Prothoracicotropic hormone modulates environmental adaptive plasticity through the control of developmental timing

During animal development, tissue growth and physiological integration of tissue function must be timed appropriately to maximize an organism's chances of survival and reproduction. In holometabolous insects, tissue growth is confined to the larval phases of development, which are punctuated by periodic shedding and resynthesis of cuticle (molts) to accommodate an increasing body size. In Drosophila melanogaster, embryonic development lasts ∼24 h, as does each of the first and second larval instar stages. During the final, third larval instar, which lasts ∼2-3 days, a great deal of growth occurs and body size becomes fixed in response to various neuroendocrine and hormonal signals that trigger metamorphosis, a process that transforms the larval body into that of the sexually mature adult fly (Mirth and Riddiford, 2007; Nijhout et al., 2014; Rewitz et al., 2013).

Both genetic and environmental factors influence larval growth rate and timing parameters (Danielsen et al., 2013; Di Cara and King-Jones, 2013; Rewitz et al., 2013). For instance, nutritional content of the food source and population density impact many life history traits resulting in numerous trade-offs that must be balanced for optimal survival (Barker and Podger, 1970; May et al., 2015; Miller and Thomas, 1958; Prout and Mcchesney, 1985). In its natural habitat, D. melanogaster utilizes rotting fruit, a short-lived food source, as its main dietary resource. As might be expected for such an ephemeral habitat, recent studies have shown that Drosophila larvae make nutritional choices that minimize developmental time. This choice impacts other life history traits such as body size, ovariole number (fecundity), viability and life span (Chippindale et al., 1993; Horvath and Kalinka, 2016; Kolss et al., 2009; Mendes and Mirth, 2016; Rodrigues et al., 2015). Despite the importance of balancing various life history traits for robust survival, the genetic mechanisms that adjust insect developmental time in response to various environmental factors are not well understood (Mirth and Shingleton, 2012; Tennessen and Thummel, 2011).

Both molting and metamorphosis require pulsed production and release of the steroid hormone ecdysone (E) from the larval prothoracic gland (PG), and classic ligature and transplantation experiments in Lepidoptera demonstrated that at least one brain-derived factor regulates E production in the PG (Kopec, 1922; Williams, 1952). Subsequent purification of the E-inducing substance from Bombyx heads identified prothoracicotropic hormone (PTTH) as a key ecdysteroidogenic inducer (Kawakami et al., 1990). PTTH is a secreted homodimer derived by processing of a precursor pro-protein and is a member of the cystine knot family of growth factors (Noguti et al., 1995). Genetic and biophysical studies have shown that it signals through the receptor tyrosine kinase Torso, which then activates a Ras/Raf/Erk MAP kinase signal transduction cascade (Jenni et al., 2015; Konogami et al., 2016; Rewitz et al., 2009). The targets of this signaling pathway are not fully identified, but are likely to include several of the E biosynthetic enzymes produced in the PG (Gibbens et al., 2011; Niwa et al., 2005).

In Drosophila, PTTH is produced by two bilateral neuroendocrine cells located in each brain hemisphere (termed PG neurons, in accordance with their innervation pattern) that project their axons contralaterally to terminate on the PG (McBrayer et al., 2007). These observations suggest that PTTH acts like a neurotransmitter for stimulating ecdysteroidogenesis rather than as a systemic hormone. However, Drosophila PTTH has a second role as a modulator of larval light sensitivity. This process involves the release of PTTH into circulation and its systemic delivery to larval light-sensing organs where it enhances larval light avoidance (Yamanaka et al., 2013b).

Genetic ablation or misspecification of the PG neurons revealed that these cells control developmental timing and body size (Ghosh et al., 2010; McBrayer et al., 2007). Loss of the PG neurons resulted in substantially delayed pupariation (by 4-6 additional days) and the production of large flies. Delayed pupariation resulted from a belated surge in the 20-hydroxyecdysone (20E) titer during the wandering stage, and both the large body size and developmental delay could be rescued by feeding 20E to third instar larvae. Although these observations suggested that PTTH is likely to be important for proper timing of the normal metamorphic transition, its exact role could not be determined since ablation/misspecification of the PG neurons would eliminate other potential factors, in addition to PTTH, that might influence endocrine activity and the timing of developmental transitions.

To unambiguously determine the role of PTTH in modulating developmental timing in response to various environmental conditions, we produced Ptth null mutants using both TALEN and gene replacement technologies. We show that these mutants produce a variable developmental delay that is influenced by both nutrient availability and population density. Careful analysis reveals that mutant animals accumulate mass at wild-type rates and eventually produce larger adults due to a longer growth phase. We find that PTTH function is required throughout the third instar larval stage since the developmental delay extends both the time to reach critical weight – a crucial nutrition-dependent developmental checkpoint (reviewed by Mirth and Riddiford, 2007) – as well as the nutrition-independent terminal growth period (Shingleton et al., 2007). Our experiments also indicate that these phenotypes result from an alteration in the kinetics of E biosynthesis. Furthermore, we demonstrate that PTTH provides animals with better adaptability to challenging environmental conditions (crowding, nutritional limitation). Lastly, manipulation of PG neuron activity demonstrates that PTTH acts in conjunction with another, as yet unknown, factor(s) to properly time larval development. Taken together, our results indicate that PTTH is a central component of a mechanism that coordinates trade-offs in several life history traits in response to environmental inputs such as nutrition and population density.

