EGFR/ARF6 regulation of Hh signalling stimulates oncogenic Ras tumour overgrowth

Multiple signalling events interact in cancer cells. Oncogenic Ras cooperates with Egfr, which cannot be explained by the canonical signalling paradigm. In turn, Egfr cooperates with Hedgehog signalling. How oncogenic Ras elicits and integrates Egfr and Hedgehog signals to drive overgrowth remains unclear. Using a Drosophila tumour model, we show that Egfr cooperates with oncogenic Ras via Arf6, which functions as a novel regulator of Hh signalling. Oncogenic Ras induces the expression of Egfr ligands. Egfr then signals through Arf6, which regulates Hh transport to promote Hh signalling. Blocking any step of this signalling cascade inhibits Hh signalling and correspondingly suppresses the growth of both, fly and human cancer cells harbouring oncogenic Ras mutations. These findings highlight a non-canonical Egfr signalling mechanism, centered on Arf6 as a novel regulator of Hh signalling. This explains both, the puzzling requirement of Egfr in oncogenic Ras-mediated overgrowth and the cooperation between Egfr and Hedgehog.

In this study, Chabu et al have investigated the mechanism by which EGFR cooperates with oncogenic Ras in tumorigenesis using the Drosophila model. The study reveals that Ras signalling upregulates the EGFR ligand, Spitz, which stimulates EGFR signalling via a non-canonical pathway via Arf6 to upregulate Hedgehog signalling. It is already known in mammalian cells that EGFR cooperates with oncogenic Ras, EGFR activates the Hedgehog pathway, and EGFR recruits Arf6-GEF to simulate Arf6 activity, however the study of Chabu et al., shows that this mechanism is conserved in Drosophila and more clearly delineates the mechanism to show that EGFR signalling leads to greater association of Arf6 with Hh and that inhibiting Arf6 results in an accumulation of Hh in Hrs-positive vesicles, which presumably then targets Hh for degradation in the lysosome. This aspect of the study seemed under-developed, and more mechanistic insight could have been obtained by some additional staining with endosomal markers. The manuscript is generally wellwritten, although would benefit from having subdivision of the Results section and headings that summarize the specific findings of each part. Furthermore, the Introduction needs to have a more comprehensive description of previous findings on Ras in tissue growth and tumorigenesis in Drosophila -at present the authors only cite their own paper from 2003, which is more to do with cooperation of RasV12 with polarity impairment. The Discussion section (needs to be labelled "Discussion") is just a summary of the results and needs to provide greater insight, for example 1) what other signalling pathways might Arf6 regulate by affecting vesicle trafficking that could be important in the RasV12 phenotype, 2) how does EGFR signalling activate Arf6-GEF -might this be through direct binding, and 3) what might be the developmental or physiological function of EGFR-Hh regulation? The quality of the data was high and mostly well quantified, although in the text further explanation is needed at several places -ie how were clones generated and were they in the eye or wing epithelium, and how was Arf6 identified in the screen alluded to in the results? Overall this is an interesting study that would be of great relevance to the signalling and cancer fields, however I believe that further tightening up of the data would improve the quality of the paper and its impact.
Specific comments: (1) Figures: The RasV12 MARCM clonal tissue seems to overtake the majority of the eye epithelium in some figures (1B, 2C) but not others (1O, 2A) -why is there variability -are larvae all staged at 5 days wandering 3rd instar and raised at the same temperature? This variability makes it important to quantify all interactions -eg the clonal area of RasV12 relative to with RasV12 with Star-RNAi, Sos-RNAi, and ras-c40e (Fig 2), to make all conclusions more robust.
