Arc controls organ architecture through modulation of Crb and MyoII

Epithelial tubular organs are essential for viability in all higher-order multicellular organisms. These organs, many of which are generated from polarized epithelial sheets, transport and exchange nutrients, wastes, and gases, and produce and secrete enzymes and hormones. Several life-threatening congenital conditions, such as esophageal atresia, pulmonary hypoplasia, and multiple forms of anorectal malformation, result from abnormal tube formation in the corresponding organs (Ioannides et al., 2010; Schittny, 2017; Wood and Levitt, 2018). Furthermore, around 90% of cancers are derived from epithelial tissues (Birchmeier et al., 1996; Hinck and Näthke, 2014). Therefore, investigating the molecular and cellular mechanisms underlying morphogenetic and homeostatic processes in epithelial tubular organs not only promotes our understanding of the biology of these essential organs but also carries crucial clinical implications.

Multiple model systems in various organisms have been used to study epithelial tubular organ development (Andrew and Ewald, 2010; Chung and Andrew, 2008; Lubarsky and Krasnow, 2003). Among them, salivary gland (SG) morphogenesis during Drosophila embryogenesis provides an excellent platform for uncovering the molecular mechanisms and cellular processes required for epithelial tube formation. SG cells are specified as a pair of two-dimensional sheets (SG placodes) located on the ventral side of the posterior embryonic head region flanking the ventral midline (Fig. 1, A and B) (Chung et al., 2014; Girdler and Röper, 2014). Formation of tubes from these placodes is initiated by invagination of a small subset of SG cells at dorsal posterior positions within the placodes (the invagination pit) (Fig. 1, I–N), followed by the consecutive and orderly internalization of the remaining precursor cells to generate completely internalized, unbranched, elongated epithelial tubes (Fig. 1, C–H) (Myat and Andrew, 2000b). During this process, SG cells undergo dynamic changes in position and shape, such as cell intercalation and apical constriction, which facilitate movement toward and internalization through the invagination pit (Chung et al., 2017; Myat and Andrew, 2000b; Sanchez-Corrales et al., 2018).

A key molecular player driving SG morphogenesis is non-muscle MyoII, a motor protein that exerts contractile force on actin filaments (actomyosin complex). As in other epithelial tissues forming tubular organs (Leptin, 1991; Sweeton et al., 1991), MyoII exhibits characteristic accumulation patterns on the apicomedial plane (apicomedial MyoII) as well as at cell–cell junctions (junctional MyoII) in SG cells (Röper, 2012). At the outer boundary of the SG placode, a contiguous myosin cable connected through multiple cells (supracellular MyoII cable) encompasses the entire placode (Röper, 2012). Several studies from our group and others have shed light on the roles of each MyoII structure during SG invagination (Booth et al., 2014; Chung et al., 2017; Sanchez-Corrales et al., 2018). For example, apicomedial MyoII drives apical constriction, which transforms the columnar SG cells into the wedge-shaped cells that facilitate internalization (Booth et al., 2014; Chung et al., 2017). Junctional MyoII promotes junctional shrinkage and extension during neighbor exchange between epithelial cells (Bertet et al., 2004; Curran et al., 2017; Levayer and Lecuit, 2013), including those of the SG (Sanchez-Corrales et al., 2018). A recent study proposes that the supracellular MyoII cable functions as a mechanical insulator, protecting neighboring tissues from the morphogenetic changes in the SG placode (Ashour et al., 2023), although this structure could also contribute to the forces driving SG internalization or stabilizing the overall apical area occupied by the SG.

Despite the central roles of MyoII in SG invagination, our current understanding of how MyoII activity and localization are regulated in SG cells is incomplete. Whereas G protein–coupled receptor (GPCR)-mediated activation of the Rho1 GTPase-Rho kinase (Rok) has been considered as a direct upstream signal to activate MyoII (Chung et al., 2017), how MyoII activity is spatially restricted and temporally coordinated for efficient cell invagination in the SG placode is largely unknown. A potential candidate that plays a role in tuning MyoII activity is the transmembrane protein Crumbs (Crb). Both crb mRNA and protein are elevated in the SG placode in comparison with surrounding epithelial cells, implying its active role in SG invagination (Kerman et al., 2008; Röper, 2012; Chung et al., 2017). Crb is well known for its functions in establishing apicobasal polarity and mediating cell–cell adhesion through homophilic interactions of its extracellular domain. The short cytoplasmic domain of Crb contains functional motifs such as a FERM-binding motif and a PDZ-binding motif to recruit downstream effectors and transduce signals (Pocha and Knust, 2013; Thompson et al., 2013). Crb has been shown to either increase or suppress MyoII activity in different cellular contexts. In ectodermal cell development and neuroblast (NB) ingression, Crb recruits a Rho GTPase guanine exchange factor Cysts to activate Rho1 and eventually MyoII (Silver et al., 2019; Simões et al., 2022). In amnioserosal cells, Crb represses MyoII activity to stabilize adherens junctions (Flores-Benitez and Knust, 2015). At the border of the SG placode, anisotropy of Crb between SG cells and surrounding epidermal cells induces the formation of the supracellular MyoII cable through accelerated membrane dissociation of Rok by Crb (Sidor et al., 2020). Whereas these examples demonstrate the complicated relationship between Crb and MyoII, it is evident that Crb is a salient regulator of MyoII function in developing epithelial tissues. Therefore, how Crb affects MyoII activity during SG invagination and how Crb activity is regulated in SG cells are pivotal questions to understand fundamental principles underlying epithelial tubular morphogenesis.

In this study, we report that the PDZ domain–containing protein, Arc, affects overall tube dimensions by regulating Crb during SG invagination. Loss of arc results in shorter, stubbier SG tubes, whereas excessive Arc dramatically elongates SG tubes. Arc colocalizes with Crb at the apical junctions (AJs) in SG cells and maintains optimal Crb protein level in the SG placode by facilitating Crb delivery to the junctional membrane. Crb reduction in arc mutant SG cells abnormally increases MyoII activation, and overexpression of Arc or Crb disrupts MyoII accumulation, indicating that Arc and Crb play an inhibitory role in MyoII activation during SG invagination.

