In multicellular organisms, development entails the progression of diverse cellular processes that need to be temporally coordinated. During Caenorhabditis elegans postembryonic development, the events of molting and cell divisions progress in parallel during four larval stages and are modulated by external cues such as temperature and nutrient availability. While seam cell divisions occur predominantly before ecdysis of the cuticle, the order of the events can change, suggesting that they are controlled by independent mechanisms. Here, we have analyzed the impact of reduced insulin signaling on molting and in stage-specific cell divisions. We find that reduced insulin signaling in the daf-2(e1370) allele delays both events but has a larger impact on the timing of cell divisions, thus increasing the desynchrony between the events. The relative delay in seam cell divisions leads to a delay in the initiation of the subsequent stage of the molting program, providing a mechanism for resynchronization of developmental processes at the beginning of the larval stages.

A lag phase in the initiation of larval stages

Analysis of postembryonic development in C. elegans has often relied on the observation of morphological changes. Here, we applied a high-throughput, quantitative method to measure development [11] that allows unveiling unknown complexities on the timing of the different larval stages [7,12,13]. This method relies on the use of the enzyme Luciferase, which is constitutively expressed in the C. elegans reporter strain sevIs1 [Psur-5::luc+::gfp] [11]. Luciferase catalyzes the oxidation of Luciferin, in a reaction that emits light [14]. We provide Luciferin in the media of the developing larvae, so that luciferin intake results in emission of light. At the beginning of the molts, larvae produce a cuticular plug that blocks the mouth, preventing the entry of luciferin. As a consequence, the emission of light decreases rapidly, generating a characteristic profile that is used to calculate the duration of each larval stage, allowing identification of intermolts (I) and molts (M) (Fig 1A) [11].

In the analysis of developmental phenotypes of the daf-2(e1370) allele, we observed a novel feature of postembryonic development. In addition to the slow development previously described for this mutant, we detected that, after the end of the molt, many larvae present a lag phase before the resumption of the sustained high signal that is a feature of the intermolts (Fig 1B, middle and right). To quantify this feature of development, we defined the periods L2lag, L3lag, and L4lag, which respectively appeared at the beginning of the L2, L3, and L4 (Fig 1C). While initially most evident in daf-2 mutants, detailed analysis of the wild-type luminescence traces also allowed to identify Lag periods (Fig 1D). To analyze whether the reduction of the luminescence signal during the lags stems from a reduced expression of the luciferase protein under the promoter of sur-5, we measured GFP in animals collected during the first molt (M1), L2lag, and the second intermolt (I2). The three categories showed similar levels of GFP expression, suggesting that the decrease of the signal is not a consequence of reduced luciferase expression (Fig 1E). Since the level of luminescence signal during L2lag is similar to that of molts, and molts involve behavioral quiescence, we tested this variable in the same categories as before. Larvae in L2lag move significantly more than those in M1, and less than larva in the second intermolt (Fig 1F). Throughout these experiments, we could visualize larvae in the L2lag stage. These larvae showed no apparent morphological differences relative to the other stages and did not present any cuticular remnants. Furthermore, to confirm that L2lag is not construct-specific, we analyzed another luciferase strain with a single-copy insertion and a different promoter driving the expression of luciferase [4]. L2lag and larval stages showed the same duration in both the wild-type and daf-2(e1370) backgrounds (S1A Fig).

We then compared the levels of lags in the wild type and the daf-2(e1370) mutant. The absence of this feature in some individual larvae is denoted by a duration of “0” hours. The lags in the wild-type strain appeared at a lower frequency (approximately in 60% of wild-type larvae versus 90% of daf-2) and with shorter duration than in daf-2 (Fig 1G). Here, we focus our analysis on L2lag, as this is the stage with the longest average duration, especially for the wild-type strain. We compared the L2lag with the previous and subsequent stages, M1 and bona fide I2, respectively, with bona fide I2 being the duration of the second intermolt starting after the L2lag period (Fig 1C). A key feature of L2lag is that, unlike molts and intermolts, the duration of these periods shows high inter-individual variability. The coefficient of variation increases around 10-fold for L2lag compared to that of the previous and following stages (Fig 1H). We also calculated, for each larva and for the different larval stages, the duration relative to the average (in fold change). This value is close to 1 for M1 and bona fide I2, reflecting the low variability in the duration of these stages, but it ranges from 0 to 4 for the L2lag stage (S1B Fig).

