Reciprocal Regulation of Shh Trafficking and H2O2 Levels via a Noncanonical BOC-Rac1 Pathway

Among molecules that bridge environment, cell metabolism, and cell signaling, hydrogen peroxide (H2O2) recently appeared as an emerging but central player. Its level depends on cell metabolism and environment and was recently shown to play key roles during embryogenesis, contrasting with its long-established role in disease progression. We decided to explore whether the secreted morphogen Sonic hedgehog (Shh), known to be essential in a variety of biological processes ranging from embryonic development to adult tissue homeostasis and cancers, was part of these interactions. Here, we report that H2O2 levels control key steps of Shh delivery in cell culture: increased levels reduce primary secretion, stimulate endocytosis and accelerate delivery to recipient cells; in addition, physiological in vivo modulation of H2O2 levels changes Shh distribution and tissue patterning. Moreover, a feedback loop exists in which Shh trafficking controls H2O2 synthesis via a non-canonical BOC-Rac1 pathway, leading to cytoneme growth. Our findings reveal that Shh directly impacts its own distribution, thus providing a molecular explanation for the robustness of morphogenesis to both environmental insults and individual variability.


Introduction
Hedgehog proteins (Hh in invertebrates and Shh or paralogs in vertebrates, collectively Hhs) are secreted morphogens playing key roles in biological processes ranging from embryonic development, proliferation, adult tissue homeostasis, and cancers [1][2][3]. Signaling by Hhs implies a complex protein journey that includes several steps incompletely characterized by, and possibly depending on physiological context [4,5].
The dual lipidation of Hedgehog ligands (palmitate at their N-terminus, cholesterol at their C-terminus after self-cleavage of the primary translation product) tends to tether them to producing cell membranes, begging the question of their mode of transport to exert their signaling function at a distance from their source [3]. Several mechanisms (not Zebrafish were maintained and staged, as previously described [41]. Experiments were performed using the standard AB wild-type strain. The embryos were incubated at 28 • C. Developmental stages were determined and indicated as hours post fertilization (hpf). The animal facility obtained permission from the French Ministry of Agriculture for all the experiments described in this manuscript (agreement No. C 75-05-12). To decrease H 2 O 2 levels, embryos were incubated in VAS-2870 (NADPH oxidase inhibitor; NOX-i) (100 nM) from Enzo Life Sciences (#BML-El395-0010, Enzo Life Sciences, Inc.; Farmingdale, NY, USA) or an equivalent amount of DMSO as a control for the duration of the time-lapse analysis. To enhance H 2 O 2 levels, D-Alanine (D-Ala, Sigma-Aldrich, St. Louis, MO, USA #A7377) (10 mM) was injected in zebrafish ventricles one hour prior to H 2 O 2 levels or filopodia analysis.

Expression Constructs, Permanent Cell Lines, and Fish Transgenic Lines
All recombinant DNA (Supplementary Table S1) were prepared by standard cloning methods. Plasmids and sequences are available on request.
Shh ligand was tagged at the position described in [48] (just down Gly198, with the intein cleavage signal aa189-198 being repeated at the C-terminus of the tag sequences) to preserve the ligand function. Tags used were: the fluorescent protein mCherry, the fluorogen-activated peptide YFAST (Plamont, 2015, exRef59), the streptavidin binding peptide (SBP) (Boncompain, 2012, exRef60), the small portion of the split nanoluciferase (HiBiT), and combinations of them.
Stable cell lines (Supplementary Table S2) were prepared under Hygromycin selection using the HeLa FlpIn-TREX cell line kindly provided by Stephen Taylor [49] and expressing the tetracycline repressor (TREX, Life Technologies).

