Medioapical contractile pulses coordinated between cells regulate Drosophila eye morphogenesis

Morphogenesis of the Drosophila retina involves changes in cell shape, size, and junction length that seem largely transient. Through this process, which unfolds over tens of hours, epithelial cells achieve their precise number, shape, and arrangement. The result is a tessellated structure composed of ∼800 nearly identical ommatidia (Cagan and Ready, 1989). This process is a model for understanding morphogenesis of differentiated cells forming a mature organ. Each ommatidium consists of four central cone cells, surrounded by two large semicircular primary (1°) cells. These are surrounded by lattice cells (LCs) and mechanosensory bristles (Fig. 1 A). Of these, the LCs exhibit a particularly intricate developmental sequence in which they intercalate to arrange in a single file between ommatidia, extra LCs are pruned, and remaining cells undergo dramatic shape changes according to their location (Cagan, 2009; Carthew, 2007; Johnson, 2021).

The initial arrangement of cells in ommatidia is controlled by differential adhesion (Bao and Cagan, 2005; Bao et al., 2010; Hayashi and Carthew, 2004). After the intercalation of LCs, certain cytoskeletal networks are preferentially associated with a subset of LCs’ contacts. Specifically, we showed that contractile junctional actomyosin networks assemble at and contract the LC–LC contacts. These contractile networks are counterbalanced by protrusive branched actin networks that assemble along the same contacts with reciprocal dynamics (Del Signore et al., 2018; Malin et al., 2022). This interplay of contractile and protrusive forces along cell contacts guides, at least in part, their movements and prevents errors in cell shaping and arrangement.

Medioapical actomyosin networks have been observed during epithelial remodeling in several processes in Drosophila, including gastrulation, germband extension, and dorsal closure (Azevedo et al., 2011; Blanchard et al., 2010; Fernandez-Gonzalez and Zallen, 2011; Martin et al., 2009; Rauzi et al., 2010; Solon et al., 2009). In those processes, medioapical actomyosin networks pulse—that is, they repeatedly assemble and disassemble. Disrupting pulsing affects cell behavior and epithelial organization. However, these early developmental processes fundamentally differ from the remodeling of the retina. They entail massive reorganization of cells that takes place over the time course of tens of minutes compared with the precise refinement of the retina, which takes tens of hours. Furthermore, the pulsation typically displays a ratcheting characteristic, in which relaxation after constriction is only partial and the cells never revert to their original shape. In contrast, in the retina, there is no obvious ratcheting characteristic to cell-shape changes.

Cell movement and shape changes, however, are clearly critical for generating the very precise organization of the retina because errors occur when they are perturbed (Blackie et al., 2020, 2021; Del Signore et al., 2018; Johnson et al., 2008; Larson et al., 2008; Letizia et al., 2019; Malin et al., 2022; Seppa et al., 2008). Thus, a major challenge is identifying mechanisms that regulate cell dynamics and understanding their impact on force generation, force transmission, and the eventual attainment of specific cell shapes. Likewise, little is known about if and how neighboring cells affect each other in this process. A medioapical actomyosin network is known to be present in the developing fly retina, but its regulation and role in coordinating cell shape changes across an epithelium remain unexplored (Blackie et al., 2020).

Here, we used Drosophila LCs to investigate the organization, dynamic properties, roles, and regulation of the medioapical actomyosin network in LC remodeling using high-resolution live imaging and laser ablation combined with genetic tests of gene function. We found a pulsatile medioapical actomyosin network, whose assembly in a ring-like structure promotes the transient, anisotropic contraction of LCs. In addition to identifying regulators of this network, our experiments surprisingly revealed the role of medioapical actomyosin networks in creating and responding to mechanical interactions between neighboring cells. Our work illuminates the operation and regulation of medioapical actomyosin networks through cells’ mechanical state and interaction between neighboring cells. Overall, our data provide evidence that biochemical and mechanical signals regulate Rho1 function to rebalance forces in the epithelium, control cell shape and cell rearrangements, and maintain tissue integrity.

The developing fly retina is a model for understanding how force-generating cytoskeletal networks control cell shape, cell-to-cell interactions, and tissue patterns. To examine the medioapical actomyosin network during lattice remodeling, we live-imaged F-actin using the actin-binding domain of Utrophin tagged with GFP (Utr::GFP), together with Spaghetti Squash (Sqh), the regulatory light chain of non-muscle Myosin II (MyoII), tagged with mCherry (MyoII::mCh). In time-lapse movies, imaging a Z-stack every 60 s, we found that F-actin and MyoII accumulated in particles that flowed from the cell surface toward the cell’s medioapical region as the apical area exhibited pulsing shape changes (Fig. 1 B). Hinted at in low-resolution imaging (Fig. 1 B) and more clearly visible at higher temporal or spatial resolution (Fig. 2 A), an actomyosin ring structure was transiently present (red arrowheads). To study how medioapical actomyosin accumulation and cell area changes relate over time, we used time-resolved Pearson’s correlation analysis. This method calculates a correlation coefficient for data pairs from two-time series, varying the time delay between them. The resulting coefficients are plotted against the time lag on a graph, showing the strength and direction of their relationship. This analysis not only detects correlations between synchronous endpoints but also between endpoints that precede or follow each other in time, revealing the temporal lag between them. Time correlation analysis confirmed that medioapical F-actin and MyoII intensity increased synchronously, with no lag, suggesting they coassemble to form a contractile actomyosin network (Fig. 1 B′ and Fig. S1, A–B′, R = 0.4400). We also observed fluctuations in the LCs’ apical area (Fig. 1 B) and asked if they correlated with fluctuations in actomyosin levels. We found a significant negative correlation between medioapical F-actin and LCs’ apical area (Fig. 1 B″ and Fig. S1 B″, R = −0.2706 at −15 s). F-actin peaked 15 s before peak apical constriction, consistent with the idea that an increase in medioapical actomyosin levels exerts tension on the apical cell perimeter leading to a reduction in the apical cell area.

The Rho1 Rho GTPase and Rho-kinase (Rok) control actomyosin network assembly and contraction (Dawes-Hoang et al., 2005; Magie et al., 1999; Mizuno et al., 1999; Mulinari et al., 2008; Warner and Longmore, 2009a, 2009b). Therefore, we examined the accumulation of GFP-tagged Rho1 and Rok reporters relative to F-actin marked with the F-actin–binding peptide Lifeact tagged with Ruby fluorescent protein (Lifeact::Ruby). We found that both Rho1 and Rok are moving and accumulating medioapically (Fig. 1, C and D, yellow arrows), suggesting that the two proteins are positioned to control medioapical actomyosin dynamics.

