key: cord-0287416-galk1eez
authors: Gandhi, Amit K.; Sun, Zhen-Yu J.; Kim, Walter M.; Huang, Yu-Hwa; Kondo, Yasuyuki; Bonsor, Daniel A.; Sundberg, Eric J.; Wagner, Gerhard; Kuchroo, Vijay K.; Petsko, Gregory A.; Blumberg, Richard S.
title: Structural basis of the dynamic human CEACAM1 monomer-dimer equilibrium
date: 2020-07-15
journal: bioRxiv
DOI: 10.1101/2020.07.14.199711
sha: ed24a34f9da92cf701031cf35c4c91092926fbb7
doc_id: 287416
cord_uid: galk1eez

Human (h) carcinoembryonic antigen-related cell adhesion molecule 1 (CEACAM1) function depends upon IgV-mediated homodimerization or heterodimerization with host ligands, including hCEACAM5 and hTIM-3, and a variety of microbial pathogens. However, there is little structural information available on how hCEACAM1 transitions between monomeric and dimeric states which in the latter case is critical for initiating hCEACAM1 activities. We therefore mutated residues within the hCEACAM1 IgV GFCC’ face including V39, I91, N97 and E99 and examined hCEACAM1 IgV monomer-homodimer exchange using differential scanning fluorimetry, multi-angle light scattering, X-ray crystallography and/or nuclear magnetic resonance. From these studies, we describe hCEACAM1 homodimeric, monomeric and transition states at atomic resolution and its conformational behavior in solution through NMR assignment of the wildtype (WT) hCEACAM1 IgV dimer and N97A monomer. These studies reveal the flexibility of the GFCC’ face and its important role in governing the formation of hCEACAM1 dimers and potentially heterodimers.

Carcinoembryonic antigen-related cell adhesion molecule 1 (CEACAM1), also referred 45 to as cluster of differentiation 66a (CD66a) and biliary glycoprotein (BGP), is a member 46 of the carcinoembryonic antigen cell adhesion molecule (CEACAM) family of glycosylated immunoglobulin (Ig) molecules (Kammerer & Zimmermann, 2010) . 48

Expressed on the surface of several cell types, CEACAM1 has been demonstrated to 49 play critical roles in morphogenesis (Huang et al., 1999) , apoptosis (Nittka et al., 2004) , 50

angiogenesis (Ergun et al., 2000) , cell proliferation (Singer et al., 2010) cell motility 51 (Ebrahimnejad et al., 2004) , fibrosis (Satoh et al., 2017) , and most recently as an 52

immunoreceptor important in mediating immune T cell tolerance (Huang et al., 2015) . 53

Human CEACAM1 (hCEACAM1) is a single pass type I transmembrane protein 54 expressed as 12 alternatively spliced isoforms that all contain an N-terminal V set fold of 55 the immunoglobulin superfamily (IgV) ectodomain followed by up to three type 2 56 constant immunoglobulin (IgC2) ectodomains (A1, B, A2), a transmembrane sequence, 57 and a signaling cytoplasmic domain. Depending on splice variation, the cytoplasmic 58 domain either includes a long (L) sequence inclusive of two immunoreceptor tyrosine-59 based inhibitory motifs (ITIMs) or a short (S) domain devoid of ITIMs (Beauchemin et 60 al., 1999 ) that impart intracellular inhibitory or non-inhibitory signals, respectively. 61 Despite the requirement of competing with CEACAM1 as a ligand, several 120 bacteria, fungi, and a virus that include E. coli (Korotkova et al., 2008) , Neisseria sp. 121 (Watt et al., 1997) , Moraxella catarrhalis (Conners et al., 2008) formation; however, there were significant differences in the structural and biophysical 134 features that distinguished hCEACAM1 IgV-HopQ heterodimerization from hCEACAM1 135 homodimerization (Bonsor et al., 2018) . One specific discriminating feature involves 136 residue N97, which has been reported to nearly abrogate hCEACAM1 137 homodimerization (K D ~ 1 mM) but does not significantly affect HopQ binding (Bonsor et 138 al., 2018) . This observation raises the question about the role of specific residues at the 139 GFCC' face in determining hCEACAM1 homophilic and heterophilic interactions, but 140 just as importantly emphasizes the need to decipher the underlying structural and Results 157

In 158 order to probe the role of the GFCC' surface in determining the monomer-dimer 159 equilibrium, we introduced alanine substitutions at residues V39, I91, N97, and E99, 160 which have been described to be important for CEACAM1 IgV homodimerization and 161 identified as naturally occurring single nucleotide polymorphisms (SNPs) sites 162

