Site-specific enzymatic polysialylation of therapeutic proteins using bacterial enzymes

Abstract

The posttranslational modification of therapeutic proteins with terminal sialic acids is one means of improving their circulating half-life, thereby improving their efficiency. We have developed a two-step in vitro enzymatic modification of glycoproteins, which has previously only been achieved by chemical means [Gregoriadis G, Jain S, Papaioannou I, Laing P (2005) Int J Pharm 300:125–130). This two-step procedure uses the Campylobacter jejuni Cst-II α2,8-sialyltransferase to provide a primer on N-linked glycans, followed by polysialylation using the Neisseria meningitidis α2,8-polysialyltransferase. Here, we have demonstrated the ability of this system to modify three glycoproteins with varying N-linked glycan compositions: the human therapeutic proteins alpha-1-antitrypsin (A1AT) and factor IX, as well as bovine fetuin. The chain length of the polysialic acid addition was optimized by controlling reaction conditions. After demonstrating the ability of this system to modify a variety of proteins, the effect of polysialylation on the activity and serum half-life of A1AT was examined. The polysialylation of A1AT did not adversely affect its in vitro inhibition activity against human neutrophil elastase. The polysialylation of A1AT resulted in a significantly improved pharmacokinetic profile when the modified proteins were injected into CD-1 mice. Together, these results suggest that polysialylated A1AT may be useful for improved augmentation therapy for patients with a deficiency in this protein and that this modification may be applied to other therapeutic proteins.

Therapeutic proteins are important tools for the treatment of numerous diseases. One major drawback to the use of proteins as therapeutics is that they often exhibit short in vivo half-lives, caused by instability, susceptibility to proteolysis, neutralization by antibodies, and/or clearance from the bloodstream. Efficacy of these drugs therefore depends on frequent administration of large doses, which leads to increased costs and a higher risk of toxicity.

The posttranslational modification of therapeutic proteins is one means of improving their circulating half-life, thereby improving their efficiency (1). Numerous strategies have been employed toward this end, including covalent modification of the proteins. Many efforts have been directed toward the chemical addition of chains of polyethylene glycol (PEG) to therapeutic proteins (reviewed in ref. 2). PEG modification of proteins can increase their circulating half-life by improving protein stability and solubility, preventing proteolytic degradation, and reducing their rate of clearance from the bloodstream. PEGylation of proteins relies on chemical conjugation of PEG chains to free amino groups or engineered Cys residues on the protein, which can lead to heterogeneously modified proteins whose activity can be adversely affected (3). An alternative method for site-specific PEGylation of proteins has been termed GlycoPEGylation, whereby PEG chains conjugated to sialic acid residues are transferred onto unoccupied, natural glycosylation sites on therapeutic proteins (4). Despite the low immunogenicity of PEG, the production of anti-PEG antibodies and the accumulation of nonmetabolized PEG in tissues remain concerns (5) and warrant exploration of alternative protein modification strategies.

The chemical addition of polysialic acid (PSA) to proteins may represent another means to improve the circulating half-life of proteins, without causing adverse effects on the patient (6). In mammals, PSA is found mostly on the brain-specific neural cell adhesion molecule (NCAM) protein (7), on CD36, and on the recently described synaptic cell adhesion molecule (SynCAM) (8, 9). PSA modification of therapeutic proteins is an attractive alternative to PEGylation, given that PSA is biodegradable and nonimmunogenic. PSA has been chemically conjugated to a few clinically relevant therapeutic proteins and was shown to improve their circulating half-life without adversely affecting their function (10, 11). Recently, site-specific chemical coupling of PSA was demonstrated on an antitumor single-chain variable region fragment (12), but this requires engineering a terminal Cys, which is not always an option for other therapeutic proteins. To our knowledge, site-specific enzymatic addition of PSA has not yet been reported.

