S. frugiperda larvae were reared on an artificial diet (Pinto et al., 2019) at 26°C ± 1°C with a 60% ± 5% relative humidity and a photoperiod of 14-h light:10-h dark. To avoid cannibalistic behavior (He et al., 2022), the larvae were separated and reared individually in plastic containers after molting into the L3 instar. Larval molting was monitored every 4 h to accurately stage the larvae, and the newly molted larvae were traced and selected for the experiment at specific developmental stages. The formed pupae were sexed and kept in the plastic containers until eclosion. After eclosion, the moths were fed a 10% honey solution for mating and oviposition.

To perform starvation treatment, five developmental stages were selected, namely, L5-48 h, L6-0 h, L6-12 h, L6-24 h, and L6-48 h, to determine the entrance time of metamorphosis and the CW. At each designated stage, approximately 15 larvae per group were weighed and then transferred to the new plastic containers containing a 2% agar–water medium to induce starvation. For the control group, larvae at the same developmental stage were fed the artificial diet. All experiment larvae were traced and weighed daily until pupation or death. The final images of the larvae in each group were captured using a Nikon stereo microscope (SMZ745T).

RNA isolation was performed with the TRIzol reagent (Thermo Fisher Scientific, Waltham, United States), according to the manufacturer’s instructions. The integrity and concentration of the RNA extracts were assessed using a NanoDrop 2000 spectrophotometer (Thermo Scientific, Wilmington, DE, United States) and an Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, United States). After the premise of guaranteed RNA quality, an equal quantity of 1.0 μg of total RNA from each sample was reverse-transcribed into cDNA using the HiScript III 1^st Strand cDNA Synthesis Kit (+gDNA wiper) (Vazyme, China) for subsequent quantitative real-time polymerase chain reaction (qRT-PCR).

qRT-PCR was performed using the AceQ Universal SYBR qPCR Master Mix (Vazyme, China) with three technical replicates on the QuantStudio 5 Real-Time PCR System (Thermo Fisher Scientific, Waltham, United States). Gene-specific primers used for qRT-PCR analysis are listed in Supplementary Table S1. The relative mRNA expression levels of the target genes were calculated with the 2^−ΔΔCT method (Livak and Schmittgen, 2001) and normalized to the expression of the housekeeping genes RPL-10 and α-tubulin (Han et al., 2021).

The PG and BCC were dissected from larvae at L4-24 h, L5-24 h, and L6-24 h. After washing with cold phosphate-buffered saline (PBS) (Coolaber, China), the tissues were fixed in 4% paraformaldehyde at 4°C for 30 min. The fixed tissues were then mounted on the microscope slide and stained with DAPI (Servicebio, China) for 1 h. The slides were visualized on an Olympus stereo microscope (SZX16-DP74) to examine the morphological changes.

To perform the RNA sequencing of PG and BCC, 15 larvae at specific developmental stages were selected as one biological replicate, and three biological replicates of each group were prepared. After the RNA isolation of the collected samples, the integrity of the RNA extracts was assessed, and the samples that met quality standards were used to construct the cDNA libraries for transcriptome sequencing analysis (Frasergen, Wuhan, China). In brief, poly(A)+ mRNA, enriched by immobilized Oligo dT, was fragmented and reverse-transcribed into double-stranded cDNA. The ends of these cDNA were blunted, and an overhang “T” was added at the 3′ end. The products were then denatured to form single-stranded DNA and looped with a bridge primer to generate the single-stranded circular DNA libraries.

After the construction of the DNA libraries, the quality was evaluated by measuring concentrations and size distributions. The concentrations were measured by qPCR, and the size distribution of the libraries was analyzed using the Agilent 2100 Bioanalyzer system (Agilent, United States). Once the library passed quality control, they were loaded onto a sequencing chip and sequenced using an MGI high-throughput sequencer. Subsequently, adapters, ambiguous nucleotides, and low-quality sequences were removed, and the obtained clean reads were aligned with the reference genome of S. frugiperda (GenBank ID: GCA_023101765.3). HISAT2 software was used to align the spliced reads with sequences in the reference genome, achieving a total mapped ratio of over 82% for the BCC and PG transcriptomes, indicating that the spliced reads belonged to S. frugiperda. In addition, multiple mapped and discordant reads were screened from the raw data and removed, and the remaining alignments were examined for the significance analysis of differential gene expression using DESeq2 software.