To better understand the dynamic properties of the medioapical actomyosin and its structure, we imaged the network at a higher temporal resolution, obtaining an image stack every 5 s (Video 1). The medioapical actomyosin peaked in intensity approximately every 6 min (Fig. 1 F; 349 ± 96 s), with a 1.4-fold increase in F-actin levels from trough to peak in each cycle (Fig. 1 G). We found that during epochs of cell contraction, F-actin particles moved toward one another and merged to form larger particles organized into a ring-like structure, corresponding to peak intensity (Fig. 1, E and E′; and Video 1). As the ring disassembled, some particles disappeared while others flowed toward cell edges and merged with the junctional network (Fig. 1 E″, yellow arrowhead; Video 1). Taken together, these observations imply that medioapical actomyosin forms a contractile network composed of interconnected nodes that pull on one another and the cell surface in a process controlled by Rho1 and Rok.

Using high-resolution Airyscan confocal imaging, we examined the structure of the medioapical actomyosin network at higher spatial resolution in time-lapse movies (Fig. 2 A). This provided increased clarity to the presence of nodes (yellow arrowheads) that form the medioapical ring-like structure (red arrowheads) and filamentous structures (blue arrowhead) that linked nodes to other nodes and the cell surface or to LC–LC contacts (orange arrowheads). This architecture suggested that the network exerts tension on the cell surface and that LCs are mechanically linked through a dynamic supracellular medioapical actomyosin network anchored, in part, at LC–LC contacts.

To test the mechanical role of the medioapical actomyosin network, we ablated the network either partially or entirely and investigated the effects. First, we asked if ablated cells remained viable and mechanically active. We found that after ablation, the network quickly recoiled toward the cell surface and then reassembled within 8 min (Fig. S2, A and A′). In contrast to apical membrane wounds, which induce purse string formation and closure (Abreu-Blanco et al., 2012; Martin and Lewis, 1992; Wood et al., 2002), our ablations did not, indicating that ablated cells remained viable and mechanically active. After ablation of 2° LCs, we found that F-actin structures recoiled toward the cell periphery (Fig. 2 B). Concomitantly, we observed a rapid anisometric relaxation of the apical cell area preferentially parallel to LC–LC contacts (Fig. 2, B and C; and Video 2). The severing and recoil of the medioapical actomyosin network increased the 2° LCs’ total area by approximately twofold, resulting from cell area expansion along the shorter axis, while the longer axis was not significantly affected (Fig. 2, B, C, and E, green double arrows). We next ablated the 3° LCs and observed a similar twofold increase in the apical cell area (Fig. 2, D and E; and Video 2). Additionally, we found that both the LC–LC (3°-2°) and 3°-1° contacts expanded, although 3°-1° contacts expanded more (Fig. 2 E, thin and thick green double arrows, respectively). Additionally, we compared the impact of ablating the medioapical actomyosin network in 2° LCs versus 3° LCs on cell contact expansion. Ablating the network in 2° LCs had a greater impact on 2°-3° contact expansion compared with the impact of ablating the network in 3° LCs. Conversely, ablating the network in 3° LCs had a greater impact on 1°-3° contact expansion compared with the impact of ablating the 2° LCs (Fig. S6). These mechanical responses imply that the medioapical actomyosin network contracts asymmetrically to preferentially shrink the LC-LC and 3°-1° contacts and guide the shape changes of the LCs. Overall, these recoil patterns indicate that in 2° LCs, the medioapical actomyosin network promotes the shortening of LC–LC contacts, while in 3° LCs, it preferentially shortens the 3°-1° contact and contributes to the shortening of LC–LC contacts (see model in Fig. 2 F).

We note a surprising non-autonomous effect of ablating the medioapical network in LCs. To study the effect on adjacent non-ablated cells, we imaged the process for several minutes after ablation. The ablation of an individual 2° LC initially resulted in the expansion of the adjacent LC (Fig. 2 G, green arrow). After this initial expansion, approximately half of the neighboring non-ablated LCs assembled a robust medioapical actomyosin network that contracted their apical cell area (red arrow; Fig. 2, G and G′, ∼52% response of N = 27). While the pulse duration in WT is ∼6 min, maximal contraction occurred 98 ± 8 s after ablating the neighboring LC, indicating that the ablation accelerated the pulse cycle. The response implies that LCs respond to the loss of tension from the adjacent cell by assembling medioapical actomyosin to restore tension.

We next investigated the effect of perturbing assembly of the medioapical actomyosin network during LC remodeling by manipulating Rho1. Expressing a dominant-negative Rho1 (Rho1^DN) caused a general increase in apical cell areas and large fluctuations in cell area over time (Fig. 3 A). In these cells, junctional and medioapical F-actin were sparse and adherens junctions (AJs) were at least partially compromised (white arrow). Nevertheless, cell area contractions were observed and they were accompanied by increases in F-actin (red brackets). This manifested as a stronger correlation between medioapical F-actin accumulation and contraction of cell area compared with wild type (WT; Fig. 3 A′). These observations provide another line of evidence for the role and importance of medioapical actomyosin networks in controlling changes in apical cell area and accumulating in response to low tension.

To test whether Rho1 was sufficient to affect medioapical actomyosin dynamics, we transiently overexpressed WT Rho1 using the GAL4/GAL80^ts system, successfully maintaining cell connectivity and tissue integrity. Rho1 overexpression altered the kinetics of medioapical actomyosin pulsing (Fig. 3, B and C; and Video 3). Rho1 overexpression also promoted the fusion and enlargement of medioapical F-actin nodes’ by approximately threefold compared with WT cells (Fig. 3, B and B′; and Video 3). There was an increased F-actin pulse frequency (245 ± 50 s in Fig. 3, C and C′, compared with 349 ± 96 s in WT in Fig. 1 F) and amplitude (4.5-fold change, Fig. 3 C″, compared with 1.4-fold change in WT in Fig. 1 G). With Rho1 overexpression, there was a robust (Fig. 3 C‴ and Video 3) correlation between F-actin and cell area (R = −0.6508 at −15 s, compared to WT shown in Fig. 1 B″, R = −0.2706 at −15 s). These results imply that Rho1 activity is tightly regulated to control the frequency and amplitude of medioapical actomyosin assembly and contractility. Furthermore, cell recoil velocities in Rho1-expressing cells increased after ablating the medioapical actomyosin network, supporting this conclusion (Fig. S5, D and E).