[rs772794650 (I91M), rs1335884800 (N97T), and rs142826356 (E99G)]. We first 163 expressed and purified WT and site-specific mutant hCEACAM1 IgV proteins using our 164 published protocols (Huang et al., 2016) and measured variations in their respective 165 thermal denaturation temperature (T M ), reflective of their stability by differential scanning 166 fluorimetry (DSF) (Figure 1A) . We observed a single T M for each protein at 25 M, 167

suggestive of a single step denaturation event despite whether the protein was 168 expected to be a hCEACAM1 IgV monomer (N97A) or dimer (WT). There was also a 169 correlation of decreasing T M with decreasing homodimerization affinity of the different 170 hCEACAM1 mutants, suggesting that weakening of hCEACAM1 IgV homodimerization 171 destabilizes hCEACAM1 IgV stability. However, the hCEACAM1 IgV N97A variant that 172 has been reported to be monomeric, with a K D approaching 1 mM (Bonsor et al., 2018) , 173 exhibited a similar melting temperature (54.09 o C) compared to WT protein (55.09 o C), 174 suggesting a unique stabilizing property of an alanine at that position and/or promotion 175 of a monomeric state. Next, we assayed the solution characteristics of each 176 (b) molecule present in the crystal asymmetric unit), Q89-Y34, S32-N97, Y34-N97, and 220 Q44-N97 were disrupted ( Figure 2B, Figure 2-figure supplement 2B) . 221 

Further, the hydrophobic interactions between two F29 residues were weaker 256 due to conformation changes and increased distance compared to WT (Figure 2B , 257 Figures 2C-D) , mutation of glutamic acid at position 99 to alanine abrogated side chain 262 to main-chain backbone interactions between E99-G41 of both hCEACAM1 molecules 263 as expected, but also abrogated the hydrogen bonded interactions between Q89-Y34, 264 N97-Y34, and Q89-N97 residues of two hCEACAM1 molecules ( Figure 2D, Figure 2 -265 figure supplement 3B). In addition, the distance between two hydrophobic V39 266 residues was slightly higher as measured by a distance of 3.9 Å between V39-β 267 carbons compared to distance of 3.7 Å in the WT ( Figure 2D The first dimer resembled a WT GFCC' face dimer with minor conformational changes 280 (Figure 3-figure supplement 5A ). More interestingly, the second GFCC' face dimer 281 comprising molecules (c) and (d) showed significant conformational differences across 282 various strands and loops (Figure 3-figure supplement 4A) , as shown by increased 283 distances between the interacting CC' loops (10 Å between β carbon of V39 residue 284 compared to distance of 3.7 Å between β carbon of V39 in the WT) (Figure 3-figure  285 supplement 4A). Further, we observed significantly weaker and a decreased network 286 of hydrogen bonded interactions within the FG and CC' loop residues in the weak V39A 287 dimer ( Figure 3B, Figure 3-figure supplement 4B ) compared to hCEACAM1 WT 288 GFCC' face dimer (PDB code 4QXW) (Figure 3-figure supplements 6A-6B) . 289

Specifically, we observed abrogation of all hydrogen bonded interactions mediated by 290 residues S32, G41, L95, and E99 of (c) molecule and residues S32, Y34, G41, Q89, dimer was observed in the resolved crystals from the two additional data sets (Not 306 reported). Interestingly for both of the V39A dimers, a significant metal ion density at the 307 4.0 σ level in a 2Fo-Fc electron density map was observed, which we refined for Nickel 308 (Ni ++ ). Here, we observed hexa-coordination of Ni ++ by residues H105 and V39 from 309 three neighboring hCEACAM1 molecules in the unit cell and crystallographic symmetry 310 mates (Figure 3-figure supplements 5B-C) . Overall previous study that examined the dynamic oligomer to monomer shift of hCEACAM1 by 457 inside out signaling and hinted at the strong propensity of monomeric CEACAM1 to 458 form oligomers (Patel et al., 2013) . Even though we observed poor NMR signals of 459 some of the resonance peaks, we were able to complete 90% of the backbone amide 460 resonance assignments for the N97A mutant (Figure 4-figure supplement 7B) . A 461 15 N/ 13 C/ 2 H triple-labeled N97A mutant hCEACAM1 protein sample was produced to 462 acquire additional NMR data sets for backbone assignments. However, per-deuteration 463 of protein that reduced NMR relaxations and increased NMR signals for the WT 464 hCEACAM1 made little improvement in the N97A mutant assignments (the reason for 465 which will be discussed below). 466