Embryo Live Imaging and Image Processing
The larvae were anesthetized in tricaine solution and embedded in low-melting agarose (0.8%). Imaging was performed with a CSU-W1 Yokogawa spinning disk coupled to a Zeiss Axio Observer Z1 inverted microscope that was equipped with a sCMOS Hamamatsu camera and a 10× (Zeiss 0.45 Dry WD: 2.1 mm) or a 25× (Zeiss 0.8 Imm DIC WD: 0.57 mm) oil objective. DPSS 100 mW 405 nm and 150 mW 491 nm lasers and a 525/50 bandpass emission filter were used for HyPer7 imaging, and DPSS 100 mW 561 nm laser and a BP 595/50 was used for mCherry imaging. Floor plate cells imaging was performed using a Zeiss LSM 980-AiryScan 2 confocal equipped with an AiryScan detector GaAsP-PMT and 25× (Zeiss 0.8 Imm WD: 0.57 mm) or a 40× (Zeiss 1.3 Oil DIC (UV) WD: 0.22 mm) oil objectives. DPSS 10 mW 488 nm and 10 mW 561 nm lasers and 517 nm and 610 nm AiryScan emission filters, respectively, were used for GFP and mCherry acquisition. AiryScan SR mode was used and AiryScan-processed images were analyzed. To calculate the HyPer ratio, images were treated as previously described [57]. For filopodia analysis, 48 hpf Tg(2.4Shha-ABC:Gal4-FF) larvae expressing Igkmb5-DAO-mCherry were used for fluorescence acquisition as described above, and one slice of each Zstack was extracted. Slices presenting the maximum filopodia number were selected for FiloQuant processing and analysis [58].

Quantitative Secretion Assay
Cells (13,000 per well) were plated on 96-well plates (Greiner Bio-one) coated with polyornithine (10 µg/mL). After 10 h, the cells were co-transfected with a plasmid expressing Shh-SBP-HiBiT (or a secreted control, SecGFP-SBP-HiBiT) under doxycycline control and a plasmid constitutively expressing the hook (Strepta, core streptavidin-KDEL provided with a signal peptide). Some of these cells received an additional plasmid constitutively expressing a membrane-bound fusion between mCherry and DAO (Lck-Che-DAO). After 24 h, the medium was removed and cells were incubated with fresh medium containing doxycycline for 2 h (to induce HiBiT fusion expression). Medium was changed and secretion was induced with biotin (100 µM final) and, after the purified large fragment of split nanoluciferase (LgBiT) protein addition to the medium, the luciferase activity was measured 1 h later with a 96-well plate luminometer (Tristar, Berthold) as described in the HiBiT assay kit (Promega). The cells were then lysed to measure the intracellular protein expression. Normalization with biotin-untreated wells enabled us to calculate the secretion index and report the secretion efficiency.

Quantitative Internalization Assay
Cells (90,000 per well) stably expressing LgBiT were plated in 24-well culture dishes. After 24 h, the medium was changed and cells were incubated for 30 min with the medium containing HiBiT fusions (taken from cells expressing Shh-HiBiT or secreted mCherry, SecCh-HiBiT for 48 h, and adjusted on protein Luciferase activity), before incubating the cells with trypsin and removing them. After centrifugation, cells were lysed and the luciferase activity of endocytosed protein was measured with a 96-well plate luminometer (Tristar, Berthold) with a HiBiT assay kit (Promega).

Quantitative Intercellular Transfer Assay
Cells (90,000 per well) were plated in 24-well culture dishes. After 10 h, the cells were co-transfected with a plasmid expressing Shh-SBP-HiBiT under the control of doxycycline and a plasmid constitutively expressing Lck-Che-DAO. After 24 h, cells were removed using trypsin and co-cultured on 96-well plates (Greiner Bio-one) coated with polyornithine (10 µg/mL) with cells stably expressing LgBiT anchored to the extracellular side of the cell surface (siL-LgBiT-mb5). After 5 h, cells were incubated with fresh medium containing doxycycline for 2 h (to induce Shh-SBP-HiBiT expression). Medium was then changed and secretion was induced with biotin (100 µM final), and luciferase activity was measured every 1 h over 4 h with a 96-well plate luminometer (Tristar, Berthold) as described in the HiBit assay kit (Promega).

