Glycobiology Advance Access originally published online on March 1, 2007
Glycobiology 2007 17(5):529-540; doi:10.1093/glycob/cwm017
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Selective clearance of glycoforms of a complex glycoprotein pharmaceutical caused by terminal N-acetylglucosamine is similar in humans and cynomolgus monkeys
4 Department of Analytical Chemistry
5 Department of Pharmacokinetics and Metabolism, Genentech Inc., 1 DNA Way, South San Francisco, CA 94080
6 Department of Clinical Pharmacology, F. Hoffmann–La Roche Ltd, CH-4070, Basel, Switzerland
1 To whom correspondence should be addressed: Tel: 650 225 6264; Fax: 650 225 3554; e-mail: jeb{at}gene.com
Received on November 17, 2006; revised on February 10, 2007; accepted on February 11, 2007
 |
Abstract |
|---|
 Top Abstract IntroductionResults and discussionConclusionsMaterials and methodsConflict of interest statementReferences |
|---|
To understand how the carbohydrate moieties of a recombinant glycoprotein affected its pharmacokinetic (PK) properties, the glycan distribution was directly assessed from serial blood samples taken during PK studies in cynomolgus monkeys and humans. The protein studied was an immunoadhesin (lenercept), containing an Fc domain from human immunoglobulin G (IgG-1) and two copies of the extensively glycosylated extra cellular domain of tumor necrosis factor receptor p55. The protein was recovered in pure form using a dual column, immunoaffinity-reversed-phase high-performance liquid chromatography method. The glycans were released and analyzed by matrix-assisted laser desorption ionization time of flight mass spectrometry (MALDI-TOF MS). Alternatively, trypsin was used to obtain glycopeptides, and these were analyzed by MALDI-TOF. The composition versus time profiles show that the distribution of glycans in the Fc domain was not altered over 10 days of circulation, consistent with their sequestration in the interior of the protein. However, the glycan composition in the receptor domain was changed dramatically in the first 24 h and then remained relatively constant. Analysis of the acidic glycans (derived exclusively from the receptor domain) showed that, in the rapid initial phase of clearance, glycans carrying terminal N-acetylglucosamine (tGlcNAc) were selectively cleared from the circulation. This phenomenon occurred similarly in humans and cynomolgus monkeys. Sialic acid content and terminal galactose showed only small changes. These data confirm the correlation of tGlcNAc and half-life of the molecule, and support the hypothesis that the mannose receptor (which can also bind tGlcNAc) causes the variable clearance of this molecule.
Key words: immunoadhesin / lenercept / terminal N-acetylglucosamine / selective glycan clearance / IgG1 Fc glycans / terminal galactose
|
Introduction |
|---|
|
TopAbstractIntroductionResults and discussionConclusionsMaterials and methodsConflict of interest statementReferences |
|---|
The metabolic clearance of glycoproteins can be caused by specific receptors which recognize certain structural features on the glycan moieties (Ashwell and Harford 1982
). The two major receptors which have received extensive study are the asialoglycoprotein receptor (Stockert 1995) and the mannose (Man) receptor (Stahl 1992). The asialoglycoprotein receptor binds to complex-type N-linked glycan structures in which two or more sialic acids are absent, wherein the underlying galactose (Gal) residues become the terminal saccharides (Rice and Lee 1993). The mannose receptor recognizes high-Man N-linked glycans and terminal N-acetylglucosamine (tGlcNAc) residues and could also be referred to as the mannose/N-acetylglucosamine (Man/GlcNAc) receptor. Animal studies of glycoprotein clearance have shown that alterations in the glycan structures can cause changes in their pharmacokinetic (PK) properties (Morell et al. 1971). Both competition studies (Achord et al. 1977) and in vitro binding studies (Stahl et al. 1978; Kéry et al. 1992) have confirmed the identities of the glycan structures responsible as those which are bound by these receptors, namely the outer portions of the monosaccharides that terminate the glycan structure.
Adequate control of the glycoform microheterogeneity of glycoprotein pharmaceuticals is critical for the production of batches possessing reproducible PK, and therefore efficacy, properties (Cumming 1991; Meier et al. 1995; Jenkins 1996). From the known properties of the clearance mechanisms above, for glycoproteins which lack high-Man structures, the main focus of assessment of microheterogeneity is often the sialic acid content (Drickamer 1991; Kronman et al. 1995). This is true, in part, because of the ready availability of methods to quantitate this parameter (Anumula 1995) and in part due to the complexity of both the glycan structures and the methods required to analyze them in detail. We have demonstrated [R. Keck et al. 2007 (submitted)] that methods which analyze only the terminal saccharide complement [i.e. sialic acid, terminal Gal (tGal), and tGlcNAc] provide sufficient information to assure PK control of a complex glycoprotein that lacks high-Man structures.The glycoprotein lenercept [(tumor necrosis factor (TNF)rp55:Fc] is an immunoadhesin (Chamow and Ashkenazi 1996) composed of the Fc domain of immunoglobulin G (IgG-1) and two copies of the extra cellular domain of the p55 TNF receptor (Figure 1) (Ashkenazi et al. 1991). It carries eight N-linked glycans and has been evaluated in rheumatoid arthritis and sepsis indications (Abraham et al. 1997). Analyses of clinical study batches showed variations in the level of galactosylation; however, the extent of sialylation of these Gal residues was essentially constant [R. Keck et al. 2007 (submitted)]. The area under the curve (AUC) of the PK profile in human volunteer trials was assessed for nine batches and was found to be dependent on, and predicted by, the tGlcNAc content, but independent of the tGal or sialic acid content [R. Keck et al. 2007 (submitted)].
