Semisynthesis and segmental isotope labeling of the apoE3 N-terminal domain using expressed protein ligation.

Apolipoprotein E (apoE) is an exchangeable apolipoprotein that functions as a ligand for members of the LDL receptor family, promoting lipoprotein clearance from the circulation. Productive receptor binding requires that apoE adopt an LDL receptor-active conformation through lipid association, and studies have shown that the 22 kDa N-terminal (NT) domain (residues 1–183) of apoE is both necessary and sufficient for receptor interaction. Using intein-mediated expressed protein ligation (EPL), a semisynthetic apoE3 NT has been generated for use in structure-function studies designed to probe the nature of the lipid-associated conformation of the protein. Circular dichroism spectroscopy of EPL-generated apoE3 NT revealed a secondary structure content similar to wild-type apoE3 NT. Likewise, lipid and LDL receptor binding studies revealed that EPL-generated apoE3 NT is functional. Subsequently, EPL was used to construct an apoE3 NT enriched with 15N solely and specifically in residues 112–183. 1H-15N heteronuclear single quantum correlation spectroscopy experiments revealed that the ligation product is correctly folded in solution, adopting a conformation similar to wild-type apoE3-NT. The results indicate that segmental isotope labeling can be used to define the lipid bound conformation of the receptor binding element of apoE as well as molecular details of its interaction with the LDL receptor.

CaCl 2 , 2 mM MgSO 4 , and 50 µg/ml ampicillin. At OD 600 = 0.6, the culture was induced with 2 mM isopropyl thiogalactopyranoside. After 6 h at 30°C, bacteria were pelleted by centrifugation at 8,000 g for 15 min and the culture supernatant collected, concentrated by ultrafi ltration, and subjected to heparin affi nity chromatography and semipreparative C 8 reversed-phase HPLC. Where indicated, apoE3 NT was expressed in M9 medium supplemented with 15 NH 4 Cl (99% purity; Cambridge Isotope Laboratories) as the sole nitrogen source. All subsequent processing and ligation steps were identical to those described above for the unlabeled protein. Isolated apoE(1-183) was then subjected CNBr digestion ( 13 ). ApoE3  was dissolved in 80% formic acid (5 mg/ml). CNBr was added to achieve a CNBr/methionine ratio >100 and the sample incubated in an oxygen-free environment for 24 h in the dark. Reactions were quenched by the addition of a 10-fold excess of deionized water and lyophilized to dryness. This procedure was repeated three times to remove residual formic acid. Complete digestion resulted in enrichment in apoE(112-183) containing the desired NT cysteine. ApoE(112-183) was further purifi ed by semipreparative C 8 reversed-phase HPLC on a Perkin-Elmer Series 200 chromatography system.

Analytical methods
Protein purity and/or ligation reaction progress was monitored by SDS-PAGE using either a 10-20% acrylamide gradient tricine gel or a fi xed 18% acrylamide slab gel. Gels were stained with Amido Black 10B. MALDI-TOF mass spectrometry was performed on a Bruker Daltronics autofl ex LRF as described previously ( 9 ). Nondenaturing gradient gel electrophoresis was performed on 4-20% acrylamide gradient Tris-glycine slab gels (Invitrogen). Samples were electrophoresed for 18-20 h at 20 mA and stained with GelCode Blue (Pierce Chemical Co.).

Circular dichroism spectroscopy
Far UV circular dichroism (CD) spectroscopy analysis was performed on a Jasco 810 spectropolarimeter. Scans were recorded between 185 and 260 nm in 20 mM sodium phosphate and 20 mM DTT, pH 7.4, at a protein concentration of 0.5 mg/ml in a 0.2 mm path length cuvette. Secondary structure content was calculated as previously described ( 16 ). For guanidine HCl-induced denaturation studies, apoE NT samples (0.2 mg/ml) were incubated overnight at a given denaturant concentration to attain equilibrium, and ellipticity was measured at 222 nm in a 1 mm path length cuvette.

