Branchless and Hedgehog operate in a positive feedback loop to regulate the initiation of neuroblast division in the Drosophila larval brain

Abstract

The Drosophila central nervous system is produced by two rounds of neurogenesis: one during embryogenesis to form the larval brain and one during larval stages to form the adult central nervous system. Neurogenesis caused by the activation of neural stem division in the larval brain is essential for the proper patterning and functionality of the adult central nervous system. Initiation of neuroblast proliferation requires signaling by the Fibroblast Growth Factor homolog Branchless and by the Hedgehog growth factor. We show here that the Branchless and Hedgehog pathways form a positive feedback loop to regulate the onset of neuroblast division. This feedback loop is initiated during embryogenesis. Our genetic and molecular studies demonstrate that the absolute level of Branchless and Hedgehog signaling is critical to fully activate stem cell division. Furthermore, over-expression and mutant studies establish that signaling by Branchless is the crucial output of the feedback loop that stimulates neuroblast division and that Branchless signaling is necessary for initiating the division of all mitotically regulated neuroblasts in the brain lobes. These studies establish the molecular mechanism through which Branchless and Hedgehog signaling interface to regulate the activation of neural stem cell division.

Introduction

Stem cells are specialized precursor cells with virtually unlimited proliferation potential. Stem cells are defined as totipotent or multipotent cells that may be able to give rise to any of the tissues of the body (totipotency) or to any of the cell types in a given tissue (multipotency). The second defining characteristic of stem cells is that they undergo asymmetric division to yield a stem cell and a progenitor which gives rise to many terminally differentiated cell-types. Because of this self-renewing capacity, pools of stem cells exist throughout life in many organs. The power of self renewal is illustrated by the repopulation of the entire spectrum of blood cells in an irradiated mouse by a single transplanted hematopoietic stem cell [1]. Given the generative capacity of stem cells, it is clear that stem cell output must be closely regulated.

Changes in stem cell proliferation occur by cell cycle arrest/activation, or by dramatically altering the rate of cell division [2]. Although a number of growth factor signals will increase the rate of cell proliferation in vivo and in vitro, it is not always clear whether the increased proliferation observed within a population of cells is due to increased cell survival, increased maintenance of stem cell identity as opposed to differentiation into post-mitotic cell-types or initiation of stem cell division by cells previously cell-cycle arrested. Clear examples of the latter are surprisingly few. For example, primitive hematopoietic stem cells appear quiescent when isolated [3–5], yet will divide rapidly both in vivo and in vitro in response to growth factors [6, 7]. Adult mammalian neural stem cells were initially identified as mitotically quiescent cells from the mouse striatum or spinal cord whose division could be stimulated in vitro by the addition of Fibroblast Growth Factor (FGF) and/or Epidermal Growth Factor (EGF) to the culture medium [8, 9]. Upon growth factor stimulation, these neural precursors underwent rounds of proliferation that produced neuronal and glial progeny while maintaining stem-cell like properties. In vivo, the presence of relatively mitotically quiescent stem cells in the subependymal zone of the mouse forebrain was inferred through studies that ablated actively dividing cells without affecting the number of stem cells that could be isolated from the treated tissue [10]. Through these and other studies, it became clear that populations of mitotically quiescent neural stem cells were present in the adult mammalian brain. However, although studies were able to demonstrate that increased physical activity or environmental stimulation could increase neurogenesis [11, 12], detailed molecular mechanisms underlying the activation of cell division in quiescent mammalian neural stem cells have yet to be elucidated.

