|Home | About | Journals | Submit | Contact Us | Français|
Lateral inhibition mediated by Delta/Notch (Dl/N) signaling is used throughout development to limit the number of initially equivalent cells that adopt a particular fate ,  and . While adjacent cells express both Dl ligand and N receptor, signaling between them ultimately occurs in only one direction. Classically this has been explained entirely by feedback: activated N can downregulate Dl, amplifying even slight asymmetries in the Dl or N activities of adjacent cells , , ,  and . Here, however, we present an example of lateral inhibition in which unidirectional signaling depends instead on Dl’s ability to inhibit N within the same cell, a phenomenon known as “cis-inhibition” , , , ,  and . By genetically manipulating individual R1/R6/R7 photoreceptor precursors in the Drosophila eye, we show that in the absence of Dl-mediated cis-inhibition, the normal direction of lateral signaling is reversed. Based on our finding that Dl in R1/R6s requires endocytosis to trans-activate but not to cis-inhibit N, we reexamine previously published data from other examples of lateral inhibition. We conclude that cis-inhibition generally influences the direction of Dl/N signaling and should therefore be included in standard models of lateral inhibition.
Each unit of the fly eye is assembled by reiterative EGF signaling , which simultaneously recruits undifferentiated cells to join the growing cluster of photoreceptor (R) neurons and induces them to transcribe Dl . The R1, R6, and R7 neuron precursors form an equivalence group: those in which N is activated adopt the R7 fate, while those in which N is not activated adopt the molecularly equivalent R1 or R6 (R1/R6) fate  and . During normal development, the first two of the precursors to be recruited receive no Dl signal and therefore adopt the R1/R6 fate. They then redundantly use Dl to activate N in the next recruit, which therefore adopts the R7 fate  and . Because both R1/R6 and R7 precursors co-express Dl and N , we hypothesized that Dl/N signaling among them might be an example of lateral inhibition and that the direction of signaling might simply be biased by the prior expression of Dl in R1/R6s (Figures 1A–1B and S1). If so, then removal of Dl from the R1 and R6 precursors should reverse the direction of signaling. To test this we used the GMR-FLP/MARCM technique to create mosaic adult retinas in which ~11% of R1/R6/R7 precursors were homozygous for a Dl null mutation and all other R cells were wild-type  and . As predicted by our hypothesis, in ommatidia in which both R1 and R6 precursors lacked Dl, the R1 and R6 precursors adopted the R7 fate, and the wild-type R7 precursor adopted the R1/R6 fate, indicating that the direction of signaling was reversed (Figures 1C, 1D, 1F, 1G and S2). In confirmation that indeed N is activated in Dl mutant R1/R6 precursors, we found that they expressed the mδ0.5-lacZ reporter of N activity  and  and did not adopt the R7 fate if their N pathway was blocked (Figure S3). To confirm that the transformation of Dl mutant R1/R6 precursors was caused by their receipt of a Dl signal from the R7 precursor, we examined ommatidia in which all three precursors were homozygous Dl mutant. Indeed, Dl mutant R1/R6 precursors did not adopt the R7 fate if the corresponding R7 precursor also lacked Dl (Figures 1E, 1H, and S2). We therefore conclude that Dl/N signaling between R1/R6 and R7 precursors is an example of lateral inhibition. The classical feedback model predicts that the prior expression of Dl in R1/R6 precursors would downregulate Dl in the R7 precursor, ensuring that the latter cannot signal back and so biasing the direction of signaling (Figure 1F). We hypothesize that the transformation of Dl mutant R1/R6s into R7s was not previously observed because the homozygous clones analyzed included Dl mutant R7s .
We noticed that the redundancy of Dl signaling from R1 and R6 exposed a discrepancy between our results and those predicted by the classical feedback model of lateral inhibition. In particular, when only one of the two R1/R6 precursors in an ommatidium is Dl mutant and the other is wild-type, the feedback model predicts that both R1/R6 precursors should nevertheless adopt the same fate, since they are both exposed to the same R7 source and therefore level of Dl (Figures 2A and 2B). If Dl in the R7 precursor remains sufficiently downregulated despite the partial reduction in N activation, then both the wild-type and the Dl mutant R1/R6 precursor should still adopt the R1/R6 fate (Figure 2A). If, instead, Dl levels in the R7 precursor are sufficiently increased by the partial reduction in N activation, then both R1/R6 precursors should adopt the R7 fate (Figure 2B). However, we found that only the Dl mutant R1/R6 precursor adopted the R7 fate, while the wild-type precursor remained untransformed, indicating that N was activated in the former but not in the latter (Figures 2C, 2D, 2F, 2G and S2). Consistent with this interpretation, the fate transformation of the Dl mutant R1/R6 precursor depended on Dl from the R7 precursor (Figures 2E, 2H and S2), and only the Dl mutant R1/R6 precursor expressed the mδ0.5-lacZ reporter of N activity (Figure S3). Because the only difference between the Dl mutant and wild-type R1/R6 precursors is their ability to express Dl, we conclude that Dl in R1/R6s autonomously represses N pathway function (Figures 2I and 2J). The only known molecular mechanism for this is cis-inhibition: recent work indicates that Dl can bind in cis the same region of N that is bound by Dl in trans, suggesting that trans Dl must outcompete cis Dl in order to activate N . Consistent with this model, we found that overexpressing full-length wild-type N in R1/R6 precursors could overcome the inhibition by cis Dl, causing many R1/R6s to adopt the R7 fate (Figure S4). The simplest explanation for our results is therefore that when one R1/R6 precursor lacks Dl, its N is no longer cis-inhibited and can therefore be activated by Dl expressed in R7, while N in the non-mutant R1/R6 precursor remains cis-inhibited.