Formation of the Drosophila SG is governed by multiple early expressed transcription factors (Chung and Andrew, 2014). Among these factors, the winged-helix FoxA family protein Fork head (Fkh) plays a major role. Fkh expression begins at the earliest stages of SG tube formation (Weigel et al., 1989a) and is sustained through tube migration and elongation (Weigel et al., 1989b). Indeed, fkh continues to be expressed throughout the lifespan of the larval gland (Cao et al., 2007) and is also expressed in the adult SG (FlyAtlas, Anatomical Expression Data). SG cells mutant for fkh fail to invaginate and remain on the embryo surface (Myat and Andrew, 2000a), suggesting that Fkh regulates expression of genes controlling the transformation of the two-dimensional SG placode into an internalized, elongated, three-dimensional epithelial tube.

Through whole-mount in situ hybridization and microarray screens to discover SG Fkh targets (Maruyama et al., 2011), we identified arc (CG6741; abbreviated as “a” in Flybase), a gene originally discovered based on the “arcing” phenotype of the adult wings in mutants (Bridges and Morgan, 1919). arc is expressed in multiple epithelia, including low-level expression in the embryonic epidermis with elevated expression in the embryonic SG, foregut, hindgut (HG), and Malpighian tubules, and in both the eye and wing imaginal discs (Liu and Lengyel, 2000). arc mRNA is expressed in the SG from the earliest stages of SG invagination through later stages of tube elongation (Fig. 2, A and B). arc expression in the SG, HG, and Malpighian tubules is lost in fkh mutant embryos (Liu and Lengyel, 2000) (Fig. 2, C and D), indicating that arc is a downstream transcriptional target of Fkh in multiple embryonic tubular epithelia. Indeed, multiple potential Fkh-binding motifs are present in the enhancer region of the arc gene (Fig. S1, A–C) (Abrams et al., 2006; Takiya et al., 2003).

To begin to study the role of arc in SG morphogenesis, we generated arc null alleles by replacing most of the coding region of arc with the white⁺ eye color gene by homologous recombination (hereafter referred to as arc^KO; Fig. S1, D and E). arc^KO flies are viable and demonstrate the same “arc” wing defect as reported with the original arc mutant (Bridges and Morgan, 1919) and with a P-element insertion allele (arc^k11011b) (Fig. 2, E and F) (Liu and Lengyel, 2000). arc mRNA was not detected in either of the two arc^KO mutant lines we generated (Fig. S1 F), confirming that both arc^KO alleles are genetically null. Since the two KO alleles are molecularly identical, all subsequent analyses were carried out with arc^KO788.

In arc^KO mutant embryos (hereafter, arc^KO indicates both maternal and zygotic loss of arc), SGs fully invaginated, migrated, and extended to form tubes (Fig. 2, G and H); the length of SGs in arc null embryos (as measured by Crb-stained lumen length), however, was significantly shorter than in WT embryos (avg. SG lumen length WT 77.8 µm versus arc^KO 68.4 µm; Fig. 2, G, H, and J). Correspondingly, overexpression of Arc in the SG (fkh-Gal4–driven GFP-tagged Arc expression) resulted in longer SG tubes (avg. length WT 77.8 µm versus fkh>GFP-Arc 94.19 µm; Fig. 2, G, I, and J). The changes in SG length with loss and overexpression of arc are not linked to changes in overall embryo length; WT, arc^KO, and fkh>GFP-arc embryos were approximately the same length (avg. WT = 415.1 µm; arc^KO = 415.9 µm; fkh>GFP-arc = 410.9 µm; Fig. 2 K). The SG shortening with loss of arc and the lengthening observed with excessive Arc were consistently observed through all embryonic developmental stages (Fig. S2, A and B). Furthermore, SG lumens were wider in arc^KO than in WT during invagination (Fig. S2, C–E). Taken together, these observations suggest that Arc affects SG tube dimensions from very early stages.

Once specified, WT SG cells cease proliferation and enter the process of polytenization (DNA replication without strand separation or nuclear or cell division) (Smith and Orr-Weaver, 1991). Likewise, there is no SG cell death during the entirety of embryogenesis and larval life in WT embryos (Cao et al., 2007; Myat and Andrew, 2000a). Thus, how the SG cells are arranged could impact overall tube dimensions. Shorter SG tubes may have more cells in circumference, whereas longer tubes may have fewer. To ask if this is the case in arc mutants, we counted the number of cells in cross sections of late embryonic SG tubes. Indeed, the shorter SGs in arc^KO embryos had more cells (avg. 9.08) in circumference than those in WT embryos (avg. 7.21) (Fig. 2, L, M, and O), whereas the longer SGs observed with overexpression of GFP-Arc had fewer cells in circumference than WT (avg. 5.63) (Fig. 2, N and O). Since the overall cell numbers in the SGs of each genotype were approximately same (Fig. 2 P), we conclude that the morphological changes observed with perturbations in arc expression result from defects in cell arrangement. Similar defects in tube dimensions were observed with arc loss in the HG, another epithelial tubular organ that expresses arc. As with the SG, HG length is significantly decreased with loss of arc (avg. HG lumen length WT 135.1 µm versus arc^KO 116.3 µm; Fig. S2, F–H). Taken together, these findings indicate that arc plays a key role in controlling the shape of epithelial organs during embryonic development by affecting cell arrangement.

To investigate how Arc affects cell arrangement during SG morphogenesis, we probed the localization of the endogenous Arc protein. Immunostaining with anti-Arc antibody combined with super-resolution confocal imaging revealed that Arc protein localizes to AJs and to punctate structures within the SG cells (Fig. 3, A and B). Interestingly, junctional Arc often colocalized with junctional Crb, especially in regions of intense staining (Fig. 3, A–B‴). The two proteins do not colocalize in the surrounding epidermal cells, where Crb has a tight apical localization and Arc localizes along the lateral membrane (Fig. 3, B–B‴, arrow). The specificity of the anti-Arc antiserum was confirmed by the absence of Arc staining in arc^KO embryos (Fig. 3, C–D″). The short cytoplasmic domain of Crb contains an essential PDZ domain–binding motif (PBM) (Klose et al., 2013). The large Arc protein isoform expressed during embryogenesis (1329 amino acid residues) contains two PDZ domains in its C-terminal half, allowing for potential direct interaction between Crb and Arc (Fig. 3 E). To determine if one or the other PDZ domain of Arc is required for colocalization with Crb, we co-expressed Crb and different PDZ-deleted versions of Arc in cultured Drosophila Schneider 2 (S2) cells (Fig. 3, F–I″). Crb was occasionally localized to sites of contact between S2 cells, which allowed us to examine localization of the co-expressed Arc proteins. We found that the first PDZ domain (PDZ1) is required for Arc colocalization with junctional Crb, but the second PDZ domain (PDZ2) is not, as shown by the lack of Arc accumulation to the Crb-enriched cell–cell contacts with Arc proteins that do not have PDZ1 (Fig. 3, F–J). The PDZ1 domain is also required for apical confinement of junctional Arc protein in the SG. When overexpressed in the SG, PDZ1-deleted Arc proteins (Arc^ΔPDZ1 and Arc^ΔPDZ1/2) were also observed in the lateral domain of SG cells where Crb is not found (Fig. 3, K–N″). This suggests that the apical localization of Arc could be, at least in part, through direct physical interactions between Arc–PDZ1 and Crb–PBM. However, we failed to detect co-immunoprecipitation of Crb and Arc from extracts of cultured cells or embryos, suggesting that either the interaction between Crb and Arc is too transient or weak for this biochemical assay and/or that the colocalization of these two proteins requires additional factors. Intriguingly, when overexpressed, Arc^ΔPDZ1 caused SG tube overelongation similar to full-length Arc, but Arc^ΔPDZ2 and Arc^ΔPDZ1/2 did not (Fig. S3, A–F). These results suggest that each PDZ domain has distinct molecular functions.

