Toll receptors mediate tissue intrinsic surveillance against aberrant cells by detecting cell fate aberrations
Anna Frey, Laurin Ernst, Friedericke Fischer, Lale Alpar, Yohanns Bellaïche, Anne-Kathrin Classen
TL;DR
Toll receptors help tissues detect and eliminate abnormal cells by comparing neighboring cell identities, linking development and tissue health.
Contribution
Long Toll receptors are identified as mediators of interface surveillance through cell fate comparisons.
Findings
Differences in long Toll receptor levels between adjacent cells trigger interface surveillance.
Long Toll receptors generate a combinatorial cell-surface code regulated by patterning pathways.
Interface surveillance operates independently of NF-κB signaling, relying on actomyosin dynamics.
Abstract
Tissue-intrinsic surveillance systems maintain tissue health by detecting and eliminating aberrant cells. One such mechanism, interface surveillance, is activated by differences in cell fate programs between neighboring cells, leading to actomyosin accumulation, JNK-signaling and apoptosis at these interfaces. Here, we identify long Toll receptors (Toll-2, Toll-6, Toll-7 and Toll-8) as mediators of interface surveillance in Drosophila imaginal discs. Using genetic mosaics and mapping of expression pattern, we show that differences in long Toll receptor levels between adjacent cells are sufficient to induce all hallmarks of interface surveillance. This response relies on the comparison of relative expression levels set by fate-specifying pathways and is thus position dependent. Specifically, long Toll receptor expression is regulated by multiple patterning pathways, generating a…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6
Fig. 7
Fig. 8- —Deutsche Forschungsgemeinschafthttp://dx.doi.org/10.13039/501100001659
- —Boehringer Ingelheim Stiftunghttp://dx.doi.org/10.13039/501100008454
- —International Max Planck Research School for Immunobiology, Epigenetics and Metabolism
- —University of Freiburghttp://dx.doi.org/10.13039/501100002714
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsDevelopmental Biology and Gene Regulation · Invertebrate Immune Response Mechanisms · Immune Response and Inflammation
INTRODUCTION
The maintenance of tissue homeostasis is essential for the survival of multicellular organisms. Evolutionarily conserved surveillance systems protect tissues against different threats, such as the emergence of aberrant or potentially oncogenic cells. ‘Cell-cell competition’, for instance, functions as a tissue-intrinsic quality control mechanism that eliminates aberrant cells with low proteostatic or metabolic fitness (Carpenter and O'Neill, 2024; van Neerven and Vermeulen, 2023; Kawai et al., 2024; Umetsu, 2022). ‘Interface surveillance’, a distinct tissue-intrinsic surveillance system, specifically protects against the emergence of cells with mutations in cell-fate-patterning pathways (Bielmeier et al., 2016; Fischer et al., 2024; Prasad et al., 2023). In Drosophila imaginal discs, mutations in cell-fate-patterning pathways (Dpp/TGF-β, Wg/WNT, Hh/Shh, JAK/STAT and Notch) or in cell-fate-specifying transcriptional regulators (Arm, Iro-C, Omb, Yki, En/Inv, Ap, Ci and Hox genes) consistently induce the recruitment of Myosin II and filamentous Actin to shared contacts between mutant and wild-type cells, leading to the segregation of cell populations in genetically mosaic tissues (Gibson and Perrimon, 2005; Shen and Dahmann, 2005; Widmann and Dahmann, 2009a,b; Pallavi et al., 2012; Gandille et al., 2010; Bessa et al., 2009; Aldaz et al., 2005; Beuchle et al., 2001; Prober and Edgar, 2000; Liu et al., 2000; Worley et al., 2013; Perea et al., 2013; Gold and Brand, 2014; Classen et al., 2009; Bell and Thompson, 2014; Organista and De Celis, 2013; Villa-Cuesta et al., 2007; Shen et al., 2010; Bielmeier et al., 2016; Fischer et al., 2024; Prasad et al., 2023; Klipa et al., 2023; Klipa and Hamaratoglu, 2019; Bosveld et al., 2016). Additionally, activation of c-Jun N-terminal kinase (JNK) signaling in cells on either side of the contact interface drives apoptotic elimination of aberrant cells (Prasad et al., 2023). Importantly, interface surveillance is highly position specific, relying on the spatial expression patterns of cell-fate-specifying pathways within the tissue to detect aberrant cells: clones expressing Cubitus interruptus (Ci) activate interface surveillance only in the posterior wing compartment, where Ci signaling is low, but not in the anterior wing compartment, where Ci is activated by Hedgehog signaling (Bielmeier et al., 2016). This observation, consistent across various cell fate patterns, demonstrates that it is the difference in cell fate between cells, rather than the specific type of cell fate itself, that initiates interface surveillance. A genetic screen from our lab identified that interface surveillance is mediated by the altered expression of at least eight cell surface molecules of the Robo, Ephrin and Teneurin families (Fischer et al., 2024). Accordingly, interface surveillance is triggered by a difference of cell surface molecules between neighboring cells, detecting altered cell surfaces based on the endogenous spatial expression pattern of the respective cell surface molecule. Moreover, expression of Robo, Ephrin and Teneurin families is regulated by cell fate programs, demonstrating how mutations that alter cell fate programs induce interface surveillance.
We were intrigued by another preliminary hit from our genetic screen identifying Toll-8 as a potential cell-surface mediator of interface surveillance. We therefore set out to further dissect the role of the Toll receptor family in interface surveillance defenses against aberrant cells. Toll receptors are recognized for their universal role in innate immunity, where they detect pathogen- or damage-associated molecular patterns leading to the activation of NF-κB transcription factors (Ming et al., 2014; Lemaitre et al., 1996; Minakhina and Steward, 2006; Anderson et al., 1985; Fitzgerald and Kagan, 2020; Anthoney et al., 2018; Imler and Hoffmann, 2001; Brennan and Anderson, 2004; Kawasaki and Kawai, 2014; Xu et al., 2000; Leulier and Lemaitre, 2008). Intriguingly, in Drosophila, Toll receptor function has been characterized beyond innate immunity. Toll receptors act as regulators of embryonic patterning and development, where their function drives precise spatial and temporal coordination of cells during morphogenesis. During germband extension, the expression of Toll-2, Toll-6 and Toll-8 in overlapping transverse stripes provides spatial cues for the planar polarization of actomyosin (Paré et al., 2019, 2014; Paré and Zallen, 2020; Tetley et al., 2016; Sharrock et al., 2022). These studies also suggest that Toll-2, Toll-6 and Toll-8 can do so by homo- and heterotypic interactions on neighboring cell surfaces, or by interaction with non-Toll receptors like Teneurins (Paré et al., 2019, 2014; Lavalou et al., 2021). Downstream, signaling by Toll receptors in these processes may not proceed through classical NF-kB signaling but rather via recruitment of Src, PI3K, GPCRs or Rho GTPases to modulate actomyosin contractility and adhesive properties between neighboring cells (Graham et al., 2019; Paré et al., 2019, 2014; Tamada et al., 2021; Lavalou et al., 2021; Kolesnikov and Beckendorf, 2007). The role of Toll receptors in actomyosin regulation during morphogenesis seems to have been evolutionarily conserved in the last common ancestor of arthropods, and modulation of actomyosin has now also been described for mammalian Toll-like receptors (TLR), highlighting the possibility that the core function of Toll receptor-dependent signaling to actomyosin may be highly conserved (Benton et al., 2016; Peterson et al., 2023). The interplay between Toll receptors, adhesion molecules and cytoskeletal components represents a fundamental paradigm in developmental biology, with potential implications for understanding both normal morphogenesis and pathological conditions.