(2) Fig 1E -its seems that RasV12 EGFR-DN results in non-cell autonomous growth and indeed there seems to be Ph3-foci around the outside of the clones in 1P-is this the case, and why is this different from RasV12 EGFR mutant mosaic eye discs ( Fig 1D)? [Fig 10 and 1P should have panels of Ph3 alone with the clones marked with dotted lines to make this easier to see.] RasV12 Arf6-RNAi and RasV12 loner-RNAi also seem to show non-cell autonomous overgrowth (Fig 3). This non-cell autonomous tissue growth effect, could be due to JNK upregulation and a secretory phenotype (Uhlirova et al., 2005, Nakamura et al., 2014)? This might be associated with the effects on Hrs accumulation seen in RasV12 Arf6-RNAi, and therefore controls should be done to show that JNK is not involved here.  Fig 4F -The Arf6-RNAi mosaic eye shown is unusual with the clone taking up all of one half of the eye, and seems inconsistent with Arf6 being important for cell proliferation -was this a common finding? Does Arf6-RNAi alone effect cell proliferation, or only in the context of RasV12? (7) Fig 4N -How was the Arf6-RNAi clone generated in the wing disc? Only eyFLP is listed in the M&M. (8) Fig 4P -did the Arf6-RNAi line also result in a wing vein ablation? The authors state that this is due to defective Hh signalling, but could it also be due to defective Ras signalling, given the role of Ras signalling in wing vein formation? Can this be rescued by expressing full-length Ci? (9) Having shown Arf6 knockdown prevents the overgrowth of RasV12 expressing tissue it would be good to show whether elevated Arf6 signalling (expression of a Arf6-GTP locked form) cooperates with RasV12 to enhance tumorigenesis via increased Hh signalling? (10) Fig 4W -the effect of the knockdown on the accumulation of Hrs vesicles and the colocalisation of Hh with Hrs was very interesting, but Hh might be expected to still be able to signal from the Multi-Vesicular Body (MVB), since Notch signalling still occurs in Escrt mutants, blocked at the MVB stage. This part of the study seemed incomplete, since other endosomal compartments, such as the early endosome, might accumulate in RasV12 tissue and be enriched with Hh, consistent with this compartment being important for promoting Hh signalling. Conversely, Arf6 knockdown might lead to greater lysosomal trafficking and degradation of Hh. Staining with early endosomal markers such as Rab5, Avl, late endosomal markers (Rab7) and lysosomal markers (Arl8) in RasV12 versus RasV12 Arf6-RNAi would provide more insight into this aspect of the study. (13) Results section: P3 -It is not very precise to refer to RasV12 expressing clones as "tumors"they are hyperplastic and have a competitive advantage, but differentiation still occurs and they are not invasive -so they should be termed "hyperplastic" or "benign tumors" so it is clear to the general reader. (14) P3 -What does "unstable clones" mean? Would be better to describe them as "small clones" or "having reduced viability". (15) P4 -The authors examine Spi, but Argos (negative regulator of EGFR signalling) has also been reported to be a target of Ras signalling -it would be interesting to see whether argos expression is also upregulated in RasV12 clones. (16) P5 -A screen is mentioned as to how Arf6 was chosen for analysis, and the Materials and Method section is referred to for details of this, however I could not find any details there? This needs to be described.
(17) P5 -The effect of Arf6 knockdown on RasV12 mammalian cell proliferation was interesting, however it would add greater relevance if the connection of Arf6 to Hedgehog signalling could be made later in the paper (after the Hh section) in these mammalian cells and correlated with the effect on cell proliferation. (18) P6 -The logic for investigating the Hh pathway because vesicle trafficking regulates Hedgehog signalling is not very convincing, since vesicle trafficking regulates many pathways including Notch. Best just to argue that they investigated the connection to Hh because of the mammalian evidence that Hh acts downstream of the EGFR. (19) P6 -The Ci antibody needs to be described more fully in the results and M&Ms to state that it is to the full-length active version of Ci. (20) P7 -More detail would be helpful in the statement -"we have previously shown that tumors co-opt developmental mechanisms to promote overgrowth" The manuscript by Chabu, Li and Xu addresses an enigma related to oncogenic Ras; despite being upstream of Ras, EGFR is required for Ras driven tumorigenesis. This observation points to involvement of a branch, downstream of EGFR, that is parallel to Ras and that contributes to activated-Ras induced phenotypes. The authors carry out a smart screen to identify this branch among the known EGFR effectors. They find that knockdown of Arf6, a Ras-related GTP-binding protein, can supress oncogenic Ras driven overgrowth. They go on to show that Arf6 works via modulation of Hh signalling and this is the novel finding of the paper. According to their model, activated Ras induces transcription of EGFR ligand Spitz, which leads to activation of EGFR and eventually Hh signalling via Arf6 which supposedly prevents/delays Hh degradation. Hh activation cooperates with and contributes to Ras-driven overgrowth. The model is neat as the circle is completed.