It has been shown that the mammalian polysialyltransferase ST8SiaII can transfer PSA to non-NCAM glycoproteins [e.g., fetuin (13)], but, to our knowledge, this enzyme has not been produced in quantities that would be required for the routine modification of therapeutic proteins. It has also been shown that the mammalian polysialyltransferases recognize sequences that confer specificity for the protein acceptor (14), which would decrease their usefulness for the general modification of therapeutic proteins. Previously, we characterized a bacterial PSA polymerase from Neisseria meningitidis group B (designated PST_NM), and showed that it could be used to modify synthetic acceptors including glycopeptides containing a mammalian O-linked glycan (15). We have now extended that study to look at modifying proteins with N-linked glycans. The objective of our study was twofold. First, we wanted to demonstrate the ability of our method to polysialylate a variety of natural N-glycan types on proteins (see below). Second, we wanted to examine the effect of polysialylation on the activity and serum half-life of the human therapeutic protein alpha-1-antitrypsin (A1AT).

A1AT is a 52-kDa glycoprotein and is the most prevalent serum serine protease inhibitor, in particular of neutrophil elastase (16). A1AT deficiency is a genetic disorder affecting an estimated 3.4 million individuals worldwide that results in a 40–90% decrease in A1AT in the serum (17). The resulting imbalance can cause emphysema and chronic obstructive pulmonary disorder, as well as a variety of other symptoms (reviewed in ref. 17). Although patients can be treated for their symptoms with bronchodialators or steroids, the only available treatment specific to A1AT deficiency is augmentation therapy, in which human plasma-derived A1AT is administered intravenously once a week (18, 19).

A major drawback to the therapy is the requirement for frequent intravenous injections for the patient lifetime, low efficacy, and high costs associated with augmentation therapy (20). Therefore, improvements to the serum half-life or function of A1AT would be a major benefit to patients. Besides being a clinically relevant therapeutic protein, A1AT is an ideal protein for enzymatic polysialylation because of its well-studied glycans and clinical need for improved pharmacokinetics.

Results

The Bacterial Sialyltransferase Cst-II Can Modify a Variety of N-Glycan Types.

In order to evaluate the generality of this method to modify a variety of N-linked glycans, we chose three proteins as acceptors based on their varying glycan composition, which include principally bi-, tri-, and tetraantennary glycans. A1AT contains three N-linked glycan sites, of which 74% are biantennary and possess a terminal α2,6-linked sialic acid. An additional 14% of the glycans are triantennary with a mixture of α2,3/6-linked terminal sialic acid residues (21). Factor IX is a coagulation factor used to treat hemophilia B, a congenital disease caused by a factor IX deficiency (22, 23). The protein contains two N-linked glycans, of which approximately 65% are tetraanntennary and 20% are triantennary glycans (24), with a mixture of α2,3/6-linked terminal sialic acid residues. Finally, bovine serum fetuin, although not a therapeutic protein, was chosen on the basis of its well-studied three N-glycan sites, of which approximately 80% are triantennary with a mixture of α2,3/6-linked terminal sialic acid residues (25, 26).

Our previous characterization of the bacterial polysialyltransferase PST_NM demonstrated the enzyme required a primer of at a least two sialic residues (15). Therefore, in order to use this enzyme we needed to generate such a primer on the N-linked glycans. The proteins were incubated with Cst-II, a bifunctional α2,3/α2,8 sialyltransferase from Campylobacter jejuni, which forms α2,8-sialic acid termini. After incubation with Cst-II, there was an increase in the apparent molecular mass of the proteins when visualized by SDS-PAGE (Fig. 1). The increase in apparent molecular mass was confirmed to be the result of modification of N-linked glycans by cleavage of the N-linked glycans with peptide N-glycosidase F (PNGase F) (Fig. S1).

Fig. 1.

Polysialylation of glycoproteins by Cst-II and MalE-Δ19PST_NM. Glycoproteins were incubated with Cst-II and/or MalE-Δ19PST_NM as described in Materials and Methods, followed by analysis by SDS-PAGE and Coomassie Blue staining. Reaction with the two enzymes results in a change in migration of the protein (disialylated-diSA) and the appearance of a high molecular mass smear (polysialylated-PSA) compared to the unmodified protein (UM). The band marked with an asterisk (*) is high molecular mass protein present in untreated fetuin. The band marked with a double-asterisk (**) is MalE-Δ19PST_NM; this reaction was purified after electrophoresis and before other assays were performed.

