Study of Glycation and Advanced Glycation on a Humanized Monoclonal Antibody

In the presented work, we reviewed the protein glycation effect on recombinant monoclonal antibodies (MAbs), and demonstrated an in vitro glycation model on rhuMAb A was a successful tool to evaluate site-specific glycation. We also implemented this model to study MAb glycation in three aspects: the formation of glycation, the change of glycation adducts under thermal stress, and the control of glycation adducts stability by formulation compositions. A protein characterization study demonstrated that the in vitro forced glycation model generated seven glycated amino acid sites on rhuMAb A. The lysine residue K49 was the preferably site, while others sites were at various low levels of glycation. Molecular dynamics (MD) analysis suggested that the high abundance of lysine 49 glycation observed on rhMAb A may be assisted by a strong electrondonating environment created by three aspartates, with D30, D31, and D105 near K49. In the thermal stress stability study, the glycation adduct hydrolysis reaction was more pronounced than the AGE formation. It was found that glycation adducts hydrolysis continued for four weeks, but AGEs formation plateaued after one week. The overall combination of both reactions caused a loss of glycation with a first order reaction rate. The structure analysis of final degradation products agreed with kinetic observation that reverse of glycation adducts was the main degradation pathway of glycated rhuMAb A at 40°C in pH 6.5 phosphate formulation buffer. Reverse of glycation adducts on K49 could be catalyst by its adjacent aspartic acids. Influencing parameters of glycation adduct degradation were also evaluated. Three formulation composition factors: pH, buffer, and oxidation control were compared. The study demonstrated the pH value was the key parameter that controls degradation pathways. pH also determined the rate constant of hydrolysis of glycated rhuMAb A. Buffer species and oxidation levels did not affect on the glycation adduct stability under the studied conditions. In addition, the effect of rhuMAb A’s initial glycation level was also evaluated. The study demonstrated that higher protein glycation levels slow down the overall hydrolysis rate of glycation adduct, presumably by providing a complex electron transferring system on protein.

xiii

LIST OF TABLES
Chapter 1

Introduction
Glycation is a known reaction which occurs between reducing sugars and amine groups on proteins, amino acids, and nucleotides. Intensive in vivo studies in humans have shown that protein glycation is implicated in the pathogenesis of multiple chronic diseases. The most common reducing sugar found in vivo is glucose, which reacts with N-terminal amino acids or with free amino groups on lysine and arginine residues. This initial reaction between glucose and amine group is often referred as a glycation adduct. Further modification of glycation adducts result in a variety of different products known as advanced glycation endproducts (AGEs) (Maillard, L.C. 1912). This leads to irreversible modifications on the proteins (Sell, D. R. and Monnier, V. M. 1989), which can subsequently cause tissue damage (Valcourt, U. et al. 2007). Protein glycation level has been known as a biomarker for diseases.
For example, individuals with diabetes have higher serum glucose concentration, are thought to be more prone to glycation than those in healthy persons. Therefore, the extent of glycation of hemoglobin in vivo is a distinctive marker for diabetes (Beisswenger, P.J. et al. 2001).
Glycation occurs in vitro as well. In the past several decades, an increasing number of recombinant humanized monoclonal antibodies (MAbs) have been developed as therapeutic agents (Figure1-1), due to their high specificity and affinity to therapeutic antigen targets (Figure 1-2). As a manufactured therapeutic protein, a MAb is subjected to many chemical modifications, including glycation throughout the manufacturing process. Chemical modifications may result in structural changes, ultimately leading to potential biological function changes such as efficacy, safety and clearance rate. Thus, understanding glycation formation, degradation and how to control these processes on MAbs is important. The following sections summarize reported studies about glycation modification on MAbs throughout their life cycle including during cell culture, storage, administration in clinics and final circulation in the body.

Mechanism of Protein Glycation and Advanced Glycation Reactions
Glycation, originally described by Maillard (Maillard, L.C. 1912), refers to the nonenzymatic reaction between reducing sugars and proteins. This process involves a series of steps starting with the reaction between the reducing carbonyl group of a carbohydrate and an amino group present on proteins, usually at the epsilon amine on lysine residues or at the N-terminus, with initial formation of an aldimine (which is known as a Schiff base) (Figure 1-3). This unstable aldimine linkage can then rearrange to form a more stable ketoamine (which is known as an Amadori product) (Bucola, R. and Cerami, A. 1992).
Although the ketoamine linkage is considerably more stable than the aldimine linkage, it has been shown to be slowly broken under relatively mild conditions Miller, A.K. et al. 2011). There are also deglycating enzymes that limit glycation of intracellular proteins (Van Schaftingen, E. et al. 2007). The nonenzymatic reverse hydrolysis reaction is straightforward, releasing glucose and its C-2 epimer mannose, without much consequence (Zhang, Q. et al. 2009). The enzymatic reverse reaction involves phosphorylation of the Amadori product, followed by release of 3-deoxyglucosone.
Even though reverse of glycation reaction is relatively straightforward, the forward reactions are more complex, generating a wide range of reactive carbonyl and dicarbonyl compounds. These reactive intermediate carbonyl compounds may react with proteins to form stable, irreversible adducts and cross-links, known as AGEs ( Figure 1-3). Oxidation and further rearrangement of the ketoamine and aldimine products can also generate AGEs.
Studies have shown that the extent of glycation of a protein depends on the rate of formation of the Amadori product and the rate of reversal or conversion to other products (Bucola, R. and Cerami, A. 1992;Cho, S.J. et al. 2007).
The study of AGEs is one of the most promising areas of research in chronic inflammation diseases today. Although the Maillard reaction, AGEs formation, has been known since 1912, it is only in the last 20 years that important works have done to elaborate the mechanism of AGEs formation. In 1988, Namiki reported the reactivity of different reducing sugars leading to AGEs formation. It was shown that compared with glucoses, aldehydes with short carbon chains (2-3 carbons) react with a model peptide, beta-alanine, faster and form AGEs more rapidly. Thus degradation of glucose to smaller reducing sugar may accelerate AGEs formation (Table 1-1, Namiki, M. 1988 Amadori product dehydration, and oxidation formed AGEs. All three pathways are thought to be trigged by oxidation (Jakus, V. et al. 2004).
Inhibitors, that prevent AGEs formation, have also been investigated by researchers. An AGEs-inhibitor is a molecule that can interrupt the covalent crosslinking of proteins and peptides by sugars or sugar derived oxidation products.
For example, it may contain a nucleophilic amino group that reacts carbonyl group on reducing sugars to compete with epsilon amine groups on lysine residues and other primary amines. Carnosine, a dipeptide, β-alanyl-L-histidine, found in long-lived tissues, such as brain, at concentrations up to 20 mM in humans, has been regarded as has been discovered whereby carnosine can react with low-molecular-weight compounds that bear carbonyl groups (aldehydes and ketones). Another example can be found in the Wolf pathway. Since the crucial step of AGEs crosslinking depends on the presence of transition metals for the formation of glycoxidation products and radicals, a family of metal chelators and radical scavengers (especially superoxide dismutase mimetics) could be regarded as AGEs-inhibitors. In addition, several natural and synthetic compounds (Figure 1-4) have been shown to be inhibitors of AGEs-formation in vitro and in vivo (Cameron, N.E. and Cotter, M.A. 1993).
In short, the chemical processes and pathways that ultimately lead to AGEs formation step by step have yet to be fully clarified. As our knowledge of AGEs expands, it is becoming apparent there are many unknown AGEs, while those that have been characterized are both complex and heterogeneous, making it an important area of research.

Glycation and AGEs in Aging and Diseases
There is an increasing amount of evidence that AGEs may play a role in Alzheimer's disease (AD), diabetes and aging, as well as other chronic disease complications. The presence of AGEs are closely related to hyperglycaemia and the patho-biochemistry of AGEs could explain many of the observed changes in diabetes related complications.
From biochemical, cell based, and animal studies, it has been demonstrated that AGEs accumulate as a function of the level of chronic hyperglycemia. It is thought that AGEs accumulation in vivo causes dysfunctional changes in extracellular matrix, abnormal receptor-mediated production of cytokines, and altered function of intracellular proteins.
In the early 80's, Monnier and Cerami (Monnier, V. and Cerami, A. 1981), the pioneers of the nonenzymatic glycation theory of aging, proposed that the AGEsmediated crosslinking of long-lived proteins contributes to the AGEs-related decline in the function of cells and tissues in normal aging (Table 1-2) (Brownlee, M. et al. 1992). Recent progress in the study of chronic diseases has confirmed that AGEs play a significant role in the evolution of vascular complications in aging, diabetes and renal failure. Pathological results showed that chronic inflammation may be triggered by endogenous trans-membrane receptor of AGEs ( RAGEs) ( Basta, G. et al. 2002).
Among the many factors proposed to be involved in the etiology or progression of AD, AGEs are a relatively new and interesting approach to unravel the mysteries of the etio-pathogenesis of this multi-factorial disease. Recent studies suggest that AGEs are active promoters of the progression of AD, not only simply a secondary epiphenomenon (Shaw, S.M. and Crabbe, M.J. 1994). However, one has to be careful not to overestimate AGEs' role in the disease process as long as only circumstantial evidence is presented. Therefore, the specific role of AGEs and which particular AGEs are involved in the disease complications needs to be established. The possibility of effective therapeutic intervention enhances the importance of detecting AGEs. Advancements in measuring extent and type of AGEs using reliable methods will help to determine the role that AGEs have in the pathogenesis of many diseases.

