An Epigenetic Priming Mechanism Mediated by Nutrient Sensing Regulates Transcriptional Output During C. elegans Development

Summary

While precise tuning of gene expression levels is critical for most developmental pathways, the mechanisms by which the transcriptional output of dosage-sensitive molecules is established or modulated by the environment remain poorly understood. Here, we provide a mechanistic framework for how the conserved transcription factor BLMP-1/Blimp1 operates as a pioneer factor to decompact chromatin near its target loci during embryogenesis (hours prior to major transcriptional activation) and by doing so, regulates both the duration and amplitude of subsequent target gene transcription during post-embryonic development. This priming mechanism is genetically separable from the mechanisms that establish the timing of transcriptional induction and functions to canalize aspects of cell-fate specification, animal size regulation, and molting. A key feature of the BLMP-1-dependent transcriptional priming mechanism is that chromatin decompaction is initially established during embryogenesis and maintained throughout larval development by nutrient sensing. This anticipatory mechanism integrates transcriptional output with environmental conditions and is essential for resuming normal temporal patterning after animals exit nutrient-mediated developmental arrests.

Graphical Abstract

eTOC Blurb

Stec et al. describe a novel gene mechanism where BLMP-1, a highly conserved transcription factor, primes its target genes by decompacting chromatin at specific loci hours prior to transcriptional activation. This mechanism regulates the transcriptional output of many key genes that control cell fate specification during post-embryonic development.

INTRODUCTION

Many developmental systems produce invariable phenotypes despite being confronted with perturbations, such as fluctuations in temperature and nutrient availability. How this robustness emerges from developmental gene regulatory networks (GRNs) and how aspects of precision and flexibility are balanced to generate robust developmental programs is a major open question in developmental biology. C. elegans serves as a unique genetic model for unraveling the mechanisms underlying precision and robustness of gene regulation during development. This is in part because the GRNs that control the essentially invariant cell lineage can generate the identical developmental outcomes in the context of diverse stochastic, environmental or genetic perturbations. A striking example of such a GRN is the so-called heterochronic pathway which controls the timing, sequence and synchrony of stage-specific patterns of cell divisions throughout post-embryonic development.¹ Analysis of this GRN indicates that it generates modular patterns of gene expression that occur in the context of the repetitive post-embryonic molting cycle which, in turn, demarcates the four successive larval stages (L1-L4). Transitions from one temporal program to the next occur through the expression of distinct microRNAs (miRNAs) that down-regulate the expression of their temporal selector gene targets in a strict dosage-sensitive manner. Ectopic or abnormally lower or higher doses of heterochronic genes (e.g. miRNAs) result in the wholesale skipping or reiteration of stage-specific developmental programs.^2–6 While the regulatory interactions that establish the sequence of stage-specific events are known, we have very little understanding of how the precise timing or levels of gene expression within this GRN are established and how both of these are buffered against perturbations to achieve robust temporal development of the animal.

High-temporal resolution transcriptomics studies have outlined a number of remarkable features associated with global gene expression patterns in developing C. elegans larva.^7,8 These studies indicate that between 10 to 20% of the post-embryonic transcriptome exhibits highly reproducible periodic expression patterns. Periodic transcription occurs in a variety of environmental conditions and is independent of life history, indicating that it is under tight genetic control. Importantly, the transcriptional rhythm follows the cycle of post-embryonic molting, a process that demarcates patterns of stage-specific developmental programs. Under various environmental conditions that modulate overall developmental pace, the timing of transcription onset scales accordingly, such that the phase of transcription onset within the molting cycle is preserved.^7,8 Currently, it is not known how these transcriptional rhythms are generated, how they are integrated into the execution of stage-specific cellular programs, and how environmental or internal cues modulate features of these transcriptional patterns to achieve robust progression through developmental programs, even after prolonged developmental arrests.

The heterochronic GRN is integrated with global aspects of transcription as each of the miRNAs in this pathway exhibits an oscillatory expression. The most promising candidate gene that integrates the rhythm of C. elegans post-embryonic molting to changes in repetitive transcriptional patterns is lin-42. The lin-42 gene encodes the C. elegans ortholog of PERIOD/Per proteins that are an essential component of the circadian clock in mice, Drosophila and humans.^9–11 Similar to PERIOD, LIN-42 functions as a transcriptional repressor and exhibits a cyclical expression pattern. In contrast to the temperature-invariant approximate 24-hour periodicity of Period, lin-42 transcription is tied to the larval molting cycle, and thus, not temperature compensated.^5,11 Hypomorphic mutations in lin-42 cause animals to prematurely execute adult-specific gene expression patterns in the L4 stage of larval development and these phenotypes are caused by the cumulative, premature over-expression of multiple heterochronic miRNAs.^5,6,12 In addition to these temporal patterning phenotypes, lin-42 mutants also exhibit lengthened and irregular molting cycles.^11,13 The mechanisms by which LIN-42 negatively regulates heterochronic miRNA transcriptional output remain unknown. Furthermore, it is not known if other regulatory components exist that antagonize LIN-42 functions to ensure normal transcriptional regulation of developmental timing genes.

Here, we leverage the developmental phenotypes of lin-42 mutations to identify BLMP-1 and ELT-3 as conserved transcription factors (TFs) that modulate the transcriptional output of many cyclically expressed genes to provide developmental robustness. We demonstrate that these TFs coordinate a plethora of post-embryonic developmental programs and we present a unique role for BLMP-1 in transcriptional priming that involves chromatin decompaction of target loci. Specifically, we show that BLMP-1 activity modulates the duration and amplitude of transcription of its targets without altering the timing of expression within the repetitive transcriptional cycle. We hypothesize that BLMP-1 functions specifically to assure proper transcriptional output in diverse environmental conditions. Consistent with this model, we demonstrate that the BLMP-1-dependent remodeling of chromatin is regulated by nutritional inputs and becomes essential when animals resume development after nutrient-mediated developmental arrests.

RESULTS

blmp-1 and elt-3 Function in a partially redundant manner to control temporal patterning during larval development.

To identify genes that antagonize the function of LIN-42 to modulate transcriptional output, we took advantage of a highly penetrant phenotype of the lin-42 loss of function allele lin-42(n0189) – the precocious expression of an adult-specific col-19::GFP reporter in both seam and syncytial hyp7 cells of the hypodermis (Figure 1A).¹⁴ We used bacterial-mediated RNAi to deplete the expression of approximately 20 C. elegans candidate TFs in lin-42(lf) animals and then scored col-19::GFP expression in F₁ progeny. These transcription factors were selected based on known strong expression in the hypodermis¹⁵ and/or previous implication in heterochronic developmental defects. The most reproducible suppressors of lin-42 precocious phenotypes were bacterial RNAi clones expressing dsRNAs that target the blmp-1 and elt-3 loci (data not shown). We validated these findings with null alleles of these genes (blmp-1(0) and elt-3(0)) which revealed a redundancy between blmp-1 and elt-3 that is distributed differentially across hypodermal cell types (Figure 1 A and B). Specifically, removal of blmp-1 activity in lin-42(lf) mutants completely suppresses the precocious col-19::GFP expression in the lateral seam cells of L4-staged animals and only partially prevents early expression in hyp7 cells (Figure 1A and B). lin-42(lf); elt-3(0) mutants were only partially suppressed for the precocious col-19::GFP expression phenotypes. In contrast, combining blmp-1(0) and elt-3(0) mutations with lin-42(lf) mutations completely alleviated precocious col-19::GFP expression in both the seam and hyp7 cells of a majority of L4-staged animals (Figure 1A and B). This cell-type specific redundancy is likely due to the partially overlapping expression patterns of these TFs in hypodermal cells throughout larval development and indicate that BLMP-1 and ELT-3 are required for lin-42(lf) phenotypes (Figure 1C).¹⁵

Figure 1. blmp-1 and elt-3 antagonize lin-42 activity to control the expression of heterochronic miRNAs.

(A) Representative col-19::GFP expression patterns in wild-type, lin-42(lf), blmp-1(0); lin-42(lf), and blmp-1(0); lin-42(lf); elt-3(0) animals.

(B) A quantification of the col-19::GFP expression phenotypes of various mutant combinations in L4- and adult-staged animals.

(C) Expression patterns of BLMP-1::GFP and ELT-3::GFP transgenes in hypodermal cells. Sn = Seam cell nuclei.

(D) SCM::GFP is expressed in lateral seam cells and can be used to monitor cell division patterns during larval development. Mid-L4-staged alg-1(0); blmp-1(RNAi) animals exhibit a supernumerary number of SCM::GFP cells.

(E) Quantification of lateral seam cell numbers of mid-L4-staged animals exposed to bacteria expressing dsRNAs complementary to elt-3, blmp-1 or both blmp-1 and elt-3. Red bars indicate mean and SEM. Asterisks indicate statistically significant differences in phenotype (p = < 0.0001) calculated from a two-tailed chi-square analysis. n = > 40 animals per genotype/RNAi condition.

(F) Lateral seam cell lineage of wild-type, blmp-1, and alg-1 mutants and experiments outlined in panel F. See also Figure S1.

Previous analysis of blmp-1(0) mutations indicate that BLMP-1 plays an important role in terminal fate specification of the hypodermis.¹⁶ We sought to determine if blmp-1 functions earlier in temporal cell fate specification when cells are proliferating. To this end, we assayed genetic interactions between blmp-1 and alg-1, the latter encoding one of the two core miRISC argonaute components. alg-1(0) mutants ineffectively process most miRNAs (including heterochronic microRNAs) and exhibit relatively mild temporal patterning defects that resemble those associated with hypomorphic alleles of let-7.¹⁷ To assay interactions, we quantified the number of seam cells at the mid L4 stage in wild-type or alg-1(0) mutant animals in conditions where we reduced blmp-1 and/or elt-3 activity using RNAi (Figure 1D, E). In wild-type animals, lateral seam cells exhibit highly reproducible cell division patterns that occur at specific stages of larval development (Figure 1F).¹⁸ Defects in this lineage can be monitored by counting seam cell numbers in the L4 larval stage. Depletion of elt-3 or blmp-1 in wild-type animals resulted in the normal number of seam cells in the L4 stage. We observed a significant increase in the number of SCM::GFP(+) seam cells in alg-1(0); blmp-1 RNAi animals (Figure 1E). This expansion of lateral seam cell number could result from inappropriate reiteration of earlier, L2-specific proliferative cell division programs, or conversion of normally asymmetric seam cell divisions in the L3 and L3-stage into symmetric ones (Figure 1F). Depletion of blmp-1 in other mildly reiterative heterochronic mutants also resulted in similar enhancement of cell lineage and col-19::GFP expression phenotypes (Figure S1). elt-3 RNAi did not significantly alter lateral seam cell numbers in alg-1(0) animals or enhance alg-1(0); blmp-1(RNAi) phenotypes. This indicates that BLMP-1 functions prior to terminal seam cell fate specification in larval development to control temporal cell fates in the hypodermis.

blmp-1 and elt-3 regulate diverse dosage-dependent developmental processes

Development of C. elegans larva results in a fixed number of cells and a stereotypical size under standard growth conditions. Mutations in a number of pathways (including nutrient signaling, TGF-beta and MAP kinase pathways) or in structural components of the cuticle alter organismal size.¹⁹ A shared feature of these regulatory and structural systems is that they function in a dose-sensitive manner wherein changes in the expression levels of individual elements within these pathways or components result in changes in animal morphology. As shown in Figure 2A, blmp-1(0); elt-3(0) animals exhibit a severe dumpy (dpy) phenotype. The dpy phenotype of blmp-1(0); elt-3(0) mutants manifests during post-embryonic development as freshly hatched blmp-1(0); elt-3(0) larvae are indistinguishable from wild-type animals (Figure 2A). blmp-1(0) single mutants exhibit less severe alterations in size and elt-3(0) animals are structurally normal (Figure 2A and B). These results demonstrate that in addition to modulating aspects of temporal cell-fate specification, blmp-1 and elt-3 cooperate to control animal morphology likely through the co-regulation of hypodermally expressed targets.

