to map bona fide gene regulatory networks and enhancers remains challenging. This study generated independent single-cell RNA-seq and single-cell ATAC-seq atlases of the Drosophila eye-antennal disc and spatially integrate the data into a virtual latent space that mimics the organization of the 2D tissue using ScoMAP (Single-Cell Omics Mapping into spatial Axes using Pseudotime ordering). To validate spatially predicted enhancers, a large collection of enhancer-reporter lines and identify ~85% of enhancers in which chromatin accessibility and enhancer activity are coupled. Next, infer enhancer-to-gene relationships were inferred in the virtual space, finding that genes are mostly regulated by multiple, often redundant, enhancers. Exploiting cell type-specific enhancers, cell type-specific effects of bulk-derived chromatin accessibility QTLs were deconvoluted. Finally, Prospero was found to drive neuronal differentiation through the binding of a GGG motif. In summary, a comprehensive spatial characterization of gene regulation is provided in a 2D tissue (Bravo González-Blas, 2020).

Cellular identity is defined by Gene Regulatory Networks (GRNs), in which transcription factors bind to enhancers and promoters to regulate target gene expression, ultimately resulting in a cell type-specific transcriptome. Single-cell technologies provide new opportunities to study the mechanisms underlying cell identity. Particularly, single-cell transcriptomics allow measuring gene expression in each cell, while single-cell epigenomics, such as single-cell ATAC-seq (Assay for Transposase-Accessible Chromatin using sequencing), serves as a read-out of chromatin accessibility. Although these technologies and computational approaches are recently evolving to include spatial information, most approaches currently target single-cell transcriptomes. It remains a challenge how to exploit single-cell epigenomic data for resolving spatiotemporal enhancer activity and GRN dynamics, both experimentally and computationally (Bravo González-Blas, 2020).

In addition, while ATAC-seq is a powerful tool for predicting candidate enhancers, not all accessible regions correspond to functionally active enhancers. For example, accessible sites can correspond to ubiquitously accessible promoters or binding sites for insulator proteins; to repressed or inactive regions due to binding of repressive transcription factors; or to primed regions that are accessible across a tissue, but become only specifically activated in a subset of cell types. Importantly, single-cell ATAC-seq has not been fully exploited to explore these aspects yet. While most scATAC-seq studies have been carried out in mammalian systems, in which enhancer testing is not trivial, Cusanovich (2018) evaluated 31 cell type-specific enhancers predicted from scATAC-seq in the Drosophila embryo, finding that ~ 74% showed the expected activity patterns (Bravo González-Blas, 2020).

Another current challenge in the field of single-cell regulatory genomics is how to integrate epigenomic and transcriptomic information. Although some experimental approaches have been developed for profiling both the epigenome and the transcriptome of the same cell, currently either the quality of the measurements, or the throughput, is still significantly lower compared to each independent single-cell assay. For example, sci-CAR (Single-cell Combinatorial Indexing Chromatin Accessibility and mRNA) or SNARE-seq (Single-Nucleus Chromatin Accessibility and mRNA Expression sequencing) on human cells achieved a median of 1,000-4,000 UMIs (Unique Molecular Identifiers) and 1,500-3,000 fragments per cell, while the coverage with non-integrative methods, such as 10x, is around 20,000 UMIs and 10,000 fragments per cell for scRNA-seq and scATAC-seq, respectively. Methods that achieve high sensitivity, such as scCAT-seq (single-cell Chromatin Accessibility and Transcriptome sequencing), are based on microwell plates rather than droplet microfluidics, making their throughput limited (Bravo González-Blas, 2020).

Given the current limitations of combined omics methods, the computational integration of independent high-sensitivity assays provides a valuable alternative. For example, Seurat and Liger have been used to integrate independently sequenced single-cell transcriptomes and single-cell epigenomes. Nevertheless, these methods require the 'conversion' of the genomic region accessibility matrix into a gene-based matrix, and how to perform such a conversion is an unresolved issue. Some studies have used the accessibility around the Transcription Start Site (TSS) as proxy for gene expression (Bravo González-Blas, 2019); others aggregate the accessibility regions that are co-accessible (i.e., correlated) with the TSS of the gene in a certain space (Pliner et al., 2018). However, promoter accessibility is not always correlated with gene expression. Furthermore, enhancers can be located very far from their target genes-upstream or downstream, up to 1 Mbp in mammalian genomes, or up to 100-200 kb in Drosophila , often with intervening non-target genes in between-and relationships between enhancers and target genes are often not one-to-one (i.e., an enhancer can have multiple targets, and a gene can be regulated by more than one enhancer) (Shlyueva et al., 2014). Enhancer-promoter interactions can also be predicted using Hi-C approaches at the bulk level (Ghavi-Helm et al , 2019); however, these methods have limited sensitivity at single-cell resolution (Bravo González-Blas, 2020).

