Pathogen-driven degradation of endogenous and therapeutic antibodies during streptococcal infections

Abstract

Group A streptococcus (GAS) is a major bacterial pathogen responsible for both local and systemic infections in humans. The molecular mechanisms that contribute to disease heterogeneity remain poorly understood. Here we show that the transition from a local to a systemic GAS infection is paralleled by pathogen-driven alterations in IgG homeostasis. Using animal models and a combination of sensitive proteomics and glycoproteomics readouts, we documented the progressive accumulation of IgG cleavage products in plasma, due to extensive enzymatic degradation triggered by GAS infection in vivo. The level of IgG degradation was modulated by the route of pathogen inoculation, and mechanistically linked to the combined activities of the bacterial protease IdeS and the endoglycosidase EndoS, upregulated during infection. Importantly, we show that these virulence factors can alter the structure and function of exogenous therapeutic IgG in vivo. These results shed light on the role of bacterial virulence factors in shaping GAS pathogenesis, and potentially blunting the efficacy of antimicrobial therapies.

Characterizations of a neutralizing antibody broadly reactive to multiple gluten peptide:HLA-DQ2.5 complexes in the context of celiac disease

Article Open access 22 December 2023

Yuu Okura, Yuri Ikawa-Teranishi, … Tomoyuki Igawa

The cGAS–STING pathway as a therapeutic target in inflammatory diseases

Article 08 April 2021

Alexiane Decout, Jason D. Katz, … Andrea Ablasser

Distinguishing features of long COVID identified through immune profiling

Article Open access 25 September 2023

Jon Klein, Jamie Wood, … Akiko Iwasaki

Introduction

Streptococcus pyogenes, also known as Group A streptococcus (GAS), is a human specific pathogen that causes a broad spectrum of both local and systemic infections^1,2,3. The global burden of GAS diseases has been estimated to ~18 million severe cases and ~500,000 deaths each year, constituting a substantial source of morbidity and mortality worldwide⁴. GAS infections display significant heterogeneity regarding tissue tropism, disease severity, and occurrence of post-infectious sequelae. For example, GAS is responsible for a variety of relatively mild and localized skin and throat infections that result in impetigo and pharyngitis. These localized infections are often self-limiting and responsive to antimicrobial treatments, but in some cases, patients may progress into life-threatening invasive conditions, including sepsis and necrotizing fasciitis^5,6. GAS infections are also linked to a wide array of severe postinfectious immune-mediated disorders, such as glomerulonephritis and acute rheumatic fever. The basis for this large disease heterogeneity is poorly understood. Defining the underlying factors that modulate bacterial virulence in different disease contexts is instrumental to the development of more sensitive diagnostics and more effective targeted therapeutics.

GAS ability to cause disease is dependent on its capacity to subvert host defenses and to evade immunosurveillance. In particular, GAS has evolved multiple mechanisms to target the structure and function of host immunoglobulin G (IgG), in order to circumvent antibody-mediated responses. For example, GAS surface proteins can sequester antibodies by binding to the IgG Fc-region, preventing bacterial opsonization and phagocytosis via Fcγ-receptor expressing immune cells⁷. This "non-immune" IgG binding by the bacteria is dependent on the local IgG concentration of specific tissue environments, suggesting that the ability of GAS to counteract adaptive immune responses might influence both pathogen tropism and disease phenotypes⁸. Furthermore, GAS expresses at least two proteases, IdeS⁹, which specifically cleaves IgGs in the hinge region, and SpeB¹⁰, a broad spectrum protease that degrades Ig antibodies, as well as other protein targets, at multiple sites. Both enzymes reduce the capacity of IgG to elicit downstream Fc-dependent effector functions. In addition, GAS secretes EndoS¹¹, a bacterial glycan hydrolase that binds to IgG and removes the conserved asparagine N-linked oligosaccharide invariably attached to the Fc heavy chains of both human and murine antibodies. Deglycosylation induces conformational changes that reduce the IgG binding affinity for Fcγ-receptors, which in turn abrogates Fc-mediated protective functions¹². The fact that GAS has evolved so many different systems to cope with host IgG responses highlights the critical importance of antibodies in mediating protection against streptococcal infections. Moreover, the expression of IgG targeting virulence factors is also a potential concern for the future development of a GAS vaccine, as well as for the evaluation of the efficacy of intravenous Immunoglobulin (IVIG) therapy, a pharmaceutical mixture of polyclonal IgG derived from thousands of individuals, that is advocated as a promising adjuvant therapy for severe streptococcal diseases^13,14,15.

