In Vivo Clearance of Alpha-1 Acid Glycoprotein Is Influenced by the Extent of Its N-Linked Glycosylation and by Its Interaction with the Vessel Wall

Alpha-1 acid glycoprotein (AGP) is a highly glycosylated plasma protein that exerts vasoprotective effects. We hypothesized that AGP's N-linked glycans govern its rate of clearance from the circulation, and followed the disappearance of different forms of radiolabeled human AGP from the plasma of rabbits and mice. Enzymatic deglycosylation of human plasma-derived AGP (pdAGP) by Peptide: N-Glycosidase F yielded a mixture of differentially deglycosylated forms (PNGase-AGP), while the introduction of five Asn to Gln mutations in recombinant Pichia pastoris-derived AGP (rAGP-N(5)Q) eliminated N-linked glycosylation. PNGase-AGP was cleared from the rabbit circulation 9-fold, and rAGP-N(5)Q, 46-fold more rapidly than pdAGP, primarily via a renal route. Pichia pastoris-derived wild-type rAGP differed from pdAGP in expressing mannose-terminated glycans, and, like neuraminidase-treated pdAGP, was more rapidly removed from the rabbit circulation than rAGP-N(5)Q. Systemic hyaluronidase treatment of mice transiently decreased pdAGP clearance. AGP administration to mice reduced vascular binding of hyaluronic acid binding protein in the liver microcirculation and increased its plasma levels. Our results support a critical role of N-linked glycosylation of AGP in regulating its in vivo clearance and an influence of a hyaluronidase-sensitive component of the vessel wall on its transendothelial passage.


Introduction
Alpha-1 acid glycoprotein (AGP; also known as orosomucoid) is a 44 kDa, heavily glycosylated protein synthesized primarily by hepatocytes and secreted into the plasma, where it circulates under normal conditions in humans at 8-30 μM [1,2]. During the acute phase response, circulating levels of human AGP are elevated over baseline values by 2-to 7-fold [3]; in rats, rabbits, and mice, the elevation of either circulating AGP or hepatic AGP mRNA ranges from 10-to 200-fold [4][5][6][7][8]. The biological function of AGP is unclear. Suggested roles for this plasma protein include the modulation of the immune response [9]; the stabilization of plasminogen activator inhibitor-1 (PAI-1) functional activity [10]; and the maintenance of normal capillary permeability and selectivity [11].
Evidence from perfusion or cell culture systems supports a role for AGP in maintaining capillary permselectivity. Perfusion studies using isolated rat skeletal muscle [12] or kidneys [13] or frog mesenteric preparations [14] initially showed that the rate of negatively charged protein loss from the perfusate (e.g., albumin or lactalbumin) was reduced in the presence of AGP. These findings lead to the hypothesis that AGP binds to the capillary vessel wall and increases its negative charge [14]. AGP binding to endothelial cells in culture has also been demonstrated [15][16][17]. The protein has furthermore been suggested to form part of the glycocalyx, a dynamic endothelial surface layer of glycosaminoglycans, 2 Journal of Biomedicine and Biotechnology proteoglycans, and adsorbed plasma proteins deduced by indirect examination, for instance as a red cell exclusion zone in intravital microscopy [18]. Most recently, transfused human AGP was shown to exert renal effects by restoring normal glomerular filtration rates in rats rendered nephrotic by puromycin aminoglycoside administration [19].
It has long been known that enzymatic removal of terminal sialic acid residues, and exposure of terminal galactose moieties, converts AGP into a high-affinity ligand of the asialoglycoprotein receptor, leading to its rapid removal from the mammalian circulation [20][21][22][23]. Relatively few studies, however, have examined the in vivo clearance of the unmodified plasma glycoprotein [24][25][26], and none have examined the consequences of complete removal of glycans on the clearance and catabolism of AGP. Since AGP is a relatively small protein of approximately 43 kDa, with 45% of its mass being contributed by its carbohydrates, we hypothesized that removal of its sugars would render it filterable in the kidneys and accelerate its clearance from the circulation [27]. In the current study, we tested this hypothesis and in addition investigated the consequences to AGP clearance of systemic disruption of the glycocalyx. 2.3. Purification of AGP from Human Plasma. AGP was purified exactly as described by Hervé et al. [28] from citrated plasma obtained from healthy volunteer donors, collected with informed consent using a protocol approved by the Hamilton Health Sciences Research Ethics Board. The plasma pool was combined, using equal volumes from each donor, aliquoted and frozen at −80 • C until use; the same five donors (4 Caucasian, 1 Black) were used for all preparations employed in this study.

