Introduction

Isthmin-1 (ISM1) is a secreted protein initially identified in Xenopus embryos isthmus organizer1. Despite broad tissue expression in mammals2, the precise functions of ISM1 remain to be fully determined. ISM1 has many diverse reported activities3,4,5,6,7. Recently, ISM1 was identified as an adipokine with insulin-like properties that mitigates diabetes via promoting adipocyte and skeletal muscle cell glucose uptake8,9,10,11. This led us to investigate the structural characteristics of this previously uncharacterized protein, aiming to elucidate how its structure may inform our understanding of its role in metabolic signaling.

ISM1 is a 60 kDa protein with a predicted disordered N-terminal region, followed by a thrombospondin type I repeat (TSR) domain, and an AMOP domain (adhesion-associated domain in MUC4 (Mucin-4) and other proteins). AMOP domains have yet to be structurally characterized. The AMOP domain consists of approximately 160 amino acids and is found in putative cell adhesion molecules including SUSD2 (Sushi domain-containing protein 2) and MUC412. In MUC4, the AMOP domain has been shown to promote pancreatic cancer angiogenesis and metastasis13. In ISM1, the AMOP is necessary and sufficient for ISM1’s reported signal inhibitory, anti-angiogenic and apoptotic functions3,14,15. Despite being a critical functional domain, the structural characteristics of the AMOP domain remain largely unexplored.

Here we show the X-ray crystal structure of ISM1, including its AMOP domain. ISM1’s AMOP domain comprises a distinct structural fold. Notably, the central core of the AMOP fold shares similarity to bacterial streptavidin. The AMOP core is conserved when compared to other predicted AMOP domains structures. However, the surface loops and termini among different AMOP domains greatly vary. Thus, the AMOP core fold allows for structural plasticity that may provide divergent interaction epitopes and different functions, among AMOP domain containing proteins. To investigate the functional importance of ISM1’s AMOP domain features we attempted structure-function analysis using the reported ISM1 responsive pre-adipocyte 3T3-F442A cell line8,9. However, contrary to these reports, we show that purified ISM1 does not demonstrate activity in this cell line. Rather, AKT phosphorylation in 3T3-F442A cells could be attributed to copurified contaminants from the Expi293 expression system.

Results

Overall structure of ISM1

Despite being highly conserved and playing important roles in angiogenesis, apoptosis, and metabolism, ISM1’s mechanism of action remains unclear. Therefore, we sought to determine the structure of ISM1—particularly its uncharacterized AMOP domain – as a first step to understand its function at a molecular resolution. We generated various constructs of ISM1 for biophysical and biochemical study (Fig. 1a). Crystals were obtained from the TL-AMOP (TSR Linker-AMOP, residues G218 – Y464). Selenomethionine-substituted protein was used for experimental phasing. The crystallographic statistics for Selenomethionine-substituted crystals were summarized in Table 1. The structure of TL-AMOP was determined to a resolution of 3.5 Å (Table 1, Fig. 1b, Supplementary Fig. 1). Within the asymmetric unit, two TL-AMOP protomers were identified. While the AMOP domains of the two protomers are symmetric, the TSR domains do not maintain the same symmetry. The TSRs have different orientations with respect to the AMOP, indicating that the TSR and AMOP do not form a rigid complex and are independent domains. Additionally, the linker region (residues D263 – N285) between TSR and AMOP domains was not visible (Fig. 1b), likely due to its high flexibility.

Fig. 1: Crystal structure of ISM1.
figure 1

a Constructs of ISM1 in the present study. TSR and AMOP domains, and the linker (L) between them were labeled. ISM1-FL is the full-length ISM1 with the amino acid sequence from Ala30 to the end. The TL-AMOP starts at Glu218, L-AMOP starts at Gly263, and AMOP starts at Ala286. b Two protomers in the asymmetric unit of TL-AMOP crystal structure are shown in cartoon with TSR in orange, one AMOP domain in green, and the other AMOP domain in dark green. The two C-mannosylated tryptophans on the TSR domain and the disulfide bonds are displayed in sticks. The linker between the TSR and AMOP domains that could not be traced in the crystal structure is represented by dash line. c The TSR domain is comprised of three anti-parallel strands stapled by three conserved disulfide bonds which are shown in stick representation. The two C-mannosylated tryptophans and the three conserved arginine residues around the tryptophan are also displayed as sticks.

Full size image
Table 1 Crystallization conditions, data collection and refinement statistics
Full size table

Structure of ISM1’s TSR

The TSR domain is a highly conserved scaffold characterized by six cysteine residues that form three disulfide bonds, securing the antiparallel strands at both ends of the domain16. Our structure confirms the presence of the expected antiparallel strand folding and disulfide bond pattern (Fig. 1c). Within the TSR domain, there is a conserved and distinctive WSXW motif, where C-mannosylation on the two Trp residues was found crucial for proper protein folding and secretion17. Our structure confirmed C-mannosylation on both Trp223 and Trp226 of ISM1 (Fig. 1c). These Trp residues are sandwiched by three Arg residues through hydrophobic interactions (Fig. 1c), similar to observations in other TSR repeats18.

