SAM1 domain of SASH1 harbors distinctive structural heterogeneity

Abstract

The sterile alpha motif (SAM) domains are among the most versatile protein domains in biology, and the variety of the oligomerization states contribute to their diverse roles in many diseases. A better understanding of the structure and dynamics of various SAM domains will provide a scientific basis for drug development targeting them. Here, we used SEC-MALS, HPLC, NMR, and other biophysical techniques to characterize the structural features and dynamics of the SAM1 domain in SASH1. SASH1 is a scaffold protein belonging to the same family as SASH3. Unlike the dimerization seen in SASH3′s SAM domain, our SEC-MALS and SE-HPLC showed that SAM1 exists primarily as a less compact monomer with a minor oligomer. NMR assignment, relaxation, and exchange experiments revealed the presence of both a disordered monomer and a more structured oligomer with multiple timescale exchange regimes in solution. Mutagenesis and SE-HPLC showed that D663A/T664K substitutions in SAM1 increased its oligomerization. In sum, this study is the first to characterize a disordered structure for a SAM domain, provides additional evidence and framework for the diversity of SAM domains, and identifies a region in SAM1 as a potential starting point to further characterize the structural mechanism of oligomerization of the domain.

1. Introduction

The scaffold protein SASH1 (SAM and SH3 domain-containing protein 1) is a tumor suppressor belonging to the protein family of the SASH1/SLy (Src homology 3 [SH3] protein expressed in lymphocytes), which also includes SASH3/SLy1 and SAMSN1 (also known as SLy2 or HACS1) (Jaufmann et al., 2021). SASH1 is also referred to as SLy3 and is ubiquitously expressed in all tissues except lymphocytes (Zeller et al., 2003). It has been shown that SASH1 plays different roles in cytoskeletal reorganization (Martini et al., 2011), tumor suppression (Chen et al., 2020; Franke et al., 2019; Lin et al., 2012; Martini et al., 2011), lung development (Coulombe et al., 2019), atherosclerosis (Weidmann et al., 2015), and dermatological pathologies (Araki et al., 2021; Cao et al., 2021; Courcet et al., 2015; Cui et al., 2020; Jaufmann et al., 2021; Shellman et al., 2015; Wang et al., 2017; Wu et al., 2020; Zhang et al., 2016; Zhong et al., 2019; Zhou et al., 2013, 2017). Organizationally, SASH1 is like SAMSN1 and SASH3 but has extended N- and C-termini (Fig. 1A). All three proteins have a conserved region containing SH3 and SAM domains (Fig. 1A), which is essential for the multiple functions of SASH1 including adhesion and migration of the cell, nitric oxide signaling, and pigmentation (Araki et al., 2021; Chen et al., 2020; Coulombe et al., 2019; Courcet et al., 2015; Cui et al., 2020; Franke et al., 2019; Lin et al., 2012; Martini et al., 2011; Shellman et al., 2015; Wang et al., 2017; Wu et al., 2020; Zhong et al., 2019; Zhou et al., 2013, 2017; Burgess et al., 2020). While it is clear that SASH1 is strongly involved in cancer and other pathologies, the molecular and biochemical mechanisms of SASH1 are still not well-defined (Jaufmann et al., 2021). Thus, a better understanding of this region on the atomic level may provide clues for drug design targeting SASH1 to treat various diseases. This study aims to investigate the first SAM (SAM1) of SASH1, which we will refer to as SAM1 from here on.

Fig. 1. Domain and SAM structural homology despite low sequence homology.

A) Domain architecture of the SLy/SASH1 family members of scaffold proteins indicates a similar region containing the SAM domain. B) Sequence alignment of the SASH1-SAM1 domain with other SAM domains from the same family shows significant sequence heterogeneity. Residue colors indicate side-chain properties: yellow, proline; blue, hydrophobic; red, positively charged; magenta, negatively charged; green, polar uncharged; orange, glycine; cyan, aromatic. The alignments were generated with Clustal X 2.0. C) Overlay of the crystal structure of the SAM domain of murine SASH3 (dark magenta; PDB: 6FXF) and the AlphaFold predicted structure of human SAM1 of SASH1 (dark grey; AlphaFold: AF-O94885-F1).

SAM domains are among the most diverse and versatile protein domains in biology and are very common in signaling proteins (Kim and Bowie, 2003). They are roughly 70 amino acids long, share a standard 5-helical structure, and are essential in mediating protein–protein interactions, signaling cascades, transcription, and DNA-repair events (DaRosa et al., 2016; Kim and Bowie, 2003; Kukuk et al., 2019; Mariotti et al., 2016; Vincenzi et al., 2020). SAM domains have also been reported to bind to RNA and lipids (Barrera et al., 2003; Green et al., 2003; Li et al., 2007). SAM dysregulation is implicated in various cancers and diseases (Li et al., 2007; Vincenzi et al., 2020). Indeed, SAM1 of SASH1 is essential in multiple bioprocesses. During lung development, the interaction between SASH1 and β-arrestin 1 in mouse endothelial cells is crucial for SASH1 to regulate alveolar epithelial cell maturation and promote pulmonary surfactant production through nitric oxide signaling (Coulombe et al., 2019). Defect in this SASH1 function leads to lung immaturity, a significant cause of respiratory distress and mortality in preterm infants. Significantly, the deletion of SAM1 disrupts the interaction of SASH1 with β-arrestin 1 (Coulombe et al., 2019). Additionally, a missense mutation in the SASH1 gene resulting in an R644W substitution in SAM1 is associated with a human pigmentation disorder (Araki et al., 2021). Thus, SAM1 could be a potential target for multiple therapeutic interventions.

SAM domains have been described as “kings of diversity” (Kim and Bowie, 2003). They can adopt many different quaternary structures, from large insoluble polymers down to dimers and monomers with unique structural mechanisms, which contribute to their array of known functions (DaRosa et al., 2016; Kim and Bowie, 2003; Knight et al., 2011; Mariotti et al., 2016; Smirnova et al., 2016; Vincenzi et al., 2020). For example, the extensive polymerization of SAM of Tankyrase is in the head-to-tail arrangement seen in most SAM domains (DaRosa et al., 2016; Kim and Bowie, 2003; Mariotti et al., 2016; Vincenzi et al., 2020). However, the SAM domain in SASH3 dimerizes in a tail-to-tail fashion and is important for SASH3’s function in immune regulation (Kukuk et al., 2019). Also, disordered or less structured states have been suggested for SAM domains, such as SAMHD1 and CASKIN2, but none have been explicitly characterized (Buzovetsky et al., 2018; Smirnova et al., 2016).