Following up on our observation that the medioapical actomyosin network appeared to be triggered to assemble by the loss of tension (Fig. 2 G and Fig. 3 A), we studied the behavior of neighboring cells with Rho1 overexpression, where there are strong changes in the apical cell area. We focused on the behavior of adjacent 2° LCs prior to completion of cell pruning. In this cell configuration, one LC shares only one contact with the neighboring LC and a second with a mechanosensory bristle, while the second LC shares a second contact with a flanking 3° cell. Thus, adjacent LCs are exposed to the minimum of possible mechanical inputs from neighboring LCs, allowing us to isolate the mechanical interactions between these cells while minimizing the relevance of other neighboring cells. Time-lapse movies showed that a 2° LC contraction coincided with a two-step response in a neighboring 2° cell (Video 4). First, the second 2° LC disassembled its actomyosin network and expanded its apical area (Fig. 3 E at 210 s). Then, it reassembled the network and contracted its apical area (Fig. 3 E at 595 s). This, in turn, coincided with the same two-step response in the first 2° LC, resulting in alternating cycles of expansion and contraction in neighboring LCs. The peak expansion of a LC occurred 15 s after the maximal contraction of the adjacent LC (Fig. 3 E′, R = −0.2020). Thus, as one cell contracted, the other expanded, and vice versa (Fig. 3, E′ and F; and Video 4). We also observed inversely coordinated cycles of expansion and contraction between adjacent anterior and posterior cone cells and 1° cells (Fig. S3, A–D, respectively). Thus, the medioapical actomyosin network not only contributes to cell-autonomous contractions but also exerts a force on adjacent LCs that sets up anticorrelated cycles of contraction and expansion in adjacent cells.

A salient characteristic of apical cell-area contraction produced by the Rho1-dependent medioapical network is that it is pulsatile. These pulsing shape changes are not ratcheting, and they oscillate for hours. To investigate the importance of the pulsatile nature of this process, we expressed a constitutively active myosin light chain kinase (MLCK^CA, Fig. 4 and Video 5), which constitutively activates MyoII using the GAL4/GAL80^ts system. We examined the AJs, sites of force transmission, in eyes that expressed MLCK^CA. We live-imaged α-Catenin tagged with GFP (α-Cat::GFP) to highlight the AJs and found that they were fragmented compared with WT (Fig. 4, B and C). A subset of the LCs failed to intercalate and clustered around mechanosensory bristles (Fig. 4 C, arrows). We also observed cells that dramatically expanded their apical area, suggesting that they failed to hold tension exerted by the medioapical actomyosin network and were pulled by their neighbors, thus compromising tissue integrity (Fig. 4 C, asterisks). Examination of these ruptures revealed filamentous actin structures linking MyoII foci across the rupture (Fig. 4 F). Furthermore, we also observed that MLCK^CA expression decreased cells’ average apical area (Fig. 4 E). Live imaging with MyoII::mCh showed a high level of MyoII along the entire cell perimeter. Additionally, MyoII was seen enriched in a continuous medioapical ring, differing from the assembly of nodes in a ring seen in WT. This ring transiently formed and disappeared (Fig. 4 F′ and Video 5) on average every 10 min (Fig. 4 F″), and its assembly correlated with cell area contraction (Fig. 4 F‴, R = −0.2126). Thus, MLCK^CA expression alters the organization and kinetics of the medioapical actomyosin network. This changes cell size, causes some cells to rupture, and results in disorganized epithelium.

Taking a complementary approach to perturbing pulsed medioapical actomyosin-dependent cell contraction, we expressed a constitutively active form of Dia (Dia^CA), which assembles linear actin filaments downstream of Rho1. We expressed Dia^CA transiently using the GAL4/GAL80^ts system. First, we fixed Dia^CA-expressing eyes and stained them for E-cad to examine the integrity of the AJs and epithelial organization. We found that the AJs were fragmented in Dia^CA-expressing eyes like in MLCK^CA-expressing eyes (Fig. 4 D and Video 5). Dia^CA expression also severely affected cell shape and arrangement in ommatidia (Fig. 4 D). We also observed intercalation defects (yellow arrow) and cavitation of cone cells (white arrows; Fig. 4 D). Unexpectedly, unlike the contraction of LCs in MLCK^CA eyes, Dia^CA expression significantly increased cells’ apical area (Fig. 4 E). To determine if Dia^CA affects medioapical actomyosin, we analyzed F-actin organization and behavior in these eyes (Fig. 4 G and Video 5). In WT, F-actin is enriched both medioapically and at LC–LC contacts, while in Dia^CA-expressing eyes, it accumulated uniformly in a thick band at the cell periphery and was missing medioapically. It is plausible that by binding efficiently to Rho1 through its RhoGTPase-binding domain, Dia^CA competitively inhibits Rok activation and MyoII phosphorylation, and thus the assembly of the medioapical actomyosin network. Overall, these observations imply that pulsing shape changes caused by medioapical MyoII and F-actin function to maintain force balance in the epithelium, affecting cell shape and arrangement. This further suggests not only that Rho1 levels must be carefully controlled, but also that a balance between network assembly and disassembly is required for epithelial remodeling and integrity.

The idea that pulsatile assembly and disassembly of the medioapical actomyosin network is necessary for morphogenesis suggests the importance of upstream regulators. We hypothesized that a Rho1 guanine nucleotide exchange factor (RhoGEF) and a Rho GTPase activating protein (RhoGAP), which respectively activate and deactivate Rho1, could regulate medioapical dynamics (Etienne-Manneville and Hall, 2002; Jaffe and Hall, 2005). To identify these putative regulators, we carried out an RNA interference (RNAi) screen for RhoGEFs and GAPs that influence eye epithelial remodeling. In this screen, we identified RhoGEF2 and RhoGAP71E based on RNAi phenotypes in epithelial remodeling.

To study RhoGEF2, we examined RhoGEF2 localization using a RhoGEF2 reporter tagged with GFP (RhoGEF2::GFP) and marked cell outlines using F-actin with Lifeact::Ruby (Mason et al., 2016). RhoGEF2::GFP accumulated medioapically in pulses (Fig. 5 A, yellow arrows) that partially overlapped with medioapical F-actin (green arrowhead). To determine if RhoGEF2 levels correlate with apical cell area, we measured the correlation between RhoGEF2 levels and cell area over time and found a weak negative correlation (Fig. 5 A′, left panel), with peak enrichment of RhoGEF2 occurring 15–20 s after peak contraction (Fig. 5 A′, right panel, R = −0.1361).