The Drosophila third-instar larval eye-antennal disc provides an ideal biological system for the spatial modeling of gene regulation at single-cell resolution. The eye-antennal disc comprises complex, dynamic, and spatially restricted cell populations in two dimensions. The antennal disc consists of four concentric rings (A1, A2, A3, and arista), each with a different transcriptome and different combinations of master regulators. For example, both Hth and Cut regulate the outer antennal rings (A1 and A2), with additional expression of Dll in A2, while Dll, Ss, and Dan/Danr are key for the development of the inner rings (A3 and arista), among others. On the other hand, a continuous cellular differentiation process from anterior to posterior occurs in the eye disc, in which progenitor cells differentiate into neuronal (i.e., photoreceptors) and non-neuronal (i.e., cone cells, bristle, and pigment cells) cell types. This differentiation wave is driven by the morphogenetic furrow (MF). Posterior to the MF, R8 photoreceptors are specified first, and then, they sequentially recruit R2/R5, R3/R4, and R7 photoreceptors and cone cells to form hexagonally packed units called ommatidia. In summary, the heterogeneity of cell types and differentiation trajectories results in diverse-static and dynamic-GRNs, which can be modeled with a combination of experimental and computational approaches (Bravo González-Blas, 2020).

This work first generated a scRNA-seq and a scATAC-seq atlas of the eye-antennal disc. Second, taking advantage of the fact that the disc proper is a 2D tissue, these single-cell profiles were spatially mapped on a latent space that mimics the eye-antennal disc, called the virtual eye-antennal disc. Next, by exploiting publicly available enhancer-reporter data, the relationship between enhancer accessibility and activity was assessed. Third, these virtual cells, for which both epigenomic and transcriptomic data are available, were used to derive links between enhancers and target genes using a new regression approach. Fourth, a panel of 50 bulk ATAC-seq profiles across inbred lines was used to predict cell type-specific caQTLs (chromatin accessibility QTLs). Finally, these findings were used to characterize the role of Prospero in the accessibility of photoreceptor enhancers. In summary, a comprehensive characterization is provided of gene regulation in the eye-antennal disc, using a strategy that is applicable to other tissues and organisms. These results can be explored as a resource on SCope and the UCSC Genome Browser (), and an R package, called ScoMAP (Single-Cell Omics Mapping into spatial Axes using Pseudotime ordering) is provided, to spatially integrate single-cell omics data and infer enhancer-to-gene relationships (Bravo González-Blas, 2020).

This work presents a semi-supervised approach to map omics data into a virtual template by extracting axial information via pseudotime ordering, available as an R package called ScoMAP. The main limitations of this approach are that (1) it can be currently only applied to 1D or 2D tissues, (2) it requires a priori information about at least one landmark between the real and the virtual cells and the direction of the axis, and (3) it assumes symmetry around the axes, meaning that other gradients may be lost as cells are spread randomly in each bin. Nevertheless, the spatial gene expression atlas resulting from the mapping of scRNA-seq accurately recapitulates known gene expression patterns and allows to generate virtual gene expression profiles for any gene, at a resolution comparable with novoSpaRc (Bravo González-Blas, 2020).

Whereas spatial inference has been reported based on scRNA-seq data, this work generates the first spatial map of a tissue from scATAC-seq data. This accessibility atlas effectively predicts enhancer-reporter activity for more than 700 enhancers from the Janelia FlyLight Project, with ~85% of enhancers showing matching accessibility and activity patterns. The remaining enhancers (~ 15%) are binding sites of the epithelial pioneer transcription factor Grainyhead, which primes these regions in all the epithelial cells without resulting in enhancer activity. Indeed, pioneer transcription factors are able to displace nucleosomes, resulting in an ATAC-seq signal; and despite that they are necessary, their binding is not sufficient for activity (Jacobs, 2018). Thus, enhancer accessibility can be achieved either by the binding of pioneer factors or through the cooperative binding of multiple TFs. These results highlight both the power of using scATAC-seq as a proxy of enhancer activity and the need for caution when dealing with pioneer factors (Bravo González-Blas, 2020).

The virtual map also acts as a latent space in which scATAC-seq and scRNA-seq data are available for each virtual cell. While experimental approaches for the simultaneous profiling of epigenome and transcriptome are emerging, these do not achieve the same throughout and sensitivity compared with the independent assays yet. Computationally, Granja (2019) has taken a similar approach, in which cells are mapped into the same latent space and for each single-cell transcriptome, the aggregate scATAC-seq profile of the closest neighbors is assigned. The resulting integrated profiles allow inferring relationships between enhancers and target genes. While Pliner (2018) has tackled this problem uniquely using scATAC-seq data, Granja (2019) used Pearson correlation between the chromatin accessibility and gene expression. This work extends this approach by also using random forest models to assess non-linear relationships. Of note, these approaches are not robust to pioneer sites, whose accessibility and activity are unpaired. For example, in the current approach a validated intronic enhancer of Atonal and Grainyhead in sca is missed, as the enhancer is ubiquitously accessible while only functional in the morphogenetic furrow, where the gene is expressed. Nevertheless, for the remaining 85% of the enhancers in which accessibility and activity are coupled, in this system, this study has been able to reconstruct novel and validated enhancer-to-target gene links (Bravo González-Blas, 2020).

