BPI–ANCA in transporter associated with antigen presentation (TAP) deficiency: possible role in susceptibility to Gram-negative bacterial infections

Abstract

Although HLA class I expression is diminished in patients with defects in the transporter associated with antigen presentation (TAP), recurrent Gram-negative bacterial lung infections are found from childhood onwards. As MHC class II-mediated responses are normal, other mechanisms that contribute to susceptibility to infections are presumed. The bactericidal/permeability-increasing protein (BPI) is a potent neutrophil antibiotic that neutralizes endotoxin efficiently. As antineutrophil cytoplasmic autoantibodies (ANCA) against BPI were found in the majority of cystic fibrosis patients and correlate with disease severity we examined the prevalence of BPI–ANCA and their contribution to susceptibility to bacterial infections in six TAP-deficient patients. Although only two patients showed ANCA in indirect immunofluorescence, BPI–ANCA occurred in five of six patients in ELISA. Purified IgG from BPI–ANCA-positive sera (five of six) inhibited the antimicrobial function of BPI in vitro. Epitope mapping revealed binding sites not only on the C-terminal but also on the antibiotic N-terminal portion of BPI, indicating that short linear BPI peptide fragments may be long-lived enough to become immunogens. In conclusion, BPI–ANCA are associated strongly with TAP deficiency. Inhibition of the antimicrobial BPI function by BPI–ANCA demonstrates a possible mechanism of how autoantibodies may contribute to increased susceptibility for pulmonary Gram-negative bacterial infections by diminished bacterial clearance.

INTRODUCTION

The human primary immunodeficiency disorder caused by mutations in transporter associated with antigen presentation (TAP) and defective surface HLA class I-expression is associated with recurrent Gram-negative bacterial infections of the respiratory tract as well as granulomatous skin lesions and skin vasculitis [1,2]. Clincally these patients present with a syndrome that resembles Wegener's granulomatosis [3–5]. However, although the underlying defect of antigen presentation by means of MHC class I would be expected to faciliate viral infections due to a disturbed interaction of CD8-positive lymphocytes with MHC class I presented viral antigens [1,2,5], TAP-deficient patients suffer from recurrent bacterial infections from early childhood onwards, similar to those in cystic fibrosis, or remain asymptomatic [6,7]. This susceptibility cannot be explained easily by down-regulation of HLA class I expression. Common bacterial strains isolated from the respiratory tract of TAP-deficient patients include Gram-negative rods, such as Pseudomonas aeruginosa, which are expected to evoke HLA class II-related immune responses [1,2,5,8] that are normal in TAP-deficient patients. We hypothesized that an additional inhibitory mechanism, rather than the primary cellular defect, accounts for the persisting compromised defence against bacterial infections. Some of the mechanisms that can impair pulmonary immune defence against bacteria have already been identified in cystic fibrosis (CF). Besides the known impeded mucociliary clearance, a high salt concentration in bronchial secretions inhibits powerful natural antibiotics secreted by epithelial cells such as beta-defensins or cathelicidins [6,9]. In addition, another antibiotic protein, the bactericidal permeability-increasing protein, is the target of antineutrophil cytoplasmic autoantibodies (BPI–ANCA) that are found in the majority of CF patients and correlate with disease severity [10,11]. As BPI is one of the most potent innate neutrophil antibiotics against Gram-negative bacteria and neutralizes their lipopolysaccharides (LPS) efficiently [12,13], the question was raised as to whether BPI–ANCA occur in TAP deficiency and interact with the antibiotic activity mediated by the N-terminal portion of BPI [14]. In this study we report on the presence of BPI–ANCA directed against the N-terminal portion of BPI as a potential cause for a diminished bacterial clearance in patients with TAP deficiency. We analysed indirect immunofluorescence (IIF), ELISA and epitope mapping for detection of these antibodies and defined immunogen epitopes recognized by the patients’ BPI–ANCA that may contribute to the inhibitory properties of these autoantibodies on the antibiotic function of BPI.

