Regulation of the RNA‐binding protein Smaug by the GPCR Smoothened via the kinase Fused

Abstract

From fly to mammals, the Smaug/Samd4 family of prion‐like RNA‐binding proteins control gene expression by destabilizing and/or repressing the translation of numerous target transcripts. However, the regulation of its activity remains poorly understood. We show that Smaug's protein levels and mRNA repressive activity are downregulated by Hedgehog signaling in tissue culture cells. These effects rely on the interaction of Smaug with the G‐protein coupled receptor Smoothened, which promotes the phosphorylation of Smaug by recruiting the kinase Fused. The activation of Fused and its binding to Smaug are sufficient to suppress its ability to form cytosolic bodies and to antagonize its negative effects on endogenous targets. Importantly, we demonstrate in vivo that HH reduces the levels of smaug mRNA and increases the level of several mRNAs downregulated by Smaug. Finally, we show that Smaug acts as a positive regulator of Hedgehog signaling during wing morphogenesis. These data constitute the first evidence for a post‐translational regulation of Smaug and reveal that the fate of several mRNAs bound to Smaug is modulated by a major signaling pathway.

The RNA interacting protein Smaug binds Smoothened and in response to Hedgehog signaling is phosphorylated reducing its mRNA repressive activity.

Introduction

A challenge in cell and developmental biology is to understand how inter‐ and intracellular signaling systems control gene expression. While post‐transcriptional regulation of numerous cytoplasmic transcripts is known to play key roles in gene expression by modulating their stability, localization, and/or translation (see for review 1), knowledge of how these processes are regulated by cell signaling remains rudimentary (see for instance, 2). Cytoplasmic mRNA regulation is mediated by RNA‐binding proteins (RBPs) and non‐coding RNAs which co‐recruit specific sets of mRNAs and protein factors that regulate mRNA fate 1. Notably, these RBPs and their bound mRNAs often form dynamic membrane‐less organelles (ribonucleoprotein (RNP) granules) with liquid droplet‐like behavior and whose assembly can be dynamically regulated in response to external and internal factors 3, 4, 5, 6, 7.

Studies of oogenesis and early embryogenesis in Drosophila melanogaster have been instrumental in the identification of the biological functions of cytoplasmic RBPs as well as the underlying molecular mechanisms (see for instance 8, 9, 10, 11, 12). The RBP Smaug plays key roles in fly early development as it is required for the correct anteroposterior polarization of the embryo and for the clearance of hundreds, if not thousands, of target mRNAs during the Drosophila maternal‐to‐zygotic transition 11, 13, 14, 15, 16, 17. Smaug is conserved throughout eukaryotes, and in mammals, there are two Smaug genes which control synapse biology, muscle growth, osteoblastogenesis, bone development, and the fate of embryonic neural precursor cells 18, 19, 20, 21, 22.

Smaug proteins bind their target transcripts via a conserved sterile alpha motif (SAM) domain that recognizes short stem/loop RNA structures 23, 24 and recruit proteins that repress translation or/and induce transcript degradation 25, 26, 27. Both in mammals and in insects, Smaug proteins form cytosolic bodies 25, 28. Notably, Vts1, the yeast SMAUG protein has recently been shown to form self‐templating condensates with prion‐like behavior. Despite important sequence divergence, this characteristic has been conserved in evolution, as it is also observed for hSmaug1 29.

While the multiplicity of Smaug's targets, mechanisms of action, and roles point to the importance of fine‐tuning its spatio‐temporal function, the molecular mechanisms that underlie its regulation are not well understood. Here, we reveal an unexpected connection between Drosophila Smaug and Smoothened, a G‐protein Coupled Receptor (SMO) required for the transduction of the Hedgehog signal (HH) 30, 31. We demonstrate using fly cells that Smaug physically interacts with SMO and that SMO activation by HH leads to Smaug phosphorylation and its recruitment to the inner surface of the plasma membrane. Using a novel assay for Smaug‐mediated repression, we show that HH/SMO signaling can both reduce Smaug protein levels and decrease its mRNA repressive activity. This latter effect relies on SMO promoting Smaug phosphorylation via the recruitment and activation of another positive regulator of HH signaling, the protein kinase Fused (FU) 32. Forcing the association between an activated form of FU and Smaug promotes Smaug phosphorylation and recapitulates all the effects of HH/SMO on Smaug. Activated, Smaug‐bound FU also suppresses the formation of cytosolic bodies of Smaug protein. Finally, using the wing imaginal disk as a model, we also demonstrate that HH downregulates the levels of smaug transcripts and upregulates endogenous mRNAs that had been identified to be bound and regulated by Smaug. We also show that smaug and fu mutants genetically interact. Together, these data constitute the first evidence that Smaug activity and ability to form cytoplasmic foci can be regulated by a signaling pathway via phosphorylation and reveal that Smaug could act as a positive modulator of HH signaling.

Results

Smaug interacts with SMO

We identified Smaug as a protein that binds to the cytoplasmic carboxy‐terminal tail of Drosophila SMO (amino acids (aa) 558–1,036, hereafter referred to as the cytotail) through a two‐hybrid screen. As activation of SMO involves the phosphorylation of the cytotail at a number of protein kinase A (PKA) sites, our screen employed a construct, “SMO^PKA‐SD cytotail” where all of modified serine (S) residues were mutated to aspartate (D), which mimics its activated state 33. This screen led to the identification of 258 hits corresponding to 38 prey genes (see Appendix Table S1). Among the 5 preys that have the highest Predicted Biological Score (PBS) 34, FU was found as expected since it is a known partner of SMO 35. Strikingly, Smaug was the most frequent prey and represented 137 of the positive hits.

We confirmed the interaction between SMO and Smaug by coimmunoprecipitation assays in Drosophila Cl8 cells, which are wing disk‐derived cells known to respond to HH 36. Full‐length wild‐type SMO or SMO^PKA‐SD tagged at their carboxy terminus with an HA epitope (SMO^WT‐HA, SMO^PKA‐SD‐HA, respectively) and full‐length, wild‐type Smaug amino terminally tagged with a Myc epitope (Myc‐Smaug^WT) were expressed either alone or together. Note that such epitope tags were previously shown not to interfere with the normal functions of SMO and Smaug, respectively 27, 35. Protein complexes immunoprecipitated with an anti‐HA antibody were analyzed by Western blot, which showed that Myc‐Smaug^WT coimmunoprecipitated with SMO^PKA‐SD‐HA as well as with SMO^WT‐HA both in the presence and in the absence of HH (Fig 1A).

Figure 1. Smaug and SMO physically interact and colocalize.

• A
  Coimmunoprecipitation of SMO‐HA and Myc‐Smaug proteins. Extracts of Cl8 cells expressing combinations of SMO^WT‐HA or SMO^PKA‐SD‐HA with Myc‐Smaug^WT in the absence or presence of HH were immunoprecipitated (IP) with anti‐HA. The input (lower panel) and the IP complexes (upper panel) were analyzed by Western blot with anti‐Myc (αMyc) or anti‐HA (αHA) antibodies. Here and in the other figures, the names of the proteins detected are indicated on the left and the molecular weights on the right, in kDa; the samples loaded for the input and the supernatants are equivalent to a tenth of the volume loaded for the IPs. * corresponds to higher molecular weight forms of SMO of unknown origin. Here and in panel F, the black arrow indicates unphosphorylated SMO^WT‐HA and the bracket indicates the phosphorylated forms of SMO^WT‐HA or SMO^PKA‐SD‐HA. All of the coIP data were independently reproduced at least twice.
• B
  Schematic representation of the domain structure of the Smaug protein. The conserved SSR 1 and 2 are shown in blue. They are separated by a 79 amino‐acid‐long non‐conserved region (here referred to as “M”). The sterile alpha motif (SAM) domain is in green and the pseudo heat analogous topology domain (PHAT domain, which increases the affinity of the SAM domain for SRE) in yellow. The dashed double‐arrow line at the top represents the smallest interacting region (called SID for smallest interacting domain) found according to the two‐hybrid screen. The truncated constructs used are presented below. The ability to interact with SMO^WT‐HA is indicated on the right: in green for yes, red for no. The full red double‐arrow line below represents the smallest SMO‐binding region (BR) that we could identify. The amino acid numbers correspond to Smaug‐PA. (http://flybase.org/reports/FBgn0016070.html). See also Fig EV1A–C.
• C
  Schematic representation of the C‐terminal “cytotail” of SMO. The PKA/CKI and FU phosphorylated regions are indicated as orange and green boxes, respectively. The FU‐binding region (BR) is indicated by a full green double arrow above. The truncated constructs used are presented below, and their ability to coIP with Myc‐Smaug^WT is indicated on the right. The full red double‐arrow line at the bottom represents the smallest Smaug‐binding region (BR) identified in this work.
• D, E
  Mapping of the Smaug interaction domain in SMO. Extracts from Cl8 cells transfected with Myc‐Smaug^WT and various regions of SMO fused to either HA (D) or to GFP (E) were immunoprecipitated with αMyc (D) or αGFP (E) and analyzed by Western blotting with αMyc (lower panel in D and upper panel in E), αHA (D, upper panel), or αGFP (E, lower panel). In: Input; Sup: supernatant. See also Fig EV1D–F.
• F
  Hyperphosphorylated forms of SMO do not interact with Smaug. Left panel: Extracts of Cl8 cells expressing SMO^WT‐HA with or without Myc‐Smaug^WT in the absence or presence of HH were IP with an αMyc antibody before analysis by Western blot with αHA (upper panel) or αMyc antibodies (lower panel). Right panel: Cl8 cells expressing SNAP‐SMO^WT with Myc‐Smaug^WT in the presence of HH were extracellularly labeled with a membrane‐nonpermeable fluorescent SNAP substrate before cell lysis and immunoprecipitation (IP) with anti‐Myc. Labeled SNAP‐SMO was directly visualized after electrophoresis. Note its presence in the IP fraction with a relative enrichment of the less phosphorylated forms compared to the hyperphosphorylated forms. See also Fig EV1G–J.
• G
  SMO and Smaug colocalize. Representative fluorescent images of Cl8 cells expressing GFP‐Smaug^WT alone (1‐2) or together with SMO^WT‐mCherry (3–3″ and 4–4″) either without (1, 3–3″) or with HH (2, 4–4″). The merge images in 3″ and 4″ show GFP‐Smaug^WT in green and SMO^WT‐mCh in red. The same results were seen with different fluorescent tags as well (Fig EV2B). The lack of effect of HH on GFP‐Smaug^WT in the absence of SMO^WT‐mCh is likely due to limiting amounts of endogeneous SMO. At least 20 cells were assayed for each condition. In the absence of HH, all co‐transfected cells exhibited greater than 90% colabeled SMO^WT‐mCherry. Scale bar (shown in G1, identical for all panels): 10 μm. See also Fig EV2.