We and others have shown that subsets of Drosophila neuroblasts are mitotically quiescent upon larval hatching, but resume cell division with characteristic timing in vivo and in vitro ( [13–15], reviewed in [16]). Two growth factor signaling systems, Fibroblast Growth Factor (FGF) and Hedgehog (Hh) are necessary for the initiation of proliferation of neural stem cells, or neuroblasts, in the Drosophila larval brain [17]. In this system, Branchless (Bnl, a Drosophila FGF homolog) and Hh signaling are modulated by the proteoglycan Trol (the Drosophila Perlecan homolog). Decreased signaling by either Bnl or Hh results in fewer neuroblasts beginning cell division at late first instar. However, whether this is due to independent parallel pathways that target different subsets of neuroblasts or interaction between the two pathways to control all regulated neuroblast division is not yet known. In this study we have found evidence of a positive feedback loop between Hh and Bnl signaling in the larval brain. We demonstrate that hh expression and signaling depend on Bnl activity, and vice versa. Both hh and bnl expression are present in the larval brain lobes upon hatching, as are expression of the Hh and Bnl response genes patched (ptc) and pointed (pnt), respectively. Use of the temperature-sensitive hh allele hh^ts2 demonstrated that the Hh-Bnl feedback loop is initiated during embryogenesis. Further studies with the hh^ts2 allele showed that decreased Hh signaling up to the first five hours of larval life inhibited neuroblast proliferation, while decreased signaling after five hours had no effect on the numbers of S phase neuroblasts. At five hours post hatching expression of the Hh response gene ptc is observed in numerous cells in the brain, but not in neuroblasts. Expression of the Bnl response gene pnt is not detectable using lacZ expression at this time. Finally, epistatic and double mutant studies also support a positive feedback loop model with Bnl signaling as the output of the pathway that activates the division of all mitotically arrested neuroblasts in the brain lobe.

Materials and methods

Genetic strains and transgenes

Flies were grown in standard medium at 25°C, unless otherwise stated. Markers and balancer chromosomes are described in flybase. Due to the variability of genetic background and their effects on neuroblast division and gene expression, crosses were designed such that sibling controls could be used in all studies.

BrdU incorporation and neuroblast counting

5-bromodeoxyuridine (BrdU) incorporation was preformed by placing 1^st instar larvae on Kankel/White medium [18] containing 0.1mg/ml BrdU from 16–20 hours post-hatching (hph). The larvae were then dissected, fixed, and labeled. BrdU visualization was observed with a primary mouse anti-BrdU antibody 1:100 dilution (BD-Biosciences) and a goat anti-mouse horseradish peroxidase secondary 1:200 dilution (Jackson ImmunoResearch). Signaling was developed using diaminobenzidine (DAB) and mounted in PBST for visualizing on a Zeiss axiophot compound microscope. Neuroblast counting was preformed by visual examination. Proliferating neuroblasts were identified by morphology of their nucleus when labeled with BrdU.

Developmental staging

Larvae were developmentally synchronized by collecting newly hatched 1^st instar larvae in one hour windows. Drosophila development takes roughly twice as long at 18°C versus 25°C. Therefore for temperature shift studies we calculated 25°C “equivalent” times at 18°C as being twice that of the developmental period at 25°C. For example to obtain a 3–4 hour old larvae at 25°C “equivalent” we assayed 6–8 hours post hatching larvae at 18°C.

RNA isolation and Real Time PCR

For RNA isolation, total RNA was extracted using Trizol Reagent (Invitrogen) according to manufacturer’s protocol. For real time PCR, total RNA was DNase (Invitrogen) treated and reverse transcribed with SuperScript First Strand RT-PCR kit (Invitrogen) using oligo(dT) primers. SYBR Green (Applied Biosystems) was used to run the reactions on a BioRad iCycler. Each sample was run in triplicate at three different concentrations. All primer set sequences are available upon request.

Statistical analysis

Standard deviation for each sample group was calculated. The two-tailed Student’s t-test was used to determine the confidence limits between experimental and control groups. Equal variance was assumed in most sample comparisons. Where we observed sample values to be more broadly distributed than normal, unequal variance was used for analysis.

Results and Discussion

Hh pathway activity is necessary and sufficient to regulate bnl expression and signaling