LDL receptor binding assay
For LDL receptor binding studies, 2.5 µg of a soluble human LDL receptor (sLDLR; NT residues 1-699) ( 17 ) in 20 mM Tris (pH 7.2), 2 mM CaCl 2 , and 90 mM NaCl was incubated with 1.5 µg Trp-null apoE3 NT labeled on Cys112 with the fl uores-portant gaps in our knowledge of this system include a detailed understanding of the receptor-active, lipid-associated conformation of the apoE ligand as well as the molecular details of its binding interaction with the receptor.
In an effort to pursue high-resolution studies of lipidbound apoE, we employed intein-mediated expressed protein ligation (EPL) (6)(7)(8). This approach, which allows for increasingly detailed analysis of structural elements within proteins, involves reconstruction of an intact protein from fragments generated separately. In this study, we used EPL to produce a semisynthetic apoE3-NT comprising the fi rst 183 residues of apoE. The results reveal an effi cient and novel approach for elucidation of the conformation of lipid-associated apoE3 NT that potentially provides a means to characterize molecular details of apoE-LDL receptor interactions.

Preparation of apoE(1-111)
Taking advantage of the lone cysteine at position 112 in human apoE3 NT together with the requirement that the desired EPL reaction includes a substrate polypeptide possessing an NT cysteine, apoE(1-111) was cloned and expressed as a Saccharomyces cerevisiae -derived vacuolar ATPase intein domain and chitin binding domain fusion protein using the pTYB1 vector (New England Biolabs) as previously reported ( 9 ). To facilitate optimal intein-mediated fusion protein cleavage ( 10 ), valine 111 was mutated to alanine using the QuikChange method (Stratagene) according to the manufacturer's instructions. Expression and purifi cation procedures for apoE(1-111) followed standardized protocols previously established for generating intein-mediated thioester-adducted proteins ( 11 ). Briefl y, saturated overnight cultures of ER2566 Escherichia coli cells harboring the vector encoding the apoE(1-111) fusion protein were inoculated into 2xYT media containing 50 µg/ml ampicillin, grown to OD 600 = 0.6, and induced with 1 mM isopropyl thiogalactopyranoside. After 6 h at 30°C the cells were pelleted by centrifugation (8,000 g for 15 min), solubilized with buffer A (20 mM Tris, 150 mM NaCl, and 1 mM EDTA, pH 8.0) containing 1% Triton X-100, and stored at Ϫ 20°C. Dissolved cell pellets were combined, passed through a microfl uidizer, sonicated, and centrifuged at 12,000 g for 20 min. Isolated clarifi ed cell extract was passed over a chitin bead column preequilibrated with buffer A containing 1% Triton X-100. The column was washed with 10 column volumes of detergent-free buffer A and fusion protein cleavage induced by addition of 2-mercaptoethanesulfonic acid (MESNA) to a fi nal concentration of 60 mM. Flow was arrested for 16-24 h at 22°C and eluted with two bed volumes of buffer A containing 5 mM MESNA. The sample was dialyzed against deionized water, lyophilized, and stored at Ϫ 20°C. ApoE(1-111)-MESNA was further purifi ed by semipreparative C 8 reversed-phase HPLC on a Perkin-Elmer Series 200 HPLC.

EPL
ApoE(1-111)-MESNA ( ‫ف‬ 13 kDa) was produced in high yield ( Fig. 2 ), and the isolated fragment was highly soluble and stable in a range of aqueous buffers at high concentration. The purity of the ‫ف‬ 8 kDa apoE(112-183) ligation substrate fragment is also shown. Upon incubation of these fragments, a time-dependent appearance of a product band with electrophoretic mobility identical to that of control apoE3 NT was observed. When excess apoE(112-183) was present, ligation efficiency increased to ‫ف‬ 70-80% under these conditions. Similar yields have been reported for ligations that employ two similarly sized fragments >8 kDa ( 24 ). Ligation progress was monitored by SDS-PAGE, and the reaction was quenched by dialysis. Following ligation, the sample was processed to remove unreacted substrate fragments. A combination of heparin affi nity chromatography and reversed-phase HPLC yielded a purifi ed semisynthetic apoE3 NT ( Fig. 2 B).

