Antibody responses to collagen peptides and streptococcal collagen-like 1 proteins in acute rheumatic fever patients

ABSTRACT

Acute rheumatic fever (ARF) is a serious post-infectious immune sequelae of Group A streptococcus (GAS). Pathogenesis remains poorly understood, including the events associated with collagen autoantibody generation. GAS express streptococcal collagen-like proteins (Scl) that contain a collagenous domain resembling human collagen. Here, the relationship between antibody reactivity to GAS Scl proteins and human collagen in ARF was investigated. Serum IgG specific for a representative Scl protein (Scl1.1) together with collagen-I and collagen-IV mimetic peptides were quantified in ARF patients (n = 36) and healthy matched controls (n = 36). Reactivity to Scl1.1 was significantly elevated in ARF compared to controls (P < 0.0001) and this was mapped to the collagen-like region of the protein, rather than the N-terminal non-collagenous region. Reactivity to collagen-1 and collagen-IV peptides was also significantly elevated in ARF cases (P < 0.001). However, there was no correlation between Scl1.1 and collagen peptide antibody binding, and hierarchical clustering of ARF cases by IgG reactivity showed two distinct clusters, with Scl1.1 antigens in one and collagen peptides in the other, demonstrating that collagen autoantibodies are not immunologically related to those targeting Scl1.1. Thus, anti-collagen antibodies in ARF appear to be generated as part of the autoreactivity process, independent of any mimicry with GAS collagen-like proteins.

Rheumatic fever is an autoimmune disease caused by a StrepA infection. Patients with rheumatic fever have antibodies that react with collagen, which may contribute to disease symptoms. But these rogue autoantibodies are not the same as the antibodies that react with parts of the StrepA bacteria, which resemble collagen. This means the rogue collagen antibodies seen in patients are most likely a consequence of the body's immune system malfunctioning and not directly caused by the StrepA infection.

BACKGROUND

Acute rheumatic fever (ARF) is a serious, post-infectious sequela of a Group A streptococcal (GAS) infection most common in children 5–15 years of age. Symptoms include fever, polymigratory arthritis and carditis, the latter of which can develop into chronic rheumatic heart disease (RHD) (Carapetis et al. 2016). Globally there are approximately 33 million people living with RHD and over 300 000 deaths attributed to this chronic condition per annum (Watkins et al. 2017). While most of the ARF/RHD disease burden occurs in low-income countries, the disease persists in indigenous populations in high-income countries, such as Māori and Pacific children in New Zealand (Bennett et al. 2021).

Acute rheumatic fever pathogenesis remains poorly understood. The prevailing hypothesis is based on molecular mimicry where the alpha helical coiled-coil structure of the GAS M-protein elicits the generation of antibodies and T-cells that cross-react with human coiled-coiled proteins (cardiac myosin, laminin and tropomyosin) (Cunningham 2012; Carapetis et al. 2016). Anti-collagen antibodies may also contribute to disease progression. GAS colonization disrupts the extracellular matrix, which may expose cryptic collagen epitopes and trigger the generation of collagen autoantibodies (Tandon et al. 2013; Karthikeyan and Guilherme 2018). In support of a collagen mediated mechanism, elevated antibody titers to collagen I, the most abundant fibrillar collagen in the human body and collagen IV, an integral part of basement membranes, have been observed in ARF (Martins et al. 2008; Dinkla et al. 2009). Selected GAS M proteins contain a peptide motif (Peptide Associated with Rheumatic Fever, PARF) that has been shown to bind and disrupt collagen fibers in vitro and may contribute to cryptic collagen epitope exposure (Dinkla et al. 2009). However, a recent analysis of over 400 ARF associated GAS strains found just 4.1% harbored the PARF motif suggesting other mechanisms also contribute to the development of collagen autoantibodies in ARF (de Crombrugghe et al. 2020).