To determine if Hh and Bnl act as independent pathways to activate neuroblast division in the larval brain, we evaluated bnl expression and pathway activity in brains with decreased Hh signaling. Our genetic studies had already demonstrated that the weak trol^b22 mutation results in decreased neuroblast proliferation when combined with a single copy of the null hh allele hh^AC [17]. Therefore we expected to observe reduced Hh signaling in trol^b22; hh^AC/+ brains. We used quantitative Real-Time PCR (qRT-PCR) to assay for expression of bnl and its target gene pointed (pnt) and to confirm decreased expression of hh and Hh signaling by monitoring the Hh target gene patched (ptc). Siblings with normal neuroblast proliferation levels were used as a control population. In animals hemizygous for trol^b22 and carrying the one copy of the null allele hh^AC (trol^b22; hh^AC/+), hh expression and signaling were significantly reduced, with a reduction in expression of both bnl and its response gene pnt (Fig. 1A, B, p < 0.01 in all cases). To eliminate the possibility that the alteration in bnl expression and signaling levels were due to the mutation in trol, we then asked if decreasing hh signaling using the temperature sensitive hh allele hh^ts2 also decreased bnl expression and signaling. Homozygous hh^ts2 animals were raised at the permissive temperature (18 °C) through embryogenesis and then transferred to the restrictive temperature (25 °C) upon larval hatching. Heterozygous hh^ts2 /+ animals treated to the same temperature regimen were used as controls. hh^ts2 homozygotes had a 75% drop in hh signaling and decreased bnl and pnt expression compared to controls (Fig. 1C, p < 0.01 in all cases). BrdU analysis showed that the hh^ts2 homozygotes also had decreased neuroblast proliferation (Fig. 1D, p < 1 × 10⁻¹⁶). These results indicate that the activity of the Hh pathway is necessary for normal levels of Bnl signaling.

Fig. 1. bnl expression and signaling respond to Hh pathway activity.

(A, C, E) Expression of hh, ptc, bnl, and pnt were quantified by qRT-PCR in 1^st instar larval brains at 19–20 hours post hatching (hph). Error bars indicate standard deviation. All samples were run in triplicate at three different concentrations. (A) in trol^b22; hh^AC/ + animals relative to trol^b22 control (C) in hh^ts2 homozygous animals relative to hh^ts2/ + heterozygotes both raised at 18°C during embryogenesis and moved to 25°C upon larval hatching (E) in hs-hh animals relative to hs-hh/ + heterozygotes both raised at 18°C during embryogenesis and moved to 25°C upon larval hatching. (B, D, F) S-phase neuroblasts/brain lobe were quantified and normalized to the average sibling control value.

We then asked whether increasing Hh signaling is sufficient to produce increased bnl expression and signaling activity. The inducible hs-hh allele was used to increase hh expression, heterozygous hs-hh animals were used as sibling controls. Brains from homozygous hs-hh animals had a significant increase in hh expression and increased Hh signaling (p < 9 × 10⁻³ for both) even over siblings with one copy of the inducible transgene (Fig. 1E). The increase in Hh signaling correlated with increased expression of bnl and pnt, as well as increased neuroblast proliferation (Fig. 1F, p < 9 × 10⁻³ for all cases). These studies demonstrate that bnl expression and activity in the larval brain is dependent on the level of Hh signaling.

Bnl pathway activity determines the level of hh expression and signaling

To ascertain if Hh signaling was similarly dependent on Bnl activity, we examined hh and ptc expression in animals with reduced Bnl signaling. We first studied animals hemizygous for trol^b22 and heterozygous for the putative null allele, bnl^P1. As expected, expression of both bnl and pnt dropped in the brains of these animals compared to controls (p < 2 × 10⁻³ for both), as did the number of BrdU labeled neuroblasts (p < 8 × 10⁻¹¹), although there was still some expression of the Bnl response gene pnt (Fig. 2A, B). Our qRT-PCR analysis show that hh and ptc message levels also decline in the trol^b22; bnl^P1 / + brains (Fig. 2A, p < 2 × 10⁻³ for both), suggesting that hh expression and signaling is dependent on Trol or Bnl activity. To establish whether the drop in hh expression and signaling were due to the mutation in trol or decreased Bnl signaling, we examined hh and ptc expression in the brains of animals homozygous for bnl^P1 but raised at 18°C to enable generation of mutant larvae. The fact that we could obtain bnl^P1 mutant larvae at 18°C but not 25°C as well as the detection of pnt expression in the bnl^P1 mutants suggests that bnl^P1 is not a true null. For this study we used bnl^P1 / + sibling animals subjected to the same temperature regimen. qRT-PCR studies confirmed a 70% drop in bnl and pnt message levels in bnl^P1 homozygotes versus control, as well as a dramatic decrease in hh and ptc expression (Fig. 2C, p < 0.01 for all). The number of BrdU labeled neuroblasts (4.4 ± 0.1) significantly decreased compared to controls (Fig 2D, p < 3 × 10⁻¹⁸). The observation that only 4–5 neuroblasts were BrdU labeled in bnl^P1 homozygotes at 20 hours post hatching suggests that all the mitotically regulated neuroblasts require Bnl signaling to begin cell division, although from this data a requirement for some minimum level of Hh signaling cannot be eliminated. This hypothesis will be addressed further in a following section. The neuroblasts labeled had cellular morphology and spatial positioning consistent with identification as mushroom body or ventral lateral neuroblasts. The four mushroom body neuroblasts and the single ventral lateral neuroblast in each brain lobe are the only neuroblasts in the larval brain that are dividing upon larval hatching and divide continuously through larval life [14, 19]. They are also the only neuroblasts in the brain lobe not affected by mutations in trol that decrease signaling by Bnl ([14, 17]. The results of the bnl mutant studies also indicate that normal levels of activity of the Bnl pathway are necessary for normal levels of Hh signaling.