Our two‐hybrid screen identified the region of Smaug between aa 74 and 291 as sufficient to bind SMO (Fig 1B). Consistent with this result, a region from aa 69 to 287 coimmunoprecipitated with SMO^WT‐HA (Figs 1B and EV1A and B). This region includes two conserved domains known as Smaug similarity regions (SSR) 1 and 2. SSR1 functions as a dimerization domain 37 and, while the role of SSR2 is unknown, a missense mutation in this domain in one of the mouse Smaug proteins, SAMD4, results in a loss‐of‐function phenotype, suggesting that it plays an important role 19. Deletion of either SSR1 or SSR2 blocks Smaug's interaction with SMO^WT‐HA, suggesting that both are necessary for this interaction (Fig EV1B and C). Deletion analysis of the SMO cytotail showed that the carboxy‐terminal 79 aa of SMO are both necessary and sufficient for its interaction with Smaug (Figs 1C–E and EV1D–F). Notably, this SMO region partially overlaps with a motif that binds the kinase FU, and is embedded in four clusters of S/T residues (green boxes in Fig 1C) whose phosphorylation is known to be induced by FU 38.

Figure EV1. Smaug–SMO interaction.

• A–C
  Extracts from transfected Cl8 cells expressing SMO^WT‐HA with different Myc‐Smaug constructs (as indicated) were immunoprecipitated with an anti‐HA (A, B) or an anti‐Myc (C) antibody prior to their analysis by Western blotting with anti‐Myc (A–C) or with anti‐HA (B, C) antibodies, respectively. In: input, IP: immunoprecipitate (beads), Sup: supernatant (left after immunoprecipitation). The samples loaded in the In and Sup lanes are equivalent to a twentieth of the volume loaded for the IP. * indicates background due to the detection of the primary antibody used for the IP. Myc‐Smaug^1‐374 and Myc‐Smaug^69‐287 interacted with SMO^WT‐HA but Myc‐Smaug^69‐199, Myc‐Smaug^121‐199, and Myc‐Smaug^121‐287 did not.
• D–F
  Extracts from transfected Cl8 cells expressing Myc‐Smaug^WT together with a set of SMO‐HA deletions (as indicated) were immunoprecipitated with an anti‐Myc antibody before Western blotting with anti‐HA (upper panels) or anti‐Myc antibodies (lower panels). SMO^Δ978‐1003‐HA and SMO^Δ978‐HA coimmunoprecipitated poorly with Myc‐Smaug, but the interaction was robust with SMO^Δ1025‐HA or SMO^Δ1004‐HA.
• G
  SMO phospho‐mutants with the corresponding phosphosites and kinases. The underlined S/T residues were mutated to D in SMO^{PKA‐SD FU‐SD} (with 21 S/T replacements by D) 38, or to A in SMO^5‐SA (with the 5S/T in the SBR being replaced). In orange: PKA sites; blue: CK1 sites; green: FU site.*: phosphosites previously identified by mass spectrometry 79, 80.
• H
  The sequence of the region of SMO that interacts with Smaug (SBR, red).The 5 S/T residues that were replaced by A in SMO^5‐SA are underlined. Four are within the 958–1,003 region required for binding to Smaug, while a fifth lies between this region and the fourth cluster of FU phosphosites 38.
• I, J
  Extracts of Cl8 cells expressing, SMO^{PKA‐SD FU‐SD}‐HA (I) or SMO^5S‐A‐HA (J) together with Myc‐Smaug, in the absence (−) or presence (+) of HH were IP with an anti‐Myc antibody before analysis by Western blot with anti‐HA (upper panel), or anti‐Myc antibodies (lower panel). SMO^{PKA‐SD FU‐SD}‐HA coIP with Myc‐Smaug. However, HH‐induced phosphorylation (presumably on other sites) precluded its interaction with Smaug. In the presence of HH, SMO^5‐SA‐HA was present mostly as phosphorylated forms that poorly coimmunoprecipitated with Myc‐Smaug. We therefore conclude that the interaction with Smaug phosphorylation is not blocked by the phosphorylation of the PKA/CKI/FU sites but by another (or others) other phosphorylation(s) outside its region of interaction with Smaug and that remains to be determined. The black arrows indicate the unphosphorylated form of SMO, and the brackets indicate the phosphorylated forms of SMO‐HA.

The activation of SMO by HH is associated with the phosphorylation of numerous sites in its intracellular C‐terminal tail, which can easily be detected as it leads to slower electrophoretic mobility 39, 40. Examination of input (In) and immunoprecipitated (IP) fractions of SMO^WT‐HA revealed that only unphosphorylated and/or partially phosphorylated SMO^WT‐HA were associated with Myc‐Smaug^WT (Fig 1F, left panel). In contrast, the most hyperphosphorylated forms of SMO remained in the supernatant (Sup) after the immunoprecipitation. This hyperphosphorylation is known to involve the sequential action of multiple kinase including the protein kinase A (PKA), the casein kinase I (CKI), and FU 38, 39. However, the loss of Smaug interaction with hyperphosphorylated SMO does not involve the phosphorylation of the PKA, CKI, or FU phosphosites, as a mutant version of SMO where all of the PKA, CKI, and FU phosphosites are mutated to D residues is still converted to a hyperphosphorylated form that inefficiently coIPs with Smaug (Fig EV1G and I). Moreover, this hyperphosphorylated form and its reduced interaction with Smaug do not depend on a putative phosphorylation of the S and T residues present in the region of SMO that binds Smaug, as their mutation into A did not allow the hyperphosphorylated SMO to interact with Smaug (Fig EV1H and J).

In summary, these data show that Smaug and SMO cytotail can physically interact both in the absence and in the presence of HH. However, the highest level of SMO hyperphosphorylation is associated with a reduction in its interaction with Smaug.

Activated SMO recruits Smaug to the plasma membrane

SMO activation is also associated with its relocalization from vesicles to the plasma membrane where it recruits an intracellular transducing complex that includes the protein kinase FU 39. We therefore analyzed whether SMO could colocalize with Smaug and affect its localization, using SMO and Smaug fusions to fluorescent tags in transfected Cl8 cells (Fig 1G). As expected 40, 41, SMO^WT‐mCherry (SMO^WT‐mCh) was present in intracellular vesicles and was relocated to the plasma membrane in response to HH (Fig EV2A). As previously described in Drosophila and others organisms 28, 29, GFP‐Smaug^WT was present in cytosolic bodies (hereafter referred to as S‐bodies and that are likely self‐templating condensates‐HA), and we found that the presence of these bodies was unaffected by HH (Fig 1G1‐2). When co‐expressed in the absence of HH, SMO^WT‐mCh and GFP‐Smaug^WT always (in 25/25 cells) strongly colocalized in cytosolic foci (Fig 1G3‐3″). Strikingly, in the presence of HH, GFP‐Smaug^WT was, instead, concentrated at or near the cell surface where it colocalized with SMO^WT‐mCh (Fig 1G4‐4″). In all (45/45) of the cells showing SMO at the cell surface, Smaug was also localized at the periphery. This change in GFP‐Smaug^WT localization depends on its interaction with SMO, as it was not observed with SMO^Δ958‐mCh, which lacks its Smaug interaction region but is still localized to the cell surface in the presence of HH (Fig EV2B). A similar relocalization of Smaug^WT at the cell surface was also observed when co‐expressed with SMO^PKA‐SD or SMO^{PKA‐SD FU‐SD} (in which the PKA and FU phosphorylated Serines (S) and Threonines (T) sites are changed to aspartic acids (D) to mimic their phosphorylation), which both accumulate at the cell surface in the absence of HH (Fig EV2C). These results seem at odds with the observation that cell surface SMO is associated with its hyperphosphorylation 33, a state that seems to prevent its interaction with Smaug. To confirm that Smaug interacts with SMO at the plasma membrane, we specifically labeled the SMO molecules that are present at the cell surface using an extracellular N‐terminal SMO fusion to the versatile SNAP tag that was labeled with a non‐permeable fluorescent substrate in intact cells (Fig EV2D). Smaug also systematically colocalized with the SNAP‐SMO^{PKA‐SD FU‐SD} present at the cell surface (Fig EV2C). We then coexpressed Myc‐Smaug with SNAP‐SMO^WT in the presence of HH, and we labeled specifically the SNAP‐SMO^WT molecules present at the cell surface before cell lysis and immunoprecipitation of Myc‐Smaug. Direct analysis after electrophoresis of the coimmunoprecipitated fraction revealed that SNAP‐SMO^WT present at the cell surface interacted with Myc‐Smaug (Fig 1F, right panel).