MATERIALS AND METHODS

Patients

Clinical data and serum samples from six patients with strongly reduced surface expression of HLA class I molecules due to defects in the TAP were analysed in this study (see Table 1) after local ethical approval and informed consent was obtained. TAP1 deficiency has been demonstrated previously in patients 3, 5 and 6 while TAP2 deficiency was present in patients 1 and 4 [5]. Patient 2 is a newly identified patient with extremely reduced HLA class I surface expression due to TAP2 deficiency, determined according to established methods [5].

Table 1.

Clinical data of TAP-deficient patients

Pat.	Age	Sex	Defect	rBPI-ANCA	Clinical presentation and histology
Pat. 1–5
1	36	f	TAP2	1 : 400	Necrotizing granulomatous lesions (ENT)
2	37	f	TAP2	1 : 1600	Retinal vasculitis
3	31	f	TAP 1	1 : 200	Skin vasculitis
4	49	f	TAP 2	1 : 200	Recurrent pulmonal infections with Pseudomonas aeruginosa, Klebsiella pneumoniae and Haemophilus influenzae
5	49	f	TAP 1	1 : 1600	Bronchiectasis
Pat. 6
6	42	f	TAP1	Negative	Necrotizing granulomatous lesions (ENT)

TAP: transporter associated with antigen processing; ENT: ear, nose and throat region. Shown is the maximum dilution in ELISA (standard 1 : 50) at which BPI–ANCA were detectable.

IIF and ELISA

Standard IIF and ELISA protocols [15,16] were applied to examine patients’ sera for the presence of ANCA. Protein specificity of ANCA was determined using recombinant BPI holoprotein (rBPI) and lipopolysaccharide binding protein (rLBP, both XOMA (US) LLC, Berkeley, CA, USA), proteinase 3 (PR3), myeloperoxidase (MPO), elastase (HLE), cathepsin G (CG), lactoferrin (LF) and lysozyme (LZ). IgG preparations of BPI–ANCA-positive sera were obtained by purification on a protein G-sepharose column (Pierce, Rockford, IL, USA). Specific binding of the purified BPI–ANCA-positive IgG preparations to the C- and N-terminus of BPI was determined by ELISA using rBPI₂₁ (residues 1–193; XOMA) and the previously described recombinant BPI-LBP fusion proteins P4160 (BPI residues 200–456 combined with N-terminus of LBP) and P4161 (BPI residues 1–200 combined with C-terminus of LBP; both XOMA) [17]. Purified IgG from goat anti-BPI-antiserum [18] served as positive control. Immunoglobulin G isotypes were determined using anti-IgG1–IgG4 antibodies (Boehringer, Ingelheim, Germany).

Epitope mapping

In order to test specific antibody binding to linear epitopes of BPI overlapping (by two amino acids) 13mer peptides covering the whole human BPI sequence [19] were prepared by automated spot synthesis and bound to cellulose membranes (Pepspots membranes, Jerini Biotools Berlin, Germany). Binding of BPI–ANCA to individual 13mer peptides was visualized by using peroxidase-labelled antihuman-IgG-antibody (Sigma, Germany) and a standard chemiluminescence kit (Boehringer, Mannheim, Germany). Cellulose-bound rBPI holoprotein served as positive control.

SDS-PAGE and Western blotting

Heat-denaturated rBPI₂₁ (boiled for 15 min at 96°C) and untreated rBPI₂₁ were applied to SDS-PAGE (15% acrylamide gel) under non-reducing conditions, transferred to a nitrocellulose membrane (Schlechter and Schuell, Germany) and probed with patients’ purified IgG preparations. Binding of the secondary antibody, an alkaline phosphatase-conjugated antihuman IgG (Sigma, Germany) was visualized by using nitroblue tetrasodium/bromochloroindolylphosphate as a substrate (Sigma, Germany).