The overall secondary structures of the N97A mutant as predicted by the 467 assigned NMR chemical shifts were similar to those of the WT protein (Figure 4-figure  468 supplement 8B), and also consistent with secondary structures observed in the crystal 469 structure of the N97A mutant. The largest backbone amide NMR chemical shift changes 470 for the N97A mutant relative to the hCEACAM1 WT were found among residues in or 471 near the GFCC' face, the homodimer interface for the WT protein ( Figure 4B, Figure 4 populations of second conformation peaks in 15 N-HSQC NMR spectra; however, the 490 peak intensities were too weak to confirm with our NMR sample conditions. 491

The X-ray crystal structure of the N97A mutant revealed a very similar structural 492 fold compared to the hCEACAM1 WT protein. Although our NMR data largely support 493 the notion that the overall structural fold of N97A in solution is very similar to the X-ray 494 structure model, there was a region of ambiguity around the FG loop. A stretch of nine 495 residues including three residues from the end of the F strand, five residues in the FG 496 loop and one residue at the beginning of the G strand could not be assigned using the 497 conventional NMR sequential connectivity method. On the other hand, there were at 498 least seven NMR peaks in the 15 N-HSQC spectra with weak or moderate intensities that 499 remained unassigned because of weak NMR cross-correlation peaks. It should be 500 noted that a two-residue fragment was tentatively assigned to K92 and S93 in the FG 501 loop based on it containing the only missing serine residue. It is unlikely that this is due 502 to protein aggregation because there were no improvements when using a triple labeled 503 protein sample that was deuterated to reduce NMR relaxation for large proteins. One 504 explanation is that the FG loop of the N97A monomer was undergoing intermediate rate 505

exchange of multiple conformational states, likely related to the dynamic monomer-506 dimer equilibrium and/or additional conformational changes. 507

In addition, 15 N NMR relaxation studies of hCEACAM1 N97A were carried out by 508 measuring the longitudinal T 1 , transverse T 2 (CPMG) and T 1 relaxation times (Figure 5-509   figure supplement 10A) . The increased R 2 /R 1 ratios (higher local  C times) for several 510 residues in the C, C', C" strands, and potentially the FG loop region, were initially 511 interpreted as reflecting local dynamic motions. However, the differences between the 512 transverse relaxation rates R 2 (CPMG) and R 1 showed the same pattern, and clearly 513 confirmed its origin to be from NMR chemical shift exchange processes (Figure 5-514   figure supplement 10B) . This is because these residues exhibited relatively large 15 N 515 chemical shift differences, such that the exchange rate from the monomer-dimer 516 equilibrium (at 300M N97A) falls within the intermediate time scale (on the order of 517 several milliseconds). These results underscore the significant effects that the N97A 518 mutation has on disrupting the extremely stable dimeric form of WT hCEACAM1, and 519 which strongly shifts the dynamic monomer-dimer exchange to a monomeric form. 520 hCEACAM1-HopQ complex (PDB code 6AW2), respectively, significant conformation 541 differences were observed in the FG loops (Figure 6-figure supplements 11A, B) . 542

Next, we determined the crystallographic Debye-Waller factor (temperature factor or B 543 factor) of the hCEACAM1 WT homodimer (PDB code 4QXW) and GFCC' face variant 544 crystal structures described above, expect for the I91A structure, which was determined 545 at too low a resolution ( Table 1) . The B factor can be used to estimate the thermal 546 motion or dynamic mobility and disorder of each atom from the spread of the electron 547 density about the atomic position (Ringe and Petsko, 1985) . We observed an overall B 548 factor of 23 Å 2 for a hCEACAM1 WT dimer (PDB code 4QXW) with similar B factor 549 values observed across the GFCC' face ( Figure 1A) where increased B factor was associated with reduced 553 thermal stability. Although the crystal structure of the N97A mutant showed the lowest B 554 factor value of 16 Å 2 , a closer examination revealed higher B factor profiles of the CC', 555 EF and FG loop residues (Figure 6A) . 556 Consistent with this hypothesis, our NMR studies of the N97A mutant revealed that 596 monomer-dimer exchange involved residues within the GFCC' face including V39, Y48, 597 Q89 and A100. 598

