(Received for publication, April 21, 1997, and in revised form, May 30, 1997)
From the Departments of Cell and Molecular Biology
and ¶ Medical Microbiology, Lund University, S-221 00 Lund, Sweden
Protein H, a surface protein of Streptococcus pyogenes interacting with the constant Fc region of IgG, is known to be released from the streptococcal surface by a cysteine proteinase produced by the bacteria. Poststreptococcal glomerulonephritis and rheumatic fever are conditions in which immune complexes and autoimmune mechanisms have been suggested to play pathogenetic roles. The present study demonstrates that addition of protein H to human serum produces complement activation with dose-dependent cleavage of C3. The activation was IgG-dependent and the result of complexes formed between IgG and protein H. These complexes were size heterogeneous with molecular masses of 400 kDa to 1.4 MDa. Using complement-depleted serum reconstituted with complement proteins, the activation by protein H was found to be dependent of the classical, but independent of the alternative pathway of complement. In contrast to results of experiments based on soluble protein H·IgG complexes, complement activation was inhibited by protein H when IgG was immobilized on a surface. The interaction between C1q and immunoglobulins represents the first step in the activation of the classical pathway, and protein H efficiently inhibited the binding of C1q to IgG immobilized on polyacrylamide beads. Protein H reduced C3 deposition on the IgG-coated beads and inhibited immune hemolysis of IgG-sensitized erythrocytes. Finally, significantly less C3 was deposited on the surface of protein H-expressing wild-type streptococci than on the surface of isogenic mutant bacteria devoid of protein H. The results demonstrate that protein H·IgG complexes released from the streptococcal surface can produce complement breakdown at the sites of infection, whereas complement activation on bacterial surfaces is inhibited. This should have important implications for host-parasite relationships. In addition, soluble protein H·IgG complexes might contribute to immunological complications of streptococcal infections.
Streptococcus pyogenes is an important human pathogen (for a review, see Ref. 1) causing suppurative infections like pharyngitis, tonsillitis, impetigo, and erysipelas. S. pyogenes is also responsible for a hyperacute and serious toxic shock-like syndrome that is sometimes associated with fasciitis and myositis. Rheumatic fever and glomerulonephritis are delayed sequelae of acute S. pyogenes infections, and several observations suggest that immunological mechanisms contribute to these conditions. Antibodies against S. pyogenes surface components have been reported to cross-react with heart sarcolemma (2) and cardiac myosin (3). Patients with acute post-streptococcal glomerulonephritis (APSGN)1 show circulating immunoglobulin (Ig) complexes (4), complement activation (4), and deposition of complement proteins in the glomeruli (5).
Complement plays an important role in defense against pathogenic microorganisms (6, 7). Several functionally interesting interactions have been described for S. pyogenes and components of the complement system. M proteins are antiphagocytic fibrous surface proteins of S. pyogenes (for a review, see Ref. 8). Members of this protein family specifically bind to the complement proteins factor H (9) and C4b-binding protein (C4BP) (10) that regulate complement activation (11-13). S. pyogenes also expresses a surface-associated peptidase, which degrades C5a (14, 15), a chemotactic fragment of C5 (16). Finally, some strains secrete protein SIC, which interacts with terminal complement proteins and inhibits complement-mediated lysis (17).
The starting point for the present investigation was the observation
that an extracellular cysteine proteinase of S. pyogenes (SCP) releases a large fragment of protein H from the surface of the
bacteria (18). Protein H is a streptococcal surface protein belonging
to the M protein family and has high affinity for the constant (Fc)
region of IgG (19, 20). The structure of protein H is schematically
depicted in Fig. 1. We report that
complex formation between protein H and IgG in the fluid phase leads to complement activation. By contrast, the interaction of protein H and
IgG on surfaces resulted in inhibition of complement function. Possible
implications for virulence and for immunological disease mechanisms are
discussed.