Structure of ISM1’s AMOP domain

The AMOP domain has a distinct fold. Overall, the ISM1 AMOP domain has 5 central antiparallel β-strands, surrounded by multiple helices. Disulfide bonds restrain both the N- and C- termini to this core, and additional disulfides stabilize the central strands together. In the orientation shown in Fig. 2a, the N-terminal and C-terminal helices are positioned on opposite sides of the central strands, and the middle helical region sits atop. The AMOP fold can be schematized as a hand holding chopsticks: the central barrel-curved strands act as the fingers, the terminal helices are the chopsticks, and the middle helical region is the thumb (Fig. 2b). This representation is functionally analogous as well: the central strands project loops (fingers) from behind the N- and C- terminal helices (chopsticks) to support them. Furthermore, the middle helices (thumb) stabilize the top portion of the terminal helices (chopsticks).

Fig. 2: Structure of the AMOP domain from ISM1.
figure 2

a Crystal structure of ISM1’s AMOP domain has three helical regions and a central strand core (green). The separate helix regions include the N-terminal helices (blue), the “thumb” (orange), and the C-terminal helix (red). The central stranded (green) projects four fingers. b Schematic of a hand holding chopsticks resembling features of the AMOP fold. c The central stranded core of the AMOP domain (orientation from perspective shown in a). d Three-dimensional alignment between the AMOP domain core and a streptavidin monomer (pink, PDB id: 1STP).

Full size image

A structural similarity search using the Dali server19 shows that the AMOP central strands overlap with a streptavidin monomer, forming a similar antiparallel β-barrel with similar topology – with a Dali Z-score of 3.3 (Fig. 2c, d). Superimposition of the main chain Cαs of the AMOP central core strands and the streptavidin β-barrel yields an R.M.S.D. of 2.95 Å for 188 atoms. The AMOP domain lacks the first and last two β strands found in streptavidin that complete the barrel. This portion of the AMOP barrel is stabilized by non-β-strand loops (Fig. 2c). The end of ISM1’s N-term helix sits at the same position as biotin occupies when in complex with streptavidin (Supplementary Fig. 2). Thus, AMOP’s central strands form approximately half of a β-barrel with structural similarity to streptavidin.

ISM1’s RKD sequence

ISM1 has been reported to bind to integrins through an RGD-like interaction20. RGD containing proteins harbor the tripeptide sequence Arginine-Glycine-Aspartate that is essential for binding to integrins21. While ISM1 does not have a canonical RGD motif, it does have a similar tripeptide sequence “RKD”, where the middle Glycine is a Lysine residue. This RKD sequence was proposed to act similar to an RGD motif and thought to direct ISM1 binding to integrins15,20. Our structure reveals that ISM1’s RKD peptide maps to the second beta strand of the AMOP domain. Notably, the RKD sequence does not fall on a flexible loop region, typical of an integrin binding RGD motif22 (Supplementary Fig. 3a). When the RKD of ISM1’s AMOP domain is directly aligned to an RGD peptide bound to integrin, significant clashes occur between the AMOP domain and the integrin receptor – indicating that the RKD of ISM1 cannot function as an integrin binding RGD motif (Supplementary Fig. 3b). In line with this, Alphafold 323 models of ISM1 binding to integrins (αVβ515,20 or α8β124) do not predict an interface involving the RKD peptide (Supplementary Fig. 4). Instead, these models converge on a different binding site within the isolated β-propeller of the integrin α subunit, one that does not involve ISM1’s RKD or the RGD binding pocket of the integrin (Supplementary Fig. 4a, c). Notably, the best ipTM scores with the isolated propeller are below 0.3, indicating a low-confidence prediction. However, when AlphaFold is used to model the interaction between ISM1 and the integrin heterodimer (αβ) extracellular region, ISM1’s AMOP again converges to the same alternative site, with improved ipTM scores exceeding 0.7 (Supplementary Fig. 4b, d). These AlphaFold models, therefore, align with our crystal structure analysis, supporting the conclusion that ISM1’s RKD likely does not function as a canonical integrin-binding RGD motif.

AMOP domain comparison

To gain further structural insight into the larger family of AMOP domains, we utilized Alphafold 225 to model the AMOP domains of ISM2, SUSD2, and MUC4. The resulting models have overall pLDDT scores over 90 except for the MUC4 AMOP which has an overall pLDDT score of 78 (Supplementary Fig. 5a–c). Comparison of the models reveals that all AMOP domains share the conserved structural core of strands (fingers), middle helices (thumb), and disulfide bonds (Fig. 3). Among AMOP domains, all the disulfide bonds (C1-C8, C3-C6, C4-C9, and C5-C7) are strictly conserved in sequence and 3-dimensional position (Fig. 3, Supplementary Fig. 5d–g). Distinctly, ISM1 and ISM2 have an additional unpaired cysteine residue (Cys303 in ISM1’s AMOP domain). In our crystal structure this cysteine is on a surface loop and does not have any obvious additional electron density (Fig. 3a). Therefore, our crystalized ISM1 is not post translationally modified at Cys303. However, the presence of a near surface cysteine in ISM1 and ISM2 may play a role in its function or regulation.

