Fc glycans of therapeutic antibodies as critical quality attributes

β1,2-xylose and α1,3-fucose

Plants are attractive hosts for the manufacture of recombinant protein therapeutics as they are relatively inexpensive systems that can be readily scaled up. However, plant-derived monoclonal antibodies contain complex N-glycans containing β(1,2)-xylose and α1,3-fucose residues not present in humans and are therefore regarded as “carbohydrate cross-reactive determinants” in IgE from sera of allergic patients (Tekoah et al. 2004). The glycan-specific IgE show 23% prevalence and have been found at levels up to 71% in individuals with multiple pollen sensitivity. However, the clinical relevance of these reactive IgE remains unclear. Strasser et al. used RNAi to successfully silence XylT, the gene coding for β1,2-xylotransferase, in tomato plants and demonstrated a patient-specific reduction in IgE reactivity in their studies (Paulus et al. 2011). Arabidopsis thaliana knockouts of the XylT (encoding β1,2-xylotransferase) and FucT (encoding α1,3 fucosyl transferase) genes have also been generated that are viable with no obvious phenotype. This represents a path forward in the production of protein biotherapeutics with human type N-glycosylation (Strasser et al. 2004). In conclusion if plants are used to produce therapeutic antibodies and the transferases are not silenced, β1,2-xylose and α1,3-fucose should be taken into account as CQAs.

N-glycolylneuraminic acid

Glycans of therapeutic proteins that are produced in SP2/0, NSO and, to a much lesser extent, in CHO cell lines are often modified with the non-human sialic acid N-glycolylneuraminic acid (Neu5Gc; NGNA) (Hokke et al. 1990; Noguchi et al. 1996). Humans do not synthesize NGNA as the gene encoding CMP-N-acetylneuraminic acid hydroxylase, which produces CMP-NGNA from CMP-N-acetylneuraminic acid (NANA), is mutated (Varki 2007). It was previously thought that healthy humans did not react to NGNA; however, more recent findings suggest that all humans have anti-NGNA antibodies, some at levels up to 0.1–0.2% of circulating IgG. This may be due to dietary exposure, and the subsequent incorporation of NGNA into human proteins (Tangvoranuntakul et al. 2003; Padler-Karavani et al. 2008). Padler-Karavani et al. (2008) suggested that the ongoing antigen–antibody reaction may generate chronic inflammation, possibly contributing to the high frequency of diet-related carcinoma and other diseases in humans.

It has been demonstrated that many humans produce antibodies against oligosaccharide structures with terminal NGNA (Zhu and Hurst 2002). The potential impact of NGNA was evaluated by Durocher and Butler (2009) who indicated that this immunogenicity can reduce efficacy due to rapid clearance of the biotherapeutic, or alternatively, prevent drug re-administration due to an undesirable immune response. As these ATAs represent the first obstacle to xenotransplantation in humans, they are clinically significant from a safety perspective (Milland and Sandrin 2006; Higgins 2010).

Biotherapeutics produced in human cell lines may also become contaminated with NGNA that is incorporated from animal-derived culture medium materials (Ghaderi et al. 2010). NGNA in cells is recycled in lysosomes and incorporated into glycoproteins (Bardor et al. 2005).

In conclusion, NGNA on Fc-glycans is potentially immunogenic and for this reason is most probably a CQA.

Terminal sialic acid

The anti-inflammatory activity of immunoglobulin has repeatedly been linked to Fc sialylation, which may lead to decreased ADCC activity and, in some cases, even affect target binding (Fukushima and Takasaki 1993; Kaneko et al. 2006; Scallon et al. 2007; Anthony, Nimmerjahn, et al. 2008; Anthony, Wermeling et al. 2008; Nimmerjahn and Ravetch 2008). In general, the amount of sialylated Fc glycans of therapeutic antibodies is very low (Jefferis 2006).

In conclusion if there is a significant amount of sialylated glycans, and ADCC is part of the MoA of the therapeutic antibody, the influence of sialylation on FcγRIIIa binding and ADCC should be determined in vitro as part of the CQA assessment.