Based on the localization and overexpression results, we asked if the PDZ domains are necessary for Arc function in SG morphogenesis. Using CRISPR/Cas9 gene editing, we introduced nonsense mutations at the N terminus of each PDZ domain in the arc gene locus to generate PDZ domain–deleted arc mutations (arc^Q776Stop and arc^T1234Stop) (Fig. 3 O). Both arc^Q776Stop and arc^T1234Stop mutant flies have the arc wing defect, demonstrating a general requirement for both PDZ domains in tissue morphogenesis (Fig. 3, P–R). Removal of both PDZ domains (arc^Q776Stop) resulted in approximately the same increases in cross-sectional nuclei number (avg. 8.67) as those observed in the arc^KO mutant (avg. 9.08), indicating that the PDZ domains are essential for Arc function in SG morphogenesis. Deletion of the PDZ2 domain alone (arc^T1234Stop) produced a milder defect (avg. 7.83), suggesting that both PDZ domains are required but may have different molecular functions in the SG (Fig. 3, S–W). Taken together, Arc localizes with junctional Crb in SG cells in a PDZ1 domain–dependent manner, and both PDZ domains are required for Arc function.

Crb plays key roles in establishing and maintaining apicobasal polarity and transducing signals in many epithelial morphogenetic processes (Bulgakova and Knust, 2009; Tepass, 2012). During invagination, SG cells elevate Crb in an Fkh-dependent manner (Chung et al., 2017). A higher Crb level in SG cells than in surrounding epithelial cells is necessary for the formation of the supracellular MyoII cable around the SG placode (Ashour et al., 2023; Röper, 2012; Sidor et al., 2020). Furthermore, perturbations in crb expression in the SG placode disrupt proper internalization of SG cells (Chung et al., 2017; Le and Chung, 2021), highlighting the importance of Crb regulation in tube morphogenesis. The colocalization of Arc with Crb prompted us to ask if Crb is affected in arc mutants. Strikingly, Crb levels in the SG placode were notably decreased with loss of arc (Fig. 4, A–B″, and D), whereas excessive Arc did not significantly increase the overall amount of Crb (Fig. 4, C–C″, and D), although it did affect its distribution (see later sections). We questioned if Crb decrease by loss of arc affects other apical components in SG cells. A Crb-associating protein Stardust (Sdt) was decreased in arc^KO SG placodes (Fig. S3, G–I), as expected. We also observed a slight decrease of DE-cadherin levels in arc^KO SG placodes, but its junctional localization remained intact (Fig. S3, J–L).

To begin to understand how Arc affects Crb levels at SG cell junctions, we did FRAP to measure the exchange of Crb proteins on the cell membrane. In the WT placode, endogenous Crb tagged with GFP exhibited a dynamic behavior as shown by the significant recovery after photobleaching (avg. half-time of recovery [T_half] = 88.23 s, avg. mobile fraction = 58.33%). In the arc^KO placode, Crb recovered to the same extent (avg. mobile fraction = 59.13%) as in WT but at a significantly slower rate (avg. T_half 143.8 s) (Fig. 4, E–I and Video 1).

We further tested how loss of arc affects apical Crb levels by examining Crb distribution through sagittal sections of SG cells. Strikingly, we found that abnormally high numbers of Crb puncta were present inside arc^KO SG cells in comparison with WT cells (Fig. 4, J–L). Whereas the Crb puncta in WT SG cells mainly localized underneath the apical surface, the Crb puncta in arc^KO SG cells scattered from apical to basal (Fig. 4, J and K). Some of these puncta located apically overlapped or closed juxtaposed with Rab11 signals, suggesting these are intracellular vesicles trafficking Crb proteins to AJs (yellow arrowheads in Fig. 4, J″ and K″) (Le and Chung, 2021). Taken together, these results suggest that Arc facilitates replenishment of Crb to cell–cell junctions in invaginating SG cells to maintain optimal Crb levels for proper SG morphogenesis.

Non-muscle MyoII exerts contractile forces on the actin cytoskeleton to drive the changes in SG cell shape and arrangement during invagination (Booth et al., 2014; Chung et al., 2017; Sanchez-Corrales et al., 2018). We find that MyoII activity is critical for proper SG cell internalization, as shown by the failure of SG cells to fully internalize with SG-specific nanobody-mediated depletion of the myosin light chain (Drosophila protein spaghetti squash, Sqh) (Fig. 5, A–B′). As in other epithelial cells undergoing morphogenetic changes (Martin et al., 2009; Simões et al., 2017), MyoII is concentrated in an apicomedial domain and in apical cell–cell junctions in invaginating SG cells (Booth et al., 2014; Chung et al., 2017; Röper, 2012). Interestingly, MyoII and Crb localize in a mutually exclusive pattern within the SG placode, suggesting possible inhibitory interactions between the two proteins (Fig. 5, C and C′). Indeed, Crb has been reported to regulate MyoII function either positively or negatively in different cellular contexts (Biehler et al., 2021; Flores-Benitez and Knust, 2015; Sidor et al., 2020; Silver et al., 2019; Simões et al., 2022), suggesting that Crb regulation by Arc in SG cells may impinge on MyoII activity during invagination.