Inhibition of antibacterial activity of rBPI and rBPI₂₁ by BPI–ANCA

According to a published protocol by Weinrauch [18], Escherichia coli DH5α (ATCC number 53868) was used to examine the effects of BPI–ANCA IgG preparations and F′ab fragments generated from these preparations (Pierce) on the antibiotic function of rBPI and rBPI₂₁. Briefly, rBPI or rBPI₂₁ was diluted to a final concentration of 0·1–100 nm/l in incubation medium and preincubated with patients’ IgG or F′ab fragments in a final concentration from 250 mg/l to 1 g/l (15 min, 37°C) before addition of bacteria. After incubation, aliquots were transferred to molten Bactoagar (Difco Laboratories, Detroit, MI, USA). After the agar solidified, bacterial viability was measured by counting the number of colony-forming units (CFU) after 24-h incubation. Purified anti-BPI IgG and F′ab fragments from goat serum, which effectively blocked the antimicrobial activity of BPI [18], served as positive control while pooled IgG from BPI–ANCA-negative healthy volunteers served as negative control. For patient 2, preabsorbtion of IgG preparations on immobilized rBPI₂₁ or Escherichia coli DH5α was carried out prior to use in the inhibition assay. In other experiments preabsorbtion of patients BPI–ANCA-positive IgG preparations was performed using the immobilized BPI-peptide 279–291 (MWG, Germany).

RESULTS

BPI–ANCA are associated with TAP deficiency

Examina-tion of patients’ sera for the presence of ANCA by IIF and ELISA gave diverging results. While IIF revealed atypical ANCA-staining patterns in only two patients, ELISA testing demonstrated the presence of BPI–ANCA at a concentration of at least 128 U/ml in all patients except patient 6 (Table 1). No other ANCA specificities were detected. BPI–ANCA were mainly of the IgG1 and IgG3 subtypes, but rarely IgG2, and never IgG4. In patients 1 and 2, BPI–ANCA were present at constant high titres over a follow-up period of 4 and 5 years, respectively. All patients suffered from chronic necrotizing granulomatous skin lesions, and patients 1, 2, 3 and 5 also suffered from frequent, severe bacterial respiratory infections with Pseudomonas aeruginosa, Klebsiella pneumoniae or Haemophilus influenzae and bronchiectasis. Patients 1 and 2 died from respiratory complications before the conclusion of this study. Patient 4, who was a first-degree relative of patient 1 and shared the same MHC I and MHC II haplotype, suffered from only three to four episodes of clinically mild bacterial infection per year and had no radiological signs of bronchiectasis. Patients 1 and 4 grew up in different European countries. Patient 6 never exhibited any signs of increased susceptibility to infection.

BPI–ANCA can inhibit the antimicrobial activity of rBPI and rBPI₂₁in vitro.

Whole IgG preparations and F′ab fragments were generated from sera of TAP-deficient patients and studied for their in vitro effect on the antimicrobial function of rBPI and rBPI₂₁ against E. coli (see Fig. 1). IgG fractions and F′ab fragments (data not shown) from patients 1, 3 and 5 reduced the antibacterial activity of both rBPI and rBPI₂₁ at least 10-fold, while IgG and F′ab preparations from patient 2 had a more than 100-fold inhibitory effect on either BPI protein. In contrast, IgG and F′ab fragments from patient 4 had no detectable effect on the antimicrobial activity at 250 mg/l, whereas an inhibitory effect was seen at higher concentrations (1–5 g/l, data not shown). In contrast, purified IgG from pooled healthy human donors’ sera exhibited no inhibition of BPI's antibiotic function at any concentration used. Because a larger amount of serum was available from patient 2, comparison of the inhibitory activity of this patient's IgG fractions before and after preabsorption on immobilized and inactivated E. coli DH5a and on immobilized rBPI₂₁ was possible. These experiments showed no effects on the inhibition of BPI by BPI–ANCA preabsorbed on immobilized E. coli DH5a, whereas preabsorption on rBPI₂₁ resulted in a loss of inhibition and ELISA activity, thereby ruling out non-specific effects or contaminants in the BPI–ANCA containing IgG preparations (Fig. 2).

Fig. 1.

