The Signal Peptide Anchors Apolipoprotein M in Plasma Lipoproteins and Prevents Rapid Clearance of Apolipoprotein M from Plasma*

Lipoproteins consist of lipids solubilized by apolipoproteins. The lipid-binding structural motifs of apolipoproteins include amphipathic α-helixes and β-sheets. Plasma apolipoprotein (apo) M lacks an external amphipathic motif but, nevertheless, is exclusively associated with lipoproteins (mainly high density lipoprotein). Uniquely, however, apoM is secreted to plasma without cleavage of its hydrophobic NH2-terminal signal peptide. To test whether the signal peptide serves as a lipoprotein anchor for apoM in plasma, we generated mice expressing a mutated apoMQ22A cDNA in the liver (apoMQ22A-Tg mice (transgenic mice)) and compared them with mice expressing wild-type human apoM (apoM-Tg mice). The substitution of the amino acid glutamine 22 with alanine in apoMQ22A results in secretion of human apoM without a signal peptide. The human apoM mRNA level in liver and the amount of human apoM protein secretion from hepatocytes were similar in apoM-Tg and apoMQ22A-Tg mice. Nevertheless, human apoM was not detectable in plasma of apoMQ22A-Tg mice, whereas it was easily measured in the apoM-Tg mice. To examine the plasma metabolism, recombinant apoM lacking the signal peptide was produced in Escherichia coli and injected into wild-type mice. The apoM without signal peptide did not associate with lipoproteins and was rapidly cleared in the kidney. Accordingly, ligation of the kidney arteries in apoMQ22A-Tg mice resulted in rapid accumulation of human apoM in plasma. The data suggest that hydrophobic signal peptide sequences, if preserved upon secretion, can anchor plasma proteins in lipoproteins. In the case of apoM, this mechanism prevents rapid loss by filtration in the kidney.

Lipoproteins consist of lipids (mainly cholesterol, phospholipids, and triglycerides) solubilized by apolipoproteins. The apolipoproteins play essential roles in controlling plasma and tissue lipid homeostasis by interacting with cellular lipoprotein receptors (the low density lipoprotein receptor, the scavenger receptor class B type-I, the ATP-binding cassette transporter AI, etc.) and enzymes (e.g. lecithin-cholesterol acyltransferase, hepatic lipase, and lipoprotein lipase) and lipid transfer protein (cholesteryl ester transfer protein and phospholipid transfer protein). However, several apolipoproteins have roles beyond lipid metabolism. Indeed, a recent shotgun proteomic approach revealed that HDL 3 on average contains Ͼ40 proteins, of which many affect complement activation or protease inhibition (1). Also, several well known HDL apolipoproteins have roles in inflammation and host defense against microbial infections (1). For instance, apoL is involved in fighting Trypanosoma infections (2), and apoJ (clusterin) has even been proposed to play a significant role in tumorigenesis (3). For apolipoproteins without any (known) effect on plasma lipid metabolism, it is conceivable that the host HDL particle mainly serves as a carrier that prevents rapid removal of its associated apolipoproteins (e.g. by filtration in the kidney) or perhaps aids in their assembly at sites of inflammation.
The structural motifs conferring the lipid-binding capacity of apolipoproteins has been most intensively studied for apoB in low density lipoprotein (4) and apoA-I in HDL (5,6). Both amphipathic ␣-helixes (7) and ␤-sheets (4) bind to the lipids. In the case of apoA-I and several other HDL apolipoproteins (e.g. apoE and the apoCs), the non-covalent binding of the amphipathic ␣-helixes are rather weak, allowing exchange of the apolipoproteins between the triglyceride-rich very low density lipoprotein and chylomicron particles and HDL (8). Indeed, some apolipoproteins are so loosely bound to HDL (e.g. apoJ (9), apoA-IV (10), and lecithin-cholesterol acyltransferase (11) that they are often stripped from the lipoprotein particles during preparative ultracentrifugation.
