Aggregation is a typical occurrence with mAbs and other therapeutic proteins. Several pathways have been proposed to describe the aggregation of proteins. They include agglomeration of monomers in their native states, aggregation of conformationally altered or chemically modified monomers, nucleation, and surface-induced aggregation [34,35,36]. Significant attention has been placed on non-native monomer aggregation, since exposed hydrophobic moieties tend to self-associate [34]. Some surfactants, osmolytes, and chaotropes induce aggregation because they denature the monomeric product, exposing more of the hydrophobic core and distorting the surface charge distribution [37]. Physical and biochemical events can also induce product degradation through enzymatic and non-enzymatic processes such as shock, light, and oxidation [38].

Physical or chemical perturbations that put a strain on the native conformation of biotherapeutic proteins such as mAbs can result in clipping or aggregation [39]. Such conditions include the presence of chaotropic chemical species, pH swings [40], shock, mechanical stress, increased concentration, and large temperature fluctuation [34,41,42]. The size, charge, and hydrophobicity of a mAb aggregate will differ from that of the native structure.

Interfacial damage can also affect the stability of a product monomer, especially at the air–liquid interface, which induces nucleation and aggregation [43]. Surface tension and physical adsorption on solid surfaces also lead to conformational changes [34,44,45]. Freezing and thawing of a product induces more aggressive fouling of virus filters [2,23,46]. Freeze–thaw-induced aggregation is due to conformational changes at the ice–water interface and by freeze concentration [34,44,47].

Reversible aggregates are usually a precursor to nucleation [48], followed by irreversible aggregation as the aggregates increase in size [49,50]. As buffer ionic strength increases, electrostatic repulsion between the mAbs decreases, whereas hydrophobic attraction between the mAb increases, often leading to product aggregation [44]. Aggregation and precipitation occur most easily at the product’s isoelectric point (pI) due to reduced electrostatic repulsion between individual product molecules [51,52].

Aggregation can occur through different pathways, resulting in aggregates that are reversible or irreversible [34,37]. mAb oligomers such as dimers, trimers, and tetramers are typically reversible [34,35]. Reversible aggregates are known as soluble aggregates, and the associated product monomers are not significantly denatured. Soluble aggregates are caused by interactions between product molecules via hydrogen bonding, electrostatic, or van der Waals forces [37,53]. These soluble aggregates can foul virus filters if their size exceeds the 20 nm size cut-off of most parvovirus filters.

Rayfield et al. [28] investigated the impact of mAb properties on virus filter filterability and showed that aggregates bigger than 17 nm were correlated to the flux decline during virus filtration [2,28]. Monoclonal antibodies typically have a hydrodynamic diameter of 9–12 nm; thus, the small oligomers can be 20 nm to as much as 50 nm in diameter. Other studies have shown that freeze-thawing of mAbs may not cause aggregation in significant amounts detectable by size exclusion chromatography due to the relatively small diameters of potential aggregates formed [46,54,55].

When product dimers and trimers undergo further aggregation, they attain a critical mass where the aggregate can no longer remain soluble. These large aggregates then precipitate out of the solution [34]. The precipitates become visible and show increased turbidity and cloudiness. These large aggregates are known as irreversible aggregates. Large, insoluble aggregates have an increased propensity to foul the separation-active layer during virus filtration. Barnard et al. investigated the principal foulant of freeze–thawed mAb solutions and found that the freeze–thaw process could induce the formation of large aggregates (>1 μm) [46]. The use of 0.1 or 0.22 µm pore size prefilters can mitigate virus filter fouling to some extent by removing these large aggregates.

Irreversible aggregation is prevalent with denatured product monomers [35,37]. Chemical degradation, such as oxidation and deamidation, alters the surface charge of product monomers and affects colloidal stability [35]. The irreversible aggregation of a product results in product loss, although even very low levels of aggregation (<1%) can cause a significant increase in filter fouling. Hawe et al. studied mAb aggregates formed during freeze–thaw- and heat-induced thermal stress [45]. Other studies show that heat denatures mAbs and leads to irreversible mAb aggregation [35,37].

HCPs feature significantly in the downstream processing of protein-based therapeutics [31]. HCPs include proteins, enzymes, and co-enzymes which emanate from the host cell used for product expression [56]. It is essential to remove HCPs from therapeutic proteins because they can elicit an immune response. There are regulatory requirements for robust HCP removal before clinical trials of drug candidates to prevent the development of anti-CHO antibodies by volunteers [57,58]. Some HCPs can co-elute with the mAbs through polishing and purification steps, either due to binding to the resin or to association with the mAb product [56,59,60,61]. Zhang et al. identified over 500 HCPs in a cell culture sample and tracked their fate through downstream processing unit operations [62]. After studying nine different mAbs, they determined that actin and clusterin were most abundant in protein A eluates [62].