Figure 2. blmp-1 and elt-3 regulate several additional morphological, behavioral, and cellular phenotypes.

(A) blmp-1(0); elt-3(0) animals exhibit a synthetic Dpy phenotype that manifests during larval development.

(B) Quantification of animal length-to-width ratio in freshly hatched larvae (left) and young adult animals (right) of the indicated genotypes (n = >20). Red bars indicate the mean and SEM.

(C) blmp-1(0) and blmp-1(0); elt-3(0) animals prematurely die after the L4-adult transition.

(D) The lethality of blmp-1(0); elt-3(0) adult animals correlates with an inability to shed the cuticle of a supernumerary molt.

(E) blmp-1(0); elt-3(0) inappropriately reanimate the expression of a mlt-10::GFP-pest transcriptional reporter during adulthood. Animals were imaged at the L4-adult transition or as gravid adults.

(F) Major components of the vulval induction pathway (see text for details). Components that are genetically involved in vulval induction are colored green whereas those involved in Notch/lateral inhibition are colored black. The let-7 family miRNA miR-84 (red) that regulates let-60 expression is temporally expressed in presumptive secondary fate vulval cells.

(G) blmp-1(0) mutations enhance the expressivity and penetrance of let-60(n1046gf) muv phenotypes. Animals of the indicated genotypes were scored for the number of vulval protrusions present on the ventral side at the young adult stage (n = >90). Brackets indicate statistically significant differences in phenotype calculated from a two-tailed chi-square analysis.

(H) Representative images depicting muv phenotypes associated with the indicated genotypes.

A majority of blmp-1(0); elt-3(0) animals precociously die two days after they reach adulthood due to the internal hatching of fertilized embryos (Figure 2C). Examination of dying blmp-1(0); elt-3(0) animals revealed that most are trapped in a partially shed cuticle (Figure 2D). Lethality is preceded by a reduction in movement and pharyngeal pumping approximately 8-10 hours after the L4-to-adult transition; a phenotype shared with other mutants that fail to post-transcriptionally down-regulate the adult expression of key genes involved in larval molting.²⁰ To determine if blmp-1(0); elt-3(0) animals inappropriately re-initiate molting programs during adulthood, we directly monitored the expression of a molting-specific GFP-pest transcriptional reporter, mlt-10::GFP-pest, in wild-type, single and double mutant animals. Wild type animals never expressed mlt-10::GFP as adults (n = 100) while a majority of blmp-1(0); elt-3(0) animals re-expressed the mlt-10::GFP-pest reporter (54%; n = 155) as adults (Figure 2E). blmp-1(0) single mutants also express mlt-10::GFP-pest within the same time period but with a much lower penetrance (18%, n = 137), consistent with previous observations using blmp-1(RNAi).²¹ These data indicate that blmp-1 and elt-3 function redundantly to halt the behavioral and physiological molting programs once animals transition into adulthood.

Development of the vulval structure in the L3 stage requires the integration of a number of interconnected signaling pathways that establish specific cell fates in vulval precursor cells (Figure 2F). One of the key initiating components in this process is let-60, encoding the C. elegans ortholog of the human Ras oncogene. Continued let-60 activation commits vulval precursor cells (VPCs) to a 1° vulval cell fate specification pattern (Figure 2F). The induction of this fate specification is sensitive to the level of let-60 activity.^22,23 For instance, the semidominant let-60(n1046gf) allele can induce ectopic vulval protrusions on the ventral surface (multi-vulva; muv) in 18% of animals harboring a single copy of this mutation (Figure 2G and H). We used this sensitized genetic context to determine whether blmp-1(0) and/or elt-3(0) mutations (or the combination of these mutations) leads to an elevation in VPC induction in let-60(n1046)/+ animals. Lack of blmp-1 activity significantly increased the penetrance and expressivity of the muv phenotype in let-60(n1046)/+ animals (Figure G and H). elt-3(0) mutations did not contribute to let-60-dependent phenotypes consistent with a lack of significant ELT-3 expression in these cell types (Figure 3G).¹⁵ Together, these results indicate that blmp-1 and elt-3 function in a partially redundant manner to regulate diverse of developmental aspects, from animal morphology, to behavior to cell fate specification.

Figure 3. BLMP-1 and ELT-3 bind similar sites in the genome and are associated with the putative regulatory regions of gens that exhibit cyclical expression patterns.

(A) A matrix of global pairwise factor co-association strengths as defined by promoter interval statistics indicate that BLMP-1 and ELT-3 bind similar genomic regions.

(B) The distribution of the peak summit positions of BLMP-1 ChIP-seq data mapped relative to the TSS of protein coding genes or to the +1 nucleotide of the DNA sequence encoding the miRNA precursor RNA.

(C) A heat map of ChIP-seq signals +/− 3 kb centered on the TSS of each of the 14,252 protein coding genes measured in the developmental RNA-seq data set.⁸ Genes are organized for each heat map in a 1:1 fashion according to their density of unique reads present in the BLMP-1 L1-stage ChIP-seq sample (see Data S1).

(D) The C. elegans larval transcriptome can be categorized into two dynamic expression classes (rising and oscillating) that are distinct from the predominantly flat expression pattern of most mRNAs (grey region; 66.9%).⁸

(E) A graph displaying the number of predicted rising or oscillating class targets for each of the 172 TFs assayed by the ModEncode Projects (data derived from a total of 265 ChIP-seq data sets assayed). Red bars indicate the median (thick) with the interquartile range (thin). Numbers in parenthesis indicate ranks of BLMP-1 or ELT-3 among all TFs (Data S2 and Figure S3).

(F) Quantification of GFP intensity for translational reporters of BLMP-1 and ELT-3 during the L4 stage.

(G) A radar chart plotting oscillation amplitude over the phase of peak expression reveals that BLMP-1 and ELT-3 bind the putative regulatory regions of periodically expressed genes in all phases of the cyclical transcriptional cycle (see Data S1).

See also Figure S2.

BLMP-1 and ELT-3 co-target overlapping gene sets

Our observations that blmp-1 and elt-3 genetically interact with lin-42 to control the transcription of cyclically expressed miRNAs (e.g., let-7, Figure 1D) and that blmp-1(0); elt-3(0) double mutants exhibit pleiotropic defects led us to hypothesize that these TFs may function more broadly to control dynamic gene expression. To define the full repertoire of BLMP-1 and ELT-3 transcriptional targets, we complemented L1-staged BLMP-1 and ELT-3 ChIP-seq data obtained in the modENCODE project with additional ChIP-seq experiments performed in extracts made from L2-, L3-, and L4-staged animals.^24,25 Analysis of these combined data indicates that the regions bound by BLMP-1 in the first larval stage are also bound in subsequent larval stages. Furthermore, the binding sites of BLMP-1 in the L1-L4 stages significantly overlap with regions also bound by ELT-3 in the L1 stage (Figure 3A). A majority of BLMP-1 peaks are enriched within 500bps upstream of the transcriptional start site (TSS) but the enrichment also extends to 5’ intergenic regions with a sharp drop-off approximately 3kb upstream of gene bodies of both protein coding and miRNA genes (Figure 3B and C). Peaks from all datasets were assigned to the nearest proximal gene (both coding and non-coding) using these parameters (Data S1A–E).

To characterize the relationship between BLMP-1, ELT-3 and their targets in more detail, we initially focused our analysis on protein coding genes whose expression can be reliably detected in developing larvae. Our analysis employed multiple, high-resolution, and genome-wide time course analyses of gene expression that parsed the larval developmental transcriptome into distinct temporal expression classes.^7,8 Both studies identified significant, overlapping sets of transcripts (between 12% and 18.9% of the transcriptome) that exhibit cyclical expression patterns. Hendriks et al. classified two types of dynamic expression: a first class of transcripts that exhibit oscillatory or cyclical patterns (osc) that are tied to the molting cycle (osc; 18.9%; 2,718 of 14,378 total genes) or a second class of genes whose expression increases monotonically throughout larval development (rising; 14.2%) (Figure 3D). To compare how BLMP-1 and ELT-3 targets may be distinguished from the targets of other TFs, we analyzed additional publicly available ChIP-seq data sets obtained by the modENCODE project. This data included the targets of 170 additional C. elegans TFs that represent diverse classes of TFs.^24,25 Using the same target assignment criteria as above, we found that most TFs are associated with a limited number of target genes that exhibit a rising expression program (median targets bound = 21 per TF; interquartile range = 9 to 54)(Figure 3E; Data S2A). This relationship was also conserved for the associations between bulk TFs and the oscillatory class of target genes (median of 43 targets per TF; interquartile range = 18 to 97). Surprisingly, we found that a limited subset of 39 TFs is disproportionally associated with a large fraction of the oscillatory class of genes (Figure 3E). Primary amongst this unique group of 39 TFs were BLMP-1 and ELT-3 (with 1124 and 738 of the 2718 cyclically expressed targets bound per TF, respectively) (Figure 3E). Both BLMP-1 and ELT-3 binding sites are statistically enriched in the promoters of cyclically expressed genes (1124 osc BLMP-1 osc targets vs. 3265 non-osc targets and 738 ELT-3 osc targets vs. 2484 non-soc targets; p < 0.0001, Chi square test with Yates correction). We performed a similar analysis of the Kim et al. RNA-seq data to determine if ELT-3 or BLMP-1 binding sites are also enriched in the promoters of cyclically expressed transcripts as defined by their temporal classifications. Our analysis found that BLMP-1 binding sites, but not ELT-3 binding sites, are enriched in the promoters of each cyclically expressed genes class and de-enriched for those that exhibit monotonic expression patterns (Figure S3 and Data S2B).

