Chinmo is the larval member of the molecular trinity that directs Drosophila metamorphosis

Based on whole-body mRNA and protein measurements in ModEncode (29), chinmo, broad, and E93 expression are associated with the three successive phases of the Drosophila life history. Chinmo immunoreactivity first appears in the CNS as the embryo is undergoing germband retraction (Stage 14; ∼11 h after fertilization) and is ubiquitously expressed by hatching (Fig. 1 B and C and SI Appendix, Fig. S1). Except for neurons, larval cells of Drosophila are polyploid and grow by cell enlargement during the three larval stages (L1 to L3), but most then degenerate early in metamorphosis. Larvae also possess clusters of diploid imaginal cells that form the imaginal discs and primordia. These do not make larval structures but proliferate during larval growth and eventually replace the larval cells to make the adult body. In both larval cells and imaginal disc cells (Fig. 1 D and E), Chinmo levels are high from hatching through the early part of L3, until the larva passes the critical weight checkpoint that is the entry into metamorphosis (∼8–10 h into L3 stage) (30). At this time, Chinmo levels begin to decline in both imaginal discs and larval cells (Fig. 1 D and E) (28) and broad isoforms appear. This change-over has been studied in the premetamorphic wing imaginal disc, where Chinmo and Broad reciprocally inhibit each other, such that the experimental suppression of one causes the elevated expression of the other (28). Broad is prominent in both larval and imaginal cells during the formation of the pupal stage and then E93 appears in both cell types as adult differentiation is under way (Fig. 1 D and E). The larval cells then degenerate while Broad protein is still present (Fig. 1E), but imaginal disc cells completely replace Broad with E93 as the adult structures differentiate (Fig. 1D).

Chinmo first appears around midembryogenesis and Chinmo mutants are embryonic lethal (26). We examined the phenotype of the chinmo¹ mutant, using the balanced line used for Chinmo MARCM (mosaic analysis with a repressible cell marker) studies (w*; chinmo¹ P {ry^+t7.2 = neoFRT}40A/CyO, y⁺) (hereafter called chinmo¹/CyO) (26) (see SI Appendix, Table S1 for Drosophila lines and abbreviations used). When this line was crossed to itself, approximately half of the larvae hatched (Fig. 2A), 90% of which showed strong Chinmo-immunoreactivity (IR) (Fig. 2 B and C; chinmo¹/CyO). Approximately one quarter of the eggs showed no obvious embryo (presumably CyO/CyO), and the remaining quarter blocked in late embryogenesis. One set of late-arresting embryos (Class I) appeared to be fully developed with gas filling their tracheal systems, while a second set (Class II) showed a gut configuration similar to a Stage 15 embryo but with sclerotized mouth hooks and obvious denticle belts. Immunostaining for Chinmo showed that many Class I embryos had the depressed levels of Chinmo expression expected for the chinmo¹/chinmo¹ homozygote (Fig. 2B). Crossing chinmo¹ to different deletion lines that removed the entire locus yielded only the Class I phenotype (Fig. 2A). Many of the Class I embryos expelled Malpighian tubule contents into the space between the embryo and the vitelline membrane, and these embryos all showed depressed levels of Chinmo (Fig. 2B). Most of these embryos had normal proportions but others were shortened along their anterior-posterior axis.

The chinmo¹ mutant was thought to be a protein null based on the lack of Chinmo-IR in homozygous chinmo¹ neuroblast clones (26), but our data suggest that it is a hypomorph. We think that the clones in the previous study lacked Chinmo expression because the mutant neuroblasts had shifted into making their late-born progeny, which are normally Chinmo-negative. Crossing chinmo¹/CyO flies to either deletion stock also produced late arrested embryos with very low levels of Chinmo-IR (Fig. 2C, chinmo¹/Df); while embryos with the overlapping deletions, which completely removed the chinmo locus, showed no Chinmo-IR (Fig. 2C, Df/Df). A few chinmo¹ homozygotes apparently hatch because we found occasional hatchlings from these crosses with very low Chinmo expression (Fig. 2B). However, we found no mid- or late-stage L1 larvae that showed low Chinmo expression, suggesting that these chinmo¹ homozygotes do not grow.