Figure EV2. Smaug‐SMO localization.

• A
  Fluorescent images of Cl8 cells transfected with SMO^WT‐mCherry (SMO^WT‐mCh), either without (1) or with HH (2).
• B
  Smaug colocalization with SMO depends on their physical interaction. Fluorescent images of Cl8 cells transfected with SMO^Δ1004‐mCh (1, 2, 3–3″, and 4–4″) or SMO^Δ958‐mCh (5, 6, 7–7″, 8–8″) alone (1, 2, 5, 6) or with GFP‐Smaug^WT (3–3″, 4–4″, 7–7″, and 8–8″); with or without HH. Merged images in 3″ and, 4″, 7″, and 8″ with Smaug in green and SMO in red. At least 10 cells were assayed for each condition.
• C, D
  Smaug colocalizes with cell‐surface‐activated SMO. Fluorescent images of Cl8 cells cotransfected with mCh‐Smaug^WT (1, 2, red in 1″ and 2″) along with SMO^{PKA‐SD FU‐SD}‐GFP (1′, green in 1″) or SNAP‐SMO^{PKA‐SD FU‐SD} (2′, green in 2″). Merged images: 1″ and 2″. For SNAP‐SMO^{PKA‐SD FU‐SD} (C), SNAP‐SMO was labeled with an extracellular fluorescent substrate. Nuclei were labeled with Hoechst. At least 20 cells were observed for each condition. Scale bar (shown in A1, identical for all panels): 10 μm. See more cells in Appendix Fig S1.

Together, these data show that SMO and Smaug physically interact and colocalize in cytoplasmic foci and that in response to HH signaling, a fraction of activated SMO that is present at the cell membrane directly recruits Smaug.

Activated SMO attenuates Smaug repressive effects and levels

The observed interaction between Smaug and SMO and the relocalization of Smaug to the plasma membrane by SMO suggest that HH/SMO might affect Smaug's ability to repress target mRNA expression. As SMO did not bind Smaug at or near the SAM domain, we speculated that it would regulate the repressive activity of Smaug rather than its ability to bind target mRNAs and chose to develop a tethering assay that allowed us to simultaneously monitor both Smaug levels and repression capacity in Cl8 cells (Fig 2A). This assay is an adaptation of dual tethering assays that exploit a peptide derived from the bacteriophage lambda λ N protein, which binds with high specificity and affinity to an RNA stem‐loop structure known as BoxB 42, 43, 44. Smaug was fused to the λN peptide, and a glucuronidase (GUS, from A. thaliana) reporter mRNA was generated carrying five BoxB stem‐loops in its 3′‐untranslated region (3′UTR). Both λN‐Smaug and GUS proteins were also fused to the SNAP self‐labeling peptide (leading to λN‐SNAP‐Smaug and SNAP‐GUS, respectively) allowing their respective levels to be simultaneously and directly quantified after fluorescent tagging and electrophoresis 45. An irrelevant construct encoding a SNAP‐GFP fusion served as an internal control for transfection and protein recovery efficiencies.

Figure 2. SMO/HH regulates Smaug levels and activity.

• A
  Smaug repression assay. The assay is based on the dual expression of a construct encoding a λN‐HA‐SNAP‐Smaug chimeric protein (abbreviated λN‐SNAP‐Smaug) and a second that is transcribed into a reporter mRNA (called SNAP‐GUS‐5BoxB) that carries a translational fusion between SNAP and GUS coding regions followed by five Box B hairpins (5BoxB) inserted in the 3′UTR. A plasmid encoding a GFP‐SNAP fusion (not shown here) was also used as a transfection and loading control.
• B, C
  Smaug downregulates reporter expression. Mean values of the relative levels of SNAP‐GUS reporter (SNAP‐GUS/GFP‐SNAP ratio, in arbitrary units) in extracts of transfected Cl8 cells in the absence (black) or presence of λN‐SNAP‐Smaug (red), λN‐SNAP (pale grey), or SNAP‐Smaug (dark grey). See Fig EV3A for the representative gel and Fig EV3B which shows that λN‐SNAP‐Smaug has no effect on a SNAP‐GUS construct that lacks the 5BoxB. N = 3 (biological replicates). See also Fig EV3A–C. Here and in (D–F): SNAP‐GUS/GFP‐SNAP ratios were set to 1 in cells expressing λN‐SNAP‐Smaug. The error bars correspond to the standard deviation of the mean (SD). Statistical analysis was done by a Kruskal–Wallis test by ranks followed by a Dunn test, and P values are as follows: black bracket < 0.05, red bracket < 0.01, and green bracket < 0.0001. Here and in D–E: 15 ng of pAct.λN‐SNAP‐smaug plasmid was used per transfection.
• D, E
  HH/SMO reduces Smaug levels and its overall negative effect on reporter expression. Mean values of the relative levels of reporter expression (determined as above) (D) or of λN‐SNAP‐Smaug (ratio λN‐SNAP‐Smaug/GFP‐SNAP) (E) in the absence of SMO constructs (red) or in the presence of SMO^WT‐HA (green), SMO^PKA‐SD‐HA (blue) or SMO^Δ958‐HA (yellow), without HH (plain boxes) or in presence of HH (striped boxes). N = 6 (independent biological replicates), except for “no SMO, no HH” and “SMO^WT with HH” conditions where n = 10 (independent biological replicates). A representative gel is shown in Fig EV3D. Note that SMO^WT‐HA or HH alone had reproducible weak effects that were observed in two independent biological triplicates but were not statistically significant. See also Fig EV3E and F.
• F
  HH/SMO reduces Smaug's intrinsic repressive ability. A similar experiment was conducted with different amounts of λN‐SNAP‐smaug expressing plasmid (ranging from 0 to 50 ng) as indicated. The mean values of the relative levels of reporter (y‐axis) are plotted against the relative levels of λN‐SNAP‐Smaug (x‐axis). SNAP‐GUS/GFP‐SNAP ratios were set to 1 in cells expressing λN‐SNAP‐Smaug in the absence of SMO^PKA‐SD/HH. The colored area represents the extent of the values observed for the relative amounts of reporter expression obtained with (blue) and without (red) SMO^PKA‐SD/HH, respectively. N = 3 (biological triplicates). See Fig EV3H and also a representative gel in Fig EV3G.

First, we validated our system by verifying that λN‐SNAP‐Smaug repressed SNAP‐GUS levels 2‐ to 2.5‐fold in a manner that depended upon both the BoxB sites in the reporter mRNA and the expression of λN‐SNAP‐Smaug (Figs 2B and C, and EV3A and B). Given the nature of our assay, the decrease in the SNAP‐GUS reporter levels could reflect a reduction in the levels of the snap‐gus‐5BoxB mRNA, of its translation or both. We therefore monitored the amount of snap‐gus‐5BoxB‐GUS mRNA by reverse transcription followed by quantitative PCR (RT–qPCR). As shown in Fig EV3C, λN‐SNAP‐Smaug reduces the amount of snap‐gus‐5BoxB‐GUS mRNA almost twofold. Although we cannot exclude additional translational regulation, this indicates that λN‐SNAP‐Smaug reduces stability of the reporter mRNA.

Figure EV3. Smaug levels and repression assay.