Fig. 3: Structural comparison of AMOP domains.
figure 3

a Crystal structure of ISM1’s AMOP domain as in Fig. 2. b AlphaFold2 predicted structure of ISM2’s AMOP domain shows almost identical 3-D structure to ISM1. Both have the unpaired cysteine (C2) and maintain a C-terminal helix. c AlphaFold2 predicted structure of SUSD2’s AMOP domain, and d AlphaFold2 predicted structure of MUC4’s AMOP domain. The AMOP domains from SUSD2 and MUC4 both lack helices on the sides of the core and have a less prominent finger 2.

Full size image

Other than the conserved disulfide bonds, the AMOP domain structures of ISM1 and ISM2 are visibly different than the predicted structures from SUSD2 and MUC4. Most obvious, the prominent parallel terminal helices (chopsticks) found in ISM1 and ISM2 are absent. Instead, a non-helical strand replaces the N-terminal helix, and the C-terminal helix is entirely absent – with the domain ending at the last disulfide. Consistent with a support role, the loop regions that stabilize the terminal helices in ISM1 are also less prominent in SUSD2 and MUC4. Interestingly, the surface lacking the C-terminal loop in SUSD2 and MUC4 is predicted to make intramolecular interactions with other domains from their Alphafold2 predictions. Thus, the structural variability that the core of the AMOP domain can accommodate may give rise to different functions through distinct epitopes.

Oligomeric state of ISM1

As previously mentioned, ISM1 packs as a dimer in the crystallographic asymmetric unit. One dimerization interface is mediated by the AMOP domain’s C-terminal helix (Figs. 1b, 4a). The C-terminal helix is specific to the AMOP domain from ISM1 and ISM2 (Fig. 3). Hydrophobic interactions involving residues Y454, I455, and F458 on the two symmetric helices contribute to the dimerization interface (Fig. 4a). Crystal packing also shows an additional dimer interface directed by the TSR domain, involving the main chain of A247-T255 (Supplementary Fig. 6). In this case, the two TSR domains are symmetric while the AMOP domains of the protomers are not (Supplementary Fig. 6).

Fig. 4: The ISM1 AMOP domain is monomeric in solution.
figure 4

a ISM1 forms a dimer in the asymmetric unit of the crystal. The dimerization is mediated by the C-terminal helix. Residues Y454, I455, F458 and Q459 contributed to the interaction are in stick representation. b SEC profiles of L-AMOP and the double mutant I455E & F458E (termed as L-AMOP-IFEE) indicate that both proteins eluted at similar volumes. c Guinier plots for L-AMOP and L-AMOP-IEFE. The mass concentration normalized natural logarithm of the scattering intensity (ln(I(q)) is plotted against q2. The normalized residues against q2 were shown below. The weight-averaged molecular mass (see Methods) of molecules were displayed and colored consistent with the Guinier plot dots.

Full size image

To better understand the potential dimerization of the AMOP domain directed by the C-terminal helix, we used a construct lacking the TSR for biophysical study (L-AMOP). We introduced mutations in the hydrophobic residues involved in the crystal dimer (I455 and F458), replacing them with acidic residue (glutamic acid), resulting in a mutant (IFEE). We compared the size exclusion chromatography (SEC) profiles of histidine-tagged L-AMOP-IFEE to L-AMOP. Both proteins eluted at the similar volumes (Fig. 4b). To further investigate the dimerization potential of the AMOP domain at high concentrations, we performed small-angle X-ray scattering (SAXS). The scattering, Kratky and P(r) plots of the his-tagged L-AMOP and IFEE mutant are shown in Supplementary Fig. 7. The Guinier plots obtained from SAXS analysis indicate that both proteins were monodisperse in solution at a concentration of 3 mg/ml (~120 µM) (Fig. 4c). The calculated masses of both proteins were below 30 kDa (Fig. 4c, Table 2, and Supplementary Table 1), corresponding to a monomeric state. These results strongly suggest that ISM1’s AMOP domain is likely to be monomeric under physiological conditions.

Table 2 Summary of SAXS data
Full size table

ISM1 does not stimulate AKT phosphorylation in preadipocytes

The 3T3-F442A preadipocyte cell line has previously been shown to respond to stimulation by IMAC (immobilized metal affinity chromatography) purified ISM18. We therefore sought to carry out structure-function studies with this cell line following the same protocol for expression and purification. We expressed various ISM1 constructs from Expi293 cells and examined their effect on AKT phosphorylation in 3T3-F442A. In line with prior studies, our IMAC purified AMOP constructs all displayed a concentration-dependent stimulation of AKT phosphorylation at S473 in 3T3-F442A cells (Fig. 5a and Supplementary Fig. 8a, b). However, IMAC purified AMOP domains expressed in insect cells showed no activity (Supplementary Fig. 8b, c). To rule out the possibility that the lack of activity of the insect-expressed proteins was due to incorrect folding, we evaluated their binding capabilities to a home-screened antibody fragment (Fab). Both insect- and mammalian-expressed proteins exhibited similar binding affinities to the Fab (Supplementary Fig. 8d), indicating that the proteins expressed from insect cells fold comparably to those expressed in mammalian cells.

