The structural basis for monoclonal antibody 5D2 binding to the tryptophan-rich loop of lipoprotein lipase

For three decades, the lipoprotein lipase (LPL)–specific monoclonal antibody 5D2 has been used to investigate LPL structure/function and intravascular lipolysis. 5D2 has been used to measure LPL levels, block the triglyceride hydrolase activity of LPL, and prevent the propensity of concentrated LPL preparations to form homodimers. Two early studies on the location of the 5D2 epitope reached conflicting conclusions, but the more convincing report suggested that 5D2 binds to a tryptophan (Trp)-rich loop in the carboxyl terminus of LPL. The same loop had been implicated in lipoprotein binding. Using surface plasmon resonance, we showed that 5D2 binds with high affinity to a synthetic LPL peptide containing the Trp-rich loop of human (but not mouse) LPL. We also showed, by both fluorescence and ultraviolet resonance Raman spectroscopy, that the Trp-rich loop binds lipids. Finally, we used X-ray crystallography to solve the structure of the Trp-rich peptide bound to a 5D2 Fab fragment. The Trp-rich peptide contains a short a -helix, with two tryptophans projecting into the antigen recognition site. A proline substitution in the a -helix, found in mouse LPL, is expected to interfere with several hydrogen bonds, explaining why 5D2 cannot bind to mouse LPL.

and surface plasmon resonance studies, that 5D2 binds to LPL's Trp-rich motif, and they went on to quantify the impact of each amino acid residue in the Trp-rich loop for 5D2 binding affinity.
Studies using 5D2 to probe LPL structure have also had a topsy-turvy history. Brunzell's laboratory reported that LPL could be detected with a "single antibody" sandwich ELISA in which 5D2 was used both to capture LPL and to detect the bound LPL (3). That observation led them to infer that LPL must be a homodimer. Additional studies with antibody 5D2 led them to infer that LPL's catalytic activity actually depends on the assembly of LPL into homodimers (3). The concept that LPL was a homodimer had already been proposed (10)(11)(12), but the immunochemical studies by Brunzell and coworkers (3) were crucial for dogmatizing this concept. The notion that LPL is active only as a homodimer was universally accepted (4,(13)(14)(15), and newer discoveries in the field were invariably interpreted within the framework of LPL being a homodimer (16,17). In recent years, however, both of the early inferences from 5D2-based immunoassays-that LPL is a homodimer and that homodimer formation is essential for catalytic activity-have proven to be incorrect. Using 5D2-based ELISAs and density gradient ultracentrifugation studies, Beigneux et al. (18) showed that freshly secreted, catalytically active LPL is monomeric. Detection of fresh LPL in a "single-antibody" 5D2 sandwich ELISA was negligible, consistent with LPL being a monomer (18). The fact that freshly secreted LPL, as well as low concentrations of purified LPL, are monomeric was confirmed by highly standardized density gradient ultracentrifugation studies (18). Subsequent small angle X-ray scattering studies and X-ray crystallography studies (19) revealed that purified LPL, when present at high protein concentrations, assumes a head-to-tail homodimer conformation (19). Two partner LPL monomers interacted reciprocally at a single site; the Trp-rich loop in the carboxyl terminus of one LPL monomer was buried in the catalytic pocket in the amino terminus of the partner LPL monomer (19). The same head-to-tail LPL dimer interactions were observed by cryo-EM in large oligomeric helical LPL fibrils (20). This head-totail homodimer conformation is incompatible with LPL activity, as it would not allow interactions of lipoproteins with the Trp-rich motif, nor would it allow triglyceride hydrolysis by the catalytic domain (18,19). The fact that 5D2 binds to LPL's carboxyl-terminal Trp-rich loop led Kristensen et al. (9) to predict that the binding of 5D2 to LPL would abolish the LPL-LPL interactions observed in the crystal structure and thereby trap LPL in a monomeric conformation. Indeed, purified preparations of LPL, even at high protein concentrations, were monomeric in the presence of 5D2 Fab fragments, and the monomeric LPL remained catalytically active against a soluble substrate (9).
Thus, the field's understanding of LPL structure had come full-circle. Older studies with 5D2 were interpreted as showing that LPL is a homodimer and that homodimers are required for catalytic activity (3). The more recent studies with 5D2 revealed that catalytically active LPL is actually monomeric (18) and that 5D2 abolishes LPL's propensity for homodimer formation (9). Now, three decades after antibody 5D2 was created, two issues require clarification. First, the structural basis for 5D2 binding to the Trp-rich loop in LPL has never been defined, nor have there been insights into why 5D2 fails to bind to mouse LPL. A second issue requiring attention is the role of LPL's Trp-rich loop in binding lipids. The Trp-rich loop is clearly important for lipoprotein binding and triglyceride hydrolysis in the context of full-length LPL (6), but evidence that the tryptophans are directly involved in lipid binding has been lacking. In the current study, we investigated both issues.