The predicted links between enhancers and target genes support that (1) the probability of an enhancer regulating a gene decreases exponentially with the distance and the number of non-intervening genes in between, as also reported by others, and (2) genes are regulated by several-and in some cases, redundant-enhancers, with a median of 22 enhancers linked to each gene. Indeed, Cannavo (2016) reported in the Drosophila embryo that ~64% of the mesodermal loci have redundant (or shadow) enhancers, of which ~ 60% contain more than one pair of shadow enhancers. In agreement, this study finds that ~80% of the genes are regulated by shadow enhancers (6,937 out of 8,307 genes), out of which ~72% are regulated by at least three shadow enhancers (4,900 out of 6,937 genes). Transcription factors are more tightly regulated, being linked with a higher number of enhancers (with an average of 13 positive links per gene) and having almost twice the number of redundant enhancers compared with non-TFs genes. As abnormalities in the expression of transcription factor genes can have more severe phenotypes compared with final effector genes, having more-and redundant-enhancers may provide evolutionary robustness. In addition, the majority of shadow enhancers are partially redundant, meaning that they can be uniquely essential on other developmental stages or tissues, or under adverse environmental conditions (Bravo González-Blas, 2020).

Of note, almost ~50% of the inferred links are negatively correlated with their target genes. While polycomb-mediated repression has been shown to reduce region accessibility, other studies suggest that, although repressed enhancers are less accessible than active enhancers, they still show accessibility compared with the non-regulatory genome (Bozek, 2019). Such effect can be observed in the embryonic eve stripe 2 enhancer, which is active (and more accessible) in the second embryonic stripe, while repressed (and less accessible) in the rest. Meanwhile, in the eye-antennal disc, where it is not active nor repressed, there is no accessibility. Thus, accessible regions do not only correspond to primed or active enhancers, promoters, and insulators, but also to repressed enhancers (Bravo González-Blas, 2020).

Several works have focused on the inference of GRNs from single-cell data, mostly exploiting scRNA-seq to infer co-expression patterns between TFs and potential target genes. In an attempt to reduce the number of false-positive targets due to activating cascade effects, SCENIC, which additionally evaluates the enrichment of binding sites for the TF around the TSS of the putative target gene, was introduced. On the other hand, other studies have exploited single-cell ATAC-seq to find target enhancers with binding sites for specific TFs. For example, chromVAR aggregates regulatory regions based on motif enrichment and then evaluates these modules on single-cell ATAC-seq data, while cisTopic (Bravo Gonzalez-Blas, 2019) performs motif enrichment on sets of co-accessible enhancers inferred from scATAC-seq profiles (i.e., topics) to find common master regulators. However, none of these approaches incorporates knowledge about the TF nor target gene expression. This study has aimed to integrate all these layers-transcription factor binding sites, chromatin, and gene expression-to infer GRNs, by deriving co-expression modules between genes and transcription factors (from the scRNA-seq data) and pruning them based on the enrichment of the TF motif in the enhancers that regulate these genes (based on the enhancer-to-target gene links derived from the integration of scATAC and scRNA-seq data). Such networks de facto have enhancers, rather than genes, as nodes (i.e., TF-Enhancer-Gene networks) (Bravo González-Blas, 2020).

As bulk profiles may mask true biological signal (due to the proportions of the different cell types), single-cell data have been used to deconvolute cell type-specific signals from bulk RNA-seq data, permitting to exploit large cohorts with bulk omics data, complemented with only one single-cell reference atlas. This study has investigated the impact of genomic variation on cell type-specific enhancers. For example, the relevance was revealed of Atonal binding sites for opening Johnston's organ precursor-specific regions and the GGG motif, previously unlinked to any transcription factor, for opening photoreceptor regions. Interestingly, Atonal has been shown to be a key transcription factor for the specification of sensory neuronsand bHLH proteins have been proposed to act as pioneer transcription factors in certain contexts, such as the mammalian family member Ascl1 (Bravo González-Blas, 2020).

The importance of the GGG motif in neuronal enhancers was evident in most of the current analyses; however, its interpretation was a challenge because the binding TFs were unknown. While yeast one-hybrid (Y1H) experiments have been previously used to reverse-engineer which transcription factors can bind a motif of interest, lowly expressed TFs may be underrepresented in the cDNA library and interactions that occur in vivo may be missed (such as those dependent of post-transcriptional modifications). This study has used a novel in vivo approach, in which the changes that overexpression of potential TF candidates causes in chromatin accessibility at the bulk ATAC-seq level were evaluated. Although this strategy allows to characterize the effects of TF overexpression directly on the tissue of interest, it also has limitations, such as the limited throughput of in vivo genetic screens (one TF per experiment, compared to dozens of TFs that can be tested by Y1H or Perturb-ATAC in vitro). This requires making a stringent selection of potential candidates that can be further bounded by the existence of compatible tools, such as UAS-TF lines. In addition, the changes in chromatin may not be direct, but these effects can be partially ruled out using external data available, such as ChIP-seq (Bravo González-Blas, 2020).

