Structural Elucidation of the Mesothelin–Mucin-16/CA125 Interaction

Summary:

Mesothelin is a cell-surface glycoprotein expressed at low levels on normal mesothelium but overexpressed on many cancers. Mesothelin has been implicated to play role/s in cell adhesion and multiple signaling pathways. Mucin-16/CA125 is an enormous cell-surface glycoprotein, also normally expressed on mesothelium and implicated in the progression and metastasis of several cancers, and directly binds Mesothelin. However, the precise biological function/s of Mesothelin and Mucin-16/CA125 remain mysterious. We report protein engineering and recombinant production, qualitative and quantitative binding studies, and a crystal structure determination, elucidating the molecular-level details governing recognition of Mesothelin by Mucin-16/CA125. The interface is small, consistent with the ~micromolar binding constant, and is free of glycan-mediated interactions. Sequence comparisons and modeling suggest that multiple Mucin-16/CA125 modules can interact with Mesothelin through comparable interactions, potentially generating a high degree of avidity at the cell surface to overcome the weak affinity, with implications for functioning and therapeutic interventions.

Graphical Abstract

eTOC blurb

Tumor antigens Mesothelin and Mucin-16/CA125, associated with cancer but with still mysterious physiological functions, natively bind avidly through multiple Mucin-16 SEA modules in the absence of glycan interactions, revealed by the crystal structure of a novel Mesothelin/SEA fusion protein determined by Rupert et al.

Introduction

Mesothelin (MSLN) is a 40 kDa glycoprotein expressed on the surface of mesothelial cells and overexpressed on several human cancers, including mesothelioma, ovarian, pancreatic adenocarcinoma, lung adenocarcinoma, and cholangiocarcinoma¹. MSLN is expressed as a ~70 kDa precursor protein that is cleaved by furin during export to release an N-terminal 31 kDa secreted fragment, called megakaryocyte-potentiating factor, leaving MSLN proper linked to the cell surface through a glycophosphatidylinositol linkage. While the three-dimensional structure of MSLN has been recently determined experimentally^2,3, and it has been established to bind Mucin-16 (Muc16), also known as carbohydrate antigen 125 (CA125)^4–7, the precise biological function of MSLN remains vague¹. Muc16, a type I transmembrane protein, is expressed in the mesothelium, upper respiratory tract epithelial cells, the cornea and conjunctiva, and ovarian and endometrial cells and has been implicated in the progression and metastasis of pancreatic and ovarian cancer. Muc16 (~22,000 residues in length) is the largest known member of the mucin family of glycoproteins (and the second largest protein in human genome) and serves as a biomarker for some cancers^8–11. Muc16 is organized structurally as a heavily O-glycosylated N-terminal domain of ~12,000 residues, an N-glycosylated domain of ≥60 tandem repeats of ~156 residues each, a membrane proximal domain of 16 tandem SEA (“sea urchin sperm protein, enterokinase and agrin”) modules of ~120 residues each, a transmembrane spanning segment, and a small (~32 residue) cytoplasmic domain. While the precise biological function of Muc16 is also unclear, Muc16 is believed to have a protective barrier function, lubricating and hydrating epithelium, and to block cellular adhesion and immunoreceptor synapse formation, mediated through the O-glycosylated domain. SEA modules are alternately proposed to function in regulating and binding carbohydrates¹². Like MSLN, the Muc16 extracellular domain (ECD) can be proteolytically cleaved at sites within the SEA block and shed from the cell surface. Both MSLN and Muc16 overexpression are associated with poorer cancer prognoses in patients.

Building on our prior studies of MSLN^2,13, we sought to detail the interaction/s between MSLN and Muc16 to further understanding of their biological function/s and development of therapeutic interventions. Two previous studies^4,7 had biochemically narrowed the interaction to the N-terminal region of MSLN proper (residues 296–359) and the 10^th and 11^th SEA modules of Muc16 (SEA10, SEA11). We confirmed binding of isolated SEA10 to MSLN directly (K_D = 0.9 μM), but the interaction of isolated SEA11 with MSLN, while apparent, was too weak to quantify (K_D > 10 μM). After several unsuccessful attempts, we determined the crystal structure of a novel MSLN–SEA10 fusion protein (d_min = 2.65Å), revealing the details of recognition. The interface is relatively small, consistent with the micromolar K_D¹⁴ (typical of cell surface protein–protein binding), and devoid of direct glycan interactions. SEA10 residues in contact with MSLN are highly conserved across multiple SEA modules, strongly suggesting that SEA2–12 would all be able to comparably bind MSLN, potentially generating a high degree of avidity for both cell surface and soluble forms of Muc16, affecting not only binding but, potentially, signaling mechanisms. The remaining SEA modules are less conserved and include substitutions that would be predicted to preclude MSLN binding.