1. Direct fluorescent imaging of an electrophoretic gel with extracts (labeled with a fluorescent SNAP substrate) of Cl8 cells that express the snap‐gus‐5BoxB reporter (encoding the SNAP‐GUS protein), in the absence (black box) or presence of λN‐SNAP‐Smaug (red), λN‐SNAP (pale grey), or SNAP‐Smaug (dark grey). GFP‐SNAP was used as a control for transfection and protein extraction efficiency to normalize the amounts of SNAP‐GUS and SNAP‐Smaug/λN‐SNAP‐Smaug. The λN‐SNAP‐Smaug, SNAP‐GUS, and GFP‐SNAP simultaneously produced in this assay have distinct molecular weights (respectively, 137, 91, and 50 kDa) and could therefore be simultaneously detected and quantified. The dashed line indicates that several irrelevant central lanes of the gel were removed.
2. Mean values of the relative levels of SNAP‐GUS reporter that lacks the five BoxB. (SNAP‐GUS/GFP‐SNAP ratio, in arbitrary units) in extracts of transfected Cl8 cells in the absence (black) or presence of λN‐SNAP‐Smaug (red). No difference was observed. Transfection and analysis conditions were as in Fig 2B and C.
3. Relative mRNA levels estimated for snap‐gus‐5BoxB in Cl8 cells with or without expression of SMO^PKA‐SD‐HA with HH. mRNA levels for snap‐gus‐5BoxB were measured using semi‐quantitative RT–PCR from Cl8 cells that express the snap‐gus‐5BoxB reporter, in the absence (black box) or presence of λN‐SNAP‐Smaug expression (red). Primer sequences used: SNAP‐Gus (Fw): 5′‐CGTGAAAGAGTGGCTGCTG‐3′ and SNAP‐Gus (Rev): 5′‐GGTTTCTACAGGACGGA CCA‐3′. mRNA levels were set to 1 in the absence of λN‐SNAP‐Smaug. Here and in Fig EV3F: N = 3 (biological triplicates), the measurements were performed three times per replicates and the results are presented as the means ± SE. The differences between groups were assessed using Student's t‐tests with Prism V8.2.1 software. A P value < 0.05 was considered to indicate a statistically significant difference*.
4. Representative gel of Fig 2D.
5. Mean values of the relative levels of SNAP‐GUS‐5BBox reporter (SNAP‐GUS/GFP‐SNAP ratio, in arbitrary units) in extracts of transfected Cl8 cells in the absence (black) or presence of SMO^WT +HH (hatched green) or SMO^PKA‐SD (blue). No λN‐SNAP‐Smaug.
6. Relative mRNA levels estimated for λN‐SNAP‐smaug—measured using semi‐quantitative RT–PCR—in Cl8 cells expressing λN‐SNAP‐smaug with or without SMO^PKA‐SD‐HA with HH. mRNA levels were set to 1 in the absence of SMO^PKA‐SD‐HA +HH.
7. Representative gel of Fig 2E (and EV3H).
8. Relative amounts of SNAP‐Smaug protein (over SNAP/GUS levels) as a function of the amounts of pAct.λN‐SNAP‐smaug‐HA plasmid with (blue) and without (red) HA‐SMO^PKA‐SD/HH. The statistical analysis was done by one‐tailed bivariate Wilcoxon rank test. N = 3. P value for each pair (with and without HH) = 0.05.

Next, we assayed the effect of HH/SMO on the expression of the reporter in the presence of λN‐SNAP‐Smaug (Figs 2D and EV3D). While SMO^WT‐HA or HH alone had no statistically significant effect, SMO^WT‐HA in the presence of HH led to an increase in SNAP‐GUS levels indicating a reduction in λN‐SNAP‐Smaug‐dependent repression. A similar effect was seen with SMO^PKA‐SD‐HA in the absence of HH. The effects of SMO^WT‐HA or SMO^PKA‐SD‐HA on the reporter reflected a reduction in its repression by λN‐SNAP‐Smaug rather than a direct, Smaug‐independent, effect as they had no significant effect on the reporter in the absence of λN‐SNAP‐Smaug (Fig EV3E). Moreover, SMO^Δ958‐HA which is unable to bind and colocalize with Smaug (see above) had no effect (despite the presence of HH; Figs 2D and EV3D). This strongly suggests that SMO needs to be associated with Smaug to negatively regulate the Smaug‐bound reporter.

The upregulation of the SNAP‐GUS reporter levels in the presence of activated SMO could be due, at least in part, to the downregulation of λN‐SNAP‐Smaug levels. We found that SMO^WT‐HA co‐expression or the presence of HH alone had no statistically significant effect on λN‐SNAP‐Smaug levels (Figs 2E and EV3D). However, SMO^WT‐HA and HH together led to a significant decrease in Smaug levels. SMO^PKA‐SD‐HA alone had a similar negative effect on λN‐SNAP‐Smaug levels which were not significantly increased by HH. Again, no effect was seen with SMO^Δ958‐HA (despite the presence of HH). The effect of SMO^PKA‐SD‐HA on λN‐SNAP‐Smaug levels was associated with a similar reduction in the levels of its mRNA that was monitored by RT–qPCR (Fig EV3F).

Finally, to assess the contribution of the reduction in λN‐SNAP‐Smaug levels to the decrease in λN‐SNAP‐Smaug repressive activity, we monitored the expression of the SNAP‐GUS reporter in response to different doses of λN‐SNAP‐Smaug, either in the absence or in the presence of SMO^PKA‐SD‐HA (with HH) (Figs 2F and EV3G and H). We transfected the cells with different amounts of the λN‐snap‐smaug expression vector (ranging from 5 to 50 ng). As expected from the above results, the amounts of λN‐SNAP‐Smaug for a given dose of vector were always lower in the presence of SMO^PKA‐SD‐HA/HH than in its absence. In both cases, the expression of the SNAP‐GUS reporter decreased exponentially in a λN‐SNAP‐Smaug dose‐dependent manner. However, for the same amount of λN‐SNAP‐Smaug protein, the levels of the SNAP‐GUS reporter were systematically higher in the presence of SMO^PKA‐SD‐HA and HH than in their absence, indicating that SMO activation reduces the intrinsic repressive activity of Smaug.

In summary these results reveal that SMO and HH signaling attenuate Smaug function in two ways: by reducing its levels and by downregulating its repressive activity. Both components depend on SMO activation and its ability to interact directly with Smaug.

SMO activation promotes Smaug phosphorylation in a FU kinase‐dependent manner

As HH signaling leads to the phosphorylation of many members of its pathway, we next tested the possibility that the effects of activated SMO on Smaug could be due to the phosphorylation of the latter. As shown in Fig 3A, SMO^WT‐GFP induced an electrophoretic mobility shift of HA‐Smaug^WT in the presence of HH, and this effect was absent either when SMO^WT‐GFP was not co‐expressed or in the absence of HH. A similar shift was seen when HA‐Smaug^WT was co‐expressed with SMO^PKA‐SD‐GFP. In all cases, treatment of the extracts with a phosphatase 33, 46, 47 abolished the slower migrating forms induced by activated SMO, demonstrating that these are phosphorylated forms of Smaug (Fig EV4A). These effects depended on the ability of SMO^WT‐GFP (and SMO^PKA‐SD‐GFP) to interact with Smaug as SMO^{PKA‐SD, Δ978}‐GFP and SMO^Δ958‐GFP were unable to induce the phosphorylation of HA‐Smaug^WT (Fig 3B). This is unlikely to be due to a lack of function of these SMO constructs since SMO^Δ978‐GFP is known to be constitutively active 35.

Figure 3. SMO/HH activation promotes the phosphorylation of Smaug.

1. HH/SMO promotes slow‐migrating forms of Smaug. Western blot analysis of Cl8 cells that transiently express HA‐Smaug^WT alone, or together with either SMO^WT‐GFP or SMO^PKA‐SD‐GFP, in the presence or in the absence of HH. GMAP serves as a loading control. Here and in the other panels, the antibodies used are indicated on the left. Here and in (B, C), the black arrows indicate the unphosphorylated form of HA‐Smaug^WT and the brackets indicate the slower migrating phosphorylated forms of HA‐Smaug^WT. NT: not transfected. See a phosphatase assay in Fig EV4A.
2. Smaug phosphorylation requires its interaction with SMO. Western blot analysis of Cl8 cells expressing HA‐Smaug^WT, in the presence of HH, in combination with SMO^Δ958‐GFP, SMO^PKA‐SD‐GFP, or SMO^{PKA‐SD, Δ978}‐GFP, as indicated.
3. FU controls Smaug phosphorylation. Extracts of Cl8 cells expressing SNAP‐Smaug with SMO^{PKA‐SD FU‐SD}‐GFP (with HH) in combination with GFP‐FU^WT, GAP‐GFP‐FU^WT, Myc‐FU^EE, or GAP‐GFP‐FU^DANA (as indicated) were labeled for the SNAP tag before electrophoresis. After quantification of the gels, the percentage (%) of phosphorylated Smaug (% of phosphorylated forms of SNAP‐Smaug total amounts of SNAP‐Smaug) was estimated. In (C–E), the mean values and SD for independent biological triplicates (N = 3) are shown in the graph at the bottom, and the statistical analysis was done by one‐tailed bivariate Wilcoxon rank test. See also Fig EV4C and D.
4. Phosphorylation of Smaug by activated FU requires the co‐expression of SMO. Mean values of the percentage of phosphorylated Smaug in extracts of Cl8 cells that express λN‐SNAP‐Smaug with GAP‐GFP‐FU^WT and SMO^WT alone or together (as indicated) N = 3 (biological triplicates). Note that the effect of GAP‐GFP‐FU^WT and SMO^WT together is lower than the one observed with GAP‐GFP‐FU^WT and SMO^{PKA‐SD FU‐SD}‐GFP together (in panel C). This could have multiple nonexclusive causes, including the fact that SMO^WT is present at much lower levels than SMO^{PKA‐SD FU‐SD} 38 and/or the involvement of another kinase. See also Fig EV4B. Note that Smaug does not coIP FU unless SMO is present see Appendix Fig S2.
5. Forcing the interaction of Smaug with activated FU leads to a SMO‐independent phosphorylation of Smaug. Mean values of the percentage of phosphorylated Smaug in Cl8 cells coexpressing λN‐SNAP‐Smaug with GFP‐FU, FU‐SBR, FU^EE‐SBR, or FU^DANA‐SBR, with or without HH, as indicated. N = 3 (biological triplicates). See also Fig EV4C and D. Note that the levels of Smaug phosphorylation observed in the presence of FU^DANA‐SBR are not lower than those seen with GFP‐FU. This suggests an incomplete inhibition of endogenous FU, which may be due to the trapping of FU^DANA‐SBR by Smaug. See a representative blot in Fig EV3C and a phosphatase assay in Fig EV4D. The mapping of the phosphosites is shown in Fig EV4E–G.