H 2 O 2 Imaging with the HyPer Probe in HeLa Cells
HyPer (H 2 O 2 ratiometric probe) fluorescence was excited with DPSS 10 mW 488 nm and 10 mW 405 nm lasers, and the corresponding YFP emission was measured using a 525/50 bandpass emission filter. Spinning-disk images were acquired using a 63× objective (63×/1.4 oil WD: 0.17 mm) on a Spinning-Disk CSU-W1 (Yokogawa) equipped with a sCMOS Hamamatsu 2048 × 2048 camera. To calculate the HyPer ratio, images were treated as previously described [57].

Quantification and Statistics
Data were analyzed using GraphPad Prism 6 and expressed as the mean ± standard error of the mean (SEM). Data distribution was first checked for normality. Comparisons between two unpaired groups were performed using the Student's t-test. For multiple conditions, ordinary one-way ANOVA followed by Tukey's multiple comparison test or two-way ANOVA followed by Sidak's or Tukey's test were performed to evaluate the significant differences. For filopodia analysis, statistical errors (SD) were estimated as √ p(1−p)/n, where p is the percentage in a class and n is the total number investigated (or √ 1/n when p = 0 or 1). The degree of significance was represented as follows: * p-value ≤ 0.05; ** p-value ≤ 0.01; *** p-value ≤ 0.001; and **** p-value ≤ 0.0001. Sample sizes and number of replicates are given in Supplementary Table S3.

