Protein L is an immunoglobulin (Ig)-binding protein produced by the Gram-positive bacterium Peptostreptococcus magnus that interacts with the variable region of Ig κ light chains. The Ig light chain-binding capacity of protein L gives it the potential to interact with cells expressing surface Ig such as B cells. The present study was performed to address the in vivo trafficking of protein L at both the organ and the cellular level. Using the powerful technique of whole-body autoradiography in a murine model system, we demonstrate specific targeting of protein L to secondary lymphoid tissues in whole-animal analysis. The observed targeting depends on the capacity to interact with murine Ig, as tissue targeting was not apparent in mice given protein H, an Ig-binding protein produced by Streptococcus pyogenes with affinity for human but not murine Ig. Tissue targeting data were combined with flow cytometry analysis, which demonstrated the capacity of protein L to target and activate B lymphocytes in vivo. B cells targeted by protein L had increased surface expression of CD86 and MHC-II, and protein L was present in vacuolar compartments of B cells. Protein L did not bind T cells or natural killer cells but had some capacity to target dendritic cells and macrophages. The data show that protein L preferentially targets secondary lymphoid organs, and activates and is internalized by B cells in vivo. Furthermore, the observed tissue and cell targeting properties require an affinity for murine Ig. These data support the potential use of this Ig-binding protein as a targeting approach to deliver agents to defined cell populations in vivo.
Several genera of Gram-positive bacteria including Staphylococcus aureus, Streptococcus pyogenes and Peptostreptococcus magnus produce surface proteins that bind immunoglobulin (Ig) molecules of several mammalian species (Boyle, 1990). Protein L produced by the Gram-positive anaerobic species P. magnus is one such Ig-binding protein (Björck, 1988). P. magnus is a member of the indigenous flora of the skin, oral cavities, the gastrointestinal tract and the urogenital tract. However, this species is also the causative agent of a variety of infections (Sutter et al., 1985). Approximately 10% of clinical isolates of P. magnus carry and express the gene encoding protein L, and the majority of these isolates are from patients with gynaecological infections. This suggests a role for protein L in the virulence of P. magnus (Kastern et al., 1990).
Protein L binds strongly to human, murine, simian, guinea pig and porcine κ light chains but does not bind bovine or canine light chains (de Château et al., 1993). The ability of protein L to bind Ig is contained within five homologous Ig-binding ‘B’ repeats, each 72–76 amino acids long, arranged sequentially in the protein following a short signal sequence and an N-terminal 76 residue ‘A’ region (Kastern et al., 1992). The B repeats are followed by two 52-amino-acid ‘C’ repeats, a putative cell wall-spanning region and a C-terminal hydrophobic membrane anchor (Kastern et al., 1992). The structure of an IgG-binding domain of protein L has been elucidated using nuclear magnetic resonance (NMR) and consists of four stranded β-sheets and a central α-helix (Wikström et al., 1994). The interactive surfaces of protein L Ig-binding domains and human Ig have been characterized using NMR spectroscopy (Wikström et al., 1995) and biochemical approaches (Beckingham et al., 2001) as well as X-ray crystallography (Graille et al., 2001). The crystal structure of a protein L Ig-binding domain complexed with a human Ig Fab fragment revealed that a single protein L domain can interact simultaneously with the light chains of two Fab molecules, supporting two Ig binding sites per protein L Ig-binding domain (Graille et al., 2001). While one interface involves interactions between the β2 strand and the α-helix of the protein L domain, the other involved the β3 and α-helix. Moreover, the structure demonstrated that the protein L domain interacted with the VL regions outside the antigen combining site (Graille et al., 2001).
