1. Structural Biology and Molecular Biophysics

Cryo-EM reconstructions of inhibitor-bound SMG1 kinase reveal an autoinhibitory state dependent on SMG8

  1. Lukas M Langer
  2. Fabien Bonneau
  3. Yair Gat
  4. Elena Conti  Is a corresponding author
  1. Max Planck Institute of Biochemistry, Germany
https://doi.org/10.7554/eLife.72353
Share this article
Cite this article
  1. Lukas M Langer
  2. Fabien Bonneau
  3. Yair Gat
  4. Elena Conti
(2021)
Cryo-EM reconstructions of inhibitor-bound SMG1 kinase reveal an autoinhibitory state dependent on SMG8
eLife 10:e72353.
https://doi.org/10.7554/eLife.72353
  2. Download BibTeX
  3. Download .RIS

Abstract

The PI3K-related kinase (PIKK) SMG1 monitors the progression of metazoan nonsense-mediated mRNA decay (NMD) by phosphorylating the RNA helicase UPF1. Previous work has shown that the activity of SMG1 is impaired by small molecule inhibitors, is reduced by the SMG1 interactors SMG8 and SMG9, and is downregulated by the so-called SMG1 insertion domain. However, the molecular basis for this complex regulatory network has remained elusive. Here, we present cryo-electron microscopy reconstructions of human SMG1-9 and SMG1-8-9 complexes bound to either a SMG1 inhibitor or a non-hydrolyzable ATP analog at overall resolutions ranging from 2.8 to 3.6 Å. These structures reveal the basis with which a small molecule inhibitor preferentially targets SMG1 over other PIKKs. By comparison with our previously reported substrate-bound structure (Langer et al.,2020), we show that the SMG1 insertion domain can exert an autoinhibitory function by directly blocking the substrate-binding path as well as overall access to the SMG1 kinase active site. Together with biochemical analysis, our data indicate that SMG1 autoinhibition is stabilized by the presence of SMG8. Our results explain the specific inhibition of SMG1 by an ATP-competitive small molecule, provide insights into regulation of its kinase activity within the NMD pathway, and expand the understanding of PIKK regulatory mechanisms in general.

Editor's evaluation

This study uses CryoEM and biochemical studies to uncover a new and potentially important conformational off-state of a key regulatory multi-subunit protein kinase, SMG1. The study was enabled by applying a small molecule ATP-site inhibitor to capture the structure. The work will be of wide interest to the signaling and structural biology communities.

https://doi.org/10.7554/eLife.72353.sa0

Introduction

Nonsense-mediated mRNA decay (NMD) is a co-translational mRNA quality control pathway central to the detection and removal of mRNAs containing premature termination codons as well as to the regulation of many physiological transcripts (Kurosaki et al., 2019; Karousis and Mühlemann, 2019). In metazoans, translation termination at a premature stop codon triggers phosphorylation of the RNA helicase UPF1, which then enables the recruitment of downstream effectors, in turn resulting in the degradation of the targeted mRNA by ribonucleases (Ohnishi et al., 2003; Yamashita, 2013; Chakrabarti et al., 2014; Nicholson et al., 2014; Okada-Katsuhata et al., 2012; Kashima et al., 2006). UPF1 phosphorylation is thus a crucial point in metazoan NMD; it occurs specifically at particular Ser-containing motifs and is mediated by the SMG1 kinase (Denning et al., 2001; Yamashita et al., 2001).

SMG1 is a member of the phosphatidylinositol-3-kinase-related kinase (PIKK) family, which also includes the ATM, ATR, DNA-PKc, and mTOR kinases (Keith and Schreiber, 1995). Despite the confounding name reflecting a possible evolutionary origin with a lipid kinase (Keith and Schreiber, 1995), and in retrospect the ability of some PIKK family members to bind inositol-6-phosphate (Gat et al., 2019), all active PIKKs are Ser-Thr protein kinases. These large, multidomain enzymes share an overall similar architecture: a C-terminal catalytic module (comprising the so-called FAT, kinase, and FATC domains) and an N-terminal α-solenoid (that typically serves as a protein-protein interaction module). Indeed, PIKKs bind to and are regulated by interacting proteins, with which they form large and dynamic assemblies. In addition, some PIKKs contain an autoregulatory element, particularly in the variable region connecting the kinase and FATC domains, that is known as the PIKK-regulatory domain (PRD) (Baretić and Williams, 2014; Imseng et al., 2018; Jansma and Hopfner, 2021). Finally, the kinase domains of PIKKs have been the target of numerous efforts in the development of specific inhibitors. Not surprisingly, given the central roles of these kinases in pathways surveilling cellular homeostasis, small molecules targeting specific PIKKs have been approved as therapeutics or are being evaluated in clinical trials (Janku et al., 2018; Zhang et al., 2011; Durant et al., 2018). Nevertheless, the high conservation of the PIKK kinase domain has complicated efforts to develop inhibitors specifically targeting only a selected member of this family of enzymes. Structural data rationalizing the binding specificity of such compounds to PIKKs are still sparse in general and entirely missing for the SMG1 kinase.

In the case of SMG1, the two interacting factors SMG8 and SMG9 have been linked to dysregulated NMD and neurodevelopmental disorders in humans (Alzahrani et al., 2020; Shaheen et al., 2016; Yamashita et al., 2009). Previous cryo-electron microscopy (cryo-EM) studies have revealed the molecular interactions that underpin the structure of the human SMG1-SMG8-SMG9 complex (Gat et al., 2019; Zhu et al., 2019) and the determinants with which it recognizes and phosphorylates UPF1 peptides with specific Leu-Ser-Gln (LSQ) motifs (Langer et al., 2020). SMG8 and SMG9 form an unusual G-domain heterodimer (Gat et al., 2019; Langer et al., 2020; Li et al., 2017). The G-domain of SMG9 binds both the α-solenoid and the catalytic module of SMG1 while the G-domain of SMG8 engages only the α-solenoid, thus rationalizing biochemical studies pointing to the crucial role of SMG9 in enabling the incorporation of SMG8 into the complex (Yamashita et al., 2009; Arias-Palomo et al., 2011). In turn, SMG8 appears to have a direct regulatory function on SMG1, as its removal or C-terminal truncation results in hyper-activation of SMG1 kinase activity (Yamashita et al., 2009; Deniaud et al., 2015; Arias-Palomo et al., 2011; Zhu et al., 2019). However, the C-terminal domain of SMG8 remains poorly defined in all available cryo-EM reconstructions (Zhu et al., 2019; Gat et al., 2019; Langer et al., 2020), hindering a molecular understanding of its regulatory role.

Another portion of the complex reported to downregulate SMG1 kinase activity is the so-called insertion domain, a large 1200-residue region connecting the SMG1 kinase and FATC domains. Removal of the SMG1 insertion domain causes hyper-activation of the kinase (Deniaud et al., 2015; Zhu et al., 2019), similarly to the effect reported for the PRDs of other PIKKs (McMahon et al., 2002; Edinger and Thompson, 2004; Xiao et al., 2019; Mordes et al., 2008). However, the SMG1 insertion domain shows no sequence similarity to the PRDs of other PIKKs and is remarkably larger in comparison. None of the current cryo-EM reconstructions of SMG1 or its complexes show density corresponding to the SMG1 insertion domain (Zhu et al., 2019; Gat et al., 2019; Langer et al., 2020), and it is thus unclear how it regulates kinase activity.

Here, we used cryo-EM and biochemical approaches to study the basis with which SMG1 is specifically inhibited by a small molecule compound (SMG1i), and in doing so we also identified the molecular mechanisms with which SMG1 is downregulated by its own insertion domain in cis and SMG8 in trans.

Results

The compound SMG1i specifically inhibits SMG1 kinase activity in vitro

We set out to study the inhibitory mechanism of SMG1i (Figure 1A), a small-molecule compound based on a pyrimidine derivative that has been reported to inhibit SMG1 catalytic activity (Gopalsamy et al., 2012; Mino et al., 2019). We expressed and purified human wild-type SMG1-8-9 complex from piggyBac transposase generated HEK293T stable cell pools, essentially as described before (Gat et al., 2019). We have previously shown that the use of a UPF1 peptide comprising the UPF1 phosphorylation site 1078 (UPF1-LSQ) allows monitoring of its specific phosphorylation by SMG1 over time using mass spectrometry (Langer et al., 2020). We made use of this method to assay the effect and potency of SMG1i (Figure 1, Figure 1—figure supplement 1). To avoid artifacts caused by ATP depletion, we performed an ATP titration and analyzed the end-point measurements (Figure 1—figure supplement 1B). As a control, we used a UPF1-LDQ peptide in which we exchanged the phosphor-acceptor Ser1078 of UPF1-LSQ to aspartic acid. Next, we repeated end-point measurements under conditions of stable ATP concentration and increasing amounts of SMG1i (Figure 1B). In this assay, we observed a significant reduction of UPF1-LSQ phosphorylation when adding SMG1i at concentrations equimolar to the enzyme (Figure 1B and Figure 1—figure supplement 1C). To corroborate these results, we repeated titration of SMG1i under conditions of stable ATP concentration using a radioactivity-based phosphorylation assay and full-length UPF1 as a substrate. Again, we observed a reduction in UPF1 phosphorylation in the presence of low micromolar amounts of SMG1i (Figure 1C).

Figure 1 with 2 supplements see all
Download asset Open asset
SMG1i specifically inhibits SMG1 kinase activity in vitro.

(A) Structure of the SMG1 inhibitor (SMG1i). (B) Titration of SMG1i using a mass spectrometry-based phosphorylation assay with 500 nM SMG1-8-9 and the indicated UPF1-derived peptides as substrates. (C) Titration of SMG1i using a radioactivity-based phosphorylation assay with 100 nM SMG1-8-9 and full-length UPF1 as a substrate. The Coomassie-stained gel is shown on top, and the corresponding radioactive signal is below. (D, E) Titration of SMG1i using a radioactivity-based phosphorylation assay with either SMG1-8-9 and full-length UPF1 (D) or mTORΔN-LST8 and GST-AKT1 (E) as a substrate. Stain-free gels are shown on top and the corresponding radioactive signal at the bottom (Ladner et al., 2004). (F) Quantification of normalized UPF1 or mTOR (auto-)phosphorylation in the presence of increasing amounts of SMG1i for both SMG1 and mTOR. For each data point, the mean is shown with standard deviations of the three replicates indicated.

