Branched-chain actin dynamics polarizes vesicle trajectories and partitions apicobasal epithelial membrane domains

In prevailing epithelial polarity models, membrane- and junction-based polarity cues such as the partitioning-defective PARs specify the positions of apicobasal membrane domains. Recent findings indicate, however, that intracellular vesicular trafficking can determine the position of the apical domain, upstream of membrane-based polarity cues. These findings raise the question of how vesicular trafficking becomes polarized independent of apicobasal target membrane domains. Here, we show that the apical directionality of vesicle trajectories depends on actin dynamics during de novo polarized membrane biogenesis in the C. elegans intestine. We find that actin, powered by branched-chain actin modulators, determines the polarized distribution of apical membrane components, PARs, and itself. Using photomodulation, we demonstrate that F-actin travels through the cytoplasm and along the cortex toward the future apical domain. Our findings support an alternative polarity model where actin-directed trafficking asymmetrically inserts the nascent apical domain into the growing epithelial membrane to partition apicobasal membrane domains.


INTRODUCTION
The prevailing polarity models place membrane-and junctionbased apicobasal core polarity cues (e.g., the partitioning defective PARs) upstream of intracellular cues during the establishment and modulation of plasma membrane polarity in epithelial cells (1,2). Once these core polarity cues have demarcated the positions and determined the identities of apicobasal membrane domains, these domains are thought to polarize intracellular processes, for instance, by providing cognate recognition sites for the vesiclebased sorting of apicobasal cargo that expands and maintains these domains (3). Domain exclusion of membrane-based core polarity cues (e.g., the PARs) is also considered a principal mechanism for modulating membrane polarity in the tissue context (4). Downstream of extracellular signals that orient cells within the epithelium, membrane-and junction-based core polarity cues are thus thought to impart the initiating polarizing event at the cell membrane, during both the establishment and modulation of epithelial membrane polarity. In these models, the process of polarized membrane biogenesis is viewed as the addition of molecules to, or the exchange of molecules within, membrane domains previously demarcated by the membrane-based polarity cues.
Recent findings have challenged these prevailing polarity models: (i) membrane-based apicobasal polarity cues (e.g., the PAR, Crumbs, and Scribble polarity complexes) appear to be dispensable for the polarity of some epithelia (5); (ii) multiple vesicle-based endo-and transcytotic-recycling components were identified that maintain or alter the position of apical membrane components and polarity cues (e.g., the PARs) in different tissues and position the lumen in tubular epithelia (6)(7)(8)(9)(10)(11)(12). These findings raised the questions: (i) what, if any, are the redundant or alternative modes of membrane polarization, and (ii) how can intracellular processes such as vesicular trafficking determine the position of polarized membrane components in the absence of demarcated target membrane domains?
We previously identified several trafficking molecules from unbiased Caenorhabditis elegans tubulogenesis screens whose perturbation reversibly changed the position of the apical membrane domain (lumen) in already polarized intestinal cells, overriding the polarity established by the core polarity cues (apicobasal membrane polarity conversion). The analysis of this polarity phenotype revealed that vesicle membrane and coat components [glycosphingolipids (GSLs), clathrin, and its AP-1 adaptor] can determine the position of the apical domain, upstream of PARs, in still dividing cells of the early-embryonic intestine (clathrin) and in postmitotic but still growing cells of the larval intestine (AP-1, GSLs) (7,10). Our accompanying article (13) demonstrates that multiple early (pre-Golgi) and late (post-Golgi) components of the anterograde (membrane-directed) biosynthetic-secretory (secretory) trafficking pathway are also required to determine the position of the apical domain in the developing C. elegans intestine, operating on both, not-yet polarized and already-polarized, expanding membranes. On the basis of these findings, we proposed an alternative mode of polarized membrane biogenesis, where the vesicle-based asymmetric insertion of the nascent apical domain into the growing membrane partitions apical and basolateral membrane domains (13). It remained unclear, however, how the secretory pathway that supplies all sides of the membrane sorts apical cargo to a not-yet polarized membrane during polarity establishment and to a basolateral (nontarget) membrane domain in already polarized cells.
Here, we focus on three components of the branched-chain actin machinery that we identified by the same apicobasal polarity conversion phenotype in the same tubulogenesis screens that identified the vesicle-based polarity cues (7,10,13). Our analysis reveals that branched-chain actin dynamics powers a vectorial F-actin network that moves from the basolateral to the future apical domain in polarizing C. elegans intestinal cells. All components of the branched-chain actin machinery and actin itself are required to route pre-and post-Golgi vesicles to the nascent apical domain during de novo polarized membrane biogenesis. These findings support the proposed alternative mode of polarized membrane biogenesis in epithelia (13) and suggest a mechanism for how anterograde vesicle trajectories can acquire long-range directionality independent of sorting to previously demarcated membrane domains.
contribute to bcAMs' polarity function, and demonstrated that bcAMs function cell-autonomously in the intestine [hence, it is presumed that the UNC-60 function is mediated by its intestine-specific isoform UNC-60A (14); fig. S1]. Confocal analysis of polarized membrane markers and functional studies in larval intestines revealed that mild interference with each of the three bcAMs changed the polarity of multiple apical, but not of basolateral, membrane components while preserving the integrity of apical junctions that secure the positions of apical and basolateral membrane domains [ fig. S1; note that excess ectopic lateral junctions form around basolateral lumens at later stages of polarity conversion (Fig. 1, X to Z); see fig. S1 for additional bcAM effects on membrane and junction biogenesis]. Transmission electron microscopy (TEM) confirmed the formation of junction-bounded ectopic basolateral lumens with microvilli in unc-60(RNAi) larval intestines with late-stage polarity conversion (Fig. 1, X to Z). All these phenotypic features copy key aspects of apicobasal polarity conversion induced by the loss of each of multiple trafficking molecules previously identified by intestinal polarity conversion in the same tubulogenesis screens (7,10,13).
We conclude that UNC-60, ARX-2, and CAP-1 are required to determine the position of the apical domain and thus membrane polarity in the C. elegans intestine. The close similarity of the bcAM-and trafficking-dependent polarity phenotypes (13) suggested that bcAMs and trafficking function together in the regulation of epithelial membrane polarity by determining the position of the apical domain on the expanding membrane [see the accompanying article (13)]. components) in expanding larval cells in the presence of intact junctions ( Fig. 2, C, D, and L), and formation of junction-bounded ectopic basolateral lumens at later stages of polarity conversion (Fig. 2, C to G; also note apical membrane blebs and apical vacuoles; see fig. S3 for glossary of terms for distinct membrane biogenesis defects). Conditional actin RNAi, induced in L1 larvae, was sufficient to change the polarity of already polarized yet still expanding membranes (Fig. 2J), demonstrating that actin directly affects polarized membrane biogenesis. Thus, actin is required to position the apical domain throughout de novo polarized membrane biogenesis, on not yet polarized membranes of dividing and moving cells during polarity establishment, and on polarized but still expanding membranes of postmitotic cells that have fixed positions within the mature but growing tissue [see the accompanying article (13) for the larval C. elegans intestine as postmitotic polarity model]. The single-cell EXC extends four arms through the animal. (W2 to W4) No EXC lumen extension (EXC is circled in Q to U and W2). Schematics indicate area shown above (boxed). (X to Z) Transmission electron microscopic (TEM) cross-sections of larval intestines (two cells are outlined by purple line in (X): lumen (L) with dense microvilli (yellow star) in (X) versus disorganized microvilli in (Y), and ectopic lumen (EL; boxed) in (Y), magnified below. (Z) EL with sparse dysmorphic microvilli (yellow star). Yellow arrows: intact apical junctions. Scaled-intensity RNAi is used throughout. Confocal images are shown in A to W4. Additional defects are described in fig. S1. See fig. S3 for intestinal morphogenesis. Scale bars, 20 μm (D to F), 10 μm (G to N and O to W1), 5 μm (W2 to W4), and 2 μm (X to Z). S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E (U to W) The actin(RNAi) polarity conversion is not enhanced by bcAMs loss. More than three replicas were analyzed (R to W). Numbers in red circles: additive values. One-way ANOVA for significance: **P < 0.01. Scaled-intensity RNAi is used throughout. Confocal images are shown (H: Nomarski). Scale bars, 10 μm, except (C and D) 50 μm, (E to G) 2 μm, and (J and K) 20 μm.
In support of a specific requirement for ACT-5 for polarity, RNAi that predominantly, although not exclusively, targets the ACT-5 3 0 UTR (designated act-5-LE17bp 3 0 UTR RNAi; LE17bp = 17 base pairs of the last exonic sequence upstream of the 3 0 UTR; fig. S2B) copied the actin(RNAi) polarity phenotype (Fig. 2K). act-5(ok1397), a large germline deletion (including 500-bp promoter) in a balanced background (i.e., in the presence of maternal act-5 product; fig. S2D) fully displaced ERM-1 from the membrane, masking possible effects on ERM-1's positioning at the membrane (Fig. 2, H and I). act-5-LE17bp 3 0 UTR RNAi and act-5(ok1397), in contrast to actin RNAi, induced larval, not embryonic, polarity defects and lethality. We concluded that actin functions in C. elegans intestinal polarity and that ACT-5 requires maternal product and/or additional actin isoforms for its polarity function.
To determine if actin operated as a downstream effector of bcAMs in polarity and if the three bcAMs affected polarity via a joined function in branched-chain actin dynamics, we carried out genetic interaction experiments, using mild RNAi conditions to capture convergent effects on polarity and avoid disrupting essential cellular functions. Triple RNAi, combining all three bcAMs, enhanced the individual unc-60, arx-2, cap-1(RNAi) polarity conversion ( Fig. 2: additive effect in M, enhancement in N and O), supporting the joined function of the three bcAMs in polarized membrane biogenesis (note that only mild RNAi is used in all genetic interaction experiments, here and below, and that only early polarity conversion on expanding L1-larval membranes is assessed; stronger depletion of unc-60, arx-2, cap-1, and actin causes a highly penetrant polarity phenotype, compounded by sterility and early lethality; fig. S2A).
We conclude that (i) actin determines the position of the apical domain (lumen), and hence polarity, on expanding membranes of embryonic and larval C. elegans intestinal cells; (ii) actin loss phenocopies the bcAM-and trafficking-dependent apicobasal polarity defect; (iii) genetic interactions in polarized membrane biogenesis between different bcAMs and between bcAMs and actin are consistent with these molecules' combined function in polarity via branched-chain actin filament dynamics. Actin thus appeared as a promising candidate cytoskeletal component to route anterograde vesicle trajectories to the nascent apical membrane domain in polarizing C. elegans intestinal cells (see Introduction and our accompanying article) (13).
bcAMs/actins and vesicle-based polarity cues determine the polarized distribution of apical PAR complex components on polarized and on not-yet polarized membranes Actin and bcAMs have conserved structural functions in cell cortex modeling that include the apical domain and junctions of the C. elegans intestine (21)(22)(23). These canonical functions of branchedchain actin dynamics take place in an epithelium already polarized by the membrane-based core polarity cues, e.g., by PAR-3, PAR-6, and PKC-3 (apical PARs), whose polarization is therefore thought to predate these activities (21). In contrast, the here-identified function of actin and bcAMs in apical domain positioning are already active at the time of intestinal polarity establishment (Fig. 1, O to R) and operational in the presence of intact junctions (Figs. 1, X to Z, and 2, E to G and L, and fig. S1). These features (early and junctionindependent activity) suggested that branched-chain actin dynamics' function in apical domain positioning was distinct from and anteceded its canonical function in apical domain modeling. Moreover, the same features apply to the previously identified vesicle-based polarity cues' function in apical domain positioning (13), suggesting that branched-chain actin dynamics-like vesicular trafficking-might act at an early step during membrane polarization, upstream of, or concomitant with, membrane-based core polarity cues.
To explore this possibility, we examined bcAMs'/actin's relationship to the apical PARs during de novo polarized membrane biogenesis in the developing intestine and tracked the polarized distribution of endogenously tagged PAR-3, PAR-6, and PKC-3, with and without ERM-1::GFP (green fluorescent protein), throughout net polarized membrane addition in embryonic and larval intestinal cells depleted of bcAMs or actin ( Mild RNAi with unc-60, arx-2, cap-1, and actin mislocalized apical PAR-6 and PKC-3 to the cytoplasm and to basolateral membrane domains of postmitotic but still expanding larval intestinal cells and, at later stages of polarity conversion, to basolateral ectopic lumens (Fig. 3, first to second row). Stronger RNAi conditions with each of these three bcAMs and with actin prevented the polarization of PAR-3, PAR-6, and PKC-3 in still dividing and intercalating early-embryonic cells, at the time when the intestine's definitive apicobasal polarity and lumen position are established (Fig. 3, third to fifth row; the expression of PAR-3, considered the earliest acting polarity cue in most epithelia, including the C. elegans intestine, weans at the larval stage) (24)(25)(26). PAR-3, PAR-6, and PKC-3 were retained in the cytoplasm of these early-embryonic cells and/or were located at all sides of a membrane that failed to be partitioned into an apical and basolateral domain. unc-60, arx-2, cap-1, and actin RNAi also perturbed intestinal intercalation and lumenogenesis, processes that depend on a polarized apical membrane [ Fig. 3; (26)]. bcAMs' and actin's effect on PARs' polarized distribution at expanding membranes of both larval and embryonic cells copied that of the previously identified vesicle-based apical polarity cues, here shown in chc-1(RNAi) early-embryonic, and let-767(RNAi) larval, intestinal cells ( Fig. 3; chc-1 encodes the clathrin heavy chain; let-767 encodes 3-ketocacyl-CoA reductase, a GSL biosynthetic enzyme) (7,10,13).
We conclude that branched-chain actin dynamics, like vesiclebased polarity cues, is required for the polarized distribution of apical PAR complex components on expanding membranes of C. elegans intestinal cells, including PAR-3, whose initial recruitment to, and polarized position at, the membrane depends on bcAMs and actin at the time of polarity establishment. Branched-chain actin dynamics, like intracellular trafficking, thus determines the position of the apical domain at an early step during membrane polarization and regulates apicobasal polarity on not-yet polarized and on already polarized but still expanding membranes (see the accompanying article) (13).
UNC-60, ARX-2, CAP-1, and F-actin concurrently shift from the basolateral to the nascent apical domain during intestinal polarity establishment bcAMs have been associated with anterograde vesicle trajectories in nonpolarized cells (27), consistent with the hypothesis that they confer directionality to anterograde trafficking in polarized cells. However, the subcellular localization of bcAMs/ACT-5 at the apical domain of the C. elegans intestine (17,21,22) seemed an unlikely position from which to guide vesicles through the cytoplasm. We expressed UNC-60::GFP, ARX-2::GFP, and CAP-1::GFP from the intestine-specific promoter elt-2 to determine the subcellular localization of these ubiquitously expressed molecules throughout intestinal development.
Next, we labeled intestinal cells with elt-2p-directed fluorescent LifeACT, an inert exogenous F-actin-binding molecule, to (i) optimize live imaging of filamentous actin (F-actin) throughout intestinal development, (ii) include all actin isoforms in the analysis, and (iii) avoid possible interference with filament assembly by actin fusion proteins or by endogenous actin-binding molecules. LifeACT reiterated the bcAMs' dynamic developmental subcellular intestinal expression pattern, also shifting from a cytoplasmic and pan-membranous location to the nascent apical domain during polarity establishment. Double labeling of LifeACT::mCherry and ARX-2::GFP highlighted the synchronous basolateral-to-apical shift of F-actin and ARX-2 ( Fig. 4, LifeACT::mCherry; ARX-2::GFP panels: A to E″).
We conclude that the developmental subcellular expression pattern of bcAMs and actin is compatible with a function of branched-chain actin dynamics in guiding vesicles through the cytoplasm to position the nascent apical domain during de novo polarized membrane biogenesis (see our accompanying article) (13). In such a function, branched-chain actin dynamics could operate from its transient location in the cytoplasm and at all sides of the membrane (future apico-basolateral membrane domains) of not yet polarized cells, rather than from its definitive location at the apical membrane domain.