Results

MSLN/Muc16 binding measurements:

To confirm prior findings, we performed quantitative SPR interaction analyses with isolated SEA10 and SEA11 modules produced recombinantly. SEA10 bound MSLN (produced as an ECD-Fc₂ construct to facilitate biosensor capture) with a reasonable K_D of 0.899(7) μM (Figures 1A and B), but SEA11, while clearly binding, did not bind strongly enough to permit SPR quantitation (estimated K_D >10 μM, Figure 1C), likely due to the substitution of a histidine for a proline (in SEA10) at position 13,739 in SEA11. Using a qualitative size exclusion chromatography (SEC) binding assay, monomeric, soluble SEA10 clearly formed stable complexes with monomeric, soluble MSLN ECD (Figure 1D) whereas monomeric, soluble SEA16, one of the most diverged SEA modules from SEA10 (32% identity), did not bind detectably (Figure 1E).

Figure 1:

MSLN/SEA interaction analyses. (A) Determination of the dissociation constant from (B) SPR equilibrium data of isolated SEA10 analyte binding to MSLN ligand captured as an Fc₂ fusion construct. [The MSLN/Muc16 K_D has been previously measured at ~5–10 nM, using cell-surface Muc16 and a MSLN ECD-Fc dimer construct⁶, reflecting a bivalent avidity consistent with the monovalent affinity reported here.] (C) SPR equilibrium data of isolated SEA11 analyte binding to MSLN ligand captured as an Fc₂ fusion construct (same analyte concentrations as in (A)). Quantitation was not performed because saturation was not achieved, though binding is apparent. (D) SEC analysis of isolated MSLN/SEA10 binding showing that this interaction is strong enough to form stable heterodimer complexes in solution. (E) SEC analysis showing that the MSLN/SEA16 interaction is too weak to form detectable complexes in solution. [For this experiment, SEA16 was produced as an Fc₂ fusion protein.] Peak elution volumes are consistent with the molecular weights of the species. (F) SEC analysis of the MSLN-SEA10 fusion protein used for crystallization showing a monomer/dimer mixture, likely in dynamic equilibrium, comparable to what is often observed with antibody single-chain Fv constructs (e.g.,¹⁷). Monomer/dimer equilibration obviously did not impede crystallization as is often also the case with antibody single-chain Fv constructs (e.g.,¹⁷). Elution volumes of M_r standards are indicated with green arrows.

MSLN/Muc16 complex crystallography:

To perform a crystallographic analysis, various complexes were screened for crystallizability, including a MSLN N-terminal fragment (MSLN^296−359)/SEA10 complex and a series of MSLN–SEA10 fusion constructs intended to overcome the fairly weak in-solution affinity and foster contacts conducive to crystallization. Fusion constructs varied the order of the moieties and linker length (N-MSLN^296−359–[19 residue linker]–SEA10-C, N-SEA10–[19 residue linker]–MSLN^296−359-C, N-SEA10–MSLN^296−359-C) with an N-MSLN^296−359–SEA10-C construct, presumed to force homodimer formation over intramolecular interactions, finally yielding diffraction-quality crystals (d_min = 2.65Å) [SEC analyses of this fusion protein (Figure 1F) show monomer/dimer equilibration.] Initial phases were determined by molecular replacement, yielding electron density maps clear enough to support model building and refinement with good final statistics (Table 1), permitting delineation of the MSLN/Muc16 interface (Figure 2, Supplementary Table 1).

Table 1:

Crystallographic Data Collection and Refinement Statistics

Data Collection
Space group	P212121
Cell dimensions (abc, Å)	58.2, 73.2, 75.1
Resolution range (Å)	50.00–2.65 (2.70–2.65)
Wavelength (Å)	1.0000
Unique reflections	9,868 (480)
Completeness (%)	99.0 (100)
Average Redundancy	5.3 (5.8)
Rmerge (%)	6.6 (62.8)
I/σ(I)	46.1 (3.7)
Structure refinement
Resolution (Å)	2.65
No. reflections	9,117
Rwork / Rfree	24.8 / 29.8
No. atoms (average B, Å2)
Protein	1,469 (69)
Others	75 (122)
r.m.s. deviations
Bond lengths (Å)	0.007
Bond angles (°)	75 (122)
Estimated Coordinate Error (Å)	0.250
Ramachandran values
Favored (%)	97.3
Allowed (%)	2.1
Outliers (%)	0.6
Molprobity score	99 (1.81)

Numbers in parentheses are for reflections in the highest resolution shell.

