Safeguarding epithelial junctions by a novel quality control pathway

Abstract

Cell junctions establish and maintain epithelial architecture despite fluctuating environmental and developmental conditions. A central question is how cells respond to challenging conditions to preserve junctional integrity. Here, we report the discovery of a previously unrecognized quality control pathway that monitors epithelial junctions (J-QC). We used the Caenorhabditis elegans epidermis as a model to investigate the DLG-1-AJM-1 complex (DAC), a junctional domain that is critical for embryonic morphogenesis. We identify two mechanisms that sustain junctional integrity: first, localized dlg-1 mRNA ensures appropriate DLG-1 protein levels at the junction; repositioning dlg-1 RNA reduces DLG-1 levels, leading to gaps between epithelial cells. Second, transcription of DAC components responds to perturbations that disrupt the DAC. This response is sequence-independent, distinguishing it from other quality control mechanisms. It is activated by perturbations of the DAC or cytoskeleton and requires the LINC complex component ZYG-12/HOOK1–3 to transduce information about junctional integrity to the nucleus. These findings define a novel junctional QC for epithelial maintenance.

Epithelial junctions, such as adherens and tight junctions, organize and compartmentalize tissues^1–3. By forming continuous contacts between neighboring cells, epithelial junctions establish an apical-basolateral boundary, maintain epithelial cohesion, and create barriers that separate the internal from the external environment⁴. However, little is known about how established junctions maintain homeostasis as epithelia experience developmental remodeling and mechanical stress. This question carries broad biological and biomedical significance, as disruption of epithelial junctions underlies pathological conditions, including the epithelial-to-mesenchymal transition during cancer progression and intestinal barrier dysfunction during aging^5–7. Junctional integrity is equally critical during embryogenesis, where junctions contribute to morphogenesis and body axis elongation in many organisms^2,8–10. During these processes, epithelial junctions are subjected to persistent mechanical and developmental stimuli, including cellular rearrangements and body elongation, which continuously challenge their integrity. Extrusion of damaged cells is one established route for preserving epithelial integrity, executed non-autonomously by neighboring cells¹¹. Mechanisms that maintain epithelia by a cell-autonomous route are largely unknown.

Here, we uncover how C. elegans embryonic epidermal cells maintain cell junction integrity during morphogenesis. Using a genetic approach combined with imaging and quantitation analyses, we find that RNA localization sustains appropriate protein levels of the junction component DLG-1 and maintains DAC architecture. In addition, we identify a novel Junctional Quality Control (J-QC) pathway that surveils the DAC and, upon perturbation, upregulates transcription of DAC components. The J-QC relies on cytoskeletal and LINC components to relay information from the junction to the nucleus. Unlike previously described QC pathways, which monitor organelle integrity¹² or the fidelity of molecular processes¹³, the J-QC surveils a subcellular structure, the junction, and couples its disruption to a compensatory transcriptional response. Together, these define a mode of cellular surveillance dedicated to maintaining a specific architectural structure within the cell.

Results

Subcellular localization of dlg-1 mRNA is required for proper DLG-1 protein levels at epidermal junctions

C. elegans, epithelial junctions share structural, molecular, and regulatory features of adherens and tight junctions present in other organisms, reflecting broad evolutionary conservation of junctional architecture and its regulatory logic^14–16. The junctional components in C. elegans are organized into two domains, the cadherin-catenin complex (CCC) and the DLG-1-AJM-1 complex (DAC)^17–19, which are critical for barrier formation and tissue morphogenesis. Loss of any of these components leads to morphogenetic defects²⁰, caused by the rupture of junctions under mechanical tension during embryo elongation, and ultimately death²¹.

Multiple transcripts coding for components in epithelial junctions preferentially localize to the junction itself during Caenorhabditis elegans morphogenesis²². To determine the roles of this subcellular localization, we redirected dlg-1 mRNA away from the junction using our recently established MS2-MCP technique for C. elegans endogenous mRNAs²³. As recent studies have shown that translation and protein folding can normally occur at the cytoplasmic side of the nuclear pore^24,25, we decided to modify our MS2-MCP system to re-localized dlg-1 mRNA at this location. Briefly, we fused the MS2 coat protein (MCP) to the nucleoporin NPP-9/Nup358 (NPP(+)) to generate an MCP::NPP-9 fusion protein (NPP(+)) as well as an unfused control (NPP(−)) Fig. 1A and S1A-F). Because NPP-9 localizes to the cytoplasmic filaments of the nuclear pore complex^26,27, this fusion protein tethered dlg-1 mRNAs carrying MS2 hairpins (MS2(+)) to the cytoplasmic face of the NPC (Fig. 1A,B). Consistent with this idea, perinuclear dlg-1 mRNAs colocalized with MCP::NPP-9 (Fig. 1B). This approach redirected endogenous dlg-1 transcripts away from the junction while keeping the mRNA configuration identical between test and control conditions.

Figure 1. Localization of dlg-1 transcripts controls junctional DLG-1 protein levels.

A. Schematic representation of an epidermal cell with the adherens junction depicted in yellow as described in Fig. S1A. dlg-1 mRNA with MS2 hairpins in its 3’UTR (MS2(+), magenta). In cyan, upper panel: MCP alone (NPP(−)) bound to dlg-1MS2 mRNA at junctions; lower panel: MCP fused to NPP-9/NUP358 (NPP(+)), bound to dlg-1MS2 mRNA at nuclear pores. B. Fluorescence images of epidermal cells of smFISH for dlg-1 mRNA (magenta), MCP (upper) or MCP::NPP-9 (lower) (cyan), and junctional AJM-1 (yellow). Merged images and zoom-ins of one cell (dotted square) are shown. Transcripts (magenta, arrowheads) localized at the junction (yellow). mRNAs (magenta) colocalize with MCP::NPP-9 perinuclearly (asterisks). Vertical arrows depict large foci within nuclei (transcription sites). Scale bars: 5 μm and 2.5 μm (zoom-ins). C. Live fluorescence images of epidermal cells show gaps (arrows) in junctions (DLG-1::GFP) for cells with re-localized dlg-1 RNA (lower) compared to controls (upper). Scale bars: 5 μm. D. Bar plot with error bars normalized to the control MS2(+); NPP(−) (dark grey) and showing the reduction of DLG-1::GFP fluorescence intensities in MS2(+); NPP(+) (green) strains, where dlg-1 transcripts are re-localized to nuclear pores. Each dot represents the maximum fluorescent intensity of a junction shared between two seam cells (three junctions from the posterior-most seam cells per embryo) marked with endogenous DLG-1::GFP. *** = p < 0.001. For raw data and statistics, see Table S3. E. Fluorescence images of epidermal cells of smFISH for dlg-1 mRNA (magenta), DLG-1::mNG protein (yellow), and merge with DNA (white). Upper panels: wild type. Lower panels: dlg-1(Δc7aa). Arrowheads depict dlg-1 transcripts (magenta) colocalizing with DLG-1 protein at the junction (yellow). Vertical arrows depict large foci within the nuclei (transcription sites). Scale bars: 5 μm. F. Live fluorescence images of epithelia for wild-type DLG-1(wt) (upper) and DLG-1(Δc7aa) (lower). Arrows depict gaps in the junctional DAC (DLG-1::mNG). Scale bars: 5 μm. G. Bar plot with error bars showing the median (magenta) and SD (black) normalized to the wild-type control (light grey). Note the reduction in DLG-1(Δc7aa)::mNG fluorescence (green). Each dot represents the mean integrated fluorescent intensity of endogenous DLG-1::mNG signal from a single Z-stack of whole embryos focusing on epidermal cells. For raw data and statistics, see Table S3.

Quantitation of junctional fluorescent intensities²³ revealed a 50% reduction of DLG-1::GFP fluorescence at the junctions of MS2(+) strains with mRNA localized to nuclear pores (NPP(+)) compared to controls (NPP(−)) (Fig. 1C,D). The variability observed in DLG-1 levels among both test and control strains likely reflects subtle embryo-to-embryo differences, but levels were clearly lower following re-localization. Normally, DLG-1 protein levels rise as embryogenesis progresses. We observed the same trend in the re-localization strains, suggesting this aspect of dlg-1 regulation remained intact, but DLG-1 never achieved the levels seen in wild-type embryos (data not shown). As expected, when dlg-1 mRNA did not contain MS2 hairpins (MS2(−)), transcripts localized normally at junctions in both NPP(−)²³ and NPP(+) backgrounds (Fig. S1B,G,H). In this configuration, DLG-1::GFP protein exhibited comparable fluorescence levels in NPP(−) and NPP(+) conditions, (Fig. S1I,J). This result indicates that the reduced DLG-1 levels observed upon transcript re-localization were not due to secondary effects caused by transgenic NPP-9 expression. We conclude that localized mRNA is critical to establish normal levels of DLG-1 protein at junctions.