Figure EV4. Characterization of Smaug phosphorylation.

• A
  Phosphatase assay. Extracts of Cl8 cells expressing either HA‐Smaug alone or with either SMO^WT‐GFP or SMO^PKA‐SD‐GFP in the presence of HH were analyzed by Western blotting. Extracts were directly put in the loading blue or treated 30 min at 30°C with buffer (as recommended by NEB), either with antiphosphatase (Phos Inhibitors, Roche) or with λ phosphatase. (λphos, from NEB) as indicated. Phosphatase, or with lambda phosphatase.
• B
  Representative gel of Fig 3D. Here, CLIP‐GUS was used as an internal control.
• C
  Representative gel of Fig 3E.
• D
  Phosphatase assay. Cl8 cells expressing SNAP‐Smaug alone or with FU^EE‐SBR were labeled in vivo with SNAP‐Cell‐TMR‐STAR at 25°C for 30 min before lysis. The extracts were heated for 15 min at 65°C to inactivate endogeneous phosphatases, buffered in the charge blue (first lane) followed by 30 min at 30°C in the absence (second line) or in the presence (third line) of λ phosphatase.
• E–G
  Mapping of the S/Ts required for the phosphorylation of Smaug induced by HH/SMO/FU signaling. Schematic representation of the eleven forms of Smaug in which the S/T was replaced by A (F). See sequence in Appendix Fig S3. The red boxes represent the regions mutated in the different constructs that are shown in E and G. See also legend of Fig 1B. Electrophoresis of extracts of Cl8 cells expressing SMO^PKA‐SD‐GFP with HH (E) or FU^EE‐SBR (G) with various forms λΝ−SNAP‐Smaug (wt or mutated as indicated) that were labeled after cell lysis. The three FU putative sites mutated are 372: SINPLCD, 394: SLGLSLE, 464: TEILDFD that fit the S(X)₅E/D consensus proposed by 53.

The kinase FU binds to the cytotail of SMO near the Smaug‐binding region (see Fig 1C). In response to HH, SMO was shown to recruit FU to the plasma membrane promoting its activation and the phosphorylation of its targets including SMO itself and downstream members of the HH pathway 38, 48, 49. This raised the possibility that FU could also be involved in the phosphorylation of Smaug upon HH/SMO activation. We therefore analyzed the effects of different forms of FU in the presence of HH and SMO^{PKA‐SD FU‐SD}‐GFP (Fig 3C): (i) GFP‐FU^WT 35; (ii) GAP‐GFP‐FU^WT, which is constitutively activated through fusion of FU to the GAP43 palmitoylated domain, which targets the protein to the plasma membrane 32, 50; (iii) Myc‐FU^EE, another activated form of FU, where Thr‐151 and Ser‐154 were both mutated to glutamate 51; and (iv) GAP‐GFP‐FU^DANA a kinase‐dead form of GAP‐FU. FU^DANA overexpression is a common tool to downregulate endogenous FU activity as it acts in a dominant‐negative fashion, likely via dimerization with endogenous FU 32, 50. We used SMO^{PKA‐SD FU‐SD} (fused to GFP, SMO^{PKA‐SD FU‐SD}‐GFP) in these experiments to avoid the effects of a known positive feedback between FU and SMO that occurs via the phosphorylation by FU of SMO's cytotail 38. To ensure precise quantification, Smaug was fused to the SNAP peptide. In the presence of SMO^{PKA‐SD FU‐SD}‐GFP alone, ~18% of SNAP‐Smaug was phosphorylated. Co‐expression of GFP‐FU^WT had no significant effect, whereas co‐expression of the constitutively activated forms GAP‐GFP‐FU^WT or Myc‐FU^EE further increased SNAP‐Smaug phosphorylation by approximately 3‐ and 2.5‐fold, respectively. In contrast, when GAP‐GFU^DANA was co‐expressed, there was a 40% reduction in phosphorylated SNAP‐Smaug (to ~12%) compared to SMO^{PKA‐SD FU‐SD}‐GFP alone. This shows (i) that the induction of Smaug phosphorylation by GAP‐GFP‐FU depends on its kinase activity and (ii) that FU activation is, at least in part, required for the full induction of Smaug phosphorylation by activated SMO.

The above data indicate that SMO promotes Smaug phosphorylation by activating FU. The ability of SMO to bind both FU and Smaug through adjacent protein regions could bring them in close proximity, thereby facilitating phosphorylation of Smaug by FU. We, therefore, tested whether a constitutively active form of FU could promote the phosphorylation of Smaug in the absence of SMO co‐expression. As shown in Fig 3D (and Fig EV4B), in the absence of HH, the percentage of phosphorylated SNAP‐Smaug was more than twice as high when GAP‐GFP‐FU and SMO‐GFP^WT were expressed together than separately. This synergic effect indicates that SMO is required for the effects of activated FU, and we then tested whether co‐expression of SMO was still necessary when we forced a direct interaction between FU and Smaug by fusing the Smaug‐binding region (SBR) present in SMO to the N‐terminus of wild‐type or mutant forms of FU (Figs 3E and EV4C). Surprisingly, FU‐SBR alone resulted in ~4× fold increase in SNAP‐Smaug phosphorylation when compared to the levels observed in the presence of GFP‐FU. A potential explanation for this activation of FU in the absence of HH is that a multimerization of Smaug may facilitate FU dimerization, an event known to be sufficient to promote its activation 52. The phosphorylation of SNAP‐Smaug was even higher with FU^EE‐SBR (with more of 70% of Smaug being phosphorylated), while it was strongly reduced in the presence of FU^DANA‐SBR (23% of phosphorylated Smaug), confirming the requirement for FU kinase activity.

Smaug contains 181 S and T (of a total of 999 amino acids). To map Smaug's phosphosites, we analyzed the effects of HH/SMO/FU on variants of smaug in which eleven large segments were, respectively, substituted by synthetic fragments in which the S/T codons were replaced by A codons (Fig EV4E–G). Two of the Smaug variants (Smaug^{274‐399 SA} in which the 38 S/T of the region 5 that spans aa 274–399 were mutated and Smaug^{400‐502 SA} in which the 15 S/T of the region 6 spanning aa 400–502 were mutated) showed a strong reduction—but not a total suppression—in the phosphorylation shift induced by FU^EE‐SBR or SMO/HH (Fig EV4F). Mutation of the 14 S/T is presented in the region between aa 372 and aa 434 a similar effect. This region contains three sequences (at positions 372, 394, and 434) that could correspond to a phosphorylation consensus site for FU (S(X)₅D/E) 53. Their simultaneous mutation led to a modest reduction in the percentage of Smaug molecules phosphorylated in the presence of FU^EE‐SBR.

Together these data show that the FU kinase promotes—likely directly although we cannot exclude an indirect effect via another kinase that would remain to be identified—the phosphorylation of Smaug in response to HH/SMO signaling. This phosphorylation requires S/T‐enriched regions located in the central region of Smaug, between the SSR1 and the SAM domain. In this process, it is likely that SMO both promotes FU activation and bridges it to Smaug.

