Vitellogenin Recognizes Cell Damage through Membrane Binding and Shields Living Cells from Reactive Oxygen Species*

Background: Vitellogenin is a central regulator of honey bee life span by largely unknown mechanisms. Results: Honey bee vitellogenin has membrane affinity that is connected to cell damage recognition and antioxidant function. Conclusion: Membrane binding documents a new molecular behavior among vitellogenins. Significance: Vitellogenins are widespread phylogenetically, and their molecular behavior is essential for fitness traits in many animals. Large lipid transfer proteins are involved in lipid transportation and diverse other molecular processes. These serum proteins include vitellogenins, which are egg yolk precursors and pathogen pattern recognition receptors, and apolipoprotein B, which is an anti-inflammatory cholesterol carrier. In the honey bee, vitellogenin acts as an antioxidant, and elevated vitellogenin titer is linked to prolonged life span in this animal. Here, we show that vitellogenin has cell and membrane binding activity and that it binds preferentially to dead and damaged cells. Vitellogenin binds directly to phosphatidylcholine liposomes and with higher affinity to liposomes containing phosphatidylserine, a lipid of the inner leaflet of cell membranes that is exposed in damaged cells. Vitellogenin binding to live cells, furthermore, improves cell oxidative stress tolerance. This study can shed more light on why large lipid transfer proteins have a well conserved α-helical domain, because we locate the lipid bilayer-binding ability of vitellogenin largely to this region. We suggest that recognition of cell damage and oxidation shield properties are two mechanisms that allow vitellogenin to extend honey bee life span.

Large lipid transfer proteins are involved in lipid transportation and diverse other molecular processes. These serum proteins include vitellogenins, which are egg yolk precursors and pathogen pattern recognition receptors, and apolipoprotein B, which is an anti-inflammatory cholesterol carrier. In the honey bee, vitellogenin acts as an antioxidant, and elevated vitellogenin titer is linked to prolonged life span in this animal. Here, we show that vitellogenin has cell and membrane binding activity and that it binds preferentially to dead and damaged cells. Vitellogenin binds directly to phosphatidylcholine liposomes and with higher affinity to liposomes containing phosphatidylserine, a lipid of the inner leaflet of cell membranes that is exposed in damaged cells. Vitellogenin binding to live cells, furthermore, improves cell oxidative stress tolerance. This study can shed more light on why large lipid transfer proteins have a well conserved ␣-helical domain, because we locate the lipid bilayer-binding ability of vitellogenin largely to this region. We suggest that recognition of cell damage and oxidation shield properties are two mechanisms that allow vitellogenin to extend honey bee life span.
Large lipid transfer proteins (LLTPs) 2 have multiple roles in animals; they are lipid transporters, inflammation suppressors, immunomodulators, and blood coagulators. Vitellogenin (Vg) is the ancient form of these proteins, with an estimated 700 million-year history (1). In most egg-laying species, Vg provides the eggs with lipids. Non-egg-laying vertebrates lack Vg but have homologs, such as apolipoprotein B (apoB), that transports cholesterol (2). Vgs are mostly produced in the liver or hepatopancreas (vertebrates, crustaceans) or fat tissue (insects), secreted into the blood, and taken up by targets, such as eggs, via receptor-mediated endocytosis (3,4).
Many blood proteins are involved in innate immunity (for a review, see Ref. 5). They recognize pathogen-associated molecular patterns (PAMPs), such as lipopolysaccharides, or endogenous damage-associated molecular patterns (DAMPs), such as phosphatidylserine (PS) (6). Complement factors are perhaps the most studied group of these pattern recognition proteins (7). Also, collectins and pentraxins identify both PAMPs and DAMPs. They, among other targets, bind to membrane protrusions of damaged cells called blebs (8,9). Like the LLTP family, Vg recognizes a variety of PAMPs and has antibacterial, antiviral, and antifungal effects studied in several fish, including salmon (Salmo salar) and zebrafish (Danio rerio) (10,11). ApoB suppresses inflammation by neutralizing both PAMPs (12,13) and necrotic cells (14). Broad-range danger recognition by individual proteins can be explained by their general affinity with exposed hydrophobic molecules and surfaces that many PAMPs and DAMPs consist of (6).
There is a conserved vitellogenin domain (Vg domain) located in the N terminus of LLTPs. The Vg domain has lipid affinity, and it appears to be important for the assembly of the lipoprotein complex (15,16). The apoB Vg domain binds to egg phosphatidylcholine (PC) monolayers and dimyristoylphos-phatidylcholine multilamellar vesicles that have been used in the studies of the apoB-lipid packing (15,17), which takes place at the endoplasmic reticulum membrane. There is further evidence of direct binding of apoB to the epithelial cell membrane via the Vg domain (18), but the mechanisms and consequences of this cell binding are unclear. The Vg domain consists of two subdomains: the N-sheet domain and the ␣-helical domain. N-sheet alone has been identified as the Vg receptor-binding region in fish (19). The role of the ␣-helical domain is less clear. Outside of the Vg domain, all known Vgs (but not apoB) have a conserved but poorly characterized von Willebrand factor type D domain (vWFD) at their C terminus (3,20).