In strains with re-localized mRNA (MS2(+); NPP(+) configuration), but not in controls, we observed gaps along the circumference of the DLG-1-marked junctions (Fig. 1C). These gaps are consistent with impaired translation or distribution of wild-type DLG-1 protein resulting from re-localization of its mRNA. Gaps were also observed for AJM-1, a second component of the DAC, suggesting biogenesis of the DAC was delayed or disturbed (Fig. 1B). These data reveal the importance of mRNA subcellular localization to maintain appropriate protein levels and distribution along the periphery of the cell, to ensure junctional integrity.

To complement the mRNA re-localization approach, we sought to perturb junctional dlg-1 mRNA localization through an independent genetic strategy. Our previous work found that the carboxy-terminal part of DLG-1 is required for mRNA localization but dispensable for protein localization in transgenic strains²². Based on these findings, we generated a dlg-1 CRISPR allele lacking seven conserved amino acids within the sequence connecting SH3 and GuK domains (dlg-1(Δc7aa); Fig. S2A) in a strain carrying an mNeonGreen (mNG) fluorescent tag at the carboxy-terminus of DLG-1 (LP598⁵⁰). Previous work demonstrated that the mNG tagged DLG-1 is fully functional and appropriately localized in the wild-type configuration²⁸. Single molecule fluorescent in situ hybridization (smFISH) analyses revealed that endogenous dlg-1 mRNAs carrying the Δc7aa mutation failed to localize properly to embryonic epithelial junctions compared to wild-type controls (Fig. 1E). To determine the consequences of disrupting dlg-1 mRNA localization through this endogenous mutation, we analyzed junction morphology by live imaging. Similar to the RNA re-localization experiment, dlg-1(Δc7aa) embryos displayed discontinuities in junctional organization, characterized by the appearance of gaps along epithelial junctions (Fig. 1E,F and S2B). Similarly to what we observed in our re-localization experiments, quantitation of DLG-1 fluorescence intensities revealed a reduction in mean integrated intensity of junctional DLG-1(Δc7aa) signal relative to controls (Fig. 1G). The reduction was less dramatic than the nuclear pore experiment, possibly because some transcripts were still localized at the junction (Fig. 1E), but the reduction was statistically significant. These experiments show that dlg-1 mRNA localization is required for appropriate DLG-1 protein levels, and re-localization disrupts junctional integrity. Thus, mRNA localization contributes to the robustness of epithelial junctions during morphogenesis.

Quantitation of transcriptional levels by DotQuant

We wondered why animals survived despite mis-localized dlg-1 mRNA, reduced protein levels, and junctional gapping. Close inspection revealed that there was an increase in dlg-1 transcription when transcripts were re-localized. For example, smFISH experiments of dlg-1 mRNA re-localized by MS2 and MCP::NPP-9 had larger and brighter foci in epidermal nuclei of MS2(+); NPP(+) strains compared to controls (Fig. 1B,E). These dots represented transcription sites, based on smFISH for intronic sequences, and their increased size and brightness suggested an upregulation in transcription²⁹. To quantify transcriptional changes in nuclei from different conditions, we developed a rapid pipeline for nuclear dot quantitation (DotQuant, https://github.com/angelo-angonezi/DotQuant). DotQuant combines automated nuclear and three-dimensional spot segmentation and fluorescence measurements in a single workflow and is supported by custom computational tools that enable fast and reproducible analysis of transcription sites across large numbers of nuclei. The pipeline provides the integrated intensity (volume × mean signal intensity) of individual transcription sites as a quantitative measure of transcriptional activity (Fig. S3A-J; Methods). DotQuant enabled quantitative comparison of transcriptional output between different genetic conditions, providing a robust approach to assess changes in dlg-1 transcriptional activity in our study.

Re-localized dlg-1 mRNA triggers compensatory upregulation of dlg-1 transcription, which supports viability

Quantitation of the integrated intensity of these transcription sites by DotQuant showed a 3.4-fold increase for re-localized dlg-1 transcripts compared to controls (Fig. 2A and S3A-J). We conclude that cells upregulate dlg-1 transcription when dlg-1 mRNAs are mis-localized.

Figure 2. Depletion of DLG-1 protein induces dlg-1 gene transcriptional.

A. Intensity of nuclear mRNA dots increases after RNA re-localization. Violin plots with overlaid notched boxplots of distributions of integrated intensities of transcription dots across strains. *** = p < 0.001. For raw data and statistics, see Table S4. B. Bar plot with error bars for embryonic viability in re-localized dlg-1 (light green NPP(+)) versus control (grey) after dlg-1 RNAi. Each dot represents the percentage of hatched and living larvae per biological replicate (x3). n.s. = p > 0.05; *** = p < 0.001. For raw data and statistics, see Table S3. C. Schematic representation of a portion of a seam cell as described in Fig. S1A. Upper panel: wild-type levels of DLG-1 protein at the junction (yellow); wild-type transcriptional levels (thin arrow) of dlg-1 gene (cyan); addition of DMSO (grey) as a negative control. Lower panel: reduced levels of DLG-1 protein at the junction (light yellow); increased transcriptional levels (thicker arrow) of dlg-1 gene (cyan); addition of 5-Ph-IAA (green) as treatment to deplete DLG-1 protein with the AID2 system. In both instances, wild-type localization of dlg-1 transcripts at the junction (magenta). D. Live fluorescence images of epidermal cells of fluorescence signal of DLG-1::EGFP::mIAA7 strain. In upper image: treatment with DMSO at t30 (control); lower image: treatment with 5-Ph-IAA-AM at t30 (green). Scale bars: 5 μm. E. Bar plot with error bars normalized to the reference t0 (dark grey) and showing the changes in DLG-1::EGFP fluorescence intensities throughout the time course experiment (light grey (−): DMSO treatment; green (+): 5-Ph-IAA-AM treatment). Each dot represents the maximum fluorescent intensity of a junction shared between two seam cells (three junctions from the posterior-most seam cells per embryo) marked with endogenous DLG-1::EGFP. ***: p < 0.001. For raw data and statistics, see Table S3. F. Fluorescence images of epidermal cells of smFISH signal for dlg-1 mRNA (magenta) and dlg-1 pre-mRNA (cyan), antibody staining of AJM-1 (yellow), and merge. From top to bottom, the time course experiments: immediately after placing embryos in M9 with DMSO (t0’); after 15 minutes in M9 with 5-Ph-IAA-AM (t15’); after 30 minutes in M9 with 5-Ph-IAA-AM (t30’); after 45 minutes in M9 with 5-Ph-IAA-AM (t45’). Example seam cell nuclei are circled with dashed lines. Arrowheads point at nuclei with larger transcription dots, found in later time points of the 5-Ph-IAA-AM treatment. Scale bar: 5 μm. G. Violin plots with overlaid notched boxplots of distributions of integrated intensities of transcription dots (channel for dlg-1 pre-mRNA, cyan) in the time course experiment shown in (E). For raw data and statistics, see Table S4. *** = p < 0.001.

We asked whether increased dlg-1 transcription buffered the phenotypic consequences of RNA re-localization. We used partial RNAi (1:10 dilution) to reduce dlg-1 mRNA levels, bringing the abundance of re-localized MS2(+); NPP(+) transcripts in line with that of localized MS2(+); NPP(−) control transcripts. Reduction of re-localized dlg-1 mRNAs reduced viability to 40% whereas untreated strains of both genotypes were completely viable (98%; Fig 2B). This result shows that the increase in dlg-1 transcripts promotes viability when mRNAs are mis-localized.

As a second test, we analyzed the dlg-1(Δc7aa) mutant, which also exhibits loss of dlg-1 mRNA localization, reduced DLG-1 protein levels, and increased dlg-1 transcription (Fig. 1E-G and S4A). We reasoned that if transcriptional upregulation contributes to survival of mutant DLG-1(Δc7aa), then partial reduction of dlg-1(Δc7aa) mRNA levels should compromise embryonic viability. Under control RNAi conditions, dlg-1(Δc7aa) embryos exhibited 80% embryonic viability compared to 99% in wild-type animals (Fig. S4B), suggesting that the conserved seven amino acids we deleted are critical for DLG-1. Upon mild dlg-1 RNAi treatment, viability was reduced to 18% in dlg-1(Δc7aa) embryos compared to 53% in controls (Fig. S4B). Thus, partial reduction of dlg-1 mRNA levels exacerbates the dlg-1(Δc7aa) mutant phenotype, supporting the conclusion that increased dlg-1 transcription aids embryonic viability when junctional dlg-1 function is compromised. Together, the data show that increased dlg-1 transcription is functionally important and contributes to embryonic viability when dlg-1 mRNA is mis-localized. These experiments demonstrate that augmented mRNA levels contribute to survival in cases where the DAC is challenged by dlg-1 mutations or re-localization. We call this transcriptional response the junctional quality control pathway or J-QC, based on these and the next experiments.