Figure 1—source data 1

Unedited images for gels shown in Figure 1.

https://cdn.elifesciences.org/articles/72353/elife-72353-fig1-data1-v2.zip

To assess the specificity of SMG1i, we tested its effect on the mTOR kinase using a similar radioactive kinase assay. In this experiment, we used a truncated mTORΔN-LST8 complex previously shown to be constitutively active and responsive to mTOR inhibitors (Yang et al., 2013). As an mTOR kinase substrate, we used a GST-AKT1450–480 fusion protein. We observed that SMG1i only affected mTOR activity in the highest concentration range, well beyond concentrations that had an observable effect on SMG1 activity (Figure 1—figure supplement 2A). We selected four different concentrations of SMG1i based on these assays (Figure 1C, Figure 1—figure supplement 2A) and repeated the radioactivity-based kinase assay for both SMG1 and mTOR in triplicates (Figure 1D and E, Figure 1—figure supplement 2B, C). Upon quantifying the levels of phosphorylation using densitometry and normalizing them to the amount of protein loaded in each lane, we found that SMG1i robustly inhibited SMG1 while mTOR activity was only weakly affected (Figure 1F, Figure 1—figure supplement 2D, E,F). In agreement with previous results (Gopalsamy et al., 2012), we concluded that SMG1i displays high potency and specificity in inhibiting the SMG1 kinase in vitro.

Cryo-EM structure of SMG1-8-9 bound to the SMG1i inhibitor

To understand the molecular basis for the SMG1i mode of action, we reconstituted purified wild-type SMG1-8-9 with an excess of SMG1i and subjected the sample to single-particle cryo-EM analysis. The resulting reconstruction reached an overall resolution of 3.1 Å, and we observed clear density for the SMG1 inhibitor bound to the kinase active site (Figure 2A, Figure 2—figure supplements 1 and 2 and Figure 2—figure supplement 3B, C). Superposition with our previously published AMPPNP- and substrate-bound model (Langer et al., 2020) revealed that SMG1i exploits essentially the same binding site as the ATP analog. However, SMG1i wedges deeper in between the N- and C-lobe of the kinase domain and is engaged in more extensive interactions (Figure 2B and Figure 2—figure supplement 3A and D). These observations identify SMG1i as a potent ATP-competitive inhibitor.

Figure 2 with 3 supplements see all
Download asset Open asset
Structural basis for selective targeting of SMG1 by the SMG1 inhibitor.

(A) Model of the SMG1-8-9 kinase complex bound to SMG1i. SMG1 is in gray, SMG8 is in blue, and SMG9 is shown in green. SMG1i is shown as a magenta model overlaid with the isolated transparent density. Approximate location of the SMG1 active site is indicated by a black circle. (B) Key interactions of SMG1i with SMG1 active site residues. Important residues located in either the N- or the C-lobe of the SMG1 kinase domain are colored gray. Other parts of SMG1 are transparent and interactions of SMG1i with SMG1 backbone are not shown. (C) Superposition of SMG1i-bound SMG1 with the mTOR active site (PDB identifier: 4JSP) over the catalytic loops of both kinases. Key SMG1 residues indicated in (B) are shown alongside the respective mTOR residues colored in green. Regions possibly accounting for preferential interaction of SMG1i with SMG1 over mTOR are circled and the relevant residues are labeled.

While many of the SMG1 residues interacting with SMG1i are conserved within the PIKK family, two contact sites appear to be specific: SMG1 Pro2213 and Asp2339 contact the sulfonamide moiety of the inhibitor while Asn2356 forms hydrogen bonds with the phenyl urea/carbamide group (Figure 2B). Superposition of the SMG1 active site in the SMG1i-bound structure with the active site of mTORΔN shows that the corresponding residues have diverged and would not be able to mediate analogous interactions with the SMG1i sulfonamide group (mTOR Thr2245 and Ser2342 at the positions of SMG1 Pro2213 and Asp2339) or with the urea group (mTOR Gly2359 at the position of SMG1 Asn2356) (Figure 2C). The same regions of the small molecule are diverging or missing in the mTOR inhibitor Torin 2 (Figure 2—figure supplement 3E). Taken together, these observations rationalize the specificity of SMG1i for SMG1 over mTORΔN that we observed in the in vitro assays (Figure 1D-F and Figure 1—figure supplement 2D, E and F).

Superposition with other PIKKs suggests the presence of potentially similar discriminatory interactions: like mTOR, DNA-PK lacks a favorable interaction site for the SMG1i urea group (Gly3943 at the position of SMG1 Asn2356) (Figure 2—figure supplement 3F). In addition, DNA-PK PRD residue Lys4019 would sterically clash with the SMG1i urea group in the inactive conformation of the DNA-PK active site. In the case of ATR, the SMG1i sulfonamide site has diverged (Gly2385 at the position of SMG1 Pro2213) and residues in the N-lobe would result in a steric clash with the urea group of the inhibitor (Lys2329 and Asp2330 corresponding to SMG1 Leu2157 and Glu2158) (Figure 2—figure supplement 3G). Finally, mTOR, DNA-PK, and ATR active sites have a different relative orientation of the kinase N- and C-lobes as compared to SMG1, changing the overall chemical environment at the SMG1i-binding site (Figure 2C and Figure 2—figure supplement 3F and G). We conclude that subtle differences in the SMG1 structure underpin the specificity of the SMG1i inhibitor.

The SMG1 insertion domain contains a PRD that blocks substrate binding in the presence of SMG8

In the reconstruction of the SMG1i-bound SMG1-8-9 complex, we observed additional density near the end of the kinase domain, protruding from the position where the insertion domain is expected to start (Figure 3, Figure 3—figure supplement 1A and B). While the lower resolution for this part of the map was not sufficient to unambiguously build an atomic model, this density clearly docks at the substrate-binding site of the kinase domain (Figure 3A, Figure 3—figure supplement 1A, B and C). Superposition with the previous cryo-EM structure of SMG1-8-9 complex bound to a UPF1 peptide (Langer et al., 2020) revealed that the additional density in the current reconstruction is mutually exclusive with the position of a bound substrate (Figure 3I, Figure 3—figure supplement 1A). This observation is in good agreement with the reported positions of the PRDs in structures of the autoinhibited forms of yeast Tel1ATM and human DNA-PK (Jansma et al., 2020; Yates et al., 2020; Chen et al., 2021). We concluded that the SMG1 insertion domain contains a PRD that can directly occupy and block the substrate-binding path in the kinase active site, analogous to other PIKKs. The SMG1 PRD would have to re-arrange upon substrate binding, as observed in the case of the inactive-to-active transition of DNA-PK (Chen et al., 2021).

Figure 3 with 4 supplements see all
Download asset Open asset
Structures of SMG1-8-9 and SMG1-9 complexes reveal that the SMG1 insertion domain can block the substrate-binding path in the presence of SMG8.

(A) Cryo-EM density of SMG1-8-9 bound to SMG1i. Density for the inhibitor is in magenta, the N-terminus of the SMG1 insertion is in red, and all other parts as indicated. (B) Cryo-EM density of SMG1-9 complex bound to the SMG1 inhibitor. Everything else as in (A). (C) Cryo-EM density of SMG1-8-9 bound to AMPPNP. Density for AMPPNP is in orange and extra density attributed to the SMG1 insertion is in red. (D) As in (C), but for the SMG1-9 complex bound to AMPPNP. Insets I–IV show close-ups of the indicated kinase active site densities superimposed with the model for the UPF1-LSQ substrate shown as blue sticks (PDB identifier: 6Z3R). All maps segmented. cryo-EM, cryo-electron microscopy.

Within the cryo-EM data set collected on an SMG1-8-9 complex incubated with SMG1i, we isolated a subset of particles missing any density for SMG8, hence representing an SMG1i-bound SMG1-9 complex (Figure 3B, Figure 2—figure supplement 1C, Figure 3—figure supplement 4). The corresponding 3.6 Å reconstruction not only showed a similar mode of interaction of SMG1i, but also several features distinct from SMG1 in isolation and the SMG1-8-9 ternary complex. Globally, the α-solenoid of SMG1 rearranges, consistent with the notion of a stepwise compaction of SMG1 upon binding first SMG9 and then SMG8 (Figure 3—figure supplement 4A and B; Melero et al., 2014; Zhu et al., 2019). Locally, the mode of interaction between SMG9 and the SMG1 α-solenoid differs between the two complexes. In the binary SMG1-9 complex, the portion of SMG9 that interacts with the N-terminal HEAT repeat helices of SMG1 changes conformation, exploiting part of a binding site that was occupied by a short N-terminal segment of SMG8 in the ternary complex (Figure 3—figure supplement 4C, D and E). Indeed, two hydrophobic residues of SMG9 (Leu456 and Leu457) take the place normally reserved for two hydrophobic residues of SMG8 (Leu351 and Leu352) (Figure 3—figure supplement 4F and G). Hence, the conformation of the SMG9 segment observed in the SMG1-9 complex is incompatible with the binding of SMG8 to this part of SMG1. Importantly, the reconstruction of SMG1-9 showed no ordered density at the active site for the PRD (Figure 3B and II, Figure 3—figure supplement 1A), which is in contrast to the ordered PRD density visualized in the SMG1i-bound SMG1-8-9 complex.

Next, we asked whether the autoinhibitory state of the SMG1 insertion domain in the SMG1-8-9 complex is due specifically to the presence of the SMG inhibitor or whether it is a more general feature. To address this question, we reconstituted purified wild-type SMG1-8-9 with an excess of AMPPNP (instead of SMG1i) and subjected the sample to a similar single-particle cryo-EM analysis routine. We obtained maps for SMG1-8-9 and SMG1-9 at overall resolutions of 2.8 Å and 3.1 Å, respectively (Figure 3C and D, Figure 3—figure supplements 2 and 3). Both reconstructions showed clear density for AMPPNP in the SMG1 active site. In the reconstruction of the ternary SMG1-8-9 complex, we observed an additional density at the substrate-binding site, in the same position as that observed when in the presence of SMG1i (Figure 3III, IV). However, in the context of AMPPNP this additional density is less prominent and does not connect to the SMG1 kinase domain. It thus appears that the SMG1 inhibitor has a stabilizing effect on the autoinhibitory conformation of the SMG1 PRD. In the reconstruction of the binary SMG1-9 complex in the presence of AMPPNP, we observed no ordered density at the substrate-binding site, consistent with the reconstruction obtained in the presence of SMG1i. We concluded that the autoinhibitory conformation of the PRD in the insertion domain of SMG1 is connected to the presence of a nucleotide in the ATP-binding site and is stabilized by the presence of SMG8.