Production of monoclonal antibody 5D2
The hybridoma for 5D2 was a gift from Dr. John Brunzell (University of Washington). The cells were adapted and cultured in a 50:50 mixture of PFHM II (Gibco) and DMEM (Gibco) media containing 10% FBS (GE Healthcare), 5% L-glutamine, 5% penicillin/streptomycin, and 5% sodium pyruvate. Medium from four T175 flasks (Corning) was harvested every 3-4 days, centrifuged to remove detached cells, and filtered through a 0.22-µm Stericup filter (MilliporeSigma). Antibodies in the medium were concentrated by precipitation with 50% ammonium sulfate and centrifugation (4000 ´ g for 30 min); the pellet was resuspended and dialyzed against 20 mM sodium phosphate buffer (pH 7.0). After centrifugation (4000 ´ g for 10 min) to remove insoluble material, the supernatant was loaded onto a Protein G-agarose (Sigma Aldrich) column. The column was washed with a minimum of 10 column volumes of 20 mM sodium phosphate buffer (pH 7.0), and antibodies eluted with 100 mM glycine buffer (pH 2.8).
The eluate (containing 5D2) was immediately neutralized with 1 M Tris (pH 9.0). The purity of the antibody was established by SDS-PAGE and Coomassie Blue staining; the binding of the antibody to human LPL was established by western blotting. Protein concentration of the 5D2 antibody was determined by the Bradford protein assay.

Mammalian expression vectors
The complete coding sequence (including the native signal peptide) of human (h), mouse (m), chicken (ck), zebrafish (z), and frog (x) LPL were cloned in frame with a carboxyl-terminus V5 tag into the pcDNA6/V5-His vector (ThermoFisher Scientific) for expression in mammalian cells under a CMV promoter (21). All plasmids were sequenced verified after cloning, and the LPL amino acid sequences started and ended as follows:

Western blot studies of 5D2 binding to LPL
CHO cells (5 ´ 10 6 ) were electroporated with an expression vector for V5-tagged human, mouse, chicken, zebrafish, and frog (Xenopus tropicalis) LPL and then seeded in a 6-well plate.
On the next day, total cell lysates were prepared in a buffer (150 mM NaCl, 1.0% IPEGAL CA-630, 0.5% sodium deoxycholate, 0.1% SDS, 50 mM Tris pH 8.0). Proteins were separated by SDS-PAGE and then transferred to a sheet of nitrocellulose membrane. LPL proteins were detected with an IRDye800-conjugated monoclonal antibody against the V5 tag (Thermo Fisher Scientific, 2 µg/ml) and IRDye680-conjugated 5D2 (4 µg/ml). The signal for each antibody was quantified with an Odyssey infrared scanner (LI-COR).

Generating 5D2 Fab fragments
5D2 Fab fragments were prepared using the mouse IgG1 Fab and F(ab¢)2 fragmentation kit (Thermo Fisher Scientific). Briefly, 5D2 (8 mg/ml) was passed through a Zeba Spin desalting column that was pre-equilibrated with digestion buffer and incubated in the presence of 25 mM cysteine with Ficin-agarose at 37°C for 5 h to create Fab fragments. The Ficin-agarose was removed by centrifugation and non-digested IgG1 and Fc fragments were removed with a Protein A-agarose spin column. Fab fragments were concentrated with a 3K centrifugal filter (Amicon) and further purified by size-exclusion chromatography with a Superdex 200 10/300 column (GE Healthcare) equilibrated with PBS on an ÄKTA Pure HPLC (GE Healthcare) at a flow rate of 0.4 ml/min. The size-exclusion fractions corresponding to the Fab fragments were collected, pooled, and purity verified by SDS-PAGE. The protein concentration of the purified Fab fragments was determined with the BCA Protein Assay (Thermo Fisher Scientific).
The ability of purified 5D2 Fab fragments to bind to GPIHBP1-bound LPL was verified with an LPL-GPIHBP1 "co-plating assay" (21). RJ-1 cells (CHO pgsA-745 cells in which the hamster gene for LPL had been knocked out) (5 ´ 10 5 cells) were electroporated with either 0.5 µg of a plasmid for S-protein-tagged versions of wild-type human GPIHBP1 or GPIHBP1-W109S. These cells were then mixed with RJ-1 cells (5 ´ 10 5 ) that had been electroporated with 0.