RESULTS
Toll receptor expression patterns provide a spatial reference map for detection of aberrant cells
To establish a framework through which we could examine a role for Toll receptors in detection of aberrant cells, we first established if and where Toll receptors are expressed in wing imaginal discs. Understanding their precise expression patterns is key to interpret interface surveillance responses caused by a mismatch in expression levels between neighboring cells. In Drosophila, phylogenetic analysis distinguishes three Toll receptor families by the length of their extracellular and cytoplasmic domains: (1) short Toll receptors, consisting of Toll-1, Toll-3, Toll-4 and Toll-5; (2) long Toll receptors, consisting of Toll-2, Toll-6, Toll-7 and Toll-8; and (3) the Toll-9 group (Fig. S1A) (Anthoney et al., 2018; Kanzok et al., 2004; Lima et al., 2021). To investigate expression patterns of these Toll receptors during wing development, we analyzed published single cell RNA-Seq data and screened available reporter lines for expression in wing discs (Floc'hlay et al., 2023). Specifically, we examined enhancer traps for Toll-2, Toll-4, Toll-5, Toll-6, Toll-8 and Toll-9, as well as GFP, YFP or Venus-tagged fusion proteins expressed from endogenous CRISPR-modified Toll-1, Toll-2, Toll-7 and Toll-8 loci (see Table S1) (Li et al., 2020a; Paré et al., 2014; Alpar et al., 2025 preprint; Iijima et al., 2020). We first focused on the long Toll receptors (Toll-2, Toll-6, Toll-7 and Toll-8). Whereas Toll-2, Toll-7 and Toll-8 are transcribed in wing discs, Toll-6 transcription was not observed (Fig. S1B) (Floc'hlay et al., 2023). Accordingly, a GAL4-MIMIC trap in Toll-6 drove expression of a UAS-GFP transgene in the brain but not in wing discs. In contrast, expression of Toll-2, Toll-7 and Toll-8 in wing discs was confirmed by imaging endogenously expressed fluorescent fusion proteins, specifically Toll-2-GFP, Toll-7-Venus, Toll-8-Venus and Toll-8-YFP. This revealed distinct expression patterns for each of the three receptors in the pouch, hinge and notum of wing discs (Fig. 1A-D; Fig. S1C-L). For Toll-2, expression patterns were recapitulated by a Toll-2pTV-GAL4, a Toll-2-LacZ line and previously reported in situ hybridization patterns (Yagi et al., 2010) (Fig. 1A; Fig. S1G,H). Similarly, for Toll-8, a GAL4 enhancer trap mirrored Toll-8-YFP expression patterns (Fig. S1I). Combined, our analysis suggests that the long Toll receptors Toll-2, Toll-7 and Toll-8, but not Toll-6, are expressed during wing disc development in distinct but partially overlapping patterns (Fig. 1K-M), which is consistent with previous studies (Casas-Tintó et al., 2017; Yagi et al., 2010). Importantly, we found that Toll-2-GFP, Toll-7-Venus and Toll-8-YFP localized to adherens junctions and basolateral surfaces of wing epithelial cells – subcellular positions implicated in contact-dependent recognition of cell fates, as well as activation of interface surveillance (Fig. 1E-J).
Toll-2, Toll-7 and Toll-8 are expressed in wing imaginal discs. (A-D) Wing discs expressing Toll-2-GFP (A), Toll-7-Venus (B), Toll-8-YFP (C) fusion proteins or Toll-6-GAL4 driving UAS-GFP (D, cyan/gray), were analyzed 102 h after egg laying (AEL). Patched (Ptc) and Wingless (Wg) (magenta) indicate the A/P and D/V boundaries. Dashed line in D shows the disc outline. (E-J) Subcellular localization of Toll-2-GFP (E,F), Toll-7-Venus (G,H) or Toll-8-YFP (I,J) (cyan/gray) in wing discs. E-cadherin (magenta/gray, top) and F-actin (phalloidin, yellow/gray, bottom) visualize apical-junctional and lateral cell surfaces, respectively. Images are derived from local z projections (see ‘Image display’ section in Materials and Methods). (K-M) Illustrations of Toll-2 (K), Toll-7 (L) and Toll-8 (M) expression patterns in third instar wing discs. Color intensity from lowest (white) to highest (orange) indicates the subjective relative fluorescence intensity. Dashed lines mark A/V and D/V boundaries. Scale bars: 100 µm in A-D; 10 µm in E-J.
We then analyzed expression of short Toll receptors (Toll-1, Toll-3, Toll-4 and Toll-5) and Toll-9. According to the scRNA-Seq atlas (Floc'hlay et al., 2023), only Toll-1 is expressed at high levels in wing discs, and Toll-9 is expressed at very low levels (Fig. S1B). Consistent with this, Toll-1-Venus (Iijima et al., 2020) was ubiquitously and uniformly expressed in wing disc tissue, and localized to cell junctions (Fig. S1M,N). Consistent with the low transcription of all other short Toll receptors in wings, the enhancer and gene-trap GAL4 lines available for Toll-4, Toll-5 and Toll-9 induced expression of a UAS-GFP in the brain but did not drive GFP-expression in the wing disc (Li et al., 2020a) (Fig. S1O-W). These data suggest that among the short Toll receptors, only Toll-1, which has established roles in innate immunity and developmental signaling, is expressed in wing discs, consistent with previous studies (Yagi et al., 2010). Combined, these findings provide a spatial reference map of Toll receptor expression patterns in wing discs, which allowed us to examine the consequences of introducing a mismatch, i.e. difference of Toll receptor expression levels on neighboring cell surfaces, and assess the activation of interface surveillance.
Differences in Toll-6 expression levels induce all hallmarks of interface surveillance in wing imaginal discs
To test the hypothesis that a difference in expression levels of Toll receptors between neighboring cells elicits interface surveillance, we employed the mosaic technique of ‘flp-out’ clones (Fig. 2A,B) (Germani et al., 2018a). This technique either enables the ectopic expression or the RNAi-mediated knockdown of Toll receptors, thereby creating strong differences in Toll receptor expression levels between clonal and neighboring wild-type cells. We monitored characteristic read-outs for the activation of interface surveillance, such as (1) actin enrichment at clone interfaces by phalloidin staining, (2) smoothening of clone interfaces by measurement of clone circularity, (3) JNK activation at clone interfaces by visualization of the JNK-reporters TRE-RFP and puc-LacZ, and (4) induction of apoptosis by staining of the cleaved effector caspase Dcp1. We first analyzed clones overexpressing short Toll receptors (Toll-1, Toll-3 and Toll-4). We did not investigate Toll-5 and Toll-9-overexpressing clones, due to missing genetic tools. Ectopic expression of Toll-1, Toll-3 or Toll-4 did not induce clone smoothening, JNK interface signaling or apoptosis (Fig. S2A-D). Notably, Toll-1 and Toll-3 induced only cell-autonomous activation of JNK in all cells within a clone, consistent with the established cell-autonomous role of at least Toll-1 in innate immune and tissue stress signaling (Fig. S2A′-C′) (Wu et al., 2015; Li et al., 2020b; Tauszig et al., 2000; Shields et al., 2022). We conclude that a mismatch in expression levels of the tested short Toll receptors does not induce cellular responses related to interface surveillance.
Differences in Toll-6 expression levels are sufficient to induce all hallmarks of interface surveillance. (A-D) Mosaic wing discs expressing UAS-GFP only (A,B, cyan/gray) or additionally UAS-Toll-6 (C,D, cyan/gray). F-actin was visualized by phalloidin staining (A,A′,C,C′, gray); JNK-pathway activation was visualized by TRE-RFP reporter activity (B,B′,D,D′, magenta/gray) and apoptosis was visualized by cleaved Dcp-1 (B,B′,D,D′, yellow/gray). Yellow arrowheads in B indicate developmentally patterned JNK-pathway activity at the A/P boundary. White arrowheads in C′ indicate Actin enrichment at clonal boundaries. Outlined areas in A,B,C,D indicate the pouch regions shown in A′,B′,C′,D′. (E) Circularity of clones expressing UAS-GFP (left, N=19 discs, n=53 clones) and UAS-Toll-6 (right, N=24 discs, n=91 clones). Data are mean±95% confidence interval (CI). P-value is indicated in the graph (two-tailed Mann–Whitney U-test). (F,G) Quantification of TRE-RFP mean fluorescence intensity in four regions in and around wild-type clones (F, N=20 discs, n=59 clones) or UAS-Toll-6-expressing clones (G, N=24 discs, n=91 clones). Regions are, from left to right (see Fig. S2E): wild-type background band (gray, WT), outer wild-type band (magenta, WT), inner clone band (magenta, GFP or Toll-6OE) and inner clone area (gray, GFP or Toll-6OE). Data are mean±95% CI. P-values are indicated in the graphs; ns, not significant [Friedman test (F); one-way ANOVA (G)]. Scale bar: 100 µm in A,B,C,D; 50 µm in A′,B′,C′,D′.
Among the long Toll receptors, we first investigated effects induced by ectopic expression of Toll-6, given that Toll-6 is not expressed in wing discs. Consequently, Toll-6-overexpressing clones create a pronounced difference in expression levels compared to surrounding wild-type cells, regardless of their position in the wing disc. Strikingly, the boundaries of Toll-6-expressing clones exhibited a smooth appearance with actin enrichment at clonal interfaces compared to wild-type clones. Additionally, Toll-6 clones displayed significant JNK interface signaling, as well as elevated levels of apoptosis near the interface and in the clone interior (Fig. 2C-G, Fig. S2E-G). Consistent with our reasoning, we found that expression of a Toll-6-RNAi construct in mosaic clones did not induce interface surveillance, as wing discs lack Toll-6 expression (Fig. S2H,I). These data demonstrate that differences in Toll-6 levels activate interface surveillance through the tissue-intrinsic detection of differences in a cell surface receptor.