The phenotypes of the chinmo¹ homozygotes and the double deletion embryos show that zygotically produced Chinmo is not needed to form the larva, but because such animals die, they cannot tell us if Chinmo is needed to maintain the larval state. Consequently, we turned to a mosaic approach in which we removed Chinmo from only some larval cells and then tracked their development in the growing animal. We used an engrailed-GAL4 driver to express a nuclear localized red fluorescent protein (nRFP) marker and chinmo RNAi constructs (SI Appendix, Table S1) in the posterior compartments of larval body segments and imaginal discs. Two independent lines, UAS-TRiP HMC05346 and UAS-TRiP HMS00036, resulted in marked depletion of Chinmo-IR in the Engrailed-positive compartments. The former gave the stronger knockdown of Chinmo-IR and was the principal line used for these studies. As seen in Fig. 2 D and E, cells that lacked Chinmo in the first larval instar showed the precocious expression of either Broad or E93, the genes that normally are expressed at the onset of metamorphosis to pupal (14, 15) or adult (20) stage, respectively (Fig. 1 D and E). However, the imaginal disc and larval cells differed in which trinity gene they expressed. Larval cells, characterized by epidermal and salivary gland cells, showed premature expression of E93 but not Broad, while the imaginal disc cells showed premature expression of Broad but not E93. We do not yet know the reasons why mutant larval cells precociously express E93 whereas the imaginal cells express Broad, but they may arise from differences in how these later trinity members interact with each other in the respective tissues. Larval and imaginal cells also differed in the timing of their response to the loss of Chinmo. Larval epidermal cells that lacked Chinmo showed precocious E93 expression by about midembryogenesis (late stage 15, shortly after dorsal closure and head involution) (Fig. 2F), shortly after the time when Chinmo first appears. In the case of Broad, by contrast, precocious expression in imaginal disc cells did not appear until the mid to late first instar (e.g., Fig. 2D). These imaginal disc cells may be basically dormant until stimulated to proliferate late in the first instar (31). This break of dormancy may be necessary for Broad expression to begin. The second chinmo-RNAi construct also caused premature expression of Broad in imaginal discs and E93 in larval cells (SI Appendix, Fig. S2).

The loss of Chinmo had profound developmental effects on both larval and imaginal disc cells. The en-GAL4 > UAS-chinmo-RNAi (UAS-TRiP HMC05346) larvae grew during the L1 stage and eventually achieved a size larger than a young L2, but they did not molt to the second instar (Fig. 3A). Some of these larvae then assumed a “tiger-striped” appearance, with two thin rings of tanned cuticle within each segmental engrailed domain (Fig. 3B). These rings correspond to tanned cuticular ridges that normally appear in the last larval cuticle as it transforms into the puparium at the start of metamorphosis (Fig. 3C). Optical cross sections through such larvae (Fig. 3D, diagrammed in Fig. 3E) show that the epidermal cells between engrailed domains make a new, L2 cuticle, while the Chinmo-negative cells within each engrailed domain did not (and instead started puparial tanning). Consequently, at the end of the first larval stage, the epidermal cells of these mosaic larvae respond to the molting surge of ecdysteroids in different ways depending on whether or not they express chinmo: the Chinmo-positive cells make a new larval cuticle (Fig. 3 D and E) while the Chinmo-negative cells are directed toward metamorphosis and puparial tanning (Fig. 3 B and E).

Analysis of the knockdown of chinmo in the salivary gland shows that Chinmo is also involved in the normal trajectory of cell growth in this tissue. At hatching, nuclear size is the same in wild-type versus Chinmo-negative regions of the salivary gland. Later in the L1 stage, the wild-type cells have grown markedly but the Chinmo-negative cells remain small (Fig. 3F).