α1,6-core fucose

Changes in the levels of afucosylated glycostructures on therapeutic antibodies may be of concern from a safety perspective. Toxicity due to off-target binding may also be a concern for antibodies with higher ADCC. Jiang et al. indicate that increased afucosylation leading to increased effector function is a potential safety concern for monoclonal antibodies with moderate or low effector function (Jiang et al. 2011).

Increased crosslinking of activating FcγRs due to higher levels of afucose may also be a safety concern due to the additional release of proinflammatory cytokines such as tumor necrosis factor (TNF)-α and IFNγ.

The lack of core fucose (i.e. the α1,6-linked fucose on the GlcNAc residue involved in the amide bond with the asparagine of the N-glycosylation site) on the carbohydrate moiety linked to the Fc region of an IgG molecule leads to a pronounced increase in ADCC via increased affinity for the FcγRIIIa expressed on immune cells such as NK cells and macrophages (Rothman et al. 1989; Umana et al. 1999; Shields et al. 2002; Niwa et al. 2004; Okazaki et al. 2004; Yamane-Ohnuki et al. 2004; Peipp et al. 2008; Chung et al. 2012). Additionally, the absence of core fucose leads to increased Fc-dependent binding to FcγRIII positive nonclassical/intermediate monocytes and macrophages, which translates into increased ADCP (Herter et al. 2014). In contrast, binding to FcγRI and II-positive and FcγRIII-negative classical monocytes, as well as ADCP-mediated by these cells, remain unchanged, irrespective of fucosylation levels (Herter et al. 2014).

While afucosylation increases ADCP in glycoengineered antibodies, the impact on ADCP due to small variations in afucosylation levels resulting from standard CHO-based manufacturing processes remains to be assessed. Finally, increased afucosylation has not been demonstrated to have any impact on CDC activity.

Studies by Junttila et al. highlighted the impact of afucosylation on pharmacological properties and half-life of an IgG1, relative to a wild-type control. In these studies, afucosylated trastuzumab displayed moderately altered pharmacokinetic dispositions compared with trastuzumab, slightly faster elimination from the circulation, and a modest reduction in half-life (Junttila et al. 2010). The authors hypothesized that the faster clearance may result from the differential bio-distribution that leads to the enrichment of antibodies to immune effector cell-rich organs due to the increased FcγRIII affinity. In conclusion concerning the scientific literature there is no clear evidence that afucosylated Fc glycans might be of concern for safety.

There is clear evidence that afucosylated Fc glycans increase ADCC and if ADCC is part of the MoA, afucosylated Fc-glycans are always likely to be a CQA concerning efficacy.

Bisecting GlcNAc

CHO cells lack the gene encoding GnTIII. Therefore antibodies produced in CHO cells lack detectable bisecting GlcNAc glycostructures (Durocher and Butler 2009). GnTIII overexpressed in antibody-producing cells catalyzes the addition of GlcNAc to N-linked oligosaccharides. The addition of bisecting GlcNAc inhibits core-fucosylation and conversion of hybrid to complex glycans (Ferrara et al. 2006).

In general, glycostructures containing bisecting GlcNAc are abundant in human polyclonal IgG and are not a considered a safety concern (Jefferis 2009a). Recently, Ritamo et al. (2013) showed that 7% of the Fc-glycans on intravenous IgG (IVIG) contain glycan structures with bisecting GlcNAc (analyzed at the glycopeptide level) (Ritamo et al. 2013). Therefore, it is unlikely that ATAs are produced against the bisecting GlcNAc epitope. In summary, bisecting GlcNac structures are not a safety issue.

Galactose

Highly galactosylated glycostructures on therapeutic antibodies may be of concern as they increase in vitro C1q-binding and CDC activity (Boyd et al. 1995; Tsuchiya et al. 1989). Increased galactosylation of CHO-derived antibodies has been shown to increase FcγRII and FcγRIIIa binding as measured using surface plasmon resonance (Ritamo et al. 2013). Additionally, it has been demonstrated that the increased galactosylation enhances ADCC activity (Jiang et al. 2011). However, the impact of galactosylation is subtle compared with the impact of afucosylation (Ritamo et al. 2013).

