Apo B100 similarities to viral proteins suggest basis for LDL-DNA binding and transfection capacity.

LDL mediates transfection with plasmid DNA in a variety of cell types in vitro and in several tissues in vivo in the rat. The transfection capacity of LDL is based on apo B100, as arginine/lysine clusters, suggestive of nucleic acid-binding domains and nuclear localization signal sequences, are present throughout the molecule. Apo E may also contribute to this capacity because of its similarity to the Dengue virus capsid proteins and its ability to bind DNA. Synthetic peptides representing two apo B100 regions with prominent Arg/Lys clusters were shown to bind DNA. Region 1 (0014Lys-Ser0160) shares sequence motifs present in DNA binding domains of Interferon Regulatory Factors and Flaviviridae capsid/core proteins. It also contains a close analog of the B/E receptor ligand of apo E. Region 1 peptides, B1-1 (0014Lys-Glu0054) and B1-2 (0055Leu-Ala0096), mediate transfection of HeLa cells but are cytotoxic. Region 2 (3313Asp-Thr3431), containing the known B/E receptor ligand, shares analog motifs with the human herpesvirus 5 immediate-early transcriptional regulator (UL122) and Flaviviridae NS3 helicases. Region 2 peptides, B2-1 (3313Asp-Glu3355), and B2-2 (3356Gly-Thr3431) are ineffective in cell transfection and are noncytotoxic. These results confirm the role of LDL as a natural transfection vector in vivo, a capacity imparted by the apo B100, and suggest a basis for Flaviviridae cell entry.


Isolation of plasma lipoproteins
Human. Highly purifi ed preparations of human plasma LDL were obtained from Invitrogen, Inc. and Athens Research and Technology, Athens, GA. LDL was also isolated from human plasma by sequential ultracentrifugation in NaBr solutions yielding the 1.019-1.05 g/ml density range of LDL fraction ( 6 ). Single donor plasma with sodium EDTA added at time of collection was obtained from Innovative Research. Typically, purifi ed preparations of LDL were dialyzed in PBS containing 10 mM MgCl 2 . LDL was than evaluated for purity using SDS-PAGE, and tested for DNA binding capacity using electrophoretic mobility shift assay (EMSA). Once DNA binding capacity of the LDL preparation was confi rmed, it was frozen drop-wise in liquid nitrogen and stored in aliquots of about 200 µL in liquid nitrogen until needed. Highly purifi ed, delipidated apo E was purchased from Athens Research and Technology, Inc.
Rats. Female Sprague-Dawley rats, 8-10 weeks old, were obtained from Harland Laboratories, Inc. Animals were housed in the National Institutes of Health-accredited facilities in the University of Texas Health Science Center and Baylor College of Medicine. All animals were treated in accordance with the Guide for the Care and Use of Laboratory Animals published by the U.S. National Institutes of Health (NIH publication no. 85-23, revised in 1996). All animal protocols were approved by the Animal Welfare Committees at the University of Texas Health Science and Baylor College of Medicine.
Rat plasma was used to purify LDL. To obtain blood, animals were fi rst sedated using inhalation anesthetic metaphane. This was followed by an injection of a combination anesthetic containing ketamine (42.8 mg/ml), xylazine (8.6 mg/ml), and acepromazine (1.4 mg/ml) in PBS. Blood was then collected by heart puncture using a 5 ml syringe containing 50 µl of 100 mM EDTA. Approximately 4-5 ml of blood was collected per animal. Samples were pooled and centrifuged at 400 rpm for 20 min. LDL was isolated using equilibrium ultracentrifugation in preformed gradient of KBr. Collected plasma, 9-10 ml from six rats, was diluted to 12 ml with saline, then the density of the sample was adjusted to 1.21 g/ml with KBr. Plasma was transferred to centrifuge LDL B/E receptor ( 24,25 ). These observations suggest possible links between LDL and viruses. Flaviviridae are single-stranded RNA viruses with a genome that encodes three structural proteins (capsid/core C protein, precursor membrane protein, and an envelope E protein) and seven nonstructural proteins (NS1, NS2A, NS2B, NS3, NS4A, NS4B, and NS5) ( 26 ). The genome is encapsulated in a sphere that includes multiple copies of the capsid/ core protein. It is therefore possible that an analog sequence of the B/E receptor ligand ( 27,28 ) is present in the structural proteins of these viruses.
In previous reports, we showed that human LDL binds DNA and RNA in vitro ( 18,20 ) and that highly purifi ed LDL can be used to transport and deliver plasmid DNA to the cell nucleus. We surmised that the capacity of LDL and VLDL to bind nucleic acids is likely based on the presence of regions in the apo B100 molecule that display sequence similarities to known nucleic acid-binding domains ( 29 ). Based on the location of arginine and lysine clusters and other motifs, fi ve potential DNA binding domains, 11 potential KH domains of the heterogeneous nuclear ribonucleoprotein K ( 30,31 ), and numerous bipartite nuclear localization signal sequences (NLS) were identifi ed in the apo B100 primary structure ( 18,20 ). One candidate DNA-binding region is con tained in the fi rst 100 N-terminal residues of the apo B100. Several sequence motifs also present in the DNA-binding domains of interferon regulatory factors ( 29 ), the interferon regulatory factor ( irf -s), are located in this region.
In this report, we explore the hypothesis that apo B100 and apo E have structural and functional similarities to viral DNA-binding proteins. We focus on two candidate nucleic acid-binding regions: N-terminus of apo B100, residues 0014 Lys-Ala 0096 (Region 1) and the section encompassing the known B/E-receptor ligand of apo B100, res idues 3313 Asp-Thr 3431 (Region 2) (32,33). Sequences of these regions are compared with viral nucleic acid-binding proteins, the capsid/core proteins. The apo B100 Region 2 sequence is compared with the NS3 proteins of the Flaviviridae viruses [Dengue ( dng ), West Nile ( wnlv ), yellow fever ( ylfv ), and hepatitis C] ( 34,35 ) and to UL122 protein from human cytomegalovirus (HCMV) ( 36 ). In addition, we expand our comparison of Region 1 to human irf proteins. Primary structure similarities strongly support these two regions of the apo B100 molecule as nucleic acidbinding domains in apo B100. We also analyze similarities between the receptor ligand region of apo E (Leu 151 -Arg 278 ) and dng capsid proteins and demonstrate DNAbinding capacity of purifi ed apo E. Our hypothesis compels us to consider that LDL and LDL-related particles, intermediate and very low density particles, are involved in transporting nucleic acids, and this capacity is imparted by specifi c regions of the apo B100 as well as apo E.
Here we present additional evidence that LDL and VLDL have the capacity to bind DNA. Further, LDL can be used to transfect a variety of cell types in vitro and in vivo. Based on the results of experimental studies utilizing synthetic peptides from Regions 1 and 2 of apo B100, we conclude that this capacity is mediated by elements in the apo B100 primary structure similar to viral proteins. plasmid DNA. CHO and NIH3T3 cells were grown on cover slips and transfected using LDL complexed to BOBO-1-pEGFP-N1 DNA as described above, except LDL was not labeled. The cover slips were recovered at different periods and were inverted over on well slides containing PBS without paraformaldehyde treatment. Cell images were obtained using a LUMAM TM EPI-Fluorescent microscope. Similarly, HeLa cells were grown in Costar® multi-well, fl at bottom polystyrene plates and transfected using solutions containing LDL complexed to BOBO-1-pCMVTNT DNA. Results of HeLa cell transfection were documented using a Zeiss Axiovert 25 microscope.
Nonlabeled LDL complexed to the nonlabeled pEGFP-N1 plasmid was used to transfect HeLa, MCF 7, CHO, and NIH/3T3 cell types as described above, and green fl uorescence protein (GFP) expression was documented using a LUMAM TM fl uorescence microscope with a GFP fi lter.
HeLa cells were used also to ascertain the transfection capacity of two sets of synthetic peptides. Cells were grown as routine, preconditioned in FBS-free DMEM for 4 h, washed thrice with PBS, and 400 µl PBS supplemented with 10 mM MgCl 2 , containing either the synthetic peptides or peptide/DNA complexes was added. Cells were incubated at 37°C as described in Methods for 30 min, then washed, and Trypan Blue dye in PBS was added.