It is therefore interesting that we also observed minor interactions between two 599 N97A monomers present in the crystal lattice through the ABED face. Further, the PDB 600 PISA studies revealed that the interfacial interaction involving the ABED loops was 601 associated with a low complexity score, raising the possibility that this face could serve 602 as an additional minor surface that facilitates CEACAM1 oligomerization. Notably an 603 ABED-mediated homodimerization interface has also been suggested in SPR binding 604 studies (Klaile et al., 2009 ) and described in another crystal structure of unglycosylated 605 hCEACAM1 WT IgV (PDB code 2GK2) (Fedarovich et al., 2006) . Although the ABED 606 surface contains three sites for carbohydrate modification, an attractive possible 607 contribution of the ABED face is to serve as a secondary homo-oligomerization site that 608 achieves relevance following trans GFCC'-initiated homo or heterodimerization by 609

propagating surface CEACAM1 clustering and downstream signal activation. 610

Another contribution of this study was our ability to fully assign the residues 611 within the NMR spectra of WT hCEACAM1 to 100% completion and provide a 612 comprehensive NMR assessment (~90% assignment) of the N97A mutant monomer. were estimated to be dimeric, did not show newly emerged peaks (Figure 5A) . In 626 addition, although some of the unassigned seven 15 NH resonance peaks appeared to 627 have stronger intensities at 16 M compared to higher protein concentrations, they did 628 not show significantly larger chemical shift changes with increasing N97A protein 629 concentrations. Perhaps these cases are more complicated and possibly involve 630 additional conformational changes besides the monomer-dimer equilibrium for these FG 631 loop residues at the GFCC' face. The higher conformational flexibility and B factor of FG 632 loops residues as described before for WT (PDB code 4QXW, 2GK2) and the 633 hCEACAM1-HopQ complex structure (PDB code 6AW2) (Figure 6A between WT and N97A mutant protein were in the same region as the N97A 15 NH 640 peaks that exhibited the largest shifting caused by changes in N97A monomer-to-dimer 641 equilibrium (Figure 5B) , the magnitude and direction of the 15 N-HSQC spectral changes 642 do not match exactly. It also does not appear that the 15 N-HSQC spectrum of N97A 643 mutant at even higher concentrations (i.e. sample with close to 100% dimer form) will 644 replicate that of the hCEACAM1 WT IgV. Therefore, it is likely that the dimer form of 645 N97A as shown in equilibrium with the monomer could be a transition state dimer akin 646

to what was observed for the V39A mutant. The local structural malleability of FG loop 647 residues is also well supported by thermal motion analysis of the N97A crystal structure 648 wherein we observed higher a B-factor at these locations. Thus, these NMR results 649 support the unique role of the N97 residue in determining the hCEACAM1 monomer-650 dimer equilibrium and the impact of the V39 hydrophobic interactions on hCEACAM1 651

Further, our high resolution structural and NMR studies are consistent with a 653 model wherein the GFCC' loops of hCEACAM1 represent the primary face involved in 654 homodimer formation (Figure 7) . When this GFCC' face is abrogated, possibly first 655 through disruption of CC' loop interactions (as observed in the V39A weak dimer GFCC' 656 face crystal structure) or when hCEACAM1 is a monomer (as observed in the N97A 657 crystal structure) in the cis state, the higher thermal motion and dynamic conformation 658 state especially associated with the CC' and FG loops contributes to monomeric 659 hCEACAM1 homophilic behavior or facilitates heterophilic interactions with its other 660 ligands in cis or trans through the GFCC' face of hCEACAM1 (Figure 7) . The 661 importance of the GFCC' face is also highlighted by the high degree of genetic 662 polymorphisms that exist there, as previously described (Huang et al., 2015) , 663

suggesting that a genetic propensity and/or other factors as discussed below may be 664 involved in regulating these processes. In the event of homophillic interactions by 665 monomeric hCEACAM1 with a neighboring hCEACAM1 monomer, the hCEACAM1 666 GFCC'-mediated homodimer is formed and becomes more thermally stable (Figure 7) . 667

The GFCC' face stabilized homodimer could subsequently participate in higher order 668 oligomer formation, which we observed in our NMR studies, possibly through minor 669 interactions mediated though the ABED face as observed in our N97A crystal structure 670 or another crystal structure of hCEACAM1 (PDB code 2GK2) (Fedarovich et al., 2006) . 671

This would enable interactions between CEACAM1 cytoplasmic tails that facilitate 672

Src-homology domain containing phosphatases. Indeed, many functional studies by 674 others have also demonstrated that the propensity of CEACAM1 to form higher order 675 (Figure 7-figure supplement 12) . hTIM-3 in particular is an (Figure 7-figure supplement 12) . These results are consistent with the 742 importance of amino acid residues such as E62 and D120 within the hTIM-3 GFCC' 743 face that determine binding to hCEACAM1 as defined by site-directed mutagenesis 744 (Huang et al., 2015) . The hTIM-3 bound crystal structure of the M6903 anti-TIM3 745 monoclonal antibody that blocks hCEACAM1-hTIM-3 interactions revealed the antibody 746