This study found that the neuronal precursor transcription factor Prospero acts as the strongest binder of the GGG motif, followed by Nerfin-1 and l(3)neo38. In fact, overexpression of each of them, but especially Prospero, results in the opening of GGG regions; and all three transcription factors, especially Pros and Nerfin-1, can bind to the GGG motif. Based on the expression of these transcription factors, it is hypothesized that Nerfin-1-and l(3)neo38-are the early binders of the GGG motif, while Pros can bind to these regions in the late-born photoreceptors, where it is expressed. In fact, Pros and Nerfin-1 have been reported to share direct targets during CNS differentiation and have been found to be key regulators during the photoreceptor and retinal differentiation in other organisms, such as zebrafish, chicken, and mammals (Bravo González-Blas, 2020).

In summary, this study provides a comprehensive and user-friendly single-cell resource of the Drosophila's eye-antennal disc. It is envisioned that these computational strategies and enhancer resources will be of value not only to the Drosophila community, but also to the field of single-cell regulatory genomics in general (Bravo González-Blas, 2020).

Nuclear Prospero allows one-division potential to neural precursors and post-mitotic status to neurons via opposite regulation of Cyclin E

In Drosophila embryonic CNS, the multipotential stem cells called neuroblasts (NBs) divide by self-renewing asymmetric division and generate bipotential precursors called ganglion mother cells (GMCs). GMCs divide only once to generate two distinct post-mitotic neurons. The genes and the pathways that confer a single division potential to precursor cells or how neurons become post-mitotic are unknown. It has been suggested that the homeodomain protein Prospero (Pros) when localized to the nucleus, limits the stem-cell potential of precursors. This study shows that nuclear Prospero is phosphorylated, where it binds to chromatin. In NB lineages such as MP2, or GMC lineages such as GMC4-2a, Pros allows the one-division potential, as well as the post-mitotic status of progeny neurons. These events are mediated by augmenting the expression of Cyclin E in the precursor and repressing the expression in post-mitotic neurons. Thus, in the absence of Pros, Cyclin E is downregulated in the MP2 cell. Consequently, MP2 fails to divide, instead, it differentiates into one of the two progeny neurons. In progeny cells, Pros reverses its role and augments the downregulation of Cyclin E, allowing neurons to exit the cell cycle. Thus, in older pros mutant embryos Cyclin E is upregulated in progeny cells. These results elucidate a long-standing problem of division potential of precursors and post-mitotic status of progeny cells and how fine-tuning cyclin E expression in the opposite direction controls these fundamental cellular events. This work also sheds light on the post-translational modification of Pros that determines its cytoplasmic versus nuclear localization (Mar, 2022).

What makes precursor cells divide a certain number of times and how the progeny cells become post-mitotic has remained enigmatic. These questions are among some of the most fundamental questions in neurobiology. The work described in this study provides a clue and indicates that Pros and Cyclin E may be some of the key players in these processes. The results indicate that the cytoplasmic, non-phosphorylated Pros becomes phosphorylated and nuclear and binds to chromatin in cells destined to divide once. It must directly or indirectly regulate gene expression in the nucleus. One such gene regulated by Pros appears to be Cyclin E. Ample data indicates that Cyclin E is essential but also sufficient to drive entry of precursor cells to the cell cycle, although within a temporal window of developmental time. The data presented in this study show that Pros regulates Cyclin E levels in opposing directions between precursor cells and their progeny. This is an elegant, yet simple mechanism by which Pros through Cyclin E confers the one-division potential to MP2, GMC4-2a, or GMCs from NB7-3, and then helps commit their progeny to a post-mitotic state. How many lineages in the CNS also utilize this mechanism remains unknown (Mar, 2022).

The situation is not an ON/OFF scenario. A clear ON/OFF scenario will also be evolutionarily prohibitive as it would negatively affect the neuronal number, plasticity, and diversity. Instead, Pros appears to augment the upregulation of Cyclin E level in the precursor enough to commit that cell to divide once. Once it divides to generate two daughters, Pros augments the downregulation of levels of Cyclin E such that progeny cells do not enter the cell cycle, but instead become post-mitotic. The evidence to support this model comes from the fact that in pros loss of function mutants, cells such as MP2 and GMC4-2a fail to divide. The levels of Cyclin E were also downregulated in MP2 in pros mutant embryos, indicating a positive role for Pros via augmenting Cyclin E expression in MP2 division. However, in pros loss of function mutants at a later developmental stage, the level of Cyclin E was upregulated, indicating a repression role for Pros in older stage embryos. Thus, a repression of Cyclin E by Pros could lead to the post-mitotic status of progeny neurons. A switch from an activator to a repressor can be achieved by partnering with different transcription regulators before and after cell division. These results are further supported by the finding that MP2 fails to divide in loss of function for cyclin E, and gain of function for cyclin E, or gain of function for pros leads to extra divisions. The penetrance of these defects is not very high, but it is thought that this is to be expected since players such as Cyclin E will be tightly regulated during development. There is also the issue of maternal deposition when loss of function mutations is in question (Mar, 2022).