When we compare the relative duration of L2lag and bf. I2 in individual larvae, we observed no correlation between these values (S1C Fig). We also checked the correlation between the duration of each of the stages of development for the wild type and daf-2. As previously shown by us and others, larvae show some degree of correlation in the duration of most stages [7,15], but the duration of the lag periods (L2lag, L3lag and L4lag) showed, in general, no correlation with the duration of any of the stages (S1D and S1E Fig). These results show that the duration of the lags does not reflect the interindividual changes in the duration of other developmental stages, that is, L2lag is not a consequence of the presence of larvae with overall faster or slower development. Indeed, for the larvae that present L2lag, if we represent the duration of the complete intermolt, strictly considering the time between molts, its duration is significantly higher than that of larvae that do not enter this stage. By subtracting the duration of the L2lags, we obtain a bona fide I2 intermolt duration that is statistically not different from the average I2 (S1F Fig). All the previous results support that L2lag is a period that is added to the (bona fide) intermolt, as opposed to a feature of the luminometry signal accommodated during the duration of the regular intermolt.

Lags occur at the beginning of larval stages, the same moment when a nutritional checkpoint was previously observed [16], and also coincide with the phase of arrest of developmental oscillations in gene expression [4]. We therefore tested the expression of the gene dpy-9 as a reporter of these oscillations [4]. We performed time-lapse microscopy of single larvae contained in 275 × 275 × 30 μm hydrogel microchambers [17] to monitor the expression of dpy-9 in the wild type and daf-2 mutants during the complete L2 stage. On average, the increase of dpy-9 signal was delayed in the daf-2 mutant (Fig 1I). However, as the duration of the L2 stage is different between the wild type and the daf-2 mutant, and among individual larvae of the same genotype, we plotted the data rescaled to the duration of each individual larva. Even after rescaling, we found that the signal increased later in daf-2 mutants (Fig 1J). This data supports the existence of delay in oscillatory gene expression at the beginning of L2 in daf-2 mutants. All results together show that the L2lag has features that are different to that of molts and intermolts, with a higher CV and a lack of correlation with the duration of other developmental stages. Furthermore, this lag could coincide with a delay in the initiation of the gene expression characteristic of the second larval stage.

The L2lag phenotype of daf-2(e1370) is partially independent from the rest of IIS

To investigate the regulation of L2lag, we focused on two relevant regulators of IIS. Most daf-2 phenotypes are mediated by the activation of DAF-16 (Fig 2A), and they are rescued by the null mutation daf-16(mu86). Thus, phenotypes observed in daf-2 mutants as constitutive dauer formation, life span extension, and the increased duration of the L2 stage, are absent in the double mutant daf-2(e1370);daf-16(mu86) [9,18,19]. As expected, we observed that both daf-16 and the double daf-2;daf-16 mutants show a duration of the second bona fide intermolt similar to the wild-type strain (Fig 2B). Therefore, the extension of this stage in the daf-2(e1370) mutant requires the function of DAF-16. When we quantified L2lag, we observed that daf-16 has similar levels to the wild-type strain, but the double mutant daf-2;daf-16 shows L2lag levels that are intermediate to the daf-2 and the daf-16 mutant (Fig 2C). The daf-16 mutation also reduced the levels of L3lag and L4lag of daf-2 (S2A and S2B Fig). These results show that part of the impact of the daf-2 mutation is independent of DAF-16. Also, it reinforces the idea that the L2lag phenotype is independent of other developmental phenotypes of daf-2(e1370).

Looking for additional regulators of the process, we focused on DAF-18, the C. elegans homologue of the tumor suppressor PTEN, which counteracts insulin signaling by dephosphorylating PIP₃, opposing the effect of AGE-1 [20] (Fig 2A). Furthermore, PTEN can dephosphorylate the Insulin Receptor (IR) and the IR inhibits PTEN [21]. We tested the effect of the deletion allele daf-18(ok480) in the developmental phenotypes observed previously. The daf-18 mutation does not provoke any changes in the duration of the second intermolt in a background wildtype for daf-2 [daf-2 (+)] and completely suppresses the daf-2 mutant phenotype [daf-2 (−)] (Fig 2D). For the regulation of L2lag, the daf-18 mutant reduces the duration measured in the wild-type background [daf-2 (+)] (wild type versus daf-18 p-value = 0.0018) and the daf-2 mutant background [daf-2 (−)] (Fig 2E). In the daf-2 mutant background, suppression by daf-18 is stronger than that by daf-16 (Fig 2E). Despite this suppression, the daf-2 mutation increases the duration of L2lag in all backgrounds (Complete statistics in S2 Data), even in the daf-16;daf-18 mutant (t test p = 0.002) (Fig 2E), which suggests that DAF-2 also signals to regulate L2lag independently of DAF-18 and DAF-16.