Analysis of 5D2-LPL binding by surface plasmon resonance
To define the species selectivity of 5D2, we used a Biacore T200 (GE Healthcare) to measure the real-time binding kinetics of synthetic 14-mer LPL peptides spanning the tryptophan-rich lipidbinding sequences from six vertebrate species. We immobilized 5D2 (5 µg/ml in 10 mM NaOAc, Data for the 5D2 Fab-LPL14 complex was collected at the 17-ID-1 beamline at NSLS II (Brookhaven National Laboratory, NY). After analyzing a single crystal by rastering, X-ray diffraction data to 2.74 Å were collected along a screw axis and processed in space group P21 with HKL2000 (29). Using Phaser (26), the first of two Fab molecules in the asymmetric unit was found by molecular replacement using the structure of the apo 5D2 Fab as the model. Two of the IgG domains of the second copy were placed in a subsequent round of molecular replacement with the first solution fixed. However, due to the difference in elbow angles, the remaining two Ig domains did not fit the electron density map. Attempts to identify a molecular replacement solution for the remaining two Ig domains (constant) using Phaser and MolRep (30) failed, possibly due to weaker electron density for these domains. As density for several b-strands was apparent in the electron density maps, Buccaneer (31) was used to autobuild several b-strands fragments into the electron density map. The complete structural model was built using COOT (32) and refined with REFMAC5 (33) at 2.71 Å resolution to an Rfactor of 20.8% and an Rfree of 24.3% ( Table 2).
Validation of the models was carried out using MolProbity (34).

Preparation of small unilamellar vesicles (SUVs)
To generate SUVs, a 25 mg aliquot of 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC from Avanti Polar Lipids) dissolved in chloroform was dried under nitrogen gas for at least 4 h. and SUVs-only) were allowed to equilibrate at 37°C for at least 1 h prior to collection of measurements. Peptide and NATA samples that did not contain SUVs were incubated at room temperature for at least 30 min prior to measurements. All reagents (NATA and buffer salts) were purchased from Sigma-Aldrich and Thermo Fisher Scientific.

UV-Vis absorption spectroscopy
The absorption spectra of LPL14 and NATA samples were acquired using an Agilent 8453 UV-Visible absorption spectrometer. The absorption features from LPL14 and NATA were by guest, on July 25, 2020 www.jlr.org Downloaded from isolated by subtraction of an SUV-only spectrum from the raw spectra of peptide and NATA. The concentrations were determined using the 280-nm molar absorption coefficient, ε280, for tryptophan (5,500 M -1 cm -1 ) and tyrosine (1,490 M -1 cm -1 ) (35) The calculated value of ε280 for LPL14 was 17,990 M -1 cm -1 based on the presence of one tyrosine and three tryptophan residues.
All absorption spectra were acquired using a 1-cm path length.

Steady-state fluorescence spectroscopy
The fluorescence spectra were acquired using a Jobin Yvon-SPEX Fluorolog FL3-11 spectrofluorometer (Horiba). All samples were held in a quartz cuvette sealed with a Teflon cap and maintained at 37°C during collection of the spectra. The samples were excited with 290 nm light along the 2 mm path, and emission was collected along the 10 mm path from 305 to 550 nm.
The excitation and emission bandpass were set to 2 nm. Spectra were collected at 1-nm increments with 0.8-sec integration time. Fluorescence spectra of SUV-only and buffer-only samples were also collected, and these background spectra were subtracted from the raw fluorescence spectra of peptide and NATA to eliminate signal from buffer and SUV scattering.

UV resonance Raman spectroscopy (UVRR)
The UVRR Ti:Sapphire laser apparatus has been described previously (36). UVRR spectra of LPL14 and NATA were acquired with a 228 nm UV excitation beam. The concentrations of LPL14 were 11 µM and 9 µM in the presence and absence of SUVs, respectively. NATA concentrations were 33 µM and 32 µM in the presence and absence of SUVs, respectively. To isolate signal from LPL14 and NATA, UVRR spectra of SUVs-only and buffer-only were also acquired and subtracted to remove contributions from vesicle and buffer scattering. All UVRR spectra were collected for 20 min. The UV power at the sample was 1.2 mW and the flow rate was set to 0.16 ml/min to ensure fresh sample with each laser pulse and avoid sample degradation.