To genetically assess Ptth function, we generated three mutants: Ptth^Delta, Ptth^120F2A and Ptth^8K1J (Materials and Methods, Fig. 1A). The deletion mutant Ptth^Delta removes the start codon and is a protein null, as shown by complete absence of immunostaining signal (Fig. S1A). By contrast, Ptth^120F2A and Ptth^8K1J are short deletions in the final exon (7 bp and 6 bp, respectively) and yield either a truncated protein or a mature protein that is shortened by two amino acids. The control strain for all Ptth^Delta studies was w^DAH, while the control for Ptth^120F2A and Ptth^8K1J was w¹¹¹⁸ (see Materials and Methods). We determined the ability of the three mutants to undergo metamorphosis by precisely timing the developmental profiles from L1 ecdysis to pupariation. As shown in Fig. 1B, larvae lacking PTTH undergo metamorphosis with a 1 day delay compared with their respective controls. Ptth^8K1J exhibits the same 1 day delay as the other Ptth mutants, suggesting that all three alleles yield a PTTH product that is non-functional. Additionally, we consider the Ptth^120F2A allele to be a genetic null since an identical 1 day delay is observed in transheterozygous (Ptth^120F2A/Ptth^Delta) animals (Fig. S1B). Since Ptth mutants show almost no effect on size and timing before third instar (data not shown), this 1 day delay occurs in the third instar larval stage. The developmental delay observed in Ptth mutants can be rescued by adding a genomic transgene containing wild-type Ptth into the Ptth^120F2A background or by using tub-Gal4 to ubiquitously express a UAS-Ptth transgene in the Ptth^Delta background (Fig. 1C). These results imply a role for PTTH in precisely timing the transition from juvenile to adult.

Male and female Ptth mutant pupae are larger than their respective controls (Fig. 1D,D′), and Ptth mutant adults show significant weight increase (Fig. 1E) indicating that the extended larval growth period of these animals increases final adult size. Consistent with these results, Ptth mutant larvae gained weight at the same rate as control larvae (Fig. S2A). Accordingly, the size of brains, fat body nuclei and salivary gland nuclei at 24 h after the L2-to-L3 transition was comparable in Ptth^Delta mutant and control animals (Fig. S2B-D). Overall, this indicates that PTTH determines final adult body size by controlling the duration of the larval growth period without affecting the bulk rate of mass accumulation in the third instar larva.

Since female body size correlates with ovary size in Drosophila (David, 1970; Bergland et al., 2008), we counted ovariole number in Ptth^Delta and control females and observed a significant increase in ovariole number (by 8.5%) in the mutant compared with the control (Fig. 1F). This indicates that, by timing the end of the larval period and the onset of metamorphosis, PTTH also influences adult physiology, such as egg production and fecundity.

Given the role of E in controlling growth and differentiation of imaginal tissues (Gokhale et al., 2016; Hackney et al., 2012; Herboso et al., 2015; Jaszczak et al., 2016), we examined whether loss of Ptth could affect the rate of imaginal disc growth. We measured wing imaginal disc size at 15, 24 and 46 h after L3 ecdysis. As previously found in mutants with reduced E levels (Parker and Shingleton, 2011), Ptth^Delta mutant discs grew significantly more slowly than controls (Fig. 2A). We did not observe any difference in cell size between the two genotypes, indicating that lack of PTTH leads to smaller discs with fewer cells (Fig. 2B). However, owing to an extended larval growth period, Ptth^Delta wing discs are 14% larger at the onset of metamorphosis and give rise to larger adult wings (Fig. 2C,D). Thus, PTTH has a dual function in regulating both imaginal tissue growth rate and larval growth duration.

Drosophila larvae pass through two developmental checkpoints during the final instar: the critical (CW) and the minimal viable weight (MVW) thresholds (Mirth and Riddiford, 2007). In third instar larvae, CW and MVW are reached concomitantly ∼8 h (on rich food) after L2/L3 ecdysis at 0.84 mg/larva (Fig. 3A). It is noteworthy that our determination of CW for wild-type w¹¹¹⁸ is nearly the same as that reported previously (McBrayer et al., 2007; Mirth et al., 2005) and virtually identical to that determined by Koyama et al. (2014) using the breakpoint method (Stieper et al., 2008). The CW for Ptth^120F2A larvae increased to 1.9 mg/larva and occurred ∼20 h after the L2/L3 transition (Fig. 3B). Interestingly, in these mutants, the MVW occurred slightly earlier than the CW and at a lighter weight (1.55-1.69 mg/larva; Fig. 3B). We also determined the length of the terminal growth period (TGP), defined as the time interval that transpires after attainment of CW to when 50% pupariation is achieved (Shingleton et al., 2007). This period was prolonged from 35 h for w¹¹¹⁸ to 47 h for Ptth^120F2A (Fig. 3C). Taken together, these results suggest that 12 h of the 24 h developmental delay in Ptth mutants occurs prior to achievement of CW and that the remainder occurs during the TGP.

Next, we examined the transcriptional timing of three E-responsive genes. Expression of the mid-L3 20E target gene Sgs3 (Fig. S3A) (Warren et al., 2006) was first detected at 21 h in wild-type larvae and rose rapidly over the next 15 h. In Ptth^120F2A, Sgs3 expression was first detected at 24 h (4 h after CW) at a very low level, increased gradually for 1 day (the mid-L3 stage), and then rose abruptly at 48 h. In both cases, white prepupae first appeared 6-8 h after peak Sgs3 expression (36 h for w¹¹¹⁸ and 59 h for Ptth^120F2A). We also examined the expression of Atet and E23, two putative E transporters that are late L3 20E targets (Hock et al., 2000; Yamanaka et al., 2015) and observed similar delays in their induction (Fig. S3B,C). Taken together, these results indicate that PTTH is likely to control E levels throughout the entire L3 stage.

Since PTTH was originally described as a factor stimulating E production in the PG (Kataoka et al., 1991; Kawakami et al., 1990), we expected that the developmental delay observed in Ptth mutants would be the result of altered kinetics in E production. Consistent with this view, feeding 20E to Ptth mutant larvae rescued the developmental delay (Fig. S1C). To more directly address the timing of E production, we determined ecdysteroid titers in staged w¹¹¹⁸ and Ptth^120F2A L3 larvae (Fig. 3D). In control larvae, ecdysteroid titers rise to a first plateau of 6-8 pg at the time of CW (8-10 h after the L2-to-L3 transition), followed by a second jump at 18-20 h and then a steady rise to a peak at wandering stage. By contrast, Ptth^120F2A mutants have lower basal ecdysteroid titers that reach a titer of ∼6 pg/animal at 20 h after L2/L3 ecdysis, which correlates with the attainment of CW in these mutant larvae. The titer then rises at a slow rate, but eventually reaches a level similar to that of the control by the time of wandering (Fig. 3D). These results demonstrate that PTTH regulates the ecdysteroid production rate during the third instar larval stage to properly time metamorphosis.