|
The work described here extended our understanding of the role that specific glycan structures play in modulating PK properties and allowed us to identify glycans that play no role in clearance. In all but one study to date, the conclusions were derived from analysis of the material injected into the animals and the correlation of that information with the measured PK properties. Herein, data are presented on the direct characterization of the glycan structures present at various times after injection. Huang et al. (2006)
took a similar approach; however, in that case the molecule was not highly glycosylated, and none of the glycans played a role in the clearance of the molecule (Huang et al. 2006). In this study, the direct characterization of the glycans was achieved by capturing lenercept from the serum samples using a dual-column affinity high-performance liquid chromatography (HPLC) method (Battersby et al. 1999) combined with a microscale glycan release method (Papac et al. 1998). Released glycans were then analyzed using the highly sensitive matrix-assisted laser desorption ionization time of flight mass spectroscopy (MALDI/TOF) technique (Figure 2). These analyses were performed on samples obtained from both cynomolgus monkey and human PK trials of several batches of drug. The results confirm that molecules bearing tGlcNAc residues in the receptor domain are selectively cleared in the initial rapid phase of the clearance and that this mechanism occurs similarly in the two species. This observation not only confirms that the monitoring of sialic acid content may not be sufficient to ensure consistent PK properties for glycoproteins but also suggests that cynomolgus monkey glycoprotein clearance mechanisms are similar to those in humans.
|
|
Results and discussion |
|---|
|
TopAbstractIntroductionResults and discussionConclusionsMaterials and methodsConflict of interest statementReferences |
|---|
During the development of lenercept as a pharmaceutical, batches were assessed for their PK properties in human and cynomolgus monkeys. The concentration of the drug was measured at various times after injection, using an immunobinding assay, and typical profiles from the cynomolgus monkey study are shown in Figure 3. Shown in this figure are data from three batches of drug which were used in clinical trials for sepsis and septic shock (Abraham et al. 1997
). Lot number M11-1/3 was manufactured using the first process (Process A), in which the cell culture yields were relatively low, while the other two lots, M11-72 and -73 were manufactured using a process (Process C) in which the cell culture yields were increased by a factor of approximately two. The purification process employed for all three lots was similar. The concentration–time profiles, in monkeys and humans, for the lots are qualitatively similar (human data not shown), that is, a rapid decrease in concentration during the first day or two ("initial" phase), followed by a slower decline until the terminal elimination phase is reached, for which the rate of decrease is similar in all three lots. Table I summarizes the PK parameters obtained for the different lots in monkeys, and shows that M11-1/3 produced a higher AUC than the other two lots, which were statistically indistinguishable from each other. The relative AUC of lots M11-72 and M11-73 (Process C), compared with lot M11-1/3 (Process A) in cynomolgus monkeys is similar to that found in the human study (data not shown), at around 70%. Also, the terminal elimination half-lives (see the Materials and methods section) of all lots were similar in both cynomolgus monkey (approximately130 h), and humans (160–170 h). The variability of the decrease in the initial phase of clearance results in a variable AUC from lot to lot.
|
|
Extensive physicochemical characterization of lenercept [R. Keck et al. 2007
(submitted)] has shown that this clearance variability is determined solely by the content of tGlcNAc. To investigate this observation in more detail, the serum samples taken for determination of drug concentration from these PK studies were used as a source for assessing the glycan status over time after injection. This was accomplished with a dual-column purification HPLC method capable of recovering lenercept from these samples in a sufficiently pure state for glycan characterization (Battersby et al. 1999). This method was optimized to ensure complete (i.e. >95%) capture of lenercept from either human or cynomolgus monkey serum or plasma samples, thereby preserving the microheterogeneity of the glycoforms. The method also provided quantitative measurement of the drug concentrations in the samples. As expected, these values were similar to those obtained using the immunobinding assay (data not shown).
An initial decrease in concentration occurs as the molecule equilibrates between the intra- and extravascular spaces (Kim et al. 1994). However, this occurs rapidly and would not be expected to vary from batch to batch. The differences in AUC values arise primarily as a result of the amplitude of the initial phase, which is presumably augmented by clearance mediated by glycoprotein receptors, while the terminal phase represents the clearance of the drug by other mechanisms. With the availability of sensitive glycan analysis tools (Papac et al. 1998) and the ability to recover drug from blood samples drawn for PK analysis (Battersby et al. 1999), we sought to understand the mechanism of the variable clearance directly (rather than extrapolating from the PK properties and the physicochemical properties of the material that was administered).
The dose of lenercept (on a milligram/kilogram basis) was substantially greater for the monkey studies (5 mg/kg) than for the human studies (0.08 or 0.12 mg/kg). Therefore, more drug was recovered from the monkey study for analysis. Complete time course profiles (9 days) were analyzed by recovering the drug from each time point from two monkeys, one injected with M11-1/3 and the other with M11-73. Once the drug was recovered from the blood sample (typically 3–10 µg), the glycans were released from the protein by immobilization on polyvinylidenedifluoride (PVDF) followed by reduction, alkylation, and digestion with peptide N-glycosidase (PNGase F) as described previously (Papac et al. 1998). The released glycans were analyzed by MALDI/TOF MS using either dihydroxybenzoic acid (DHB), for positive mode analysis of neutral glycans, or trihydroxyacetophenone (THAP) for negative mode analysis of acidic (i.e. sialylated) glycans.
Presented in Figure 4 are the spectra obtained with negative mode mass spectrometic analysis of the glycans present in circulation either 5 min (upper trace) or 3 days (lower trace) after injection of M11–73 into a cynomolgus monkey. The peaks are labeled according to a four-digit code (associated structures are shown in Figure 5) in which 0000 designates the tri-Man core present in all complex-type N-linked glycans. The first digit signifies the number of branches (or antennae) present as determined by the number of GlcNAc residues attached to the Man residues; the second digit signifies the presence or absence of a core fucose; the third digit signifies the number of Gal residues and the fourth digit represents the number of sialic acid residues in the glycan. For some designations there are multiple structures possible.
|
|
Labeled in bold and underlined are glycan structures which have one or more exposed (i.e., terminal) GlcNAc residues (Figure 4). These can be seen to be considerably reduced in the 3-day sample compared with the 5-min sample. The 5-min sample had a very similar glycan profile to the material before injection (data not shown). In both the spectra, the data are displayed with the y-axis scaled to 100% for the most abundant glycan namely, the 2122 species (i.e. biantennary, bisialylated glycan). The glycans in Figure 4 are exclusively from the receptor domain, since there are no acidic glycans in the Fc (see Figure 10). Only signals for glycans containing at least one tGlcNAc residue (such as structures 2111 and 3122) are selectively reduced, suggesting that this specific structural feature is recognized by a clearance receptor. At 3 days these glycans have decreased significantly, but not completely to zero.