Binding of 5D2 to LPL
We analyzed, by surface plasmon resonance (SPR), the binding affinity of immobilized 5D2 to a 14-mer Trp-rich human LPL synthetic peptide (LPL14; KSDSYFSWSDWWSS) and 14-mer peptides corresponding to the same Trp-rich loop in mouse, bovine, rat, chicken, and megabat ("flying fox," Pteropus vampyrus) LPL. The binding affinity of 5D2 to the human and bovine LPL peptides was very high (KD of 0.19 and 0.78 nM, respectively) (Table 1, Fig. 1). The binding affinity of 5D2 for the chicken LPL peptide was also very high, despite the fact that 5 of 14 residues differed from the human sequence ( Table 1). The binding affinity for the rat LPL peptide was modestly reduced, likely due to a Ser-to-Arg substitution in the Trp-rich loop ( Table 1). In an earlier study (9), changing that serine in human LPL (Ser-416) to Ala reduced the binding affinity for 5D2. By SPR, 5D2 did not bind to the Trp-rich peptide from mouse LPL, where Ser-418 is replaced with a Pro (Table 1, Fig. 1). 5D2 also did not bind to the Trp-rich LPL peptide from megabat (Table 1, Fig. 1), which has the same Ser-to-Pro substitution. In earlier SPR studies (9), replacing Ser-418 in human LPL with Ala reduced the affinity of 5D2 binding to a synthetic LPL peptide.
We used western blotting to assess 5D2 binding to V5-tagged full-length human, mouse, chicken, zebrafish, and frog LPL (Fig. 2). We observed avid 5D2 binding (relative to the binding of a V5 antibody) to human LPL, but there was no binding of 5D2 to mouse, zebrafish, or frog LPL (Fig. 2). All three of those species contain substitutions in the WSDWW motif (417-421) that is important for 5D2 binding (2,9). We did detect 5D2 binding to full-length chicken LPL ( Fig. 2), but the binding was lower than we would have expected from the SPR studies ( Table 1, Fig. 1). This discrepancy suggests that the strength of 5D2 binding to full-length LPL differs from the 14-mer LPL peptide (LPL14). Such a conclusion is consistent with findings from Kristensen et al. (9), who observed that 5D2 binds to the properly folded carboxyl-terminal domain of human LPL (residues 340-475) with greater affinity than to a 14-mer human LPL synthetic peptide.
Given that affinity of 5D2 binding to synthetic peptides and properly folded LPL can differ, we examined 5D2 binding to full-length mouse and human LPL. We transfected CHO cells with by guest, on July 25, 2020 www.jlr.org Downloaded from a V5-tagged full-length mouse LPL vector or with a mutant vector in which Pro-418 had been changed to Ser (the residue found in human LPL). By immunocytochemistry, 5D2 failed to bind to full-length wild-type mouse LPL but bound avidly to the mutant mouse LPL with the p.Pro418Ser substitution (mLPL-P418S) (Fig. 3). In parallel, we tested 5D2 binding to full-length human LPL and a mutant human LPL containing a p.Ser418Pro substitution. 5D2 bound avidly to the wild-type human LPL, but there was no binding to the mutant human LPL with the p.Ser418Pro substitution (hLPL-S418P) (Fig. 3).

Structure of the 5D2 Fab fragment bound to the human LPL14 peptide
We purified Fab fragments of 5D2, verified that they bound GPIHBP1-bound LPL (Fig. S1), and then used X-ray crystallography to solve the Fab structure, both in the presence and absence of the 14-mer LPL peptide (LPL14). The structure revealed a typical Fab tertiary fold comprised of four Ig domains, two from the light chain and two from the heavy chain and contained one Fab molecule per crystallographic asymmetric unit. In the Apo 5D2 structure, indexing and scaling of diffraction images for the apo 5D2 structure was complicated by anisotropic diffraction and elongated spots leading to elevated Rwork and Rfree factors (Table 2); however, the electron density maps were easily interpretable and allowed for an unambiguous assignment of main-chain and side-chain atoms. Interestingly, the electron density maps for the Apo 5D2 structure showed evidence for two conformations in the side chain of W69 of the heavy chain (part of the antigen binding site) (Fig. S2). Apart from differences in the IgG hinge angle, the Apo and LPL14-bound 5D2 structures were very similar; the variable domains superimpose with a root-mean-square (RMS) deviation of 0.61 Å in main-chain atom positions (Fig. S3). The 5D2-LPL14 complex contained two molecules per crystallographic asymmetric unit which are almost identical to each other, with an RMS deviation of 0.31 Å in main-chain atom positions within the variable domains.
LPL14 was located in the canonical antigen-binding site at the interface between the heavy and light chains and shaped at its periphery by residues from the complement-determining region (CDR) loops (Fig. 4). The LPL14 peptide was located parallel to the axis of the antigen recognition site, with a hydrophobic face comprised of side chains from F415, W417, W420, and W421 oriented towards the light chain and a polar face (comprised of side chains from S416, S418, D419, by guest, on July 25, 2020 www.jlr.org