Figure 2:

MSLN/Muc16 SEA module binding interfaces. (A) Multiple sequence alignment of SEA modules 2–12 in human Muc16 shown as a sequence logo generated with WebLogo¹⁸. Residues making contacts with MSLN based on the SEA10 complex crystal structure are marked with arrows, red denoting side-chain contacts and orange denoting main-chain contacts. N-linked glycan sites occur at asparagines 8, 29, and 45 in the alignment. Views of (B) the MSLN residues contacting SEA10 and (C) the SEA10 residues contacting MSLN are highlighted from the crystal structure. Contacting residues are shown in a licorice stick rendering against a molecular surface representation, colored by electrostatic potential (blue: positive, red: negative), of the cognate molecule in the complex. Residues are numbered and colored by atom type: oxygen: red, nitrogen: dark blue, MSLN carbons: light blue, and SEA10 carbons: green. Images were generated with MacPyMOL¹⁹.

MSLN/Muc16 complex structure:

The asymmetric unit of the crystal contains a single MSLN–SEA10 fusion protein with each moiety folding independently (Figure 3). The peptide backbone at the junction between the two moieties is well-ordered as a 3₁₀ helix. The MSLN moiety makes crystal contacts to three adjacent symmetry mates through their SEA10 modules. One interface (Figures 2 and 3) is significantly larger (592 vs. <311 Ų), has a better calculated binding energy (−6.8 vs. >−2.6 kcal/mole), excludes all but one water, and is structurally reasonable, meeting criteria for assignment as the physiological interface among observed crystal contacts.

Figure 3:

MSLN/Muc16 SEA module crystal structure and comparisons to prior structures. (A) the crystal structure of the MSLN^296−359–SEA10 fusion protein is shown in a ribbon representation. Two molecules are shown from neighboring unit cells in the crystal with the physiologically relevant binding inface circled in red. Moieties in the fusion protein are colored as indicated, with the 3₁₀ helix at the fusion juncture indicated. N- (blue circles) and C-termini (red circles) are also indicated, when visible. N-glycan sites are numbered and two partially resolvable N-glycans are shown as licorice stick representations. (B) The MSLN^296−359 moiety from the fusion protein structure is shown superimposed on the crystal structure of the MSLN ECD (gray ribbon; from 8CXC.pdb²) with the positions of residues Y318, W321, and E324 highlighted in yellow and H354 marked with a magenta asterisk. Note the conservation between MSLN structures and the distance between highlighted sequence positions and the MSLN/SEA10 interface (yellow circle). The SEA10 moiety of a symmetry-related fusion protein is shown docked onto the MSLN^296−359 moiety as observed in the structure, with modeled N-glycan saccharides shown as orange spheres. N-termini of both are indicated with blue spheres and the C-terminus of the SEA10 moiety is indicated with a red sphere. Green arrows indicate where neighboring SEA modules would be positioned, showing the clearance available for the rest of Muc16. A red arrow marks the sole, glancing, non-polar contact between SEA10 (at D13,541) and the docked MSLN ECD (at M407), involving a flexible loop in MSLN which varies between MSLN isoforms². Fusion moieties are colored as indicated. Inset: A view focused on the MSLN/SEA10 binding site in the same orientation as in (B) is shown with the SEA10 moiety replaced with the amatuximab V_LV_H cassette from 8CXC.pdb² shown as a molecular surface colored in purples. Note the overlap between SEA10 and amatuximab binding sites. (C) Superposition of the experimentally determined and AlphaFold-Multimer²⁰ predicted structures of the MSLN/SEA10 complex are shown as ribbon representations, colored as indicated. Note that the entire MSLN ECD was used for AlphaFold docking, which identified only a single SEA10 binding site, though only the fragment corresponding to the MSLN^296−359 moiety in the crystallized fusion protein is shown. Images were generated with MacPyMOL¹⁹.