Fig. 5: ISM1 does not activate AKT phosphorylation in 3T3-F442A cells.
figure 5

a IMAC purified TL-AMOP expressed from Expi293 cells shows stimulation of AKT phosphorylation at S473 in a dose-dependent manner in the preadipocyte 3T3-F442A cell line. Phosphorylation of endogenous AKT at S473 (pAKT473) and total AKT were assessed by immunoblotting and quantified. A representative blot is shown and quantification of three independent biological repeats are presented as mean ± s.d. b Activity of IMAC plus SEC purified TL-AMOP is diminished. A representative blot is shown and quantification of three independent biological repeats are presented as mean ± s.d. c IMAC pulldown of media from Expi293 cells either expressing ISM1-FL protein or not expressing recombinant protein (Expi293med) as a negative control. Both ISM1-FL and the negative control exhibited activity on 3T3-F442A AKT phosphorylation in a dose-dependent manner. A representative blot is shown and quantification of two independent biological repeats are presented as mean ± s.d. d SEC profiles of ISM1-FL (black line) and Expi293med (gray dash line) are shown. Fractions were assessed by SDS-PAGE and combined as frac1 (10.4–11.4 mL) and frac2 (15.9–16.9 mL). e The activity of the fractions of ISM1-FL and Expi293med on 3T3-F442A AKT phosphorylation indicates pure ISM1-FL protein (ISM1-FL F1) has limited activity but F2 of both ISM1-FL and Expi293med have potent activity. A representative blot is shown and quantification of four independent biological repeats are presented as mean ± s.d. f The activity of twin-strep tagged ISM1-FL purified by Strep-Tactin resin (ISM1-FLts) compared to ISM1-FL purified by IMAC on 3T3-F442A AKT phosphorylation. A representative blot is shown and quantification of three independent biological repeats are presented as mean ± s.d. Source data are provided as a Source Data file.

Full size image

To rule out the possibility that a co-purifying contaminant from Expi293 expression is responsible for the observed activity, we further purified our ISM1 constructs. After IMAC, we additionally purified our ISM1 constructs with SEC. The SEC purified TL-AMOP demonstrated a significantly diminished ability to stimulate AKT phosphorylation in 3T3-F442A cells (Fig. 5b). Whereas the activity of the smaller L-AMOP was retained, but it required higher concentrations for maximal stimulation (Supplementary Fig. 8e). These findings are consistent with an active contaminant that co-elutes with the smaller L-AMOP construct but is separated from the larger TL-AMOP.

To investigate whether the observed 3T3-F442A AKT activation was indeed due to an active contaminant, we compared full length ISM1 (ISM1-FL) purified from Expi293 cells to Expi293 conditioned media (that was not transiently transfected and does not express any ISM1) but went through the same purification procedure. Similar to the shorter constructs, ISM1-FL displayed activity after IMAC but prior to SEC (Fig. 5c). Importantly, IMAC of Expi293 conditioned media also exhibited potent activity stimulating pAKT without any ISM1 present (Fig. 5c). After SEC, the fractions with pure ISM1-FL (Fraction 1) lost most of its activity. ISM1-FL is the largest construct at ~65 kDa but runs even larger on SEC due to its non-globular N-terminal region and overall linear shape. However, fractions corresponding to the size of ~25–30 kDa (similar to the smaller L-AMOP construct on SEC) from both ISM1-FL and Expi293 conditioned media (Fraction 2) robustly activated AKT phosphorylation (Fig. 5d, e). These findings are consistent with an active contaminant related to the expression system that co-purifies with His-tagged ISM1 through IMAC.

To further demonstrate that the active contaminant co-purifies through IMAC specifically, we generated an ISM1-FL with a twin-strep tag. We purified twin-strep tagged ISM1-FL (ISM1-FLts) protein with Strep-Tactin resin and tested its activity on AKT phosphorylation. Purified ISM1-FLts could not stimulate AKT phosphorylation in 3T3-F442A cells (Fig. 5f). Taken together, these results indicate that ISM1 is not the component responsible for activating AKT phosphorylation in 3T3-F442A preadipocytes. Rather, contaminants co-purified through IMAC and eluting between 25 and 30 kDa from SEC are responsible for ISM1’s reported activity.

To identify the contaminants responsible for the AKT activity we analyzed the fractions of Expi293 cell cultured media by mass spectrometry (MS). Among the various components detected, PDGFA and GDF15 emerged as primary hits in fraction 2, while absent in fraction 1 (Fig. 6a). Both PDGFA and GDF15 function as dimeric proteins with a molecular size at 25–30 kDa. We re-analyzed the MS data of the IMAC purified recombinant ISM1 in Jiang et al. study8, and identified both of these proteins. Additionally, through western blot analysis, we validated the presence of PDGF-A and GDF15 exclusively in fraction 2 of Expi293 cell-cultured media (Fig. 6b, c). We then tested to see if PDGF-AA could stimulate AKT phosphorylation in 3T3-F442A cells. At concentrations below 1 nM, recombinant PDGF-AA robustly stimulated phosphorylation of AKT (Fig. 6d). These results collectively demonstrate that contaminants, rather than ISM1, exhibit activity on pre-adipocytes.