One robust feature of cyclically expressed mRNAs is that the timing of expression within each larval stage occurs in the same phase of each transcriptional cycle. This expression/phase relationship is preserved even when animals are grown at lower or higher temperatures which lengthens or shortens the periodicity of these patterns, respectively.^7,8 Because BLMP-1 and ELT-3 are constitutively expressed through each larval stage (Figure 3F), we aimed to determine if the subset of BLMP-1 and ELT-3 targets that exhibit cyclical expression patterns were concentrated in any particular phase of expression during each larval stage. To compare these patterns, we plotted the peak phase (in degrees/radians of a repetitive cycle) and expression level (log2 change in expression as an increasing radius from the center) for each of the cyclically expressed transcripts on a radial chart (Figure 3G). ELT-3 and BLMP-1 targets were distributed across each sector of the repetitive temporal pattern and were not statistically enriched for any phase of expression. The distributed associations of these TFs and their varied targets combined with the constitutive expression of BLMP-1 and ELT-3 throughout each larval stage suggest that BLMP-1 and ELT-3 are associated with their targets throughout the transcriptional cycle regardless of transcriptional class of the target gene or their timing of expression.

Conserved sequences of the upstream regulatory regions of lin-4 are necessary and sufficient for high amplitude expression in hypodermal tissues

Many of the heterochronic phenotypes associated with a null mutant of lin-4, lin-4(e912), can be partially rescued by a high-copy transformation of a 693bp genomic fragment that contains approximately 500bp of upstream putative regulatory sequence.²⁶ Analysis of the RNAs that are produced from the lin-4 locus indicate transcription of pri-lin-4 is initiated at least five TSS upstream of the lin-4 gene.²⁷ Of these TSSs, three occur upstream of the minimal rescuing fragment, suggesting that additional regulatory information may be required to control normal aspects of lin-4 transcription. These upstream regions are also associated with the active histone modifications H3K4me3 and open chromatin as measured by ATAC-seq assays (Figure 4A).^28,29 We compared the syntenic sequences of five additional nematode species to determine if regions of the F59G1.4 ninth intron were conserved through evolutionary selection. This analysis revealed at least three discrete regions upstream of each lin-4 gene that were disproportionately conserved when compared to similarly sized intronic and intragenic regions (Conserved Elements (CE) A, C and D). Examination of our ChIP-seq data indicate that each of these regions is enriched for BLMP-1 binding (Figure 4A, Data S1A–D). We did not detect a strong enrichment of ELT-3 bound to the putative lin-4 regulatory sequences though multiple predicted ELT-3 consensus binding site sequences are found in this region (Figure 4A).

Figure 4. Conserved elements in the lin-4 promoter are both necessary and sufficient for high-amplitude expression in hypodermal tissues.

(A) A browser track overlaying the predicted BLMP-1 consensus binding sites, the BLMP-1::GFP ChIP-seq signal from L1-staged larvae, and conservation of these sequence features found in C. elegans, C. brenneri, C. tropicalis, C. remanei, C. briggsae and C. japonica near the lin-4 locus and promoters used in the expression analysis. Green bars indicate H3K4me3 ChIP-seq data tracks (indicative of active promoters near transcriptional start sites). Black bars indicate the localization of open chromatin as measured from ATAC-seq experiments.²⁸

(B) Representative measurements of full-length lin-4::GFP-pest expression levels in lateral seam cells. The fluorescent intensity of five seam cells per animal were averaged and 8-15 animals were measured per indicated L4 morphological substage.

(C) Representative fluorescence and DIC images of the full-length lin-4::GFP-pest transcriptional reporter in the hypodermal cells of developing L4-staged animals. L4-stage numbers correspond to the defined sub-stages of vulval morphological development (Sn = seam cell nuclei, h = hyp7 cell nuclei, and m = muscle cell nuclei).³⁰

(D) Measurements of seam cell expression levels for the various lin-4::GFP-pest reporter constructs outlined in panel A at the L4.5 stage of development. (n= 5-10 animals per transgenic strain)

(E) Representative images of each transcriptional reporter outlined in D. Individual labels defining cell types are the same as in panel C.

(F) Position Weight Matrix of the consensus BLMP-1 binding site calculated from ChIP-seq data outlined in Figure 3A and two sequences located in the minCE element that conform to this consensus.

(G) Gel shift experiments demonstrating that a GST fusion protein harboring the 5 ZnF domains of BLMP-1 (but not GST alone) can bind to the two sequences outlined in panel F.

(H) Comparison of the GFP-pest expression patterns of gst-5::GFP-pest or (2xminCE)gst-5::GFP-pest in late L1-stage animals. The pulsatile expression of the (2xminCE)gst-5::GFP-pest expression correlates with a single pulse in the middle of each larval stage. For the developmental time course, a minimum of forty animals expressing gst-5::GFP-pest or (2xminCE)gst-5::GFP-pest were scored. See also Figure S2 and S4.

To test the functional relevance of these elements, we constructed a transcriptional GFP-pest reporter that harbors a SL2 splice acceptor sequence directly upstream of the GFP open reading frame (Figure 4A; Full Length (FL)). This feature enables each reporter to generate a single, defined mRNA transcript from each of the various candidate TSSs. This reporter is robustly expressed in a variety of cell types including the hypodermis, pharynx, body wall muscle, and ventral neurons on the surface of larva. Expression of the lin-4::GFP-pest reporter occurs once per larval stage (similar to other characterized transcriptional reporters) in these cell types indicating that this sequence contains the information to drive temporally and spatially restricted expression in multiple cell lineages. To characterize the temporal expression patterns in detail, we measured the levels of lin-4::GFP-pest expression in lateral seam cells at each of the morphologically defined stages of L4 development.³⁰ These experiments revealed that GFP-pest expression is highly pulsatile with expression peaking at the L4.5 and L4.6 stages (Figure 4B and C). Expression in other cell types (neuronal, pharynx, muscle and hyp7 cells) also peak during these larval substages with different expression trajectories (Data not shown). Specifically, muscle cells exhibit a broader (longer) transcriptional pulse compared to those observed in seam cells. Expression in hyp7 and neuronal cell types exhibit a shorter pulse of transcription either muscle or seam cells (Figure 4B).

We then compared the expression patterns we observed using the full-length construct to previously published reporters or variants of the full-length reporter that lack individual or combinations of the conserved sequence elements outlined above. Each of the reporters express pulses of GFP-pest expression in similar cell types as the full-length version.^5,27 However, peak expression in the hypodermal cells was typically 5-20-fold higher from the full-length reporter compared to the other derivative reporter genes (Figure 4D). Deletion analyses of the full-length construct indicate that CE A, harboring sequences that bind BLMP-1 in vivo, are primarily responsible for this increased transcriptional output as reporters that lack this element exhibit residual expression to a similar level as the minimal rescuing fragment (Figure 4D and E).

Using the ChIP-seq data sets for BLMP-1::GFP, we identified a consensus sequence for BLMP-1 that is enriched in the CE A regulatory sequence (Figure 4A and C). A recombinant GST translational fusion of the five BLMP-1 zinc fingers (ZnFs), GST-BLMP-1(ZnFs), but not GST alone bound several sequences found in CE A element that conform to this consensus. This binding was diminished by the addition of cold competitor DNA harboring consensus BLMP-1 binding sites but not by a similar sized, non-specific target DNA (Figure 4G). To determine if these elements function as a transcriptional enhancer, we cloned a 138bp fragment from the CE A element, called minimal CE (minCE) (Figure 4A and H), into the 5’ regulatory sequences of the gst-5 gene encoding the C. elegans ortholog of human HPGDS (hematopoietic prostaglandin D synthetase). RNA-seq experiments demonstrate that gst-5 mRNAs are expressed throughout larval development in a variety of tissues (including the hypodermis) and exhibit a non-cyclical or flat expression pattern.^7,8,15 A transcriptional reporter bearing 2.9kb of the gst-5 upstream regulatory sequences drives a low level GFP-pest expression in a number of larval cell types including the pharynx, intestine, neurons and hypodermis and this expression pattern remained below the threshold of imaging during normal development (Figure 4H). In contrast, the transcriptional reporter that included two copies of the minCE in the reverse orientation (reversed compared to the orientation within the lin-4 enhancer element) drove high amplitude, periodic expression of GFP-pest exclusively in hypodermal cells (Figure 4H). Expression of the (2xminCE)gst-5::GFP-pest reporter in blmp-1(0) mutant animals resembled the expression of the simple, gst-5::GFP-pest reporter in wild-type animals (Figure S4), consistent with the hypothesis that this minimal sequence from the lin-4 upstream region functions as both a spatial and temporal enhancer.

BLMP-1 controls the transcriptional output of its targets

A large fraction of BLMP-1 and ELT-3 targets are expressed in a cyclical fashion (approximately 26% and 23%, respectively) (Figure 3D and Data S2A). We reasoned that targets that exhibit this highly dynamic form of gene expression could provide an opportunity to determine the effects of BLMP-1 and/or ELT-3 binding on transcriptional dynamics in more detail. To quantify changes in expression dynamics, we obtained or constructed integrated transcriptional GFP-pest reporters that fit three criteria. First, we identified candidate genes that are predicted to be targets of BLMP-1 and/or ELT-3 as measured by ChIP-seq (Figure 3 and Figure S2). Second, we selected genes that are predicted (by RNA-seq experiments) to be dynamically expressed in hypodermal tissues.¹⁵ Third, we selected a set of genes whose expression peaks are predicted (again, by RNA-seq experiments) to occur at a variety of phases within the molting cycle.⁸ Based on these criteria, we constructed or obtained multicopy integrated transcriptional GFP-pest reporters for (2xminCE)gst-5, moe-3, mlt-10, and ZK180.5 and a single copy version of a transcriptional GFP-pest reporter for C02E7.6 integrated on chromosome II. To characterize the temporal expression patterns of these reporters, we assembled a series of measurements that spanned the morphologically defined stages of L4 development. As shown in Figures 5A–C, each reporter is expressed in a pulsatile expression pattern with (2xminCE)gst-5 and C02E7.6 reporters exhibiting a peak of expression in the mid-L4 stage, and the moe-3 transcriptional reporter reaching maximal amplitude near the end. We next aimed to determine how mutations in blmp-1 and/or elt-3 altered these expression patterns. GFP-pest expression levels in seam cells were then quantified in each genetic context at a developmental stage where reporter expression peaks in wild-type animals. As illustrated in Figure 5A and B, blmp-1(0) mutations dampen the expression of both (2xminCE)gst-5 and C02E7.6 reporters by approximately 1.7-fold. In contrast, elt-3(0) mutations had little effect on the transcriptional output of either of these two reporters. Furthermore, the decrease in seam cell expression for the (2xminCE)gst-5 and C02E7.6 reporters was not further decreased in blmp-1(0); elt-3(0) double mutants, consistent with a lack of ELT-3 binding sites in these promoters. In contrast, examination of moe-3::GFP-pest expression in various mutant backgrounds indicate that activity of both blmp-1 and elt-3 are required to maintain high transcriptional output for this gene in the hypodermis (Figure 5C).