In order to determine the extent of the modification by Cst-II, N-linked glycans liberated from the proteins were analyzed by capillary electrophoresis (CE)-MS, in negative ion detection mode. Because the fragment ions at m/z 290.2 and 581.3 correspond to the presence of monosialic acid (Neu5Ac) and disialic acid (Neu5Ac-Neu5Ac), respectively, the sialylation patterns can be screened using precursor ion monitoring MS experiments. The analysis of native A1AT revealed a mixture of sialylated biantennary and triantennary glycans (Table S1) with additional fucosylated biantennary and triantennary glycans. After incubation with Cst-II, a mixture of modified glycans of sizes consistent with the addition of one to three Neu5Ac residues was observed (Table S1).

A number of N-linked glycans were identified in the native factor IX sample, whose masses are consistent with tri- and tetraantennary glycans capped by single Neu5Ac residues (Table S1). After incubation with Cst-II, mass values consistent with the addition of one to two Neu5Ac residues (triantennary) and one to four Neu5Ac residues (tetraantennary) were observed. Factor IX also contains two O-linked glycosylation sites, but we did not see any significant modification of these O-linked glycans. MS analysis of fetuin N-linked glycans revealed a mixture of triantennary and biantennary glycans capped with a single Neu5Ac residue (Table S1). Following incubation with Cst-II, triantennary glycans with an additional 1–3 Neu5Ac residues and biantennary glycans with an additional 1–2 Neu5Ac residues were observed. No change in the size of O-linked glycans was detected following reaction with Cst-II compared to unmodified fetuin.

The glycans predicted by the MS analysis of unmodified proteins are consistent with published data on the structure of the glycans (21, 25). Together, these results indicate that Cst-II is able to modify a variety of glycan structures including bi-, tri-, and tetraantennary glycans with either α2,3- or α2,6-linked terminal sialic acids. These glycans have been designated “disialyl” (diSA)-proteins in the remainder of the manuscript to indicate the addition of at least one Neu5Ac residue resulting in two terminal Neu5Ac residues per glycan arm.

Cst-II–Primed Glycoproteins Are Polysialylated by MalE-Δ19PST_NM.

Following the priming reaction with Cst-II (which is necessary for the activity of PST_NM), A1AT (or fetuin and factor IX) was incubated with MalE-Δ19PST_NM as described in Materials and Methods. This reaction resulted in the conversion of diSA-A1AT to polysialylated A1AT (PSA-A1AT), as shown by the disappearance of the diSA-A1AT acceptor and the appearance of a smear of high molecular mass (HMM) material in a Coomassie-stained SDS-PAGE gel (Fig. 1). The presence of PSA in the HMM material was verified by digestion with bacteriophage K1F endosialidase (Fig. S1). Numerous variables were manipulated in attempts to optimize the overall conversion, including varying concentrations of MalE-Δ19PST_NM, acceptor protein, CMP-Neu5Ac, and incubation times. For the purpose of modifying therapeutic proteins, it may be preferable to generate a product with limited PSA chain length. A comparison of polysialylation reactions at different diSA-A1AT concentrations showed that using high acceptor concentrations resulted in shorter chain lengths of PSA (Fig. S2), which is consistent with previous data from assays with colominic acid as the acceptor (27).

We then determined the PSA chain length by liberation and derivatization of the PSA on proteins using 1,2-diamino-4,5-methylenedioxybenzene (DMB) followed by chromatographic separation as described by Inoue and Inoue (28). PSA chains released from the proteins showed a diverse size distribution (from 4 to > 70 Neu5Ac residues), although favored chain lengths varied between the proteins. Approximately 53% of the PSA-A1AT chains analyzed were less than 20 Neu5Ac residues long (Fig. 2), compared to 75 and 70% for PSA-fetuin and PSA-factor IX, respectively (Fig. S3). Although labeling conditions were used to minimize PSA hydrolysis, some hydrolysis is inevitable from this technique during sample preparation and analysis (28), possibly leading to an underrepresentation of longer chain lengths in all samples. The results should therefore be considered to be relative to one another, rather than absolute values. Regardless, the results of the PSA sizing are consistent with the SDS-PAGE analysis. For example, in the PSA-A1AT sample in Fig. 1, a significant amount of material has increased approximately 20 kDa (approximately equivalent to the mass of 66 Neu5Ac residues). With three biantennary sites available for sialylation, this would mean that the average chain length would be 11, which is consistent with the PSA sizing data.

Fig. 2.