Fab binds with TIM-3 residues E62 and D120 at atomic level resolution (Zhang et al., 747 2020) . Similarly, hydrogen deuterium exchange mass spectrometry (HDxMS) studies 748 revealed that the anti-hTIM-3 monoclonal antibody 2E2 binds the GFCC' face of hTIM-3 749 with blockade of hCEACAM1 binding in transfected cells based upon flow cytometry 750 (Sabatos -Peyton et al., 2018) . Importantly, amino acid residues such N42, R43, Q44, 751 G47, and Q89 of hCEACAM1 that mediate hTIM-3 binding as defined by site-directed 752 mutagenesis studies are also located within the GFCC' face and involved in the 753 formation of hCEACAM1 homodimers (Figure 7-figure supplement 12) . This 754 highlights the strong competition between formation of high affinity homodimers and 755 lower affinity heterodimers and the critical need to take the monomer-dimer equilibrium 756 into account in considering and designing experimental conditions to detect hCEACAM1 757 interactions with its various ligands, a note of caution given the recent studies of others 758 (De Sousa Linhares et al., 2020) . 759

Consistent with our findings, many recent biophysical studies, including the 760 crystal structure of a hCEACAM1-HopQ complex and small angle x-ray scatting studies, 761 support the importance of the GFCC' face and monomer-dimer equilibrium in the 762 binding of hCEACAM1 to its various ligands (Figure 7-figure supplement 12) . In the 763 case of HopQ, its higher affinity for hCEACAM1 (K D ~23-279 nM) relative to that 764 associated with hCEACAM1 homodimerization (K D ~450 nM) resulted in successful 765 competition for binding to hCEACAM1 GFCC' face of another monomer. Furthermore, 766 the crystal structure that defines this ability of HopQ to achieve the formation of a high 767 affinity heterodimer is through its competence in interacting with the V39, I91 and N97 768 residues of hCEACAM1 (Figure 7-figure supplement 12) . Similarly, hCEACAM1 769 GFCC' face residues F29, Q44, I9I and E99 have also been shown to be critical to the 770 binding of OPA proteins and Afa/Dr adhesins (Figure 7-figure supplement 12) . Thus, 771 these recent findings further extend the role of GFCC' face residues not only in the 772 formation of hCEACAM1 homodimers, but also in the interactions with many different 773 heterophilic ligands including TIM-3, CEACAM5, HopQ, Opa proteins and Afa/Dr 774 adhesins in immune-regulation and immune-evasion that are exploited by neoplastic 775 cells and microbial pathogens (Figure 7-figure supplement 12) . 776

In summary, our biophysical and structural studies by crystallography and in 777 solution by NMR support a model wherein the GFCC' face is highly dynamic and seeks 778 thermal and energetic stability through formation of dimers (either homodimers or 779 heterodimers) that lock in a structurally favorable state. As such, CEACAM1 prefers to 780 be in a dimeric state that specifically stabilizes the CC' and FG loops, making the 781 GFCC' face the major interaction site for formation of homodimers and heterodimers. A 782 caveat of these studies is that they were performed with unglycosylated proteins that 783 may affect the ABED face; however, studies reporting a role of glycosylation in 784 disrupting homodimerization have been retracted (Zhuo et al., 2020 , Zhuo et al., 2016 . 785

That said, given the location of the carbohydrate side-chain modifications of CEACAM1 786 along the ABED face, the mutational analyses performed here and its implications still 787 have substantial physiologic merit. In addition to understanding the structural 788 mechanisms that underlie the formation of CEACAM1 monomers that are amenable to 789 interactions with another CEACAM1 molecule or its potential heterophilic partners, we 790 are hopeful that our high-resolution structural studies and hCEACAM1 monomer-dimer 791 model may also be useful in considering therapeutic targeting of hCEACAM1 792 interactions with its various ligands. 793

The hCEACAM1 WT IgV and mutant (V39A, I91A, N97A, E99A) proteins were 796 expressed and purified using our published protocols (Huang et al., 2016) . 797

Differential scanning fluorimetry was performed using a QuantStudio 6 (Life 799 Technologies) RT-PCR instrument with the excitation and emission wavelengths set to 800 587 and 607 nm, respectively. Assay buffer was 10 mM HEPES pH 7.4, 150 mM NaCl. 801