The Ig light chain-binding capacity of protein L should allow it to interact with and potentially influence the function of cells expressing surface Ig. Consistent with this, it has been shown previously that protein L can interact with human and murine lymphocytes (Axcrona et al., 1995), and can activate human basophils and mast cells (Patella et al., 1990; Genovese et al., 2000). Moreover, polyvalent forms of protein L cause lymphocyte proliferation (Axcrona et al., 1995). The observations that protein L affects lymphocyte proliferation and binds Ig molecules, which are present on the surface of B cells and act as antigen-capturing receptors, provide the basis for exploiting this molecule as a delivery system to target B cells. Thus, the present study was performed to address the in vivo trafficking of protein L in a murine model using whole-body autoradiography and flow cytometry to characterize tissue and cellular targeting of protein L respectively. This is the first demonstration of specific targeting of an immunoglobulin-binding protein to secondary lymphoid tissues in whole-animal analysis. In addition, the flow cytometry analysis demonstrates the preferential capacity of protein L to target and activate B lymphocytes in vivo. Together, these data underscore the potential use of protein L as a delivery system to target linked agents to defined cell populations and tissues in vivo.
Protein L is detected in splenic white pulp and peripheral lymph nodes without accumulation in non-lymphoid tissues
To determine the organs targeted in vivo after administration of the Ig κ light chain-binding protein, protein L, autoradiographic studies on whole-animal sections following intravenous administration of [125I]-protein L were performed. Sections prepared from whole animals sacrificed between 15 min and 48 h after administration showed that radiolabel was primarily found within the spleen and lymph nodes (Fig. 1A and 2). Initially, at 15 min and 4 h after administration, radioactivity was found in the blood and in the splenic marginal zone (Figs 2 and 3). The marginal zone of the spleen forms the border between the red and white pulp of this organ and contains dendritic cells, a specialized population of ‘marginal zone’ macrophages and marginal zone B cells (Kraal, 1992; De Smedt et al., 1996; Steinman et al., 1997; Crowley et al., 1999; Karlsson et al., 2003). At these early time points, the level of radioactivity detected within the white pulp was low (Figs 2 and 3). However, between 16 and 48 h after administration, radiolabel was primarily found in the splenic white pulp and in the lymph nodes (Fig. 1A and 2; Table 1). The temporal detection of protein L in the splenic marginal zone and the white pulp (Figs 2 and 3) suggests initial interaction of protein L with marginal zone cells such as B cells or dendritic cells. The detection of protein L in the white pulp at later time points could result from movement of free protein L from the marginal zone into the white pulp. Alternatively, protein L could be bound to marginal zone B cells or dendritic cells that move from the marginal zone to the white pulp (De Smedt et al., 1996; Reis e Sousa and Germain, 1999; Karlsson et al., 2003; Lo et al., 2003). The detection of radioactivity over time appeared to be accumulation of radioactivity to certain parts of the splenic pulp and lymph nodes, resulting in a retention of the radiolabel in these tissues as a whole and thus a slower rate of decline than in other tissues (Table 1). Maximum blood ratios in splenic white pulp peaked 24 h after administration of radiolabelled protein L, while the blood ratio in the lymph nodes increased further between 24 and 48 h after injection (Table 1). At 24 and 48 h after administration, a large part of the radioactivity was excreted in the urine. No accumulation of radioactivity was apparent in other non-lymphoid organs including the liver, lung, thymus and bone marrow (Fig. 1A and 2). Furthermore, a decrease in the blood ratio for these tissues was apparent between 16 and 24 h, and diminished further at 48 h (Table 1).
Table 1. . Blood ratio of radioactivity in mice injected with [ 125 I]-protein L or [ 125 I]-protein H.
. The blood ratio in tissues of mice injected with either [ 125 I]-protein L or [ 125 I]-protein H denoted by L and H, respectively, sacrificed at the indicated time after injection is shown. The blood ratio is defined as the ratio of the concentration of radioactivity in the tissue to the concentration of radioactivity in the blood. Protein L binds murine κ light chains ( de Château et al., 1993 ), while protein H, a bacterial Ig-binding protein that binds human and rabbit Ig, but not murine Ig, was used as a negative control ( Gomi et al., 1990 ; Åkesson et al., 1994 ). The proteins were endotoxin free as determined by the LAL assay. After sacrifice, mice were immediately frozen in liquid nitrogen. Frozen carcasses were embedded in a gel of aqueous carboxymethyl cellulose, frozen in hexane, cooled with dry ice and sectioned sagittally. Sections were freeze dried, placed on imaging plates, exposed and analysed using a bioimaging analysis system.
