Many neural progenitors, including Drosophila mushroom body (MB) and projection neuron (PN) neuroblasts, sequentially give rise to different subtypes of neurons throughout development. A novel BTB-zinc finger protein, named Chinmo (Chronologically inappropriate morphogenesis), governs neuronal temporal identity during postembryonic development of the Drosophila brain. In both MB and PN lineages, loss of Chinmo autonomously causes early-born neurons to adopt the fates of late-born neurons from the same lineages. Interestingly, primarily due to a posttranscriptional control, MB neurons born at early developmental stages contain more abundant Chinmo than their later-born siblings. Further, the temporal identity of MB progeny can be transformed toward earlier or later fates by reducing or increasing Chinmo levels, respectively. Taken together, it is suggested that a temporal gradient of Chinmo (Chinmo^high --> Chinmo^low) helps specify distinct birth order-dependent cell fates in an extended neuronal lineage (Zhu, 2006).

Neural progenitors often generate distinct subtypes of neurons in an invariant temporal sequence during development. For example, cortical progenitors produce neurons that will occupy different layers of the neocortex in a stereotyped temporal order. Orderly generation of distinct cell types is also observed in the retina, spinal cord, and hindbrain. However, the molecular mechanisms governing the specification of neuronal cell fates based on birth order/timing, especially in vertebrates, remain largely unknown (Zhu, 2006).

Several studies suggest that the sequential production of different neuronal subtypes reflects temporal changes in neural progenitors. Heterochronic transplant or coculture experiments have yielded evidence supporting two distinct models elaborating such temporal changes. In the 'progressive restriction model,' neural progenitors progressively lose their intrinsic capabilities to produce different types of cells: young progenitors generate specific progeny in response to changes in extrinsic cues, and older progenitors appear intrinsically limited to generate only later-born cell fates. Evidence for the progressive restriction model is mainly derived from heterochronic transplantation experiments, in which early cortical progenitors can develop into any later-born cell fates if transplanted into an older cortex, but not the converse. Analogous studies, however, reveal a different scenario for retina progenitor cells (retinoblasts). Heterochronic coculture or transplant experiments show that the type of cells that the donor retinoblasts generate is independent of the developmental age of the host environment. These observations have led to proposal of the 'competence model', which suggests that progenitor cells pass through a series of distinct competence states to produce different types of progeny sequentially. Although progenitors are more influenced by the environment in the progressive restriction model, both models acknowledge intrinsic differences in progenitor cells of different ages (Zhu, 2006 and references therein).

The Drosophila embryonic ventral nerve cord (VNC) has been an informative model system for studying the specification of neuronal temporal identity. In each hemisegment of the Drosophila VNC, 30 neural progenitors, neuroblasts (Nbs), undergo a series of asymmetric divisions, with each division generating a variety of distinct neurons and/or glia in a fixed temporal sequence. In many lineages, individual Nbs and their progeny can be uniquely identified based on their spatial positions and expression of specific cell markers. Interestingly, temporal cell fate specification in most embryonic Nb lineages appears to be governed by a common mechanism involving the sequential expression of a set of transcription factors, Hunchback (Hb) --> Kruppel (Kr) --> POU domain transcription factors (Pdm) --> Castor (Cas). These transcription factors are transiently expressed in the Nbs in the above sequential order, and the temporal window of this expression for each transcriptional factor averages about one-cell-cycle. These factors are inherited by the postmitotic cells, which are born within these temporal gene expression windows, and in the daughter cells they act to specify temporal identity (Brody, 2002 and Isshiki, 2001). Together, these suggest that intrinsic changes in Nb gene expression at different developmental stages play an important role in specifying the temporal identities of Nb progeny (Zhu, 2006 and references therein).

Cellular mechanisms governing temporal cell fate specification are also critical during postembryonic development of the Drosophila brain. By the unparalleled ability to genetically label single neurons at specific developmental time points, MARCM (mosaic analysis with a repressible cell marker) has been used to demonstrate the presence of stereotyped neuronal temporal identities in a number of Drosophila postembryonic neuronal lineages, including the mushroom body (MB) and projection neuron (PN) lineages. In a MARCM-based screen, novel BTB-zinc finger protein, named Chinmo (Chronologically inappropriate morphogenesis), was identified based on defects in neuronal development in the Drosophila olfactory learning/memory center, the MBs. Chinmo's loss-of-function phenotypes were examined in multiple postembryonic neuronal lineages; mutant neurons born at early developmental stages often adopted the fates of late-born neurons of the same lineage. Interestingly, possibly due to differential translation of chinmo messages, early-born neurons contain more Chinmo than their later-born siblings; and a reduction or an increase in the Chinmo levels can accelerate or delay fate transitions in MB neuronal temporal identity. Taken together, a novel mechanism is suggested for the specification of neuronal temporal identity: postmitotic neurons born at early developmental stages contain more abundant Chinmo than later-born neurons in the same lineage, and such temporal gradients of Chinmo (Chinmo^high --> Chinmo^low) specify temporal cell fates in an extended neuronal lineage (Zhu, 2006).

