Insulated Switches: Dual-Function Protein RalGEFRGL-1 Promotes Developmental Fidelity

The C. elegans vulva is an excellent model for the study of developmental biology and cell–cell signaling. The developmental induction of vulval precursor cells (VPCs) to assume the 3°-3°-2°-1°-2°-3° patterning of cell fates occurs with 99.8% accuracy. During C. elegans vulval development, an EGF signal from the anchor cell initiates the activation of RasLET-60 > RafLIN-45 > MEKMEK-2 > ERKMPK-1 signaling cascade to induce the 1° cell. The presumptive 1° cell signals its two neighboring cells via NotchLIN-12 to develop 2° cells. In addition, RasLET-60 switches effectors to RalGEFRGL-1 > RalRAL-1 to promote 2° fate. Shin et al. (2019) showed that RalGEFRGL-1 is a dual-function protein in VPCs fate patterning. RalGEFRGL-1 functions as a scaffold for PDKPDK-1 > AktAKT-1/2 modulatory signaling to promote 1° fate in addition to propagating the RasLET-60 modulatory signal through RalRAL-1 to promote 2° fate. The deletion of RalGEFRGL-1 increases the frequency of VPC patterning errors 15-fold compared to the wild-type control. We speculate that RalGEFRGL-1 represents an “insulated switch”, whereby the promotion of one signaling activity curtails the promotion of the opposing activity. This property might increase the impact of the switch on fidelity more than two separately encoded proteins could. Understanding how developmental fidelity is controlled will help us to better understand the origins of cancer and birth defects, which occur in part due to the misspecification of cell fates.


Introduction
The C. elegans vulva is an excellent model in which to study cell-cell signaling. Six specialized equipotent epithelial cells, the vulval precursor cells (VPCs), are arranged along the ventral midline. During the L3 stage, the VPCs are induced to form the pattern of 3 • -3 • -2 • -1 • -2 • -3 • cell fates. The 1 • and 2 • cells go through three rounds of division and proceed to form the vulva; an epithelial tube connects the uterus with the external environment. The 3 • cells divide once and then fuse with the surrounding epithelium. This pattern occurs with 99.8% accuracy [1,2]. A recent study from Shin et al. (2019) revealed that this developmental fidelity is imposed in part by a signaling protein, RalGEF RGL-1 , that encodes dual functions [2]. The study concluded that RalGEF RGL-1 orchestrates opposing 1 • -and 2 • -promoting modulatory cascades to help promote developmental fidelity. However, the study hints at a far more important interpretation of bifunctional RalGEF RGL-1 .
Three models are used to frame our understanding of C. elegans VPC fate patterning: (1) the sequential induction model, (2) the graded signal model, and (3) the mutual antagonism model. In the sequential induction model, induction occurs in a stepwise fashion: the 1 • fate is induced first, then the 2 • fate. The EGF LIN-3 ligand, produced by the anchor cell (AC) in the ventral gonad, activates EGFR  in the nearest VPC, which, in turn, activates Ras LET-60 to trigger activation of the Raf LIN-45 > MEK MEK-2 > ERK MPK-1 canonical MAP kinase cascade to induce 1 • fate [3][4][5] Figure 1). This signal is the core cascade for inducing 1 • fate: each component is necessary and sufficient for 1 • cell fate. Induced 1 • cells synthesize DSL ligands, which via lateral signaling induce the two neighboring VPCs via Notch  to assume 2 • fate [6][7][8]. This second signal is the core cascade for 2 • fate: Notch LIN-12 is necessary and sufficient to induce 2 • fate. The induced presumptive 1 • cell synthesizes redundant Notch ligands to laterally signal Notch  activation to induce neighboring cells to assume 2 • fate. In the presumptive 1 • cell, activation of EGFR LET-23 causes internalization and degradation of Notch LIN-12 . In the presumptive 2 • cells, Notch LIN-12 transcriptional target