The three bcAMs and actin interact with vesicle-based apical polarity cues in apical domain positioning
The bcAM/actin-dependent intestinal polarity defect phenocopies the trafficking-dependent polarity defect ( Fig. 5A; see the accompanying article) (13). Three lines of evidence further supported the hypothesis that branched-chain actin dynamics regulates membrane polarity via trafficking: (i) the previously noted reversibility of the act-5(RNAi)-dependent intestinal defects and larval lethality (not consistent with a structural morphogenesis defect) (17) To directly address the question if bcAMs/actins regulate polarity via trafficking, we examined genetic interactions between cytoskeleton-and vesicle-based polarity cues. Combining temperaturesensitive mutants (at the permissive temperature), balanced mutants (in the presence of the balancer) and mild RNAi conditions, we assessed interactions between animals with minor or no polarity defects (minimal depletion of either type of polarity cue). This approach avoided masking effects on membrane polarity (caused by strong depletion of these pleiotropic molecules) and allowed for the assessment of synthetic polarity phenotypes (induced by the combined depletion of cytoskeleton-and vesiclebased polarity cues).
CHC-1, the heavy chain of the post-Golgi vesicle coat clathrin, previously shown to function as polarity cue in the C. elegans intestine (10,13), was examined in the temperature-sensitive mutant chc-1(b1025ts) that arrests as embryo and fails to establish intestinal polarity at temperatures above 22°C but is viable and largely without polarity defects at 15°C. RNAi with bcAMs and actin was titrated in a wild-type background to produce (i) ≤ 1 / 2 progeny with early-embryonic lethality and intestinal polarity defects at 22°C (Fig. 5D), (ii) < 1 / 2 progeny with larval basolateral ERM-1 mislocalization at 15°C (Fig. 5E), and (iii) no progeny with ectopic basolateral lumen formation at 15°C (Fig. 5F). chc-1(b1025ts) enhanced the bcAM/actindependent early-embryonic lethality and larval polarity conversion at the restrictive and permissive temperatures, respectively (Fig. 5, D, E, and G). All double chc1(b1025ts) bcAM(RNAi) conditions generated a synthetic ectopic lumen phenotype at 15°C (Fig. 5, F and G).
GSLs, Golgi-and post-Golgi endomembrane-based lipid raft components, also previously shown to function as polarity cues in the C. elegans intestine (7,13), were examined in the let-767(s2819); sDp3 mutant that exhibits the polarity conversion/ectopic lumen phenotype in the absence of the duplication (sDp3) but appears wild type in its presence. RNAi with bcAMs was titrated to induce no or minimal basolateral ERM-1 displacement in a wildtype background. Double let-767(s2819) bcAM(RNAi) conditions generated a synthetic polarity phenotype in larval intestines (Fig. 5, H and J).
RAB-1 is an early secretory pathway molecule, located on pre-Golgi and Golgi endomembranes, whose depletion also induces polarity conversion in the C. elegans intestine (13). actin RNAi, titrated to generate no polarity defect on its own, enhanced the basolateral mislocalization of ERM-1 on expanding membranes of larval intestines mildly depleted of RAB-1 (Fig. 5, I and K).
We conclude that the three components of the branched-chain actin machinery and actin itself functionally interact in apical domain positioning with the previously identified pre-Golgi-, Golgi-, and post-Golgi-based polarity cues (see the accompanying article) (13).