The opposing role of Pros in cyclin E regulation, depending upon the developmental stage, is meant to switch from facilitating a single division of precursors to facilitating a post-mitotic commitment of progeny cells. De-repression of Cyclin E alone in progeny cells in late-stage embryos either in the pros mutant or by over-expression of Cyclin E in post-mitotic progeny cells does not appear to be sufficient to make them re-enter the cell cycle. It may be that the Cyclin E level was not high enough in those stages of development, or the process that makes cells post-mitotic involves additional players. Thus, elevating Cyclin E alone at the 'Cyclin E-insensitive' stage may not be enough to make them re-enter cell cycle. Additionally, differentiation genes also begin to express in progeny cells, and then there is the physical process of differentiation that gets underway with neurites sprouting and axon forming. These structural changes in post-mitotic neurons could also prevent them, in addition to new gene expression programs, from re-entering the cell cycle despite elevated levels of Cyclin E. Furthermore, in 12 hpf or older pros mutant embryos, there is a general up-regulation of Cyclin E, not only in the MP2 lineage, but also in other cells in the nerve cord. The consequence of this upregulation, such as if those lineages produce extra cells, is not known. This question is currently being examined. It is also not known if Pros plays a similar role in type II NBs in the embryonic nervous system or during neurogenesis of the adult brain (Mar, 2022).

These results also argue that there may not be a dedifferentiation of cells in pros mutants as previously thought, at least not in every lineage. Pros, with its chromatin localization in cells committed to a differentiation pathway, appears to control many genes. Cyclin E alone, at least in earlier stages of development, is sufficient to make cells divide or not divide depending on the levels, and Pros fits in this Cyclin-E-mediated model by its ability to regulate cyclin E expression. In how many lineages Pros confers the single-division potential to precursor cells is not clear. It is not known if the asymmetrically localized cytoplasmic Pros in NBs has any role in cell division or if it simply is a mechanism to segregate Pros to GMCs. In any event, Pros is unlikely to regulate Cyclin E in NBs (other than MP2/NBs that have single division potential) (Mar, 2022).

Finally, these results are consistent with previous finding that in embryos mutant for a gene called midline, MP2 undergoes multiple self-renewing asymmetric divisions. Pros was cytoplasmic in MP2 in midline mutants [24], which further indicates that the single division potential of MP2 correlates with a nuclear/chromatin-bound Pros. A recent paper indicated that Pros remodels H3K9me3+ pericentromeric heterochromatin by recruiting Heterochromatin Protein 1 during neuronal differentiation (Liu, 2020). This conclusion is consistent with the supposition that Pros augments post-mitotic commitment and neuronal differentiation of progeny cells, and regulation of Cyclin E and modulating heterochromatin are essential to these developmental events. Neuronal differentiation is a complex and evolutionarily crucial process for survival; therefore, it is not surprising that various mechanisms will augment the process as a shared phenomenon. A partial redundancy for a gene or a pathway is a common theme during neurogenesis or development (Mar, 2022).

Early Studies of Prospero

How do two cells, the progeny from a single cell division, develop different fates? This is the fundamental question of developmental biology. prospero and numb provide clues to one possible mechanism.

When Prospero and Numb proteins are first made, they congregate in the cell's cytoplasm, but unlike many proteins they are not distributed uniformly throughout the cell. Rather, they form a cresent pattern at one end of the cell, just under the cell membrane, between the membrane and the cell's centrioles (Hirota, 1995, Spana, 1995a and 1995b, and Knoblich, 1995). The cell is said to be polarized; that is, one end is somehow different from the other end.

Is the centriole the immediate cause of Prospero and Numb localization? One might consider that the two centrioles are not identical, and that one directs Prospero and Numb localization? But there are other elements involved in cell polarity, and their involvement must be considered. The cytoskeleton and the cell membrane are also involved. Perhaps the cell membrane acts as a tether for Prospero and Numb. In fact, the region directly under the cell membrane, the cortex is known to be highly structured with subcortical cytoskeletal elements.

Action of Inscuteable, a cytoskeleton associated protein, suggests an involvement of cytoskeleton in providing cues for Prospero and Numb localization. Inscuteable localizes opposite the pole to which Prospero and Numb localize. In inscuteable mutants both Prospero and Numb form cresents, but the position of both the spindle and the position of Prospero and Numb is disrupted, that is, Prospero and Numb do not localize preferentially localize. Cytochalasin D, which disrupts Actin filaments, eliminates Inscuteable cresents and results in incorrect Prospero crescent positioning, while treatment with colcemid, which destroys microtubules (and consequently the spindle) does not effect Inscuteable localization (Kraut, 1996).