Reduction in DLG-1 protein triggers the dlg-1 transcriptional increase

We hypothesized that increased dlg-1 transcription responded either to mRNA re-localization per se, or to the consequent reduction in DLG-1 protein at the junction. To distinguish between these possibilities, we degraded DLG-1 protein while leaving dlg-1 mRNA intact and localized appropriately (Fig. 2C). DLG-1 tagged with both mIAA7 and EGFP enabled us to induce and visualize DLG-1 degradation upon addition of the auxin analog 5-Ph-IAA-AM^30,31 (Fig. 2C,D and S5A). In a time-course experiment, DLG-1::EGFP::mIAA7 fluorescence at the junction decreased by 50% within 15 minutes of treatment and 80% within 30–45 minutes, compared to time 0 (Fig. 2D,E). smFISH revealed that despite reduction of DLG-1 protein, junctional dlg-1 mRNA localization was maintained (Fig. 2F, magenta and merge).

dlg-1 transcription increased up to 15-fold after 45 minutes of treatment compared to controls (Fig. 2F,G and S5B). Whereas control cells had no or one transcriptional dot within cells, we observed many nuclei with two active loci after DLG-1 degradation, and these appeared within 30 minutes of treatment. Individual transcription dots were also brighter, compared to controls, suggesting that dlg-1 loci were more active after DLG-1 degradation compared to controls. These findings support the idea that epithelial cells monitor DLG-1 protein levels and respond with a compensatory increase in dlg-1 transcription to restore junctional protein abundance. Degradation of DLG-1 protein for 45 minutes or longer compromised the overall structure of the junction. Junctions no longer formed a continuous belt around the cell but instead appeared discontinuous, as visualized by AJM-1 (Fig. 2F, yellow, and S5C). This phenotype resembled the phenotypes observed after mRNA re-localization as well as published dlg-1 loss-of-function contexts³² (RNAi or hypomorphic mutants). These data show that DLG-1 protein, not junctional dlg-1 mRNA, mediates the J-QC pathway.

A junctional quality control (J-QC) pathway selectively monitors DAC integrity

Some QC systems monitor via sequence complementarity (e.g., transcriptional adaptation^33–35) whereas others are dedicated to a specific membranous organelle^12,13,36. We asked whether the transcriptional upregulation of dlg-1 reflected surveillance of a structure, namely the epithelial junction, or was specific for the dlg-1 sequence. To distinguish between these possibilities, we analyzed a second component of the DAC called AJM-1 (Fig. 3A); dlg-1 and ajm-1 lack sequence similarity at the RNA or protein level. smFISH experiments coupled with quantitation of dlg-1 transcription sites revealed significant upregulation of dlg-1 in ajm-1 mutants compared to a wild-type control (Fig. 3B,C). Conversely, 5-Ph-IAA-AM-mediated depletion of DLG-1 protein or RNAi-mediated inactivation of dlg-1 transcripts led to a significant upregulation of ajm-1 transcription (Fig 3D,E). We observed upregulation in developing epidermal cells, particularly in dorsal epidermal cells (Fig. 3D,E and S6A,B). Because dlg-1 and ajm-1 share no primary sequence similarity, these results indicate that the transcriptional response reflects the compromised integrity of the DAC epithelial junction rather than sequence complementarity. As predicted by this model, loss of ajm-1 did not show additive effects upon simultaneous DLG-1 depletion, consistent with both perturbations acting in the same pathway. We refer to this pathway as the junctional quality control (J-QC), to distinguish it from previously described processes. We note that the nonsense-mediated mRNA decay pathway is required for transcriptional adaptation, but was dispensable for the J-QC, indicating that these are distinct transcriptional pathways (Fig. S6C).

Figure 3. The DAC is a quality control checkpoint for the junctions.

A. Simplified schematic representation of the two sub-complexes listing their components constituting C. elegans adherens junctions. An apical cadherin-catenin complex (CCC) in blue with HMR-1/E-cadherin, HMP-1/α-catenin, and HMP-2/β-catenin. Below the CCC, the DLG-1-AJM-1 complex (DAC) in yellow, named after its components. B. Fluorescence images of epidermal cells of smFISH signal for dlg-1 mRNA (magenta) and dlg-1 pre-mRNA (cyan), and merge. Upper panel: wild-type background. Lower panels: ajm-1 mutant background. Example nuclei of seam and ventral epidermal cells are circled with dashed lines. Scale bar: 5 μm. C. Violin plots with overlaid notched boxplots of distributions of integrated intensities of transcription dots (channel for dlg-1 pre-mRNA) in wild type versus ajm-1 mutant backgrounds shown in (B). *** = p < 0.001. For raw data and statistics, see Table S4. D. Fluorescence images of epidermal cells of smFISH signal for ajm-1 pre-mRNA (cyan), and DLG-1::EGFP (yellow). Upper panel: EV RNAi. Lower panels: dlg-1 RNAi. Example nuclei of seam and ventral epidermal cells are circled with dashed lines. Arrowheads show transcription dots of dorsal epidermal cells. Scale bars: 5 μm. E. Violin plots with overlaid notched boxplots of distributions of integrated intensities of transcription dots for ajm-1 pre-mRNAs in 5-Ph-IAA-AM-untreated (DLG-1 +) versus treated (DLG-1 −) DLG-1::EGFP::mIAA7 embryos. *** = p < 0.001. For raw data and statistics, see Table S4. F. Violin plots with overlaid notched boxplots of distributions of integrated intensities of transcription dots in multiple conditions. n.s. = p > 0.05; *** = p < 0.001. For raw data and statistics, see Table S4. G. Fluorescence images of epidermal cells of smFISH signal for hmr-1 mRNA (magenta) and hmp-1 mRNA (cyan), and merges with DLG-1::EGFP (yellow). Upper panel: EV RNAi. Lower panels: dlg-1 RNAi. Example nuclei of seam and ventral epidermal cells are circled with dashed lines. Scale bars: 5 μm.

We next tested whether perturbation of alternative domains within epithelial junctions could trigger the J-QC. The cadherin-catenin complex (CCC) localizes apically to the DAC and is a separate compartment from the DAC (Fig. 3A). RNAi-mediated reduction of CCC components hmr-1/E-cadherin or hmp-1/α-catenin did not induce transcriptional upregulation of dlg-1, in contrast to ajm-1 (Fig. 3F and S7A-C). Thus, disruption of the CCC alone was insufficient to activate the J-QC.

Next, we asked whether the CCC was required for J-QC triggered by DAC disruption. To test this idea, we combined hmr-1 or hmp-1 RNAi with AID2-mediated depletion of DLG-1 protein (Fig. S7B,C). Reduction of hmr-1 or hmp-1 did not alter dlg-1 transcriptional upregulation compared to DLG-1 depletion alone (Fig. 3F and S6A). Thus, CCC components are not necessary for the J-QC response. Conversely, downregulation of dlg-1 mRNA did not induce transcriptional upregulation of genes encoding CCC components or apical polarity regulators (par-6 and pkc-3) (Fig. 3G and S7D). RT-qPCR analysis of intronic and exonic sequences in wild type and ajm-1 mutant embryos confirmed selective upregulation of dlg-1, but not CCC components, in response to ajm-1 disruption (Fig. S7E). These results indicate that the CCC and DAC, although adjacent within epithelial junctions, behave autonomously and without detectable cross-regulation. Thus, general cellular disturbances cannot account for dlg-1 transcriptional induction, but rather, the effect is specific. In sum, the DAC employs a selective J-QC mechanism that responds precisely to perturbations affecting DAC integrity.