The SMG1 insertion domain can block overall access to the kinase active site and interacts with the SMG8 C-terminus

The findings above suggested a possible role for SMG8 in stabilizing an autoinhibited state of SMG1. Indeed, previous work implicated SMG8 and, in particular, its C-terminal domain in the downregulation of SMG1 activity (Zhu et al., 2019; Arias-Palomo et al., 2011; Deniaud et al., 2015; Yamashita et al., 2009). SMG8 has a modular domain organization: the N-terminal G-domain is followed by a helical stalk that protrudes into solvent with well-defined density, but the remaining C-terminal region (amounting to about 45% of the molecule) is flexible and poorly resolved in the published cryo-EM studies (Gat et al., 2019; Zhu et al., 2019; Langer et al., 2020). By further processing of the SMG1i-bound SMG1-8-9 data, we could achieve improved density for the C-terminal domain of SMG8, showing a knob-like feature directly connected to the end of the stalk and in turn connecting to a larger globular mass (Figure 4A). Concurrently, we observed an additional density extending from the location of the PRD that we attributed to the SMG1 insertion domain (see below). On one side, this density wraps along the catalytic module of SMG1, occludes the access to the kinase active site (Figure 4A), and reaches toward the IP6-binding site (Figure 4A and C). On the other side, the density projects into solvent and approaches the globular SMG8 C-terminal region. Although the low resolution of this part of the map did not allow model building, the close proximity suggested a physical contact between the SMG1 insertion domain and the SMG8 C-terminal domain. We proceeded to test the interplay between these two parts of the SMG1-8-9 complex.

Figure 4 with 4 supplements see all
Download asset Open asset
The SMG1 insertion domain can block overall access to the kinase active site.

(A) Cryo-EM map after 3D variability analysis filtered by resolution and segmented. Two different views displaying extra density for SMG8 C-terminus (dark blue) and SMG1 insertion domain. The location of the SMG1 kinase active site is indicated by an orange circle. (B) Inter and intra cross-links of the SMG1-8-9 kinase complex. Proteins are colored as in (A). Intra-links shown in Figure 4—figure supplement 1B, C are in black, the inter-link shown in Figure 4—figure supplement 1D is in blue, and inter-links between SMG1 insertion and SMG1 C-terminus are in red. (C) Zoom-in highlighting positions of Lys residues (magenta spheres) crosslinking to the C-terminus of the SMG1 insertion domain. A table listing the cross-linked residues is shown below. Cross-links found in only one of two samples are in italics. (D) Coomassie-stained SDS-PAGE analysis of pull-down experiment showing an interaction between SMG1 insertion domain (SMG12427–3606) and SMG8 C-terminus (SMG8728–991). cryo-EM, cryo-electron microscopy.

Figure 4—source data 1

Table with cross-linking data used in Figure 4.

https://cdn.elifesciences.org/articles/72353/elife-72353-fig4-data1-v2.xls
Figure 4—source data 2

Unedited images for gels shown in Figure 4.

https://cdn.elifesciences.org/articles/72353/elife-72353-fig4-data2-v2.zip

We first used cross-linking mass spectrometry (XL-MS) analysis (Figure 4B, Figure 4—figure supplements 1 and 2). Upon treatment of purified SMG1-8-9 with Bis(sulfosuccinimidyl)suberate (BS3), we detected intra and inter cross-links consistent with parts of the complex for which an atomic model was available and measuring well within the expected range of distance for BS3 cross-links (Figure 4B, Figure 4—figure supplements 1 and 2). We also observed reproducible intra cross-links between the end of the SMG1 insertion and the catalytic module, in particular the FAT (residues 823–1938) and kinase (residues 2080–2422) domains (Figure 4B and C). These cross-links mapped in close proximity to the aforementioned density extending from the PRD. We thus attribute this density to the C-terminus of the SMG1 insertion (Figure 4A, B and C). By combining our low-resolution EM density and our cross-linking data with an AlphaFold prediction of SMG8 (Jumper et al., 2021), we were able to produce a more complete model of SMG1-8-9, now containing large parts of the SMG8 C-terminus (Figure 4—figure supplement 3). Importantly, we detected inter cross-links between the SMG1 insertion domain and the SMG8 C-terminal domain (Figure 4B), consistent with the proximity of the EM densities (Figure 4A) and previously reported XL-MS data (Deniaud et al., 2015).

Next, we tested whether these parts of the complex can interact directly in biochemical assays with recombinant proteins. We purified SMG1 insertion domain (SMG1 residues 2427–3606) with an N-terminal TwinStrep-tag from transiently transfected HEK293T cells and SMG8 C-terminus (SMG8 residues 728–991) from Escherichia coli. Since the structural analysis is indicative of a dynamic or weak contact, we performed pull-down assays in low-salt conditions. We observed that the SMG1 insertion domain indeed co-precipitated the SMG8 C-terminus in these conditions (Figure 4D). As a control, an unrelated TwinStrep-tagged protein (PP7 coat protein) did not pull down SMG8 C-terminus under the same conditions (Figure 4D). As anticipated, increasing the salt concentration had a detrimental effect on the SMG8 C-terminus—SMG1 insertion co-precipitation (Figure 4—figure supplement 4A).

Taken together, these experiments indicate the existence not just of proximity but of a direct physical link between the SMG1 insertion domain and SMG8 C-terminus, rationalizing the observed dependency of SMG1 autoinhibition on the presence of SMG8 (Figure 3).

Conclusions

In this manuscript, we reported a reconstruction of the SMG1-8-9 complex bound to an SMG1 inhibitor. We showed that this compound binds to the ATP-binding site within the kinase active site. By comparison with structural data of related kinases, we identified two functional groups within the small molecule that possibly confer specificity to SMG1. Using the SMG1-related mTOR kinase as an example, we confirmed the specific action of the inhibitor biochemically.

In addition to the SMG1 inhibitor, we observed density for the N-terminal part of the SMG1 insertion domain in the SMG1-8-9 reconstruction. This part of the insertion domain acts as a PRD and occupied the substrate-binding path within the SMG1 active site. Consistent with data for other PIKKs, a PRD within the SMG1 insertion domain can therefore restrict access to the kinase active site.

The structure of an SMG1i-bound SMG1-9 complex reconstructed from the same data set as SMG1i-bound SMG1-8-9 did not show ordered density for the SMG1 PRD in the SMG1 active site. This suggested that an interaction between the SMG1 insertion and SMG8 was important for stabilizing the autoinhibited state of the complex. Consistently, our biochemical analysis indicated a direct physical link between parts of the SMG1 insertion domain and the C-terminal domain of SMG8. While autoinhibition was not observed in previously published apo-structures, our reconstructions of SMG1-9 and SMG1-8-9 bound to AMPPNP showed that blockage of the substrate-binding path was also possible when the SMG1-8-9 complex active site was bound to a nucleotide. Accordingly, the described autoinhibition was not a phenomenon limited to the inhibitor-bound structures. We conclude that at least two layers of regulation of SMG1 kinase activity exist: first, access to the SMG1 active site can be restricted in cis by a PRD within the SMG1 insertion domain. Second, this autoinhibition is stabilized in trans by the SMG8 C-terminus.

How is the autoinhibited complex activated? The position of the SMG1 insertion domain in the inhibitor-bound SMG1-8-9 complex reconstruction raised the possibility that the autoinhibited complex might not be able to interact with its substrate, UPF1 (Figures 3I, III, 4A). To test this hypothesis, we incubated TwinStrep-tagged SMG1-8-9 with increasing concentrations of SMG1i before adding UPF1 and performed a pull-down experiment (Figure 4—figure supplement 4B). The efficiency of UPF1 co-precipitation was unaffected by the presence of SMG1i, suggesting that either UPF1 can overcome blockage of the SMG1 active site by the SMG1 insertion domain or has secondary binding sites independent of the observed SMG1 autoinhibition. Indeed, superposition of the autoinhibited complex reconstruction with a previously reported negative-stain map of a cross-linked UPF1-bound SMG1-8-9 complex raised the possibility that UPF1-binding and SMG1 autoinhibition might be able to occur in parallel (Figure 4—figure supplement 4C). An alternative explanation would be that optimal positioning of UPF1 with respect to the SMG1 kinase will allow the unstructured N- and C-terminal ends of UPF1 to efficiently compete with the SMG1 insertion domain for binding to the SMG1 active site in vivo. This would be mimicked by the high protein concentrations used in in vitro assays (such as Figure 4—figure supplement 4B).

Whether SMG8 or an SMG8-9 complex dissociates from SMG1 at any point in the kinases’ catalytic cycle during NMD in vivo is an outstanding question. Early biochemical experiments have suggested that SMG8 plays a crucial role in recruiting SMG1 complex to the NMD-competent messenger ribonucleoprotein particle (Yamashita et al., 2009). Given that the SMG1-8-9 complex actively phosphorylated UPF1 in in vitro experiments using purified proteins and our observation that SMG1i binding did not impair the interaction of SMG1 with UPF1, it is tempting to speculate that UPF1 itself might be sufficient to overcome the effect of SMG8 on stabilizing SMG1 autoinhibition within an SMG1-8-9 complex. In this model, nucleotide binding and kinase activation could occur in two distinct steps during NMD: upon ATP binding, the kinase complex would adopt an autoinhibited conformation and only be licensed for specific phosphorylation of its target by the substrate itself (Figure 5). In addition, autophosphorylation of the kinase complex may play a crucial role in kinase activation.

Figure 5
Download asset Open asset
Hypothetical model of SMG1 kinase regulation.

Structural and biochemical data suggest different layers of regulation on SMG1 kinase activity. Upon ATP binding (orange), the SMG1 kinase adopts an autoinhibited conformation (step 1)—mediated by the concerted action of the SMG1 insertion domain in cis and the SMG8 C-terminus in trans (dotted arrow). The presence of the correct substrate and likely other factors/cues (indicated by X) then trigger the release of the autoinhibition and activity of the kinase toward UPF1 (step 2).