The FU kinase downregulates Smaug repressive activity

The induction of Smaug phosphorylation by HH/SMO signaling relies, at least in part, on the recruitment and activation of FU. This raised the possibility that FU could also mediate the negative effects exerted by HH/SMO signaling on Smaug. To test this, we further analyzed the effects of FU^EE‐SBR on Smaug levels and repressive capacity. As shown in Figs 4A and EV5A, increasing levels of the fu ^EE ‐SBR expression plasmid upregulated the percentage of λN‐SNAP‐Smaug phosphoforms in a dose‐dependent manner, starting from a background of ~5% in the absence of fu ^EE ‐SBR plasmid to almost 70% with 60 ng. In parallel, fu ^EE ‐SBR expression upregulated the levels of SNAP‐GUS reporter expression (Fig EV5A and B). This effect required the presence of λN‐SNAP‐Smaug (Fig EV5C). The expression of the reporter correlated with the fraction of phosphorylated λN‐SNAP‐Smaug (Fig 4B). This indicates that the phosphorylation of λN‐SNAP‐Smaug induced by activated FU could reduce its repressive activity on the SNAP‐GUS reporter. In addition, fu ^EE ‐SBR expression also downregulated λN‐SNAP‐Smaug protein levels (Figs 4C and EV5A). This latter effect did not vary with the dose of fu ^EE ‐SBR and did not depend on the percentage of phospho‐Smaug induced by fu ^EE ‐SBR. Moreover, the kinase‐inactive form, FU^DANA‐SBR, also downregulated Smaug levels (Fig EV4D and E), which indicates that FU regulates Smaug levels independently of its kinase activity.

Figure 4. FU downregulates Smaug levels and repressive effect.

• A, C
  Effect of FU^EE‐SBR on Smaug phosphorylation and levels. Extracts of Cl8 cells transfected with λN‐SNAP‐smaug (25 ng) and SNAP‐GUS‐5BoxB (300 ng) plasmids in the absence or with different amounts of FU^EE expressing plasmid (ranging from 15 to 60 μg, as indicated) were analyzed by electrophoresis before quantification as above. See a representative gel in Fig EV5A. The % of phosphorylated SNAP‐Smaug protein (determined as in Fig 3) (A) and relative levels of λN‐SNAP‐Smaug protein (C) (determined as in Fig 2) are represented as a function of the levels of FU^EE‐SBR. The mean values and SD for independent biological triplicates (N = 3) are shown in the graph at the bottom. The statistical analysis was done by one‐tailed bivariate Wilcoxon rank test with a P value = 0.05 (*). Here, CLIP‐GUS was used as an internal control.
• B
  Effect of FU^EE‐SBR on Smaug activity. The relative levels of reporter expression obtained in the same experiment as in (A, C) are represented as a function of the % of phosphorylated Smaug. Color of data points corresponds to ng of transfected FU^EE‐SBR in (A–C). See also Fig EV5B–D.
• D
  Effect of FU^EE‐SBR on Smaug subcellular localization. Representative fluorescent images of Cl8 cells expressing GFP‐Smaug^WT (1), FU^EE‐SBR‐mCherry (2), FU^DANA‐SBR‐mCh, (3) or expressing GFP‐Smaug^WT together with FU^EE‐SBR‐mCherry (4–4′’) or GFP‐Smaug^WT together with FU^DANA‐SBR‐mCh (5–5′’). The merge images in 4″ and 5″ show GFP‐Smaug^WT in green and FU^EE‐SBR‐mCh or FU^DANA‐SBR‐mCh in red. (6) In the presence of FU^EE‐SBR, 274 of a total of 294 cotransfected cells show a diffuse localization of Smaug and FU and no foci. In the presence of FU^DANA‐SBR, 100% of the cotransfected cells (n = 241) show Smaug and FU distributed in bodies and colocalized. 250 ng of plasmid was used for each construct. Scale bar (shown in D1, identical for all panels): 10 μm.

Figure EV5. Effect of FU^EE‐SBR on Smaug levels and activity.

• A, B
  Forcing the interaction of Smaug with activated FU reduces Smaug negative effect on reporter expression. Extracts of Cl8 cells expressing λN‐SNAP‐Smaug and the SNAP‐GUS‐5BoxB reporter in the absence or with different amounts of FU^EE‐SBR expressing plasmid (ranging from 15 to 60 μg) as indicated were analyzed by electrophoresis (representative gel in A) before quantification (B) as above. N = 3 (biological triplicates). The relative reporter levels of the SNAP‐GUS (calculated as in Fig 2) are represented as a function of the levels of FU^EE‐SBR levels. The mean values and SD for independent biological triplicates (N = 3) are shown in the graph at the bottom. The statistical analysis was done by one‐tailed bivariate Wilcoxon rank test with a P value = 0.05 (*).
• C
  Effect of FU^EE‐SBR on the relative levels of λN‐SNAP‐Smaug. Relative mRNA levels estimated for λN‐SNAp‐smaug—measured using semi‐quantitative RT–PCR—in Cl8 cells with or without expression of FU^EE‐SBR.
• D
  Forcing the interaction of Smaug with activated FU leads to a downregulation of Smaug levels independently of Fused's kinase activity. Relative mean values of the total λN‐SNAP‐Smaug levels in Cl8 cells coexpressing GFP‐FU, FU‐SBR, FU^EE‐SBR, or FU^DANA‐SBR, with or without HH, as indicated. N = 3.
• E
  Effect of SMO/HH on other endogeneous Smaug target mRNAs in Cl8 cells. Quantification by RT–qPCR of Smaug's targets mRNAs in Cl8 cells that express FU^EE‐SBR (black), FU^EE (dark grey), FU^DANA‐SBR (light grey), or SMO^PKA‐SD‐HA (with HH) (checkered) was reported with respect to a GFP control. mRNA levels were set to 1 for the GFP control. Mitochondrial protein encoding mRNA is in blue.

In conclusion, these data show that the FU kinase induces the phosphorylation of Smaug and negatively regulates Smaug by at least two means: It attenuates Smaug's mRNA repressive activity in a phosphorylation‐dependent manner and reduces Smaug levels independent of its kinase activity.

The FU kinase suppresses S‐body formation

We next addressed whether FU^EE‐SBR has an effect on Smaug's subcellular distribution (Fig 4D). We co‐expressed GFP‐Smaug^WT with FU^EE‐SBR tagged with mCherry. Strikingly, the presence of FU^EE‐SBR‐mCh completely suppressed the formation of GFP‐Smaug^WT bodies, leading to a more uniform distribution. The effect was due to the kinase activity of FU^EE‐SBR‐mCh, as the kinase‐dead form of FU did not affect S‐body formation. Whereas FU^EE‐SBR‐mCh alone displayed diffuse cytoplasmic localization both in the presence or in the absence of GFP‐Smaug, the kinase‐dead FU^DANA‐SBR‐mCh was mostly found to colocalize with the S‐bodies. Note that the drop in Smaug levels described above is not the cause of the absence of S‐bodies in the presence of FU^EE‐SBR‐mCh as both FU constructs have a similar effect on Smaug levels.

The FU kinase upregulates Smaug endogenous targets

We next examined whether HH/SMO/FU could also regulate endogenous Smaug target mRNAs in Cl8 cells. We assessed the effect of FU^EE‐SBR on the endogenous levels of thirteen Smaug mRNA targets that include the following: (i) its well‐characterized mRNA target Hsp83, (ii) five mRNAs that were among the top 15 Smaug targets from early embryo data (as defined by their enrichment in Smaug IPs and the extent of stabilization and translational upregulation in smaug mutants compared to wild type; Chen et al 13 and Tadros et al 15, and (iii) eight mRNAs that were shown to bind Smaug and encode mitochondrial proteins 54. These latter mRNA were chosen as HH signaling was reported to regulate mitochondrial function. FU^EE‐SBR significantly upregulated the levels of eighth of these mRNAs (Figs 5A and EV5E). These effects require the kinase activity of FU and its binding to Smaug (to a lesser extend) as FU^DANA‐SBR and FU^EE had no or little effect, respectively. Importantly, in almost all cases that we tested SMO ^PKA‐SD (with HH) recapitulated the effects of FU^EE‐SBR: It increases the levels of ND‐23, ND‐75, Rpn1, and RFeSP that were affected by FU^EE‐SBR but had no significant effect on ND‐42 and SdhA transcripts that were also not significantly affected by FU^EE‐SBR.

Figure 5. HH signaling and Smaug control each other in vivo .

1. Effect of FU^EE‐SBR and HH/SMO on endogeneous Smaug target mRNAs in Cl8 cells. Relative mRNA levels estimated using semi‐quantitative RT–PCR for various Smaug's targets in Cl8 cells expressing FU^EE‐SBR (black), FU^EE (dark grey), or FU^DANA‐SBR (light grey) were reported with respect to levels of a GFP control. mRNA levels were set to 1 for the GFP control. The list of genes and the sequences of the primers used are shown in Appendix Tables S2 and S3. See also Fig EV5E. Here and in Fig EV5D: RNA encoding mitochondrial proteins are written in blue. N = 3 (biological triplicates) and the quantifications were repeated 3 times on each sample. The results are presented as the means ± standard error. The differences between groups were assessed using Student's t‐tests with Prism V8.2.1 software. A P value <0.05 was considered to indicate a statistically significant difference*.
2. Effect of HH on endogeneous targets of Smaug in the wing imaginal disk. Relative levels of Smaug's targets mRNAs in MS1096; UAS‐hh or hh ^ts and their respective controls (as in Fig EV5X). N = 3, with 40 disks for each phenotype.
3. Effect of smaug mutation on [fu1] wing class distribution. Percent distribution of phenotypic classes (defined in Appendix Fig S4) in fu ¹ males in presence of zero (fu ¹ ), one (fu ¹;;smaug ⁴⁷ /+), or two copies (fu ¹;; smaug ⁴⁷ /smaug ⁴⁷) of the smaug ⁴⁷ allele. The wings were double‐blindly classified in “weak” and “strong” according to the LV3‐V4 defects (see Fig EV6B). *Chi‐square statistical analysis gave a P of 0.015 for fu ¹ and fu ¹;;smaug ⁴⁷ /+ distributions and of 0.0024 when comparing fu ¹;;smaug ⁴⁷ /+ to fu ¹;; smaug ⁴⁷ /smaug ⁴⁷.