No structural or biochemical studies have been reported specifically on the SAM1 domain of SASH1 to date. However, two studies are available that offer insight into the possible structure and function. One high throughput study of human SAM domains did include SAM1 and SAM2 but only concluded that neither forms a large polymer visible by electron microscopy (Knight et al., 2011). This suggests that neither SAM1 nor SAM2 is involved in large polymerization events but may be involved in other roles. The other study is the only one that has characterized a SAM domain of the Sly/SASH1 family (Kukuk et al., 2019). This study shows that SASH3-SAM dimerizes utilizing an electrostatic ridge predominantly along helix 5. Though homologous domains, the SAM domains of the SLy/SASH1 protein family have diverse sequences with a percent identity of 43.08 % to 61.54 % among the members (Fig. 1B). Despite this sequence heterogeneity, an AlphaFold (Jumper et al., 2021) structural prediction of SASH1-SAM1 (AF-O94885-F1) from the Uniprot database (ID: O94885; residues 629–72), suggests a high structural similarity to SASH3-SAM (Fig. 1C). Sequence alignment of SASH1-SAM1 and SASH3-SAM showed a high degree of chemical conservation between these two domains at the nine most important residues of the electrostatic ridge (Table S1). In particular, one residue, R262^SASH3, is suggested to contribute to two important electrostatic interactions in the dimerization event (Kukuk et al., 2019). Sequence alignment (Fig. 1B) indicates that R262^SASH3 is homologous to R644^SASH1, which has been reported as a causal mutation in the pigmentation disorder dyschromatosis. Combined, these studies suggest that SAM1 may be involved in a helix 5-to-helix 5 dimerization event similar to that of SASH3-SAM and motivated us to study and characterize the similarities or differences with the SAM domain from SASH3.

Here, SEC-MALS, HPLC, NMR, and other biophysical techniques allowed us to detect the presence of two solution species of SAM1 domain and quantify their ratios. We characterized the dynamics of both species and detected different exchange timescale regimes. Together, our data showed that in solution, SAM1 exists as a disordered or loosely structured monomer that can weakly oligomerize.

2. Results

2.1. SEC-MALS and SE-HPLC show the presence of a predominant monomeric and a less abundant oligomeric state

SAM1 monomer is predicted to be 9.7 kDa. In size exclusion purification on fast performance liquid chromatography (SE-FPLC), SAM1 eluted earlier than expected for a folded monomeric protein (Fig. 2A). In fact, SAM1 eluted significantly earlier than a folded 7 kDa protein reference, HN-S CTD, as revealed by an overlay of the elution profiles (Fig. 2A). This suggested some degree of oligomerization for SAM1 and/or the presence of disorder that decreases the diffusion coefficient. Gel electrophoresis in SDS verified the purified SAM1 domain as a single band, closest to the size of monomeric protein (Fig. S1).

Fig. 2. Probing the oligomerization of SAM1.

A) Preparatory SE-FPLC overlays of SAM1 (red) and the 7 kDa HN-S CTD reference (blue) show the increased effective size of SAM1 according to the hydrodynamic radius. Elution was performed using a HiLoad 16/600 Superdex 75 pg column in the NMR buffer at 5°C. B) SEC-MALS plot shows that the major species is monomeric with a molecular weight of 9.7 kDa, and a second, smaller population of an oligomeric species has a molecular weight of 40.4 kDa, suggestive of a tetramer. C) Stacked analytical SE-HPLC chromatograms from top to bottom of the protein standard, 7 kDa reference of HN-S CTD, SAM1, and SAM1 with the D663A/T664K substitution. A dashed line marks the location of the 13.7 kDa ribonuclease A to distinguish between theoretical monomer and dimer. The protein standard is the Millipore Sigma Protein Standard Mix 15–600 kDa (Thyroglobulin – 660 kDa, Gamma-globulin – 150 kDa, Albumin – 66 kDa, Ribonuclease A – 13.7 kDa, and para-aminobenzoic acid – 0.14 kDa). HPLC was performed using the AdvanceBio SEC column (4.6×300mm ID, 2.7-μm particle size, 300 Å pore size). Isocratic elution was conducted in the NMR buffer at a flow rate of 0.1 mL/min at ambient (25 °C) temperatures.

To determine if SAM1 was in a higher oligomeric state, we ran SEC-MALS (Fig. 2B). The SEC-MALS results show that the primary peak has a molecular weight of 9.7 kDa, consistent with a monomer size. It also indicates the presence of a second minor peak at 40.4 kDa with a population of 1.1–1.9 %, consistent with a tetramer. These highly reproducible results suggest the presence of a predominantly monomeric state and a second minor oligomeric state.

This was further supported by analytical size-exclusion high-performance liquid chromatography (SE-HPLC) (Fig. 2C). The 7 kDa HN-S CTD and the 13.7 kDa ribonuclease A were used as references. In agreement with SEC-MALS, the chromatogram showed a minor peak at a retention time (RT) of 34.4 min (~1% peak area), which corresponds to tetramer oligomerization, and a major peak that corresponds to the monomeric unit at 37.7 min (Fig. 2C). The monomer eluted earlier than the 13.7 ribonuclease A (RT = 38.8 min) indicating that the monomer is in a less compact state. To test the presence of these states at higher sample concentration (as will be the case with NMR, see below), we ran native polyacrylamide gel electrophoresis PAGE (Fig. S2), which again showed two species with different sizes. Although we cannot accurately quantify the population of the larger unit, it is clearly larger than 50 %. We attribute the population differences to different conditions in the experimental setups.