Antibodies (MAbs)
Therapeutic MAbs, like other therapeutic proteins, are enormously complex drugs. They are typically produced in mammalian tissue culture cells through recombinant DNA technology (Goetze, A.M. et al. 2010). MAbs are a fast growing therapeutic modality in the pharmaceutical industry due to their higher specificity, lower toxicity, and greater in vivo predictability than small molecule pharmaceuticals (Banks, D.D. et al. 2009). However, unlike small molecule drugs, protein drugs are generally more susceptible to various conformational instabilities which can potentially influence immunogenicity and biological activity (Banks, D.D. et al. 2009).
As of May 2008, there were 15 recombinant MAbs and fusion proteins approved in the US for human therapeutic use (Table 1-3), with the majority approval for oncology and immunological indications (Zider, A. and Drakeman, D.L. 2010 galactose. Cells are separated from the cell culture fluid using centrifugation followed by filtration. The harvested cell culture fluid is purified across multiple column chromatography steps, typically in series which may include for example: a protein A affinity column, an ion exchange column, and a hydrophobic interaction column and/ or one more additional ion exchange column or other chromatography column(s).
Virus inactivation and filtration operations are also designed and executed during the downstream processing. The purified MAb product is eventually formulated and filled into either drug product vials or pre-filled syringes. Subsequently, the DP is stored under recommended storage conditions before being shipped for clinical use.
Health authority approval for these protein drugs requires chemical and biological analyses designed to determine product identity, purity, and potencywhich is not a trivial task for an approximately 1,300-amino-acid, 150-kDa heterodimer structure. The resulting analytical description encompasses the protein sequence, size, and charge as well as the degree of heterogeneity associated with each parameter.
To complete the physiochemical and biological characterization, forms of variants need to be identified and their impact on potency needs to be determined. Typical recombinant protein modifications include asparagine deamidation, aspartate isomerization, methionine oxidation, protein aggregation and fragmentation, and protein glycation.
Control of micro-heterogeneity within defined analytical specifications has been used in quality control laboratories to guarantee consistent product quality during current good manufacturing process (cGMP) (Goetze, A.M. et al. 2010). Defining and adhering to the appropriate range of specifications ensures drug safety and efficacy to patients.

Therapeutic Monoclonal Antibodies
Recombinantly produced MAbs are exposed to reducing sugars during mammalian cell culture after they are secreted into the cell culture medium that contains sugars. Sugars are present in medium as energy and carbon sources for the cells, and as a way to modulate glycosylation. For example, galactose may be added to the cell culture medium to control the level of galactosylation (Clark, K.J. et al. 2005).
In addition, standard process parameters of cell culture such as temperature (~37°C physiological condition), pH (~ pH 7.4 physiological conditions), time (10-14 days), and reactant (amine and sugar) concentrations, affect both the glycation reaction kinetics and extent.
Under typical batch-feed manufacturing conditions for MAbs, the cell culture medium is maintained at approximately 10,000-fold molar excess reducing sugar per mole protein product for the cell culture duration, indicating that the bimolecular glycation reaction kinetics would be pseudo first-order for the protein (Quan, C . et al. 2008). However, each MAb molecule has more than 100 primary amines, and the cell culture medium also contains significant levels of free amino acids and other compounds containing potentially reactive amines, thereby reducing the sugar molar excess for the glycation reaction. The protein glycation reaction extent is thought be governed mostly by the reactivity of accessible amino groups, although different sugars have different reactivity. For example, galactose is 5-fold more reactive than glucose (Bunn, H.F. and Higgins, P.J. 1981). In the artificial environment with complex chemical components present in cell culture medium and feeding conditions, In the past several decades, the primary sites of glycation have been analyzed for a number of proteins under a variety of conditions. On some proteins, glycation is evenly distributed, while on other proteins, the N-terminus (Holmquist, W.R. and Schroeder, W.A. 1964) and specific lysine residues , Gadgil, H.S. et al. 2010, Miller, A.K. et al. 2011 can have a greater degree of modification.
Glycation distribution difference between studied protein molecules indicates that sequence or structure may determine the specificity of glycation reactions. In the past, it was thought that the apparent pKa and exposure of the amine attachment site determines the degree of Schiff base formation. In addition, the proximity of amino acids capable of proton abstraction helps catalyze the formation of the ketoamine, or Amadori product (Venkatraman, J. et al. 2001. Buffer components that can bind near the reaction site and abstract protons from sugars, such as phosphate, can also serve as a catalyst to increase the rate of formation of stable products . Thus, the protein amino acid sequence and reaction conditions can all affect the site specificity of glycation reactions. A MAb contains many primary amine groups. In addition to four N-terminal amino acids, most MAbs have approximately 80 Lys residues, giving MAbs many potential sites for glycation (Lapolla, A. et al. 2000).
Historical data has shown that a typical MAb can be modified by glucose that results in very low abundances of glycation, with sites distributed over different amino acid residues. Often, the abundance of glycation product can be below the detection limit of many analytical methods. For example, a study by Quan and coworkers showed that MAbs produced in a 10 day glucose containing cell culture were approximately 18% glycated overall including both light chains, and both heavy chains. The quantification of this MAb's overall glycation is based on percentage of glycated MAb separated from total MAb by boronate affinity chromatography (Quan, C. et al. 2008). Within this 18 % population of glycated MAbs, the glycation modification appeared as mono-glycation on 8 lysine residues on half of the molecule with low frequency (1-12%) at each of the eight lysine residues. (Quan, C. et al. 2008).
These glycated species were separated from nonglycated material based on an approximately 0.05-pI charge unit difference and appeared as an acidic variant form in charge-based assays (Quan, C. et al. 2008). Because of the disperse locations and low abundance of glycation, chromatography methods did not produce good charge-or hydrophobicity-based separation to detect glycation modifications, therefore different glycation modification could not be detected. However, mass spectrometry and boronate affinity chromatography can be used to detect glycation. Both methods coalesce multiple glycation adducts into a single-unit modification, allowing separation between glycated and unglycated protein species, and specific detection on large and complex molecules such as MAbs , Quan, C. et al. 2008).
In recent years, researchers have reported site specific glycation on MAbs (Miller, A.K. et al. 2011, Gadgil, H.S. et al. 2010 conditions. Intensive studies have been carried out to study the effect of glucose to protein ratio and glucose feed strategies on MAb glycation in the hope of optimizing the cell culture condition to control the total glycation level of the MAb (Yuk, I.H. et al. 2011). This finding contrasts with historical data that have suggested that glycation sites are typically located randomly at all accessible lysine residues distributed over the entire molecule ).
To date, only a few publications had proposed a mechanism of site specific glycation on MAbs. One theory is that site specific glyaiton is due to proton transfer with assistance from nearby amino acids. Sometimes,a histidine residue is hypothetically involved ( Quan, C. et al. 2008). In other cases, a aspartic acid residue is thought to be involved . However, the exact mechanism of site specific glycation on MAbs is till yet to be determined. This uncertainty enhances the interest of study in the formation and control of glycation in MAbs.

Glycation in MAbs Drug Product Stability and Administration
Protein glycations may occur during formulation, packaging, long term storage, or clinical administration steps, where sugars are commonly used as excipients in liquid or lyophilized formulations of therapeutic proteins (Chang, B.S. and Hershenson, S. 2002). This modification of glycation is one of the primary reasons that formulation scientists tend to avoid using reducing sugars (glucose, lactose, fructose, maltose) in formulations.
Even for non-reducing disaccharides, such as sucrose, there is still a potential to generate reducing sugars by in situ hydrolysis, resulted in subsequent glycation modification. For example, sucrose can undergo hydrolysis -e.g. at low pH -to glucose and fructose. Smales, et al. demonstrated this hydrolysis for sucrose-based formulations undergoing viral inactivation at elevated temperatures (Smales, C.M. et al. 2000 and. Similarly, glycation in sucrose-based formulations has been observed during storage studies (Fischer, S. et al. 2008, Gadgil, H.S. et al. 2007, although, at elevated temperatures and acidic pH. These extreme conditions are not within typical pharmaceutical formulation conditions, therefore the risk of protein glycation induced by sucrose hydrolysis in protein therapeutic is very low during the storage shelf life. Finally, for delivering a drug to patients, sugars are often used in IV (intravenous) solutions to prevent precipitation of proteins, attain isotonicity of drug and provide an energy source for patients. Since some of the sugars commonly used for these purposes are reducing sugars (5% dextrose), the spontaneous glycation of were 70% glycated after a 14 day incubation with 5% dextrose at ambient temperature. (Fischer, S. et al. 2008 been reported yet (Vrdoljak, A. et al. 2004). Nevertheless, it is important to understand the glycation effect on MAbs during IV delivery, as drug compatibility with administration route is the final controllable step for ensuring the MAb's safety and efficacy to patient.