Noncovalent labeling of plasmid DNA with BOBO-1
Labeling of DNA with BOBO-1™-iodide was accomplished according to the methods described by the vendor (Molecular Probes dimeric cyanine nucleic acid stains) with minor modifi cation. Briefl y, 3 µl of 10 µM dye (1 mM stock solution diluted 1:100 with ethanol) was added to 10 g of DNA at a concentration of 0.5 g/ l in PBS, and the solution was incubated at ambient temperature for 1 h.

Labeling of Lipofectin TM and LDL with CM-DiI
Labeling of Lipofectin TM and LDL with Cell Tracker™ CM-DiI was performed according to the methods described by the vendor (Molecular Probes ). Transfection agents were labeled by adding 1 µl of 20 µg CM-DiI/ml in ethanol (stock solution) to 100 µL of Lipofectin (undiluted reagent) or LDL (1.5 mg/ml by protein), and the mixture was incubated for 1 h at ambient temperature. Unreacted dye was removed using a Sephadex G-25 column.

Liposome preparation
Transfection Reagent 1 (10 mg, Avanti Polar Lipids, Inc.) was dissolved in 2 ml of a sterile buffer solution of 0.9% NaCl, 5.0% glucose, and 10% sucrose. The suspension was placed in a 37°C water bath for 10 min and vortexed to disperse the opaque lipid vesicles. Small unilamellar vesicles (SUV) were then formed by sonicating the mix for about 3 min. The transparent SUV mix was concentrated to approximately 11 mg/ml using an Amicon/ Microcon fi lter concentrator with a YM-3 membrane (Sigma, Inc.). The solution was then sterilized through a 0.22 micron fi lter (Millipore, Inc.).

Transfection solutions
All LDL preparations used in this study were shown to bind DNA in EMSA prior to cell transfection experiments. Plasmids pCMV ␤ -Gal, pEGFP-N1, pEGFP-N2, phMGFP, pGL2-Control, and pCMVTNT ® bound LDL in a similar manner.
LDL. For cell transfection, purifi ed LDL was added to the microfuge tube containing DNA and PBS with 10 mM MgCl 2 . The mixtures were then incubated at 37°C for 30 min before use. Typically, 20-40 µg of LDL protein was complexed with 1.0 µg DNA. tubes of SW 40 Ti swinging bucket rotor (Beckman), 4 ml per tube, overlaid with KBr solutions of the following densities: 1.063 g/ml (3.0 ml), 1.02 g/ml (3.0 ml), and 1.006 g/ml (2.5 ml). Samples were next centrifuged at 39,000 rpm for 60 h at 4°C. Seven fractions were collected from each tube, starting from the top of the centrifuge tube, and similar fractions from different centrifuge tubes were pooled then dialyzed against PBS. Protein concentration was routinely determined using modifi ed Lowry method ( 18 ), and SDS-PAGE (2-12%) was performed. LDL fraction was identifi ed by presence of apo B100 band.

EMSA
An aliquot of plasmid DNA digested using restriction enzymes was placed at the bottom of a microfuge tube; next, a buffer solution containing 25 mM TA, pH 7.6, and 10 mM MgCl 2 was added. An aliquot of purifi ed lipoprotein (LDL, VLDL, apo E) or synthetic peptides was then added; the cocktail was stirred gently for less than 5 s and incubated for 30 min at 37°C. Sample loading buffer (Bio-Rad) was then added at a 1:5 (v/v) ratio to the polypeptide-DNA mix. Next, each sample was mixed and subjected to electrophoresis using 0.5-0.8% agarose gels in TA buffer at 100 V. Aliquots of nucleic acid, synthetic peptides, and lipoprotein were each analyzed in separate lanes as controls. DNA bands were visualized using ethidium bromide. Peptides and proteins in lipoprotein particles were visualized using Coomassie brilliant blue R-250 (CBB).