Overall, the data presented is of good quality and the findings are of interest to a wide readership. I would recommend publication in Nature Communications given the second half of the story, the connection between Arf6 and Hh signalling, is taken a bit further and the model is better supported.
Specific comments: -The feedback regulation of Spi and its contribution to Ras driven overgrowth is neatly demonstrated. Is transcriptional regulation of Spitz by EGFR signalling shown for the first time here? Are examples of feedback regulation of its ligands via the EGFR pathway known in other systems? Please discuss.
-How important is Arf6 as an EGFR effector? Can it supress activated-EGFR driven overgrowth? -What is the role of Arf6 in vesicle trafficking? -Can the authors comment more on the effect described in Suppl Fig. 3. Is this a regional effect? Which part of the disc are we looking at? -Colocalization statements are difficult to make as only a small percentage of the signal actually overlap, but it might still be functionally relevant. Is there a difference between the RasV12 clones and wt tissue with respect to Hh and Afr6 colocalization?
-Genetic data showing that Hh signalling is important for the growth of RasV1 clones is very nice (Fig. 4R) while the claim that RasV12 leads to upregulation of Hh activity is less convincing. While I trust the authors conclusion is most likely correct, I would like to see better data points supporting this conclusion.
-In Figure 4, are panels H-L showing the same images as B-F? It is confusing! Would be nice to show the ci channel alone for rasV12 experiment in B and H. This part of the paper is the least convincing part. Can the authors use other assays than immunostaining for Ci to measure Hh activity? -Are these effects specific to the eye discs or seen in other discs as well? Clones are by nature more difficult to interpret as they are randomly distributed. A simpler assay would be to induce RasV12 in a large area instead of clones, for example with Nub-G4 in the wing or a dorsal-eye-gal4 in the eye.
-It is nice that Arf6 RNAi treatment supresses Ci levels in both eye and wing discs. Does RasV12 induce Ci in the wing? -Can Arf6 overexpression stabilize Hh and increase Hh signalling? -I disagree with the statement that "abrogating EGFR function in RasV12 cells supresses ARF6's ability to pull-down Hh" on page 9 as still a good amount of Hh can be pulled downed with ARF6.
-Does ARF6 knock-down or overexpression influence Hh localization to Hrs-positive vesicles in a wild-type background? -The authors should use the same font size in all figures and panels throughout.

Chabu et al, Nat Comm 2016
This manuscript presents evidence that oncogenic Ras acts in conjunction with EGFR signaling to drive cell proliferation in a Drosophila tumor model, and in a panel of human tumor cell lines. The data suggest that EGFR signaling is necessary for optimal proliferation even in the presence of oncogenic Ras, which is uncoupled from the activated receptor. The authors show that knockdown of either Arf6 or one of its guanine nucleotide exchange factors (Loner/Schizo) attenuates growth induced by oncogenic Ras, and that this may be due to altered trafficking of Hedgehog (Hh). The findings, particularly related to a potential role for Arf6 in Hh signaling, are novel and would be of interest to the cancer signaling community. However, there are several siginificant problems with the manuscript in its current form. These include: 1. Many of the images shown do not appear to match the corresponding quantitative data, and call into question how quantitation was performed. For example, in Fig. 2, panels A and C both represent control imaginal discs expressing oncogenic Ras. However, the staining pattern in A is very similar to the patterns in D, E and F, which supposedly represent attenuated growth caused by loss of Spitz (D), star (E) or ligand binding to EGFR (F). In addition, panel C is massively overexposed, while D appears to be underexposed relative to all the others. If A and C represent equivalent conditions, the error bars in the corresponding quantitation should be huge, and they are not. No methodology is provided describing how this quantitation was done. Since the entire study is based on quantitation of clone size, the data are impossible to interpret. 