Molecular Cloning and Expression of Lipid Transfer Inhibitor Protein Reveals Its Identity with Apolipoprotein F*

Published studies demonstrate that lipid transfer inhibitor protein (LTIP) is an important regulator of cholesteryl ester transfer protein (CETP) activity. Although LTIP inhibits CETP activity among different lipoprotein classes, it preferentially suppresses transfer events involving low density lipoprotein (LDL), whereas transfers involving high density lipoprotein as donor are less affected. In this study, we report the purification of LTIP and the expression of its cDNA in cultured cells. Purification of LTIP, in contrast to other published protocols, took advantage of the tight association of this protein with LDL. Ultracentrifugally isolated LDL was further purified on anti-apoE and apoA-I affinity columns. Affinity purified LDL was delipidated by tetramethylurea, and the tetramethylurea-soluble proteins were separated by SDS-polyacrylamide gel electrophoresis. The protein migrating at a molecular mass of ∼33 kDa was excised from the gel and its N-terminal amino acid sequence determined. The 14-amino acid sequence obtained showed complete homology with the sequence deduced for apolipoprotein F (apoF) cDNA isolated from Hep G2 cells. On Western blots, peptide-specific antibodies raised against synthetic fragments of apoF reacted with the same 33-kDa protein in LTIP-containing fractions purified from LDL and from lipoprotein-deficient plasma. In contrast to that previously reported, apoF was shown to be associated almost exclusively with LDL, identical to the distribution of LTIP activity. The cDNA for apoF was cloned from a human liver cDNA library, ligated into a mammalian expression vector, and transiently transfected into COS-7 cells. Conditioned media containing secreted apoF demonstrated CETP inhibitor activity, whereas cells transfected with vector alone did not. This CETP inhibitor activity was efficiently removed from the media by nickel-Sepharose, consistent with the 6-His tag incorporated into recombinant apoF. By Western blot, the 6-His-tagged protein had a molecular weight slightly larger than native apoF. The CETP inhibitor activity of recombinant apoF possessed the same LDL specificity, oleate sensitivity, and dependence on lipoprotein concentration as previously noted for LTIP. We conclude that LTIP and apoF are identical.

One of us (R. E. Morton), in collaboration with D. B. Zilversmit, was the first to identify and partially characterize a novel protein in human plasma that inhibits cholesteryl ester transfer protein (CETP) 1 activity in vitro (1). Subsequent reports by Son and Zilversmit (2) and by Nishide et al. (3) have confirmed and extended the initial characterization of this factor as a unique ϳ30,000-35,000 molecular weight acidic glycoprotein. This protein, LTIP, equally inhibits the CETPmediated transfer of TG and CE (1,2). The mechanism of inhibition has not been studied in detail; however, limited studies by us and others (1,2,4) suggest that LTIP disrupts the association of CETP with lipoproteins.
In contrast to the effects of other apoproteins that have been shown to possess CETP inhibitory activity when assayed at high concentrations in vitro (2,(5)(6)(7)(8), LTIP shows lipoprotein specificity in its inhibitory capacity. LTIP preferentially suppresses lipid transfers involving LDL (1.019 Ͻ d Ͻ1.063 g/ml) (9,10) while having the least effect on transfers between VLDL (d Ͻ1.006 g/ml) and HDL (1.063 Ͻ d Ͻ1.21 g/ml). In mass lipid transfer assays with whole plasma, LTIP supplementation leads to decreased efflux of CE from LDL to VLDL and increased efflux of CE from HDL to VLDL, resulting in HDL that are better substrates for lecithin-cholesterol acyltransferase (10). We have proposed that this happens because plasma TG, not CETP, is rate-limiting for net lipid transfer in normal plasma (11,12). Thus, in the presence of LTIP, less TG is expended by transfer to LDL, allowing transfer from VLDL to HDL to be increased. Among control subjects, LTIP activity levels in whole plasma correlate negatively with the rate of transfer between VLDL and LDL and with HDL size (9). Recent studies have substantiated these observations in a patient population deficient in LTIP activity (13). Additionally, we have shown that CETP itself has little preference for lipid transfer from either LDL or HDL but, rather, the preferential transfer of CE from HDL seen in normal plasma is a function of LTIP activity (14). High, yet physiological, concentrations of free fatty acids can inactivate LTIP (15), suggesting that LTIP activity can be regulated by dietary status. Together, these data strongly suggest that LTIP alters the overall pattern of lipid transfer events between lipoproteins by functioning as a traffic policeman, impeding lipid fluxes between some lipoproteins while allowing others to continue. Therefore the effectiveness of lipid transfer activity in plasma is determined by both CETP and LTIP activities.