Mutations in heterochronic genes impact L2lag

L2lag represents a novel feature of the timing of development. To investigate the relationship of L2lag with other developmental events, we focused on the heterochronic gene pathway, that controls the timing of cell divisions during postembryonic development. We first addressed the effect of heterochronic genes that impact stage-specific larval events. lin-14 mutants or RNAi knockdown leads to skipping the L1-specific cell fates. At the level of seam cell divisions, this means that the first intermolt to occur after hatching shows the characteristic symmetric cell divisions of the wild-type L2 stage. lin-28 mutants or RNAi leads to skipping the L2 cell fates (reviewed in [2]) (Fig 3A). We previously tested the effect of these RNAi treatments in the duration of the larval stages and did not observe major alterations on the duration of the stages in the wild type [7]. When we tested the effect of these mutations in L2lag, we observed that lin-14 RNAi led to a significant reduction compared to the control RNAi, in both the wild type and daf-2. This result could suggest that L2lag is also dependent on the L2 symmetric division. However, lin-28 RNAi treatment which does not affect L1 events but leads to skipping L2-specific cell fates, did not alter the duration of L2lag (Fig 3B). These results suggest that L2lag is not related to the L2-specific symmetric seam cell divisions, but it might be a consequence of L1-specific events.

Next, we investigated the impact of another heterochronic gene, lin-42, in L2lag. lin-42 mutants show precocious terminal differentiation [22]. However, unlike lin-14 and lin-28, lin-42 is expressed periodically during larval molts and its function impacts all larval stages regulating the reiterative process of molting [23–25] (Fig 3C). Two lin-42 mutant alleles reduced the duration of the L2lag, but the mutation in ok2385 had a stronger effect than n1089 (Fig 3D). In summary, our results show that the duration of L2lag is impacted by genes of the heterochronic pathways that control the timing of seam cell divisions and molting.

daf-2(e1370) increases desynchrony between ecdysis and seam cell divisions

To investigate the relationship between the timing of both molting and cell divisions in the daf-2 mutants, we focused on seam cells. Seam cells divide asymmetrically at the beginning of each larval stage, and show an additional symmetric division at the L2 stage, prior to the asymmetric one [1]. Specifically, we monitored larvae from hatching to the second larval stage. We performed time-lapse microscopy of single larvae contained in 275x275x30 μm hydrogel microchambers [17]. This setup allows direct observation of hatching and ecdysis as markers of the molting process. Using strains with the reporter matIs38, we monitored seam cell divisions of the V lineage. First, we measured the timing of the first two seam cell divisions, hatching and the first ecdysis (Fig 4A). We analyzed these markers in the wild type, the daf-2(e1370) mutant, and the double mutant daf-2;daf-18. The first seam cell division (div 1) took place at a similar time relative to hatching for the wild type and daf-2 but the following event, ecdysis 1 (ecd 1), was delayed in the daf-2 mutant (Fig 4B). In the wild type, the second seam cell division (div 2) occurred mostly around ecdysis for V1-V4, but V6 usually divides after ecdysis. Div 2 is also delayed in the daf-2 mutant, but to a larger extent than ecdysis 1. The differential impact of the daf-2 mutation on the different events creates relative desynchronization between ecdysis 1 and div 2 (Fig 4D). The daf-18 mutant showed a level of desynchrony similar to that of the wild type (S3A Fig). However, in the daf-2 background, the daf-18 mutation accelerated divisions so that, in the daf-2;daf-18 mutant, most divisions occur before ecdysis (Fig 4B and 4D). The average timing of the last divisions (V6 lineage) in the wild type occurred at 0.65 hours after ecdysis. For the daf-2 mutant, this event happened 2.55 hours after ecdysis (t test; p < 0.0001) and for the daf-2;daf-18 mutant at 0.22 hours after ecdysis (t test; p = 0.04). These results show that reduced insulin signaling impacts the molting and cell divisions differently, leading to a desynchronization between them. This desynchronization occurs around the transition from L1 to L2, the moment when L2lag takes place. To understand whether seam cell delay (relative to ecdysis) could be related to the L2lag phenotype, we compared their durations. We plotted the L2lag duration against the V6 div 2–ecd 1 interval for wild type, daf-2, and daf-2;daf-18 animals. We observed a strong positive linear correlation between these variables across genotypes (R² = 0.9981, p = 0.0391), demonstrating that variation in the duration of the L2lag could reflect changes in the timing of seam cell divisions relative to ecdysis (Fig 4E). We also performed the same analysis for the following larval transition, that is, at the end of L2 when ecdysis 2 (ecd 2) and fourth seam cell divisions (div 4) take place (Fig 4C). Overall, we observe the same effect of the daf-2 and daf-18 mutations, but the desynchrony is less pronounced than at the previous stage (Fig 4F), consistently with the overall low levels of L3lag (S3B Fig).

In these experiments, we could observe pumping after ecdysis, independently of the state of seam cell division. Furthermore, we measured the length of the larvae before and after ecdysis, while monitoring seam cell division, and observed that larvae grew at a constant rate after ecdysis, independently of the timing of cell divisions (S3C Fig). For the same group of larvae, the growth rate was similar between ecdysis and div 2, and after div 2 (S3D Fig).

