Mapping the sites of the lipoprotein lipase (LPL)–angiopoietin-like protein 4 (ANGPTL4) interaction provides mechanistic insight into LPL inhibition

Cardiovascular disease has been the leading cause of death throughout the world for nearly 2 decades. Hypertriglyceridemia affects more than one-third of the population in the United States and is an independent risk factor for cardiovascular disease. Despite the frequency of hypertriglyceridemia, treatment options are primarily limited to diet and exercise. Lipoprotein lipase (LPL) is an enzyme responsible for clearing triglycerides from circulation, and its activity alone can directly control plasma triglyceride concentrations. Therefore, LPL is a good target for triglyceride-lowering therapeutics. One approach for treating hypertriglyceridemia may be to increase the amount of enzymatically active LPL by preventing its inhibition by angiopoietin-like protein 4 (ANGPTL4). However, little is known about how these two proteins interact. Therefore, we used hydrogen–deuterium exchange MS to identify potential binding sites between LPL and ANGPTL4. We validated sites predicted to be located at the protein–protein interface by using chimeric variants of LPL and an LPL peptide mimetic. We found that ANGPTL4 binds LPL near the active site at the lid domain and a nearby α-helix. Lipase lid domains cover the active site to control both enzyme activation and substrate specificity. Our findings suggest that ANGPTL4 specifically inhibits LPL by binding the lid domain, which could prevent substrate catalysis at the active site. The structural details of the LPL–ANGPTL4 interaction uncovered here may inform the development of therapeutics targeted to disrupt this interaction for the management of hypertriglyceridemia.

Lipoprotein lipase (LPL) 2 is a key enzyme for regulating plasma triglyceride levels (1,2). LPL hydrolyzes triglycerides contained within lipoprotein particles in the capillary endothelium to facilitate the passage of free fatty acids into surrounding tissues (3,4). Patients with loss-of-function mutations in LPL have extremely high levels of plasma triglycerides, or hypertriglyceridemia, that can cause complications such as pancreatitis, eruptive xanthomas, and lipemia retinalis (5). Although LPL deficiency is rare, moderate hypertriglyceridemia is common with more than three million cases diagnosed every year in the United States (6). Despite the frequency of hypertriglyceridemia, treatment options are limited to diet and exercise, statins, or fibrates (7). Historically, the development of novel triglyceride-lowering therapeutics was curbed due to inconclusive epidemiological associations between hypertriglyceridemia and cardiovascular risk (8,9). However, recent genetic studies have demonstrated that hypertriglyceridemia is an independent risk factor for cardiovascular disease and should be addressed in the clinic (10 -12).
LPL is first synthesized in the parenchymal cells of heart, skeletal muscle, and adipose tissue. LPL is then secreted and transported to the vascular endothelium where it is anchored to the walls of the vessel lumen (13). Here, LPL is regulated by activators and inhibitors to precisely control the delivery of free fatty acids to either oxidative tissues for catabolism or adipose tissue for storage (1, 14 -16). One family of inhibitory proteins, known as angiopoietin-like proteins (ANGPTL3, -4, and -8), are up-regulated in response to nutritional cues to inhibit LPL in a tissue-specific manner (17). ANGPTL8 is up-regulated in the postprandial state and interacts with ANGPTL3 to cooperatively inhibit LPL within oxidative tissues (17). Inhibition of LPL in oxidative tissues, while simultaneously leaving LPL active in adipose tissue, directs free fatty acids to adipose tissue for storage (15). Conversely, in the fasted state ANGPTL4 is up-regulated to inhibit LPL in the adipose tissue (18). Inhibition of LPL in adipose tissue, while simultaneously leaving LPL active in oxidative tissue, directs free fatty acids to these tissues for catabolism (15).
Human genome-wide association studies have identified individuals with loss-of-function mutations in either ANGPTL3 or ANGPTL4 as having a unique lipid profile that significantly decreases their risk for developing cardiovascular disease (19 -21). Loss-of-function mutations in ANGPTL3 results in hypobetalipoproteinemia-2, which is characterized by low levels of plasma triglycerides, HDL, and LDL cholesterol (22,23). Individuals with hypobetalipoproteinemia-2 appear to have few negative health effects and have motivated the development of novel antisense oligonucleotide therapeutics that silence ANGPTL3 expression (22,24). By comparison, inactivating mutations in ANGPTL4 result in a slightly different lipid profile with low levels of plasma triglycerides, high levels of HDL cholesterol, and no apparent effect on LDL cholesterol (19,25,26). Previous studies using either antibody to silence ANGPTL4 in monkeys or the genetic loss of ANGPTL4 in mice recapitulate low levels of plasma triglycerides seen in humans with loss-of-function ANGPTL4 mutations (25,27). However, both ANGPTL4 knockdown monkeys and knockout mice inexplicably developed severe mesenteric lymphadenitis, or inflammation in the abdominal lymph nodes, when placed on a high-fat diet (25,27,28). Individuals with loss-of-function mutations in ANGPTL4 do not have increased rates of lymphadenopathy, and the animals do not develop this phenotype on standard chow or low-fat diets (25,26).
ANGPTL4 was originally thought to be an adipokine given its robust expression in adipocytes and hepatocytes, but it has since been reclassified with roles in angiogenesis, wound repair, kidney function, tumorigenesis, redox regulation, and energy homeostasis (29). Structurally, ANGPTL4 is composed of a coiled-coil N-terminal domain and a C-terminal fibrinogenlike domain that is selectively cleaved by proprotein convertases in certain tissues (30,31). Adipocytes primarily secrete fulllength ANGPTL4, and hepatocytes secrete the cleaved N-and C-terminal domains (29,32). The N-terminal domain, herein referred to as nANGPTL4, is known to potently inhibit LPL, whereas the C-terminal domain is thought to be involved in other roles like angiogenesis or wound repair (33)(34)(35). Given the diverse roles for ANGPTL4, it is not surprising that complete silencing results in a severe diet-induced pathology (29,36). The disparity between an advantageous phenotype in human variants and a pathological phenotype in animals suggests additional research is needed to understand the precise function of each ANGPTL4 species.
We set out to characterize how nANGPTL4 binds LPL and use this information to block the interaction using a small peptide mimetic of an LPL-binding site. Using both hydrogendeuterium exchange MS (HDX-MS) and site-specific mutagenesis, we determined that nANGPTL4 binds near the active site of LPL at the lid domain (residues 224 -238) and a nearby ␣-helix (residues 89 -102). Moreover, a peptide mimetic of the LPL lid domain is capable of binding nANGPTL4 and protects fulllength LPL from inhibition. Our data suggest nANGPTL4 reversibly inhibits LPL by binding the lid domain to prevent substrate catalysis by the active site.