Branched-chain actin dynamics confers apical directionality to pre-and post-Golgi vesicles during de novo polarized membrane biogenesis
If branched-chain actin dynamics asymmetrically routes vesicles with apical membrane components to the growing membrane to insert the nascent apical domain [ Fig. 6A; proposed polarity model; compare (13)], the apical directionality of biosynthetic vesicles should depend on bcAMs and actin during de novo membrane biogenesis and polarity establishment. The polarization of vesicles (and nuclei) toward the future apical domain was indeed noted to predate membrane polarity establishment in the C. elegans intestine (26,28). It remains difficult in any in vitro or in vivo system to visually track the directionality of biosynthetic trafficking, since polarized cargo becomes successively enriched from the endoplasmic reticulum (ER) to the plasma membrane, while vesicle-membrane and coat components are consecutively recruited and shed (29). To map wild-type vesicle trajectories through polarity establishment in C. elegans intestinal cells, we directed RAB-11::GFP and RAB-10::GFP, marking vesicles with presumed apical versus basolateral directionality (30), to the intestinal E-cell lineage by the earlyacting intestine-specific promoter elt-2. Consistent with RAB-11's proposed function in apical domain positioning (13), RAB-11::GFP was recruited to vesicles in intercalating early-embryonic intestinal cells, and GFP + vesicles became enriched at the apical plasma membrane during polarity establishment. In contrast, the spatiotemporal subcellular localization of RAB-10::GFP + vesicles remained largely unaltered, with mild enrichment at all, specifically at basolateral, membrane domains (Fig. 6B).
We concluded that UNC-60, ARX-2, CAP-1, and actin are required for the apical positioning of multiple pre-and post-Golgi vesicles during net polarized membrane expansion, including vesicles marked by polarity cues (e.g., RAB-11, CHC-1, and SEC-23) (13) and LROs, previously noted to move to the apical domain in the C. elegans intestine (31). The results were consistent with the hypothesis that branched-chain actin dynamics asymmetrically directs anterograde vesicle trajectories from the ER to the apical membrane domain during this domain's biogenesis [ Fig. 6, A and E; compare (13)]. The findings could also suggest that distinct and presumed specialized vesicle populations are co-opted for the (F) Spatiotemporal coexpression of ARX-2 with RAB-11 + , RAB-10 + , and RAB-5 + vesicles during intestinal polarity establishment. Merged images are measured by the Eyedropper tool: overlap = yellow (right; high magnification). Cytoplasmic overlap (early-embryonic intestine, left), but no vesiclebased ARX-2 overlap (some vesicles indicated by long arrows), except with apical membrane-close RAB-11 + vesicles (full arrowhead). Confocal/confocal-Nomarski overlay images of embryonic or larval intestinal cells are shown. Labeling by translational fluorescent fusion proteins, except for autofluorescent lysosome-related organelles (blue). elt-2p or vha-6p restrict the ubiquitous expression of vesicle components to the intestine. Scale bars, 10 μm. delivery of apical membrane components during de novo polarized membrane biogenesis.
To explore if bcAMs might assemble actin on specific vesicles for forward propulsion, we also searched for colocalization of bcAMs and post-Golgi vesicle markers during intestinal development. This analysis failed to detect any direct overlap with the tested vesicle markers, except for ARX-2::RFP colocalization at the apical membrane with RAB-11::GFP + vesicles ( Fig. 6F and fig. S4, C and D). These findings suggested that bcAMs, rather than propelling specific vesicles, e.g., via actin plumes, might direct entire vesicle populations by powering broader actin structures.

ACT-5 functionally interacts with other actin isoforms in intestinal membrane polarity
To understand how ACT-5/actin might route vesicles through the intestinal cytoplasm to the apical domain, we returned to the question if ACT-5, thought to be exclusively localized at the apical domain (17), was the only actin in the C. elegans intestine and was therefore mediating actin's polarity function. Several lines of evidence suggested that this was not the case:  (17)].
We generated isoform-specific ACT-1, ACT-2, ACT-3, and ACT-4 translational GFP fusions, in addition to ACT-5::GFP, and analyzed their developmental expression pattern. This analysis revealed that (i) ACT-1, ACT-2, ACT-3, and ACT-4 are also expressed in the intestine (in addition to their extra-intestinal expression; Fig. 7 and fig. S5); (ii) ACT-5-as well as ACT-1, ACT-2, and ACT-3-is also located in the cytoplasm and at future basolateral membranes in the early-embryonic intestine before and during polarization; (iii) ACT-5-and ACT-1, ACT-2, and ACT-3-undergoes a synchronous basolateral-to-apical polarity shift during intestinal polarity establishment (ACT-4::GFP is only detected in the mature intestine).
To determine if ACT-5 was even necessary for polarity, we titrated act-5 3 0 UTR RNAi conditions up, avoiding strong ACT-5 depletion expected to displace ERM-1 from the membrane and thus impede the ability to assess ERM-1's polarized position at the membrane. Although unable to induce polarity conversion in wild-type, act-5 3 0 UTR RNAi mispositioned ERM-1 to the basolateral membrane in an RNAi-sensitive rrf-3 background (table S1, Fig. 7). This background also augmented the penetrance of apicobasal polarity conversion in act-5-LE17bp(3 0 UTR-RNAi) and actin(RNAi) larval intestines. We concluded that act-5 is indeed necessary, but that additional actin isoforms may be needed, for actin's full polarity function. The results also validated the 3 0 UTR RNAi approach.
All three alleles failed to mislocalize ERM-1::GFP to the basolateral intestinal membrane, yet caused mild, isoform-specific apical domain biogenesis defects ( fig. S8). We concluded that, if isoforms other than ACT-5 function in polarity, these isoforms operate redundantly, require ACT-5, and/or require more than one non-ACT-5 isoform for their polarity function.
The genetic analysis of actins has been complicated by large families of almost identical, redundant actin isoforms and the susceptibility of filament assembly to imbalances in actin stoichiometries. To deplete non-ACT-5 isoforms only mildly yet distinguish isoform-specific functions, we trialed a combinatorial 3 0 UTR-directed RNAi approach and targeted the non-ACT-5 isoforms act-1, act-2, act-3, and act-4 separately and together (transcripts are closely similar, while 3 0 UTRs diverge; fig. S6A). All single, double, and triple combinations of act-1, act-2, act-3, and act-4 3 0 UTR RNAi, with or without enhanced RNAi conditions (rrf-3), failed to mislocalize ERM-1 to the basolateral membrane or to generate obvious other phenotypes (table S1, Fig. 7). In marked contrast, quadruple act-1 + act-2 + act-3 + act-4 3 0 UTR RNAi caused embryonic and L1-larval arrest with pronounced body morphology defects, yet also failed to affect polarity. However, quintuple act-1 + act-2 + act-3 + act-4 + act-5 3 0 UTR RNAi induced intestinal polarity conversion (act-5 3 0 UTR RNAi alone does not induce polarity conversion, see above) and enhanced polarity conversion in intestines treated with act-5 3 0 UTR RNAi in an rrf-3 RNAi-sensitive background or with act-5-LE17bp 0 UTR RNAi ( Fig. 7; note that act-5 produces >85% of intestinal actin, suggesting that the effect of act-1, act-2, act-3, and act-4 3 0 UTR RNAi on the total amount of actin is minor; fig. S6D and Fig. 7). These results revealed that non-ACT-5 actin isoforms (i) contribute to ACT-5's function in intestinal polarity, (ii) likely require ACT-5 for their function in polarity, and (3) function redundantly in intestinal polarity and extraintestinal morphogenesis or require several isoforms for these functions. The results also validated a multiple (up to quintuple) RNAi approach.
We conclude that ACT-5 is required for intestinal polarity, but that at least two more, almost identical actin isoforms may be needed in addition to ACT-5 for actin's full polarity function. The concomitant basolateral-to-apical shift of ACT-5 and of ACT-1, ACT-2, and ACT-3 in polarizing cells furthermore suggested that these different actin isoforms might contribute to an F-actin structure that confers apical directionality to vesicle trajectories during polarity establishment.