Signaling from DAF-2, which is modulated by DAF-18, converges on the transcription factor DAF-16/FOXO, which contributes to cell arrest upon nutrient removal during development [16]. We investigated whether DAF-16 also plays a role in delayed cell division of the daf-2(e1370) mutants in food-replete conditions. We assessed activation of DAF-16 by analyzing the subcellular localization of DAF-16::GFP in the wild-type and daf-2(e1370) mutant backgrounds around the first molt. While the wild-type background showed mainly cytoplasmic localization of DAF-16, the daf-2(e1370) mutant showed intermediate/nuclear localization (S3E Fig). To link DAF-16 activation in daf-2(e1370) to delayed seam cell division in fed conditions, we performed time-lapse experiments with the wild type, daf-2 and daf-2;daf-16 mutants. The daf-16 mutation accelerated seam cell divisions of daf-2(e1370) similarly to their impact on L2lag (S3F and S3G Fig). However, the acceleration from the daf-16 mutation was not as strong as that provoked by the daf-18 mutation, as the average timing of the last divisions (V6 lineage) of the daf-2;daf-16 (0.35 hours) was close to that of the wild type (0.12 hours, t test p = 0.3). In summary, daf-16 and daf-18 mutations reduce desynchrony between ecdysis and seam cell divisions of the daf-2(e1370) similarly to their effect on L2lag and other daf-2 phenotypes.

We then analyzed another cellular process that takes place around the end of the L1 stage, endomitosis of intestinal cells. During endomitosis, cells enter the M phase of the cell cycle but do not undergo cytokinesis. Endomitosis occurs during L1 in 14 of the 20 intestinal cells, producing binucleated cells [26]. The mutation daf-2 does not delay the timing of this event relative to ecdysis (S3H Fig), suggesting that this mutation only impact the timing of some cellular events.

Desynchrony between developmental events results in L2lag

L2lag and the desynchrony between cell divisions and molting coincide in time, and their magnitude correlates at the population levels when comparing genotypes. Next, we investigated whether the two processes correlate in individual larvae by analyzing the state of seam cells during L2lag. We combined the reporter used for quantification of development by luminometry with a reporter for seam cell division (matIs38). We then initiated development of daf-2 mutants in a luminometer, followed the luminescence profile in real time until the end of the first molt, and collected larvae from three groups. The first group showed a luminometry profile corresponding to the L2lag, with decreasing signal after ecdysis. The second group showed a profile corresponding to a regular intermolt, in this case I2, as determined by a stable, high signal after ecdysis. The third group showed a profile corresponding to a regular intermolt after a period of L2lag (Fig 5A). For each of the larvae, we then scored the division of seam cells V1-V4 and V6. The L2lag group showed a significantly lower fraction of larvae with divided seam cells than the I2 groups, regardless of whether they had previously passed through L2lag (Fig 5A). Specifically, only 1 of 9 larvae of the L2lag category had completed all 10 seam cell divisions (V1-V4, V6; left and right). This result suggests that L2lag could indeed be a consequence of the desynchrony between the molting and cell division processes. Additionally, we measured luminescence over the L1 to L2 transition in a microscope, coupled to imaging of both ecdysis and seam cell divisions. We confirmed that ecdysis always occurs linked to the increase of signal after approximately two hours of reduced luminescence signal. Then, larvae that reached ecdysis before completing seam cell division showed, in general, lower levels of luminescence (S4 Fig). When the different categories of larvae were averaged, there was a clear difference in bioluminescence between animals that had already finished seam cell divisions and those that did not. In particular, the larvae that had not completed divisions before ecdysis showed a reduction in the signal that, after approximately 2.5 hours reached again the levels of luminescence measured after ecdysis (Fig 5B). This result suggests that delayed developmental events, evidenced by seam cell divisions, lead to entry into L2lag.

To test this hypothesis, we experimentally tested conditions that modified the timing of seam cell divisions relative to ecdysis. We expected that advancing seam cells divisions would reduce L2lag and delaying seam cell divisions would increase it. We first advanced seam cell divisions relative to ecdysis by growing the larvae on Escherichia coli HT115, typically used for RNAi. HT115 accelerates the timing of postembryonic development [27], but its role on seam cell division has not been assessed. While feeding with HT115 might impact additional processes, we could establish that HT115 significantly advanced seam cells divisions (div 2 relative to ecd 1) compared to OP50-1, in both the wild type and daf-2(e1370) (Fig 5C), without impacting the timing of ecdysis (S5A Fig). This led to a situation where most divisions occur before ecdysis (Fig 5D). Indeed, when we measure L2lag, we found a significant reduction in both the wild type and daf-2(e1370) (Fig 5E), again, without altering the duration of L1 measured by luminometry (S5B Fig).