Downloaded from
and S422) oriented towards the heavy chain (Fig. 5). The N-terminus of LPL14 arches over the top of the heavy chain CDR-H3 loop, with the side chain of D412 in LPL14 folding back to form a hydrogen bond with the guanidinium group of the R121 side chain in the 5D2 heavy chain. A Cterminal a-helical turn in LPL14 (residues 416-421) was nestled into the base of the antigen binding site (Fig. 5).
The interaction of LPL14 with 5D2 is biased towards heavy chain contacts. A total of 193 intermolecular interactions between LPL14 and the heavy chain were identified within a 4.5 Å limit by CONTACT from the CCP4 suite (38), whereas only 37 interactions were identified between LPL14 and the light chain. In addition, 454 Å 2 of the peptide surface area was buried by the heavy chain, whereas only 131 Å 2 was buried by the light chain, as determined by the PISA server (39). Electron densities for LPL14 residues K410, S411, and S423 were absent and were not included in the model.
W417 and W421 of LPL14 (within the a-helical turn) extend into a cleft in the antigenrecognition site, with the LPL14 side chains forming edge-to-face p-stacking interactions with light chain W112 (Fig. 6) and additional hydrophobic interactions with the side chains of H54, W69, and H118 of the heavy chain (Fig. 6) and L117 of the light chain. The indole nitrogen of LPL W417 forms a hydrogen bond with the main-chain carbonyl of heavy chain N122, and the indole nitrogen of LPL W421 forms a hydrogen bond with the backbone carbonyl of W112 of the light chain (Fig. 5). S418 in LPL14 is located within the a-helical turn of the peptide (Fig. 6), and both the main-chain nitrogen and side-chain hydroxyl are within hydrogen bonding distance of the side-chain carboxylate of heavy chain D120 (Fig. 5).
Finding a short a-helix in LPL14 was in line with predictions regarding the structure of the peptide. We submitted the sequence of the LPL14 peptide to the PEP-FOLD3 server (40), and five models were returned. All five had an amino-terminal region of undefined secondary structure and a carboxyl-terminal a-helix similar to that observed in the LPL14-5D2 crystal structure (Fig. S4).
The top model was very similar to the crystal structure, with an RMS deviation for backbone atoms of residues in the a-helical region (residues 416-421) of 0.13 Å (Fig. S4). While the side chains by guest, on July 25, 2020 www.jlr.org Downloaded from of F415, W417, and W421 were observed as different conformers in the crystal structure and the PEP-FOLD3 models, the-a carbons were nearly identical.
There are eight atom pairs in LPL14 and 5D2 common to both molecules in the asymmetric unit that are within hydrogen bonding distance of each other (Table S1). Of the polar atoms within hydrogen bonding distance, only one pair is found between the LPL peptide and the light chain (LPL W421 indole nitrogen with light chain W112 carbonyl). One pair includes LPL14 main chain atoms and 5D2 side chain atoms, and three pairs include LPL14 side chain atoms and 5D2 main chain atoms. There is one hydrogen bond between backbone atoms of both chains-heavy chain D120 carbonyl oxygen to LPL W417 amide nitrogen. The remaining three potential hydrogen bonds are found between pairs of side chain atoms. The network of hydrogen bonds extends across the span of the peptide that was visualized, with the exception of LPL14 residues Y414, D419, and W420 (where the side chains point away from the binding site) and F415 (which cannot form a side-chain hydrogen bond). Of note, framework residue I52 of the heavy chain, which is present in only 0.74% of Kabat database IgG sequences (www.bioinf.org.uk/abs/seqtest.html), forms extensive van der Waals contacts with several LPL14 residues (W417, S418, W421) (Fig. 6).
Electron densities for LPL residues K410, S411, and S423 were absent, suggesting that they are not particularly important for 5D2 binding.
Earlier SPR studies found that converting F415, W417, or W421 to an Ala reduced 5D2 binding affinity by more than three orders of magnitude (9). All three aromatic side chains contribute significantly to both the intermolecular hydrophobic packing interactions in the antigen recognition site (Fig. 5, Fig. 6) as well as to intramolecular hydrophobic packing that shapes the aromatic face of the peptide. In the same SPR study (9), changing LPL S418 or W420 to Ala decreased binding by 50-fold. The p.Ser418Ala substitution would eliminate a hydrogen bond between the side-chain hydroxyl of Ser-418 and the side-chain carboxylate of heavy chain D120 (Fig. 5). The W420A substitution would eliminate hydrophobic packing with the phenyl group of F415.
The conformation of the peptide itself appears to be stabilized by a network of hydrogen bonds which are conserved between the two copies in the asymmetric unit, except for that formed with by guest, on July 25, 2020 www.jlr.org Downloaded from the main-chain carbonyl atom of Ser-418 (Table S2). The carboxyl-terminal a-helical turn is stabilized by main-chain hydrogen bonds formed between residues S416 and W420 and between W417 and W421. Our experiments demonstrated that 5D2 does not bind to mouse LPL (which contains a Pro at residue 418 rather than a Ser), either in the context of the 14-mer synthetic LPL peptide or full-length LPL (Fig. 1, Fig. 3). Also, when Ser-418 in human LPL is replaced with a Pro, the binding of 5D2 is abolished (Fig. 3). These observations are explained by our crystal structure. Both the backbone nitrogen and side-chain hydroxyl of Ser-418 in human LPL14 are within hydrogen bonding distance of the side-chain carboxylate of heavy chain D120 (Fig. 5).
Substitution of Ser-418 with a Pro would eliminate crucial hydrogen bonds and introduce a hydrophobic side chain which would clash with the side chain of D120. Also, the side-chain hydroxyl of Ser-418 is within hydrogen bonding distance to the side-chain hydroxyl of Ser-416, thereby contributing to the internal order of the bound LPL14 structure (Fig. 5). While the Pro at residue 418 interferes with 5D2 binding, there is no reason to suspect that it would interfere with the role of the Trp-rich loop in enzyme function. Interestingly, the PEP-FOLD3 server (40) predicted a short a-helix in the mouse 14-mer LPL peptide, with one side enriched in hydrophobic amino acids (similar to the crystal structure and the PEP-FOLD3 prediction for human LPL14) (Fig. S5).