The cytoskeleton acts as a bridge between the DAC and transcriptional regulation

We wondered how aberrant junctions signal to nuclei to alter gene expression. Given the involvement of junctions in cytoskeletal organization, we hypothesized that the cytoskeleton would be required to relay information from the junction to the nucleus. To minimize pleiotropic effects of cytoskeletal disruption, we took advantage of mutants with mild or inducible alterations in cytoskeletal organization. Specifically, we used a viable loss-of-function triple mutant to disrupt actin (act-1,2,3(lf)), and two thermo-sensitive mutants to disrupt the microtubule-based cytoskeleton: one for microtubules (tba-1(ts)), and one for dynein heavy chain (dhc-1(ts)), which drives minus-end-directed transport along microtubules. All three mutants showed a significant increase in dlg-1 transcription compared to controls after a brief inactivation (two hours; Fig. 4A,B and S8A). dhc-1(ts) transcriptional levels were the highest of the three mutants, suggesting a pivotal role for the cytoskeleton in general, and microtubule-mediated movement in particular, for junctional homeostasis (Fig. 4A,B and S8A). We also found that auxin-mediated removal of DLG-1 protein showed no significant difference in dlg-1 transcriptional signal when the cytoskeleton was disrupted (Fig. 4C,D and S8B). The lack of an additive effect upon simultaneous DLG-1 depletion and cytoskeletal disruption suggests that cytoskeletal components, microtubule-dependent retrograde transport, and DLG-1 function within the same pathway to mediate J-QC transcription.

Figure 4. The cytoskeleton and mechanical forces are involved in transmitting the signal from the junction to the nucleus.

A. Fluorescence images of epidermal cells of smFISH signal for dlg-1 mRNA (magenta), dlg-1 pre-mRNA (cyan), and merge with DLG-1::EGFP (yellow). All the strains shown in this panel contain endogenous dlg-1 tagged with EGFP and mIAA7, but in different backgrounds and have been treated with DMSO (grey) for 30’. From top to bottom: wild type (wt); act-1,2,3 hypomorphic mutant (act-1,2,3(lf)); tba-1 thermo-sensitive mutant (tba-(ts)); dhc-1 thermo-sensitive mutant (dhc-(ts)). Animals were grown on NGM plates with OP50 bacteria as food at 15°C, shifted for 2 hours at 24°C. Embryos were then collected, split into two: one half treated with DMSO (grey) for 30 minutes in tubes. Example nuclei of seam and ventral epidermal cells are circled with dashed lines. Scale bars: 5 μm. B. Violin plots with overlaid notched boxplots of distributions of integrated intensities of transcription dots (channel for dlg-1 pre-mRNA) in the four strains shown in (A). n.s. = p > 0.05; ** = p < 0.01; *** = p < 0.001. For raw data and statistics, see Table S4. C. Fluorescence images as described in (A). In this case, embryos were treated with 5-Ph-IAA-AM (green) for 30’. Example nuclei of seam and ventral epidermal cells are circled with dashed lines. Scale bar: 5 μm. D. Violin plots with overlaid notched boxplots of distributions of integrated intensities of transcription dots (channel for dlg-1 pre-mRNA) in the three strains shown in (C). n.s. = p > 0.05. For raw data and statistics, see Table S4. E. Fluorescence images of epidermal cells of smFISH signal for dlg-1 mRNA (magenta) and dlg-1 pre-mRNA (cyan), and merge with DLG-1::EGFP (yellow). The strain shown in this panel is the wild type (N2), on two RNAi treatments: upper panel: EV; lower panel: mel-11. Example nuclei of seam and ventral epidermal cells are circled with dashed lines. Scale bars: 5 μm. F. Violin plots with overlaid notched boxplots of distributions of integrated intensities of transcription dots (channel for dlg-1 pre-mRNA) shown in (E). *** = p < 0.001. For raw data and statistics, see Table S4.

In a converse experiment, we induced actomyosin hypercontraction within developing embryonic epidermal cells using a mutation in the myosin-associated phosphatase orthologue mel-11/PPP1R12A (protein phosphatase 1 regulatory subunit 12A). Reduction of mel-11 levels via RNAi induced dlg-1 transcriptional upregulation (Fig.4E,F and S9A), indicating that J-QC is responsive to cytoskeletal disruptions, leading to transcriptional upregulation. Because mel-11-induced hypercontraction affects both the microtubule network and the actin cytoskeleton, we cannot determine which cytoskeletal network is primary^37,38 In sum, these data suggest that the cytoskeleton is an integral component of the J-QC pathway controlling dlg-1 transcription. The lack of an additive effect upon simultaneous disruption of DLG-1 and the cytoskeleton indicates that junctional and cytoskeletal integrity are monitored through the same genetic pathway.

ZYG-12 mediates the J-QC response

We next considered how the non-nuclear compartment of the cytoskeleton communicates to the nucleus. The linker of nucleoskeleton and cytoskeleton (LINC) complex transduces signals between the cytoplasm and nucleus, ultimately remodeling gene expression during cell migration and development^39–43. We therefore tested whether LINC components were required for the J-QC response.

LINC complexes consist of two interacting transmembrane proteins in the perinuclear space: a SUN (Sad-1-UNC-84) domain-containing protein of the inner nuclear membrane that connects to the nuclear lamina, and a KASH (Klarsicht, ANC-1, Syne homology) protein of the outer nuclear membrane that ultimately links the nucleus to the cytoskeleton. C. elegans possesses three SUN/KASH combinations that connect the cytoskeleton to the nucleus (Fig. 5A): SUN-1/ZYG-12 and UNC-84/UNC-83, which associate with microtubules through the motor protein dynein, and UNC-84/ANC-1, which binds actin^44,45. These complexes have been studied in the context of nuclear migration and chromosome dynamics during germline and embryonic development in C. elegans^46,47, but their roles in signal transduction and transcription remain unknown.

Figure 5. ZYG-12 is required to upregulate dlg-1 in the J-QC mechanism.

A. Simplified schematic representation of the three known LINC complexes in C. elegans. From left to right (KASH/SUN): ZYG-12/SUN-1, UNC-83/UNC-84*, and ANC-1/UNC-84. ZYG-12 (red), UNC-84* (orange with asterisk), and ANC-1 (dark blue) have been tested to cover the three respective complexes. B. Fluorescence images of epithelial cells of smFISH signal for dlg-1 mRNA (magenta) and dlg-1 pre-mRNA (cyan), DLG-1::EGFP::mIAA7 fluorescent signal (yellow), and merge. Embryos were treated with DMSO or 5-Ph-IAA-AM treatment for 30 minutes. From top to bottom: strain containing DLG-1 tagged with EGFP and mIAA7 (black vertical line on the side), treated with DMSO (DLG-1) (first raw) or with 5-Ph-IAA-AM (DLG-1(−)) (second raw); strain containing DLG-1 and ZYG-12, both tagged with green fluorophores and AID tags (red vertical line on the side), treated with DMSO (DLG-1/ZYG-12) (third raw) or with 5-Ph-IAA-AM (DLG-1(−)/ZYG-12(−)) (fourth raw); strain containing DLG-1 tagged with EGFP and mIAA7 in an unc-84 mutant background constituted by an in-frame deletion (allele e1174) affecting specifically the interaction with UNC-83 and not ANC-1⁹³ (orange vertical line on the side), treated with DMSO (unc-84;DLG-1) (fifth raw) or with 5-Ph-IAA-AM (unc-84;DLG-1(−)) (sixth raw); strain containing DLG-1 tagged with EGFP and mIAA7 in an anc-1 mutant background (dark blue vertical line on the side), treated with DMSO (anc-1;DLG-1) (seventh raw) or with 5-Ph-IAA-AM (anc-1;DLG-1(−)) (eighth raw). Example nuclei are circled with dashed lines. Arrowheads point at example nuclei with large transcription dots. Scale bars: 5 μm. C. Violin plots with overlaid notched boxplots of distributions of integrated intensities of transcription dots for dlg-1 pre-mRNAs in the different conditions. Data normalized to the control strain, DLG-1 tagged with mIAA7 and treated for 30 minutes with DMSO. n.s. = p > 0.05; *** = p < 0.001. For raw data and statistics, see Table S4. D. Violin plots with overlaid notched boxplots of distributions of integrated intensities of transcription dots for dlg-1 pre-mRNAs in the different conditions. Data normalized to the control strain, DLG-1 tagged with mIAA7 and treated for 30 minutes with DMSO. * = p < 0.05. For raw data and statistics, see Table S4. E. Live fluorescence images of epidermal cells and zoom-ins (dotted magenta squares) for DLG-1::EGFP::mIAA7 signal upon 5-Ph-IAA-AM-mediated depletion of DLG-1 alone (DLG-1(−), upper panels) or both DLG-1 and ZYG-12 (DLG-1(−)/ZYG-12(−), lower panels). The images are the same as in Fig. 5B, but with higher contrast to focus on gaps at junctions (arrows). Scale bars: 10 μm (left panels) and 5 μm (right panels).