Our efforts in the past to isolate proteins that inhibit CETP have consistently supported the conclusion that a single pro-tein is responsible for most, if not all, of the inhibitory activity observed in whole human plasma. Purification of LTIP has been hampered by its tendency to associate with other plasma proteins, most particularly apoD (10), during purification. Interference with CETP/LTIP assays by low levels of many dissociating agents commonly used to disrupt protein-protein interactions has greatly limited the methods that can be used to dissociate LTIP-protein complexes for further purification. In this report we have developed an alternative approach to LTIP purification that takes advantage of our observations that LTIP activity is primarily associated with LDL in plasma (10) and that a portion of this LTIP activity is retained on LDL after its isolation by standard ultracentrifugal techniques. We report here the isolation of LTIP from human LDL and the cloning and expression of its cDNA.
Lipoprotein Isolation and Radiolabeling-Fresh human plasma from the Blood Bank of the Cleveland Clinic Foundation was the source of VLDL, LDL, and HDL. Lipoproteins were isolated by sequential ultracentrifugation (16), extensively dialyzed against 0.9% NaCl, 0.01% EDTA, 0.02% NaN 3 , pH 8.5, and stored at 4°C. Lipoproteins were quantitated based on their total cholesterol content. In some instances, lipoproteins were radiolabeled with [ 3 H]CE by a lipid-dispersion method (1). ApoA-I was isolated from fresh human plasma by a combination of ultracentrifugation, HDL delipidation, and chromatofocusing of human apoHDL (17).
CETP and LTIP Isolation from Lipoprotein-deficient Plasma-Partially purified CETP was isolated from lipoprotein-deficient human plasma by hydrophobic and ion exchange chromatography as described previously (18) and stored in 0.27 mM EDTA, pH 7.4. LTIP was partially purified from lipoprotein-deficient plasma as described previously (1,10).
CETP and LTIP Assays-Assays for CETP and LTIP activity were performed according to our previously published methods (19 -21). CETP activity was routinely measured in assays of lipid transfer between radiolabeled LDL ([ 3 H]CE LDL) and unlabeled HDL (10 g of cholesterol each). At the end of the assay, LDL and HDL were separated by the phosphate-manganese procedure (19,20). CETP activity was determined from the fraction of label transferred to the acceptor and expressed as percent k t (19).
CETP inhibitor activity was measured in the assays described above or in assays where lipid transfer was measured between radiolabeled LDL ([ 3 H]CE LDL) and biotinylated LDL (21). After incubation, these lipoproteins were separated by incubation with avidin-Sepharose. Inhibitor activity is expressed as the percent inhibition of a standard amount of CETP added to the assay.
Antibodies and Immunoadsorbants-Goat anti-human apoA-I and anti-apoE IgG (Calbiochem) were purified from antisera by affinity chromatography on protein G-agarose as detailed by the manufacturer (Pierce). Eluted IgG-containing fractions (25-28 mg) were pooled, dialyzed versus 0.1 M NaHCO 3 , 0.5 M NaCl, pH 8.3, overnight, and then coupled to 10 ml of CNBr-Sepharose (Amersham Pharmacia Biotech) following the manufacturer's protocol. Synthetic fragments of apoF were prepared by Fmoc (N-(9-fluorenyl)methoxycarbonyl) solid phase chemistry utilizing a multiple antigen peptide carrier core; antibodies to this antigen were elicited in rabbits (Bio-Synthesis, Inc., Lewisville, TX). Goat anti-apoF antiserum, prepared as described previously for apoF isolated from delipidated HDL by ion exchange chromatography (22), was the gift of Dr. Petar Alaupovic.
Purification of LTIP from LDL-Citrated human plasma was obtained from the Blood Bank of the Cleveland Clinic, and LDL was isolated by sequential ultracentrifugation (50,000 rpm, 50.2 Ti rotor, 10°C, 18 h) as the 1.019 Ͻ d Ͻ 1.063 g/ml fraction (16). LDL was extensively dialyzed versus 0.9% NaCl, 0.02% EDTA, pH 8.5. LDL was further purified by chromatography on two affinity columns. Tandem columns of anti-apoA-I-Sepharose (1 ϫ 19 cm) and anti-apoE-Sepharose (1 ϫ 10 cm) were packed and equilibrated at room temperature with 0.9% NaCl, 0.01% EDTA, pH 7.4. LDL (5 ml, 15-18.5 mg protein) was applied to the columns and eluted by 0.9% NaCl, 0.01% EDTA, pH 7.4, at a rate of 2 ml/h. Eluted fractions containing LDL, as determined by cholesterol assay, were pooled and subsequently extracted by TMU as described below. After elution, the columns were recycled for subsequent use by extensive washing with 0.9% NaCl, 0.01% EDTA (pH 11, adjusted with NH 4 OH), followed by the same buffer at pH 7.4.