The Trp-rich 5D2 peptide (LPL14) interacts directly with lipids
Spectroscopy revealed tryptophan-lipid interactions between LPL14 and SUV bilayers. The absorption spectrum of LPL14 at 250-320 nm primarily reflects LPL14's three tryptophan residues. In the presence of SUVs, the absorption spectrum exhibited a 1-nm red shift from 280 to 281 nm, relative to LPL14 in buffer (Fig. 7). This shift was only observed in LPL14, and not the model tryptophan compound NATA. This 1-nm shift for LPL14 indicates that the tryptophan residue(s) are in a more hydrophobic environment in the presence of SUVs (41). This change in environment from aqueous to lipid bilayer is corroborated by fluorescence spectra. The tryptophan residues in LPL14 exhibited a fluorescence maximum at 358 nm in buffer and shifted to 347 nm in the presence of SUVs, and there was a concomitant increase in the fluorescence intensity (i.e., quantum yield) (Fig. 8). This 11-nm blue shift and increase in the quantum yield upon incubation with SUVs indicates that the tryptophan residues are in a more hydrophobic lipid environment (42).
Vibrational spectroscopy in the form of UVRR supported the absorption and fluorescence findings. UVRR is an ideal tool because excitation at 228 nm enhances the signal from tryptophan without obfuscation from other residues or buffer. The UVRR bands reflect different properties of the indole side chain. The tryptophan W7 Fermi doublet band at 1300 to 1400 cm -1 is particularly informative because it is a sensitive marker for local hydrophobicity, and the ratio of the peak intensities of 1366 to 1345 cm -1 (RFD) increases when tryptophan is in a hydrophobic environment (36). The RFD value for LPL14 in the presence of SUVs is 1.6, which is significantly larger than the values in buffer and for NATA (Fig. 9). This finding supports the notion that the tryptophan residue(s) in LPL14 directly interact with the more hydrophobic environment of the lipid bilayer.

DISCUSSION
We defined the structure of 5D2 bound to a 14-mer peptide (LPL14) corresponding to the Trprich loop within the carboxyl terminus of human LPL. Eleven of the 14 LPL residues were identified in the electron density map. The carboxyl-terminal portion of the peptide formed a short amphipathic a-helix that was positioned in the cleft of the antigen recognition site, with the hydrophobic face (three tryptophans and a phenylalanine) oriented towards the light chain and a polar face (three serines and an aspartic acid) pointing towards the heavy chain. In our crystal structure, a short a-helix was observed in the LPL14 peptide bound to 5D2. However, at this point, it is unclear whether this a-helix is an invariant feature in native LPL or whether it simply represents one of several transient conformations that is detected and then trapped by 5D2 binding.
Favoring the latter possibility is the observation that this segment in native LPL (specifically, the peptide FSWSNW) is not protected from deuterium uptake in hydrogen-deuterium exchange studies unless the LPL is bound to Fab 5D2 (9). If the a-helical conformation were an invariant feature of native LPL, we would have expected to find that the backbone LPL amide hydrogens were protected from deuterium uptake (even in the absence of bound 5D2). The Trp-rich loop in full-length LPL was visualized in a recent cryo-EM LPL structure (20) and appeared to show an a-helical conformation. However, that conformation was presumably a consequence of the repetitive head-to-tail homodimer LPL configuration in the cryo-EM structure, where the Trp-rich loop is buried in the catalytic pocket of a partner LPL molecule. In our X-ray crystal structure of the LPL-GPIHBP1 complex (19), LPL's Trp-rich loop was also buried in the catalytic pocket of a partner LPL molecule and not clearly visualized. In considering the latter structure, there is little reason to believe that GPIHBP1 binding would have changed conformation of the LPL's Trp-rich loop, simply because the Trp-rich loop and the binding site for GPIHBP1 are located on opposite sides of the LPL molecule (19). 5D2 binds avidly to GPIHBP1-bound LPL, as shown in earlier studies (9,43) and in Fig. S1 of the current study.
Our studies demonstrated that LPL's Trp-rich loop is a crucial part of the 5D2 epitope, but it is possible that other LPL sequences influence, either directly or indirectly, the strength of 5D2 binding. First, by SPR, 5D2 bound with higher affinity to the carboxyl-terminal region of human by guest, on July 25, 2020 www.jlr.org