In Drosophila, most Nbs repeatedly undergo asymmetric divisions to produce multiple rounds of two postmitotic neurons following the transient derivation of ganglion mother cells (GMCs). In an extended multicellular lineage, MARCM permits labeling of single-cell/two-cell clones that consist of the progeny born shortly after induction of mitotic recombination. By examining single-cell/two-cell clones generated at different developmental stages, it has been demonstrated that postembryonic development of the Drosophila MBs involves sequential generation of four distinct subtypes of neurons (γ → α′/β′ → pioneer α/β → α/β (Zhu, 2006).

To elucidate the molecular mechanisms underlying temporal cell fate specification, a screen was performed for mutations that affect sequential generation of distinct MB neurons in mosaic organisms. By using MARCM with a ubiquitous MB GAL4 driver (GAL4-OK107), one can generate and selectively visualize clones of MB neurons that are homozygous mutant for a specific chromosome arm within an otherwise heterozygous organism. A mature MB Nb clone, induced shortly after larval hatching (ALH), normally generates progeny that populate all the five MB lobes. The γ lobe (weakly labeled by the mAb 1D4) is derived from MB neurons born before the midthird-instar stage, the α′ and β′ lobes (negative for the mAb 1D4) consist of projections from late larval-born MB neurons, and the α and β lobes (strongly labeled with the mAb 1D4) are pupal-born MB neurons' derivatives. It was reasoned that mutations in genes autonomously controlling MB neuron temporal identity would cause predictable changes in the cellular composition of Nb clones and that one should be able to identify such abnormalities via analysis of subtype-specific axon projections. For example, a mutation in a gene required for early temporal fates might lack γ neurons. Conversely, mutations in genes mediating the acquisition of late-born fates might show defects in α/β production. 3500 mutagenized chromosome arms were screened and recovered one such MB neuronal temporal identity mutant, l(2L)MB523, was found whose full-sized Nb clones specifically lack larval-born MB γ and α′/β′ neurons (Zhu, 2006).

In contrast with wild-type controls, MB Nb clones homozygous for l(2L)MB523 generated at newly hatched larval (NHL) stage contain only one vertical and one horizontal axon bundle despite presence of many cell bodies. Colocalization was observed of l(2L)MB523 mutant axon bundles with the strong 1D4-positive α and β lobes, and mutant Nb clones contain ectopic fascicles of axons that also have high affinity for the mAb 1D4. These results suggest that most of the axons in the l(2L)MB523 mutant Nb clones likely differentiate as the α/β type of MB projections. In addition, there appears to be far more projections of the pioneer α/β type, as evidenced by prominent neurite staining on the dorsal/upper portion of the proximal two thirds of the mutant β lobe. Pioneer α/β neurons are normally born before α/β neurons (born within 6 hr before pupal formation [BPF]). Unlike other MB neurons, each pioneer α/β neuron extends only one primary dendrite into the calyx, and these dendrites, even in cells with different clonal origins, analogously project to the anterodorsal side of the calyx. Besides, their medial axon projections selectively extend along the upper surface of the β lobe but fail to reach the lobe terminus (Zhu, 2006).

Furthermore, in contrast with GAL4-OK107-labeled clones that have comparable sizes irrespective of genotypes, GAL4-201Y and GAL4-NP21 labeled much fewer cell bodies in l(2L)MB523 mutant Nb clones. Given that both GAL4-201Y and GAL4-NP21 selectively label γ neurons, this observation further suggests that early-born MB neurons in l(2L)MB523 clones have acquired later-born cell fates (Zhu, 2006).