Characterization of semisynthetic apoE3 NT
Mass spectrometry analysis of the ligation product (21,278 Da) and WT recombinant apoE3 NT (21,197 Da) agree well with the calculated masses of these proteins based on their amino acid composition (21,270 Da and 21,192 Da, respectively). To examine whether the ligation product folds in solution to adopt a conformation similar to native apoE3 NT, far-UV CD spectroscopy was performed ( Fig. 3 ). Control and ligated apoE3 NT spectra were overlapping with dual minima at 208 and 222, corresponding to an ␣ -helix content of 64%, consistent with earlier reports on apoE3 NT ( 12,25,26 ). Guanidine HClinduced denaturation studies ( Fig. 4 ) revealed that control apoE3 NT and ligation product were stable in solution, yielding native to unfolded transition midpoints in the range of 2-2.5 M guanidine HCl. Based on this, we conclude that, following ligation, EPL generated apoE3 NT folds in solution to adopt a stable conformation characteristic of apoE3 NT.

LDL receptor binding activity
Control WT apoE3 NT and semisynthetic apoE3 NT were complexed with DMPC to generate rHDL ( 18 ). Native PAGE analysis revealed comigration of WT and ligated apoE3 NT rHDL ( Fig. 5 ). To examine the relative ability of semisynthetic apoE3 NT rHDL to serve as an LDL receptor ligand, a fl uorescence resonance energy transferbased competition binding assay ( 17 ) was performed. As shown in Fig. 6 , both WT apoE3 NT•DMPC and EPL apoE3 NT•DMPC induced a concentration-dependent decrease in AEDANS fl uorescence intensity (excitation 280 nm) that is characteristic of binding to sLDLR. Thus, WT apoE3 NT and EPL apoE3 NT rHDL are able to compete with AEDANS Trp-null apoE3 NT•DMPC for binding to sLDLR. cent probe N -(iodoacetyl)-N -(5-sulfo-1-naphthyl) ethylenediamine (AEDANS) and complexed with dimyristoylphosphatidylcholine (DMPC) to form reconstituted HDL (rHDL) particles as above ( 18 ). Interaction between AEDANS-Trp null apoE3 NT•DMPC and sLDLR was detected by fl uorescence resonance energy transfer between excited Trp residues in the sLDLR and the AEDANS moiety covalently attached to Trp null apoE3 NT•DMPC ( 17 ). Unlabeled wild-type (WT) apoE3 and EPL apoE3 NT were complexed with DMPC (5:1 lipid to protein weight ratio) as described by Weers, Narayanaswami, and Ryan ( 18 ) and employed as competitor ligands in receptor binding assays. In this assay, the ability of the unlabeled competitor ligands to bind sLDLR is measured as a function of changes in AEDANS fl uorescence intensity (excitation 280 nm; emission 470 nm) after 10 min of incubation at 25°C (fi nal sample volume 300 µl). Fluorescence measurements were made on a Perkin-Elmer LS50b luminescence spectrometer.  15 N and 1 H dimension, respectively. Data were processed using NMRPIPE ( 19 ) and analyzed using NMRVIEW ( 20 ). Spectral assignment of Gly residues was achieved by comparison of the chemical shifts of Gly resonances in uniformly 15 N-labeled apoE3 NT ( 21 ).