Figure 5. blmp-1 modulates the amplitude and transcriptional duration of their cyclically expressed targets.

(A-C) Quantification of GFP-pest expression from the (2xminCE)gst-5, C02E7.6 and moe-3 transcriptional reporters in L4-staged animals. Top panel for each reporter indicates the average fluorescent intensity of GFP-pest expression in the lateral seam cells for each morphologically defined L4 stage ((2xminCE)gst-5 = L4.5, C02E7.6 = L4.5, and moe-3 = L4.9) (n = 7-14 animals)(Sn = seam cell nuclei).

(D) Merged DIC and fluorescence micrographs (after worm axis straightening³¹) showing reporter expression at peak intensity in the L2 stage.

(E) Estimation plot⁷⁰ showing the duration of second larval stage (L2) for WT (N=92), blmp-1(0) (N=46), elt-3(0) (N=15) and blmp-1(0); elt-3(0) animals (N=50). Upper row: Grey circles indicate L2 duration, measured by observing cuticle shedding (molts) during long-term live imaging using a microfluidics device (Keil et al., 2016). (see Methods). Black error bars indicate mean and standard deviation. Lower row: Mean bootstrapped difference values compared to WT and 95% confidence intervals of bootstrap distributions (black error bars). See also Figure S5.

(F) ZK180.5::GFP-pest reporter intensity, averaged over the entire animal throughout development as a function of time for WT (light red, N=23, dark red average) and blmp-1(0) mutant (light blue, N=13, dark blue average). Time courses are scaled such that molts (dashed grey lines) of individual animals align with average WT molting timings.

(G) Inset of (F), showing the fluorescence intensity during the L2 stage of development.

(H) L2 peak fluorescence intensity (left most), onset phase of reporter expression (left middle), peak phase of reporter expression (right middle) and Output (AOC = area under the curve) (right most) for the ZK180.5::GFP-pest reporter in WT (red circles, N=21) and blmp-1(0) mutant (blue circles, N=13). Black triangles indicate mean bootstrapped difference values, blue error bars indicate 95% confidence intervals of bootstrapped difference distribution. See also Figure S6.

(I,J) As (F, G) but for mlt-10::GFP-pest (WT, N=7; blmp-1(0), N=5). See also Figure S6.

We next sought to correlate dynamic changes in gene expression to other aspects of larval development in more detail. To accomplish this, we performed long-term, live imaging of developing larvae in a microfluidic device³¹, enabling us to continuously monitor gene expression dynamics in individual animals and also correlate this dynamics with other cellular and developmental milestones. We focused on two of the most strongly expressed reporter genes, mlt-10pro::GFP-pest and zk180.5pro::GFP-pest, both exclusively expressed in the hypodermis (Figure 5D) and their respective putative regulatory elements bound by BLMP-1 and ELT-3 (Figure 3 and Figure S6). First, we compared overall developmental pace between WT, blmp-1(0), elt-3(0) and blmp-1(0);elt-3(0) animals in the microfluidic chambers by assessing the time between the successive ecdysis (cuticle shedding). While both WT and mutant animals displayed variability of larval stage duration within each genotype, neither blmp-1(0) nor elt-3(0) mutations (or combination of these mutations) affected the average duration to a substantial degree (Figure 5E and Figure S5). However, our live imaging experiments revealed a partially penetrant larval lethality phenotype, in particular in the L3 stage, that was not apparent on standard, solid NGM plates.

The fact that the larval stage durations are largely preserved between WT and blmp-1(0)/elt-3(0) mutants allowed us to directly compare average quantitative temporal features of transgene gene expression dynamics such as the time within each transcriptional cycle in which transcription begins, peak phase (the point within each cycle in which the peak expression is reached), and the cumulative accumulation of GFP-pest expression (output) between genetic backgrounds. To this end, we extracted individual whole-animal fluorescence time course measurements and rescaled them such as to align the timing of molting events of each animal (see Methods) to the average WT molting times (Figure 5F). Consistent with other blmp-1- or blmp-1; elt-3-dependent reporters, cyclical transcription of ZK180.5::GFP-pest and mlt-10::GFP-pest was maintained throughout larval development indicating that blmp-1 (or elt-3) are not essential for generating pulsatile patterns of transcription (Figure 5F–J and Figure S6). The primary difference between the expression trajectories of wild type and mutant animals was the drastic reduction in the amplitude of individual traces in blmp-1(0) mutants. Surprisingly, the reduction of transcriptional output was not accompanied by any substantial change in the timing of transcriptional induction (Figure 5H and J and Figure S6). Ultimately, the reduction of transcriptional output (as measured by total cumulative GFP-pest expression during each pulse) correlates with a reduction in the relative duration of expression in mutant animals (Figure 5H and J and Figure S6). These results indicate that BLMP-1 functions to amplify the baseline pulsatile transcriptional dynamics that are likely generated by other gene regulatory mechanisms. With respect to the (2xminCE)gst-5::GFP-pest reporter (see above), our findings suggest that the 138bp minCE fragment may contain the regulatory sequences required by these mechanisms to generate oscillatory transcription but the resulting oscillation amplitude is strongly dependent on BLMP-1 binding.

The lin-4 locus adopts a BLMP-1-dependent, active chromatin structure in hypodermal cells

Previous studies indicate that BLMP-1 physically and functionally interacts with multiple chromatin remodeling factors (e.g., SWI/SNF subunits) and that BLMP-1 binding sites are enriched in open chromatin regions (Figure S2).^28,29,32 We hypothesized that BLMP-1 may function to potentiate transcription through chromatin remodeling near its binding sites which would potentiate the binding of additional TFs. This model would account for the dependency of the diverse transcriptional targets of BLMP-1 that differ in their temporal pattern of expression (phase) during the repetitive transcriptional cycle and would be mechanistically similar to the activities of other TFs (including pioneer factors) that remodel chromatin accessibility prior to the onset of major transcriptional induction in C. elegans.^33–35

We therefore sought to measure how BLMP-1 activity may alter chromatin dynamics at a specific locus with spatiotemporal resolution. We utilized an established chromosome-tagging strategy that has been employed to visualize both the localization of transgenic arrays and dynamics of chromatin compaction in living animals.^33,36,37 Transgenes containing the lin-4(FL)::mCherry-pest ORF flanked by tandem copies of the bacterial lac operator (lacO) sequences were integrated to a single chromosomal site and then crossed into a strain harboring a soluble, ubiquitously expressed GFP-tagged lac repressor (LacI) fusion that recognizes lacO sequences with high affinity.³⁵ The enabled us to monitor the lin-4 locus in live animals while simultaneously monitoring transcriptional output of the lin-4 enhancer/promoter (Figure 6A). Animals harboring these transgenes exhibit two foci in each somatic nucleus indicating the stable, single integration site of the lin-4::mCherry/LacO transgene. The intranuclear GFP-LacI foci found in freshly hatched L1-staged animals can be grossly categorized into two basic classes: Class 1) a punctate or “ball-like” pair of GFP-LacI foci that are found in a majority of somatic cells and Class 2) pairs of “puffed” foci that are found in the nuclei of hypodermal cells (Figure 6B). This lineage-specific decompaction of the lin-4 locus is surprising for multiple reasons. First, lin-4 loci appear to be decompacted only in hypodermal tissues where BLMP-1 is normally expressed. Second, chromatin decompaction is typically associated with active transcription in a variety of systems^38–40 and the decompaction of the lin-4 loci in freshly hatched L1-stage animals precedes any detectible lin-4::mCherry-pest expression (Figure 6B).

Figure 6. The lin-4 locus is maintained in an open chromatin state in hypodermal cells in a blmp-1-dependent manner.

(A) Schematic of the transgenic array used for visualization of the lin-4 locus in living animals.

(B) Representative image of L1-staged animals expressing the GFP-lacO-labeled lin-4 locus.

(C) Analysis of the expression pattern of an endogenously GFP-tagged translational fusion of BLMP-1 indicating that expression begins in hypodermal cells around the bean stages and remains stably expressed throughout embryonic development.

(D) Pictographs depicting the compression level of the lin-4 locus/GFP-LacI foci in differentially staged wild-type and blmp-1(0) mutant embryos. Yellow arrows indicate the location of hypodermal nuclei.

(E) Quantification of the percentage of wild-type or blmp-1(0) late-stage (pretzel) embryos that exhibit decompacted lin-4 loci in hypodermal cells.

(F) Close-up images of representative seam cell and neuronal nuclei from L3-staged wild-type and blmp-1(0) mutant larva. GFP images are the maximum intensity projections obtained for each relevant nuclei.

(G) Quantification of the sizes of the nuclear GFP-LacI foci in various cell types of wild-type and blmp-1(0) mutant animals. For each cell type, GFP-LacI areas were quantified from greater than 18 individual nuclei from each genotype. Boxes and median line indicate the interquartile range. Whiskers cover 10-90th percentile. Circles represent outliers. Brackets indicate statistically significant differences in “puffs size” calculated from a two-tailed chi-square analysis.

To test whether BLMP-1 expression correlates with target loci decompaction, we performed two types of analyses. First, we characterized when BLMP-1 expression begins in embryos. BLMP-1::GFP can be detected as early as the bean stage (approximately 365 minutes post fertilization) in most developing hypodermal cells (Figure 6C). This tissue-specific expression pattern remains throughout the rest of embryonic development (Figure 6C). We next aimed to determine how the timing of BLMP-1 expression correlates with the decompaction level of the lin-4 loci by examining changes in GFP-LacI foci. From the onset of GFP-LacI expression to the early gastrulation stage of embryogenesis, most somatic lin-4 GFP-LacI foci are decompacted, consistent with a general lack of condensed chromatin and reduced histone H3 lysine 9 (H3K9) methyltransferase activity in early embryos.^40–43 By the end of gastrulation (approximately 330 minutes post fertilization) the foci in most somatic cells are equally condensed consistent with the onset of large-scale heterochromatin formation in somatic cells and a reduction in transcriptional plasticity.^40–42 At roughly the same time as the onset of BLMP-1 expression in embryonic hypodermal cells, the GFP-LacI foci begin to decompact in posterior seam cells (Figure 6D). The cell type-specific decompaction of lin-4 loci in hypodermal tissues is complete for a majority of wild-type animals by the end of embryogenesis while the same foci in other somatic cell types remain tightly compressed (Figure 6D). GFP-LacI foci dynamics were similar in wild-type and blmp-1(0) animals during early embryogenesis. In contrast to the late embryonic decompaction of lin-4 loci in wild-type animals, hypodermal GFP-LacI foci (and foci in other somatic nuclei) in blmp-1(0) mutants remained condensed (Figure 6D and E). We conclude that BLMP-1 is required for decompaction of the lin-4 loci in hypodermal cells during embryogenesis. Furthermore, this decompaction occurs at a point in development where lin-4 transcription is being actively repressed by multiple FLYWCH proteins⁴⁴ and transcriptional activity is undetectable with our fluorescent transcriptional reporter; though minimal transcription during this period cannot be excluded.