Impact of MAb Glycation on Its Biological Activities
Glycation of MAbs is of special interest due to the possible influence on the functionality of antigen binding, Fc effector function, and neonatal Fc receptor (FcRN) binding. Quan, et al. 2008, studied the effect of glycation on in vitro potency of a MAb,

Antigen Binding
where the second most susceptible site to glycation was located within the second complementarity-determining region (CDR) on the heavy chain (HC CDR2 K65). The glycated and unglycated MAb were separated by and fraction collected from boronate affinity chromatography. In vitro biological activity were assayed for each fraction.
Even though there was glycation at an active binding site, there was little observable effect on bioactivity for the glycated boronate fraction, compared to unglycated. The glycated MAb showed similar binding activity to a biologically functional ligand.
Although glycation at the 10% level at this one residue (K65) within one CDR was insufficient to affect potency, additional increase in glycation levels may ultimately be expected to interfere with the ligand-binding ability of the MAb. Also, the effect of glycation modification impact depends on the location of modified lysine, e.g. whether it is directly involved in antigen binding (Quan, C. et al. 2008).
Researchers are also interested in glycation impact on the stability of antibodyantigen complex. For example, Kennedy and coworkers reported antibody-antigen complex study based on surface plasmon resonance, in 1994. Although glycated MAbs had similar association constant (k ass ) with their antigen, compared with unglycated MAbs, the glycated MAbs have significantly higher dissociation constant (k dis ) with their antigen, indicating that the glycated MAbs dissociate with their antigen faster. Therefore, glycation lowered the affinity ( Kd= k dis /k ass ) of MAbs to their antigens ( Kennedy, D.M. et al. 1994). Because of the different approaches and detection techniques used, the results from different studies were not coherent, especially regarding the influence of glycation on the ability of MAbs to bind antigens and induce the complement cascade.

Fc effector Function, and FcRN Binding
Forced glycation studies using high pH conditions have shown that Fc functions of human antibodies, such as protein A binding and complement activation, can be impaired by glycation (Dolhofer-Bliesener, R. and Gerbitz, K.D. 1990).
However, under physiologic conditions, binding activity of the Fc is less sensitive, presumably due to lower levels of glycation. Goetze, et al. 2012 reported glycation of MAbs can result from incubation with a reducing sugar during circulation in vivo. The contributions to total glycation from the labile aldimine and more stable ketoamine forms were separately monitored via the use of selective sodium borohydride reduction. Since the Fab protein sequences are versatile, the Fab susceptibility to glycation, differs between antibodies. A number of studies have focused primarily on the conserved Fc regions (Goetze, A.M. et al. 2012).
An attempt was made to model physiological glycation rates observed using a simple in vitro system. The impact of Fc glycation on function was also explored using surrogate binding assays that probed the interaction between MAbs and FcγRIIIa, neonatal Fc receptor (FcRn) and immobilized protein A (Goetze, A.M. et al. 2012).
Upon injection of a recombinantly produced human therapeutic IgG into humans, changes in the glycation levels could be observed as a function of circulation time (Goetze, A.M. et al. 2012

Conclusions
Protein glycation reactions can occur to recombinant MAbs during many manufacturing process steps such as cell culture and storage, as well as during clinical administration to a patient and circulation in human body. Cell culture process is the initial place generating glycated recombinant MAbs. Clinical administration is the final step for controlling MAb glycation before introduction into patients.
Glycation modification results in structural heterogeneity in the MAbs. It changes MAbs charge, target binding capacities, and degradation rates (Zhang, Q. et al. 2009). However, the site specific glycation mechanism and effect of glycation impact on efficacy, safety and stability of MAbs are not well understood.
Development of sensitive and specific methods to monitor advanced glycation products is important to fully understand and monitor glycation during the manufacturing process and clinical administration of MAbs. Improved and more specific assays are also needed to assess the impact of novel protection strategies aimed at designing glycation inhibitors for therapeutic applications (Zhang, Q. et al. 2009).
In the quality-by-design (QbD) paradigm for manufacturing MAbs, the decision to monitor and control glycation for a given MAb would be a thorough evaluation process based on the impact of manufacturing parameters on MAb glycation modification and the effect of glycation on MAb's biological function for safety and efficacy (Quan, C. et al. 2008).

Radical Formation (Intensity of ESR Spectra)
Not Determined 2109 ++++ In the past decades, there has been an increasing number of recombinant humanized monoclonal antibodies (MAbs) developed as therapeutics due to their high specificity and affinity to therapeutic antigen targets. As manufactured proteins, MAbs have the potential to be chemically modified during manufacturing process.
MAbs are typically manufactured via a 10-14 day cell culture process. In this process a MAb is produced in Chinese Hamster Ovarian cells (CHO) then secreted outside the cell, into the culture medium, which contains glucoses as the cell's nutritional carbon source. This protein-glucose contact results in MAbs glycation (Quan, C. et al. 2008. In most MAbs, the glycation level is <15% and widely distributed over the entire MAb molecule. In recent years, researchers have also reported site specific glycation on MAbs (Miller, A.K. et al. 2011, Gadgil, H.S. et al. 2010). Zhang's group, in particular, has published a significant amount of work describing a site specific glycation on lysine K49 of a rhuMAb A molecule. The glycation level on this particular lysine K49 residue can be as high as 50% under typical cell culture conditions. Intensive studies have been carried out to study how glucose to protein ratio, and glucose feed strategies can be altered to optimize the cell culture condition. The ultimate goal is control the total glycation level of rhuMAb A (Yuk, I.H. et al. 2011).
In the artificial environment with complex chemical components present in cell culture media and feeding conditions, the glycation kinetics and extent on MAbs can vary with cell culture conditions Yuk, I.H. et al. 2011).
As the protein sequence differs between MAb molecules, the degree of glycation would also be different, even under similar conditions. The uncertainty of controlling MAbs glycation enhances the interest of study in the mechanism site-specific of glycation in MAbs.
To date, only a few publications have proposed a mechanism of site-specific glycation on MAbs. The proposed glycation mechanism suggests that nearby amino acids may facilitate proton transfer. Sometimes, a histidine residue is thought to be involved (Quan, C. et al. 2008), in other cases an aspartate residue may be involved .
With a great amount of mutation work on rhuMAb A, Zhang proposed that an adjacent aspartate residue D31 is the local catalyst for site specific glycation in rhuMAb A. D31 is spatially located 11 Å to lysine residue K49 based on the crystal structure of rhuMAb A. The mutation of this D31 to other amino acids may reduced the glycation level of K49. However, in Zhang's study, there were several cases where K49 lysine was still glycated up to 20% after replacing D31 with different amino acids.
In order to further examine the mechanism of site specific glycaiton on K49 of this rhuMAb A, we employed an in vitro forced glycation model to generate close to 100% glycated rhuMAb A. We also performed comprehensive analytical characterization on the glycated rhuMAb A produced using this model system. Then, we applied molecular dynamics simulations to determine the local environment of the primary modified lysine K49 by mimicking the rhuMAb A in aqueous solution.

II. 1. Materials and Reagents
A regular lot of rhuMAb A (

II.2. In Vitro Glycation of rhuMAb A
The rhuMAb A in PBS buffer, pH 7.4, was further glycated in vitro. rhuMAb A was buffer exchanged into glycation buffer with dialysis prepared in 15 ml cassette with 10kD membrane. The final in vitro glycation solution contains 50mg/ml rhuMAb A with 1200 mM D-glucose, in 20 mM sodium phosphate buffer, pH 7.4. The in vitro glycation solution was kept in an incubator at 37 °C for 24 hours.
The in vitro glycation solution was immediately buffer exchanged using a 15 ml dialysis cassette with 10kD membrane into phosphate buffer, pH 6.5. The samples were then stored at -70°C until analysis.

II.3. Boronate Affinity HPLC
Early stage glycation adducts often can be identified by a boronate affinity HPLC method to determine the overall glycation level on a whole protein, using a TSK gel boronate 5PW column 7.5x75mm. The mobile phase A was 50mM EPPS, 10mM Tris, 200mM NaCl, at pH 8.7, and mobile phase B, 500mM sorbitol added mobile phase A. The mobile phase flow rate 1 mL/min and the column temperature was maintained at 40°C with a UV and fluorescence detector. The gradient for HPLC elution required optimization for rhuMAb A. )

II. 4 Enzymatic Digestion of rhuMAb A
Several enzymatic treatments were performed on rhuMAb A samples before they were analyzed for reverse phase-HPLC, or mass spectrometric analyses, to determine the amino acid sequence and structural integrity of rhuMAb A.
PNGase F Treatment. PNGase F was used to remove the oligosaccharides in the Fc portion of a MAb. The rhuMAb A samples were buffer exchanged into 50 mM Tris buffer, pH 7.5, using a NAP-5 column. The protein concentration was adjusted to 2.5 mg/mL. PNGase F was added in an enzyme-to-substrate ratio of

II.5. Reverse Phase-HPLC with Electron Spray Ionization/ Mass Spectroscopy (RP-HPLC-ESI/MS)
The molecular masses of the intact, PNGase F-treated, and reduced rhuMAb A samples were determined by using a LC/ESI-MS setup coupled with an Agilent

II.7. Protein Chip Matrix Assisted Laser Desorption Ionization-Time-Of-Flight mass spectrometry (MALDI-TOF)
Isolated tryptic peptides were spotted individually on a SCOUT 384 multiprobe plate with a CHCA matrix and analyzed in the positive mode using a Bruker Ultraflex-I MALDI-TOF-TOF mass spectrometer (Bremen/Leipzig, Germany).
Selected precursor ions were subjected to collision-induced dissociation with argon as the collision gas. LIFT mode was used to analyze all fragment ions in single sweep, and data were processed with Flex Analysis software.