F-actin moves from the basolateral to the apical domain during polarity establishment
To visually identify a filamentous actin (F-actin) structure suitable to give direction to vesicles during membrane polarization in the developing C. elegans intestine, we searched for F-actin assemblies with documented roles in directional trafficking. Neither by phalloidin, nor by fluorophore fusions with ACT-5 or other actin isoforms, nor by LifeACT, could we detect obvious actin cables that might serve as long-range tracks to move vesicles via atypical myosins (32). However, short-range, apical membrane-close Factin tracks, implicated in apical secretion in other epithelia (33,34), might not have been resolved by conventional microscopy. To image the submembranous apical cytoskeleton at molecular resolution, we examined L1 larvae by stochastic optical reconstruction microscopy (STORM) and used phalloidin to detect all actins (Fig. 8, A to C). This approach resolved single actin filaments in apical microvilli at nanometer resolution (average, 26 filaments/microvillus; Fig. 8A), extending high-resolution confocal and immunoelectron microscopy studies (35). Double-color STORM revealed that ERM-1 traces actin filaments into the microvilli (Fig. 8, A to C), but neither phalloidin + nor ERM-1 + tracks were found to emanate Unexpectedly, no such belt, nor junctional F-actin, was identified. Instead, double-color STORM of F-actin and IFB-2 revealed that actin filaments were rooted in a peri-lumenal intermediate filament belt (Fig. 8B).
bcAM-powered F-actin networks can produce force with and without myosins, and actin filament turnover (treadmilling) can self-organize to create directional momentum that propels membranes forward (19). We next examined actin's ability to determine its own polarity, i.e., its polarization to the apical membrane. We took advantage of the transgenic ACT-5::GFP, expressed from a construct missing the act-5 3 0 UTR (act-5p::ACT-5::GFP::unc54-3 0 UTR; Fig. 8, D to I). The depletion of endogenous ACT-5 by act-5 3 0 UTR RNAi (figs. S2B and S6A) interfered with the polarity S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E of transgenic act-5p::ACT-5::GFP::unc54-3 0 UTR, whose expression was not decreased since it misses the act-5 3 0 UTR ( Fig. 8D; expression levels were, conversely, increased; see fig. S8M and legend for further discussion of this only partially rescuing transgene). In contrast, actin RNAi, used as control, effectively removed act-5p::ACT-5::GFP::unc54-3 0 UTR (Fig. 8E). Both act-5 3 0 UTR and act-5-LE17bp 3 0 UTR RNAi mislocalized act-5p::ACT-5::GFP::unc54-3 0 UTR to the cytoplasm and to basolateral membrane domains (Fig. 8, F and G). Moreover, single and combinatorial act-1, act-2, act-3, and act-4 3 0 UTR RNAi also mislocalized ACT-5::GFP (although not ERM-1::GFP, see above) to the basolateral membrane, to the cytoplasm, and to cytoplasmic and cortex-associated patches (Fig. 8H). Similarly, the single and combined depletion of UNC-60, ARX-2, and CAP-1 displaced ACT-5::GFP to basolateral membrane domains and to a mesh-like structure in the cytoplasm (Fig. 8I). We concluded that ACT-5, other actin isoforms, and the three bcAMs are nonredundantly required for ACT-5's own polarization to the apical domain in the developing intestine. Furthermore, non-ACT-5 actin isoforms are required to polarize ACT-5. Together, these findings were consistent with a bcAM-powered vectorial F-actin network that moves through the cytoplasm and along the cortex to position the nascent apical domain, and that might combine several, almost identical actin isoforms with ACT-5. However, bcAMs/actins could contribute in other ways to the positioning of the apical domain, a domain that also supports the de novo nucleation of F-actin that could be displaced to basolateral domains during polarity conversion.
To search for a putative vectorial F-actin network and track its dynamics during cell polarization, we used positional recording and time-lapse imaging of LifeACT, the most sensitive and actin isoform unbiased F-actin tracer (36).
In not yet polarized cells of the pre-intercalation embryonic intestine, high-resolution confocal microscopy of LifeACT::GFP revealed a cytoplasmic F-actin mesh, enriched at all sides of the membrane (Fig. 8J) To directly address whether F-actin physically moved across the polarizing cell, we generated photoactivatable LifeACT::PA-GFP and photoconvertible LifeACT::Dendra2 fusion proteins and expressed them in the developing intestine ( Fig. 9 and fig. S9). Activated LifeACT::PA-GFP (green) or converted LifeACT::Dendra2 (red) only detect F-actin generated before, but none generated after, activation or conversion. We examined the location of the activated GFP and the converted Dendra2 during intestinal development in single intestinal cells and in the whole organ by time-lapse confocal microscopy (compare Fig. 10, A to G″, for time-lapse images of nonmodulated LifeACT::GFP). The photomodulated fluorophores traced the movement of LifeACT through the cytoplasm and along the cortex to the apical domain in polarizing early-embryonic, and in postmitotic but still expanding late-embryonic/ larval, intestinal cells (Fig. 9, fig. S9, and movies S1 and S2). Activated LifeACT::PA-GFP (green) and converted LifeACT::Dendra2 (red) did not disperse randomly to other areas of cellular F-actin, as demonstrated by the counter labeling of newly generated Factin by LifeACT::mCherry (red) or nonconverted LifeACT::Den-dra2 (green), respectively. We conclude that F-actin physically moves toward the nascent apical domain throughout de novo polarized membrane biogenesis in single cells of the developing intestine. These findings are consistent with a scenario where the apical shift of a dynamic F-actin network confers directionality to biosynthetic vesicle trajectories that insert the nascent apical domain into the growing membrane [see our accompanying article (13)].

Branched-chain actin dynamics determines apicobasal membrane polarity
Here, we identify UNC-60/cofilin, ARX-2/Arp2/3 component, CAP-1/capping (designated bcAMs), and actin itself as intracellular polarity cues that act upstream of the membrane-based core polarity cues in the process of epithelial membrane polarization. Branchedchain actin dynamics is required to establish and maintain the position of the apical domain (lumen) throughout de novo polarized membrane biogenesis in the developing C. elegans intestine: The loss of any of its component results in apicobasal polarity conversion in the expanding larval intestine and in a failure to partition polarized membrane domains in the early-embryonic intestine during polarity establishment. We suggest that bcAMs operate in this polarity function from a transient cytoplasmic and pan-membranous location in cells poised to polarize, where they power a basolateral-to-apical F-actin shift that asymmetrically drives anterograde vesicle trajectories to the nascent apical domain. A scaled-intensity RNAi approach allowed us to separate actin's role in polarized membrane biogenesis from its many other essential cellular functions. The independent identification of all three components of branched-chain actin dynamics in unbiased tubulogenesis screens by the same apicobasal polarity conversion phenotype strongly endorses the requirement of branched-chain actin dynamics for this specific polarity function.

Actin positions the apical membrane domain by polarizing intracellular trafficking
The following evidence suggests that bcAMs and actin function in polarity by conferring apical directionality to anterograde vesicle trajectories to position the apical domain, rather than by their canonical functions in apical membrane or junction modeling: (i) the bcAM/actin-dependent polarity defect phenocopies the polarity defect induced by interference with vesicle-based polarity cues that direct biosynthetic trafficking to the apical domain (13); (ii) bcAMs/actins genetically interact with vesicle-based apical polarity cues in polarity; (iii) all components of branched-chain actin dynamics are required to direct pre-and post-Golgi vesicles to the apical domain during de novo polarized membrane biogenesis; (iv) both systems (actin dynamics and trafficking) are required to polarize PAR-3, PAR-6, and PKC-3 during polarity establishment; (v) the components of both systems, branched-chain actin dynamics (bcAMs; several actin isoforms; the actin network itself ), and vesicles shift from the cytoplasmic and basolateral to the apical domain during polarity establishment; (vi) a bcAM-driven F-actin network physically moves to the nascent apical domain during net polarized membrane expansion. We therefore propose that branched-chain actin dynamics determines membrane polarity by asymmetrically directing the biosynthetic-secretory pathway from the ER to the plasma membrane to insert the nascent apical domain, thereby partitioning the growing membrane into apical and basolateral domains. This proposition supports the idea that the directional delivery of bulk membrane can determine epithelial membrane polarity (13) and provides a mechanism for a directional regulation of membrane delivery that is independent of cargo sorting to already polarized target domains [see our accompanying article (13)].