We also tried to delay the cell divisions by treating developing larvae with hydroxyurea (HU), which prevents DNA synthesis by replication fork arrest. Since the delay in divisions could also feedback to the molting process, we aimed for concentrations of HU that allowed complete development and did not increase the duration of the first larval stage. HU at 4 and 8 mM allowed progression through postembryonic development (S5C Fig), and only 4 mM permitted developmental progression without significant delay of the first larval stage (S5D Fig). As before, treatment with HU might impact other processes. To confirm the impact of HU on the timing of seam cell divisions relative to ecdysis, we followed these two features on the larvae treated with 4 mM HU. The treatment with HU significantly delays divisions relative to 0 mM HU (Fig 5F), without impacting the timing of ecdysis (S5E Fig). This provokes that many cell divisions occur after ecdysis (Fig 5F–5G). Indeed, at 4 mM HU the L2lag was significantly prolonged compared to the control condition (Fig 5H). For both HT115 feeding and treatment with HU, we plotted the L2lag duration against the timing of seam cell division, finding a significant positive correlation (R² = 0.9339, p-value = 0.0064) between their durations across conditions (Figs 5I and S5F). These results show that varying the relative desynchrony between ecdysis and cell divisions modulates the duration of L2lag.

Interestingly, this also explains our findings on how lin-42(ok2385) mutation impacts L2lag. Previous research showed that the lin-42(ok2385) increased the duration of the L1 stage but showed a timing of seam cell divisions similar to the wild-type, resulting in advanced cell divisions relative to ecdysis [23,24], which is consistent with the reduced L2lag we find in this mutant (Fig 3D). This suggests that the effect of LIN-42 on L2lag is likely related to its molting clock function [25].

Finally, we asked whether increased L2lag is a general principle for other interventions that affect the duration of development. C. elegans development shows a strong temperature dependence [28]. We focused on this environmental cue as, despite the slowed developmental rate at lower temperature, the events of cell divisions and ecdysis occurred at the same time relative to the duration of the total development [6]. This means that reduction of speed due to temperature does not involve increased desynchrony between ecdysis and cell divisions. Since we previously characterized the regulation by temperature of postembryonic development for temperature between 10 and 27 °C [7], we could extract information about L2lag from that dataset. Despite the change in developmental rate, L2lag showed similar values at all temperatures tested (Fig 6A). When analyzed in time-lapse experiment, the L1 stage (hatch to ecd 1) was about 3 hours faster at 22.5 °C than at 20 °C (Fig 6B), but the timing of seam cell division relative to ecd 1 was similar between conditions (Fig 6C), consistent with the lack of differences in L2lag. Likewise, L2lag is similar between 12 and 22 °C for the daf-2 mutants (Fig 6D), as despite of acceleration of development (Fig 6E), temperature did not impact overall seam division timing (Fig 6F). This means that seam cell divisions are not per se more susceptible to environmental interventions that reduce developmental rate. Additionally, culture of daf-2(e1370) mutants at 22 °C leads to dauer formation, an alternative state to L3 stage. The similar levels of L2lag in the daf-2 mutant at 20 and 22 °C, as well as the similar levels of seam cell desynchrony, confirm that L2lag does not correlate with dauer formation. Consequently, increased L2lag is neither a trivial nor a direct consequence of slow development.

Taken together, our results support the idea that the initiation of larval stages can be delayed when events of the previous stage did not complete. As we have observed, the alteration in the order of developmental events is tolerated by the system. However, such a mechanism would prevent that desynchrony provoked by delayed events amplifies over the course of development.

Caenorhabditis elegans

C. elegans strains were cultured following standard methods [36]. Animals were grown and maintained on nematode growth media (NGM) with a lawn of E. coli OP50-1 at 20 °C. Experimental temperature was 20 °C, except when stated otherwise. All strains used are listed in S1 Table.

Bacterial strains

For maintenance plates and experiments, the food source was E. coli OP50−1. The only exceptions are Figs 3B, 5C, 5E, S5A, and S5B where we used E. coli HT115.

Luminometry of single worms

To analyze the developmental timing of single worms, we used the bioluminescence-based method previously described [11]. To obtain age-synchronized eggs, 10–15 hermaphrodites were transferred to a fresh NGM plate, allowing them to lay eggs for one hour. For experiments with six conditions or less, synchronized eggs were individually placed in wells of a 96-well plate containing 100 μl of S-basal (5 μg/ml cholesterol) with 200 μM Luciferin. After placing all embryos, 100 μl of S-basal with 20 g/l E. coli OP50-1 or HT115 was added to each well. For experiments with more than 6 conditions, the eggs were placed individually in wells of a 384-well plate containing 25 μl of S-basal (5 μg/ml cholesterol) with 400 μM Luciferin, adding 25 μl of S-basal with 20 g/l E. coli OP50-1 (or HT115 for Figs 5E and S5B) per well after all eggs were transferred. In both cases, samples from different strains or conditions were alternated to avoid local effects. Plates were sealed with a gas-permeable membrane and placed in a luminometry reader (Berthold Centro XS3) localized inside a cooled incubator (Panasonic MIR-254). Luminescence was recorded for 1 s (for 96-well plate) or 0.7 s (for 384-well plate) at 5-min intervals until reaching adulthood.