Hydrogen-deuterium exchange MS reveals four nANGPTL4-binding sites on bovine LPL
To answer the question of how nANGPTL4 mediates inhibition of LPL, we used HDX-MS to identify residues at the inter-face of this protein-protein interaction. The rate of deuterium exchange between amide hydrogens is used to probe protein secondary structure and allosteric conformational changes induced by binding interactions (37,38). We performed HDX-MS on bovine LPL (bLPL) alone, nANGPTL4 alone, and a combination of bLPL and nANGPTL4. Samples were incubated in deuterated buffer for 0, 1, 5, and 10 min, quenched, and digested for immediate analysis by reverse-phase chromatography coupled to MS. Between bLPL and nANGPTL4, we identified 710 unique peptides (Table S1).
The difference in deuterium uptake for bLPL was measured under ϪnANGPTL4 and ϩnANGPTL4 conditions and is represented using a heat map to highlight areas with decreased exchange (dark blue) and increased exchange (red) (Fig. 1A). Peptide sequences that were statistically significantly different (p value Ͻ 0.05) between the ϪnANGPTL4 and ϩnANGPTL4 conditions are represented with percent deuterium uptake curves (Fig. 1A) and colored based on differential deuterium exchange on a computational model of LPL generated by  (Fig. 1B). Because the crystal structure of LPL is not known, we have also included two additional computational models of LPL generated by Hayne et al. (39) in Fig.  S1, A and B. A decrease in deuterium exchange in the complex condition suggests that those peptide sequences are at the binding interface. We identified four potential nANGPTL4-binding sites on bLPL at residues 17-26, 89 -102, 224 -238, and 290 -311 (residue 1 is the first amino acid after signal peptide cleavage, Fig. 1, A  and B). Herein, these sites will be referred to as the N-terminal ␤-turn, lid-proximal helix, lid, and the furin cleavage region, respectively. Interestingly, there were several sites within bLPL that increased in exchange: residues 63-72, 75-83, 103-113, 133-146, and 166 -180 (after signal peptide cleavage) (Fig. S1C). Regions of increased deuterium exchange correspond to more structurally dynamic regions (38). It is unclear at this time whether regions with increased deuterium exchange result from bLPL unfolding as suggested by Mysling et al. (40) or other causes, such as allosteric changes (38,41).
To address the possibility of nANGPTL4-mediated irreversible unfolding of bLPL, we utilized a kinetic assay. Reversible, noncompetitive inhibition can be distinguished from irreversible inactivation by measuring bLPL activity at different concentrations under V max conditions in the presence and absence of nANGPTL4. By plotting V max versus the total concentration of bLPL, the data fit a linear equation with the slope equal to the k cat (V max ‫؍‬ k cat [E total ], where [E total ] is the total enzyme concentration) (42). For a reversible, noncompetitive inhibitor, both the uninhibited and inhibited conditions intersect the origin, and the slope of the inhibited condition is less than the uninhibited control by the quantity 1 ϩ ([I]/(K i ), where [I] is the concentration of inhibitor and K i is the inhibition constant (Fig.  S2A). For an irreversible inhibitor, the slope of the inhibited condition is the same as the uninhibited control. However, the line does not intersect the x axis at the origin, but rather at a position on the x axis that is equivalent to the amount of enzyme that is irreversibly inactivated (Fig. S2B). Plotting V max versus the total concentration of active bLPL shows that both the inhibited and control conditions intersect the origin (Fig.  S2C). Additionally, the slope of the inhibited plot decreased Site-specific interactions between LPL and ANGPTL4 from 1.097 to 0.541. These plots indicate reversible, noncompetitive inhibition (Fig. S2A).
We identified two sites with decreased deuterium exchange in nANGPTL4: residues 15-33 and 95-106 (after signal peptide cleavage) (Fig. 2). Residues 15-33 are in agreement with previous reports detailing a peptide sequence necessary for nANGPTL4 inhibition of LPL, and site 95-106 has not been previously reported (34,35). There were no sites on nANGPTL4 that increased in deuterium exchange in the presence of bLPL.