For thermal stability measurements, the temperature scan rate was fixed at 1 °C/min. was calculated for each protein samples through computation of a temperature 808 derivative for each respective melting curve that was then processed with a peak fitting 809 algorithm, applying a sigmoidal baseline and fitting the peak to determine the T m and its 810 standard error. our hCEACAM1 WT crystal structure (PDB code 4QXW) and many rounds of structure 846 refinement were done with simultaneous model building using Refmac (Murshudov, 847 Vagin, & Dodson, 1997) and COOT (Emsley et al., 2010) . The Fo-Fc map at 3.0 σ level 848 (derived from the initial model) showed significant positive Fo-Fc map density where 849 residue A39 in the V39A refinement model, residues A91 in the I91A refinement model, 850 residue A97 in the N97 refinement model , and residues A99 in the E99A refinement 851 model were not fitted, respectively ( Figure supplements 13A-D) . 852

The crystallographic twining and the significant metal electron density near the 853 and their two symmetry mates 0100-100 (S1) and 02000000 (S2) with the strategy that 861 showed optimal R/Rfree and revealed hexadentate interactions of Ni ++ with three His105 862 sidechains and three carbonyl groups of Val106 residues (Figure 3-figure  863 supplements 5B-C). The (Figure supplements 13E-H) 25°C on a 700MHz Agilent DD2 spectrometer equipped with a cryogenic probe. The 893 data were processed using NMRPipe (Delaglio et al., 1995) and Iterative Soft 894

Thresholding reconstruction approach (istHMS) (Hyberts et al., 2012) and analyzed by 895 CARA (Keller, R 2004 increments of 15ms or 20ms) up to 60ms or 100ms were used for anti TROSY 908 measurements. The 1D spectral region between 9.5ppm and 8.7ppm were integrated to 909 extract the pro and anti TROSY relaxation rates. Data were processed and analyzed 910 using the Bruker Topspin program. 911

NMR relaxation experiments were carried out using a 0.3mM 15 N/ 13 C double-912 labeled N97A mutant hCEACAM1 IgV protein sample at 25°C on a 700MHz Agilent 913 DD2 spectrometer equipped with a cryogenic probe. The T 1 relaxation times were 914 determined using antiphase inversion recovery delays of 10ms, 250ms, 500ms, 750ms 915 and 1s. The T 2 relaxation times were determined using Carr-Pursell-Meiboom-Gill 916 (CPMG) pulse train with  value of 625s, and delays of 10ms, 30ms, 50ms, 90ms and 917

150ms. The T 1rho relaxation times were determined with spin-locking field strength of 918 1.5kHz, and delays of 10ms, 30ms, 50ms, 90ms and 130ms. Data were processed 919 using NMRPipe and analyzed by CARA and Microsoft Excel programs. The errors of 920 the relaxation times were estimated from fitting routines. The average value of R 2 /R 1 921 (=T 1 /T 2 ) relaxation rate ratio is 9.4. This corresponds to a molecular rotational 922 correlation time of 7.9ns ( ), in agreement with the TRACT 923 measurement results after taking into account the homodimer population. The NMR 924 chemical exchange process for some of the N97A residues were not sufficiently 925 suppressed by the T 2 CPMG pulse train (with  = 625s), versus the more efficient 926 rotation frame T 1 spin-locking, resulting in artificially higher R 2 /R 1 and R 2 /R 1 ratios. 

The A91 residues as observed in the I91A crystal structure is shown by stick representation. (B)

The hydrogen bonded interactions between the residues described in the Figure 2B 1312 and V106 from molecule (a) and its two symmetry mates 0100-100 (S1) and 02000000 (S2) 1313 with bound Ni ++ (green sphere) as observed in the V39A crystal structure. Residues H105,

V106 from molecule (a), H S1 105, V S1 106, from symmetry molecule a_0100-100, and H S2 105, 1315 V S2 106 from symmetry molecule a_02000000, are highlighted in stick representation. These 1316 residues make interactions with Ni ++ , whereas nitrogen of three histidine rings of residues H105 1317 (2.35 Å) and three carbonyl oxygen of residues V106 (4.53 Å) participates in hexa-coordinated 1318 interactions with Ni ++ . Similar interactions were also observed with molecule (c) and its 

The binding between hCEACAM1 and hTIM-3 as shown by NMR, SPR and ELISA exhibits a K D 

959 iMOSFLM: A new graphical interface for diffraction-image processing with 960 MOSFLM

Redefined nomenclature for members of the 964 carcinoembryonic antigen family

Proceedings of the National 969 Academy of Sciences of the United States of America

The Helicobacter pylori adhesin protein HopQ exploits the 973 dimer interface of human CEACAMs to facilitate translocation of the oncoprotein 974

Fusobacterium spp. target human CEACAM1 via the trimeric autotransporter 977 adhesin CbpF

T cell immunoglobulin mucin-3 crystal structure reveals a 981 galectin-9-independent ligand-binding surface