We next analyzed the transcriptional timeline of seven E biosynthetic enzymes during the third instar stage of w¹¹¹⁸ and Ptth^120F2A (Fig. S4). The profiles of nvd and spok were similar, exhibiting a delayed rise in the mutant background, achieving only 50-65% of the w¹¹¹⁸ peak level just prior to wandering (Fig. S4A,B). shd and phm displayed a similar slow rise in expression in the Ptth mutant background but reached approximately the same peak levels as the control (Fig. S4E,F). By contrast, although expression of dib also showed a slow rise, it was induced to less than 50% of wild-type levels (Fig. S4C).

sro and Cyp6t3 revealed noticeably different transcription profiles compared with the control. In the Ptth mutant, sro expression peaked in mid-L3 at approximately twice the wild-type level and remained high during the entire extended developmental period before declining in early pupae (Fig. S4G). The expression profile of Cyp6t3 was drastically reduced on normal food, but remained near normal levels on richer media (Fig. S4D; data not shown). Thus, expression of the E biosynthetic genes is not coordinately regulated in the Ptth mutant since transcripts are either upregulated, unchanged, downregulated or dependent on nutrition relative to wild type.

It has previously been suggested that the ‘black box’ step involving the enzymes Spo/Spok, Sro and Cyp6t3 might be rate-limiting during E biosynthesis (Lafont et al., 2005; Ou et al., 2011; Rewitz et al., 2009). To examine if alterations in the expression profiles of a single enzyme were responsible for the developmental delay, we individually overexpressed some of these genes (spok, Cyp6t3), as well as dib and nvd using a PG-specific driver (p0206-Gal4) in a Ptth mutant background, but observed no rescue of developmental delay induced by Ptth loss (Fig. S4H-K). Assuming that all of the UAS constructs provide high-level overexpression, these data suggest that no individual enzyme is rate-limiting as a result of Ptth loss. Rather, it is likely that the partial reduction in several enzymes, an as yet unidentified component, or another regulated step is responsible for the slower kinetics of E production and the developmental delay.

We have previously suggested that PTTH-expressing neurons provide a growth signal to the PG, since hypomorphic mutations in the giant gene induce stochastic unilateral innervation of the PG resulting in an asymmetrically sized PG in which the innervated side is significantly larger than the non-innervated portion of the gland (Ghosh et al., 2010). To determine whether PTTH is the trophic signal, we examined PG nuclear size of innervated and non-innervated PG lobes derived from Ptth^120FA2, giant^E1 double-mutant larvae. Whereas the nuclei of innervated lobes were ∼33% larger in giant^E1 PGs, the innervated lobes of Ptth^120FA2, giant^E1 double mutants displayed no difference in the size of nuclei compared with non-innervated lobes. Moreover, both innervated and non-innervated lobes of Ptth^120F2A, giant^E1 double mutants showed approximately the same nuclear size as the non-innervated PG lobe from giant^E1 mutants (Fig. 4A,A′). This indicates that loss of Ptth leads to the reduced size of the non-innervated PG lobe and demonstrates that PTTH is the trophic signal to the PG.

Next, we inspected glands from larvae in which Ptth expression is lost in only one hemisphere of the brain using a Gal4 driver with variegated expression in PG neurons (NP0394-Gal4) in combination with UAS-Ptth-RNAi. As shown in Fig. 4B, Ptth knockdown causes a reduction in lobe as well as nuclear size compared with the contralateral side that has activated PTTH signaling. This confirms that PTTH acts as a PG growth factor. Moreover, PTTH secretion onto one PG lobe cannot compensate for its loss in the other lobe, suggesting that PTTH acts in a local manner.

In most cases, we observed that each PG cell is in contact with at least one varicosity (Fig. S5A). Surprisingly, these varicosity structures co-express PTTH (Fig. 4C′) with synaptic markers such as NC82 (Bruchpilot) (Fig. 4C″, Fig. S5B), Csp (Fig. S5C) and Syb-GFP (Fig. 4C). When populations of PG cells with and without PTTH input are intercalated in the same lobe, the PG cells with PTTH input are larger and express a higher level of Dib than immediate adjacent cells that do not receive PTTH (Fig. 4D,D′). We next generated PG cell clones expressing RNAi targeting the PTTH receptor Torso and observed that PG cells depleted of Torso also contain smaller nuclei (Fig. 4E). These results reveal that synaptically released PTTH acts autonomously through its receptor Torso to control PG cell size.

A common adaptive strategy that insects use in response to an adverse environment is to adjust the time of larval development at the expense of adult size (Rodrigues et al., 2015). To examine whether PTTH plays a role in adapting individual life history traits to changes in environmental conditions, we evaluated the ability of Ptth^Delta larvae to undergo metamorphosis under three diets with different nutrient content depending on their levels of inactivated yeast extract: from 5% (poor) to 17% (normal) and 34% (rich) (see Materials and Methods). When grown on poor food, Ptth^Delta larvae were delayed 1 additional day compared with controls (Fig. 5A), with a slight increase in mortality (from 6% to 15%, Fig. 5B). Therefore, concomitant reduction of nutrient signaling and PTTH pathways caused additive effects that could compromise the survival of the organism. We also investigated whether an artificial increase in PTTH levels generated using ubiquitous tub-Gal4 to drive expression of UAS-Ptth-HA could rescue the additional delay caused by growing larvae in low nutrients. As shown in Fig. 5C, PTTH was unable to rescue the developmental delay caused by low nutritional input. Moreover, Ptth^Delta larvae grown on rich food were still delayed by at least 20 h (Fig. 5A). Thus, PTTH cannot compensate for a low nutritional input and vice versa. These results suggest that several ecdysteroidogenic signals, including PTTH, serotonin (Shimada-Niwa and Niwa, 2014) and nutrient response IIS/Target of Rapamycin (TOR) signaling (Caldwell et al., 2005; Lavalle et al., 2008; Mirth et al., 2005; Walkiewicz and Stern, 2009) are timing metamorphosis in a non-redundant manner and may have additive effects in controlling the adaptive response to the environment.