Cells
All cell types used in these studies were obtained from the American Type Culture Collection Organization. Cells were kept in liquid nitrogen until needed and then grown according to American Type Culture Collection protocols.
Several cell types, NIH/3T3 (mouse embryonic fi broblast cell line, CRL-1658™), CHO (chinese hamster ovary, CCL-61™), MCF7 (human breast adenocarcinoma cell line, HTB-22™), Hep G2 (human hepatocyte carcinoma cells, HB-8065™), and HeLa (human cervix epithelial adenocarcinoma cell line, CCL-2™), were used for LDL-and apo B100 synthetic peptide-mediated transfection. Cells were grown and maintained in media as follows: HeLa, Hep G2, MCF7, and NIH/3T3 cell types were in DMEM supplemented with 10% FBS, 100 units penicillin G-sodium, 100 units/ml streptomycin sulfate, and 250 ng/ml amphotericin B. CHO cells were in RPMI-1640 containing 10% FBS, L -glutamine, and 100 units penicillin G-sodium, 100 units/ml streptomycin sulfate, and 250 ng/ml amphotericin B. Cells were maintained at 37°C in an atmosphere of 5% CO 2 in a humidifi ed incubator. Typically, cells were grown in culture plates (with or without glass cover slip in wells) to 60-70% confl uence ( 20 ). Prior to transfection, the medium was removed, and cells were washed thrice with PBS. Cells were then incubated for a minimum of 2 h but not more than 4 h in FBS-free medium.
Dual label experiments were conducted using Hep G2 cells. The cells were grown overnight as described above on FBS-coated cover slips to enhance attachment. Cells were next washed with PBS and incubated in FBS-free medium for 4 h, then incubated for 3 h in 200 µl of transfection solution containing FBS-free DMEM, 10 mM MgCl 2 , and preformed complexes of BOBO-1labeled pCMV ␤ -Gal plasmid DNA (3 µg), and either 15 µL of CM-DiI-labeled Lipofectin TM or 60 µg of CM-DiI-labeled LDL. The cell-coated cover slips were then removed, washed in PBS, and fi xed in 4% paraformaldehyde for 10 min at 4°C. Each cover slip was then inverted over a well of a hanging drop slides containing PBS and viewed using an Olympus Model BH-2 fl uorescent microscope.
Similar methods were used to study LDL-mediated transfection of CHO, NIH3T3, and HeLa cells using BOBO-1-labeled closed with a wound clip. All animals were injected with mixtures containing 200 µg of pGL2-Control plasmid DNA (Promega, Inc.). The control animal, rat1, was injected DNA in Tris-EDTA buffer. Test rat2 was injected with pGL2-Control complexed with LDL (1:5) in 10 mM Tris (pH 7.5), 40 mM NaCl, 1 mM EDTA, 1 mM DTT, and 4% glycerol. Rat3 was injected with pGL2-Control complexed with the liposome (1:10, as described above) and rat4 was with cocktail composed of pGL2-Control: LDL: liposome (1:1:7). Animals were anesthetized with ketamine/xylazine and decapitated 24 h after the transfection injections were administered. Tissue samples were collected from heart, lung, liver, brain, kidney, and spleen, weighed, and immediately frozen in liquid nitrogen. Liquid N 2 was added to the sample (25 mg of each tissue, except 10 mg of spleen was used) in a cold mortar and ground to a fi ne powder using a chilled pestle. Tissues were digested with Proteinase K for 24 h at 56°C; DNA was then extracted from each tissue using the QIAamp DNA Mini kit (QIAGEN, Inc.) as described in the QIAGEN Handbook of Protocols. Presence of the luciferase gene was confi rmed via PCR using reagents from Roche Biochemicals, Inc., and a primer set, 5 ′ agcaactgcataaggctatg and 3 ′ gttggtactagcaacgcact obtained from Genosys Biotechnologies, Inc. Reactions were performed in a Perkin Elmer DNA Thermal Cycler Model 480.

Fluorescent microscopy
Microscopy of Hep G2 cells and animal frozen tissue sections was performed using an Olympus model BH-2 fl uorescence microscope equipped with a Hamamatsu 3 CCD model C5810 camera. The LUMAM TM EPI-Fluorescence microscope (LOMO America, Inc.) equipped with an Optronics MacroFire® 2.0 CCD camera and the Zeiss Axiovert 25 microscope with the Optronics MicroFire CCD camera were used to obtain all other cell images. The following fi lter sets were used: a fl uorescein (FITC/TRITC) no. 51004V2 cube, a Chroma HQ-GFP NB 710 cube no. 41020 (Chroma Technology, Brattleboro, VT), and an Olympus dichroic DM500 (BP490) fi lter.

Image processing
Adobe Photoshop CS version 8 software and full-frame images of the same fi eld of view of equal size and dimensions were used to create merged overlay images. Images obtained at emission spectrum of 505-515 nm were used as background, and bright light images or red region images were overlaid at 50% transparency.

Sequence alignment
Selection of protein sequences was based on the association of LDL/VLDL and Flaviviridae viruses ( 24,25 ) in human plasma, observations regarding colocation of apo B100 and HCMV DNA in arterial wall ( 37,38 ), our previous reports on the similarity of apo B100 to irf proteins, and on preferential binding of LDL to CMV promoter-containing plasmids ( 20 ). The apo E LDL receptor ligand sequence spanning residues Leu 151 to Arg 278 was also included. All protein sequences were obtained from the National Institutes of Health protein database and saved as simple text documents. The capsid protein sequences for dng versions 1-4, hcv , wnlv , ylfv virus, bovine viral diarrhea virus, and NS3 helicase sequences were obtained from their respective polyprotein fi les in the same database. Rather than identifying all clusters of grouprelated amino acids, we fi rst focused our efforts on two types of clusters. The letter symbols for amino acids arginine (R), lysine (K), histidine (H), proline (P), cysteine (C), and glycine (G) were highlighted in each protein sequence fi le to locate basic amino acid clusters such as the apo E LDL receptor ligand "RKXRKR", and motifs common to nucleic acid binding domains as well as structurally important motifs such as "PG" and "GXXG". Se-Lipofectin TM . Ten microliters of reagent in 100 µl culture medium was combined with 100 µl culture medium containing 10 µg DNA. The solution was then stirred gently and incubated for at least 30 min at room temperature.
Liposome. Transfection Reagent 1 was prepared as described above. The SUV preparation was combined with plasmid DNA (10:1) in TE buffer and incubated at ambient temperature for 5 min before use.
Synthetic peptides. Amino acid sequences of the peptides are listed in Table 1 (Region 1) and Table 3 (Region 2). Peptides were received in lyophilized form and 1 mg/ml solutions were prepared in PBS/10 mM MgCl 2 by vortexing. Solutions were frozen drop wise in liquid nitrogen and stored at Ϫ 80°C until use. Aliquots were thawed and sterilized using a 0.2 micron fi lter immediately prior to use. In EMSA, plasmid DNA was mixed with 3, 6, and 12 µg of peptide individually or in a mixture representing a region of apo B100. Peptides were fi rst mixed in equal amounts (v/v) and incubated at 37°C for 30 min. The peptide mix was then added to the DNA in a microtube after which PBS/MgCl 2 buffer was added and the cocktail was mixed by repeated drawing into the pipette tip.