To determine the role of RhoGEF2 in eye epithelial development, we investigated its effects on epithelial remodeling and actomyosin dynamics. Loss of RhoGEF2 function in genetically marked mutant cell clones led to severe defects in the arrangement and shape of ommatidia cells (Fig. S4, G–G″). Broad depletion of RhoGEF2 by RNAi led to weaker defects that included loss of LCs and formation of rosettes that can result from defects in LC intercalation (Fig. S4, E and F). To test if RhoGEF2 controls medioapical actomyosin dynamics, we examined F-actin and MyoII dynamics in RhoGEF2 RNAi-expressing clones (Fig. 5, B–B‴ and Video 6). We found a decrease in medioapical actomyosin levels in the clones (Fig. 5 B′ and Video 6) compared with WT counterparts (Fig. 5 B″, red arrows). There was a weaker peak correlation (at −15 s) between MyoII accumulation and apical cell area contraction in RhoGEF2 RNAi clones compared with WT counterparts (Fig. 5 B‴). These results provide evidence that RhoGEF2 activates Rho1 to control the assembly of the medioapical actomyosin network and is essential for accurate morphogenesis of the eye.

We also investigated the impact of RhoGEF2 overexpression on these dynamics. We found that RhoGEF2 overexpression induced robust and more frequent changes in the medioapical actomyosin network that coincided with changes in the cell apical area (Fig. 5, C and C′; and Video 6). Actomyosin assembled into a dense network that correlated with apical area contraction (red arrow) and then disassembled (green arrow), resulting in apical area relaxation. The assembly of the medioapical actomyosin network strongly correlated with apical cell area contraction (Fig. 5, C′ and D, R = −0.4076 at −15 s) than in WT (Fig. 1 B″ and Fig. S1 B″, R = −0.270 at −15 s). Furthermore, overexpression of RhoGEF2 accelerated F-actin pulse frequency (Fig. 5 C″, 215 ± 56 s) compared with WT (Fig. 5 E, 349 ± 96 s) and enhanced the amplitude of F-actin accumulation (Fig. 5 C‴, 1.9-fold) compared with WT (Fig. 5 F, 1.4-fold). Thus, RhoGEF2 overexpression increased the amplitude and frequency of medioapical actomyosin assembly and phenocopied the Rho1 overexpression phenotypes. Cell recoil velocities in RhoGEF2 expressing cells increased after ablating the medioapical actomyosin network, supporting this conclusion (Fig. S5, C and E; and Fig. S6).

To localize RhoGAP71E, we live-imaged a RhoGAP71E reporter tagged with GFP (RhoGAP71E::GFP; Denk-Lobnig et al., 2021). RhoGAP71E::GFP was enriched at the cell surface and subtly accumulated medioapically during cell area contractions (Fig. 6 A, yellow arrow). RhoGAP71E levels peaked 10 s after maximal LC contraction and decreased during apical area expansion (Fig. 6, A–A″, R = −0.1399, blue arrow).

To determine if RhoGAP71E controls medioapical actomyosin dynamics, we first examined the effects of RhoGAP71E loss on LCs’ remodeling. Eliminating RhoGAP71E function or expressing RhoGAP71E RNAi in whole eyes produced flies without eyes. However, using the FLP/FRT (Flippase/Flippase Recognition Target) technique, we were able to recover sparse RhoGAP71E mutant cells (Fig. 6 B). The RhoGAP71E 1° and 2° mutant cells had smaller apical areas compared with WT counterparts, suggesting increased actomyosin assembly and tension (Fig. 6 B′). To confirm this, we examined levels of the phosphorylated activated form of MyoII (p-MyoII) in clones expressing RhoGAP71E RNAi (Fig. 6 C). We found an increase in p-MyoII levels in the clones (Fig. 6 C′) compared with WT cells (Fig. 6, C″ and C‴). We also live imaged clones expressing RhoGAP71E RNAi (Fig. 6 D and Video 7). Opposite to the effect of RhoGEF2 depletion on cell shape, we found a decrease in the apical cell area in cell clones expressing RhoGAP71E RNAi (Fig. 6 D′) compared with WT cells (Fig. 6 D″). The increased assembly of the medioapical actomyosin network in the RhoGAP71E RNAi-expressing cells increased the correlation between MyoII and cell area contraction (Fig. 6 D‴) and led to increased MyoII pulse frequency compared with WT (Fig. 6 D‴, 212 ± 69 s and 387 ± 177 s, respectively), similar to results with RhoGEF2 overexpression. These results indicate that RhoGAP71E inhibits medioapical actomyosin network assembly and thereby regulates the kinetics of pulsing shape changes during LC’s remodeling.

We also examined the effects of RhoGAP71E overexpression on medioapical actomyosin and apical cell area fluctuations. RhoGAP71E overexpression expanded the apical cell area and compromised the ability of a subset of cells to hold tension resulting in their expansion (Fig. 6, E–F′, asterisk). We also observed a loss of continuity of AJs marked with α-Cat::GFP in highly expanded cells (Fig. 6 F, yellow arrowhead). These results suggest that RhoGAP71E overexpression reduced medioapical tension, allowing the apical cell perimeter to relax. However, we were not able to detect a decrease in cell recoil velocities after ablating the medioapical actomyosin network in RhoGAP71E expressing cells compared with WT (Fig. S5 B). This may reflect compensatory uncoordinated contractions as observed in Rho^DN eyes, which may mask the effect in this assay. Although medioapical F-actin fluctuated in these eyes (Fig. 6 G and Video 7), it appeared to be less robust, and network assembly did not correlate with cell area contraction as in WT eyes (Fig. 6 G′).

Finally, we examined the interaction between RhoGAP71E and RhoGEF2 by coexpressing the two proteins in the retina. While RhoGAP71E expression alone expanded the apical cell area and led to cell intercalation defects, co-expressing RhoGEF2 with RhoGAP71E suppressed these phenotypes (Fig. S4). Together, these findings provide evidence that RhoGAP71E modulates medioapical actomyosin dynamics and tension by inhibiting Rho1 activity medioapically and antagonizing RhoGEF2 function. Thus, RhoGAP71E and RhoGEF2 regulate the Rho1 GTPase cycle and the overall levels of Rho1 activity to control medioapical actomyosin assembly and contractility, which govern cell shape and arrangement. Moreover, they regulate the kinetics of pulsing shape changes essential for morphogenesis.