Fig. 2. hh expression and signaling respond to Bnl pathway activity.

(A, C, E) Expression of hh, ptc, bnl, and pnt were quantified by qRT-PCR in 1^st instar larval brains at 19–20 hph. Error bars indicate standard deviation. All samples were run in triplicate at three different concentrations. (A) in trol^b22; bnl^P1/ + animals relative to trol^b22 control (C) in bnl^P1 homozygous animals raised at 18°C during embryogenesis and moved to 25°C upon larval hatching relative to wildtype controls (E) in UAS-bnl + / + hs-gal4 animals relative to UAS-bnl animals both raised at 18°C during embryogenesis and moved to 25°C upon larval hatching. (B, D, F) S-phase neuroblasts/brain lobe were quantified and normalized to the average sibling control value.

We then investigated whether increased Bnl signaling is sufficient to increase hh expression and signaling. Up regulation of Bnl signaling was accomplished by using a hs-GAL4 construct to drive expression of a UAS-bnl transgene. Animals were maintained at 18 °C to minimize expression from the hs promoter and activity of the GAL4 transcription factor during embryogenesis. Upon larval hatching, animals were transferred to 25 °C to induce bnl expression. Increased expression and activity of bnl were confirmed by qRT-PCR and correlated with increased expression of both hh and ptc (p < 0.01 for all) as well as increased numbers of BrdU labeled neuroblasts (Fig. 2E, F, p < 1 × 10⁻¹⁹). Taken together, these results indicate that hh expression and activity is dependent on the level of Bnl signaling in the larval brain.

Decreasing both Bnl and Hh signaling causes a further decrease in neuroblast proliferation

The question then arose as to whether the relative ratio of Hh and Bnl signaling was critical for activation of neuroblast division (i.e. hh^null / hh⁺: bnl^null / bnl⁺ would work as well as hh⁺ / hh⁺: bnl⁺ / bnl⁺), or if the absolute level of signaling (compared to wildtype controls) is important. To address this issue, we evaluated the amount of neuroblast division in trol^b22 mutant animals heterozygous for both hh^AC and bnl^P1 and compared it to that in trol^b22, trol^b22; hh^AC/+ and trol^b22; bnl^P1 / + animals (Fig. 3A). Statistical analysis showed that the decrease in numbers of BrdU labeled neuroblasts in the trol^b22; bnl^P1 + / + hh^AC brains was significantly more than in either single heterozygote (p < 9 × 10⁻¹¹). qRT-PCR showed roughly equivalent decreases in expression of both the ligands and their target genes (Fig. 3B). The increased severity of the neuroblast proliferation phenotype in the double hh bnl heterozygote indicates that maintenance of the overall magnitude of Bnl and Hh signaling is essential for normal neuroblast proliferation.

Fig. 3. Maintenance of both Hh and Bnl signaling is required for normal neuroblast proliferation.

(A) S-phase neuroblasts/brain lobe were quantified and normalized to the average sibling control value. Expression of hh, ptc, bnl, and pnt were quantified by qRT-PCR in 1^st instar larval brains at 19–20 hph (B) in trol^b22; bnl^P1 +/ + hh^AC animals relative to trol^b22 control. Error bars indicate standard deviation. All samples were run in triplicate at three different concentrations.