A vectorial F-actin network, powered by bcAMs, confers apical directionality to vesicles
The classical mode of actin-guided vesicle movement is the MyoVdependent forward propulsion of vesicles along long-range cytoplasmic F-actin cables with predetermined directionality (32). Branched-or straight-chain actin dynamic-dependent apical trafficking, guided by short, apical domain-associated actin tracks,

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E
has also been described: the Arp2/3-dependent trafficking of the Notch ligand Delta into apical microvilli of Drosophila sensory organ precursors (37); the formin-dependent apical trafficking that expands but does not position the lumen in Drosophila tubular epithelia (33); apical secretion in the mouse pancreas (34). We failed to identify F-actin cables for directed vesicle propulsion, nor did we find apical domain-linked actin tracks in the C. elegans intestine, but instead, we identified a dynamic vectorial F-actin network. Straight-or branched-chain powered F-actin networks can also move vesicles and organelles and may operate with greater flexibility than predetermined tracks, responding to different modes of vesicle dynamics. For instance, the formin-dependent propulsion of RAB-11 + vesicles to the cortex of nonpolarized mouse oocytes (38) can be directionally altered by vesicle density to support asymmetric spindle positioning (39). Mitochondria, whose directional movement during cell division is Arp2/3-dependent in yeast (40), can both be "shuffled" and spatially directed in HeLa cells by a combination of a subcortical actin meshwork and organelle-based actin clouds with the ability to transform into comet tails (41). We suggest that the bcAM-powered F-actin network in the C. elegans intestine transiently acquires apical directionality in polarizing cells, with apical membrane-biosynthetic vesicles connecting to it via those vesicle components previously identified as apical polarity cues [e.g., COPI, COPII, and clathrin coat and coat assembly components (13)].

Cytoplasmic actin structure and dynamics Actin mesh
The vectorial F-actin network identified in polarizing cells of the C. elegans intestine is of a mesh-like structure that appears better suited to moving groups of vesicles than single vesicular carriers, typically shuttled along F-actin cables by myosins (32) or directly propelled by F-actin plumes, comets (42), contractile coats, or rings (43). Unlike the formin-dependent static and isotropic Factin mesh of similar appearance that maintains the polarity of microtubules in Drosophila melanogaster oocytes (44), the bcAM-dependent F-actin mesh in C. elegans intestinal cells is dynamic and anisotropic. It remains to be shown whether this mesh is contractile, and, if so, if it is retracted from or pushed to the apical domain and if its contractility is myosin-dependent. Contractility could depend on unc-60/cofilin activity alone, as shown in starfish oocytes, where the disassembly-driven contractility of a nuclear actin mesh becomes asymmetrically determined by cortex affixation (45,46). Little is known about the role of highly similar actin isoforms in such polymeric F-actin structures. We here find that several other, almost identical, intestinal actin isoforms recapitulate the basolateral-toapical shift of ACT-5, the dominant C. elegans intestinal actin, and are nonredundantly required to optimize F-actin's directional shift and its polarity function. These actin isoforms may therefore be functionally relevant structural F-actin mesh components. Actin treadmilling Treadmilling, powered by branched or straight-chain filament assembly modifiers, is the conserved mode of actin's self-organization into vectorial F-actin networks (18,19), used for directed forward movement from bacteria and parasitic fungi to human cells (47). bcAM-powered treadmilling can directly generate force on intracellular organelles (48,49), and it drives lamellipodial membrane protrusions to direct cell migration, where it strictly depends on stoichiometries between bcAMs, especially Arp2/3 and capping proteins (20). We find the stoichiometric requirement of the distinct roles of UNC-60/cofilin, ARX-2/Arp2/3, and CAP-1/capping in filament assembly/disassembly reflected in their genetic interactions during polarized membrane biogenesis in C. elegans intestinal cells, consistent with a shared function of these three bcAMs in polarity via actin treadmilling. However, a physical basolateral-to-apical Factin shift, endorsed by the movement of photomodulated LifeACT fluorophores, suggests that filaments also assemble into a dynamic higher-order polymeric F-actin structure, yet to be characterized [treadmilling itself, with continuous disassembly at one filament end and reassembly at the other end, only gives the appearance of filament movement (19)].

Actin biomechanics
Cytoplasmic actin flow or streaming, first linked to organelle movement in plants (50), is another actin-dependent mechanism to polarize intracellular processes. F-actin flow dynamics have taken center stage over canonical leading-edge F-actin dynamics (treadmilling) in directed cell migration (51), where gradients of actin network compression and destruction, regulated by myosin and cofilin, were shown to operate in the cytoplasm. In zebrafish oocytes, directional bulk F-actin dynamics induces phase segregation of ooplasm and yolk granules that in turn induces F-actin comet assembly at the yolk granules, which pushes them into the opposite direction (52). Actin clouds, phases, gel condensation and fluidation, and passive and active diffusion have been characterized as biomechanical forces that can combine with branchedchain actin or actomyosin dynamics to confer directionality to organelles (53). The Formin2-and MyoVb-dependent dynamics of an F-actin mesh combined with vesicles, for instance, can create a gellike cytoplasmic pushing mechanism ("active diffusion") that positions organelles as large as nuclei in mouse oocytes (54,55) and centers spindles in the mouse zygote to allow for the switch from asymmetric to symmetric cell division (56). The biomechanical analysis of distinct dynamic F-actin assemblies (flow, mesh, patches) in polarizing C. elegans intestinal cells may reveal yet another mode of actin's uncanny ability to generate spatial organization and directional movement within a cell.

Cortical actin dynamics
Our findings suggest that F-actin moves coincidentally through the cytoplasm and along the cortex from the basolateral to the emerging apical domain of polarizing C. elegans intestinal cells. Rho-dependent cortical actomyosin dynamics is a key polarity cue in nonepithelial cells, and it also determines polarity in the C. elegans one-cell embryo (zygote) (57), where it restricts PAR-3, PAR-6, and PKC-3 to the anterior cortex (58). Anterior-posterior PAR asymmetries and polarity itself are thought to be first determined at the cortex (59,60), with cortical F-actin flows able to asymmetrically selfpropagate and later self-align along the equatorial cortex during cytokinesis (59,61). Cortical polarity establishment by membrane domain boundary definition via cross-inhibition of anterior-posterior PARs is a process considered to be principally mirrored by PARs and other apicobasal membrane-based polarity cues in epithelia (62,63). A cortical (apical) actomyosin network was recently also implicated in the polarization, albeit not in the positioning, of the apical domain in epithelial cells (64).
In contrast to the well-characterized process of cortical PAR polarization, it is not clear how PARs are recruited toward the cortex and whether this recruitment itself is polarized. Although not considered drivers of polarity, intracellular processes, such as vesicle-and actin-based dynamics, are known to affect C. elegans zygotic polarity: Vesicle asymmetries are present (65); Arp2/3-dependent recycling maintains PAR-6 at the anterior cortex (9); and Arp2/3 dependent cytoplasmic actin dynamics moves the male pronucleus to the anterior domain (66). Modeled on the analysis of zygotic PAR polarity, the analysis of C. elegans intestinal polarity has remained focused on boundary definitions at the cortex, including the polarized positioning of apical PARs. However, unlike anterior PARs in the zygote (restricted to the anterior from an initial pan-membranous location), apical PARs (and junctions) are directly recruited as foci to the nascent apical domain and its lateral membrane boundaries in the intestinal cells (15,24,26,67). It is not clear how lateral PAR foci are consolidated into a contiguous apical PAR lining (and how apical spot junctions are consolidated into lateral junctions). If directly recruited from the cytoplasm, lateral PARs and apical spot junctions might be the results of "imprecise" delivery. On the other hand, F-actin, here observed to shift along the cortex to the nascent apical domain, could also provide the postulated cytoskeletal support for a regulated lateral-to-apical shift of PAR-3, currently considered to initiate polarity in the C. elegans intestine (15,24). A concomitant function of cytoplasmic and cortical F-actin dynamics in apical PAR polarization would provide robustness to membrane polarity establishment in the epithelial tissue context.