Fig. 6: Expi293 derived PDGFA and GDF15 copurify with ISM1 through IMAC.
figure 6

a Heat map of the secreted proteins identified by mass spectrometry. The two columns are fraction 1 (F1) and fraction 2 (F2) of Expi293 conditioned media as in Fig. 5d, respectively. Peptides of PDGFA and GDF15 in fraction 2 are highlighted. b 1 µg total protein of fractions 1 (F1) and 2 (F2) of the Expi293 cell-cultured media were loaded for WB detection of GDF15. A representative blot of three independent biological repeats is shown. c 10 µg total protein of fractions 1 (F1) and 2 (F2) of the Expi293 cell-cultured media were loaded for WB detection of PDGFA. A representative blot of three independent biological repeats is shown. d Both fraction 2 (F2) and recombinant purified PDGFAA show potent activity on AKT phosphorylation in 3T3-F442A cells. Data are mean ± s.d. of three independent measurements. A representative blot is shown and quantification of three independent biological repeats are presented as mean ± s.d. Source data are provided as a Source Data file.

Full size image

Discussion

As a conserved secreted protein with multiple reported functions, understanding the structural basis of ISM1’s function is important. We successfully determined the crystal structure of ISM1, including its structurally distinct AMOP domain. The AMOP domain has core strands forming approximately half of a β-barrel. Loops and helices around the core vary among AMOP domains, suggesting that the core of the AMOP domain is the basis for building up distinct epitopes for different functions.

The RKD peptide on the second β-strand has been reported to be important for integrin αvβ5 binding and angiogenesis inhibition20. However, our docking demonstrates that in its modeled position it would be incompatible with integrin binding. RKD containing peptides derived from ISM1 also target cell-surface GRP78 and induced apoptosis15. While our structural analysis is consistent with the RKD β-hairpin functioning as a binding epitope, further investigations are needed to elucidate its specific binding partners. Notably, and in line with our docking, a recent in silico AlphaFold screen to identify potential receptors for ISM1 did not identify integrins26.

ISM1 was recently identified as an adipokine that regulates glucose uptake and insulin sensitivity in adipose tissues and skeletal muscle cells8,9. In our study, we used this activity assay (AKT phosphorylation in the 3T3-F442A preadipocyte cells) for ISM1 structure-function analysis. Surprisingly, we found that pure ISM1 alone was unable to stimulate AKT phosphorylation. Instead, we find that a contaminant or contaminants present in the Expi293 cell cultured media are responsible for 3T3-F442A AKT phosphorylation. The active contaminants bind to the IMAC resin used to purify His-tagged ISM1 and elute from SEC at a range of 25–30 kDa. Thus, ISM1 that is not purified subsequent to IMAC retains these active contaminants. Additionally, truncations of ISM1 purified by SEC that fall within the 25–30 kDa size range also retain the active contaminant(s) as it overlaps with the contaminant’s elution profile. Whereas ISM1 that is purified by a different resin (Strep-Tactin) or purified from insect cells does not retain the contaminant responsible for activity on 3T3-F442A cells. Together this indicates that 3T3-F442 cells do not respond to ISM1 via AKT signaling.

Using mass spectrometry, we identified PDGFA and GDF15 among the proteins that co-purify through IMAC and elute at ~25 kDa on SEC (fractions active on 3T3-F442 cells). Importantly, both PDGFA and GDF15 have well established roles in metabolism27,28. In addition to stimulating 3T3-F442 cells, IMAC-purified ISM1 has been reported to exhibit a distinct dual metabolic function, regulating glucose uptake while suppressing lipid accumulation when administered pharmacologically in vivo8. Notably, GDF15 has been described to have similar dual roles, increasing insulin sensitivity and decreasing liver lipogenesis, thereby preventing hepatic steatosis29. The functional overlap between that proposed for ISM1 and co-purifying contaminants GDF15 and PDFGA must be considered when studying the metabolic effects of ISM1. Consequently, future studies must carefully distinguish the impacts of ISM1 from those of PDGFA, GDF15, and other potential signaling molecules.

In conclusion, ISM1 is a pleiotropic factor with diverse reported functions (e.g., craniofacial patterning, hematopoiesis, apoptosis, lung homeostasis, metabolism, and renal branching) as demonstrated by knockout and knockdown experiments5,6,8,24,30,31. Our findings reveal that ISM1’s AMOP domain has a distinct structure with similarity to bacterial streptavidin. The core half beta-barrel serves as a platform capable of displaying various epitopes – well suited to different functions for AMOP domains. Our structure does not capture the N-terminal region of ISM1 (residues 31–216), which lacks predicted domains and is suggested to be intrinsically disordered by AlphaFold. While this region may not adopt a defined structure in isolation, it could play an important but yet unidentified role in ISM1’s function. Further investigations are required to elucidate ISM1’s specific binding partners and mechanism of action. Additionally, it is crucial to resolve potential confounding effects from co-purifying contaminants to better establish the physiological and pathophysiological roles of ISM1.