Taken together, our structures will help to improve the design of selective PIKK active site inhibitors, for example, specific for SMG1. Our data suggest a unifying structural model for concerted action of the SMG1 insertion domain in cis and SMG8 in trans to tune SMG1 kinase activity, shedding light on the intricate regulation of this kinase complex in metazoan NMD.

Materials and methods

Key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Gene (Homo sapiens)SMG1Shigeo Ohno labUniprot Q96Q15
Gene (H. sapiens)SMG8Shigeo Ohno labUniprot Q8ND04
Gene (H. sapiens)SMG9Shigeo Ohno labUniprot Q9H0W8
Cell line (H. sapiens)HEK293TATCC
Strain, strain background (Escherichia coli)BL21 Star (DE3) pRAREEMBL Heidelberg Core FacilityElectrocompetent cells
Peptide, recombinant proteinUPF1-LSQ (peptide 1078) and derivativeIn-house as described in doi: https://elifesciences.org/articles/57127
Chemical compound, drugSMG1 inhibitorRobert Bridges, Rosalind Franklin University of Medicine and Science, and the Cystic Fibrosis Foundation
Chemical compound, drugAMPPNPSigma-Aldrich
Chemical compound, drugATPSigma-Aldrich
Software, algorithmSerialEMhttps://bio3d.colorado.edu/SerialEM/
Software, algorithmFocushttps://focus.c-cina.unibas.ch/ wiki/doku.phpv1.1.0
Software, algorithmRELIONdoi: 10.7554/eLife.42166RELION 3.0
Software, algorithmCryosparcdoi: 10.1038/nmeth.4169Cryosparc2
Software, algorithmUCSF ChimeraUCSF, https://www.cgl.ucsf.edu/chimera/
Software, algorithmUCSF ChimeraXUCSF, https://www.rbvi.ucsf.edu/chimerax/
Software, algorithmCOOThttp://www2.mrc-lmb.cam.ac.uk/personal/pemsley/coot/
Software, algorithmPHENIXhttps://www.phenix-online.org/PHENIX 1.17
Software, algorithmMolprobityDuke Biochemistry, http://molprobity.biochem.duke.edu/
Software, algorithmPyMOLPyMOL Molecular Graphics System, Schrodinger LLCPyMOL 2.3.2

Protein expression and purification

Request a detailed protocol

TwinStrep-tagged SMG1-8-9 complex was expressed and purified as described before (Gat et al., 2019; Langer et al., 2020). The source cells were authenticated by genotyping (Eurofins) and tested negative for mycoplasma contamination (LookOut Mycoplasma PCR Detection Kit).

N-terminally TwinStrep-tagged SMG1 insertion domain (SMG12427–3606) was expressed and purified essentially identical to the SMG1-8-9 full-length complex, with the exception that lysis and affinity purification was carried out in a buffer based on 2× phosphate-buffered saline (PBS). The protein was frozen in gel filtration buffer at concentrations ranging from 15 to 40 µM depending on the purification.

Untagged full-length UPF1 expressed in E. coli was purified as reported previously (Langer et al., 2020).

N-terminally TwinStrep-tagged full-length UPF1 was expressed using a stable pool of HEK293T cells as described for SMG1-8-9. Approximately 400×106 cells were lysed with a Dounce homogenizer in buffer containing 1× PBS, 5 mM MgCl2, 1 µM ZnCl2, 10% (v/v) glycerol, 1 mM DTT supplemented for lysis with Benzonase and DNase I as well as EDTA-free Complete Protease Inhibitor Cocktail (Roche). The lysate was cleared by centrifugation at 25,000 rpm for 30 min at 10°C. The supernatant was filtered and applied to a StrepTrap HP column (Sigma-Aldrich). After passing 25 column volumes of lysis buffer without supplements, the column was further washed with 25 column volumes of buffer Hep A (20 mM HEPES/NaOH pH 7.4, 85 mM KCl, 2 mM MgCl2, 1 µM ZnCl2, 10% (v/v) glycerol, and 1 mM DTT) and bound protein was eluted directly onto a HiTrap Heparin column (GE Healthcare) using buffer Hep A supplemented with 2.5 mM Desthiobiotin. After washing the Heparin column with 10 column volumes of buffer Hep A, bound protein was eluted using a gradient increasing the salt concentration to 500 mM KCl over 100 ml while collecting fractions (AEKTA prime FPLC system, GE Healthcare). Fractions were analyzed by SDS-PAGE and those containing pure full-length TwinStrep-UPF1 were collected, concentrated to 30–80 µM, and flash-frozen without buffer exchange using liquid nitrogen until further usage.

The SMG8 C-terminus (SMG8728–991) was N-terminally fused to a 6xHis-SUMO tag and expressed in E. coli BL21 STAR (DE3) pRARE cells. Bacteria were grown in TB medium at 37°C shaking at 180 rpm. At an OD of 2, cultures were cooled down to 18°C and overnight protein expression was induced by adding IPTG. After harvesting (6000 rpm, 10 min), bacteria were lysed by sonication in a buffer containing 20 mM Tris-HCl pH 7.5, 400 mM NaCl, 10 mM Imidazole pH 7.5, 1 mM ß-mercaptoethanol, Benzonase, DNase I and 1 mM PMSF. After centrifugation (25,000 rpm, 30 min, 10°C) and filtration, the cleared lysate was applied to a HisTrap FF column (Cytiva) and the column was washed with 30 column volumes of wash buffer (20 mM Tris-HCl pH 7.5, 400 mM NaCl, 40 mM Imidazole pH 7.5, and 1 mM ß-mercaptoethanol). Bound protein was eluted with wash buffer supplemented with 340 mM Imidazole pH 7.5 and the elution was combined with His-tagged SUMO protease for cleavage of the N-terminal tag and dialyzed overnight against 20 mM Tris-HCl pH 7.5, 400 mM NaCl, and 1 mM ß-mercaptoethanol. The dialyzed, protease treated sample was passed again over a 5 ml HisTrap FF column to remove cleaved tags and protease, the flow-through was diluted to 150 mM NaCl and 10°C (v/v) glycerol and applied to a HiTrap SP HP column (Cytiva). After washing with 10 column volumes of 20 mM Tris-Cl pH 7.5, 150 mM NaCl, 10°C (v/v) glycerol, and 1 mM ß-mercaptoethanol, proteins were eluted using a gradient increasing the salt concentration to 500 mM NaCl. The peak fractions containing SMG8728–991 were pooled, concentrated, and injected onto a Superdex 75 10/300 GL (Cytiva) equilibrated with 20 mM HEPES/NaOH pH 7.4, 200 mM NaCl, and 1 mM DTT. Again, the peak fraction containing SMG8728–991 was pooled and concentrated to 100–200 µM and flash-frozen in gel filtration buffer using liquid nitrogen. mTORΔN-LST8 and GST-AKT1450–480 were prepared as described before (Gat et al., 2019).

Kinase assays

Request a detailed protocol

The mass spectrometry-based kinase assays were carried out as previously reported (Langer et al., 2020). For experiments involving SMG1i or Torin 2, the assembled reactions were incubated in the absence of ATP for 30 min at 4°C. All reactions involving SMG1i were performed with equal amounts of DMSO present.

Kinase assay based on the use of γ-32P-labeled ATP was carried out as reported. For all assays, 100 nM of kinase complex and 1 μM of substrate were used. All reactions were supplemented with equal amounts of DMSO and incubated without ATP for 30 min at 4°C before starting the experiment. After 30 min at 30°C, phosphorylation reactions were stopped by adding SDS-containing sample buffer and samples were analyzed using an SDS-polyacrylamide gel supplemented with 2,2,2-trichlorethanol (TCE, Acros Organics) for stain-free visualization (Ladner et al., 2004). The TCE-tryptophan reaction was started by exposing the protein gel to UV light for 2 min and proteins were visualized and quantified using a Bio-Rad gel imaging system and the Image Lab software (version 6.1, Bio-Rad) or by Coomassie-staining. Subsequently the gel was washed several times with water to remove residual γ-32P-labeled ATP and phosphoproteins were detected using autoradiography and an Amersham Typhoon RGB biomolecular imager (GE Healthcare). Phosphoprotein signal (P-UPF1 or P-mTORΔN) was quantified by densitometry using Fiji and normalized to the total amount of protein determined using TCE. Independent triplicates of each condition were normalized to the DMSO-only sample and plotted using Prism (GraphPad).

Cross-linking mass spectrometry

Request a detailed protocol

For cross-linking mass spectrometry, 1 μM of SMG1-8-9 complex was incubated with 0.5 mM BS3 for 30 min on ice in a buffer containing 1× PBS, 5 mM MgCl2, and 1 mM DTT. The reaction was quenched by adding 40 mM Tris-HCl pH 7.9 and incubating for 20 min on ice. The sample was spun for 15 min at 18,000×g. For denaturation of the crosslinked proteins, 4 M Urea and 50 mM Tris were added to the supernatant and the samples were sonicated using a Bioruptor Plus sonication system (Diogenode) for 10×30 s at high intensity. For reduction and alkylation of the proteins, 40 mM 2-cloroacetamide (CAA, Sigma-Aldrich) and 10 mM Tris (2-carboxyethyl)phosphine (TCEP; Thermo Fisher Scientific) were added. After incubation for 20 min at 37°C, the samples were diluted 1:2 with MS grade water (VWR). Proteins were digested overnight at 37°C by addition of 1 µg of trypsin (Promega). Thereafter, the solution was acidified with trifluoroacetic acid (TFA; Merck) to a final concentration of 1%, followed by desalting of the peptides using Sep-Pak C18 1 cc vacuum cartridges (Waters). The elution was vacuum dried.