Chinmo is also involved in the early growth of the imaginal discs. For example, the wing imaginal disc is an invaginated sac of epidermis that shows little distinction between the anterior and posterior compartments at hatching. The wing discs of en-GAL4 > UAS-nRFP, UAS-chinmo-RNAi larvae form normally and look like those of wild-type at hatching. Proliferation begins in the wing disc in the late first instar (31), and those from en-GAL4 > UAS-nRFP larvae expand as a relatively uniform epithelium (Fig. 3G), while those from en-GAL4 > UAS-nRFP, UAS-chinmo-RNAi larvae show cells within the RFP domain that are abnormally thickened and disorganized. In extreme cases, these latter cells ball-up and are extruded into the disc lumen (Fig. 3H). Considering that these cells are expressing Broad, we suspect that they prematurely shift into a premetamorphic mode, although we have not yet found an independent molecular marker to support this conclusion.

The life history of holometabolous insects, with their complete metamorphosis, involves a progression through three distinct body forms comprising the larva, the pupa, and the adult, respectively. Single genes, broad and E93, oversee the formation of the pupal and adult stages, respectively (9, 19). Although these genes are transcription factors, they also are involved in altering the chromatin landscape, opening up regions of DNA relevant to their respective life-stage and closing down regions relevant to the other two (32, 33). This function has been elegantly shown for the actions of E93 in opening up the chromatin landscape needed for the formation of the adult and for closing the chromatin around the broad gene (32, 34). Until now, a larval specifier—and hence the first member of the hypothesized larval-pupal-adult specifying trinity of master genes (12)—has remained elusive. Our data suggest that chinmo is the first member of the trinity, overseeing the maintenance of the larval stage in holometabolous insects. As a member of the BTB gene family that includes broad, Chinmo, like Broad, may mediate maintenance of the larval state through chromatin modification as well as by direct control of gene expression.

Timing of chinmo expression is consistent with that expected for a larval control gene (Fig. 4A). It appears around embryonic stage 14, during germband retraction and dorsal closure. This time of appearance is significant because the extended germband stage, when body segmentation is complete and the primordia for appendages and the various organ systems have been established, is similar in both hemi- and holometabolous insects (35). In direct developers, this stage is followed by the production of the imaginal body plan, while in insects with a larval stage development is redirected to making the larval body. Chinmo expression then continues at high levels until “critical weight,” the growth-related checkpoint that allows the preparation for metamorphosis (30, 36). After this checkpoint, larval tissues, such as the fat body (37) and the salivary glands (38), shift to making proteins necessary for metamorphosis, and imaginal discs switch from simple nutrient-dependent growth to the morphogenetic programs that results in disc patterning (36). For both larval cells (Fig. 1E) and imaginal discs (Fig. 1D) (28), the down-regulation of chinmo after the critical weight checkpoint is accompanied by the up-regulation of broad. In the discs, it has been shown that this change from Chinmo to Broad is essential for the discs to shift from self-renewal growth that allows disc regeneration in response to injury to differentiative growth that can support only simple wound healing (28).

The late embryonic blockage seen when the chinmo locus is completely removed through overlapping deletions shows that a larva can form in the absence of zygotically derived chinmo transcripts. However, Chinmo appears necessary to maintain larval function: salivary gland cells lacking Chinmo show retarded growth (Fig. 3F) and Chinmo-negative epidermal cells respond to the L1 ecdysteroid surge by trying to make a puparium rather than by making a second instar cuticle (Fig. 3 D and E). In young imaginal discs, the lack of Chinmo causes their cells to form a thickened epithelium that resembles that seen as a prelude to metamorphosis (Fig. 3H). We do not yet know whether these effects of the lack of Chinmo are direct responses to the loss of this factor or are indirect responses caused by the precocious appearance of E93 and Broad in larval and imaginal cells, respectively.