lip-1 is expressed. lip-1 encodes ERK phosphatase (MKP) to quench ERK activation in presumptive 2 • cells. During vulva fate patterning, Ras LET-60 switches effectors from canonical Raf LIN-45 to non-canonical RalGEF RGL-1 > Ral RAL-1 that promotes 2 • fate. The dual functions of RalGEF RGL-1 are shown with purple arrows, indicating that RalGEF RGL-1 promotes both 1 • and 2 • fate via two modulatory signaling cascades: (1) RalGEF RGL-1 promote 1 • fate via a non-canonical, GEF-independent activity (scaffold for PDK PDK-1 and Akt AKT-1 in the modulatory 1 • -promoting PI3K AGE-1 cascade) (light blue); and (2) RalGEF RGL-1 promote 2 • fate via canonical GEF-dependent activation of Ral RAL-1 (light rose). Necessary and sufficient cascades are in darker blue and darker maroon colors. The graded signal model is based on the observation that the position of isolated VPCs within the EGF LIN-3 gradient can dictate their ultimate fate [9,10]. Furthermore, variation in the dose of EGF  or the activation of EGFR LET-23 modulates VPC fate selection [11,12]. Taken together, these observations indicate that VPCs are influenced by spatially graded EGF LIN-3 . However, two lines of evidence suggest that sequential induction-1 • , then 2 • -takes precedence over the graded signal. First, mutations disrupting either the well-known EGF LIN-3 > Ras LET-60 > MAP kinase or Notch cascades result in absent 1 • or 2 • cells, respectively. Second, while disruption of the EGF LIN-3 > Ras LET-60 > MAP kinase results in a vulvaless phenotype, i.e., no induction of any vulval fates [5], the disruption of Notch  results in missing 2 • cells but the continued presence of 1 • cells [7]: 2 • cells are never observed in the absence of Notch LIN-12 but induction of 1 • fate is not compromised. To restate, 1 • fate can be induced in the absence of 2 • -promoting signal, but 2 • cells cannot be induced in the absence of 1 • -promoting signal. The sequential induction and graded signal models were not reconciled for a long time [13].
We were able to reconcile these two models by discovering the mechanism by which the graded signal is interpreted: effector switching by Ras LET-60 ( Figure 1). Oncogenic Ras has been known for decades to use different oncogenic effectors: Raf, PI3K and RalGEF [14,15]. Of these, RalGEF signaling through Ral and further downstream is poorly understood [16]. We found that during the patterning of VPC fates, Ras LET-60 switches effectors, from Ras LET-60 > Raf LIN-45 promoting 1 • fate to Ras LET-60 > RalGEF RGL-1 > Ral RAL-1 promoting 2 • fate [17]. Ras LET-60 signaling through Raf LIN-45 is the core signaling cascade for inducing 1 • fate. Conversely, Ras LET-60 signaling through RalGEF RGL-1 > Ral RAL-1 is a modulatory signaling cascade that functions in support of the core induction of 2 • fate by Notch  . Apparently, the core Ras LET-60 > Raf LIN-45 (1 • -promoting) and Notch LIN-12 (2 • -promoting) cascades provide the skeleton of VPC patterning, while the graded signaling is overlaid to improve the fidelity of this process [18]. Ral RAL-1 propagates this 2 • -promoting signal by activation of a MAP4K > p38 MAP kinase signaling cascade [19]. The mechanism by which Ras LET-60 switches between utilization of Raf LIN-45 and RalGEF RGL-1 effectors remains unknown. Mirroring the 2 • -promoting Ras LET-60 > RalGEF RGL-1 > Ral RAL-1 modulatory signal in support of Notch LIN-12 , an PI3K AGE-1 > PDK PDK-1 > Akt AKT-1/2 modulatory signal promotes 1 • fate in support of the core Ras LET-60 > Raf LIN-45 1 • -promoting signal [2,20]. Activation upstream of PI3K AGE-1 is unknown: it could be another Ras LET-60 input, or it could be EGFR or InsR DAF-2 . To summarize, to properly pattern the VPCs at high fidelity, each vulval contributing cell fate type-1 • and 2 • -requires a core, essential signaling cascade and a modulatory, inessential signaling cascade.