Differences in Toll-8 expression levels induce interface surveillance – a response dependent on position within the endogenous Toll-8 expression pattern
To understand if long Toll receptors commonly function in interface surveillance, we turned to analyze Toll-8. In contrast to Toll-6, Toll-8 is expressed in a complex expression pattern, with high expression levels in the hinge region and a stripe in the notum (Fig. 3A). Toll-8-expressing clones were recently described to recruit actomyosin to clonal interfaces via GPCR-signaling (Lavalou et al., 2021), yet other aspects of possible interface surveillance responses have not been analyzed. Furthermore, the complexity of the Toll-8 expression pattern opens an opportunity to ask if clones with upregulated or downregulated Toll-8 expression levels would only induce pattern-specific interface surveillance responses in regions where Toll-8 expression is, respectively, low or high, or in other words, whether only a pattern-specific mismatch of Toll-8 can induce interface surveillance.
Differences in Toll-8 expression levels induce all hallmarks of interface surveillance in a Toll-8 expression pattern-specific manner. (A) Illustration of Toll-8 expression in the wing disc. Black outlines illustrate the positions of regions shown in C-H,O-R within the context of the endogenous Toll-8 expression pattern. (B) Wing disc with UAS-GFP-expressing wild-type clones (cyan) (see also Fig. S3A,D,E). (C,E-H) Wing disc with UAS-Toll-8-expressing clones (cyan). Outlined areas in C indicate regions depicted in E-H, where clones are in areas of low (E,F) or high (G,H) endogenous Toll-8 expression. See A for an illustration of Toll-8 expression and positional information (also see Fig. S3B). (I) Circularity of UAS-GFP-expressing clones (left, WT, N=16 discs, n=35 clones) and UAS-Toll-8-expressing clones, analyzed in the pouch (middle, region of low endogenous Toll-8, N=20 discs, n=69 clones) and notum (right, region of high endogenous Toll-8, N=7 discs, n=10 clones). Data are mean±95% CI. P-values are indicated in the graph; ns, not significant (Kruskal–Wallis test). (J,K) Quantification of mean fluorescence intensity of TRE-RFP in four regions in and around UAS-Toll-8-expressing clones in the pouch (low endogenous Toll-8 expression; J) and notum (high endogenous Toll-8 expression; K). Regions are defined as in Fig. 2F,G. (J) N=20 discs, n=69 clones; (K) N=7 discs, n=10 clones. Data are mean±95% CI. P-values are indicated in the graphs; ns, not significant [one-way ANOVA (J); Friedman test (K)]. (L) Circularity of UAS-GFP-expressing clones (left, wild type, N=15 discs, n=49 clones) and UAS-Toll-8-RNAi-expressing clones, analyzed in the pouch (middle, region of low endogenous Toll-8, N=11 discs, n=26 clones) and hinge (right, region of high endogenous Toll-8, N=11 discs, n=38 clones). Data are mean±95% CI. P-values are indicated in the graph (one-way ANOVA). (M,N) Quantification of mean fluorescence intensity of TRE-RFP in four regions in and around UAS-Toll-8-RNAi-expressing clones in the pouch (low endogenous Toll-8 expression; M) and hinge (high endogenous Toll-8 expression; N). Regions are defined as in Fig. 2F,G. (M) N=11 discs, n=24 clones; (N) N=11 discs, n=37 clones. Data are mean±95% CI. P-values are indicated in the graphs; ns, not significant [Friedman test (M); one-way ANOVA (N)]. (D,O-R) Wing disc with UAS-Toll-8-RNAi-expressing clones (cyan). Areas outlined in D indicate regions depicted in O-R showing clones in regions of low (O,P) or high (Q,R) endogenous Toll-8 expression. See A for an illustration of Toll-8 expression and positional information. White arrowheads in Q indicate Actin enrichment at clonal boundaries (see also Fig. S3C). Phalloidin visualizes F-actin (B-D,E,G,O,Q); TRE-RFP (B,F,H,P,R; magenta/gray) visualizes JNK-pathway activity and cDcp-1 (B,F,H,P,R; yellow/gray) visualizes apoptosis. Scale bars: 100 µm in B-D; 25 µm in E-H,O-R.
Indeed, Toll-8-expressing clones located in high Toll-8-expressing domains of the hinge and stripe in the notum did not activate interface surveillance responses, as indicated by the lack of JNK interface signaling, absence of actin recruitment to interfaces and low clone circularity. In contrast, Toll-8-expressing clones in the pouch, which were surrounded by cells with low endogenous expression levels of Toll-8, activated interface surveillance responses, characterized by high JNK interface signaling, high clone circularity and actin recruitment (Fig. 3B,C,E-K; Fig. S3A-H). Importantly, a Toll-8-RNAi construct, which reduced Toll-8-YFP expression levels (Fig. S3I), induced the expected reverse responses. Specifically, Toll-8-RNAi-expressing clones only activated interface surveillance if located in high Toll-8-expressing domains of the hinge or notum, but not if located in low Toll-8-expressing domains of the pouch (Fig. 3A,D,L-R; Fig. S3J). In conclusion, our findings demonstrate that a pattern-specific mismatch in Toll-8 expression levels induces interface surveillance, highlighting the importance of local expression patterns in activating this difference-driven cellular response.
Differences in Toll-2 and Toll-7 expression levels induce interface surveillance redundantly, and this depends on the position within endogenous Toll-2 and Toll-7 expression patterns
To investigate whether pattern-specific activation of interface surveillance is a general feature of long Toll receptors, we analyzed the closely related Toll receptors Toll-2 and Toll-7. We found that Toll-2- or Toll-7-expressing clones strongly activated interface surveillance in wing discs. Specifically, we observed smoothening of clone interfaces, actin enrichment and JNK activation at the clone interface, as well as apoptosis. Importantly, clones expressing Toll-2 or Toll-7 induced the strongest interface surveillance responses in the pouch compared to the hinge, consistent with a pattern-specific response of clones in the context of low endogenous expression levels in the pouch and high endogenous expression levels in the hinge (Fig. 4A-M; Fig. S4A,B).
Differences in Toll-2 and Toll-7 expression redundantly trigger interface surveillance, depending on their endogenous expression patterns. (A) Illustration of Toll-2 (left) and Toll-7 (right) expression in the wing disc. Asterisks indicate pouch-hinge-fold. (B,D) Comparison of circularity between clones expressing UAS-GFP (N=20 discs, n=59 clones) and UAS-Toll-2 (N=23 discs, n=101 clones) (B) or UAS-GFP (N=16 discs, n=37 clones) and UAS-Toll-7 (N=24 discs, n=72 clones) (D). Data are mean±95% CI. P-values are indicated in the graphs (two-tailed Mann–Whitney U-test). (C,E) Quantification of mean fluorescence intensity of TRE-RFP in four regions in and around clones expressing UAS-Toll-2 (C, N=23 discs, n=101 clones) or UAS-Toll-7 (E, N=24 discs, n=72 clones). Regions are defined as in Fig. 2F,G. Data are mean±95% CI. P-values are indicated in the graphs (Friedman test). (F-M) Wing discs carrying clones expressing either UAS-Toll-2 (F-H,L; cyan) or UAS-Toll-7 (I-K,M; cyan). Phalloidin visualizes F-actin (H,K; gray), TRE-RFP visualizes JNK-pathway activity (G,J; magenta/gray) and cDcp-1 visualizes apoptosis (L,M; yellow/gray). Areas outlined in F-M indicates clones shown on the right, representing regions with low (top) or high (bottom) endogenous Toll-2 or Toll-7 expression. White arrowheads in H indicate F-actin enrichment at clonal interfaces. Yellow dashed lines indicate the pouch-hinge fold. See A for illustrations of Toll-2 and Toll-7 expression patterns, and positional context. (N,O) Wing discs with clones concurrently expressing UAS-Toll-2-RNAi and UAS-Toll-7-RNAi (cyan). Phalloidin visualizes F-actin (N; gray) and TRE-RFP visualizes JNK-pathway activity (O; magenta/gray). White arrowheads in N indicate F-actin enrichment at clonal interfaces. Areas outlined in N and O indicate clones shown in N′,N″ and O′,O″, respectively, representing regions with high (N′,O′) or low (N″,O″) endogenous Toll-2 and Toll-7 expression. (P,Q) Quantification of mean fluorescence intensity of TRE-RFP in four regions in and around clones co-expressing UAS-Toll-2-RNAi and UAS-Toll-7-RNAi in the hinge (P, high Toll-2/-7 expression; N=5 discs, n=19 clones) or the pouch (Q, low Toll-2/-7 expression; N=5 discs, n=11 clones). Regions are defined as in Fig. 2F,G. Data are mean ± 95% CI. P-values are indicated in the graphs (one-way ANOVA). Scale bars: 100 µm in lower magnification images; 25 µm in higher magnification images.