TMU extraction of LDL was performed by a modification of the method Kane et al. (23). Briefly, in a disposable glass tube, affinity purified LDL (600 g of protein) was adjusted to 1 ml with 0.9% NaCl, 10 mM EDTA, pH 9, and incubated at 37°C for 10 min. Pre-warmed TMU (0.75 ml) was added to each 1-ml sample, mixed gently, and incubated at 37°C in a sealed tube until the apoB-lipid pellicle floated at the top of the tube (2-3 h). After filtration through a glass wool plug, the TMU-soluble proteins were dialyzed against 0.9% NaCl, 0.01% EDTA, pH 7.4, buffer overnight and concentrated about 10-fold (Centricon 10, Amicon, Beverly, MA).
Electrophoresis and Immunoblot Analyses-Pre-made 7.5% SDS-PAGE gels (Integrated Separation Systems, Natick, MA) were pre-run for 1 h (20 mA) as recommended by the manufacturer. Samples were combined with 5ϫ SDS-PAGE sample buffer (0.3125 M Tris-HCl, 25% ␤-mercaptoethanol, 50% glycerol, 0.5 mg/ml bromophenyl blue, pH 8.3) and heated for 5 min at 100°C. Samples were applied and gels developed at 20 mA for 1.5 h. Proteins were visualized by silver staining (Dai-ichi silver stain kit, Integrated Separation Systems) or Coomassie Blue (Sigma). For Western blot analyses, proteins in SDS-PAGE gels were electrotransferred to nitrocellulose at 4°C overnight as described by Towbin et al. (24). Membranes were blocked with 5% dry milk in phosphate-buffered saline, reacted with the indicated antibody followed by reaction with peroxidase-conjugated secondary antibody. Reaction products were visualized by reaction with ECL (Amersham Pharmacia Biotech) and exposure to x-ray film.
When detecting apoproteins on intact lipoproteins, either in serum or in isolated form, samples were fractionated by one-dimensional agarose gel electrophoresis (25). Lipoproteins were transferred to nitrocellulose, which was backed with wetted filter paper, by passive adsorption for 15 min. Blots were blocked with 5% powdered milk, 2% calf serum, and 1% BSA in 0.9% NaCl, 0.1% EDTA, pH 7.6. Buffers for the blocking step and all subsequent steps contained 1% BSA in 0.9% NaCl, 0.1% EDTA, pH 7.6, the latter being important to prevent lipid oxidation in the nitrocellulose-bound lipoproteins. Immune complexes were visualized by reaction with ECL as indicated above. Alternatively, gels were dried and lipoproteins visualized by staining with Fat Red 7B (25).
Cloning of ApoF cDNA-Based on a 14-amino acid N-terminal sequence obtained for the 33-kDa protein from LDL, a 26-nucleotide primer (5Ј-GAGGACTGTGAGAATGAGAAGGAGCA-3Ј) was synthesized (Genosys, The Woodlands, TX) and used to screen a human liver cDNA library by 3Ј-rapid amplification of cDNA ends PCR with Marathon-Ready cDNA (CLONTECH, Palo Alto, CA). The single, 1.2-kilobase pair product was ligated into a pCR 3.1-Bi plasmid by TA cloning (Eukaryotic TA Cloning Kit, Invitrogen, Carlsbad, CA) and transformed into Escherichia coli (TOP10FЈ) according to the methods described by the manufacturer (Invitrogen). DNA sequencing was performed with a BigDye Terminator Cycle Sequencing Kit using the ABI PRISM 377XL DNA sequencer (26). Based on homology of sequenced clones with apoF (27), additional gene-specific primers (see "Results") were prepared to permit PCR cloning of the cDNA fragment of apoF encoding for amino acids 147-308, which corresponds to the complete sequence of the secreted protein. A 492-base pair HindIII/XbaI restriction fragment was subcloned into a mammalian expression vector, pSecTag (Invitrogen), to permit the synthesis and secretion of apoF with six C-terminal histidine residues.
Express of ApoF cDNA in COS-7 Cells-COS-7 cells were transfected by the LipofectAMINE method (Life Technologies, Inc.). Briefly, in six-well culture plates, 1 ϫ 10 5 cells were seeded per well in 2 ml of the appropriate growth medium supplemented with 5% fetal bovine serum and incubated at 37°C in a CO 2 environment until the cells were 50% confluent. Each well received DNA (1 g) and 5 l of LipofectAMINE in a final volume of 1 ml of serum-free medium. After incubating with DNA complexes for 24 h, cells were washed and incubated in serum-free media for 48 h. The media were collected, briefly centrifuged, concentrated on Centricon 10 membranes (Amicon), and CETP inhibitory activity in the conditioned media determined.