We have previously reported that GAS infections are associated with IgG glycan degradation in both humans and mice¹⁶. Interestingly, the degree of deglycosylation was found to be more pronounced at the local site, in patients with superficial skin and throat infections, compared to circulating IgG in the plasma of systemically infected septic patients. These differences suggest that EndoS activity might be modulated by signals derived from the tissue environment and/or associated with progression of invasive disease. However, the exact mechanisms that regulate the activity of EndoS are poorly understood, as well as their contribution to clinically relevant transitions, such as the switch from a local to a life-threatening systemic infection. The interaction of EndoS with other IgG-targeting mechanisms, and their combined impact on therapeutic interventions, such as IVIG therapy, also remain unaddressed. In this study, we provide evidence for the critical role of the host microenvironment in regulating the ability of GAS to induce changes in host IgG homeostasis during an ongoing infection. Homeostatic imbalance is mediated by the simultaneous secretion of IdeS and EndoS into circulation, which in turn results in widespread proteolytic and glycan degradation of both endogenous murine IgG, as well as exogenously administered therapeutic antibodies. More importantly, IgG degradation is modulated by the route of infection, highlighting the critical role of the host microenvironment in shaping both GAS virulence and the therapeutic efficacy of antibody-based antimicrobial treatments.

Results

Streptococcal EndoS deglycosylates all IgG subtypes in a murine model of disseminating GAS infection

We have previously reported the accumulation of deglycosylated IgG in the body fluids and tissues of GAS-infected patients, and shown that this can be partially recapitulated in a murine model of disseminating GAS infection¹⁶. In this previous study, IgG deglycosylation was monitored through targeted selected reaction monitoring (SRM) mass spectrometry (MS), resulting in the absolute quantification of the most abundant truncated glycoform of murine IgG1 at 24 h post infection (p.i.), a single N-acetylglucosamine (GlcNAc) linked to the amide nitrogen of the asparagine residue 297. The specificity of this EndoS-mediated phenotype was confirmed by control infection with isogenic mutant bacteria. However, the high sensitivity but reduced scope of this approach, limited by the scarce availability of commercial glycopeptide standards, came at the expense of losing track of potential changes occurring on other IgG isoforms. To generate a more comprehensive view of all ~30 different glycoforms that normally modify the 4–5 different IgG subtypes circulating in mouse plasma, we took advantage of the versatility of label-free glycoproteomics analysis, using data-dependent acquisition (DDA) mass spectrometry and stepped high-energy collisional dissociation (sHCD), which effectively resolves site-specific IgG glycopeptide differences¹⁷. Here, C57BL/6 J mice were subcutaneously inoculated with 2 × 10⁵ colony forming units (CFU) of the highly virulent M1 strain AP1, which causes a local skin infection that becomes fully systemic by 24 h, and results in ~50% mortality by 36 h p.i. To assess the extent and time-dependency of the IgG deglycosylation in this model, plasma from infected mice was collected at 0 h, 12 h, 24 h and 36 h after inoculation. Plasma IgG was purified on protein G columns using an automated liquid handling platform, digested into peptides, and subjected to glycoproteomics analysis.