The initiation of the Hh-Bnl feedback loop observed in the larval brain occurs during embryogenesis

Thus far, our studies are consistent with a positive feedback loop between Hh and Bnl that regulates the level of growth factor expression similar to that observed in the chick limb bud. The output of this loop then regulates the activation of neuroblast proliferation. Since Hh- and Bnl-dependent neuroblast proliferation does not begin until 8–10 hours post hatching, we asked when Hh and Bnl are first expressed in the larval brain. Larvae were collected in one hour increments and the amount of hh and bnl message in the larval brain evaluated. Both hh and bnl are expressed upon larval hatching, and the level of expression does not significantly change during the first four hours of larval life (Fig. 4A). We then asked if the level of Hh or Bnl signaling activity also remained constant during the first few hours of larval life. ptc and pnt show much more dynamic temporal pattern of expression than the level of pathway ligand (Fig. 4B). Since both Hh and Bnl are stimulating expression of their response genes within an hour of larval hatching, the feedback loop could be initiated during embryogenesis, prior to larval hatching.

Fig. 4. Initiation of the Hh-Bnl feedback loop occurs during embryogenesis.

Canton-S larval brains were collected in 1 hr increments from 0–20 hph at 25°C and qRT-PCR used to quantify (A) hh and bnl expression (B) ptc and pnt expression. Expression of hh, ptc, bnl, and pnt were quantified by qRT-PCR in (C) hh^ts2 homozygous larval brains from 0–1 hph raised at 25°C throughout embryogenesis compared to larval brains from 0–1 hph raised at 18°C throughout embryogenesis. Error bars indicate standard deviation. All samples were run in triplicate at three different concentrations.

If the Hh-Bnl feedback loop is initiated during embryogenesis, then decreasing Hh signaling only during embryogenesis should result in lowered bnl expression and Bnl signaling in the larval brain immediately upon hatching. To test this hypothesis, we used the temperature-sensitive hh allele, hh^ts2, and raised mutant animals at the restrictive (25°C) or permissive (18°C) temperature throughout embryogenesis. Both experimental and control plates were then placed at the permissive temperature (18°C) and newly hatched larvae collected in a two hour window. This experimental design resulted in decreased Hh signaling only during embryogenesis in the experimental samples. Larval brains were dissected and assayed for levels of hh, ptc, bnl and pnt expression. Our qRT-PCR results show that the levels of both bnl and its response gene pnt decreased in the larval brains of hh^ts2 animals raised at the restrictive temperature compared to animals raised at the permissive temperature (p < 0.01 for all). The expected decrease in expression of hh and the Hh response gene ptc in experimental samples compared to controls was verified by qRT-PCR (Fig. 4C). This result demonstrates that lowering Hh signaling during embryogenesis results in decreased Bnl production and signaling and is consistent with the hypothesis that the Hh-Bnl feedback loop in the brain is already operational by larval hatching.

Bnl is epistatic to Hh for the proliferation of regulated neuroblasts in the larval brain lobe