Enriched peptides were loaded onto a 30 cm analytical column (inner diameter: 75 µm; packed in-house with ReproSil-Pur C18-AQ 1.9 µm beads, Dr. Maisch GmbH) by the Thermo Easy-nLC 1000 (Thermo Fisher Scientific) with buffer A (0.1% (v/v) Formic acid) at 400 nl/min. The analytical column was heated to 60°C. Using the nanoelectrospray interface, eluting peptides were sprayed into the benchtop Orbitrap Q Exactive HF (Thermo Fisher Scientific)(Hosp et al., 2015). As gradient, the following steps were programmed with increasing addition of buffer B (80% Acetonitrile, 0.1% Formic acid): linear increase from 8% to 30% over 60 min, followed by a linear increase to 60% over 5 min, a linear increase to 95% over the next 5 min, and finally maintenance at 95% for another 5 min. The mass spectrometer was operated in data-dependent mode with survey scans from m/z 300to 1650 Th (resolution of 60k at m/z=200 Th), and up to 15 of the most abundant precursors were selected and fragmented using stepped Higher-energy C-trap Dissociation (HCD with a normalized collision energy of value of 19, 27, and 35). The MS2 spectra were recorded with dynamic m/z range (resolution of 30k at m/z=200 Th). AGC target for MS1 and MS2 scans were set to 3×106 and 105, respectively, within a maximum injection time of 100 and 60 ms for the MS1 and MS2 scans, respectively. Charge state 2 was excluded from fragmentation to enrich the fragmentation scans for cross-linked peptide precursors.

The acquired raw data were processed using Proteome Discoverer (version 2.5.0.400) with the XlinkX/PD nodes integrated (Klykov et al., 2018). To identify the crosslinked peptide pairs, a database search was performed against a FASTA containing the sequences of the proteins under investigation. DSS was set as a crosslinker. Cysteine carbamidomethylation was set as fixed modification and methionine oxidation and protein N-term acetylation were set as dynamic modifications. Trypsin/P was specified as protease and up to two missed cleavages were allowed. Furthermore, identifications were only accepted with a minimal score of 40 and a minimal delta score of 4. Otherwise, standard settings were applied. Filtering at 1 % false discovery rate (FDR) at peptide level was applied through the XlinkX Validator node with setting simple. A file listing identified cross-links can be found as Figure 4—source data 1.

Pull-down assays

Request a detailed protocol

Pull downs including the SMG8 C-terminus were performed using Strep-TactinXT Superflow beads (IBA) equilibrated with 20 mM HEPES/NaOH pH 7.4, 50 mM NaCl, 10% (w/v) glycerol, 1 mM DTT, and 0.1% (w/v) NP-40 substitute (Fluka) unless stated otherwise. 1.5 μM TwinStrep-tagged protein was combined with 15 μM untagged protein in the buffer described above and incubated at 4°C for 30 min before adding equilibrated resin. The complete reaction was incubated for 1.5 hr, beads were washed four times with 20× resin volume of buffer, and bound protein eluted by adding buffer supplemented with 50 mM Biotin.

To analyze the effect of SMG1i binding to SMG1 on the interaction between the SMG1-8-9 kinase complex and UPF1, 0.12 μM TwinStrep-tagged SMG1-8-9 was combined with DMSO, or the respective amounts of SMG1i in buffer containing 1× PBS, 5 mM MgCl2, 1 mM DTT, and 0.1% (v/v) NP-40. After incubation for 30 min at 4°C, 0.4 μM of untagged UPF1 was added and the reactions were combined with MagStrep ‘type3’ XT beads (IBA). Samples were incubated for 30 min before washing four times with 20× resin volume of buffer and precipitation of bound proteins using SDS-containing sample buffer.

For all pull-down assays, input and elution samples were analyzed by SDS-PAGE and stained with Coomassie.

Cryo-EM sample preparation, data collection, and data processing

Request a detailed protocol

Grids were prepared as described before (Langer et al., 2020), with the difference that SMG1-8-9 was incubated with either 4 μM SMG1i or 1 mM AMPPNP for 30 min on ice in 1× PBS, 5 mM MgCl2, and 1 mM DTT before adding 0.04% (v/v) n-octyl-ß-D-glucoside and plunging using an ethane/propane mixture and a Thermo Fisher FEI Vitrobot IV.

Cryo-EM data were essentially acquired as reported previously using a Thermo Fisher FEI Titan Krios G3 microscope equipped with a post-column GIF (energy width 20 eV). The Gatan K3 camera was used in counting mode and data were acquired using SerialEM (Mastronarde, 2005) and a beam-tilt based acquisition scheme. The nominal magnification during data collection for both data sets was 105,000×, corresponding to a pixel size of 0.8512 Å at the specimen level. The SMG1i sample was imaged with a total exposure of 89.32 e−/Å2 evenly spread over 4 s and 40 frames. The AMPPNP sample was imaged with a total exposure of 60.99 e−/Å2 evenly spread over 5.7 s and 40 frames at CDS mode. For both data collections, the target defocus ranged between –0.5 and –2.9 μm.

Movies were pre-processed on the fly using Focus (Biyani et al., 2017), while automatically discarding images of poor quality. Picked candidate particles were extracted in RELION 3.1 (Zivanov et al., 2018; Scheres, 2012). After two rounds of reference-free 2D classification, particles were imported to CryoSPARC v2 (Punjani et al., 2017) for further processing. After additional sorting in 2D, particles were iteratively classified in 3D to separate SMG1-8-9 and SMG1-9. Particles were distributed over multiple batches for all classification steps. To visualize highly flexible parts of the SMG1 insertion domain in the SMG1i-bound complex, the particle stack containing SMG1-8-9 data was submitted to 3D variability analysis into six clusters. Filtering according to local resolution was employed to better visualize parts of the complex with different resolutions after refinement. Following Ctf refinement, a mask was generated based on the more rigid SMG1 body to focus refinement of the cleaned SMG1i-bound SMG1-8-9 particles on the inhibitor-bound SMG1 active site. For details refer to Figure 2—figure supplement 2 and Figure 3—figure supplement 3. 3D FSC curves were calculated using the 3D FSC online application (Tan et al., 2017). Map versus model FSCs were calculated within the PHENIX software suit (Adams et al., 2011; Liebschner et al., 2019).

Model building

Request a detailed protocol

Model building was carried out using Coot (version 0.8.9.2) (Emsley et al., 2010; Emsley and Cowtan, 2004) and iterative rounds of real-space refinement in the PHENIX software suit (Adams et al., 2011; Liebschner et al., 2019) and was based on our previously published models of SMG1-8-9 (PDB identifiers: 6SYT, 6Z3R). Geometric restraints for the SMG1i molecule were calculated with eLBOW (Moriarty et al., 2009). For details see Supplementary file 1. Structure visualization and analysis were carried out using UCSF ChimeraX (version 1.2.5) (Goddard et al., 2018) and PyMOL (version 2.3.2).

To include the AlphaFold derived SMG8 C-terminus model, we downloaded the predicted model for full-length SMG8 from the AlphaFold Protein Structure Database (Jumper et al., 2021). We removed all parts with a per-residue confidence score below 50 and superimposed the prediction with our EM data-derived model for the SMG8 N-terminal half (residues 4–559) using Coot. After fusing the roughly positioned AlphaFold-predicted SMG8 C-terminal half (560–991) to our SMG1i-bound SMG-1-8-9 model (PDB identifier: 7PW4), we rigid body-fitted the SMG8 C-terminus into the EM density (resulting in PDB 7PW5). It is worth noting that the somewhat different orientation of the SMG8 C-terminal domain with respect to the N-terminal part of the molecule observed when comparing the EM density to the AlphaFold model before rigid body-fitting is in agreement with the flexibility of this domain generally present in the cryo-EM data. An overview of refinement statistics is shown in Supplementary file 1.

Data availability

Models have been deposited in the PDB under the accession codes 7PW4 (SMG1-8-9 bound to SMG1i), 7PW5 (SMG1-8-9 bound to SMG1i, with SMG8 C-terminus), 7PW6 (SMG1 body bound to SMG1i), 7PW7 (SMG1-9 bound to SMG1i), 7PW8 (SMG1-8-9 bound to AMPPNP) and 7PW9 (SMG1-9 bound to AMPPNP). The corresponding EM maps have been deposited in the EMDB under the accession codes EMD-13674, EMD-13675, EMD-13676, EMD-13677, EMD-13678 and EMD-13679.

The following data sets were generated
    1. Langer LM
    2. Conti E
    (2021) RCSB Protein Data Bank
    ID 7PW4. Human SMG1-8-9 kinase complex bound to a SMG1 inhibitor.
    https://www.rcsb.org/structure/7PW4
    1. Langer LM
    2. Conti E
    (2021) RCSB Protein Data Bank
    ID 7PW5. Human SMG1-8-9 kinase complex bound to a SMG1 inhibitor with predicted SMG8 C-terminus.
    https://www.rcsb.org/structure/7PW5
    1. Langer LM
    2. Conti E
    (2021) RCSB Protein Data Bank
    ID 7PW6. Human SMG1-8-9 kinase complex bound to a SMG1 inhibitor - SMG1 body.
    https://www.rcsb.org/structure/7PW6
    1. Langer LM
    2. Conti E
    (2021) RCSB Protein Data Bank
    ID 7PW7. Human SMG1-9 kinase complex bound to a SMG1 inhibitor.
    https://www.rcsb.org/structure/7PW7
    1. Langer LM
    2. Conti E
    (2021) RCSB Protein Data Bank
    ID 7PW8. Human SMG1-8-9 kinase complex bound to AMPPNP.
    https://www.rcsb.org/structure/7PW8
    1. Langer LM
    2. Conti E
    (2021) RCSB Protein Data Bank
    ID 7PW9. Human SMG1-9 kinase complex bound to AMPPNP.
    https://www.rcsb.org/structure/7PW9
    1. Langer LM
    2. Conti E
    (2021) Electron Microscopy Data Bank
    ID EMD-13674. Human SMG1-8-9 kinase complex bound to a SMG1 inhibitor.
    https://www.ebi.ac.uk/pdbe/entry/emdb/EMD-13674
    1. Langer LM
    2. Conti E
    (2021) Electron Microscopy Data Bank
    ID EMD-13675. Human SMG1-8-9 kinase complex with AlphaFold predicted SMG8 C-terminus, bound to a SMG1 inhibitor.
    https://www.ebi.ac.uk/emdb/error/entry/EMD-13675
    1. Langer LM
    2. Conti E
    (2021) Electron Microscopy Data Bank
    ID EMD-13676. Human SMG1-8-9 kinase complex bound to a SMG1 inhibitor - SMG1 body.
    https://www.ebi.ac.uk/emdb/error/entry/EMD-13676
    1. Langer LM
    2. Conti E
    (2021) Electron Microscopy Data Bank
    ID EMD-13677. Human SMG1-9 kinase complex bound to a SMG1 inhibitor.
    https://www.ebi.ac.uk/emdb/error/entry/EMD-13677
    1. Langer LM
    2. Conti E
    (2021) Electron Microscopy Data Bank
    ID EMD-13678. Human SMG1-8-9 kinase complex bound to AMPPNP.
    https://www.ebi.ac.uk/emdb/error/entry/EMD-13678
    1. Langer LM
    2. Conti E
    (2021) Electron Microscopy Data Bank
    ID EMD-13679. Human SMG1-9 kinase complex bound to AMPPNP.
    https://www.ebi.ac.uk/emdb/error/entry/EMD-13679

References

    1. Emsley P
    2. Cowtan K
    (2004) Coot: model-building tools for molecular graphics
    Acta Crystallographica. Section D, Biological Crystallography 60:2126–2132.
    https://doi.org/10.1107/S0907444904019158
    1. Emsley P
    2. Lohkamp B
    3. Scott WG
    4. Cowtan K
    (2010) Features and development of Coot
    Acta Crystallographica. Section D, Biological Crystallography 66:486–501.
    https://doi.org/10.1107/S0907444910007493

Decision letter

  1. Philip A Cole
    Senior and Reviewing Editor; Harvard Medical School, United States
  2. Kacper B Rogala
    Reviewer

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Decision letter after peer review:

Thank you for submitting your article "Cryo-EM reconstructions of inhibitor-bound SMG1 kinase reveal an autoinhibitory state dependent on SMG8" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by Reviewing/Senior Editor (Philip Cole). The following individual involved in review of your submission has agreed to reveal their identity: Kacper B Rogala (Reviewer #1).