. The values for protein L-injected mice at 1 h, 4 h and 24 h are mean values of two animals at each time point.
. Note that no difference in radioactivity was measurable in the red or white pulp of mice given [ 125 I]-protein H at all time points.
. tWP denotes the blood ratio to the total white pulp.
. WP max denotes the blood ratio to the area of the white pulp giving the most intense signal.
. tRP denotes the blood ratio to the total red pulp.
. Inguinal and axillary lymph nodes.
The experiment was repeated three times with similar results.
No significant radioactivity was apparent in muscle or brain 48 h after injection. At this time, the blood ratio for each of these tissues was 0.05. Similarly, little radioactivity was apparent in the eye. In contrast, accumulation of radioactivity was apparent in the kidney cortex 4 h after administration (Table 1). However, the tissue to blood ratio of radioactivity in this organ decreased at subsequent time points. This suggests that complexes between plasma Ig and protein L are formed and deposited in the kidneys and are subsequently cleared. Indeed, gel filtration chromatography performed on plasma samples from mice injected with [125I]-protein L showed that protein L–Ig complexes in the molecular mass range of 5 × 105−5 × 106 Da were present (data not shown). This supports the view that protein L–Ig complexes were formed in vivo under the experimental conditions used and is consistent with previous findings that protein L can bind soluble Ig and form immune complexes (Berge et al., 1997). Thus, radiolabelled protein L targeted primarily to lymphoid tissues as assessed by high ratios of tissue to blood radioactivity in lymph nodes and splenic white pulp. In contrast, radioactivity was not retained in non-lymphoid organs.
To assess the requirement for binding of host (murine) Ig for protein L to target lymphoid organs, mice were injected with [125I]-protein H. Protein H is an Ig-binding protein with specificity for human IgG heavy chains and Fc fragments (Åkesson et al., 1990; Gomi et al., 1990). Unlike protein L, however, protein H does not bind mouse IgG (Åkesson et al., 1990; 1994; Gomi et al., 1990). Autoradiographic analysis of whole-animal sections prepared from [125I]-protein H-injected mice showed no accumulation of radiolabel in the spleen or lymph nodes (Fig. 1B). Similar to the data for protein L, no accumulation of radioactivity was apparent in other organs including the liver, kidney cortex, lung, thymus or bone marrow (Table 1), and radioactivity was excreted in the urine. Together, these data suggest that the ability of protein L to target the splenic white pulp and lymph nodes results from its capacity to bind host Ig.
Protein L rapidly and preferentially targets B cells in vivo: an ex vivo study at the spleen level
To define the cell population(s) targeted by protein L in vivo, mice were injected with either protein L–fluorescein isothiocyanate (FITC) or bovine serum albumin (BSA)–FITC, and spleens were harvested 0.5, 1, 2, 4 or 12 h after injection. Splenocytes were stained with monoclonal antibodies (mAbs) to detect surface expression of B220, CD4, CD8, CD11c, NK1.1 or F4/80 and analysed by flow cytometry (Fig. 5). Within as little as 30 min after administration of protein L–FITC, gated B220+ splenocytes showed significant FITC staining, with 8% of the B220+ population staining positive (data not shown). The population of B220+FITC+ cells reached a maximum ≈ 2 h after injection, with 60% of gated B220+ cells also staining for FITC (Fig. 5). As 95% of murine Ig light chains are κ and only 5% are λ (Max, 1999), the maximum fraction of B cells detected in the experimental conditions used is not limited by κ to λ light chain usage. No B220+FITC+ double-positive cells were detectable in mice injected with BSA–FITC above the background level of B220+FITC+ staining observed in splenocytes from mice injected with PBS alone at either 2 h after injection (Fig. 5) or the later time points (data not shown). B220+FITC+ splenocytes from mice injected with protein L–FITC were most readily detected between 2 and 4 h after administration, with double-positive cells also being detectable but diminishing in number rapidly at 12 h after injection. At this time point, 2–3% of gated B220+ cells were also FITC+. Similar results were obtained using intraperitoneal (Fig. 5) or intravenous (data not shown) administration of the proteins.