While artificial overexpression of Dl can prevent N activation in many contexts , , ,  and , the only previous loss-of-function evidence that cis-inhibition occurs during normal development remains the example of Drosophila wing boundary formation, during which cis-inhibition does not influence the direction of signaling but is instead required to keep all signaling off . The possible role of cis-inhibition in lateral inhibition has been largely ignored (although see , , and ). To test our cis-inhibition model further, we wanted to examine the consequence of eliminating N activity specifically from R7 precursors: while the feedback model predicts that this would cause a reversal in the direction of signaling, causing R1/R6 precursors to adopt the R7 fate, the cis-inhibition model predicts instead that N in R1/R6 precursors would remain cis-inhibited, resulting in all three precursors’ adopting the R1/R6 fate. For technical reasons it is not possible to create N mutant R7 precursors  (and see Experimental Procedures), but Tomlinson and Struhl  used an exclusively repressive form of Su(H) to reduce activity of the N transduction pathway specifically in R7s. Consistent with the cis-inhibition model, they found that 10–20% of such R7s adopted the R1/R6 fate but reported no transformation of the corresponding R1 and R6 precursors into R7s . To confirm and extend this result, we used an alternative strategy, taking advantage of the fact that Dl ligand must be endocytosed in order to trans-activate N, a process that in R1/R6 cells requires the E3 ubiquitin ligase Neuralized (Neur) ,  and . By contrast, a mutant form of the N ligand Serrate that cannot be endocytosed retains the ability to cis-inhibit N when artificially overexpressed . We therefore tested whether by removing Neur from both R1 and R6 we could eliminate Dl’s ability to trans-activate N in R7 without affecting its hypothesized cis-inhibitory activity. The feedback model predicts that the neur mutant R1 and R6 precursors will adopt the R7 fate for the same reason that Dl mutant R1/R6 precursors do so, that is, because Dl in the R7 precursor is no longer downregulated by activated N (Figure 3A). By contrast, the cis-inhibition model predicts that Dl in the neur mutant R1/R6 precursors will still cis-inhibit N, causing all three cells to adopt the R1/R6 fate (Figure 3B). We found that neur mutant R1/R6 precursors never adopted the R7 fate, despite failing to activate N in the R7 precursor (Figures 3C–3E and S2). We therefore conclude that Dl cis-inhibits N even in neur mutant R1/R6s.
We have so far shown that Dl-mediated cis-inhibition is required in the R1/R6 precursors to prevent their transduction of a Dl signal from the R7 precursor. Because R7 does not express Dl until approximately three hours after the R1/R6s have begun to express the transcription factor Seven-up (Svp), previously shown to be necessary and sufficient to specify the R1/R6 fate (Figures 1A and S1) ,  and , we hypothesized that cis-inhibition prevents a reversal of the R1/R6 precursors’ choice of fates. Alternatively, it was possible that loss of Dl from R1/R6 precursors might delay their differentiation or accelerate the R7 precursor’s expression of Dl. To distinguish these possibilities, we examined the timecourse of Dl mutant R1/R6 precursor development. To avoid the complex, pleiotropic effects that are caused by removing Dl earlier in eye development, we used, as before, GMR-FLP to induce mitotic recombination during the final cell division that generates the R1/R6/R7 precursors, resulting in mosaic eye discs in which ~11% of R1/R6/R7s were homozygous Dl mutant but all other R cells were heterozygous. As a consequence of this specificity, however, marker proteins are inherited by and perdure in both mutant and non-mutant cells, preventing us from using conventional labeling techniques to distinguish these genotypes in L3 eye discs. Instead, we deduced the timecourse of Dl mutant R1/R6 fate choice as follows. In mosaic eye discs containing Dl mutant R1/R6/R7 precursors, we found that all R1/R6 precursors initially expressed Svp (Figures 4A and S5). However, once R7s expressed Dl (row “3” in Figures 1, ,4,4, S1, and S3–S6), ~14% of ommatidia contained R1/R6s that instead expressed the R7-specific transcription factor Prospero (Pros)  (Figure 4A; see Figure S5 for an explanation of our quantification). By contrast, in wild-type mosaic discs, all R1/R6 precursors continued to express Svp for approximately 12 hours and never expressed Pros (data not shown). We therefore conclude that Dl mutant R1 and R6 precursors initially select the R1/R6 fate but that, upon exposure to Dl from R7, their N is activated and they instead adopt the R7 fate. Consistent with this interpretation, we find that R1/R6 precursors in Dl mosaic eye discs first expressed the mδ0.5-lacZ reporter of N activity at the same time that the R7 precursors first expressed Dl (row “3”; Figure S3). In transitioning from the R1/R6 to the R7 fate, Dl mutant R1/R6 precursors do not revert to a common precursor state; instead they temporarily expressed both Svp and Pros simultaneously (Figures 4A and S5), a combination of transcription factors that is never observed in any wild-type cell in the eye. These results indicate that Dl-mediated cis-inhibition prevents the trans-differentiation of R1/R6s directly into R7s.