We also examined pattern-specific responses in the inverse scenario by expressing Toll-2 and Toll-7 RNAi-constructs, which successfully reduced Toll-2-GFP and Toll-7-Venus expression levels in control experiments. However, we did not observe the induction of interface surveillance responses (Fig. S4C-G). The high degree of sequence similarity between Toll-2 and Toll-7 may allow them to compensate for each other in interface surveillance (Umetsu, 2022), similar to compensation observed for Robo2 and Robo3 receptors (Fischer et al., 2024). Indeed, co-expression of Toll-2 and Toll-7 RNAi-constructs within the same clone induced robust interface surveillance in the hinge, where both Toll-2 and Toll-7 are most highly expressed (Fig. 4N-Q). In conclusion, our findings demonstrate that differences in Toll-2 and Toll-7 expression levels between neighboring cells activate interface surveillance responses.
Differences in expression levels of long Toll receptors between neighboring cells induce recruitment of actomyosin regulators
To confirm that long Toll receptor-mediated interface surveillance is molecularly similar to interface surveillance activated by differences in cell fate between neighboring cells or by differences in the expression of cell surface receptors, such as Robo2 (Bielmeier et al., 2016; Fischer et al., 2024), we analyzed the organization of actomyosin regulators at clonal interfaces. Consistent with these previous studies, we observed that clonal interfaces with differences in Toll-6 or Toll-8 expression levels recruit non-muscle Myosin II (Sqh), the Rho-kinase (ROK) and the Ena/VASP family member Ena (Fig. 5; Fig. S5). These findings recapitulate molecular changes activated by differences in cell fate or in expression of other receptors, such as Robo2 and Robo3 (Bielmeier et al., 2016; Fischer et al., 2024). Importantly, they also invoke the recruitment of Myosin II to heterologous contact sites of cells expressing different long Toll receptors observed during early embryo development, suggesting that regulation of actomyosin may be a core function of long Toll receptors (Paré et al., 2014).
Differences in long Toll receptor expression levels induce recruitment of actomyosin regulators. (A-F) Visualization of non-muscle Myosin II via Sqh-GFP (A,D; cyan/gray), the Rho-kinase Rok via Rok-mNeonGreen (B,E; cyan/gray) or the actin polymerization regulator Ena via antibody staining (C,F) (cyan/gray) in wing discs with UAS-Toll-6-expressing clones (A-C; magenta) or UAS-Toll-8-expressing clones (D-F; magenta). Phalloidin visualizes F-actin (A-F; gray) and anti-E-cadherin staining shows adherens junctions (A-C; gray). Scale bars: 20 µm.
Cell fate-patterning pathways regulate expression of long Toll receptors, establishing a Toll surface code in wing discs
Our results demonstrate that differences in long Toll receptor expression levels activate interface surveillance, thereby phenocopying the observed response to differences in neighboring cell fates. Cell fate programs regulate the expression of hundreds of genes, including those encoding cell-surface proteins (Pope and Medzhitov, 2018; Steinberg, 2007; Tsai et al., 2022). Consequently, Toll receptor genes include regulatory regions extensively bound by transcription factors that are central to cell-fate specification and differentiation (Mod et al., 2010). We thus hypothesized that cell fate-specifying pathways regulate the expression of long Toll receptors, and that aberrant cell fates cause differences in receptor expression levels, which contributes to detection of these cells by interface surveillance. We thus induced mosaic expression of Forkhead (Fkh), a conserved transcription factor required for salivary gland specification, or Eyeless (Ey), a master regulator of eye development, neither of which are normally expressed in wing discs. We also expressed a constitutively active Thickveins (Tkv^CA^) receptor, an activator of Dpp signaling, which is highly active in the central pouch and necessary for wing patterning. Previous studies demonstrate that misexpression of these factors can drive dramatic re-specification of cell fates and alterations of entire gene expression programs in imaginal discs (Halder et al., 1995; Zecca et al., 1995; Tripathi and Irvine, 2022). Importantly, mosaic clones misexpressing Ey, Fkh or Tkv^CA^ strongly activate all hallmarks of interface surveillance in a pattern-specific manner (Bielmeier et al., 2016; Fischer et al., 2024; Prasad et al., 2023). We then asked if expression of Toll-2-GFP, Toll-7-Venus and Toll-8-YFP was altered in these clones. Importantly, we found that Fkh-, Ey- and Tkv^CA^-expressing clones exhibited aberrant up- and downregulation of all three receptors in distinct spatial domains of the disc (Fig. 6, Fig. S6). For instance, Fkh-expressing clones generally downregulate Toll-2 and upregulate Toll-8. Fkh downregulates Toll-7 in the hinge, whereas it upregulates Toll-7 in the pouch (Fig. 6A-C,G-I). In contrast, Ey upregulates Toll-2 and Toll-7 in the pouch, but has no effect on Toll-8 expression (Fig. S6A-D). These observations suggest that even when a transcription factor is not endogenously expressed in a tissue, it can both positively and negatively regulate different Toll receptors in a domain- and receptor-specific manner. This highlights a potential integration between transcription factors, receptor types and spatial patterning, thereby creating combinatorial opportunities for interface surveillance. Supporting this conclusion, we found that Tkv^CA^-expressing clones upregulate Toll-2 and Toll-7 in the notum. Yet, Tkv^CA^ downregulates Toll-2, Toll-7 and Toll-8 in the peripheral pouch, mirroring the low receptor expression normally found in the high Dpp/Tkv-signaling domain of the central pouch, where Tkv^CA^-expressing clones maintain wild-type receptor levels (Fig. 6D-F,J-K′). This suggests that Dpp/Tkv-signaling normally represses expression of Toll-2, Toll-7 and Toll-8 in the central pouch and that coupling endogenous Dpp/Tkv-signaling directly to long Toll receptor expression provides a mechanism by which Tkv^CA^-expressing clones can be recognized outside the high Dpp/Tkv-signaling domain during development.
Cell fate-patterning pathways regulate the expression of long Toll receptors in wing discs. (A-F) Wing discs expressing Toll-2-GFP, Toll-7-Venus or Toll-8-YFP and either UAS-Fkh (A-C) or UAS-TkvCA (D-F) to induce ectopic activation of cell-fate-related gene expression programs. Areas outlined in A-F highlight regions with Toll-receptor deregulation (A′-E″,F″) or no effect (F′). Compare to L-N for changes and to G-K′ for quantifications. (G-K′) Quantification of changes in mean Toll-2-GFP (G,G′,J), Toll-7-Venus (H,H′) or Toll-8-YFP (I,I′,K,K′) fluorescence intensity inside clones expressing UAS-RFP (wild type; WT), UAS-Fkh (FkhOE) or UAS-TkvCA (TkvCA) relative to surrounding wild-type cells in distinct domains (as indicated above the graphs). Data are mean±95% CI. P-values are indicated in the graphs (two-tailed Mann–Whitney U-tests). Sample sizes were as follows: (G) N=8 discs and n=43 clones for WT, N=7 discs and n=37 clones for Fkh; (G′) N=7 discs and n=26 clones for WT, N=8 discs and n=76 clones for Fkh; (H) N=5 discs and n=39 clones for WT, N=8 discs and n=12 clones for Fkh; (H′) N=5 discs and n=99 clones for WT, N=6 discs and n=38 clones for Fkh; (I) N=3 discs and n=13 clones for WT, N=3 discs and n=21 clones for Fkh; (I′) N=3 discs and n=38 clones for WT; N=3 discs and n=61 clones for Fkh; (J) N=7 discs and n=16 clones for WT, N=9 discs and n=46 clones for TkvCA; (K) N=3 discs and n=38 clones for WT, N=3 discs and n=143 clones for TkvCA; (K′) N=3 discs and n=13 clones for WT, N=3 discs and n=42 clones for TkvCA. (L-N) Illustration of Toll-2 (L), Toll-7 (M) and Toll-8 (N) expression in the wing disc. Scale bars: 100 µm in lower magnification images; 25 µm in higher magnification images.