In yeast the site of the previous bud determines the cell polarity. Does the same thing happen in the fly? It seems not. Instead Inscuteable localizes to the apical cortex of ectodermal cells and neuroblasts (the apical direction is towards the surface and is opposite to the basal pole), irrespective of the orientation of their previous division, suggesting that the localization responds to the apical-basal polarity.

It appears that there must be an organizer to provide positional information for both the orientation of the mitotic spindle and asymmetric localization of Numb and Prospero (Knoblich, 1995). The dynamics of Inscuteable suggests that the microfilamental network, organized with respect to apical-basal polarity, acts to organize both the Numb and Prospero by acting through Inscuteable. Likewise Inscuteable acts to organize the microtubule network and thus the plane of cell division. For more information an apical-basal polarity, see the Crumbs site.

Relevant to this discussion is a structure found oogenesis called the fusome. Fusomes consist of cytoskeletal proteins, alpha-Spectrin, ß-Spectrin, Hu-li tai shao (an adducin-like protein) and Ankyrin. Of particular interest is the association of fusomes with the pole of the mitotic spindle (Lin, 1995). During the first cystoblast (cystoblasts are derived from germ line stem cells) division, fusome material is associated with only one pole of the mitotic spindle, revealing that this division is asymmetric. During the subsequent three divisions, the growing fusome always associates with the pole of each mitotic spindle that remains in the mother cell, and only extends through the newly formed ring canals after each division is completed. The association of fusome proteins with the mitotic spindle indicates a direct interaction between cytoskeletal components and only one pole of the mitotic spindle. Surely this must have something to do with the underlying mechanism of asymmetric cell division.

As a cell divides, Prospero and Numb are asymmetrically distributed to the two progeny. The cell receiving the two proteins has a different fate from that of its sister. In the cell receiving both Prospero and Numb ( the ganglion mother cell), Prospero migrates from its cytoplasmic position into the nucleus, where they assumes the role of transcription factor, thus establishing the neural fate of the cell in which it is found. Numb function is to antagonize Notch signaling, and has been shown to directly interact with the cytoplasmic domain of Notch (Guo, 1996). The second progeny, the one that does not receive Prospero or Numb, is in effect, a new neuroblast. This cell will continue to divide, generating new ganglion mother cells and new neuroblasts (Knoblich, 1995, Rhyu, 1995 and Spana, 1995) A special sequence in Numb and Prospero protein is responsible for localization of the protein in the cortex, immediately adjacent to the cell membrane (Hirata, 1995). With what protein does Numb and Prospero interact? An answer to this question will still not shed light on how apical-basal polarity is established, but it will help in understanding how Prospero and Numb become asymmetrically distributed.

A similar asymmetrical system of Prospero distribution operates in other cells expressing prospero, such as the precursors of glia (Spana, 1995).

GENE STRUCTURE

Bases in 5' UTR - 1089

Exons - two

Bases in 3' UTR - 1097

PROTEIN STRUCTURE

Amino Acids - 1403

Structural Domains

The Prospero transcript is alternatively spliced to encode two proteins: ProsL protein (1403 amino acids, predicted 165 kDa) and ProsS protein (1374 amino acids, predicted 160 kDa). The extra 29 amino acids in ProsL are at the beginning of the homeodomain (Chu-LaGraff, 1991).

Prospero has a novel homeodomain and three PEST domains which confer susceptability to rapid degradation. There is also a CAX (opa) repeat, a common transcription activation domain (Doe, 1991 and Vaessin, 1991).

Prospero promotes neural differentiation in Drosophila, and its activity is tightly regulated by modulating its subcellular localization. Prospero is exported from the nucleus of neural precursors but imported into the nucleus of daughter cells, which is necessary for their proper differentiation. Prospero has a highly divergent putative homeodomain adjacent to a conserved Prospero domain; both are required for sequence-specific DNA binding. The structure of these two regions consists of a single structural unit (a homeo-prospero domain), in which the Prospero domain region is in position to contribute to DNA binding and also to mask a defined nuclear export signal that is within the putative homeodomain region. It is proposed that the homeo-prospero domain coordinately regulates Prospero nuclear localization and DNA binding specificity (Ryter, 2002).

The C-terminal 163 amino acids of Drosophila Prospero (residues 1241-1403) were overexpressed, purified, and crystallized. The X-ray structure was determined at 2.05 Å resolution. Despite classification as highly divergent based on sequence comparisons, residues 1241-1303 were predicted to be capable of assuming a canonical homeodomain structure. In fact, structural analysis shows that this region does assume an overall fold very similar to homeodomains but with one striking difference. In every homeodomain structure determined to date, the so-called DNA recognition helix is either at the extreme C terminus of the protein or leads into a flexible linker of variable length that connects to another essentially independent domain. In Prospero, the recognition helix (alpha₃) connects the putative homeodomain and the Prospero domain as a single structural unit. The region corresponding to the Prospero domain can be described as a four-helix bundle (alpha₃–alpha₆). Residues 1314-1326 are disordered and residues 1368-1370 and 1391-394 form short 3₁₀ helices. While interactions between the homeo- and Prospero domain regions occur primarily within the hydrophobic core, the C-terminal residues of the Prospero domain region make additional contacts with the homeodomain region by extending into a cleft between helices alpha₁ and alpha₂ (Ryter, 2002).