Temperature regulation

We measured the temperature in the plate using a data logger Thermochron iButton DS1921G (Maxim Integrated). The temperature in the plate is ~3.5 °C higher than that set at the incubator, due to the production of heat from the luminometer. All temperatures shown in the data correspond to the temperature experienced by the larvae.

Live time-lapse imaging of single worms

To study gene expression, progression of molting and seam cells divisions of single worms we adapted a system for time-lapse microscopy in [17]. To prepare the acrylamide-gel chambers, we used micropatterned silicon wafer with microchambers dimensions of 275 × 275 × 30 μm. The microchambers were prepared by first mixing a polyacrylamide solution containing 10% acrylamide/bis-acrylamide (29:1), 1% ammonium persulfate and 0.1% TEMED. The liquid solution was pipetted into a compartment composed by the master wafer glued to a methacrylate frame. After filling all the compartment, a glass microscope slide was placed on top to seal the compartment and let it to solidify for 0.5-2 hours. The resulting polyacrylamide gel is washed with ddH₂O at least 4 times and stored in ddH₂O.

To perform the experiments, the polyacrylamide microchamber used is previously stabilized in M9 during 1 hour prior to the experiment. To prepare the long-term imaging, a methacrylate frame is stuck to a microscope slide. The M9-stabilized microchamber is then placed on top of the slide, within the frame. To place embryos in the polyacrylamide chambers, we proceeded as previously described [37]. We obtained age-synchronized embryos by transferring 10–15 gravid hermaphrodites to fresh NGM plates and letting them to lay eggs for one hour. Using an eyelash, eggs were individually transferred to the wells of the microchamber (1 egg per well). After eggs were transferred, we collected some OP50-1 from NGM plates using the eyelash and added it to wells as food source. The only exception is for Figs 5C and S5A, that we used OP50-1 or HT115. The compartment was filled with M9 and sealed with cover slides.

We performed the time-lapse imaging in a Leica scope M205 FCA with a motorized stage, using GFP excitation/emission filters. Time-lapse was performed by taking pictures every 20 min for each worm in a Z-stack of 40 μm divided in 11 steps. Worms were photographed in two different channels (white light channel and GFP channel) with exposure times of 50 ms for both cases. The developing larvae worms were recorded for approximately 40 hours. For S3F Fig, we used strains that contain the reporter sevIs1 in addition to matIs38. Since sevIs1 expresses GFP ubiquitously, we visualized seam cells using the mCherry signal from matIs38. For this experiment, we used the Leica M205 FCA scope and a Nikon Eclipse Ti2 microscope.

Analysis of time-lapse microscopy

For the analysis of the expression of dpy-9p::GFP, we selected the best-focused Z step for every time point (GFP channel). Using ImageJ software, we calculated the expression by obtaining the average GFP intensity of 3 middle-body cells after subtracting the background. We normalized the GFP intensity levels to the average values of the first 3 hours before ecdysis. For Fig 1J, we smoothed the signal by applying a centered moving average over 1 hour (5 time points). For the analysis of molting progression, we reported hatching when we found any part of the larvae outside of the eggshell. Likewise, ecdysis is marked when the larvae partially or completely exited the old cuticle.

For the analysis of seam cell divisions, we reported each specific seam cell (V1–V4, V6) as divided when at least one of the sides presented evidence of cytokinesis. For the analysis of intestinal cell endomitosis, we reported the timing of initiation of nuclear divisions. Unlike seam cell division, intestinal cell endomitosis occurs with high synchrony throughout the larva.

Seam cell division analysis at specific larval stages

We analyzed the state of seam cell division at specific developmental stages. To do so, we perform the luminometry assay as described above (see “Luminometry of single worms”) using worms combining the bioluminescence (sevIs1) and the seam cell fluorescence-tagged reporter (matIs38). After approximately 20 hours of initiating, we monitored the larval developmental progress continuously. At a single time point, we collected larvae that were in the different categories to be analyzed. During each replicate, we extracted only 4–5 larvae from the complete 96-well plate to maintain all the processing time of the samples lower than 30 min. We collected animals that, at the time of collection, were within 0.5–1 hour from the previous transition, that is, from the end of M1 for animals in the categories “L2lag” and “I2”, and from the end of L2lag for the larvae in the category “I2 after L2lag”. Individual worms were isolated and washed from food with M9. Finally, worms were immobilized using 10 mM Levamisol diluted in M9 and placed in agarose pads. Larvae with seam cell divisions were identified using a Leica scope M205 FCA and mCherry excitation/emission filters.