H 2 O 2 Affects Shh Trafficking in HeLa Cells
To analyze the mechanism by which H 2 O 2 levels could affect Shh trafficking, we used a cellular system enabling accurate quantification of this process. HeLa cells were chosen to avoid complex feedback loops: these cells do not express endogenous Shh (nor the other two members of the family: Indian Hh, Ihh or Desert Hh, Dhh) or respond to the canonical Shh signaling pathway, but otherwise express pathway components known to be involved in Shh trafficking (lipidation, access to the extracellular space, endocytosis, and delivery to receiving cells, Supplementary Table S4). Shh trafficking from producing to receiving cells is a circuitous journey. While some aspects are still a matter of debate, there is no doubt that cis-endocytosis in producing cells can be part of the process [4,7]. Making use of Shh constructs (mouse sequence) tagged according to [48] to preserve the qualitative properties of the protein, we first verified that HeLa cells recapitulated the overall traffic of Shh ( Figure 1A-D). To distinguish different steps of the Shh journey ( Figure 1A), we exploited a double tag combining a fluorogen-activating peptide (YFAST) and a classical fluorescent protein, the monomeric Cherry (mCherry). Contrary to mCherry, in which chromophore takes time to mature, YFAST fluoresces instantly after the fluorogen addition [59], allowing Shh detection from the beginning of its journey in the endoplasmic reticulum (ER) (step 1 in Figure 1A), which is too early for mCherry detection. However, YFAST is sensitive to pH and cannot fluoresce in endosomes (step 2 in Figure 1A), contrary to mCherry. When finally sent for the receiving cells (step 3 in Figure 1A) Shh should be detectable via both tags. The transfection of HeLa cells with such a construct indeed confirmed the usefulness of this cell line for our purposes (Figure 1B). At a steady state, the majority of cells display diffuse green staining, including the ER, as well as large red spots about the size of endosomes ( Figure 1B). HeLa cells with cellular protrusions associated with yellow puncta were found, indicating detection via both tags of Shh en route to receiving cells ( Figure 1B'). The end of the journey was also easy to image by co-culturing transfected and untransfected HeLa cells. As shown in Figure 1C,C', when transfected and untransfected cells were nearby, the ShhmCherry could be detected in the untransfected cells, and the arrangement of cytoplasmic protrusions between the two cells suggested a transfer occurring via filopodia.
To reach an adequate level of precision in the analysis of the H 2 O 2 effect on Shh trafficking, we needed to synchronize its secretion. We thus made use of the Retention Using Selective Hooks (RUSH) system [60], where Shh, fused to the streptavidin-binding peptide (SBP), is retained in the ER of cells expressing streptavidin fused to the KDEL retrieving signal until the addition of biotin. As shown in Figure 1D, before the biotin addition (t = 0), Shh-SBP-AFP (Shh-SBP is fused to an AutoFluorescent protein, mCherry, here) was hooked in the ER, and the biotin addition allowed us to calibrate the timing of the Shh journey using real-time imaging. It takes approximately 30 to 40 min for the bulk of Shh to reach the Golgi (very similar to many secreted proteins in these conditions, e.g., Wnt3a [61]), secretion is abundant from 1 h, localization in endosomes is conspicuous between 2 and 3 h, and detection in (or at the surface of) non-producing cells lasts up to 5 h and then disappears. HeLa cells thus display typical features of Shh trafficking and can be used to analyze the potential effects of modifying H 2 O 2 levels if we have the means to rigorously quantify Shh in different compartments at different steps of the process. In addition, we recently developed [62] quantitative assays to track the different steps of a protein journey by combining the RUSH [60] and HiBiT systems [63] (HiBiT assay, Promega) ( Figure 2). The HiBiT system is constituted by a split luciferase: when in the same compartment, the two luciferase moieties, HiBiT (Sbi in plasmid names) and LgBiT (G/Lbi in plasmid names) may spontaneously assemble and restore the luciferase activity that can be measured by the substrate addition ( Figure 2B,D,F). These assays are inducible, quantitative, and specifically adapted to protein trafficking. We then set up this quantitative assay to separately analyze each step of Shh's journey. This strategy allowed us to determine the optimal time frame for analyzing each step of Shh trafficking in cell culture, has not yet matured) and an abundant red vesicular signal (mCherry detected in the endosomes, where the pH prevents YFAST detection) and in filopodia where both tags are detected (B'). (C) In co-culture, pairs of Shh-producing (mCherry signal) and non-producing cells (both decorated by the binding of the WGA lectin to cell coat) often display filopodia between them. Large red dots in the non-producing cell indicate Shh transfer to recipient cells, and their position suggests delivery via the filopodia (white arrows in the enlarged inset in (C')). (D) Shh-Sbp-mCherry secretion synchronized with the RUSH system displays classic timing: Shh is correctly hooked in the endoplasmic reticulum before biotin addition, reaches the Golgi 30-40 min after biotin addition, is secreted at approximately 1 h, can be easily detected within endosomes of the producing cell at 2 h, and can be visualized in non-producing cells at 4 h. Scale bars, 10 µm.