ApoM (188 amino acids) was discovered in 1999 by Xu and Dahlback (12). In plasma, apoM is exclusively found in associ-ation with lipoproteins (mainly HDL). Recent studies of mice with genetically modified apoM expression showed that apoM affects plasma HDL metabolism and development of atherosclerosis (10,13), but apoM also has antioxidant effects, and its physiological roles remain to be determined (10). Analysis of the tertiary structure of apoM by in silico homology modeling has revealed that apoM belongs to the lipocalin protein superfamily (14). Like in other lipocalins, the structure of apoM contains eight antiparallel ␤-sheets forming a small lipid-binding pocket. Recent experimental data confirmed the lipocalin structure of apoM and showed that it can bind retinoic acid and retinol (15). Human apoM can be glycosylated at amino acid Asn-135 surrounded by the glycosylation signal Asn-Glu-Thr (12). Hence, Western blotting of SDS-PAGE gels with reduced human plasma samples typically reveals both glycosylated apoM and a non-glycosylated apoM of ϳ25 and ϳ20 kDa (9,12). The predicted structure of apoM lacks external amphipathic motifs that would explain the lipoprotein association of apoM. NH 2 -sequencing of plasma apoM has shown that it is secreted to the blood with its NH 2 -terminal signal peptide (12). This is unusual because intracellular cleavage of the NH 2 -terminal signal peptide by signal peptidases in the endoplasmic reticulum conventionally is seen as a prerequisite for cellular release of proteins to plasma. We are, however, aware of at least two other plasma proteins with preserved NH 2 -terminal signal peptides, i.e. paraoxonase-1 (PON-1) (16) and haptoglobin-related protein (HRP) (17). Like apoM, PON-1 and HRP are mainly found in a subfraction of HDL (although not the same HDL subfraction as that with apoM (9)), and in vitro studies suggest that the signal peptide in PON-1 can mediate its association with lipoproteins (16). Based on these findings and the highly hydrophobic nature of the signal peptide, we hypothesized that the signal peptide serves as an anchor for apoM in plasma lipoprotein particles.
To test this idea and explore the impact of the signal peptide on the metabolism of plasma apoM, we generated transgenic mice that express a mutated human apoM Q22A cDNA in the liver. The cDNA encodes a mutated human apoM Q22A protein where glutamine in position 22 is replaced by alanine. The Gln-22 3 Ala substitution results in secretion of apoM without its hydrophobic signal peptide (15). The mice expressing apoM Q22A were compared with transgenic mice expressing wild-type human apoM.

MATERIALS AND METHODS
ApoM-Tg and apo MQ22A -Tg Mice-Transgenic (Tg) mice expressing a truncated human apoM protein without the signal peptide were generated. Site-directed mutagenesis resulting in substitution of amino acid Gln-22 with Ala in human apoM (apoM Q22A ) was done previously (15). Secretion of apoM Q22A cDNA from transfected HEK293 cells was assessed as described (15) by Western blotting (9) and ELISA (18). A 774-bp fragment of the apoM Q22A -encoding cDNA was amplified with Pfu DNA polymerase system (Stratagene, AH Diagnostic) and primers (pcDNA3-apoMQ22A-51:5Ј-gcggccgcactggcggccgttactagtggat-3Ј and pcDNA3-apoM Q22A -31:5Ј-gcggccgcgcagtaggtgtccaccatgttcc-3Ј) and cloned into a PCR-Blunt II-TOPO vector (Invitrogen A/S). The ApoM Q22A encoding cDNA was subsequently cloned into a pGEMAlbSV40 vector (obtained from Ragnar Matsson, Lund University) between a murine albumin promotor/enhancer sequence and an SV40 intron/ poly(A) sequence. The correctness of apoM Q22A encoding cDNA was verified by DNA sequencing (primers will be given by request). The 4446-bp albumin-apoM Q22A -SV40 fragment was purified from the plasmid by restriction digestion with Nsi and ApaI, agarose gel electrophoresis and the QIAEX II gel extraction kit (Qiagen-Nordic) and used for pronuclear microinjections into fertilized mouse oocytes (C57/Bl X CBA) at the Transgenic Core Facility, University of Lund, Sweden. Human apoM Q22A transgenic founders were identified by real-time PCR amplification (LightCycler, Roche Diagnostic) of a 175-bp fragment of human APOM from tail DNA with primers h-apoM-51 and h-apoM-31 (10). The human apoM-transgenic mice expressing wild-type human apoM (apoM-Tg) were described recently elsewhere (as apoM-Tg N mice (10)). Both apoM-Tg and apoM Q22A -Tg mice were backcrossed with C57Bl/6 mice for at least five generations before the present studies.
The mice were housed at the Panum Institute (University of Copenhagen) and fed standard chow (Altromin 1314, Brogaarden). Blood samples were taken from the venous plexus in the orbital cavity into Na 2 EDTA-or heparin-containing tubes and kept on ice. Plasma was obtained by centrifugation at 3000 rpm for 10 min at 4°C and stored at Ϫ80°C or Ϫ20°C. All procedures were approved by the Animal Experiments Inspectorate, Ministry of Justice, Denmark.