Quantification of GFP and movement at specific larval stages

To collect larvae in the different stages, we proceeded as explained above. Then, for quantification of GFP, we place individual larvae in a drop of approximately 10 μl of S-basal on a microscopy slide and added 1 μl of 10 mM Levamisol. We acquired pictures using a Leica scope M205 FCA and GFP excitation/emission filters. We manually segmented the complete larvae and measured the average pixel intensity with Image J. For the quantification of movement, we placed the larvae on NGM plates, measured the tracks left by the movement of the larvae after approximately 30 min, and then normalized to the exact time for each larva.

Luminescence recording by time-lapse imaging

For luminescence measurement in microchambers, we prepared the acrylamide chambers as previously reported, but the gels are stabilized in M9 containing 2.5 mM Luciferin. After placing the eggs, the compartment is filled with the same solution. The system is transferred to a microscope Nikon Ti2 Eclipse equipped with a camera for transmitted light and RFP (Nikon DS-Qi2) and an additional camera with high-sensitivity CCD sensors (iKon-M 934 CCD; Andor). For bioluminescence, we integrated the signal from 1-min exposure. To obtain the average bioluminescence signal, we use the ImageJ software. As larvae move over the 1-min exposure, we draw a region of interest (ROI) that frames the entire well to integrate the signal from the complete acquisition time, then obtaining the average bioluminescence value per pixel. After this, we subtracted the background calculated by drawing a ROI where the worm is absent, then obtaining the average background value. In this experiment (Fig 5B), we used strains that contain the reporter sevIs1 for luminescence and matIs38 to track seam cell divisions. Since sevIs1 expresses GFP ubiquitously, we visualized seam cells using the signal in the RFP channel from matIs38. For white-light and RFP, we acquired in a Z-stack of 40 μm divided in 11 steps. For white-light, exposure time was 70 ms and for RFP, exposure time was 200 ms.

Hydroxyurea treatment

To test the effect of cell division blocking chemicals on the development, we treated worms with HU specifically during the postembryonic development. To do so, we performed the luminometry assay as described above (see “Luminometry of single worms”). The difference is that eggs were placed in wells containing 90 μl of S-basal (5 μg/ml cholesterol) with 200 μM Luciferin and after placing them, we added 90 μl of S-basal with 20 g/l E. coli OP50-1 to each well, so the final volume is 180 μl per well. Twelve hours after the beginning of the luminometry recording, the plate was unloaded from the luminometer. For non-treated, we added 20 μl ddH₂O per well. For the 4 mM HU treatment, 40 μl of the 1 M HU stock was diluted in 960 μl ddH₂O and 20 μl of that mix was added to each well. For the 8 mM HU treatment, 80 μl of the 1 M HU stock was diluted in 920 μl ddH₂O and 20 μl of that mix was added to each well. After adding the HU, the plate was sealed with a new gas-permeable membrane and placed again in a luminometry reader.

For HU treatment in microchambers, we proceeded as previously described for time-lapse imaging, but non-treated worms and 4 mM HU-treated worms were placed in separated chambers. The two chambers were placed on different slides but imaged together using a Leica scope M205 FCA. For the 4 mM treatment, the gel was stabilized in M9 with 4 mM HU, and after placing the eggs, the compartment was filled up with the same solution.

Statistical analysis

For statistics, we have used the averages of independent biological replicates to avoid the inflated N value from using individual animals in luminometry experiments [38]. Before calculating the averages, we applied a filter for outliers using ROUT (Q = 0.2%) on Graphpad Prism 10. For Fig 3D, we used iterative Grubbs’, as ROUT removed all the values different from 0. The number of larvae and number of independent replicates performed are shown in S2 Table. For the data on Fig 1G, we used two-way ANOVA to test both the effects of genotype and larval stage, followed by Fisher’s LSD to reveal significant differences between genotypes at each stage. We used unpaired two-tailed t test to compare the means of two groups in Figs 5C, 5E, 5F, 5H, S5A, S5B, S5E, 6B, 6C, 6E and 6F. For the rest of the experiments, we used one-way ANOVA to compare more than two groups. After one-way ANOVA, we either perform Dunnett’s multiple comparison to compare the average of each genotype/condition to the average of the wild type, or Tukey’s multiple comparison to compare the average of each genotype to all the other genotypes. The type of analysis that applied to each panel is detailed in the figure legends. For clarity, some plots show only relevant comparisons, but all the results from statistical testing are shown in S1 Data. In all cases, * means p < 0.05, ** p < 0.01, *** p < 0.001, and **** p < 0.0001. Graphs and statistics were performed on Graphpad Prism 10.