2. The authors use Cubitus interruptus (Ci) expression as a readout of Hh signaling. While Ci expression is indeed higher in some clones expressing oncogenic Ras, this is not true for all of them (Fig. 4B). How do the authors interpret this? 3. Arf6 is activated by a number of different GEFs. Although knockdown of Loner/Schizo does appear to weakly attenuate the growth of clones expressing oncogenic Ras (Fig. 3D), it is possible that it is also activated by other GEFs (e.g. Steppke). This should be examined more thoroughly, especially since they cite a paper describing a role for Stepkke in EGFR signaling (Hahn et al, ref 27). 4. In the context of human cell lines, are all of the Ras-dependent lines also dependent on Hh signaling? Where does Hh come from under these conditions? If autocrine Hh signaling occurs in every case, this should be demonstrated. 5. In Fig. 4U, the authors use co-precipitation to suggest an interaction between Arf6 and Hh. However, there are two problems with this figure. First, total Hh should be shown in addition to tubulin. The difference in association +/-EGFR could simply be due to differences in overall expression level. Second, according to their Methods section, co-precipitation was performed from cell homogenates in the absence of detergent. Thus the two proteins may not interact at all, but are simply present in the same membranes. 6. The authors use colocalization of Hh with the endosomal protein Hrs as an indicator of entry into the degradative pathway. However, Hrs is present on early endosomes, and the image shown in Fig. 4W more likely represents accumulation of Hh in early, non-degradative endosomes in the absence of Arf6. How would this affect Hh signaling? Where is Patched (Hh receptor) under these conditions? Related to point #5, are Hh levels increased or decreased under these conditions? Where is the Hh under control conditions, if not associated with Hrs? And again, no description is provided of how quantitation of colocalization was performed. How many cells, how many discs were imaged to obtain the data shown in Fig. 4Z? Colocalization with bona fide late endosomal markers (mannose-6-phosphate receptors, cathepsins, LAMPs) would be necessary to make the authors' point here. This variability makes it important to quantify all interactions -eg the clonal area of RasV12 relative to with RasV12 with Star-RNAi, Sos-RNAi, and ras-c40e (Fig 2), to make all conclusions more robust.

Response to Reviewers' Comments
Response: The reviewer is correct. The difference in clone size between Figures (1B, 2C) and Figures (1O, 2A) is because they are from different stages. Growth analyses for clones were conducted at wandering 3 rd instar larval stage following standard practice ( Figures 1B and 2C). Immuno-staining for assessing signaling, however, were carried out at early 3 rd instar stage when clones are in rapidly growing phase and can be compared with surrounding wild-type tissue ( Figures 1O and 2A). Mutant and wild-type control animals were raised at the same condition. Clone size from animals at the same stage is quantified. We apologize for not making clear that growth analysis and immune-staining were from different time points and have now specified this in all relevant figure legends.
(2) Fig 1E -it seems that RasV12 EGFR-DN results in non-cell autonomous growth and indeed there seems to be Ph3-foci around the outside of the clones in 1P-is this the case, and why is this different from RasV12 EGFR mutant mosaic eye discs ( Fig 1D) Response: The apparent difference between tissues with Ras V12 , EGFR-DN clones ( Figure   1D) and tissues with Ras V12 , Egfrmutant clones ( Figure 1E) are because the previous two images had different scale bars in the old Figure 1. The tissues with either type of clones behave similarly and both show non-autonomous overgrowth. We have now used images with the same scale bars to avoid the confusion (new Figure 1D, 1E).
Following the reviewer's suggestion, we have now shown PH3 channel alone with the clones marked by dotted lines (Figure 1Q, 1R).
We agree that determining whether JNK signaling mediates EGFR-DN effects on Ras V12 growth is important. We tested this by directly blocking JNK signaling in Ras V12 , EGFR-DN clones and asked whether this would rescue growth and found that it did not. This   The authors state that this is due to defective Hh signalling, but could it also be due to defective Ras signalling, given the role of Ras signalling in wing vein formation? Can this be rescued by expressing full-length Ci?
Response: ARF6-RNAi does have a weak wing vein ablation phenotype ( Supplementary   Fig. 6). To test whether Hh signaling can rescue ARF6-DN vein phenotype, we coexpressed ARF6-DN and full-length Ci (Ptc>ARF6-DN, Ci ACT ) as the reviewer suggested.