LLTPs become easily oxidized in blood (for a review, see Ref. 21). Excessive oxidation of human apoB and its lipids is linked to atherosclerosis, which is an arterial response to chronic inflammation (22). In the honey bee (Apis mellifera), on the contrary, oxidative carbonylation of Vg is linked to beneficial effects. Vg titer positively correlates with the honey bee's oxidative stress tolerance, because the antioxidant property of Vg shields other hemolymph (the insect blood) molecules from reactive oxygen species (ROS) (23). According to the free radical theory of aging, the balance between ROS and antioxidants is essential in inhibition of chronic inflammation and agingassociated diseases (24). Generally, antioxidants include proteins that are located either in intra-or intercellular fluid or on membranes, where they can be permanently (25) or temporarily (26) attached.
Honey bee Vg appears to have also other life-extending effects. It supports immune cell viability (27) and suppresses risky foraging behavior performed by the essentially sterile helper (worker) bees of the colony (28). Excessive buildup of Vg (up to 100 mg/ml) in hemolymph is characteristic of an extremely long lived worker phenotype called winter bees that can survive at least 8 months, in contrast to the 4 -8 weeks of a normal worker life span (29). Atypical changes in the workers' seasonal Vg expression can furthermore be used as an indicator of the disease syndrome colony collapse disorder, which in recent years has caused a dramatic loss of commercial honey bee populations worldwide (30).
Typically, honey bee Vg is a monomeric ϳ180-kDa protein, but the N-sheet of the Vg domain can dissociate in a probably phosphorylation-regulated process in fat tissue, leaving a 150-kDa putatively receptor-free Vg fragment remaining (31,32). However, although honey bee Vg is a molecularly, physiologically, ecologically, evolutionarily, and commercially interesting protein, little is known about how this molecule confers its beneficial effects.
We hypothesized that affinity to hydrophobic surfaces, similarly to many serum proteins and the Vg domain in apoB, could be one of the shielding mechanisms of Vg in the honey bee. If this ROS scavenger protein had some affinity to the cell membrane, it could benefit cells under oxidative stress. On the other hand, affinity to DAMP-containing dying cells could link Vg to the anti-inflammatory function of apoB. To assess our hypothesis, we investigated Vg localization in honey bee tissues and the interaction of Vg with healthy, apoptotic, and dead insect cells. We also assayed Vg binding to membrane-mimicking PC and PS/PC bilayers. Vg was found to interact with tissues, cells, and bilayers, and the binding to dead cells and PS/PC was enhanced compared with healthy cells and PC. Furthermore, we used limited proteolysis, homology modeling, and tryptophan fluorescence to understand the binding mechanism and tested the potential of Vg to disrupt PS-liposomes. Finally, we tested whether Vg on the cell surface can affect the oxidative stress tolerance of live cells.

EXPERIMENTAL PROCEDURES
The Antibody, Gels, and Western Blots-The polyclonal Vg antibody produced in rabbit (Pacific Immunology, Ramona, CA) has been tested before (29,31,33). This antibody was used for all imaging and Western blotting. The gel and Western blot reagents were purchased from Bio-Rad. The Western protocol for all of the blots was as follows. The blotted nitrocellulose membrane was incubated with TBS containing 0.5% Tween with 2.5% fat-free milk powder overnight. The membrane was incubated for 1 h with the Vg antibody and for 1 h with a horseradish peroxidase-conjugated secondary antibody prior to imaging using an Immun-Star kit. All gels and blots were imaged, and band intensities were measured using a ChemiDoc XRSϩ imager (Bio-Rad).
Material Sampling, Protein Purification, and Cell Lines-Handling of colony sources, honey bee tissue dissection, Vg extraction, and Vg purification was performed as described elsewhere (31). Single ion exchange purification was used for hemolymph vitellogenin (hlVg) purification. Fat body vitellogenin (fbVg) was purified from fat body protein extract (31) by GenScript with size exclusion chromatography and ion exchange steps (31), and the purification was completed with a Superdex 75 column (Sigma-Aldrich). The ␣-helical (residues 416 -778) and vWFD (residues 1444 -1615) domains were produced by GenScript; the vWFD DNA was subcloned into pUC57 vector, and the ␣-helical domain was subcloned into pCold trigger factor vector. The N-terminal hexahistidine tag of the fragments was used for one-step affinity purification. The ␣-helical domain is a fusion construct with trigger factor that is a highly soluble protein with chaperone activity (34), and separation of the ␣-helical fragment from its fusion partner by HRV 3C cleavage was not feasible due to fragmentation of the ␣helical domain.
We used Sf9 or High Five cells in the in vitro experiments. Sf9 is a pupal Spodoptera frugiperda cell line, and High Five is a parental line of Trichopulsia ni (see provider's information; Invitrogen). Most experiments were performed with High Five cells due to their shorter doubling time.