Collagen-like sequences have been detected in various bacterial and viral pathogens (Rasmussen, Jacobsson and Björck 2003), including GAS, which produces cell surface proteins with a collagenous domain known as streptococcal collagen-like (Scl) proteins. These proteins comprise a globular, N-terminal variable (V) domain that is projected away from the cell surface by an elongated collagen-like (CL) domain. The CL domain has a homo-trimeric triple helical structure with the repeating amino acid sequence of Gly-Xaa-Yaa, which broadly resembles human collagen (Lukomski et al. 2017). Human collagens are also comprised of a repeating Gly-Xaa-Yaa sequence, but harbor a high proportion of hydroxyprolines (∼38%) at the Y position, which is a major contributor to the thermal stability of the triplex helix (Mohs et al. 2007). GAS, like all prokaryotes, lacks the prolyl hydroxylase enzyme to convert proline to hydroxyproline and as such there is an absence of hydroxyprolines at the Y position in Scls. Nevertheless, Scl-CL triple-helices have thermal stability comparable to human collagen, and can be recognized by human collagen-binding integrin receptors (Caswell et al. 2008). There are two major types of GAS Scl proteins, Scl1 and Scl2. Though scl1 and scl2 genes are found in all strains tested, differences in expression regulation mean that Scl1 proteins are produced by all GAS strain, while Scl2 is variably expressed (Lukomski et al. 2017).

Previous studies have shown that Scl1 and Scl2 proteins elicit an IgG response in patients with invasive GAS disease and pharyngitis (Hoe et al. 2007), as do short peptides derived from the region between the CL domain and the cell wall anchor motif in patients with ARF (Chaudhary et al. 2017). However, studies that comprehensively examine the IgG response to Scl V and CL domains and compare this to collagen reactivity to investigate ARF pathogenesis are lacking. As such, the aims of this study were 2-fold: firstly, to explore antibody responses to triple helical type-I and type IV collagen peptides and Scl proteins in ARF patients; and secondly, to determine if there was any relationship between the antibody reactivity to human collagen and the Scl proteins given the similarities in triple helical structure. Previous studies that have examined collagen autoantibodies in ARF have made use of material purified from calf skin and human placenta (Martins et al. 2008; Dinkla et al. 2009). These preparations of collagen are likely to be heterogeneous and contain other extracellular matrix proteins. Here, synthetic collagen I and IV mimetic peptides have been utilized that offer the advantage of pure, homogeneous material that self-assembles into triple helical structures (Okuyama et al. 2012; Xiao et al. 2015).

MATERIALS AND METHODS

Study subjects

Human sera samples were obtained from the Rheumatic Fever Risk Factors (RF RISK) study conducted in New Zealand between 2014–2017 (Baker et al. 2019). This study included sera from participants recruited in the first 18 months (September 2014–April 2016) of enrolment. All participants (or their proxies) provided written informed consent and the protocols had appropriate ethical board approval (HDEC 14/NTA/53). Acute rheumatic fever patients were diagnosed according to the New Zealand modification of the Jones criteria (Atatoa-Carr et al. 2008) and healthy control samples were obtained from participants matched for age, ethnic identification, area-based measures of sociodemographic deprivation (as assessed by the New Zealand Deprivation Index score (Salmond et al. 2006)) and geographic area (district health board region in which the subject resides).

Recombinant Scl proteins

Recombinant streptococcal collagen‐like proteins (rScl) were produced in Escherichia coli using the Strep‐Tag II expression and purification system (IBA‐GmbH) as described (Xiao et al. 2015; Lukomski and McNitt 2020). Escherichia coli strains carrying pASK-IBA2 plasmids encoding full-length (FL) Scl1.1 protein (rScl1.1-FL), collagen-like (CL) domain (rScl1.1-CL) and N-terminal variable (V) domain (rScl1.1-V) from the emm1 GAS genotype (GenBank accession number AF252861) were fused to a C-terminal Strep-Tag II octapeptide motif (WSHPQFEK) as reported (Han et al. 2006) and utilized for periplasmic expression followed by affinity chromatography purification with Strep-Tactin Sepharose.