Semisynthesis rationale
The goal of the present semisynthesis strategy was to generate an apoE3 NT protein specifi cally isotope labeled in the receptor recognition segment of the protein (i.e., residues 130-172). The overall design strategy is outlined in Fig. 1 . Since EPL requires an NT cysteine nucleophile to initiate protein ligation ( 11,22 ), the ε 3 isoform of apoE NT, which contains a lone cysteine at position 112, is well suited for EPL. To begin, we independently generated a C-terminal thiol-stabilized apoE(1-111) fragment and an apoE(112-183) fragment. This choice eliminated the need for mutagenic introduction of a potentially disruptive cysteine elsewhere in the protein and employed an apoE fragment (i.e., residues 112-183) that encompasses the entire LDL receptor recognition sequence. To generate the apoE(1-111) fragment, thiol-induced, intein-mediated cleavage of a protein chimera was performed ( 9 ). On the other hand, CNBr cleavage of a modifi ed intact apoE3 NT generated the apoE(112-183) fragment. Ligation of these fragments to reconstruct intact apoE3 NT, with a peptide bond between residues 111 and 112 and without extraneous tags or tails, involved incubating the fragments in the NT are missing from the spectrum of segmental isotopelabeled apoE3 NT. This result is anticipated because the apoE3 NT sequence contains three Gly in the segment from 1-111 and seven in the segment from 112-183. Figure 8 also depicts representative examples of Gly resonances that are virtually coincident in the two samples (e.g., Gly173 and Gly182) and resonances that are slightly shifted between the two samples (e.g., Gly113 and Gly127). The high degree of resonance overlap between segmental and uniformly 15 N-labeled apoE3 NT confi rms the expected relatedness of the two proteins and documents the reduced complexity afforded by segmental isotope labeling. A similar simplifi cation is observed in the region corresponding to the Gln side chain amino group spectral pairs ( 1 H ppm 6.0-8.0 to 15 N ppm 108-113) due to a reduction in the number of Gln from 18 in uniformly 15 Nlabeled samples to fi ve in segmentally 15 N-labeled apoE3 NT. Whereas spectral crowding makes determination of the 18 spectral pairs diffi cult, the observation of doublet pairs for the fi ve glutamines in segmentally labeled apoE3 NT is evident.

DISCUSSION
The objective of this study was to employ EPL methods to generate a functional apoE3 NT for structural analysis of lipid-bound apoE by NMR spectroscopy. This approach