Histology-Dissection, tissue preparation, and immunohistochemistry were largely performed as described previously (29). In brief, tissue samples of mature nest bees (age Ͻ27 days) were fixed in paraformaldehyde (4% in PBS), embedded in London Resin White (Electron Microscopy Science, Hatfield, PA), and cut with a Reichert Jung ultra-microtome (Leica, Wetzlar, Germany; section thickness, 1-2 m). Mounted sections were rinsed with PBS-NTx (0.25% Triton X-100 in PBS), preincubated with 2% bovine serum albumin (BSA) (Sigma-Aldrich) in PBS-NTx for 60 min, and incubated overnight with the anti-Vg antibody (1:100). After washing in PBS-NTx, a Cy5-conjugated anti-rabbit antibody (Jackson ImmunoResearch, West Grove, PA; 1:400) and the nuclear stain 4Ј,6-diamidino-2-phenylindole (DAPI; 1:1,000 from 0.5 mg/ml stock; Sigma-Aldrich) were coapplied. Finally, sections were rinsed in PBS-NTx and cleared in glycerol (30% in PBS). To rule out false positives, controls that were not incubated with the primary antibody were included. Two controls and two test samples were prepared for each of five individuals. Confocal images were acquired on a Leica TCS SP5 laser-scanning confocal microscope, using a ϫ63 oil immersion objective (numerical aperture ϭ 1.4). Image stacks (z ϭ 2 m, ⌬z ϭ 0.5 m) or single optical sections (control versus test comparisons) were viewed and processed in ImageJ version 1.44b (National Institutes of Health).
Membrane Protein Immunoblot-The membrane protein extraction protocol was modified from Refs. 35 and 36. The sample was kept at 0 -4°C, and buffers contained a protease inhibitor mixture (Roche Applied Science). Five frozen diutinus bee abdomens, gut and ovary removed, were homogenized (as in Ref. 31) and centrifuged at 800 ϫ g for 10 min in order to exclude exoskeleton and nuclei. The supernatant was centrifuged at 30,000 ϫ g for 20 min. The resulting supernatant was filtrated (0.2-m syringe filter; Pall Corp., Port Washington, NY) and stored as the cytosolic fraction. The membrane-containing pellet was washed three times with 1 ml of HBS, centrifuged at 30,000 ϫ g for 20 min, and suspended in HBS containing 0.1% Triton X-100. 24 g of the cytosolic and the membrane proteins were applied on a 7.5% SDS-polyacrylamide gel and blotted.
Cell Binding Assay-The Sf9 cell test was modified from Ref. 37. Centrifugations were 5 min at 2,000 ϫ g in room temperature, and the wash volume was 0. The samples were stored in 100 l of 2% paraformaldehyde in PBS and analyzed the following day with LSRII (BD Biosciences). To assist the analysis, there was a cells-only sample, a single-color control for each fluorophore, and a Vg-negative sample with secondary antibody for monitoring background staining.
Immunocytochemistry-6.8 ϫ 10 5 High Five cells were grown on chamber slides (LabTek) overnight. The cells were incubated with 5.3 M fbVg in PBS (volume 50 l) purified from honey bee abdomens (see sample purification) for 1 h in 27°C and washed with HBS. Cells were incubated with 10 l of annexin V Alexa Fluor 647 conjugate (Invitrogen) in 200 l of HBS in the dark followed by HBS wash. One sample set was stained for 5 min with 10 l of propidium iodide (PI) before wash to stain necrotic and late apoptotic cells. Cells were fixed with 2% paraformaldehyde for 10 min and washed twice with ice-cold PBS. 1% BSA in PBS was used for 30-min blocking. The Vg antibody (1:5000) 30-min incubation was followed by three PBS-BSA washes, and the 1:200 Alexa Fluor 488 goat anti-rabbit IgG (Jackson ImmunoResearch) 30-min incubation was followed by three PBS-BSA washes. The fluorescence images were acquired on a Zeiss Axiovert200M fluorescence microscope (Thornwood, NY) using ϫ20 and ϫ40 magnification dry objectives, excitation of 492 and 622 nm, and emission of 535 and 668 nm.
We used BSA as a control for nonspecific binding to dead cells. 5 ϫ 10 5 Sf9 cells killed by freezing (Ϫ80°C) in 50 l of PBS were incubated with 2.65 M hlVg or BSA in a total 100-l volume for 1 h in 27°C and washed twice with PBS. Fixing, blocking, and antibody treatments were as above, but peptone (Biokar Diagnostics) was used instead of 1% BSA. The anti-BSA primary antibody (produced in rabbits; 1:1000 dilution) was purchased from Invitrogen. The samples were mounted with ProLong Gold reagent with DAPI (Invitrogen). The secondary antibody control was a BSA-treated sample, where primary antibody incubation was omitted. The cells were imaged with a Zeiss Axio Imager M2. All images were processed in ImageJ.