Determination of PSA chain length by DMB derivatization and ion exchange analysis. PSA was released and derivitized with DMB as described in Materials and Methods. The chromatogram shows the analysis of PSA chains from A1AT modified at high protein concentrations. The arrow shows the location of a chain of six Neu5Ac residues, as measured by chromatography of a Neu5Ac hexamer (Fig. S3).

Polysialylation Does Not Adversely Affect A1AT Activity in Vitro.

In order to determine if polysialylation had an effect on the function of A1AT as a protease inhibitor, it was tested in a reaction with its target, human neutrophil elastase and a synthetic peptide cleavable by elastase. The product formed by the action of elastase was detected at A₄₁₀ over the course of 1 h, and the activity of the enzyme was calculated. A 60-min incubation with unmodified A1AT resulted in 45.6 ± 23.7% inhibition of elastase compared to the activity of the enzyme without the addition of an inhibitor (Fig. 3). Similarly, the addition of diSA-A1AT or PSA-A1AT inhibited the activity of elastase 83.2 ± 12.1% and 58.6 ± 13.5%, respectively. These results suggest that the modification of A1AT does not adversely affect its function as an elatase inhibitor and statistical analysis suggests that the modifications may actually improve activity, especially in the case of the diSA-A1AT (ANOVA, p < 0.05).

Fig. 3.

In vitro elastase inhibition by A1AT, diSA-A1AT, and PSA-A1AT. A1AT (and modified derivatives) were incubated with a natural target of A1AT, human neutrophil elastase in the presence of a synthetic peptide substrate, N-methoxy-succinyl-Ala-Ala-Pro-Val-p-nitroanilide. Cleavage of the peptide by elastase results in a color change, detectable at A_410 nm, which was monitored over 1 h. Data are mean ± SD from three assays performed in triplicate. ▪ is the reaction with no A1AT; ▴ is the reaction with A1AT diluted to give approximately 50% inhibition over the course of the assay; ▾ is the reaction with diSA-A1AT; ♦ is the reaction with PSA-A1AT.

Polysialylation of A1AT Leads to an Improved Pharmacokinetic Profile and Affects Biodistribution in Mice.

The pharmacokinetic profiles of A1AT and the modified versions of the protein fit well with the biphasic two-compartmental IV-bolus model (Fig. 4). diSA-A1AT and PSA-A1AT exhibited approximately a 6.5- and 18-fold, respectively, greater bioavailability (area under the curve) compared to unmodified A1AT (Fig. 4). This was due to an increased beta (terminal) half-life of 13 h for diSA-A1AT and 27 h for PSA-A1AT, compared to 5 h for A1AT. Clearance rates from the blood were significantly reduced for diSA-A1AT (2.11 ± 0.51 mL h^-1) and PSA-A1AT (0.75 ± 0.085 mL h^-1), compared to A1AT (14.33 ± 4.77 mL h^-1). Together, these results indicate that sialylation of A1AT improved the pharmacokinetic profile of the protein.

Fig. 4.

Pharmacokinetic profile of modified glycoproteins. Pharmacokinetic profiles of Cy5.5-labeled A1AT, diSA-A1AT, and PSA-A1AT in mice. Mice were intravenously injected with 200 μg of labeled protein, and blood was collected at various time points. Data are mean ± SEM for three mice per group. Levels of the Cy5.5-labeled proteins in blood samples were measured by a fluorescent plate reader (excitation/emission: 670/690 nm) and their concentration interpolated from a standard curve of known concentrations of the labeled protein diluted in blood. Unmodified A1AT, ▽; diSA-A1AT, □; PSA-A1AT, ○.

In vivo optical imaging was used to evaluate the tissue biodistribution of A1AT and the modified derivatives. Full-body dorsal scan of animals injected with Cy5.5-labeled proteins indicated a prolongation of Cy5.5 fluorescence up to 48 h for diSA-A1AT and up to 72 h for PSA-A1AT, compared with only 24 h for A1AT (Fig. 5). These results are in support of the circulation half-life results obtained during pharmacokinetic experiments. A half-life of 5 h for A1AT is expected to clear completely from the circulation within 24 h (approximately 5 half-lives); similarly, diSA-A1AT and PSA-A1AT are expected to clear completely from the blood compartment within 48 h and approximately 96 h, respectively. From the dorsal scan, it was apparent that all A1AT constructs localized to the kidneys, liver/spleen area, and lymph nodes to varying degrees. Ventral scans, however more clearly illustrated the accumulation in the liver and clearance of free Cy5.5 dye and Cy5.5-labeled protein via the bladder (Fig. 5). Cy5.5 is expected to be cleaved during metabolism and then cleared though the kidney, while any Cy5.5-labeled protein below the kidney cutoff of approximately 60 kDa will also be cleared through the kidney. The PSA-A1AT construct exhibited a greater accumulation compared to diSA-A1AT and A1AT in the liver as a result of its larger size and longer circulating half-life.