Downloaded from
LPL (residues 340-475) than to the LPL14 synthetic peptide (9). Second, by SPR, the KD for 5D2 binding to the human and chicken 14-mer peptides were similar, but by western blotting 5D2 binds more avidly to full-length human LPL than to full-length chicken LPL.
Studies by Williams et al. (7) revealed that mutating W420 and W421 in the carboxyl terminus of LPL markedly reduced LPL's ability to bind lipoproteins. Subsequent studies by Lookene et al. (6) revealed that mutating W420 and W421 markedly reduced triglyceride hydrolysis, suggesting that the tryptophans were either directly involved in binding lipids or that the tryptophan mutations had simply disrupted LPL conformation. Our current studies support the former possibility. When we incubated LPL14 with small unilamellar vesicles (SUVs), the peak emission spectra by fluorescence spectroscopy shifted from 358 to 347 nm, whereas no shift occurred when SUVs were added to the model compound N-acetyl-tryptophanamide (NATA). These findings implied direct interactions of the hydrophobic Trp-rich loop with lipids, which is consistent with the fact that tryptophans often participate in the binding of proteins to membrane bilayers (44). Moreover, ultraviolet resonance Raman (UVRR) spectra indicated that the tryptophans in LPL14 interacted directly with lipids. The hydrophobicity of LPL's tryptophan-rich loop also underlies the propensity of LPL, at high concentrations, to form head-to-tail homodimers (19). In those homodimers, the interaction of the Trp-rich loop with the catalytic pocket shields the hydrophobic tryptophans from the aqueous environment (19). The binding of 5D2 to LPL also shields the tryptophan-rich loop and thereby prevents the formation of homodimers (9 Birrane (e-mail: gbirrane@bidmc.harvard.edu; phone: 617-667-0025). As noted in Figure 2, the LPL11 (PDB ID 6WN4) have been deposited -coordinates for APO 5D2 (PDB ID 6WT3) and 5D2 in the PDB. Fo and Fc are the observed and calculated structure factors, respectively, and k is a scaling factor. The summation is over all measurements. d Rfree is calculated as Rcryst using 5% of the reflections chosen randomly and omitted from the refinement calculations. e Bond lengths and angles are root-mean-square (RMS) deviations from ideal values.
The coordinates for APO 5D2 (PDB ID 6WT3) and 5D2-LPL11 (PDB ID 6WN4) have been deposited in the PDB.  Species selectivity of 5D2 binding to LPL peptides, as judged by surface plasmon resonance. Shown here are sensorgrams displaying the real-time binding profile of five 14-amino acid LPL peptides to 5D2. The five peptides correspond to the tryptophan-rich loop in the LPL from different mammalian species. Shown here are the sensorgrams recorded by single-cycle kinetics for two-fold dilutions (0.5 nM to 8 nM) of synthetic peptides for human LPL (LPL, cyan), cow LPL (green), and rat LPL (gray), along with the corresponding kinetic fits to the experimental data (thin black lines). Also shown are the sensorgrams for mouse LPL (blue) and megabat (Pteropus vampyrus) LPL (red) measured for two-fold dilutions spanning from 8 nM to 128 nM. We omitted the sensorgram for chicken LPL because the data were recorded at higher peptide concentrations, complicating direct comparisons to the data generated with human, bovine, and rat LPL.   (21), mLPL, ckLPL, zLPL, and xLPL. On the following day, total cell extracts were prepared from the transfected cells for western blot analysis (B). After probing the blot with an IR800-labeled anti-V5 antibody (21) and IRDye680-labeled 5D2 (B), the signal for each antibody was quantified with an Odyssey infrared scanner (LI-COR). The ratio of the 5D2 signal to the V5 antibody signal was plotted in the bar graph shown in panel C. Fig. 3. Introducing a p.S418P mutation into human LPL (hLPL) abolished 5D2 binding, whereas introducing a p.P418S mutation into mouse LPL (mLPL) makes it possible for the protein to bind 5D2. CHO cells were transfected with expression vectors for wild-type mLPL, mLPL-P418S, wild-type hLPL, or hLPL-S418P; all LPL proteins contained a carboxyl-terminal V5 epitope tag. Cells were plated onto glass cover slips. After 24 h, cells were washed three times (10-min each) with PBS/Ca 2+ /Mg 2+ ; fixed with 3% paraformaldehyde (PFA); and permeabilized with 0.2% Triton X-100. Cells were then incubated for 1 h with an Alexa 647-conjugated antibody against the V5 tag (red) and Alexa 488-conjugated 5D2 (green). After washing the cells, they were fixed with 3% PFA, stained with DAPI (blue), and mounted onto slides. Images were recorded with an LSM 700 confocal fluorescence microscope (Zeiss). Scale bar, 50 µm. showing hydrogen bonds (dashed cyan lines) between the Fab fragment and the LPL14 peptide (depicted as sticks and colored by atom; C, colored according to chain; O, red; N, blue). The side chains of LPL residues D412, S413, S416 in the unstructured region of LPL14 and S418 in the ahelical region form a polar interface that is oriented towards the heavy chain (purple ribbons). The N-terminal region of the LPL14 peptide points out of the antigen binding site and arches over the CDR-H3 loop, with LPL D412 forming hydrogen bond with the R121 side-chain guanidinium of the heavy chain. Ne1 in the five-membered pyrrole ring of W417 and W421 forms hydrogen bonds with the main chain carbonyl groups of N122 (heavy chain) and W112 (light chain), respectively.