Human LPL HL chimeric molecules are insensitive to inhibition by nANGPTL4
To validate potential nANGPTL4-binding sites on LPL, we mutagenized human LPL in regions with decreased deu-terium exchange and screened for nANGPTL4 inhibition in vitro (Fig. 3). LPL shares 44% sequence identity with hepatic lipase (HL), which has been putatively shown to be resistant to ANGPTL4 inhibition, although the data surrounding this issue have been contradictory (43)(44)(45). Here, we show that HL is not inhibited by nANGPTL4, even at a 1:4000 molar excess (Figs. 4B and 5A). Therefore, we used the corresponding HL sequences to generate LPL HL chimeras in two regions near the active site: the lid-proximal helix and the lid (Fig. 3). Previous studies have shown that the lipase lids alone dictate substrate specificity, and the chimeric lid mutants of LPL, HL, and endothelial lipase alter both substrate specificity and enzyme kinetics (46 -48). If nANGPTL4 binds the lid of LPL, it could explain why ANGPTL4 specifically inhibits LPL Sites of deuterium exchange were identified using LC-MS-MS after incubation of bLPL, nANGPTL4, or a combination of bLPL and nANGPTL4 in 99.9% deuterated solution for 0, 1, 5, and 10 min. A, heat maps were generated using percent deuterium uptake values for all peptides with a standard deviation of less than 0.2 Da. Deuterium uptake curves are shown for peptides that were statistically significantly different between the bLPL apo (black) and complex state (gray) at each time point. Sites of differential deuterium exchange were determined using unpaired two-tailed Student's t test with cutoff of p Ͻ 0.05. B, computational model of dimeric LPL generated by Hayne et al. (Original model, see Ref. 39). Peptide sequences with differential deuterium exchange in the complex state at every time point are colored blue for decreased exchange and red for increased exchange. For simplicity, one monomer unit is represented as a cartoon model (gray), and the other unit is a surface model (white). Panels are zoomed in on the lid-proximal helix, lid, and furin cleavage region (blue).