The data contained in these MS spectra can be converted to yield information on the terminal saccharide content of the sample. When MALDI-TOF MS is used for glycan analysis it can be shown that within each spectrum, the peak area distribution is a quantitative reflection of the molar distribution of the individual glycans (Harvey 2002). Accordingly, it is possible to reduce the data from each such spectrum into three values; the moles of terminal saccharide per mole of polypeptide for GlcNAc, Gal, or sialic acid (see the Materials and methods section). We have previously shown that there is good agreement between this method and direct methods of measuring the terminal sugars with R > 0.9 (Papac DI, Briggs JB, Chin E, Nayak N, Lerner L, Jones AJS, unpublished data). By measuring these three values for each time point, it is possible to interpret the changes in the key parameters which characterize glycan distribution with respect to potential glycan-mediated clearance mechanisms (Figure 6). This was performed for a complete PK profile from two monkeys, each injected with a different batch of lenercept. The results are shown in Figures 6A, B, and C. In these figures, the glycan composition at the first data point (5 min) is essentially unchanged from the starting material (data not shown). It should be noted that these plots indicate "composition over time" rather than "concentration over time". Shown in Figure 6A is the sialic acid content of the acidic glycans over time, and this indicates that there is only a small change over the 10 days of the study. The similar sialic acid content of Process A and Process C materials is confirmed at the initial time point (the first time point is 5 min—time zero is the injection time). This was similar to that measured before injection (data not shown). The modest increase in sialic acid in Process C material in the initial rapid phase is a reflection of the decrease in tGlcNAc content (Figure 6C) since the tGal content does not change significantly (Figure 6B).
|
These composition versus time profiles were calculated (in terms of moles of saccharide per mole of protein) with the simplifying assumption that all the receptor domain glycans are acidic. While this was necessary due to the limited material available, acidic glycans do comprise the majority (80–90%) of the receptor domain glycans (Papac DI, Briggs JB, Chin E, Nayak N, Lerner L, Jones AJS, unpublished data), and the Fc domain (neutral) glycans do not contribute to these profiles (see Figure 7).
Figure 6B shows a similar picture for the tGal content of the acidic glycans for the same lots. The higher Gal and tGal content of Process A material is also confirmed, but neither changes significantly over the whole clearance process. The fact that a higher tGal content (of Process A material) does not result in faster clearance by the asialoglycoprotein receptor mechanism (Stockert 1995) is probably related to the fact that glycans bearing three tGal residues (3130, 4141), which are the forms strongly bound by that receptor (Rice and Lee 1993), represent less than 1% of the total receptor glycans (data not shown). In addition, not only is the percentage of glycans bearing three tGal residues low but also the overall content of tGal is low.
In Figure 6C, the data for the terminal GlcNAc content are shown. Here, the composition of the glycans changes dramatically. The lower initial level of tGlcNAc is confirmed for Process A material, relative to Process C material, although both drop in the initial phase. The levels do not drop to zero in either case, suggesting that some receptor domain tGlcNAc is not accessible to the clearance receptor, and this occurs to a similar extent in both materials. The significant change in composition that occurs during the initial phase, where the drug concentration is dropping rapidly, is consistent with the selective clearance of molecules bearing these glycans being the major cause of the additional decline in concentration beyond the glycan-independent distribution into the extra vascular space for the initial phase. It cannot be ruled out that certain lenercept molecules are cleared through the asialoglycoprotein receptor pathway, but it can be inferred that this is only a minor fraction. These data directly support the observed correlation between tGlcNAc content and AUC, and with no correlation observed for tGal or sialic acid [R. Keck et al. 2007 (submitted)].
Previously it was shown through characterization of purified lenercept material that the Fc domain (obtained by papain digestion) of this immunoadhesin contains almost exclusively biantennary neutral glycans, while the three sites in the receptor domain contain mainly acidic glycans (Papac DI, Briggs JB, Chin E, Nayak N, Lerner L and Jones AJS, unpublished data). With sufficient material it was possible to assemble a complete description of the glycans present in the injected material (Figure 7). However, there was insufficient material to accomplish this on the samples recovered from serum at the serial time points. Nonetheless, the neutral glycans from the whole molecule could be analyzed. A majority of the signals from the neutral glycans released from the whole molecule will therefore be derived from the Fc domain (Figure 7). The fact that the Fc glycans account for approximately 63% of the total neutral glycans (24% in the Fc portion and 14% in receptor domain) meant that they dominated the MALDI TOF mass spectrum and prevented us from directly assessing the fate of the lower abundant neutral glycans of the receptor domain. Figure 6D shows the three major biantennary neutral glycans (from the whole molecule), as composition versus time profiles. After some initial changes in which G0 (two tGlcNAc residues) decreases and G1 (one tGlcNAc residue) and G2 increase, the pattern is stable for the whole period up to 10 days. The initial changes are consistent with the decrease in neutral, receptor domain glycans, carrying tGlcNAc which are cleared by the same mechanism as for acidic glycans with tGlcNAc. Thus, after the first 12–24 h, all lenercept molecules with accessible tGlcNAc residues have been removed from circulation and the remaining clearance profiles are similar for all lots; this is consistent with the curves in the drug concentration profiles seen in Figure 3, becoming parallel after approximately 12–24 h. The long half-life of these remaining molecules is comparable with those observed for antibodies and isolated Fc domains (Kim et al. 1994) and is presumably driven by clearance mechanisms other than those mediated by specific glycan receptors.