Primary Hepatocyte Cell Culture-Primary hepatocytes were isolated from ApoM-Tg N , ApoM Q22A -Tg, and wild-type mice (21). Briefly, mice were perfused with 5 ml of liver perfusion medium (Invitrogen) followed by 5 ml of digestion medium containing 4 mg/ml collagenase type I (C2624, Sigma-Aldrich). The liver was then placed in a sterile tube with 5 ml of digestion medium at 37°C for 20 min before being placed in ice-cold L-15 cell medium (Invitrogen). Subsequently, the tissue was homogenized with a Pasteur pipette, passed through a 70-m nylon filter, and washed with hepatocyte washing medium (Invitrogen). Centrifugation between washes was done at 50 ϫ g for 5 min. The cells were resuspended in 5 ml of F-12/Dulbecco's modified Eagle's medium (Invitrogen) with 1% penicillin/streptomycin, 10% bovine calf serum, and 2 mM L-glutamine (G6392, Sigma-Aldrich) and grown on 6-well collagen-coated plates (152034, Nunc). The concentration of human apoM in cell culture medium was measured after 48 h. Lipoproteins (d Ͻ 1.21 g/ml) were separated from non-lipoproteins in plasma or concentrated (ϫ4) cell culture medium by adjusting 20 l of sample to d ϭ 1.21 g/ml with NaBr and ultracentrifugation in a TLA-100 ultracentrifuge (Beckman) using 200-l tubes at 15°C for 4 h and 100,000 rpm.
Plasma Turnover and Tissue Elimination of Human apoM  -To study the plasma turnover of apoM without the signal peptide, recombinant apoM  or wild-type human apoM were injected intravenously into recipient wildtype mice, and the plasma turnover was assessed by measuring human apoM plasma concentrations with ELISA. Recombinant human apoM 22-188 was produced as described (18), and protein concentration was measured with the Pierce BCA protein assay kit (Bie and Berntsen A-S) using bovine serum albumin as a standard (9). For each recipient mouse, apoM  (100 g) in 148 l of phosphate-buffered saline was mixed with 90 l of plasma from a wild-type mouse and injected intravenously into a wild-type mouse (n ϭ 3); control mice (n ϭ 3) received 148 l of phosphate-buffered saline mixed with 90 l of plasma from human apoM-Tg mice that overexpress human apoM 10-fold (10). Blood samples were taken after 5, 30, 120, and 1400 min for measurement of human apoM. To assess the lipoprotein association of human apoM, lipoproteins (d Ͻ 1.21 g/ml) were separated from proteins in plasma samples taken after 30 min as described, above and the human apoM concentration was determined in the fractions by ELISA.
To determine tissue uptake and degradation, we labeled apoM  or albumin with 125 I-TC and injected it into wildtype mice. The 125 I-TC moiety is trapped inside cells upon lysosomal degradation of the labeled protein, enabling in vivo studies of degradation sites of plasma proteins (22). First, tyramine cellobiose (TC) (50 nmol) was labeled with 37 MBq 125 I in tubes coated with 10 g of Iodogen (Invitrogen) (23). Then, the 125 Ilabeled TC was activated and coupled to 100 -120 g of apoM  or albumin at pH 9 -10. To separate the unbound 125 I and 125 I-TC from 125 I-TC-albumin and 125 I-TC-apoM  , the mixture was passed over a PD-10 column (Amersham Biosciences). The labeling efficiency was 34% for apoM  and 22% for albumin, and 91 and 86% of the 125 I in the doses were precipitated with 15% trichloroacetic acid for apoM  and albumin, respectively. The integrity of the labeled proteins was examined by 12% SDS-PAGE, and visualization of the radioactivity was done with a Fujix BAS 200 bioimaging analyzer (Fuji Photo Film).
The concentration of 125 I-TC-labeled proteins was determined by the BCA Pierce kit (Pierce), and 125 I-TC-apoM  or 125 I-TC-albumin (10 g, 3.2-5.3 ϫ 10 6 cpm) was injected into a tail vein of wild-type mice (n ϭ 2 ϫ 3) in a total volume of 330 l. Blood samples were taken 10 min and 6 h after injection before the mice were anesthetized and perfused with phosphate-buffered saline. Plasma samples and biopsies from liver, kidney, lung, spleen, eyes, heart, brain, gut, and gall bladder were placed in 1 ml of phosphate-buffered saline and counted in a 1470 automatic gamma counter (PerkinElmer Life Sciences) for 30 min.