Collagen peptides

Collagen I and collagen IV mimetic peptides were synthesized using the Biotage (Uppsala, Sweden) Initiator + Alstra Automated Microwave Peptide Synthesizer according to published protocols (Madhan et al. 2008; Okuyama et al. 2012; Xiao et al. 2015), with sequence details shown in Table S1 (Supporting Information). The collagen I peptide forms a homotrimer, while for collagen IV two separate peptides (designated A and B) are mixed in a ratio of 2A:1B for formation of heterotrimeric structure. Collagen I peptide and collagen IV peptide B were synthesized with an N-terminal biotin and the predicted mass of each peptide was verified using an Agilent Technologies (Santa Clara, CA) 6120 mass spectrometer employing electrospray ionization in the positive mode. Triple helical conformation was confirmed by circular dichroism spectroscopy (Jasco-J815 spectropolarimeter) as previously described (Parmar et al. 2012; Xiao et al. 2015).

Enzyme-linked immunosorbent assays (ELISA)

For the Scl proteins ELISA were performed using Strep-Tactin coated plates (IBA-GmbH) that capture antigens via the Strep-Tag II tag as previously published (Hoe et al. 2007). Briefly, proteins (rScl1.1-FL, rScl1.1-CL and rScl1.1-V) were diluted to 0.5 µM in Phosphate Buffered Saline (PBS) for coating. After washing, serum was added at 1:200 in 0.1% Tween 20 in PBS (PBST) supplemented with 5% skim milk powder and incubated for 1 h at room temperature. Following a further three washes, human IgG binding was detected with a 1:3000 dilution of goat anti-human IgG conjugated to horse radish peroxidase (Santa Cruz, Texas, USA). The 3,3′,5,5′-Tetramethylbenzidine (TMB) was used as a substrate, reactions were stopped with 1M Hydrochloric acid and absorbance read at 450 nm.

For the biotin-tagged collagen mimetic peptides ELISA were performed using Streptavidin coated plates (G-biosciences, Missouri, USA). Peptides were added at 10 µg/mL in PBS for 2 h at room temperature (collagen I) or overnight at 4°C (collagen IV). After washing, serum was added at a 1:200 dilution and incubated for 2 h at room temperature (Collagen I) or overnight at 4°C (Collagen IV). Human IgG binding was detected as described above. Human sera previously shown to have no reactivity with rScl protein or collagen were used as negative controls and samples were considered positive if absorbance values were greater than the mean of the negative controls plus three standard deviations.

Statistical analysis

Differences between the ARF cases and healthy controls were assessed by non-parametric unpaired Mann-Whitney U test. Correlations were calculated using Spearman's r value. These analyses were performed in GraphPad Prism (Version 8.0) and a P-value of ≤0.05 was considered significant. Hierarchical clustering using Euclidean distance with average linkage method was carried out using Morpheus (https://software.broadinstitute.org/morpheus).

RESULTS

The 72 participants comprised children diagnosed with ARF according to the New Zealand modification of the Jones criteria (n = 36) and matched healthy controls (n = 36). Demographics are shown in Table 1. All participants identified as either Māori or Pacific and the median age range was 10.5–11.0-years-old, reflecting the peak incidence age for ARF in New Zealand (Gurney et al. 2016).

Table 1.

Demographic characteristics of Acute Rheumatic Fever (ARF) patients and matched healthy controls.

Characteristic	Healthy control	ARF
Number, n	36	36
Age, median (range), years	10.5 (5.0–18.0)	11.0 (7.0–16.0)
Male sex (%)	22 (61)	23 (64)
Ethnicity
Māori (%)	11 (31)	14 (39)
Pacific (%)	25 (69)	22 (61)
ARF manifestations
Carditis (%)	–	29 (81)
Arthritis (%)	–	25 (69)