NMR studies of segmental isotope-labeled apoE3 NT
A goal of apoE3 NT semisynthesis is the generation of a segmental isotope-labeled protein for heteronuclear multidimensional NMR spectroscopy. Previous studies in our laboratory have shown that a 58-residue peptide fragment of apoE3 NT corresponding to residues 126-183 fails to adopt secondary structure in buffer yet forms an extended ␣ -helix in the presence of 50% (v/v) trifl uoroethanol cosolvent or the micelle forming lipid dodecylphosphocholine ( 27,28 ). In the present approach, we sought to extend this work by examining the structure of this region of the protein within the context of an intact NT domain. By culturing E. coli in minimal media containing 15 NH 4 Cl as the sole nitrogen source, effi cient labeling was achieved ( 12 ). 15 N-labeled WT apoE3 NT was then used as starting material for production of labeled apoE(112-183). Subsequent ligation with MESNA-apoE(1-111) resulted in generation of a semisynthetic apoE3 NT that harbors 15 N isotope only in residues 112-183. For comparison, two-dimensional 1 H- 15 N HSQC spectra were obtained for uniformly 15 N-labeled and segmental 15 N-labeled apoE3 NT at pH 7.4 ( Fig. 7 ). In this HSQC experiment, the observed cross-peaks correlate the chemical shift of amide protons with the amide nitrogen for each of the 15 N-labeled amino acids in the polypeptide chain. From this data, it can be seen that segmental isotope-labeled semisynthetic apoE3 NT resonances coincide with a subset of resonances manifest by uniformly 15 N-labeled apoE3 NT, thereby affording signifi cant simplifi cation of the NMR spectrum for EPLgenerated apoE3 NT. The HSQC resonances for segmental isotope-labeled apoE3 NT were well dispersed with minimal spectral crowding outside the central region of the spectrum. In contrast, uniformly 15 N-labeled apoE3 NT contained a much higher resonance density arising from overlapping cross-peaks, consistent with a greater number of labeled residues present in this sample. Based on peak intensity, clustering, and their position in the spectrum, it is likely that peaks around 1 H ppm 8.0 and 15 N ppm 125-132 ppm are the result of contaminating denatured protein. Further examination of the HSQC signature using complementary NMR experimentation may clarify the nature of these aberrant peaks.
Examination of overlaid HSQC spectra in the 1 H ppm 7.6-9.2 and 15 N ppm 104-111 region highlights differences between the segmental and uniformly 15 N-labeled proteins ( Fig. 8 ). Assignment of glycine resonances, which predominate this spectral region, reveals that three of the 10 Gly cross-peaks present in uniformly 15 N-labeled apoE3 Fig. 1. Design strategy for EPL-mediated semisynthesis apoE3 NT. ApoE3 NT protein fragments designed for EPL were generated as separate recombinant preproteins, cleaved by an intein-mediated reaction (residues 1-111) or CNBr chemical cleavage (residues 112-183) and ligated to produce a semisynthetic apoE NT. M denotes MESNA, used to induce intein-mediated cleavage of the apoE(1-111)•intein•chitin binding domain (CBD) fusion protein. fragment encompassing residues 112-183. In the fi nal ligation product, two amino acids substitutions were introduced to facilitate the design. Val111 was converted to Ala to promote intein-mediated cleavage of the precursor chimera, and Met125 was converted to Ala to prevent unwanted CNBr cleavage at that site.
Characterization studies of the ligation product provided evidence that semisynthetic apoE3 NT folded in solution to adopt an amphipathic ␣ -helix bundle that possesses intrinsic lipid binding activity and is able to function as an LDL receptor ligand. The yield of ligation product was up to 70%, suggesting this method can be used to generate large quantities of specifi cally labeled protein. The versatility of the EPL approach relates to the ability to introduce labels or modifi cations in specifi c regions or sites within a protein while the other portion remains unaffected. For example, by introducing any number of modifi cations, including unnatural amino acids ( 30,31 ), biophysical probes ( 32 ), posttranslational modifi cations ( 33 ), or isotope labels ( 34,35 ), into one fragment, EPL can be used to study structural and functional aspects of proteins that might otherwise be diffi cult to address. Indeed, EPL applications for the study of posttranslational modifi cations of proteins have proved particularly useful for their ability to create large quantities of proteins with modifi cations that would otherwise be diffi cult to achieve by traditional recombinant or protein purifi cation methods ( 22 ). By modifying one or both fragments prior to ligation, EPL can produce proteins that contain unique characteristics that allow for the study of protein function as it pertains to alternative conformational state transitions ( 24,36 ), posttranslational affects ( 37 ), and protein-protein interactions ( 24,32 ). Given that the only requirements of EPL are the presence of an active cysteine at position 1 of the C-terminal fragment and a leaving group moiety on the NT fragment ( 11 ), modifications made elsewhere in the fragments will be imparted to the