Animal transfection experiments
Two separate animal studies were conducted. In one study at Baylor College of Medicine, 12 rats were used to determine GFP expression in different tissues after human LDL-and rat LDLmediated transfection. Prior to transfection, both human LDL and rat LDL were shown to bind DNA in a similar manner using EMSA (data not shown). Animals were separated into three sets. Transfection cocktails were introduced via multiple ports, including those typically associated with viral infections such as HIV and HSV. Each animal was given an intravenous injection in the femoral vein and subcutaneously; also, inoculants were introduced into the peritoneal cavity, applied into the pharynx, nasal cavity, and rectum. Animals in Set 1 (control) were inoculated with linearized pEGFP-N1 plasmid DNA in which the HCMV I.E. promoter sequence was interrupted by digestion with Hin dIII, 5 µg of DNA in 100 µL of PBS/10 mM MgCl 2 per site. Set 2 animals were inoculated with a preformed complex of purifi ed rat LDL and pEGFP-N1 plasmid, linearized using Stu I. Set 3 animals were inoculated using a cocktail mix containing human LDL and linearized pEGFP-N1 plasmid (100 µg of LDL protein and 5 µg of DNA in 100 µL of PBS/10 mM MgCl 2 per site). One animal from each set was euthanized as described above (Isolation of Plasma Lipoproteins section) on days 2, 5, and 7; tissues were excised and immobilized in O.C.T. compound and frozen using liquid nitrogen. The immobilized tissue samples were sectioned using a cryomicrotome. Sections (5-8 µm thick) were fi xed for 30 min in 4% paraformaldehyde and analyzed for expression of GFP by fl uorescent microscopy. Animals scheduled for day 9 study were spared, because GFP expression was not observed in tissues harvested on days 5 and 7.
In separate experiments performed at UTHSC, four animals were transfected with either pGL2-Control alone or pGL2-Control complexed to LDL, cationic liposome (Transfection Reagent 1), or to a mixture of LDL and liposome and distribution of the luciferase gene in rat tissues was determined by PCR.
Rats were anesthetized with ketamine/xylazine (200 mg/10 mg per 1 kg animal weight injected intraperitoneally) and a small incision was made through the skin of the inner thigh to expose the femoral vein. Injection of plasmid DNA was made with a 30 gauge needle directly into the femoral vein. After the injection, pressure was applied to the puncture to avoid leakage and was continued until coagulation occurred. The skin incision was

Sequence comparison analysis
Highlighting group-related amino acids in sequence fi les discussed in Methods revealed the location of clusters and motifs that were then used to align multiple sequences.

Region 1, the N-terminal sequence of apo B100
Capsid proteins of the fl aviviruses and DNA binding domains of irf proteins are rich in arginine and lysine amino acids. Arg/Lys-rich clusters are located near the N-termini of apo B100, hcv , ylfv , dng 1 -4 , and irf proteins 1, 2, 5, 6, 8, and 9. In the sequences of irf proteins 3, 4, and 7 and wnlv and bvdv , Arg/Lys clusters are located distal to the fi rst 20 N-terminal residues.
Flaviviruses are thought to gain cell entry via the LDL B/E receptor ( 24,25 ). The well-established B/E receptor ligand region in apo E is an Arg/Lys-rich cluster, 0141 LRKLRK 0146 quences containing these features were then aligned and grouprelated amino acids in multiple sequences were highlighted to further assess similarity. Two regions in the apo B100 sequence were selected for these analyses: Region 1 spans the N-terminus from Glu 0013 to Pro 0140 and Region 2 includes residues Asp 3313 to Arg 3500 . Each of these apo B100 regions contains at least one basic amino acid cluster. Different options for multiple sequence alignment were also evaluated using T-COFFEE (http://www. bioinformatics.nl/tools/t_coffee.html).
Sequence comparison was also performed using two recognized algorithms for sequence similarity searches, SSEARCH and PSI-BLAST. The entire sequences of Regions 1 and 2 of apo B100, as well as their fragments and expanded versions, were used as query sequences. SSEARCH of UniProtein Knowledgebase, and UniProtKB/SwissProt databases was performed at the EBI site (http://www.edi.ac.uk/Tools/fasta33/index.html) using all available BLOSUM matrices (BLOSUM 50, 62, and 80) and default or weaker gap penalties (e.g., BLOSUM 62 has -11/-1 for gap opening/gap extension as default setting). PSI-BLAST searches were conducted at the NCBI website (http://blast. nchi.nlm.nih.gov/Blast.cgi). Databases of various sizes were searched, each with various search settings such as scoring matrices (BLOSUM 45, 62, and 80) and PSI-BLAST thresholds. Several iteration steps (3-5 times) were performed until the search did not result in any new additional fi nds.

DNA-binding
Present studies expand data on the capacity of LDL to bind HCMV promoter-containing plasmids and to transfect cells in vitro ( 18,20 ); also, the latter property is newly demonstrated in vivo. In Fig. 1 (upper panel), DNAbinding capacity of human lipoproteins, VLDL, LDL, and HDL, is illustrated. LDL samples from two donors are shown (donor A, lanes 1-5, and B, lanes 6-10); both samples had similar mobility in CBB-stained gel (middle panel) and bound the pCMV ␤ -Gal plasmid in almost identical fashion (upper panel). VLDL, although a larger apo B100-containing particle, displays higher electrophoretic mobility compared with LDL due to differences in both lipid and protein contents ( 6,39 ). Results shown in Fig. 1 (lanes 14-15, donor B only) indicate that VLDL binds to plasmid DNA in a concentration-dependent manner. Further, the mobility pattern for VLDL in the EMSA differs from the pattern seen for LDL. At lower VLDL concentrations, bound DNA is seen in a band of higher electrophoretic mobility (lanes 12-13); the shift toward lower mobility region is observed at increased VLDL protein/ DNA ratio (lanes [14][15] in the upper panel. In separate experiments, high purity, delipidated apo E was used to determine whether it contributes to the VLDL/DNA binding. Apo E binding to plasmid DNA ( Fig. 1 , lower panel) suggests that both apo B100 and apo E may determine the observed EMSA pattern for these particles.
Only apo B100-containing lipoproteins bound DNA in our studies. HDL, which lacks apo B100, did not infl uence plasmid mobility, lanes 16-18. Protease-treated LDL lost its capacity to bind DNA in EMSA (not shown). Hence, apo B100 is essential for DNA binding to lipoproteins.