In addition, we recently developed [62] quantitative assays to track the different steps of a protein journey by combining the RUSH [60] and HiBiT systems [63] (HiBiT assay, Promega) ( Figure 2). The HiBiT system is constituted by a split luciferase: when in the same compartment, the two luciferase moieties, HiBiT (Sbi in plasmid names) and LgBiT (G/Lbi in plasmid names) may spontaneously assemble and restore the luciferase activity that can be measured by the substrate addition ( Figure 2B,D,F). These assays are inducible, quantitative, and specifically adapted to protein trafficking. We then set up this quantitative assay to separately analyze each step of Shh's journey. This strategy allowed us to determine the optimal time frame for analyzing each step of Shh trafficking in cell culture, i.e., secretion ( Figure 2C), endocytosis ( Figure 2E), and delivery ( Figure 2G). These quantitative results are consistent with our direct fluorescence microscopy observations ( Figure 1D) and were combined with redox modulation tools to test the hypothesis of the redox regulation of Shh trafficking.
i.e., secretion ( Figure 2C), endocytosis ( Figure 2E), and delivery ( Figure 2G). These quantitative results are consistent with our direct fluorescence microscopy observations (Figure 1D) and were combined with redox modulation tools to test the hypothesis of the redox regulation of Shh trafficking.  First, we studied the effects of H 2 O 2 on the primary Shh secretion. To increase H 2 O 2 levels, we expressed inducible D-aminoacid oxidase (DAO), which produces H 2 O 2 in the presence of D-Alanine (D-Ala) and is not expressed in HeLa cells [41,55,64]. For cells expressing DAO and treated with 10 mM D-Ala, we observed a specific reduction in Shh secretion, not observed with a secreted GFP (secGFP) construct taken as a control ( Figure 3A,B). This result is in close agreement with our previous observations: after oxidative treatment, cells showed reduced Shh secretion, and a pool of Shh was trapped in the Golgi apparatus [16]. Conversely, reduction of H 2 O 2 levels by the direct addition of Catalase (CAT ext ) in cell culture [41], enhanced Shh but not SecGFP secretion ( Figure 3A,C). Next, we applied the same treatments to study Shh endocytosis on cells incubated with the conditioned media of Shh-expressing cells. Compared to the control conditions (secreted mCherry, SecCh, and conditioned media), enhancing H 2 O 2 levels with DAO (by the addition of D-Ala) stimulated the Shh endocytosis ( Figure 3D,E). Conversely, the reduction in H 2 O 2 levels with CAT ext treatment reduced Shh endocytosis ( Figure 3D,F). Finally, we studied the effects of H 2 O 2 level modulation on Shh delivery to recipient cells in the co-culture assay. Cells expressing DAO and treated with D-Ala (but not untreated cells) demonstrated increased Shh delivery to recipient cells ( Figure 3G,H), while a reduction in H 2 O 2 levels with CAT ext had the opposite effect ( Figure 3G,I). Altogether, these results indicate that physiological variations in H 2 O 2 levels impact Shh trafficking in HeLa cells. This raises the interesting possibility that the heterogenous H 2 O 2 distribution could polarize Shh secretion and endocytosis in vivo. We thus decided to look more carefully at the spatial and temporal variation of H 2 O 2 distribution in the zebrafish embryo during pattern formation in the central nervous system, and whether its experimental manipulation would modify aspects of this patterning. First, we studied the effects of H2O2 on the primary Shh secretion. To increase H2O2 levels, we expressed inducible D-aminoacid oxidase (DAO), which produces H2O2 in the presence of D-Alanine (D-Ala) and is not expressed in HeLa cells [41,55,64]. For cells expressing DAO and treated with 10 mM D-Ala, we observed a specific reduction in Shh secretion, not observed with a secreted GFP (secGFP) construct taken as a control ( Figure  3A,B). This result is in close agreement with our previous observations: after oxidative treatment, cells showed reduced Shh secretion, and a pool of Shh was trapped in the Golgi apparatus [16]. Conversely, reduction of H2O2 levels by the direct addition of Catalase (CAText) in cell culture [41], enhanced Shh but not SecGFP secretion ( Figure 3A,C). Next, we applied the same treatments to study Shh endocytosis on cells incubated with the conditioned media of Shh-expressing cells. Compared to the control conditions (secreted mCherry, SecCh, and conditioned media), enhancing H2O2 levels with DAO (by the addition of D-Ala) stimulated the Shh endocytosis ( Figure 3D,E). Conversely, the reduction in H2O2 levels with CAText treatment reduced Shh endocytosis ( Figure 3D,F). Finally, we studied the effects of H2O2 level modulation on Shh delivery to recipient cells in the coculture assay. Cells expressing DAO and treated with D-Ala (but not untreated cells) demonstrated increased Shh delivery to recipient cells ( Figure 3G,H), while a reduction in H2O2 levels with CAText had the opposite effect ( Figure 3G,I). Altogether, these results indicate that physiological variations in H2O2 levels impact Shh trafficking in HeLa cells. This raises the interesting possibility that the heterogenous H2O2 distribution could polarize Shh secretion and endocytosis in vivo. We thus decided to look more carefully at the spatial and temporal variation of H2O2 distribution in the zebrafish embryo during pattern formation in the central nervous system, and whether its experimental manipulation would modify aspects of this patterning.