To test if SUN/KASH complexes were required for the J-QC, we paired DLG-1 degradation (the J-QC trigger) with loss of SUN/KASH components. Specifically, we disrupted the SUN-1/ZYG-12 and UNC-84/UNC-83 complexes, which connect the nucleus to microtubules, as well as the UNC-84/ANC-1 complex, which links the nucleus to actin^44,45. To circumvent nonspecific effects by LINC component mutations, we used conditional (AID2 for zyg-12; thermo-sensitive unc-84) or loss-of-function (anc-1) alleles to test their function during morphogenesis. We found that loss of ZYG-12, but not UNC-84 or ANC-1, prevented the compensatory increase in dlg-1 transcription observed upon DLG-1 depletion (Fig. 5B-D and S10A-E). Thus, ZYG-12 is required to induce the J-QC.

Our model suggests that loss of the J-QC will enhance phenotypes where the DAC has been compromised. We performed a double depletion of DLG-1 and ZYG-12 to disrupt the DAC and inhibit the J-QC response simultaneously. As expected, double inactivation produced worse junctional defects, including reduced DLG-1 protein and gapping along the DAC compared to DLG-1 inactivation with an active J-QC system (Fig. 5E). These results suggest a specific role for ZYG-12 in the J-QC response and underscore the importance of this system in responding when the DAC is compromised.

Discussion

In this study, we identify two complementary processes that cooperate to ensure epithelial junctions remain intact. First, we show that subcellular localization of dlg-1 mRNA promotes local DLG-1 protein accumulation at junctions; loss of mRNA localization leads to reduced DLG-1 and discontinuity within the DAC (gaps). Second, we uncover a novel quality control pathway, the junctional QC (J-QC), that triggers compensatory transcriptional upregulation of DAC components when the junction is compromised, and a signal is transduced to the nucleus by ZYG-12 (Fig. 6). Phenotypically, loss of the J-QC also leads to epithelial gaps, underscoring its importance. These findings suggest that epithelial junctions maintain homeostasis through the coordinated action of local protein production and adaptive transcriptional feedback. We suggest that spatially localized mRNAs enrich key structural proteins at sites of junction assembly, while the J-QC pathway adjusts transcriptional output in response to junctional integrity rather than sequence complementarity. These dual mechanisms ensure tissue integrity in the face of junctional perturbations.

Figure 6. Model of the J-QC pathway.

Proposed pathway in which the DAC components (yellow) and the cytoskeleton (green) negatively regulate ZYG-12 (red), which in turn positively regulates the expression of DAC genes (blue). Simplified schematic representation of the pathway within the cell. Left panel: under wild-type conditions, functional junctional DAC (yellow, #1), cytoskeleton (green, #2), and ZYG-12 (red, #3) maintain a basal level of transcription of junctional components (blue thin arrow). Middle panel: loss of DAC or cytoskeletal function triggers transcriptional upregulation of DAC genes (blue thick arrow). Right panel: transcriptional upregulation of DAC genes does not occur when ZYG-12 is lost in addition to the DAC.

Conservation of DLG-1 orthologs and implications for disease

Our findings may be relevant to DLG-1 orthologues (DLG1) in other species and cell types. DLG-1 belongs to a highly conserved family of scaffolding proteins whose orthologs perform essential functions at cell-cell contacts across metazoans, including epithelial junctions⁴⁸, neuronal synapses⁴⁹, and immunological synapses in T-cells^50,51. In several tumor types, total DLG1 protein is elevated, even if it is depleted from cell-cell contacts and redistributed to the cytoplasm. In cervical cancer, for example, DLG1 protein is both increased and preferentially cytoplasmic ⁵² and in in colorectal cancer, elevated cytoplasmic DLG1 protein correlates with reduced overall survival⁵³. These patterns may reflect a J-QC–like response: cells sense the loss of junctional DLG1 and upregulate transcription in response, but when the additional protein fails to reach the junction, it drives further upregulation. High levels of cytoplasmic DLG1 correlate with poor prognosis in these cancers, suggesting this regulatory loop may be functionally important. Understanding the J-QC may offer a means to dismantle DLG1 mis-regulation in diseased states.

mRNA localization as a component of junctional homeostasis

mRNA localization is widespread across species and cell types, with hundreds to thousands of localized transcripts identified in diverse biological systems^22,54,55. In the best-understood paradigm, cis-regulatory “zip-code” elements, located in the 3’UTR, are recognized by RNA-binding proteins (RBPs) that direct the transcript to its destination^56,57. RBP-bound transcripts are transported as part of larger, condensate-like ribonucleoprotein granules along microtubules, exemplified by kinesin-dependent transport of oskar mRNA⁵⁸. This mechanism is particularly useful in situations where mRNA localization is temporally decoupled from translation, such as quiescent germ cells or highly polarized neurons. Classical studies established that maternally deposited mRNAs, such as bicoid and oskar, direct spatially restricted translation and protein localization after the onset of embryogenesis^59,60, and this concept has since expanded across cell types and developmental stages^{54,55,61–64}. Epithelia are another tissue rich in localized RNAs, but one where translation and mRNA localization are coupled temporally^22,65–68. Here, we focus on epithelial dlg-1, which relies on a co-translational mechanism for mRNA localization rather than a UTR-based zip-code²². Currently, the best studied examples of translation-dependent mRNA localization are proteins that reside in membranous organelles such as the ER or mitochondrion, where translation kinetics and the size of the protein influence mRNA localization^69,70. This pathway allows proteins to pass the membrane surrounding the organelles, but it is not binary, with a cohort of mRNA translated away from its destination. This observation raises the question of how important mRNA localization is for structures that lack a membrane. Our data with dlg-1 show that mRNA localization has a large influence on integrity of the junction, by ensuring adequate levels of DLG-1 protein.

The functional consequence of localization differs between transcripts and species. In mouse melanoma cells, Kif1c mRNA localization is dispensable for total protein abundance but required for directed cell migration, reflecting a role in cellular polarity rather than protein levels⁵⁷. Similarly, C. elegans erm-1 mRNA localization is not required for protein abundance, but rather for intestinal morphogenesis²⁵, implying that localization has roles beyond protein abundance. dlg-1 transcripts represent the converse case: mRNA re-localization reduces total junctional DLG-1 protein, indicating that localization safeguards protein abundance at the site of junction assembly, but is not required for polarity per se. The DAC is incorporated into the junction only after the CCC and core polarity regulators are already established^19,21. Therefore, junctional mRNA localization likely serves a reinforcing, structural role rather than the patterning function classically described to localized maternal transcripts. We suggest dlg-1 mRNA remains translationally competent when directed to pores²⁵, but that one or more events occur: i) the translated protein is not efficiently delivered to the junction; ii) specialized regulators selectively augment DLG-1 translation at the junction. Either of these hypotheses is consistent with the co-translational requirement for dlg-1 mRNA localization²²; iii) re-localized dlg-1 transcripts or protein may undergo rapid degradation. Selective degradation has been described for other developmentally regulated transcripts through RBP/deadenylation- or miRNA-dependent mechanisms^71–73. Thus, mRNA localization is a critical mode of regulation that ensures newly produced DLG-1 accumulates efficiently at the junction.

dlg-1 orthologs (DLG1) are also subject to subcellular mRNA localization in vertebrate neurons^74,75. This preferential subcellular localization does not depend on the DLG1 3'UTR⁷⁵, suggesting a translation-dependent mode of localization analogous to what we discovered for C. elegans dlg-1 in epithelia²². This observation suggests that translation-dependent mRNA localization, rather than RBP-mediated transport, may represent an evolutionarily conserved feature of DLG1 transcripts. Because localization occurs while translation is already in progress, both the transcript and the protein synthesis machinery are brought to the specific subcellular locus. This mechanism allows fast local production of DLG1 protein and is particularly advantageous for structures that require rapid assembly or reinforcement, such as epithelial junctions and neuronal or immunological synapses. The work we presented here suggests that vertebrate neurons may also depend on mRNA localization to establish appropriate levels of DLG1 at the synapse. Our work provides a functional explanation for the conserved localization of dlg-1 ortholog mRNAs, suggesting that they similarly support local DLG1 protein accumulation at synapses and other vertebrate cell-cell contacts.