Apo E similarity to Flaviviridae capsid proteins
A sequence alignment comparison of the apo E sequence containing the B/E receptor ligand, residues Leu 151 to Arg 278 , and the sequences of the four versions of dng was performed to identify analog elements. Results are presented in Table 2 . The dng1 and dng3 proteins have three discrete Arg/Lys-rich cluster in their N termini, each containing a copy of the K/R-XX-K/R motif (underlined in columns 1 and 2). Additional analog sequences are highlighted in bold in column 2. A potential bipartite NLS sequence was identifi ed in the carboxy-terminal region of the apo E molecule (column 3, Table 2 ).
In summary, the N-terminal region of apo B100, the LDL B/E receptor ligand region of apo E, capsid proteins of the several fl aviviruses, and the DNA binding domain regions of the irf proteins appear to be analogs of each other, which suggests similar functions. For apolipoproteins, these similarities mean potential DNAbinding and nuclear translocation capacities. In turn, similarity of Flaviviridae capsid proteins to apolipoproteins may greatly facilitate ability of these viruses to penetrate cells.

Region 2, the putative helicase domain of apo B100
The apo B100 sequence extending from Asp 3313 to Ser 3458 was predicted to bind DNA and mediate cell entry due to the presence of both Arg/Lys-rich clusters and the accepted B/E receptor ligand sequence, 3353 KLEGTTRL-TRKRGLKLA 3369 . Based on the apparent preferential binding of LDL to plasmids containing the HCMV IE2 promoter ( 20 ), this region of the apo B100 sequence was compared with the sequence of the HSV5 UL122 protein, which binds to the CMV promoter ( 36 ). The apo B100 Region 2 sequence was also compared with the NS3 helicase of the Flaviviridae viruses, dng1 , wnlv , and ylfv , because an Arg/ Lys-rich cluster was located within the NS3 sequence by highlighting of the arginine and lysine residues in their precursor polyprotein sequences. The sequence alignment comparison of the NS3 helicase domains of the HSV5 UL122 and Flaviviridae NS3 to the proposed analog sequence in apo B100 is presented in Table 3 . Apo B100 (residues 3316-3458) and UL122 (residues 276-420) share motifs KSS, LLSSSSSV, and GTTR. In column 3, a comparison of arginine/lysine-rich clusters of the NS3 helicases, e.g., 1685 REAIKRKLRT in dng1 , and the analog sequence in apo B100 LDL B/E receptor ligand, 3359 RLTRKRGLK, is shown. In the NS3 helicases of dng , wnlv , and ylfv , the K XX R motif is part of the Walker A motif ( 34,46 ), GXG K TR R . Analog sequences occur as GTT R LT R in apo B100 and as ASTGP R KK K in UL122 (Walker motifs are associated with nucleotide binding in kinases and helicases) ( 47 ). Flanking the Walker A motifs in the viral helicases are sequences of two ␤ -sheet structures, e.g., QITVLDL and LRTAVLAP in wnlv. These are conserved in analog sequences SSSVIDA and LKLATALS in apo B100. The hydrophobic stretch, 3367 KLATALSLSNKFV, which follows the receptor ligand in apo B100, may be a legitimate analog to the sequence 1691 KLRTLILAPTRVV in dng1 and other Flaviviridae helicases ( 35 ). The sequence alignments in columns 4 and 5 ( 27 ). The critical residues in binding/docking to the receptor are lysine residues 0143 and 0146, which are located on the hydrophilic side of an amphipathic helix ( 27 ). Replacing either lysine with arginine (as R XXK or KXX R ) does not affect binding affi nity; however, arginine substitution of both sites ( R XX R ) reduces binding by almost 70% ( 27 ).
The R/K-X-X-R/K is a multi-functional motif (including RXXR) essential in both protein-protein interactions, including nuclear entry (NLS sequences) and proteinnucleic acid interactions (40)(41)(42)(43). A comparison of the R/K clusters in Region 1 of apo B100, fl aviviral capsid proteins, and irf DNA binding domains is shown in Table 1 . The N-terminal cluster of apo B100 contains three renditions of the R/K-XX-R/K motif as R XX H , K XX R , and H XX K (see column 1); apo E N-terminus has H XX K , two R XX R , and one K XX K ; there are three copies in hcv ( K PQ R , R KT K , K TK R ); one as R XX R in ylfv ; and all versions of Dengue capsid proteins are replete with this motif, bipartite NLS sequences, and the R/K-X-R/K-X-R/K motif.
Arginine, in R X R X R and R XX R motifs, is the predominant basic amino acid in the N-terminal clusters of the irf proteins. Only the R XX R low affi nity B/E receptor ligand motif is present in the N-termini of DNA binding domains of irfs 5 , 6 , 8 , and 9 . The high-affi nity B/E receptor ligand motifs ( 27 ) K XX K , K XX R , and R XX K , absent in the N-terminal R/K clusters of irf proteins, are present in their DNA binding and NLS motifs located in the C-terminal region of these domains ( 44 ).
The motif K/R-XXX-K/R shown in column 3 and present in the sequences of apo B100 Region 1, bvdv , hcv , and irf s is repeated several fold in the irf proteins. Another motif that appears in most of these proteins is the ⌽ ⌽ ⌽ -K/R motif ( ⌽ , hydrophobic residue); it is also highlighted in column 3. In irf s, these motifs have been shown to interact with the backbone phosphate moieties of the nucleic acid ( 45 ).
Analogs of the K/R-XXX-K/R motif, shown in column 4, are part of the NLS sequence located before the metalbinding sequence of the irf proteins ( 45 ). The apo B100 sequence, NPEGKALLK is very similar to the sequence QPEGRAWAQ in hcv capsid protein. In irf s, the NPEG appears as EPDP and is separated from the K/R-XXX-K/R motif by a tripeptide sequence, KTW, missing in apo B100, and two viral capsid proteins, bvdv and hcv . An analog of the apo B100 sequence, K KTKNS EEF, is contained in the bvdv as T KSKNT QDG and as " NKSS EF" in irf 9.
A PG motif is also located near the N termini of apo B100, hcv , and irf proteins 1, 2, 4, 5, 6, 8, and 9 but not 3 and 7 (column 2, Fig. 1 ). The PG motif was used as a reference point for alignment and served in locating clusters of polar amino acids, also highlighted in column 2.
In hcv , the capsid/core protein is cleaved within a short stretch of hydrophobic amino acids, 0175 SIFLLALLS 0183 ( 26 ). Analog sequences are also present in other fl aviviruses. Three motifs contained in this region of the hcv capsid protein, shown in column 5, are also apparent in apo B100, proteins included in Tables 1 and 3 , and these resulted in the detection of viral sequences only.