Next, we tested whether HH signaling affected the function of Smaug in vivo using as a model the wing imaginal disk, whose development is controlled by HH. We tested the effects of HH on five of the mRNAs that were regulated by FU^EE‐SBR and/or HH/SMO^PKA‐SD in the Cl8 cells (Gapdh2, COX7, ND‐23, ND‐75, and Rpn1). As shown in Fig 5B, overexpression of hh in the whole wing pouch (using the UAS/GAL4 system) strongly (from twofold to ninefold, depending on the mRNAs) increased the levels of all these transcripts (compared to a control that expresses a GFP under the GAL4 driver) and conversely, the inactivation of HH activity using a thermosensitive allele (hh ^ts2) at restrictive temperature significantly downregulated their level when compared to a wild‐type strain raised at the same temperature). Notably, these effects revealed opposing responses of smaug mRNA levels to manipulations of HH as they are upregulated when HH levels are increased and reduced (from x to y folds) when HH is inactivated.

In summary, our in vivo data demonstrate that in both transfected Cl8 cells and fly, HH/SMO signaling decreases the levels of the smaug mRNA and downregulates the levels of several of its mRNA targets.

smaug and fu genetically interact during wing morphogenesis

Taken together our results suggest that Smaug may be involved in the regulation of HH signaling. smaug loss‐of‐function mutants are viable and display no obvious adult phenotypes 55. However, since the effects of loss of another FU‐binding partner and target, SU(FU), are only visible in the absence of FU kinase activity 56, we tested the effect of the smaug ⁴⁷ mutation in combination with fu ¹, an allele that encodes a kinase‐dead form of FU 56. fu ¹ leads to a mild defect in wing patterning characterized by a narrowing of the intervein between longitudinal veins 3 and 4 (LV3‐4). As this fu ¹ phenotype is somewhat variable, we categorized it as “weak” when the fusion of LV3‐4 was mainly visible in the proximal region (with only a limited distal fusion near the wing margin) and as “strong” when the fusion was more extensive and/or when anastomoses were present between the two LVs (Appendix Fig S4). As shown in Fig 5C, fu ¹ /Y; smaug ⁴⁷ /smaug ⁴⁷ double‐mutant males displayed a strong LV3‐4 fusion phenotype more than twice as often as the fu ¹ /Y or fu ¹ /Y; smaug ⁴⁷ /+ control males. This indicates that a smaug loss‐of‐function mutation genetically enhances the effect of the loss of activity of the FU kinase, suggesting that Smaug can positively contributes to HH signaling to pattern the wing.

Discussion

This work uncovers an unexpected interplay between HH/SMO signaling and the RBP Smaug in Drosophila. In brief, using wing‐cultured cells Cl8 cells as a “test tube”, we identified and reconstituted a “HH/SMO/FU pathway” that regulates Smaug. We show that it promotes its phosphorylation, downregulates its levels, and reduces its repressive activity. All these effects rely on the GPCR SMO acting as a scaffold bridging Smaug with the protein kinase FU which promotes Smaug's phosphorylation. Forcing the interaction between Smaug and activated FU recapitulates the effects of HH/SMO on Smaug. Moreover, it reduces the endogenous levels of its target mRNAs and inhibits the formation of the cytoplasmic S‐bodies. Importantly, HH/SMO/FU also regulates the endogenous targets of Smaug both in cultured cells and in in vivo. Finally, smaug and fu genetically interact to modulate wing patterning, revealing a so far unsuspected role of Smaug in Drosophila development as a positive modulator of HH signaling.

Based on these data, we propose that HH regulates via SMO and FU the fate of mRNAs bound to Smaug, leading to positive modulation of HH signaling. These data support the following model (presented in Fig 6). In the absence of HH, Smaug forms S‐bodies and promotes the decay of its target mRNAs. It also binds SMO which is present on intracellular trafficking vesicles. In response to HH, SMO activates FU and this promotes the phosphorylation of Smaug leading to the disappearance of the S‐bodies and release of the bound mRNAs, leading to their accumulation. Similarly, a transient dissolution of the Smaug/SAMD4 bodies was reported in the post‐synaptic region of rat neurons upon stimulation of specific receptors 18, 57, 58, 59.

Figure 6. Model.

In the absence of HH, Smaug is associated, likely as S‐bodies (in red) to inactive SMO present on intracellular vesicles and leads to the repression (blue line) of its target mRNA(s) (called here xmRNA) via the recruitment of proteins (not represented here) that promotes their decay. Upon HH reception, activated and phosphorylated (yellow P circles) SMO accumulates at the plasma membrane where it relocates both the kinase FU and Smaug. This leads to FU activation which in turn phosphorylates (green full circle) SMO and Smaug. This phosphorylation of Smaug would reduce its negative effect (dashed blue line) favoring the release of its target mRNAs—possibly via the dissolution of the S‐foci—and leading to their accumulation. Finally, when the highest levels of SMO phosphorylation are reached (red P circle), phosphorylated Smaug is released, leading to attenuation of the effects of HH/SMO/FU on Smaug.

A novel reporter to monitoring Smaug repressive effects

We developed a novel Smaug repression assay that allows—thanks to the SNAP tag—the simultaneously evaluation of Smaug repressive activity and levels. This permits the evaluation of variations in Smaug levels on the global changes in reporter expression. This reporter also allows the specific monitoring of changes in Smaug's ability to repress mRNAs, independently of its ability to bind them. Smaug has been shown to act on its target mRNAs either through an inhibition of their translation or by promoting their degradation after deadenylation, depending on the proteins that it recruits 11, 27, 60. Here, Smaug has comparable effects on the protein levels and the mRNA levels of the reporter, which indicates that this experimental system mainly monitors the effect of Smaug on the decay of its reporter mRNA.

HH/SMO/FU signaling regulates the RBP Smaug

Modulation of RBP functions can be achieved by controlling their stability, their binding to target mRNAs or to protein co‐regulators, their subcellular localization; and/or ability to form membraneless organelles by liquid–liquid phase separation 3, 4, 5, 6, 7, 61. Several of these mechanisms appear to be involved here.

First, HH/SMO/FU signaling controls Smaug's intrinsic repressive action on its target mRNAs, independently of any effect on its levels. This is likely due to a suppression of Smaug promoting the destabilization of mRNAs as HH/SMO/FU leads to a reduction in the levels of its targets mRNA. This effect of HH/SMO/FU requires the kinase FU, which promotes Smaug phosphorylation, likely directly (but possibly via another kinase that would remain to be determined). It is also associated with loss of Smaug cytoplasmic bodies that are also triggered by FU. Of note, the regions involved in this phosphorylation contain low‐complexity regions (LCR) rich in S/T residues. LCR regions control phase transitions in many RBP 6, 62 due to their disordered nature which favors condensate formation via dynamic multivalent interactions 63. An attractive possibility is that phosphorylation of Smaug in these LCRs could prevent phase transition by introducing negative charges.

Second, HH/SMO/FU also downregulated Smaug levels both in cultured cells and in in vivo. This does not require the kinase activity of FU and is associated with a strong reduction in its transcript levels. This reduction is unlikely to reflect a reduced transcription as it was also seen with a smaug transcript generated under an exogenous promotor from a plasmid. Given these data, it probably reflects an increase in the decay of Smaug mRNA by a mechanism that remains to be identified. Note that a kinase‐independent function of FU has previously been described that probably involves the regulatory C‐terminal part, which could act as a scaffolding protein 64, 65. In the present case, it could recruit factors that would promote the destabilization of the smaug transcript.

HH and regulation of mRNA fate

This work strengthens the corpus of emerging connections between HH signaling and the post‐transcriptional fate of mRNAs. Most of these studies have focused on the effects of HH signaling on translation in mammalian cells (reviewed in 66). Thus, Sonic Hedgehog (Shh) has been shown to have general and specific effects on translation via the activation of key components of translation (for review, see ref. 67) or via the specific regulation of RBPs. For instance, similarly to its effect on Smaug, Shh controls the phosphorylation of the RBP zipcode‐binding protein‐1 (ZBP1) during axon guidance 68.