2.2. D663A/T664K, but not R644W and D663R/A685E, shift the equilibrium towards oligomerization

Next, we tested the effect of substitutions in the presumed oligomerization interface on the abundance of these two species. We tested substitutions of R644W, D663R/A685E and D663A/T664K. R644W was chosen because this amino acid substitution had been implicated in a pigmentation disorder, and the arginine is conserved at that location in SASH3-SAM and considered necessary for SASH3 dimerization (Fig. 1B) (Araki et al., 2021; Kukuk et al., 2019). The two double substitutions, D663R/A685E and D663A/T664K, were chosen because their positions in the predicted AlphaFold SAM1 structure (Fig. S3) align with positions in the strongly polymerizing Tankyrase family SAM domains that disrupted polymerization (DaRosa et al., 2016; Mariotti et al., 2016). Neither R644W nor D663R/A685E changed the structural compactness as assessed by SE-HPLC, with the oligomerization peak in the range of 1–3 % of the entire peak area in both cases and no increase in oligomer retention times (data not shown). Results here suggest an alternate disruptive mechanism of R644W in disease other than oligomerization disruption and that neither D663 nor A685 are important for oligomerization or oligomer disruptions. Nevertheless, the D663A/T664K double substitution showed a substantial increase in the oligomerization peak on SE-HPLC (Fig. 2C). Contrary to our expectations, the oligomerization peak increased from 1 % of the peak area in the wild-type SAM1 to more than 50 %. With the previous substitutions of D663R/A685E showing that D663 is not an essential residue in the interaction, T664K is likely introducing an electrostatic interaction between monomers, thereby shifting the solution equilibrium towards the oligomeric state.

As this D663A/T664K substitution series in Tankyrase 1 (TNKS1) disrupted oligomerization, we hypothesized that this mutation introduced an electrostatic interaction with an acidic residue that is not present in the TNKS1 sequence. Alignment of SAM1 and TNKS1-SAM showed that E682^SASH1 replaced G1074^TNKS1 at the N-terminus of helix 5. This part is significantly involved in oligomerization, as well as the region the substitution is located at, according to the broadly canonical SAM homo-oligomerization/polymerization head-to-tail model (DaRosa et al., 2016; Knight et al., 2011; Mariotti et al., 2016; Vincenzi et al., 2020). We hypothesize that this substitution introduces an electrostatic interaction with T664K that may offer a glimpse into understanding the oligomerization mechanism of SAM1.

2.3. NMR backbone assignment of SAM1 reveals the presence of a monomeric disordered and a folded oligomeric state

To characterize the protein dynamics and structure on the atomic level, we used various NMR experiments. Our initial 2D ¹⁵N-¹H heteronuclear single-quantum coherence (HSQC) spectrum at 25°C had fewer resonances than expected and significant overlap of undispersed peaks (Fig. S4). When we shifted the measurement temperature to 5°C, more peaks upfield and downfield of the central cluster in the ¹H dimension appeared, suggesting that the protein was stabilized at lower temperatures. This HSQC (Fig. 3) showed a peak dispersion, which implied the presence of more folded protein (‘dispersed peaks’). On the other hand, multiple peaks clustered in the central amide region indicate the presence of disordered/less compact states (‘undispersed peaks’) (Fig. 3).

Fig. 3. The SAM1 NMR spectrum has both undispersed and dispersed peaks corresponding to disordered and more folded species.

¹⁵N-¹H HSQC spectrum of SAM1 and adjacent residues (627–703) at 5°C showing the backbone N-HN chemical shift assignments of a set of undispersed peaks as obtained from 3D experiments. Average R_1ρ relaxation rates of assigned undispersed peaks (green box) and unassigned dispersed peaks (blue boxes covering 9.0 – 9.6/7.3 – 7.8 ppm in the ¹H dimension) are 9.6 s⁻¹ and 20.3/22.9 s⁻¹, respectively.

Next, we used a standard suite of 3D NMR experiments (HNCACB, CBCAcoNH, HNCO, HNcaCO) to assign the backbone resonances of the SAM1 (Cavanagh, 2007). Unexpectedly, none of the dispersed peaks in the HSQC (between 8.8 and 9.6 ppm) had any corresponding peaks in any of the 3D NMR experiments (Fig. S5). This indicates higher relaxation rates for these residues, which cause faster signal decay. This effect is particularly prominent in the 3D experiments that are less sensitive than the 2D HSQC due to the longer magnetization transfer pathways, relaxing the signals below the noise level. The elevated relaxation rates are incompatible with a monomeric state.

In contrast, the undispersed peaks were typically also present in the 3D spectra (Fig. S5), and we obtained 71 % backbone assignment distributed all over the protein sequence (BMRB entry: 51486). Given this distribution rather than clustering to specific protein regions, the dispersed peaks are unlikely to originate from the same protein state. Instead, it strongly suggests the presence of two different states of SAM in solution, one disordered and monomeric and a second one at least partially folded but in a multimeric state. We note that there are fewer peaks in the dispersed region than there are residues in SAM1. However, the predicted structure features mostly α helix, which typically produces peaks in the non-dispersed region. Therefore, most dispersed peaks originate from non-helical residues.

To identify secondary structure based on the undispersed peaks, we computed the residue-specific Secondary Structure Propensity (SSP) (Marsh et al., 2006) score from HN, N, Cα, Cβ, and CO chemical shift data (Fig. 4A). The SSP score ranges from + 1 for a fully formed helix to −1 for a fully formed sheet and is ~ 0 for loops and disordered residues. Scores larger than 0 were obtained for residues 638–646, 661–666, 671–679, and 691–694. Even though these regions correspond to α-helical segments in the predicted AlphaFold structure (Table S2), none of these SSP scores was above 0.23, indicative of a disordered protein adopting partial α-helical structure rather than a folded protein. To confirm this disorder, we used the Chemical shift Secondary structure Population Inference (CheSPI) (Nielsen and Mulder, 2021), a more recent and refined algorithm to analyze populations of secondary structure elements. CheSPI provides relative populations of the following class elements: extended, helix, non-folded, and turn. CheSPI, in agreement with SSP, shows a high level of disorder for the assigned residues (Fig. 4B). In addition, we used the TALOS+ program, which establishes an empirical relationship between the assigned chemical shifts, and the backbone torsion angles ϕ and ψ and order parameter (S²) that quantifies the fluctuation amplitude of bond vectors (Shen et al., 2009). The calculated secondary structure propensities and the order parameters confirm the disorder for all the assigned peaks based on chemical shift values (Fig. S6). Overall, these results indicate a disordered protein with favorable relaxation properties and another at least partially highly ordered species that suffers from faster relaxation rates resulting in peak disappearance in the 3D NMR experiments.