We next aimed to directly monitor any dynamic changes in “puffs” sizes within the transcriptional cycle of lin-4 during larval development and determine if these changes are dependent on BLMP-1 activity. In head ganglion cells, derived from a cell lineage that never express the lin-4pro::mCherry-pest transgene, the two lin-4pro::mCherry-pest loci are constitutively compacted (Figure 6F and G). To measure changes in cells that do express lin-4 miRNAs, we focused our analysis on developing lateral seam cells for multiple reasons. First, seam cells constitutively express BLMP-1 and dynamically express the lin-4pro::mCherry-pest reporter at specific phases of each larval stage (Figure 1B and Figure 4A). Second, in contrast to hyp7 syncytial cells that endoreduplicate their nuclear DNA content at each division resulting in a stable ploidy of 4n, seam cells replicate their nuclear DNA once per division at specific, defined phases of the larval molt cycle.⁴⁵ To ensure accuracy in both the timing of lin-4::mCherry-pest expression and seam cell divisions, we specifically measured the compaction status of the lin-4::mCherry-pest loci in late L3-staged animals where morphological features of the vulval cell lineage, gonad morphology, and mCherry-pest expression could be used to compare animals of both genotypes at similar developmental stages. We found that the GFP-LacI “puffs” size lateral seam cells with undetectable lin-4pro::mCherry-pest expression was typically three times larger than those found in neurons (Figure 6F and G). Later, during the L3 lethargus, when lin-4pro::mCherry-pest is expressed, we observed that the average size of the GFP-LacI “puffs” further increase in size (approximately an additional 2-fold when compared to seam cells lacking lin-4pro::mCherry-pest expression) (Figure 6F and G). This suggests that the lin-4 locus is constitutively decompacted (compared to the same locus in neuronal cells) and the process of transcribing this locus further relaxes chromatin architecture. We then asked if chromatin decompaction at these loci was dependent on BLMP-1 activity by examining the same features in blmp-1(0) animals. A comparison of the GFP-LacI “puff” sizes in head neurons indicates that they are equally compressed in wild-type and blmp-1(0) animals (Figure 6F and G). In contrast, GFP-LacI foci, in both non-transcribing and lin-4 expressing cells, are significantly reduced in size compared to those observed in stage-matched wild-type animals (Figure 6F and G). These experiments demonstrate that the chromatin structure of the lin-4 locus is dynamically remodeled in seam cells and that BLMP-1 is required for decompaction of the lin-4 locus in hypodermal cells prior to high-amplitude transcription in wild-type, larval animals.

BLMP-1 and ELT-3 are required to resume normal temporal patterning after nutrient deprivation

Development of C. elegans larvae in laboratory conditions is both rapid and continuous due to an abundant, controlled food supply and standardized growth conditions. In contrast, the natural environment in which C. elegans larvae normally live is far more diverse and dynamic leading to a more punctuated developmental trajectory.⁴⁶ Acute food removal induces a defined developmental diapause that can occur immediately after each larval molt.⁴⁷ The diapause is part of a developmental checkpoint that is distinct from dauer development and controlled by insulin signaling.⁴⁷ When insulin signaling becomes rapidly inactivated (i.e. during acute starvation), animals finish the current larval stage, including somatic cell divisions and the molt, but then arrest all somatic cell movements and divisions, and suspend the subsequent expression of molting-specific genes.⁴⁷

To determine if global aspects of cyclical transcription are altered during acute starvation, we monitored the expression of four transcriptional reporters that exhibit distinct phases of expression (Figure 5). These reporters include targets of BLMP-1 and ELT-3 ((minCE)gst-5::GFP-pest, ZK180.5::GFP-pest, and mlt-10::GFP-pest) as well as a transcriptional reporter of blmp-1, since ChIP-seq data indicates that it is expressed in a cyclical expression pattern^7,8 and BLMP-1 activity regulates its own expression through direct positive feedback (Figure S7). We grew populations of L1-staged animals under nutrient-rich conditions for 22 hours and then one half of each sample was allowed to continue development in nutrient replete conditions while the other half was acutely starved (Figure 7A). Acute food removal results in greater than 90% of transgenic animals arresting development with undivided vulval precursor cells after 24 hours indicating that larval development has been paused.⁴⁷ During the starvation period, we monitored GFP-pest expression in fed and starved populations. In contrast to the cyclical expression of each reporter under nutrient replete conditions, GFP-pest expression for each of the reporters was extinguished during starvation conditions with similar kinetics (Figure 7B). This transcriptionally inactive state was maintained throughout the starvation period. Re-fed animals resumed normal, cyclical expression patterns within a few hours after food exposure (Figure 7B). These experiments demonstrate that oscillatory transcription is inhibited by the starvation-induced developmental diapause and that wild-type animals can re-initiate these transcriptional patterns in a nutrient-dependent manner.

Figure 7. Dynamic gene expression is halted during nutrition-mediated developmental arrest and BLMP-1 and ELT-3 are essential for the recovery of normal temporal patterning after starvation.

(A) A schematic of the starvation experiment used to measure nutrient-depended changes in the expression patterns of cyclically expressed genes. See also Figure S7.

(B) Continuous cycling of gene expression is arrested during starvation conditions and is reinitiated in a coordinated manner when animals resume development. For each timepoint, 70-200 animals were scored.

(C) BLMP-1::GFP expression is down-regulated during starvation conditions and expression is rapidly resumed when animals reinitiate development. See also Figure S7 for L1 starvation data.

(D) The lin-4 locus is compacted during starvation conditions in lateral seam cells. Data for each condition was quantified as in Figure 6G. See also Figure S7

(E) A schematic of the starvation-mediated arrest and re-feeding experiment used to measure the ability of animals to resume normal temporal patterning after starvation.

(F) The resumption of normal temporal patterning after starvation requires blmp-1 and elt-3 activity. Defects in temporal development were monitored by measuring the col-19::GFP expression phenotypes in re-fed animals after they resumed growth for 24hours after the indicated period of starvation. (n = >100 per condition)

(G) Representative pictographs of the col-19::GFP phenotypes in wild-type and blmp-1(0); elt-3(0) mutants after animals have recovered from 5 days of starvation.

(H) A diagram outlining the expression features of a single transcriptional cycle in wild-type and blmp-1(0) animals.

(I) A model for how BLMP-1 and ELT-3 function to regulate transcriptional output during larval development.

(J) A diagram outlining the changes of blmp-1 mRNA and BLMP-1 protein levels during normal grown and during starvation conditions.

BLMP-1 expression is actively regulated post-translationally by a conserved F-box protein, DRE-1, that mediates the ubiquitination of target proteins which are degraded by the proteasome.¹⁶ Given that blmp-1 is cyclically transcribed and its transcription is inhibited in nutrient depleted conditions (Figure 7B), we hypothesized that the constitutive expression of BLMP-1 during normal development may be altered during starvation. To test this, we repeated the nutrition-mediated arrest experiments outlined above in strains harboring BLMP-1::GFP and ELT-3::GFP translational fusions. The expression patterns of each translational fusion were compared to a constitutively expressed monomeric GFP transgene (sur-5::GFP) to control for potential changes in GFP half-life under starvation conditions. Surprisingly, BLMP-1::GFP expression was diminished during starvation with >50% of animals loosing detectible BLMP-1::GFP expression after 12 hours of starvation. In contrast to the down-regulation of BLMP-1::GFP, most starved animals retained ELT-3::GFP and soluble GFP expression throughout the starvation period with only a mild reduction after 24 hours (Figure 7C). As with the transcriptional reporter for blmp-1 expression, BLMP-1::GFP expression was detectible again in most animals within 2 hours or re-feeding and resumed normal, pre-starved expression levels after 5 hours (Figure 7C). Similar changes in blmp-1 mRNA and BLMP-1 protein expression were seen in freshly hatched and starved L1 (Figure S8). We conclude that BLMP-1 expression, both transcriptionally and post translationally, is regulated by nutrient availability.

To examine if BLMP-1-dependent chromatin remodeling is also modulated by nutrient-dependent signaling, we monitored compaction of lin-4 locus in animals that have entered the acute starvation-dependent developmental diapause. At the mid-L3 stage (first VPC division), a developmental time point that precedes lin-4 transcription during the L3 stage the lin-4 locus was decompacted (Figure 7D). We then rapidly removed the bacteria food source to induce the L4-specific developmental diapause. After 22 hours of starvation, >90% of animals arrested at the invagination stage of vulval development (L4.1-L4.2) which precedes fusion between the anchor cell and vulval precursor cells on the ventral surface of the L4 larva.³⁰ Analysis of lateral seam cell GFP-LacI puffs sizes in starved animals demonstrated that the lin-4 loci in seam cells were dramatically compacted when compared to the same loci in continuously developing animals (Figure 7D). Re-plating of these starved animals lead to the rapid decompaction of the lin-4 locus at a timepoint with BLMP-1 expression has resumed (Figure 7D). Interestingly, the level of decompaction is greater in recovering animals that those found in non-starved, continuously developing animals (Figure 7D). We also monitored lin-4::mCherry/LacO loci in the seam cells of starved, L1-stage animals and also found that loci compacted in these starvation conditions (Figure S7). In sum, these experiments indicate that dynamic chromatin rearrangements are modulated by nutrient sensing.

Changes in BLMP-1 expression and the dramatic compaction and decompaction of BLMP-1 target chromatin during starvation and recovery, respectively, suggests that BLMP-1 could play an essential role in modulating transcriptional output of critical developmental genes in dynamic environments. To determine whether animals that lack blmp-1 and/or elt-3 can robustly adapt to starvation, we repeated the food experiments as described in Figure 7A but maintained the animals in starvation conditions for increasing periods of time. At the end of the starvation period, animals were re-plated on normal food and monitored for developmental phenotypes by scoring changes in col-19::GFP expression when the animals had reached adulthood (Figure 7E). In these experiments, wild-type and elt-3(0) larvae re-initiate normal temporal patterning after prolonged bouts of starvation (Figure 7E). In contrast, a minor fraction of blmp-1(0) mutants failed to express col-19::GFP in hyp7 cells, a phenotype associated with weak reiterative heterochronic mutants.^17,48 The relatively mild blmp-1(0) phenotypes were dramatically enhanced in blmp-1(0); elt-3(0) double mutants. These results indicate that BLMP-1 and ELT-3 are essential for resuming normal temporal patterning after continuous development has been interrupted by starvation.