To map l(2L)MB523, a lethal mutation was identified within the 22A6–22B1 cytogenomic region, based on a series of complementation tests. It was subsequently found that l(2L)MB523 belongs to the complementation group already consisting of one P element insertion, l(2)04111^K13009, and one EMS mutation, l(2)04111^M33. To determine which gene is affected in the l(2)04111 complementation group, the lethal P insertion line was characterized. Viable revertants were obtained following induction of P element excision, ascribing the lethality to disruption of an essential gene by a P element in l(2)04111^K13009. Then the P element's insertion site was determined by inverse PCR, and it was found that one P element is inserted in the first exon of the gene CG31666 in l(2)04111^K13009. CG31666 encodes a predicted BTB-zinc finger protein of 604 amino acids, with one BTB domain (amino acid 22–128) near the N terminus and two C₂H₂-type zinc fingers (amino acid 517–540 and 545–568) in the C-terminal region. Sequencing of CG31666's predicted exons and their splicing junctions further allowed identificaiton of a missense mutation in l(2)04111^M33, converting an evolutionally conserved aromatic amino acid (Phe88) to Isoleucine in the BTB domain. This missense mutation apparently blocks CG31666 gene's function because the same mutation abolished ectopic CG31666's gain-of-function phenotypes (Zhu, 2006).

To examine whether CG31666 is indeed expressed in the developing MBs and whether its expression is compromised by the unidentified l(2L)MB523 mutation, a peptide antibody was generated that specifically recognizes CG31666. Immunostaining with the anti-CG31666 antibody revealed nuclear localization of CG31666 in wild-type MB neurons at the mid-third-instar stage. In contrast, no CG31666 immunoreactivity was detected in l(2L)MB523 mutant MB clones. These results demonstrate CG31666 expression in the larval MBs and indicate the l(2L)MB523 as likely a protein null allele of CG31666 (Zhu, 2006).

This study shows that Chinmo levels in MB Nb progeny are predictive of daughter cells' temporal fates. Examinations and manipulations of Chinmo levels in vivo reveal how Chinmo^high → Chinmo^low may help govern the γ → α′/β′ → pioneer α/β → α/β order of MB neurogenesis. First, prospective γ neurons contain more abundant Chinmo than prospective α′/β′ neurons, and Chinmo expression apparently drops to zero in prospective α/β neurons. Second, MB Nbs repeatedly produce prospective γ neurons during the first 3 days of larval development; and, interestingly, gradients of Chinmo also exist among the sequentially derived prospective γ neurons. Third, most of the subsequently derived progeny likely become null for Chinmo after removal of chinmo from a young MB Nb, which explains why most of the larval-born MB neurons adopt the Chinmo null temporal fate and develop into pioneer α/β neurons in chinmo mutant MB Nb clones. Fourth, possibly due to perdurance of Chinmo after clone induction, most of single-cell clones of chinmo mutant MB neurons adopt their next temporal fates, reflecting likely presence of different residual Chinmo in the single-cell clones that are generated at different developmental stages. Chinmo perdurance may result from transcription of chinmo before the GMC → postmitotic neuron division, as suggested by presence of abundant chinmo transcripts in the larval MB Nbs. Fifth, MB Nbs precociously produce late-type neurons in a chinmo¹/+ heterozygous background but prolong the production of early-type neurons in the organisms that carry extra copies of chinmo. Such gene dosage affects the absolute amounts of Chinmo protein but probably leave the gradients largely unaltered, as suggested by their subtle phenotypes in shifting the timing of temporal cell fate transition. Interestingly, if Chinmo protein expression could be maintained at a high level through development (e.g., with GAL4-OK107-mediated induction of UAS-chinmo-3′UTR, MB progenitors appear remaining young and continuously yield the first-derived subtype of MB neurons until the end of neurogenesis. Taken together, the specification of distinct MB temporal cell fates on the basis of Chinmo's levels of expression suggests a novel mechanism by which time/birth order-dependent cell fates are determined (Zhu, 2006).

Posttranscriptional regulation is potentially used to control gradient expression of chinmo. Although changing chinmo's gene dosage might modulate the absolute amounts of Chinmo protein, multiple independent lines of evidence do support the notion that gradients of Chinmo are primarily derived from differential translational control in postmitotic neurons. Remarkably, such differential translation phenomena can later be recapitulated and used to tell the temporal identity of individual postmitotic neurons. The fact that these characteristics are stably inherited by every postmitotic neuron is in great contrast to the mechanism thought to operate in the embryonic CNS. In the embryo the regulatory networks governing the successive Hb → Kr → Pdm → Cas expression are likely to operate only transiently during neurogenesis. This argues against the possibility that Chinmo gradients are directly generated by some timing machinery. Instead, postmitotic cells likely acquire the ability to synthesize specific amounts of Chinmo based on birth order/timing following their initial temporal cell fate specification. Levels of Chinmo subsequently govern birth order/timing-dependent subtype-specific terminal differentiation in individual postmitotic neurons. Interestingly, Chinmo is also selectively required for proper specification of early temporal cell fates, at least, in the anterodorsal lineage of PNs. Taken together, these Chinmo studies so far suggest that temporal cell fate specification may generally take place step by step, and that a temporal gradient of Chinmo is possibly the intermediate master for specifying distinct birth order/timing-dependent neuronal morphologies (and, likely, subtype-specific molecular features as well) (Zhu, 2006).