Enzymatic Modification of AGP.
The terminal sialic acid chains of purified AGP were removed by treatment with Clostridium perfringens neuraminidase, using 5 units of enzyme per μg of AGP in 50 mM sodium citrate, pH 6.0, overnight at 37 • C. Entire N-linked glycan chains were removed from AGP using Peptide: N-Glycosidase F (PNGase F). For electrophoretic use, AGP (0.02 mg) was denatured by boiling in 0.5% SDS, 80 mM dithiothreitol for ten minutes, then reacted with 2500 units PNGase F at 37 • for one hour, after dilution into 1% NP-40, 50 mM sodium phosphate pH 7.5. For radiolabeling and in vivo use, 0.2 mg AGP in 5.0 M urea was incubated for one hour at room temperature, then diluted 1 : 1 with 10,000 units of PNGase F in 0.1 M sodium phosphate pH 7.5, and incubated overnight at 37 • C, prior to dialysis against phosphate buffered saline.

Expression and Purification of Recombinant AGP.
Plasmids pPICZ9ssAGPH6 and pPICZ9ssAGP-N(5)Q-H6 were separately used to transform Pichia pastoris X33 yeast to Zeocin resistance, and resulting cell lines were cultured and Journal of Biomedicine and Biotechnology 3 induced with methanol; conditioned media was neutralized and secreted recombinant proteins were purified by nickel affinity chromatography, as previously described [30].

In Vivo Clearance of AGP in Rabbits.
Purified plasmaderived, enzymatically treated, or recombinant forms of AGP were iodinated using the Iodogen method [31] and injected into the marginal ear vein of rabbits, as previously described [32]. At timed intervals (0.083, 0.5, 1, 2, 4, 6, 8, 24, 48, 72, 96, and 168 hours or until the recovered radioactivity declined lower than 0.1% of the injected dose) blood samples were taken from the marginal vein of the other ear, centrifuged to obtain plasma, and the trichloroacetic acid-precipitable radioactivity was determined by γ counting. For each rabbit, semilogarithmic substrate disposition graphs of residual plasma radioactivity against time were generated, and analyzed using PK Solutions version 2.0 software (Summit Research Services, Montrose, CO, USA). Briefly, the software was employed to fit the data using noncompartmental (area) methods based on the trapezoid rule. For this purpose, radioactive doses were converted into pmoles using the measured specific activity of radiolabeling, and molecular weights of 43, 33, and 23 kDa for pdAGP, PNGase-AGP, and rAGP N(5)Q, respectively. The specific activity of radiolabeling was similar for all three proteins, ranging from 3.94 to 7.05 X 10 6 cpm/μg. In some experiments, the tissue distribution of radioactivity 30 minutes after intravenous injection of radioiodinated AGP or derivatives was measured as described [33]; briefly, organs were excised, rinsed in ice cold saline, weighed, and either a portion or the entire organ, depending on its size, was γ counted. Results were expressed as a fraction of the total radioactive dose injected. In other experiments, either 4.0 g of α-methyl-D-mannopyranoside (α-methyl-mannoside) in 20 mL sterile saline, or 0.54 g of D-galactose in 10 mL sterile saline, was injected intraperitoneally immediately prior to intravenous injection of radioiodinated protein in an effort to block glycan-dependent uptake via hepatic receptors.

In Vivo AGP Clearance in Mice.
Clearance of radiolabeled human AGP in C57Bl/6 mice was followed as previously described [30,34] by measuring acid-precipitable plasma radioactivity in plasma samples. Plasma radioactivity was reported as percentage of the injected dose, using as a dilution factor the plasma volume divided by the sample volume. Plasma volume was calculated as the product of the weight of the mouse in grams X 0.078 mL blood volume per gram body weight X (1-the hematocrit), using a hematocrit of 0.44 (US NIH reference values from http://oacu.od.nih.gov/ARAC/). In some experiments the mice were treated with Streptomyces hyaluronidase (Sigma-Aldrich, St. Louis, MO, USA) by intravenous injection of 47 U (to deliver 30 U/mL estimated plasma volume) of the enzyme in 0.1 mL sterile saline coinjected with iodinated human AGP.