Fig. 4. Secondary structure calculations indicating the disordered nature of SAM1 with low helical propensity.

A) SSP score: secondary structure propensities derived from chemical shift values of ¹HN, ¹⁵N, Cα, Cβ, and CO. An SSP score of + 1 indicates a fully formed α-helix, while − 1 indicates a fully formed β-sheet. Regions 638–646, 661–666, 671–679, and 691–694 exhibit transient helical propensity. B) CheSPI populations: the plot shows populations of helix (red), turn (green), non-folded (grey), and extended conformation (blue).

To verify the shift in equilibrium towards more oligomerization for SAM1 D663A/T664K as inferred by SE- HPLC, we collected the ¹⁵N-¹H HSQC of this variant. The spectrum shows a dramatic reduction of the overlapped, clustered (undispersed) peaks in favor of more dispersed peaks, especially upfield. This indicates a reduction of the population of the disordered monomer and an increase of folded species in the case of the variant compared to the wild type (Fig. 5).

Fig. 5. Overlay of ¹⁵N-¹H HSQC spectra of WT and D663A/T664K SAM1 shows an increase in peak dispersion and a reduction in disorder for the variant.

The 2D HSQC of D663A/T664K variant (green) has fewer overlapped peaks in the middle of the spectrum but more dispersed peaks upfield and downfield of that region than the WT SAM1 (red). This indicates less presence of disorder and more compact/folded protein for the variant. The expressed SAM1 included SAM1 domain (633–697) plus a few extra amino acids (627–703) of SASH1.

2.4. NMR relaxation experiments confirm the presence of disordered protein and a slower tumbling second state

To characterize the residue-specific dynamics and oligomerization in more detail, we performed various ¹⁵N NMR relaxation experiments on SAM1. We investigated different timescale contributions to the dynamics ranging from pico- to milliseconds (ps – ms) by measuring longitudinal ¹⁵N relaxation (R₁), spinlock ¹⁵N relaxation (R_1ρ), and ¹⁵N-¹H heteronuclear NOE (Het-NOE). We then used R₁ and R_1ρ to calculate residue-specific R₂ (see Methods section). Using the Lipari-Szabo model (Lipari and Szabo, 1982) to describe overall and residue-specific internal motion, R₁, R_1ρ, and R₂ depend on the overall tumbling time (τ_c) of the protein and an order parameter (S²) quantifying the sub-nanosecond fluctuation amplitude of individual H–N bonds. In addition, R₂ may be enhanced by a term R_ex if conformational exchange on the micro- to millisecond (μs – ms) timescale is present. Since our setup for the R_1ρ measurement quenches contributions from exchanges slower than 100 μs and we calculated R₂ from R₁ρ, the detected R_ex is only sensitive to sub 100 μs motions. In addition, we directly measured R₂, whose R_ex contribution is sensitive to the entire μs – ms timescale.

The average R₁, R_1ρ and R₂ (derived from R₁ and R_1ρ) values for the assigned residues are 1.27 s⁻¹, 9.6 s⁻¹ and 10.0 s⁻¹, respectively (Fig. 6). Given the previous indication for disordered protein and therefore the presence of substantial supra-nanosecond motion, we used the R_ex-independent R₁ and Het-NOE data to estimate the average τ_c and S² to be 3 ns and 0.57, respectively (for details on the simulation of the relaxation rates see Methods section). Such a τ_c value is typical for an intrinsically disordered protein (~2.5 ns at 298 K and somewhat larger at the 278 K used here). Similarly, S² is much lower than values of more than 0.8 expected for a globular protein. Then we used τ_c and S² to back- calculate R₂ without R_ex contribution to be ~ 3.6 s⁻¹. The discrepancy of greater than 6 s⁻¹ to the measured value of 10.0 s⁻¹ indicates significant μs-ms timescale conformational exchange throughout the entire protein. The directly measured R₂ average is even larger (12.1 s⁻¹ ) possibly suggesting some minor additional motion on the ms timescale. To independently confirm the presence of such μs – ms motion, we used Carr-Purcell-Meiboom-Gill (CPMG) experiments that are most sensitive to motions on the timescale of 100 μs – 10 ms. Indeed, many of the residues indicate the presence of such motion (Fig. S7).

Fig. 6. NMR relaxation experiments on SAM1.

A) ¹⁵N longitudinal relaxation rates R₁, B) ¹⁵N spin-lock relaxation rates R_1ρ, and C) ¹⁵N-¹H heteronuclear NOE enhancements are plotted versus the residue numbers. All measurements were conducted at 900 MHz field strength. The magenta boxes indicate the region spanning residues 674–689. The larger-than-average values of R_1ρ and Het-NOE and smaller-than-average R₁ for this region demonstrate smaller motional amplitudes on the sub-nanosecond timescale and structural compaction and/or exchange on the μs-ms timescale. The dotted lines show the average relaxation rates.

Notably, the region spanning residues 674–689 features lower-than-average R₁ (1.13 s⁻¹) and higher-than-average R_1ρ and R₂ (derived from R₁ and R_1ρ) rates (10.7 s⁻¹ and 11.5 s⁻¹). This region is part of the N- terminus of helix 5 and an adjacent linker and is known to be involved in head-to-tail oligomerization (DaRosa et al., 2016; Mariotti et al., 2016; Vincenzi et al., 2020). This irregularity can be explained either by the presence of slower effective tumbling within this region due to increased compactness and/or by conformational exchange on the μs-ms time scale. The Het-NOE values also show a minor increase in this region, which indicates that the H–N bond vectors for these residues indeed experience smaller amplitudes of sub-nanosecond motion.