The S. pyogenes strain AP1 used in this
study is the 40/58 strain from the World Health Organization
Collaborating Center for References and Research on Streptococci,
Institute of Hygiene and Epidemiology, Prague, Czech Republic. To
generate an isogenic mutant of AP1 lacking surface-bound protein H, the
plasmid pBK37 was used. To generate pBK37 an internal PCR fragment,
covering about of sph, the gene coding for protein H, was ligated into the XbaI site in pJRS233 (21), a derivative of the temperature-sensitive shuttle vector
pG+host4 (22). The PCR fragment was generated using
oligonucleotides hybridizing with nucleotides 330-350 and 1151-1132
(also 1025-1006) in the published sequence of sph (20),
using AP1 chromosomal DNA as template. S. pyogenes strain
AP1 was transformed with pBK37 by electroporation as described
previously (23). To generate insertionally inactivated mutants the
procedure of Perez-Casal et al. (21) was followed. Single
colonies were established and screened for lack of IgG binding.
Southern blot analysis was used to confirm the integration of the
plasmid into sph. The sph gene was shown to be
interrupted and the adjacent gene on the chromosome, emm1,
coding for M1 protein (24), was shown to be intact in the mutant, as
analyzed by PCR. This procedure resulted in a mutant, BM27.6, with a
truncated sph gene, which lacks the COOH-terminal part
necessary for integration into the cell wall (25). The Phadebact typing
test (Pharmacia Biotech Inc., Uppsala, Sweden) established that the
mutant was a S. pyogenes strain. The occurrence of protein H
on the bacterial surface was investigated in binding experiments. Thus,
AP1 or BM27.6 bacteria were incubated with 125I-labeled IgG
in 200 µl of phosphate-buffered saline (PBS) containing 0.02%
NaN3 and 0.05% Tween 20, for 30 min, and washed in the
same buffer. Following centrifugation the radioactivity of the pellets was measured. Dilutions of the bacteria revealed that the binding of
IgG to BM27.6 was >1000 times lower than the binding of IgG to AP1.
AP1 bacteria were grown in Todd-Hewitt broth (Difco) supplemented with
0.2% yeast extract, and BM27.6 bacteria in this medium containing 1 µg/ml erythromycin (Sigma). Bacteria were heat-killed by incubation for 5 min at 80 °C.
Recombinant protein H was obtained by expression
in Escherichia coli of a PCR-generated DNA fragment
corresponding to the mature protein H except for the hydrophobic
membrane-spanning region (see Fig. 1). Thus, PCR was performed using
oligonucleotides synthesized from nucleotides 451-474 and 1374-1351
in the published sequence of protein H (20) using AP1 chromosomal DNA
as the template. PCR products resulting from amplification with these primers were ligated into the high expression vector pHD389 (26) as
described (27). Genomic DNA preparation, ligation, and transformation procedures were as described (28). Plasmids were transformed into
E. coli strain LE392 (28). Protein H was purified from cell
lysates as described (19). Ig-binding proteins were purified as
described previously: M1 protein (24), protein L (29), and protein G
(30). Protein A (31) was from Pharmacia. Protein PAB from
Peptostreptococcus magnus was purified as described (32). The isolation of C1q, factor D, and properdin was done as described (33-35). Myeloma IgG was isolated from human sera. Sheep erythrocytes were sensitized (36) with rabbit antiserum (Swedish National Bacteriological Laboratory, Stockholm, Sweden). IgG and IgM fractions of the rabbit antiserum were isolated by gel filtration (Ultrogel AcA22, LKB, Bromma, Sweden). F(ab)2 fragments prepared
from anti-C3c antibodies were from Dako (Glostrup, Denmark). Gelatin
was from Bio-Rad, and human polyclonal IgG and albumin were from
Sigma.