To examine the cellular distribution of labeled apoM 22-188 , 1-m Epon sections were obtained from kidneys fixed by immersion in 4% formalin 6 h after injection of labeled protein.
The tissue was postfixed in OsO 4 , dehydrated, and embedded in Epon 812. The Epon sections were processed for autoradiography using Ilford K2 emulsion, exposed for 1-4 weeks, and developed in Kodak D19. The sections were examined in a Leica DMR microscope equipped with a Leica DFC320 camera. Images were transferred by a Leica TFC Twain 6.1.0 program and processed using Adobe PhotoShop 8.0.
Exclusion of Kidney Elimination of Plasma Proteins-To exclude kidney elimination of plasma proteins, apoM Q22A -Tg (n ϭ 6), apoM-Tg N (n ϭ 3), and wild-type mice (n ϭ 3), animals were anesthetized, and the kidney was dissected free as described previously (24) to enable bilateral ligation of the kidney artery. The musculofascial and skin incisions were then sutured, and the animals were given buprenorphine (0.001 mg/10 g of body, subcutaneously) for analgesia while kept anesthetized for another 4 h. Blood samples were taken before and 1 and 4 h after the kidney artery ligation. Q22A in Transgenic Mice-To examine the impact of the signal peptide sequences on the metabolism of apoM, we used a mutated human apoM Q22A cDNA driven by albumin enhancer-promoter sequences to make Tg mice. The Gln-22 3 Ala substitution introduces a signal peptidase cleavage site and results in secretion of a truncated apoM without the signal peptide from HEK293 cells (Fig. 1a). Twelve founders were generated by pronuclear injection of the transgene into fertilized mouse eggs. One transgenic line with an estimated ϳ32 copies of the transgene was bred for further studies and compared with Tg mice expressing wild-type human apoM (10). The apoM Q22A -Tg mice thrived well, bred normally, and were indistinguishable from wild-type mice.

Expression of ApoM
The liver expression of human apoM mRNA in the apoM Q22A -Tg mice was ϳ40% of that in apoM-Tg mice (Fig.  1b). However, on Western blotting of mouse plasma with a human apoM-specific polyclonal antibody, apoM was only detectable in the apoM-Tg and not in apoM Q22A -Tg mice (Fig.  1c). Of note, the polyclonal anti-human apoM antibody recognized recombinant apoM 22-188 as well as wild-type human apoM (Fig. 1c). This implies that the lack of apoM in apoM Q22A -Tg mouse plasma does not reflect the lack of reactivity of the antibodies toward truncated apoM. In agreement with the Western blotting results, human apoM could not be detected in the plasma from apoM Q22A -Tg mice with a human apoM-specific sandwich ELISA when the plasma was diluted as little as 50 times. The ELISA uses two different monoclonal antibodies to human apoM and reacts well with both wild-type human apoM and recombinant apoM without the signal peptide sequences (Fig. 1e). The detection limit is ϳ0.3 nM. Hence, the concentration of human apoM in apoM Q22A -Tg mice is Ͻ15 nM. In contrast, the concentration of human apoM is ϳ2 M in the human apoM-Tg mice. The absence of detectable apoM Q22A in plasma was also seen in analyses of eight additional apoM Q22A -Tg founders and thus was not dependent on the site of transgene integration in the mouse genome. Overex-pression of human apoM had no discernable effect on the plasma levels of endogenous mouse apoM (Fig. 1d).