Next, we tested animal survival in response to a different environmental stress, larval crowding, which is frequently observed in natural populations of Drosophila (Bubli et al., 1998). Larval crowding is a multifactorial stressor that, in addition to nutrient restriction, results in inevitable exposure to high concentrations of larval products such as nitrogenous metabolic waste (Joshi et al., 1996a,b; Schering et al., 1984) and pheromones (Mast et al., 2014). Interestingly, a recent report found a strong negative correlation between development time and fitness at high density (Horvath and Kalinka, 2016), meaning that fast developers have higher fitness. To determine if PTTH plays an adaptive role in this condition, L1 larvae were collected and grown at increasing densities in tubes (L1/t) of normal food. As shown in Fig. 6A, Ptth^Delta mutant and w^DAH control animals equally survive until adulthood when grown at standard densities (40 L1/t and 150 L1/t). However, survival of Ptth^Delta animals was drastically reduced at high density (43% at 350 L1/t and 16% at 500 L1/t), compared with control animals (72% at 350 L1/t and 50% at 500 L1/t), with most animals dying at late larval stages (data not shown). This indicates that PTTH is required for optimal survival under crowded as well as nutrient restricted conditions, possibly by fine tuning the duration of larval development. Moreover, the Ptth^Delta mutant ecloses with a 3-4 day delay in the 500 L1/t condition (Fig. 6B). Therefore, without PTTH, animals cannot properly adapt larval development time and survival to an environmental stress, such as crowding or nutrient restriction.

Because we observed a drastic effect on larval survival in crowded conditions, we tested the effects of this stressor on adult size and ovariole number in Ptth mutants and controls. Whereas w^DAH flies showed strong size adaptation to crowding (73% size reduction in 500 L1/t conditions, see Fig. 6C), Ptth^Delta mutants were less plastic (52% reduction). This, and the additional delay observed in crowded conditions, mean that Ptth^Delta adult flies weighed more than 200% of controls at 500 L1/t density, whereas at 40 L1/t the weight difference between Ptth mutants and controls was only 20% (Fig. 6C). A similar reduced plasticity in ovariole number was found in Ptth^Delta females exposed to high population density (Fig. 6D). Overall, these results indicate that PTTH participates in a general mechanism of adaptation to environmental stress, allowing animals to adjust adult size and fecundity to maximize survival.

The developmental delay that we observe with Ptth loss (1 day on rich food) is significantly less than that observed when PG neurons are ablated or misspecified (5-6 days on rich food; see also Ptth>grim in Fig. 7B) (Ghosh et al., 2010; McBrayer et al., 2007). In addition, the CW of the Ptth mutant (1.9 mg at 20 h after L3 ecdysis, Fig. 3B) is less than that of the PG neuron-ablated animals (2.6 mg at 72 h after L3 ecdysis, Fig. 7G). A simple explanation for these differences would be the existence of an additional ecdysteroidogenic signal(s) produced by the PG neurons. To assess this possibility, we manipulated neuronal activity of PTTH-producing neurons in control or Ptth^120F2A animals and measured the effect on developmental transitions. When control PG neurons were inactivated by selectively expressing either Kir2.1 (Hodge, 2009) or tetanus toxin (TetxLC) (Popoff and Poulain, 2010), we observed a 17-24 h developmental delay (Fig. 7A), similar to the delay produced by Ptth loss. However, inactivation of PG neurons in Ptth^120F2A larvae induced an additional delay of ∼24 h (Fig. 7B).

In order to activate PG cells, we used the targeted expression of TrpA1 (Hamada et al., 2008) and activated the channel by a temperature shift at various times after the L2/3 transition. As shown in Fig. 7C,E, shifting before or near the time of CW results in the maximal developmental acceleration of ∼17 h. Shifting after CW is still able to accelerate development, but in a graded manner. This is consistent with our data above describing a continuous requirement for PTTH signaling throughout the third instar period. Remarkably, we also see a developmental acceleration of 17 h when TrpA1 is activated in the PG neurons of Ptth^120F2A (Fig. 7D,F). This acceleration is able to rescue most of the delay caused by Ptth loss (Fig. 7F). These data indicate that at least one other factor may be released from the PG neurons to activate pupariation in the absence of PTTH. The existence of this additional signal could explain the different delays and CWs observed in response to PTTH loss compared with PG neuron ablation.

A central problem that all organisms face is to adjust their development to variation in environmental conditions. This adaptive phenotypic plasticity impacts life history traits, enabling a single genotype to produce an optimized output in the face of fluctuating conditions in its natural habitat. In both vertebrates and invertebrates, neuroendocrine signaling is thought to be a central component of the mechanism that responds to environmental cues to allow phenotypic plasticity (Denver, 1997; Gotthard and Nylin, 1995; Lessells, 2008), but the molecular circuitry remains ill-defined. In this report, we demonstrate that the insect neuropeptide PTTH is a key brain-derived signal that modulates hormone production to alter developmental timing in response to both nutritional cues and population density. In its absence, larvae spend additional time feeding and growing larger, which, in the face of a time-limited food supply such as rotting fruit, is likely to be detrimental to the long-term survival of these individuals. Indeed, recent studies suggest that, given a choice, Drosophila larvae regulate their macronutrient uptake (protein versus carbohydrate) to ratios that result in the shortest developmental time (Rodrigues et al., 2015). Short developmental time results in trade-offs with other life history traits such as body size and ovariole number that are maximized by longer developmental intervals and larger bodies (Rodrigues et al., 2015), as we observed for Ptth mutants.