ligated protein product that allow for discriminating functional studies to be performed. complements a recent study by Weisgraber and colleagues who reported an X-ray structure of apoE4 bound to dipalmitoylphophatidylcholine at 10 Å resolution ( 29 ). While the apoE4 lipid complexes reported in this study were LDL receptor competent, the limited resolution did not afford discernable secondary structure or provide direct evidence of the detailed determinants of the receptor active conformation. The design adopted in this study involved joining apoE fragments together via a peptide bond to create an intact NT domain that is similar to WT apoE3 NT. Despite subtle differences, the LDL receptor binding curves for WT apoE3 NT•DMPC and EPL apoE3 NT•DMPC were consistent with binding activity. Whereas one fragment (apoE residues 1-111) was conveniently generated following intein-mediated cleavage of a precursor protein chimera, CNBr cleavage of a modifi ed apoE3 NT yielded a   a strategy to specifi cally label the region of interest. A similar approach has been employed to investigate apoE interdomain conformational adaptation using a stable isotope-enriched C-terminal domain ( 39 ). This differs from the goal of this study, which aims to investigate conformational adaptation of the receptor binding region of the NT domain upon association with lipid to confer LDL receptor binding activity.
In earlier studies, we characterized a 58-residue peptide fragment of apoE corresponding to residues 126-183. Since this peptide was not present in the context of the intact NT domain, it failed to adopt a native conformation in solution ( 13 ). Whereas structural data were obtained in the presence of a helix-inducing cosolvent ( 27 ) or a lipid mimetic detergent ( 28 ), it is uncertain if the In this study, our goal was to generate a segmental isotope-labeled apoE3 NT. The ability to introduce stable isotopes specifi cally into the receptor binding region of this protein affords new opportunities for detailed structural analysis. Because apoE3 NT is biologically active in a lipidassociated state, characterization of its structural properties in this environment is desired. While such studies are feasible by NMR spectroscopy, the spectra of helix-rich apolipoproteins undergo signifi cant spectral compression upon complexation with lipid ( 38 ). While previous NMR spectroscopy studies generated a resonance assignment of lipid-free apoE3 NT ( 21 ), spectral complication associated with lipid complexation requires an alternative approach for ascertaining details of the LDL receptor active conformation. In order to decrease spectral overlap, we adopted   absence of the remainder of the domain infl uences the properties of the peptide. Thus, in order to examine the structure of this segment of the protein within the context of the intact NT domain, the present EPL approach was pursued. Whereas such methods are not required for smaller apolipoproteins (e.g., apoC-1, apoC-II, or apoC-III), intact apoE NT in the presence of lipid is suffi ciently large that spectral crowding and overlap will hamper resonance assignment and high-resolution structure determination ( 40,41 ).
It is evident from the characterization studies presented that the EPL-generated apoE3 NT adopts a native conformation. Indeed, the segmental 15 N-labeled apoE3 NT gave rise to an HSQC spectrum that largely coincides with the corresponding spectrum of uniformly 15 N-labeled apoE3 NT. Overall, >50% of the residues comprising the 112-183 15 N-labeled fragment directly overlay with corresponding residues in uniformly 15 N-labeled apoE3 NT, strongly indicating that the two proteins adopt similar conformations and display similar chemical environments for key residues. Many other nonoverlapping residues are closely positioned to corresponding residues in the overlaid HSQC, confi rming their chemical relatedness. Despite close monitoring of the protein conditions between samples, small changes in protein conformation due to slight differences in pH and temperature can cause deviations in the resulting HSQC spectra, which may explain the observed small shifts in some peaks. Additionally, slight deviations observed in three of the segmentally labeled glycines (Gly113, Gyl 120, and Gly127; see Fig. 8 ) may be ascribed to their proximity to site-specifi c mutations introduced at positions 111 and 125 to facilitate EPL. While this may explain the observed chemical shift deviation for these residues, it also confi rms that the two amino acid substitutions present in EPL-generated apoE3 NT do not induce major changes to the overall chemical environment and/or structural architecture of the protein.
Taken together, the results obtained indicate EPL has been successfully applied to construction of an intact, functional apoE3 NT. Furthermore, it is apparent that segmental isotope labeling offers a feasible approach to simplify NMR spectra of apoE3 NT for study of its lipid-bound, receptor active conformation. Currently, efforts are underway to evaluate segmental isotope-labeled apoE3 NT in complex with lipid.
V.R. is a senior research associate at the National Fund for Scientifi c Research (Belgium).