To further investigate how cells’ mechanical state affects Rho1 activation, we used the Rho1 binding domain of Anillin tagged with GFP (AniRBD::GFP) to visualize the active GTP-bound Rho1 during cell area pulsing (Munjal et al., 2015). In WT cells, this sensor accumulated junctionally and medioapically across the apical cell area and at LC–LC contacts (Fig. 7 A and Video 8). In Rho1-overexpressing retina, this sensor accumulated to high levels and formed aggregates preferentially in LCs (Fig. 7 B). Therefore, we examined the reporter’s dynamics relative to cell area fluctuations in cone cells where it minimally aggregated. In WT cone cells, cell contractions and changes in AniRBD were subtle and correlations peaked at −30 s before maximal contraction (Fig. 3 C). In Rho1-overexpressing retina, we found that during the early expansion of cone cells, Rho1 activation was very low in the entire apical region. During mid to late expansion, Rho1 activation gradually increased first at the cell periphery and then also medioapically (Fig. 7 B, right panel, blue and yellow arrowheads, respectively). Finally, during contraction, Rho1 activation continued to increase junctionally and medioapically, concentrating in puncta (blue arrowhead). Although peak correlation was at 10 s after peak contraction, AniRBD began to accumulate before peak expansion. Time correlation analysis revealed a strong correlation with contraction in Rho1-expressing cells (Fig. 7 C). The onset of Rho1 activation during apical area expansion suggests that low tension triggers Rho1 activation to promote contraction and counterbalance cell area expansion.

We also examined the localization of the Rho1 inhibitor RhoGAP71E::GFP in WT and Rho1-overexpressing LCs to determine how the cells’ mechanical state relates to the onset of Rho1 inhibition. While in WT eyes, RhoGAP71E::GFP accumulated in pulses that correlated weakly with cell area contraction (Fig. 6, A and A′), in Rho1-overexpressing cells, it accumulated in strong pulses that correlated strongly with cell area contraction (Fig. 7, D and D′, red arrowheads and brackets). During cell expansion, RhoGAP71E::GFP accumulated medioapically at low levels (bottom panel, green arrowheads, and brackets), while during late contraction, it accumulated strongly medioapically in punctate structures that then dissipated during early apical area expansion. Strong RhoGAP71E accumulation during apical area contraction suggests that high tension promotes RhoGAP71E recruitment to the apical cell area to inhibit Rho1 and counterbalance contraction.

Overall, our findings show that biochemical and mechanical interactions orchestrate medioapical contractile pulses that shape the Drosophila eye. Rho1 activity affects actomyosin contractility and, thereby, the mechanical states of the cells. In turn, these mechanical states feedback to affect Rho1 signaling (Fig. 7 E). Our discoveries indicate that both molecular and mechanical feedback loops coordinate the mechanical behavior of neighboring cells to rebalance forces in the epithelium and promote tissue remodeling.

During epithelial remodeling of the fly retina, at least two dynamic actomyosin networks operate at the apical region of the cells. The first is a junctional network that cyclically contracts and expands the LC–LC contacts (Del Signore et al., 2018; Malin et al., 2022). Here, we describe a second mechanism essential for accurate morphogenesis: a non-ratcheting pulsatile medioapical actomyosin network controlled by Rho1 and its effectors Dia and MyoII. This network is established by flows of F-actin from the cell surface toward the medioapical region where it assembles into actomyosin nodes that generate anisotropic contractile force with the highest force predicting the cell edges that will eventually shorten. This force is transient, and upon release, the actomyosin nodes remodel, disassemble, or merge with the junctional network. Force generated by contractile actomyosin is transmitted between cells, and the loss of force from a neighbor often prompts medioapical actomyosin assembly after initial disassembly and cell area relaxation. More strikingly, pulses of medioapical contraction and release are reciprocally synchronized between adjacent 2° LCs, revealing that these Rho1-regulated networks are integral to biomechanical feedback coupling. That is, they appear to trigger and be triggered by forces between neighboring cells and coordinate their activity. Further supporting this conclusion, we found activation of Rho1 as cells expanded and negative Rho1 regulator RhoGAP71E accumulated as cells contracted. Finally, we note that all manipulations that disrupted the kinetics of medioapical pulsing resulted in developmental errors. These observations expand our understanding of these players’ roles in embryonic undifferentiated tissue development to a fully differentiated tissue where they work over different time scales to control cell shape determination.

This is the first investigation of the pulsing features and regulation of medioapical actomyosin networks in retina LCs. Overexpressing Rho1 dramatically revealed that the induced rapid cell area oscillations were inversely coordinated in neighboring 2° LCs: one cell contracted while its neighbor relaxed and vice versa. Although it is known that cells can sense and adapt to mechanical forces (del Rio et al., 2009; Gaertner et al., 2022; Mueller et al., 2017; Spadaro et al., 2017; Yonemura et al., 2010), the current observations expand our understanding of how forces work in this system. Our findings suggest that neighboring LCs and the medioapical actomyosin networks do not merely strike a balance of mechanical forces but are continually adapting to changing forces over time, generating coordinated pulsing changes in cell shape. This is supported by our observation that laser ablation of the medioapical actomyosin network in one cell was often associated with a subsequent actomyosin accumulation and contraction of the neighboring cell following an initial relaxation. More specifically, our results provide evidence that mechanical cell-to-cell interactions can activate or inhibit Rho1 activity and actomyosin contractility and that cells respond to tension and stretching under normal physiological conditions by modulating Rho1 function. Thus, Rho1-dependent medioapical actomyosin networks appear integral to feedback loops that coordinate pulsing in neighboring cells.

Not only was anti-correlated pulsing between neighboring cells detected with Rho1 overexpression, but it was also apparent in WT eyes (Video 1). However, the correlation between actomyosin accumulation and cell area contraction in WT cells had a wide distribution (Fig. S1, B″ and C′). This suggests that the pulsing is highly tuned to work in balance with other ongoing cellular processes that may provide permissive conditions, such as the net force exerted by neighboring cells. The anti-correlated pulsing of neighboring 2° LCs is a unique feature contrasting with other examples of pulsatile contraction. For example, in cell intercalation during germband extension, adjacent cells constrict simultaneously, resulting in loss of contacts and rearrangement of the epithelium (Fernandez-Gonzalez and Zallen, 2011; Rauzi et al., 2010; Sawyer et al., 2011; Vanderleest et al., 2018). In apical constriction, it has been observed that when one cell undergoes ratcheting contraction, ratcheting of nearby cells becomes more likely to stabilize changes in cell shape and facilitate invagination (Xie and Martin, 2015). Likewise, the late phase of dorsal closure involves collective apical area contraction of amnioserosa cells. In contrast, the early phase exhibits preferential anti-correlated changes in cell shape and medial myosin accumulation in neighboring cells. RhoGEF2 and Rho1 affect these dynamics, with Rho1 acting both autonomously and nonautonomously (Azevedo et al., 2011; Saravanan et al., 2013; Solon et al., 2009), suggesting that the mechanism we discovered may be used in other contexts. Overall, these findings suggest that although Rho1-regulated medioapical actomyosin constriction generally appears sensitive to forces exerted across the epithelium, the cell’s specific response to forces imposed by neighbors is a control point with a variable outcome.