The Moraxella adhesin UspA1 binds to its human CEACAM1 receptor by a 986 deformable trimeric coiled-coil

TIM-3 and CEACAM1 do not interact in cis and in 991 trans

NMRPipe: A multidimensional spectral processing system based on UNIX pipes

CEACAM1 enhances invasion and migration of 997 melanocytic and melanoma cells

Features and development 1001 of Coot

CEA-related cell adhesion molecule 1: a potent angiogenic factor and a 1005 major effector of vascular endothelial growth factor

Structure of the N-1009 terminal domain of human CEACAM1: binding target of the opacity proteins during 1010 invasion of Neisseria meningitidis and N. gonorrhoeae

High resolution X-ray and NMR structural study of 1016 human T-cell immunoglobulin and mucin domain containing protein-3. Scientific 1017 Reports

Abstract 2753: The molecular basis of blocking the TIM-3 checkpoint 1020 with the LY3321367 mAb in cancer immunotherapy

Carcinoembryonic antigen-1023 related cell adhesion molecule (CEACAM)-binding recombinant polypeptide confers 1024 protection against infection by respiratory and urogenital pathogens

Essential role of biliary 1027 glycoprotein (CD66a) in morphogenesis of the human mammary epithelial cell line 1028 MCF10F

Erratum: Corrigendum: CEACAM1 regulates TIM-3-1033 mediated tolerance and exhaustion

CEACAM1 regulates TIM-3-mediated tolerance and 1036 exhaustion

Application of iterative soft thresholding for fast reconstruction of NMR data non-1039 uniformly sampled with multidimensional Poisson Gap scheduling

Coevolution of activating and inhibitory 1042 receptors within mammalian carcinoembryonic antigen families

Optimizing the process of nuclear magnetic resonance spectrum 1045 analysis and computer aided resonance assignment. Swiss Federal Institute of 1046 Technology Zurich

CEACAM1 structure 1048 and function in immunity and its therapeutic implications

The CEACAM1 N-terminal Ig domain mediates cis-and 1052 trans-binding and is essential for allosteric rearrangements of CEACAM1 1053 microclusters

Binding of Candida albicans to human CEACAM1 and 1058 CEACAM6 modulates the inflammatory response of intestinal epithelial cells

The CEACAM1 N-terminal Ig domain 1062 mediates cis-and trans-binding and is essential for allosteric rearrangements of 1063 CEACAM1 microclusters

Helicobacter pylori exploits human CEACAMs via HopQ for adherence and 1067 translocation of CagA

Binding of Dr adhesins of Escherichia coli to carcinoembryonic antigen 1071 triggers receptor dissociation

Binding of Dr adhesins of Escherichia coli to 1076 carcinoembryonic antigen triggers receptor dissociation

Inference of Macromolecular Assemblies from 1079 Crystalline State

Effective rotational correlation times 1082 of proteins from NMR relaxation interference

The PD-1/PD-L1 complex resembles the antigen-binding Fv domains 1086 of antibodies and T cell receptors

Helicobacter pylori adhesin HopQ disrupts trans dimerization in 1090 human CEACAM s

Refinement of macromolecular 1092 structures by the maximum-likelihood method

The human 1095 tumor suppressor CEACAM1 modulates apoptosis and is implicated in early 1096 colorectal tumorigenesis

Processing of X-ray diffraction data collected in 1098 oscillation mode

Inside-out signaling promotes dynamic changes in the 1102 carcinoembryonic antigen-related cellular adhesion molecule 1 (CEACAM1) 1103 oligomeric state to control its cell adhesion properties

Nonuniform Sampling 1106 for NMR Spectroscopy

Blockade of Tim-3 binding to phosphatidylserine and 1110 CEACAM1 is a shared feature of anti-Tim-3 antibodies that have functional 1111 efficacy

Identification of an atypical monocyte and committed progenitor 1114 involved in fibrosis

Protein backbone and sidechain torsion angles predicted 1116 from NMR chemical shifts using artificial neural networks

Fast, scalable generation of high-quality protein multiple sequence 1120 alignments using Clustal Omega

1123 Deregulation of the CEACAM expression pattern causes undifferentiated cell 1124 growth in human lung adenocarcinoma cells

Crystal structure of murine sCEACAM1a[1,4]: a coronavirus receptor in the 1127 CEA family

Signaling by epithelial 1131 members of the CEACAM family -Mucosal docking sites for pathogenic bacteria

A highly conserved tyrosine of Tim-3 is phosphorylated upon stimulation by 1135 its ligand galectin-9

The n-domain of the human CD66a adhesion molecule is a target for opa 1141 proteins of neisseria meningioitis and neisseria gonorrhoeae