Antibodies against rScl1.1 were measured in the sera of ARF patients and matched healthy controls by ELISA. The Scl1.1 protein (emm1 GAS) was selected as a representative from the ubiquitously expressed Scl1 protein family (Lukomski et al. 2017). Acute rheumatic fever patients had significantly elevated levels of anti-Scl1.1 IgG compared with controls (P < 0.0001) when the full-length protein is used as antigen (Fig. 1A). To investigate whether the variable (V) domain or collagen-like (CL) domain were targeted by antibodies, ELISA were next performed with rScl1.1-V and rScl1.1-CL as antigens. Interestingly, there was a significant elevation in anti-Scl1.1-CL IgG in ARF patients compared with controls (Fig. 1B and P < 0.001), but no elevation in anti-Scl1.1-V IgG (Fig. 1C), suggesting the collagen-like domain is responsible for the Scl1 antibody reactivity in ARF patients. Indeed, there is a highly significant correlation between anti-rScl1.1-FL and anti-rScl1.1-CL IgG (Fig. 1D, Spearmans's r 0.807 and P < 0.0001), and a complete lack of correlation between anti-rScl1.1-FL and anti-rScl1.1-V IgG (Spearmans's r 0.107 and P = 0.555). Furthermore, in competition ELISA with selected ARF patients, both the full-length protein and CL domain inhibited serum IgG binding to rScl1.1-FL, while the V domain showed no inhibition (Figure S2, Supporting Information).

Figure 1.

Anti-Scl IgG response in patients with ARF and matched healthy controls. Bars represent mean ± SD and the P-values were determined by Mann-Whitney U test (***P < 0.001 and ^****P < 0.0001). The dashed line represents the negative control plus three SD values. Antigens tested were full-length rScl1.1-FL (A), rScl1.1-CL domain (B) and rSCl1.1-V domain (C). Correlation of IgG responses in ARF cases only (n = 36) between full-length rScl1.1 and rScl1.1-CL domain (D) where Spearman's r = 0.8070.

Circulating anti-collagen antibodies were assessed by ELISA using biotin-tagged collagen I and collagen IV mimetic peptides. Acute rheumatic fever patients had significantly elevated levels of both anti-collagen I and IV serum antibodies compared with controls (Fig. 2A and B, P < 0.001), in keeping with previous reports (Martins et al. 2008; Dinkla et al. 2009). No significant difference in the levels of anti-collagen antibodies were observed in ARF patients with carditis or arthritis, compared to those without these symptoms (data not shown). However, this is likely due to an insufficient number of participants without these symptoms (Table 1), limiting sub-group analysis. The levels of anti-collagen I and IV antibodies were moderately correlated (Fig. 2C, Spearmans's r 0.4475 and P <0.01). However, there was no relationship between the presence of anti-collagen antibodies and antibodies specific for the Scl protein, exemplified by the lack of correlation between anti-collagen I or anti-collagen-IV IgG with anti-Scl1.1-FL IgG (Fig. 2D and E). This is supported by competition ELISA performed with selected ARF patients, which showed that neither the collagen-I or collagen-IV peptide inhibited serum IgG binding to rScl1.1-FL, in contrast to the Scl1.1-CL domain (Figure S2, Supporting Information). Finally, hierarchical clustering of ARF cases by IgG reactivity to rScl1.1 FL, rScl1.1-CL domain, collagen-I and IV using a non-correlation-based approach (Euclidean distance method) shows the cases fall into two distinct clusters with Scl1.1 antigens in one and both collagens in the other (Fig. 2F), further illustrating these antibodies are not immunologically related.

Figure 2.

Anti-Collagen IgG response in patients with ARF and matched healthy controls. (A and B) Bars represent mean ± SD and the P-values were determined by Mann–Whitney U test (***P < 0.001 and ^****P < 0.0001). The dashed line represents the negative control plus three SD values. Antigens tested were collagen I (A) and collagen IV (B). (C, D and E) A correlation of IgG response in ARF cases only (n = 36) between collagen I and IV (C), full-length rScl1.1 and collagen I (D), full-length rScl1.1and collagen IV (E). Spearmans's r values as shown. (F) Heat-map of IgG responses in ARF cases to collagen I, collagen IV, full-length rScl1.1 and rScl1.1-CL domain. Hierarchical clustering indicated two distinct clusters as shown by dendrogram and orange/mustard squares. Values below respective cut-offs (negative control plus three SD values for each antigen) are shown in white, with a scale from minimum (light blue) to maximum (dark blue) row values represented by color scale.