These differences can be explained by the integration of Ey, Fkh or Tkv^CA^ activity into the distinct cell type-specific transcriptional networks of the pouch, hinge and notum, such that the transcriptional context of a cell determines which cell-surface molecule is expressed or not (Beira and Paro, 2016; Ruiz-Losada et al., 2018; Tripathi and Irvine, 2022). We conclude that fate-specifying patterning pathways alter expression of the long Toll receptors Toll-2, Toll-7 and Toll-8, thereby driving complex changes to a ‘cell-surface code’. Given that deregulation of each long Toll receptor is individually sufficient to induce interface surveillance, the combined deregulation of different long Toll receptors by one patterning pathway would facilitate the detection of aberrantly specified cells across various spatial positions in the wing imaginal discs. These observations align with previous findings about at least eight other cell surface receptors, like Robo2, functioning in interface surveillance (Fischer et al., 2024).
Oncogenic Ras but not classical cell-cell competition alters the Toll receptor cell-surface code
Cell fate changes are frequently observed in cancer (Hanahan, 2022). To better understand the relevance of long Toll receptors in the detection of oncogenic cells affected by cell fate changes, we turned to the analysis of oncogenic mutations induced by Ras^V12^. Ras signals through the EGF/ERK pathway and regulates both proliferation and wing fate patterning (De Celis, 2003). Although Ras^V12^-expressing cells are highly resistant to apoptosis, they do induce actin enrichment and JNK signaling at clonal interfaces, suggesting that, while they cannot be eliminated, they are still detected by interface surveillance (Prasad et al., 2023; Fischer et al., 2024; Bosveld et al., 2016). Indeed, Ras^V12^-expressing cells express an aberrant complement of interface surveillance-inducing molecules, such as Robo2 and Ten-m (Fischer et al., 2024). To determine if Ras^V12^-expressing cells additionally alter Toll receptor expression, we monitored Toll-2-GFP, Toll-7-Venus and Toll-8-YFP in Ras^V12^ mosaic discs. Indeed, Ras^V12^-expressing clones de-regulated the expression of all these Toll receptors (Fig. 7A-D; Fig. S7A,B). This observation is consistent with a model where a single fate-specifying aberration can alter the expression of multiple cell surface molecules, thereby driving complex changes to a ‘cell-surface code’.
RasV12 but not classical cell-cell competition deregulates the expression of Toll-7 and Toll-8 in the wing disc. (A,B) Wing discs expressing Toll-7-Venus (A) or Toll-8-YFP (B) and wild-type clones (UAS-RFP, magenta). Patterns are illustrated in A′ and B′, respectively. (C,D) Wing discs with oncogenic UAS-RasV12-expressing clones (magenta) in the background of Toll-7-Venus (C) or Toll-8-YFP (D) (cyan/gray). Areas outlined in C,D highlight regions of Toll-receptor deregulation (C′,D′) or no effect (D″) upon expression of RasV12. Compare to A-C for changes. (E-H) Wing discs expressing Toll-7-Venus (E,G; cyan/gray) or Toll-8-YFP (F,H; cyan/gray) together with clones expressing UAS-Myc (E,F; magenta) or UAS-Hpo (G,H; magenta) to induce super-competitor or loser cell phenotypes, respectively. Compare to A-C for changes. Scale bars: 50 µm.
To investigate if Toll receptors may also be utilized by other tissue-intrinsic error-correction mechanisms that recognize aberrant cells, we analyzed genetic models of classical cell-cell competition. In cell-cell competition, the comparison of proteostatic or metabolic fitness between neighboring cells drives the elimination of less fit ‘loser’ cells by fitter ‘winner’ cells (Baumgartner et al., 2021; Blanco et al., 2020; Baker et al., 2019; Baillon et al., 2018; Ochi et al., 2021; Kale et al., 2018; Lee et al., 2018; Ji et al., 2019). In fact, previous studies link different Toll receptors to cell-cell competition (Germani et al., 2018b; Byun et al., 2019; Alpar et al., 2018; Hof-Michel et al., 2024; Meyer et al., 2014; Katsukawa et al., 2018). To test if altered long Toll receptor expression is associated with winner or loser cell states, we analyzed Toll-2, Toll-7 or Toll-8 expression in imaginal discs containing mosaic clones with a Myc-expressing winner or a Hippo/Warts-expressing loser genotype. We observed that neither Myc-expressing nor Hippo/Warts-expressing clones showed alterations in long Toll receptor expression patterns predictive of winner or loser state (Fig. 7E-H; Fig. S7C,D). These experiments demonstrate that cell-cell competition does not rely on altered expression of long Toll receptors to deduce winner or loser state, in agreement with a model where interface surveillance is molecularly and mechanistically distinct from classical cell-cell competition.
So far, our combined observations provide evidence that: (1) individual long Toll receptors are regulated by cell fate patterning pathways; (2) each cell fate patterning pathway affects the expression of multiple long Toll receptors; and (3) a mismatch in any of the long Toll receptors is sufficient to induce interface surveillance. We now aimed to demonstrate that the long Toll receptors are also functionally necessary for the detection of aberrant cells. However, based on previous studies (Fischer et al., 2024), we anticipated that genetically targeting a single Toll receptor is not sufficient to prevent aberrant cell detection, as many receptors are deregulated in aberrant cells, and deregulation of any receptor can independently drive interface surveillance. In order to test this, we designed genetic necessity experiments. First we co-expressed a Toll-8-RNAi or Toll-8 overexpression construct in clones with aberrant cell fates, with the goal to restore Toll-8 expression levels to match those of surrounding cells. However, Tkv^CA^-expressing clones, which downregulate Toll-8 in the peripheral pouch, still induced clone smoothing, actin enrichment and apoptosis upon Toll-8 co-expression (Fig. S7E), likely due to the concurrent downregulation of Toll-2 and Toll-7. Similarly, attempts to restore low expression levels of Toll-8 in Fkh-expressing clones of the pouch by co-expressing a Toll-8-RNAi construct failed to reduce interface surveillance responses (Fig. S7F), likely due to the concurrent upregulation of Toll-7 and the interface surveillance-competent Robo2 (Fischer et al., 2024). Furthermore, double knock-down of Toll-2 and Toll-7 in Ey-expressing clones failed to reduce interface surveillance in the pouch (Fig. S7G,H), where both receptors are upregulated by Ey, but also high Robo2 is observed (Fischer et al., 2024). To additionally test whether removing a long Toll receptor from both aberrant and surrounding wild-type cells could suppress interface surveillance, we expressed UAS-RNAi constructs under the control of en-GAL4 for Toll-2, Toll-7 or Toll-8, all of which are downregulated in the peripheral pouch by Tkv^CA^. Consequently, all cells in the posterior compartment lacked a Toll receptor at the time of introducing Tkv^CA^-expressing clones under the control of LexA/LexO. Yet removing Toll-2, Toll-7 or Toll-8 from all posterior cells failed to affect the interface surveillance responses activated by Tkv^CA^-expressing clones (Fig. S8), likely due to the remaining differences in the other two Toll receptors. In summary, our findings indicate that, while individual long Toll receptors are regulated by cell fate patterning pathways and differences in their expression levels can induce interface surveillance, the suppression of this surveillance requires more-comprehensive functional-genetic approaches beyond targeting just one or two receptors within the complex cell surface code mediated by now at least 12 molecules.
Long Toll receptors act independently of NF-κB activity in interface surveillance
To understand how long Toll receptors may act in interface surveillance, we asked if canonical Toll signaling may play a role. The fly NF-κB homologues Dorsal (dl) and Dorsal-related immunity factor (Dif) are required for innate immune function but have not been implicated in morphogenetic functions (Iijima et al., 2020; Nüsslein-Volhard, 2022; Lavalou et al., 2021; Benton et al., 2016; Kleve et al., 2006; Tamada et al., 2021; Paré et al., 2014, 2019; Kolesnikov and Beckendorf, 2007; Umetsu, 2022; Graham et al., 2019; Tetley et al., 2016; Sharrock et al., 2022). Accordingly, we find that expression of Toll-6 with a mutated TIR domain normally required for downstream signaling to NF-κB, or expression of Toll-6 or Toll-8 that lack the entire cytoplasmic domain required for NF-κB signaling, still activated all characteristic features of interface surveillance (Fig. 8). This observation is particularly striking for Toll-6, which is not endogenously expressed in the wing disc, thereby eliminating any potential compensation by full-length endogenous Toll-6. These observations indicate that the primary function of long Toll receptors in interface surveillance is mediated by yet uncharacterized interactions via their extracellular domains, rather than by activation of canonical Toll-signaling on their cytoplasmic side.