Small interference RNAs (siRNAs) efficiently induce gene silencing in other lepidopterans (Fu et al., 2022). The siRNA sequences targeting shadow were designed using siDirect 2.1 (https://sidirect2.rnai.jp/) based on the verified sequences (accession number: LOC118269113; sequences listed in Supplementary Table S1). These siRNAs were chemically synthesized by Shanghai GenePharma Co., Ltd. (Shanghai, China) and dissolved in diethyl pyrocarbonate-treated water (Milli-Q grade) to a concentration of 4 μg/μL. Forty larvae were randomly selected for each group, with a total of 80 larvae used in the experiment. When the larvae molted into L4-0 h, a volume of 1 μL of the siRNA solution was injected into the ventral part of the intersegment membrane between the first and second thoracic segments using a micro-syringe. The negative control (NC)-siRNA was used as a control. The injection was repeated 24 h later on the same individuals to enhance the effectiveness of the treatment. The treated larvae were maintained separately in the laboratory under standard conditions and observed daily for potential phenotypic changes induced by target gene silencing. Meanwhile, 15 PGs in L4-44 h RNAi larvae and 10 PGs in L4-68 h RNAi larvae were dissected to assess the RNAi efficiencies at specific developmental stages.

During the late larval instars, the CW is a crucial size assessment event that determines whether a larva is ready to progress to the next developmental stage or requires further growth (Mirth and Riddiford, 2007). However, the CW of S. frugiperda has not been previously measured. To establish a precise weight range for the CW, starvation treatment was carried out to mimic the size assessment process of S. frugiperda. Hatching larvae were fed the artificial diet until they developed into the specific stages selected for the treatment. Subsequently, these larvae were provided with a water diet consisting of 2% agar until they either began the metamorphic molt or succumbed to starvation. The water diet was used as a supply of water to keep the larvae hydrated and prevent death from thirst. Generally, larvae in the laboratory cease growth and enter a wandering phase between 48 and 72 h after molting into the L6 instar, approximately 60 h before pupation (Figure 1A). During starvation treatment, the larvae could hardly initiate metamorphosis or form pupae until they developed into L6-12 h. However, most of the experiment larvae at L6-12 h failed to complete pupation and died with a delay of 1–2 days (Figure1 A). Notably, half of the larvae at L6-24 h (8/16) managed a normal time course to metamorphosis, albeit at a smaller size (Figure1 B). Almost all larvae at L6-48 h (14/15), after an additional 24 h of feeding, initiated metamorphosis at normal times, suggesting that they had attained the post-critical weight (Figure1 A). Thus, the threshold size, leading to a timely metamorphic molt, appears to be ranged amongst the larvae at L6-24 h. It is worth noting that the weight loss of the larvae occurred after 1–2 days of starvation (Supplementary Table S2), which may depend on the amount of food remaining in their guts. Therefore, the last measured weight before the onset of weight loss of the starved larvae was selected to estimate the CW. The precise weight range for the CW is estimated to be between 0.348 g and 0.367 g.

The CW represents a developmental “check-point” that, once reached, triggers an endocrine cascade leading to the cessation of feeding and the initiation of metamorphosis (Nijhout and Callier, 2015). It has been reported that this metamorphic checkpoint is associated with a decrease in JH titers, driven both by the inhibition of JH biosynthesis and the emergence of JH-specific esterases in the hemolymph (Mirth and Riddiford, 2007; Nijhout and Callier, 2015). However, larvae experiencing delayed metamorphosis did not immediately cease their growth but instead showed a gradual decrease in weight before finally undergoing metamorphosis. This suggests a degree of flexibility in the endocrine system during this transition.

To further explore this, a developmental gene expression profile was carried out. The expression of the pupal specifier gene, BR-C, was transiently detected before the L6 instar and significantly increased from L6-24 h. BR-C expression peaked at the onset of pupation, coinciding with the pupal development (Figure 2A), indicating its conserved role in S. frugiperda. Myo maintained a basal expression level before L4-24 h but sharply increased at L6-24 h and reached the top until the completion of pupation (Figure 2B). This developmental gene expression profile implied that the period between L4-24 h and L6-24 h might be associated with some expected physiological changes necessary for the initiation of metamorphosis.

Morphology of the main endocrine organs was also identified to seek out more clues at this period. The major components of the neurosecretory system—BCC and PG—were dissected from the L4, L5, and L6 instar larvae for morphological characterization. These organs showed a significant increase in size as development progressed through the late larval instars. However, aside from the increase in size, no notable morphological changes were observed (Figures 2C–G).

To explore molecular changes associated with metamorphosis, RNA-seq of BCC and PG from larvae at L4-24 h and L6-24 h was further done. The raw data on RNA-seq were uploaded to the NCBI under the accession number PRJNA1173108. Using a p-value threshold of 0.05 (FDR or adjusted p-value) and an absolute log₂ (fold change) value of 1.0, 2,453 and 2,622 differentially expressed genes (DEGs) in the BCC and PG were identified, respectively (Figures 3A, D). Among them, 1,167 DEGs in the BCC and 1,351 DEGs in the PG were upregulated in the L6-24 h versus L4-24 h group, while 1,286 DEGs in the BCC and 1,271 DEGs in the PG were downregulated (Figures 3B, E).