A structural comparison of the homeodomain region (HD) alone reveals that, while possessing a highly divergent class sequence, the structure assumes a fold most similar to the Drosophila Engrailed homeodomain. A detailed residue-by-residue analysis was possible. A majority of the core residues seen in the homeodomain region are either invariant or conserved hydrophobic with respect to the Engrailed homeodomain sequence or homeodomain consensus sequence, indicating that this region possesses a core consistent with a standard homeodomain hydrophobic core. Variations occur primarily at the loop between alpha₁ and alpha₂ and in the middle of alpha₃. In these regions, residues that would be on the surface of a typical homeodomain are replaced by bulky hydrophobic groups that interact with the hydrophobic core of the Prospero domain. Within the loop, Phe1261 and Trp1262, along with the invariant Phe1260, create a very hydrophobic turn that protrudes into this core region. In the middle of alpha₃, Phe1298, Tyr1299, Tyr1300, and Met1303 contribute to the hydrophobic core of the Prospero domain four-helix bundle (Ryter, 2002).

The Prospero domain region (PD) is composed of a four-helix bundle (alpha₃-alpha₆). Consecutive helices are antiparallel and all helices interact via ridges in grooves. The up-down-up-down topology is the simplest found among four-helix bundles and in this case is right turning. The central residues of this bundle, representing the alpha-carbons in closest proximity, are Ala1307 of alpha₃ and Phe1361 of alpha₅. The ~50° crossing angle between the helix axes is also the most common angle found in globular proteins. The core consists of hydrophobic residues throughout that are either invariant or highly conserved throughout the Prospero/Prox1 protein family. As noted above, a number of residues within the homeodomain region also contribute to this hydrophobic core (Ryter, 2002).

The homeodomain and the Prospero domain regions do not just touch each other; rather, they interact quite extensively. If the two regions are considered as separate entities, then 22% (1499 Ų) of the surface area of the homeodomain region contributes to the interface with the Prospero domain region and, vice versa, 35% of the surface of the Prospero domain region contributes to the same interface. These fractions are significantly higher than the average value of 15% that is observed for the surface area involved in the interfaces of typical protein dimers. It confirms that the homeodomain and Prospero domain regions combine to form a distinct structural unit (Ryter, 2002).

There are extensive hydrophobic core contacts that occur between the HD and PD regions. A hydrophobic cleft on the HD is formed between Trp1262 and Tyr1274 of the strand just C-terminal to alpha₁ and alpha₂, respectively. SeMet1259 forms the floor of this cleft, with Val1263 and Val1270 flanking one end. Seated in the cleft are Val1389, Pro1390, and Phe1393 of the PD region. This cleft is flanked at the opposite end by a hydrogen-bonding interaction between His1252 and Asp1277. Phe1275, of the strand C-terminal to alpha₂, positions Lys1255 of alpha₁ in an orientation optimal for a number of hydrogen-bonding interactions. These include direct interactions with backbone carbonyls of residues 1274 and 1393, thereby creating contacts between HD alpha₁ and alpha₂, as well as the PD region (Ryter, 2002).

There are interesting similarities (see 1mijA at Structural Neighbours in PDB90 and structural alignments) between the homeo-prospero domain (HPDP structure) and that of the MATa1/MATalpha2 homeodomain heterodimer bound to DNA (Li, 1998). In the HPD structure, the C terminus of the PD region (residues 1388-1396) extends into a cleft between alpha₁ (residues 1250-1259) and alpha₂ (residues 1268-1274) of the HD region. This rather lengthy segment includes a 3₁₀ helix (residues 1391-1394) and contacts the HD region on the face opposite its DNA binding surface. This interaction between HD and PD is stabilized by hydrophobic interactions, as well as a number of hydrogen bonds. In the homeodomain heterodimer complex between MATa1 and MATalpha2, the C terminus of MATalpha2 also extends away from the body of the domain and similarly makes hydrophobic and hydrogen bond interactions within a groove formed between alpha₁ and alpha₂ of MATa1. This binding groove on the homeodomain MATa1 corresponds to that within the HD region of Prospero. In addition, the polypeptide chain that binds within the respective grooves in both cases adopts an alpha-helical or 3₁₀-helical conformation (Ryter, 2002).