Methods

Recombinant protein expression and purification

Codon-optimized cDNAs for the human ISM1 were synthesized (IDT). The sequence was incorporated into a mammalian cell expression plasmid or insect cell expression plasmid followed wnt3 signal peptide (MEPHLLGLLLGLLLGGTRVLAG) and an octa-histidine tag linked with a Factor Xa cut site (GHHHHHHHHGSSTSNGTIEGRSLD). For the twin-strep tagged ISM1, an additional twin-strep tag (SRGSAWSHPQFEKGGGSGGGSGGSAWSHPQFEK) was added at the C-terminus. ISM1 truncations were conducted by Phusion Site-Directed Mutagenesis Kit (Thermo Fisher Scientific). Mutation of I455E and F458E was introduced by site directed mutagenesis using the QuickChange Kit (Agilent Technologies).

ISM1 TL-AMOP protein for crystallography study was expressed from a stable 293 T cell line and purified from the FreeStyle 293 Expression Medium (Thermo Fisher Scientific, cat # 12338026). The selenium-methionine substituted protein expression was referred to Barton et al.32 and Li et al.33. In brief, the cells were washed with PBS before supplied with Expi293 methionine-free medium (Thermo Fisher Scientific, cat# A4096701) and starved for 8 h. Then, seleno-L-methionine (Sigma-Aldrich, cat# S3132) dissolved in PBS at a stock concentration of 30 g/L was added to the cell culture to a final concentration of 100 mg/L. Three days after seleno-L-methionine incorporation, the media were collected for selenium-methionine substituted protein purification. The other proteins were purified from the Expi293 media of transient transfected Expi293F cells or EX-CELL 405 media (Sigma-Aldrich, cat# 14405 C) of infected Hi5 cells. Media were harvested 3–4 days post-infection for Expi293F cells and 2–3 days post-infection for Hi5 cells and flowed directly over Ni-Penta™ Agarose-Base Resins (Marvelgent Biosciences, cat# 11-0227) by gravity. The resins were then washed with 15 mM Imidazole (pH 8.0), 100 mM NaCl, and bound proteins were eluted with 300 mM Imidazole (pH 8.0), 100 mM NaCl. For further purification by SEC, eluted proteins were loaded onto a HiLoad 26/600 superdex 200 pg column (Cytiva Life Sciences, cat# 28989336) equilibrated in 20 mM HEPES (pH 7.5) with 100 mM NaCl or Superdex 200 Increase 10/300 GL size exclusion column (Cytiva Life Sciences, cat# 28990944) equilibrated in the same buffer. Fractions were evaluated by SDS-PAGE then concentrated in a Centricon 10 kDa spin concentrator (MilliporeSigma, cat# UFC901024 and UFC801024) and saved at 4 °C.

Crystallography

Crystals of ISM1 TL-AMOP were grown at 16 °C by mixing equal volumes of protein (3–6 mg/mL) and reservoir solution (0.9–1.0 M NaMalonate (pH 7.0), using hanging-drop method. Crystals were cryoprotected in 2 M NaMalonate (pH 7.0) then shot at the Advanced Photon Source (APS) beamline NE-CAT. Four diffraction datasets from selenium-methionine substituted protein were combined for single-wavelength anomalous diffraction (SAD) analysis to get intial phases. NCS averaging generated phases capable of chain tracing. However, the resolution did not allow unambiguous fitting. After the Alphafold predicted protein structure database was released25, we were able to identify specific domains of ISM1 TL-AMOP, then solved the structure by fitting the separate domains using Phaser, and by building the missing loops in Phenix software34. We then iteratively fit and refined with Phenix software34.

AlphaFold models

The AlphaFold model of ISM2 (Uniprot ID: Q6H9L7) AMOP was downloaded from the AlphaFold Protein Structure Database with the identifier AF-Q6H9L7-F1-v4, of which residues D400 – Y571 were used for analysis. The AlphaFold model of SUSD2 (Uniprot ID: Q9UGT4) AMOP was downloaded from the AlphaFold Protein Structure Database with the identifier AF- Q9UGT4 -F1-v4, of which residues P285 – Y441 were used for analysis. The AlphaFold model of MUC4 (Uniprot ID: Q99102) AMOP was generated using the AlphaFold3 Server with residues P4555 – Y4676.