Other Analytical Methods-Protein was quantitated by the method of Lowry et al. (28) as modified by Peterson (29), with BSA as standard. Total cholesterol of lipoproteins was assayed by a colorimetric, enzymatic method using Cholesterol 100 reagent kit (Sigma). For amino acid analyses, proteins were transferred to polyvinylidene difluoride membranes by electrophoresis in transfer buffer (10 mM CAPS, 10% methanol, pH 11.0) at 10°C for 2 h (30). Transferred proteins were visualized by brief staining with 0.1% Coomassie Blue (Sigma) in 50% methanol, and bands of interest were excised. N-terminal protein sequence analysis was carried out with an Applied Biosystems Procise, model 492 protein sequencer fitted with a 140c microgradient system, 785A programmable absorbance detector, and a 610A (version 2.1) data analysis system (31, 32).

RESULTS
LTIP Purification from LDL-To facilitate the purification of LTIP by an approach that minimizes the potential for apoD contamination (10), we devised a protocol to purify this protein from LDL, which contains most all LTIP activity in native plasma (10). Based on the recovery of LTIP activity in the d Ͼ1.21 g/ml fraction after ultracentrifugation, compared with that in lipoprotein-deficient plasma prepared by dextran sulfate-manganese precipitation (which dissociates all LTIP from lipoproteins (9)), we have estimated that ϳ30% of plasma LTIP activity is retained in this lipoprotein fraction after isolation. Although LDL only retains a small portion of its original LTIP, this approach provides a means of avoiding apoD contamination since apoD resides on apoA-I-containing lipoproteins (33).
Beginning with isolated LDL, the purification scheme involved affinity adsorption of minor contaminating proteins, apoE and A-I, with solid phase antibodies to these apoproteins, followed by TMU extraction (23) of the affinity purified LDL. The lipid-free, TMU-soluble proteins from LDL exhibited inhibitory activity. This purification scheme typically resulted in ϳ530-fold purification of CETP inhibitor activity with ϳ8% recovery (Table I). The CETP inhibitor activity in the TMUsoluble LDL fraction manifested the same preference for lipid transfers involving only LDL and suppression by sodium oleate as we have previously observed for LTIP in the CM-cellulose fraction of LTIP isolated from lipoprotein-deficient plasma (Table II). In contrast to the inhibitory activity in the CM-cellulose and LDL-TMU fraction, apolipoprotein A-I, which suppresses CETP activity under some conditions (2,15), was unaffected by 10 M oleate (data not shown). The TMU-soluble fraction from LDL was further resolved by SDS-polyacrylamide gel electrophoresis (Fig. 1, 3rd lane). A slightly diffuse staining band near 33 kDa was observed in the LDL-derived proteins, but neither VLDL nor HDL contained detectable protein of similar molecular weight, consistent with the near-exclusive association of LTIP with LDL. This molecular mass compares well with that determined for LTIP (30 -35 kDa) based on the elution of LTIP activity on gel filtration under dissociating conditions (1). Although other small molecular weight proteins (Ͻ10,000 molec-ular weight, presumably apoCs) were present in the LDLderived fraction, these proteins were not found in LTIP prepared from lipoprotein-deficient plasma (Fig. 1, 2nd lane), suggesting that these contaminants do not contribute to the LTIP activity measured in the LDL protein fraction. Assuming the inhibitor activity in the TMU-soluble LDL fraction is due to the 33-kDa component, 35 pmol of inhibitor protein was sufficient to suppress CETP activity by 50% in the standard LDL to LDL transfer assay, which contains 33 pmol of lipoprotein. This equates to an IC 50 of ϳ0.05 M.
