Mutations in SKI in Shprintzen-Goldberg syndrome lead to attenuated TGF-β responses through SKI stabilization

Shprintzen-Goldberg syndrome (SGS) is a multisystemic connective tissue disorder, with considerable clinical overlap with Marfan and Loeys-Dietz syndromes. These syndromes have commonly been associated with enhanced TGF-β signaling. In SGS patients, heterozygous point mutations have been mapped to the transcriptional corepressor SKI, which is a negative regulator of TGF-β signaling that is rapidly degraded upon ligand stimulation. The molecular consequences of these mutations, however, are not understood. Here we use a combination of structural biology, genome editing and biochemistry to show that SGS mutations in SKI abolish its binding to phosphorylated SMAD2 and SMAD3. This results in stabilization of SKI and consequently attenuation of TGF-β responses, in both knockin cells expressing an SGS mutation, and in fibroblasts from SGS patients. Thus, we reveal that SGS is associated with an attenuation of TGF-β-induced transcriptional responses, and not enhancement, which has important implications for other Marfan-related syndromes.


INTRODUCTION 41
Shprintzen-Goldberg syndrome (SGS) is a multi systemic connective tissue disorder.

50
The TGF-β family of ligands comprises the TGF-βs themselves, Activins, Nodal, bone 51 morphogenetic proteins (BMPs) and growth differentiation factors (GDFs), and they play 52 pleiotropic roles in embryonic development and tissue homeostasis. In addition, their signaling 53 is deregulated in diverse pathologies (Miller and Hill, 2016). They exert their action by binding 54 to type I and type II serine/threonine kinase receptors at the cell surface (TGFBR1 and 55 TGFBR2 respectively, for the TGF-bs) (Massague, 2012). In the resulting ligand-bound 56 heterotetrameric receptor complex, the type II receptor phosphorylates and activates the type 57 I receptor, which in turn phosphorylates the intracellular mediators, the receptor-regulated 58 SMADs (R-SMADs). Once phosphorylated, the R-SMADs (SMAD2 and SMAD3 in the case 59 of TGF-b, Activin and Nodal) associate with the common mediator of the pathway, SMAD4.

60
The resulting heterotrimeric complexes accumulate in the nucleus where they interact with 61 other transcriptional regulators to activate or repress target gene expression (Massague, 62 2012). Two highly-related co-repressors, SKI and SKIL (formerly known as SnoN) act as 63 negative regulators in the pathway (Deheuninck and Luo, 2009); see below).

64
The role of deregulated TGF-b signaling in Marfan-related syndromes is controversial. 7 been shown to occur naturally in the human colorectal cancer cell lines CACO-2 and SW948, 169 and have lost the ability to bind phosphorylated R-SMADs (De Bosscher et al., 2004). In 170 addition, we used the crystal structure of the MH2 domain of SMAD4 and the SAND domain 171 of SKI, to design two mutations Ala433-->Glu (A433E) and Ile435-->Tyr (I435Y), that would 172 be expected to abolish SMAD4 binding to SKI and SKIL (Wu et al., 2002).

173
We confirmed that these SMAD4 mutants behaved as expected in the rescue cell lines 174 by testing their interaction with SKIL and R-SMADs by immunoprecipitation. As endogenous 175 RNF111 triggers SKIL degradation in TGF-β/Activin dependent manner, the stable SMAD4-176 expressing HaCaT rescue cell lines were incubated with the proteasome inhibitor, MG-132 for 177 3 h prior TGF-β stimulation, to block SKIL degradation. As predicted, the D351H and D537Y 178 SMAD4 mutants had lost their ability to bind SMAD2 upon TGF-b induction, but retained the 179 interaction with SKIL. By contrast, A433E and I435Y SMAD4 mutants were unable to bind

185
Having demonstrated that these mutants behaved as designed, we asked which were 186 able to mediate TGF-b-induced SKIL degradation, using three different assays. In a Western 187 blot assay using nuclear extract, we found that reintroduction of WT SMAD4 in SMAD4-null 188 cells caused a 50% reduction in SKIL levels in TGF-b-induced cells compared to those treated 189 with SB-431542 ( Figure 2B). However, none of the four SMAD4 mutants could rescue TGF-

199
Reintroduction of WT EGFP-SMAD4 conferred the ability to degrade SKIL upon TGF-β 200 treatment, whereas none of the mutant SMAD4s were able to rescue SKIL degradation ( Figure   201 3, arrows). Thus, all three assays demonstrate that a ternary R-SMAD-SMAD4 complex is 202 absolutely necessary for TGF-b-induced SKIL degradation, as is the ability of SMAD4 to 203 interact with SKIL itself. This suggests that within a canonical activated ternary SMAD

219
The SGS mutations discovered so far mostly cluster within this 11-45 region of SKI

232
In both cases, the SGS mutations prevented interaction of the SKI peptide with the SMAD2 233 MH2 domain.

234
We next used a peptide array to gain a better understanding of which amino acids can 235 be tolerated at the positions found to be mutated in SGS and to determine which other amino 236 acids in this region of SKI are essential for the R-SMAD interaction. The SKI peptide 237 corresponding to amino acids 11-45 was synthesized as an array on a cellulose sheet such 238 that each residue in the sequence between residues 19 and 35 was substituted with all 19 239 alternative amino acids ( Figure 4E). The array was probed with a recombinant PSMAD3-9 SMAD4 trimer, generated by co-expressing the SMAD3 and SMAD4 with the TGFBR1 kinase 241 domain in insect cells. The PSMAD3-SMAD4 complex was then detected using a 242 fluorescently-labeled SMAD2/3 antibody. Eight residues are intolerant to almost any amino 243 acid substitution (Thr20, Leu21, Phe24, Ser28, Ser31, Leu32, Gly34 and Pro35). Strikingly, 244 six of these residues are the amino acids known to be mutated in SGS patients, and the array 245 results readily explain why these residues are mutated to a number of different amino acids in 246 SGS ( Figure 4E). In addition, Thr20 and Phe24 are also crucial residues for binding the