Although we were unable to assign the dispersed peaks in the HSQC spectrum due to their absence in the 3D spectra, we were able to fit R_1ρ relaxation rates for a group of the dispersed peaks (Fig. 3). With averages of R₁ and R_1ρ of 0.95 s⁻¹ and 20.3 s⁻¹, R₂ results in 20.9 s⁻¹. The directly measured R₂ average from the peaks that we were able to evaluate is in reasonable agreement with 17.3 s⁻¹, suggesting no substantial motion on the ms timescale. Overall, the R₂ values are approximately twice as large as those of the non-dispersed peaks and quantitatively explain the disappearance of the peaks in 3D spectra. Given the clear indication of a fully formed local structure, it is unlikely that a further R_ex contribution causes this increase. Instead, it is predominantly caused by an increased overall tumbling time τ_c, likely with an accompanying increase in S². Assuming an S² value of 0.85 as typical for folded proteins, this suggests a τ_c of ~ 14 ns. This is significantly larger than the value expected for a folded monomer. Thus, the species giving rise to the dispersed peaks is in a multimeric state. We note that an alternative explanation for the large τ_c is an increased hydrodynamic radius brought about by disordered parts of an otherwise monomeric state. However, that would also be the case for the state causing the non-dispersed peak set. In addition, all other methods presented above indicate the presence of multimeric states.

To probe motion on the ~ 0.5 – 50 ms timescale, we used ¹⁵N Chemical Exchange Saturation Transfer (CEST). A series of 2D ¹⁵N-¹H HSQCs spectra were erecorded with a weak 5 Hz ¹⁵N radio-frequency field scanning the relevant ¹⁵N frequency range (Vallurupalli et al., 2012). Typically, only the peaks corresponding to major states are visible. Still, if the B₁ field matches the Larmor frequency of a second, possibly low-populated state, the peak intensity of the major state is disturbed. Generally, the CEST profiles of SAM1 do not indicate the presence of such exchange between states. However, there is a signature for exchange between a major and a minor conformation for some residues, mainly 686 and 689, which are located within helix 5 (Fig. 7). Using the CHEMEX (Vallurupalli et al., 2012) software to perform a global fit for the CEST profiles, we obtain an exchange rate (k_ex = k_major/minor + k_minor/major) of 275 s⁻¹ and a population of the minor state of 1.57 %. Together with the relaxation data, this indicates that this protein region undergoes motion on various slow timescales between μs to ms.

Fig. 7. Conformational exchange of SAM1 probed by ¹⁵NCEST NMR spectroscopy.

¹⁵N CEST frequency scanning profiles are shown for residues 686 and 689. The CEST intensities plotted on the y-axis are quenched when a weak ¹⁵N radio-frequency field (x-axis) resonates with the ¹⁵N chemical shift of a potential second state. Both profiles have, in addition to the peak intensity alteration at the frequency of the visible peak (significant dips at 14.5 and 119.0 ppm), a second minor dip (121.5 and 122.0 ppm), indicating the presence of a second lower populated state exchanging with the major state at a rate of ~ 275 s⁻¹ and with a population of 1.57 %. CEST measurements were conducted at 900 MHz field strength.

Finally, we sought to set a limit to the timescale of exchange between the monomer and multimer species. To that end, we performed ZZ-exchange NMR experiments (Mittermaier and Kay, 2009) for both SAM1 WT and SAM1 D663A/T664K. If the exchange was faster than the mixing time (100 ms in our setup) spectrum would feature cross peaks connecting the individual peaks of the same atom in the monomeric and multimeric species. Neither the spectrum of WT nor that of D663A/T664K SAM1 showed the presence of such peaks (Fig. S8), indicating that the exchange minor state of the monomer detected by CEST is not the monomeric state.

In summary, the relaxation experiments show that SAM1 exists as a highly disordered monomer with a slight tendency to form α helices and an at least partially folded oligomer. The monomer, despite its disorder, features a more compact segment that coincides with the oligomerization interface.

2.5. CD data shows the presence of helical structure(s) in SAM1

For independent confirmation of helical propensity in SAM1, we collected circular dichroism (CD) spectra at three different temperatures (5, 25, and 35 °C). Proteins with predominantly α-helical structure show prominent negative bands at 222 and 208 nm and a positive band at 193 nm, while disordered proteins have very low ellipticity above 210 nm and negative bands near 195 nm (Fasman, 1996). The spectrum of SAM1 corresponds to a mixture of disordered and helical content (Fig. 8) as there are negative values at 208 nm and a negative shoulder at 222 nm, although not to an extent expected for a fully helical protein. These signatures of residual helical structure agree with our findings from NMR spectroscopy. The helical content is present over the entire temperature range of 5 – 35 °C, similar to previous reports on other intrinsically disordered proteins (Celestino et al., 2019).

Fig. 8. Circular Dichroism shows helical content in SAM1.

Far-UV CD measurements across three different temperatures (5, 25, and 35°C) reveal the presence of helical content across all three temperatures. The molar ellipticity ratio of 222 to 208 nm for each temperature reading is shown in parenthesis in the legend and indicates minimal change in the helical signature.

3. Discussion

SASH1 has been implicated in a range of cellular processes, including metastasis suppression (Franke et al., 2019), cellular adhesion and migration (Martini et al., 2011), and lung development (Coulombe et al., 2019). Mutations in SASH1 have been associated with cancers (Courcet et al., 2015) and pigmentation diseases (Araki et al., 2021; Cao et al., 2021; Courcet et al., 2015; Cui et al., 2020; Lin et al., 2012; Shellman et al., 2015; Wang et al., 2017; Wu et al., 2020; Zhang et al., 2016; Zhong et al., 2019; Zhou et al., 2013, 2017 ). Yet, there has been no work to study the atomic-level structure and dynamics of SASH1. SAM domains are among the most versatile and essential protein domains in biology (Kim and Bowie, 2003). However, the diversity of their oligomerization has not been well understood, which makes it challenging to target them for drug development. To better understand SASH1 and SAM domains in general, we characterized the structural features and dynamics of SAM1 of SASH1.