Discussion

Features of the oscillatory expression program impose constraints on modulating transcriptional output

Oscillatory patterns of gene expression are a ubiquitous feature of developmental systems where they orchestrate repetitive biological functions and/or anticipate recurring environmental conditions.⁴⁹ While most GRNs that program cyclical patterns of gene expression utilize a basic negative feedback structure, specific features of their architecture define the precision of the timekeeping mechanism and the ability to modulate and/or preserve features of dynamic transcription. For instance, the periodicity of circadian gene expression persists in the absence of external cues over a wide range of temperatures, yet changes in lighting or nutrition can alter the phase of gene expression or the amplitude of clock target gene transcription, respectively.^50,51 The genetic control of plant root formation also employs a complex feedback loop that generates cyclical transcriptional cycles that induce the formation of lateral roots at periodic positions along the radial surfaces of the developing primary root. The transcriptional output of this GRN is modulated by a hormone, auxin, that amplifies oscillatory transcription in a dose-dependent manner.⁵² Exposure to light is required for maintaining the oscillatory transcriptional patterns at the pre-branch sites.⁵³ Water availability can also modulate the ability to the root clock to induce lateral root founder cell development. Uneven water distribution on the surface of the developing primary root leads to a suppression of LR development on the dryer root surface. Importantly, changes in water distribution do not alter the cycling of core root clock components indicating the environment can control clock output at the phenotypic level.⁵⁴

While the molecular components of the C. elegans expression oscillator are not fully known, a detailed characterization of the systems-level behavior of this GRN indicates that two features are hard-wired. First, the biological oscillator is directly tied to the molting cycle and gated by genetically-regulated developmental checkpoints.^8,47,55 Under replete nutritional conditions, transcriptional cycling is continuous.⁸ Removal of food either at hatching or just prior to the termination of the molt results in a regulated arrest of the developmental clock at a precise point within the transcriptional cycle (Figure 7B).⁸ Re-initiation of the transcriptional clock (either from starved L1-stage animals, dauer arrest, or the nutrition-mediated arrest points at each larval stage) occurs at this same point.⁸ These features indicate that once initiated, the transcriptional cycle is modular in nature and progresses until the next checkpoint. Secondly, the periodicity of thousands of cyclically expressed transcripts scales inversely with temperature over a wide dynamic range and the relative timing of expression for individual genes within the transcriptional cycle is phase-locked.^7,8

Because the accordion-like, modular structure of this GRN constrains the relative timing of transcriptional events within the cycle, other features of cyclically transcribed genes are likely modulated independently of the clock. This adaption of transcriptional output to various environments would be especially important for genes that are expressed at determinant phases within the transcriptional cycle and function in dosage-sensitive manners. This feature is exemplified by the heterochronic miRNAs that exhibit pulsatile transcriptional patterns at each larval stage and also function to control discrete transitions of temporal cell fate at specific developmental milestones (molting).⁵⁶ It had been previously shown that LIN-42, the C. elegans Period ortholog, functions to limit transcriptional output of this class of genes by negatively regulating the overall duration of transcription within each larval stage.^5,6,12 In this manuscript, we present evidence that BLMP-1 functions to antagonize LIN-42 activities by increasing the duration of transcription for cyclically expressed genes. As with lin-42, blmp-1 is not essential for the generation of cyclical expression patterns but plays a modulatory role in controlling the duration and amplitude of transcription (Figure 7G). How the interplay between these two antagonistic modulators of gene expression achieves precise tuning of gene-dosage to confer robustness and adaptivity to post-embryonic developmental programs is an exciting route for further investigation given that BLMP-1 targets the lin-42 locus (Data S1A–D).

BLMP-1 couples chromatin de-compaction and the regulation of transcriptional output

In C. elegans, BLMP-1 is known to physically interact with chromatin remodeling components (SWSN-1, a core component SWI/SNF of the complex (orthologous to human SMARCC1/BAF155) and HAM-3, an accessory subunit of the SWI/SNF complex) suggesting that BLMP-1 may direct these activities in a sequence-specific manner.³² Mouse Blimp-1 has been shown to alter transcriptional activity of target genes by recruiting chromatin remodeling and histone modifying complexes.⁵⁷ The physical associations are conserved between the human ortholog of BLMP-1, hBlimp1, and the hSWI/SNF complex; suggesting additional features of BLMP-1 function may be shared. Here, we characterize a molecular role for BLMP-1 in regulating the transcriptional output of a number of cyclically expressed genes in C. elegans through a mechanism involving the decompaction of chromatin near its target gene loci. In contrast to the chromatin decompaction that occurs during normal transcriptional activation^38–40, we demonstrate that BLMP-1-dependent decompaction is temporally separated from the activation of the target gene expression and can occur even when transcription is actively repressed by other mechanisms (e.g. for the lin-4 gene during embryogenesis (Figure 6B and D).⁴⁴ We hypothesize that this anticipatory priming mechanism constitutively opens chromatin loci near BLMP-1 binding sites. These “primed” loci would then be accessible to additional TFs whose binding capacity may be normally restricted by nucleosome complexes. Once bound, these factors would induce the phased patterns of gene expression that define the transcriptional cycle for each target gene (Figure 7 G). Because BLMP-1 binding sites are associated with a multitude of genes that are expressed in diverse phases of the oscillatory transcriptional pattern (Figure 3), we speculate that the BLMP-1-dependent “priming” activity is not limited to a single partner TF but likely facilitates the association of several TFs that function at distinct phases of the transcriptional cycle. In the absence of BLMP-1, many of these factors would exhibit a reduced capacity to bind to their cognate binding sites. As a result, blmp-1(0) mutants exhibit broad changes in transcriptional dynamics. The diversity of these changes would result in pleiotropic phenotypes whose only common feature would be a change in gene dosage across multiple GRNs. Notably, while our analysis focused on genes whose overall expression is upregulated in the presence of BLMP-1, the mechanism we propose by which BLMP-1 modulates target transcription is equally capable of inhibiting transcription of target genes (by increasing chromatin accessibility for transcriptional repressors).

The molecular mechanism that we propose for BLMP-1 in temporal gene regulation is analogous to the role that the TF Zelda (ZLD) plays in establishing normal spatial gene regulation during D. melanogaster embryogenesis. In this system, ZLD is ubiquitously expressed throughout the embryo where it functions as a pioneer factor to rearrange chromatin near its binding sites.^58–60 Spatially ubiquitous ZLD activity facilitates the transcriptional activation of genes that are co-targeted by the transcription factor Dorsal (DL). Unlike ZLD, DL is expressed in a dorsal-ventral gradient that extends from the dorsal surface of the embryo to the lateral edge. DL functions in a dosage-dependent manner to generate a graded expression of target genes.^61,62 Mutation of ZLD binding sites in DL-dependent reporter transgenes results in the dramatic compression of reporter gene expression along the dorsal-ventral axis.⁶³ Importantly, zld mutations do not alter DL distribution along the dorsal-ventral axis suggesting that ZLD activity primes DL targets by reorganizing chromatin architecture that would otherwise impair DL binding and function. As with mutations of BLMP-1 in the context of temporal gene expression, zld mutations additionally alter the duration of DL target expression.⁶³

Regulation of BLMP-1 expression in response to nutrient availability coordinates transcriptional output

We hypothesize that the re-initiation of transcription after starvation elicited by food is essential for animals to re-format the chromatin landscape near their target genes after quiescence and to rapidly modulate the transcriptional output of genes that control temporal patterning. In this manuscript we also show that blmp-1 expression is regulated by nutrient availability and is important for the coordination of transcriptional output in a variety of conditions. Under nutritionally replete growth conditions, blmp-1 transcription is both cyclical and auto-regulatory. While blmp-1 expression at the mRNA level is highly dynamic, BLMP-1 protein expression is highly stable and maintained at relatively constant levels during continuous growth.^7,8,16 Starvation elicits two main changes in blmp-1 expression that directly lead to a reduction in the priming activity. First, the cyclical transcription of blmp-1 (as with most other cyclically expressed genes) is arrested when animals pause development at a starvation-induced developmental checkpoint. Second, BLMP-1 expression is rapidly curtailed during starvation. BLMP-1 expression is likely dampened by the combined reduction in transcription and the constitutive activity of DRE-1 that functions independently of starvation.¹⁶ This reduction of BLMP-1 expression is coincident with a re-compaction of chromatin near BLMP-1 target genes. The establishment of non-permissive chromatin architecture may be a common feature of quiescent cells as diverse cell types (from yeast to human cells) exhibit these changes. It has also been previously suggested to control inappropriate transcription in the context of development.^64–69 The regulation of BLMP-1 expression through nutrient availability is also important for the coordination of gene expression after starvation as blmp-1(0) and blmp-1(0); elt-3(0) animals fail to resume normal temporal patterning after starvation.

Finally, we hypothesize that the active control of BLMP-1 levels/activities may be important for the adaption of global transcriptional patterns to the environment. The separate genetic control of transcriptional timing and transcriptional output would enable gene expression dynamics to be optimized to specific conditions and ensure developmental robustness. Evidence that the levels of BLMP-1 protein are actively regulated are derived from the observation that dre-1(lf) phenotypes result in an inappropriate over-expression of BLMP-1 and a coincident increase in precocious developmental phenotypes.¹⁶ The active control of BLMP-1 protein expression may play an important role in the adaptation of transcriptional output to different temperatures where the periodicity of the transcriptional cycles and developmental pace changes dramatically.^7,8 This homeostatic strategy, mediated by controlling the expression of a single pioneer transcription factor, could buffer differences in transcriptional dynamics across the transcriptional cycle by modulating chromatin accessibility.

STAR METHODS

LEAD CONTACT AND MATERIALS AVAILABILITY

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Christopher M. Hammell (chammell@cshl.edu). Plasmids generated in this study are available upon request made to the Lead Contact. C. elegans strains generated in this study are available upon request made to the Lead Contact.

Data and Code Availability

A public repository containing MATLAB scripts used for figures and statistical analyses in this paper can be found at https://github.com/wolfgangkeil/Stec_et_al_2020_code.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

C. elegans maintenance and genetics

C. elegans strains were maintained on standard media at 20°C and fed E. coli OP50.⁷¹ A list of strains used in this study is provided in the Key Resources Table. Some strains were provided by the CGC, which is funded by NIH Office of Research Infrastructure Programs (P40 OD010440). blmp-1(tm548) was obtained from Shohei Mitani the National BioResource Project (NBRP) at the Tokyo Women’s Medical University.