Liposome Preparation-Two kinds of liposomes were prepared: PC and PC ϩ PS in a 1:1 mixture. Egg yolk PC and porcine brain PS (Avanti Polar Lipids, Alabaster, AL) mixed 1:1 with PC were dissolved in chloroform, which was evaporated, resulting in formation of a lipid film. Samples were lyophilized for 2 h to remove traces of chloroform and resuspended in HBS (for surface plasmon resonance (SPR)) or in 10 mM Hepes, 100 mM NaCl, including 12.5 mM 8-aminonaphtalene-1,3,6-trisulfonic acid and 45 mM p-xyelene-bis-N-pyrimidium bromide, pH 7.0 (for leakage), and left overnight in the dark at room temperature. The hydrated phospholipid solutions were then frozen and thawed seven times, followed by extrusion (11 times) with the Avanti Mini-Extruder using a membrane of 0.1-m pore size. The leakage-intended liposomes were filtered with a Sephadex G-75 column to remove unencapsulated 8-aminonaphtalene-1,3,6-trisulfonic acid and p-xyelene-bis-N-pyrimidium bromide. Determination of phospholipid concentration was based on the determination of total phosphorus using a protocol from Avanti Polar Lipids (38). The liposome size was determined by dynamic light scattering (Zetasizer Nano): PC 102 nm and PS 108 nm for SPR, and PC 122 nm and PS 191 nm for leakage.
SPR-The L1 chip was conditioned with 40 mM octyl glucoside (30 s; flow rate, 30 l/min) using the Biacore T200 instrument (GE Healthcare). The chip was fully regenerated by injecting 2:3 isopropyl alcohol/NaOH (30 s, 30 l/min) after each sample injection. Two kinds of liposomes were used: PC and PC ϩ PS in a 1:1 mixture (the latter hereafter referred to as PS). Concentrations of 0.5 mM PC and 1.2 mM PS liposomes were captured on separate channels of the chip (30 s, 10 l/min), and unbound structures were washed with 10 mM NaOH, followed by a stabilization time of 300 s. The binding temperature was 25°C, whereas the sample compartment was kept at 4°C. The liposome surface coverage was verified according to the Biacore user manual's instructions using BSA that binds on the naked chip but has low affinity toward a liposome-covered chip. The following protein concentrations of hlVg, fbVg, the ␣-helical fragment, and vWFD were used: 0, 100, 500, 1,500, and 2,500 nM. After each sample injection (60 s, flow rate 30 l/min, dissociation time 600 s), the chip was fully regenerated.
Liposome-combined Proteolysis and LC-MS/MS-PC (size 120 nm) and PS (99 nm) liposomes were incubated with fbVg (see "Liposome Preparation" and "Material Sampling, Protein Purification, and Cell Lines") for 1 h at room temperature in the dark and digested with 25 g of chymotrypsin for 30 min on ice with the following setup: 1 l of 0.1 M liposomes and 8 l of 0.7 mg/ml fbVg, the total volume adjusted to 12 l with HBS. The controls were fbVg without liposomes or protease, fbVg with PC or PS, and fbVg with chymotrypsin. The samples were run on a 4 -20% SDS-polyacrylamide gel, and band intensities were measured. The protease-resistant band in the liposome samples was excised and analyzed using LC-MS/MS; the gel pieces were washed, treated with 10 mM DTT (Amersham Biosciences) for cysteine reduction, and alkylated using iodoacetamide (Sigma-Aldrich), followed by trypsin digestion. After digestion, the dried samples were dissolved in 11 l of 0.1% formic acid, and 6 l was used for injection. The samples were analyzed on a 4000 QTrap (Applied Biosystems/MDS SCIEX, Concord, Canada).
Tryptophan Fluorescence-Tryptophan fluorescence of fulllength hlVg and the ␣-helical domain was measured in the presence and absence of PS liposomes at 25°C on a Cary Eclipse fluorescence spectrophotometer (Agilent Technologies). The measurement was repeated after a 30-min incubation time. The excitation wavelength was set to 295 nm, and the emission spectra were acquired at 310 -400 nm using 5-nm slit widths and a scan rate of 600 nm/min. Protein-free blanks were run in parallel with the samples and subtracted from the data.
Extrinsic Fluorescence Measurement-The fluorescence spectra of the leakage assay using the hlVg and fbVg samples were recorded at 25°C on a Cary Eclipse fluorescence spectrophotometer (Agilent Technologies). Excitation of 8-aminonaphtalene-1,3,6-trisulfonic acid was performed at 355 nm, and its subsequent emission was measured at 510 nm at a scan rate of 600 nm/min. 10-and 20-nm slit widths in the excitation and emission pathways, respectively, were used. A 10-mm quartz cuvette was filled with 1 mM PS liposomes (made of a 1:1 mixture of PC and PS), and 2.5 M sample was added. The fluorescence was measured every 0.25 min over a time range of 20 min. To estimate the maximum attainable fluorescence, Triton X-100 was added (a final concentration of 2 mM) to ensure disruption of the liposome bilayer at the end of the time scan. The leakage was evaluated as a percentage of complete release of the encapsulated contents with Triton X-100. The initial intensity (at 0 min) of the liposome and sample solution was set to 0% leakage, and the final intensity after adding Triton X-100 corresponded to 100% leakage.
Oxidative Stress Assay-All centrifugation steps were 800 ϫ g, 5 min, and 24°C. 1 ϫ 10 7 High Five cells were suspended in PBS (control), 5.3 M BSA, or 5.3 M fbVg in PBS (volume 50 l) and incubated for 1 h in 27°C under gentle agitation. The cells were washed twice with 0.5 ml of PBS. The pellet was suspended in 100 l of culture medium (Invitrogen). The cells were transferred to 0.4-ml chamber slides (Lab-Tek) that contained 0.3 ml of medium and a final concentration of 1.5 mM H 2 O 2 . After a 1.5-h incubation, 10 l of PI in 2ϫ saline-sodium citrate buffer was added for a final concentration of 5 M. The differences between PI-stained cells of the total cell count in four replicates were analyzed by the 2 test. The cells were imaged with a Zeiss Axiovert200M fluorescence microscope using bright field for total cell count and fluorescence filter for dead cell count (excitation 510 nm, emission 630 nm).