Blood was drawn from healthy volunteers. The blood was
allowed to clot at room temperature for 60 min. After another 120 min at 4 °C tubes were centrifuged and the serum collected and stored in
aliquots at 80 °C or used for the preparation of C1q, factor D,
and properdin (C1qDP)-depleted serum as described (37). Briefly, serum
was dialyzed against 50 mM phosphate buffer, pH 7.3, containing 2 mM EDTA, and NaCl to yield a final conductance
of 13 millisiemens/cm and was applied to a Bio-Rex 70 column (Mesh size
200-400, Bio-Rad). The material not bound to the column was dialyzed
against 0.1 M Tris-HCl, pH 7.4, containing 0.4 M NaCl and 2 mM EDTA, and applied to a column
of Sepharose-4B coupled with anti-properdin antibodies. The filtrate
was dialyzed against veronal-buffered saline (VBS) (5 mM
sodium 5,5-diethylbarbiturate, pH 7.4, 145 mM NaCl)
containing 0.15 mM Ca2+ and 0.05 mM
Mg2+ (VBS2+) and stored in aliquots at
80 °C. Before dialysis, some of the C1qDP-depleted serum was used
for preparation of IgG-depleted C1qDP-depleted serum. This reagent was
prepared by absorption of immunoglobulins from the serum on a column of
Sepharose-4B coupled with protein LG (38). The material not bound to
the column was dialyzed against VBS2+ and stored in
aliquots at
80 °C. The IgGC1qDP-depleted serum contained no
detectable IgG (<0.6 µg/ml). Another C1qDP-depleted serum was
prepared using buffer with a conductance of 12.5 millisiemens/cm instead of 13 millisiemens/cm during the initial ion exchange chromatography step. The final preparation contained no detectable C4BP
or factor H, whereas other complement proteins (C1r, C1s, C2, C4, and
C3) were present at concentrations that were similar to those found in
the other C1qDP-depleted sera. This reagent, C4BPC1qDP-depleted serum,
was used in some C3 deposition experiments. All plasma protein
concentrations were determined by electroimmunoassay (37).
The capacity of the serum and the serum reagents to support fluid-phase cleavage of C3 during incubation with bacterial proteins was investigated with crossed immunoelectrophoresis (37, 39). Normal serum, C1qDP-depleted serum, and IgGC1qDP-depleted serum were used at concentrations of 40%. IgG, C1q, factor D, properdin, and the bacterial proteins were added to yield final reaction mixture volumes of 100 µl. The plasma proteins were added at physiological concentrations (IgG, 10 mg/ml; C1q, 70 µg/ml; factor D, 1 µg/ml; and properdin, 25 µg/ml) with regard to undiluted serum. Following incubation for 30 min at 37 °C, 5 µl 0.2 M EDTA was added and the samples were put on ice, centrifuged, and subjected to 0.6% agarose gel electrophoresis. The gel corresponding to one lane was transferred to the second-dimensional plate. Separated proteins were electrophoresed into a gel containing anti-C3 antiserum. Immunoprecipitates were stained with Coomassie Brilliant Blue.
Double immunodiffusion according to Ouchterlony was performed in 1%
agarose (SEA-KEM, FMC, Rockland, ME) in PBS. Diffusion was allowed to
proceed for 48 h at 4 °C. The gels were immersed in PBS for
24 h and in distilled water for 60 min, dried, and stained with
Coomassie Brilliant Blue. Gel filtration experiments were performed on
an FPLC Superose-6 column (Pharmacia) equilibrated with PBS containing
0.02% NaN3. The flow rate was 0.25 ml/min, and 0.5-ml
fractions were collected. Before SDS-PAGE (40) fractions were
concentrated 10 times by precipitation in 10% trichloroacetic acid
(Sigma) and boiled in sample buffer containing 2% SDS and 5%
-mercaptoethanol.
Polyclonal human IgG was coupled to polyacrylamide beads (Immunobeads, Bio-Rad) according to the producer's instructions. 125I-Labeled C1q (33) in 0.1 ml of VBS and 0.1% gelatin, 0.1 ml of Immunobeads coupled with IgG, and 0.2 ml of various Ig-binding proteins in the same buffer were mixed and incubated overnight at 20 °C. Two ml of the same buffer containing 0.01 M EDTA was added, beads were spun down, washed, and the radioactivity of the pellets was measured. A concentration of beads resulting in submaximal binding (60%) of 125I-C1q to IgG was consistently used in these experiments. Data points represented the mean of duplicate determinations in a single experiment. All experiments were performed at least three times. The same procedure was followed in experiments with albumin-coupled polyacrylamide beads. In similar experiments AP1 bacteria (2 × 108 cells) were preincubated with an excess of IgG for 30 min at 37 °C. The cells were washed, radiolabeled C1q was added, and the radioactivity of the pellets was measured.