In this context, the present work reveals an additional layer of regulation of HH signaling during wing development that involves the regulation of a RBP by SMO and FU to control mRNA fate, independently of transcription. Indeed, more than half of the mRNAs targeted by Smaug that we tested are also regulated by HH/SMO/FU. Importantly, all the transcripts upregulated by HH/SMO are also upregulated by activated FU when this kinase is artificially tethered to Smaug (and vice versa) but not in the absence of its binding to Smaug. These data strongly indicate that HH/SMO/FU controls Smaug transcripts via their action on Smaug. Strikingly, most of the mRNA targets of Smaug regulated by HH/SMO/FU are not directly linked to signaling processes, but many are involved in proteasome function and metabolism, especially members of the oxidative respiratory chain of the mitochondria. Of note this may be related to the fact that Shh has been shown to affect energy metabolism and mitochondrial function 69, 70. Given that it has been shown that Smaug controls the decay of hundreds, if not thousands of targets, the non‐canonical, Smaug‐mediated effects that we report here likely affect a wide range of cellular processes.

In conclusion, our data both strengthen the view that Smaug may act as a hub for integration of external information, as has been found for P‐bodies and stress granules (for review, see 71, 72). It also contributes to an emerging paradigm in which signaling and GPCRs can control the fate of mRNAs in the cytoplasm 73.

Materials and Methods

Immunoprecipitation

Coimmunoprecipitation in cell lysates: For input (In), 50 μg of protein extract was mixed with loading buffer and frozen in liquid nitrogen before storage overnight at −80°C. For each IP, 1 mg of protein extract was mixed with 2 μg of antibody against the protein tag: mouse anti‐HA 12CA5 (Sigma‐Aldrich) or rabbit anti‐Myc 51c (Euromedex) in 500 μl reaction volume of lysis buffer (RIPA buffer with “Complete EDTA‐free antiprotease mix” (Roche)) and Phostop (Roche). IPs were incubated in 1.5‐ml tube on a rotating wheel overnight at 4°C. 20 μl Pre‐washed Protein A/G Magnetic beads (ThermoScientific) were added for 2 h at 4°C per IP. The beads (IP) were washed three times with cell lysis buffer and re‐suspended in 50 μl of 2× Laemmli sample buffer (Bio‐Rad) before heating at 95°C for 3 min. For the GFP fusions, coimmunoprecipitation was done using anti‐GFP nanobodies cross‐linked to NHS resin (1 μg/μl from ChromoTek) and beads were pelleted by centrifugation before loading.

Western‐blotting and immunodetection

Cl8 cells were cultured as previously described in 2% CFS (Hyclone). Transient transfections employed the TransIT‐Insect Reagent (Mirus) using a total of 0.5–1 μg of total plasmid DNA for 2–4 μl reactant, respectively. At 48 h post‐transfection, cells were washed twice in 1× PBS. After centrifugation, the pellet was lysed in RIPA buffer with the “Complete EDTA free antiprotease mix” (Roche) and the phosphatase inhibitor mix Phostop (Roche), before centrifugation (12,000 rcf) for 10 minutes at 4°C, and then mixed with Laemmli sample buffer (Bio‐Rad) and 0.1M DTT. Protein concentrations were estimated with the Bradford Ultra reagent (Expedeon). For direct immunodetection, 60 μg of protein was warmed 5 min at 25°C before loading on a 10% Anderson gel (ratio acrylamide/bis‐acrylamide = 77) 74. Gels were run for 90 min at 150 volts in a Miniprotean (Bio‐Rad) apparatus. The subsequent steps were performed as in Ref. 38. Primary antibodies were as follows: 1:1,000 rat monoclonal anti‐HA (Roche), 1:5,000 rabbit anti‐GFP (Torrey Pines Biolabs), 1:2,000 rabbit anti‐GMAP (Sigma, gift from Laurent Ruel), and 1:1,000 mouse anti‐Myc (clone 4A6, Millipore). Secondary antibodies conjugated with HRP were as follows: anti‐rat (JacksonImmuno), anti‐mouse (Sigma), and anti‐rabbit (JacksonImmuno). The enhanced chemiluminescence detection system (ECL Select, Amersham) was used on a LAS‐3000 imager (Fujifilm).

SNAP‐Smaug electrophoresis and phosphorylation assay

48 h after transfection, Cl8 cells were lysed in 1% Triton, 50 mM Tris pH 8, 150 mM NaCl, 1 mM DTT with complete EDTA‐free antiprotease mix (Roche) before labeling for 30 min at 30°C with 1 μM SNAP‐Cell TMR STAR (NEB) alone or together with 1 nM CLIP‐Cell TMR STAR (NEB) in 0.66% Triton, 50 mM Tris pH 8, 150 mM NaCl, and 1 mM DTT. Samples were run on a 10% Anderson gel. All gels of a biological triplicate were run and scanned together on a Typhoon (GE). Images were analyzed and quantified using Imagelab (Bio‐Rad) software. Statistical analyses were done using the one‐tailed Kruskal–Wallis rank test using GraphPad Prism.

Cell fluorescent imaging

5 × 10⁵ Cl8 cells were plated 24 h (in 24‐well plates) before transfection with 250 ng of each plasmid as indicated. In single transfections, pAct.GAL4 was added to ensure a total DNA concentration of 0.5 μg. Cells were analyzed 48hr after transfection. Extracellular SNAP labeling was done by incubation with an extracellular fluorescent substrate SNAP‐Surface 488 (NEB) (1.66 μM in Cl8 medium) for 10 min at room temperature before being briefly rinsed 3 times in PBS, fixed for 15 min in 4% PFA, and then washed with PBS three times. Images were taken with a CSU‐W1 (Yokogawa‐Andor) spinning disk Leica DMI8 microscope with a 63× oil‐immersion objective.

Repression assay

Unless indicated otherwise, the following plasmid concentrations were used for transfection: 300 ng pAct.SNAP‐GUS‐Stop‐5BoxB, 15 ng pAct.λN‐SNAP‐smaug, 50 ng pAct.GFP‐SNAP, 50 ng pAct.smo‐GFP (or pAct.smo ^PKA‐SD‐GFP), and 50 ng pAct.hhN; the total level of DNA was adjusted to 500 ng using pAct.GFP. 48 h after transfection, cells were lysed in 1% Triton, 50 mM Tris pH 8, 150 mM NaCl, 1mM DTT with complete EDTA‐free antiprotease mix (Roche) before labeling for 30 min at 37°C with 1.66 nM SNAP‐Cell Oregon Green (NEB) in 0.5% Triton, 50 mM Tris pH 8, 150 mM NaCl, and 1 mM DTT. At least three independent biological experiments were performed each time. All gels of a triplicate experiment were run and scanned together. Images were analyzed and quantified using Imagelab (Bio‐Rad) software. After quantification, the λN‐SNAP‐Smaug and SNAP‐GUS were normalized to the levels of GFP‐SNAP. Statistical analyses were done using the Kruskal–Wallis rank test followed by the Dunn test using GraphPad Prism.

Relative mRNA quantification

Transfections were done as for the repression assays (see Materials and Methods). Transfected cells were washed in the RNeasy Kit buffer and stored at −80°C until use. Total RNA was isolated from transfected Cl8 cells using RNeasy Kit (NucleoSpin RNA, Macherey‐Nagel), and cDNA was synthesized using Superscript III Reverse Transcriptase Kit (Invitrogen). Extraction from wing imaginal disks was performed using xx imaginal disks from third‐instar larvae. Quantitative RT–PCR was run in 10 μl reactions in a real‐time PCR system (Analytik Jena) using SYBR‐Green qPCR Mix (EurobioGreen Lo‐ROX, Eurobio). qPCRsoft (3.0) software was used for analyzing cycle threshold (Ct) values. Fold change in RNA levels (expressed as 2^−ΔC) was normalized to the expression of the Tubulin gene. RT–qPCR was performed in triplicate on each of three independent biological replicates. The sequence of the primers used is given in Appendix Table S3.

Drosophila strains and genetics

Strains were as follows: w1118, MS1096 (chr. X 75, UAS hh 76, hh ^ts 77, fu ¹ 78, and smaug⁴⁷ 13. Flies were raised at 25°C unless otherwise indicated. The w, fu ¹ /Y;;smaug ⁴⁷ /smaug ⁴⁷ males were obtained by crossing w, fu ¹/Y, or FM6;; smaug ⁴⁷/TM6 Tb, Sb flies together. The w, fu ¹ /Y;;smaug ⁴⁷ /+ males were the progeny from a cross between fu ¹/FM6;; smaug ⁴⁷/TM6 Tb, Sb females with w ¹¹¹⁸ males. The w, fu ¹/+;; TM6 Tb, Sb/+ females progeny from this latter cross were mated to w ¹¹¹⁸ males to obtain the w, fu ¹/Y;; +/+ males. Fly wings were put in 70% ethanol before mounting in Hoyer's medium.

See supplemental materials and methods in Appendix.

Author contributions

IB, AP, and MS conceived and supervised the experiments. IB, AP, and RAH wrote the manuscript, AP secured funding. GA, MG‐A, CA, LB, SM, MS, and FQ performed experiments. LB and MS performed the statistical analysis. IB and MS prepared the figures. GLB, RAH, HDL, and CCS provided reagents, expertise, and feedback.

Conflict of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Supporting information

Appendix

Expanded View Figures PDF

Review Process File

EMBO Reports (2020) 21: e48425