To ask if Arc affects MyoII activity, we compared MyoII distribution throughout the SG placode in WT and arc^KO embryos (Fig. 5, D–F). In WT embryos, MyoII intensity is the strongest in both the apicomedial plane and in junctions in cells closest to the invagination pit; this intensity declines with increasing distance from the pit (Fig. 5, D, G, and I). Whereas arc^KO SG cells present the same pattern in MyoII accumulation (highest at/near the invagination pit and lowest in the SG cells farthest from the pit), the steepness of the MyoII gradient throughout the placode is significantly reduced (Fig. 5, E and G–J). Thus, without Arc, MyoII activity in the SG placode is abnormally elevated in cells several diameters away from the invagination pit. As expected from the abnormal MyoII activation, we found that apical constriction in arc^KO SG cells is defective. Instead of a coordinated constriction pattern—smaller apical area in only SG cells close to the invagination pit (Fig. 5, K and K′)—arc^KO SG cells quite distant from the invagination pit have abnormally small apical areas (Fig. 5, L and L′). Surprisingly, the size distribution of apical areas of arc^KO versus WT SG cells is no different (Fig. 5, M and N); it is only the spatial distribution of cells with smaller apical areas that is changed with loss of arc (Fig. 5, K′ and L′). Cells with the smallest apical area are distributed more anteriorly in arc mutants than in WT (Fig. 5, O–R). These findings propound a molecular pathway wherein Arc promotes Crb localization. In turn, Crb attenuates MyoII activity to limit apical constriction to the subset of cells closest to the invagination pit, thus potentially limiting the number of SG cells that invaginate at any given time. Our findings also suggest that other factors affect the range distribution of apical area sizes in the invaginating SG.

To test the model wherein Arc-dependent increases in Crb limit MyoII activity, we examined the localization of these proteins in SG placodes overexpressing Arc. Intriguingly, whereas Arc overexpression did not overtly increase the total amount of Crb in invaginating SG cells (Fig. 4 D), it did alter Crb localization. In WT SGs, Crb localizes exclusively to AJs between cells, predominantly at bicellular junctions (Fig. 6, A–A‴). To our surprise, Crb also localized to diffuse punctate structures in/near the apical surface of Arc-overexpressing SG cells (Fig. 6, B–B‴). We detected similar ectopic Crb accumulation in diffuse punctate structures in or near the apical surface membrane in SG cells overexpressing Crb, although Crb levels are also increased at cellular junctions (Fig. 6, C–C‴). When overexpressed, Arc also accumulates to very high levels in apical-medial puncta in addition to its junctional localization, although Arc and Crb do not colocalize in these apical-medial puncta (Fig. S4). These observations indicate that excessive Arc (as well as excessive Crb) results in the delivery of Crb protein to cell membrane locations beyond its normal distributions.

We next asked if MyoII is also affected by Arc or Crb overexpression since loss of arc resulted in decreases in Crb levels and increases in MyoII activation (Figs. 4 and 5). As described earlier, MyoII normally forms large trabeculate networks on the apicomedial plane and near AJs in SG placode cells close to the invagination pit (Fig. 5, C and C′; and Fig. 6, D–D‴). In many SG placodes overexpressing Arc, MyoII formed smaller, less contiguous aggregates (Fig. 6, E–E‴). A similar, more exacerbated pattern of MyoII accumulation was observed in SG cells overexpressing Crb (Fig. 6, F–F‴). Although apicomedial MyoII was evident in SG cells overexpressing Crb, junctional MyoII was nearly gone. Measuring MyoII intensity clearly demonstrated that Arc or Crb overexpression reduced apicomedial and junctional MyoII accumulation (Fig. 6, J and K). Thus, the ectopic accumulation of Crb, whether through overexpression of Arc or of Crb, disrupts MyoII activation, which is required for apical constriction. Supporting this notion, we found that Arc- and Crb-overexpressing SG placode cells have similarly large apical areas and frequently fail to form an invagination pit at the right developmental stage (stage 11) (Fig. 6, B and C). To further corroborate that Arc regulates MyoII activity through Crb, we asked if reducing Crb level could rescue MyoII activation in Arc-overexpressing SG cells (Fig. 6, G–I‴). Indeed, both apicomedial and junctional MyoII accumulation in crb heterozygous SG cells overexpressing Arc recovered to levels observed in WT cells (Fig. 6, L and M). These findings support a model wherein Arc acts through Crb to attenuate MyoII activation.

The changes in Crb and MyoII caused by Arc perturbation and the consequent defects in apical constriction prompted us to ask if altering Crb or MyoII activity affects SG tube morphology in a manner similar to loss or misexpression of Arc. We genetically manipulated the activities of Crb and MyoII specifically in the developing SG and probed SG tube dimensions by counting the nuclei number in cross sections.

As reported by others (Tepaβ and Knust, 1990), we found that crb null mutant embryos exhibit severe developmental defects from very early stages, which makes examination of SG morphogenesis impossible. To circumvent this obstacle, we attempted to reduce Crb function specifically in SG cells by expressing a truncated Crb protein that contains only the transmembrane and intracellular cytoplasmic domains (Crb^intra; [Klebes and Knust, 2000]). The Crb^intra protein localizes to the entire SG cell membrane (everywhere except where septate junctions form), unlike endogenous Crb, which is confined apically. Importantly, with Crb^intra expression, the endogenous Crb protein also localizes to the basolateral membrane, thereby potentially reducing Crb levels and function in the apical domain (Fig. S5). Overexpression of Crb^intra resulted in shorter SGs with increased cross-sectional nuclei number (avg. 8.95), very similar to the SG phenotypes observed in arc^KO embryos (avg. 9.08) (Fig. 7, A–C′ and G). Furthermore, hyperactivation of MyoII by overexpression of a myosin light chain kinase (Kim et al., 2002) also generated phenotypes similar to loss of arc: shorter glands with more nuclei in cross section (avg. 8.98) (Fig. 7, D, D′, and G). On the other hand, depletion of MyoII increased tube length and dramatically reduced the cross-sectional nuclei number (avg. 3.98), the same phenotypes as observed in embryos overexpressing Arc in the SG (avg. 5.63) (Fig. 7, E–F′ and G). In summary, we have shown that loss of Arc, decreased Crb activity, and increased MyoII activity generate shorter SG tubes with more cells in circumference, whereas gain of Arc and decreased MyoII activity result in abnormally long SGs with very few cells in circumference (Fig. 7 H).