Why does Dl in the R7 precursor not similarly cis-inhibit N? One possibility was that R7 expresses Dl too late to prevent activation of N by Dl in R1 and R6. Alternatively, the timing of Dl expression might be unimportant and instead unknown factors make N in R7 precursors resistant to or Dl in R7 precursors incapable of cis-inhibition. To distinguish these models, we examined whether premature expression of Dl in R7 precursors could cis-inhibit N. As no known promoters drive expression specifically in the R7 precursor prior to recruitment of R1 and R6, we used an insertion of a Gal4-containing P element into the lozenge (lz) locus to drive expression of UAS-Dl approximately simultaneously in all cells within the pool from which R1/R6 and R7 precursors are recruited  (Figure 4B). We found that this premature Dl expression caused many R7 precursors to become R1/R6s (36%, 92/254; Figure 4C) and so conclude that N in R7s can be cis-inhibited. In confirmation that premature rather than increased expression of Dl was responsible, we used PM181-Gal4 to drive expression of UAS-Dl specifically in R7s shortly after their recruitment . Indeed, overexpressing Dl in this way did not prevent R7 precursors from adopting the R7 fate (Figure 4D). We conclude that it is the timing of Dl expression that allows Dl-mediated cis-inhibition to create unidirectional signaling from R1 and R6 to R7 and propose the following model of R1/R6/R7 specification (Figure 4E). The first two of these cells to be recruited by EGF signaling are somehow prevented from receiving or transducing the Dl signal present in R2–R5 and R8 and so express Svp and adopt the R1/R6 fate. EGF also causes these cells to express Dl, which cis-inhibits their N. The next cell to be recruited by EGF is immediately exposed to Dl from R1 and R6 before expressing sufficient Dl to cis-inhibit N. N is therefore activated in this third recruit, which therefore fails to express Svp and instead adopts the R7 fate. While EGF also causes the third recruit to express Dl, the latter cannot trans-activate the already cis-inhibited N in R1 or R6.
While the direction of signaling from the R1/R6 to R7 precursors can thus be explained entirely by ordered Dl expression and its consequent cis-inhibitory and trans-activation effects, it remains possible that other mechanisms including feedback also influence the levels of N activation in R1/R6 and R7 precursors. The antibody mAb323, which recognizes a subset of N targets, has recently been reported to label R1/R6s , suggesting either that cis-inhibition does not completely prevent activation of N by Dl from R7 or that R1 and R6 precursors can transduce a Dl signal that is present before their N is cis-inhibited. One possibility is that R1/R6 precursors can transiently receive the Dl signal expressed earlier by R2–R5 or R8; consistent with this, we note that a small proportion of Dl mutant R1/R6 precursors still adopt the R7 fate even when Dl has been removed from the corresponding R7 precursor (Figure S2). The transcription factor Roughened eye has recently been shown to inhibit transcription of N target genes in R1/R6 precursors and may therefore contribute to protecting R1/R6s from this early source of Dl . We have also found that Dl is transcribed at a higher level in R1/R6 precursors than is reached in R7 precursors (Figure S1A), a difference that cannot be explained by their order of recruitment alone. One possibility is that activation of N in R7 does partially downregulate Dl; indeed, we found that ectopically expressing activated N in R1/R6s and R7s caused a modest decrease in Dl levels in all three precursors, although blocking the N pathway in R7s did not increase Dl levels (Figure S6). Factors that affect the levels of Dl’s cis and trans activities, perhaps by regulating endocytosis, would also be predicted to influence the outcome of signaling between R1/R6 and R7 precursors. In summary, many mechanisms may converge to ensure that signaling during lateral inhibition is unidirectional. We provide the first evidence that ligand-mediated cis-inhibition is one such mechanism.