So far, our results show that Bnl and Hh signal in a positive feedback loop that is initiated prior to larval hatching. We next asked if the activity of both signaling pathways was necessary for neuroblast proliferation or if over-expression of one signal could rescue a deficit in the other signal. The inducible hs-hh allele was used to increase hh expression in a bnl^P1 homozygous mutant. Animals were maintained at the 18°C to minimize expression from the hs promoter and activity of the GAL4 transcription factor during embryogenesis. Upon larval hatching, animals were transferred to 25°C to induce hh expression. Increased expression and activity of hh were confirmed by qRT-PCR (p < 1 × 10⁻⁵), and bnl expression and activity mirrored that of bnl^P1 homozygotes, indicating that mis-expression of hh does not bypass or suppress the bnl mutant phenotype (Fig. 5A). Note that since we are comparing hs-hh; bnl^P1/bnl^P1 animals to bnl^P1/bnl^P1 controls the expression level of bnl is unchanged upon expression of hh (a value of 1.0) and the level of the Bnl response gene pnt actually decreases slightly. More strikingly, over-expression of hh does not rescue neuroblast proliferation in the bnl^P1 mutant (Fig. 5B). As noted previously when analyzing bnl^P1 homozygotes, we observed only 4.4 ± 0.1 neuroblasts labeled with BrdU in hs-hh ; bnl^P1 samples (indicative of the mushroom body and ventral lateral neuroblasts, which are not affected by Hh or Bnl signaling), compared to approximately 20 observed in normal controls and the 30–35 observed in hs-hh animals alone. The failure of hh over-expression to overcome the effects of a bnl null mutation confirms the hypothesis that all the mitotically regulated neuroblasts require Bnl signaling to initiate cell division. This data also suggests that Bnl activity is the signaling output of the positive feedback loop. To confirm this conclusion, up-regulation of Bnl using the same hs-GAL4/UAS-bnl expression system described previously was examined in an hh^ts2 homozygous animal. Animals were again maintained at the 18°C to minimize expression from the hs promoter and activity of the GAL4 transcription factor during embryogenesis. Upon larval hatching, animals were transferred to 25°C to induce bnl expression. Increased expression and activity of bnl were confirmed by qRT-PCR (Fig. 5C, p < 1 × 10⁻³ for both). BrdU incorporation studies demonstrated that over-expression of bnl in a hh^ts2 homozygote resulted in significant increase (p < 8.7×10⁻⁵) in the number of S phase neuroblasts over that normally observed in hh^ts2 homozygotes at the restrictive temperature (Fig. 5D, 14.8 BrdU labeled neuroblasts/brain lobe ± 0.3). In fact, over-expression of bnl in an hh^ts2 homozygote produced an over-proliferation phenotype (29.4 BrdU labeled neuroblasts/brain lobe ± 0.2) compared to the normal 20 neuroblasts/brain lobe observed in wild-type controls. Over-expression of bnl in an hh^ts2 homozygote also resulted in higher levels of hh expression and signaling activity as indicated by ptc expression (Fig. 5C, p < 1 × 10⁻³). This is likely due to the fact that the hh^ts2 allele is not a functional null at the restrictive temperature, and thus is capable of a low level of signaling and participation in the Hh-Bnl feedback loop. Participation in the feedback loop is indicated by values of hh and ptc expression significantly greater than 1.0 (p < 1×10⁻³ for both) when comparing expression of hh and ptc in hs-GAL4/UAS-bnl; hh^ts2 animals compared to hh^ts2 controls. However, the hh^ts2 allele still impairs hh function and decreases feedback, thus decreasing the amount of Bnl produced in hs-GAL4/UAS-bnl; hh^ts2 animals (bnl expression in hs-GAL4/UAS-bnl animals is 2.4 ± 0.4 fold higher than in hs-GAL4/UAS-bnl; hh^ts2 animals while Hh signaling activity as monitored by ptc is 3.8 ± 0.8 fold higher). Therefore we expect the number of BrdU labeled neuroblasts in hs-GAL4/UAS-bnl animals to be higher than in hs-GAL4/UAS-bnl; hh^ts2 animals, as is observed (34.6 ± 0.7 versus 29.4 ± 0.2). Taken together, these studies indicate that over-expression of hh cannot overcome a null mutation in bnl to rescue neuroblast proliferation, but over-expression of bnl can overcome a mutation in hh to rescue neuroblast proliferation. These results establish that Bnl signaling activity is the essential output of the Hh-Bnl feedback loop that regulates the activation of neuroblast proliferation.

Fig. 5. Bnl is epistatic to Hh for activation of proliferation in the regulated neuroblasts of the larval brain lobe.

(A, C) Expression of hh, ptc, bnl, and pnt were quantified by qRT-PCR in 1^st instar larval brains at 19–20 hph. Error bars indicate standard deviation. All samples were run in triplicate at three different concentrations. (A) in hs-hh; bnl^P1 animals relative to bnl^P1 control both raised at 18°C during embryogenesis and moved to 25°C upon larval hatching (C) in UAS-bnl hh^ts2/ hs-gal4 hh^ts2 animals relative to UAS-bnl hh^ts2/ hh^ts2 control both raised at 18°C during embryogenesis and moved to 25°C upon larval hatching. (B, D) S-phase neuroblasts/brain lobe were quantified and normalized to the average sibling control value.