The Reviewing Editor has drafted this to help you prepare a revised submission. Although there are a substantial number of points listed below, the vast majority relate to manuscript presentation and writing issues rather than requested additional experimentation.

Essential revisions:

1. A significant concern with the manuscript is that there is no clear way to test the autoinhibition model. UPF1 binding itself is proposed to relieve autoinhibition, meaning that activity towards UPF1 would not be a readout of autoinhibition. One possible prediction of the model would be that activity of the complex toward a short peptide including the UPF1 phosphorylation sites should be low due to occlusion of the SMG1 catalytic cleft. Addition of an UPF1 truncation mutant lacking the C-terminal tail would be expected to activate the SMG1 complex toward phosphorylation of a peptide substrate. The SMG1-SMG9 dimer is modestly more active on UPF1 than the full SMG1-SMG8-SMG9 heterotrimer (see the Deniaud et al. paper cited in the text). However, using a peptide substrate, that difference in activity would be expected to us to be much larger. It would make the model more powerful if such ancillary enzymatic evidence were included.

2. Figure 1. Phosphorylation assays presented in this figure are missing an important control. And that is the reaction without the enzymes – for both SMG1 and mTOR. It is difficult to judge the extent of SMG1i inhibition if we cannot see the baseline with and without the enzymes. In general, monitoring of the AKT1 peptide phosphorylation with this method seems rather noisy, and this reviewer is unsure how relevant that specificity data is in the absence of proper controls. The same method presented in the authors' earlier paper (Gat et al., 2019, PMID: 31792449) seemed to be of higher quality.

3. Figure 1 is presented as providing evidence that SMG1i is selective for SMG1 over mTOR, and that it works by an ATP competitive mechanism. None of the experiments in Figure 1 or the associated supplementary figures actually investigate whether the compound is ATP-competitive; this was established later by the cryo-EM structure. It is worth noting that a non-selective analog bound at the ATP binding site when co-crystallized with PI3K-γ (reported in the paper that described SMG1i). Furthermore, in that original work, SMG1i was reported to be >400-fold selective for SMG1 over mTOR, and it was also tested against a number of other kinases. From this standpoint Figure 1 should be acknowledged as being confirmatory to that prior work rather than a new result, and it could be moved to a supplemental figure. There are also some technical issues with those experiments – the mTOR kinase assays used for replicates cover only a narrow concentration range and have high background; it's not clear why the full dose response shown in Figure 1 S2A wasn't simply repeated three times. The SMG1 activity assays were also performed at high kinase concentration, so the true potency of the compound cannot be determined. The original paper reported sub-nM inhibition of SMG1.

4. PRD density in the apo vs inhibitor/AMPPNP-bound mode of the SMG1-8-9 complex. The authors do not seem to offer any visual cues or thoughts to better understand why the PRD density would only rigidify in the presence of the inhibitor or ATP analogs. And within that group – why would the PRD density be more pronounced in the presence of the inhibitor versus the ATP analog? Are there any conformational changes in the SMG1-8-9 complex upon binding of the inhibitor/AMPPNP? Can the authors trace any specific residue/domain movement and rationalize this observation?

5. Methods. For the references cited in the methods, please use the primary articles on which the technique is developed rather than a subsequent paper that cites the original reference.

6. The presented structural data is clearly of high quality but additional figure panels covering the process of cryo-EM data processing in greater detail should be produced:

a. Representative micrograph of collected cryo-EM data.

b. 2D classes of their cryo-EM data.

c. Angular distribution plot for each deposited reconstruction in addition to their 3D FSC plot.

d. FSC plots for map vs. model.

7. Crosslinking-MS analysis on the SMG1-SMG8-SMG9 complex (as well as the complex including UPF1) was also previously done, and provided largely similar results to those reported in Figure 4B. Though not the intention of the authors, the use of bold lines for the highlighted crosslinks suggests that those specific interactions were identified at higher confidence or provided a higher signal; just keeping with the color scheme would be sufficient for the purpose of drawing attention to those crosslinks. In addition to the groove that approaches the IP6 binding site where additional density is seen in the inhibitor bound complex, the C-terminus of the insert region also crosslinks at multiple points within the catalytic domain itself. Can the authors reconcile the observed crosslinks to both regions?Reviewer #1 (Recommendations for the authors):

Claims/Observations:

(1) The SMG1i compound is specific to SMG1 because it engages unique residues in the active site (unique to SMG1 versus other PIKKs). Beyond the structural rationale provided by solid cryo-EM work, the authors attempted to strengthen this claim with in vitro kinase assays.

This reviewer believes that this data is of high value, but some controls for the in vitro kinase assays are missing and should be addressed to support this claim. Please see the detailed comments section.

(2) Structural rationale for the autoinhibitory function of the insertion domain of SMG1 and the C-terminal section of SMG8 – on the overall activity of the SMG1-8-9 complex.

Putative density belonging to the insertion domain (specifically its N-terminal section) is found in the active site of SMG1 – occluding the previously-mapped SQ substrate binding site, and explaining the previously-reported autoinhibitory function of this domain. Corresponding density has been observed in other PIKKs, and is normally referred to as PRD for PIKK regulatory domain. The authors were unable to unambiguously assign the amino-acid register for the putative PRD density due to its rather weak resolution. Their cross-linking mass spectrometry work confirmed that the insertion domain makes many contacts with the FATKIN domain of SMG1, and an extra few cross-links with the C-terminus of SMG8. in vitro binding assays with purified SMG1-insertion-domain and SMG8-C-terminus confirmed direct binding between these two regions of the complex, offering an explanation as to why the PRD section of the insertion domain is only rigidified in the context of the full SMG1-8-9 complex, and not the sub-complex lacking SMG8. They conclude that SMG8 stabilizes the PRD inside the SMG1 substrate binding site.

Given the wide range of methods applied to support this claim, this reviewer finds the evidence rather compelling. There are some points that should be addressed – either to clarify some statements or to expand on them. Please see the detailed comments section.Reviewer #2 (Recommendations for the authors):

Langer et al., present cryo-EM reconstructions of the PIKK family SMG1-9 and SMG1-8-9 kinase complexes bound to a SMG1 inhibitor or to an ATP analogue at sub-4 Å resolutions. Together, these structures show (1) the molecular interactions that define the specificity of a SMG inhibitor and (2) reveal an autoinhibitory function of the SMG1 insertion domain. Additional biochemical and cross-linking mass spectrometry work shows interactions between the SMG1 insertion domain and the SMG8 C-terminus. It clarifies how the SMG1 insertion domain and the SMG8 mediate autoinhibition of SMG1. Additionally, the authors show that the employed inhibitor has a high specificity for the inhibition of SMG1 in vitro.

The manuscript is written very clearly. The authors' data is overall well-presented and represents the first structure of SMG1 bound to an inhibitor. The presented structural data is of high quality and appears plausible. Together, the authors' claims are supported by the presented data but additional figures and expansion of the methods section to present their cryo-EM data analysis in greater detail will increase overall clarity.

Reviewer #3 (Recommendations for the authors):

The SMG1-SMG8-SMG9 kinase complex phosphorylates the protein UPF1 to promote nonsense-mediated mRNA decay, an essential co-translational quality control mechanism for degradation of messages with premature stop codons. This manuscript from Langer et al. describes cryo-EM structures of the SMG1-SMG8-SMG9 complex bound to a selective ATP-competitive small molecule inhibitor. The structure provides a rationale for why the compound is selective for SMG1 over related kinases such as mTOR and DNA-PK, which should prove useful in furthering inhibitor design and development for this class of kinases. Intriguingly, the inhibitor bound structure revealed additional density absent from prior reconstructions of the complex arising from am "SMG1 insertion region" C-terminal to its catalytic domain. While this density could not be fit to an atomic model, a portion of it occludes the catalytic cleft, suggesting that it mediates cis-autoinhibitory regulation of the kinase. Furthermore, this region appears to contact the SMG8 subunit, and the added density is absent from cryo-EM reconstructions of an SMG1-SMG9 complex lacking SMG8. Both the SMG1 intramolecular contacts and SMG1-SMG8 intermolecular contacts were supported by crosslinking-mass spectrometry analysis, which appeared largely similar to a previously reported analysis using the same method. These observations support a model by which SMG8, through direct interactions with the SMG1 insertion region, has a key role in autoinhibition of the kinase. These studies are important for the field in that how the SMG1 complex is regulated has previously been obscure. A weakness of the study is this structural hypothesis could not be independently tested. Indeed, given that the SMG1 insertion region was not observed at high resolution, it would not be possible to design mutants to disrupt key interactions and examine their impact on SMG1 kinase activity. The authors hypothesize that these autoinhibitory interactions are relieved by substrate binding, which is consistent with prior cryo-EM characterization of the SMG1-SMG8-SMG9-UPF1 complex. However, this potential substrate activation mechanism, while plausible, is not directly addressed in the manuscript.

https://doi.org/10.7554/eLife.72353.sa1

Author response

Essential revisions:

1. A significant concern with the manuscript is that there is no clear way to test the autoinhibition model. UPF1 binding itself is proposed to relieve autoinhibition, meaning that activity towards UPF1 would not be a readout of autoinhibition. One possible prediction of the model would be that activity of the complex toward a short peptide including the UPF1 phosphorylation sites should be low due to occlusion of the SMG1 catalytic cleft. Addition of an UPF1 truncation mutant lacking the C-terminal tail would be expected to activate the SMG1 complex toward phosphorylation of a peptide substrate. The SMG1-SMG9 dimer is modestly more active on UPF1 than the full SMG1-SMG8-SMG9 heterotrimer (see the Deniaud et al. paper cited in the text). However, using a peptide substrate, that difference in activity would be expected to us to be much larger. It would make the model more powerful if such ancillary enzymatic evidence were included.