Site-specific interactions between LPL and ANGPTL4
rather than HL, as these lid sequences have only 32% identity (43).
Purified LPL and variants, LPL HL lid or LPL HL helix , were assayed for nANGPTL4 inhibition by measuring the rate of substrate hydrolysis in the presence of increasing concentrations of nANGPTL4. Although LPL was sensitive to nANGPTL4 inhibition with a K i of 1.7 Ϯ 0.74 M, both LPL HL chimeric mutants appeared to be resistant to nANGPTL4 inhibition ( Table 1).
The LPL HL helix had a slight sensitivity to nANGPTL4 with an estimated K i of 15 Ϯ 7.0 M (Fig. 4C). Unless 50% enzyme inhibition occurred, the reported K i values could only be estimated. LPL HL lid showed a similar resistance to nANGPTL4 inhibition as HL, such that a K i value could not be calculated (Fig. 4, B and D). These data suggest that both the lid and lid-proximal helix are necessary for nANGPTL4 binding LPL. Although it is possible that the mutations altered the three-dimensional struc- Heat maps of nANGPTL4 and nANGPTL4 ϩ bLPL were generated using percent deuterium uptake values for peptide sequences with a standard deviation of less than 0.2 Da. Percent uptake curves were generated for sites that were statistically significantly different between the nANGPTL4 apo and complex state. Sites of differential deuterium exchange were determined using unpaired two-tailed Student's t test with cutoff of p Ͻ 0.05. Figure 3. Human LPL and HL construct schematic. LPL (light gray) and HL (dark gray) sequences were aligned and defined according to potential nANGPTL4binding sites identified by our HDX-MS experiments using bLPL. The corresponding sequences in human LPL and HL are defined as the lid-proximal helix (residues 87-100 or 101-115) or the lid (residues 216 -239 or 232-255). Chimeric mutants were generated using overlap extension PCR.

Site-specific interactions between LPL and ANGPTL4
ture of the lipase, both LPL HL chimeric molecules displayed no obvious catalytic defects (Table 1).
We next tested whether the furin protease cleavage region was necessary for nANGPTL4 inhibition (residues 288 -309) (Fig.  S3A). Furin protease recognizes the consensus sequence "RXX(K/ R)" present in LPL at residues 294 -297, resulting in inactive Nand C-terminal cleavage fragments (49,50). In HL, the furin consensus sequence is disrupted by a proline residue and protected from proteolytic cleavage in this region (residues 314 -334).
We first exchanged LPL residues 288 -309 for the corresponding HL sequence (Fig. S3A). As expected, this mutation abolished furin cleavage in LPL but also abolished LPL enzymatic activity (Fig. S3, B and C). It is important to note that the proline residue in the corresponding HL sequence at position 294 may compromise the three-dimensional structure of LPL and possibly result in catalytic inactivation. Therefore, we split the full sequence into two parts: 288 -299 and 300 -309 (Fig. S3A). Similar to LPL HL288 -309 , LPL HL288 -299 was not enzymatically active. Although LPL HL300 -309 does not contain the R294P mutation, its enzymatic activity could not be detected either (Fig.  S3B). Western blot analysis revealed that nearly the entire population of LPL HL300 -309 was cleaved by furin protease (Fig. S3C) (49). Because we were unable to measure lipase activity for any of the LPL HL furin cleavage chimeras, we cannot determine whether residues 288 -309 are directly involved in nANGPTL4 binding.

Human HL LPL chimeric proteins are sensitive to inhibition by nANGPTL4
To further validate sequences needed for nANGPTL4 inhibition of LPL, we aimed to "sensitize" HL to nANGPTL4 inhibition by generating human HL chimeric mutants containing the corresponding LPL sequences for the lid-proximal helix and the lid (Fig. 3). Similar to LPL HL chimeric mutants, all HL LPL chimeric mutants were generated using overlap extension PCR, whereby each sequence of interest was exchanged for the corresponding LPL sequence (Fig. 3).
HL LPL lid and HL LPL helix were kinetically similar to HL with average V max and K m values of 2.0 mol/s and 0.6 M, respectively, using the DGGR lipase substrate (Table 1). However, unlike HL, HL LPL helix and HL LPL lid were sensitive to nANGPTL4 inhibition with an estimated K i of 32.7 Ϯ 5.6 and 45.5 Ϯ 2.4 M, respectively (Fig. 5, B and C). HL LPL lid ϩ helix appeared to be the most sensitive to inhibition with an estimated K i of 22 Ϯ 8.7 M (Fig. 5D). Taken together, these data show that the LPL lid and the lid-proximal helix provide binding sites for nANGPTL4.