Fig. 6.
Hydrophobic interactions of LPL14 residues F415, W417, S418, W421, and S422 (green) with the 5D2 Fab (heavy chain, purple; light chain, khaki). Side chains are represented as sticks and colored by atom (O, red; N, blue; C, colored according to chain). The LPL14 peptide is oriented parallel to the axis of antibody binding. The side chains of F415, W417, S418, and W421 are directed towards the antigen binding site and make hydrophobic contacts predominantly with the 5D2 heavy chain. LPL-S418 makes a hydrophobic contact with W69 and Y71 of the heavy chain, and LPL-S422 interacts with heavy chain K78. There is a minor interaction between LPL W421 and N115 of the light chain. W112 of the light chain interacts with LPL W421 and W417. L117 of the light chain also interacts with LPL W421. The side chain of I52 forms hydrophobic interactions with the side chains of LPL W417 and W421. Fig. 7. Absorption spectra of LPL14 in buffer with and without small unilamellar vesicles (SUVs). Also included are spectra of the model compound N-acetyl-tryptophanamide (NATA) in buffer with and without SUVs. Spectra were normalized to equal absorbance at the peak wavelengths of 280 or 281 nm. The spectra of NATA with and without SUVs overlap. Concentrations for LPL14 were 9 µM (buffer) and 11 µM (SUVs); concentrations for NATA were 32 µM (buffer) and 33 µM (SUVs). The cuvette pathlength was 1 cm. The 1 nm shift in absorption maximum from 280 to 281 nm for LPL14 in the presence of SUVs is consistent with interaction of LPL14 with the lipid bilayer.   9. Ultraviolet resonance Raman (UVRR) spectra of tryptophan residues in LPL14 in buffer with (A) and without SUVS (B) and model compound NATA in buffer with (C) and without SUVs (D). Spectra were normalized to match the intensity of the W16 band. Concentrations for LPL14 were 9 µM (in buffer) and 11 µM (with SUVs), and for NATA were 32 µM (in buffer) and 33 µM (with SUVs). The excitation wavelength was 228 nm (1.2 mW at the sample). Data were collected for 20 min for each spectrum. Prominent tryptophan UVRR peaks, including their energies, are labeled. Expanded view of the Fermi doublet region from 1300 to 1400 cm -1 and the calculated ratio (RFD) of the 1366 cm -1 to 1345 cm -1 bands are shown above each spectrum. Gaussian fits to the doublet are shown as red and blue curves, and the sum of the Gaussians is shown as the dashed overlay. The high value of RFD = 1.6 for LPL14 in the presence of SUVs indicates the tryptophan residue(s) interact with the bilayer.