Toll receptors act independently of classical NF-κB-signaling in interface surveillance. (A-C″) Wing discs carrying clones expressing UAS-Toll-6 (A-A″), UAS-Toll-6-TIR-dead (B-B″) or UAS-Toll-6Δcyto (C-C″) (cyan) to induce either ectopic expression of full-length Toll-6, Toll-6 with a non-functional TIR domain or Toll-6 lacking the cytoplasmic domain, respectively. White arrowheads in A″,B″,C″ indicate Actin enrichment at clonal interfaces. (D-E″) Wing discs carrying clones expressing UAS-Toll-8 (D-D″) or UAS-Toll-8Δcyto (E-E″) (cyan) to induce overexpression of full-length Toll-8 or expression of a cytoplasmically truncated Toll-8, respectively. TRE-RFP visualizes JNK-pathway activity (A,A′,B,B′,C,C′,D,D′,E,E′, magenta/gray); phalloidin visualizes F-actin (A″,B″,C″,D″,E″, gray). White outlines in A-D highlight regions shown in A′-E″.(F) Activation of different cell-fate patterning pathways (light- and dark-gray nuclei, cell fate 1 and 2) is integrated in a cell type-specific manner (top and bottom row, spatial domains A and B) to drive differential expression of cell-surface molecules, such a s Toll-2 and Toll-8 (magenta and cyan, cell-surface molecules A and B). Thus, the combination of long Toll receptors displayed at the cell surface depends on the signal integration and fate specification at any position within the imaginal disc (right-hand schematic). The differences in long Toll receptors expression levels between neighboring cells induces all hallmarks of interface surveillance (green lightning), including JNK activation, actomyosin recruitment and apoptosis. Scale bars: 50 µm.
This conclusion is further supported by an analysis of the NF-κB homologues Dl and Dif. While Dif levels are regulated by JNK signaling (Floc'hlay et al., 2023), and Dif was thus elevated at clonal interfaces upon activation of interface surveillance (Fig. S9A-D), we could not detect any effect in Toll-6, Toll-8 or Tkv^CA^-expressing clones, when Dif function was removed by co-expression of a functional Dif-RNAi either within clones or within the entire posterior compartment (Fig. S9G-K). In addition, we found no evidence for regulation of Dorsal during interface surveillance (Fig. S9E,F). Our results are consistent with a model where NF-κB activation is not required for interface surveillance, supported by similar observations during early embryonic morphogenesis (Iijima et al., 2020; Nüsslein-Volhard, 2022; Lavalou et al., 2021; Benton et al., 2016; Kleve et al., 2006; Tamada et al., 2021; Paré et al., 2014, 2019; Kolesnikov and Beckendorf, 2007; Umetsu, 2022; Graham et al., 2019; Tetley et al., 2016; Sharrock et al., 2022).
DISCUSSION
Our study demonstrates that long Toll receptors are each individually sufficient to induce interface surveillance via detection of expression level differences between neighboring cells, phenocopying the responses induced by the presence of cells with aberrant cell fate programs. In contrast to classical ligand-mediated Toll receptor signaling, it is the difference in the expression levels of the long Toll receptors Toll-2, Toll-6, Toll-7 and Toll-8 at cell boundaries, rather than their absolute expression levels within aberrant cells, which is sufficient to induce all hallmarks of interface surveillance (Fig. 8F). These hallmarks, which are identical to those typically induced by cells with aberrant patterning and cell-fate specification pathways, include recruitment of actin regulators and Myosin II to clonal interfaces, as well as JNK activation at the interface and apoptosis. Importantly, these responses are highly pattern specific. For example, Toll-8 induced interface surveillance is contingent on local Toll-8 expression levels, such that interface surveillance is activated only at clonal interfaces with strong differences in Toll-8 expression. As such, interface surveillance operates within four distinct endogenous Toll receptor expression patterns, enabling the recognition of diverse aberrant cells in different spatial domains. This function is specific to the long Toll-receptor subgroup, and not to other members of the Toll receptor family, thereby delineating functional distinction among Toll receptors. While our data demonstrate that individual long Toll receptors are each sufficient to induce interface surveillance, a limitation of our study is the absence of a direct genetic test of their necessity in recognizing aberrant cells, likely due to redundancy arising from the combinatorial deregulation of multiple cell surface molecules.
We demonstrate that expression of long Toll receptors is regulated by cell fate specification pathways, where any long Toll receptor is regulated by multiple cell fate specification pathways, and that each cell fate specification pathway regulates multiple long Toll receptors. Consequently, the composition of Toll receptors on a cell surface reflects the cumulative activity of a transcriptional network specific to each cell type. Previously, we identified cell-surface molecules of the Eph, Robo or Teneurin families as mediators of interface surveillance (Fischer et al., 2024). We thus propose a model for a cell-surface code, consisting of distinct expression patterns of multiple cell-surface molecules regulated by cell-fate specification pathways. Future research needs to address if this is reflected in normal imaginal disc development, and if distinct cell fates establish unique cell-surface profiles using a combinatorial code of different long Toll receptors, thereby providing information about the spatial position of cells. Our findings suggest the importance of a spatial code of Toll receptor for the detection of aberrant cell fates, which provides an opportunity of integrating developmental fate specification with tissue error correction mechanisms. Importantly, in a normally developing tissue, the graded expression patterns of cell fate established by morphogen gradients would prevent the appearance of strong differences between neighboring cells and thereby prevent the activation of interface surveillance. In contrast, alterations in developmentally aberrant or pre-cancerous cells would activate interface surveillance, thereby correcting mistakes and acting as a tumor suppressor mechanism.
Our results reveal distinct apoptotic patterns in tissues containing Toll receptor-misexpressing clones. These differences likely arise from multiple pro-apoptotic signals acting at and away from the interface. Specifically, our previous work has shown that JNK activation in aberrant and wild-type cells at the interface sensitizes both neighboring cells to apoptosis (Prasad et al., 2023). However, attenuation of JNK signaling only partially suppresses cell death, suggesting the contribution of additional mechanisms. These will certainly include mechanical compression, which can arise from the geometry of interface curvature as well as from population growth dynamics (Valon et al., 2025; Bielmeier et al., 2016). Of note, in a normally developing tissue, cells of different fates meet at lineage boundaries, such as the anterior-posterior compartment boundary in wing discs. Importantly, these boundaries are generally straight, thereby reducing local neighborhood connectivity that induces JNK in each cell (Prasad et al., 2023) and eliminating curvature induced compression (Valon et al., 2025), which, combined, would strongly reduce interface surveillance-mediated apoptosis at these naturally occurring cell-fate boundaries. Other effects that influence apoptotic patterns in Toll receptor-misexpressing clones are buckling of the epithelial sheet in the clone interior (Bielmeier et al., 2016; Prasad et al., 2023), mechanical fluctuations at the interface and differential propagation of the resulting tension within cell populations (Schoenit, Monfared et al., 2025), or long-range apoptotic signaling triggered by the elimination process itself (Perez-Garijo, Fuchs et al., 2013). Our boundary-resolved analyses show that Toll-2- and Toll-8-expressing clones generate apoptotic patterns consistent with signals acting at clonal interfaces, whereas wild-type encircled Toll-6 and Toll-7 clones exhibit elevated clonal death without an interface-correlated pattern. These results suggest that, despite the common activation of interface actomyosin and JNK signaling, the different long Toll receptors may also engage receptor-specific pathways of cell elimination, and possibly JNK-activation, a view supported by their different functions, ligands and effectors (McIlroy et al., 2013; Foldi et al., 2017; Akhouayri et al., 2011; Ding et al., 2022; Kong et al., 2022; Mishra-Gorur et al., 2019; Nakamoto et al., 2012; Lavalou et al., 2021; Tamada et al., 2021; Brutscher and Basler, 2025). Accordingly, previous studies on the function of long and short Toll receptors during classical cell-cell competition found that signaling by Toll-1, Toll-2, Toll-3, Toll-6, Toll-8 and Toll-9 triggers NF-κB-mediated apoptosis of loser cells (Blanco et al., 2020; Baker et al., 2019; Baillon et al., 2018; Baumgartner et al., 2021; Lee et al., 2018; Ochi et al., 2021; Kale et al., 2018; Ji et al., 2019; Alpar et al., 2018; Katsukawa et al., 2018; Hof-Michel et al., 2024; Meyer et al., 2014; Li et al., 2020b; Wu et al., 2015; Umetsu, 2022; Byun et al., 2019; Germani et al., 2018b). Upstream, these cell-cell competition processes require interaction between Toll receptors and Spätzle (Spz) ligands. The downstream pro-apoptotic signaling cascades vary, where Myc and Minute models generally depend on either canonical NF-κB pathway components or the Sarm-Rel axis (Meyer et al., 2014; Alpar et al., 2018; Germani et al., 2018b), while scrib models relay to Hpo signaling, either via a Toll-6/Spectrin or via a Toll-1/Serpin axis (Katsukawa et al., 2018; Kong et al., 2022). Thus, the family of Toll receptors is involved in both cell-cell competition and interface surveillance. Yet the differences in function may be rooted in the specific use of long or short Toll receptors, activation of downstream effectors like NF-κB and Rok/Myosin II, or the use of ligands like Spz.