LPL lid peptide blocks nANGPTL4-mediated inhibition of LPL
Given that mutating the LPL lid abolished nANGPTL4 inhibition, we wanted to know whether the LPL lid domain alone could compete with full-length LPL for binding nANGPTL4. Therefore, we synthesized a peptide composed of the LPL lid sequence (CNIGEAIRVIAERGLGDVDQLVKC). The native LPL lid was circularized via a disulfide bond between two cys-

Site-specific interactions between LPL and ANGPTL4
teine residues at the termini of the lid motif. The synthetic lid peptide was tested in cyclic form (disulfide-bonded) or linear form (NEM-capped). Correct synthesis and circularization of the peptide were verified by MS, as shown in Fig. S4, A and B. We tested whether the peptide could prevent LPL inhibition by nANGPTL4 by incubating the inhibition reaction with increasing concentrations of either cyclic and linear NEM-capped lid peptide. Interestingly, the cyclic peptide protected full-length LPL from nANGPTL4 inhibition in a dose-dependent manner, whereas the linear NEM-capped peptide did not (Fig. 6, A and  B). Presumably, the cyclic peptide adopts a structure similar to the native lid domain. Isothermal titration calorimetry (ITC) was used to confirm that the cyclic lid peptide binds nANGPTL4 with micromolar affinity. These data suggest that two lid peptides bind each nANGPTL4 tetramer with a K d of ϳ57 M (Fig. 6C). Although the ITC shows that the lid peptide binds nANGPTL4 with relatively low affinity, these data confirm that the LPL lid is an important site for nANGPTL4 binding.

Discussion
Hypertriglyceridemia is an independent risk factor for cardiovascular disease with relatively limited treatment options and, therefore, an unmet need for new therapies (6,10). Given LPL has a direct influence on plasma triglyceride levels, it is a promising target for novel triglyceride-lowering therapeutics. One approach for targeting LPL is to prevent its inhibition by ANGPTL4. Previous reports describing ANGPTL4 knockout animals demonstrated significantly reduced triglyceride levels with little negative side effects under standard chow diets (25,27). However, the safety of ANGPTL4 knockdown has been called into question due to severe lymphadenopathy observed in animals placed on high-fat diets (27,28). Intriguingly, humans with loss-of-function mutations in ANGPTL4 do not appear to be at an increased risk for lymphadenopathy (25,26). Therefore, additional research is needed to assess whether specifically blocking ANGPTL4 inhibition of LPL can be a viable treatment for hypertriglyceridemia. Our goal was to study the interaction of nANGPTL4 with LPL to inform the development of specific inhibitors for this interaction.

Table 1 Enzyme kinetics for each LPL and HL chimeric variant
Michaelis-Menten curves were generated for each lipase in the presence of increasing concentrations of nANGPTL4 over multiple fluorescent substrate concentrations. LPL and LPL HL chimeric mutants have similar V max and K m values. LPL HL lid and LPL HL helix have an increased resistance to nANGPTL4 inhibition, relative to LPL. HL and HL LPL chimeric mutants have similar V max and K m values, except HL LPL lid ϩ helix has a significantly lower V max than HL. HL LPL chimeric mutants have increased sensitivity to nANGPTL4, relative to HL. Average values of at least three biological replicates Ϯ S.D. RFU indicates relative fluorescent units.

Site-specific interactions between LPL and ANGPTL4
The exact mechanism for LPL inhibition by ANGPTL4 has been controversial. nANGPTL4 was first shown to irreversibly inhibit LPL by converting active dimers into inactive monomers (52,53). Our previous data support an alternative model, whereby nANGPTL4 inhibits LPL in a reversible, noncompetitive mechanism (54,55). The difference in observed mechanisms of inhibition may depend on the initial stability of the LPL in the assay. Purified LPL is notoriously unstable in standard buffer conditions and is prone to thermal denaturation (54). Therefore, many in vitro LPL activity assays and purification protocols use a bile acid, deoxycholate, to stabilize LPL (40, 50, 54 -58). In the presence of deoxycholate, nANGPTL4 inhibits LPL by a reversible, noncompetitive mechanism (40,54). Because LPL is known to have enhanced thermal stability within the biological fluids that contain bile acids, like serum and milk, we used them to better mimic physiological conditions in our kinetic assays (58,59).
We have identified both the lid and lid-proximal helix as key nANGPTL4-binding sites. Removing these sites from LPL abolished nANGPTL4 binding and protected the enzyme from inhibition. Alternatively, cloning these sites into HL sensitized this resistant lipase to nANGPTL4 inhibition. Our data suggest nANGPTL4 binds the lid and lid-proximal region of LPL, which may prevent proper lid function and inhibit substrate catalysis. There are conflicting reports whether nANGPTL4 can inhibit LPL in the presence of natural chylomicron or very LDL substrates (58,60). Here, the synthetic substrate DGGR was needed to enable precise measurements of LPL inhibition kinetics. Natural lipoprotein substrates have different properties than DGGR, so our work does not address this conflict.
Our HDX-MS data also identified a third potential binding site for nANGPTL4 on LPL at residues surrounding its furin cleavage site (residues 288 -309). LPL is inactivated by furin protease cleavage into the N-and C-terminal domains at the consensus sequence "RAKR" (residues 294 -297). Although we were unable to generate enzymatically active chimeric mutants at this site, we have shown that furin-resistant LPL point mutants, R297N and R297N/S298C, are sensitive to nANGPTL4 inhibition (50). These data suggest residues Arg-297 and Ser-298 may not be necessary for nANGPTL4-mediated inhibition of LPL. Recent reports have shown that ANGPTL4 enhances furin cleavage of LPL in adipocytes (49). Given that ANGPTL4 is also cleaved by the furin protease, it is possible that ANGPTL4 binds LPL intracellularly to recruit the furin protease for cleavage of both ANGPTL4 and LPL.
Finally, we demonstrate that a small peptide containing the LPL lid sequence can protect full-length LPL from nANGPTL4 inhibition. Although this interaction is not yet sufficiently strong, it serves as a starting point for development of potent peptide inhibitors for nANGPTL4 that specifically abolish LPL inhibition and leave both full-length and cleaved forms of ANGPTL4 intact to perform their other functions.