The SG of Drosophila is a well-suited model for uncovering the molecules and mechanisms underpinning the morphogenetic processes of epithelial tube formation. Here, we report on Arc, a large PDZ domain–containing cytosolic protein that functions downstream of Fkh to regulate the overall dimensions of at least two embryonic tubular organs. With loss of arc, we observe shorter SGs and shorter HGs, with no corresponding change in overall embryo length. In the SG, where cell number is known to be invariant, we find that the length change in arc nulls reflects the arrangement of cells, with more SG cells in circumference and fewer SG cells along the length of the tube. Correspondingly, overexpression of Arc results in longer tubes with fewer cells in circumference. Our data support a model wherein Arc functions during early stages of SG internalization to promote Crb delivery to the apical plasma membrane. In turn, Crb modulates MyoII activity to control the number of cells that change shape and internalize at any one time, thus templating overall tube dimensions from the earliest stages of SG morphogenesis (Fig. 7 H).

Pools of MyoII are critical for multiple aspects of morphogenesis. Apicomedial MyoII provides forces perpendicular to the AJs to pull these junctions inward at points of contact (Martin et al., 2009). Junctional MyoII drives the corresponding shrinkage of AJs for apical constriction and for neighbor exchange through the formation of T1 junctions or multicellular rosettes (Bertet et al., 2004; Blankenship et al., 2006; Curran et al., 2017). Both apical constriction and neighbor exchange are critical to SG internalization, as these processes control which cells internalize in what order, as well as how many cells internalize at a given time, thus controlling overall tube geometry (Chung et al., 2017; Myat and Andrew, 2000a; Myat and Andrew, 2000b; Myat and Andrew, 2002; Sanchez-Corrales et al., 2018; Sánchez-Corrales et al., 2021). Here, we show that Crb and MyoII exhibit complementary junctional localization within the SG placode prior to and during SG internalization: where Crb is high, MyoII is low, and vice versa (Fig. 5 C). Related complementary patterns of Crb and MyoII accumulation are observed in SG cells at the boundary with the surrounding non-SG epithelium; junctional Crb is high where SG cells contact other SG cells but low where SG cells contact surrounding epithelial cells since, within the SG, Crb levels are much higher than in the surrounding epithelia (Chung et al., 2017; Kerman et al., 2008; Röper, 2012). A molecular pathway involving Crb’s recruitment of Par6, Cdc42, and Pak1 has been described for how Crb inhibits the membrane association of Rok (Sidor et al., 2020), a major activator of MyoII activity that acts through phosphorylation of the myosin regulatory light chain (Yoneda et al., 2005). Thus, where Crb is high, junctional Rok and, consequently, junctional MyoII are low. Correspondingly, MyoII is high in the boundary junctions, leading to the formation of the supracellular MyoII cable surrounding the gland primordia (Röper, 2012). The mutual exclusivity between Crb and MyoII within and around the SG placode positions Crb as a key player in limiting MyoII activity.

Crb protein levels in epidermal cells are maintained by active endocytosis and recycling (Bajur et al., 2019). During apical constriction, actomyosin-driven contractions of the apical surface impact junctional Crb (and AJ-associated proteins) by driving endocytosis and the disassembly of apical complexes (Bertet et al., 2004; Cavanaugh et al., 2020; Levayer and Lecuit, 2012). This turnover of apical complexes has been nicely demonstrated during NB ingression, a process wherein individual isolated NBs undergo pulsatile apical contractions that serially decrease apical area through a ratcheting-type process (Simões et al., 2022; Simões et al., 2017), much like occurs in mesodermal cells during gastrulation (Martin et al., 2009) and in SG cells during invagination (Booth et al., 2014; Chung et al., 2017). In NBs, apical area contraction and expansion during this ratcheting are accompanied by decreases and increases in junctional Crb levels, respectively. Importantly, the amplitude of apical constriction is dependent on the endocytosis of Crb and sorting of Crb for either degradation or recycling back to the junction. Loss of retromer-mediated recycling function reduces both Crb-containing apical vesicles and Crb at AJs. Loss of retromer-mediated recycling also accelerates apical shrinkage of NBs. This finding indicates that junctional Crb provides resistance to MyoII-mediated cell shape changes (Simões et al., 2022). Endocytosis of junctional Crb is normally limited by a PDZ domain–containing binding partner of Crb, Sdt. With loss of Sdt, Crb is rapidly and completely endocytosed (Knust et al., 1993). Indeed, the completion of NB ingression—delamination of NB cells from the surrounding epithelia—requires the degradation of Sdt and the consequent depletion of junctional Crb (Simões et al., 2022). Taken together, these studies indicate that junctional Crb stabilizes AJs, functioning in a manner opposite to that of apicomedial and junctional MyoII, but in keeping with known functions of Crb in maintaining/expanding AJs in all primary epithelia (Tepass et al., 1990), including in the SG (Chung and Andrew, 2014; Chung et al., 2017) (this paper). This antagonistic relationship between Crb and MyoII in the SG and in the NB (Simões et al., 2022; Simões et al., 2017) highlights the importance of Crb regulation in epithelial morphogenesis.

Since Arc colocalizes with Crb at AJs and increases the rate of Crb recovery to these junctions based on our FRAP experiments, we propose that Arc facilitates the recycling of Crb to the apical surface to keep junctional levels of Crb appropriately high to counter MyoII activity. Although we could not capture such interactions biochemically, Arc-dependent Crb recycling could, nonetheless, be mediated through direct interactions between PDZ1 of the Arc protein and the PBM of Crb, which is located at the C terminus of Crb’s small intracellular domain. Upon arrival at the apical surface, Arc may release Crb to allow for binding to Sdt, the PDZ domain–containing Crb-interacting protein that stabilizes junctional Crb (Tepass and Knust, 1993). This role for Arc would explain the relative depletion of junctional Crb in the absence of arc and the delivery of Crb to ectopic sites (apical-medial membrane) when Arc is overexpressed. Our C-terminal deletions removing either one or both PDZ domains from the endogenous Arc protein indicate that both domains are required for WT Arc function. Whereas the PDZ1 domain is necessary to limit Arc to the apical domain of SG cells, the PDZ2 domain is required for elongating the SG tube when Arc is overexpressed, suggesting that this domain may function in the efficient delivery of Crb to its site of action.