For this study, we focused on understanding the structural states and dynamics of SAM1 in solution. Our multiple analyses indicate that SAM1 exists as two distinct species: a predominant disordered monomer and a minor, at least in part folded, oligomer species. Our SEC data showed that the hydrodynamic radius of SAM1 abundant species was larger than that of a monomer suggesting a small oligomer (di-trimer) or a disordered extended monomer (Fig. 2A and 2C). SE-HPLC also showed a small amount (~1% of the peak area) of larger oligomeric species (Fig. 2C). SEC-MALS determined the molecular weights of the major peak to be monomeric with a size of 9.7 kDa (98.4 % mass fraction) and the minor oligomer peak to be a likely tetramer with a size of 40.4 kDa (1.5 % mass fraction) (Fig. 2B). At higher protein concentration, the population of the minor state is significantly higher, as determined from a native gel run. These methods indicate that there is a weak oligomerization preference, and the predominant monomer is less compact than a truly globular protein. This agrees with another study suggesting that SAM1 is not a polymer large enough to be visible by electron microscopy (Knight et al., 2011).

The 2D ¹⁵N-¹H HSQC spectrum showed distinct and clearly dispersed peaks at 5°C, indicating the presence of a more folded protein in solution. The spectrum at 25°C features less dispersed peaks, which indicates that the protein experiences increased conformational exchange, resulting in peak broadening or loss of structure in the oligomeric state. The same behavior upon temperature increases from 5 to 25°C has been previously demonstrated for the SAM domain in CASKIN2 (Smirnova et al., 2016). Recent computational studies of SAM domains EphA2 and Ship2 (Li et al., 2022) showed that quenching of structural fluctuation at decreasing temperature lowers the entropic penalty for dimerization. Similarly, multimerization may be prevented in SAM1 by intrinsic flexibility at a higher temperature. However, at lower temperatures, fluctuations are reduced such that the protein adopts a more compact structure that, in turn, allows oligomerization. Analyzing dispersed and undispersed sets of NMR peaks using chemical shifts and relaxation rates at 5 °C, we identified the presence of both disordered monomer and at least partially folded oligomer. While we could not assign the resonances of the dispersed peaks due to a large transverse relaxation rate corresponding to a tumbling time of ~ 14 ns, we achieved a 71 % backbone assignment for the disordered species of SAM1. Chemical shift-based secondary structure propensity scores from multiple algorithms further supported that the assigned peaks belong to a disordered protein and that some regions adopt a partial helical structure. Circular dichroism confirmed the presence of both helical and disordered characters over the temperature range from 5 °C to 35 °C. The helical propensity may be partially attributed to transient helical formation in the monomer and the folded oligomeric state. Indeed, a large portion of the dispersed NMR peak set falls into the frequency range that coincides with disordered protein. This is expected for α helix, where peaks would overlap with disordered protein peaks, but not for β sheet. In support of this, both AlphaFold structure prediction and comparison to other SAM domains of known structure indicate a dominant presence of α helix.

The NMR relaxation rates and CEST profiles for residues 674–689 that deviate substantially from the average pattern of the entire SAM1 suggest increased compactness in this region and conformational exchange on the μs-ms timescale. This region is involved in both the dimerization interface of SASH3 as well and the oligomerization interface of most other SAM domains. The elevated rigidity of the monomer is likely required for efficient binding to other SAM domains. The conformational exchange in the μs-ms time regime then is either an exchange between different monomeric conformations or an exchange between monomeric and dimeric states (not to be confused with the multimeric states detected by SE-HPLC and -FPLC, MALS, native gel and the second set of NMR peaks). All these observations offer a glimpse at the potential initial seeding event of the oligomerization process of SASH1-SAM1. Once dimerization has occurred, secondary structure forms, and higher-order oligomers are assembled. This process, however, must be much slower than the initial encounters since we observe two separate species in all our experiments. Additional studies will be needed to test this hypothesis.

To further understand the mechanism of oligomerization, we utilized mutagenesis combined with SE-HPLC and NMR. Based on the sequence homology, SAM1 D663A/T664K mimics oligomer-disrupting substitutions used to solubilize TNKS1 SAM (DaRosa et al., 2016). These substitutions were predicted to remove an electrostatic interaction and add a repulsive electrostatic interaction. Surprisingly, SAM1 D663A/T664K drastically shifts the monomer to a predominant oligomerized species in solution. This drastic change may come from introducing a basic lysine side chain at position 664 on helix 3. Sequence alignment with TNKS1 SAM suggests a potential interaction with E682 on the N-terminal of helix 5. Though unexpected, this indicates that the oligomerization event occurs in a head-to-tail fashion, much like the canonical SAM polymerization events, and does not involve tail-to-tail dimerization as seen in SASH3-SAM (DaRosa et al., 2016; Kukuk et al., 2019; Mariotti et al., 2016). These results show that D663A/T664K alters the oligomerization state of SAM1, offering a potential starting point to further characterize the structural mechanism of oligomerization of the domain.

Nonetheless, these data must be observed in the context of study limitations. There is no published structure of SAM1, and all mutations were based on using a computationally predicted model of the structure aligned both structurally and sequentially with known SAM domains to derive the relative importance of residues in binding. A mutagenesis library would be encouraged to validate the proposed binding mechanism without structural data on the more folded oligomer. This study provides the first step in unraveling the molecular mechanism of SAM1 and how it may be a therapeutic target. The logical next steps would be to study any structural changes in SAM1 due to interaction with the C-terminal region of β-arrestin 1 and map the binding site on SAM1, as well as to interrogate if the R644W substitution seen in a pigmentation disorder has any effect on the binding affinity.

SAM domains have traditionally been portrayed as highly structured and rigid domains with set 4–5 helical structures, but SAM domains are known to be very diverse (Kim and Bowie, 2003; Vincenzi et al., 2020). For example, both SAM1 of CASKIN2 and the SAM of SAMHD1 have shown disordered characteristics in NMR spectroscopy and X-ray crystallography (Buzovetsky et al., 2018; Smirnova et al., 2016). In addition, CASKIN2 and AIDA-1 (ANS1B) have tandem SAM domains that bind with each other either inter- or intramolecularly (Kurabi et al., 2009; Smirnova et al., 2016). Interestingly, both contain one SAM domain that is highly structured and one that is poorly structured or unstable unless bound to the other SAM domain (Kurabi et al., 2009; Smirnova et al., 2016). SASH1-SAM2 does have an NMR structure deposited in the PDB but without a supplemental paper (PDB access number: 2DL0). This structure suggests that SAM2 of SASH1 is well expressed, well folded in solution, and monomeric while our data indicates that SAM1 has a major disordered state and a minor more structured state. If what was found in CASKIN2 and AIDA-1 applies, then SAM1 and SAM2 are likely involved in self association such as a self-regulation mechanism as in AIDA-1 or in complex formation as in CASKIN2. Further studies will be needed to fully address this.