As expected, the data confirmed that GAS infection triggers profound time-dependent IgG glycan remodeling in murine plasma, which results in the generation of truncated glycoprotein products derived from the Fc-region of several murine IgG subclasses and containing a single GlcNAc, with or without a core fucose (Fuc) (Fig. 1a–e). However, infection with an isogenic EndoS mutant GAS strain reduced the truncated glycopeptides to baseline levels confirming that deglycosylation was mainly driven by the EndoS activity (Fig. 1c–f, g). In uninfected conditions, murine IgG was modified with complex-type biantennary N-linked glycans with variable degrees of galactosylation, fucosylation and sialylation, that slightly differed across IgG subclasses, with IgG2b and IgG2c containing more galactose than IgG1 and IgG3 (Fig. 1h–k). Confirming previous reports, murine IgG had trace levels of bisecting GlcNAc^18,19, and terminal sialylation was overwhelmingly accounted for by N-glycolylneuraminic acid (NeuGc) monosaccharides, as opposed to the exclusive expression of N-acetylneuraminic acid (NeuAc) of human glycoproteins, due to the inactivation of cytidine monophospho-N-acetylneuraminic acid hydroxylase (CMAH) during human evolution²⁰. In contrast, the predominant IgG glycoforms of infected plasma were truncated glycoprotein products, a cleavage pattern that is consistent with the catalytic mode of action of streptococcal EndoS (Fig. 1h–k). Truncated glycopeptides accounted for a small fraction of the total glycopeptide intensity of each IgG subclass by 12 h (IgG1: 2.5%, IgG2b: 4.0%, IgG2c: 4.4% and IgG3: 11.7%). However, these products significantly accumulated by 24 h (IgG1: 46.5%, IgG2b: 68.4%, IgG2c: 66.0% and IgG3: 88.2%), resulting in an almost complete deglycosylation of the entire murine IgG pool by 36 h p.i. (IgG1: 87.1%, IgG2b: 97.2%, IgG2c: 94.3% and IgG3: 94.2%).

**Fig. 1: Glycosylation analysis of murine IgGs during subcutaneous GAS infection.**

Multiple strategies to directly measure EndoS in infected mouse plasma, including targeted and untargeted mass spectrometry analysis, were unsuccessful, suggesting that the enzyme circulates at low levels. However, the presence of EndoS can be probed by enzymatic assays due to the efficient catalytic activity of the enzyme. To assess the direct involvement of EndoS in the IgG phenotype observed, we performed overnight digestion of uninfected human and murine plasma with recombinant EndoS in vitro. EndoS digestion resulted in a similar deglycosylation pattern as the one observed in vivo, with ~79% of human and ~93% of mouse IgGs being hydrolyzed into EndoS truncated glycoprotein products (Supplementary Fig. 1). The remaining intact IgG glycoforms were mostly assigned to high-mannose N-linked glycans, which are known to be resistant to EndoS activity²¹. Next, we passed infected plasma (36 h p.i.) through Protein G columns and the flow-through containing the non-IgG glycoproteins was collected and subjected to glycoproteomics to determine whether other glycoprotein substrates were also targeted by the deglycosylation activity triggered by GAS. Glycoproteomics analysis identified 246 glycopeptides derived from 48 abundant plasma glycoproteins with a broad spectrum of complex-type bi- and tri-antennary N-glycan structures. Most glycoproteins were decorated with terminal NeuGc modifications, although a small fraction of NeuAc-containing glycans as well as high-mannose structures were also observed (Supplementary Fig. 2 and Supplementary Table. 1). Importantly, the typical signature of EndoS mediated glycan truncation was only observed on IgG, confirming the high substrate specificity that has been reported and structurally elucidated for EndoS²². Taken together, the data is consistent with GAS-infection triggering the time-dependent deglycosylation of all murine IgG subtypes, which results in the plasma accumulation of truncated glycoprotein products with the typical cleavage pattern of streptococcal EndoS activity.

IdeS activity contributes to alterations in IgG homeostasis

To understand whether glycan degradation is paralleled by other changes in the structure and function of IgGs and/or other murine proteins, the plasma samples taken at different time points (0 h, 12 h, 24 h, and 36 h p.i.) were further subjected to quantitative proteomics analysis. The proteomics data revealed that disease progression was associated with significant changes in the levels of several proteins linked to coagulation and inflammation, and with alterations in the abundance of factors involved in cellular and humoral responses (Supplementary Fig. 3). Principal Component Analysis (PCA) stratified the plasma proteome changes into two distinct stages, one early stage (0 h and 12 h) and one late stage (24 h and 36 h) (Fig. 2a). The early stage was driven by the activation of the acute phase response and other inflammatory proteome changes, whereas the late stage was characterized by the plasma accumulation of several markers of tissue damage, neutrophil proteins, and Proteoglycan 4 (PRG4), a protein that has recently been reported to accumulate both in the plasma and on the injured vasculature of a murine model of Staphylococcus aureus (S. aureus) sepsis^23,24.