Following transfer to polyvinylidene difluoride membranes, the 33-kDa band was excised and its N-terminal amino acid sequence determined. N-terminal sequence analysis yielded a single sequence consisting of SLPTEDXENEKEQA. A search of sequence data bases revealed complete homology of this 14amino acid sequence with amino acids 147-160 of human apoF (27). Amino acid 147 corresponds to the N-terminal amino acid of secreted apoF, which is derived from the cleavage of a larger precursor protein. No other proteins of similar sequence were identified. ApoF has a calculated isoelectric point of 4.11, which is near the acidic pI we and others (1, 2) previously reported for partially purified LTIP under dissociating conditions.  The CETP inhibitor activities in CM-LTIP (CM-cellulose fraction of lipoprotein-deficient plasma) and LDL-LTIP (TMU-extracted LDL proteins) were compared. Inhibitor activity was determined for each fraction in assays where LDL/biotin-LDL or LDL/HDL were the donor/ acceptors. Also, in LDL/biotin-LDL assays, the effect of oleate on the expression of inhibitor activity was determined. See "Experimental Procedures" for experimental details. Proteins were visualized by silver staining. Note: the major band in the CM-cellulose fraction of LTIP is apoD as determined by N-terminal amino acid sequencing. Thus, the novel protein derived from LDL is slightly smaller than apoE and slight larger than apoD.
Anti-ApoF Antibodies-Peptides fragments of apoF, selected based on surface probabilities and lack of homology with other proteins in published data bases, were synthesized and injected into rabbits to raise antibodies against selected regions of apolipoprotein F. Antibodies to fragment 1 (anti-apoF Ab1, amino acids 1-14 of mature, secreted apoF, i.e. SLPTEDCENEKEQA) and fragment 2 (anti-apoF Ab2, amino acids 94 -107, i.e. QYYQDQKDANISQP) reacted with the same 33-kDa band on SDS-PAGE/immunoblots of two distinct preparations of CETP inhibitor activity (CM-cellulose fraction and the LDL-TMU fraction, Fig. 2).
Antibodies to fragments of apoF could also be used to detect this apoprotein in serum resolved by one-dimensional agarose electrophoresis and subsequently blotted onto nitrocellulose (Fig. 3, data shown for anti-apoF Ab2 only). ApoF was detected almost exclusively on LDL in whole serum (Fig. 3). Minor immunoreactivity was noted with VLDL, whereas HDL in whole serum (Fig. 3) or in isolated form (data not shown) did not react with anti-apoF Ab2. Notably, the distribution of apoF in serum is the same as the distribution of LTIP activity in plasma determined by gel filtration (10). Although the high sample concentrations used in these blots permitted us to assess whether low levels of apoF were associated with other lipoproteins, they do not reflect the relative apoF content of these fractions due to the nonlinear response in this dose range. Based on the immunoreactivities of serially diluted serum and isolated LDL fractions, we determined that isolated LDL contains ϳ25% of the apoF originally on this lipoprotein in serum, with the reminder lost to the d Ͼ1.21 g/ml fraction during centrifugation (data not shown). This closely mirrors the recovery of LTIP activity in the d Ͼ1.21 g/ml fraction of plasma compared with the LTIP activity in lipoprotein-deficient plasma prepared by divalent cation precipitation (see above); divalent cations effectively dissociate all LTIP from lipoproteins (9, 34).
This lipoprotein distribution for apoF is contradictory to that previously reported by Alaupovic and co-workers (22). With a goat antiserum against apo"F" provided to us by Dr. Alaupovic, the major antigen for this antibody was recovered in the lipoprotein-free fraction (d Ͼ1.21 g/ml) of plasma, with minor reactivity detected in HDL and LDL (data not shown). This antigen distribution is distinct from that obtained with antibodies prepared against pure apoF fragments (Fig. 3), clearly showing that these two antisera are primarily directed against different proteins. Data from Day et al. (27) also suggest that this goat antibody reacts with a number of plasma proteins. We interpret these data to mean that the original antisera against apoF (that prepared by Alaupovic et al.) is not specific for apoF, although it does detect apoF under certain conditions. Based on our data with antibodies directed against synthetic fragments of apoF, we conclude that apoF is primarily an LDL-associated apoprotein although minor amounts may be recovered in other lipoproteins after ultracentrifugation.
It is important to note that although anti-apoF Ab2 antibodies could react with apoF on intact LDL when the lipoprotein was adsorbed onto nitrocellulose, they did not react with apoF on LDL in solution. That is anti-apoF Ab2 antibodies (nor anti-apoF Ab1 antibodies) were not able to mediate the immunoadsorption of LDL-associated apoF even when added in apparent excess (data not shown). Consistent with this, anti-apoF peptide antibodies were not able to block LTIP activity nor could they mediate immunoadsorption of inhibitory activity to a significant degree. This indicates that the antigenic sites for these antibodies are not accessible on native apoF.