Early signaling by Hh is required to stimulate normal neuroblast division

Our studies show that the Hh-Bnl feedback loop is established during embryogenesis to trigger neuroblast proliferation during first instar. We next wanted to ask when signaling by the feedback loop is necessary in order to obtain normal neuroblast proliferation. To address this question, we designed a series of temperature shift experiments with the hh^ts2 allele to determine when normal levels of Hh activity are required to produce normal neuroblast proliferation (Fig. 6). Animals homozygous for hh^ts2 were raised at the permissive temperature of 18°C through embryogenesis and then shifted to the restrictive temperature of 25°C at different times during first instar. Times for temperature shift and BrdU incorporation were calculated such that shifts took place at the equivalent of one hour timepoints post hatching at 25°C and labeling was carried out at the equivalent of 16–20 hours post hatching (hph) 25°C (see Materials and Methods for details). Animals heterozygous for hh^ts2 that underwent the same temperature shift regimen were used as controls. When animals were shifted to the restrictive temperature upon larval hatching, hh^ts2 homozygotes showed decreased neuroblast proliferation at 16–20 hph (Fig. 6A, B, p < 1 × 10⁻¹⁵). This remained true through shifts at the equivalent of 1–2 hph, 2–3 hph, 3–4 hph (not shown) and 4–5 hph (Fig. 6C, D, p < 1 × 10⁻⁵ for all). However, when samples were shifted to the restrictive temperature at the equivalent of 5–6 hph, the number of BrdU labeled neuroblasts at the equivalent of 16–20 hph are indistinguishable between experimental and control animals (Fig. 6E, F, p < 0.4). Thus in order to exhibit normal levels of neuroblast proliferation, animals had to be subjected to normal levels of Hh signaling (and thus normal function of the Hh-Bnl feedback loop) until 5–6 hph, approximately 10 hours prior to BrdU assay.

Fig. 6. Normal Hh signaling is required until 5–6 hours post hatching to obtain normal neuroblast proliferation.

The number of BrdU labeled neuroblasts at 16–20 hours post hatching 25°C “equivalent” were quantified in (A) homozygous hh^ts2 animals or (B) control hh^ts2/ + heterozygotes raised at 18°C during embryogenesis and moved to 25°C upon larval hatching; (C) homozygous hh^ts2 animals or (D) control hh^ts2/ + heterozygotes raised at 18°C from embryogenesis through 4–5 hours post hatching 25°C “equivalent”, then shifted to 25°C; and (E) homozygous hh^ts2 animals or (F) control hh^ts2/ + heterozygotes raised at 18°C from embryogenesis through 5–6 hours post hatching 25°C “equivalent”, then shifted to 25°C.

We then asked where Hh and Bnl signaling is active at 4–5 hours post hatching to determine which cells are responding to the Hh-Bnl loop to activate neuroblast division. We followed Bnl activity using a pnt-lacZ line (Fig. 7A, C) and Hh signaling activity using a ptc-lacZ line (Fig. 7B, D). pnt-lacZ expression was not detectable within the larval brain (Fig. 7C), while expression of ptc-lacZ was visible in numerous cells at 5 hours post hatching (Fig. 7D). The spatial pattern of expression of β–galactosidase staining did not change appreciably between 0–1 and 4–5 hour post hatching for either pnt-lacZ or ptc-lacZ (data not shown). Close inspection of the nuclear size and morphology of the lacZ expressing cells indicate that the cells responding to Hh or Bnl signaling are not neuroblasts. This was confirmed by double labeling with BrdU in ptc-lacZ samples at 16–20 hph (Fig. B’). The combined histochemistry data suggest that neuroblasts respond indirectly to the ongoing activity of the Hh-Bnl feedback loop at a specific time in development.

Fig. 7. pnt and ptc expression in the larval brain.

(A) pnt-lacZ expression in a 1^st instar larval brain at 20 hph. (B) ptc-lacZ expression in a 16–20 hr BrdU labeled brain at 20 hph, inset (B’) shows close up of brain lobe in panel B where the arrowhead indicates a BrdU labeled neuroblast and the arrow indicates a lacZ positive cell. Expression of (C) pnt-lacZ and (D) ptc-lacZ in a first instar brain at 5 hph.