A structural QC mechanism for epithelial junctions

Our findings identify a novel QC system, the J-QC, that acts in epithelial cells to buffer junctional disruption. With the J-QC, cells couple the detection of structural perturbations of a defined subcellular entity to transcriptional responses to restore homeostasis. The J-QC is distinct from other QC mechanisms, which assess either organelle functionality or, alternatively, molecular fidelity. In the first instance, QC systems surveille membranous organelles such as mitochondria⁸, chloroplasts¹² and the endoplasmic reticulum¹³ (ER). These systems commonly monitor the accumulation of damaged or misfolded proteins and restore organelle homeostasis through selective degradation mechanisms ^76–78. In the second instance, systems monitor molecular processes such as DNA repair⁷⁹, nonsense-mediated mRNA decay⁸⁰ (NMD), transcriptional adaptation^33,34, or the unfolded protein response⁸¹ (UPR). Transcriptional adaptation, for example, responds to NMD-mediated mRNA degradation by compensatory transcriptional induction of paralogous genes³⁵. The importance of QC pathways is revealed by their wide ability to allow animals to survive compromised states, ranging from neurodegenerative disorders (e.g., Alzheimer’s disease) to disruptions in metabolism or protein-folding (e.g., type 2 diabetes and cystic fibrosis)^35,82–85. Conceptually, the J-QC expands the scope of QC systems from monitoring proteins, RNAs, or membranous organelles to the maintenance of cellular architecture within epithelia.

A key feature of the J-QC is its remarkable specificity. Both trigger and response are restricted to the DAC: perturbation of DLG-1 or AJM-1 activates transcription of DAC genes, whereas disruption of the adjacent CCC or polarity regulators does not. This selectivity argues against a general stress response^86,87 and instead suggests that epithelial cells distinguish between different junctional modules. Within the context of an organism, individual epithelial cell subtypes (e.g., the epidermis) may have evolved additional mechanisms to restrict transcriptional responses to specific genes, rather than activating a broad stress program. The precise transcriptional response sets the J-QC apart from other described forms of junctional surveillance, which act at the protein level but without altering gene expression: loss of tension across the CCC in mammalian models, for instance, triggers a conformational change in α-catenin and local recruitment of vinculin to reinforce the junction^88,89. The J-QC instead couples a structural defect in one specific junctional module to a dedicated transcriptional program, revealing a module-selective layer of surveillance. Such specificity may allow epithelial cells to preserve tissue integrity while avoiding broad stress responses that could alter other aspects of development. More generally, our findings suggest that QC mechanisms can act not only on damaged proteins or organelles, but also on the composition of functional cellular structures whose precise organization is required during morphogenesis.

One possible explanation for the distinction between the DAC and the CCC is that the two junctional domains fulfill distinct roles. The cadherin-based CCC is established early during embryogenesis and provides the adhesive framework that maintains epithelial cohesion, whereas the Discs large-based DAC is built later, coinciding with the onset of morphogenesis and body elongation^19,21,32. During this developmental transition, epithelial junctions experience continuous remodeling and variable mechanical forces^32,90. A dedicated QC mechanism acting on the DAC could therefore represent a mechanism to reinforce junctions and maintain tissue integrity under variable mechanical load.

The J-QC operates independently of previously described RNA-mediated transcriptional compensatory pathways. The closest system is perhaps transcriptional adaptation in which mutated mRNAs induce transcriptional upregulation of paralogous genes^33,34. However, J-QC is distinct from transcriptional adaptation since it does not depend on sequence similarity between the trigger and the responder, and it does not rely on nonsense-mediated decay components or premature translational termination. J-QC is also distinct from promoter-directed RNA activation (RNAa), which relies on transcriptional activation mediated by small RNA complementary to promoter regions^91,92. Together, these observations indicate that the J-QC is not driven by RNA sequence or RNA metabolism but is a new type of surveillance system that monitors the state of the epithelial junction itself.

The LINC component ZYG-12 mediates the transcriptional response in the J-QC pathway

The LINC complex was first genetically defined in C. elegans through the identification of SUN- and KASH-domain proteins required for nuclear anchorage and migration^93–95. These studies established the LINC complex as a conserved functional unit that physically connects the cytoskeleton to the nucleus³⁹. In vertebrate models, LINC complexes have been implicated in force transmission, particularly in nuclear positioning and mechanotransduction⁸⁷. Actin-associated LINC complexes, such as nesprin-1/2 coupled to SUN2, participate in mechanotransduction pathways that regulate gene expression in response to changes in cellular tension and extracellular environment^96,97. By contrast, microtubule- and dynein-associated LINC modules have primarily been linked to nuclear and chromosome positioning⁹⁸, including the C. elegans KASH-domain protein ZYG-12^99–102. However, whether distinct LINC modules can transmit specific information about cellular architecture to the nucleus to regulate transcription remains unclear. Our findings identify a previously unrecognized function for ZYG-12 in coupling epidermal junction defects to transcriptional upregulation of DAC components in the J-QC. Rather than representing a generic consequence of disrupting nuclear–cytoskeletal connections, this response requires a specific LINC module, suggesting that distinct LINC assemblies may have specialized roles in linking cellular architecture to nuclear responses.

ZYG-12 is a conserved member of the HOOK family of dynein adaptors. It combines a HOOK domain predicted to bind dynein light chain¹⁰³, with a KASH domain that allows it to span the outer nuclear membrane and bind other components of the LINC complex. Together, these features enable ZYG-12 to couple microtubule-based transport through dynein to the nucleoplasm. Consistent with a role for dynein-dependent transport in this pathway, we showed that altering the function of dynein heavy chain (DHC-1) also affects the transcriptional response in the J-QC pathway, further supporting a functional connection between ZYG-12 and dynein-dependent transport processes. Interestingly, disruption of the actin cytoskeleton also activated the J-QC in addition to the microtubule- and dynein-associated LINC module. One possible explanation is that perturbation of the actin network indirectly alters microtubule organization. Indeed, actin and microtubules are extensively interconnected, and changes in actomyosin organization or contractility have been shown to remodel microtubule architecture¹⁰⁴. Stil, we cannot completely exclude the possibility that both cytoskeletal systems may influence the J-QC directly.

Our analyses suggest a genetic pathway in which junctional DAC components and the cytoskeleton negatively regulate ZYG-12, which in turn positively regulates the expression of DAC genes (Fig. 6). Genetically, this model is supported by our finding that ZYG-12 is required for the transcriptional upregulation of DAC genes following disruption of either the DAC or the cytoskeleton, placing ZYG-12 downstream of these structural components and upstream of the transcriptional response. At the molecular level, our data are consistent with a model in which the integrity of the DAC and cytoskeleton is continuously monitored through a ZYG-12-dependent LINC module. Perturbation of this structural connection would then alter the signaling state of the pathway, leading to compensatory transcription of DAC genes. This model provides a unifying explanation for why diverse perturbations affecting junction integrity or cytoskeletal organization converge on the same transcriptional response, whereas disruption of adjacent junctional modules, such as the CCC, does not. We propose that the J-QC pathway functions to restore physiological levels of DAC components and thereby support epithelial biogenesis and embryonic elongation¹⁰⁵.

Methodology development

Our experiments relied on proven MS2/MCP methodology to re-localize dlg-1 mRNA. Fusing MCP with nuclear pore components provides an alternative to RBP mutation or deletion strategies, which can induce pleiotropic effects unrelated to localization itself. Our approach is conceptually related to CRISPR-based tools such as CRISPR-TO¹⁰⁶, CRISPR-TO technology uses dCas13 to reposition transcripts, which requires suitable guide sequences within the 3'UTR and assembling a multi-component complex that can interfere with gene expression independently of relocalization¹⁰⁶. Our system is simpler as it requires only MS2 hairpins and a fused MCP protein, which can be targeted to any subcellular structure for which a suitable anchor protein exists. This system does not depend on a pre-existing localization signal or its cognate RBP, representing an advantage for transcripts like dlg-1 that lack one. Importantly, re-localization was achieved directly in a live, developing organism rather than in cultured cells, making it possible to link mRNA localization directly to epithelial morphogenesis in the context of embryonic development rather than only to cellular phenotypes.

Quantitative analysis of transcription sites remains a challenge, particularly when comparing transcriptional activity across multiple genetic or experimental conditions. Although several tools have been developed for RNA spot detection and quantification^108,110, DotQuant (https://github.com/angelo-angonezi/DotQuant) provides a streamlined workflow that integrates automated nuclear and three-dimensional spot segmentation and fluorescence measurements to enable rapid and reproducible measurement of transcription site intensity across large numbers of nuclei. While developed here for the quantitation of transcription sites in C. elegans embryos, the modular design of DotQuant can be adapted to other types of fluorescent puncta and imaging datasets from different systems. Furthermore, future incorporation of automated nuclear segmentation based on nuclear markers will extend its applicability by enabling fully automated selection and analysis of nuclei across different tissues and model systems.