Antibacterial activity of rBPI and rBPI₂₁ against E. coli DH5α is inhibited by patients’ BPI–ANCA in vitro. Effects of purified IgG preparations from BPI–ANCA-positive patients in a constant concentration of 250 µg/ml on the antibiotic activity of rBPI (a) and rBPI₂₁ (b) against E. coli DH5α were measured as described in Methods. Pooled human IgG served as negative control, purified IgG from goat anti-BPI antiserum served as positive control. The results shown represent the mean of at least three experiments. For clarity, error bars have been left out; variation was <20% in all cases.

Fig. 2.

Inhibition of antibacterial activity of rBPI₂₁ against E.coli DH5α by BPI–ANCA IgG can be reversed by preabsorption on rBPI₂₁. Preabsorption of BPI–ANCA IgG from patient 2 was performed using either an rBPI₂₁ coated (8 µg/ml), an E.coli DH5α coated (heat-inactivated, 10⁶/ml) or an uncoated ELISA plate prior to use in the growth inhibition assay. Results show the mean of two experiments.

In TAP-deficient patients BPI–ANCA recognize epitopes on the C- and N-terminal parts of the BPI molecule involved in the antimicrobial activity of BPI

The antibiotic function of BPI has been demonstrated previously to depend on the N-terminus rather than the C-terminus of BPI. We next studied by ELISA the domain-specificity of the patients’ BPI–ANCA-positive IgG preparations using recombinant proteins containing the N-terminus (P4161, rBPI₂₁) or C-terminus (P4160) of BPI as target antigens (see Table 2). BPI–ANCA-positive IgG preparations of patients 1, 2, 3 and 5 recognized only the C-terminus (P4161), while both the C- and N-terminus of BPI (P4161, rBPI₂₁, P4160) were recognized by patient 4. None of the patients’ IgG preparations cross-reacted with immobilized rLBP. These results indicate that in ELISA, using the above-mentioned proteins, BPI–ANCA recognize the C-terminal more often than N-terminal domain of BPI. To analyse further the diverging results between recognition of the C-terminal part in ELISA and the inhibitory in vitro effects of patients’ BPI–ANCA IgG preparations on the antimicrobial function of BPI, mediated by the N-terminal portion, an epitope mapping (13mers peptides, two amino acids overlapping) covering the whole human BPI sequence was performed (see Fig. 3). It revealed that both IgG preparations of all TAP-deficient patients (including patient 6 who was negative in ELISA) and to some extent pooled healthy controls contained a diverse range of polyclonal BPI–ANCA with different binding specificities not only to epitopes within the C- but also the N-terminus of BPI. Peptide 345–355 was recognized by all IgG preparations, suggesting that these autoantibodies had no inhibitory effect on the antibiotic function of BPI. Conversely, peptide 279–292 was recognized only by IgG preparations of TAP-deficient patients 2, 3 and 5, which contained inhibitory BPI–ANCA (Fig. 1). However, preabsorption of these patients’ IgG on immobilized peptide 279–292 and competitive inhibition assay had no detectable effect on either their inhibitory activity of BPI function or their binding to rBPI. Short linear N-terminal BPI peptides were recognized only by four TAP-deficient patients but not by pooled healthy donors’ IgG. Amino acid sequences within the previously defined antibiotic and endotoxin neutralizing regions of BPI (residues 17–36, 82–108 and 146–159) [20,21] were recognized specifically by patients 2 and 3, while BPI–ANCA of patients 4 and 6 bound to peptides immediately upstream and downstream to these sequences, respectively. Although BPI–ANCA-positive IgG of patients 1 and 5 inhibited the antibiotic activity of both rBPI₂₁ and rBPI (Fig. 1), they did not bind to any of the overlapping N-terminal 13-mer peptides used in our epitope mapping. Western blot analysis gave consistent results, as binding to heat-denatured rBPI₂₁ was detected only for IgG-preparations of patients 2, 3 and 4 (see Fig. 4). Together these results demonstrate the existence of BPI–ANCA specific to non-conformational, short linear epitopes of BPI in TAP-deficient patients. They further indicate that BPI–ANCA against conformation-dependent N-terminal epitopes may be involved in mediating inhibition of the antibiotic function of BPI.