The MATa1/MATalpha2 homeodomain complex is required for cell type-specific transcriptional repression in yeast. The Prospero HPD is also required for regulating transcription, but may have an additional function in controlling the subcellular localization of Prospero. Demidenko (2001) utilized molecular dissection of the HPD with expression of chimeric proteins in mammalian and insect cultured cells to show that residues 1248–1261, encompassing alpha₁, contain a nuclear export signal (NES). In the absence of any portion of the PD region, Prospero is exported from the nucleus via the Exportin pathway and subsequently found in the cytoplasm, while presence of this region blocks nuclear export and allows Prospero to accumulate in the nucleus. The HPD structure strongly suggests that it is the extreme C terminus of the PD region that sterically prevents access to the nuclear export signal (Ryter, 2002).

Interestingly, previous analysis of Prospero during asymmetric cell division has shown its phosphorylation state correlates with subcellular localization. Cytoplasmic Prospero is highly phosphorylated compared to nuclear Prospero, which raises the possibility that phosphorylated HPD may have a more open conformation in which the NES is exposed (Ryter, 2002).

In a structural comparison, the structure of the HD region was shown to be most similar to the Engrailed homeodomain. It is therefore possible to use the structural superposition matrix to model the HPD structure in complex with DNA. Three regions of potential protein-DNA contact are suggested. (1) The DNA recognition helix alpha₃ appears to contact the major groove of the DNA; (2) the N-terminal arm appears to contact the minor groove of the DNA, and (3) unexpectedly, the PD region appears to contact the DNA backbone via residues within or close to the N terminus of helix alpha₆. Engrailed homeodomain residues Asn51 (Pros HD Asn1294) and Arg53 (Pros HD Arg1296) are invariant and presumably contact the DNA in a similar fashion. In the Engrailed structure, Ile47 (Pros HD Lys1290) and Lys50 (Pros HD Ser1293) also contact the bases in the major groove to provide binding site discrimination. Pros HD Glu1297 appears poised in the major groove available for possible water-mediated contacts. Additionally, there are a number of residues that are not only invariant between the Engrailed homeodomain and the Prospero HD region, but are also positioned to make similar contacts with the DNA phosphate backbone as seen in the Engrailed homeodomain structure. In particular, Engrailed homeodomain Arg53 (Pros HD Arg1296) and Tyr25 (Pros HD Tyr1265) are predicted to interact directly with one strand of the DNA backbone in a similar fashion, while Trp48 (HD Trp1291) could interact with the opposite strand (Ryter, 2002).

In the Engrailed homeodomain structure, the N-terminal arm (residues 5-9) fits into the DNA minor groove, supplementing the contacts made by the recognition helix. In the Prospero HPD-DNA model, the N-terminal arm has a somewhat different alignment but could easily move so as to contact the DNA. A sequence comparison reveals predominantly conservative differences between the Engrailed and the Prospero sequences in the N-terminal arm, except at Engrailed homeodomain Arg5 (Pros HD Ser1245), a position shown to be important for binding site discrimination (Ryter, 2002).

This structural comparison also suggests that the Prospero HD region may contact DNA as well. Lys1380, -1376, and -1375, found within or close to the N terminus of the alpha₆ helix, are all poised to potentially interact with the DNA phosphate backbone. The basic nature of this region is further manifest in the negative electrostatic potential of the molecular surface. Tyr1379 and Ser1373 could possibly interact with the DNA backbone either directly or via hydrogen bond interactions. All these residues are conserved to a high degree in the Prospero/Prox class homeobox proteins. Of course, final confirmation of the HPD-DNA interactions discussed above must await the structure determination of the HPD structure in complex with its DNA binding site (Ryter, 2002).

Therefore, instead of forming an essentially independent unit, the Prospero domain is shown to join together with the homeodomain to form a larger structural unit that has been named the 'homeo-prospero domain'. Model building suggests that this larger structural unit serves in part to align the Prospero domain region on the DNA target. Also, the Prospero domain region is positioned in such a way that it is able to mask a defined nuclear export signal that is within the homeodomain region (Ryter, 2002).

Subcellular localization of the transcription factor Prospero is dynamic. For example, the protein is cytoplasmic in neuroblasts, nuclear in sheath cells, and degraded in newly formed neurons. The carboxy terminus of Prospero, including the homeodomain and Prospero domain, plays roles in regulating these changes. The homeodomain has two distinct subdomains, which exclude proteins from the nucleus, while the intact homeo/Prospero domain masks this effect. One subdomain is an Exportin-dependent nuclear export signal requiring three conserved hydrophobic residues, which models onto helix 1. Another, including helices 2 and 3, requires proteasome activity to degrade nuclear protein. Finally, the Prospero domain is missing in pros(I13) embryos, thus unmasking nuclear exclusion, resulting in constitutively cytoplasmic protein. Multiple processes direct Prospero regulation of cell fate in embryonic nervous system development (Bi, 2003).

date revised: 20 October 2023

Home page: The Interactive Fly © 1995, 1996 Thomas B. Brody, Ph.D.

The Interactive Fly resides on the
Society for Developmental Biology's Web server.