ApoF Cloning and Expression-Based on the 14-amino acid sequence obtained for the 33-kDa band shown above, we prepared a 26-nucleotide primer (see "Experimental Procedures") and screened a human liver cDNA library by 3Ј-rapid amplification of cDNA ends PCR. The single, 1.2-kilobase pair product was ligated into the pCR3.1 vector and transformed into E. coli. The nucleotide sequence of isolated clones was homologous with the published sequence for the apoF cDNA cloned from Hep G2 cells (27). The nucleotide sequence of liver apoF was 95.6% homologous to Hep G2 cDNA; the bulk of this nonidentity was due to the absence of one of the two 32-nucleotide repeats present in the 3Ј-untranslated region of the Hep G2 sequence (base 1545-1608). Subsequently, additional primers were prepared to permit PCR cloning of a cDNA fragment of apoF coding for amino acids 147-308, corresponding to the complete sequence of secreted apoF (Fig. 4). This fragment was subcloned into a mammalian expression vector, pSecTag, to FIG. 2. Western blot of inhibitor fractions with anti-apoF peptide antibodies. CETP inhibitor-containing preparations isolated from LDL or from lipoprotein-deficient plasma by phenyl-Sepharose and CM-cellulose chromatography (CM, i.e. standard LTIP preparation) were electrophoresed on SDS-PAGE gels and blotted onto nitrocellulose. Samples were incubated with the indicated anti-apoF peptide antiserum or preimmune serum (1:100 dilution). Immune complexes were reacted with anti-rabbit IgG peroxidase-conjugated IgG (1:1000 dilution) and visualized by enhanced chemiluminescence. Similar amounts of CETP inhibitor activity were applied for CM and LDL inhibitor fractions (42 and 72% inhibition, respectively).

FIG. 3. Immunoreactivity of sera and lipoprotein fractions.
The indicated samples were electrophoresed on agarose gels, blotted onto nitrocellulose, and incubated with isolated anti-apoF peptide IgGs (apoF Ab1 ϩ Ab2, 125 l of serum equivalent each) or IgG from nonimmune sera (250 l of serum equivalent). Immune complexes were reacted with anti-rabbit IgG peroxidase-conjugated IgG (1:5000 dilution) and visualized by ECL. LDL and d Ͻ1.21 samples were concentrated 2.4and 1.9-fold relative to starting plasma, respectively. Org, origin.
permit the synthesis and secretion of apoF with six C-terminal histidine (6-His) residues.
After transfection into COS-7 cells and secretion, media from cells transiently transfected with vector containing apoF cDNA expressed inhibitor activity that suppressed CETP in a dosedependent manner, whereas cells transfected with the vector (Fig. 5A) or LipofectAMINE alone (data not shown) did not markedly affect CETP activity. Treatment of conditioned media with nickel-agarose to bind the 6-His-tagged recombinant apoF removed Ͼ93% of the inhibitor activity. The proteins eluted from the nickel-Sepharose matrix also demonstrated CETP inhibitor activity (Fig. 5B). The eluted protein fraction was Ͼ95% pure recombinant apoF as determine by SDS-PAGE and immunoblot analysis (Fig. 5C). The apparent molecular weight of recombinant apoF was slightly higher than that of native apoF (ϳ35,000 versus ϳ33,000), consistent with the additional amino acids incorporated into the expressed protein.
Recombinant apoF showed the same specificity for lipid transfer assays involving only LDL and inactivation by oleate as previously noted for LTIP (10,15) (Fig. 6 A and B). Although the ratio of inhibitor activities in the LDL/LDL and LDL/HDL assays tended be lower for recombinant apoF (5.8 Ϯ 1.8 (n ϭ 3) versus 7.4 Ϯ 1.8 (n ϭ 4) for apoF and CM-LTIP, respectively) this difference was not significant (p ϭ 0.47). Based on its protein concentration, determined by comparison of nickel-Sepharose-purified recombinant apoF with known proteins on protein-stained SDS-PAGE gels, we estimate a specific activity of 14% inhibition/g of protein, resulting in an IC 50 for recombinant apoF of Ͻ0.15 M. This is comparable to the 0.05 M IC 50 value determined for purified, native apoF (LDL-TMU fraction). Like CM-LTIP, the inhibitory activity of recombinant apoF was strongly and inversely related to the assay content of lipoproteins (Fig. 6C). Recombinant apoF equally suppressed the capacity of CETP to mediate TG and CE transfers (ratio of TG/CE inhibition ϭ 0.95 Ϯ 0.06, mean Ϯ S.D., n ϭ 5), as was the case for CM-LTIP (0.99 Ϯ 0.10, mean Ϯ S.D.), indicating that inhibition is achieved by blocking all CETP function, not by modifying the capacity of CETP to interact with specific lipid substrates. DISCUSSION Our current understanding of LTIP derives from studies mostly performed on the CM-cellulose fraction of partially purified LTIP isolated from lipoprotein-deficient plasma. Based on measurements of LTIP activity, we and others (1)(2)(3)10) have shown that LTIP is an acidic glycoprotein (pI Յ 4) of 30,000 -35,000 molecular weight that is associated almost exclusively with LDL. This LDL-specific association is consistent with the observation that LTIP has greater suppressive activity toward CETP activity involving LDL compared with other lipoprotein substrates (9, 10). These properties of LTIP, based on its activity, are important criteria for comparing LTIP with CETP inhibitory activity isolated from plasma by other approaches.