Several reports indicate that a high degree of galactosylation promotes increased binding to C1q leading to enhanced activation of the complement system and CDC activity (Hodoniczky et al. 2005; Jefferis 2009a; Nimmerjahn et al. 2007). Similar effects have been observed for the glycoengineered therapeutic antibody obinutuzumab in vitro, although CDC is not considered to be a relevant in vivo activity. Several reports state that the extent of terminal galactosylation does not affect ADCC in standard IgG1 antibodies (Boyd et al. 1995; Shinkawa et al. 2003; Hodoniczky et al. 2005). Another report found that terminal galactosylation has a positive effect on FcγRIIIa binding (Houde et al. 2010). We have recently demonstrated enhanced ADCC upon enzymatic hypergalactosylation of four different monoclonal antibodies produced using standard CHO manufacturing processes (Tejada et al. submitted; Thomann et al. 2015). We also showed that elevated levels of terminal galactose have no effect on ADCC of two different glycoengineered therapeutic antibodies (Thomann et al. 2015), indicating that the extent to which terminal galactose modulates ADCC depends on the background level of afucosylation. Lastly, we quantitatively compared the effects of galactosylation and afucosylation in the context of glycan heterogeneity to demonstrate that while galactose can influence ADCC activity, afucosylation remains the primary driver of this activity.

Karsten et al. (2012) recently reported an anti-inflammatory activity of IgG1 mediated by Fc galactosylation and association of FcγRIIb and dectin-1. The authors concluded that high N-glycan galactosylation of IgG1 molecules promotes cooperative signaling of the FcγRIIb with dectin-1, resulting in an inhibitory signaling pathway that blocks proinflammatory effector functions.

In conclusion if CDC is part of the MoA for a therapeutic antibody the impact of galactosylation levels reflecting the manufacturing experience should be considered as it relates to efficacy during CQA assessment. If ADCC is part of the MoA of the therapeutic antibody, the impact of galactosylation should be determined by in vitro studies as part of CQA assessment.

High mannose glycostructures

No data or direct evidence in the literature supports the immunogenicity of high mannose glycan structures (Pacis et al. 2011). Studies by Zhou et al. reported enhanced FcγRIIIa binding and ADCC in antibodies with oligomannose-type glycans generated using kifunensine (Yu et al. 2012). Kanda et al. (2007) generated core fucose-lacking human IgG1 antibodies with three different N-linked Fc oligosaccharides, namely, a high-mannose, hybrid and complex type, using the same producing clone. They demonstrated that the mannosylated glycostructures could induce ADCC via FcγRIIIa binding but to a lesser extent than the hybrid and complex structures. The same holds true for CDC and C1q binding. More recently, Yu et al. (2012) confirmed that high mannose glycoforms exhibited higher FcγRIIIa binding and ADCC activity as well as decreased C1q binding and CDC activity.

In general, core fucose is absent in the highly mannosylated glycoforms generated. Therefore, it remains unclear if the observed increases in FcγRIIIa binding and ADCC activity are due to the presence of multiple mannose moieties, or the absence of core fucose. The reduced C1q binding and CDC activity observed in these studies may also reflect the lack of terminal galactose in these highly mannosylated forms.

Recent studies show clear evidence for selective clearance of oligomannose forms of IgG1 and IgG2 most likely via a mannose receptor-mediated mechanism (Goetze et al. 2011; Alessandri et al. 2012; Yu et al. 2012). Goetze et al. (2011) used IgG1 and IgG2 in human studies to demonstrate that Man5 isoforms were preferentially removed from circulation, independent of the route of administration. In conclusion if there is a significant amount of mannosylated glycans and if ADCC is part of the mode of action of a therapeutic antibody, mannosylation may be a CQA concerning efficacy. Additionally, there is clear evidence that oligomannose bearing IgGs are faster cleared than others and therefore may be a CQA concerning PK/PD.