This interface (Figure 2, Supplementary Table 1) is largely hydrophobic and centers around a protruding proline residue (343) from MSLN that fits into a small hydrophobic pocket on SEA10). This pocket is formed from side-chains of V13,510, T13,506, T13,507, L13,490, and M13,496, and includes hydrogen bonds to the MSLN backbone from Muc16 K13,503 and Y13,584. Muc16 G13,513 and the side-chains from L13,514 and P13,583 provide additional hydrophobic contacts. Muc16 P13,583 reciprocally protrudes into a hydrophobic groove adjacent to MSLN P343 forming a proline “yin-yang” (☯) arrangement. Hydrophobic contacts from MSLN contributing to the interface include the side-chains of I316, F317, A341, T345, and Y346. Additional polar interactions contributing to binding include hydrogen bonds between MSLN Y346 and the backbones of Muc16 N13,489 and A13,543, and MSLN E347 and Muc16 N13,489, and a salt bridge between MSLN E313 and Muc16 R13,509. One water molecule is embedded in the interface, packing against the side-chain of MSLN Y346 and making a bridging main-chain hydrogen bond from MSLN F344 to Muc16 N13,489. No significant differences or conformational changes are observed within the MSLN moiety of the fusion protein when superimposed on MSLN ECD structures (Figure 3B).

The SEA10 sequence includes three potential N-glycan sites. Two of these (N13,484 and N13,505) showed clear electron density permitting partial modeling of N-glycans (Figure 3) while the third (N13,521) lacks any apparent non-protein density (Supplementary Figure 1). The modeled glycans do not interact with or block MSLN, and the third, 17 Å distant from the nearest MSLN residue, likely would not if glycosylated.

MSLN/Muc16 complex modeling:

Partly to assess the current state-of-the-art of protein structure prediction, which had been remarkably successful with the structure of the MSLN ECD on its own², the MSLN ECD/SEA10 complex was modeled with AlphaFold-Multimer¹⁵. A single binding site was predicted recapitulating all the salient details of the crystal structure to a remarkable degree (Figure 3C). The superposition RMSD on Cα atoms (59 from MSLN, 121 from SEA10) was 0.68 Å (!). Indeed, modeling even accurately predicted residue rotamer usage at the binding interface. Extending this analysis, AlphaFold was used to model MSLN complexes with all 16 SEA modules. AlphaFold predictions showed that SEA modules 2–12 would bind MSLN using the same interaction mode as observed in the MSLN–SEA10 complex structure, with comparable predicted binding scores (0.89–0.92). No other binding sites were predicted.

Discussion

We report protein engineering and recombinant production, qualitative and quantitative binding studies, and a crystal structure determination elucidating the molecular-level details governing recognition of MSLN by Muc16. Confirming prior studies, SEA10 (K_D = 0.9 μM) and SEA11 (K_D = >10 μM) were demonstrated to directly interact with MSLN, though the more divergent SEA16 module did not bind appreciably. The noticeable reduction in SEA11 affinity can be explained by the Pro>His substitution at position 13,739 in SEA11, which corresponds to one of the key “yin-yang” prolines, handily serving as a serendipitous “mutation” to further confirm our interface assignment (Supplementary Figure 2). Using the crystal structure as a guide, SEA module sequence comparisons (and AlphaFold modeling) strongly suggest that SEA2–12 should also bind comparably to MSLN, but the remaining SEA modules should not. SEA modules 2–12 have ≥79% pairwise sequence identity with marked conservation of MSLN-contacting residues (Figure 2A). The remaining SEA modules (1, 13–16) exhibit lower pairwise sequence identities with the predicted binding modules and have substitutions that would likely impede binding with MSLN, confirmed for SEA16.

Alanine mutations made in the MSLN sequence to narrow the Muc16 binding site⁷ in advance of the complex structure included Y318A (abolished Muc16 binding), W321A and E324 (decreased binding), and H354A (no effect on binding). However, none of these residues make contacts with SEA10 in the complex structure (Figure 3B), suggesting that any effects on binding were indirect, possibly through altered or disrupted folding. It is reasonable to infer effects on folding by these mutations, based on available MSLN structures. Y318 is buried, inaccessible to binding partners, overlaid by the side-chains of K306 and E322. The side-chain of W321, while accessible, packs intimately against the side-chains of E322 and K319. E324 salt-bridges to K320. Alanine substitutions at any of these three positions could reasonably affect folding. The side-chain of H354, while accessible and not involved in any interactions predicted to affect folding, does not approach any atom of the SEA10 moiety closer than 9Å, consistent with our assignment of the physiological interface. As had been previously noted⁷, binding of amatuximab, an antibody which binds the immunodominant N-terminal domain of MSLN^2,3,16, would be predicted to sterically block SEA module binding clearly, based on our structure (Figure 3B, inset), further supporting our interface identification.