We have also postulated the mutual antagonism model [21]. In response to initial patterning of VPC fates by the two core signaling cascades, Ras-Raf-MAP kinase and Notch, a series of additional signals and transcriptional changes in signaling molecules occur. Collectively, these changes appear to reinforce initial patterning events and/or reduce conflicting signals within cells. Perhaps VPCs thereby avoid aberrant or ambiguous cells fates that would compromise the function of the vulva and hence decrease reproductive fitness. For example, in presumptive 1 • cells, the Notch LIN-12 receptor is internalized and degraded, thus preventing conflicting 2 • -promoting signals in a VPC that is assuming 1 • fate [22,23]. Conversely, presumptive 2 • cells express ERK LIP-1 phosphatase (DUSP or MAP kinase phosphatase "MKP" in Figure 1) as a Notch LIN-12 transcriptional client gene to quench inappropriate ERK MPK-1 activation in these cells, thus preventing conflicting 1 • -promoting signals in VPCs that are assuming 2 • fate [24]. A series of other regulatory events presumably contribute similarly to preventing contradictory signals, and thus are likely to increase the developmental fidelity of the system (reviewed in [21]).
Here, we reflect on a study that suggests an additional, fourth model that increases developmental fidelity, which we term "insulated switches" in signaling. Specifically, Shin et al. (2019) show that i) RalGEF RGL-1 is a dual-function protein in VPC fate patterning, ii) that the two functions are genetically separable, iii) that RalGEF RGL-1 likely functions as a scaffold for PDK PDK-1 > Akt AKT-1/2 signaling in addition to propagating the Ras LET-60 signal through Ral RAL-1 , and iv) that deletion of RalGEF RGL-1 does not alter the relative balance of 1 • and 2 • fates in VPC patterning but v) increases the frequency of patterning errors 15-fold about the wild-type control [2]. In other words, RalGEF RGL-1 orchestrates the two opposing modulatory signals in VPC fate patterning: Akt promotion of 1 • fate and Ral promotion of 2 • fate. That both activities are embodied in the same signaling molecule raises thought-provoking questions about the functional significance of this concurrence. We speculate that both signals embodied in the same protein represent an "insulated switch," whereby the promotion of one signaling activity curtails the promotion of the opposing activity and vice-versa, perhaps thus increasing the impact of the switch on fidelity more than could two separately encoded proteins. This model also casts the frequently described phenomenon of "pathway cross-talk" in an interesting new light.

Two Antagonistic Functions of RalGEF RGL-1
The dual function of RalGEF RGL-1 was uncovered serendipitously: the deletion of Ral RAL-1 reduced 2 • -promoting signaling while the deletion of RalGEF RGL-1 caused no effect on patterning [2]. These results were paradoxical. Since RalGEF RGL-1 functions as an intermediary between Ras LET-60 and Ral RAL-1 [25], we expect the deletion of either RalGEF RGL-1 or Ral RAL-1 to cause the same signaling defect. A natural hypothesis, eventually confirmed, was that RalGEF RGL-1 plays opposing and genetically separable roles in VPC fate patterning. Answering this question was technically demanding. Since the VPCs induced at the mid-L3 stage comprise approximately one-ten thousandth of the volume of the animal, relatively minute and transient changes in conventional signaling biomarkers like phospho-ERK or phospho-Akt are not detectable. Furthermore, no transcriptional reporters have been described as readouts of the modulatory cascades of Ras LET-60 > RalGEF RGL-1 > Ral RAL-1 or PI3K AGE-1 > PDK PDK-1 > Akt AKT-1/2 during patterning of VPC fates, precluding a quantitative readout of fluorescent reporters of activity of those cascades. Consequently, the authors used phenotypic analysis to dissect these signals, by counting VPCs whose fate were transformed from 3 • to 1 • or 3 • to 2 • [2,17].