Key Resources Table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-GFP	Abcam	ab290
Bacterial Strains
E. coli OP50	CGC	N/A
BL21(DE3)	Novagen	69450-3
C. elegans TRX Factor RNAi library	Source Biosciences	73
Chemicals, Peptides and Recombinant Proteins
pGex-3x	GE Lifesciences	28954654
GST-BLMP-1(ZnFs)	this study	GST-BLMP-1(ZnFs)
Critical Commercial Assays
MinElute PCR purification kit	Qiagen	28004
NEBuilder HiFi DNA Assembly Master Mix	New England Biolabs	E2621L
Sheared salmon sperm DNA	Invitrogen	AM9680
Amicon Ultra-4 Centrifugal Filter Unit	EMD Milipore	UFC801008
Experimental models: Organisms/Strains
Figure 1
lin-42(n1089); maIs105 [col-19::GFP]	5	VT2210
blmp-1(tm548); lin-42(n1089); maIs105 [col-19::GFP]	this study	HML261
lin-42(n1089); maIs105 [col-19::GFP]; elt-3(gk122)	this study	HML467
blmp-1(tm548); lin-42(n1089); maIs105 [col-19::GFP]; elt-3(gk122)	this study	HML468
unc-119(ed3); wgIs75 [elt-3::TY1::EGFP::3xFLAG + unc-119(+)]	CGC	OP75
unc-119(ed3); wgIs109 [blmp-1::TY1::EGFP::3xFLAG + unc-119(+)]	CGC	OP109
wIs51 [SCM::GFP]	this study	HML243
wIs51 [SCM::GFP]; alg-1(gk214)	this study	HML307
Figure 2
wild type (Bristol)	CGC	N2
blmp-1(tm548) I	this study	HML293
elt-3(gk122) X	CGC	VC144
blmp-1(tm548)I; elt-3(gk122) X	this study	HML630
mgIs49[mlt-10::gfp-pest; ttx-3::gfp]	CGC	GR1395
blmp-1(tm548); mgIs49[mlt-10::gfp-pest; ttx-3::gfp]	this study	HML274
blmp-1(tm548);mgIs49[mlt-10::gfp-pest; ttx-3::gfp];elt-3(gk122)	this study	HML474
mgIs49[mlt-10::gfp-pest; ttx-3::gfp]; elt-3(gk122)	this study	HML475
let-60(n1046)/tmC5	this study	HML922
let-60(n1046)	this study	HML923
let-60(n1046)/tmC5; elt-3(gk122)	this study	HML924
blmp-1(tm548); let-60(n1046)/tmC5	this study	HML925
let-60(n1046); elt-3(gk122)	this study	HML927
let-60(n1046)/tmC5; blmp-1(tm548); elt-3(gk122)	this study	HML928
blmp-1(tm548); let-60(n1046)	this study	HML929
blmp-1(tm548); let-60(n1046); elt-3(gk122)	this study	HML931
Figure 3
unc-119(ed3); wgIs75 [elt-3::TY1::EGFP::3xFLAG + unc-119(+)]	CGC	OP75
unc-119(ed3); wgIs109 [blmp-1::TY1::EGFP::3xFLAG + unc-119(+)]	CGC	OP109
Figure 4
unc-119(ed3); cshEx208[lin-4(minimal)SL2nlsnls-GFP-pest]	this study	HML839
unc-119(ed3); cshEx212[lin-4(long_deltaA)SL2nlsnls-GFP-pest]	this study	HML843
unc-119(ed3); cshEx217[lin-4(long_delta_C)SL2nlsnls-GFP-pest]	this study	HML848
unc-119(ed3); cshEx219[lin-4(long_deltaD)SL2nlsnls-GFP-pest]	this study	HML850
unc-119(ed3); cshIs120[lin-4long::SL2nlsnls::GFP-pest, ttx-3::GFP]	this study	HML815
unc-119(ed3); cshEx206[lin-4(long_deltaA+D)SL2nlsnls-GFP-pest]	this study	HML875
unc-119(ed3); cshEx84[Pgst-5::GFP-pest; ttx-3::GFP; unc-119(+)]	this study	HML382
unc-119(ed3); cshEx85[P(2xPCE)gst-5::GFP-pest; ttx-3::GFP; unc-119(+)]	this study	HML383
Figure 5
unc-119(ed3); cshIs83(4xBC)(moe-3::GFP-pest); blmp-1(tm548)	this study	HML728
unc-119(ed3); cshIs83(4xBC)(moe-3::GFP-pest); elt-3(gk122)	this study	HML729
unc-119(ed3); cshIs83(4xBC)[moe-3::GFP-pest]; blmp-1(tm548), elt-3(gk122)	this study	HML730
unc-119(ed3); cshIs83(4xBC)[moe-3::GFP-pest]	this study	HML731
cshIs70[ZK180.5::GFP ; ttx-3::GFP] 6XBC	this study	HML620
blmp-1(tm548); cshIs70[ZK180.5::GFP-pest; ttx-3::GFP; unc-119(+)]	this study	HML697
cshIs70[ZK180.5::GFP-pest; ttx-3::GFP; unc-119(+)]; elt-3(gk122) X	this study	HML558
blmp-1(tm548); cshIs70[ZK180.5::GFP-pest]; elt-3(gk122)	this study	HML698
mgIs49[mlt-10::gfp-pest; ttx-3::gfp]	CGC	GR1395
blmp-1(tm548); mgIs49[mlt-10::gfp-pest; ttx-3::gfp]	this study	HML274
blmp-1(tm548);mgIs49[mlt-10::gfp-pest; ttx-3::gfp]; elt-3(gk122)	this study	HML474
mgIs49[mlt-10::gfp-pest; ttx-3::gfp]; elt-3(gk122)	this study	HML475
cshIs106[P(2xPCE)gst-5::GFP-pest]	this study	HML817
cshIs106[P(2xPCE)gst-5::GFP-pest]; blmp-1(tm548)	this study	HML942
cshIs106[P(2xPCE)gst-5::GFP-pest]; elt-3(gk122)	this study	HML943
cshIs106[P(2xPCE)gst-5::GFP-pest]; blmp-1(tm548), elt-3(gk122)	this study	HML830
cshIs56[C02E7.6::GFP-pest]	this study	HML539
blmp-1(tm548); cshIs56[C02E7.6::GFP-pest]	this study	HML537
cshIs56[C02E7.6::GFP-pest]; elt-3(gk122)	this study	HML538
blmp-1(tm548); cshIs56[C02E7.6::GFP-pest]; elt-3(gk122)	this study	HML959
Figure 6
blmp-1(tm548); cshIs116[lin-4p::mCherry w/ LacO-ISceI]; otIs593[let-858::GFP::LacI; ttx-3::mCherry]	this study	HML891
cshIs116[lin-4p::mCherry w/LacO-ISceI]; otIs593[let-858::GFP::LacI; ttx-3::mCherry]	this study	HML892
cshIs41[BLMP-1::GFP]	this study	HML396
Figure 7
blmp-1(tm548); maIs105[col-19::GFP]	this study	HML260
maIs105[col-19::GFP]; elt-3(gk122)	this study	HML458
blmp-1(tm548); maIs105[col-19::GFP]; elt-3(gk122)	this study	HML459
maIs105[col-19::GFP] V	48	VT1367
blmp-1(tm548); cshIs116[lin-4p::mCherry w/ LacO-ISceI]; otIs593[let-858::GFP::LacI; ttx-3::mCherry]	this study	HML891
cshIs116[lin-4p::mCherry w/LacO-ISceI]; otIs593[let-858::GFP::LacI; ttx-3::mCherry]	this study	HML892
Recombinant DNA
GST-BLMP-1(ZnFs)	this study	pCMH1417
mcs-nls-GFP-pest	this study	pCMH1225
Plin-4(long)::GFP-pest	this study	pCMH1359
lin-4::GFP-pest	this study	pCMH1595
Plin-4(FL)-SL2-2xnls-GFP-pest	this study	pCMH1954
Plin-4(minimal)-SL2-2xnls-GFP-pest	this study	pCMH1982
Plin-4(ΔA)-SL2-2xnls-GFP-pest	this study	pCMH1974
Plin-4(ΔC)-SL2-2xnls-GFP-pest	this study	pCMH1978
Plin-4(ΔD)-SL2-2xnls-GFP-pest	this study	pCMH1980
Plin-4(ΔAΔD)-SL2-2xnls-GFP-pest	this study	pCMH2000
256x LacO sites	this study	pCMH1972
Simple_Plin-4-SL2-2xnls-mCherry	this study	pCMH1984
ttx-3::GFP co-injection plasmid	Hobert Lab	ttx-3::GFP
pMCS5	Mo Bi Tec	pMCS5
unc-119(+)	H17	pIF9
mCherry ORF	this study	pCC318
Pgst-5::GFP-pest	this study	pCMH1418
P(2xPCE(TINY))gst-5::GFP-pest	this study	pCMH1379
Pmoe-3::GFP-pest	this study	pCMH1545
pZk180.5::GFP-pest	this study	pCMH1591
C02E7.6::GFP-pest	this study	pCMH1590
Blmp-1 CRISPR	this study	pCMH1406
Blmp-1 CRISPR guide	This study	pCMH1410
Software and Algorithms
ImageJ	Rasband,W.S (NIH)	https://imagej.nih.gov/ij/
Graphpad Prism	GraphPad	https://www.graphpad.com/scientific-software/prism/
Zen	Zeiss	https://www.zeiss.com/microscopy/us/products/microscope-software/zen.html
SnapGene	Insightful Science	snapgene.com

METHOD DETAILS

RNAi feeding

RNAi by feeding was performed using E. coli HT115 expressing double stranded RNA corresponding to blmp-1 or elt-3, or containing a control dsRNA expression plasmid that does not contain sequence corresponding to any C. elegans gene.^72,73 To prevent contamination by E. coli OP50, gravid adult animals were added to RNAi plates individually after removing cotransferred bacteria. Unless otherwise noted, F1 progeny were analyzed for RNAi-induced phenotypes following 2-4 days.

Transgenesis

Cloning was performed using the Gibson Assembly Master Mix (New England Biolabs, #E2611L). Extra-chromosomal arrays were generated using ttx-3::GFP as a co-injection marker and unc-119 rescue to select transformants. Gamma ray mutagenesis was used to induce integration of extrachromosomal arrays.⁷⁴

GFP tagged alleles of blmp-1

We amplified two 0.5kb-long homology arms from N2 genomic DNA flanking the blmp-1 gene and the GFP open reading frame using standard PCR. These fragments were used with Gibson cloning to insert them into a pMCS5 backbone. The resulting plasmid, pCMH1406, was used in combination with pCMH1410 and standard protocols⁷⁵ to integrate GFP at the blmp-1 locus. The resulting animals were characterized phenotypically for alae defects that are associated with blmp-1(0). The GFP-tagged allele was wild type.