A key finding of our study is that the combination of arranging a large number of SEA modules in tandem with a binding orientation (the middle of the SEA module binding the N-terminal end of MSLN as extended out from the module termini as possible, Figure 3B) likely enables a single Muc16 protein to simultaneously engage multiple MSLN molecules on a target cell surface. This binding mode generates both a high level of avidity, dramatically increasing the effective affinity of soluble Muc16 for MSLN expressing cells, and an efficient clustering/capping of MSLN on a cell surface. Both outcomes affect potential MSLN signaling mechanisms in response to Muc16 ligation. Our results also suggest that potential anti-Muc16 anti-tumor agents blocking this interaction would need to engage multiple SEA modules to be effective but provide a clear map for targeting both SEA modules and MSLN. Our results, however, are based on isolated SEA module interactions with MSLN and do not address potential interactions between other parts of Muc16 and MSLN or other cell-surface components, or altered contexts, such as tumor-expressed variants of MSLN or Muc16.

STAR Methods

RESOURCE AVAILABILITY

Lead contact:

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Roland K. Strong (rstrong@fredhtuch.org).

Materials availability:

Vectors generated in this study incorporate proprietary elements with restrictions on distribution. Coding inserts are available from the lead contact upon request.

Data and code availability:

The diffraction data and crystal structure of the Muc16/MSLN fusion protein has been deposited in the RCSB Protein Databank (PDB) under accession code 8VM1. This paper analyzes existing, publicly available data deposited in the RCSB PDB under accession codes 8CXC and 7SA9. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request. This paper does not report original code.

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Microbe strains:

E. coli DH5α (ZymoResearch T3009)

Cell lines:

HEK293T (ATCC CRL-3216, RRID:CVCL_0063)

FreeStyle 293-F (Thermo Fisher R79007, RRID:CVCL_D603)

METHOD DETAILS

Protein expression and purification:

Recombinant SEA and MSLN fusion proteins were produced using the Daedalus lentiviral transduction platform²¹. cDNAs encoding fusion constructs were codon optimized for human cells (Genscript), synthesized (Genscript), and subcloned into optimized lentiviral vectors²¹ (Supplementary Table 2). Lentiviruses were produced in HEK293T cells (ATCC CRL-3216, RRID:CVCL_0063) and used to transduce FreeStyle 293-F cells (Thermo Fisher R79007, RRID:CVCL_D603) in Freestyle media (Thermo Fisher 12338018). Isolated SEA modules and the MSLN-SEA10 fusion protein included Tobacco Etch Virus (TEV) protease-cleavable N-terminal Siderocalin (Scn) fusion partners to improve yields²¹ (protein sequences are detailed in Supplementary Table 3). An intact MSLN precursor-antibody Fc fusion protein (MSLN-Fc₂) was produced to facilitate sensor chip capture; SEA16 was also produced as an antibody Fc dimer. Proteins were purified by immobilized metal chelate affinity chromatography (IMAC, on Ni-NTA agarose, Qiagen 30210) and SEC (on Superdex 200 columns, Cytiva) after TEV cleavage. For IMAC, culture supernatants were filtered and brought to 300 mM NaCl, loaded onto HisTrap HP columns (Cytiva #17524701), washed with phosphate buffered saline (PBS (Gibco 14190–144) supplemented to 300 mM NaCl and 10 mM imidazole, and eluted with PBS (300 mM NaCl) plus 250 mM imidazole. Analytical SEC analyses were run in PBS. Proper folding and purity were confirmed by comparative reduced/non-reduced PAGE and analytical SEC analyses (e.g., Figure 1).