Homophilic adhesion of human CEACAM1 involves N-terminal 1145 domain interactions: structural analysis of the binding site

Acta 1150 Crystallographica Section D: Biological Crystallography

Automated data collection for macromolecular 1153 crystallography

Identification and characterization of M6903, an antagonistic anti-TIM-3 1156 monoclonal antibody

The Tim-3 ligand galectin-9 negatively regulates T helper type 1 1160 immunity

Glycosylation Alters 1164

Dimerization Properties of a Cell-surface Signaling Protein, Carcinoembryonic 1165 Antigen-related Cell Adhesion Molecule 1 (CEACAM1)

Correction: 1168 Glycosylation alters dimerization properties of a cell-surface signaling protein, 1169 carcinoembryonic antigen-related cell adhesion molecule 1 (CEACAM1)

E19A crystal structure. The weaker hydrophobic interactions mediated between two V39 1250 residues are shown by fewer pointers on the hydrophobic arc

molecules (c) and (d) residues of the V39A mutant crystal structure are shown by dashed lines

The residues of molecules (c) and (d) are shown in bold and italics underlined, respectively

The superimposition of the V39A mutant molecules (c, d, colored yellow) and WT molecules (a

The A39 and V39 residues of the weak V39A 1286 dimer and WT dimer are shown by yellow stick/dashed arrows and green stick/solid arrows, 1287 representation, respectively. The CC' loops are labeled as described above for WT and weak 1288 V39A dimer, and weaker hydrophobic interactions and conformation changes in the F29 1289 residues of the weak V39A dimer are shown by

D) The arc/stick representation of weaker hydrophobic interactions by F29, and I91 1291 residues as observed between molecules (c) and (d) in the formation of weak V39 dimer of the 1292 V39A crystal structure. The weaker hydrophobic interactions mediated are shown by fewer 1293 pointers on the hydrophobic arc. in binding with second Ni ++ as observed in the crystal structure. (C)

GFCC'C' face dimer that mimics WT dimer. In contrast, molecules (c) and (d) make a weak 1324 V39A dimer where CC' loops are apart and show large conformation differences compared to 1325 WT GFCC

The molecule (a) residues are 1349 labeled in bold, and molecule (b) residues are labeled in italics and underlined. The hydrogen 1350 bonded interactions across the GFCC' face by residues described above are shown by dashed 1351 lines. (B) The hydrophobic interactions of GFCC' face F29, I91, and V39 residues are shown by 1352 arc/point representations. The molecule (a) and (b) residues are labeled as described above. 1353 (C) Seven asymmetrical hydrogen bonded interactions (dashed lines) mediated by N97A 1354 residues of molecule a in red and molecule (b) in magenta as observed in the hCEACAM1 WT 1355 homodimer structure (PDB code 4QXW). The N97 residue from molecule (a) makes three 1356 hydrogen bonded interactions with residues S32, Y34, and Q44 of molecule (b) and N97 residue from molecule (b) makes four hydrogen

Figure supplement 7 15 N HSQC spectra of hCEACAM1 IgV WT and N97A mutant proteins

Assigned 15 N-HSQC spectrum of hCEACAM1 IgV WT protein (blue)

adhesins also target this GFCC' face for cellular invasion and immune evasion. Specially, the 1466

V96 binds hCEACAM1 with higher K D of 1467 23-279 nM than K D of homodimer formation and abrogate GFCC' face. rounds of model building and refinement. Figures E-H showing superimposition of 1504 2Fo-Fc map on the final refined model and residues are shown by stick representations

The final refined 1506 model of each mutant was reverse mutated to the respective residue present in the WT, 1507 whereas A91I (panel I), A99E (panel J), A39V (panel K), and A97N (panel L) mutations were 1508 done in the I91A, E99A, V39 and N97A mutants coordinates, respectively and one round of 1509 refinement cycle was performed. The observed negative Fo-Fc map at 3.0 σ level for each 1510 mutant clearly shows negative density around reverse mutation site and validates proper model 1511 building and refinement of the hCEACAM1 mutants

The residues are numbered 1539 with conserved residues shown in red and non-conserved residues shown in black. Asterisk (*) 1540 indicates similar residue conserved across all family members, Period (.) indicates conservation 1541 of weakly similar residues, and a colon (:) indicates conservation of strongly similar residues

The authors thank all the beamline scientists and the support staff of LS-CAT beamlines 936

mutation on the superimposed I91A mutant and WT structure is shown by stick representation.

The CC and CC' loops involved in the formation of GFCC'C' face I91A and WT dimer are