In conclusion, the SAM1 domain of the scaffold protein SASH1 exists as a disordered monomer that samples more folded oligomerized states up to a tetramer (Fig. 9). This monomer has a much lower affinity for homo-oligomerization than most known SAM domains, opening the possibility for a self-regulatory role of SASH1 through a potential SAM2 interaction. Notably, SAM1 is also important for SASH1′s function via the interaction with other proteins such as β arrestin. This dynamic behavior introduces a novel potential drug target for regulating SASH1′s functions in tumor suppression, dermatological pathologies, and/or lung development. Further, the D663A/T664K substitutions, which can induce SAM homo-oligomerization, may provide a basis for designing peptides or small molecules to modulate these interactions as a therapeutic strategy.

Fig. 9. SAM1 domain of SASH1 does not follow the traditional portray of SAM domains.

SAM1 has an abundant disordered monomeric nature which has much lower affinity for a more folded oligomeric state. NMR data shows an increase in compactness of H5 which is involved the oligomerization interface of most other SAM domains. Possibly, SAM may sample a more compact intermediate state before oligomerization. The D663A/T664K mutation drastically shifts the monomer to a predominant oligomerized species and potentially induces canonical SAM homo-oligomerization.

4. Materials and methods

4.1. Protein expression and purification

The cDNA for expression of the WT human SAM1::6xHis domain (UniProt ID: O94885; residues 627–703), R644W, D663R/A685E, and D663A/T664K substituted variants were cloned into pMAL-c4x-1-H (RBS) vector and purchased from GenScript^®. The vector has a Tobacco Etch Virus nuclear-inclusion-a endopeptidase (TEV protease) cleavage site between the MBP tag and the protein construct.

All bacterial expression constructs were transformed into the E. coli strain BL21 (LEMO21-DE3). Expression was induced with 0.4 mM IPTG at an OD600 of 0.7 and grown overnight at 20 °C. Cells were harvested by centrifugation at 5,000x g for 20 min. For ¹⁵N- and/or ¹³C-labeled protein production, M9 minimal media was used inoculated with ¹⁵N-ammonium chloride and ¹³C-glucose to produce uniformly labeled protein.

For all the constructs, the cells were lysed using low-imidazole binding buffer (20 mM HEPES, 200 mM NaCl, 1 mM EGTA, 1 mM MgCl₂, 1 mM NaN₃, 20 mM imidazole, pH 7.3), disrupted by sonication and cleared by centrifugation at 30,900x g. The lysate was purified using a HisTrap FF column (Cytiva), where the protein was eluted using high-imidazole buffer (20 mM HEPES, 200 mM NaCl, 1 mM EGTA, 1 mM MgCl₂, 1 mM NaN₃, 200 mM imidazole, pH 7.3). The eluted protein was incubated with TEV-protease at 4 °C overnight and then dialyzed against the low-imidazole binding buffer. MBP was removed by running the protein over the HisTrap FF column for a second time. Further purification was done on a size-exclusion HiLoad 16/600 Superdex 75 pg (Cytiva) in NMR buffer (20 mM NaP, 100 mM NaCl, 1 mM DTT, 0.02 % NaN₃, pH 7.0).

4.2. Size exclusion – High-Performance liquid chromatography (SE-HPLC)

Analytical runs were carried out on an Agilent 1200 series liquid chromatography (Dwell volume = 1.05 mL). An isocratic gradient of aqueous NMR buffer (20 mM NaP, 100 mM NaCl, 1 mM DTT, 0.02 % NaN₃, pH 7.0) was used at a flow rate of 0.1 mL/min and room temperature with a detection wavelength of 280 nm. The runs were carried out on an AdvanceBio SEC column (4.6×300 mm ID, 2.7-μm pore size, 300 Å pore size) from Agilent Technologies. 10 μL injections of each component of the Millipore Sigma Protein Standard Mix 15–600 kDa (Thyroglobulin – 660 kDa, Gamma-globulin – 150 kDa, Albumin – 66 kDa, Ribonuclease A – 13.7 kDa, and pare-aminobenzoic acid – 0.14 kDa), HN-S CTD, and SAM1 were conducted in isolation or standard spikes with SAM1 in a 1:1 or 1:2 ratio by volume to confirm elution order and retention times.

4.3. Size exclusion chromatography – Multi-Angle light scattering (SEC-MALS)

Protein samples were filtered through a 0.22 um filter and spun down immediately prior to sample loading. Proteins were resolved by size using size exclusion chromatography with NMR buffer on a TSK-GEL G2000SW_XL column powered by a Shimadzu HPLC LC-20AD pump. Eluents were then monitored by a Wyatt Dawn EOS 18-angle light scattering detector combined with an in-line Wyatt Optilab DSP refractive index monitor.

4.4. NMR spectroscopy

For backbone resonance assignment of SAM1, ¹⁵N-¹H HSQC, HNCACB, and CBCA(co)NH spectra (Cavanagh, 2007) were recorded on a triple-resonance Varian 900 MHz NMR cryo-probe spectrometer. HNCO and HN(ca)CO (Cavanagh, 2007) spectra were recorded on a BRUKER 600 MHz spectrometer with cryoprobe. Measurements were conducted at 5 °C using a ¹³C/¹⁵N-labeled sample in 20 mM NaP and 100 mM NaCl, 1 mM DTT, and 0.02 % NaN₃, at pH 7.0 with 345 μM protein concentration in a standard 5 mm Shigemi tube. The 3D spectra were acquired with a non-uniform sampling (NUS) scheme generated by the NUS@HMS scheme generator software (Hyberts et al., 2012) with 1,024 complex data points in the direct dimension and 30 % sampling of the original 96 and 80 points in the indirect ¹³C and ¹⁵N dimension, respectively. The spectral widths were 16 ppm (¹H), 35 ppm (¹⁵N), 14 ppm (C=O), and 70 ppm (Cα/Cβ), the interscan delay 1.7 s, and the number of scans was 16 for all experiments. The NUS-acquired data were reconstructed using the hmsIST software (Hyberts et al., 2012). Zero-filling was achieved by the addition of 256 points in each indirect dimension. A solvent subtraction function was applied in the direct dimension.