LDL-mediated transfection of cells in culture
Different cell types, including HeLa, Hep G2, CHO, and NIH3T3, were transfected using LDL mixed with BOBO-1labeled plasmid DNA containing the HCMV promoter as described in Methods. Typically, LDL-mediated transfection of HeLa, Hep G2, and NIH3T3 cells occurred rapidly after upregulation of the B/E-receptor via incubation of cells in FBS-free medium for 2-4 h; fl uorescence was observed within minutes after LDL/BOBO-1-DNA mix is added to the medium. Almost 100% of HeLa cells shown in Fig. 2A were transfected within 30-45 min, and BOBO-1 fl uorescence is seen predominantly in cell nuclei. In CHO cells, transfection occurred over an extended period for hours, not minutes ( Fig. 2B ). Also, fl uorescence appeared to remain in the cytoplasm with little to none in nuclei. Few to no cells were transfected in all cell types by naked DNA ( Fig. 2C ).
In Hep G2 cells, two fl uorescence dyes (CM-DiI for transfection agents, BOBO-1 for pCMV ␤ -Gal DNA) were used to compare Lipofectin-and LDL-mediated transfection ( Fig. 2D-K ). All images were taken at 3 h after transfection. This time point was chosen for optimal transfection results using Lipofectin, which was delayed compared with weakly suggest similarities between these proteins. Notably, the DEAD motif typical of helicases is not conserved in apo B100 but may be represented as DFNS.
Our method for sequence comparison analysis identifi es weakly similar sequences that may be analogs of known proteins and therefore may perform similar functions. This algorithm reveals that apo B100 Region 1 sequence shares about 46% similarity and 22% identity with the DNA binding domains of the irf proteins and about 25% similarity and <10% identity with the viral capsid proteins. Sequence similarities between apo E and the irf proteins were <25%. The apo E sequence spanning 0155 Leu and 0248 Asp is 55% similar to the fi rst 100 N-terminal residues of dng1.
Similarity searches using SSEARCH and PSI-BLAST algorithms were performed to probe the protein sequence databases for potential homologies and to further substantiate our observations. Although these algorithms fi nd several matches to RNA-and DNA-binding proteins in the SwissProtein database, the E-value range for these matches is 1 < E < 10 and is therefore not considered highly significant. All statistically signifi cant homologs belong to several families of lipoprotein transport proteins of animals, and no signifi cant viral homologs were revealed. Similar searches were performed with input sequences of the viral these experiments. Fluorescence was not seen in controls where naked DNA was used (not included). Merged overlays of brightlight and fl uorescence images in Fig. 2L show GFP expression in MCF7 cells at 2 h after transfection. Loci of apparent GFP synthesis are visible on the periphery of nuclei. GFP expression was seen in <10% CHO cells examined (Fig. 2M). Intense and uniform distribution of GFP expression was seen in NIH/3T3 cells at 18 h (Fig. 2N) and in HeLa cells at 1 h (Fig. 2O) and 18 h (Fig. 2P) post transfection. Nuclei devoid of GFP are clearly visible (Fig. 2P). In summary, images in Fig. 2 unambiguously demonstrate successful LDL-mediated intracellular delivery and nuclear translocation of dyelabeled DNA as well as GFP expression in transfected cells in vitro.

Animal studies
Rats inoculated with rat or human LDL/pEGFP-N1 mixtures via multiple ports were euthanized on days 2, 5, and 7 (Methods). However, GFP expression was not observed in tissues harvested on days 5 and 7. Photographs taken LDL-mediated DNA delivery. Uptake of label was not seen in cells given DNA only (not shown). Extracellular location of fl uorescence can be seen in cells treated with labeled Lipofectin alone (Fig. 2D). These extracellular aggregates may result from Lipofectin interaction with cell debris or may form before its cell entry. In contrast, signal from labeled LDL is typically seen associated with cells (Fig. 2H). Micrographs E-G and I-K illustrate cellular uptake of dual labeled Lipofectin/DNA and LDL/DNA complexes, respectively. Red fl uorescence of labeled Lipofectin (Fig. 2E) in the cytoplasm is more intense and not in discrete loci as seen for labeled LDL (Fig. 2I). Little or no red fl uorescence is present in the nuclear spaces. BOBO-1 green fl uorescence is seen in nuclei and cytoplasm of cells transfected with Lipofectin/DNA (Fig. 2F) and LDL/DNA (Fig. 2J). Dual color images (Fig. 2G, K) further emphasize colocation and separation of dyes suggested by the monochrome images.
LDL-mediated delivery of pEGFP-N1 DNA to cells was tested further by monitoring GFP expression in different cell types ( Fig. 2 , right panel, L-P). No dyes were used in acids to nonpolar, hydrophobic amino acids ( 4 ). The N terminus, residues 0012 thru 0160 (Region 1), and the known B/E receptor ligand region, residues 3313 Asp thru Ser 3458 (Region 2), meet these criteria.