Enzymatic HCPs (host cell proteases) can clip or denature product monomers, expose hydrophobic residues and charged moieties, and alter the product’s biophysical properties. Denatured product monomers with exposed residues induce virus filter fouling by adsorptive processes in addition to mAb–mAb and mAb–HCP association. Host cell proteases have been reported to result in the fragmentation of mAb products, with increased susceptibility to nucleation and aggregation [63]. However, proteases themselves are probably not principal foulants of virus filters, since virus filtration occurs towards the end of downstream processing, where only trace amounts of non-mAb impurities may be detected [19,64].

HCP diminishes the biotherapeutic quality of biotherapeutic products and increases downstream processing costs. If HCPs are not sufficiently removed, they could potentially induce flux decay during virus filtration. HCPs have a range of biophysical properties, such as pI (2–11) and mass (10–200 kDa), which can be used to separate the HCP from the biotherapeutic [65,66]. Protein A chromatography significantly reduces HCPs in the mAb product due to high selectivity for the Fc region of mAbs [59]. Several studies reported that the propensity of different HCPs to bind and co-elute with mAbs from protein A columns vary from mAb to mAb [60,67]. Problematic HCPs are many and include lipoprotein lipase, nidogen-1, clusterin, histones, keratins, phospholipases, ribosomal proteins, and serine proteases [65].

Endotoxins or lipopolysaccharides (LPS) are contaminants that can enter the process through growth media or other cell culture additives used in mammalian cell cultures. LPS are produced by Gram-negative bacteria, commonly used in recombinant DNA production [68,69,70]. Endotoxins are commonly found contaminants in mammalian cell-derived therapeutics [71]. LPS are complex molecular conjugates of an amphiphilic component (lipid A) and a polar polysaccharide component [69,70]. The isoelectric point of LPS ranges from 1 to 4 [66]. LPS removal techniques that have been reported include two-phase extraction, affinity chromatography, and ion exchange chromatography [70].

LPS have been reported to have a high affinity for some biotherapeutic proteins [71]. LPS and therapeutic proteins can form micellar aggregates, complicating the removal process and potentially carrying over into the virus filtration step [70,72]. Phosphorylated moieties of LPS electrostatically bind with the carboxyl moieties of amino acids in the biologic of interest [70,72]. Solutions of 0.5 M arginine have been shown to promote LPS clearance during polishing steps [72].

LPS have molecular masses ranging from 3–40 kDa, which vary due to their polysaccharide chain lengths [66,70]. Endotoxins can co-elute with mAbs and Fc-fusion proteins from protein A resins by molecular conjugation through hydrophobic and electrostatic interactions, ultimately causing problems during virus filtration. Endotoxin-contaminated mAb streams have an increased propensity to cause virus filter fouling. Removal of endotoxins through ion exchange polishing steps increases the virus filtration capacity of virus filters.

The charge variant profile is a critical quality attribute of mAbs [73] and Fc-fusion proteins. Charge variants in mAbs can result from post-translational modifications (PTMs), such as deamidation of asparagine, C-terminal lysine variants, and glycosylation [74,75,76,77]. Glycans are mostly polar, hydrophilic oligosaccharides that can induce micro-differences in the surface charge of a glycoprotein. Negatively-charged glycans incorporating phosphorylated mannose and sialic acid can introduce micro-heterogeneities. Charge heterogeneity is observed in isoelectric-focusing electropherograms of most glycoproteins [74]. Acidic and basic variants of glycoproteins such as mAbs and Fc-fusion proteins can have different glycan profiles [78,79]. Meyer et al. [72] reported that specific charge variants of a mAb candidate were aggregation prone. Acidic variants of this mAb showed more pronounced hydrophobicity [80].

The net charge and surface charge distribution of glycoproteins change with buffer pH [76,81,82]. The pI of a protein is the pH value at which the net charge is zero [51]. For most mAbs, the pI ranges from 6.5–9.5 [51]. There is more biochemical variability with Fc-fusion proteins. A protein will be net negatively charged when the buffer pH is above the pI and positively charged when the buffer pH is below the pI [83]. Exposed surface residues on a glycoprotein can become protonated or deprotonated depending on the buffer pH, thereby inducing localized charged groups [84]. Charged moieties due to glycosylation, phosphorylation, and other PTMs affect the net charge of glycoproteins and their interactions with other product monomers and virus filtration membranes [83].