In the embryo, heterologous interactions between the long Toll receptors themselves on neighboring cell surfaces drives the polarized recruitment of Src, PI3K, GPCRs or Rho GTPases to modulate actomyosin contractility and adhesive properties (Graham et al., 2019; Paré et al., 2019, 2014; Tamada et al., 2021; Lavalou et al., 2021; Kolesnikov and Beckendorf, 2007). Consequently, expression patterns of Toll receptors modulate affinity at contact surfaces between different ectodermal lineages, thereby driving their spatial partitioning and organization during tissue morphogenesis (Iijima et al., 2020; Nüsslein-Volhard, 2022; Lavalou et al., 2021; Benton et al., 2016; Kleve et al., 2006; Tamada et al., 2021; Paré et al., 2014, 2019; Kolesnikov and Beckendorf, 2007; Umetsu, 2022; Graham et al., 2019; Tetley et al., 2016; Villedieu et al., 2023; Sharrock et al., 2022; Alpar et al., 2025 preprint). Similarly, expression patterns of Toll receptors control cell-cell interactions by regulating axon targeting of different neuronal cell lineages during nervous system development (Ward et al., 2015; Guo et al., 2022; Li et al., 2020a; McIlroy et al., 2013). Combined, our data suggests that these observations are based on similar molecular mechanisms. Importantly, the role of Toll receptors in regulating actomyosin seems to be evolutionarily conserved and has now also been described for mammalian Toll-like receptors (Benton et al., 2016; Peterson et al., 2023). Overall, our results support the concept of a Toll receptor surface code, composed of the unique combination and expression levels of long Toll receptors in each wing disc cell, which together provide positional cues that adjacent cells can interpret and use to drive affinity, signaling and apoptotic decisions. Our findings highlight the specificity of aberrant cell detection which may represents a broader use of ‘self’ and ‘non-self’ recognition engrained in Toll receptor function, not just in innate immunity, but also in cell-cell competition, interface surveillance and embryonic morphogenesis.
MATERIALS AND METHODS
Drosophila genetics
Drosophila melanogaster fly stocks (see also Table S1) were maintained on standard fly food (10 l water, 74.5 g agar, 243 g dry yeast, 580 g cornflour, 552 ml 407 molasses, 20.7 g Nipagin and 35 ml propionic acid) at 18°C-22°C. Mosaic analysis of wing discs was performed by using the Flip-out system (act or tub>GAL4/UAS, CoinFLP-LexA/LexO). Due to genetic limitations, the CoinFLP experiments contained two GAL4-drivers, the Actin5C-FRT/STOP-GAL4 and the en-GAL4, resulting in mosaic GAL4 expression in the total disc and GAL4 expression in the posterior compartment. Larvae were dissected at wandering 3rd instar stage or as indicated (30 h or 48 h after heat-shock). Experimental crosses were raised on food prepared according to Bloomington formulation (175.7 g Nutry-Fly, 1100 ml water 20 g dry yeast, 1.45 g 409 Nipagin in 15 ml ethanol and 4.8 ml propionic acid) at 25°C. After an egg-laying period of 72 h, Flippase (flp) expression was induced by heat shock (HS) at 37°C for 5.5-14 min. Larvae from experimental crosses were raised at 18°C, 21°C or 25°C and dissected after 24-48 h, unless stated otherwise. For expression pattern visualization, parental flies were allowed to lay eggs for 8 h and larvae were dissected at 80 h and 102 h after egg laying (AEL). Our experimental design did not consider differences between sexes unless for genetic crossing schemes.
Immunohistochemistry and imaging
3rd instar larvae were dissected and inverted in 1× phosphate-buffered saline (PBS). Cuticles with attached wing imaginal discs or CNS were fixed in 4% paraformaldehyde/PBS for 15 min at 22°C. Samples were washed in 0.1% Triton X-100 in PBS (PBT). Primary antibody incubation was performed in 400 µl PBT overnight at 4°C, using the following antibodies (see also Table S1): mouse anti-dorsal (1:10, DSHB, 7A4), mouse anti-Ena (1:100, DSHB, 5G2), mouse anti-Ptc (1:20, DSHB, Apa1), mouse anti-Wg (1:100, DSHB, 4D4-s), rabbit anti-Dcp-1 (1:400, Cell Signaling, 9578S), rabbit anti-GFP (1:400, ThermoFisher, G10362), rabbit anti-mNeonGreen (1:2000, Abcam, ab321887), rat anti-E-cadherin (1:100, DSHB, DCAD2-s), rat anti-RFP (1:2000, Chromotek, 5F8), mouse anti-β-gal (1:2000, Promega, Z3783), rabbit anti-HA (1:200, Invitrogen, 715500) and rat anti-HA (1:250, Roche, 3F10). Secondary antibody and dye incubation was performed in 400 µl PBT for 2 h at 22°C with the following antibodies and dyes: goat anti-rabbit Alexa Fluor 488 (1:500, Invitrogen, A11008), goat anti-rabbit Alexa Fluor 555 (1:500, Invitrogen, A21428), goat anti-rat Alexa Fluor 555 (1:500, Invitrogen, A21434), goat anti-mouse Alexa Fluor 555 (1:500, Invitrogen, A21422), goat anti-rabbit Alexa Fluor 647 (1:500, Invitrogen, A21244), goat anti-mouse Alexa Fluor 647 (1:500, Invitrogen, A32728), donkey anti-rat Alexa Fluor 555 pre-adsorbed (1:500, Abcam, AB150154), donkey anti-mouse Alexa Fluor 647 pre-adsorbed (1:500, Abcam, AB150111), DAPI (0.25 ng/µl, Sigma), Phalloidin-Alexa Fluor 405 (1:2000, Abcam-ab176752) and Phalloidin-Alexa Fluor 555 (1:500, Sigma-P1951). Samples were mounted with SlowFade Gold Antifade Reagent (10-15 µl) and imaged using Leica TCS SP8 confocal microscopes with a 40× or 63× oil objective.
Image quantification and statistical analysis
Figure display
Images were processed using Fiji 2.14 and figure panels were assembled using Affinity Designer 2. Figure panels either display individual channels at the same z position or individual channels from different z positions of a confocal stack for better visualization of spatially distinct phenotypes. Specifically, apical projections were used to display junctional actin and basal z sections were used to display Dcp-1. JNK-pathway activity by TRE-RFP reporter activity was mostly displayed in lateral z sections. Z-stack position of the channel displaying the corresponding cytoplasmic/nuclear RFP/GFP/LacZ/HA was chosen accordingly. Overlaid panels were assembled with the z-stack position used to display the individual channels in gray values. Visualization of actin enrichment was performed using the LocalZProjector plugin (https://gitlab.pasteur.fr/iah-public/localzprojector, v1.5.4) for Fiji to project the curved tissue surface represented in several z-stacks of a 3D image stack into a 2D plane by using phalloidin or E-Cadherin staining as a reference plane. For subcellular localization of Toll-2, Toll-7 and Toll-8, staining of the adherens junction marker E-cadherin was used to project apical surfaces using the local z projector, and lateral F-actin sections were extracted by applying an offset relative to the adherens junction projection. To derive schematic representations of Toll-like receptor expression patterns used throughout the manuscript, characteristic points of reference, such as hinge folds, A/P and D/V boundaries were used to transfer the expression domains manually into a wing disc template in Affinity Design using Procreate (Version5.2.9) and Inkscape (0.92.4). The grading of fluorescence intensity is based on the overall impression of intensity across the wing disc and is presented as a subjective approximation. Thus, the schemes provide a general reference for areas with the (subjective) highest and lowest expression, but they lack precise details, particularly in the hinge folds, due to the limitations of representing the expression pattern within a highly dynamic 3D tissue in a 2D format. In Fig. 3D, Fig. S1I, Fig. 6A,D and Fig. S3C′, composite images were generated by combining separately acquired images of the notum and pouch from the same z-plane of a single wing disc.
Image analysis
For all measurements, one wing disc was dissected from each larva. Unless stated otherwise, individual clones on the wing discs obtained were considered to be an independent sample for statistical analysis. Control and corresponding experimental samples were processed together, and were always imaged in parallel, using the same confocal settings.