DISCUSSION
LDL and VLDL transport lipids in the plasma and are characterized by apo B100 as their major protein. We have demonstrated plasmid DNA-binding capacity for these lipoproteins and that DNA bound to LDL is transferred to the cell nucleus. Our results confi rm that LDL is an effective transfection vector in vivo and in vitro. Lp (a), another apo B 100-containing particle that may be present in small quantities in the LDL fraction with buoyant density < 1.050 g/ml ( 48 ), was not shown to possess DNA-binding capacity ( 20 ). Apo (a), the massive, hydrophilic molecule covalently linked and noncovalently attached to apo B100 to form Lp (a), contains only one Arg/Lys-rich cluster in the kringle domain in the single copy of kringle type 11 and two in the putative serine protease domain ( 49,50 ). Kringle types 1-10 in apo (a) are similar to plasminogen kringle 4 ( 49 ) and are devoid of basic amino acid clusters and motifs suggestive of nucleic acid binding or nuclear translocation. The noncovalent interactions between apo (a) immediately after collection of tissue sections on day 2 are shown in Fig. 3 . In the top row, GFP fl uorescence is seen in the basal cell layer of the stratifi ed squamous epithelium of the esophagus ( Fig. 3 A). No fl uorescence is visible in the lamina propria region located at the lower portion of the image. GPF fl uorescence is also seen in the spinal cord ( Fig. 3C ), lung ( Fig. 3 E), liver ( Fig. 3 G), and heart ( Fig. 3 I) in animals given rat LDL/pEGFP-N1 plasmid cocktail. Fluorescence was not detected in samples from both large and small intestine, brain, and dorsal aorta (images not included). The intense fl uorescence seen in the spiral and tubular structures in the lung image ( Fig. 3 E) and the endothelium of the hepatic vein ( Fig. 3 G) was a matter of concern, because these structures are characterized by elastic laminae, which are known to autofl uoresce. However, images of tissues obtained from control animals treated only with the Hin dIII-cut pEGFP-N1 ( Fig. 3B , esophagus; D, brain; F, lung; H, liver; J, heart) showed no or low intensity fl uorescence. In the rat inoculated with human LDL/pEGFP-N1 complex, intense fl uorescence is visible in the myocardial tissue ( Fig. 3K ). Results obtained from other tissues from this animal were inconclusive.
Separate experiments were conducted to compare tissue distribution of the luciferase gene delivered as pGL2-Control DNA alone or complexed to LDL, cationic liposome, or to a mixture of LDL and liposome. Each rat was given one injection of DNA/transfection agent mix at a single site (see Methods). DNA extracted from each tissue was analyzed by PCR using luciferase-specifi c primers ( Fig. 3 , lower panel). A qualitative comparison of these results is shown in Table 4 . Briefl y, pGL2-Control vector alone appears to transfect liver, brain, and spleen at barely detectable levels. Whether the DNA binds to LDL particles present in the rat's plasma was not determined in these studies. LDL-mediated delivery was highly effective in the kidney and spleen, somewhat effective in brain and heart, but negative results are seen in the lungs and liver. The difference between results obtained in LDL-mediated transfection with GFP gene and the luciferase gene, both driven by the CMV promoter, may be attributed to different inoculation methods (multiple sites of inoculation/application in the former study vs. single site injection in the latter experiment). The liposome-plasmid complex was highly effective in liver and visibly effective in heart, transfected lung, kidney, and spleen at lower levels, but was negative in brain. The LDL/Liposome mix was highly effective in delivery of the plasmid to the heart and lung, less effective in spleen, only slightly effective in brain, and negative in liver and kidney. These preliminary results confi rm that LDL particles are useful transfection vectors in vivo.