Although we do not yet know all the various inputs that PG neurons integrate, their extensive dendritic arbors (Ghosh et al., 2010; McBrayer et al., 2007) suggest that they are capable of processing a multitude of extrinsic and intrinsic signals. For example, Ilp8 (Dilp8), an imaginal disc damage-induced signal, is received by a set of LGR3-positive neurons that in turn contact PTTH dendrites to reduce PTTH production or release to prolong larval development and thereby facilitate damage repair (Colombani et al., 2015; Garelli et al., 2015; Jaszczak et al., 2016; Vallejo et al., 2015). Likewise, in larval and maturing pupal brains, a subset of central circadian clock neuron terminals are in close proximity to the PG neuron dendritic field, and recent studies have shown that PTTH helps coordinate the activity of the central brain clock with the peripheral clock found in the PG (Selcho et al., 2017). This circuit ensures circadian rhythmicity during pupal development, which both influences the timing of pupal development and the observed rhythmicity of adult eclosion that may aid in reproductive success (Kouser et al., 2013). PTTH signaling has also been shown to enhance larval light avoidance at the time of pupation via an E-independent mechanism (Yamanaka et al., 2013a). This process ensures pupation in dark environments, which is likely to aid in pupal survival by reducing dehydration and predation. These examples suggest that the PG neurons, and PTTH signaling in particular, facilitate coordination of developmental processes with environmental conditions to maximize species fitness.

In holometabolous insects, feeding during the larval stages is especially important for their developmental program since they must accumulate sufficient energy reserves to fuel metamorphosis during the non-feeding pupal stage. As a consequence, various nutritional checkpoints such as CW and MVW have evolved to ensure that larvae do not enter metamorphosis before sufficient fuel has been stored. Two studies have put forward different mechanisms for determining CW. In the first, the process of larval molting is proposed to produce a transient starvation state leading to nuclear accumulation of a FoxO-USP complex that inhibits E production until sufficient feeding relieves this constraint and CW is achieved (Koyama et al., 2014). In the second paradigm, nutrient-dependent TOR signaling in the PG is thought to regulate endocycle entry, and DNA copy number is found to correlate with attainment of CW (Ohhara et al., 2017). Our examination of Ptth mutants clearly demonstrates that PTTH signaling also contributes to setting the CW checkpoint. All three processes share a common feature in that they affect the rate of basal E production, providing support for the notion that CW represents the attainment of a minimal threshold level of E that enables nutrient-independent completion of larval development and commencement of metamorphosis (Rewitz et al., 2013). It is interesting that the low basal E titer caused by Ptth loss does not influence larval growth rate, whereas lower basal levels of E produced by other methods have been reported to enhance larval growth rate (Boulan et al., 2013; Colombani et al., 2005; Moeller et al., 2017). We are presently investigating the mechanism responsible for these differences.

Whether and how these three molecular mechanisms intersect remains to be determined, but it is interesting to note that in our present study we demonstrated a role for PTTH in regulating nuclear size. Thus, PTTH signaling might modulate E production in conjunction with TOR by determining the number of DNA templates available for transcription of E biosynthetic enzymes. Alternatively, PTTH might act more directly to influence transcriptional control of E biosynthetic components. Numerous transcription factors, including EcR, Mld, Usp, FoxO, Br, Ftz-f1, HR4, Vvl and Ouija board (Danielsen et al., 2014; Komura-Kawa et al., 2015; Koyama et al., 2014; Moeller et al., 2013; Ou et al., 2011; Parvy et al., 2005) have been implicated in transcriptional regulation of one or more E biosynthetic enzymes, although none has been shown to be a direct target of Rolled (Erk) phosphorylation, the major downstream mediator of PTTH signaling. One factor of particular interest, however, is the nuclear receptor HR4, since it has been shown that knockdown of its expression in the PG accelerates development, while overexpression retards development (Ou et al., 2011). Furthermore, during the first half of the third instar stage, HR4 cycles in and out of the nucleus with a periodicity that depends on Ras/Erk signaling. The proposed mechanism posits that HR4 negatively regulates one or more E biosynthetic genes, including Cyp6t3, and that periodic PTTH signals enable the production of transient small E pulses, each of which might represent passage through a specific developmental checkpoint that controls timing of metamorphosis (Ou et al., 2011; Rewitz and O'Connor, 2011; Rewitz et al., 2013). Although this mechanism is an attractive possibility, we find no rescue of developmental timing when certain biosynthetic enzyme genes, including Cyp6t3, are individually overexpressed. These observations indicate that either there is an as yet unidentified PTTH-regulated rate-limiting component, or that it is the combined reduction in several enzymes that results in slower E production. Lastly, in Lepidoptera, PTTH has been shown to stimulate translation as well as transcription and this might also come into play when considering how a slow rise in E titer is produced (Keightley et al., 1990; Kulesza et al., 1994; Rybczynski and Gilbert, 1995).

One of the more unexpected findings of this study is that Drosophila PTTH appears to act autonomously for regulating E biosynthesis within the PG, whereas a previous study demonstrated that PTTH in the hemolymph acts non-autonomously, and in a Torso-dependent manner, to regulate larval light avoidance (Yamanaka et al., 2013b). In addition, a recent study showed that Torso expression in the fat body is required for proper insulin signaling and body size control although the ligand involved was not identified (Jun et al., 2016). Since PTTH reaches the PG through the circulatory system in Lepidoptera, it is unlikely that there is an intrinsic functional need to regulate E biosynthesis through direct innervation of the PG, and indeed we found that overexpression of PTTH in the fat body was able to rescue the developmental delay and body size phenotypes of Ptth null mutants (data not shown). This apparent dichotomy with respect to autonomous and non-autonomous requirements in different tissues might be explained by the unusual biophysical features of Torso, which binds PTTH asymmetrically resulting in anti-cooperativity (Jenni et al., 2015). Thus, receptor levels determine the nature of the signal output: high receptor levels stimulate a transient burst of enhanced signal, whereas low receptor levels produce a low amplitude, sustained signal. In addition, high receptor levels may facilitate intermolecular disulfide bond formation between Torso monomers within the transmembrane domain and thereby stabilize a preformed dimeric receptor that is primed for activation by ligand binding (Konogami et al., 2016). Perhaps innervation of the gland by PG neurons facilitates clustering of receptors at the synapses, stimulating production of preformed receptor dimers near the PTTH release sites and resulting in a high-level signal in the PG cells that is necessary for rapid induction of E biosynthetic enzymes. By contrast, the non-autonomous effect of PTTH on Bolwig's organ and class IV multidendritic neurons might result from low receptor concentrations leading to the sustained low-level signal needed to modulate light avoidance. It is interesting to note in this regard that we cannot detect endogenous Torso protein by immunolocalization in Bolwig's organ or class IV multidendritic neurons, but can readily detect it in PG cells, consistent with the idea that receptor levels may influence the type of signal produced by PTTH in different tissues.