|
The data from the cynomolgus monkey PK studies described here are very similar to the results obtained in human trials. However, in humans a small difference was observed for AUC ratios between the two Process C lots (R. Keck et al. 2007 [submitted]). This was not observed in the cynomolgus monkey PK study and was probably due to the smaller sample size used in this study (there were six animals per group in the cynomolgus monkey study, while in the human PK trials there were 8–20 individuals per group). With this small difference in mind, recovered samples provided an opportunity to address, at the molecular level, the comparability of monkey and human clearance mechanisms for this drug. We, therefore, recovered the lenercept from some human PK samples to address whether the underlying mechanism was qualitatively and/or quantitatively similar. The lower dose used in the human trials (0.08–0.12 mg/kg), compared with the dose used in the monkey studies (5 mg/kg) necessitated a modification of the purification method (Battersby et al. 1999
) and resulted in smaller amounts being recovered (typically 1 µg or less as opposed to 10 µg or less in monkeys). Also, fewer samples were taken in the human PK studies than in the cynomolgus monkey studies. The signal to noise ratio in the MALDI TOF masss spectra was consequently lower (data not shown) than for those from the monkeys. Figure 8 shows tGlcNAc data from four patients (and two monkeys), two of whom received lot M11-47 (Process B). This lot had the highest content of tGlcNAc. One patient received lot M11-72 (Process C) with an intermediate level of tGlcNAc, and the remaining patient received lot M11-1/3 (Process A), with the lowest level of tGlcNAc. The glycan composition at the first data point (5 min) is essentially unchanged from the starting material (data not shown) and confirms the rank order of the tGlcNAc values measured for these lots [R. Keck et al. 2007 (submitted)]. Also the similarity of the profiles from the two patients receiving lot M11-47 demonstrates the reproducibility of the data obtained. The profiles also show that in humans, as in cynomolgus monkeys, the glycoforms bearing accessible tGlcNAc in the receptor domain are selectively removed in the early phase of clearance, and that there is a fraction which is not cleared and, therefore, long-lived in the circulation. The stability of these values (and those in Figure 6) after the early phase suggests that the clearance of the long-lived molecules after about 12 h does not involve glycan-mediated mechanisms. This is consistent with the value of the terminal half-life being similar to that of antibodies and isolated Fc domains (Kim et al. 1994). We were therefore interested in understanding what caused some tGlcNAc to be cleared by the clearance receptor(s) and others to be excluded from that pathway. This lack of influence of normal Fc glycosylation (Kim et al. 1994; Wright and Morrison 1994) on antibody clearance is discussed further below.
|
The data presented so far have been derived from the glycan populations from the whole molecule, with the acidic glycans coming exclusively from the receptor domain and the neutral glycans coming predominantly from the Fc. The ability to manipulate sub-microgram amounts of protein for glycan analysis suggested that microscale peptide mapping, coupled with the high sensitivity of MALDI-TOF MS, might provide information on the glycans at specific sites on the molecule. The manipulation of microgram amounts of protein for glycan analysis by the use of PVDF membranes was described above and a similar approach was used to develop a reliable peptide mapping procedure for these amounts. The glycopeptide peaks (or clusters) were collected and either analyzed directly by MALDI-TOF MS or treated as for the intact molecule, i.e. they were immobilized on PVDF and subjected to PNGase F digestion to release the glycans for analysis (Figure 2). While the amount of information obtained in this way was limited by the amount of protein and the additional manipulations, the method provided some useful insights into the clearance mechanisms in the cynomolgus monkey. There are three potential N-linked glycosylation sites in the receptor. These are at Asn25, Asn116 and Asn122. Digestion of lenercept with trypsin releases these sites with Asn25 on tryptic fragment T5, and Asn116 and Asn122 on peptide T16. Tryptic fragment T5 (containing Asn25) is a relatively small tryptic peptide and thus did not bind significantly to the PVDF membrane. As a consequence, the glycans could not be released and therefore were analyzed as glycopeptides only. THAP was used as the matrix because this matrix was shown to have good sensitivity for acidic glycopeptides (Papac et al. 1996
).
The glycopeptide distribution from the asparagine-25 glycosylation site is shown in Figure 9A and B. These figures show that good quality mass spectra could be obtained by analyzing these glycopeptides directly. All peaks were assigned to glycans except for peaks at m/z of approximately 3220 and 3310 (indicated by asterisks) which were identified as co-eluting, nonglycosylated peptides. Figure 9C and D show data obtained by re-immobilizing the larger glycopeptides and assessing the released glycans as described earlier for the intact protein. Figure 9A presents the data from the 5-min time point, and Figure 9B presents the data from the 3-day time point of a monkey injected with lot M11-73. Glycopeptides with tGlcNAc are indicated by bold, underlined labels. The comparison of these spectra show two major points: first, there are relatively few glycans at the Asn25 site with tGlcNAc (the most abundant being 2100 and 2111) and, secondly these tGlcNAc-bearing structures are completely cleared by the 3-day time point, whereas the distribution of the other glycans seems relatively unaltered. The other two receptor glycosylation sites (Asn116 and Asn122) were found on the same tryptic peptide and could not be easily cleaved for individual analysis. However, this peptide was sufficiently large to be efficiently immobilized on PVDF and after PNGase F treatment, the free glycans were analyzed as above. Glycan distributions at Asn116 and Asn122 at 5-min and 3-days are shown in Figure 9C and D, respectively. From Figure 9C it can be seen that the glycans bearing tGlcNAc (bold and underlined) are in greater abundance than similar glycans on Asn25 (Figure 9A), and that by day three (Figure 9D) a significant fraction of them have not been cleared, even though they contain the moiety known to bind to the Man/GlcNAc receptor. The glycan compositions at these sites, at the 5-min time point, confirm the results obtained from the characterization of the injected material (data not shown), with Asn25 having approximately three-fold lower tGlcNAc content than at Asn116 and Asn122. Receptor peptide T16 not only has two glycosylation sites (Asn116 and Asn122) but also a higher tGlcNAc content per glycan than Asn25, and, although we have identified the glycans on peptide T16, we have not been able to determine the glycan distributions between the two sites independently. However, an examination of the crystal structure of the receptor (Banner et al. 1993) and the sequence of the immunoadhesin construct indicate that both sites on peptide T16 are in close proximity to each other (and possibly to the hinge region of the molecule), suggesting steric hindrance, and that this steric hindrance may be the cause of lower galactosylation, i.e. higher tGlcNAc content during synthesis of the molecule in Chinese hamster ovary (CHO) cells.