II.8. Molecular Dynamics (MD) Simulations
To compare K49's property with all lysine residues on this rhuMAb A, molecular dynamics simulations were performed to analyze the movements of lysine residues by calculating their motions in aqueous solution. Based on the IgG1 protein framework, a generic model was built which includes common intra-molecule forces for a MAb. rhuMAb A's particular protein amino acid sequence was applied to this model to generate a molecule specific 3D structure in aqueous solution. Lysine's solvent accessible surface area (SASA) was determined by AREAIMOL (CCP4 supported Program), which is based on a "rolling ball" algorithm (Lee, B., and Richards, F.M., 1971). This algorithm uses a sphere of water as a particular radius to 'probe' the surface of the rhuMAb A molecule then calculate the exposure area of a particular lysine.
Furthermore, molecular dynamics was applied to determine the lysine's local environment via a computer simulation of physical movements of atoms and molecules. In a given time, the atoms on this rhuMAb A were allowed to interact with and move in the aqueous solution, resulting a view of the motion of the atoms.
In these simulations we performed explicit water (with 11780 water molecules) molecular dynamics (MD) simulations of a set of FAb fragments and calculated different properties. Focus was placed on lysine residues to identify the properties of lysine residues which might correlate to experimentally observed glycation formation propensities in rhuMAb A. This approach is different from the empirical methods discussed above, because we calculated dynamic properties of the protein that could not be determined from the static structure.
the system to a temperature bath of 300 K using simulation algorithm (Bussi, G., Donalido, D., Parrinello, M., 2007) at pH7.4. Histidines on rhuMAb A were unprotonated. During equilibration, the volume was initially kept constant by coupling the system to a pressure bath at 1.0 atm (Bernedsen, H.J.C., Postma, J.P.M. et al. 1984). Following equilibration, the simulations were kept at constant pressure.
Following equilibration, a set of trajectories with 75ns duration was performed.
Trajectories from MD simulations were then analyzed with various tools available in AMBER program suite.  III. Results

III.1. Boronate Affinity HPLC
We determined total glycation level of rhuMAb A by using boronate affinity chromatography (BAC). Boronate affinity chromatography is based on the formation of a five-member ring structure between a sugar's cis-hydroxl groups and boronate ion's negative charge on the column (Figure 2-1). The affinity to sugar ensures boronate capture the glycated protein under alkaline condition. After loading onto a boronate affinity column, protein samples were eluted with a gradient of a competing sugar, sorbitol. Unglycated proteins were eluted at ~2.4 min, while glycated proteins were detected at ~25 min. The elution conditions used here did not provide resolution of proteins with different sugar to protein ratios. Therefore, the overall glycation level was defined as percent peak area of the retained peak on boronate affinity chromatograms.
From this study result, glycated rhuMAb A from a regular glucose batch feed process (batch feed protein to glucose concentration, 1.5g/L titer: 8g/L glucose feed) was 42.0% based on the integration of the retained peak area, as shown in Figure

III.2. RP-HPLC-ESI/MS
Because BAC cannot determine the glucose to protein ratio, Electron Spray Ionization/ Mass spectrometry (ESI/MS) was applied to determine the approximate numbers of glucose per rhuMAb A on the forced glycated rhuMAb A sample. In order to measure the mass, rhuMAb A was reduced, then separated into

III.3. Tryptic Peptide Map
The next question to be answered was which amino acid sites on the primary sequence level were glycated. Identification of the glycated amino acid sites was attempted by tryptic peptide mapping. Tryptic peptides were generated by enzymatic digestion of rhuMAb A. The digested peptide fragments were separated on RP-HPLC, resulting in a unique peptide elution profile of each protein faction (unglycated and glycated fractions) from BAC. By comparing peptide-map profile of the glycated and unglycated samples, glycated peptide fragments were detected.
MS/MS analysis was performed on the putative glycated peptide fragments to assign the peptide sequence and identify the amino acid site of modification. All the other sites were glycated at very low levels, less than 5% modified. This sitespecific result agrees with published work on non-forced glycation of rhuMAb A by ).

Dynamic (MD) Simulations
What special property does K49 have that allows such a high extent of modification by glycation? Does it happen to be more solvent accessible, which increases the probability of reaction with glucoses? Or is there a special local electrostatic environment?
The solvent accessible surface area (SASA) of each lysine residue was determined using AREAIMOL calculation. The summary of average SASA on each lysine residue is shown in Table 2-2. It is notable that K49 was not the most solvent accessible lysine residue on rhuMAb A.
Also, the local environment of glycated lysine residues, especially the potential aspartate to lysine pairing effect, was evaluated. There were three major results: the average distance between D31 to K49 was 11.8Å, which agrees with Zhang's report. However, D31 to K49 is not the only spatial adjacent (D:K) pair in rhuMAb A. There were five other (D:K) pairs of a distance ~5.5 Å. Interestingly, only two of the five lysine residues were glycated. More importantly, lysine K49 was surrounded by three aspartates. They were light chain D30, D31 and another aspartate on heavy chain, D105. D30 and D31 had similar spatial distance to K49.
However, the D105 was only 5.4 Å to K 49, a much closer spatial distance than D30 or D31 had (Figure 2-6). These three aspartate residues also had different solvent accessible surface areas. As shown in Table 2-3, D30 and D31 had the highest SASA value among all aspartate residues that were paired with lysine residues. On the other hand, D 105 had the lowest SASA value.

IV. Discussion
From the characterization study result, in vitro forced glycation can successfully produce a nearly 100% glycated rhuMAb A. The ESI/MS result indicated that the forced glycation of rhuMAb A forms a heterogeneous mixture, with different extents of glucose modification. To understand the complexity of the fully glycated rhuMAb A, we calculated the possible glucose to rhuMAb A ratio, based on different scenarios of glucose modification on light chains and heavy chains. By combining possible glucose ratio on both heavy chains and both light chains, we estimated that the average intact rhuMAb A may contain from one to ten glucoses per molecule, with the majority containing at least one glucose on the light chain.
The forced glycation model of rhuMAb A was also proven to mimic the site specific glycation in typical cell culture conditions, meaning the most susceptible glycation site is still light chain lysine, K49. Since this forced glycation model provides high abundance of glycated rhuMAb A in a much rapid manner than a 10-14 day cell culture, it becomes a great tool to investigate site specific glyation reaction on MAbs and proteins.
In the past, researchers had shown that glycation specificity is not governed by the pKa of an amine group, rather by the internal micro-environment of a protein.
Neighboring amino acids can often involve a catalyst effect. Watkins, et al. first reported specificity of glycation on RNase in 1985. They reported that certain lysine residues, especially the ones close to the RNase active site, were easy to be glycated.
On the other hand, the surface lysine residues were relatively inactive. Watkins also proposed that both the Schiff base equilibrium concentration and the rate of Amadori rearrangement at lysine residues were important in determining the specificity of RNase glycation . For site specific glycation in MAbs, Quan, et al. had observed histidine presence near a lysine residue, and proposed that a histidine residue may act as a proton transfer which could help to glycate nearby lysines (Quan, C. et al. 2008).
In the case of rhuMAb A glycation specificity, Zhang's group proposed an aspartate catalyst effect in 2008. The proposed mechanism was based on intensive mutation work, suggesting that light chain aspartate D31 is the essential catalyst for K49 site-specific glycation phenomenon. However, even with mutations of D31 to other amino acids, such as threonine, lysine K49 was still glycated to a relatively high level, ~ 25% compared to ~ 40% before mutation . Since Zhang's estimation of spatial distance of D31 and K49 was based on static crystal structure, it does not represent the lysine residue's activity and movement in a dynamic liquid environment.
In our study, molecular dynamics was used to evaluate K49's dynamic motions in aqueous solution. Two parameters were calculated 1) solvent accessible surface area, and 2) the distance between aspartate residues and lysine residues on this rhuMAb A's Fab region.
The presented work demonstrated: Lysine K49 does not exhibit the highest solvent accessible surface area (SASA) compared with other lysine residues. Rather K49 is on the lower rank of exposed lysine residues, indicating that site-specific glycation on rhuMAb A is not determined by solvent accessibility. Among five lysine residues, which have aspartate spatially close to them, only two lysines (K49 and K75) were significantly glycated. This indicates that the presence of aspartate near by lysines does not necessarily catalyze glycation either. Finally, a total of three aspartates are thought to be involved in K49 microenvironment. However, the distances and angles between these aspartate residues and K49 do not favor formation of hydrogen bond. D30 and D31 are located 11 Å from K49, which is too far to form hydrogen bonds with K49. D105 is only 5.3 Å from K49, which is still too far to form a hydrogen bond. Therefore, it is unlikely that these aspartate residues react or bond to lysine K49 directly to promote glycation. Instead, the three aspartate residues together are thought to provide a strong electron-donating environment around K49 at physiological pH.
How would this catalysis work? Do the aspartate residues directly react with glucose rather than lysine? We are inspired by knowledge learned from nature, especially the enzymatic mechanism of two enzymes in sugar metabolism: Aldose B and Triose phosphate isomerase. Aldose B has an active site involving an aspartate residue D33, and a lysine residue K229. These two residues act as mediators to shuttle the electron on the carbonyl group and the hydroxyl group on sugar. Garrett, R.H and Crisham, C.M., 2010). Triose phosphate isomerase also contains a basic glutamate residue E165 and an acidic proton-donating histidine residue H95 to convert the -CHO aldehyde to a -C=O ketone group of sugar  1) The forced glycation model is successful to analyze glycation modification from the whole protein level to the primary amino acid level.
2) The forced glycated rhuMAb A molecule is shown to be a mixture of species with various levels of glycation with complex structure modification.
3) Site-specific glycation on K49 is not due to spatial proximity of D31. Instead, a strong electron-donating environment created by three near by aspartates: D30, D31, and D105 may facilitate K49's specific glycation.    Table 2-3: Summary of lysine to aspartate pairs with average spatial distance between lysine and aspartate.
Unveiling a glycation hot spot in a recombinant humanized monoclonal antibody.