Together, these findings indicate that Rho1 dynamics must be tightly controlled in space and time to regulate actomyosin dynamics and epithelial morphogenesis. Using a genetic screen, we identified RhoGEF2 and RhoGAP71E as upstream regulators of Rho1 whose perturbation yielded altered cell shape and arrangement while the integrity of the adherens junctions was maintained. Although we previously identified these genes in a screen for epithelial folding in leg joint morphogenesis and they are both known regulators of Rho1 in other systems, we were surprised to identify these genes because the characteristics of the medioapical actomyosin network in the LCs are unique (Azevedo et al., 2011; Denk-Lobnig et al., 2021; Fox and Peifer, 2007; Greenberg and Hatini, 2011; Häcker and Perrimon, 1998; Mason et al., 2016; Mulinari et al., 2008). In these other cases involving RhoGEF2 and RhoGAP71E, there is typically ratcheting shape change, irreversible loss of contacts, and/or epithelial folding, none of which characterize changes in the retina at this phase of development. Thus, identifying RhoGEF2 and RhoGAP71E in this study is important because it reveals that their participation is not linked to a particular morphological outcome. Since pulsing appears regulated by forces imposed by neighboring cells, our results also suggest the hypothesis that mechanical forces could, in turn, regulate RhoGEF2 and/or RhoGAP71E. This idea is supported by the observations that Rho1 is activated in expanding cells under low tension and RhoGAP71E is recruited to the medioapical region in contracting cells under high tension (Fig. 7 E). A related feedback mechanism may control vesicle secretion in Drosophila larval salivary glands, where Rho1 initiates actin coat assembly and vesicle contraction. In response, RhoGAP71E accumulates on vesicles and inhibits Rho1, leading to coat disassembly and termination of the contraction cycle (Segal et al., 2018). In sum, these results provide a starting point for the revelation of the regulatory networks that guide these intermediaries to respond to cell’s mechanical state.

Furthermore, our data indicate that tuning of pulsing dynamics by Rho1 and its regulators is crucial for the cells to maintain their shape and position in the lattice and prevent catastrophic tissue ruptures during this period of eye development. However, it is interesting to note that with constitutively active MLCK, we continued to observe pulsing accumulation of MyoII, even though the tissue architecture was highly disrupted. This observation is consistent with the idea that a core of pulsing dynamics may arise from the intrinsic biophysical properties of contractile actomyosin networks, wherein as F-actin filaments are exposed to high tension, they spontaneously disassemble (Haviv et al., 2008; Munjal et al., 2015). However, our results would also support the idea that pulsing characteristics are finely tuned to work in balance with other processes occurring in the epithelium to maintain tissue structure and generate the appropriate developmental pattern.

A salient feature of retinal development is the continual movement and shape changes of cells, even though it takes hours for persistent changes to unfold. Our results suggest that continuous cycling of the nucleotide state of the Rho1 GTPases drives Rho1-dependent regulation of the cytoskeleton that underlies components of these movements. Thus, these are active, energy-dependent processes that cells invest in, despite the apparent transience of shape changes seen over the short term. There are various explanations for why medioapical actomyosin pulsing exists in other systems, for example, to enable apical constriction by ratcheting or to utilize the relaxation phase to remove membrane and junctional proteins (Jewett et al., 2017; Miao et al., 2019). Here, in the Drosophila retina, it appears medioapical actomyosin pulsing coordinates the mechanical behavior of neighboring cells and functions to maintain mechanical tissue integrity. Previously, we described oscillations of junctional networks of LC–LC contacts that occurred with a periodicity of about 15 min (Del Signore et al., 2018; Malin et al., 2022). Here, pulsing of medioapical actomyosin networks has a periodicity closer to 6 min. While we do not yet know how these processes interplay, it is intriguing to note the existence of multiple sinusoidal functions that could distribute mechanical forces and speculate on how their actions are eventually integrated to generate the precise shaping of the eye.

In summary, we identified a pulsatile medioapical actomyosin network in LCs that exerts tension on the apical cell surface in a pattern that predicts the final shape changes of the cells. This network assembles into a ring of nodes connected by F-actin filaments that exert tension on one another and the cell surface. RhoGEF2 activates Rho1, actomyosin assembly, and cell area contraction, while RhoGAP71E inhibits Rho1 and promotes actomyosin disassembly and cell area expansion. Low tension in expanding cells triggers Rho1 activation and cell area contraction while rising tension in contracting cells triggers RhoGAP71E recruitment, Rho1 inhibition, and cell area expansion. Contraction of one cell promotes actomyosin disassembly and area expansion of neighboring cells, which in turn triggers Rho1 activation and contraction. Together, these intracellular and cell-to-cell interactions manifest in inversely synchronized cell-to-cell oscillations of actomyosin and apical cell area (Fig. 7 E). Our research suggests that during epithelial remodeling, when the cells are engaged in the pursuit of stable forms, finely tuned pulsing of medioapical actomyosin networks functions in balancing the forces between epithelial cells. Regulated by mechanical and biochemical signals, Rho1 plays a crucial role in this delicate balancing act.

We employed a RhoGEF2::GFP BAC transgene inserted at the VK33 attP transgene landing site to examine RhoGEF2 dynamics during LCs remodeling (Mason et al., 2016). We expressed UAS-RhoGEF2 protein with an N-terminal T7 tag (RRID:BDSC_9387) to examine the RhoGEF2 overexpression phenotypes, and FRT42D rhogef2^e037764 (Kyoto-114511) to generate genetically marked mutant clones. We employed a RhoGAP71E::GFP CRISPR insertion (Denk-Lobnig et al., 2021) and RhoGAP71E::GFP driven by the ubiquitin promoter (this study) to examine RhoGAP71E protein distribution and dynamics, a WT UAS-RhoGAP71E (this study) to examine overexpression phenotypes, and RhoGAP71E^j6B9 FRT80B (Kyoto-111395) to generate genetically marked mutant clones.