**Fig. 2: IdeS mediates changes in murine IgG homeostasis in a subcutaneous model of local to systemic GAS infection.**

Surprisingly, in addition to the expected decrease in the molecular weight of the IgG heavy chains due to deglycosylation (Fig. 2c–g), disease progression in the mouse model was also associated with a significant reduction in the levels of IgG3 (Fig. 2b). Quantification of the individual IgG subtypes in plasma and liver tissues showed that the reduction in IgG concentration was in fact restricted to IgG3 (Fig. 2h), a phenotype that was specifically linked to GAS infection since S. aureus bacteremia did not induce changes in IgG levels, neither in the plasma nor in the organs (Fig. 2i). The hinge region of murine IgG3 is reportedly susceptible to proteolytic digestion by streptococcal IdeS²⁵. It has also been reported that IdeS degradation of human IgGs is rapidly followed by IgG clearance from circulation²⁶. Based on these previous observations, we hypothesized that IdeS may also be responsible for the reduced levels of IgG3 in the mouse model. Overnight digestion of human and mouse plasma with recombinant IdeS revealed that unlike the general proteolytic susceptibility of human IgGs, IgG3 is the only murine IgG expressed by C57BL/6 J mice that can be partially cleaved in the hinge region in vitro (Supplementary Table. 2). This differential susceptibility is linked to distinct amino acid sequence variations in the hinge regions of the IgG subtypes in humans vs mice (Fig. 2j). To determine whether IdeS activity is responsible for the observed reduction of IgG3 in the GAS model, we used mass spectrometry to quantify the levels of the expected IdeS specific proteolytic reporter fragments of the IgG hinge regions in mouse plasma over the course of infection. Intact IgG3 hinge region peptides were readily measured in uninfected as well as infected samples up to 12 h p.i. (Fig. 2k). However, their abundances significantly decreased by 24 h and 36 h p.i., paralleled by an increase in IdeS-specific cleavage products. These data suggest that IdeS might indeed be responsible for the proteolytic degradation and systemic reduction of IgG3 in this mouse model of disseminating GAS infection. More importantly, they highlight fundamental differences between mouse and human IgGs in terms of their amino acid sequences and their sensitivity to bacterial proteolytic factors. Collectively, our results demonstrate that disease progression in the GAS model is paralleled by profound alterations in the expression of multiple plasma proteins over time, and a significant IdeS- and EndoS-dependent remodeling of IgG, resulting in proteolytic degradation of IgG3 and complete deglycosylation of the Fc region of all murine IgG subtypes. Finally, although the plasma levels of IdeS and EndoS were below the limit of detection of the mass spectrometry assays, both enzymes were unambiguously identified in murine skin at the site of infection (Supplementary Fig. 4), providing direct evidence for their in vivo expression during the ongoing infection.

EndoS and IdeS mediate degradation of therapeutic antibodies in vivo

Both EndoS and IdeS are highly conserved across all GAS genomes sequenced to date, indicating that these enzymes might be important for GAS virulence. In previous work, we have demonstrated that EndoS expression interferes with the therapeutic efficacy of active immunization against the protective streptococcal antigen M1¹⁶, but its impact on the efficacy of passive immunization remains unexplored. Given the observed glycan and proteolytic degradation of endogenous IgG in the mouse model, we hypothesized that upregulation of EndoS and IdeS activities might also result in cleavage and inactivation of exogenously administered IgG. To test this hypothesis, we probed the ability of plasma derived from GAS infected mice to degrade IVIG, a pharmaceutical-grade IgG mixture derived from thousands of human donors, and therefore representative of different IgG subclasses and glycosylation patterns. We have previously shown that IVIG contains antibodies that mediate GAS opsonophagocytosis, at least in vitro²⁷. Also, based on their distinct amino acid sequences in the Fc-region, murine and human IgG can be separately quantified using mass spectrometry, facilitating the analysis of complex samples containing IgG from both species. As shown in Fig. 3a–c, EndoS truncated human Fc-glycopeptides were abundantly detected in infected mouse plasma (36 h p.i.) after ex-vivo overnight incubation with human IVIG, indicating that the circulating levels of EndoS were sufficient to robustly deglycosylate exogenous human antibodies.