An alternative possibility is that two synergistic signals, one of which is localized in the PG, are required to produce rapid induction of E biosynthetic gene transcription in the PG, but two signals are not required for light avoidance. We provide evidence that at least one other signal from the PG neurons is essential for proper timing of E production. Local loss of this signal in cells lacking a varicosity could be sufficient to prevent high-level induction of E biosynthetic enzymes in those cells, while release of PTTH from remaining varicosities could provide sufficient systemic levels of PTTH to stimulate light avoidance. It is interesting to note with respect to the two-signal model that, in a tissue culture assay system, Torso activation can be stimulated by the synergistic interaction of Trunk, a second Torso ligand involved in the specification of terminal embryonic cell fates, and Tsl, the product of the torso-like locus (Amarnath et al., 2017).

Although the data presented here provide strong evidence that PTTH is a key environmentally responsive developmental timing signal in Drosophila, our manipulation of PG neuron activity, coupled with previously reported ablation studies (Ghosh et al., 2010; McBrayer et al., 2007), clearly suggest the presence of at least one additional timing factor, the release of which is activity dependent. Whether this factor also signals through Torso or another receptor is unresolved, but it is interesting to note that spätzle-like proteins show some structural features common to both PTTH and Trunk, the two known Torso ligands (Casanova et al., 1995). The presence of multiple timing signals is perhaps not surprising since redundancy is likely to provide both robustness to the circuit and flexibility for output responses. Recent characterization of Ptth mutants in Bombyx also showed alterations in developmental timing during larval stages (Uchibori-Asano et al., 2017). However, there was much more variability in the timing delay phenotype, as well as significantly enhanced lethality, compared with what we observe in Drosophila, perhaps indicating that alternative signals are either not present or less effective in compensating for Ptth loss in Bombyx.

Plasmids TAL248NN and TAL249NN (in the pT3TS-delta152, +63 backbone) were digested with Ecl136II, and 1 µg was transcribed using the T3 mMessage mMachine Kit (Ambion). The 5′-capped mRNAs were phenol-chloroform extracted and precipitated with isopropanol. These left and right Ptth TALEN monomers were dissolved in 30 µl 0.1× PBS (DEPC treated) and mixed to make a 30 µl solution with each mRNA at a final concentration of 400 µg/ml in 0.1× PBS. An aliquot was injected into w¹¹¹⁸ embryos (Duke Injection Services) and individual G0 adults were crossed to balancer line sna^Sco/CyO* (BDSC #3198). Once larva were observed in the vial, the G0 fly was removed and frozen at −80°C for subsequent single-fly DNA extraction with squishing buffer (Gloor et al., 1993). Genomic DNA from 75 fertile G0 flies was analyzed by PCR with primers PTTH TAL 01F and PTTH TAL 03R (Table S1) incorporating the LightCycler 480 High Resolution Melting Master (Roche) and using the Tm calling program. Five mosaic G0 lines were selected based on an altered Tm peak: either wider, displaying a shoulder, or the presence of two Tm peaks. Ten to fifteen individual balanced F1 males from the mosaic candidates were backcrossed to w*;sna^Sco/Cyo* females and then self-crossed to obtain homozygous stocks. These homozygous stocks were again analyzed by Tm calling and the region of interest sequenced. Subsequent analyses used w¹¹¹⁸ as the control to the Ptth^120F2A and Ptth^8K1J mutants since that was the line originally injected and so should adequately serve as a control.

Two distinct lesions were found at the targeted Ptth site, and no additional changes were observed in these lines when the complete gene was sequenced. The lesions identified comprise: (1) a 7 bp deletion (ΔCGTTTGC between nucleotides 465 and 471 of the PTTH-RF coding sequence) in line Ptth^120F2A; and (2) a 6 bp deletion (ΔATGCCG between nucleotides 461 and 466 of the PTTH-RF coding sequence) in line Ptth^8K1J. The Ptth^120F2A allele generated a frameshift distal to N154 of PTTH-PR that continued out of frame for 24 amino acids before reaching a stop codon. This mutation removed much of the predicted mature ligand portion of PTTH. The 6 bp deletion found in line Ptth^8K1J leads to a deletion of N154 and A155 and a V156I substitution (Fig. 1A). A third allele, Ptth^8M9A, which deleted 5 bp (ΔCGTTT) between nucleotides 465 and 469 of the PTTH-RF coding sequence, was not used in this work.

To generate the Ptth^Delta mutant, the accelerated homologous recombination technique was used as described (Baena-Lopez et al., 2013). Two homology arms were amplified from genomic DNA using the primers listed in Table S1, which incorporate NotI or BglII restriction sites for cloning. The resulting PCR products were digested and cloned into pTV-Cherry vector (2.7 kb corresponding to the 5′ arm with NotI and 2.3 kb for the 3′ arm with BglII). This Ptth targeting vector, containing the appropriate homology arms, was introduced at random genomic locations by P-element-mediated transformation (BestGene). Transformants (not necessarily mapped or homozygosed) were crossed to hs-FLP, hs-I-SceI flies (BDSC #25679) and the resulting larvae were heat shocked at 48 h and 72 h after egg laying for 1 h at 37°C. Approximately 200 adult females with mottled red eyes (indicating the presence of the targeting vector and the transgene carrying hs-FLP and hs-I-SceI) were crossed to ubiquitin-Gal4[3×P3-GFP] males and the progeny screened for the presence of red eyes. The ubiquitin-Gal4[3×P3-GFP] transgene was subsequently removed by selecting against the presence of GFP in the ocelli. Red-eyed flies resulting from these crosses were subjected to PCR with primers PTTH forward and reverse (Table S1). One line was confirmed to be a Ptth null homologous recombinant, referred to herein as Ptth^Delta. Sequencing results confirmed a deletion of 604 bp in the Ptth^Delta allele that included the putative ATG and signal peptide (Fig. 1A).