Intravital Microscopy.
Anesthetized mice (20-30 g body weight) were cannulated via the right jugular vein, the abdomen opened, and a lobe of the liver placed on the heated Plexiglas stage of a Leica DMI 6000 B confocal microscope. The liver microvasculature was photographed using a Hamamatsu C9100-12 back-thinned EMCCD camera attached to a Leica DMI 6000 B confocal microscope using the 63X objective. Either 15 μg biotinylated bovine hyaluronic acid binding protein (bHABP) complexed to streptavidinconjugated Alexa Fluor 647 (Associates of Cape Cod, Falmouth, MA, USA) or 40 μg Griffonia simplicifolia 1B 4 (GS-1B 4, ) lectin directly conjugated to Alexa Fluor 488 (Invitrogen, Carlsbad, CA, USA) was injected through the cannula into the vasculature. Baseline images were obtained over the next 15 minutes, then hyaluronidase was injected as described above, and images were acquired for an additional hour. In some experiments purified human AGP (165 mg/kg body weight) was injected after bHABP:streptavidin Alexa Fluor 647 administration and baseline image capture and 15 minutes prior to hyaluronidase treatment. Fluorescent intensity was quantified using ImageJ software (http://rsb.info.nih .gov/ij/).

Ligand
Blotting. The ability of various forms of AGP to react with digoxigenin-labeled lectins was assessed following SDS-PAGE and electrical transfer of proteins to nitrocellulose; blots were analyzed using a DIG Glycan Differentiation Kit as suggested by the manufacturer (Roche Diagnostics, Mississauga, ON, Canada).

Statistical
Analysis. Data are reported as the mean ± the standard deviation in all cases. Statistical analysis was performed using GraphPad Instat version 3.06 software (Graph-Pad Software Inc, San Diego, CA), using one-way Analysis of Variance (ANOVA) for multiple comparisons and Student's t-tests for paired comparisons. Nonparametric methods were employed if data did not pass tests of normality and equivalence of standard deviation; further details are provided in Figures 1-6