C3 Deposition AssayPolyclonal human IgG coupled to
polyacrylamide beads was incubated with bacterial proteins in 0.2 ml of
PBS containing 0.25% gelatin, 0.25% Tween, and 0.25% bovine serum
albumin for 20 min, washed twice in VBS2+ containing 0.1%
gelatin, and resuspended in 100 µl of the same buffer. 100 µl of
20% serum was added, and the samples were incubated at 37 °C for 20 min. Two ml of cold VBS containing 10 mM EDTA were added.
The tubes were centrifuged and the pellets washed. The pellets were
resuspended in 200 µl of PBS containing 0.25% Tween and 30,000 cpm
of F(ab)2 anti-C3c antibodies 125I-labeled
with the chloramine-T method (41). After 3 h of incubation, 2 ml
of the same buffer was added, the tubes were centrifuged, and the
radioactivity of the pellets was measured. The same procedure was
followed in experiments with albumin-coupled polyacrylamide beads. In
other experiments the polyacrylamide beads were exchanged for 100 µl
of AP1 or BM27.6 bacteria (2 × 105 cells), omitting
the blocking step described for the beads. Data points represent the
mean of duplicate determinations of a single experiment. Experiments
were performed at least three times.
Optimally sensitized (IgG or IgM) sheep erythrocytes (2 × 109 cells/ml) were incubated for 20 min with the bacterial proteins. The cells were centrifuged at 175 × g for 7 min, washed once in VBS2+, and resuspended in the same buffer to 2 × 109 cells/ml. This suspension (100 µl) was incubated with an equal volume of 20% serum for 20 min at 37 °C. Three ml of cold VBS containing 10 mM EDTA were added, and the samples were centrifuged at 175 × g for 7 min. The absorbance of the supernatants was measured at 541 nm. Data points represent the mean of duplicate determinations in a single experiment. Experiments were performed at least three times.
At the streptococcal surface a large IgGFc-binding fragment of protein H is released by SCP (18). To investigate if soluble protein H is capable of activating complement, purified protein H was added to serum. C3 breakdown was then studied by crossed immunoelectrophoresis. M1 protein is expressed at the bacterial surface together with protein H. M1 protein also binds to IgG, but compared with protein H, the affinity is much lower (24). Studies of M1 protein, and of strongly Ig-binding proteins (staphylococcal protein A, streptococcal protein G, and peptostreptococcal protein L) were included in the experiments.
Fig. 2 shows that proteins A, G, H, and L
all activate complement in a dose-dependent manner, when
added to human serum. M1 protein, however, had no effect also at the
highest concentration tested (35 µM). C3 conversion could
result either from activation of the classical or the alternative
pathway or from a combination of both. To discriminate between these
alternatives a C1qDP-depleted serum was utilized. Reconstitution of the
serum with C1q (Fig. 3E) restored complement activation by
protein H to the level obtained with intact normal serum (Fig.
3A). Reconstitution with
factor D and properdin had no effect (Fig. 3D). The results
demonstrated that fluid-phase activation of C3 by protein H is mediated
through the classical pathway.
The critical importance of IgG for protein H-induced complement
activation in serum was shown with an IgGC1qDP-depleted serum (Fig.
4). No C3 cleavage was seen in this serum
when protein H and complement proteins were added. The capacity to
support protein H-induced complement activation was re-established with
polyclonal or monoclonal IgG.
Protein H Forms Soluble C1q-binding Complexes with IgG
In
contrast to proteins A, G, and L, protein H did not form distinct
precipitation arcs with human IgG, when tested by double diffusion in
agarose (Fig. 5). Diffuse staining was
observed between the IgG- and protein H-containing wells, indicating
the formation of heterogeneous complexes. Gel filtration experiments
were performed to demonstrate and analyze soluble protein H·IgG
complexes. Protein H was incubated with IgG at the equimolar
concentrations shown to result in complement activation (Fig. 2).
Proteins A, G, and L formed insoluble precipitates when incubated with
IgG in solution; no visible precipitates were formed between protein H
and IgG. Protein H, IgG, and a mixture of the two proteins were each
subjected to gel filtration on a Superose-6 column (Fig.