Altogether, our data, considered in light of published studies on Crb and MyoII in the SG as well as in other systems undergoing similar morphogenetic behaviors, provides a working model for how Arc affects tube geometry (Fig. 8). SG internalization begins in a dorso-posterior position in the SG placode, likely due to the localized delivery of Fog or its GPCR receptors (e.g., Smog) (Le and Chung, 2021; Sánchez-Corrales et al., 2021; Vishwakarma et al., 2022). Upon receptor engagement, RhoGEF2, activated downstream of the receptor-associated G proteins, activates Rho1 GTPase and subsequently apicomedial Rok (Barrett et al., 1997; Levayer et al., 2011). In turn, localized Rok activates apicomedial Myosin, which drives apical constriction by pulling inward on the AJs (Booth et al., 2014; Chung et al., 2017). Based on the studies of NBs (Simões et al., 2017), we propose that apicomedial MyoII flows into the nearby AJs to shrink them through the endocytosis of Crb and AJ components (Bertet et al., 2004). The anisotropic flow of MyoII into the junctions results in the local endocytosis of Crb, creating domains of high MyoII and low Crb. This initial asymmetry is reinforced since Crb negatively regulates membrane-associated Rok (Sidor et al., 2020), leading to the downregulation of activated MyoII in areas of higher Crb. Anisotropy in MyoII accumulation is important since SG cells must also rearrange with respect to one another so that the right number of SG cells internalize at any given time in the process (Sánchez-Corrales et al., 2021). Importantly, it is the variance in levels of AJ-associated MyoII that drives cell rearrangements (Curran et al., 2017). Thus, MyoII drives apical contractions and decreases in AJ size. In concert with Crb, MyoII also drives directional cell rearrangement, with Crb stabilizing contacts between neighboring cells and MyoII destabilizing these contacts.

We propose that the increases in apical-medial and junctional MyoII observed in cells several rows away from the invagination pit in arc mutants are a direct consequence of the reduced levels of plasma membrane localized Crb. We further propose that the consequent increases in MyoII support increases in both apical constriction and cell rearrangement, resulting in more SG cells internalizing during early stages of tube formation. We observe that overexpression of Arc results in increased levels of apicomedial Crb and decreases in both apicomedial and junctional MyoII. In the apicomedial domain, Crb may interfere with Rok localization and/or activity. Reduced activation of MyoII in the apicomedial domain would consequently decrease junctional MyoII due to decreased flow. The consequence of these decreases in MyoII activation is that fewer cells undergo apical constriction and rearrangement, leading to the formation of tubes with fewer cells invaginating at any given time, with consequent longer tubes with a subset of cells that sometimes fail to internalize and remain on the embryo surface.

One interesting aspect of this study and related previous reports is that the transcription factor Fkh activates expression of two players in SG invagination that support mechanically opposing functions. Fkh activates fog expression in the SG (Chung et al., 2017). In turn, Fog activates MyoII through a GPCR signaling pathway that activates Rok in the apicomedial domain of invaginating SG cells (Chung et al., 2017). Rok-activated apicomedial MyoII drives apical constriction by pulling AJs inward. Independently, Fkh activates arc expression to ensure efficient delivery of Crb to AJs. At these junctions, Crb recruits the protein complex that phosphorylates Rok and decreases its membrane association, thus limiting Rok’s ability to activate junctional MyoII. These molecular pathways can explain the anisotropy in junctional MyoII and Crb distribution observed in SG cells. There, the two proteins have opposing activities, with MyoII driving junction disassembly and Crb stabilizing junctions. Thus, these Fkh downstream targets, through their opposing actions, fine-tune the morphogenetic behaviors required to build epithelial tubes of proper dimensions (Fig. 8). We fully expect that the identification of additional Fkh target genes and investigation of their potential roles in modulating MyoII and Crb functions will lead to a deeper understanding of the molecular and cellular principles governing epithelial tubular architecture.

The fly lines used in this study include Oregon R (as WT, RRID:DGGR 109612), w¹¹¹⁸ (as WT, RRID:BDSC_3605), fkh H99 (Myat and Andrew, 2000a), fkh-Gal4 (Henderson and Andrew, 2000), Crb::GFP-A (Huang et al., 2009), Sqh[ki]eGFP (Sqh-eGFP^29B; [Proag et al., 2019]), UAS-GFP-Crb^intra (Pellikka and Tepass, 2017), Sqh-mCherry (RRID:BDSC_99923), crb^11A12/TM3 (RRID:BDSC_3448), UAS-Crb (RRID:BDSC_5544), UAS-MLCK^ct (RRID:BDSC_37527), and UAS-Nslmb-vhhGFP (RRID:BDSC_58739) (Bloomington Drosophila Stock Center).

Two knockout alleles of the arc gene (arc^KO757 and arc^KO788) were generated by homologous recombination (Fig. S1) (Rong and Golic, 2001). Both alleles are identical molecularly and phenotypically, so are referred to as arc^KO in the main text. The PDZ domain deletion alleles, arc^Q776Stop and arc^T1234Stop, were generated by CRISPR/Cas9-mediated mutagenesis (Kanca et al., 2022).

Flies carrying the UAS constructs were generated using P-element–mediated germline transformation (Rainbow Transgenic Flies Inc.). Expression levels of Arc protein from UAS constructs by fkh-Gal4 were assessed by immunofluorescence staining with anti-Arc antibody. By the UAS/Gal4 expression, Arc staining in the SG was increased by about 1.5-fold at room temperature and by about 5.5-fold at 29°C. The phenotypic analyses of Arc overexpression were done at 29°C. It should be noted that fkh-Gal4 also drives gene expression in other non-SG tissues, including in the hemocytes, CNS, and anterior head ectoderm. Given the complementary phenotypes in the SG of arc loss and overexpression, we expect that it is the SG-driven expression that is causing the tube overelongation.

To generate the knockout construct for the arc gene, ∼4 kb of genomic DNA both upstream and downstream of the two largest coding exons of arc were amplified by PCR and cloned into the pW25 vector (RRID:DGRC_1166) (Gong and Golic, 2004) on either side of the white⁺ (w⁺) coding region (Fig S1 D). Transgenic lines were generated, and a line with an insertion of the construct on the third chromosome was subsequently crossed to a line expressing I-Sce1 under the control of a heat shock promoter. Progeny lines in which the w⁺ marker had moved to the second chromosome (where arc localizes) were selected and analyzed by PCR for w⁺ insertion into the endogenous arc gene (Fig S1 E).

To generate deletions of the PDZ domains in the endogenous arc gene, we attempted the method of Kanca et al. (2022) to cleanly delete each PDZ domain. This process includes synthesizing a DNA clone encoding two guide RNAs targeting the 5′ and -3′ regions of each PDZ domain and 200 bp immediately upstream and downstream of the corresponding PDZ domain, which should have yielded clean deletions of each PDZ domain. Although the guide RNAs worked well to cut the genomic DNA at the 5′ ends of each PDZ domain, homologous recombination did not occur with any of the lines that were generated. We did, however, succeed in generating lines by nonhomologous end joining that introduced premature stop codons, deleting either one or both PDZ domains (arc^Q776stop) or only PDZ-2 (arc^T1234stop).