Abstract
An in vitro forced glycation model was applied to evaluate glycation product stability, degradation pathway, and the degradation products of glycated rhuMAb A.
During an accelerated stability study, upon thermal stress at 40°C for four weeks, glycated rhuMAb A went through two reactions: hydrolysis of glycation adducts (Schiff base and Amadori product) and formation of advanced glycation products.
While hydrolysis reaction was more pronounced than the AGEs formation. Kinetically, it was found that hydrolysis of glycation adduct continued for four weeks, but AGEs formation plateaued after one week. The overall combination of both reactions caused glycated rhuMAb A to lose affinity to boronate resin, with a first order reaction rate. A set of analytical characterization on final degradation products structures was performed. The results demonstrated that the unglycated form of rhuMAb A, which was generated by hydrolysis of glycation site at K49 is the main final degradation products after thermal stress. This observation agreed with kinetic observation that reverse hydrolysis of both the Schiff base and Amadori product is the main degradation pathway of glycated rhuMAb A at 40°C in pH 6.5 phosphate formulation buffer. The reversibility of glycation adduct on K49 may be catalyzed by adjacent aspartic acids.

I. Introduction
The common manufacturing process of a MAb involves production in mammalian chinese hamster ovarian cells (CHO) by cell culture for 10-14 days.
During this process a MAb is expressed in CHO and secreted into the cell culture medium, which contains glucoses, where the MAb can undergo glycation. Thus the purified MAb contains a certain level of glycation (Quan, C. et al. 2008. Several studies have been performed to understand the cell culture conditions that control the total glycation level (Yuk, I.H. et al. 2011). However, little knowledge is available regarding impact of glycation on most final purified MAbs. In particular, the stability and degradation pathway of glycation adducts under pharmaceutical drug product conditions are not well understood. This is due to wide distribution of glycation sites on a large protein, low level of glycation on each site, and lack of a robust analytical technology capable of quantify glycation level and changes.
As described in previous study (Chapter 2) , a forced glycation rhuMAb A model was developed and shown to be a successful tool to evaluate amino acid susceptibility to glycation. Fully glycated rhuMAb A was produced from this accelerated in vitro model. This highly glycated product provides an opportunity to understand the glycation adduct's degradation pathways and to perform further structural characterization.
The glycation reaction has multiple steps and produces multiple intermediate The forced glycation model for rhuMAb A described in previous study was implemented to investigate the stability and degradation pathways of glycated rhuMAb A under pharmaceutical storage conditions. To understand the destiny of Schiff base (aldimine) and Amadori product (ketoamine), we focused on glycation adduct stability in formulation buffer in the absence of exogenous carbonyl reagents.
An accelerated stability study of glycated rhuMAb A upon thermal stress was performed. The samples were incubated at 40°C for 4 weeks at pH 6.5 in 20 mM sodium phosphate. The total level of glycation adduct (% protein glycated) was monitored by boronate affinity HPLC (BAC) with UV detection at 280 nm as previously described. In addition, the glycation adduct's glucose to protein ratio was monitored under reduction conditions by LC-ESI/MS. Fluorescence spectroscopy and BAC with fluorescence detection were used to monitor the formation of AGEs upon thermal stress. The total amount of AGEs was measured via quantification of overall polyclonal antibody binding to different AGEs species. Relative intensity of AGEs during a thermal stability study was reported.

II. 1. Materials and Reagents
A

II.2. In Vitro Glycation of rhuMAb and Stability
The

II. 4 Enzymatic Digestion of rhuMAb A
Several enzymatic treatments were performed on rhuMAb A samples before they were analyzed for reverse phase-HPLC, or mass spectrometric analyses, to determine the amino acid sequence and structural integrity of rhuMAb A.
PNGase F Treatment. PNGase F was used to remove the oligosaccharides in the Fc portion of a MAb. The rhuMAb A samples were buffer exchanged into 50 mM Tris buffer, pH 7.5, using a NAP-5 column. The protein concentration was adjusted to 2.5 mg/mL. PNGase F was added in an enzyme-to-substrate ratio of 1:600 (w/w). The digestion was performed at 37°C overnight (15 h

II.5. Reverse Phase-HPLC with Electron Spray Ionization/ Mass Spectroscopy (RP-HPLC-ESI/MS)
The molecular masses of the intact, PNGase F-treated, and reduced rhuMAb

II.6. Tryptic Peptide Mapping
Peptides profiles of tryptic-digested rhuMAb A were analyzed by an RP-HPLC using a Jupitor C18 2.0x250 mm column at 45°C and UV detection at

Flight mass spectrometry (MALDI-TOF)
Isolated tryptic peptides were spotted individually on a SCOUT 384 multiprobe plate with a CHCA matrix and analyzed in the positive mode using a Bruker Ultraflex-I MALDI-TOF-TOF mass spectrometer (Bremen/Leipzig, Germany).
Sele8cted precursor ions were subjected to collision-induced dissociation with argon as the collision gas. LIFT mode was used to analyze all fragment ions in single sweep, and data were processed with Flex Analysis software.

II. 8. Fluorescence Spectroscopy
Fluorescence emission spectra were measured using a Fluoromax-4 spectrofluorometer by Horiba Jobin Yvon ( Edison NJ, USA). The excitation wavelength was 350 nm, emission wavelength was 440nm. A circulating water bath set at 25 ± 1°C controlled the sample cell temperature during the measurements. Both spectrometric analyses were performed using cuvettes (12.5 × 12.5 × 36 mm) from Fisher Scientific (New Lawn, NJ, U.S.A.).

II.9. Total AGEs Enzyme-linked Immunosorbent (ELISA) assay
AGEs ELISA is an enzyme immunoassay for rapid detection and quantification of AGEs-protein adducts.

III.1. BAC
The thermally stressed samples exhibited a rapid loss of total glycation level over 4 weeks at 40°C, as shown in Figure 3-2. The glycation level dropped from 98.5% at initial time, to 69.6% after one week, 60.5% after two weeks, and 32.8% after four weeks. The decrease of glycation is a linear plot on natural log vs. time, indicating this is a first order reaction for glycation loss (Figure 3-3).

III.2. LC-ESI/MS
The decrease of glycation can occur because of the Schiff base reversibility, that resulting in a loss of glucose or formation of AGE products that alter the cishydroxal structure. Reduced ESI/MS analysis confirmed the glucose loss on both light chains and heavy chains (Figure 3-4). No new mass species were detected on the mass spectrometry profile besides known glycated and un-glycated rhuMAb A peaks.
Most importantly, there was an obvious change on light chain mass spectrometry profile. The glycated light chain intensity decreased after thermal stress, while the un-glycated light chain peak intensity increased. Unglycated rhuMAb A finally became the main light chain peak in mass spectrometry profile (Figure 3-4). This indicates most glycated light chains went to the unglycated form, demonstrating the reversibility of glycation modification under the thermally stressed condition at pH 6.5.

III.3. Fluorescence Spectroscopy and BAC
While thermal stress reduced the amount of glycation, can it also introduce more advanced glycation products? Formation of pentosidine-like AGEs can be measured by fluorescence spectroscopy at excitation 350nm/emission 440 nm (Dutta, U. et al. 2005). After thermal stress, fluorescence spectroscopy and boronate affinity chromatography with fluorescence detection, were utilized to monitor fluorescent AGEs. Fluorescent analysis was performed on bulk property for solution of glycated rhuMAb A at 1 mg/ml (Figure 3-5). The result showed a jump of emission intensity between initial time zero and one week after thermal stress. After one week the relative emission intensity plateaued out. This indicates the formation of pentosidine like fluorescent AGEs occurred rapidly and stabilized after one week.
Fluorescence traces on BAC showed two peaks with fluorescent signal at 350/440nm (Figure 3-6). Peak 1, at retention time 2.4 min, overlapped with unglycated rhuMAb A. Peak 2, at retention time 2.7 min, is a new peak formed upon thermal stress (Figure 3-6). The intensity of both peaks increased during the thermal stress, peak 1 being the major peak. However, the total peak area based on fluorescence signal was very low, making it hard to quantify the exact amount of AGEs. Figure 3-7 shows the comparison of the UV and fluorescence signals in flow through portion on BAC, while the fluorescence signal increased linearly with time, the size of the peak at 280nm increased. This was due to the continuous increasing amount of rhuMAb A in the flow-through portion. After normalizing the fluorescence peak area by the protein amount, the normalized fluorescence signal/protein amount is consistent with spectroscopy measurements, the amount of AGEs detected on BAC increase in the first week of thermal stress, then plateaued (Figure 3-8).
Therefore, BAC with tandem-UV 280nm -fluorescence 350/440nm detection is capable of monitoring both glycation level change and AGEs formation.