Integration of FGF and Hh signaling

Our studies have demonstrated that Hh and Bnl act in a positive feedback loop in the larval brain to control the onset of neuroblast proliferation (Fig. 8). The feedback loop acts at the transcriptional level, such that Hh signaling activity is essential to control the level of bnl expression and vice versa. Our double mutant analyses showed that an absolute level of signaling by both Bnl and Hh are required to maintain normal neuroblast activation, rather than other possible models that would suggest a certain balance of signaling activity (for example more Bnl than Hh) is sufficient regardless of the exact magnitude of signaling activity. The discovery that Bnl signaling is the critical output of the feedback loop suggests that the main function of Hh signaling is to maintain the proper level of Bnl production and signaling. Furthermore, the observation that only the mushroom body and ventral lateral neuroblasts continue to divide in bnl null mutants regardless of the level of Hh signaling indicates that all the regulated neuroblasts, both optic lobe and central brain sets, require the input of the Bnl pathway to enter S phase. Thus the Hh-Bnl feedback loop appears to control cell cycle progression in all the mitotically arrested neuroblasts that begin cell division in first instar.

Fig. 8. Model of Bnl-Hh positive feedback loop.

Other developmental events that require Hedgehog and FGF signaling have used those pathways in different manners to achieve their goals. For example, in the mouse ventral telencephalon, Hedgehog and FGF/MAPK signaling operate as two independent pathways. FGF signaling is independent of Sonic Hedgehog (SHH) and does not affect expression of either SHH itself or its target gene and effector GLI1 [20]. Other systems have shown a linear dependence of FGF expression on SHH signaling and vice versa. During budding morphogenesis in the mouse lung Hedgehog signaling inhibits expression of FGF10 but up-regulates FGF7 [21–23]. In the Xenopus eye, expression of Banded Hedgehog increases expression of FGF8 [24]. In the zebrafish forebrain inhibition of Hh signaling decreases expression of FGF3, FGF8 and FGF19 [25]. Hedgehog also regulates FGF expression in the zebrafish mid/hindbrain [26]. However, in the zebrafish forebrain HH expression requires FGF signaling. Inhibition of both FGF3 and FGF8 expression resulted in a downregulation of SHH [27]. Alternatively, the HH and FGF pathways can integrate at the level of intracellular components. FGF has been shown to induce expression of GLI2, a transcription factor and HH signaling effector in ventroposterior development in zebrafish [28].

Of course the classic example of FGF and SHH interplay is the development of the chick limb bud [29]. In this system, several FGFs set up a signaling center at the tip of the bud that turns on expression of SHH in the posterior limb mesenchyme. In turn, SHH signaling is required for maintenance of FGF4, FGF9 and FGF17 expression in the bud tip. This function of SHH occurs through the expression of Gremlin, an inhibitor of Bone Morphogenetic Protein signaling [30]. Gremlin inhibition of Bone Morphogenetic Protein signaling prevents down-regulation of the FGFs. Thus a positive feedback loop exists between SHH and FGFs, mediated by Gremlin.

The model of the Hh-Bnl feedback loop proposed here is most similar to the classic SHH-FGF feedback loop described in the vertebrate limb bud. We do not yet know whether the regulation of bnl expression by Hh signaling is direct or if it is mediated by another signaling pathway such as the Gremlin/Bone Morphogenetic Protein connection that operates in the limb bud [31, 32]. However, we have already shown that like the distinct domains of FGF and SHH in the limb bud [29], bnl and hh expression also occur in distinct regions of the brain lobe [17]. The fact that the Hh-Bnl feedback loop is activated during embryogenesis, but that the first regulated neuroblasts do not enter S phase until 8–10 hours after larval hatching [14, 18, 33, 34] also suggests that additional events must take place downstream of Bnl signaling to permit mitotically arrested stem cells to transit through G1 to S phase [13, 35]. One such possibility is exposure to the steroid hormone ecdysone, which is necessary during first larval instar for the initiation of neuroblast division a few hours later [15]. Both SHH and FGF2 have been shown to be necessary for the division of different subsets of neural stem cells in many different vertebrate and mammalian models and in multiple contexts [36, 37]. This is the first time that the interactions between these two pathways necessary to stimulate the reactivation of stem cell division in quiescent neural stem cells have been elucidated. The next challenge will be to determine whether different molecular mechanisms tying these two signaling pathways are used for different developmental decisions such as progeny cell fate, initiation of cell division and maintenance of stem cell identity.