Fig. 5.

Optical imaging (biodistribution) of modified glycoproteins in mice. In vivo optical imaging of the biodistribution of Cy5.5-labeled A1AT, diSA-A1AT, and PSA-A1AT injected intravenously in mice. (A) Prospective in vivo dorsal and ventral images of the whole animal body at various time points (up to 72 h) after intravenous injection of Cy5.5-labeled A1AT (Left), diSA-A1AT (Center), and PSA-A1AT (Right). (B) Ex vivo optical images of various organs (brain, heart, lung, kidney, spleen, liver, and muscle) at 72 h after intravenous injection of Cy5.5-labeled A1AT (Left), diSA-A1AT (Center), and PSA-A1AT (Right), followed by saline perfusion and animal sacrifice. Each group was assessed in three separate experiments.

At the end of the imaging protocol, 72 h after injection of the Cy5.5-labeled proteins, ex-vivo fluorescence concentration was quantified (Fig. 5). The liver and kidney organs contained the highest fluorescence concentration in the body for all of the A1AT constructs. The fluorescence concentration in the liver for PSA-A1AT [15.76 ± 1.19 arbitrary fluorescence units (AU)] was higher than for both diSA-A1AT (12.85 ± 1.43 AU) and A1AT (9.97 ± 1.35 AU). However, in the kidney, the fluorescence concentration for A1AT (9.68 ± 2.25 AU) was higher than for both PSA-A1AT (3.64 ± 0.50 AU) and diSA-A1AT (5.29 ± 0.99 AU). These results suggest that for PSA-A1AT and diSA-A1AT, the liver is the predominant route of elimination. In contrast, for A1AT both the liver and kidney are equally major routes of elimination. We did not observe any abnormal uptake of modified A1AT in other organs.

Discussion

Here, we describe a two-step enzymatic polysialylation of glycosylated acceptor proteins. The process begins with the addition of a Neu5Ac primer onto existing glycans on the acceptor protein by Cst-II, a bifunctional α2,3/2,8 sialyltransferase. Based on reactions with synthetic acceptors, Cst-II activity is more efficient on acceptors already containing an α2,3-linked Neu5Ac (15, 29), suggesting it may preferentially modify Neu5Ac-terminated glycans on the protein acceptor. The Neu5Ac cap on eukaryotic N-glycans can either be α2,3- or α2,6-linked. Cst-II is known to transfer α2,8-linked Neu5Ac to α2,3-linked Neu5Ac, and it has very good but somewhat reduced activity on α2,6-linked Neu5Ac, as indicated in studies using small oligosaccharides (30). Here, we have demonstrated that Cst-II can modify glycoproteins that contain either exclusively α2,6-linked termini (A1AT), or a mixture of α2,3/6-linked termini (factor IX and fetuin) efficiently. Furthermore, the results of our mass spectrometry analysis suggest that more than one antennae of a multiantennary N-glycan is accessible for modification. In summary, we have demonstrated the utility of this method to prime proteins with N-linked glycans with various branches and both α2,3- and α2,6-terminal sialic acids.

We did not demonstrate priming on the O-linked glycans from fetuin or factor IX, and we will need to examine proteins that only contain O-linked glycans (like interferon α2B, or granulocyte colony-stimulating factor). We have previously shown (15) that we could modify a glycopeptide with α2,3-sialyl-T antigen, and we are optimistic that such a glycan may be a substrate on a protein that only contains O-linked glycans.