However, these animals do not eclose, making the analysis impossible. depleted Ras V12 cells. We could not co-stain tissues with anti-Hh and anti-Arl8 (or anti-Rab7) antibodies because these antibodies were raised in the same species. Instead we used the GMR-GAL4 driver to express GFP-labeled full-length or a version of Hh corresponding to its secreted and active form (Hh-GFP and Hh-N-GFP, respectively) in cells expressing either Ras V12 alone or co-expressing ARF6-DN or ARF6-RNAi, and stained tissues against Arl8. We found that Hh-GFP preferentially localizes to clusters of Arl8-positive vesicles in Ras V12 , ARF6-DN tissues (Figure 5c, 5c', and 5g). Similarly, we expressed Rab7-GFP in Ras V12 , ARF6-DN cells, stained against Hh, and found that Hh predominantly localizes to Rab7-GFP vesicles in these cells compared to controls (Ras V12 cells) (Figure 5d, 5h vs. 5i, 5m). Moreover, expression of ARF6-RNAi in Ras V12 cells resulted in the redistribution of Hh-N-GFP to Arl8 endosomes (Figure 5j, 5n vs. 5k, 5o).
Furthermore, expression of ARF6-RNAi in otherwise wild-type cells similarly showed Hh localization to Rab7 endosomes ( Supplementary Fig. 7 e, f, and h). Consistent with this, patched (Hh receptor) localizes to Rab7 or Arl8 endosomes in ARF6-RNAi or ARF6-DN expressing cells (Fig. 5l, 5p, and Supplementary Fig. 6 g-j). Together with our previous results with Hrs, our data strongly argue that ARF6 normally prevents the trafficking of Hh to the degradation pathway. These new data have now been included in the manuscript (pages 12 and 13, lines 241-258). (13) Results section: P3 -It is not very precise to refer to RasV12 expressing clones as "tumors" -they are hyperplastic and have a competitive advantage, but differentiation still occurs and they are not invasive -so they should be termed "hyperplastic" or "benign tumors" so it is clear to the general reader.
Response: We agree and have now clearly stated that Ras V12 clones give rise to "hyperplastic tumors" in the introduction before simply referring them as tumors. This information is on page 4, lines 58-59.
(14) P3 -What does "unstable clones" mean? Would be better to describe them as "small clones" or "having reduced viability".
Response: We agree and are now describing Egfr clones as small clones (page 4, line 67).
(15) P4 -The authors examine Spi, but Argos (negative regulator of EGFR signaling) has also been reported to be a target of Ras signaling -it would be interesting to see whether argos expression is also up-regulated in RasV12 clones.
Response: We stained tissues containing Ras V12 clones against Argos and found that Argos protein level was not elevated in Ras V12 clones compared to surrounding wild-type tissues (see below).

Argos expression in Ras V12 clones
A-C) Images showing eye disc containing Ras V12 clones (green) stained with DAPI and anti-Argos antibodies (C). Individual clone channel is shown in (B).
(16) P5 -A screen is mentioned as to how Arf6 was chosen for analysis, and the Materials and Method section is referred to for details of this, however I could not find any details there? This needs to be described.
Response: We did not describe or mention a screen in the manuscript. ARF6 is a result of a candidate approach.
(17) P5 -The effect of Arf6 knockdown on RasV12 mammalian cell proliferation was interesting, however it would add greater relevance if the connection of Arf6 to Hedgehog signalling could be made later in the paper (after the Hh section) in these mammalian cells and correlated with the effect on cell proliferation.
Response: We examined changes in Gli1 (Hh signaling transcriptional target) protein levels following ARF6 RNAi knockdown and found that ARF6 RNAi reduced GL1 levels in multiple lung cancer cell lines (Fig. 4q). We then directly blocked Gli1 activity in these cells with GANT61, a specific Gli1 small molecule inhibitor 3,4 and found that it suppresses growth (Fig. 4r), similar to ARF6 knockdown. Thus, similar to the effects of ARF6 knockdown in flies, ARF6 knockdown blocks Hh signaling in mammalian cells and this effect correlates with growth inhibition. These exciting new data have been added in Figure 4q, r, and in the text (page 11; lines 213-219).