Structural alignment

The alignment between ISM1’s AMOP and streptavidin used the command super in PyMol (2.5.4) to superimpose the main chain of the β-strands of AMOP (residues A49 – D64 and R80 – N104, 449 atoms) and streptavidin β-barrel (residues A38 – D61 and T71 – V97, 443 atoms). To analyze if ISM1’s AMOP can mimic an RGD containing proteins to bind to integrin via RKD residues, the crystal structure of integrin in complex with fibronectin, a typical RGD containing protein ligand of integrin, 7NWL, were selected for analyzing. The RKD residues of ISM1 were aligned to the RGD residues of fibronectin in the ligand groove of integrin using default align command in PyMol (2.5.4). The structural comparison of AMOP domains also used the default align command in PyMol (2.5.4). To show the pLDDT of AlphaFold predicted models in PyMol, command prompts: ‘run https://raw.githubusercontent.com/cbalbin-bio/pymol-color-alphafold/master/coloraf.py’ and ‘coloraf (object_name)/sele/enable’ were used.

Small-angle X-ray scattering (SAXS)

SAXS data were recorded at 4 °C on a Rigaku BioSAXS-2000nano 2D Kratky block camera system with a Rigaku 007HF rotating anode source and a Rigaku HyPix-3000 HPAD CCD detector, with 90-min exposures using SAXLab version 4.0.2 (Rigaku). The concentration for L-AMOP WT and IEFE was 3 mg/ml (~120 µM) in 20 mM HEPES pH 7.5, 100 mM NaCl. Data were reduced, and matched buffer was subtracted using BioXTAS RAW version 2.1.135 to yield the corrected scattering profile in which intensity (I) is plotted as a function of q (q  =  4πsinθ/λ, where 2θ is the scattering angle). Extrapolation of the Guinier fit to the y axis intercept results in the estimation of I(0), which is proportional to the weight-averaged molecular mass of molecules in a SAXS sample: ln I(q) = ln I(0) – (Rg2/3)q2, where Rg is the radius of gyration. Details are summarized in the Supplementary Table 1 following the SAXS analysis guidelines36.

Cell stimulation assay

3T3-F442A cells (Sigma-Aldrich, cat# 00070654-1VL) was cultured in DMEM (Gibco, cat# 12491023) supplied with 10% New Calf Serum (Sigma-Aldrich, cat# N4637-500ML), 2 mM L-Glutamine (Gibco, cat# A2916801) and 1% penicillin/streptomycin (Sigma Aldrich, cat# P4333-100ML) in 6-well plates until reaching 80% confluence after which cells were serum starved with DMEM/F12 (1:1) medium (Gibco, cat# 11320082) overnight. The starved cells were stimulated with ISM1 proteins at 37 °C for 5 min.

Western blot

After stimulation, cells were immediately kept on ice and washed with ice-cold PBS twice. Cell lysis buffer (20 mM Tris-HCl pH 8.0, 150 mM, NaCl 0.5% NP-40) supplied with phosphatase and protease inhibitor cocktail (Cell Signaling Technology, cat# 5872S) was added to the wells and cells were scrapped and transferred into ice-cold EP tubes and lysed on ice for 30 min. After the lysis, the solutions were centrifugated at 21,130 × g for 15 min and the supernatants were collected. The concentrations of the total proteins were measured by Bradford assay (Thermo Fisher Scientific, cat# 23236).

For detecting the AKT and pAKT, 50 μg total proteins were loaded on 4–20% gradient gel for SDS-PAGE. BLUEstain 2 protein ladder (GoldBio, cat# P008-500) was used as the molecular weight ladder. For detecting the PDGF and GDF15, 5 ug (for PDGF detection) or 1 µg (for GDF15 detection) proteins of fraction 1 and 2 of Expi293 cell culture media were loaded for SDS-PAGE. After the protein electrophoresis, proteins were transferred to PVDF membrane (EMD Millipore, cat# IPVH00010) in the transfer buffer (25 mM Tris, 192 mM glycine, 20% methonal (v/v)) using 90 V constant voltage for 45 min. The membranes were then blocked with 5% non-fat milk (ChemCruz, cat# H1823) in TBST buffer (20 mM Tris-HCl pH8.0, 150 mM NaCl, 0.1% Tween 20) for 1 h at room temperature followed by incubating with the primary antibodies 1 h at room temperature. The dilutions for the primary antibodies were: AKT (Cell Signaling Technology cat# 4691S), 1:1000; p-AKT473 (Cell Signaling Technology cat# 4060S), 1:1000; PDGF-AA (R&D Systems cat# MAB2211-SP, 1:50); GDF15 (Cell Signaling Technology cat# 79996S, 1:1000). Then, the horseradish peroxidase (HRP) conjugated goat anti-rabbit IgG antibody (Cell Signaling Technology) was diluted 1:5000 and incubated with the membranes at room temperature for 1 h. The membranes were developed with Amersham ECL prime western blotting detection reagent (Cytiva Life Sciences, cat# RPN2236) and chemiluminescence scanned using ChemiDoc (Bio-Rad). For PDGFAA detection, the SuperSignal™ West Atto Ultimate Sensitivity Substrate kit (Thermo Fisher Scientific, cat# A38554) was utilized. The bands were quantified using ImageJ 1.51 h. For an example of presentation of full scan blots, see the section of Source Data of Blots in the Supplementary Information File.