The Ptth^Delta allele was backcrossed into the w^DAH standardized genetic background for at least ten generations before any experimental procedures. Thus, we used w^DAH as the correct background control for Ptth^Delta.

D. melanogaster release 6 (FlyBase Consortium) was used to define the DNA, RNA and amino acid coordinates of Ptth. Three start sites are identified in FlyBase, and we used the cDNA identified as G.

The genomic fragment of HA-tagged PTTH from pMBO2318 (McBrayer et al., 2007) was excised as a NotI-EcoRI fragment and placed in a pattB vector (pattB-QF-sv40, Addgene #24367) which was digested with NotI and EcoRI in order to drop out the QF fragment. This plasmid, pMBO2604, was inserted into the attP2 site of y¹ w^67c23; P[CaryP]attP2 (BDSC #8622) by injection (GenetiVision) and phiC31 integration. The transformant line was designated Ptth WT-attP2-2A3. The spok>Gal4 line was made by inserting a 1.4 kb upstream fragment of spok (Komura-Kawa et al., 2015) into the Pelican-Gal4 vector (from M.B.O.'s lab.). Transgenic line spok-Gal4 1L3 was derived after injection into w¹¹¹⁸.

Animals were reared at 25°C on a variation of standard cornmeal food (CMF, BDSC), in which brewer's yeast was substituted for baker's yeast and Tegosept was added for mold control. Live baker's yeast granules were shaken liberally onto the surface. Some timing experiments were performed on ‘normal’ fly food, which contains per liter: 17 g inactivated yeast powder, 83 g corn flour, 10 g agar and 4.6 g Nipagin; for ‘rich’ or ‘poor’ food, the amount of inactivated yeast powder was adjusted to 34 g/l or 5 g/l, respectively.

The following fly stocks were used: P0206-Gal4 (Colombani et al., 2005), UAS-spok 4A2 (Komura-Kawa et al., 2015), UAS-dib 20A2 (NotI-Asp718I dib cDNA in pUAST), UAS-nvd (Yoshiyama et al., 2006), spok-Gal4 1L3 and UAS-Cyp6t3 (Ou et al., 2011), fb-Gal4 (Tom Neufeld, University of Minnesota, USA), UAS-Ptth HA 96L3 (McBrayer et al., 2007), gt^E1;UAS-GFP;Feb²¹¹-Gal4, UAS-mCD8-GFP (Siegmund and Korge, 2001), Ptth-Gal4 45A3, 117bA3 (McBrayer et al., 2007), Ptth-Gal4 45A3, 117bA3; UAS-grim (McBrayer et al., 2007), UAS-TrpA1/TM6B,Tb (BDSC #26264), UAS-Kir2.1 on the 3rd chromosome (BDSC #6595), UAS-TetxLC on the 3rd chromosome (BDSC #28997), UAS-TetxLC.IMP on the 3rd chomosome (BDSC #28841), UAS-dORKΔC on the 3rd chromosome (BDSC #6586), NP0394-Gal4 [Drosophila Genetic Resource Center (DGRC) Kyoto, Japan #103604], UAS-Ptth-RNAi (Vienna Drosophila Resource Center (VDRC) #102043], UAS-torso-RNAi_GD (VDRC #36280), UAS-Synaptobrevin-GFP (BDSC #60677) and Ay-Gal4 (act-FRT-stop-FRT-Gal4) (BDSC #3953). Recombinant or combination stocks were made using standard genetic methods.

All developmental timing experiments were performed at 25°C in 12 h light/dark cycle conditions. One- to four-hour time collections of embryos laid on apple juice plus yeast paste plates were aged for 20 h, at which point freshly ecdysing L1 larvae were transferred to vials. Each vial contained 30, 40, 150, 350 or 500 L1 larvae. The time and date of pupariation were scored every 1-5 h during the light cycle and the time in hours from L1 ecdysis to puparium formation was determined using the time duration calculator (www.timeanddate.com). Data from four vials were compiled and ordered by progressive pupariation time and cumulative percentage pupariation and subsequently analyzed in GraphPad Prism using non-linear regression curve fit or in Microsoft Excel.

Male and female pupae were photographed and measured with ImageJ. Pupal length was measured along the medial line between anterior and posterior, not including the anterior and posterior spiracles. Pupal width was measured along the axial line. Pupal volume was determined using the formula for a prolate spheroid: (π/6)W²×L, where W is width and L is length. Sexed adults were weighed in groups of three to five flies or individually using a Mettler Toledo XP26 microbalance.

Tissues dissected from larvae in 1× PBS (137 mM NaCl, 2.7 mM KCl, 4.3 mM Na₂HPO₄, 1.47 mM KH₂PO₄, pH 8) at the indicated hours after egg deposition were fixed in 4% formaldehyde (Sigma) in PBS for 20 min at room temperature, washed in PBS containing 0.1% Triton X-100 (PBT), blocked for 2 h in PBT containing 10% FBS (PBS-TF), and incubated overnight with primary antibodies at 4°C. The next day, tissues were washed, blocked in PBS-TF, and incubated with secondary antibodies for 2 h at room temperature. After washing, tissues were mounted in Vectashield containing DAPI (Vector Labs) for staining of DNA. Fluorescence images were acquired using a Leica SP5 DS (20× and 40× objectives) and processed using Adobe Photoshop CS5 or ImageJ.

Primary antibodies were anti-PTTH (guinea pig, 1:200; P.L.), anti-NC82 (mouse hybridoma, 1:50; DSHB), anti-Dib [rabbit, 1:500; Parvy et al. (2005)] and anti-GFP (rabbit polyclonal, 1:500; Abcam). Secondary antibodies were Alexa Fluor 555 goat anti-guinea pig, Alexa Fluor 488 sheep anti-rabbit, AlexaFluor 488 donkey anti-rabbit (1:500; Abcam) and Cy3 sheep anti-mouse (1:200; Jackson ImmunoResearch).

Four-hour egg collections were performed on agar plates, and after 20 h L1 larvae were collected and reared at 40 animals per tube at 25°C. Larvae that ecdysed from L2 to L3 over a 2 h period were transferred onto new food plates and allowed to feed until they reached the appropriate age. After immunohistochemistry of larval tissues, the image of wing discs was acquired using a Leica SP5 DS, and the disc and nucleus area were measured using ImageJ.