|
We suggest that steric constraints during glycosylation (glycan processing), which results in lower galactosylation at Asn116 and Asn122, and consequently higher levels of tGlcNAc, also prevent the Man/GlcNAc receptor from access to some of these tGlcNAc residues, and thereby protects them from clearance. Whether this occurs at both sites or only one cannot yet be determined. Such accessibility constraints have been proposed to account for the lack of effect on clearance by glycans in the Fc domains of antibodies (Wright and Morrison 1997
). Additionally, steric effects between glycans themselves have been shown to occur (Ferrara et al. 2006). Thus, the close proximity of glycans on Asn116 and Asn122 may allow for an intramolecular glycan interaction. Such an interaction could prevent binding to the Man/GlcNAc receptor and/or alter their affinities for the receptor. In addition to the higher levels of tGlcNAc found on tryptic peptide T16, there is a modest proportion of a triantennary structure with two tGal residues (3131 in Figure 9C) as a result of incomplete sialylation. This glycan (highlighted by $ in Figure 9D) decreases relative to those glycans which do not have tGlcNAc, i.e. 2121, 2122, 3132, and 3133 in Figure 9C. The decrease is approximately 50% over the 3-day interval, consistent with only a low affinity of such triantennary structures for the asialoglycoprotein receptor and consequently, a minor role for tGal in the clearance.
The biochemical characteristics of the Fc domain of this immunoadhesin have been assessed [R. Keck et al. 2007 (submitted)], and it was found to carry glycans similar to those found in unmodified IgG molecules (Wright and Morrison 1997), with nonsialylated biantennary structures with two tGlcNAcs (2100, G0, approximately 80%) or one tGlcNAc (2110, G1, approximately 15%) accounting for more than 95% of the Fc glycans. The peptide carrying this glycosylation site was collected from the peptide separation and analyzed by MALD/TOF MS using THAP as the matrix, and the mass spectrometer in the negative mode. The spectra obtained from one animal are shown in Figure 10, upper panel, and demonstrate good signal-to-noise ratios and the predominance of the G0 (m/z = 2640) and G1 (m/z = 2800) forms, i.e. with two and one tGlcNAc residues, respectively, each with minor sodium and potassium adduct peaks and a small but definite peak for the form with two tGal residues, G2 (m/z = 2960). Figure 10, lower panel, presents the data for two animals (one injected with M11-1/3 and the other with M11-72). The data were processed as in Figure 6C, i.e. with moles of tGlcNAc per mole of Fc content as the y-axis. These values are derived solely from the Fc glycosylation site carried on tryptic peptide T28 recovered after PVDF tryptic digest of lenercept recovered from PK samples. The amounts of G0 (two tGlcNAc residues) and G1 (one tGlcNAc residue) change modestly over the first two time points and then become nearly constant. The consistency of amounts from later time points suggests that the small initial drop in tGlcNAc is indeed real (Figure 10, lower panel). However, this decrease is probably not due to clearance mediated through the Fc glycans since we have previously correlated galactosylation in the Fc with galactosylation in the receptor domain (Papac DI, Briggs JB, Chin E, Nayak N, Lerner L, Jones AJS, unpublished data) and therefore believe that it is more likely that this clearance is due to a "bystander" effect i.e. molecules carrying high tGlcNAc in the receptor will also carry higher tGlcNAc in the Fc because they were synthesized when the cells were galactosylating less efficiently. Their removal due to clearance mediated by the glycans in the receptor domain will consequently reduce the tGlcNAc in the Fc at the same time. The data are, therefore, still consistent with previous observations that normal glycan structures, or even the absence of glycosylation in the Fc (Kim et al. 1994), have no effect on clearance of the Fc domain through glycan receptor mechanisms. Indeed, these data, obtained directly by following the glycan composition over time during circulation, provide definitive support for this conclusion. A summary of the glycan distribution and type found on lenercept is given in Figure 7.
|
From the seminal work of Ashwell on glycan-mediated clearance (Ashwell and Morell 1974
; Ashwell and Harford 1982), it became clear that the asialoglycoprotein receptor is very efficient at rapidly removing many, but not all, glycoproteins bearing tGal from circulation. This resulted in particular attention being paid to the sialic acid content, and its consistency, in the manufacture of pharmaceutical glycoproteins. The clearance behavior of follicle-stimulating hormone (Mulders et al. 1997) provides an example of a predictive correlation between PK properties and sialic acid content, while Kronman et al. (1995) concluded that the more relevant correlation was with the extent to which sialic acid residues were absent. The unexpected observation that transferrin retained a long half-life in the asialo form (Morell et al. 1968) was subsequently explained by the strict geometric requirements of the receptor (Rice and Lee 1993) and the weak binding of biantennary glycans, which are the predominant glycans on transferrin (Spik et al. 1975). In the initial studies, the clearance of asialo-agalactoproteins (i.e. carrying only tGlcNAc), created as a result of galactosidase treatment, was much slower, reverting almost to that of the fully sialylated protein (Morell et al. 1968; van den Hamer et al. 1970). The clearance of asialo-agalactoprotein could be competitively reduced by mannan and other high-Man structures (Achord et al. 1977). Thus, a new receptor was identified in the liver (Stockert et al. 1976) as distinct from the asialoglycoprotein receptor, and is now commonly referred to as the mannose receptor (Stahl 1992; Taylor 1993) or the Man/GlcNAc receptor. It is considered to play a role in both innate (Taylor 1993) and adaptive immunity (Stahl and Ezekowitz 1998) since large proportions of high-Man structures are rare in mammalian glycoproteins, but common in bacteria, fungi, and parasites. As a result, there are few recent studies on mammalian glycoproteins in which tGlcNAc residues are considered, although there are in vitro studies that clearly show it can bind to the receptor (Stahl et al. 1978).