To understand these approaches, we need to understand the use of sensitized genetic backgrounds to reveal pathway activity through the use of genetic principles of parallelism and epistasis [26][27][28]. Single mutations abrogating the function of RalGEF RGL-1 or Ral RAL-1 conferred no obvious VPC patterning defects in an otherwise wild-type animal, consistent with a modulatory signal rather than the necessary and sufficient core signals [17,19]. Shin et al. (2019) used a constitutively activating mutation in Ras LET-60 , let-60(n1046gf), which harbors a moderately activating G13E mutation, to provide a "sensitized" genetic background that responds to genetic perturbations of modulatory proteins. let-60(n1046gf) confers 3 • -to-1 • cell fate transformations, resulting in ectopic 1 • lineages (also called "multivulva," or "Muv"). Mutations that reduce the function of genes that act downstream of Ras LET-60 suppress the induction of let-60(n1046gf)-dependent ectopic 1 • cells (e.g., [29,30]). Mutations that increase the activity of the PI3K AGE-1 > PDK PDK-1 > Akt AKT-1/2 cascade increase the frequency of let-60(n1046gf)-dependent ectopic 1 • cells, while a reduction in this cascade function reduces induction of ectopic 1 • cells [20]. In contrast, based on the principle of mutual antagonism -2 • signals are antagonistic to 1 • signals and vice versa -a reduction in 2 • -promoting signals increases the induction of ectopic 1 • lineages [2,17]. Such analyses take advantage of the principles of genetic parallelism: the genetic perturbation of cascades that function in parallel to let-60(n1046gf) 1 • -promoting activity partially enhance or suppress the induction of ectopic 1 • cells, as assayed by counting induced cells [2,17,19]. Shin et al. (2019) were able to unravel this signaling puzzle using these concepts of parallelism and epistasis. First, the authors defined genetic tools to selectively ablate the 2 • -promoting signal of RalGEF RGL-1 > Ral RAL-1 without altering the other properties of these proteins. The R139H missense mutation in Ral RAL-1 and R361Q putative GEF dead mutation in RalGEF RGL-1 selectively disrupt 2 • signaling in the VPCs and enhance 1 • induction in the let-60(n1046gf) background [2,19]. These results indicate that RalGEF RGL-1 promotes 2 • fate via the canonical GEF-dependent activation of Ral RAL-1 . In an experiment that bypassed the GEF requirement of RalGEF RGL-1 with activated Ral, the effect of deletion of RalGEF RGL-1 on the 1 • promoting function could be detected. Therefore, RalGEF RGL-1 performs an additional GEF-independent function that antagonizes its canonical function.
To further investigate the non-canonical GEF-independent activity of RalGEF RGL-1 in promoting 1 • fate, epistasis experiments were performed that point to the deletion of RalGEF RGL-1 suppressing the effect of deletion PTEN DAF-18 and constitutively activated PDK PDK-1 , and partially suppressing the effect of constitutively activated Akt AKT-1 . These results are consistent with the model that RalGEF RGL-1 functions as a scaffold for PDK PDK-1 and AKT AKT-1 in the modulatory 1 • -promoting PI3K AGE-1 cascade. This observation is consistent with mammalian studies showing that mammalian RalGDS physically interacts with PDK and Akt as a scaffold [31,32]. However, these observations are also consistent with RalGEF RGL-1 functioning in parallel to PI3K AGE-1 > PDK PDK-1 > Akt AKT-1 . Thus, RalGEF RGL-1 likely balances the functions of two modulatory cascades: canonical GEF-dependent activation of Ral RAL-1 and promotion of 2 • fate, and non-canonical GEF-independent scaffolding of PDK PDK-1 and Akt AKT-1 to promote 1 • fate.
To circumvent the requirement for a sensitized background, the authors counted patterning errors in the wild type vs. two strains with deleted RalGEF RGL-1 , which requires the scoring of many more animals than used in sensitized backgrounds (1200 per genotype). Strikingly, the baseline error rate in VPC patterning was 15-fold higher in the two deletion mutants than in the wild type (P < 0.0001). Notably, environmental perturbations had no effect on different mutant backgrounds [2]. These observations indicate that the RalGEF RGL-1 function is not to defend against environmental variability but to decrease the variability in the system, i.e., fluctuations in signaling output or "noise".