Fly lines from the Bloomington Drosophila Stock Center: (1) Rho1::GFP (RRID:BDSC_9528), (2) UAS-Rho1.N19 (RRID:BDSC_58818), (3) UAS-Rho1 (RRID:BDSC_28872), (4) UAS-MLCK.Ct (RRID:BDSC_37527, 37528), (5) UAS-Dia.CA (RRID:BDSC_27616), (6) UAS-T7.RhoGEF2 (RRID:BDSC_9387), (7) UAS-Lifeact::Ruby (RRID:BDSC_35545), (8) sqh-GFP::Rok (RRID:BDSC_52289), (9) GMR-GAL4 (RRID:BDSC_9146, 84247), (10) y w; Actin>y+>GAL4, UAS-GFP (RRID:BDSC_4411), (11) w; Actin>CD2stop>GAL4, UAS-mRFP (RRID:BDSC_30558), (12) y w; α-Cat::GFP^gfstf (RRID:BDSC_59405), (13) y w hsFLP; UAS-GFP^nls (RRID:BDSC_9431), (14) w; tub-GAL4; FRT40A, tub-GAL80^ts (RRID:BDSC_86315), (15) w; FRT80B, Ubi-GFP (RRID:BDSC_1620), (17) UAS-RhoGEF2-RNAi (RRID:BDSC_34643), and (18) UAS-RhoGAP71E-RNAi (RRID:BDSC_32417). α-Cat::Venus^CPTI002596 (115551) and RhoGAP71E^j6B9 FRT80B (111395) and FRT42D RhoGEF2^e037764 (114511) were obtained from the Kyoto Stock Center. Fly lines generated in the lab for Del Signore et al. (2018) and Malin et al. (2022): (1) UAS-Lifeact::Ruby; GMR-GAL4, (2) UAS-Lifeact::GFP; Sqh-Sqh::mCherry, GMR-GAL4, and (3) UAS-Lifeact::Ruby; Sqh-Rok::GFP, GMR-GAL4. Additional stocks used: (1) Sqh-Sqh::mCherry, (2) Sqh-UtrABD::GFP, Sqh-Sqh::mCherry (Martin et al., 2009), (3) GMR-GAL4, UAS−α-Cat::GFP (Larson et al., 2008), (4) RhoGAP71E::GFP (Mason et al., 2016), (5) RhoGEF2::GFP (Mason et al., 2016), and (6) AniRBD::GFP (Munjal et al., 2015).

We generated several new reporters driven by the UAS and the ubiquitin promoter, including Ubi-RhoGAP71E::GFP and UAS-RhoGAP71E::GFP. The RhoGAP71E open reading frame (ORF) was amplified from cDNA clone LD04071. The PCR products were inserted into the pENTR plasmid by Topo cloning. All expression clones were generated by the Gateway technology using the Drosophila Gateway Vector Collections (obtained from the Drosophila Genomics Resource Center) using the Clonase II reaction to fuse the ORFs in frame with a desired fluorescent protein. Transgenic flies carrying these transgenes were established by standard methods by BestGene, Inc.

GMR-GAL4 was used to broadly express UAS-transgenes in the eye (Wernet et al., 2003). The FLP/FRT (Xu and Rubin, 1993) and Mosaic Analysis with a Repressible Cell Marker techniques (Lee and Luo, 2001) were used to generate genetically marked clones by FLP-mediated mitotic recombination and the FLP-Out/GAL4 technique to express desired transgenes in genetically marked clones. hsFLP; Ubi-GFP, FRT80B was used to generate RhoGAP71E mutant FLP/FRT clones, and hsFLP; FRT42D Ubi-GFP was used to generate the RhoGEF2 mutant clones. Mitotic and FLP-Out clones were induced by a heat shock for 30 min at 34°C.

White prepupae (0 h after puparium formation or APF) were selected and aged on glass slides in a humidifying chamber at 25°C. Pupal eyes were dissected in phosphate-buffered saline (PBS), fixed for 35 min in 4% paraformaldehyde in PBS, and stained with antibodies in PBS with 3% BSA, 0.3% Triton X-100, and 0.01% sodium azide. Primary antibodies used were rat anti-E-cad (1:100, #DCAD2; DSHB), mouse anti-Dlg (1:500, #4F3; DSHB), and guinea pig anti-Sqh1P (Zhang and Ward, 2011). The following ThermoFisher conjugated secondary antibodies were used at 1:150: Alexa 405 (RRID:AB_221604), Alexa488 (RRID:AB_143165, AB_2534069), and Alexa647 (RRID:AB_2535805). The following Jackson ImmunoReaserch conjugated secondary antibodies were used at 1:150: Cy3 (RRID: AB_2632516, AB_2632516) and Cy5 (RRID: AB_2338713, AB_2338713).

Flies were selected for imaging at 0 h APF and aged at 25°C unless described otherwise. Pupae were mounted in a slit created in an agarose block with eyes facing the coverslip after the operculum for each pupa was removed. The agarose block was surrounded by a Sylgard 184 gasket prepared in the lab and capped with a custom-built humidified chamber. Time-lapse imaging was performed on a Zeiss LSM800 inverted laser scanning confocal microscope with Airyscan. Images were taken every 5, 30, and 60 s as noted with an optimal pinhole (1 AU) using a 63×, 1.4-NA, plan Apochromat immersion objective, 0.7 μm per optical section with a 10–50% overlap between sections, at a scan speed of 6–7 with no averaging in Zeiss Zen Blue 2.6 software.

Laser nano-ablations to probe tissue mechanics were performed using a Zeiss LSM880 NLO with a near-infrared InSight X3 two-photon tunable laser (680–1,300 nm) using a 740 nm multiphoton excitation. Samples were imaged with Utr::GFP and Sqh::mCh to identify the medioapical actomyosin network for ablation using a 63× oil immersion objective (NA 1.4). A small field size of 4 × 4 pixels was ablated in the center of either 2° or 3° LCs using a 30–40% laser output, at a scan speed of one, with one iteration. Images were collected every second for 250 s in a frame of 1,584 × 1,584 pixels to calculate cell recoil velocity or for 10 min to examine cell-to-cell mechanical interactions in Zeiss Zen Black 2.3 software. Ablations were performed 6 s after the beginning of each experiment.