H2O2 Impacts Filopodial Formation and Shh Cellular Targets in the Embryonic Spinal Cord
Shh can be delivered to the receiving cells via filopodia, and it has been proposed that the number and length of filopodia impact Shh distribution [5,9,[67][68][69][70][71]. We used a double-transgenic fish 2.4Shha-ABC:Gal4FF; Shh-mChe<UAS>GFP-farn with bidirectional UAS-driven expression to visualize both Shh and plasma membranes in the MFP. Shh-mCherry was indeed detected along and at the tip of filopodia, as shown at 48 hpf ( Figure 5A,B'). To test whether physiological modifications of H2O2 levels could impact

H 2 O 2 Impacts Filopodial Formation and Shh Cellular Targets in the Embryonic Spinal Cord
Shh can be delivered to the receiving cells via filopodia, and it has been proposed that the number and length of filopodia impact Shh distribution [5,9,[67][68][69][70][71]. We used a doubletransgenic fish 2.4Shha-ABC:Gal4FF; Shh-mChe<UAS>GFP-farn with bidirectional UASdriven expression to visualize both Shh and plasma membranes in the MFP. Shh-mCherry was indeed detected along and at the tip of filopodia, as shown at 48 hpf ( Figure 5A,B'). To test whether physiological modifications of H 2 O 2 levels could impact filopodia, we expressed DAO at the level of the plasma membrane in MFP cells (2.4Shha:Gal4; UAS:Igk-mb5-DAO-mCherry), and we injected D-Ala into the spinal cord canal at 48 hpf to enhance H 2 O 2 levels in Shha-expressing cells ( Figure 5C). We then visualized the filopodia using the fluorescence of the membrane-bound mCherry ( Figure 5D) and quantified the number of filopodia per cell ( Figure 5E) as well as the length of the filopodia ( Figure 5F). Interestingly, enhancement of H 2 O 2 levels induced increases in both the number and the length of filopodia ( Figure 5E,F) Figure S1) from 30 hpf to 48 hpf. This reduction in H 2 O 2 levels induces a decrease in the filopodia number in MFP (Supplementary Figure S1). We then analyzed the distribution of OPCs in the embryonic spinal cord at 72 hpf in (Olig2:EGFP) larvae ( Figure 5G-I). A small reduction in H 2 O 2 levels (approximately 10%) was sufficient to enhance by a factor of 2 the number of OPC ( Figure 5I), known to depend on Shh activity [66,72], without affecting shh expression (Supplementary Figure S2). Thus, small modifications in H 2 O 2 levels not only modified the number and length of filopodia in vivo but also altered a Shh-dependent switch.