RESULTS
Detecting Vg in Honey Bee Tissues and Cell-binding Experiments-We performed immunohistochemistry and Western blot experiments to assess Vg localization in honey bee tissues and to initially test Vg-cell binding. The high resolution confocal micrographs in Fig. 1 represent sections of n ϭ 5 individuals for abdominal (Fig. 1, A-C and G-L) and head tissue samples (Fig. 1, D-F). As described previously (29), Vg immunoreactivity (VgIR) is found in the vesicles of fat body trophocytes (Fig. 1,  A-C). In a subset of these vesicles, we found intense VgIR at the vesicular membranes (Fig. 1B). As expected for an abundant hemolymph protein, VgIR was scattered across different tissue sites (Fig. 1, C, E, and F) apart from the known localization within the cytosol of the cells of the fat body (Fig. 1, A-C) and the hypopharyngeal glands (Fig. 1D). In particular, intense VgIR was detected at boundary layers (e.g. between single cells in hypopharyngeal glands (Fig. 1D) or between syncytial cells of muscle fibers (Fig. 1, E and F)). Our controls for autofluorescence, detector noise, and nonspecific labeling by the secondary antibody were all negative (Fig. 1, H, J, and L; n ϭ 5), as compared with test samples recorded with similar microscopy settings (Fig. 1, G, I, and K; n ϭ 5). An immunoblot (Fig. 1M) shows Vg in both the cytosolic and membrane fraction of abdominal honey bee tissue samples. A 150-kDa Vg fragment that has lost the N-sheet domain (31) is found in both fractions. Finally, Vg binding to standard insect Sf9 cells was positive, including the 150-kDa fragment (Fig. 1N). These experiments suggest that honey bee Vg can interact with cells and membranous struc-tures and that this interaction is independent from the putative receptor-binding N-sheet domain that is missing from the 150-kDa Vg fragment.
Vg Binding Is Enhanced to Dead Cells-Next, we explored the idea of an apoB-like anti-inflammatory role of Vg by measuring Vg binding to healthy, apoptotic, and dead cells using cultured insect High Five cells and FACS (n ϭ 5). The Vg-incubated cells were separated into dead and live ( Fig. 2A), and live cells were further separated into apoptotic and healthy cells (Fig. 2B). The fluorescence intensity statistics are shown in Fig. 2C. The median fluorescence intensities of the healthy, apoptotic, and dead cell populations were found to differ (Kruskal-Wallis analysis of variance: H ϭ 9.5, df ϭ 2, p ϭ 0.0087). Vg binding to dead cells was ϳ25 times higher than to live cells measured by median fluorescence intensity; this was assessed to be a significant difference by Mann-Whitney U post hoc pairwise tests (the values for both healthy and apoptotic cells versus dead cells were Z ϭ 2.6112, p ϭ 0.0079). The binding difference between healthy and apoptotic cells was not significant (Z ϭ 0.5222, p ϭ 0.6905). The antibody labeling was estimated to be successful because there was low antibody background signal (Fig. 2D) and a significant increase in Alexa-488 fluorescence in the Vg-incubated samples (n ϭ 5) compared with the antibody control (n ϭ  3) (Mann-Whitney U test; Z ϭ 2.2361, p ϭ 0.03571). In brief, all cell populations bound Vg, but the binding was significantly higher in the case of dead cells.
To obtain detailed information about the Vg localization on a damaged cell surface, we performed fluorescence imaging using High Five and Sf9 cells that were incubated with Vg. The antibody background was low (Fig. 3A (No Vg incubation) and the bottom panel in Fig. 4 (no primary antibody (Secondary antibody control)). We observed strong Vg localization on the late apoptotic cells (Fig. 3B) that were identified by their morphology (small, irregular) and by their strong binding of our positive control protein, annexin V. Furthermore, we found Vg bound to cell blebs that are signs of a damaged cell (Fig. 3B). Also, Vg bound to necrotic cells (Fig. 3C) that were identified by morphology (large, rough surface) and PI staining. In the absence of Vg, the apoptosis marker annexin V, which binds to PS, bound to cell blebs (Fig. 3D), but this binding appears to be inhibited in the presence of Vg (Fig. 3B). Thus, some of the annexin V-negative cells, determined to be healthy in the FACS experiment, might be damaged and blebby. In Fig. 4, another serum protein, BSA, was used as a negative control for dead cell binding. BSA was binding very weakly compared with the markedly positive Vg signal on the dead cells' surface.

SPR Measurement of Lipid Bilayer Binding-Because
Vg shows differentiated cell surface binding, a question opens as to what mediates the binding. To test whether the phospholipid bilayer in itself can contribute to the affinity, we tested honey bee Vg binding to PC and PS/PC 1:1 mixture liposomes (the mixture is hence called PS) using an SPR technique. PC is a major lipid on insect cell membranes (44). Because Vg appears to selectively bind more strongly to dead cells and the PS-rich cell blebs (see above) compared with healthy cells, PS-liposomes were used as a simple model of damaged or dead cell membrane.