Multiple mechanisms may act independently and/or in concert to simultaneously and/or sequentially govern neuronal temporal identity. For instance, a non-Chinmo mechanism is apparently required for proper specification of distinct subtypes of MB α/β neurons; and gradients of Chinmo may operate collaboratively with additional mechanisms to increase neuronal diversity in the PN lineages. It is speculated that Hb → Kr → Pdm → Cas and gradients of Chinmo may complement each other and, respectively, mediate quickly changing temporal cell fates and control slowly progressing cell fate transitions. Interestingly, Chinmo is also enriched but not uniformly expressed in the developing embryonic nervous system, suggesting its possible involvement in regulating cell fate transitions during embryonic neurogenesis as well. In sum, systematic analysis of Chinmo functions promises to shed new light on how numerous distinct neuron types can be derived from a limited number of progenitors during development of the complex nervous system (Zhu, 2006).

The timing mechanisms responsible for terminating cell proliferation toward the end of development remain unclear. In the Drosophila CNS, individual progenitors called neuroblasts are known to express a series of transcription factors endowing daughter neurons with different temporal identities. This study shows that Castor and Seven-Up, members of this temporal series, regulate key events in many different neuroblast lineages during late neurogenesis. First, they schedule a switch in the cell size and identity of neurons involving the targets Chinmo and Broad Complex. Second, they regulate the time at which neuroblasts undergo Prospero-dependent cell-cycle exit or Reaper/Hid/Grim-dependent apoptosis. Both types of progenitor termination require the combined action of a late phase of the temporal series and indirect feedforward via Castor targets such as Grainyhead and Dichaete. These studies identify the timing mechanism ending CNS proliferation and reveal how aging progenitors transduce bursts of transcription factors into long-lasting changes in cell proliferation and cell identity (Maurange, 2008).

Initially investigated was whether distinct temporal subsets of neurons are generated throughout the larval CNS. Chinmo and Broad Complex (Br-C), two BTB-zinc finger proteins known to be expressed in the postembryonic CNS, are distributed in complementary patterns in the central brain, thoracic, and abdominal neuromeres at the larval/prepupal transition stage at 96 hr (timings are relative to larval hatching at 0 hr). Chinmo is expressed by early-born neurons located in a deep layer, whereas Br-C marks later-born neurons in a largely nonoverlapping and more superficial layer. The deep Chinmo⁺ layer comprises most/all neurons born in the embryo plus an early subset of those generated postembryonically. Thoracic postembryonic neuroblasts undergo the Chinmo --> Br-C switch at ~60 hr such that they have each generated an average of 15 Chinmo⁺ cells expressing little or no Br-C and 39 Chinmo^- Br-C⁺ cells by 96 hr. The Chinmo⁺ and Br-C⁺ neuronal identities can be recognized as distinct cell populations on the basis of an ~2-fold difference in cell-body volume. This equates to an average cell-body diameter for Chinmo⁺ neurons of 4.5 μm, compared to only 3.6 μm for Br-C⁺ neurons. Plotting cell diameter versus deep-to-superficial position within postembryonic neuroblast clones reveals an abrupt decrease in neuronal size at the Chinmo --> Br-C transition. Together, these results provide evidence that most, if not all, postembryonic neuroblasts sequentially generate at least two different populations of neurons. First they generate large Chinmo⁺ neurons and then they switch to producing smaller Br-C⁺ neurons (Maurange, 2008).

To begin dissecting the neuronal switching mechanism, the functions of Chinmo and Br-C were investigated, but neither factor was found to be required for the transition in cell identity and cell size. Next it was asked whether a temporal transcription factor series related to the embryonic Hb --> Kr --> Pdm --> Cas sequence might be involved. Cas is known to be expressed in the larval CNS (Almeida, 2005), and this study shows that many different thoracic neuroblasts progress through a transient Cas⁺ phase during the 30-50 hr time window. Thoracic neuroblasts transiently express another member of the embryonic temporal series, Svp, during a somewhat later time window, from ~40 to ~60 hr. These results indicate that postembryonic Cas and Svp bursts are observed in many, but probably not all, thoracic progenitors and that their timing varies from neuroblast to neuroblast (Maurange, 2008).