Molecular cloning
Human LPL, HL, and chimeric mutants were cloned into pCDNA5/FRT/TO vectors (ThermoFisher Scientific) containing lipase maturation factor 1 (LMF1) with an internal ribosome entry site (IRES) for co-expression. All constructs contain a 6ϫ polyhistidine-tag on the C terminus, except the WT HL

Site-specific interactions between LPL and ANGPTL4
construct has no tag. All chimeric mutants were generated using overlap extension PCR. Cloning of pET16B nANGPTL4 was previously described (54). All plasmid sequences are available upon request.

Cell culture and protein expression
LPL, HL, and chimeric mutant plasmids were stably integrated into FlpIn TM T-REx TM HEK293 cell lines (Thermo-Fisher Scientific) using the manufacturer's instructions. Each stable cell line was maintained in Dulbecco's modified Eagle's medium, 10% fetal bovine serum, 1% penicillin/streptomycin, and 1% L-glutamine (growth media). At 70% confluence, cells were given Dulbecco's modified Eagle's medium, 1% fetal bovine serum, 1% penicillin/streptomycin, 1% L-glutamine, 10 units/ml heparin, and 2 g/ml tetracycline (expression media) to induce the expression of lipase. Expression media for all LPL species and HL was collected and flash-frozen in liquid nitrogen every 24 h for 6 days. Expression media for all HL variants was collected every 48 h, bound to nickel-Sepharose Excel TM resin (GE Healthcare), washed, and flash-frozen in liquid nitrogen for Ϫ80°C storage until further purification as described below.

Protein purification
bLPL-bLPL was purified from fresh bovine milk as described previously (51). Briefly, chilled bovine milk was centrifuged and filtered using glass wool to remove milk fat. Solid NaCl was added to the skim milk to 0.34 mol/liter, and bLPL was purified using heparin-Sepharose 6 Fast Flow resin (GE Healthcare).
HL-Filtered expression media loaded onto two tandem 1-ml HiTrap heparin-Sepharose high-performance columns (GE Healthcare) were washed with 20 ml of HL heparin buffer (20 mM Na 2 HPO 4 , pH 7.2, 500 mM NaCl, 10% glycerol) and eluted along a linear salt gradient from 500 to 2000 mM NaCl in HL heparin buffer. Fractions containing HL were pooled, concentrated using Amicon ultracentrifugal units (Millipore), aliquoted, and flash-frozen for Ϫ80°C storage until use.

Quantification of active lipase
Quantification was performed as described previously (55). Briefly, ActivX TM TAMRA-FP serine hydrolase probe (ThermoFisher Scientific) was used to label active LPL and HL variants where bovine LPL of known concentration was used as a protein standard. The relative purity of all purified protein can be seen by both Coomassie stain and TAMRA-FP labeling in Fig. S5, A-D.