Since apoM Q22A is secreted from transfected HEK293 cells, we reasoned that the absence of apoM Q22A in the apoM Q22A -Tg mouse plasma most likely did not reflect impaired secretion from the mouse liver. To test this assumption, primary hepatocytes from apoM Q22A -Tg, apoM-Tg, and wild-type mice were grown in tissue culture plates. Analysis of the medium by ELISA (Fig. 2) and Western blotting (not shown) showed that human apoM was secreted into the medium from both apoM Q22A -Tg and apoM-Tg hepatocytes. The concentration of apoM Q22A in the medium was ϳ40% of that of wild-type human apoM. Hence, the relative rate of secretion of apoM Q22A and wild-type human apoM from the cultured hepatocytes was similar to the relative amount of apoM Q22A and wild-type human apoM mRNA expression in the liver. The virtual absence of apoM Q22A in plasma despite ample secretion from the liver suggested that the lack of the signal peptide might cause rapid clearance of apoM Q22A from plasma because apoM Q22A cannot associate with plasma lipoproteins.  (15). Human plasma (0.5 l) and concentrated (ϳϫ40) cell culture medium from transfected HEK293 cells (3 l) were analyzed by Western blotting after SDS-PAGE using an anti-human apoM antibody from BD Biosciences. b, ethidium bromide-stained 2% agarose gel showing human apoM mRNA expression in apoM Q22A -Tg and apoM-Tg mouse livers; wild-type mouse liver was included as negative control. c, Western blot of plasma from apoM Q22A -Tg, apoM-Tg, and wild-type mice with a polyclonal human apoM antibody. Human plasma (1, 0.5, or 0.25 l of a plasma pool with ϳ23 g/ml apoM) and recombinant apoM apoM  (50, 25, or 5 ng/lane) were included as controls to document that the antibody reacts with both wild-type human apoM (h-apoM) and apoM without the signal peptide. d, Western blot of mouse apoM (m-apoM) in plasma (0.5 l) from apoM-Tg, apoM Q22A -Tg, and wild-type mice using a polyclonal mouse apoM antibody. Human plasma was included as negative control. e, the relationship between the concentration of human apoM (in the well) versus resulting absorbance in the human apoM ELISA. The results were obtained with a dilution series of human plasma (open circles) or recombinant apoM

(filled circles).
Impact of the Signal Peptide on Plasma Metabolism of apoM-To explore the impact of the signal peptide on the plasma metabolism of apoM, we produced a recombinant truncated human apoM  without the signal peptide (corresponding to amino acids 22-188 of wild-type human apoM) in Escherichia coli and injected it (100 g) into wild-type mice. ELISA measurements showed that apoM  was rapidly cleared from plasma with only ϳ5% of the injected dose remaining in plasma 2 h after the injection (Fig. 3a). We also examined the plasma metabolism of wild-type human apoM by injecting plasma from apoM-Tg mice into wild-type mice; the clearance of wild-type human apoM was much slower than that of apoM  with ϳ50 and ϳ5% of the injected wild-type human apoM remaining in plasma after 2 and 24 h, respectively (Fig. 3a). To analyze the lipoprotein association of apoM  , lipoproteins were separated from the remaining plasma proteins by ultracentrifugation at d ϭ 1.21 g/ml using plasma collected 30 min after injection of apoM  or wild-type human apoM particles. ApoM  was recovered in the protein fraction, whereas wild-type human apoM predominantly was in the lipoprotein fraction (Fig. 3b).
Due to the small size (ϳ20 kDa) and the lack of lipoprotein association of apoM  , we suspected that it would be rapidly removed by filtration in the kidney. We could, however, not detect any human apoM in the urine of apoM Q22A -Tg mice by ELISA or Western blotting, suggesting that if the kidney was a primary site for removal of apoM without the signal peptide, then the filtered apoM might be taken up in the proximal tubule rather than being excreted into the urine. Indeed, we previously showed that the endocytotic receptor megalin, which is highly expressed on the luminal side of the proximal tubule epithelium, can bind and internalize apoM (26). To explore the tissue sites for uptake and degradation of apoM 22-188 , we labeled apoM  with 125 I-TC and injected it into wild-type mice. Parallel studies were done with human albumin. 125 I-TC label-   JULY 4, 2008 • VOLUME 283 • NUMBER 27 ing did not affect the integrity of apoM  or albumin as both proteins migrated as one band with the expected size in SDS-PAGE gels (Fig. 4a). Similar to non-labeled apoM  , 125 I-TC-apoM  was rapidly removed from plasma (Fig. 4b). Six hours after injection, the main portion of the 125 I-TC-apoM  was recovered in the kidney, whereas 125 I-TC-albumin was recovered in the plasma, liver, and kidney (Fig. 4c). To examine which cells had taken up 125 I-TC-apoM  , sections of the recipient mouse kidneys were subjected to autoradiography. The vast proportion of 125 I in the kidney was seen in proximal tubule epithelial cells, whereas only a few grains were seen in glomerular cells or distal tubules (Fig. 4d). The rapid plasma elimination and uptake by kidney proximal tubule cells of human apoM  were reiterated when using 125 I-TC-labeled recombinant mouse apoM without its signal peptide sequences (data not shown).