Shh Regulates H 2 O 2 Levels and Filopodial Growth via a Non-Canonical Route
After demonstrating that H 2 O 2 impacts Shh trafficking ( Figure 3) and signaling outcome ( Figure 5), we wondered whether this was integrated into a larger H 2 O 2 -Shh feedback loop, as first observed during adult zebrafish caudal fin regeneration [17,33]. If this were the case, Shh's interaction with cells should itself impact H 2 O 2 levels. We first observed that HeLa cells co-expressing HyPer and Shh-mCherry exhibited higher levels of H 2 O 2 than cells expressing HyPer and mCherry ( Figure 6A,B), suggesting that Shh per se affected H 2 O 2 balance. Interestingly, untransfected neighbors of cells expressing Shh-mCherry also displayed increased H 2 O 2 levels compared to neighbors of cells expressing only mCherry ( Figure 6B). This observation suggests a paracrine effect of Shh on H 2 O 2 levels. Moreover, when HyPer was addressed to the plasma membrane via the myristoylation and palmitoylation signal from the Lck tyrosine kinase (Lck-HyPer), filopodia bearing Shh were readily identified, and high levels of H 2 O 2 consistently co-localized with Shh ( Figure 6C). To quantitively assess the impact of Shh on H 2 O 2 levels, we treated a stable cell line expressing Lck-HyPer with Shh-containing conditioned medium ( Figure 6D). This treatment induced a global increase in H 2 O 2 levels ( Figure 6E,F), which was accompanied by filopodia growth, as quantified by their cumulative length ( Figure 6G).
Finally, we investigated the molecular pathway used by Shh to increase H 2 O 2 levels. Smo involvement was highly unlikely, as its mRNA was not detected in our RT-qPCR experiments (Supplementary Table S4). Indeed, cyclopamine (Shh-i), a Shh antagonist that specifically binds Smo, did not affect basal H 2 O 2 levels and did not inhibit the Shhinduced H 2 O 2 burst ( Figure 6H,I). In contrast, treatment with Nox-i efficiently inhibited the Shh-mediated H 2 O 2 increase, demonstrating the involvement of NADPH oxidase (NOX) enzymes ( Figure 6I). Rac1 was an attractive candidate to bridge Shh and NOX in our system. It is ubiquitous, activates NOX2 (the only NOX complex present in HeLa cells, Supplementary Table S4), and is well known to regulate cytoskeleton dynamics [73]; in addition, its interplay with Shh and NOX1 was previously observed in a different system [74]. We used a Rac1 inhibitor (Rac1-i) to evaluate its involvement in H 2 O 2 production induced by the Shh treatment. As shown in Figure 6J, Rac1-i itself had no effect on H 2 O 2 levels in untreated cells but was able to block the effect of Shh treatment. Rac1 thus sits at the crossroads of Shh activity on HeLa cells, leading both to H 2 O 2 production via the NOX/SOD (superoxide dismutase) complex and to the cytoskeleton remodeling that ultimately enables filopodium formation. A potential pathway to mediate Rac1 activation by Shh was recently published [75]: Shh-mediated axon guidance in the spinal cord depends on BOC receptor stimulation, leading to ELMO-Dock release and Rac1 activation. In addition, it was recently demonstrated that BOC is necessary for OPC formation in zebrafish [66] and we hypothesized it could be the receptor involved in the H 2 O 2 enhancement by Shh. As BOC (Supplementary Table S4) and ELMO2 and 3 as well as several DOCK proteins (Human Protein Atlas [76]) are expressed in HeLa cells, we tested this hypothesis using a pan-Dock inhibitor (CPYPP, DOCK-i). As shown in Figure 6J, DOCK-i per se did not affect H 2 O 2 levels but blocked the H 2 O 2 increase induced by Shh. Moreover, simultaneous treatment with DOCK-i and Rac1-i had no cumulative effect ( Figure 6J), suggesting that they belong to the same pathway. In summary, these experiments demonstrate that Shh induces a local increase in H 2 O 2 levels in HeLa cells via a non-canonical route involving activation of Rac1, most likely via BOC and the Elmo-Dock pathway. As thoroughly discussed in Acevedo and Gonzalez-Billault [73], this leads to NOX2 activation along the redox branch of Rac1 activity and membrane protrusion growth along the actin cytoskeleton branch. Very recently, Shh cytoneme formation in rodent cells was demonstrated to depend on another feedback loop, including a Disp-BOC co-receptor complex [71].

Discussion
In the present study, we first proved that three main steps of Shh traffic are sensitive to physiological variations of H2O2 levels in Hela cells: primary secretion is inversely correlated to peroxide content while endocytosis and cell-to-cell transfer are directly correlated to it.