Experimental workflow for quantification of TRE-RFP and cleaved Dcp-1 intensity
Due to differences in overexpression intensity of the used constructs, different experimental workflows were implemented to aim for clones large enough for ROI segmentation (see below). UAS-Toll-6 and UAS-GFP genotypes were dissected 30 h after heat-shock (AHS) and reared at 25°C, UAS-Toll-2 and UAS-GFP genotypes were dissected 48 h AHS and reared at 18°C. UAS-Toll-8, UAS-Toll-7 and UAS-GFP genotypes were dissected 40 h AHS and reared at 18°C. The TRE-RFP signal to assess JNK-activity was boosted by using an anti-RFP antibody in all samples. To assess apoptosis, we used antibody staining for cleaved Drosophila caspase 1 (cDcp-1). For both TRE-RFP and Dcp-1, average fluorescence intensity within specific regions was measured in arbitrary units, using an 8-bit image where intensities scale between 0 to 255. These regions were segmented according to the criteria described below.
Quantification of TRE-RFP intensity
To automate the analysis of the obtained images, we first wrote a macro code in Fiji to quantify TRE-RFP intensity in defined regions of interest [outer band (OB), background band (BB), inner band (IB) and inner area (IA)], ensuring accurate and reproducible analysis. First, the GFP signal was used to generate a clonal mask. One z-slice in the lateral cell domain was selected, and a region of interest (ROI) for quantification (e.g. pouch, hinge or notum) was defined using the polygon tool. ROIs for quantification were defined to test context specific TRE-response and clone morphology: UAS-Toll-2, UAS-Toll-6, UAS-Toll-7 and UAS-GFP clones were quantified in the hinge and pouch, UAS-Toll-8 and UAS-GFP clones were quantified in (1) the pouch, an area with low endogenous Toll-8 expression, and (2) the notum, an area with high endogenous Toll-8 expression. To improve thresholding, Contrast Limited Adaptive Histogram Equalization (CLAHE) with default parameters and Gaussian Blur (Radius=2) were applied. After thresholding GFP-expressing clones, the image was converted to a mask. Individual clones were selected and saved as clone ROIs using the Analyse Particles function (size=10 µm-infinity). For each clone, an outer band (4 µm) was created using the Make Band function, and the clone areas were subtracted from the outer bands to ensure no overlap. Inner areas were generated by negatively enlarging (−4 µm) the clonal ROIs. Inner bands were created by performing an XOR operation between the inner area and the clonal ROI, ensuring non-overlapping inner bands. Background bands were generated by initially enlarging the clonal ROI by 4 µm, representing the clonal ROI plus the outer band. A second band (12 µm) was created around the clonal ROI plus outer band to encompass an area extending further from the outer band. Any overlaps between the newly created background bands and previously defined bands were removed using Boolean operations to ensure that the background bands included only areas that were truly in the background and did not overlap with the areas occupied by clones and their immediate surroundings. Finally, all generated bands and clonal ROIs were restricted to the previously defined ROI representing the area of quantification using AND operations. The generated bands were named sequentially and accurately represented the background band (BB), outer band (OB), inner band (IB) and inner area (IA) of individual clonal ROIs. TRE-RFP intensity was measured in the generated areas using the measurement command. For further analysis, the generated data were exported to Excel. Interface surveillance is a clone size-dependent process in which smaller clones are more prone to elimination due to enhanced JNK signaling and potentially stronger contractile forces at the interface (Prasad et al., 2023). To reflect similar interface surveillance dynamics in all measurements and to prevent disproportionate effects due to clone size differences, clones with an inner area (IA) larger than 80 µm² were identified and selected for subsequent statistical analysis.
Quantification of clone circularity
Circularity of GFP-positive clones was quantified from the thresholded clone masks generated within the Fiji macro for quantification of TRE-RFP intensity using the Analyze Particles function. Circularity was calculated by Fiji as 4π×area/perimeter^2^, where 1.0 indicates a perfect circle and values approaching 0 reflect increasingly elongated or irregular shapes. Measurements were derived directly from binary clone masks after thresholding and hole filling, and exported together with area and intensity data for statistical analysis.
Quantification of cDcp-1 intensity
To quantify cDcp-1 intensity, a sum projection of three basal z-slices was generated for both the GFP -channel, which contains the clones, and the cDcp-1 channel. The sum projection of the GFP channel was utilized to create a clonal mask, which facilitated the measurement of cDcp-1 intensity in the corresponding sum projection of the cDcp-1 channel. To ensure accurate and reproducible analysis, a macro code was written for Fiji . Clonal segmentation, thresholding and data processing in Excel were performed following the same workflow previously described for the quantification of TRE-RFP intensity.
Quantification of Toll-2-GFP, Toll-7-Venus and Toll-8-YFP expression in UAS-RFP, UAS-TkvCA, UAS-Fkh and UAS-Ey clones
To quantify changes in Toll-2-GFP, Toll-7-Venus and Toll-8-YFP expression in Fiji, four contiguous sections through basolateral cell surfaces of the disc were selected. Spatial domains of interest for quantification (e.g. pouch, hinge and notum) were defined using the polygon tool. The ROI segmentation described above was then applied to each z-section, excluding the CLAHE plug-in and Gaussian Blur functions, to measure receptor fluorescence intensity within the outer band (OB) and inner band (IB) ROIs. Clone centroids were used to compute Euclidean similarity and distance indices in Python, which enabled automatic tracking of individual clones across the four z-sections, resulting in up to four independent measurements per clone. To minimize artifacts in receptor fluorescence intensity caused by clone deformation or tissue invagination, data points were filtered based on actin level changes: z-sections showing more than a 20% difference in actin intensity between OB and IB regions were excluded. Additionally, clones smaller than 40 µm², larger than 1200 µm² or detected in only one of the four z-sections were excluded from analysis. For each clone, the final data point used for statistical analysis was taken from the most apical z-section that satisfied all filtering criteria.
Statistical analysis
GraphPad Prism (version 9.5.0) was used for statistical analysis and generation of graphs. The number of samples (n/N) and the specific statistical tests used are detailed in the associated figure legends. A significance level of P<0.05 was used for all tests. Data are presented as mean±95% confidence interval. All data were tested for normality and homogeneity of variances. In cases where data did not meet the assumptions of parametric tests, non-parametric tests were used instead to ensure robust and accurate analyses.
Supplementary Material
10.1242/develop.205006_sup1Supplementary information
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1Akhouayri, I., Turc, C., Royet, J. and Charroux, B. (2011). Toll-8/Tollo negatively regulates antimicrobial response in the Drosophila respiratory epithelium. P Lo S Pathog. 7, e 1002319. 10.1371/journal.ppat.100231922022271 PMC 3192845 · doi ↗ · pubmed ↗
- 2Aldaz, S., Morata, G. and Azpiazu, N. (2005). Patterning function of homothorax/extradenticle in the thorax of Drosophila. Development 132, 439-446. 10.1242/dev.0158515634705 · doi ↗ · pubmed ↗
- 3Alpar, L., Bergantiños, C. and Johnston, L. A. (2018). Spatially restricted regulation of Spatzle/toll signaling during cell competition. Dev. Cell 46, 706-719.e 5. 10.1016/j.devcel.2018.08.00130146479 PMC 6156939 · doi ↗ · pubmed ↗
- 4Alpar, L., Pigache, R., Leroy, A., Ech-Chouini, M., Pelletier, S. and Bellaïche, Y. (2025). Drosophila Toll-2 controls homeotic domain formation and timing of folding. bio Rxiv, 2025.01.07.631521. 10.1101/2025.01.07.631521 · doi ↗
- 5Anderson, K. V., Jürgens, G. and Nüsslein-Volhard, C. (1985). Establishment of dorsal-ventral polarity in the Drosophila embryo: genetic studies on the role of the Toll gene product. Cell 42, 779-789. 10.1016/0092-8674(85)90274-03931918 · doi ↗ · pubmed ↗
- 6Anthoney, N., Foldi, I. and Hidalgo, A. (2018). Toll and Toll-like receptor signalling in development. Development 145, dev 156018. 10.1242/dev.15601829695493 · doi ↗ · pubmed ↗
- 7Baillon, L., Germani, F., Rockel, C., Hilchenbach, J. and Basler, K. (2018). Xrp 1 is a transcription factor required for cell competition-driven elimination of loser cells. Sci. Rep. 8, 17712. 10.1038/s 41598-018-36277-430531963 PMC 6286310 · doi ↗ · pubmed ↗
- 8Baker, N. E., Kiparaki, M. and Khan, C. (2019). A potential link between p 53, cell competition and ribosomopathy in mammals and in Drosophila. Dev. Biol. 446, 17-19. 10.1016/j.ydbio.2018.11.01830513308 PMC 6642609 · doi ↗ · pubmed ↗