Results
The in vivo clearance of AGP treated with neuraminidase to remove the terminal sialic acid residues has been widely investigated, to a greater extent than that of the unmodified protein [21,24,35,36]. We first ensured that our purified AGP behaved as expected following this enzymatic modification. As shown in Figure 1(a), neuraminidase treatment increased the electrophoretic mobility of AGP and eliminated its recognition by Sambucus nigra agglutinin (SNA), which binds to sialic acid linked to galactose. Five minutes after injection, 45.5 ± 4.0% of the injected AGP was detected in plasma, compared to 0.77 ± 0.36% of neuraminidasetreated AGP, a difference of 82-fold. By fifteen minutes, the proportion of AGP remaining in the circulation was unchanged for native AGP (42.0 ± 5.5%) but had further declined to 0.25 ± 0.5% in the case of the neuraminidase-treated form (Figure 1(b)).
Enzymatic and recombinant DNA methods were used to reduce or eliminate the number of N-linked glycan chains Table 1: Pharmacokinetic parameters relating to AGP clearance in rabbits.
attached to AGP. Denaturation of AGP was required to obtain any significant reduction in its extent of glycosylation using PNGase F, which removes glycan chains in their entirety at their point of attachment to asparagine residues on the polypeptide backbone of AGP. Enzymatic deglycosylation following denaturation of AGP with urea or SDS was compared. As shown in Figure 2, transient denaturation with urea allowed the partial removal of 0-5 glycans per AGP molecule; full denaturation with SDS removed all glycans.
That the most rapidly migrating PNGase F reaction product of AGP was fully deglycosylated was indicated by its comigration both with SDS and PNGase F-treated AGP, and with a form of recombinant AGP mutated to abrogate Nlinked glycosylation (rAGP N(5)Q) (Figure 1(a)), and by the loss of recognition of all of these putatively deglycosylated species with SNA lectin (data not shown). In keeping with its intermediate state of deglycosylation, PNGase F-treated AGP exhibited more rapid clearance in rabbits than unmodified AGP, but was retained in the circulation to a greater extent than rAGP N(5)Q; these differences were apparent as early as 30 minutes after injection (Figure 2(b)). Electrophoresis and autoradiography of postinjection plasma samples revealed that the most rapidly migrating deglycosylated forms of AGP (e.g., "1" versus "2" in Figure 3(b)) were lost from plasma more rapidly than less rapidly migrating forms (e.g., "3" or "4" in Figure 3(b)). In terms of organ distribution, both PNGase F-treated AGP and rAGP N(5)Q were found in the kidney to a greater extent than native AGP, the effect being more pronounced for the fully deglycosylated form (Figure 3(d)).
The differences in clearance of the three forms of AGP were apparent not only on substrate disposition versus time graphs, such as those shown in Figure 2, in which the proteins were compared on the basis of residual radioactivity, but also following pharmacokinetic analysis, in which the proteins were compared based upon their dose in pmoles and their residual plasma concentration in pmol/mL (Table 1). PNGase-AGP and rAGP N(5)Q exhibited decreased areas under the clearance curve, decreased mean retention times, and decreased half-lives, but the decreases were only statistically significant for rAGP N(5)Q. The volumes of distribution and clearance rates were significantly increased for both altered forms (e.g., mean CL was increased 4.9-fold for PNGase-AGP and 45-fold for rAGP N(5)Q, P < 0.001 versus pdAGP in both cases).
We produced recombinant nonglycosylated AGP N(5)Q in a Pichia pastoris yeast expression system [29]. The disappearance of injected radiolabeled rAGP-WT from the rabbit circulation, which like rAGP-N(5)Q, was C-terminally hexahistidinylated to facilitate purification, but which unlike the mutated AGP, retained all five sites of N-linked glycosylation, was also examined. As shown in Figure 4(a), rAGP-WT was much more rapidly removed from the circulation than its mutant rAGP-N(5)Q counterpart. Thirty minutes after injection of equal doses of either radiolabeled rAGP-WT or rAGP N(5)Q, the residual plasma concentration of the latter exceeded the former by more than 10-fold (Figure 4(a); 2.2 ± 0.5% versus 26 ± 3%, P < 0.0001 by unpaired t-test, Welch corrected). Coadministration of α-methylmannoside, but not galactose, significantly reduced the loss of rAGP-WT from the circulation (to 6 ± 1%, P < 0.05 by Kruskal-Wallis test); neither sugar significantly altered the concentration of rAGP N(5)Q in the circulation, suggesting that the yeast glycans contained terminal mannose residues (Figure 4(a)). Electrophoresis showed that rAGP-WT migrated more rapidly than pdAGP, consistent with its modification with shorter N-linked glycans than in plasma-derived AGP (Figure 4(b)), and ligand blotting showed that rAGP-WT, but not plasmaderived human AGP, bound Galanthus nivalis agglutinin (GNA), which is a lectin specific for glycans terminating in mannose residues (Figure 4(b)). In contrast, rAGP-N(5)Q was unreactive with all lectins, consistent with its intended lack of glycosylation (data not shown).
Intravenous administration of hyaluronidase prior to injection of radiolabeled AGP changed both its recovery and distribution in mice for several hours after treatment. Initially, the recovery of AGP rose from 80.2 ± 5.8 in the absence of hyaluronidase to 96.8 ± 6.0 (P = 0.0007 by twotailed unpaired t-test); as seen in Figure 5, the elevation of AGP in the plasma in hyaluronidase-treated mice remained detectable until 3 hours after treatment. The area under the observed curve (t = 0 to t = 6 hours) was increased from 250 ± 30%-hr to 300 ± 30%-hr (P = 0.037 by two-tailed unpaired t-test) by hyaluronidase treatment. We next sought independent confirmation that hyaluronidase was having its intended effect on the murine vasculature, using intravital confocal microscopy of livers of mice treated with or without hyaluronidase. Treatment with the enzyme reduced sinusoidal vessel staining with fluorescently labeled HABP but left vascular binding by fluorescent GS-1B 4 unchanged ( Figure 6, compare (a) to (b) and (d) to (e)). The distribution of HABP was also altered by administration of 3.3 mg intravenous human AGP (10-20 times endogenous AGP levels) without hyaluronidase treatment. This intervention left 5-to 6-fold more HABP in the plasma column than in the absence of AGP treatment (Figures 6(c) and 6(f)). Although hyaluronidase treatment altered the clearance and distribution of AGP, and AGP injection increased HABP retention in plasma, AGP did not bind to immobilized hyaluronan in vitro in microtiter plates, unlike HABP, which demonstrated avid binding (data not shown).