We used two types of native samples purified from a large set of honey bees: full-length Vg purified from hemolymph (hlVg) and predominantly fragmented Vg (the previously described 150-kDa fragment) purified from fat body (fbVg). In addition, two domains were produced in Escherichia coli: the ␣-helical and vWFD domain, which both localize to the 150-kDa fragment of Vg. The vWFD domain produced in E. coli did not bind to the liposomes (Fig. 5), but the sensograms of all the other samples indicate that the Vg molecules become permanently associated with the bilayers (Fig. 5A). In every case, the binding response to PS exceeded that of PC. The full-length hlVg had the highest affinity to both bilayers. The predominantly 150-kDa fbVg fragment had slightly lower affinity than the fulllength hlVg. The ␣-helical domain alone, moreover, showed a similar response to fbVg (Fig. 5, A and B), particularly when bound to PS. S 0.5 is the protein concentration that yields half of the maximal response. This binding affinity value was calculated for hlVg based on the affinity curve of the five measure- A, the top row shows a ϫ20 magnification of High Five insect cells that were incubated with Vg. Intense Vg signal mostly co-localizes with intense annexin V signal. The bottom row is an antibody background control that was not incubated with Vg. Vg and annexin V images are shown separately, and the last image is their overlay with a differential interference contrast (DIC) image. B, ϫ60 magnification of Vg-and annexin V-bound cells. The thick arrow points to the shriveled late apoptotic cell with strong Vg and annexin V co-localization colored cyan. Vg also binds heavily to cell membrane irregularities called blebs (small arrows). C, vitellogenin binds to necrotic cells (the swollen cell in the middle). PI (red) was used in this sample to mark the nuclei of dead cells. D, in the absence of Vg, annexin V binds to cell blebs.

FIGURE 4. Vg binding to dead Sf9 cells compared with a BSA control. Killed
Sf9 insect cells were incubated with Vg or BSA, followed by a treatment using rabbit anti-Vg or rabbit anti-BSA polyclonal primary antibody. The secondary antibody control was not incubated with a primary antibody, and, as expected, marked fluorescence was absent in these samples. All samples were treated with Alexa-488 (green) secondary antibody and DAPI. In contrast to Vgincubated cells, the controls incubated with BSA did not exhibit marked positive staining, suggesting that binding of BSA was negligible. Scale bar, 20 m. ments (Fig. 5B). hlVg S 0.5 with PC was 367 Ϯ 80, and with PS it was 219 Ϯ 26. The given S.E. is based on the curve fitting.
Limited Proteolysis of Vg in the Presence of Liposomes-In addition to SPR, another approach was used to verify Vg-lipo-some interaction: limited proteolysis in combination with liposomes. This method also provides a rough insight into the protein region that interacts with the bilayers, because this region can become protected against proteolysis. fbVg, incubated with PC or PS, was enzymatically digested and run on gel (Fig. 6A). In the presence of liposomes, a ϳ120-kDa putatively membraneembedded part of Vg was less degradable. LC-MS/MS hits of this segment span amino acid residues 135-1613 (Fig. 6C). This region contains the ␣-helical domain (residues 416 -765). In order to quantify the gel experiment and to see if there is any difference between Vg bands in the PC and PS samples, the intensities of the protease-resistant gel bands were measured (n ϭ 6). PC and PS sample band intensities differed from the corresponding, nearly zero, background gel region in the liposome-free control (Kruskal-Wallis analysis of variance; H ϭ 10.75, p ϭ 0.005; Fig. 6B) but not from each other (Mann-Whitney post hoc test; Z ϭ 0.16, p ϭ 0.873; Fig. 6A). For reference, we measured a gel region in the range of ϳ25 kDa (Fig. 6A), where no intensity differences were found (Kruskal-Wallis analysis of variance; H ϭ 0.047, p ϭ 0.977; Fig. 6B), indicating FIGURE 6. A ϳ120-kDa Vg segment is resistant to proteolysis in the presence of liposomes. A, a gel showing how a part of Vg (the band marked with black arrows) is not digested by chymotrypsin in the presence of PC or PS liposomes. The gray arrows show two bands, whose intensity in all lanes was used as a reference. S, molecular weight standard. Vg, untreated, purified Vg from honey bee abdomen. Control, no liposomes. PS, Vg with PS liposomes. PC, Vg with PC liposomes. B, the intensity of the bands marked with black arrows in A is significantly higher than the intensity of the corresponding gel region in the controls but does not differ between PS and PC. The reference bands (gray columns; gray arrows in A) showed no significant difference. The error bars show the S.D. of six band intensities. C, the LC-MS/MS hits (red) of the gel-extracted honey bee Vg segment that resists chymotrypsin digestion in the presence of liposomes (black arrows in A). The ␣-helical domain is highlighted gray. The most reliable peptide hits with high Mascot ion score (Ͼ75) are underlined. that the digestion is generally similar in the control and the liposome samples with the exception of the ϳ120 kDa band. This experiment adds to the SPR data, indicating that a part of Vg is able to integrate into both PC and PS liposomes, and this part contains the ␣-helical subdomain.