The data presented in this report strongly support the identity of LTIP and apolipoprotein F. Similarities in molecular weight, isoelectric point, lipoprotein distribution in plasma, and co-recovery in the d Ͼ1.21 fraction after ultracentrifugation provide strong evidence for such a conclusion. This conclusion is further strengthened by our finding that two preparations of LTIP activity isolated by independent protocols each contain apoF. And finally, recombinant apoF manifests CETP inhibitory activity that has specificity for inhibiting transfer reactions involving LDL and is suppressed by fatty acids as has been reported for LTIP (9,10,15,21). Based on these several lines of evidence, we conclude that LTIP is apoF. In assigning this identity, we follow the nomenclature of Day et al. (27) where apoF is defined as the C-terminal 162 amino acids derived from the cleavage of a 308-amino acid precursor. Given the location of the putative signal peptide, it appears likely that this cleavage occurs extracellularly, although this has not been determined.
The identification of LTIP as apoF finally assigns a function to this little understood apoprotein. We suggest that this function of apoF is unique among apoproteins since the reported CETP inhibitory activities of better studied apolipoproteins have not been borne out by further studies in more physiological settings. For example, although purified apoA-II was FIG. 4. Nucleotide sequence of Hep G2 apoF cDNA and a clone isolated from a human liver cDNA library. The nucleotide sequence of apoF cDNA fragment amplified by PCR from a human cDNA library is compared with the published sequence of a portion of the apoF cDNA from Hep G2 cells (27). PCR primers are indicated by the underlined sequences. Nucleotides indicated in bold are unique to the clone and were integrated into primers to aid subcloning. Numbers in parentheses note the amino acid numbers for secreted apoF, which correspond to amino acids 147 and 308 of the full-length protein. The stop codon, noted by * in the native sequence, was deleted when cloned into pSecTag.
shown in several studies to inhibit CETP activity in in vitro assays (5-7), CETP activity is not decreased in transgenic mice co-expressing apoA-II and CETP (35), and HDL particles containing apoA-I only or both apoA-I and apoA-II are equal in their CETP substrate capacity (36). In other studies, pure apoA-I, -C-I, and -C-III inhibited CETP activity with IC 50 values of 0.5 to 0.8 M (2, 7). However, co-expression of apoC-III in CETP transgenic mice does not decrease plasma CETP activity in these mice (35), and co-expression of human apoA-I in CETP transgenic mice leads to increased, not decreased, CETP activ- FIG. 5. CETP inhibitor activity in conditioned media. COS-7 cells were transfected with the indicated vector by the LipofectAMINE method. After 24 h, cells were washed and incubated in serum-free media for 48 h. CETP inhibitory activity in the conditioned media was assayed as described under "Experimental Procedures." A, dose response of CETP inhibitor activity in conditioned media from cells expressing vector ϩ apoF (closed squares) or vector alone (closed circles). Media were concentrated 10-fold for assay of inhibitor activity. Values are mean Ϯ S.D. B, nickel-Sepharose-retained proteins (100 l) were eluted and assayed for CETP inhibitor activity. Condition media (45 ml) were adjusted to 500 mM NaCl and treated with 5 ml of nickel-Sepharose. Following a 150 mM imidazole wash, proteins were eluted with 500 mM imidazole and concentrated 30-fold relative to the starting media. Values are mean Ϯ S.D. C, nickel-Sepharose purified proteins were applied to SDS-PAGE gels and stained for protein or electrotransferred and immunoblotted with anti-apoF Ab2 antisera. Lanes 1 and 3, nickel-Sepharose-purified proteins from apoF-transfected cells; lanes 2 and 4, nickel-Sepharose-purified proteins from cells transfected with vector alone.