L1 larvae were collected 24 h after egg deposition (4 h egg collections) and reared at 40 animals per tube at 25°C. Adult flies of the appropriate genotype were collected and stored in ethanol and mounted in a Euparal solution (ROTH). Images of dissected wings were acquired using a Leica MZ16 FA fluorescence stereomicroscope and DFC 490 digital camera. Wing areas were measured using ImageJ.

Newly eclosed flies were maintained in vials on standard food until the time of dissection (3-4 days after eclosion). Ovaries were dissected in cold PBS, and ovarioles were teased apart and counted under a dissecting microscope.

A variation of published methods (McBrayer et al., 2007; Mirth et al., 2005) was used. Larvae were raised at 25°C in continuous light conditions. Newly ecdysed L1 larvae, as described above, were transferred at intervals to CMF in 35 mm diameter plates and aged for 48 h. Freshly ecdysed L3 larvae were transferred to a new CMF plate at 1 h intervals at a density of less than 30 L3 larvae per plate, and timing initiated. Every 2 h starting at L3 ecdysis (T0) and continuing until 24 h for w¹¹¹⁸ and 48 h for Ptth^120F2A, ten feeding larvae were removed, individually weighed and placed in starvation medium [1.5% Bacto Agar (BD Diagnostics) in water]. Puparium formation was quantitated and timed, as well as eclosion rates. The critical period was defined as the number of hours after starvation when ≥90% of the larvae pupariated and did so without a delay. The CW was the average weight of those ten larvae at the critical time.

Thirty freshly ecdysed Ptth^120F2A L3 larvae, grown at 25°C under 12 h light/dark conditions, were washed with water and transferred to a small CMF plate for additional aging. After 20 h, larvae were washed and transferred to a vial supplemented with either 20-hydroxyecdysone (Cayman Chemical; dissolved in 95% ethanol, final concentration 0.2 mg/ml) or 95% ethanol (same volume as 20-HE). The control w¹¹¹⁸ larvae were directly transferred to vials upon L3 ecdysis. Once seeded with L3 larvae, the vials were returned to 25°C under 12 h light/dark and monitored, as well as timed, for pupariation.

To show that PTTH is the trophic signal to the PG, females of genotype gt^E1/FM, act-GFP; Ptth^120F2A/CyO, act-GFP were crossed to Ptth^120F2A; Feb²¹¹-Gal4, UAS-mCD8-GFP males. CNS plus RG complexes from GFP-negative wandering larvae were dissected, fixed and mounted. About 25% of animals had asymmetric innervation. For the same purpose, females of the variegated driver NP0394-Gal4 were crossed to UAS-Ptth-RNAi and the offspring larvae were grown on food containing per liter: 8.5 g inactivated yeast powder, 83 g corn flour, 10 g agar and 4.6 g Nipagin. CNS plus RG complexes were dissected, fixed and stained with anti-PTTH antibody. About 15% of animals had asymmetric Ptth silencing.

Mitotic clones marked by GFP were generated using hsFLP; Act>CD2>Gal4, UAS-nGFP in combination with UAS-torso-RNAi. Clonal induction was performed for 5 min at 37°C in the second instar.

Larvae of the indicated genotypes were first synchronized at L2/L3 molting, and then vials of 25-30 larvae were further reared at 25°C. Vials were transferred to 29°C at progressive times to activate the TrpA1-expressing neurons. Pupariation was monitored by counting the number of pupae at consecutive time points.

Timed samples were raised at 25°C under continuous light. Forty newly ecdysed L1 larvae, precisely timed on apple juice collection plates, were transferred at 1 h intervals to 35 mm diameter plates containing CMF with surface granules of live baker's yeast. Larvae continued to develop for ∼48 h and were then monitored for morphological features characteristic of ecdysis to the third instar stage. These freshly ecdysed L3 larvae were transferred to a fresh food plate, and timing of L3 age in hours commenced. For qRT-PCR, larvae of defined ages were removed from the food, washed in DEPC-treated water, homogenized in Trizol (Invitrogen), and stored at −80°C. For E titers, larvae of defined ages were removed from the food, washed twice in water, dried on a Kimwipe, weighed in bulk for biological replicates, and stored at −80°C.

Total RNA was purified from Trizol, transcript RNA was isolated using the RNeasy Mini Kit (Qiagen), and reverse transcribed with the SuperScript III First-Strand Synthesis Kit (Invitrogen). qRT-PCR reactions used LightCycler 480 SYBR Green I Master Mix on a LightCycler 480 instrument (Roche). Primers are listed in Table S1. All reactions were performed in duplicate. Expression (arbitrary units) was determined relative to RpL23 using the formula (2^−ΔCp)×100,000 (or a factor that gives an expression value of 10-500 arbitrary units). The average of four biological replicates ±s.e.m. was plotted.

Biological replicates of 4-15 larvae were homogenized twice in methanol and cleared by centrifugation to give a final combined volume of ∼360 µl. Duplicate 100 µl samples were dried and resuspended overnight in 50 µl EIA buffer (Cayman Chemical). Cayman Chemical reagents for 20-HE ELISA were used as per manufacturer's recommendation with two changes: in some experiments, the E antibody L2-B (kind gift of Jean-Paul Delbecque, Aquitaine Institute for Cognitive and Integrative Neuroscience, Bordeaux, France) was substituted for the Cayman antiserum, and in other cases where there was a low E titer because of small sample size, the 20-HE AChE tracer was diluted up to 1:5 in the reaction. The standard curve was determined using GraphPad Prism software, non-linear regression curve fit, asymmetric sigmoidal, 5PL, X is log (conc). Per animal, E titers were adjusted for the volume used in the assay.

We thank the Bloomington Drosophila Stock Center (BDSC) for numerous Drosophila lines and Aidan Peterson and Derya Deveci for comments on the manuscript; Colby Starker, Michelle Christian and Dan Voytas for help in designing and making TALENs for the Ptth null mutant production; Anna Petryk and Michael Jarcho for producing UAST-sad, phm and dib constructs; and Thomas Phil for technical support.