Surface Plasmon Resonance (SPR)

Surface Plasmon Resonance (SPR) experiments were performed at 25°C on a Biacore T100 instrument (Cytiva) using a running buffer of 10 mM HEPES, pH 7.4, 150 mM NaCl, 3 mM EDTA, 0.05% Tween-20 with 0.1 mg/mL bovine serum albumin (added after immobilization). Goat anti-human IgG, Fcγ fragment specific antibody (Jackson ImmunoResearch 109-005-098) at 50 μg/mL in 10 mM sodium acetate, pH 5.5 was amine coupled to 2 flow cells of a Series S CM4 chip (~2600 response units (RUs)). MSLN-Fc₂ was captured on one flow cell of amine-coupled anti-human IgG Fcγ (~68 RUs). Serial two-fold dilutions of uncleaved Muc16 SEA10 and SEA11 were prepared in running buffer at a concentration range of 40 μM down to 36 nM along with a corresponding PBS buffer volume matched dilution series. Muc16 was injected at 50 μL/min over both the captured MSLN-Fc₂ and anti-human IgG Fcγ alone (reference) surfaces for two minutes with three minutes of dissociation, going from low to high concentration with each concentration preceded by its buffer matched control. Regeneration was accomplished with buffer flow alone.Double-referenced data were analyzed in Scrubber 2.0c software (BioLogic Software) using the closest blank for referencing and a steady-state 1:1 interaction model.

Crystallography:

Crystals of the MSLN-SEA10 fusion were obtained by vapor diffusion, with drops of protein at ~8 mg/ml in 25 mM PIPES (pH=7.2), 150 mM NaCl, 1 mM EDTA mixed with a well solution of 100 mM Bis-Tris (pH=6.5), 250 mM NH₄Cl, 12% w/w polyethylene glycol (M_r = 3,350). Crystals were cryopreserved in well solution plus 15% v/v glycerol, and data were collected at the Advanced Light Source (Berkeley, CA), beamline 5.0.2, and processed with HKL-2000²² (Table 1). Initial phases were determined by molecular replacement using PHASER²³ with coordinate set 8CXC.pdb (for MSLN) and 7SA9.pdb (for SEA10) as search models^2,5. Density maps were clear enough to permit partially building two N-glycans (Supplementary Figure 1). Iterative rounds of alternating positional refinement and model building, using the programs REFMAC²⁴ and COOT²⁵, including placing ordered solvent molecules, were followed by a final round of TLS refinement²⁶. Residues or side-chains that did not exhibit 2F_obs-F_calc electron density when contoured at 0.7σ were removed or truncated to the Cβ atom. Polder electron density maps²⁷ were generated with PHENIX²⁸. The physiologically-relevant interface was identified with PISA analyses²⁹. The quality of the final model was assessed using Molprobity³⁰.

QUANTIFICATION AND STATISTICAL ANALYSIS

Crystallographic data collection and refinement statistics are shown in Table 1. No other statistical analyses or software were used.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Oligonucleotides
See Supplementary Table 2
Antibodies
Goat anti-human IgG, Fcγ fragment specific antibody	Jackson ImmunoResearch	109-005-098
Chemicals, peptides, and recombinant proteins
FreeStyle 293 Expression Medium	Thermo Fisher	12338018
phosphate buffered saline (PBS)	Gibco	14190-144
Ni-NTA agarose	Qiagen	30210
Deposited data
Human Mesothelin ectodomain crystal structure	Reference2	PDB: 8CXC
Human MUC16 SEA5 crystal structure	Reference5	PDB: 7SA9
Human MSLN296−359-SEA10 fusion protein crystal structure	This work.	PDB: 8VM1
Experimental models: Cell lines
HEK293T	ATCC	ATCC CRL-3216 RRID:CVCL 0063
FreeStyle 293-F	Thermo Fisher	Thermo Fisher R79007 RRID:CVCL D603
Software
Scrubber 2.0c	BioLogic Software	http://www.biologic.com.au/scrubber.html
MacPyMOL	Reference19	https://pymol.org
PHASER	Reference23	https://www-structmed.cimr.cam.ac.uk/phaser_obsolete/
REFMAC	Reference24	https://www2.mrc-lmb.cam.ac.uk/groups/murshudov/content/refmac/refmac.html
COOT	Reference25	https://www2.mrc-lmb.cam.ac.uk/personal/pemsley/coot/
PHENIX	Reference28	https://phenix-online.org
PISA	Reference29	https://bio.tools/pisa
Molprobity	Reference30	http://molprobity.biochem.duke.edu

Highlights.

• Crystallography reveals the details of the Mesothelin/Mucin-16 SEA10 interaction.

• Mesothelin/SEA10 binding is not mediated by glycan interactions.

• Multiple SEA modules likely bind comparably to Mesothelin generating high avidity.