HDX-MS
HDX labeling-bLPL and nANGPTL4 alone were diluted in 50 mM sodium phosphate buffer, pH 7.2, to 8 and 16 M, respectively, and a condition with 8 M bLPL and 16 M nANGPTL4 was created prior to deuterium labeling. An automated HDX labeling LEAP robot was used for deuterium labeling. Briefly, 5 l of each protein sample were diluted with 45 l of 50 mM sodium phosphate in H 2 O for time 0 injections or 45 l of 50 mM sodium phosphate in D 2 O (99.9% Sigma) for deuterium uptake injections at 1, 5, and 10 min. Deuterium exchange was quenched with 45 l of pre-chilled 0.3% formic acid in water with 1 mM TCEP at the corresponding times. Quenched samples were subjected to on-line pepsin digestion using Waters Enzymate BEH pepsin column (2.1 ϫ 30 mm ϫ 5 m) at 200 l/min at 20°C for 2 min (pH 2.5, 0.05% formic acid in water). Sample processing was managed within Waters HDX Acquity Module (Waters).

Site-specific interactions between LPL and ANGPTL4
LC-MS-MS-Downstream LC-MS-MS was acquired in duplicate using the nanoAcquity UPLC system (Waters) coupled to a Waters Synapt G2 high-resolution accurate mass tandem mass spectrometer via an electrospray ionization source. The sample was first trapped on a VanGuard 2.1 ϫ 5-mm BEH C18 1.7-m trapping column (40 l/min at 5% v/v water/acetonitrile, 0.1% formic acid). Analytical separation was performed using a 1.7-m Acquity BEH130 C18 1 ϫ 100-mm column (Waters Corp.) using a linear gradient from 5% acetonitrile, 0.1% formic acid to 35% acetonitrile, 0.1% formic acid over 7 min at 40 l/min. Trapping and analytical separation were all performed at 1.0°C with the HDX manager. Data collection on the Synapt G2 mass spectrometer was performed in an ion-mobility enabled data-independent acquisition mode of acquisition alternating between a 1.0-s low energy scan from m/z 50 to 2000 followed by a 1.0-s high energy scan from m/z 50 to 2000. Every 30 s, a low energy lock-mass scan of 400 fmol/l Glu-Fibronectin peptide was performed.
Data analysis-All nondeuterated, time 0, data files were imported and searched within ProteinLynx Global Server (version 2.5.2) against a SwissProt_Human database containing bLPL and ANGPTL4 sequences. Nonspecific enzyme rules were selected with Ϯ10 ppm mass accuracy tolerances and 5% retention time relative standard deviation criteria. ProteinLynx Global Server peptide score of at least 6.5 was included in the final analysis for a final peptide false discovery rate of Ͻ 2.0%. DynamX HDX Data Analysis Software (version 2.0, Waters) was used to calculate deuterium uptake for all identified peptide sequences. To identify residues of differential uptake on bLPL as a function of nANGPTL4, or vice versa, the calculated uptake values were subjected to an unpaired two-tailed Student's t test calculation at each time point for every condition. Only sites with significant differential uptake with a p value Ͻ 0.05 were selected for manual interrogation to ensure correct identification of peaks with no interference from neighboring peaks and to assess the global uptake distribution across all time points. Deuterium uptake was represented in a heat map using percent deuterium uptake values for all peptides with a standard deviation of less than 0.2 Da at each time point. Heat maps for bLPL alone and bLPL ϩ nANGPTL4 were generated using 393 and 350 unique peptides, respectively. Heat maps for nANGPTL4 alone and nANGPTL4 ϩ bLPL were generated using 520 and 505 unique peptides, respectively. All raw data used in our analysis can be found in Table S1.