6A). Protein H tends to form
multimers (42), which explains the broad peak. The protein H·IgG
mixture gave rise to soluble complexes of predicted molecular masses
ranging from 400 kDa to 1.4 MDa. Analysis by SDS-PAGE (Fig.
6B) showed that the complexes contained both protein H and
IgG. Finally, when radiolabeled C1q was added to the mixture of protein
H and IgG, the label appeared together with protein H·IgG complexes
(Fig. 6C), reflecting C1q interaction with the complexes.
Inhibition of C1q Binding to Immobilized IgG by Soluble Protein H
Soluble protein H and other Ig-binding bacterial surface
proteins were tested for their capacity to interfere with binding of
radiolabeled C1q by IgG-coated polyacrylamide beads. Proteins A, G, and
H all bind to the C2-C
3 interface region of IgG (43-46), whereas
the binding site for C1q is in the C
2 domain (47). On a molar basis
proteins A and H inhibited the binding of radiolabeled C1q with the
same efficiency as unlabeled C1q. Protein G unexpectedly increased the
uptake of radiolabeled C1q by solid-phase IgG (Fig. 7). The enhanced binding in the presence
of protein G was not due to an interaction between protein G and C1q.
Thus, C1q did not bind to protein G immobilized on Sepharose (not
shown). M1 protein and the Ig light chain-binding protein L had no
effect on C1q binding to IgG (Fig. 7). Control experiments with
albumin-coated polyacrylamide beads showed that background binding of
radiolabeled C1q was low (<5%) in the assay system.
Inhibition of C3 Deposition and Immune Hemolysis by Soluble Protein H
IgG-coated polyacrylamide beads were incubated with human serum
and protein H. C3 deposition was measured with radiolabeled anti-C3c
F(ab)2. Protein H completely inhibited C3 deposition (Fig.
8). The same effect was seen with protein
A, whereas protein G enhanced deposition, and proteins L and M1 had no
effect (Fig. 8). Besides their IgGFc-binding activity, proteins A and G
have weak affinities for Fab fragments of IgG (48, 49). Judging from
control experiments performed in the absence of serum this did not
influence the results. In other control experiments polyacrylamide beads coated with albumin were incubated with IgGC1qDP-depleted serum
reconstituted with C1q. In this case, C3 deposition was at background
level, and no effect was seen with any of the Ig-binding proteins (not
shown).
Protein H partially inhibited immune hemolysis of IgG-sensitized sheep
erythrocytes (Fig. 9). Protein A blocked
hemolysis completely, whereas protein G caused a
dose-dependent increase of the hemolytic activity (Fig. 9).
The effect of protein G is in line with the observation that the
protein enhanced binding of C1q to IgG and surface deposition of C3
(see above). None of the proteins affected immune hemolysis of
IgM-sensitized sheep erythrocytes (not shown).
C3 Deposition on Intact Streptococci
No binding of
radiolabeled C1q was recorded in experiments with protein H-expressing
AP1 streptococci preincubated and saturated with human polyclonal IgG,
indicating that Fc regions are not exposed in IgG molecules bound to
the bacterial surface. Consistent with this finding, incubation of the
AP1 streptococci with serum gave little or no C3 deposition (Fig.
10). An isogenic mutant (BM27.6) devoid
of protein H was generated by insertional inactivation of the protein H
gene in AP1. The C3 deposition on the mutant bacteria was comparatively
high (Fig. 10A). Results obtained with C1qDP- or
IgGC1qDP-depleted sera reconstituted with C1q or C1q and IgG (Fig.
10B) showed that C3 deposition depended on C1q and IgG,
i.e. the classical pathway. C4BP, a down-regulator of the classical pathway, is known to interact with members of the M protein
family (10), including protein H (50). For this reason, C3 deposition
experiments were repeated with a C1qDP-depleted serum devoid of C4BP.
With the conditions used, C4BP did not appear to influence C3
deposition on the AP1 or BM27.6 streptococci (Fig. 10C). In
conclusion, the results were fully compatible with findings using
immune hemolysis and IgG-coated polyacrylamide beads (Figs. 7, 8, 9) and
strongly suggested that surface bound protein H has an
anti-opsonizing effect due to interference with IgGFc-C1q
interactions.