An alternative mode of epithelial membrane polarization
This study, in conjunction with the accompanying article (13), proposes a model for apicobasal membrane polarity where the actindirected, vesicle-based delivery and insertion of the apical domain (here identified by the apical membrane identity marker ERM-1) partitions apicobasal membrane domains and positions apical membrane-and junction-based polarity cues (e.g., PAR-3, PAR-6, and PKC-3; Fig. 10). This proposed model places intracellular polarity cues-bcAMs/actin (this study); multiple components of biosynthetic-secretory vesicles (the accompanying study)-upstream of membrane-based polarity cues within the process of de novo polarized membrane biogenesis. Early-acting intracellular polarity cues, as identified here, raise the possibility that apical PARs, as well as junction components, can be polarized in epithelia by their recruitment from the cytoplasm. PARs might be directly delivered by vesicles or by other modes of polarized transport, or they might be secondarily recruited by components of the newly inserted apical membrane (e.g., by GSLs or their local enzymes) (7,68), in which case they could travel either through the cytoplasm or along the cortex.
The establishment of epithelial membrane polarity is a multistep process that depends on reinforcement, requiring a succession of molecular cues that position, assemble, and expand the membrane, and build its domain-specific microdomains (2). In a model of membrane polarization via apical domain insertion, apical PARs could operate as early reinforcers of apical polarity by acting as platforms for the recruitment of submembranous apical molecules that in turn act as platforms at later stages [e.g., ERMs at the stage of apical microvilli assembly (69)]. Apical PARs could also reinforce apical polarity by securing the insertion of an expanding apical domain via their canonical functions in (i) separating the emerging apical from the basolateral domain by apico-lateral junction assembly (70), (ii) defining apicobasal domain borders within the signaling network of mutual inhibition of apical and basal polarity cues (63), and (iii) tethering the polarized biosynthetic vesicle trajectories to the expanding apical domain (71).
A role of vesicular trafficking as intracellular polarity cue is supported by observations in nonepithelial, flat-epithelial, and tubularepithelial tissues from worms to humans, where various trafficking molecules continue to be identified by their contribution to the positioning of the anterior, apical, or lumenal membrane domain, respectively (6)(7)(8)(9)(10)(11)(12). Perturbations of some of these molecules induce phenotypes closely resembling the here-observed apicobasal polarity conversion phenotype of the C. elegans intestine. For instance, a Rab11-dependent transcytotic-recycling defect causes polarity inversion in Madin-Darby canine kidney (MDCK) cysts that mirrors polarity conversion in the C. elegans intestine, which can also be induced by depleting RAB-11 (6,8,13). The identification of additional, foremost secretory trafficking components as apical polarity cues in the C. elegans intestine could suggest that recycling (MDCK) and secretion (C. elegans intestine) converge on an actinguided anterograde vesicle trajectory that traverses the recycling compartment to position the apical domain [see the accompanying article (13)].
The role of actin-guided secretory trafficking as a polarity cue has precedence in single-cell organisms (62,72). ACT1, the single actin of budding yeast, mediates polarized cell division by directing the secretory pathway to the bud site to insert the nonpolarized membrane for the daughter cell. The formin-dependent actin cables that route secretory vesicles to the bud site can also operate along the yeast cortex (73). Actin-guided secretion is likewise required for CDC-42-dependent spontaneous single-cell polarization in yeast (74) and for polarity site relocation during yeast mating (75). However, unlike the bcAM-powered basolateral-to-apical Factin shift that appears to drive vesicles toward the apical domain in C. elegans intestinal cells, the prior polarization of cortical actin in yeast "pulls" vesicles toward itself on actin cables that were previously radiated into the cytoplasm. In a similar "pulling" movement, the polarized Arp2/3-rich cortical actin cap in mouse oocytes maintains the spindle position by generating an actin flow that results in "retrograde" cytoplasmic streaming toward itself (76).
It remains to be determined how the actin-dependent anterograde vesicle trajectory is oriented in epithelial cells and what might trigger its asymmetric orientation. The midbody, a remnant of cell division, serves as landmark for apical transcytosis during polarity inversion in MDCK cysts (77). In the C. elegans intestine, the midbody, although not a candidate for the cortex-based lateral-to-apical movement of PAR-3 (15,67), might orient vesicle trajectories directly or indirectly via actin, a process that could predate or accompany the cortical PAR movement (see above, cortical actin dynamics). The midbody moves to and remains at the apical membrane in the early-embryonic C. elegans intestine, where RAB-11 + vesicles can be detected before PARs are polarized (78). Extracellular signals-e.g., basement membrane-derived extracellular matrix (ECM) signals-are conserved triggers for epithelial polarity that can directly or indirectly confer directionality to Factin dynamics or vesicle trajectories (79). In MDCK cells, ECM signals initiate polarity inversion and induce the basolateral endocytosis of future apical membrane components, including ERMs/ ezrin (80). No ECM-derived polarity signals have yet been identified in the C. elegans intestine (26,81), but multiple basement membrane components have now been shown to be expressed in the early-embryonic intestine (82). Lateral cortical polarity complexes, consisting of PAR-3 and the junction component HMR-1/cadherin, were recently suggested as intra-intestinal, albeit extracellular, polarity triggers in the C. elegans intestine (83). They must, however, like PARs and junctions, themselves be polarized to apical and apico-lateral domains, respectively (discussed above).
The here-proposed alternative mode of membrane polarization may be conserved at the earliest stages of cell polarization in mammals. Polarity in the eight-cell mouse embryo is established by the asymmetric insertion of the apical domain, an event likewise characterized by the apical recruitment of PARs and ERMs/ezrin (84). Moreover, the process of apical domain insertion is developmentally regulated by Tfap2c and Tead4, transcription factors that induce the expression of regulators of branched-chain actin dynamics such as Arp2/3 complex components. While Rho-dependent cortical actin dynamics was found to laterally expand the apical domain along the embryo cortex, actin dynamics was also noted to be required for the recruitment of apical membrane components toward the apical domain, a process not yet understood (85). It is therefore tempting to speculate that branched-chain actin dynamics also establishes polarity in the mammalian embryo by asymmetrically routing anterograde vesicle trajectories to the nascent apical domain. Recent reports on the polarization of human pluripotent stem cells via intracellular lumenogenesis by the "apicosome" (86) could furthermore suggest that the here proposed alternative mode of membrane polarization is conserved from C. elegans to humans.

Experimental model
Extended Methods for the in vivo analysis of polarized membrane biogenesis in C. elegans tubular epithelial cells are provided in (87,88).

C. elegans strains, culture conditions, and genetics
Wild-type (N2 Bristol) and mutant C. elegans strains were cultured, and genetic crosses were performed using standard methods (89). Worms were generally maintained at 20°to 22°C (unless otherwise noted) on Nematode Growth Medium (NGM) plates seeded with Escherichia coli OP50 (90). The list of strains used in this study is provided in table S2.

RNA interference General
Methods have been described in our accompanying article (13). Briefly, RNAi was carried out by feeding worms E. coli HT115 (DE3), producing double-stranded RNA (dsRNA) of the gene of interest, as previously described (7,91). Here, before seeding worms onto RNAi plates, animals were washed three times with M9 (89) and swirled on the plate in drops of carbenicillin solution (500 mg/ml) to avoid any OP50 contamination. For standard RNAi, bacterial feeding clones were inoculated from LB plates into 1 ml of LB liquid medium containing ampicillin (50 μg/ml) and incubated for 8 to 18 hours at 37°C. Cultured RNAi bacteria (200 μl) were seeded onto agar plates supplemented with 2 mM isopropyl-β-D-thiogalactopyranoside (IPTG) and carbenicillin (25 μg/ml). dsRNA was induced at room temperature for at least 6 hours before picking four to six L4 larvae onto each RNAi plate. Most bacterial clones were derived from the Ahringer genome-wide RNAi feeding library (J. Ahringer, Welcome Trust/Cancer Research UK Gurdon Institute, Cambridge, UK). The integrity of all RNAi clones was verified by sequencing.

Scaled-intensity RNAi
Rationale. Most genes examined in this study have essential, pleiotropic, maternal-effect, and dose-dependent functions in basic cellular processes (e.g., in intracellular trafficking and cellular morphogenesis). Germline mutant or strong tissue-specific (e.g., intestinal) loss-of-function conditions mask these genes' specific effects on polarized membrane biogenesis by disrupting these basic cellular processes. RNAi was the method of choice for lossof-function studies of these genes since it can be titrated to produce a range of mild to severe phenotypes. It also effectively targets maternal RNA.
Procedures. RNAi conditions were empirically determined for any given gene and experiment by modulating IPTG concentrations in RNAi plates, diluting the RNAi clone of interest with different amounts of mock RNAi bacteria, choice of age of the parental strain in which RNAi was induced (L2 to adult), and using the RNAi-sensitive strain rrf-3(pk1426) (87). In addition, conditions were varied by temperature (15°C, 22°C, and 25°C) and time of interference, i.e., RNAi was induced either in parents (evaluating the F1 progeny), in larvae (evaluating the same generation; conditional larval RNAi), or in adults. Conditional larval RNAi was carried out by bleaching 30 to 50 gravid adults in one drop bleaching solution (a 1:4 mix of 10 M NaOH and household sodium hypochlorite) on the edge of an RNAi plate and allowing hatched larvae to crawl to the bacterial lawn (87). Conditional RNAi was also induced at other stages of development by transferring L1, L2, L3, and L4 larvae or adult animals to RNAi plates and scoring the same generation. Appropriate controls were added to ensure that RNAi was effective when induced at later time points during development (e.g., by assessing the ability of gfp RNAi to remove fluorescence in a GFP-expressing strain). All experiments were repeated three or more times for each dataset. See table S3 for specific conditions used in each experiment.