The rate of patterning errors changed from 0.2% in the wild type to 3.0% upon loss of both RalGEF RGL-1 functions. Most animals still patterned the VPCs correctly using the core signaling cascades of Ras LET-60 > Ral RAL-1 > MAP kinase and Notch LIN-12 . However, this 15-fold change in rate of developmental error reveals the function performed by modulatory cascades in patterning VPCs fates. These cascades likely reinforce initial commitment to fates analogously to the mechanisms of mutual antagonism discussed above. If both activities were encoded separately, a GEF intermediate between Ras LET-60 and Ral RAL-1 on the one hand and a scaffold for PDK PDK-1 on the other, we would never take note of the phenomenon. It is the dual function of RalGEF RGL-1 in the process that draws our attention as unusual. Does the bifunctionality of RalGEF RGL-1 have special mechanistic significance? We next discuss the possible meanings of this phenomenon.

Speculative Model: The Dual Function RalGEF RGL-1 Works as an "Insulated Switch"
We are intrigued by the observation that opposing functions are encoded in the same protein. How is such a phenomenon different mechanistically from the same activities being encoded in two separate proteins, e.g., a signal transducer and a scaffold? In simple linear views of signaling, the phenomenon of the bifunctional protein would not matter. However, we speculate that the observed mechanisms are not a meaningless coincidence. We hypothesize that RalGEF RGL-1 is an "insulated switch": both activities in the same protein are important to reduce signaling noise ( Figure 2).
We envision the insulated switch being a mechanism whereby the activity promoting one signal antagonizes the other signal, and vice versa. We foresee two potential mechanisms by which an insulated switch could reduce signaling noise: (1) cell biological or (2) steric hindrance. In a cell biological mechanism, the putative RalGEF RGL-1 insulated switch is recruited to discrete subcellular compartments for each activity that does not support contribution to the opposing activity. For example, recruitment of RalGEF RGL-1 to the plasma membrane by upstream activated Ras LET-60 may sequester RalGEF RGL-1 away from the subcellular address in which RalGEF RGL-1 would function as a scaffold for PDK PDK-1 and Akt AKT-1 , while scaffolding would sequester away from potential activation by Ras LET-60 . Conversely, in a steric hindrance or physical interference model, Ras LET-60 binding to RalGEF RGL-1 would interfere with scaffolding, while scaffolding blocks RalGEF RGL-1 accessibility to Ras LET-60 , to thereby activate Ral RAL-1 . In the insulated switch model, we imagine that both activities in the same protein increase the fidelity of VPC fate patterning, compared to a situation in which the same activities were encoded in different proteins. RalGEF RGL-1 functions in two modulatory cascades to promote both 1 • and 2 • fate. RalGEF RGL-1 is a scaffold for PDK PDK-1 and Akt AKT-1 in the modulatory 1 • -promoting PI3K AGE-1 cascade (light blue). And RalGEF RGL-1 promotes 2 • fate via canonical GEF-dependent activation of Ral RAL-1 (light rose). A pushmi-pullyu animal is used as a metaphor for RalGEF RGL-1 protein that has dual function in promoting both 1 • and 2 • fate.
In molecular pathways we tend to think linearly; linearity is built into the word "pathway." For instance, we tend to think about RalGEF RGL-1 as only a GEF of Ral RAL-1 and a scaffold is only a scaffold. Yet many larger proteins perform multiple functions. They bind to targets other than those involved in their canonical function. Canonical functions are usually identified accidentally: what did we find first? It is hard to break out of that chronological bias, which leads to a conceptual bias.