Discussion
In the present study, we first proved that three main steps of Shh traffic are sensitive to physiological variations of H 2 O 2 levels in Hela cells: primary secretion is inversely correlated to peroxide content while endocytosis and cell-to-cell transfer are directly correlated to it.
There are several possible ways by which H 2 O 2 could affect these three steps. Our results demonstrated a clear accumulation of Shh in the Golgi apparatus upon increase in H 2 O 2 levels, and this was not a general shutdown of the secretion machinery, as control proteins unrelated to Shh were still secreted to the same extent. Up to now and apart from the oxidative distress response in the ER, not much is known about how secretion might be influenced by physiological redox fluctuation. Two cases of non-conventional secretion have been demonstrated to be influenced by oxidation: periredoxins 1 and 2 (Prdx1, Prdx2) [77] and the transcription factor Engrailed-2 [41] and fusion of secretory vesicles during the conventional secretion of insulin was also shown to be affected [78], but the mechanisms are still elusive and the cellular contexts bear no clear relationship with the Shh primary secretion studied here. It remains to be determined whether Shh is preferentially blocked in the Golgi, or more actively retrieved from downstream compartments, and the mechanisms will deserve subsequent analysis.
Concerning the endocytic step, it has long been established that endocytosis may favor redox signaling in "redosoxomes" [79]; however, the converse effect of redox potential on endocytosis has been little studied, mostly in the context of distress [80,81] or in the context of plant physiology [82]. Of note, the thioredoxin-like protein TXNL1 [80] is able to convert redox changes into modification of GDI ability to capture Rab5, thus modulating endocytosis, and this represents an interesting hypothesis to study the mechanism of Shh endocytosis stimulation by H 2 O 2 .
Another possibility linking the effect of H 2 O 2 signaling on Shh secretion and internalization would be a modification of either the expression of the receptor (Ptch) and/or co-receptors (Boc and Cdon) or their affinity for Shh. The change in expression cannot be precluded but is not the most likely explanation because the effect of H 2 O 2 modulation on Shh traffic kinetics is much faster than the time needed for new receptor molecules to reach the cell surface. A change in the affinity of pre-existing receptors is an interesting hypothesis, given that Ptch, Boc, and Cdon (all expressed by HeLa cells, Supplementary  Table S4) have been demonstrated to be endocytosed together with Shh, and found with Shh along filopodia [67,83].
The delivery step is itself a complex operation subdivided into distinct substeps, which could not be dissected in our experiments. Our results confirm that filopodia growth and density are sensitive to the redox state, but the exact place taken by filopodia and/or exosomes as well as the topology of delivery are still too uncertain to allow reasonable hypotheses on the mechanisms allowing its adjustment by H 2 O 2 . In particular, it will be necessary to clarify whether the transfer occurring at the end of filopodia takes place by the local budding of ectosomes, local release of exosomes, or by trogocytosis. In any case, cytoskeleton dynamics [84] and membrane proteins, such as integrins or disintegrins and metalloproteins [85], are already known to be susceptible to regulation by the oxidation level of their environment.
We extended these results by demonstrating that Shh distribution is also dynamic in vivo, at a time when the levels of H 2 O 2 vary rapidly in time and space in the embryo (this study and [86]), and that experimentally modifying these levels modulates filopodia formation and leads to changes in the patterning of the central nervous system. Signaling filopodia or cytonemes are a common route for the secretion of morphogens [9,87], and it was recently demonstrated that several morphogens (Fgf2, Wnt8a, Wnt3a, Shh) are involved in a regulatory loop with cytoneme formation and growth, independent of their canonical pathway [71,88].
Finally, we discovered a feedback loop by which physiological levels of H 2 O 2 adjust the trafficking of Shh, which, in turn, enhances H 2 O 2 levels. It is worth noting that, in a different context (cross-talk between Shh and H 2 O 2 during fin regeneration) we identified another type of feedback between Shh and H 2 O 2 signaling, based on mechanisms completely unrelated to the BOC-Rac1 axis [17]. As a consequence of this feedback, local and timely H 2 O 2 production may polarize the journey of Shh by modifying the rates of Shh secretion and endocytosis as well as the regulation of filopodia. We speculate that H 2 O 2 could integrate the spreading of morphogens in a developing tissue, with different morphogens responding differentially to specific concentrations of H 2 O 2 and, in turn, modifying H 2 O 2 levels. Indeed, we already know that H 2 O 2 impact on trafficking differs between Engrailed [41] and Shh (this study). This interplay between morphogens and redox signaling is also likely to have a role in various pathologies [19].

Conclusions
In a broader perspective, we propose that environmental insults or individual genetic variations inducing very subtle differences in H 2 O 2 levels could impact morphogen distribution, resulting in inter-individual differences in organism patterning or disease susceptibility.

Supplementary Materials:
The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/antiox11040718/s1; Figure S1: NOX-i treatment induces a reduction of H2O2 levels as well as a reduction of filopodia length in the MFP; Figure S2: Shha expression is not affected by NOX-i treatment; Table S1: Plasmids used in this study; Table S2: Stable HeLa cell lines used in this study; Table S3: Sample sizes and number of replicates; Table S4: Shh pathway member expression in cell lines.