To determine Svp function, thoracic neuroblast clones were generated homozygous for svp^e22, an amorphic allele. In ~53% of svp^e22 neuroblast clones induced in the early larva (at 12-36 hr), the Br-C⁺ neuronal identity is completely absent, all neurons express Chinmo, and there is no sharp decrease in neuronal size. The proportion of lineages failing to generate Br-C⁺ neurons rises to ~70% when clones are induced in the embryo and falls to only ~7% with late-larval (65-75 hr) induction. This is consistent with a previous finding that Svp bursts are asynchronous from neuroblast to neuroblast. The expression and clonal analyses together demonstrate that a progenitor-specific burst of Svp is required in many lineages for the switch from large Chinmo⁺ to small Br-C⁺ neurons (Maurange, 2008).

Thoracic neuroblast lineages homozygous for a strong cas allele, cas²⁴, show no obvious defects in the Chinmo --> Br-C transition when induced at 12-36 hr. However, since Cas is expressed in many postembryonic neuroblasts before their first larval division, it can only be removed by inducing clones during embryonic neurogenesis. Such cas²⁴ clones generate supernumerary Chinmo⁺ neurons and completely lack Br-C⁺ neurons at 96 hr, although this switching phenotype is restricted to only ~16% of thoracic neuroblasts. Constitutively expressing Cas blocks the Chinmo --> Br-C switch in a similar manner, with a frequency dependent upon whether thoracic UAS-cas clones are induced during embryogenesis (~47%), at early-larval (~10%) or at late-larval (0%) stages. This indicates that the response to Cas misexpression decreases as neuroblasts age. Together, the expression and loss- and gain-of-function analyses demonstrate that Chinmo and Br-C are negative and positive targets, respectively, of Cas and Svp. They also strongly suggest that progression through transient Cas⁺ and Svp⁺ states permits many postembryonic neuroblasts to switch from generating large to small neurons (Maurange, 2008).

To investigate whether Cas and Svp regulate neural proliferation as well as neuronal fates, the effector mechanism ending neurogenesis in the central brain and thorax was identified. In these regions, most neuroblasts cease dividing in the pupa at ~120 hr. Correspondingly, neurogenesis in all regions of the wild-type CNS ceases before the adult fly ecloses such that no adult neuroblasts are detected. In contrast to the central abdomen, blocking cell death by removing Reaper, Hid, and/or Grim (RHG) activity in the central brain and thorax does not prevent or delay pupal neuroblast disappearance. However, time-lapse movies of individual thoracic neuroblasts at ~120 hr reveal an atypical mitosis that is much slower than at ~96 hr, producing two daughters of almost equal size. This is largely accounted for by a reduction in the average diameter of neuroblasts from 10.4 μm at 96 hr to 7 μm at 120 hr, as GMC size does not vary significantly during this time window. The end of this atypical progenitor mitosis temporally correlates with reduced numbers of Mira⁺ cells and disappearance of the M phase marker phosphorylated-Histone H3 (PH3), indicating that it marks the terminal division of the neuroblast (Maurange, 2008).

Next whether late changes in basal complex components might underlie loss of neuroblast self-renewal was addressed. At 120 hr, it was found that Mira becomes delocalized from the cortex to the cytoplasm and nucleus of many interphase neuroblasts. In metaphase neuroblasts, Mira fails to localize to the basal side of the cortex, although it does selectively partition into one daughter during telophase. This late basal restoration resembles the 'telophase rescue' associated with several apical complex mutations. Pros, the Mira-binding transcription factor and GMC-determinant, is not detectable in the neuroblast nucleus at 96 hr, but at 120 hr a burst of Pros was observe in the nucleus of many Mira⁺ cells of intermediate size indicative of neuroblasts in the interphase preceding the terminal mitosis. Clones lacking Pros activity contain multiple Mira⁺ neuroblast-like cells. It was found that they do not respect the ~120 hr proliferation endpoint and even retain numerous dividing Mira⁺ progenitors into adulthood. In addition, GAL80^ts induction was used to induce transiently the expression of a YFP-Pros fusion protein well before the normal ~120 hr endpoint. Under these conditions, YFP-Pros can be observed in the nucleus of neuroblasts, most Mira⁺ progenitors disappear prematurely, and neural proliferation ceases much earlier than normal. Together, these results provide evidence that most neuroblasts terminate activity in the pupa via a nuclear burst of Pros that induces cell-cycle exit. These progenitors are referred to as type I neuroblasts to distinguish them from the much smaller population of type II neuroblasts that terminate via RHG-dependent apoptosis (Maurange, 2008).