S1 Fig. Relative to Fig 1.

(A) Duration of I1, M1, L2Llag, bona fide I2 and M2 for wild type and daf-2 mutant with the luminescence reporters sevIs1 and xeSi296. (B) Duration of M1, L2lag and bona fide I2 relative to the average duration of each stage for the same larvae shown in Fig 1H. (C) Relative duration of L2lag and bf. I2 for each of the larvae in B. (D, E) Matrix heatmap of the correlation between the duration of all larval stages in 58 individual larvae of wildtype (D) and 57 individual daf-2(e1370) mutant larvae (E). (F) Diagrams showing the luminescence profile and durations analyzed, for wild-type larvae without and with L2lag (left), and duration of the corresponding periods (right). In A and F, we used One-way ANOVA followed by Tukey’s multiple comparisons to test differences between strains or developmental stages. Differences are marked as * p < 0.05, ** p < 0.01, *** p < 0.001 and **** p < 0.0001. The data represented in all plots can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003867.s001

(TIFF)

S2 Fig. Relative to Fig 2.

Duration of L3lag (A) and L4lag (B) for the wild type, daf-2, daf-16, and daf-2;daf-16. Colored circles show values for individual larvae and black dots show the average for each experiment. For both panels, we performed One-way ANOVA followed by Tukey’s multiple comparisons to test differences between all strains. Only significant differences are shown, **** p < 0.0001. The data represented in all plots can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003867.s002

(TIFF)

S3 Fig. Relative to Fig 4.

(A) Timing of seam cell division (div 2) relative to ecdysis (ecd 1) for the wild type, daf-18 and daf-2;daf-18. This data set is shown overlaid to the data presented in Fig 4D. (B) Duration of L3lag for the wild type, daf-2 and daf-2;daf-18. (C) Quantification of size relative to the timing of ecdysis. Gray lines show the values of individual larvae and red line shows the average (and 95% CI). Gray dots show the timing of seam cell division for each animal. (D) Size increase per time interval between ecd 1 and div 2 (red) and in a similar period of time after div 2. (E) Representative images of DAF-16::GFP in the wild type and daf-2(e1370) backgrounds. (F) Timing of seam cell division (div 2) relative to ecd 1 for the wild type, the daf-2 mutant, and the double mutant daf-2;daf-16. Colored circles show values for individual larvae. (G) Correlation between L2lag and the timing of V6 seam cell division relative to ecdysis (div 2-ecd 1). (H) Timing of intestinal cell endoreplication relative to ecdysis, for the wild type and daf-2. Statistics reflect the result of unpaired t test. For B and F, we performed One-way ANOVA followed by Tukey’s multiple comparisons to test differences between strains. Only significant differences are shown, ** p < 0.01, **** p < 0.0001. The data represented in all plots can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003867.s003

(TIFF)

S4 Fig. Relative to Fig 5.

Luminescence profile from microscopy experiments for larvae categorized as seam cell divisions (div 2) ending after or before ecdysis (ecd 1). Red arrowheads signal ecdysis and black arrowheads mark V6 division, except when otherwise stated. The data represented in all plots can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003867.s004

(TIFF)

S5 Fig. Relative to Fig 5.

(A) Timing of ecdysis 1 (first release of the cuticle observed in time-lapse experiments) relative to the beginning of imaging in time-lapse experiments, for the wild type and daf-2(e1370) growing on OP50-1 or HT115. (B) Duration of the complete L1 stages, defined as the time from hatching to the end of the first molt, as inferred from the bioluminescence signal in a luminometer, for the same conditions as in A. (C) Fraction of animals that reach adulthood after treatment with different concentration of HU from L1, in four independent experiments. (D) Duration of the stage L1 after treatment with different concentrations of HU in four independent experiments. We performed One-way ANOVA followed by Dunnett’s multiple comparisons to test differences between each of the treatments and the untreated larvae. Differences are marked as ** p < 0.01, **** p < 0.0001. (E) Timing of ecdysis 1 (first release of the cuticle observed in time-lapse experiments) relative to the beginning of imaging in time-lapse experiments, for the wild type on 0- or 4-mM Hydroxyurea. Statistics reflect the result of unpaired t test. (F) Correlation between L2lag and the timing of V6 seam cell division relative to ecdysis (div 2-ecd 1) for all conditions and genotypes tested. Statistics reflect the result of unpaired t test. Differences are marked as ** p < 0.01, **** p < 0.0001. Gray or colored circles show values for individual larvae and black dots show the average of each experiment. The data represented in all plots can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3003867.s005

(TIFF)