Homology Modeling-Because the ␣-helical part of Vg appears to be able to interact with lipid bilayers, we modeled it in order to better understand the structural basis of this interaction. Our homology model shows a total of 49 positively charged residues, 34 of which point directly to the solvent (Fig.  7A). The side facing the binding cavity of Vg is less charged, as expected because Vg carries hydrophobic molecules in this cavity (Fig. 7B). The secondary structure prediction program PSIPRED (43) assigned the domain exclusively to helices with connective loops that support the accuracy of the helix locations in the model (data not shown). Residues 579 -603 could not be modeled because they are absent in the template (lamprey lipovitellin); PSIPRED prediction suggests that this missing sequence is a random coil (data not shown). Based on our alignment of Vgs of insects and vertebrates and the related human proteins, we assign the loop as insect-specific (Fig. 7C). Some of the amino acid residues in the loop are well conserved in insects: His 587 , Tyr 590 , Pro 591 , Val 592 , His 593 , Gly 596 , Arg 597 , Leu 598 , and Lys 601 (highlighted light gray in Fig. 6C).
Tryptophan Fluorescence-For detection of possible conformational changes upon liposome binding, we measured tryptophan fluorescence of full-length Vg and the ␣-helical domain. Vg (without its signal peptide) contains 17 tryptophan residues at various locations in the sequence, three of which are located in the ␣-helical domain (Fig. 8). If the chemical environment of these sites underwent major structural changes, a shift in the protein's tryptophan fluorescence spectrum would be expected. However, no shift was observed for either of the samples (Fig. 8). The fluorescence intensity was quenched in the liposome-bound samples compared with the liposome-free samples, as expected.
The Effect of Vg Binding Measured by Fluorescence Leakage and Oxidative Stress Tests-After the demonstrations of Vgcell and Vg-bilayer binding, we tested two possible biological functions of the binding to bilayers/cells. First, Vg binding to PS-enriched dying cells might cause disruptive membrane rearrangements that facilitate cytolysis (45). Therefore, we tested the effect of hemolymph and fat body purified native Vg on the integrity of PS liposomes. We used an extrinsic fluorescence leakage assay, which monitors the release of a fluorophore and its quencher encapsulated in vesicles. The fluorescence signal was not notably changed until the final 2.5 M protein concentration (the same as used in SPR); thus, Vg-mediated disruptive membrane activity found no support in the experiment (Fig.  9A). Second, Vg binding to healthy cells could be connected to the antioxidant properties of Vg. In our ROS test, Vg-incubated insect cells (High Five) were found to be more tolerant against H 2 O 2 -induced oxidative stress than any of our two controls (Fig. 9, B and C). The controls were cells incubated with buffer and cells incubated with BSA. An overall 2 test between the Vg-treated samples and the double controls was significant ( 2 ϭ 54.0932, df ϭ 2, p ϭ 1.794eϪ12). Vg enhanced the survival of the cells significantly compared with the buffer control ( 2 ϭ 33.6257, df ϭ 1, p ϭ 6.68eϪ09) and the BSA control ( 2 ϭ 46.3796, df ϭ 1, p ϭ 9.743eϪ12). The two controls did not differ ( 2 ϭ 0.7335, df ϭ 1, p value ϭ 0.3918).

DISCUSSION
Many serum proteins, including LLTPs, have affinity to exposed hydrophobic molecules and surfaces that often signal danger. This work shows for the first time that honey bee Vg has similar properties. Immunohistochemistry of honey bee tissues and Western blotting show that Vg is associated with cells and membranous structures in healthy tissues. Further FACS and imaging experiments detailed that Vg affinity with dead cells and membrane blebs is enhanced. We suggest that direct Vg binding to cell membrane phospholipids is involved because SPR and limited proteolysis in combination with liposomes support Vg-bilayer binding. The binding can be linked to the antioxidative and putative anti-inflammatory defense mechanisms of Vg in the honey bee.
Vgs have reproductive (egg yolk precursor) and non-reproductive (behavior, immunity) roles. The N-sheet domain is the predicted receptor binding site of Vgs and presumably essential for the uptake of the protein by oocytes (19). In honey bees, the N-sheet is under strong negative selection (i.e. mutational change is not permitted) (46), yet our observations (Figs. 1 (M and N), 5, and 6) suggest that Vg is able to adhere to lipid bilayers without receptor interaction. In the current literature, only one study documents a Vg-membrane association, and this interaction was shown for reproductive cells. In the amphibian  The fluorescence signal at 0 min was set to 0%, and 100% leakage corresponds to the signal of liposomes fully disrupted by Triton X-100. Vg samples do not cause notable leakage, which means that the integrity of PS liposomes is not affected by Vg. B, insect cells were coated with purified Vg, carefully washed, and subjected to 1.5 mM H 2 O 2 for 1.5 h. Total cells were counted using differential interference contrast (DIC) microscopy, and PI staining was used as a cell death indicator. C, counts of healthy (white) and PI-stained (red) cells of n ϭ 4 experiments, where cells were incubated with either buffer, BSA, or Vg. Vg-coated cells showed significant reduction in PI staining compared with the controls. Discoglossus pictus, lipovitellin (i.e. cleaved Vg in eggs) appears at the surface of each egg at a particular sperm binding site, where lipovitellin is involved in sperm-egg interaction (47). Our data of Vg on somatic cells, including muscle cells, of functionally sterile honey bee workers suggests that Vg-cell interaction can be non-reproductive. The 150-kDa Vg fragment (Vg without the N-sheet) has been (and still is) under considerable positive selection, not only in honey bees but in several Apis species (46). The reasons for the positive selection of the Vg sequence subsequent to the N-sheet have been purely speculative (46,48), but we suggest that membrane affinity could be involved in this evolution.