Whether the postembryonic pulses of Cas and Svp in type I neuroblasts are implicated in scheduling their subsequent cell-cycle exit was tested. Remarkably, it was observed that many svp^e22 clones induced at early-larval stages retain a single Mira⁺ neuroblast at 7 days into adulthood. The persistent adult neuroblasts in a proportion of these clones also express the M phase marker, PH3, indicating that they remain engaged in the cell cycle and, accordingly, they generate approximately twice the normal number of cells by 3 days into adulthood. Furthermore, superficial last-born neurons in these over proliferating svp^e22 adult clones are Chinmo⁺ Br-C^- indicating a blocked Chinmo --> Br-C transition. Similar phenotypes are obtained in some UAS-cas and cas²⁴ clones. This analysis demonstrates that stalling the temporal series not only inhibits the late-larval switch to Br-C⁺ neuronal identity but also prevents the pupal cell-cycle exit of type I neuroblasts (Maurange, 2008).

To test the regulatory relationship between Pros and the temporal series, svp^e22 clones were examined at pupal stages. It was found that mutant interphase neuroblasts fail to switch on nuclear Pros at 120 hr, although svp^e22 GMCs express nuclear Pros as normal. This likely accounts for why adult clones lacking Svp retain only a single neuroblast, whereas those lacking Pros contain multiple neuroblast-like progenitors. Importantly, these results demonstrate that nuclear Pros acts downstream of the temporal series in type I neuroblasts. Together, the genetic and expression analyses of Svp and Pros show that the temporal series triggers a burst of nuclear Pros in type I neuroblasts, thus inducing their cell-cycle exit (Maurange, 2008).

To determine how the temporal series is linked to the cessation of progenitor divisions, two transcription factors expressed in neuroblasts in a temporally restricted manner were examined. Dichaete (D), a member of the SoxB family, is dynamically expressed in the early embryo and is required for neuroblast formation. Consistent with the previous studies, it was observed that most or all embryonic neuroblasts progress through a transient D⁺ phase, but those in the lateral column of the ventral nerve cord initiate expression after their medial and intermediate counterparts. D subsequently becomes repressed in ~85% of neuroblasts during late-embryonic and postembryonic stages. Grainyhead (Grh) is first activated in neuroblasts in the late embryo and is required for regulating their mitotic activity during larval stages (Almeida, 2005, Bray, 1989, Brody, 2000, Cenci, 2005). Blocking early temporal series progression in the embryo, either by persistent Hb or loss of Cas activity, prevents most neuroblasts from downregulating D and also from activating Grh at late-embryonic and postembryonic stages. As forcing premature Cas expression leads to precocious D repression and Grh activation, both factors are likely to be regulated by Cas rather than by a later member of the temporal series. These results demonstrate that transient embryonic Cas activity permanently switches the expression of Grh on and D off. They also identify Grh and D as positive and negative targets, respectively, of the temporal series in many neuroblasts (Maurange, 2008).

Loss of Grh activity in thoracic neuroblasts (here defined as type I neuroblasts) leads to their reduced cell-cycle speed and disappearance during larval stages. At 96 hr, it was observed that 65% of grh³⁷⁰ type I neuroblasts are smaller than normal (~6.4 μm in diameter), delocalize Mira from the cortex to the cytoplasm and nucleus, and strongly express Pros in the nucleus. These events, reminiscent of the 120 hr terminal cell cycle of wild-type progenitors, show that Grh is required to prevent the premature cell-cycle exit of type I neuroblasts. Given this finding, and that Cas is required to activate Grh, the question arises as to how some neuroblasts lacking embryonic Cas activity are able to continue dividing into adulthood. However, cas²⁴ neuroblasts persisting in adults all retain Grh, suggesting that they may derive from clones that lacked only the last of the two Cas pulses observed in some embryonic neuroblasts, perhaps retaining enough Cas activity to support Grh activation but not later progression of the temporal series (Maurange, 2008).

To determine if late temporal series inputs, after embryonic Cas, are also required to maintain the long-lasting postembryonic expression of Grh, svp^e22 clones were induced in early larvae. Although type I neuroblasts in these mutant clones have a stalled temporal series, they retain postembryonic Grh expression through to adult stages. Thus, two sequential inputs from the temporal series are required for type I neuroblasts to undergo timely Pros-dependent cell-cycle exit. First, embryonic Cas activity switches on sustained Grh expression, inhibiting premature nuclear Pros and permitting continued mitotic activity. Second, a late postembryonic input, requiring Svp, counteracts this activity of Grh by triggering a pupal burst of nuclear Pros (Maurange, 2008).