In the case of lenercept, however, a strong dependence of AUC on tGlcNAc content, presumably mediated by the Man receptor, was established [R. Keck et al. 2007 (submitted)], and directly confirmed in this work. While the magnitude of this dependence was somewhat unexpected, it can be explained by the following. For molecules that lack glycans recognized by the two major receptors described earlier (e.g. for many antibodies, for isolated Fc domains, and for lenercept in the terminal phase), the long half-life is probably modulated by another receptor, the FcRn receptor (Raghavan and Bjorkman 1996). Indeed for lenercept, there is evidence to show that the typical species scaling effect on clearance does not apply, possibly because of its long half-life (Richter et al. 1999). The observed correlation of AUC and tGlcNAc content is driven primarily by variability of the amplitude of the initial phase when these structures are being cleared, and this is complete by 12–24 h (compare Figures 3 and 6C). At this point, the curves become parallel, but the AUC from that point on is still a substantial fraction of the total AUC. The magnitude of this effect is also likely to be amplified by the fact that lenercept has six glycosylation sites which may contribute to the clearance such that the whole molecule is cleared as a result of carrying one tGlcNAc, regardless of the status of the other glycans on the same molecule. For example, 5% of the glycans bearing tGlcNAc spread over six sites could be responsible for clearing 30% of the protein. On the other hand, the mannose receptor, containing multiple carbohydrate recognition domains, each of which can bind only a single saccharide, results in variable affinities for glycoproteins due to a clustering effect (Stahl and Ezekowitz 1998). A given receptor molecule could therefore bind to more than one glycan on an individual glycoprotein molecule. It has been shown (Noormann et al. 1998) that the protein structure may also influence the conformation of a high-Man structure to enhance its affinity for the receptor by a similar mechanism. For lenercept, only a small contribution to the initial clearance phase may be ascribed to the asialoglycoprotein receptor. We have shown that the major factor affecting the variability of the clearance of lenercept, in both cynomolgus monkey and humans, is the tGlcNAc content, and suggest that this is mediated by a specific receptor, presumed to be the Man/GlcNAc receptor. This conclusion emphasizes the need to identify glycans that are not only capable of binding to clearance receptors but are located such that binding is not sterically hindered, and for monitoring the status of all three terminal saccharides on glycoprotein pharmaceuticals, especially those designed to have a long half-life, to ensure consistency of drug performance from batch to batch.
|
Conclusions |
|---|
|
TopAbstractIntroductionResults and discussionConclusionsMaterials and methodsConflict of interest statementReferences |
|---|
Recovery of sufficiently pure glycoprotein pharmaceuticals from PK samples can be achieved if circulating concentrations are sufficiently high (in the µg/mL range). The methodology developed also allows such a method to be used for quantitation purposes. Analysis of such samples for glycan characterization is possible due to the high sensitivity of the MALDI TOF MS instrument and suitable matrices. Manipulation of low microgram amounts of protein for glycan release or peptide mapping is facilitated by the use of PVDF microwell plates. These methods allow direct examination of the role of specific glycans during clearance of a heterogeneous population of glycoforms. Microscale peptide mapping allows an assignment, on the basis of glycosylation site, of the glycans involved.
For lenercept we conclude that molecules bearing accessible tGlcNAc moieties in the receptor domain are selectively cleared in the early phase of clearance. Some tGlcNAc moieties in the receptor domain are not cleared, probably due to limited access to the clearance receptor. These same steric constraints may have originally limited the galactosylation at these sites (Asn116 and/or Asn122). Other tGlcNAc moieties appear to be completely accessible (Asn25). The magnitude of the tGlcNAc effect directly confirms the basis for the predictive value (for AUC) of overall tGlcNAc content of the batches measured before injection. The glycans in the Fc domain are not accessible to the clearance receptors and play no role in the clearance of the molecule. Clearance of lenercept in humans and cynomolgus monkeys is similarly mediated, in both magnitude and mechanism, by selective removal of molecules bearing tGlcNAc in the receptor domain. Because they are present at such low levels in the receptor glycans, glycans with three terminal Gal moieties do not play a significant role in the variability of AUC from batch to batch. The receptor responsible for tGlcNAc-mediated clearance of lenercept is probably the Man/GlcNAc receptor.
|
Materials and methods |
|---|
|
TopAbstractIntroductionResults and discussionConclusionsMaterials and methodsConflict of interest statementReferences |
|---|
Various batches of lyophilized clinical-grade lenercept, from three different cell culture processes, identified by process and/or batch number, were used in these studies. One of the groups in the cynomolgus monkey study received material which was not lyophilized but was maintained as a concentrated "bulk" drug substance. Lenercept was recovered in pure form from plasma samples by a dual-column method as previously described in detail (Battersby et al. 1999
). Briefly, samples were diluted and applied to a 100 µL immunoaffinity column with an immobilized anti-TNF receptor monoclonal antibody; after appropriate washing, the column was eluted with acid and the eluate was captured on a 16 µL reversed-phase HPLC column. This column was then eluted with an acetonitrile gradient and the lenercept peak was collected, neutralized, and frozen until analyzed. Glycan analysis of the purified samples was performed as previously described (Papac et al. 1998). Briefly, protein samples (0.5–10 µg) were immobilized on separate PVDF membranes in wells of a microtiter plate and reduced and alkylated. Glycans were released with PNGase F, converted to the reducing form with dilute acetic acid, desalted by cation exchange, and analyzed by MALDI-TOF MS using DHB [(2,5-dihydroxybenzoic acid) 5-methoxysalicylic acid, 95/5 (w/w)] as the matrix for the positive mode and THAP (2,4,6-trihydroxyacetophenone) as the matrix for the negative mode with a PerSeptive BioSystems Voyager instrument. Mass spectra are displayed as detector counts versus calculated mass-to-charge ratio (m/z), while peak areas were determined by summing the counts for each peak in the profiles, and normalized to the percentage of total area from all glycans. From the observed mass of a particular glycan and the knowledge of the glycans produced by CHO cells (Spellman et al. 1989), the glycan structures were assigned (see Figure 5). For each glycan structure, the numbers and types of terminal saccharides (i.e. sialic acid, Gal, or N-acetylglucosamine) can be simply counted by inspection. For each glycan structure, those numbers of terminal saccharides were multiplied by the fractional abundance of that glycan and the resulting values for each terminal type were then summed. This provides an overall count of "terminal saccharide residues per glycan" in the glycan population. Conversion to "moles terminal saccharide per mole of lernercept polypeptide" requires multiplication by the number of glycan sites under consideration, e.g. three for acidic glycan data, i.e. glycans of one arm of the receptor. These calculations were accomplished with the simplifying assumption that all three receptor sites are 100% occupied by acidic glycans. We have previously shown that these numbers are a good approximation of moles of terminal sugar per mole of glycan since we observed good agreement between this method and direct methods (R > 0.9) of measuring the terminal sugars (Papac DI, Briggs JB, Chin E, Nayak N, Lerner L and Jones AJS, unpublished data).