DISCUSSION

Acute rheumatic fever is an immune sequela of a GAS infection, with multiple prior GAS infections thought to prime the immune system for a loss of tolerance (Lorenz et al. 2021), such that an elevation in GAS specific antibodies in ARF patients is anticipated as a consequence of these prior infections. This study shows that GAS collagen-like proteins are immunogenic during infection as previously reported (Hoe et al. 2007; Chaudhary et al. 2017), with ARF cases having significantly elevated Scl1.1 antibody compared with controls. By examining IgG reactivity to full-length Scl1.1 protein as well as domain constructs, the majority of antibody binding could be attributed to the collagen-like (CL) domain. This might be expected based on sequence homology given the Scl1-CL domain is relatively conserved between strains and comprises recurring Gly-Xaa-Yaa (GXY) triplets. Indeed, an analysis of triplet repeats in 60 Scl variants from diverse GAS strains found there were six GXY triplets (GEA, GPA, GKD, GEK, GPQ and GET) that accounted for ∼50% of all triplets observed within the collagen-like domains (Han et al. 2006). In contrast, limited immunoreactivity was detected against the non-collagenous variable (V) domain in ARF patients in this study, and the sequence of this domain varies significantly with strain type (McNitt et al. 2018). The GAS strains epidemiologically associated with contemporary ARF in our setting and globally are highly diverse with no single emm-type dominating (Williamson et al. 2015; de Crombrugghe et al. 2020). Thus, it is likely the ARF patients examined have been exposed to broad range of V domains, in keeping with the relative lack of reactivity with the emm1 V domain observed.

The elevated anti-collagen antibodies detected in ARF patients in this study are consistent with previous reports (Martins et al. 2008; Dinkla et al. 2009). In contrast to the heterogenous collagens preparations used in prior studies, collagen mimetic peptides were utilized as antigens in this study. These peptides mimic the triple helical structure of native collagens but lack the higher order complexity of collagens extracted from the extracellular matrix (Xu and Kirchner 2021). While this may reduce the detection of autoantibodies specific for complex collagen-epitopes, the peptides ensure highly uniform collagen antigens are the basis for the immunoassays, increasing assay robustness. The increased IgG reactivity to both collagen I and IV peptides provides further evidence that antibodies to collagen molecules are generated during ARF pathogenesis. However, the lack of correlation between anti-Scl1.1-CL domain antibodies and collagen antibodies suggest these responses are independent. In contrast to the proposed molecular mimicry between coiled-coiled GAS M proteins and human cardiac proteins as a basis for autoantibody generation (Cunningham 2012; Carapetis et al. 2016), our data suggest a lack of mimicry or cross-reactivity between the GAS CL domain and human collagen. This is likely driven by the inherent differences in both sequence and triple helix stabilization between collagenous bacterial proteins and human collagen, and in particular the absence of hydroxyprolines at the Y position within Gly-Xaa-Yaa repeats in Scl proteins (Lukomski et al. 2017). Thus, anti-collagen antibodies in ARF appear to be generated as part of the auto-reactivity process, rather than in response to a GAS infection per say, in line with a recent proposal that collagen antibodies arise via epitope spreading following initial valve damage (Karthikeyan and Guilherme 2018).

In summary, this study has shown that while antibodies to both GAS collagen-like proteins and human collagen are present in patients with ARF, these antibody responses are unrelated. The elevation in collagen antibodies, independent of any mimicry with GAS collagenous domains, points to a role for extracellular matrix disruption and epitope spreading in ARF pathogenesis that is worthy of further investigation.

Conflicts of Interest

None declared.