Analysis of actin isoforms by 3 0 UTR RNAi
Rationale. The genetic analysis of actin has been complicated by actin's essential and pleiotropic functions, the sensitivity of filament assembly to any changes in actin stoichiometries, and the large families of almost identical and often redundant actin isoforms. Lossof-function analyses of actin isoforms typically fail to produce effects if targeting single isoforms. On the other hand, loss-of-function strategies that remove several or all actins induce sterility/early lethality when used globally or severe cellular morphogenesis defects when used in a tissue-specific manner. The mutational analysis of actins has largely recovered dominant alleles that typically induce dominant negative, but also neomorphic, changes [e.g., a dominant act-2 germline mutation, but not an act-2 germline deletion, produces embryonic morphogenesis defects in C. elegans (92)]. We trialed a 3 0 UTR-directed RNAi approach to (i) test the loss of function of various single and multiple combinations of isoforms without disrupting basic cellular functions and (ii) separately target almost identical isoforms by RNAi (act-1, act-2, act-3, act-4, and act-5 transcripts are closely similar, while 3 0 UTRs diverge; fig. S6A).
Procedures. The 3 0 UTR of each isoform was targeted starting from the stop codon ( fig. S6A). The Clone Mapper tool excluded any possible "off-target" regions (93). The desired (specific) 3 0 UTR region was generated by polymerase chain reaction (PCR) and inserted (Gibson method) (94) into the multiple cloning site of the L4440 backbone (a modified version of the pBlueScript plasmid) and sequence-validated. The plasmid was transfected into the E. coli strain HT115 (DE3), which carries a defective ribonuclease (RNase) III and an IPTG-inducible T7 polymerase gene, to generate dsRNA (91).

Genetic interactions (double mutant/RNAi analysis) Rationale
To bypass the inability to capture convergent genetic interactions between maternal-effect early lethal genes via full null alleles, we assessed the ability of mild double mutant/RNAi conditions to enhance a minimal phenotype or to generate a synthetic phenotype. Using the scaled-intensity RNAi approach, conditions were empirically determined for each gene that induced no phenotype on their own (or that only generate a minimal phenotype) and combined with temperature-sensitive or balanced mutant alleles that appeared wild-type in the presence of a balancer or at the permissive temperature.

Procedures
For detailed procedures, please refer to the correspong sections, "C. elegans strains, culture conditions, and genetics" and "RNA interference."

In vivo analysis of polarized membrane biogenesis in single cells
The expanding, single-layered postmitotic intestine of the late C. elegans embryo or early (L1) larva was used as an in vivo model for the analysis of de novo polarized (apical) membrane biogenesis. It allows for the separation of polarized membrane biogenesis from polarized cell division and migration that occur concomitantly during polarized tissue morphogenesis [see our accompanying article (13); see (fig. S3) for net apical membrane expansion in the C. elegans embryonic and larval intestine]. To distinguish membrane biogenesis defects from sequelae of preceding tissue morphogenesis defects, perturbations were introduced after completion of intestinal morphogenesis (e.g., by conditional larval RNAi). To distinguish defects in polarized membrane addition from defects in polarized membrane maintenance, the effects of perturbations introduced in the expanding larval intestine were compared to the effects of perturbations introduced in the no-longer expanding adult intestine. See above, RNAi, for procedures.

Measurement of apical membrane expansion in the developing intestine
Using confocal images, the length of the apical (lumenal) membrane of the intestine was measured at different stages of development, from the embryo to the adult, with the help of the "Annotation and Measurement tool: polyline length tool" provided in the NIS-Elements of the confocal microscope. The results were computed by a simple statistics table and exported into an Excel sheet. Triplicates of 15 animals each were examined. Image pixels: 512 × 512. Objectives used: for embryo and L1: 60×; for L2 to L3 larvae: 20×; for L4 larvae and adults: 10×.

DsRed feeding
Methods have been described in our accompanying article (13). DsRed HT115 RNAi bacteria were generated with a DsRed-expressing plasmid. Animals were fed on plates containing RNAi bacteria targeting the gene of interest and control RNAi bacteria for 2 days. More than 50 animals were transferred to plates containing a 1:1 mixture of gene-specific and DsRed-containing RNAi bacteria at least 15 hours before evaluation.

Temperature shift experiments
Experiments were performed in a temperature-controlled incubator or temperature-controlled room. Same stage animals were evaluated at different time points at 15°C versus 22°C to account for slower development at the lower temperature.

Fluorescent fusion proteins
All strains with fluorescently labeled fusion proteins are described in table S2. The subcellular localization of most fusion proteins used in this study to identify polarized membrane domains and junctions was previously confirmed by us and others by various labeling procedures (e.g., antibody staining, chemical staining, germline knockins). Many were also internally controlled by a panel of different transgenes, fluorophores, and by both N-and C-terminal fusions.
Although not critical for their use as markers, most were also shown to be functional by the rescue of the corresponding mutant phenotype (see text and figure legends for references).

Rationale for generating exogenously tagged fusion proteins and extrachromosomal transgenes
The bcAMs UNC-60, ARX-2, CAP-1, and the five actin isoforms were exogenously tagged to (i) avoid the risk of generating phenotypic effects by modifying these molecules' germline loci, given the sensitivity of actin filament assembly to minor changes in actin modulation and stoichiometries; (ii) support a high-resolution subcellular analysis of these ubiquitously expressed molecules by restricting their expression to the intestine (actin isoforms were expressed from their own promoters to assess whether they were expressed in the intestine); and (iii) achieve optimal expression levels by modulating transgene copy number (see procedures below). All bcAMs were previously shown to be expressed in the intestine (references in text). They were used in this study to determine subcellular localization changes during development. Fig. 4 and fig. S4C document identical expression of exogenously and endogenously tagged fusions for ARX-2::GFP and ARX-2::RFP and demonstrate the superior resolution of subcellular structures labeled by the exogenously tagged ARX-2::GFP. Procedures Methods were previously described (87,88).
Promoters. To examine expression from an early time of intestinal development, most fusion proteins were directed to the intestine by the elt-2 promoter, exclusively expressed in the clonal intestinal Elineage (95)(96)(97). To optimize expression and facilitate cloning procedures, the length of this promoter was empirically determined for each construct and varied over a range of 600 to 5000 bp (table S2).
Cloning. Genes were tagged at their 3 0 or 5 0 ends and cloned in frame with GFP, tagRFP, or other fluorophores, using standard cloning procedures or Gibson Assembly (94). We generated all bcAMs (UNC-60, ARX-2, CAP-1) and several vesicle-associated markers (e.g., RAB-11) as N-terminal and C-terminal fluorescent fusions to confirm subcellular localization. Promoters (either elt-2p or the endogenous promoter) were amplified by PCR from wild-type genomic DNA. Some cDNAs were amplified from their respective cDNA plasmid clones (gifts from Y. Kohara, National Institute of Genetics, Mishima, Japan). GFP and tagRFP DNA fragments were amplified from ppD95.75 and pPD284, respectively [(98); Addgene]. For 3 0 UTRs, unc-54, elt-2, or the endogenous 3 0 UTR of the gene were used, as indicated. Plasmids were either cloned by standard procedures or generated using the PCR-stitching method (99). All recombinant plasmids and PCR-stitched chimeric DNAs were sequence-verified.
Transgenesis. To optimize expression levels, we generated extrachromosomal transgenes by germline transformation using standard procedures (100), as it allowed for testing a range of different DNA concentrations. With the goal to use the lowest concentration possible to achieve good subcellular expression, a concentration of 1 to 2 ng/μl DNA was used for germline transformation for most genes, with the expectation to generate low transgene copy numbers (101,102). All strains were examined for the absence of overexpression-induced artifacts. To generate transgenic lines with extrachromosomal arrays, DNAs were microinjected into either wild-type, unc-119, or other mutant C. elegans gonads for germline transformation, with or without the pRF4 plasmid that encodes a mutant collagen [rol-6(su1006)] marker, and/or the unc-119 + rescuing construct, using standard techniques (100,101). Multiple lines were generated for each transgene. Primers for cloning are available on request. See table S2 for strain genotypes.