Peptide production
Peptide synthesis was carried out using solid-phase peptide synthesis on a Biotage Initiatorϩ Alstra microwave peptide synthesizer using rink amide ChemMatrix resin (Biotage). Peptidic scaffolds were synthesized from the C to the N terminus with a C-terminal amide functional group using standard Fmoc (fluorenylmethoxycarbonyl) chemistry. The building blocks are composed of commercially available N-␣-Fmoc-L-amino acids. Briefly, the resin was initially wetted and swollen with 1:1 mixture of dichloromethane and DMF. Fmoc deprotection was achieved using incubation of the resin with 20% (v/v) piperidine/DMF for 3 and 10 min each, with a DMF wash at the end of each incubation step. The N-␣-Fmoc-L-amino acids (5 eq) were coupled using 1-[bis(dimethylamino)methylene]-1H-1,2,3-triazolo[4,5-b]pyridinium 3-oxide hexafluorophosphate (5 eq) as the coupling agent and diisopropylethylamine (10 eq) as the base. Coupling steps were carried out at 75°C for 5 min each using microwave irradiation. After the final deprotection to obtain a primary amine at the N terminus of the synthesized peptide, the resin was washed with dichloromethane and dried. The synthesized peptide was cleaved from the resin using 94 -95% (v/v) TFA, 2.5% (v/v) water, and 2.5% (v/v) triisopropylsilane while incubating at 37°C for 1 h followed by incubation at room temperature for 30 min. Resin-cleaved peptide was precipitated by the dropwise addition of the cleavage reaction mixture into cold diethyl ether. The precipitate was separated by centrifugation, and the supernatant was decanted. Dried peptide was dissolved in DMSO and diluted with a mixture of 50% (v/v) acetonitrile/water. The crude peptide was then purified using semi-preparative HPLC. Each fraction was analyzed using the UV trace at 220 and 280 nm and LC-MS. Fractions containing pure peptide were collected and lyophilized. The solid peptide was dissolved in DMSO and then diluted to a peptide concentration of 200 M using 0.1 M ammonium bicarbonate, pH 8.0. This step allowed it to circularize via intramolecular disulfide bond formation. Completion of circularization was verified by LC-MS, and the peptide was lyophilized and stored at Ϫ20°C until used.
NEM capping of lid peptide-LPL lid peptide purified as described above was dissolved in DMSO and diluted to a final concentration of 0.4% DMSO and 0.1 M TCEP using 1ϫ PBS, pH 7.8. The solution was incubated at 37°C for 15 min to reduce disulfide bonds. Following TCEP reduction, 0.1 M NEM was added to the reaction and incubated at 37°C for 60 min. The capped peptide was recovered by HPLC, verified by LC-MS, and lyophilized for storage at Ϫ20°C.
Inhibition of LPL, HL, and chimeric mutants by nANGPTL4 -Inhibition assays using purified lipase and lipase variants by nANGPTL4 were carried out essentially as described (2). LPL and bLPL are equivalently inhibited by nANGPTL4. Briefly, nANGPTL4 was added to dilute LPL, HL, or chimeric mutants at a final volume of 70 l/reaction. The reaction was initiated by the addition of 30 l of varying concentrations of fluorescent substrate DGGR (Sigma) in anzergent 3-16 (Affymetrix), to a final volume of 100 l. Samples were shaken in a Spectramax M5 plate reader at 37°C for 5 s, and substrate hydrolysis was measured by fluorescence excitation at 529 nM, emission at 600 nM, and a filter of 590 nm.

Site-specific interactions between LPL and ANGPTL4
using simultaneous nonlinear regression with the program Mathematica (Wolfram Research).
Inhibition of LPL by nANGPTL4 in the presence of an LPL lid peptide-LPL inhibition by nANGPTL4 was measured using the DGGR fluorescent assay described above in the presence of the LPL lid peptide. nANGPTL4 (0 or 0.4 M) was incubated with increasing concentrations of cyclic lid peptide (0, 20, 50, or 100 M) or linear NEM-capped peptide (0, 30, 90, or 160 M) for 10 min at room temperature. After incubation, the mixture was added to dilute LPL in a final volume of 70 l/reaction and incubated for an additional 5 min at room temperature. LPL activity was measured upon the addition of 30 l of DGGR for a final reaction volume of 100 l. Final assay buffer concentrations were 20 mM Tris, pH 8.0, 150 mM NaCl, 1 mM deoxycholate, 0.2% fatty-acid free BSA, 0.01% anzergent 3-16, 10 M DGGR, 0.8% DMSO, 2.5 nM LPL, 0 or 0.4 M nANGPTL4, and 0, 20, 50, or 100 M cyclic lid peptide or 0, 30, 90, 160 M linear NEM-capped peptide.
Isothermal titration calorimetry-Lyophilized lid peptide was resuspended in DMSO and diluted in nANGPTL4 dialysis buffer (1ϫ PBS, pH 7.8) to a final concentration of 3087 M and 4% DMSO. Purified nANGPTL4 was at a concentration of 338 M and brought to 4% DMSO to exactly match the peptide buffer. Isothermal titrations were performed using the MicroCal Auto-iTC200 (Malvern Panalytical). Injections of 1.5 l of peptide were added in 180-s intervals at 10°C. Heat of dilution was measured by titrating the peptide into a blank solution and subtracted from our data before a one-site curve fitting was performed using MicroCal Auto-iTC200 software. Stoichiometry (N), association, and dissociation constants (K a and K d ) and enthalpy change (⌬H) were obtained directly from the data.