Microscopy

Images were acquired with a Zeiss Axio Observer microscope equipped with Nomarski and fluorescence optics as well as a Hamamatsu Orca Flash 4.0 FL Plus camera. An LED lamp emitting at 470 nm was used for fluorophore excitation. For single images, animals were immobilized on 2% agarose pads supplemented with 100mM Levamisole (Sigma). For long-term imaging methods, see Microfluidics and Long-term Imaging section.

Chromatin immunoprecipitation

BLMP-1 short reads were first clipped off adapter sequences. Reads of minimum 22bp were mapped to UCSC C. elegans genome (ce10) using bowtie program⁷⁶, looking for unique alignments with no more than 2 mismatches. MACS program (v1.4)⁷⁷ was used for peak calling with significant p-value cutoff equals 1e-5. Target annotations were based on WormBase (version 220) using customized R scripts and Bioconductor packages. We defined the promoter region as upstream 3kbp to downstream 300bp around transcription start site (TSS). Peaks located in promoter regions were annotated to their closest TSS sites for both coding and non-coding genes. These potential targets were then overlapped to sets of oscillatory genes identified in previous mRNA-seq-based studies^7,8 and the two-tailed Fisher’s test was used to calculate the p-values associated with these classifications. ELT-3 (Embryo and L1-staged) raw data were downloaded from modENCODE and processed in the same way as BLMP-1. Peaks from each CHiP-seq dataset were compared to identify common sites with at least one base pair overlapping using BEDTools.⁷⁸ All ChIP-seq mapping graphs and images were produced in R by customized scripts.

Microfluidics and long-term imaging

For microfluidics experiments, early to mid L1-staged animals were isolated through a 1-2 hours timed egg-lay approximately 12-14 hours at 20°C before an experimental time course. Animals were mounted into the microfluidic device as previously described.³¹ During the imaging process, animals were constantly fed NA22 E. coli suspended in S medium. Imaging was done with a 40x, 1.3NA objective using an Andor Zyla 4.2 scMOS camera. The temperature was kept constant at 20°C both, at the objective and the microfluidic device using a custom-built water-cooled aluminum ring (for objective) and custom-built aluminum stage inset that was directly coupled to a thermal Peltier device.³¹

Native gel shifts

For recombinant GST-BLMP-1(ZnFs) protein expression, an approximately 400bp PCR fragment (with BamHI ends) was subcloned into the BamHI sites of pGex-3X. The resulting plasmid, pCMH1417, or pGEX-3X alone, was transformed into BLM21(DE3) E. coli. Cultures of these bacteria were grown to an O.D. of 0.5 A600 in LB+amp and induced with 0.2mM IPTG for 2 hours at 37°C. Cell pellets were resuspended in PBS and then sonicated and clarified by centrifugation. Fusion proteins were isolated on glutathione-agarose and eluted with glutathione. The resulting fusion proteins were then dialyzed in 10 mM Tris 7.5, 50 mM NaCl, 0.01% NP-40 1 mM DTT overnight and snap frozen in liquid nitrogen. For gel shifts with free DNA, 5’ IRDye (IRDye700 or IRDye800)-labeled and unlabeled oligos were obtained from IDT (Coralville, Iowa). The sequences of these oligos are the following:

• lin-4 site 1 sequence (+) CTTTCTCCTTCACTTTCTCTCTCTCGGATCACCAG

• lin-4 site 1 sequence (−) CTGGTGATCCGAGAGAGAGAAAGTGAAGGAGAAAG

• lin-4 site 2 sequence (+) ATGGGAAAATAGAAAGAGAAAATGAGAGAAAGGAT

• lin-4 site 2 sequence (−) CTTTCTCTCATTTTCTCTTTCTATTTTCCCAT

• non-specific sequence (+) GGCTCTTCCGGTCGTGCAAGAGGAAATACAAAAAA

• non-specific sequence (−) TTTTTTGTATTTCCTCTTGCACGACCGGAAGAGCC

Complementary oligos were annealed in TE to a final concentration of 0.1 pmol/μL. 0.1 pmol of these probes were incubated with recombinant GST or GST-BLMP-1(ZnFs) in 10 mM Tris pH 7.5, 50 mM KCl, 1 mM DTT, 0.1μg/μL poly (dIdC) 0.25% Tween 20 for 30 minutes at 20°C (in dark chamber) and then run in a 4% native polyacrylamide gel containing 50mM Tris pH 7.5, 0.38 M glycine and 2mM EDTA in 1x TBE buffer. Gels were imaged using a Licor Odyssey Imager (Lincoln, Nebraska).

Chromatin decompaction assays

To analyze the compaction state of the lin-4 locus, we exploited the LacI/LacO approach as previously described.³⁷ The arrays containing lin-4 were generated by co-injecting the following components: a) 2ng/μL of linearized pIF9 (unc-119(+)), b) 3ng/μL of linearized pCMH1972 (256x LacO sites), 3ng/μL of pCMH1984 (lin-4pro::SL2::mCherry-pest), and 90ng/μL of sheared salmon sperm DNA (Invitrogen). Transgenic lines produced in this manner (e.g. cshIs116) were integrated using standard procedures and outcrossed into a strain harboring otIs593 (ttx-3::mCherry; let-858pro::GFP::Lacl)(Gift of O. Hobert). For imaging nuclear spots assay of either eggs or staged L3/L4 molt animals, Z-stacks (0.3μm spacing) were collected using three channels: DIC, GFP (150ms exposure) and mCherry (60ms exposure). Both acquisition and post-acquisition processing was performed using ZEN Pro software (Zeiss). Maximum-intensity projections of the planes that span the nuclei of interest were obtained separately for seam and head area. The projections were then exported to Fiji⁷⁹ and manual traces of areas of the nuclei with GFP intensity above background were made to measure locus compaction/decompaction as described.³³

Food removal assays

Starvation experiments were performed as previously described.⁴⁷ Briefly, embryos were isolated from gravid adult animals using a standard, hypochlorite treatments and allowed to hatch in M9 buffer. L1 staged/arrested animals were plated onto NGM plates seeded with OP50 E. coli at a concentration lower that 2,500 animals per 60 mm petri dish. Animals were then grown for the indicated amount of time at 20°C unless otherwise noted. Assessment of developmental age was made by direct observation of gonad growth or by specific cell lineage landmarks of lateral seam cells or developing vulval precursor cells. Acute starvation was induced by washing animals off the petri dishes and washing them with M9 for nine times. After the final was animals were resuspended in M9 buffer to a concentration of ~1 animal per μL of sterile M9. Animals were maintained in these starvation conditions at 20°C for the indicated times. Animal development was re-initiated by plating animals onto fresh OP50 E. coli-seeded NGM plates at 20°C.

Quantification Of Gene Expression using GFP reporters

Fluorescence intensities were quantified using ImageJ custom-written MATLAB scripts. software. For measurement of reporter gene expression for time course experiments, we performed dual channel imaging, visualizing gene expression patterns by visualizing GFP-pest expression and overall worm morphology with Nomarski microscopy (DIC). We straightened each three-dimensional image stack using a previously published algorithm³¹ based on a worm backbone which was manually outlined using the DIC channel. The resulting straightened worm z-stacks were then cropped in xy to a region of interest. To deconvolve the straightened, cropped GFP-pest reporter z stacks, we measured the point-spread function (PSF) of our optical setup using red (580/605nm) fluorescent 200nm beads (ThermoFisher Scientific). Using the obtained PSFs, we deconvolved stacks using the classic maximum likelihood estimation (CMLE) algorithm with standard parameters (refractive index of imaging medium: 1.338) in the Huygens Essential software (Huygens Essential 3.7.1) by Scientific Volume Imaging (SVI).

Statistical Analysis

All statistical analyses were performed in MATLAB or GraphPad Prism 9, either using built-in functions (chi squared, Student’s t tests) or custom-written scripts (bootstrap permutation test). Differences between two groups are judged to be statistically significant when p < 0.05 by Student’s t test or bootstrap permutation test, where appropriate. Mean and SEM were calculated and plotted using GraphPad Prism or custom scripts. T-tests analyses are based on the normal distribution theory. We did not test the assumption of normality because even when the raw data are not exactly normally distributed, these analysis methods are very robust to departures from normality as long as the sample sizes are larger than approximately 30. For Figure 3, changes in translational reporter expression for BLMP-1::GFP and ELT-3::GFP were quantified as previously described. ³¹ Each circle in panel F represents the average GFP fluorescence for 3 hypodermal cells per animal for each translational reporter. For Figure 4, n = number of animals scored in each assay. GraphPad Prism software was used to calculate Mean expression level for each time point measured. Each circle represents the average GFP fluorescence for 3 seam cells per animal. For Figure 5, GraphPad Prism software was used to calculate Mean expression levels. For A-C, each circle represents the average GFP fluorescence for 3 seam cells per animal. For E, each circle represents the duration from one ecdysis event to the next in the L1 to L2 molt cycle. Image analysis was performed as in Keli et al. 2016.³¹ Estimation plots were calculated as previously describted.⁷⁰ For Figure 6, panel E, n = number of animals scored in each assay. For Panel G, GFP-LacI areas were quantified from greater that 18 individual nuclei from separate animals for each genotype. GraphPad Prism software was used to plot image data and box plots show data within the 10-90^th percentile with whiskers. For Figure 7, n = number of animals scored in each assay. GraphPad Prism software was used to calculate Mean. For panel D, GFP-LacI areas were quantified from greater than 20 seam cell nuclei from separate animals in each condition.

Supplementary Material

1

Data S1. Summary of ChIP-seq experiments for BLMP-1::GFP and ELT-3::GFP, related to Figure 3. Tables include gene assignments for L1-L4 BLMP-1 (Data S1A-D) and L1 ELT-3 ChIP-seq data sets (Data S1E).

2

Data S2. Comparison of ChIP-seq data to mRNA expression classes as defined by Hendriks et al.⁸ and Kim et al.⁷ time course data sets, related to Figure 3. Tables include gene class assignments (flat, oscillating and rising) of ModEncode ChIP-seq data (Data S2A) and assignments of BLMP-1 ChIP-seq data sets to expression cluster data (Data S2B).

Highlights.

• BLMP-1 functions as a pioneer factor to decompact chromatin at target gene loci

• BLMP-1-dependent priming controls the output of transcriptional pulses

• Nutritional availability regulates BLMP-1 expression and activity

• Animals require BLMP-1 activity to recover from starvation conditions