The areas of 2° and 3° LCs were measured in Fiji immediately before and 180 s after ablation to calculate the fold expansion of the apical cell area. Ratios were calculated in Microsoft Excel by dividing the area after ablation by the area before ablation. LC–LC (2°-3°), 2°-1°, and 3°-1 contact lengths for both 2° and 3° LCs were also measured before ablation and 180 s after ablation. Ratios were plotted in GraphPad Prism 9 and compared with normalized contact length before ablation.

We measured the mean signal intensity of F-actin (using Utr::GFP), MyoII (using MyoII::mCh), RhoGEF2::GFP, RhoGAP71E::GFP, and apical cell area at 28 h APF in time-lapse movies with a time resolution of 5 s. Using the Fiji polygon selection tool, we manually traced individual apical perimeters of 2° LCs. Regions of interest (ROIs) were generated, and minimum, maximum, and mean average signal intensities were obtained for each cell and normalized against the background. Using a custom R script, we measured the time-shifted Pearson’s cross-correlations (time windows from ±8 min) between the mean signal intensities of the fluorescent reporter and cell area using 20–30-min-long movies. The data were smoothed using an “R” kernel regression smoother and presented as the average Pearson’s correlation of individual cells and the standard error of the mean. A one-sample t test was performed on the average changes in R-values for apical cell area versus RhoGEF2::GFP, apical cell area versus RhoGAP71E::GFP, apical cell area versus Utr::GFP, apical cell area versus Sqh::mCh, and apical cell area versus MyoII in Act>RhoGAP71E RNAi- and Act>RhoGEF2 RNAi-expressing cells with a theoretical mean of 0. Pulse duration was obtained after plotting the apical cell area against time. Distances between the maxima were obtained and averaged. Amplitudes were obtained by calculating the distance between the maxima and minima to obtain an average amplitude or fold-change of signal intensity. For the cell-to-cell comparisons, R-values were generated for the correlation between areas of adjacent 2° LCs and between the area of one cell and actin intensity in the neighboring cell. R-values were averaged and a one-sample t test was performed against a theoretical mean of 0. To calculate the initial area recoil velocity (V₀), we measured the area of the 2° LCs after ablation for a period of 60 s in Fiji, which changed linearly during this time interval. Subsequently, we plotted the area against time in GraphPad Prism 9 and computed the V₀ by averaging the slopes of the fitted lines.

LCs were selected for analysis if all adjacent cells expressed the RNAi, or, for the control group, if all adjacent cells were WT. For each 2° LCs, circular ROIs were used to measure the signal intensity at the medioapical region from which the background was subtracted to calculate the average signal intensity in RhoGAP71E RNAi or RhoGEF2 RNAi compared with WT LCs. A t test for normally distributed data and a Mann–Whitney test for non-normally distributed data were used to compare the groups.

Fig. S1 shows medioapical actomyosin network assembly imaged at a high spatial and temporal resolution that negatively correlates with cell area changes. Fig. S2 shows the reassembly of medioapical F-actin network after ablation. Fig. S3 shows negatively correlated cytoskeletal and mechanical coupling between anterior and posterior cone cells and 1° cells. Fig. S4 shows that RhoGEF2 overexpression rescues cell shape and rearrangement defects induced by RhoGAP71E overexpression and RhoGEF2 RNAi and clonal phenotypes. Fig. S5 shows the recoil velocity of 2° LCs that overexpress RhoGAP71E, RhoGEF2, and Rho1 compared with WT. Fig. S6 shows that ablation of medioapical actomyosin in 2° LCs preferentially expands LC–LC contacts, while ablation in 3° LCs preferentially expands 3°-1° contacts. Video 1 shows medioapical F-actin network dynamics in WT LCs with nodes forming a ring and then fusing or remodeling. Video 2 shows anisometric apical area relaxation of 2° and 3° LCs after laser ablation of their medioapical actomyosin network. Video 3 shows that Rho1 overexpression accelerates medioapical F-actin dynamics and cell area fluctuations of LCs compared with WT. Video 4 shows that medioapical actomyosin network assembly and cell area fluctuations are inversely coordinated between neighboring 2° LCs. Video 5 shows that constitutive F-actin and MyoII assembly disrupt medioapical actomyosin organization and dynamics. Video 6 shows that RhoGEF2 promotes medioapical F-actin and MyoII dynamics in LCs. Video 7 shows that RhoGAP71E inhibits medioapical F-actin and MyoII dynamics in LCs. Video 8 shows that in Rho1-expressing eyes, Rho1 is initially activated in expanding cells through the use of the AniRBD::GFP Rho1 sensor, whereas the Rho1 inhibitor RhoGAP71E tagged with GFP (RhoGAP71E::GFP) accumulates in contracting cells.

All data needed to evaluate the conclusions in the paper are present in the article and/or the supplementary materials. Key image data were deposited in the Tufts University Dataverse with the DOI https://doi.org/10.7910/DVN/LGL8PS. Image data and other supporting data of this study are available from the corresponding author upon reasonable request. Requests for reagents should be directed to and will be fulfilled by the lead contact, Victor Hatini ([email protected]).

We thank J. Treisman (New York University, New York, NY), A. Martin (Massachusetts Institute of Technology, Cambridge, MA), and J. Zallen (Memorial Sloan Kettering Cancer Center, New York, NY) for generous gifts of flies, the Bloomington Drosophila Stock Center, the Vienna Drosophila Research Center, and the Kyoto Stock Center for flies, the Developmental Studies Hybridoma Bank, R. Ward (Case Western Research University, Cleveland, OH), and J. Treisman for generous gifts of antibodies, and G. Rong and the Institute for Chemical Imaging of Living System at Northeastern University for assistance with multiphoton imaging. We thank K.G. Commons for her critical reading of the manuscript and editorial suggestions and Paul Hatini (Roche Pharmaceutical, Allston, MA) for writing the R script to perform the time-resolved Pearson’s correlation analysis and plot the results.

This work was supported by a grant from the National Institutes of Health to V. Hatini (R01 GM129151).

Author contributions: Conceptualization: V. Hatini; Investigation and Resources: C. Rosa-Birriel, J. Malin, and V. Hatini; Formal Analysis: C. Rosa-Birriel; Writing Original Draft: V. Hatini; Manuscript Review and Editing: C. Rosa-Birriel, J. Malin, and V. Hatini; Supervision: V. Hatini; Funding Acquisition: V. Hatini.