To confirm the accuracy of the RNA-seq findings, 22 transcripts were randomly selected from each transcriptome for qRT-PCR analysis. Out of the 12 genes selected from the BCC transcriptomes and 10 from the PG transcriptomes, expression measurements were successfully obtained. The results of qRT-PCR were consistent with those of RNA-seq for seven genes in the BCC transcriptome and six genes in the PG transcriptome, confirming the reliability of the RNA-seq analysis (Figures 3C, F).

JH and 20E are the two key hormones that coordinate the metamorphic processes, acting antagonistically (Truman and Riddiford, 2019). To investigate the changes in these two hormones around the metamorphic phase, transcripts of 20E pathways were first identified in the transcriptomes, including those with insignificant changes (Figure 3D). The Halloween genes, which are involved in 20E biosynthesis—spook, phantom, disembodied, and shadow—were downregulated in L6-24 h compared with L4-24 h in the PG transcriptomes. In addition, another Halloween gene, shade, which is expressed in the peripheral nervous system and involved in the conversion of ecdysone to its active form, 20E, was also downregulated in L6-24 h compared with L4-24 h in both transcriptomes.

The downward trends of these Halloween genes implied a reduction in 20E synthesis, which was consistent with the observed downregulation of the transcript of EcR and its downstream effectors, ecdysone-inducible protein 75 (E75) and ecdysone-induced protein 78C (E78C), in L6-24 h PG transcriptomes. However, the transcript levels of the second partner of the ecdysone receptor, Ultraspiracle (USP), and one of its downstream effectors, ecdysone-induced protein 74 (E74), were upregulated in the PG transcriptomes. Interestingly, the expression of cytochrome P450 protein 18A1 (CYP18A1), a key cytochrome P450 enzyme involved in the inactivation of 20E (Guittard et al., 2011), was sharply reduced in the BCC transcriptomes upon developing into L6-24 h. Additionally, PTTH and its receptor torso, which stimulated PG glands to produce ecdysone, were detected as insignificant in the BCC transcriptomes, further supporting the inactivation of 20E synthesis at this stage.

Transcripts of JH pathways were also identified in the transcriptomes (Figure 3H). As the critical rate-limiting enzyme for JH biosynthesis, JH acid o-methyltransferase (jhamt) controls JH levels in most insects. It was only detected in the BCC transcriptomes and downregulated in L6-24 h compared with L4-24 h in the BCC transcriptomes. Three enzymes that generate the JH direct precursor methyl farnesoate, namely, cytochrome P450 protein 15A1 (CYP15A1), farnesol dehydrogenase (FOHSDR), and farnesyl diphosphate phosphatase (FPPP), showed a lower expression level in the BCC transcriptomes at L6-24 h, implying that JH biosynthesis was not activated at the time. Meanwhile, the transcripts of JH esterase (Jhe) and JH epoxide hydrolase (JHEH), two enzymes that efficiently degrade JH, were barely detected in either transcriptome, which hinted that JH clearance in these organs had not been initiated yet.

The downward trends of 20E and JH clearance transcripts in L6-24 h compared with L4-24 h might provide insights into the physiological status in the late larval instars. To seek more details of the status, functional annotation was performed in both transcriptomes, and some enriched DEGs in the transcriptomes could be clustered in a few Gene Ontology (GO) terms, which showed a strong tendency in the expression pattern.

In the BCC transcriptomes, almost all DEGs identified in the GO term “Ligand–receptor activity” were downregulated in L6-24 h compared to L4-24 h in the BCC transcriptomes. This included receptors such as dopamine receptors (DRD1N and DRD2), octopamine receptors (Octalpha and Octbeta), thyrotropin-releasing hormone receptors (TRHR), neuropeptide receptors (NPSIFR2), and G-protein-coupled receptors (MTH4, DopEcR, MTH3, and GPR84). In addition, members of the nuclear receptor superfamily regulated by 20E, such as fushi tarazu factor-1 (FTZ-F1), hormone receptor 3 (HR3), and hormone receptor 38 (HR38), were also identified in this term (Figure 4A). “Hydrolase and peptidase activity” term, on the other hand, enriched all upregulated DEGs in the BCC transcriptomes, comparing L6-24 h and L4-24 h, and some of them are peptidases, like aminopeptidase (ANPEP), zinc carboxypeptidase (CPO), pro-aminopeptidase (pepP), and carboxypeptidase Q (CPQ). Others like trypsin II, chicken fucosyltransferase 1 (CFT-1), alkaline C, kallikrein 1 (KLKB1), and serine protease 7 (Sp7), are serine-type endopeptidases participating in protein degradation (Figure 4B).