Table 2.

Serological data of TAP-deficient patients

		ELISA
Pat.	IIF	PR3-/MPO-ANCA	rBPI-ANCA	rBPI21-ANCA	P4160-ANCA	P4161-ANCA	rLBP-ANCA
1	Negative	Negative	1 : 400	Negative	1 : 400	Negative	Negative
2	aANCA	Negative	1 : 1600	Negative	1 : 1600	Negative	Negative
3	Negative	Negative	1 : 200	Negative	1 : 200	Negative	Negative
4	aANCA	Negative	1 : 200	1 : 100	Negative	Negative	Negative
5	Negative	Negative	1 : 1600	Negative	1 : 1600	Negative	Negative
6	Negative	Negative	Negative	Negative	Negative	Negative	Negative

aANCA: atypical antineutrophil cytoplasmic autoantibodies; IIF: indirect immunofluorescence technique; LBP: lipopolysaccharide binding protein; MPO: myeloperoxidase; PR3: proteinase 3. Shown is the maximum dilution in ELISA (standard 1 : 50) at which BPI–ANCA were detectable. Besides rBPI and the N-terminal fragment rBPI₂₁ recombinant fusion proteins containing the N-terminus of LBP and the C-terminus of BPI (P4160) or vice versa (P4161) as well as LBP were used as described in Methods.

Fig. 3.

Linear BPI epitopes recognized by BPI–ANCA IgG of TAP-deficient patients. Binding of patients’ IgG to cellulose-bound peptides was performed as described in Methods. Numbering of amino acid sequence is according to the sequence of rBPI published by Beamer et al. [19]. Shown are residues 1–456. Regions recognized by IgG from BPI–ANCA positive patients’ and goat anti-BPI antiserum are shown as horizontal bars. Regions that retain endotoxin neutralizing and antibiotic activity are shown in bold italics.

Fig. 4.

rBPI₂₁-ANCA of TAP-deficient patients recognize conformational epitopes. SDS-PAGE and Western blotting with purified IgG preparations from BPI–ANCA-positive patients and untreated or boiled rBPI₂₁ was carried out as described in Methods. Lane 1: IgG from patient 2; lane 2: IgG from patient 3; lane 3: IgG from patient 4; lane 4: IgG from patient 5; lane 5: IgG from patient 1; lane 6: rBPI₂₁ only; lane 7: pooled human IgG; lane 8: IgG from patient 6.

DISCUSSION

In our study we could demonstrate that in TAP-deficient patients BPI is the target of autoantibodies that occur in patients with a severe pulmonary phenotype. Whereas the BPI–ANCA-positive patients 1–5 suffered from recurrent bacterial pulmonary infections with Ps. aeruginosa, Klebsiella pneumoniae and Haemophilus influenzae and granulomatous skin lesions patient 6, who was BPI–ANCA-negative, presented with an almost asymptomatic course of the disease as described recently for two TAP-deficient patients [7]. An apparent paradox of patients with TAP deficiency is their high susceptibility to Gram-negative bacterial rather than viral infections, which affect mainly the respiratory tract from childhood onwards. Chronic colonization of the lungs by Gram-negative bacteria such as Ps. aeruginosa suggest impeded pulmonary defence mechanisms that result in a similar pulmonary phenotype, as in CF patients [1,2,5,22,23]. Although the reason for the diminished bacterial clearance in the respiratory tract of TAP-deficient patients is not understood fully, present data indicate that low numbers of pulmonary macrophages, tissue damage due to activated NK cells that lyse cells with low MHC class I expression and possibly a disturbed presentation of bacterial antigens by means of MHC class I as well as immunosuppressive therapy may contribute to the impaired pulmonary immune response [5,24,25]. As a net effect there is a persisting colonization with Gramnegative bacteria such as Pseudomonas that causes lung destruction by a constant endotoxin stimulus evoking an inflammatory cytokine response as well as recruitment and activation of neutrophils and macrophages [26–28]. One effective counter-strategy against these proinflammatory mechanisms is BPI, which is delivered by activated neutrophil granulocytes and mucosal epithelial cells [12,29], and by means of its potent antibiotic and endotoxin-neutralizing abilities facilitates bacterial clearance and dampening of the lipopolysaccharide-triggered inflammation locally and systemically [29,30]. In CF, the previously shown association of BPI–ANCA with a decreased lung function, Pseudomonas colonization and a more severe structural lung damage has already suggested a pathophysiological role [10,11,30]. For some CF patients, inhibition of opsonophagocytosis mediated by the C-terminal portion revealed one mechanism showing how BPI–ANCA can compromise immune defence [11,31]. Our data show in addition that BPI–ANCA of TAP-deficient patients can inhibit the antibiotic function mediated by the N-terminal part of BPI and may thus contribute to a diminished Gram-negative bacterial clearance.