While including cis-inhibition in theoretical models of lateral inhibition bolsters their ability to generate unidirectional signaling , there has been no previous evidence that cis-inhibition normally affects lateral inhibition. Our results suggested that we might detect the influence of cis-inhibition on other examples of lateral inhibition by comparing Dl and neur loss-of-function phenotypes: if cis-inhibition does not bias the direction of signaling then these phenotypes should be identical. We have found two such published comparisons. First, in the fly eye, Dl in the R3 precursor is normally upregulated in response to positional information and so activates N in the R4 precursor , ,  and . Loss of Dl specifically from the R3 precursor reverses the direction of Dl signaling  and . By contrast, loss of neur specifically from the R3 precursor causes both precursors to become R3s, indicating that N is activated in neither  and . It was previously proposed that Dl within a neur mutant R3 precursor may retain residual activity and so can trans-activate N in the R4 precursor enough to downregulate Dl but not enough to specify the R4 fate  and . However, given our results, a more parsimonious explanation is that Dl normally cis-inhibits N in the R3 precursor. Similarly, presumptive sensory organ precursors (SOPs) in the fly wing use Dl to activate N in surrounding cells and thereby prevent them from also becoming SOPs . While nearly all Dl mutant cells that are adjacent to wild-type SOPs become non-SOPs, indicating that their N has been activated (29/30 = 97%) , a substantial proportion of neur mutant cells adjacent to wild-type SOPs instead become SOPs themselves (9/33 = 27%) , indicating that their N cannot be activated despite being adjacent to Dl signaling cells. Again, the simplest explanation is that their N remains cis-inhibited by Dl. We suggest that cis-inhibition of N by endogenous Dl likely plays a general role in regulating the direction of signaling and should be included in standard models of lateral inhibition.
Delta/Serrate/Lag-2 (DSL) signals are used throughout development and adulthood. Our work highlights the importance of protecting cells from inappropriately transducing the DSL signals that inevitably surround them. The R1/R6 fate choice is tenuous: despite having already expressed the R1/R6-specific transcription factor Svp for approximately three hours, R1 and R6 remain vulnerable to receiving a Dl signal that directly switches their developmental program. This plasticity may be specific to binary fate decisions determined by lateral DSL signaling or may be a common feature of cell fate choices that are specified by the failure to receive a signal. Abnormally high levels of N activation are associated with a variety of cancers as well as other pathologies . Our results suggest that such activation may in some cases be caused by the loss of DSL ligand.
The R1, R6, and R7 precursors are recruited from a pool of equipotent cells generated by the so-called second mitotic wave (SMW) . We used the GMR promoter to express FLP recombinase and thereby induce FRT-site-specific recombination specifically during the SMW, resulting in mosaic animals in which ~11% of R1s, R6s, and R7s were homozygous for a given chromosome arm; any combination of one, two, or three of these cells could be homozygous in a given ommatidium  (data not shown). Homozygous cells were specifically labeled by the MARCM technique using act-Gal4 and UAS-mCD8-GFP, which labels cell bodies but is excluded from photoreceptor rhabdomeres  and ; Gal80 is specifically expressed in non-mutant cells, where it prevents Gal4-driven expression of GFP . Because R1/R6/R7 precursors inherit Gal80 protein from their heterozygous parents, homozygous mutant cells do not express GFP until approximately 12 hours after puparium formation (data not shown); they are therefore unmarked in the experiments depicted in Figures 4A, S3A–S3D, and S5. Because Dl and neur are not transcribed prior to the SMW  and , homozygous mutant R1/R6 and R7 precursors are predicted to lack wild-type protein. However, homozygous N mutant R1/R6/R7 precursors created by GMR-FLP inherit wild-type N from their heterozygous parents, preventing the use of this technique to remove N from R7. We used the DlRevF10 and neur1 null alleles for all analyses presented but found that the DlB2 and neurA101 alleles caused identical phenotypes (Figure S2). Tissues were dissected, fixed, and stained as described previously . Confocal images were collected on a Leica SP2 microscope and analyzed with Leica or ImageJ software.
We thank H. Ruohola-Baker, G. Struhl, E. Lai, S. Bray, A. Martinez Arias, D. Ready, S. Britt, P. Gergen, C. Doe, Y. Hiromi, T. Isshiki, N. Baker, the Bloomington Drosophila Stock Center, and the Developmental Studies Hybridoma Bank for flies and antibodies. We also thank C. Doe, K. Guillemin, J. Eisen, A. Tomlinson and C. Desplan for discussion or comments on the manuscript. This work was funded by a Burroughs-Wellcome Career Development Award to T.G.H. and by an institutional NIH NRSA Genetics Training Grant (5-T32-GM007413) to A.C.M.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.