In the PG transcriptomes, there was a cluster of DEGs identified in “proteasome activity” that were downregulated in L6-24 h compared with L4-24 h. Most of them were 26S proteasome subunits, and several were E3 ubiquitin–protein ligases, including SIAH1, ring finger protein 34 (RNF34), ring finger protein 5 (RNF5), potassium channel modulatory factor 1 (KCMF1), ariadne-1 (ARIH1), ring box protein (RBX1), and ring finger protein 81 (RNF181) (Figure 4C).

Most enzymes involved in 20E and JH syntheses decreased as larvae developed into the CW, suggesting an endocrine switch linked to metamorphosis. The inhibition of JHAMT or other biosynthetic enzymes in the late larval stages induces precocious metamorphosis in B. mori (Daimon et al., 2012; Daimon et al., 2015), and the phenotypes resulting from disrupted JH signaling by the knockdown of Met or Kr-h1 have been well-described in the larvae of B. mori and T. castaneum (Daimon et al., 2015; Konopova and Jindra, 2007), reinforcing the idea that JH biosynthesis is required to maintain the normal larval development in the late larval stages. However, the molecular mechanism of 20E synthesis during this stage has not been fully discussed. It has been proven that the expression of disembodied and shadow was restricted to PG as the critical rate-limiting enzymes of 20E synthesis (Gilbert, 2011; Pan et al., 2021). Moreover, shadow could be efficiently knocked down in lepidopterans (Peng et al., 2019). To better understand the biological process, siRNAs were used to knock down the shadow gene, which was highly and specifically expressed in PG (Figure 5A). The siRNAs were injected into newly molted L4-0 h larvae to suppress shadow expression, and a second injection was performed after 24 h to enhance the knockdown effect. Typically, the L4 instar lasts 48–56 h under laboratory conditions, and control larvae molted into L5 at approximately 50 h in the experiment. Thus, L4-44 h was chosen for sampling. The PGs from 15 injected larvae were dissected for total RNA isolation at this time; then, the knockdown efficiencies were confirmed via qRT-PCR. As shown in Figure 5B, the knockdown efficiency exceeded 50%, significantly inhibiting the growth of the injected larvae. Twelve out of 15 shadow RNAi larvae extended their growth in L4 for an additional 24–72 h compared to the control larvae. Ten RNAi larvae at L4-68 h were subsequently sampled. Interestingly, compared to the control larvae at L4-44 h, many of the shadow RNAi larvae at both L4-44 h and L4-68 h displayed a melanized phenotype (9/15), potentially due to the interactions between the hormone and melanin pathway, and most RNAi larvae experienced developmental arrest after the injection (Figure 5C). Finally, fewer than 20% of injected larvae (2/15) survived to molt into the L5 instar.

The differential expression patterns of genes associated with development were also analyzed to assess the impact of shadow depletion (Figure 5D). EcR expression decreased sharply and stayed below 50% of its control level, implying that the 20E signal was significantly reduced after the silencing of shadow. The expression level of jhamt was also affected by the shadow RNAi and decreased to 20% of its control level. Interestingly, the transcript of the JH-inducible gene, krüppel-homolog 1 (Kr-h1), was temporarily activated in L4-44 h but returned to a lower level 24 h after the knockdown of the shadow gene. The potential “metamorphosis initiation factor” Myo was inhibited and maintained at a low level after shadow knockdown. To better understand the molecular mechanisms behind the melanized phenotype observed in the shadow RNAi larvae, the expression levels of dopa decarboxylase (DDC), pro-phenoloxidase 1 (PPO1), and pro-phenoloxidase 2 (PPO2) were measured. Remarkably, after knocking down the shadow gene, the expression of the PPO1 gene increased approximately 10-fold. At the same time, PPO2 was downregulated after knocking down the shadow gene. Interestingly, the expression of the DDC gene fluctuated in the control larvae, suggesting potential instability in the melanin production pathway. Previous studies in Ae. aegypti found that JH suppression does not induce the melanized phenotype in the late larval stages (Zhu et al., 2019). In this study, after knocking down the shadow gene, the differentially expressed jhamt and Kr-h1 genes indicated that the antagonistic regulation between JH and 20E is out of control (42). This dysregulation contributed to the upregulation of the PPO1 gene, resulting in a melanized larval phenotype. Additionally, the newly reported larval master gene, chronologically inappropriate morphogenesis (chinmo) (Reynolds, 2022), was temporarily inhibited after shadow RNAi, which might support the slow growth of the shadow RNAi larvae.