Despite undergoing premature cell-cycle exit, it was noticed that grh mutant neuroblasts can still generate both Chinmo⁺ and Br-C⁺ neurons. Thus, Grh is not required for the neuronal Chinmo --> Br-C switch. Conversely, neither Chinmo nor Br-C appears to be required postembryonically for neuroblast cell-cycle exit. In summary, the properties of both neuroblasts and neurons are regulated by downstream targets of the temporal series (Maurange, 2008).

Next whether the temporal series and its targets also function in type II neuroblasts, which terminate via RHG-dependent apoptosis rather than Pros-dependent cell-cycle exit, was tested. Focus was placed on one identified type II neuroblast in the central abdomen, called dl, which undergoes apoptosis at 70-75 hr and produces only a small postembryonic lineage of ~10 neurons. It was observed that the dl neuroblast expresses bursts of Cas (~45 to ~60 hr) and Svp (~62 to ~65 hr) and sequentially generates Chinmo⁺ (~7 deep) and Br-C⁺ (~3 superficial) neurons. Loss of Cas or Svp activity, or prevention of temporal series progression in several other ways all lead to a blocked Chinmo --> Br-C transition, a failure to die at 70-75 hr, and the subsequent generation of many supernumerary progeny. These results show that the temporal series performs similar functions in type I and type II neuroblast lineages, regulating both the Chinmo --> Br-C neuronal switch and the cessation of progenitor activity. Next, the mechanism linking the temporal series to RHG-dependent death of type II neuroblasts was tested. As for type I neuroblasts, dl progenitors in cas²⁴ clones, induced in the embryo, fail to activate Grh and repress D. Moreover, if grh activity is reduced, or if D is continuously misexpressed, dl progenitors persist long after 75 hr. Therefore, both of these early Cas-dependent events are essential for subsequent type-II neuroblast apoptosis. However, in contrast to members of the temporal series themselves, persistent misexpression of their target D does not block the Chinmo --> Br-C switch in the dl lineage. Thus, the D^- Grh⁺ state of type I and type II neuroblasts, installed via an embryonic Cas pulse, appears to be necessary for progenitor termination but not for Chinmo --> Br-C switching. Nevertheless, dl neuroblasts stalled at a postembryonic stage in svp^e22 and UAS-cas clones retain the D^- Grh⁺ code yet still fail to undergo apoptosis. Therefore, as with type I cell-cycle exit, timely type II apoptosis requires both embryonic Cas-dependent and postembryonic Svp-dependent inputs from the temporal series (Maurange, 2008).

Finally, the regulatory relationship between the temporal series and AbdA, a Hox protein transiently expressed in postembryonic type II (but not type I) neuroblasts and required for their apoptosis was dissected. dl neuroblasts lacking postembryonic Svp activity or persistently expressing Cas still retain AbdA expression yet do not die. This suggests that AbdA is unable to kill neuroblasts unless they progress, in a Svp-dependent manner, to a late Cas^- temporal state. To test this prediction directly, use was made of the previous finding that ectopic AbdA is sufficient to induce neuroblast apoptosis, albeit only within a late time window. Constitutive AbdA-induced apoptosis is efficiently suppressed by persistent Cas, not only in type II but also in type I neuroblasts. This result demonstrates that, in order to terminate, type II neuroblasts must progress to a late Cas^- state, thus acquiring a D^- Grh⁺ Cas^- AbdA⁺ code. It also suggests that AbdA is sufficient to intercept progression of the temporal series in type I neuroblasts, inducing an early type II-like termination (Maurange, 2008).

This study has found that the Drosophila CNS contains two distinct types of self-renewing progenitors: type I neuroblasts terminate divisions by cell-cycle withdrawal and type II neuroblasts via apoptosis. Despite these different exit strategies, both progenitor types use a similar molecular timer, the temporal series, to shut down proliferation and thus prevent CNS overgrowth. These findings demonstrate that the temporal series does considerably more than just modifying neurons; it also has multiple inputs into neural proliferation. The identification and analysis of several pan-lineage targets of the temporal series also begins to shed light on the mechanism by which developmental age modifies the properties of neuroblasts and neurons. Two targets, Chinmo and Br-C, are part of a downstream pathway temporally regulating the size and identity of neurons. Two other temporal series targets, Grh and D, function in neuroblasts to regulate Prospero/RHG activity, thereby setting the time at which proliferation ends. The temporal series regulates both cell proliferation and cell identity; a feedforward mechanism is proposed for generating combinatorial transcription factor codes during progenitor aging (Maurange, 2008).