This point addresses two different albeit connected aspects: autoinhibition and activation.

The auto-inhibition mechanism is the focus of this manuscript.

There have been several papers suggesting the presence of negative regulation based on biochemical data but the molecular mechanism for these observation had remained unclear. For example, a negative regulatory effect of the SMG1 insertion domain has been reported in Deniaud et al., 2015 and in Zhu et al., 2019. Our observation from the cryo-EM analyses that the SMG1 insertion domain directly blocks the substrate binding path within the SMG1 active site rationalizes the negative regulation observed in biochemical data. In addition, autoinhibition by the PRD/insertion domain has been observed structurally in related kinases (see for example Stakyte et al., 2021 (human ATM) or Chen et al., 2020 (human DNA-PKc)). Furthermore, Arias-Palomo et al., 2011, Deniaud et al., 2015 and Zhu at al., 2019 described a negative regulatory effect of SMG8. Our structural data rationalize this effect: e.g. the dependency of SMG1 autoinhibition on the presence of SMG8 in the complex, the observation of close proximity of the SMG1 insertion domain and the SMG8 C-terminus as well as our discovery of a direct interaction between these parts of the complex via pull-downs using purified proteins. Overall, this allows us to integrate the existing biochemical data into a model for the concerted action of the SMG1 insertion domain and the SMG8 C-terminus to induce an autoinhibited state of the SMG1-8-9 kinase complex.

The activation mechanism is a very interesting aspect and we agree with the Reviewer/s that this remains unclear. However, the activation mechanism is not the focus of this manuscript. We briefly discuss the possible relationship between autoinhibition and activation at the end of the manuscript, to point out unanswered questions and future avenues of research.

We are grateful for the suggestions of the Reviewer/s on how to tackle the activation aspect of the process, though there are technical issues that complicate such experiments.

1) Our data indicate that ideal phosphorylation motifs bind the SMG1 active site more tightly than the SMG1 insertion domain (high resolution observed for substrate peptide in Langer et al., 2020, as compared to medium resolution observed for SMG1 insertion domain in the current manuscript). Therefore, depending on its concentration relative to the kinase, an ideal peptide will be able to compete with the insertion domain autoinhibition in this in vitro setup without any additional activation. Higher peptide concentrations might simply mimic the effect of optimal positioning of the UPF1 N- and C-terminal regions with respect to the SMG1 kinase in an in vivo NMD complex, independent of direct interaction of SMG1-8-9 and UPF1. We have included an additional sentence to our discussion to stress this possibility:

"An alternative explanation would be that optimal positioning of UPF1 with respect to the SMG1 kinase in vivo will allow the unstructured N- and C-terminal ends of UPF1 to efficiently compete with the SMG1 insertion domain for binding to the SMG1 active site. This would be mimicked by the high protein concentration used in in vitro assays (such as Figure 4 Supp. 3 B)".

2) Our SMG1-8-9 purifications contain a mixture of SMG1-8-9 and SMG1-9 complexes, as evidenced by the reconstructions reported in the current manuscript and we were so far not able to separate these subpopulations during purification. The presence of a significant amount of SMG1-9 in our purifications means that any effect observable in the assay suggested by the Reviewer/s would be greatly diluted. Further aggravating this issue, overexpression of only SMG1 and SMG9, but not SMG8, in HEK293T cells results in preparations that still contain endogenous SMG8 (see Author response image 1).

Author response image 1
Download asset Open asset
SDS-PAGE analysis of complexes purified from cells transfected to stably express either TwinStrep-SMG1-8-9 or TwinStrep-SMG1-9, as indicated.

Note the co-purification of endogenous SMG8 from the cell line not transfected with SMG8.

In general, understanding of how SMG1-8-9 activation is regulated will require insights into the molecular interplay between SMG1-8-9, UPF1, other NMD factors and a terminating ribosome. This is an area of future studies that goes beyond the scope of this manuscript.

2. Figure 1. Phosphorylation assays presented in this figure are missing an important control. And that is the reaction without the enzymes – for both SMG1 and mTOR. It is difficult to judge the extent of SMG1i inhibition if we cannot see the baseline with and without the enzymes. In general, monitoring of the AKT1 peptide phosphorylation with this method seems rather noisy, and this reviewer is unsure how relevant that specificity data is in the absence of proper controls. The same method presented in the authors' earlier paper (Gat et al., 2019, PMID: 31792449) seemed to be of higher quality.

We thank the Reviewer/s for pointing this out. We have added three new experiments to address this point:

We have included assays with a "no kinase" control for both SMG1 and mTOR (Figure 1 Supp. 2 D and E).

As a further control, we have included the mTOR inhibitor Torin 2 into this experiment. This assay was carried out in triplicates for both SMG1 as well as mTOR allowing quantification (Figure 1 Supp. 2 F).

In addition, we have also repeated the mTOR triplicate experiment shown in Figure 1 E and Figure 1 Supp. 2 C to improve the quality of the gels. We would also like to draw the attention of the reviewer to the Figure captions and the Methods section, where we describe that we quantify mTOR autophosphorylation to avoid problems with the weaker AKT1 peptide signal.

3. Figure 1 is presented as providing evidence that SMG1i is selective for SMG1 over mTOR, and that it works by an ATP competitive mechanism. None of the experiments in Figure 1 or the associated supplementary figures actually investigate whether the compound is ATP-competitive; this was established later by the cryo-EM structure.

We agree with the Reviewer/s. We have removed claims concerning ATP competition in context with Figure 1 and introduce this observation when describing the structural data.

It is worth noting that a non-selective analog bound at the ATP binding site when co-crystallized with PI3K-γ (reported in the paper that described SMG1i). Furthermore, in that original work, SMG1i was reported to be >400-fold selective for SMG1 over mTOR, and it was also tested against a number of other kinases. From this standpoint Figure 1 should be acknowledged as being confirmatory to that prior work rather than a new result, and it could be moved to a supplemental figure.

We agree with the Reviewer/s that the data on SMG1i compound presented in Figure 1 are confirmatory, although we point out that this is actually the first time that in vitro raw data using purified components are shown (in Gopalsamy et al., 2012, there is a Table reporting affinities, but there are no in vitro raw data nor gels showing the quality of enzymes used for the characterization). The data on SMG1i compound presented in Figure 1 are meaningful in the context of this manuscript, as they provide the biochemical evidence for the feasibility of structural characterization using this experimental system. Since these data set the stage for the structural analysis, we would prefer to keep these data in main figure, while specifying that they confirm/validate previous reports:

"In agreement with previous results (Gopalsamy et al. 2012), we concluded that SMG1i displays high potency and specificity in inhibiting the SMG1 kinase in vitro."

There are also some technical issues with those experiments – the mTOR kinase assays used for replicates cover only a narrow concentration range and have high background; it's not clear why the full dose response shown in Figure 1 S2A wasn't simply repeated three times. The SMG1 activity assays were also performed at high kinase concentration, so the true potency of the compound cannot be determined. The original paper reported sub-nM inhibition of SMG1.

As detailed in our response to point 2, we have repeated the mTOR triplicate experiment and now provide higher quality gels with less background in Figure 1 E and Figure 1 Supp. 2 C. The goal of this experiment was to confirm the specificity of SMG1i for SMG1 over mTOR qualitatively in our experimental setup before proceeding to structural analysis. The concentration range tested in triplicates was chosen based on the titration experiments shown in Figure 1 C and Figure 1 Supp. 2 A. It is beyond the scope of this study to provide a comprehensive characterization of the pharmacological properties of SMG1i.

4. PRD density in the apo vs inhibitor/AMPPNP-bound mode of the SMG1-8-9 complex. The authors do not seem to offer any visual cues or thoughts to better understand why the PRD density would only rigidify in the presence of the inhibitor or ATP analogs. And within that group – why would the PRD density be more pronounced in the presence of the inhibitor versus the ATP analog? Are there any conformational changes in the SMG1-8-9 complex upon binding of the inhibitor/AMPPNP? Can the authors trace any specific residue/domain movement and rationalize this observation?

Overall, we observed better defined density for the entire kinase active site when either a nucleotide analogue or the inhibitor is present as compared to our previously published apo-structure (Gat et al., 2019). We believe that the extended interface between the inhibitor and SMG1 reaching closer towards the PRD (as compared to AMPPNP) might further contribute to stabilization (Figure 2 Supp. 3 D).

We did not detect conformational changes upon ligand binding to the active site. Due to the limited resolution of the PRD density, we cannot interpret interactions at the residue level.

5. Methods. For the references cited in the methods, please use the primary articles on which the technique is developed rather than a subsequent paper that cites the original reference.

We included additional citations where we noticed we had not cited the primary articles on which the technique is developed, such for Relion and Coot for example.

6. The presented structural data is clearly of high quality but additional figure panels covering the process of cryo-EM data processing in greater detail should be produced:

a. Representative micrograph of collected cryo-EM data.

b. 2D classes of their cryo-EM data.

c. Angular distribution plot for each deposited reconstruction in addition to their 3D FSC plot.

d. FSC plots for map vs. model.

We have now split Figure 2 Supp. 1 into two Figures (Figure 2 Supp.1 and Figure 3 Supp. 2) and included map vs. model FSC plots as well as angular distribution plots for each reconstruction. In addition, we included 2D class averages calculated from SMG1-8-9 and SMG1-9 particle stacks in Figure 2 Supp. 2 A to illustrate differences between the complexes in two dimensions. We did not include a micrograph or 2D class averages originating from initial classifications as these have been shown before for our SMG1-8-9 preparation (Gat et al., 2019, Langer et al., 2020). We point out that this manuscript was indeed written as a Research Advance to be connected to Langer et al., 2020.