III.4. Peptide Mapping
What are the degradation end-products of the amino acid level after thermal stress? The possible degradation products from thermally stressed, glycated rhuMAb A were studied using tryptic peptide mapping. By comparing the time zero control to thermally stressed samples, several peptide fragments were found after thermal  Table 3-1 summarizes the degradation products on each glycated amino acid site. The MS/MS analysis on these peptides identified two major degradation products. One is the non-fluorescent AGE: carboxyl-methyl lysine (CML) modified peptides which eluted around 40 min on peptide map (Figure 3-9). The other is the unglycated rhuMAb A, which eluted around 79 min (Figure 3-10). The presence of these two products confirms that glycated rhuMAb A underwent both pathways: reversal of glycation and formation of advanced glycation endproducts.
The abundance of unglycated rhuMAb A was greater than that of the CML modified rhuMAb A. It is notable that majority of glycation reversal occurred on K49, other glycation site appeared to be more stable, maintaining consistent glycation levels after thermal stress.

III.5. Total AGE ELISA
AGE ELISA is based on binding of a family of polyclonal antibodies against AGEs. Total AGEs was measured by using polyclonal antibodies to capture variety of different AGEs structures, including fluorescent structures and non-fluorescent structures (Figure 3-11). The ELISA result showed an increase of relative intensity of AGEs upon thermal stress. This confirmed AGEs formation during the 40°C incubation. The increasing trend was similar to fluorescence and BAC results, where a rapid AGEs formation was detected in the first week then plateaued (Figure 3-12).

Degradation pathway of glycated rhuMAb A
The thermal stability study demonstrated that glycated rhuMAb A can undergo two possible parallel reactions (Figure 3-13). The glycated rhuMAb A either reverses back to the unglycated rhuMAb A, or moves forward to form AGEs. It is important to note that 1) the glycation adduct hydrolysis reaction is more pronounced than the AGE formation, and 2) glycation adducts hydrolysis continued for four weeks, but AGEs formation plateaued after one week. What could be the scientific rationale behind these observations?
Amadori rearrangement is favored at pH greater than pH 7.0, where the C=N double bond can shift to a C=O double bond. At pH 6.5, the rearrangement from Schiff base to Amadori product is not at the optimum condition, and the most of early glycation adducts should be in the Schiff base form. Since the Schiff base is thermodynamically unstable in liquids at neutral pH, the hydrolysis reaction is favored. Thus, the reversal to unglycated amino acid on protein is the predominant reaction and it proceeds towards completion. Second, the amount of carbonyl groups (C=O group) in Amadori products restrained further reaction to advanced glycation, because advanced glycation needs continuous attack of carbonyl group to amine group. Without additional free glucose in the formulation, the only carbonyl group available to initiate further advanced glycation reaction is the ketone in Amadori products or its degradation bi-carbonyl products. Since amount of precursor Amadori product is limited, the amount of ketone carbonyl available for advanced glycation reaction is also limited. In addition, the ketone group in the Amadori product is less reactive than the regular aldehyde group in free glucose. Therefore, under the studied pharmaceutical storage condition, the total amount and reactivity of carbonyl groups were limited, and the AGEs formation reaction is not the dominant reaction. This is why AGEs formation reached plateau after one week.

AGEs Structures
The low level AGE products formed after thermal stress was confirmed to be a mixture. The fluorescent pentosidine-like ring structures and straight chain CML structures were both indentified (Figure 3-14). However, due to the limitations of analytical methods, it is not possible to quantify the exact amount of fluorescent pentosidine-like AGEs, or to further characterize their structures. The straight chain CML can be identified and quantified by peptide map with MS/MS analysis. CML was less than five percent of the final degraded rhuMAb A.

Degradation at Modified Sites
Although we did not study the glycation loss of each site during thermal stress kinetically, the glycation reversibility of each site can still be evaluated, based on the final glycation level at each site that were summarized in the Table 3-1. K49 and heavy chain N terminal had the highest glycation level initially, but these two sites also have the most rapid reversibility of glycation adduct. As we described in Chapter 2, three aspartic acids catalyze the glycation on K49. Based on catalyst nature of the K49 microenvironment, the nearby aspartic acids can also facilitate the reverse reaction: hydrolysis of glycation adducts. The reason for glycation reversal on the heavy chain N -terminal is yet to be known.
The unique site specific glycation nature of rhuMAb A makes it a great candidate to study MAb structure modification by glycation and glycation adduct stability. The following conclusions can be drawn from the studied performed: 1) The forced glycation model is successful to analyze glycation chemical modification on whole protein levels and primary amino acid levels.
2) BAC with tandem-UV 280-fluorescence 350/440nm method is a great tool to monitor both glycation level change and AGEs formation.
3) Upon thermal stress, the forced glycated rhuMAb A showed a rapid and continuous hydrolysis reaction, and a slow and limited advanced glycation reaction.
The overall combination of both reactions caused glycated rhuMAb A to lose affinity to boronate resin with a first order reaction rate.
4) The structure analysis of final degradation products agreed with kinetic observation that reversibility of Schiff base is the main degradation pathway of glycated rhuMAb A at 40°C in pH 6.5 phosphate formulation buffer. 5) Reversibility of glycation adduct on K49 is catalyzed by its adjacent aspartic acids.  a) at control t0 rhuMAb A is 98.5% glycated; b) at 1week 40°C rhuMAb A is 69.6% glycated; c) at 2weeks 40°C rhuMAb A is 60.5% glycated; d) at 4weeks 40°C rhuMAb A is 32.8% glycated.        Reference: Sell, D. R., Nagaraj, R. H., Grandhee, S. K., Odetti, P., Lapolla, A., Fogarty, J. and Monnier, V. M. (1991), Pentosidine: A molecular marker for the cumulative damage to proteins in diabetes, aging, and uremia. Diabetes /Metabolism Reviews 7: 239-251. Abstract A forced glycation model on rhuMAb A and boronate affinity chromatography technology were applied to evaluate the effect of drug product compositions on the degradation of glycation adduct. Three formulation composition factors were studied, such as pH, buffer, and oxidation control.
The results demonstrate that the pH is the key parameter to control degradation pathways, as pH also determines the hydrolysis rate constant of glycated rhuMAb A. The unique structure of rhuMAb A forms a rich electron donating environment around K49. This unique environment determines the pH dependence of glycation adduct's hydrolysis rate. Buffer species and oxidation levels do not impact the glycation adduct stability under the studied pharmaceutical conditions. In addition, the effect of rhuMAb A's glycation level was also evaluated. The study demonstrated that higher protein glycation levels slow down the overall hydrolysis rate of the glycation adduct, presumably by providing a complex electron transferring system on protein.

I. Introduction
In previous studies, it was shown that rhuMAb A can be used as a model to study site-specific glycation and glycation adduct degradation pathways. In addition, the boronate affinity chromatography (BAC) coupled with UV and fluorescence detection is suitable to monitor both glycation level change and AGEs formation.
Previous protein characterization work and stability study demonstrated that the glycation adducts of rhuMAb A degraded via two pathways, under thermal stress in a typical pharmaceutical formulation at pH 6.5. The main reaction was hydrolysis to reform unglycated rhuMAb A, while a minor reaction was the formation of AGEs.
To understand which parameters control the stability of glycation adducts (Schiff base or Amadori product) during the final drug product storage, we used glycated rhuMAb A as a MAb model and BAC as the analytical tool to investigate the effect of four key parameters on glycation adduct stability, in a set of thermal stress studies. These parameters included three important factors for formulation compositions: pH conditions, buffer species, oxidative conditions. The final factor, relevant to the protein, was the initial glycation level, which is based on rhuMAb A's ( the active pharmaceutical ingredient) production process.

pH conditions
Since both formation of Schiff bases and rearrangement of Amadori products are pH depended, we investigated the overall effect of pH on the stability of glycated rhuMAb A. Six different pH conditions were tested between pH 4.5 and pH 7.0, which is a typical pH range for protein therapeutics formulations. One drug product lot, containing 36% glycated rhuMAb A, derived from common cell culture was used in the pH study. The rhuMAb A drug product was buffer exchanged into six different pH conditions, including 20mM sodium acetate buffer at pH 4.5 and 5.0, and 20mM histidine/histidine hydrochloride buffer at pH 5.5, 6.0, 6.5, and 7.0. In all samples, the final protein concentration was 0.5mg/ml with 8% trehalose, and 0.04% polysorbate 20. Watkins, et al. had reported that phosphate ions could bind to the active site of RNase and serve as a local catalyst for Amadori rearrangement of lysine residues near the RNase's active site. They established the buffer effect on the kinetics, and specificity of glycaiton of RNase. . To explore the buffer effect on glycated rhuMAb A, we tested its stability in three buffers: • phosphate, to test the effect of phosphate ions on stabilizing glycated rhuMAb A; • arginine, to test its inhibition effect on AGEs formation; • histidine as a typical formulation control.