This study has found that the temporal series regulates a widespread postembryonic switch in neuronal identity. Most, if not all, type I and type II neuroblasts first generate a deep layer of large Chinmo⁺ neurons and then switch to producing a superficial layer of small Br-C⁺ neurons. Two lines of evidence argue that this postembryonic neuronal switch is likely to be regulated by a continuation of the same temporal series controlling early/late neuronal identities in the embryo. First, the postembryonic Chinmo --> Br-C neuronal switch is promoted by the transient redeployment of two known components of the embryonic temporal series, Cas and Svp. Second, this switch remains inhibitable by misexpression of the other embryonic temporal factors such as Hb. Since both Cas and Svp are expressed somewhat earlier than the neuronal-size transition, it is likely that they promote bursts of later, as yet unknown, members of the temporal series that more directly regulate Chinmo and Br-C. Although neuronal functions for both BTB zinc-finger targets have yet to be characterized, a progressive early-to-late decrease in postmitotic Chinmo levels is known to regulate the temporal identities of mushroom-body neurons (Zhu, 2006). The current results now suggest that this postmitotic gradient mechanism may be linked to, rather than independent from, the temporal series (Maurange, 2008).

Type I neuroblasts in clones lacking postembryonic Cas/Svp activity or retaining an early temporal factor, fail to express nuclear Pros during pupal stages and thus continue dividing long into adulthood. These overproliferating adult clones each contain only a single neuroblast, sharply contrasting with adult clones lacking Brat or Pros, in which there are multiple neuroblast-like progenitors (Bello, 2006). Hence, manipulations of the temporal series and its progenitor targets offer the prospect of immortalizing neural precursors in a controlled manner, without disrupting their self-renewing asymmetric divisions (Maurange, 2008).

This study demonstrates that type I and type II neuroblasts must progress through at least two critical phases of the temporal series in order to acquire the D^- Grh⁺ Cas^- combinatorial transcription factor code that precedes Pros/RHG activation. The early phase corresponds to embryonic Cas activity switching neuroblasts from D⁺ Grh^- to D^- Grh⁺ status. The equally essential, but less well-defined, late postembryonic phase of the temporal series requires transition to a Cas^- state and a late Svp burst. For type I neuroblasts, Grh and a late Cas^- temporal identity are both required for timely expression of nuclear Pros and subsequent cell-cycle withdrawal. For type II neuroblasts, these two inputs are also necessary for RHG-dependent apoptosis, with the additional requirement that D must remain repressed. Although the temporal series and its targets are similarly expressed in type I and type II neuroblasts, only the latter progenitors undergo a larval burst of AbdA. This AbdA expression is likely to be the final event required to convert the D^- Grh⁺ Cas^- state, installed by the temporal series, into the D^- Grh⁺ Cas^- AbdA⁺ combinatorial code for RHG-dependent apoptosis. This code prevents type II neuroblasts in the abdomen from reaching the end of the temporal series and accounts for why they generate fewer progeny and terminate earlier than their type I counterparts in the central brain and thorax (Maurange, 2008).

The data in this study support an indirect feedforward model for neuroblast aging. Key to this model is the finding that, although members of the temporal series are only expressed very transiently, some of their targets can be activated or repressed in a sustained manner, as observed for Chinmo/Br-C in neurons and also for Grh/D in neuroblasts. In principle, this indirect feedforward allows aging progenitors to acquire step-wise the combinatorial transcription factor codes modulating cell-cycle speed, growth-factor dependence, competence states, and neural potential. Like Drosophila neuroblasts, isolated mammalian cortical progenitors can sequentially generate neuronal fates in the correct in vivo order (Shen, 2006). These studies suggest that it will be important to investigate whether the transcription factors controlling this process also regulate cortical proliferation and whether their targets include BTB-zinc finger, Grh, SoxB, Prox, or proapoptotic proteins. Some insect/mammalian parallels seem likely, since it is known that Sox2 downregulation and Prox1 upregulation can both promote the cell-cycle exit of certain types of vertebrate neural progenitors (Dyer, 2003, Graham, 2003). Thus, although insect and mammalian neural progenitors do not appear to use the same sequence of temporal transcription factors, at least some of the more downstream components identified in this study might be functionally conserved (Maurange, 2008).