Honey bee Vg binding to dead and damaged (blebbing) cells suggests that Vg can be an anti-inflammatory actor similar to many other plasma proteins (6). The mechanism of dead cell recognition can be explained, to some extent, by the enhanced affinity of Vg for the PS bilayer in the SPR measurements. The affinity difference was not revealed by SDS-PAGE (Fig. 6), which might be due to the known shortcomings of SDS-PAGE, in particular the poorer resolution of signal differences in quantitative analysis. PS is exposed on the membrane of apoptotic and dead cells. PS is found concentrated on blebs (49), where our imaging data show high Vg occupation and inhibition of the binding of our positive control, annexin V. Similar annexin V inhibition is recorded for other bleb-binding serum proteins (50). However, Vg does not appear to make a difference between healthy and early apoptotic cells, the latter being clearly stained by annexin V. Annexin V-PS affinity is affected by PS percentage, phospholipid clustering and immobilization, calcium strength, membrane curvature and rigidity, and protein clustering on the membrane (51). Vg might be sensitive to such factors, and perhaps the protein does not favor the conditions on the membrane of early apoptotic cells. Because LLTPs are broad spectrum pattern recognition receptors, it is possible that Vg binds to additional molecules that are exposed on dead cells or blebs, such as DNA, proteins, or glycans. However, further studies are needed to verify the binding mechanism and to reveal the downstream effects of Vg-damaged cell recognition. Our leakage assay results indicate that Vg binding on PS-rich bilayer does not damage the membrane. Potentially, Vg binding could promote the clearance of damaged cells, as is seen in several plasma proteins (9,(52)(53)(54).
The Vg domain (the N-sheet in combination with the ␣helical subdomain) association with lipid surfaces has been described previously in human apoB (15,17). We narrow down this property to the ␣-helical part in the honey bee. The positively charged solvent-facing side of the ␣-helix is probably attracted by the negative charge of especially the PS of blebbing or otherwise compromised cellular membranes. For example, arginine residues are important for PS binding of annexin V (55). Our fluorescence data do not suggest major structural rearrangements upon liposome binding for the full-length Vg or the ␣-helical domain. A likely explanation is that the binding is mediated by the loops between the helices or by some of the 15 helices with no tryptophan. Both Vg and the domain show typical changes in tryptophan fluorescence emission intensity when measured bound to liposomes.
Because the phospholipid affinity of Vg domain is now reported for as phylogenetically distant proteins as human apoB and honey bee Vg, it is tempting to interpolate this property to other LLTPs. However, the relatively well conserved Vg domain has great variation among species at two sites. One is the linker between the N-sheet and the ␣-helix. This linker is elongated and heavily phosphorylated in most insects compared with vertebrate Vgs (32), and its length and the number of phosphorylation sites varies tremendously within insects (3). The other variable site was identified in this study; there is a loop in the middle of the ␣-helical that is well conserved in insects but absent in higher animals. These structural differences within the Vg domain can have an effect on affinity or binding selectivity of LLTPs.
Vg binding to healthy cells is visible in our histology and FACS data and is supported by Vg binding to PC liposomes. In the images of cultured cells (Fig. 3), however, Vg binding to healthy cells was hardly visible and was largely overshadowed by the bright signal of the blebs and dead cells. Even if Vg binds only weakly to cultured cells, we show that Vg shields these cells against H 2 O 2 damage. Previous studies have shown that honey bees with a higher Vg titer in hemolymph are more resistant to oxidative stress, and Vg is the major target of oxidative carbonylation (23). Our results strengthen this view. The membrane might be the optimal location for Vg to provide direct shielding for the cell against a ROS attack because lipids are chemically prone to oxidation (56).
Chronic inflammatory responses are connected to severe health conditions prevalent in the Western world (cardiovascular diseases, diabetes, Alzheimer disease, obesity, and cancer (57,58)), and in all these, apoB is involved. This makes LLTPs exceedingly interesting proteins for medical research. The long lived wintertime honey bee workers accumulate Vg in hemolymph and fat tissue, and Vg is thought to play a central role in the extended life span of the winter bees (23, 28, 29, 59 -62). We show how Vg not only helps cells to battle ROS but might also inform the body about tissue damage. Vgs, although extensively studied, in particular in fish (20), have not traditionally been viewed from the inflammatory perspective. Our study highlights that the honey bee offers a simple model system for understanding the complicated physiological effects of LLTPs.