**Fig. 3: EndoS drives deglycosylation of human IVIG.**

To investigate whether EndoS-driven deglycosylation of IVIG might also occur in vivo, mice were pretreated with IVIG followed by subcutaneous challenge with GAS (Fig. 3d–h). Plasma samples were collected at 36 h p.i., and subjected to IgG enrichment, followed by glycoproteomics analysis. The results showed that EndoS abundantly deglycosylates the administrated IVIG in vivo during GAS infection, hydrolyzing ~50–70% of the N-glycans on both human IgG1 and IgG2 (Fig. 3d, e, g). In contrast, infection with an isogenic AP1 EndoS mutant strain¹¹ prevented the occurrence of EndoS specific truncated glycopeptides and resulted in maintained intact glycan structures on both human IgG1 and IgG2 (Fig. 3f, g), similar to the glycan structures observed in the uninfected controls. These results demonstrate that EndoS is responsible for the IgG deglycosylation phenotype observed in this model. Notably, despite the marked differences in IgG glycan modifications in the infection with wildtype bacteria compared to isogenic mutant, the total IVIG levels in plasma remained unchanged (Fig. 3h), suggesting that glycosylation does not regulate the half-life of circulating human IgG during infection. Finally, plasma samples were also subjected to mass spectrometric quantification to assess the presence of IgG products consistent with IdeS proteolytic processing of human IVIG. Accordingly, we identified abundant IdeS cleaved peptides of human IgG in the plasma samples of infected IVIG-pretreated mice, but not in uninfected animals (Supplementary Fig. 5), confirming that both glycosidic and proteolytic mechanisms can simultaneously act on exogenously administered antibodies, leading to substantial remodeling of circulating IVIG.

GAS virulence in the disseminating infection model is not attenuated by EndoS inactivation or IVIG treatment

Both preclinical and clinical studies have shown variable results regarding the therapeutic benefits of IVIG therapy to treat GAS infections^13,14,28. In our model of disseminating GAS infection, we found that IVIG treatment did not confer protection against the wildtype AP1 strain or the EndoS isogenic mutant. IVIG pre-treatment had a negligible impact on disease progression independently of the bacterial strain used for infection, with both treated and untreated animals displaying significant weight loss from 24 h p.i.(Fig. 4a). Furthermore, the absence of EndoS had no impact on the development of leukopenia or bacterial load in distant organs, independently of IVIG pre-treatment (Fig. 4d–f). Additionally, the plasma proteome changes triggered by the infection with the EndoS mutant were virtually indistinguishable from the changes induced by wildtype bacteria (Fig. 4b, c). These results suggest that despite the significant IgG perturbations induced by GAS infection, genetic inactivation of EndoS or pre-treatment with IVIG is not sufficient to prevent disease progression in this mouse model. In an attempt to explain these results, we investigated whether the used IVIG preparation contained IgG antibodies against GAS antigens, and if antigen-specific IgG could activate cellular signaling through Fcγ-receptors. To this end, we used enzyme-linked immunosorbent assay (ELISA) to show that IVIG contains IgG against several GAS-specific antigens (Fig. 4g–i), and a luciferase reporter cell system, to further demonstrate that these antibodies were capable of robustly activating cellular signaling through Fcγ-receptor IIa (CD32) and IIIa (CD16) in an antigen-specific manner (Fig. 4j, k). As expected, this activity was completely abrogated by pre-incubating IVIG with either recombinant IdeS or EndoS, highlighting the ability of both enzymes to block Fcγ-dependent functions, such as antibody-dependent phagocytosis (ADP) and antibody dependent cellular cytotoxicity (ADCC). Based on these results, we concluded that IVIG, despite containing functional GAS-specific IgGs with the capacity to trigger downstream effector functions, did not result in therapeutic benefit, at least in this infection model.