7. Crosslinking-MS analysis on the SMG1-SMG8-SMG9 complex (as well as the complex including UPF1) was also previously done, and provided largely similar results to those reported in Figure 4B.

There are several reasons why we decided to carry out an independent XL-MS analysis:

1) to the best of our knowledge, raw data of the previously published cross-linking MS experiments (Deniaud et al., 2015) are not publicly available, and we therefore decided to obtain raw data ourselves.

2) We wanted to carry out cross-linking experiments in buffer conditions identical to the ones we used for structure determination (for the best possible correlation).

We will upload a spreadsheet with the identified cross-links and other relevant data to accompany the manuscript.

In the revised manuscript, we could actually go a step further, as AlphaFold (Jumper et al. 2021) became available and predicts the fold for the SMG8 C-terminal domain. The AlphaFold structural prediction of the SMG8 C-terminal domain fits remarkably well in the features of the cryo-EM density and allows to map crosslinks we had detected with this portion of the molecule. Thus, our XL-MS and EM data, in combination with the AlphaFold prediction allows us to provide a more complete model of the SMG1-8-9 complex, as illustrated in the newly added Figure 4 Supp. 3.

Though not the intention of the authors, the use of bold lines for the highlighted crosslinks suggests that those specific interactions were identified at higher confidence or provided a higher signal; just keeping with the color scheme would be sufficient for the purpose of drawing attention to those crosslinks.

We have changed the display in Figure 4 B and do now highlight cross-links only by color to avoid misinterpretation.

In addition to the groove that approaches the IP6 binding site where additional density is seen in the inhibitor bound complex, the C-terminus of the insert region also crosslinks at multiple points within the catalytic domain itself. Can the authors reconcile the observed crosslinks to both regions?

There are indeed more cross-links that support our assignment of the additional density to the SMG1 insertion C-terminus region: K2363 and K2370 of the SMG1 kinase domain cross-link to multiple residues of the SMG1 insertion C-terminus and are in close proximity to the observed additional density, albeit located in regions closer to the C-terminal end of the insertion domain (see Author response image 2). We chose to highlight the residues shown in Figure 4 C since this is where the additional density is the strongest and it allows us to illustrate the proximity to the IP6 binding site.

Author response image 2
Download asset Open asset
Residues of the SMG1 kinase domain (K2363, K2370) that cross-link to the SMG1 insertion C-terminus are observed in close proximity to the additional density (red).

Related to Figure 4 C.

There are two residues within the SMG1 kinase site that also cross-link to the SMG1 insertion C-terminus, but are not located in direct proximity to the observed additional density. We think that this reflects the generally flexible nature of the SMG1 insertion domain, which is also indicated by the low-resolution of the visualized densities.

https://doi.org/10.7554/eLife.72353.sa2

Article and author information

Author details

  1. Lukas M Langer

    Max Planck Institute of Biochemistry, Martinsried, Germany
    Contribution
    Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Validation, Visualization, Writing - original draft, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-9977-2427

  2. Fabien Bonneau

    Max Planck Institute of Biochemistry, Martinsried, Germany
    Contribution
    Data curation, Formal analysis, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-8787-7662

  3. Yair Gat

    Max Planck Institute of Biochemistry, Martinsried, Germany
    Contribution
    Investigation, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-2338-9384

  4. Elena Conti

    Max Planck Institute of Biochemistry, Martinsried, Germany
    Contribution
    Conceptualization, Project administration, Resources, Supervision, Writing - original draft, Writing - review and editing
    For correspondence
    conti@biochem.mpg.de
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-1254-5588

Funding

Boehringer Ingelheim Fonds

  • Lukas M Langer

Max-Planck-Gesellschaft

  • Elena Conti

European Commission (ERC Advanced Investigator Grant EXORICO)

  • Elena Conti

Deutsche Forschungsgemeinschaft (SFB1035)

  • Elena Conti

Deutsche Forschungsgemeinschaft (GRK1721)

  • Elena Conti

Deutsche Forschungsgemeinschaft (SFB/TRR 237)

  • Elena Conti

Novo Nordisk Foundation (Exo-Adapt Center)

  • Elena Conti

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

The authors are grateful to Robert Bridges, Rosalind Franklin University of Medicine and Science, and the Cystic Fibrosis Foundation for providing the SMG1 inhibitor. Daniel Bollschweiler and Tillman Schäfer at the MPIB cryo-EM facility for help with EM data collection and Barbara Steigenberger and Elisabeth Weyher at MPIB biochemistry core facility for carrying out mass spectrometry. The authors thank Christian Benda and J Rajan Prabu for maintenance and development of computational infrastructure for EM data processing, Marcela Cueto for TwinStrep-PP7CP, Daniela Wartini for help with tissue culture, and Courtney Long and members of the lab for help with the preparation of the manuscript. This work was supported by funding from the Max Planck Gesellschaft, the European Commission (ERC Advanced Investigator Grant EXORICO), the Novo Nordisk Foundation (Exo-Adapt Center) and the German Research Foundation (DFG SFB1035, GRK1721, SFB/TRR 237) to EC and a Boehringer Ingelheim Fonds PhD fellowship to LL.

Senior and Reviewing Editor

  1. Philip A Cole, Harvard Medical School, United States

Reviewer

  1. Kacper B Rogala

Version history

  1. Preprint posted: July 28, 2021 (view preprint)
  2. Received: July 28, 2021
  3. Accepted: October 25, 2021
  4. Accepted Manuscript published: October 26, 2021 (version 1)
  5. Version of Record published: November 15, 2021 (version 2)

Copyright

© 2021, Langer et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,958
    Page views
  • 288
    Downloads
  • 12
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Lukas M Langer
  2. Fabien Bonneau
  3. Yair Gat
  4. Elena Conti
(2021)
Cryo-EM reconstructions of inhibitor-bound SMG1 kinase reveal an autoinhibitory state dependent on SMG8
eLife 10:e72353.
https://doi.org/10.7554/eLife.72353

Categories and tags

Research organism

Further reading

    1. Structural Biology and Molecular Biophysics

    Structure of substrate-bound SMG1-8-9 kinase complex reveals molecular basis for phosphorylation specificity

    Lukas M Langer, Yair Gat ... Elena Conti
    Short Report Updated

    PI3K-related kinases (PIKKs) are large Serine/Threonine (Ser/Thr)-protein kinases central to the regulation of many fundamental cellular processes. PIKK family member SMG1 orchestrates progression of an RNA quality control pathway, termed nonsense-mediated mRNA decay (NMD), by phosphorylating the NMD factor UPF1. Phosphorylation of UPF1 occurs in its unstructured N- and C-terminal regions at Serine/Threonine-Glutamine (SQ) motifs. How SMG1 and other PIKKs specifically recognize SQ motifs has remained unclear. Here, we present a cryo-electron microscopy (cryo-EM) reconstruction of a human SMG1-8-9 kinase complex bound to a UPF1 phosphorylation site at an overall resolution of 2.9 Å. This structure provides the first snapshot of a human PIKK with a substrate-bound active site. Together with biochemical assays, it rationalizes how SMG1 and perhaps other PIKKs specifically phosphorylate Ser/Thr-containing motifs with a glutamine residue at position +1 and a hydrophobic residue at position -1, thus elucidating the molecular basis for phosphorylation site recognition.

    1. Biochemistry and Chemical Biology
    2. Structural Biology and Molecular Biophysics

    The neuronal calcium sensor NCS-1 regulates the phosphorylation state and activity of the Ga chaperone and GEF Ric-8A

    Daniel Muñoz-Reyes, Levi J McClelland ... Maria Jose Sanchez-Barrena
    Research Article

    The Neuronal Calcium Sensor 1, an EF-hand Ca2+ binding protein, and Ric-8A coregulate synapse number and probability of neurotransmitter release. Recently, the structures of Ric-8A bound to Ga have revealed how Ric-8A phosphorylation promotes Ga recognition and activity as a chaperone and guanine nucleotide exchange factor. However, the molecular mechanism by which NCS-1 regulates Ric-8A activity and its interaction with Ga subunits is not well understood. Given the interest in the NCS-1/Ric-8A complex as a therapeutic target in nervous system disorders, it is necessary to shed light on this molecular mechanism of action at atomic level. We have reconstituted NCS-1/Ric-8A complexes to conduct a multimodal approach and determine the sequence of Ca2+ signals and phosphorylation events that promote the interaction of Ric-8A with Ga. Our data show that the binding of NCS-1 and Ga to Ric-8A are mutually exclusive. Importantly, NCS-1 induces a structural rearrangement in Ric-8A that traps the protein in a conformational state that is inaccessible to Casein Kinase II-mediated phosphorylation, demonstrating one aspect of its negative regulation of Ric-8A-mediated G-protein signaling. Functional experiments indicate a loss of Ric-8A GEF activity towards Ga when complexed with NCS-1, and restoration of nucleotide exchange activity upon increasing Ca2+ concentration. Finally, the high-resolution crystallographic data reported here define the NCS-1/Ric-8A interface and will allow the development of therapeutic synapse function regulators with improved activity and selectivity.

    1. Cell Biology
    2. Structural Biology and Molecular Biophysics

    Baited reconstruction with 2D template matching for high-resolution structure determination in vitro and in vivo without template bias

    Bronwyn A Lucas, Benjamin A Himes, Nikolaus Grigorieff
    Research Advance

    Previously we showed that 2D template matching (2DTM) can be used to localize macromolecular complexes in images recorded by cryogenic electron microscopy (cryo-EM) with high precision, even in the presence of noise and cellular background (Lucas et al., 2021; Lucas et al., 2022). Here, we show that once localized, these particles may be averaged together to generate high-resolution 3D reconstructions. However, regions included in the template may suffer from template bias, leading to inflated resolution estimates and making the interpretation of high-resolution features unreliable. We evaluate conditions that minimize template bias while retaining the benefits of high-precision localization, and we show that molecular features not present in the template can be reconstructed at high resolution from targets found by 2DTM, extending prior work at low-resolution. Moreover, we present a quantitative metric for template bias to aid the interpretation of 3D reconstructions calculated with particles localized using high-resolution templates and fine angular sampling.