Control on Oxidative Conditions
Oxidation involved pathways are the keys for initiating AGEs formation. In this study, one drug product lot with 15% glycated rhuMAb A was placed three different oxidation conditions: • regular drug product vial with air in the head space; • drug product vial with N 2 in the head space; • drug product vials with phenol added as an oxidation scavenger. histidine buffer, pH5.8 with 0.02% polysorbate 20.

Glycation level:
As previously reported, cell culture conditions impacted the final glycation level of rhuMAb A (Yuk, I.H. et al. 2011). It is unknown if glycation adduct hydrolysis under in vitro storage would vary with initial glycation levels, so we compared hydrolysis of glycated rhuMAb A among three different initial glycation levels: 13%, 42% and 99%. Final samples contained 20mg/ml protein in 20mM histidine buffer at pH 6.5.

Stability set up:
After samples were prepared under each condition, samples were stressed thermally at 40°C for four weeks. After stresses, samples were analyzed to evaluate the stability of glycated rhuMAb A. All the collected time points were analyzed with boronate affinity chromatography with UV and fluorescence detection (BAC/UV/FLuo) assay. Total glycation level and AGEs formation were monitored by boronate affinity chromatography.

II.3. Boronate Affinity Chromatography (BAC)
High performance liquid chromatography (HPLC) with solvent delivery system were used, where the analyte separation were performed with an Agilent 1100 series HPLC equipped with a thermostatic auto-sampler.
Early stage glycation adducts often can be identified by a BAC method using A TSK gel boronate 5PW column 7.5x75mm. The mobile phase A was 50mM EPPS, 10mM Tris, 200mM NaCl, at pH 8.7, and mobile phase B was 500mM sorbitol in mobile phase A. The mobile phase flow rate 1 mL/min and the column temperature will be maintained at 40°C with UV detector at 280 nm and fluorescence detector at 350/440nm. The gradient for HPLC elution had been optimized for this rhuMAb A.
Due to the difficulty for quantification of fluorescence signal, the exact amount of fluorescent AGEs cannot be determined, only relative fluorescence intensity is used to compare different stability conditions.

II. 4 Enzymatic Digestion of rhuMAb A
Several enzymatic treatments were performed on rhuMAb A samples before they were analyzed for reverse phase-HPLC, or mass spectrometric analyses, to determine the amino acid sequence and structural integrity of rhuMAb A.
PNGase F Treatment. PNGase F was used to remove the oligosaccharides in the Fc portion of a MAb. The rhuMAb A samples were buffer exchanged into 50 mM Tris buffer, pH 7.5, using a NAP-5 column. The protein concentration was adjusted to 2.5 mg/mL. PNGase F was added in an enzyme-to-substrate ratio of 1:600 (w/w). The digestion was performed at 37°C overnight (15 h).
Tryptic Digestion. Prior to tryptic digestion, rhuMAb was reduced with DTT and then alkylated with IAA. Typically, 250 µL of antibody sample (2 mg/mL) was mixed with 20 µL of 1 M DTT in 730 µL of 6 M guanidine, 50 mM Tris, pH 8.0.
The mixture was incubated at 37 °C for 1 h. It was then cooled to room temperature, and 50 uL of 1 M (IAA) in 1 M NaOH was added for carboxymethylation. The alkylation reaction was incubated at room temperature in the dark for 15-20 min. The residual IAA was quenched by the addition of 10 µL of 1 M DTT. The reduced and carboxymethylated rhuMAb A was then buffer exchanged into the digestion buffer, which contained 25 mM Tris, 1 mM CaCl 2 , pH 8.3, using a PD-10 column (Sephadex G-25 medium, GE Healthcare). Trypsin was added at an enzyme-tosubstrate ratio of 1:50 (w/w). The solution was mixed briefly and incubated in a 37°C water bath for 5 h. The digestion was terminated by adding 0.3% (v/v) trifluoroacetic acid (TFA) to the solution. The digest was then stored at -70°C until analysis.

II.5. Tryptic Peptide Mapping
Peptides profiles of tryptic-digested rhuMAb A were analyzed subsequently by an RP-HPLC using a Jupitor C18 2.0x250 mm column at 45°C and UV detection at 214nm. The flow rate was at 0.25ml/min, in mobile phase A: 0.1% TFA in water and B: 0.09% TFA in acetonitrile. The gradient was optimized for this rhuMAb A.
By comparing the peptide maps of time zero and stressed glycated rhuMAb A, the distinctive peptide factions on degraded glycated rhuMAb A were collected for further characterization by MALDI-TOF MS/MS sequencing.

mass spectrometry (MALDI-TOF)
Isolated tryptic peptides were spotted individually on a SCOUT 384 multiprobe plate with a CHCA matrix and analyzed in the positive mode using a Bruker Ultraflex-I MALDI-TOF-TOF mass spectrometer (Bremen/Leipzig, Germany).
Sele8cted precursor ions were subjected to collision-induced dissociation with argon as the collision gas. LIFT mode was used to analyze all fragment ions in single sweep, and data were processed with Flex Analysis software.

pH conditions:
BAC-UV 280nm to measure total glycation level change The pH study revealed that glycation adduct stability is greatly dependent on pH conditions. By comparing affinity chromatograms of the final time point (40°C 4weeks) at different pH conditions, we found that as pH increased, the rhuMAb A samples lost the affinity to boronate ligand upon thermal stress (Figure 4-1), indicating loss of glycation adduct. The total remaining glycation level is summarized in Table 4-1. It shows that samples stored at pH 4.5 had the highest remaining glycation level, 32.0% compared to 36% at time zero. From pH 5.5 to 6.0, remaining glycation level dropped from 22.0% to 8.7%, while pH 7.0 had the lowest remaining glycation level, only 5.6%.
The graph of glycation level vs. pH, confirms that the remaining glycation level is inversely dependent on pH values (Figure 4-2). More importantly, the graph shows a non-linear relationship between pH and remaining glycation level. There were three pH responding sections, in which the glycation adduct's stability responded to pH change differently. From pH 4.5 to 5.5, the glycation adduct level change remained fairly mild. After thermal stress, the remaining glycation level differed about 5% per 0.5 pH unit, which is a modest response to pH change. From pH 5.5 to pH 6.0, this region showed the most drastic relation ship between on the remaining glycation level and pH. In this narrow pH range, with only 0.5 pH unit increases, the remaining glycation level dropped 13.3%. From pH 6.0 to 7.0, the final glycation level decreased rapidly. In this range, the remaining glycation level was less than 8.7 %. However, the difference of remaining glycation level between pH 6.0 and pH 7.0 is small, only approximately 2% per 0.5 pH unit. Therefore pH 5.5 to pH 6.0 is the most sensitive pH range for glycated rhuMAb stability.

BAC-Fluorescence 350/440nm to measure AGEs formation
In contrast to loss of glycation adducts, fluorescence traces on BAC showed two peaks with fluorescence signals at 350/440nm, which represent formation of AGEs. Peak 1, at retention time 2.4 min, overlapped with unglycated adducts. Peak 2, at retention time 3.6 min, is a new peak formed upon thermal stress. (Figure 4-3) Comparing the fluorescence signals of different pH conditions after thermal stress, we found that the intensity of fluorescence signal increased with increasing pH (Figure 4-4). It is notable that, at pH 4.5, 5.0, 5.5 and 6.0, only peak 1, the unglycated peak was observed. At pH 6.5 and pH 7.0, both peak 1 and peak 2 were observed.
This characterization result indicates that a new pentosidine-like AGEs-protein species was formed upon thermal stress at higher pH conditions. However, even at pH 6.5 and 7.0, the majority of fluorescence signal intensity was attributed to peak 1, the unglycated rhuMAb A. AGEs formation demonstrated by peak 2, represented a relatively small portion of total fluorescence intensity.
Comparing the pH dependence of total glycation level (Figure 4-2) and pH dependence of AGE formations (Figure 4-4), it is notable that below pH 6.5, the most pronounced fluorescent peak was the hydrolysis product, unglycated rhuMAb A. This indicates that at low pH range, pH 4.5 to pH 6.0, the hydrolysis of glycation adducts was the main degradation pathway. At higher pH range, pH 6.5 to 7.0, both peak 1 and peak 2 were formed, but the intensity of peak 2 was small compared to peak 1. This observation reveals AGE formation plays a minor role in the degradation pathway for rhuMAb A. Overall, the preferred degradation pathways of glycation adducts were driven by the pH conditions.