Using mass spectrometry analysis of the glycans, we can monitor the extent of the Cst-II reaction by determining the number of additional Neu5Ac residues added. Although it would be difficult to confirm 100% conversion with this method, it provides more insight than SDS-PAGE analysis, because the change in migration of the protein is so small. Monitoring the extent of the priming reaction is important because prolonged incubations of the protein acceptor with Cst-II, as well as increased concentration of Cst-II in the reaction mixture, can lead to sialidase activity as was observed in prolonged incubations, or with excessive amounts of enzyme (31). Therefore, conditions must be chosen that allow extensive priming while minimizing the back reaction. Despite these challenges inherent in the priming reaction, we observed the complete consumption of the disialyl-proteins after treatment with MalE-Δ19PST_NM when the acceptor concentrations were high. This suggests that, at least with these proteins, we were able to create a minimum of one PSA chain on each of the acceptor molecules.

It is equally important to be able to control the polysialylation reaction, as well as the priming reaction. Our data suggest that varying the concentration of the acceptor protein in the polysialylation reaction affects the size distribution of the added PSA chains. At low acceptor protein concentrations, we see a range of PSA chain lengths, including very large chains whose length exceeds 70 Neu5Ac residues (Fig. S3). In contrast, at high acceptor protein concentrations, a narrower size range of shorter PSA chains is added. This ability to vary the length of the PSA chains being added to the acceptor proteins may be valuable for the modification of therapeutics, as it may be possible to select conditions that allow sufficient PSA modification for improved stability and circulation of the therapeutic protein, without adversely affecting the drug’s activity and tissue penetrability/interaction. Given that our approach selectively modifies existing glycans on the therapeutic protein, we believe these modifications are less likely to adversely affect activity compared with existing chemical coupling approaches.

Our in vitro assay of A1AT with human neutrophil elastase showed that the PSA modification did not negatively affect the activity of the protein. We were able to modify another therapeutic protein, human factor IX, but we were unable to obtain in vivo serum half-life data because we lacked sufficient quantities of the protein to perform all of the optimization, labeling, and subsequent enzyme assays and animal work. Naturally, the suitability of this method for the modification of other therapeutic proteins will need to be determined for each acceptor protein.

In humans, the half-life of naturally occurring glycosylated A1AT is about 4–5 d (32) and is administered as a weekly intravenous infusion in A1AT-deficient patients. In rodents, the half-life of glycosylated A1AT is much shorter, with values in the literature ranging from 3–20 h (33–35). Biologics often demonstrate complex and species-specific pharmacokinetic profiles compared to small molecules due to their unique physiochemical properties (36). In this study, the half-life of injected A1AT in mice was in the order of hours, but through polysialylation of existing glycan structures, the half-life was extended to days without any loss of in vitro protease inhibitor activity. Because the A1AT protein used in this study was derived from human plasma, the gain in half-life obtained in mice through polysialylation is expected to result in an increased half-life in humans. In vivo optical imaging proved useful in confirming pharmacokinetic data and that the protein modifications did not produce any abnormal uptake of the protein in nonmetabolic/clearance organs.

In conclusion, prolongation of the half-life for A1AT by the enzymatic addition of short PSA chains to the existing glycan structure may have significant benefit in the treatment of A1AT-deficient patients by both extending the dosing interval and reducing the amount of protein administered during therapy. These attributes could allow for better patient compliance of long-term dosing and a reduction in the costs associated with repeated, life-long drug administration (20).

Materials and Methods

Materials.

A1AT was purchased from rPeptide. Recombinant human factor IX and CMP-Neu5Ac were a kind gift of NEOSE Technologies Inc.

Construction, Expression, and Purification of MalE-Δ19PST_NM and Cst-II.

The amplification of pst_NM from chromosomal DNA of N. meningitidis 992B, and subsequent cloning into pCWmalE-thrombin, were reported previously (15). The construct used in this study expresses a PST_NM derivative that lacks 19 residues from the N terminus, resulting in increased solubility compared to MalE-PST_NM (Fig. S4). A derivative lacking 32 amino acids from the N terminus was also tested. Oligonucleotide sequences for generation of the truncated protein are listed in Table S2. MalE-Δ19PST_NM was expressed in Escherichia coli AD202 (E. coli Genetic Stock Center, CGSC no. 7297) and purified using the conditions described previously for MalE-glycosyltransferase fusions (37), with the following changes. The enzyme was purified from the supernatant of a cell lysate (centrifuged at 27,000  × g) on a 5-mL Heparin HiTrap HP column (GE Healthcare) with a linear gradient of 0–60% using the following buffers: A, PBS pH 7.4; B, PBS pH 7.4 + 2 M NaCl. Cst-II was expressed and purified as previously described (37, 38), with the following changes. The supernatant of a cell lysate (centrifuged at 27,000  × g) was chromatographed on a 5-mL HiTrap Q HP column (GE Healthcare) with a linear gradient of 0–50% using the following buffers: A, 20 mM Tris-HCl, pH 8.3; B, 20 mM Tris-HCl, pH 8.3 + 1 M NaCl. The activity of the two enzymes was verified by testing activity on fluorescently coupled small molecule acceptors as described previously (15, 39, 40).