252
To discover why these eight amino acids were so crucial for R-SMAD binding, and also to 253 understand why SKI and SKIL only recognize phosphorylated R-SMADs, we solved the crystal 254 structure of the SKI peptide (amino acids 11-45) with a phosphorylated homotrimer of the 255 SMAD2 MH2 domain. We confirmed using SEC-MALLS that the phosphorylated SMAD2 MH2

263
The N-terminal helix of SKI packs against helix 3 of SMAD2, and the C-terminal portion of the 264 SKI peptide, which contains the critical Gly34 and Pro35, forms a sharp turn that is stabilized 265 by pi-stacking coordination between Phe24 of SKI, Trp448 of SMAD2 and Pro35 of SKI ( Figure   266 5B). Moreover, the NE1 of the Trp448 side chain forms a H-bond to the main chain carbonyl 267 group of Gly33, which in turn positions Pro35 for the interaction with Trp448 ( Figure 5B).

268
Furthermore, Glu270 in SMAD2 provides a pocket, which has a negatively charged base that 269 ties down SKI Gly34 through hydrogen-bonding to its main chain amides. Other key 270 interactions involving amino acids identified above as crucial for binding include the main chain 271 carbonyl of Ser31, which forms a hydrogen bond to the ND1 of Asn387 in helix 3 ( Figure 5C), 272 and the hydroxyl group of SKI Thr20, which forms a hydrogen bond with the Gln455 at the 273 end of helix 5 of SMAD2, and is nearly completely buried in the interface ( Figure 5D). The two

302
We used CRISPR/Cas9 technology with a single-stranded template oligonucleotide to 303 knock in the Pro35 à Ser (P35S) mutation into HEK293T cells, and we efficiently generated 304 a number of homozygous clones (Figure 6 Supplement 1A). In three independent clonal cell 305 lines carrying the P35S SKI mutation, the binding to endogenous phosphorylated SMAD2 was 306 severely compromised, compared with WT SKI ( Figure 6A). The binding to SMAD4, however, 307 was unchanged in the mutant cell lines, as the SGS mutations do not affect the SKI SAND 308 domain which is responsible for SMAD4 binding ( Figure 6A). To assess the impact of the 309 P35S SKI mutation on the SKI and SKIL degradation, cells were treated with Activin for 1 or 11 level is almost entirely degraded after 1 h of Activin treatment ( Figure 6B). The presence of 313 mutated SKI had no effect on the Activin-induced degradation of SKIL in these lines ( Figure   314 6B). Thus, the P35S mutation renders SKI completely resistant to ligand-induced degradation.

315
To determine if SGS mutations had the same effect in SKIL, we introduced a G103V 316 mutation into SKIL, corresponding to the SGS mutation G34V in SKI (referred to as SKIL 317 ΔS2/3). Transfection of G103V SKIL in HEK293T cells led to reduction of SMAD2 binding in 318 parental cells (Figure 6 Supplement 1B). The residual binding was mediated via SMAD2's 319 interaction with SMAD4, as it was lost in the SMAD4 knockout cells (Figure 6 Supplement 1B).

320
Binding of SMAD4 in absence or presence of signal was unaffected by the mutation. As 321 observed above for SKI, the SGS mutation in SKIL led to resistance to Activin-induced 322 degradation ( Figure 6 Supplement 1C), indicating that the R-SMAD interaction was essential.

323
In addition, we made a version of SKIL with mutations in the SAND domain (R314A, T315A,

324
H317A and W318E) that rendered it unable to interact with SMAD4 (referred to as SKIL ΔS4).

325
This mutant was also not degraded upon Activin stimulation ( Figure

396
In this study, we have resolved the mechanism of action of SKI and the related protein SKIL, 397 which has allowed us to elucidate the molecular consequences of SKI mutations in SGS. Our 398 proposed mechanism of how SKI acts as a transcriptional repressor of TGF-β/Activin signaling 399 in health and disease is illustrated in Figure 7D, and our results suggest that the mechanism 400 of action of SKIL is equivalent. In the absence of ligand stimulation, in both healthy and

430
SMADs, which also requires the L3 loop of SMAD4. Thus, the authors concluded that the 431 mechanism whereby SKI (and by analogy, SKIL) inhibited TGF-b/Activin signaling was by 432 binding to the activated R-SMADs and SMAD4 in such a way as to disrupt the phosphorylated 433 R-SMAD-SMAD4 complexes required for transcriptional activation (Wu et al., 2002). This 434 mechanism has been supported by the observation that overexpression of SKI and SKIL 435 inhibits TGF-b-induced functional responses (Luo, 2004). The two mechanisms are 436 fundamentally different. In the first, SKI and SKIL are constitutive repressors that need to be 437 degraded to allow pathway activation. In the second, SKI and SKIL act as inducible repressors,

477
Our structural data also elegantly explain why SKI and SKIL only bind to 478 phosphorylated SMAD2 and SMAD3 in the context of an activated SMAD trimer, and not to 479 monomeric SMAD2 or SMAD3. The key residue for this discrimination is Trp448 in SMAD2,                      Table. 685 A peptide array was generated using peptides corresponding to SKI amino acids 11-                                         . * * :