Determination of the hydrolysis rate constant
We can calculate the rate constant of hydrolysis at different pH conditions based on the following two assumptions: the hydrolysis reaction rate was is much greater than AGEs formation rate at all pH conditions; and the hydrolysis reaction was a first order reaction, as we learned in Chapter 3.
The rate constant k can be derived from this function

k=ln([A 0 ]/[A n ])/t n .
Where, k is the reaction's rate constant and n is the particular time point, e.g. four weeks etc. The half-life of glycation adducts at each condition can be calculated as: Where, t 1/2 is the half-life of glycated rhMAb A. It means at this time point, the concentration of glycation adducts equals to half of the initial concentration of glycation adduct. The parameter a is the molar ratio of reactant for glycation adduct.
In our study, a =1. Therefore, pH conditions determine the rate of hydrolysis of glycation adducts.
BAC can measure the amount of both Schiff base (aldimine) and Amadori product (ketoamine). The total glycation level is the sum of these two species in which cis-hydroxyal groups are still attached to rhuMAb A molecules.
Historically, Schiff base is known to be thermally unstable, and to undergo  , et al. 1963). Their study demonstrated a transition of the rate limiting step. As pH conditions decreased, rate-determine step changed from Schiff base formation/ hydrolysis under neutral to alkaline pH (step 1), to rate-determine amine attack / decomposition of carbinol-amine under acidic pH (step 2). They also found that the rate of hydrolysis markedly decreased in strong acid, which agrees with the results in our study. Furthermore, Jencks's group showed that under more acidic conditions, a large fraction of Schiff base exists as the conjugate acid. The hydrolysis rates decrease with decreasing pH for Schiff bases indicating an electron-donating functional group on the molecule. The changes in rate with pH are correlated with the conversion of Schiff bases to their conjugate acids.
We can determine the pKa of glycation adduct of rhuMAb A by plotting % glycated rhuMAb A vs. pH, or k obs hydrolysis vs. pH ( Figure 4-6). The plateau area at pH 5.75 represents the maximum glycation level change upon pH changes. pH 5.75 is the pKa value of glycation adduct in Schiff base /conjugate acid form (Figure 4-6).
The other aspect of glycation adduct stability is the Amadori rearrangement.
After dehydration, the Amadori rearrangement is favored at alkaline conditions ( Figure 4-7) at pH >7.0. The Amadori product is considered more stable than the Schiff base. However, there are studies  that have shown the ketoamine linkage between glucose and protein to be reversible. In our particular study subject of glycated rhMAb A, the major change in glycation was observed on K49, the preferred glycation site. The observed rate of hydrolysis decreased with decreasing pH. This pH-dependent hydrolysis indicates the electron donor-enriched environment around lysine K49. The catalyst effect of three aspartic acid residues surrounding K49 facilitates the formation of the Amadori product (ketoamine) on K49. They also promotes K49 Amadori product reverse to Schiff base (aldimine), then Schiff base reverse to unglycated rhuMAb A.
Therefore, between pH 4.5 and 5.5, the attack from nitrogen on the lysine amine group to the carbon on carbonyl group of glucose is the rate-limiting step, while adding a water to Schiff base to form conjugate acid is favored. The BAC was able to detect both Schiff base and its conjugate acid. Since the conjugate acid is stable in lower pH conditions (pH <5.5), overall glycation level of rhuMAb A is maintained. At slightly higher pH, pH 5.5 to 6.0, around the pKa of the conjugate acid of Schiff base (pH 5.75), the amounts of base and conjugate acid were equal. A big change of Schiff base is needed to shift the pH of Schiff base/ conjugated acid system. Thus, we observed at pH 5.5 to pH 6.0, there was a significant difference on the remaining glycation level. BAC detected glycation adducts were mainly in Schiff base form. In addition to rapid loss of glycation adducts, at pH 6.0 to 7.0, Amadori rearrangement occurred leading to a minor advanced glycation reaction. The BAC detected glycation adducts are both Schiff base and Amadori products. Hence, the overall main reaction in the studied pH conditions, from pH 4.5 to pH 7.0, is the reverse of glycation adduct by Schiff base hydrolysis.
We can conclude that pH condition controls both the direction and rate of Schiff base hydrolysis. Below the pKa of the Schiff base, glycation adducts are relatively stable. However, at the pKa and above, Schiff base hydrolysis becomes the preferred reaction, with Amadori rearrangement as a secondary reaction, leading to loss of glycation adducts and an increase in prevalence of AGEs.

Buffer species
After 4 weeks of thermal stress at 40°C, no difference on BAC was found among three buffer systems: histidine, arginine and phosphate buffers. Using the same approach as described in the previous section, we calculated the degradation rate constant in three buffers. The result demonstrated similar rate constants among histidine, arginine, and phosphate buffer systems, with 0.10 week -1 , 0.11 week -1 , and 0.13 week -1 , respectively. This observation indicates that these buffers did not cause any difference on the reversibility of glycation adducts (Figure 4-8, Table 4-3). In addition, the trends of AGE formation in three buffers detected by BAC were parallel at fluorescence 350nm/440nm (Figure 4-9). Peptide mapping results confirmed that even at the amino acid level, the degradation products are similar in three buffers ( Figure 4-10). Glycated rhuMAb A was shown to degrade to wild type rhuMAb A (>80%) while a small portion (<10%) degraded to carboxyl-methyl-lysine (CML) (Figure 4-10), The relative proportion of these two degradation products were at similar levels in all three buffers after thermals stress at 40ºC for 4 weeks. Watkins, et al. has reported the catalysis effect of phosphate buffer on the kinetics, and specificity of glycation on RNase . By binding the RNase active site, phosphate ion was localized on RNase, creating a negatively charged environment. This increase of electron density helped nearby lysine residue to form Amadori product during glycation. However, based on the structure and function of the studied rhuMAb A, there is no phosphate ion binding site to keep phosphate ion localized on protein. From the characterization work from previous study and published work by Zhang , K49 was still the primary susceptible site in both phosphate and sodium bicarbonate buffers. Presence of phosphate ion did not catalyze the glycation on this rhuMAb A. From the presented stability result in this study, phosphate buffer did not show any effect on glycated adduct hydrolysis, either. Thus, we hypothesize that phosphate ions cannot serve as a catalyst in the glycation reaction of this rhuMAb A or be involved in the glycation adduct hydrolysis without the active site to ensure its localization on rhuMAb A. Therefore, phosphate ions did not affect hydrolysis rate of glycated rhuMAb A.
On the AGE formation, arginine was thought to be able to slow down the advanced glycation reaction, by competing with rhuMAb A's own amine groups, to react with any carbonyl group on glycated rhuMAb A. However, in the presented study, arginine did not affect AGEs formation on glycated rhuMAb A. One possible reason is that advanced glycation reactions were minor at the studied pH condition, pH 6.5.

Oxidation Control:
The oxidation controlled study showed very similar hydrolysis rate constants for all conditions. Nitrogen (N 2 ) purged vials resulted in the slowest rate constant, at 0.2%/week; then regular air purged vials at 0.22%/week; followed by phenol, an oxidation scavenger, contained vials at 0.23%/week (Figure 4-11 and Table 4-3).
This indicated the degradation of the glycated rhuMAb A did not slow down at pH 6.5 in absence of oxidative reagent.

Initial Glycation Level
Table 4-4 summarizes the rate constants for all three initial glycation levels.
The hydrolysis rate dropped from 0.46 to 0.39 week -1 for the glycated rhuMAb A species at 13% to 42% respectively. When the protein was 99% glycated, the glycated adduct degradation rate decreased to only 0.10 week -1 , with a half life of ~ 7 weeks (Figure 4-12, Table 4-4). From BAC fluorescence signal, the AGEs formation was fastest at the high glycation level ( Figure 4-13).
Furthermore, analysis of size heterogeneity at different glycation levels was done by size exclusion chromatography. The result demonstrated that at higher glycation level, more protein aggregates were formed. It is a evidence that protein to protein interaction was more pronounced at higher glycation levels ( Figure 4-14).
As previous study showed (Chapter 2), the heavily glycated rhuMAb is a mixture with complex structural modifications. With higher total glycation level, not only K49 was glycated to a higher level, but other lysine residues were also modified. different levels. Table 4-4 summarizes the total glycated levels at each sites detected, 13% one (one site K49 only); 42% (6 sites, K49 level primary) ; 99% (7 sites, K49 primary about 90% as in chapter 2) ).
In the forced glycation model, lysine residues buried inside the protein were modified without the strong catalysis effect provided by neighboring aspartic acids.
These much stable glycation sites may lead to further AGEs formation, as well as protein-protein interaction to provide a complex protein network. Thus, we hypothesize that the complexity and heterogeneity of final glycation states contributes to a slower overall hydrolysis rate. The presence of more enduringly glycated sites may provide a complex electron -transfer network on rhuMAb A that allows electron transferring within the molecule, stabilizing the Schiff base.

IV. Conclusions
1) Among the studied formulation parameters, pH is a key parameter to control glycation adduct degradation pathways, it also determines the hydrolysis rate constant of glycated rhuMAb A.
2) The unique structure of rhuMAb A creates a rich electron donating environment around K49, which determines its site specific glycation and contributes to the pH dependence of hydrolysis rate.
3) Buffer species and oxidation levels do not affect the glycation adduct stability under the studied conditions. 4) Initial protein glycation level does affect the overall hydrolysis rate of glycation adduct on rhuMAb A. This may be related to protein-protein interaction involving electron transferring system among different glycated sites. The pH dependence on glycated rhuMAb stability at 40°C pH 5.5 -6.0 showed to be the most sensitive pH range for glycation change.        The rate of hydrolysis fits a linear regression, indicating a first order reaction. At higher glycation level, hydrolysis rate was slower.     Table 4-4: Summary of rate constant of glycated rhuMAb A degradation at 40°C in three different initial glycation levels.All samples were in 20mM histidine buffer pH 6.5.