Signal Peptide Anchors apoM in Plasma Lipoproteins
The turnover studies using recombinant apoM  suggested that the absence of human apoM in the plasma of apoM Q22A -Tg mice reflects rapid removal of apoM Q22A in the kidney and that human apoM would accumulate in the plasma of the apoM Q22A -Tg mice if kidney elimination were prevented. To test this idea, we excluded kidney degradation of plasma proteins by bilateral ligation of the renal arteries in anesthetized apoM Q22A -Tg, apoM-Tg, and wild-type mice and measured human apoM in plasma with ELISA. Indeed, human apoM became detectable in plasma of the apoM Q22A -Tg mice already 1 h after eliminating kidney excretion (Fig. 5), which strongly supports the conclusion that apoM can be secreted from the liver without its signal peptide but that the lack of signal peptide results in rapid elimination of the truncated plasma protein by the kidney. The concentration of human apoM in apoM-Tg mice did not increase after bilateral ligation of the kidney arteries (Fig. 5).

DISCUSSION
The NH 2 -terminal signal peptide plays important roles in the intracellular processing of proteins. It can determine protein folding (e.g. EspP (27)) and direct proteins into specific secretory pathways (e.g. pro-opiomelanocortin (28)) or cellular compartments (e.g. PB2 of the influenza virus is directed to the mitochondria (29)). Also, the hydrophobic nature of the signal peptide can anchor proteins in the cell membrane. Failure of cleavage of the signal peptide of proteins that are normally secreted can result in intracellular degradation (30). In the present study, however, human wild-type apoM and human apoM Q22A appeared to be equally effectively secreted from primary mouse hepatocytes. The results suggest that the signal peptide is responsible for apoM's lipoprotein association and is crucial to prevent rapid clearance of apoM from plasma. The present in vivo findings are in accord with a recent in vitro observation by Axler et al. (25), suggesting that apoM Q22A fails to associate with HDL in the culture medium when expressed in HEK293 cells.
ApoM shares its preservation of the NH 2 -terminal signal peptide sequences with PON-1 and HRP (16,17). In plasma, all three proteins are mainly in the HDL fraction. The apoM-containing HDL fraction is not enriched in PON-1 and HRP (9), and HRP is mainly part of an HDL subfraction named the trypanosome lytic complex, which also contains apoA-I and apoL (17,31). Thus, the signal peptides do not anchor the three proteins to the same HDL subfraction. Ovalbumin (32) and PAI-2 (33) are also secreted without cleavage of their signal peptides. However, in those cases, the signal peptide sequences are not at the NH 2 terminus of the proteins, and the hydrophobic residues are embedded in the three-dimensional structures of the proteins (34). Hence, neither ovalbumin nor PAI-2 is lipoprotein-bound.
In silico sequence analyses using the signalP software predict that the preservation of the signal peptides in apoM and PON-1 results from the lack of a signal peptide cleavage site and, for both proteins, single amino acid substitutions can result in removal of the signal peptide prior to secretion (15,16). In the case of HRP, however, the signalP software predicts that the signal peptide can be cleaved, and in vitro expression in 293 cells of an HRP encoding cDNA showed that the major fraction of HRP indeed is secreted from the cultured cells without the signal peptide (35). Nevertheless, plasma HRP does contain the signal peptide. The present data provide a possible explanation for the apparent discrepancy; even if only a subfraction of HRP is secreted with the signal peptide sequences, the anchoring of HRP with the signal peptide in HDL may prevent clearance from plasma, whereas HRP without the signal peptide (ϳ35 kDa) predictably will be rapidly removed from plasma.
The physiological role of apoM remains to be elucidated. Genetic elimination or overexpression of apoM in mice affects plasma lipid metabolism (10,13), and in healthy humans, the plasma apoM concentration is positively associated with total cholesterol concentration (18). Nevertheless, apoM may have roles beyond lipid metabolism; HDL carries multiple proteins with roles in inflammation and innate immune responses, and HDL has pronounced anti-inflammatory effects in animal models (1,36). Thus, it is possible that apoM, only being present in ϳ5% of the plasma HDL particles (9), may also have effects unrelated to lipid metabolism. For instance, apoM protects against Cu 2ϩ -induced lipid oxidation (9, 10) and binds retinol and retinoic acid in vitro (15). The latter observation is explained by the lipocalin structure of apoM where a small hydrophobic binding pocket enables binding of small lipophilic molecules (14). It is interesting to note that apoM is also found in a highly conserved major histocompatibility complex class III cluster in primitive organisms such as fish, indicating conservation of the cluster for more than 450 million years (37), further suggesting that apoM may have essential biological functions. The present data document that the signal peptide in apoM serves an important function by anchoring the apolipoprotein to lipoproteins, thus inhibiting the elimination of apoM in the kidney and maintaining stable apoM concentrations in plasma.