Phage-displayed Fragment Antigen-binding (Fab) Selection

This was performed essentially as described before37,38. In brief, the Fab library (library F)37 was added to Nunc MaxiSorp 96-well immunoplates (Thermo Fisher Scientific, cat# 430341) coated with streptavidin and incubated at room temperature for 2 h to eliminate non-specific phage binders or unwanted epitope binders. The phage solution was then transferred to the other plates coated with human ISM1 protein and incubated at room temperature for 2 h. The plates were rinsed with phosphate-buffered saline (PBS) (pH 7.4) with 0.05% tween-20 and eluted with 100 mM HCl. The eluted phages were neutralized with 1 M Tris (pH 11), then propagated for the next round of selection until five rounds of selection. For rounds three, four, and five, the phages were diluted and infected bacteria omnimax and sprayed on LB plates supplied with carbenicillin. Colonies were picked and cultured. The media containing phages were directly analyzed by enzyme-linked immunosorbent assay (ELISA). For the ELISA, Nunc MaxiSorp 384-well immunoplates (Thermo Fisher Scientific, cat# 464718) were first coated with ISM1, streptavidin as a negative control, and anti-Flag M2 antibody (Sigma-Aldrich) as a positive control overnight at 4 °C. The media were then diluted and split into the protein-coated wells. Colonies that specifically bind to ISM1 over the negative control were subjected to DNA sequencing analysis. Unique colonies were cloned into RH2.2 p441 vector for Fab expression39. The Fabs were purified by protein A agarose resin (Goldbio) and eluted in 50 mM NaH2PO4, 100 mM H3PO4, 140 mM NaCl, pH 2.8. Then the eluents were neutralized by adding 1 M Tris (pH 11) to a final concentration of 100 mM. The Fabs were further purified by SEC with a Superdex 200 Increase 10/300 GL column equilibrated in 20 mM HEPES (pH 7.5) with 100 mM NaCl.

Bio-layer interferometry

Using an Octet RED96e instrument (Sartorius, Cat# Octet RED96e), penta-His sensors (Sartorius, 18–5120) were first loaded with Fab to a response unit of about 1 nm followed by quenching in 10 ng/mL biocytin (Sigma-Aldrich, cat# B4261) for 120 s and equilibrated in the working buffer (20 mM HEPES (pH 7.5), 100 mM NaCl) for 30 s. The sensors were then dipped into indicated concentrations of ISM1 proteins for 60 s, then dissociated in the working buffer for 60 s. Equilibrium (plateau) response was plotted against concentration (GraphPad Prism 10.0.2).

Mass spectrometry

Two replicates of gel samples were prepared for mass spectrometry. Twenty microgram of fraction samples F1 and F2 of Expi293 conditioned media after SEC were loaded on 4–20% gradient SDS-PAGE and ran for 5 min at 120 V. F1 is the control for F2. Bands were cut out followed by destaining with 30% acetonitrile and 100% methanol. Reduction and alkylation were carried out using 10 mM dithiothreitol (DTT) for 1 h at 56 °C, followed by 20 mM iodoacetamide (IAA) in darkness for 45 min at room temperature. The samples were then digested with trypsin (Promega) at a concentration of 10 ng/μL overnight at 37 °C. The purification of the digested peptides was performed using a C18 column (MacroSpin Columns, NEST Group INC). One microgram of the peptides was utilized for total proteome analysis.

The samples were measured by data-independent acquisition (DIA) MS method as described previously40,41,42, on an Orbitrap Fusion Tribrid mass spectrometer (Thermo Scientific) coupled to a nanoelectrospray ion source (NanoFlex, Thermo Scientific) and an EASY-nLC 1200 system (Thermo Scientific, San Jose, CA). A 150-min gradient was used for the data acquisition at the flow rate at 300 nL/min with the column temperature controlled at 60 °C using a column oven (PRSO-V1, Sonation GmbH, Biberach, Germany). The DIA-MS method consisted of one MS1 scan and 33 MS2 scans of variable isolated windows with 1 m/z overlapping between windows. The MS1 scan range was 350–1650 m/z and the MS1 resolution was 120,000 at m/z 200. The MS1 full scan AGC target value was set to be 2E6 and the maximum injection time was 100 ms. The MS2 resolution was set to 30,000 at m/z 200 with the MS2 scan range 200–1800 m/z and the normalized HCD collision energy was 28%. The MS2 AGC was set to be 1.5E6 and the maximum injection time was 50 ms. The default peptide charge state was set to 2. Both MS1 and MS2 spectra were recorded in profile mode. DIA-MS data analysis was performed using Spectronaut v1843,44,45 with directDIA algorithm by searching against the SwissProt downloaded human fasta file (UP000005640). The oxidation at methionine was set as variable modification, whereas carbamidomethylation at cysteine was set as fixed modification. Both peptide and protein FDR cutoffs (Qvalue) were controlled below 1% and the resulting quantitative data matrix were exported from Spectronaut. All the other settings in Spectronaut were kept as Default. The mass spectrometry raw data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository46 with the dataset identifier PXD062353.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.