Reply to Werner and Nimmerjahn letter to the Glyco-Forum

In a Letter to the Glyco-Forum (Werner and Nimmerjahn 2022), Anja Werner and Falk Nimmerjahn argue our studies reporting that ablation of ST6Gal1 in B cells (Jones et al. 2016) and/or hepatocytes (Oswald et al. 2022) failed to impact overall IgG sialylation are flawed. They further point to their findings claiming to show that IgG sialylation in the absence of B cells is insignificant and restricted to the F(ab) domain (Schaffert et al. 2019), ultimately concluding that B cells are responsible for IgG sialylation, not a B cell-extrinsic mechanism as we and others have proposed (Jones et al. 2012, 2016). I believe that Nimmerjahn and colleagues not only ignore important experiments in our prior study that directly address their central concern (Fig. 3 in Oswald et al. 2022) but also misinterpret their own data and misunderstand the implicit expectations that accompany the extrinsic sialylation model.

Werner and Nimmerjahn are correct in pointing out that the B cell-specific knockout of ST6Gal1 (BcKO) mouse we created and used in 2 studies is not 100% complete (Jones et al. 2016; Oswald et al. 2022). Indeed, no Cre-Lox murine model is perfect and we were specifically worried that any remaining ST6Gal1⁺/SNA⁺ cells could account for the lack of difference in IgG sialylation. In our original study (Jones et al. 2016), we immunized and boosted wild-type and BcKO mice and then sorted the antigen-primed B cells by SNA fluorescence-activated cell sorting to remove any lingering SNA⁺ cells and to isolate a purified SNA⁻ population. These cells were adoptively transferred into B cell-deficient recipients and the recipients boosted with antigen. We found that sialylation of both bulk (Fig. 3B in Jones et al. 2016) and antigen-specific IgG (Fig. 3D in Jones et al. 2016) was indistinguishable from controls in which recipients were given donor SNA⁺ B cells from immunized wild-type animals. This approach eliminated potential contaminating ST6Gal1⁺ B cells from confounding our results and helped to ensure that the IgG isolated was primarily from donor cells due to their antigen-primed state. In strong support of our in vivo findings, we have also discovered that not only do primary murine B cells produce predominantly asialylated IgG ex vivo, but that in cultured antibody-producing cells, IgG sialylation appears to be severely limited by an intracellular trafficking pattern that limits exposure to the ST6Gal1 enzyme itself (Glendenning et al. 2022).

In their 2019 article (Schaffert et al. 2019), Nimmerjahn and colleagues injected both B cell-deficient and ST6Gal1 global knockout mice with either 10-mg neuraminidase-treated human IVIg or 1-mg monoclonal murine antibody cross-reactive against human and mouse TRP1. After 2, 4, and 6 days, IgG was harvested and their associated N-glycans analyzed. Remarkably, their data showed clear mono- and di-sialylation events on these antibodies in a B cell-independent and ST6Gal1-dependent fashion. They demonstrated that the addition of sialic acids is time-sensitive, still increasing after 6 days, and represented over 2% of the total glycans in some experiments—which translates into as much as 200 μg of IgG having been sialylated in vivo in just a few days (Schaffert et al. 2019). These data robustly support our model of B cell-extrinsic IgG sialylation.

In a subsequent analysis of 2 samples using a different analytical platform, they also claimed that all of the changes were localized to glycans in the F(ab) domain (Schaffert et al. 2019). While it is conceivable that fully exposed F(ab)-localized N-glycans could be more efficient substrates for ST6Gal1 compared with Fc-localized glycans, the ability of ST6Gal1 to sialylate Fc glycans has been shown both in vitro (Barb et al. 2009) and in vivo (Pagan et al. 2018). There is no basis to claim or expect that only F(ab) glycans are accessible to ST6Gal1—if IgG is sialylated by ST6Gal1 outside of the B cell, both sites are available. The lack of detectable Fc glycan sialylation was likely due to their inadequate experimental design. First, their data are based on 2 replicates, which is underpowered. Second, the purity of their F(ab) fragments was neither checked nor reported. Third and most importantly, the analysis was limited to a single time point at day 4. Indeed, their entire study does not account for the time needed to reach equilibrium, as partially evidenced by the still increasing sialylation after 6 days (Figs 3A and B and 5A in Schaffert et al. 2019). As a result, their findings are at best an underestimate of the degree of glycan remodeling possible.

The B cell-extrinsic IgG sialylation model requires that upon release, IgG is somehow exposed to the appropriate conditions to allow ST6Gal1 to add α2,6-linked sialic acids (e.g. Neu5Gc and Neu5Ac in mice) (Lewis et al. 2022; Oswald et al. 2022) to IgG glycans. Since we have now ruled out the plasma microenvironment itself (Oswald et al. 2022), it must follow that the bulk of IgG will not be exposed to these conditions simultaneously, making sialylation and the achievement of equilibrium to 10–15% total sialylation a relatively slow and iterative process, and dependent on the absence of inflammatory stimuli. Consistent with this interpretation, mono-sialylated IgG glycans progress to di-sialylated species between 2 and 8 weeks postimmunization in mice (Fig. 9 in Oswald et al. 2022). Four days is simply not enough time, especially if the Fc glycans are at all less efficient as ST6Gal1 substrates.

Finally, Nimmerjahn and colleagues assume in their study that the injection of either desialylated human IgG at 400 mg/kg or desialylated autoantibody (antihuman and antimouse TRP1) at 40 mg/kg is completely inert with respect to the extrinsic IgG sialylation pathway. There is no evidence that these are inert; however, there is evidence that desialylating an autoantibody prior to injection can be pathogenic and cause an inflammatory response (Rademacher et al. 1994). Given that decades of research show that inflammation leads to decreases in sialylation, among other changes in IgG glycans (reviewed in Cobb 2020), this is another reason to believe that their results mimicking or even inducing inflammation with injected desialylated IgG (Schaffert et al. 2019) yield underestimated degrees of extrinsic sialylation.

In summary, our findings and the model of B cell-extrinsic IgG sialylation speak for themselves. I will leave others to make their own judgments based on the primary literature as I continue to hope that the mechanism(s) underlying IgG sialylation will someday be discovered. I believe that it holds exciting promise for a new class of anti-inflammatory therapies.

Funding

This Glyco Forum response was supported by a National Institute for General Medical Science grant to BAC (GM115234).

Conflict of interest statement: None declared.