FIG. 6. Properties of CETP inhibitory activity in COS-7 conditioned media. A, the CETP inhibitor activity expressed in assays measuring transfer from LDL to biotin-LDL (LDL/LDL) and LDL to HDL (LDL/HDL) was determined. Samples are as follows: LTIP (CM-cellulose fraction) and recombinant apoF (rApo F, conditioned media from COS-7 cells transiently expressing apoF). For clarity, activity values are expressed as that measured in 50 l (15 g protein) of LTIP or that in 1 ml of 10ϫ concentrated media for recombinant apoF. LDL specificity is calculated as the ratio of these two activities (LDL/LDL/LDL/HDL). These results are representative of 4 and 3 determinations for LTIP and recombinant ApoF, respectively. B, inactivation of inhibitor activity by sodium oleate. LTIP (CM-cellulose fraction) and recombinant ApoF inhibitor activities were assayed by the LDL to biotin-LDL assay in the presence of the indicated amount of oleate. Data are the mean Ϯ S.D. for five (LTIP) or three (rApoF) determinations. C, dependence of inhibitor activity on the assay concentration of lipoproteins. LTIP (CM-cellulose fraction) and recombinant ApoF inhibitor activities were determined in assays measuring lipid transfer from LDL to HDL. The assay content of lipoproteins was varied as indicated while maintaining an LDL to HDL ratio of 1. Data are the mean Ϯ S.E. for 2-4 determinations. ity (37). Consistent with this finding, in other studies isolated apoA-I, apoA-II, and phospholipid transfer protein have been shown to stimulate CETP activity under some assay conditions (38 -41). The baboon plasma CETP inhibitor (N-terminal 38 amino acids of apoC-I) reported by Kushwaha et al. (8) also suppresses CETP activity in vitro, but with an IC 50 value of ϳ100 M (7) it seems unlikely that it has significant in vivo function in human plasma where apoC-I concentrations are Յ15 M (42). Overall, the capacity of various known apoproteins to stimulate and/or inhibit CETP activity has been highly assay-dependent and not supported by in vivo studies. Additionally, effects of these isolated apoproteins have not been demonstrated in more physiologic assays that approximate the conditions of whole plasma. By comparison, the inhibitory activity of LTIP/apoF has been shown in a wide variety of assays using isolated lipoproteins, synthetic liposomes, and unfractionated plasma as substrate. The estimated IC 50 value for LTIP/apoF is 0.05-0.15 M, much lower than that reported for other apoproteins.
In contrast to that previously reported, we demonstrate here that apoF is an LDL-associated protein and that little, if any, is associated with HDL in serum. The observation that apoF can be isolated from HDL suggests that some of this apoprotein may redistribute to this lipoprotein fraction with ultracentrifugation. Since the majority of LDL-associated apoF is lost during ultracentrifugation, the recovery of apoF in HDL under some centrifugation conditions seems probable. The near-exclusive association of LTIP/apoF with LDL and the observation that LTIP preferentially suppresses transfer reactions with LDL suggests that lipoprotein association is important to its function. This is consistent with limited kinetic analysis which suggests that LTIP binds to lipoproteins and inhibits CETP by preventing its association with the lipoprotein surface (4), and with our current data and with published data (1,2,4) showing that inhibitor activity is inversely related to assay lipoprotein concentration. With native LTIP/apoF (LDL-TMU fraction), we observed here that a 1:1 mole ratio of inhibitor to lipoprotein is sufficient to suppress CETP activity by 50%. This clearly suggests that there are a limited number of sites where LTIP/apoF and CETP compete for interaction on the lipoprotein surface. The plasma concentration of apoF has been reported to be 27 g/ml (0.82 M) (43), which is similar to a 70 g/ml (ϳ2 M) value that can be estimated from our purification data. Thus, at a typical plasma LDL concentration of 1.87 M (4), apoF is sufficiently abundant to reside on most LDL particles. This suggests that most LDL particles in plasma may support CETP activity at a reduced level due to the competition of apoF and CETP for residence on the lipoprotein surface.
The coupling of lipoprotein association and inhibitory function suggests that LTIP activity could be altered if its binding to lipoproteins is prevented or altered, perhaps by changes in lipoprotein composition. In preliminary studies, we have observed that a portion of LDL-associated LTIP/apoF is lost to the VLDL fraction when plasma is incubated at 37°C overnight in the presence of lecithin-cholesterol acyltransferase inhibitors. 2 The elucidation of lipoprotein properties that facilitate LTIP binding and the metabolic and pathologic conditions that alter this binding are under investigation.
In summary, we present data demonstrating that apoF possess the same biochemical and functional properties as LTIP. The weight of these data leads to the conclusion that apoF is LTIP. This the first report of a function for this minor apoprotein. Based on our understanding of LTIP from in vitro studies of its functional properties (1-4, 9, 10, 13-15), we suggest that apoF is an important regulator of cholesterol transport between plasma lipoproteins which directly affects the sterol content of individual lipoprotein fractions. The advances presented in this report will permit a more direct testing of this conclusion than previously possible.