Hydrophobic interaction is the preferential association of non-polar residues in aqueous media [85]. Amino acids with non-polar side chains are typically hydrophobic, e.g., valine, leucine, proline, and tryptophan. Polar amino acids such as arginine impart hydrophilic attributes to glycoproteins [86]. When hydrophobic amino acids are surface-exposed on a glycoprotein, hydrophobicity increases. Hydrophobic amino acids tend to be buried in the globular core of most glycoproteins. The hydrophobicity of a protein is also affected by the buffer pH and the protein’s charge state [87]. When the buffer pH is close to the pI of the protein, the protein is the most hydrophobic [85]. Denaturation and unfolding of glycoproteins can lead to variants with a higher fouling propensity on virus filters.

The glycan appendages of glycoproteins also contribute to the final stable conformation, and glycan variation can introduce minor hydrophobicity variations. Careful handling and mild changes in formulation conditions will reduce the formation of conformational variants, which could foul virus filters or induce product aggregation.

Monoclonal antibodies and Fc-fusion proteins consist of amino acids in specific sequences that form secondary, tertiary, and quaternary structures. Sequence variants arise due to genetically unprogrammed amino acid substitutions, omissions, or insertions during biosynthesis [88]. Sequence variants result in macro-heterogeneities with biomolecular differences from the desired product [88]. Sequence variants possess different affinities to substrates [89] due to surface charge and hydrophobicity dissimilarities. The amino acid sequence of a glycoprotein determines its hydrophobicity, conformation, and charge, amongst other properties [88].

The primary structure (amino acid sequence) of a glycoprotein can determine intermolecular, monomeric association, and aggregation propensity [49]. Even minor sequence differences can cause conformational differences leading to product variants with different biophysical attributes and virus filter fouling propensity. Inadvertent substitution of hydrophilic amino acids with hydrophobic amino acids or vice versa in the polypeptide sequence amplifies sequence variants.

mAbs, antibody fragments, bispecific antibodies, and Fc-fusion proteins are expressed in mammalian cells such as Chinese Hamster Ovary (CHO) cells for pharmaceutically relevant glycosylation profiles [16]. Flynn et al. reported that a typical CHO cell culture batch of mAbs has three major glycan species present, and they are G0F, G1F, and G2F [90]. These three dominant glycan structures are dependent on cell lineage and culture parameters [91]. E. coli expresses mostly insoluble, non-glycosylated variants [75]. Hybridomas offer a rapid expression template for initial product manufacture [16,92,93].

During cell culture and harvesting operations, expressed glycoproteins are usually not uniformly glycosylated [38,94,95]. Glycoproteins are expressed with a range of glycosylation profiles depending on cell culture conditions [96,97,98,99,100,101]. Micro-heterogeneity of glycoproteins can occur as a result of differences in glycosylation and other post-translational modifications. Variations in appended glycans introduce charge heterogeneity to the product monomer and determine the glycoprotein’s native fold state, aggregate susceptibility, and stability [102,103,104,105]. These product variants can affect the performance of virus filters.

Glycans are hydrophilic oligosaccharide moieties typically appended to glycoproteins in the cell during glycoprotein synthesis [106]. Glycans assist proper folding of the polypeptide chain before product secretion [74,94,107]. Most therapeutic proteins are glycoproteins. Glycoforms of protein products introduce structural heterogeneity, which affects their affinity to substrates, their stability, and other physicochemical characteristics of these therapeutic proteins [106,108,109]. Even in the same cell culture batch, a range of glycoforms occur [106,110,111]. Glycoforms occur due to skipped glycosylation sites on the glycoprotein or differences in the structure of appended glycans [94].

Glycan type and abundance can alter the product’s biophysical properties. Several studies have looked at the stability of different mAb glycoforms. These results show that aggregation is more prevalent in unglycosylated mAbs since glycans modulate aggregation [106,112]. Furthermore, a study showed that in terms of physical stability between pH 4–6, di-glycosylated IgG1-type mAbs were the most stable, and mono-glycosylated IgG1 was the least stable [113]. Post-translational modification can strongly affect the pI of a glycoprotein [74,106]. Variations in the pIs of product variants can affect hydrophobic and electrostatic interactions.

This entry is adapted from 10.3390/bioengineering9040155