Specific fusion proteins
ERM-1::GFP. ERM-1 is used in this study as an apical membrane identity marker [see our accompanying article (13)]. To avoid any possible disturbance of the erm-1 germline locus that might interfere with apical membrane biogenesis, an integrated transgenic line is used in this study (VJ610, table S2). This strain was previously characterized, is devoid of overexpression artifacts, and accurately tracks ERM-1's subcellular localization, confirmed by a spectrum of different transgenic strains (high and low copy number, fused to different fluorophores), labeling approaches, rescue of the erm-1 mutant phenotype, and an ERM-1::GFP germline knock-in (7,10,103,104). GFP, mCherry, or tagRFP fusions with components of endomembranes or vesicle membranes and coats. GFP, mCherry, or tagRFP fusions with components of endomembranes or vesicle membranes and coats are directed to the intestine either by the elt-2 (this study) or vha-6 promoter (the latter strains are gifts from B. Grant; table S2).
ACT-3::GFP. The translational ACT-3::GFP fusion was not compatible with survival beyond the embryonic stage, although robustly expressed in a variety of different lines, even when the plasmid was introduced at low concentration (1 ng/μl).
LifeACT::GFP. To minimize the risk of interference with actin filament assembly, the exogenous LifeACT was chosen over endogenous small F-actin-binding molecules such as the C-terminal actin binding sites of C. elegans ERM-1 or of the closely related Drosophila MOE (9,105). LifeACT strains were backcrossed four times and integrated using ultraviolet (UV) irradiation to establish nonmosaic transgenic lines (106).
Photo-convertible Dendra2 and photoactivatable PA-GFP fusions. A pFG102 vector was constructed, containing 566 bp of elt-2 promoter followed by LifeACT and Dendra2 or PA-GFP, the elt-2 3 0 UTR, and a kanamycin selection marker.
PA-GFP. Initially described by Patterson and Lippincott-Schwartz (108), PA-GFP was previously used by Mijalkovic et al. (109) in C. elegans. We engineered a photoactivatable GFP using Genewiz DNA block and Gibson cloning to introduce all four known PA-GFP mutations (L64F, T65S, V163A, and T203H) into the Fire vector pPD95.75 GFP. The LifeACT coding sequence (encoding only 17 amino acids) is separated by a flexible linker (encoding seven inert amino acids) from the protein, and synthetic introns are optimized for brighter expression in C. elegans (Fig. 9D). Constructs were microinjected, and transgenic strains were generated, as described above.

Immunohistochemistry
Methods have been previously described (87,88). Briefly, L1 larvae were collected in M9 medium (89) onto slides coated with 1 to 2% poly-L-lysine (Sigma, P5899), covered with overhanging coverslips, and then permeabilized by flash freezing in liquid nitrogen and subsequent flicking off of the coverslip. Fixation was performed by sequential incubation in methanol and acetone at −20°C. Immunofluorescent staining was carried out as described [procedures are demonstrated in (87)]. For MH33 staining, slides were exposed to the first antibody (1:10 dilution) overnight at 4°C, washed, and then exposed to the secondary antibody for 1 hour at room temperature. Permount (Fisher, SP15-100) was used as a coverslide mounting medium.

Confocal and dissecting microscopy
Methods are described in our accompanying article (13). Briefly, differential interference contrast (Nomarski) and confocal images were acquired using a Nikon Eclipse-Ti inverted microscope equipped with a C2 confocal system. Most confocal images were obtained with a 63× objective. Exposure to fluorescent light was S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E minimized to avoid bleaching, and images were obtained within minutes of mounting. Images were captured as single sections or a series of sections along the z axis with differing thicknesses (generally 0.1 to 1.0 μM). For multichannel images, individual channel intensity was adjusted and the samples were scanned sequentially to exclude the possibility of bleed-through between channels. Confocal imaging parameters, such as pinhole size and laser intensity, were empirically determined based on fluorophore intensity and experimental setting (avoiding phototoxicity, photobleaching, bleedthrough). Deconvolution software was only used in Fig. 8, where indicated, and images were not further edited except for adjustment of brightness and contrast (Adobe Photoshop).

The eyedropper tool
To measure the intensity of the merged channels of two fluorophores, we took advantage of the eyedropper tool in Adobe Photoshop 2022. The eyedropper tool provides color information and displays the intensity value of each RGB (red, green, blue) band of any single pixel under the pointer. In addition, it displays the measured color of that particular pixel in a large circle that is normally indistinct to the eye. The values in the info panel are the original color values without any adjustment. The eyedropper tool's instruction is provided on the Adobe website as Histogram and Pixel Value.

Fluorescence intensity measurement during time-lapse imaging of activated GFP
Fluorescence intensity of the entire image was measured using ImageJ (110) software. Since all images recording photoactivated PA-GFP during one time-lapse imaging experiment were taken with the identical confocal image acquisition criteria and settings (e.g., laser intensity, gain, pixels, and pinhole), background noise subtractions were not performed. Results were reported as integrated density. Triplicates of 15 animals were analyzed for each set.

Transmission electron microscopy (TEM)
Methods have been previously described (13). Briefly, larvae were washed off in standard M9 medium (89) and collected into 1.5ml Eppendorf tubes. They were then fixed in 2.5% glutaraldehyde and 1.0% paraformaldehyde in 0.05 M sodium cacodylate buffer (pH 7.4) plus 3.0% sucrose. Before fixation, the cuticles were "nicked" with a razor blade in a drop of fixative under a dissecting microscope to allow the fixative to penetrate. After an initial 2-hour fixation at room temperature, the specimens were transferred into fresh fixative and stored overnight at 4°C. Specimens were rinsed several times in 0.1 M cacodylate buffer and then postfixed in 1.0% osmium tetroxide in 0.1 M cacodylate buffer for 2 hours on ice. After postfixation, specimens were rinsed several times in 0.1 M cacodylate buffer and then embedded in 2.0% agarose in phosphate-buffered saline (PBS) for ease of handling. The agarose blocks were dehydrated through a graded series of ethanol to 100%, dehydrated briefly in 100% propylene oxide, and pre-infiltrated overnight on a rocker in a 1:1 mixture of propylene oxide:Eponate resin (Ted Pella, Redding, CA). The following day, the agarose blocks were infiltrated in 100% Eponate resin for several hours, then embedded in flat molds in fresh Eponate resin, and allowed to polymerize for a minimum of 24 hours at 60°C. Thin sections were cut on a Leica UC7 ultramicrotome and collected on formvar-coated grids, poststained with uranyl acetate and Reynold's lead citrate, and viewed in a JEOL 1011 TEM at 80 kV equipped with an AMT digital imaging system (Advanced Microscopy Techniques, Danvers, MA).

Stochastic optical reconstruction microscopy Mounting
Coverslips, not glass slides (length, 50 mm; thickness, 0.16 to 0.19 mm; Fisherbrand, 12-544-EP), were used to mount the worms. Thirty microliters of 2% poly-L-lysine (Sigma, p5899) was added per two coverslips, sandwiched, separated, and air-dried for 30 min. Five microliters to 10 μl of 1× PBS was added, and 50 to 100 animals were mounted onto the slide (in 1× PBS). The second coverslip (length, 22 mm; thickness, 0.13 to 0.17 mm; Fisherbrand, 12-542-BP) was added on top, and with gentle pressure, all animals were flattened. The sandwich was placed on a −80°C metal block (using liquid nitrogen) and frozen for 5 min. Once frozen, one coverslip was flicked off.

Photo-conversion (Dendra2)
Mounting. Early bean, bean, comma, or 1.5-fold stage embryos were selected and transferred onto agarose pads. A square-shaped 3 mm × 3 mm agar pad (5% noble agar) was created with enough O 2 ventilation to avoid hypoxia. Four to five transgenic embryos were mounted onto this pad into a 1-μl drop of M9 and covered with a standard glass coverslip.
Photo-conversion. Photo-conversion was carried out using a laser scanning confocal microscope equipped with a 63× objective. Before conversion, a z-stack image was taken using fluorescein isothiocyanate (FITC) and tetramethylrhodamine isothiocyanate (TRITC) channels and customizing parameters to avoid bleedthrough between the channels (TRITC laser wavelength: 488.0, Photomultipliers High Voltage (PMT HV): 70, PMT offset: 0, and FITC laser wavelength: 561.0, 6.0, PMT HV: 78, PMT offset: −15, S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E and the Transmission Detector (TD) light channel PMT HV: 89, PMT offset: 0). The region of interest (ROI) was selected using the ROI tab, and stimulation was applied (UV 405 nm, C2plus stimulation as 0.36; Near Diffraction (ND) limited stimulation: 1 s, 3 loop). Another postconversion z-stack image was taken at minute 0, followed by images at desired time intervals during further embryonic and larval development.

Photo-activation (PA-GFP)
Equipment and mounting were as above (photoconversion). Before photoactivation, an image of the worm was acquired as described above for photoconversion. Next, using a 405-nm UV laser, the ROI was photoactivated using the customizable ROI bleaching tool of the laser scanning confocal acquisition software. Here, optimal photoactivation was achieved using the 405-nm laser at 10% power. Worms were imaged immediately following photoactivation. The image acquisition settings (e.g., exposure, gain, laser power, and binning) were strictly maintained between intervals in time-lapse imaging experiments.
Images were not further edited to remove movement artifacts (background of activated GFP or converted Dendra2 outside the targeted area) induced by rapid embryo movement within the eggshell during later stages of embryonic development.

Statistics
Statistical analyses were performed by GraphPad Prism 9.3.0 (Mac) software. All values are mean ± SEM of three or more independent experimental data sets. P values were calculated by analysis of variance (ANOVA) or Student's t test, as indicated in the legend of each figure. n (sample size) is indicated in the text and in the figure legends.