(18) P6 -The logic for investigating the Hh pathway because vesicle trafficking regulates Hedgehog signalling is not very convincing, since vesicle trafficking regulates many pathways including Notch. Best just to argue that they investigated the connection to Hh because of the mammalian evidence that Hh acts downstream of the EGFR. Response: We thank the reviewer for suggesting this experiment. We ectopically expressed EGFR in the Ptc stripe in the wing discs and found that it causes overgrowth compared to controls, as expected ( Supplementary Fig. 6a and 6b). Expression of ARF6-RNAi significantly suppressed the overgrowth phenotype ( Supplementary Fig. 6c), indicating that ARF6 is an important effector of EGFR. This information has been added on pages 7 and 8, lines 136-138.
2) What is the role of Arf6 in vesicle trafficking?
Response: ARF6 regulates the formation of carrier vesicles to positively or negatively control vesicle trafficking in epithelial cells endosomes 5--7 . Our data show that ARF6 prevents Hh from entering the degradation pathway. This is supported by our additional data showing that Hh predominantly localizes to degradation pathway endosomes (Rab7/Arl8-positive) in ARF6 knockdown cells. This information has been added to the discussion section on page14; lines 283-288.
3) Can the authors comment more on the effect described in Suppl Fig. 3. Is this a regional effect? Which part of the disc are we looking at?
Response: We found that clones of Ras V12 cells show elevated Hh protein levels compared to surrounding wild-type cells in eye discs irrespective of the position of the clones. We have replaced the images with lower magnification images to help readers better assess the effect of Ras V12 on Hh expression (Supplementary Fig. 3).

5) Genetic data showing that Hh signalling is important for the growth of RasV12 clones
is very nice (Fig. 4R) while the claim that RasV12 leads to upregulation of Hh activity is less convincing. While I trust the authors conclusion is most likely correct, I would like to see better data points supporting this conclusion.
Response: In addition to the initial image and the Western blot (Fig. 4a) showing elevated Ci levels in clones and in tissue lysates, respectively, we have now included an additional image to show elevated Hh signaling in Figure 4c. In a separate experiment suggested below by the reviewer, ectopic expression of Ras V12 using nub-GAL4 showed ectopic Ci activation in wing discs ( Supplementary Fig. 5), providing further evidence that oncogenic Ras stimulates Hh signaling. Ci immunostaining is routinely used and represents the standard assay for assessing Hh signaling status. In addition to Ci immunostaining, we have also used a complementing biochemical approach and found that Ci protein levels were elevated in Ras V12 cells compared to control from Western blot experiments (Fig. 4a). 7) Are these effects specific to the eye discs or seen in other discs as well?
Clones are by nature more difficult to interpret as they are randomly distributed. A simpler assay would be to induce RasV12 in a large area instead of clones, for example with Nub-G4 in the wing or a dorsal-eye-gal4 in the eye. It is nice that Arf6 RNAi treatment supresses Ci levels in both eye and wing discs. Does RasV12 induce Ci in the wing?
Response: We followed the reviewer's suggestion and expressed Ras V12 in wing discs using nub-Gal4 and stained against active Ci to assess Hh signaling. Ci activation is restricted to the anterior compartment in wild-type discs. In contrast, expression of Ras V12 clones were conducted at wandering 3rd instar larval stage following standard practice ( Fig. 2c-i). Immuno-staining for assessing signaling, however, were carried out at early 3rd instar stage when clones are in rapid growing phase and can be compared with surrounding wild-type tissue (Fig. 2a). All clone size quantification data are derived from analyzing discs of wandering 3rd instar animals. We apologize for not making this clear and have now labeled this in the figure legends (page 25, lines 617 and 621).
In addition, panel C is massively overexposed, while D appears to be underexposed relative to all the others. If A and C represent equivalent conditions, the error bars in the corresponding quantitation should be huge, and they are not. No methodology is provided describing how this quantitation was done. Since the entire study is based on quantitation of clone size, the data are impossible to interpret.
Response: The image in Figure 2c is not overexposed. Both Figures 2c and 2d were acquired under identical imaging conditions. The difference in fluorescence intensity comes from the fact that Ras V12 clones (Fig. 2c) are significantly overgrown compared to Ras V12 , spi double mutant clones (Fig. 2d). On the other hand, Figure 2a and 2c do not represent equivalent conditions. As mentioned above, growth analyses for clones were conducted at wandering 3rd instar larval stage following standard practice (Fig. 2c).