Discussion
In this study we explored the effects of altered, diminished, or absent glycosylation of AGP on its in vivo clearance. Although neuraminidase-treated AGP (also called asialoAGP or asialoorosomucoid) has been widely employed for many years as an injectable ligand and competitor for the hepatic asialoglycoprotein receptor in vivo [21,24,35,36], the clearance of glycosylated AGP has been less extensively examined. Regoeczi et al. [24] reported a terminal catabolic halflife of 68.5 hours for iodinated human AGP injected into a single 3.8 kg rabbit; this value is similar to our finding of 59 ± 4 hours (mean ± SD, n = 6, range 53-65 hours, weight range 2.52-2.79 kg). Some of the difference may derive from our use of noncompartmental modeling and pharmacokinetic software, as opposed to the compartmental model and manual calculation employed in the older report; alternatively, our use of lighter rabbits may also be a contributing factor. In this regard, others employing noncompartmental analysis of AGP clearance in rats reported steady state weight-adjusted volumes of distribution similar to that observed in rabbits in this study. For pdAGP in rabbits, we observed 160 ± 20 mL/kg, while others reported 150 mL/kg [37] and 90 ± 30 mL/kg [26]. Taken together, these comparisons suggest Depicts the residual acid-precipitable radioactivity remaining in the plasma of rabbits after intravenous injection of the indicated radiolabeled proteins as a function of time; for each protein, plasma radioactivity is presented as the percentage of the injected radioactive dose recovered in plasma at the indicated time after injection. Each point is the mean ± SD of 6 determinations; in most instances error bars are smaller than the symbol that denotes the mean. that our clearance findings with respect to pdAGP are reasonable in light of previous reports in the literature and form a sound basis for comparison to the novel deglycosylated forms of AGP that we generated.
We used two strategies to reduce or eliminate entire Nlinked glycans from AGP: enzymatic treatment with PNGase F, or mutation of the five asparagine residues to which the glycans are ordinarily attached. The enzymatic strategy was complicated by the resistance of AGP to deglycosylation in the absence of treatment with SDS. We elected to employ urea, since it could be removed by dialysis, promoting refolding, as opposed to SDS, which binds tightly to proteins and cannot readily be removed. PNGase F treatment was only partially effective in altering AGP clearance, likely due to the presence of some unaltered, multiply glycosylated AGP in the mixture; however, the least glycosylated forms of AGP in that mixture were most rapidly removed from the circulation. A more convincing difference between native AGP and nonglycosylated AGP was achieved via the mutational strategy; nonglycosylated rAGP-N(5)Q exhibited a 15-fold decrease in area under the clearance curve (extrapolated to infinity) and a 45-fold increase in clearance rate compared to plasma-derived AGP. Although it was not possible to control the rAGP-N(5)Q experiments with wild-type recombinant AGP made in the same experimental system, due to the high mannose-type glycosylation typical of yeast, our results were consistent with previous data. These data were obtained by investigators who employed metabolic labeling of rat hepatocytes in the presence of an inhibitor of N-linked glycosylation (tunicamycin) and immunoaffinity purification to obtain nonglycosylated rat AGP [38]. Although the difficulty in obtaining radiolabeled AGP by this route prevented replication and kinetic analysis in this study, the 50% dose retention time was noted to decrease from 100 minutes to 5 minutes for nonglycosylated AGP, as compared with native AGP, consistent with our findings, as was increased renal clearance of nonglycosylated AGP [39].