The major conclusion of this investigation is that protein H activates complement when the molecule is part of soluble complexes with IgG, but prevents the activation of complement when associated with the bacterial surface. These effects should both provide selective advantages to S. pyogenes and help to explain how Ig-binding surface proteins contribute to the virulence (51-53) of this important human pathogen.
It has been demonstrated that an IgG-binding fragment of protein H is
released from the streptococcal surface by SCP (18). This fragment
covers the entire surface-exposed part of the molecule. The fragment
used throughout this study has a similar size excluding only the
bacterial cell wall-associated COOH-terminal region (Fig. 1). IgG is
one of the most abundant soluble extracellular human proteins, and the
high affinity between protein H and IgG, 1.6 × 109
M1 (19), suggests that protein H in
vivo is always complexed with IgG. When protein H·IgG complexes
are released by SCP, the data of this study show that the complexes are
soluble and capable of activating the classical complement pathway.
This activation will lead to breakdown of complement in the vicinity of the bacteria, thus preventing assembly and activation of complement at the bacterial surface. On the other hand, complement activation will generate C5a and thereby attract phagocytic cells. From the bacterial point of view, this should be an unwanted effect of the release of protein H·IgG complexes by SCP. However, SCP efficiently and simultaneously releases also a large and biologically active fragment of a C5a peptidase (18) associated with the streptococcal surface (14). This enzyme cleaves and destroys C5a as a chemoattractant for polymorphonuclear leukocytes (15).
To initiate complement activation away from the bacterium and to inactivate the C5a that is generated represents sophisticated and rational microbial defense mechanisms. The inhibition of complement activation at the surface of protein H-expressing bacteria should also be beneficial for the microbe. Despite that large amounts of IgG are bound to the bacterial surface (54) through Fc-protein H interactions, the blocking of C1q binding to IgGFc by protein H was found to be highly effective. As a result, the deposition of opsonic C3 fragments was significantly lower in protein H-expressing S. pyogenes than in the isogenic mutant devoid of protein H. The blocking of the C1q-binding region of IgGFc by surface-associated protein H is in contrast to the soluble IgG-protein H complexes. Here the stoichiometry of complexes apparently results in the exposure of IgGFc regions capable of interacting with C1q and subsequently activation of the classical pathway of complement.
In contrast to protein H and staphylococcal protein A, protein G of group C and G streptococci unexpectedly enhanced the binding of radiolabeled C1q to IgG. The binding site(s) in IgGFc for protein H has not yet been defined by x-ray crystallography or NMR. In the case of proteins A and G, it was demonstrated that apart from a shared binding region, the proteins have unique binding sites in IgGFc. Compared with protein A, the unique protein G-interacting site is located away from the predicted C1q-binding site (43, 47, 55), which explains why protein G does not inhibit the binding of C1q to IgGFc. Protein G was not found to interact with C1q, and it remains unclear why complexes between protein G and IgG appear to have higher affinity for C1q than IgG itself.
Protein H is expressed by S. pyogenes strains of the M1 serotype. This serotype is associated with severe complications of suppurative S. pyogenes infections, i.e. the toxic shock-like syndrome, rheumatic fever, and APSGN (for references, see Ref. 56). The data of this study raise the possibility that release of complement-activating IgG-protein H complexes from the bacterial surface could be involved in development of these complications. Like human IgG, rabbit IgG has affinity for protein H (20). When two rabbits were given protein H intravenously (0.5 mg at times 0, 12 h, and 24 h), the animals developed anuria and died within 48 h,2 demonstrating the toxic property of protein H. It is tempting to speculate that protein H·IgG complexes may cause localized inflammation and tissue damage as a result of deposition in for instance the heart and the kidneys. In APSGN, the patients typically develop renal symptoms 1-3 weeks after a throat or skin infection with S. pyogenes. Glomerular deposition of protein H·IgG complexes in the course of infection could be an interesting mechanism by which antigen is "planted" in a target organ. In this case, a classical immune response directed against protein H and/or protein H·IgG complexes could then initiate local pathophysiological events in the glomeruli, explaining the latency period following the triggering infection. These and other possible consequences of the molecular host-microbe interactions described here are currently under investigation.