Cst-II and MalE-Δ19PST_NM Modification of Therapeutic Proteins.

A1AT was modified by incubating the protein with Cst-II using the following reaction conditions: 5 mg/mL A1AT, 50 mM sodium phosphate pH 8.0, 10 mM MgCl₂, 5 mM CMP-Neu5Ac, 600 mU/mL Cst-II for 30 min at 37 °C. Reactions with bovine fetuin, type III (Sigma-Aldrich) and factor IX were carried out in the same manner, with the optimizations to reaction components as described in SI Materials and Methods. The resulting diSA-protein was purified from the reaction mixture by chromatography of a 2-mL reaction on a 5-mL HiTrap Q HP column, as described for Cst-II above. Fractions containing the modified protein were pooled and desalted using an Amicon Ultra-15 device (10,000 molecular weight cutoff) (Millipore), followed by dialysis against PBS at 4 °C overnight. Samples were analyzed by SDS-PAGE and visualized with Coomassie staining. diSA-A1AT was incubated with purified PST_NM using the following reaction conditions: 5 mg/mL diSA-A1AT, 50 mM sodium phosphate pH 8.0, 10 mM MgCl₂, 10 mM CMP-Neu5Ac, 0.05 mg/mL MalE-Δ19PST_NM for 1 h at 30 °C. Other diSA-proteins were incubated similarly, with the optimizations to reaction conditions as described in SI Materials and Methods. Purification of PSA-A1AT from the glycosyltransferase and reaction components was achieved by chromatography of a 0.5-mL reaction on a 1-mL HiTrap Heparin HP column (GE Healthcare) in PBS pH 7.4 (a gradient was not necessary because the product flows through the column, whereas PST_NM remains bound). The purified protein was then dialyzed against PBS at 4 °C overnight.

CE-MS Glycan Profiling of Proteins.

An aliquot of 50 μg purified diSA-proteins (or unmodified proteins resuspended in PBS) was either treated with PNGase F to release the N-glycans or subjected to β-elimination to release O-linked glycans (41, 42). The resulting N- and O-linked oligosaccharides were loaded on PGCs and then washed with water (3.0 mL) to remove buffer and salts. Oligosaccharides were first eluted with 25% MeCN in 0.1% TFA and dried for further analysis using CE-MS in negative ion mode. CE-MS spectra were acquired using the 4000 Q-Trap mass spectrometer (Applied Biosystems/MDS Sciex) with a capillary electrophoresis interface. Precursor ion scans at m/z 290.2 and m/z 581.3 were used to identify glycans containing Neu5Ac Neu5Ac-Neu5Ac residues, respectively.

DMB Derivatization and Chromatography of PSA-A1AT.

Polysialylated proteins were analyzed using a reaction with the PSA labeling reagent DMB (Sigma-Aldrich), as described by Inoue and Inoue (28). Reactions were incubated in the dark for 48 h at 10 °C, in order to minimize PSA cleavage. Reactions were stopped by the addition of one-fifth of the reaction volume of 1 M NaOH, incubated overnight at 4 °C to hydrolyze the lactones formed, and then stored at -80 °C. DMB-derivatization reactions were separated on a DNAPac PA-100 column using an AKTA FPLC system equipped with a Waters 474 scanning fluorescence detector set at excitation and emission wavelength 373 and 448 nm, respectively. Elution was performed with a segmented linear gradient of 1 M NaNO₃, with steps at 2, 3, 10, 20, 25, 35, and 100% over 120 min total running time at 1 mL/ min, according to Inoue et al (43). Colominic acid (Sigma-Aldrich) was labeled to serve as a control for PSA. A DMB-labeled Neu5Ac hexamer (a gift from H.J. Jennings, National Research Council of Canada, Ottawa, ON, Canada) was also applied to the column to indicate the elution time of material containing six Neu5Ac residues, allowing the length of the PSA chains to be estimated (Fig. S3).