Trypsin digestion
For some samples of recovered lenercept, the protein was digested with trypsin to permit an analysis of the glycans present on specific sites in the molecule. The predicted tryptic fragments are numbered sequentially from the N-terminus. Lenercept samples collected from the reversed-phase column were concentrated under vacuum to remove the acetonitrile and reduced with 10 mM dithiothreitol (DTT) in buffer (0.36 M Tris, 3.2 mM ethylenediaminetetraacetic acid, pH 8.6) at 37°C for 1 h. The reduced samples were carboxymethylated by the addition of 35 mM iodoacetic acid (IAA) and incubated in the dark at ambient temperature for 20 min. The alkylation reaction was quenched by the addition of 50 mM DTT. The reaction mixture was applied to a ProSorb cartridge (PE/ABI) containing a methanol-activated PVDF membrane. The membrane was removed and placed into a small sialylated polypropylene tube. It was washed with water, blocked with 200 µL of 1% PVP360 for 90 min, and rinsed with water. The membrane was then incubated overnight at 37°C with trypsin (Modified trypsin from Promega) in 75 µL 0.1 M Tris–HCl, pH 8.0, containing 10% (v/v) acetonitrile (to ensure elution of peptides from the membrane after cleavage by trypsin). Trypsin was present at 1/20 of the mass (determined by the peak area of the reversed-phase peak during the recovery step) of lenercept. The tryptic peptides were separated on a Hewlett-Packard 1090M HPLC system equipped with a diode array detector. The digests were loaded onto a Zorbax 300SB, C8, 250 mm x 2 mm HPLC column maintained at 50°C. The column was equilibrated with 100% solvent A [0.05% aqueous trifluoroacetic acid (TFA)] at a flow rate of 0.2 mL/min. After sample injection, the column was held at 100% A for 3 min then the peptides were eluted with linear gradients to 15% solvent B (0.05% TFA in acetonitrile) in 30 min, to 29% solvent B in another 40 min, to 35% solvent B in the following 30 min, then to 60% solvent B at 108 min after the beginning of the run. The column was re-equilibrated in 100% solvent A for 30 min. The peptides were detected at 214 nm. The lenercept glycopeptides were preparatively collected and stored at –20°C for further mass spectral analysis. The glycan distributions on collected peptides were analyzed either directly, using THAP as matrix for glycopeptide analysis, or as glycans after PNGase F-mediated release, as described earlier for the intact protein samples.
Cynomolgus monkey PKs study
Female cynomolgus monkeys (2.7–5.0 kg) were quarantined for at least 5 weeks before study and were housed in cages with water and a certified commercial laboratory diet available ad libitum. Diet was supplemented with fresh produce. Lenercept was given as an intravenous bolus over a 10- to 30-s period into a saphenous vein, and the dosing catheter was flushed with approximately 3 mL of saline to ensure complete dose delivery. Blood samples (approximately 1.5 mL) were collected at 5 and 30 min, 1, 2, 4, 6, 8, 12, 24, and 36 h and on days 3, 4, 6, 8, 10, 12, 15, and 18 after dosing. The plasma was frozen for subsequent analysis. Plasma concentrations of lenercept were determined with an immuno-binding assay with a lower limit of quantification equal to 0.6 ng/mL. An overview of the analytical procedures used to compile the carbohydrate composition of lenercept is shown in Figure 2.
Data analysis
Dosing solutions were confirmed using an immunoassay. Plasma concentration versus time data were normalized to a 5 mg/kg dose (based on actual dose concentrations) and analyzed using noncompartmental models. Measured plasma concentrations decreased after day 10 due to a host antibody response (Richter et al. 1999. Therefore, data after day 10 were omitted from analysis of the "AUC" (area under the plasma concentration versus time curve) and elimination half-life data. Truncated AUC was determined by the linear trapezoidal method from 5 min to day 10. Elimination half-life was estimated from day 3 to day 10 by a linear regression program (Statistical Consultants Inc. 1986). Peak concentration (Cmax) and time of peak concentration (Tmax) were the observed values. PK parameters were compared among groups using analysis of variance (ANOVA 1992).
|
Conflict of interest statement |
|---|
|
TopAbstractIntroductionResults and discussionConclusionsMaterials and methodsConflict of interest statementReferences |
|---|
None declared.
|
Footnotes |
|---|
2 Present address: Myriad Pharmacueticals, Inc., Salt Lake City, UT 84108
3 Present address: Amgen Inc., One Amgen Center Drive, Thousand Oaks, CA 91320-1799
|
Abbreviations |
|---|
AUC, area under the curve; CHO, Chinese hamster ovary; DHB, 2,5-dihydroxybenzoic acid; DTT, dithiothreitol; Fuc, fucose; Gal, galactose; GlcNAc, N-acetylglucosamine; IAA, iodoacetic acid; IgG, immunoglobulin G; MALDI-TOF, matrix-assisted laser desorption ionization-time of flight; Man, mannose; MS, mass spectrometry; NeuAc, N-acetyl neuraminic acid; PK, pharmacokinetic; PNGase F, peptide N-glycosidase F; PVDF, polyvinylidenedifluoride; RP-HPLC, reversed-phase high-performance liquid chromatography; TFA, trifluoroacetic acid; THAP, 2,4,6-trihydroxyacetophenone; TNF, tumor necrosis factor.
|
References |
|---|
|
TopAbstractIntroductionResults and discussionConclusionsMaterials and methodsConflict of interest statementReferences |
|---|
Abraham E, Glauser MP, Butler T, Garbino J, Gelmont D, Laterre PF, Kudsk K. p55 tumor necrosis factor receptor fusion protein in the treatment of patients with severe sepsis and septic shock, a randomized controlled multicenter trial. J Am Med Assoc (1997) 277:1531–1538.[Abstract]
Achord DT, Brot FE, Sly WS. Inhibition of the rat clearance system for agalacto-orosomucoid by yeast mannans and by mannose. Biochem Biophys Res Commun (1977) 77:409–415.[CrossRef][ISI][Medline]
ANOVA. Statview v. 4.1 Abacus Concepts. (1992) Berkeley. CA.