Immuno-staining for assessing signaling, however, were carried out at early 3rd instar stage when clones are in rapid growing phase and can be compared with surrounding wild-type tissue (Fig. 2a). We apologize for not making clear and have now labeled this in the figure legends (page 25, lines 617 and 621). While Ci expression is indeed higher in some clones expressing oncogenic Ras, this is not true for all of them (Fig. 4B). How do the authors interpret this?
Reviewer: The reviewer is correct, not all clones show high Ci levels. In fact, Hh signaling was particularly high in bigger, more proliferative clones. This could be because bigger clones would produce more Spi and Hh in the clones' milieu allowing ARF6 to drive a more robust Hh signaling in these cells. We have included this in the discussion (page 14, lines 291-293).
3. Arf6 is activated by a number of different GEFs. Although knockdown of Loner/Schizo does appear to weakly attenuate the growth of clones expressing oncogenic Ras (Fig. 3D), it is possible that it is also activated by other GEFs (e.g. Steppke). This should be examined more thoroughly, especially since they cite a paper describing a role Response: We examined whether Hh signaling is required in human cancer cell lines and found that it is. We directly blocked Gli1 activity in multiple lung cancer cell lines with GANT61, a specific Gli1 small molecule inhibitor 3,4 and found that it suppresses the growth of all the tested cell lines. Hh comes from the cancer cells, as Hh expression is upregulated in lung cancer cells 11 . This exciting new data has been added in Figures 3g,   4q and in the text (page 11; lines 213-219). Fig. 4U, the authors use co-precipitation to suggest an interaction between Arf6 and Hh. However, there are two problems with this figure. First, total Hh should be shown in addition to tubulin. The difference in association +/-EGFR could simply be due to differences in overall expression level. Second, according to their Methods section, coprecipitation was performed from cell homogenates in the absence of detergent.

In
Thus the two proteins may not interact at all, but are simply present in the same membranes.
Response: We apologize for inadvertently omitting to indicate that the lysis buffer contained detergent (.01% TritonX-100). However, we repeated the experiment with detergent-containing lysis buffer again and blotted for total Hh protein in the preprecipitation lysates and in the post-precipitation eluates. Although the total protein concentrations in the lysates was much lower this time because of the laborious dissections it requires to obtain highly concentrated lysates from larval discs and the time constraints on resubmission, we observed that EGFR knockdown diminished ARF6's ability to co-precipitate with Hh. This experiment is included in Supplementary Figure   4a Where is Patched (Hh receptor) under these conditions? Related to point #5, are Hh levels increased or decreased under these conditions? Where is the Hh under control conditions, if not associated with Hrs? And again, no description is provided of how quantitation of colocalization was performed. How many cells, how many discs were imaged to obtain the data shown in Fig. 4Z? Colocalization with bona fide late endosomal markers (mannose-6-phosphate receptors, cathepsins, LAMPs) would be necessary to make the authors' point here.
Response: Following the reviewer's suggestion, we used late endosomal markers (Arl8 and Rab7) to test whether ARF6 knockdown causes trafficking of Hh to the degradation pathway, as suggested by Hh protein localization to Hrs vesicles in ARF6 depleted Ras V12 cells. There is no available working Drosophila cathepsins or mannose-6-phosphate receptors antibodies for immuno-staining. We used anti-Rab7 and anti-Arl8 antibodies instead, as suggested by the other reviewers. We could not co-stain tissues with anti-Hh and anti-Arl8 (or anti-Rab7) antibodies because these antibodies were raised in the same species. Instead we used the GMR-GAL4 driver to express GFP-labeled full-length or a version of Hh corresponding to its secreted and active form (Hh-GFP and Hh-N-GFP, respectively) in cells expressing either Ras V12 alone or co-expressing ARF6-DN or ARF6-RNAi, and stained tissues against Arl8. We found that Hh-GFP preferentially localizes to clusters of Arl8-positive vesicles in Ras V12 , ARF6-DN tissues (Figure 5c, 5c', and 5g).
Together with our previous results with Hrs, our data strongly argue that ARF6 normally prevents the trafficking of Hh to the degradation pathway. These new data have now been included in the manuscript (pages 12 and 13, lines 241-258).
For the Hh/Hrs colocalization study, we analyzed 27 wild-type or 32 Ras V12 , ARF6-RNAi cells across 3 and 5 discs, respectively. The quantification in Figure 4z