Methods

Nematode culture

All animal strains were maintained as previously described (Brenner, 1974) at 20°C, unless stated otherwise. For experiments with temperature-sensitive strains, animals were grown at 15°C and shifted to the temperatures and times described in the corresponding figure legends and/or methods. RNAi experiments were performed as previously described except for the growing temperature which was set to 24°C, instead of 25°C, and starting from a synchronized population of molting L3-L4 animals left to grow until adulthood and lay eggs. Embryos were collected for further processing. For a comprehensive list of alleles, transgenic lines, and RNAi clones¹⁰⁷, see Table S1.

SUN/KASH alleles description

To perturb SUN-1/ZYG-12 function during morphogenesis, we used a strain carrying AID- and GFP-tagged ZYG-12 and induced degradation as previously described⁹⁹, to bypass the early embryonic lethality of sun-1 and zyg-12 mutants. To perturb the UNC-84/UNC-83 complex, we used the temperature-sensitive unc-84(n369ts) allele (marked with an asterisk for clarity, unc-84(ts)*), which contains an in-frame deletion that specifically disrupts UNC-84 interaction with UNC-83 while preserving interaction with ANC-1⁹³. To perturb the UNC-84/ANC-1 complex, we used the anc-1(e1873) mutant strain⁹⁵. These perturbations were combined with the AID2 system for DLG-1::EGFP::mIAA7 degradation (to induce the J-QC) to analyze the J-QC response.

Generation of transgenic lines

The transgenic strain where MCP was paired to NPP-9 was generated with the mosSCI technique¹⁰⁹. Insertions were verified by genotyping, sequencing, and phenotypic readout. The transgenic sequences (promoter: pes-10, coding sequence: 3xFLAG fused to two MCP sequences, two mCherry sequences²³, and the endogenous npp-9 sequence (exons and introns); 3’UTR: tbb-2) were assembled with the NEBuilder^® HiFi DNA Assembly Cloning Kit (New England BioLabs, cat#E5520), using pCFJ150 as a vector to insert the transgene into the ttTi5605 locus on chromosome II (strain name: EG6699). The plasmid containing the transgene is called pCT3.82 and is available upon request. NEB^® 10-beta Competent E. coli (High Efficiency) (NEB, cat#C3019) competent cells were used for transformation. Full plasmid sequencing from Microsynth was used to confirm the sequences were correct. Plasmids were purified with the NucleoBond Xtra Midi Plus (Macherey-Nagel, cat#740412.10) and single-copy integrated lines were generated and verified following the mosSCI protocol¹⁰⁹.

The dlg-1(Δc7aa) embryos strain was generated with the CRISPR/Cas9 technique¹¹¹ and consisted of a deletion of seven amino acids in the sequence linking SH3 and GuK domains of DLG-1 (aa:727–733). Two CRISPR RNAs (crRNA1: GAATGGGAACTTTCTGGAGA; crRNA2: CGGTCGGTACTTTTGACGAA; IDT), a single-stranded oligodeoxynucleotide as a repair oligo (ssODN, TGGAAAACCGAAGAGGAAGCCGCAGTCAGCTTTCTGTCAAAAGTACCGACCGGCTCAACGATCTTAATGA; IDT), Cas9 enzyme (Alt-R^® S.p. Cas9 Nuclease V3, IDT, cat#1081058), and tracrRNA (IDT, cat#1072533) were used to generate the deletion strain on an already available allele where DLG-1 protein was tagged with mNG (LP598²⁸). Genotyping was used to screen rollers, and further sequencing to verify correct deletion (Fw oligo: TGGATTGGAAAACCGAAGAG; Rv oligo: TGCTCAACTGCCTGGTAGG).

smFISH

smFISH primary probes design was performed as previously described^22,112. smFISH experiments were adapted from previously described protocols^22,112. Adults from plates with a lot of laid embryos were removed by washing them off plates with distilled water (two times with 1 ml). The embryos and the leftover bacteria lawn were collected after adding 1 ml water and mechanically released by gentle rubbing with a gloved finger. Embryo suspensions were transferred to 1.5-ml microcentrifuge tubes and washed two times with distilled water by brief centrifugation (5 seconds, short pulse) to remove remaining worms, bacteria, and debris. Glass slides (Epredia^™ Epoxy Diagnostic Slides, Fisher Scientific, cat#10393881) were coated with 1 μl poly-L-lysine and air-dried prior to use. Approximately 50 μl of embryo suspension was applied to each slide and allowed to settle for approximately 30 seconds. Excess liquid was removed, and embryos were immediately treated with fixative solution by adding 50 μl of freshly prepared 3.7% formaldehyde in PEM buffer for 5 minutes at room temperature in a humidity chamber. After removal of the fixative (though ensuring embryos did not dry), a small drop (2–3 μl) of the removed fixative was placed on a coverslip, which in turn was immediately applied to the slide. Slides were transferred onto a metal plate placed on dry ice and frozen for at least 30 minutes. Coverslips were removed by freeze cracking. Embryos were transferred in different coplin jars containing the following reagents: ice-cold methanol for 5 minutes, PBS for 1 minute at room temperature, 3.7% formaldehyde in PEM buffer for 5 minutes at room temperature in a second fixation step. Samples were briefly washed at room temperature in coplin jars as follows: 1 minute in PBS, 10 minutes in PBS-T, 20 minutes in PBS-T, and 5 minutes in PBS. Slides were dried but the well containing the embryos, which were with 100 μl hybridization solution for 1 hour at 37°C in a humidified chamber. Afterward, the blocking solution was removed and replaced with 100 μl hybridization solution containing smFISH probe(s) (2 μl per 100 μl hybridization solution for each FLAP duplex probe set, or 1 μl for Stellaris probes). Samples were incubated for at least 4 hours or overnight at 37°C in a dark, humidified chamber. Primary antibody against AJM-1 (MH27¹¹³, Developmental Studies Hybridoma Bank – DSHB) was added during this step when required. Following hybridization, samples were washed twice with wash buffer (WB) – 100 μl per wash by pipetting up and down on the well containing the sample. Where applicable, secondary antibodies (Goat anti-Mouse IgG (H+L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor^™ 488; Invitrogen, cat#A-11029) or nanobodies (GFP-Booster ATTO488, Chromotek, cat#gba488; RFP-Booster ATTO594, Chromotek, cat#rba594) diluted in WB were added and incubated for 1 hour at 37°C in the dark. Nuclei were stained with Hoechst (1:50 in WB) for 15 minutes at room temperature in a humidity chamber in the dark, followed by three washes in WB − 100 μl per wash by pipetting up and down on the well containing the sample. Samples were mounted in 12 μl of VECTASHIELD^® Antifade Mounting Medium (Vector Laboratories; cat#H-1000–10) and sealed with nail polish prior to imaging and stored at 4°C in the dark.

Composition of buffers and solutions: Fixative solution (3.7% formaldehyde in 1xPEM buffer): formaldehyde solutions were prepared fresh from 37% stock. For on-slide fixation, 50 μl 37% formaldehyde was mixed with 100 μl 5xPEM buffer and 350 μl H₂O. For the second fixation in a coplin jar, 5 ml of 37% formaldehyde was mixed with 10 ml 5xPEM buffer and 35 ml H₂O. 5xPEM buffer (pH 6.8): 500 mM PIPES, 5 mM EGTA, 5 mM MgCl₂. PBS-T: PBS with 0.1% (v/v) of Tween-20. Wash buffer (WB): 10% (v/v) of formamide and 2xSSC in H₂O (usually prepared 10 ml and stored at 4°C). Hybridization solution: 10% (w/v) of dextran sulfate, 10% (v/v) of formamide, and 2xSSC in H₂O; always prepared fresh.

For dlg-1, ajm-1, hrm-1, hmp-1, par-6, and pkc-3 mRNAs and dlg-1 pre-mRNA we used probes previously described^22,23. For the sequences of the newly made smFISH probes (ajm-1 pre-mRNA), see Table S2.

Microscopy

Images were acquired on embryos at the same stage (comma stage, at the onset of morphogenesis) and cell type (epidermal cells - dorsal-ventral and seam) throughout the manuscript for consistency.

For live imaging experiments, embryos were treated as previously described²². A widefield ZEISS Axio Imager M2 equipped with a ZEISS Colibri as LED light source, Hamamatsu ORCA flash 4.0 camera, and ZEN 2.6 software (blue edition) were used for capturing images for live imaging experiments. All images were processed in OMERO (https://www.openmicroscopy.org/omero/) and processed in the same manner within datasets. Figures were assembled in Adobe Illustrator (https://www.adobe.com/).