Evaluation of A1AT Activity.

In order to assess the activity of A1AT, the protein was incubated with its target, human neutrophil elastase (Sigma-Aldrich) in the presence of a synthetic peptide substrate for elastase, N-methoxy-succinyl-Ala-Ala-Pro-Val-p-nitroanilide (Sigma-Aldrich) as follows: 100 mM Tris pH 8.0, 0.3 mM N-methoxy-succinyl-Ala-Ala-Pro-Val-p-nitroanilide, 8.4 nM A1AT, 8.4 nM elastase. Several concentrations of the reaction components were tested in order to determine the optimal reaction conditions. The product was detected by the release of nitroaniline and monitoring the reaction at A₄₁₀ for 1 h.

Pharmacokinetic Experiments.

All animal procedures were approved by the National Research Council Canada—Institute for Biological Sciences Animal Care Committee and were in strict compliance with the recommendations of the Canadian Council of Animal Care. A1AT, diSA-A1AT, and PSA-A1AT were labeled with Cy5.5 succinimidyl ester (GE Healthcare) using methods recommended by the manufacturer. Labeling was optimized such that each protein had a dye/protein ratio ranging from 1–2. Cy5.5-labeled protein (200 μg) was injected via the tail vein in CD-1 mice and subjected to fluorophore-based pharmacokinetic analysis as described previously (44). Blood samples (50 μL) were collected after A1AT injection at different time points (30 min, 1 h, 2 h, 3 h, 4 h, 24 h, 48 h, and 72 h) by creating a small nick in the tail vein and then stored on ice in heparin-coated tubes. Levels of Cy5.5-labeled proteins were quantified in the blood samples using a fluorescent plate reader with excitation 670 nm and emission 690 nm. The concentration of protein in each sample was derived from standard curves consisting of a range of concentrations of the labeled proteins diluted in blood. Pharmacokinetic parameters were calculated using the WinNonlin pharmacokinetic software package version 5.2 (Pharsight Corporation). A two-compartment, IV-Bolus model was selected for pharmacokinetic modeling, because it best represented the actual data. This model is described by the following equation: C(t) = A exp(-αt) + B exp(-βt), where C(t) represents the concentration of agent in serum; A and B represent the zero time intercept of the alpha phase and beta phase, respectively; and α and β are disposition rate constants, α > β. The area under the serum concentration-time curve was calculated with the equation AUC_0–∞ = D/V/K₁₀, where D is dose given, V is apparent distribution volume, and K₁₀ is elimination rate constant. Total clearance was determined from the equation Cl/F = D/AUC_0–∞.

In Vivo Near-Infrared Fluorescence Imaging of A1AT Derivatives in CD-1 Mice.

Cy5.5-labeled A1AT (71.82 μg), diSA-A1AT (123.39 μg), and PSA-A1AT (100 μg) were injected via the tail vein in CD-1 mice. To account for small differences in Cy5.5 labeling of the proteins, the amount of protein injected was adjusted to ensure that the three proteins had the same initial fluorescence level. In vivo imaging studies were performed using a small-animal time-domain eXplore Optix MX2 preclinical imager (Advanced Research Technologies) as described previously (45, 46). Imaging was conducted at 1.5 h, 4 h, 24 h, 48 h, and 72 h after injection. Mice were anesthetized with isofluorane and then positioned on an animal stage in a chamber that allows for maintenance of gaseous anesthesia. In all imaging experiments, a 670-nm pulsed laser diode with a repetition frequency of 80 MHz and a time resolution of 12 ps light pulse was used for excitation. The fluorescence emission at 700 nm was collected by a highly sensitive time-correlated single photon counting system and detected through a fast photomultiplier tube. The data were recorded as temporal point-spread functions, and the images were reconstructed as fluorescence concentration-depth maps using a time-domain algorithm contained within the ART Optix Optiview analysis software 2.0 (Advanced Research Technologies). At the end of the experiment, animals were perfused with heparanized saline, and dissected organs were imaged ex vivo.