**Fig. 4: GAS virulence in the subcutaneous infection model is not impacted by IVIG treatments or EndoS-inactivation.**

The route of bacterial inoculation impacts both GAS infection and IVIG therapeutic efficacy

Despite this apparent lack of therapeutic benefit of IVIG in our infection model, we noticed that both the severity of the infection, as well as the accumulation of IdeS and EndoS cleavage products, became more pronounced towards 24 h p.i., concomitantly with the transition from a contained local, to a systemic infection. We hypothesized that the host tissue environments (skin vs blood) might have an impact on GAS infection, for example by regulating bacterial virulence and susceptibility to treatment. To test this notion, mice were inoculated with the AP1 strain intraperitoneally (IP), to bypass the local skin stage of the infection, and to recreate a single stage systemic infection, as opposed to a two-stage local to systemic infection model. IVIG pre-treatment was also added to one group to evaluate the impact of the route of infection on the therapeutic efficacy of IVIG. Blood and tissues were harvested at 24 h p.i., and subjected to mass spectrometry-based proteomics and glycoproteomics analysis. These analyses revealed the absence of an imbalance in IgG homeostasis in the single stage systemic IP model compared to the two-stage subcutaneous model, with no observable differences in Fc-glycan patterns (Fig. 5a–c) or IgG3 levels (Fig. 6g) between uninfected and infected animals. Additionally, there was no difference in the glycan pattern of the injected human IVIG (Fig. 5d). To rule out endogenous inhibitory factors in plasma that might interfere with the activity of the bacterial enzymes, we look for the presence of human antibodies against EndoS and IdeS in the IVIG formulation. Although both anti-EndoS and anti-IdeS antibodies were detected in IVIG, their partial degradation in the subcutaneous model already indicated that they were of insufficient titers or activity to neutralize the enzymes (Supplementary Fig. 6). Moreover, the total absence of neutralizing murine antibodies triggered by the infection was also ruled out by ELISA measurements, which is also consistent with the rapid disease progression and the lack of adaptive immune responses in both models. Additionally, spiking in recombinant EndoS into IP infected blood resulted in abundant glycan degradation, ruling out the presence of other endogenous inhibitors (Supplementary Fig. 7). On the other hand, this lack of EndoS and IdeS enzymatic activity was consistent with a reduction in the mRNA expression of both EndoS and IdeS in the infected splenic tissues of the IP model compared to the subcutaneous model, pointing towards reduced enzyme expression linked to the route of infection as the most likely explanation for the absence of enzymatic activity in the IP model (Fig. 5e).

**Fig. 5: EndoS is not catalytically active in the murine model of intraperitoneal (IP) GAS infection.**

**Fig. 6: Murine GAS infection and IVIG therapeutic efficacy is modulated by route of infection.**

In sharp contrast to the results obtained in the two-stage subcutaneous model, pre-treatment of the single stage IP model with IVIG substantially reduced the levels of splenic colonization, with a similar trend also observed for other organs such as the liver, but without having a significant effect on weight loss (Fig. 6a–c). Further therapeutic benefit in the single stage IP model was evidenced by a significant reduction in leukocytopenia by 24 h p.i. (Fig. 6d). Plasma proteome analysis of infected animals revealed that multiple proteins were differentially regulated by the IVIG treatment compared to the untreated infected group (Fig. 6e). These proteome changes were linked to an increase in the plasma levels of liver-derived intracellular metabolic proteins that can leak out to plasma due to inflammation and tissue damage (Fig. 6f). This effect on the plasma proteins was selective since IVIG treatment did not affect the induction of acute phase proteins or vascular activation (Fig. 6h), but significantly reduced the levels of circulating markers of liver damage (Fig. 6i), pointing towards a potential hepatoprotective role for IVIG in this model. In conclusion, our results suggest that the transcriptional status of the bacteria is impacted by the route of infection, which directly implicates the tissue microenvironment in the modulation of GAS pathogenesis through the regulation of the expression of specific virulence factors such as IdeS and EndoS in vivo (Fig. 7). More importantly, the data also highlights how the progression of GAS infection, as well as the therapeutic efficacy of antibody-based treatments, are both fundamentally shaped by the complex interplay between host and pathogen factors.

Salmonella Treatment