We assessed residual secondary structure using the secondary structure propensity (SSP) score program developed by Forman-Kay and co-workers with the re-referencing algorithm for ¹³Cα and ¹³Cβ shifts (Marsh et al., 2006). The method combines different chemical shifts into a single residue-specific secondary structure propensity score. Our input shifts were those of ¹HN, ¹⁵N, ¹³CO, ¹³Cα, and ¹³Cβ. The same chemical shifts were used as input for CheSPI (Nielsen and Mulder, 2021), a more recent algorithm to analyze populations of secondary structure elements. In contrast to SSP, CheSPI can distinguish structured regions other than α-helix and β-sheet from true disorder due to its multi-parameter output. CheSPI analysis was performed using the Python script provided by Frans Mulder (https://github.com/proteinnmr/CheSPI). TALOS+ analysis was performed via exporting the assignment as a compatible shift list and running it on the published software, which is part of the NmrPipe package (Delaglio et al., 1995; Shen et al., 2009).

¹⁵N R₁, R_1ρ, and R₂ relaxation rate constants and ¹⁵N-¹H heteronuclear NOE enhancements of SAM1 were measured on a triple-resonance Varian 900 MHz NMR cryo-probe spectrometer at 5 °C. The sample was ¹⁵N-labeled with all other conditions the same as the ones used for the resonance assignment. The relaxation sampling time points were 300, 500, 700, and 1000 ms for R₁, 10, 30, 50, 70, 90, 110, 130, 150, and 220 ms for R_1ρ, and 10, 30, 50, 70, 90, 110, 130, 170, 210 and 230 ms for R₂. During the R_1ρ relaxation time, a ¹⁵N spin-lock field of 1500 Hz strength was applied. R₂ was calculated from R₁ and R_1ρ using R₂ = R_1ρ + (R_1ρ − R₁)*tan²(θ), where θ = tan⁻¹(2πΔν/γ_NB₁), Δ_ν is the resonance offset, |γ_NB₁|/2π is the strength of the spin-lock field B₁ and γ_N the gyromagnetic ratio of the ¹⁵N spin. ¹⁵N-¹H heteronuclear NOE enhancements were determined from two spectra recorded in the presence and absence of ¹H saturation in an interleaved manner. The saturation time applied on ¹H was 2 s.

The ¹⁵N-CEST experiment (Vallurupalli et al., 2012) was performed at 5 °C on VARIAN 900 MHz spectrometer with cryoprobe using the same sample as for the relaxation experiments. A weak ¹⁵N B₁ field of 5 Hz strength was applied during a 330 ms relaxation time. Ninety-two data sets were obtained corresponding to a chemical shift range of 103–133 ppm. In all experiments, ¹H decoupling was achieved using a 90_×240_y90_x composite pulse to reduce the decoupler side bands. Each 2D data set was recorded with 32 scans and 30 % NUS sampling of a total of 92 complex points in the indirect dimension. The recorded spectral widths are 16 ppm and 31 ppm in the ¹H and ¹⁵N dimensions, respectively. A global fit of the exchange rate k_ex was performed using the Python program CHEMEX (Vallurupalli et al., 2012).

¹⁵N- CPMG profiles (Millet et al., 2000) were measured at 5 °C on VARIAN 900 MHz spectrometer with cryoprobe using the same sample as for the other relaxation experiments. The 2D spectra were recorded as a pseudo-3D spectrum using 10 different inversion pulse frequencies ν_CPMG (100–1000 Hz). The relaxation delay used was set to 40 ms. A two-state exchange model was assumed depending on the populations of the states, the exchange rate between them, and the chemical difference between them. The parameters were fitted using Monte Carlo simulations with an in-house written Matlab script.

ZZ-exchange NMR experiments were measured at 5 °C on BRUKER 600 MHz spectrometer with cryoprobe (Mittermaier and Kay, 2009) using 100 ms mixing time. The 2D spectra were acquired with 1024 complex data points in the direct dimension and 128 points in the indirect ¹⁵N dimension. The spectral widths were 16 ppm (¹H), and 35 ppm (¹⁵N), the interscan delay was 1.9 s, and the number of scans was 32 for all experiments.

All spectra were processed and analyzed with NmrPipe (Delaglio et al., 1995), CcpNmr (Vranken et al., 2005), and SPARKY (Lee et al., 2015) software.

4.5. Simulation of relaxation rates

Simulations of ¹⁵N relaxation rates included the dipole–dipole interactions with H^N and H^α spins and ¹⁵N CSA and using the full Lipari-Szabo expressions for the spectral density functions at all relevant frequencies (Lipari and Szabo, 1982).

4.6. Circular dichroism (CD)

CD was used to monitor changes in secondary structure over a range of temperatures. CD spectra and the mean residue molar ellipticities of 0.5 mg/mL SAM1 were determined using a Jasco J0815 spectropolarimeter (Jasco Inc. Easton, MD, USA) at 5, 25, and 35 °C in the NMR buffer. SAM1 was loaded into a 0.1 cm quartz cell, and the ellipticity was scanned from 185 to 250 nm. Molar ellipticities were calculated using the equation m°*M/(10*L*C), where m° is the millidegree value, M is the molecular weight gm/mole, L is the path length in cm, and C is the concentration in gm/liter.

4.7. Sequence alignment

Sequence alignment was performed using Clustal_X (Larkin et al., 2007)(Larkin et al., 2007) and visualized using Jalview (Waterhouse et al., 2009)(Waterhouse et al., 2009). Clustal_X performed k-tuple distance calculation and progressive alignment. The three aligned SAM domain sequences have the Uniprot references: SAM1: O94885, Sly2-SAM: Q9NSI8, and SASH3-SAM: O75995.

Data availability statement

The ¹H, ¹³C, and ¹⁵N chemical shift assignments have been deposited in the BioMagResBank database (https://www.bmrb.wisc.edu) under the accession number 51486.

Data availability

Data will be made available on request.