Synthetic peptide studies
Synthetic peptides of apo B100 Regions 1 and 2, presented in Tables 1 and 3 , were obtained to evaluate the role of apo B100 in nucleic acid-binding, transport, and delivery. Two additional criteria were important in selecting these regions: presence of an apo E receptor ligand analog sequence and solubility (hydrophilicity and hydrophobicity), based on ratio of charged and polar amino and known nucleic acid-binding domains of transcriptional regulators ( irf proteins and capsid/core proteins of Flaviviruses) provide an explanation for the capacity of LDL to bind DNA and RNA (17)(18)(19)(20)(21)(22)(23). These sequence similarities and experimental results strongly suggest that the fi rst 160 residue N-terminal region of apo B100 and LDL receptor ligand region of apo E have functions of nucleic acid bind-and apo B100 that probably involve lysine residues on the apo B100 ( 50 ) may render nucleic acid binding regions inaccessible. Hence, the Lp (a) particle may not have a role in the transport and delivery of nucleic acids.
LDL transfection capacity is likely based on apo B100, which provides the particle with DNA binding and cell entry potential. Sequence similarities in apo B100, apo E, show luciferase-specifi c 800 bp band in heart, liver, lung, brain, kidney, and spleen, respectively. Rats were transfected using pGL2-Control plasmid alone, or cocktails of rat LDL/pGL2-Control, Liposome/pGL2-control, and rat LDL/Liposome/pGL2-Control plasmid. DNA was extracted from rat tissues on day 2 after transfection and assayed using PCR with luciferase-specifi c primers 5 ′ agcaactgcataaggctatg and 3 ′ gttggtactagcaacgcact. Rats were transfected in vivo with pGL2-Control plasmid, alone (column 2) or complexed to LDL (column 3), liposome (column 4), or to LDL/liposome mix (column 5). Presence of luciferase gene was determined by PCR in heart, liver, lung, brain, kidney, and spleen tissues (column 1) of the rat on day 2 after transfection. GFP expression in tissues of the rat transfected with pEGFP-N1 plasmid/LDL mix (column 6) was determined on day 2 and was based on green fl uorescence in frozen section of different tissues. The data are illustrated in Fig. 3. * Inoculant was administered at multiple sites. regulate transcription (52)(53)(54)(55). In the fl aviviruses, once the polyprotein is expressed, the capsid proteins are cleaved by an endopeptidase at one of two dipeptide motifs of small hydrophobic amino acids such as Leu-Leu ( 26 ). Analogs of these dipeptide motifs are also present in a like-region of apo B100 as Ile-Ile and Leu-Leu. It is interesting to consider a similar mechanism is employed to release the N-terminal region of apo B100 for similar functions. Another property reported for the wnlv capsid/ core protein is cytotoxicity ( 52,(56)(57)(58). This effect was also observed with the B1-1/B1-2 peptides mix on HeLa cells. One possible scenario would be that the N-terminal region of apo B100 is cleaved so that it may compete with viral capsid proteins for the nucleic acid and/or kill infected cells to suppress the infection process.
The synthetic peptides of the apo B100 region 2 were shown to bind DNA in EMSA. Region 2 sequence is similar to HSV5 UL122 helicase and Flaviviridae NS3 helicases and contains several analog motifs present in these viral proteins. Surprisingly, these peptides were ineffective in delivery of DNA to the cells despite the presence of the know B,E-receptor-binding ligand, possibly due to blocking of the receptor ligand site by DNA or inadequate conformation for docking with the receptor.
In summary, apo B100 and apo E, constituent proteins in LDL and VLDL, share many similarities with the fl aviviral proteins that may be construed as analogs. Signifi cant viral homologs were not found in the databases using recognized search algorithms and therefore lessen the possibility that these apolipoproteins are viral in origin. These apolipoproteins may have evolved as part of the im mune system to address and/or eliminate viral (microbial) nucleic acids from circulation. Alternatively, we can consider also that fl aviviruses may have evolved to mimic apo B100 and apo E domains to gain cell entry and evade the immune system of the host. Viral genomes are known to evolve faster than the ing. Synthetic peptides covering part of this apo B100 region, Region 1, bind plasmid DNA and substantiate this function to a degree, with the caveat that ligand specifi city and peptide conformation are undetermined. Region 1 peptides also mediate cell entry for plasmid DNA, presumably via the B/E receptor and/or through the mechanism(s) used by Arg/Lys-rich cell-penetrating peptides ( 51 ). NLS sequence motifs are characterized by Arg/Lys-rich clusters, as in the DNA binding domains of most irf proteins and in the capsid/core proteins of hcv and other Flaviviridae viruses. Hence, apo B100 Region 1 contains motifs that impart nucleic acid binding, mediate cell uptake, and are apparently involved in transferring DNA into the cell nucleus.
Several studies suggest that Flaviviridae viruses may gain cell entry via the B/E receptor (24)(25)(26)52 ). This mechanism is thought to involve the E1 and E2 structural proteins of the virus ( 26 ). Although one LRK motif occurs in envelope proteins of dng1 and dng3 , our examination of the primary structures of dng1 envelope proteins did not reveal a complete apo E receptor ligand Arg/Lys-rich cluster similar to those contained in its capsid protein. However, four additional analogs of the K/R-XX-K/R are contained within the sequence spanning 0561 Gly to 0695 Leu of the dng1 polyprotein sequence. In hcv and the four versions of dng , Arg/Lys-rich clusters containing multiple copies of the K/R-XX-K/R motif similar to those in the apo E receptor ligand sequence are located near the N termini of the capsid proteins. In wnlv and ylfv , the motifs are also located in the capsid proteins but are distal to the N terminus. It is therefore possible that cell entry is mediated by capsid proteins as well, and the almost ubiquitously expressed LDL B/E receptor may enhance viral infectivity and viability.
Flaviviral capsid proteins are multi-functional proteins, i.e., they bind nucleic acids, translocate nucleic acids to cell nucleus, have a role in dimerization of viral RNA, and ) of apo B100, as well as combinations of peptides derived from the same region. Amino acid sequences of the peptides are listed in Table 2 (Region 1) and Table 3 (Region 2). Electrophoresis was performed in a 0.8% agarose and TA buffer. A: lane 1: plasmid only; lanes 2, 3, and 4: plasmid plus 3, 6, and 12 µg, respectively, of peptide B1-1 ( 0014 Lys-Glu 0054 ); lanes 5, 6, and 7: plasmid plus 3, 6, and 12 µg, respectively, of peptide B1-2 ( 0055 Leu-Ala 0096 ); lanes 8, 9, and 10: plasmid plus 3, 6, and 12 µg, respectively, of the peptide B2-1 ( ing to speculate that LDL may be also binding in vivo to circulating nucleic acids in plasma, which are present under normal conditions and are elevated in several diseases including acute infl ammation and cancer (59)(60)(61)(62). The role played by LDL and VLDL in the dissemination of hcv genomic RNA has been established by other investigators ( 16,17,19 ), and binding of LDL to genomic DNA was reported previously ( 18,20,23 ). Combined with the capacity of LDL to transfect different cell types and tissues, these observations suggest that perhaps LDL, and possibly VLDL, play a similar role in the mechanisms of metastatic disease. human genome. This ability paired with typically large population sizes can result in a fast selection of viral protein patterns that are able to mimic LDL receptorbinding motifs.
We surmise that the transport of nucleic acids by LDL may occur naturally and may be an essential function. Lipids may mediate the conformation state of the apo B100 for addressing a wide spectrum of nucleic acids as well as interactions with proteins. A function of apo B100-and apo E-containing particles may be in removing viral or other microbial nucleic acid debris from circulation. It is interest- Dye-labeled plasmid DNA complexed to apo B100-derived peptides was used to transfect HeLa cells grown in DMEM medium as described in Methods. All frames show a merged overlay image of the fl uorescence micrograph and the corresponding phase contrast version. Frames A thru C show cells transfected using cocktails of synthetic peptides B1-1 and B1-2 ( Table 2 ) derived from N-terminal region of Apo B100 mixed with 1.0 g BOBO-1-labeled phMGFP plasmid DNA: A: 1.5 nmol of B1-1 and 1.9 nmol of B1-2 plus plasmid; B: 5.9 nmol of B1-1 and 7.5 nmol of B1-2 plus labeled plasmid; C: 15 nmol of B1-1 and 19 nmol of B1-2 plus labeled plasmid. All transfection experiments were conducted in 400 l of DMEM. Enlarged cut-out images in inserts D-H show intracellular loci of label. All images were obtained using a Zeiss Axiovert 25 scope and the Optronics MicroFire CCD camera. Bottom panel: Cytotoxic effects of peptides B1-1 and B1-2 on HeLa cells are demonstrated using Trypan Blue dye . Cytotoxicity caused by mixtures of synthetic peptides B1-1 and B1-2, 4 µg each peptide (frame I) and 10 µg each peptide (frame J), added to preconditioned HeLa cells in DMEM. In contrast, no cytotoxicity is seen for HeLa cells treated similarly with mixtures containing the same quantities of synthetic peptides B2-1/ B2-2, frames K and L, respectively. These studies offer a new perspective in our understanding of the pathogenic nature of apo B100 lipoproteins. Changes in lipid content, chemical modifi cations of apo B100, accumulation of lipo-viral debris, and other changes affecting nucleic acid-binding capacity of LDL may alter functionality of these particles.
LDL and peptides derived from the apo B100 have great potential as nucleic acid transport and delivery vehicles with applications in gene replacement therapies, DNA vaccines, rapid creation of transgenic animals, and cell bioreactors in the production of pharmaceutical proteins. LDL and VLDL are the only delivery nanoparticles that can be used in autologous formulations. Unlike viral or synthetic vectors and cell penetrating peptides ( 63 ), autologous LDL will not elicit an immune response. LDLs are easily purifi ed and can be readily stored for long periods in liquid nitrogen. These naturally occurring particles can be exploited in the development of accurate gene replacement formulations.