The striking feature of the removal of Chinmo is the precocious expression of the two key metamorphic genes, broad and E93 (summarized in Fig. 4A). A switch from Chinmo and Broad expression at critical weight has been known for a while for both young neurons (27) and imaginal discs (28). These two genes reciprocally inhibit each other’s expression around this time (28) and this transcriptional shift is reinforced by the Let-7 family of microRNAs that stabilize broad transcripts and destabilize those of chinmo (39). We find, though, that a requirement for Chinmo to suppress broad expression extends back to the L1 stage when the imaginal discs begin their proliferation. The function of Chinmo in suppressing E93 expression occurs even earlier, being evident from midembryogenesis (Fig. 2F), around when Chinmo first appears. We do not know if this is also a reciprocal relationship in which E93 can suppress expression of chinmo.

The gene network that supports the progression from larva to pupa to adult is depicted in Fig. 4B. As proposed by Williams and Kafatos (12), each of the master genes strongly influences the expression of the others. The known interactions are all inhibitory, as detailed above, although there is not yet experimental evidence to support the inhibition of E93 by Broad or of chinmo by E93. Once established, each node of the network is inherently stable due to these inhibitory interactions but extrinsic factors, largely supplied by hormones, such as the ecdysteroids and juvenile hormone, and morphogens, like myoglianin (40, 41), can shift the animal from one node to the next, thereby allowing a metamorphic transition.

What, then, is the role of the trinity in insects that show direct development (Fig. 4C)? The function of E93 is the same in insects that show direct development and those that have complete metamorphosis. E93 is required at the end of somatic growth to transition into a reproductive stage (19–21). The somatic growth of direct developing insects occurs through the nymphal stages which display the species-typical body plan throughout. In insects with complete metamorphosis, though, somatic growth occurs in a modified larval form, and the species-typical body plan organizes at the end of growth with formation of the pupa. In both cases, Broad is associated with the appearance of the species-typical body plan. In the latter case of complete metamorphosis, Broad appears toward the end of growth and is essential for generating the pupa. In direct developing insects, by contrast, broad appears early in embryogenesis (17, 18), and suppression of embryonic broad by maternal RNAi in the milkweed bug, Oncopeltus fasciatus, and the cockroach, Blattella germanica, results in a substantial percentage of embryos that do not progress beyond the extended germband stage (17, 18). This blockage point is important because it is where direct development and metamorphic development diverge with the former progressing to its species-typical form and the latter diverted to making a larval form. Chinmo then functions in Drosophila to maintain this larval form and to inhibit expression of both broad and E93 until the end of the growth period. The developmental consequences of the relationship of chinmo with broad are elegantly shown by their functions in Drosophila imaginal discs where they provide the molecular gate that allows the discs to shift from simple growth to patterning and cell determination at critical weight (28). The only data on chinmo in direct developing insects are transcriptomic data in embryos of B. germanica (42): chinmo transcripts are prominent very early in embryogenesis but then decline between embryonic day 1 and 2 as the embryonic germ-band forms and broad transcripts increase. This timing and temporal relationship with Broad is consistent with a scenario in which Chinmo supports early embryonic growth and antagonizes the shift to embryonic patterning and cell determination. It will be crucial to understand how chinmo and broad interact in insects with direct development and what may have changed in this interaction to allow the maintenance of Chinmo and direct embryonic development away from making the species-typical form to making a larva instead.

We thank Dr. Chris Q. Doe, University of Oregon, and Dr. Nicholas Sokol, Indiana University, for generously providing antibody reagents. Other antibodies were obtained from the Developmental Systems Hybridoma Bank (Iowa City) and the Drosophila stocks from the Bloomington Drosophila Stock Center. We thank Drs. Richard Strathmann, Victoria Foe, Jason Hodin, Rachel Merz, Brian Clark, and Barbora Konopova for critical comments on the manuscript. Funding was a gift from the Howard Hughes Medical Institute to the University of Washington to support the research of J.W.T.