Anumula KR. Rapid quantitative determination of sialic acids in glycoproteins by high-performance liquid chromatography with sensitive fluorescence detection. Anal Biochem (1995) 230:24–30.[CrossRef][ISI][Medline]
Ashkenazi A, Marsters SA, Capon DJ, Chamow SM, Figari IS, Pennica D, Goeddel DV, Palladino MA, Smith D. H. Protection against endotoxin shock by a TNF receptor immunoadhesin. Proc Natl Acad Sci USA (1991) 88:10535–10539.
Ashwell G, Harford J. Carbohydrate-specific receptors of the liver. Annu Rev Biochem (1982) 51:531–554.[CrossRef][ISI][Medline]
Ashwell G, Morell AG. The role of surface carbohydrates in the hepatic recognition of circulating glycoproteins. Adv Enzymol (1974) 41:99–128.[CrossRef]
Banner DW, D'Arcy A, Janes W, Gentz R, Schoenfeld H-J, Broger C, Loetscher H, Lesslauer W. Crystal structure of the soluble human 55 kd TNF receptor-human TNFß complex: implications for TNF receptor activation. Cell (1993) 73:431–445.[CrossRef][ISI][Medline]
Battersby JE, Vanderlaan M, Jones AJS. Purification and quantitation of tumor necrosis factor receptor immunoadhesin using a combination of immunoaffinity and reverse phase chromatography. J Chromatogr B (1999) 728:21–33.[CrossRef]
Chamow SM, Ashkenazi A. Immunoadhesins: principles and applications. Trends. Biotechnol (1996) 14:52–60.
Cumming DA. Glycosylation of recombinant protein therapeutics: control and functional implications. Glycobiology (1991) 1:115–130.
Harvey DJ. Matrix-assisted laser desorption/ionization mass spectrometry of carbohydrates and glycoconjugates. J Am Soc Mass Spectrom (2002) 226:1–35.
Drickamer K. Clearing up glycoprotein hormones. Cell (1991) 67:1029–1032.[CrossRef][ISI][Medline]
Ferrara C, Stuart F, Sondermann P, Brunker P, Umana P. The carbohydrate at Fc
RIIIa Asn-162. An element required for high affinity binding to non-fucosylated IgG. J Biol Chem (2006) 281:5032–5036.
Huang L, Biolsi S, Bales K, Kuchibhotla U. Impact of variable domain glycosylation on antibody clearance: an LC/MS characterization. Anal Biochem (2006) 349:197–207.[CrossRef][ISI][Medline]
Jenkins N. Role of physiology in the determination of protein heterogeneity. Curr Opin Biotechnol (1996) 7:205–209.[CrossRef][ISI][Medline]
Keck R, Nayak N, Lerner L, Raju S, Ma S, Schreitmueller T, Chamow S, Moorhouse K, Kotts C, Jones A.J.S. Characterization of a complex glycoprotein whose variable metabolic clearance in humans is dependent on terminal N-acetylglucosamine content. (2007).
Kéry V, Krepinsky JJF, Warren CD, Capek P, Stahl PD. Ligand recognition by purified human mannose receptor. Arch Biochem Biophys (1992) 298:49–55.[CrossRef][ISI][Medline]
Kim J-K, Tsen M-F, Ghetie V, Ward ES. Identifying amino acid residues that influence plasma clearance of murine IgG1 fragments by site-directed mutagenesis. Eur J Immunol (1994) 24:542–548.[ISI][Medline]
Kronman C, Vellan B, Marcus D, Ordentlich A, Reuveny S, Shafferman A. Involvement of oligomerization, N-glycosylation and sialylation in the clearance of cholinesterases from the circulation. Biochem J. (1995) 311:959–967.[ISI][Medline]
Meier W, Gill A, Rogge M, Dabora R, Majeau GR, Oleson FB, Jones WE, Frazier D, Miatkowski K, Hochman PS. Immunomodulation by LFA3TIP, an LFA3/IgG1 fusion protein: cell line dependent glycosylation effects on pharmacokinetics and pharmacodynamic markers. Therap Immunol (1995) 2:159–171.[Medline]
Morell AG, Gregoriadis G, Scheinberg IH, Hickman J, Ashwell G. The role of sialic acid in determining the survival of glycoproteins in the circulation. J Biol Chem (1971) 246:1461–1467.
Morell AG, Irvine RA, Sternlieb I, Scheinberg IH, Ashwell G. Physical and chemical studies on ceruloplasmin V. Metabolic studies on sialic acid-free ceruloplasmin in vivo. J Biol Chem (1968) 243:155–159.
Mulders JWM, Derksen M, Swolfs A, Maris F. Prediction of the in vivo biological activity of human recombinant follicle stimulating hormone using quantitative isoelectric focusing. Biologicals (1997) 25:269–281.[CrossRef][ISI][Medline]
Noormann F, Barrett-Bergshoff MM, Rijken DC. Role of carbohydrate and protein in the binding of tissue-type plasminogen activator to the human mannose receptor. Eur J Biochem (1998) 251:107–113.[ISI][Medline]
Papac DI, Briggs JB, Chin EH, Jones AJS. A high throughput microscale method to release N-linked oligosaccharides from glycoproteins for matrix assisted laser desorption ionization time of flight mass spectrometry. Glycobiology (1998) 8:445–454.
Papac DI, Wong A, Jones AJS. Analysis of acidic oligosaccharides and glycopeptides by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry. Anal Chem (1996) 68:3215–3223.[Medline]
Raghavan M, Bjorkman PJ. Fc receptors and their interactions with immunoglobulins. Annu Rev Cell Dev Biol (1996) 12:181–220.[CrossRef][ISI][Medline]
Rice KG, Lee YC. Oligosaccharide valency and conformation in determining binding to the asialoglycoprotein receptor of rat hepatocytes. Adv Enzymol (1993) 66:41–83.[CrossRef]
Richter WF, Gallati H, Schiller CD. Animal pharmacokinetics of the tumor necrosis factor receptor-immunoglobulin fusion protein lenercept and their extrapolation to humans. Drug Metab Dispos (1999) 27:21–25.