As the obvious discrepancies in patient 2 between ELISA results and inhibition of the antimicrobial activity indicate, BPI–ANCA that recognize the amino-terminal BPI portion, especially in low quantities, may be easily missed. Routine IIF is negative in more than half of them due to standard fixation procedures [15,16]. Although ELISA is more sensitive to detect BPI–ANCA, it appears to be not sensitive enough to detect BPI–ANCA directed against epitopes within residues 1–193, probably because anionic-rich areas involved in the initial interaction with bacteria or endotoxin may either be shielded when immobilized on an ELISA plate or become available for autoantibodies only after binding to bacterial surfaces. Epitope mapping reveals these epitopes, but misses conformational epitopes whose blocking by BPI–ANCA may mediate inhibition of BPI's antibiotic activity. As a result there is only a poor correlation between the different methods of detection and characterization of BPI–ANCA.

The similarity in pulmonary phenotype between TAP-deficient and CF patients and the high prevalence of BPI–ANCA in both diseases [10,11,32] raises the question for common mechanisms that lead to the generation of these autoantibodies. In chronic mucosal inflammation [33], a loss of tolerance for mucosal antigens, especially microorganisms, has already been described [34]. We could show that BPI–ANCA of four TAP-deficient patients could recognize partially or completely a small C-terminal amino acid sequence of BPI (AESDRLVGELRLDR) that is highly homologous to the outer membrane protein OprC PA 3790 of Ps. aeruginosa (AERDRLIGGLRLDR) [35]. Both the presence of Pseudomonas as well as the permanent endotoxin-mediated neutrophil activation and degranulation in the lung provide locally a constant amount of small BPI and bacterial peptide fragments by neutrophil protease cleavage [26,36]. Thus, based on our findings we hypothesize that proteolytic degradation of BPI at active sites of infection may set free small linear BPI peptides that by means of their proximity to bacterial antigens and persistence elicit antibody responses by reactive B-lymphocytes over time that spread over the whole BPI molecule.

Once BPI–ANCA are generated and circulate in the bloodstream, plasma exudation, an established mechanism of lung tissue to react on bacterial provocations [37,38], allows transition of immunoglobulins in the alveolar space and makes possible binding of BPI–ANCA to BPI. The antibiotic activity of the BPI holoprotein can be neutralized by patients’ BPI–ANCA locally, as demonstrated by our data and blocking experiments on epithelial cells with goat anti-BPI antibodies [29]. Moreover, binding of patients’ BPI–ANCA to the C-terminal portion could not only impair opsonophagocytosis [31], but also convert the neutral delivery and disposal of BPI/LPS complexes to monocytes to an inflammatory process by activating them through their Fc receptors [39]. Although the quantitative role of BPI in the alveolar space is, as yet undetermined, the above-mentioned mechanisms of inhibition of BPI functions may contribute to a diminished bacterial clearance and endotoxin neutralization and create a chronically inflammed alveolar environment that faciliates further NK cell mediated damage in the lung [24].

In summary, we demonstrate that BPI–ANCA are strongly associated with TAP deficiency and can inhibit the antimicrobial function of BPI against Gram-negative bacteria in vitro and may contribute to a diminished local bacterial clearance and a perpetuating pulmonary inflammation.