These ideas lead us to the concept of "cross talk" in signal transduction. Researchers frequently describe one signaling axis communicating with another signaling axis, and this phenomenon is frequently referred to as "cross talk." However, many such events are orphans, or stand-alone observations that are not followed up on to investigate the biological consequences of the activity. However, this phenomenon is frequently described in the context of the "canonical pathway" vs the novel cross-talk function, which represents an intrinsic bias. We see this in the naming of RalGEF RGL-1 itself: had it been originally discovered as a scaffold for PDK PDK-1 and Akt AKT-1 signaling, its name would be very different than "RalGEF RGL-1 ". If we think about these signaling systems as signaling networks instead of pathways, then these dual-functioning proteins are not unusual, but instead a key feature of signaling. Mechanistically, this dual-functional property of proteins at the network level could encode far more richness into signaling networks. Combinatorial functionality produces far more complexity. How do we generate, or build, one of the most complicated animals on the planet with the same number of genes as we use to generate one of the simplest animals? The answer probably is combinatorial functions of proteins: dual functionality (or more than two: many). Importantly, we also can get increased developmental fidelity with this special feature of the dual-functioning protein.
We speculate about the role of the insulated switch: can a dual-functional protein reduce noise more than two single-function proteins, i.e., a signaler and a scaffold? Can the same protein simultaneously promote one activity while also working to reduce the opposing activity? Figure 2 shows the pushmi-pullyu from the children's story Dr. Doolittle, which serves as a metaphor for the RalGEF RGL-1 protein that can function to promote both a primary and secondary fate. How would we test this? Each activity would need to be separately mutated by creating two different genes in the animal so that they are no longer linked. One is only a scaffold, one only a signaler, but both are expressed at same levels and in same cells. This approach might break the insulated signal in RalGEF RGL-1 , so the signaling activities are still encoded but the switch no longer exists. Since we do not know where the scaffolding activity is encoded in RalGEF RGL-1 , this approach might not be technically feasible.
We further speculate that proteins in the signaling world that can be identified as obviously multifunctional could be insulated switches. One such protein is PLC-epsilon, an isozyme of the phospholipase C (PLC) family that functions both in the transduction of signal and as a shared effector protein in Ras-, Rho-, and heterotrimeric G protein -mediated signaling. Thus, crosstalk between heterotrimeric and small GTPase signaling pathways is sensed and mediated by multifunctional protein PLC-epsilon (reviewed in [33]). Another interesting example is the SOS protein, which provides dual GEF catalytic activity for both Ras and the small GTPase Rac, a member of the Rho branch of the Ras superfamily. The SOS protein contains Dbl homology (DH, also called RhoGEF) and Pleckstrin homology (PH) domains. The tandem DH-PH domain cluster comprises the majority of GEFs for Rho family GTPases. The catalytic DH domain mediates the GDP/GTP exchange on Rho GTPases, while the PH domain modulates the activity of the DH domain (e.g., promoting membrane association, facilitating GTPase subtract binding, or control intramolecular interactions; reviewed in [34])

Conclusions
The patterning of the VPC fate in C. elegans occurs with 99.8% accuracy. Many layers of signaling network orchestrate together to generate high fidelity by minimizing mis-specified or ambiguous fates. This accuracy is generated by the combination of (1) sequential induction, (2) graded signal, (3) mutual antagonism, and (4) the orchestration of two modulatory cascades promoting 1 • and 2 • signaling by the dual functioning protein RalGEF RGL-1 . The study by Shin et al. (2019) revealed this new, latter mechanism by using genetic manipulations, the combined principles of epistasis and parallelism. Cancer often develops from increased noise and error rate, misspecification or variable specification of cells. Birth defects may also be initiated from increased noise during development. Therefore, the signaling networks that reinforce the fidelity of cell fate patterning will provide us with important clues to understand the disease's origins. For organisms with thirty trillion cells, more noise reduction becomes increasingly important. In mammals, there are four RalGEF/RalGDS-encoding genes. Would these operate as insulated switches with four-fold redundancy? There might be more combinatorial properties, such as scaffold specificity for different PDK and Akt isoforms or splice variants. We conclude that insulated switches could provide additional layers of fidelity and buffering of signaling networks against stochasticity during development. Funding: This research was funded by NIH grants R01GM121625 and R21HD090707 to D.J.R.

Conflicts of Interest:
The authors declare no conflict of interest.