For imaging of fixed samples, in figures 1B and 2E a widefield microscope FEI “MORE” with total internal reflection fluorescence (TIRF), equipped with a Hamamatsu ORCA flash 4.0 cooled sCMOS camera, and a Live Acquisition 2.5 software were used for capturing images. smFISH pictures were then deconvolved with the Huygens software. In the rest of the figures, we used a Nikon Ti2 with perfect focus system (PFS), equipped with a CSU-W1 Yokogawa Spinning Disk Field Scanning Confocal System, a Hamamatsu ORCA-Fusion sCMOS camera, and a NIS elements with JOBS software to acquire images. All images were processed in OMERO (https://www.openmicroscopy.org/omero/), and figures were assembled in Adobe Illustrator (https://www.adobe.com/).

Image analysis and quantitation

Quantitation of fluorescence signal from the junctions via live imaging: fluorescently labelled DLG-1 protein was quantified by analyzing the intensity (max) of five junctions connecting seam cells per embryo and removing the background signal (min). Fluorescent signal quantitation was performed in Fiji.

Quantitation of integrated intensities of transcription dots in smFISH experiments were performed via DotQuant (https://github.com/angelo-angonezi/DotQuant). We selected nuclei of interest following two approaches. In a cellpose-based approach, nuclei were segmented using Cellpose-SAM (https://doi.org/10.1101/2025.04.28.651001). Nuclei of interest were manually selected in Napari by identifying and recording the corresponding segmentation object IDs (https://napari.org/stable/) (Fig. S3A-B). Single-nucleus image crops were generated from the raw input images using the spatial coordinates defined by the selected segmentation masks. In a second, ROI-based approach regions of interest (ROIs) were manually defined around nuclei locations in Fiji using the ROI Manager tool (https://imagej.net/software/fiji/) (Fig. S3C-F). For each image, ROI coordinates were exported as a .csv file describing a three-dimensional bounding box specified by five parameters: center coordinates (X, Y, Z), width, and height (Fig. S3G). Single-nucleus image crops were generated using a custom Python script that integrated the raw image data with the corresponding ROI coordinate files. For each ROI, a fixed axial depth of 10 Z-slices (±5 slices from the ROI Z-center) was used (Fig. S3H). Cropped nuclei were post-processed using a Difference of Gaussians (DoG) filter to enhance small, high-intensity puncta. FISH signals were segmented by intensity thresholding of the DoG-filtered images (a different set of thresholds was applied to pre-mRNA and mRNA per dataset due to varying intensities between channels). Individual dot volumes were calculated from the voxel-based binary masks by converting each segmented object into a surface mesh and computing its volume using the Python trimesh library (https://pypi.org/project/trimesh/). In nuclei containing more than two detected dots, only the two largest objects (by volume) were retained for downstream analysis. To maintain a consistent dot count per nucleus, one or two empty dots (volume = 0.0, intensity = 0.0) were added to nuclei containing fewer than two detected dots. For intensity quantification, the segmentation masks obtained in the previous step were applied to the original images to extract the pixels corresponding to each individual dot. The mean intensity was then calculated for each dot. Background intensities used for normalization were calculated as the mean intensity within each single-nucleus crop, excluding pixels belonging to the dot masks. Integrated intensity was defined as the product of each dot rescaled mean intensity and the corresponding mesh volume.

Statistical analyses and data visualization

Data analysis and visualization were performed in R version 4.5.1 (https://www.R-project.org/) using the packages ggplot2 (https://ggplot2.tidyverse.org) and dplyr (https://dplyr.tidyverse.org). The Shapiro–Wilk test was used to check normality of the data and verified with normality diagnostic plots for such a test. Homogeneity of variances was evaluated using Levene’s test. For normally distributed data with equal variances, a one-way ANOVA was performed. For normally distributed data with unequal variances a pair-wise t-tests with Holm correction for multiple comparisons was used (Fig. 1D,G, 2B,E). For non-normally distributed data, group differences were assessed using the Kruskal-Wallis test. When the Kruskal-Wallis test was significant, it was followed by pairwise Wilcoxon rank-sum tests with continuity correction (P value adjustment Benjamini-Hochberg for multiple testing correction) to test pairs (Fig. 2A,G, 3C,E,F, 4B,D,F, 5C,D). Bar plots were used for normally distributed data: the bar represents the median, whiskers the standard deviation (SD), diamonds the mean, and black dots individual data points. Violin plots with overlaid notched box plots were used for non-normally distributed data: median (central thick line), interquartile range (IQR, box), and approximate 95% confidence interval (whiskers). Notches represent the confidence interval (CI) around the median. Magenta diamonds indicate the mean, and dots represent outliers. Statistical significance is indicated above bars or notched boxes as specified in each figure legend. For raw data, sample size (n), mean, median, and SD from each experiment, see Table S3 (live samples) and Table S4 (fixed samples).

mRNA re-localization

The dlg-1 mRNA re-localization experiments were performed by adapting the MS2-MCP technique. Specifically, dlg-1 mRNA with or without MS2 hairpins in its 3’UTR was paired to MCP fused or not to the nuclear pore protein NPP-9/NUP358. The four transgenic combinations were all in an NMD-deficient background (smg-2/UPF1, allele e2008) to prevent degradation of dlg-1 mRNA containing MS2 hairpins²³. NPP-9 is on the cytoplasmic surface of the nuclear pore, and dlg-1 mRNAs containing MS2 hairpins were trapped in this subcellular location by the MCP::NPP-9 protein fusion upon exiting the nucleus. In this way, dlg-1 mRNA could not reach its normal location at the apical junction. Experiments on the four strains were performed at 24°C to allow basal expression of MCP from the heat-shock promoter as previously described²³.

Protein degradation experiments with AID2 system

The DLG-1 degradation experiments were performed following the AID2 system that takes advantage of a modified TIR1 enzyme from Arabidopsis thaliana (AtTIR1(F79G)) and the eggshell-permeable 5-phenyl-indole-3-acetic acid acetoxymethyl ester (5-Ph-IAA-AM), a modified synthetic analog of the auxin hormone³⁰. Laid embryos from the C. elegans strain containing the TIR1 enzyme and mIAA7-EGFP-tagged DLG-1³¹ were collected in an Eppendorf tube and resuspended in M9. Some embryos from this sample were used for live imaging and fluorescence quantitation analyses. The remaining sample was then split in half into two separate Eppendorf tubes: DMSO (Dimethyl sulfoxide, Sigma Aldrich, cat#D4540) was added to the control sample, and 5-Ph-IAA-AM (Tocris Bioscience^™ 5-Ph-IAA-AM, Fischer Scientific, cat#18786115) (resuspended in DMSO) was added to the treated sample for a final concentration of 50 μM. The samples were placed on a rotating wheel and one third of embryos from both samples (DMSO and 5-Ph-IAA-AM) were collected at 15, 30, and 45 minutes for the time course experiment. These pools of embryos were then split and used for both live imaging for fluorescence quantitation analyses, and to prepare slides to be processed for smFISH experiments. In the other AID2 experiments, including the double DLG-1/ZYG-12 degradation, embryos were collected after a 30-minute treatment.

Real-time quantitative PCR (RT-qPCR)

Total RNA was isolated from pelleted embryos using the RNeasy Plus Mini Kit (Qiagen, cat#74134) following the manufacturer’s protocol. Residual genomic DNA was removed by treatment with DNase I (Thermo Fisher Scientific, cat#18060–015). RNA concentration and purity were assessed spectrophotometrically. First-strand cDNA synthesis was performed using the SuperScript^™ IV First-Strand Synthesis System (Thermo Fisher Scientific, cat#18091050) with random hexamer primers, using DNase-treated RNA as input. cDNA samples were diluted prior to quantitative PCR analysis. Real-time quantitative PCRs were performed using PowerUp^™ SYBR^™ Green Master Mix (Applied Biosystems, cat#A25742). Reactions were performed using 5 ng of cDNA per reaction and gene-specific primers at a final concentration of 500 nM in a final reaction volume of 10 μL. Amplification was performed on a real-time PCR system under standard cycling conditions, followed by melt-curve analysis to confirm amplification specificity. No-template controls were included in all experiments. Relative gene expression levels were determined using the comparative Ct (ΔΔCt) method. Ct values for target genes were normalized to those of a reference housekeeping gene (ama-1) to obtain ΔCt values. ΔΔCt values were calculated relative to the control condition, and fold changes in gene expression are shown in Fig. S7E. For reagents and raw data from the RT-qPCR experiments, see Table S5.

Glossary

AID

auxin-inducible degron

CCC

cadherin-catenin complex

DAC

DLG-1-AJM-1 complex

MCP

MS2 coat protein

NMD

nonsense-mediated mRNA decay

NPP

nuclear pore protein

QC

quality control

J-QC

junctional quality control

RT-qPCR

real-time quantitative PCR

smFISH

single molecule fluorescent in situ hybridization