Having shown that appropriately terminated N-linked glycans keep AGP from exhibiting accelerated clearance either via glomerular filtration or via hepatic receptors specific for abnormal glycosylation, we turned to the issue of AGP interaction with the vessel wall. Intravenous administration of hyaluronidase to hamsters led to a decrease in the FITC-Dextran 70 permeation zone defined using intravital microscopy [40], one consistent with degradation of the glycocalyx, a surface layer of glycosaminoglycans, proteoglycans, and adsorbed plasma proteins [18,41] thought to contain AGP [42,43]. While hyaluronidase treatment of rabbits would have been ideal in the context of this study, we found such treatment impractical in this larger laboratory animal. In mice, we found that hyaluronidase treatment decreased the initial clearance of AGP, an effect consistent with an AGP-glycocalyx action as a normal part of its distribution. Hyaluronidase treatment appeared to be specific in that HABP binding to liver sinusoidal vessels was reduced, but that of GS-1B 4 , an α-galactose-specific lectin that mainly binds to animal endothelial cells [44], was unaffected. AGP administration also increased the amount of HABP found in the plasma column within liver sinusoids. This increase could have occurred due to AGP binding to hyaluronan in the glycocalyx, or its binding to a component of the glycocalyx that affected the access of HABP to hyaluronan. We favour the second explanation because we were unable to detect AGP-hyaluronan binding in vitro; nevertheless, we cannot exclude the possibility that this interaction occurs in vivo. Although our primary reason for examining the liver microcirculation was technical in that this organ is amenable to intravital microscopy [45], it has also been shown that this microcirculatory bed is greatly enriched in hyaluronan as opposed, for instance, to those of lung or heart [46]. Our results support the suggestion that AGP is a component of the glycocalyx, as previously suggested [17,42,47,48].
From a biotechnological point of view, our results also provide novel information with respect to targeting of recombinant AGP to either renal or hepatic destinations. AGP is a member of a protein family known as the lipocalins [49]. Lipocalins bind many drugs and metabolites within a central binding pocket of the polypeptide. Interestingly,  Dose in plasma (%) Figure 5: The disappearance of injected AGP from the plasma of mice following intravenous injection is altered by systemic hyaluronidase treatment. Radioiodinated AGP was injected into mice with (AGP + hy'ase) or without (AGP) concurrent hyaluronidase treatment, and acid-precipitable radioactivity, shown as the percentage of the dose remaining in plasma, was followed over time, n = 6±SD.
in this regard lipocalin libraries have been described with engineered binding properties [50]. Our biological distribution results raise the possibility that engineered AGPs with defined ligand binding specificities could be directed to renal or hepatic destinations using either the N(5)Q mutant, or mannosylated rAGP-WT produced in yeast. In the latter instance, producing AGP (e.g., for conjugation to DNA, as has been previously reported for gene targeting [51,52]) or an AGP fusion protein in Pichia pastoris could provide hepatic targeting in a single step, without the need for neuraminidase treatment. This study had several limitations. It would have been preferable to use homologous rabbit AGP or murine AGP in rabbits and mice, respectively, but our efforts to isolate rabbit AGP by adapting a protocol designed for the purification of rat AGP [53] were not successful, and it was impractical to purify sufficient quantities of murine AGP from murine plasma for our purposes. Our study was also limited to human AGP purified from the plasma of healthy individuals, and clearance was similarly investigated in healthy animals. As such, our results do not address the possibility that AGP glycosylation or clearance properties are altered during inflammation, as has been suggested [54][55][56]. Our use of plasma purified from a small pool of healthy donors to purify AGP was aligned with the concept that this material might be a future therapeutic product of plasma fractionation [57,58].
Our results show that N-linked glycosylation of AGP is a critical contributor to its plasma residency, in that reducing the number of glycans accelerated clearance; this was most clearly demonstrated for a mutant AGP completely lacking N-linked glycans. In this regard, AGP resembles antithrombin, in which mutation of any one of four sites of N-linked glycosylation has been shown to lead to accelerated clearance in vivo in rabbits [59], but not albumin, in which either naturally occurring [60] or engineered sites of N-linked glycosylation accelerate the clearance of the plasma protein [33].  Glycosylation thus contributes to maintaining a circulating pool of AGP for exchange into the glycocalyx, where it may contribute to the maintenance of vascular permeability.