To make the expression constructs of the arc gene (UAS-Arc, UAS-GFP-Arc, UAS-Flag-Arc, UAS-Flag-Arc^ΔPDZ1, UAS-Flag-Arc^ΔPDZ2, and UAS-Flag-Arc^ΔPDZ1/2), the Gateway cloning system was used (http://flybase.org/reports/FBrf0179058). To specifically delete each PDZ domain from the arc ORF, the Takara In-Fusion cloning system was used following the manufacturer’s instructions (Takara). The individual alleles were identified by genomic DNA PCR amplification and sequencing. The UAS lines expressing full-length or deleted versions of arc were generated by amplifying the arc ORF from cDNA GH21134 and cloning it into the pENTR-D/TOPO vector (RRID::Addgene_164959) and then recombining the ORF into the Gateway vectors (http://flybase.org/reports/FBrf0179058) for expressing either UAS-tagged (pTGW; RRID:DGRC_1075 or pTFW; RRID:DGRC_1115) or untagged (pTW; RRID:DGRC_1129) versions of the Arc protein.

Whole-mount mRNA in situ hybridization was performed as described (Lehmann and Tautz, 1994) using an antisense digoxygenin-labeled arc RNA probe generated from cDNA GH21134.

For immunostaining, embryos were fixed with 4% formaldehyde (Fisher Chemical) in PBS and devitellinized with methanol (Fisher Chemical). For α-Arc antibody staining, embryos were fixed by heat treatment as described (Liu and Lengyel, 2000; Peifer, 1993). For F-actin staining with phalloidin, the vitelline membrane was removed by hand after fixation with formaldehyde-saturated heptane (Sigma-Aldrich). The dilution ratios of primary antibodies follows: chicken α-GFP (1:500, RRID:: AB_300798; Abcam), mouse α-Crb (1:10, Drosophila Studies Hybridoma Bank [DSHB], RRID::AB_528181), mouse α–alpha-spectrin (1:2, RRID:: AB_528473; DSHB), mouse α-Arc (1;2,000, a gift from J.A. Lengyel/V. Hartenstein, University of California, Los Angeles, CA, USA; Liu and Lengyel, 2000), mouse α-Sdt (1:400, [Bulgakova and Knust, 2009]), rabbit α-GFP (1:500, RRID:: AB_221569; Molecular Probes), rabbit α-Flag (1:200, RRID:: AB_11232216; Proteintech), rabbit α-CrebA (1:100, [Fox et al., 2010]), rabbit α-Rab11 (1:500, [Le and Chung, 2021]), rat α-E-Cadherin (1:20, RRID:: AB_528120; DSHB), and guinea pig α-Sage (1:200, RRID:: AB_2632603, [Fox et al., 2013]). Alexa Fluor–conjugated (A-11001, A-11004, A-11008, A-11011, A-11039, A-11073, A-21235, A-21247; Invitrogen) or biotin-conjugated (715-065-150; Jackson ImmunoResearch) secondary antibodies were used at 1:200. Alexa Fluor 488– or Alexa Fluor 568–conjugated phalloidin (A12379 and A12380; Thermo Fisher Scientific) was used at 1:200.

For time-lapse live imaging, embryos expressing GFP-tagged endogenous Crb (Crb::GFP-A; [Huang et al., 2009]) were collected and dechorionated in 50% bleach. After rinsing in water, embryos were placed onto double-sided tape (3 M) on a microslide and covered in a layer of Halocarbon oil 700/27 (2:1; Sigma-Aldrich) with a coverslip. FRAP and time-lapse image acquisition were carried out on an LSM 700 Meta confocal microscope (Zeiss). Regions of interest (ROIs) in an SG placode were manually selected and imaged in five frames to record the original fluorescent intensity (prebleach). Then the ROIs were quickly bleached to less than ∼20% of their original intensity (time point 0) and subsequently imaged with 5-s time intervals (postbleach). The fluorescent intensities of the pre- and postbleached ROI were measured using the imageJ program. The GraphPad Prism software (GraphPad Prism version 10.2.3 for Windows, GraphPad Software, https://www.graphpad.com/) was used to determine the maximum recovery level (the percentage recovery to the pre-bleached level, mobile fraction) and the T_half in a kinetic curve fit with an exponential association equation.

S2 cells (CVCL _Z232), a gift from the laboratory of Elizabeth Chen, University of Texas Southwestern, Dallas, TX, USA, were cultured in Schneider’s medium (Gibco) supplemented with 10% FBS (Gibco) and penicillin/streptomycin (Sigma-Aldrich). To express a protein of interest, the UAS vector encoding the protein and the ubiquitin-Gal4 construct were transfected with Effectene (Qiagen) according to the manufacturer’s protocol. For immunostaining, cells were fixed on a glass coverslip with 4% formaldehyde in PBS, washed in PBS with 0.1% Triton X-100 (PBST) and PBST with 0.2% BSA (PBSTB 0.2% BSA) consecutively. The primary antibodies were used in PBSTB with the dilution ratios that follows: rabbit α-Flag (1:1,000; Proteintech) and mouse α-Crb (1:50; DSHB). Secondary Alexa Fluor 488–, Alexa Fluor 568–, or Alexa Fluor 647–conjugated antibodies (Invitrogen) were used at 1:500. DAPI (0.1 μg/ml; Thermo Fisher Scientific) was used to stain nuclei.

DIC images of embryos were obtained on an Axiophot microscope (Zeiss) with a Plan-NeoFluor 20×/0.50 lens equipped with the ProgRes CapturePro software (Jenoptik). Methyl salicylate (Sigma-Aldrich) was used for embryo mounting. The whole adult fly images were taken by an Iphone13 (Apple) on the Nikon SMZ800 microscope for magnification. Fluorescent images of embryos and S2 cells were obtained on an LSM 700 Meta confocal microscope with a Plan-NeoFluor 40×/1.3 NA oil or Plan-Apochromat 63×/1.4 NA oil DIC objective or an LSM 880 Airyscan (Zeiss) with a Plan-Apochromat 63×/1.4 NA oil objective. The Zen software (Zeiss) was used for image acquisition. Prolong Gold antifade mountant (Thermo Fisher Scientific) was used as the imaging medium. All images were taken at room temperature. The obtained images were processed using Image J (http://imagej.nih.gov/ij/), Imaris (Bitplane), or Photoshop CS (Adobe).

To measure the lumen length of the SG and HG, embryos were immunostained with an α-Crb antibody. The lumen length was manually measured by tracking Crb signal in each tube using the Image J program. To count the number of SG nuclei in cross sections, entire SGs immunostained with α-Sage, α-Crb, and α–α-spectrin were imaged in multiple Z focal planes. The cross-sectional images were reconstructed from the Z focal images using the ImageJ program. The nuclei at four different cross sections spaced optimally in one SG were counted and averaged to calculate the nuclei number for the SG. The GraphPad Prism software was used to calculate average and deviation values, to generate the graphs, and to perform the statistical test (two-tailed unpaired Student’s t test).

To examine localization of Crb and Arc at contacts between S2 cells, S2 cells expressing Crb and Flag-tagged Arc proteins were immunostained with anti-Crb and anti-Flag antibodies. Cell–cell contact areas were line-scanned based on Crb accumulation between cells. Fluorescent signals of Crb and Flag staining from the linescan were used to calculate Pearson’s correlation coefficient to assess co-localization of Crb and Arc. The GraphPad Prism software was used to calculate average and deviation values, to generate the graphs, and to perform the statistical test (two-tailed unpaired Student’s t test).

To quantify Crb amount in SG placodes, embryos were immunostained with anti-Crb. Phalloidin was used to stain F-actin. Anti-Sage staining was used to define SG cells. The fluorescent intensity from an entire placode was measured and normalized by area. Quantifying Sdt and DE-cadherin levels was done in the same way using the antibodies detecting each protein. The GraphPad Prism software was used to calculate average and deviation values, to generate the graphs, and to perform the statistical test (two-tailed unpaired Student’s t test).

To quantify MyoII intensity in apicomedial and junctional area, respectively, embryos carrying the GFP-tagged endogenous sqh gene were immunostained with anti-GFP antibody. The fluorescent intensity from the GFP staining from apicomedial or junctional area in individual SG cells was manually measured using the ImageJ program. The slope values of MyoII intensity through SG cells were computed using the Microsoft Excel program. The GraphPad Prism software was used to calculate average and deviation values, to generate the graphs, and to perform the statistical test (two-tailed unpaired Student’s t test).

To count Crb-harboring vesicles in invaginating SG cells, embryos were immunostained with anti-Crb. Rab11 was stained with anti-Rab11 antibody (Le and Chung, 2021) to mark the apical plane and endocytic vesicles. Crb-positive vesicles >0.2-µm diameter were counted through 30 Z focal images (0.22-µm distance between each plane) taken laterally for each SG placode. The GraphPad Prism software was used to calculate average and deviation values, to generate the graphs, and to perform the statistical test (two-tailed unpaired Student’s t test).

The apical area of cells in an SG placode was determined by cell segmentation and calculation as described previously (Chung et al., 2017). Briefly, E-cadherin images in SG placodes from one to three focal planes were integrated into one image by maximum intensity projection. The range of SG cells was determined by α-Sage staining. Cell segmentation based on the E-cadherin signals, the subsequent apical area calculation, and the colored map generation were performed by the Bitplane Imaris program version 10 (https://www.bitplane.com/imaris/imaris). All analyzed SG placodes were at the “deep invagination” state in stage 11, in which SG cells under the invagination pit had internalized >2 µm. Percentage distribution and cumulative percentage distribution of cells with a different apical area (bin width = 2) was performed using the GraphPad Prism software. P values were calculated using the Mann–Whitney U test (percentage of cells) and using the Kolmogorov–Smirnov test (cumulative percentage of cells). For the spatial distribution analysis of SG cells, the x and y coordinates of each cell in an SG placode were determined using the Imaris software and used as the positions along the AP and DV axes from the center of the invagination pit at (0,0). Standard deviations for the positions along the AP or DV axis were calculated using the Microsoft Excel program and used to compare relative dispersion of SG cells. The GraphPad Prism software was used to calculate average and deviation values, to generate the scattered dot graphs, and to perform the statistical test (two-tailed unpaired Student’s t test).

Fig. S1 shows the Fkh binding on the arc gene and generation of arc null alleles by homologous recombination. Fig. S2 shows that the arc affects SG tube dimension and HG length. Fig. S3 shows that the PDZ2 is required for elongated SG tube phenotype by Arc overexpression, and Sdt and DE-cadherin were decreased in arc^KO SG placodes. Fig. S4 shows that the overexpressed Arc localizes apicomedially and junctionally in invaginating SG cells. Fig. S5 shows that the overexpressed Crb^intra mislocalizes endogenous Crb protein. Video 1. shows the FRAP time-lapse movies of GFP-tagged endogenous Crb in WT and arc^KO embryos. Table S1 shows the oligos used in this study.

The data for the Fkh ChiP seq experiments can be accessed through this link: https://zenodo.org/records/15555276.

We thank members of the Andrew Lab who critically evaluated the science at all stages and Ashleigh Shoemaker and Katherine Hutchings, who provided direct feedback on the manuscript. We thank the Bloomington Drosophila Stock Center (NIH P40OD018537) for the fly stocks that were provided and used in this study. We thank Flybase for information on arc gene structure and the Gateway vector sequence information (Flybase, 1995). We thank the Developmental Studies Hybridoma Bank, created by the NICHD of the NIH and maintained at The University of Iowa, Department of Biology, Iowa City, IA, USA 52242, for the mouse α-Crb and α–alpha-spectrin antibodies. We thank Review Commons for their reviewing service and the reviewers who provided very thoughtful critiques of our original manuscript. We thank The Johns Hopkins Medicine Microscope Facility for use of the AiryScan microscope (NIH S10 OD023548).

We gratefully acknowledge support for this work from NIH RO1 DE013899 and NIH RO1 GM145873.

Author contributions: J.H. Kim: conceptualization, data curation, formal analysis, investigation, methodology, validation, visualization, and writing—original draft, review, and editing. R. Maruyama: investigation and methodology. K. Kim: formal analysis and investigation. D.A. Vertrees: investigation. P. Paul: investigation and methodology. K. Britson: investigation. N.R. Laughner: data curation, formal analysis, resources, validation, and visualization. D.J. Andrew: conceptualization, funding acquisition, investigation, methodology, project administration, resources, supervision, visualization, and writing—review and editing.