As discussed above, protein L can bind soluble Ig and form immune complexes (Berge et al., 1997), and protein L–Ig complexes were formed in mice injected with protein L. We thus investigated whether the observed binding of protein L to CD11c+ and F4/80+ cells in vivo was mediated by the binding of protein L–serum Ig complexes to the Fcγ receptors on these cells. To address this, flow cytometry analysis was performed on splenocytes from FcRγ–/– mice injected with protein L. FcRγ–/– mice are genetically disrupted in the γ subunit that is a common component of the high-affinity IgG receptor (FcγRI) and the low-affinity IgG receptor (FcγRIII) expressed by dendritic cells and macrophages (Takai et al., 1994; 1996; Pulendran et al., 1997). FITC+CD11c+ cells were still apparent in FcRγ–/– mice injected with protein L, whereas no such cells were detected in the spleen of FcRγ–/– mice injected with BSA–FITC (Fig. 5B). These data suggest that the observed binding of protein L to CD11c+ splenocytes in vivo occurs via mechanism(s) that do not rely on the expression of FcγRI or FcγRIII by these cells.
We next addressed whether the FITC+ B cells detected by flow cytometry contained internalized protein L–FITC. Thus, splenocytes were stained to detect B220 2 h after administration of protein L–FITC, and confocal laser scanning fluorescence microscopy was performed. FITC fluorescence was detected within B220+ cells as distinct punctate staining close to the cell surface, suggesting that the internalized material is concentrated in vacuolar compartments at this time point (Fig. 6). Some cells also appeared to have FITC fluorescence on the cell surface with no intracellular fluorescence detectable, while other B220+ cells had no apparent FITC fluorescence. An interaction with surface Ig on the B220+ cells was required to detect FITC fluorescence in vacuolar compartments, as no detectable FITC fluorescence was apparent either on the surface or within peripheral vesicles of B220+ cells from mice given BSA–FITC (data not shown).
B cells interacting with protein L in vivo display increased surface expression of MHC-II and CD86
B lymphocytes are the cells of the immune system responsible for antibody production. B cell internalization of protein antigen via surface Ig concomitant with their activation, which can be assessed by increased surface expression of MHC-II and the co-stimulatory molecules CD80 and CD86, for example, is a critical step in initiating an antibody response (Kosco-Vilbois et al., 1993; Lenschow et al., 1994; Constant et al., 1995). To examine the activation state of B cells after protein L administration, spleens were harvested 4, 8, 12, 16, 24 or 48 h after injection, and splenocytes were double stained for expression of B220 and CD80, CD86, CD40, MHC-I or MHC-II. Flow cytometry analysis of double-positive cells showed no apparent influence of protein L on surface expression of CD80 or MHC-I on gated B220+ cells at these time points (data not shown). In contrast, surface expression of CD86 and MHC-II increased on B220+ cells from the spleen as early as 4 h after protein L injection (Fig. 7A and 8). This was apparent from the twofold increase in mean fluorescence intensity of CD86 and MHC-II on B220+ cells from protein L-injected mice compared with those that received BSA (Fig. 7A). A similar upregulation of these surface molecules, reflected in increased mean fluorescence intensity, was apparent on B220+ cells from inguinal lymph nodes of protein L-injected relative to BSA-injected mice (data not shown). A slight but reproducible increase in mean fluorescence intensity of CD40 (Fig. 7A) and CD69 (data not shown) expression on B220+ splenocytes from protein L-injected mice was also apparent at this time. Flow cytometry analysis of splenocytes 8 h after injection gave results identical to those observed at 4 h. This indicates that surface expression of CD86 and MHC-II reached a plateau by 4 h or peaked between 4 and 8 h after protein L injection. B220+ splenocytes examined > 8 h after protein L injection had reduced expression of CD86 and MHC-II relative to that observed on cells examined < 8 h after administration (data not shown), with CD86 expression being reduced more rapidly than MHC-II expression. By 48 h after injection, surface expression of both CD86 and MHC-II was back to background levels, that is, the level of expression observed on splenocytes from mice injected with BSA (Fig. 7B). This was apparent from the similar mean fluorescence intensity values of CD86, MHC-II and CD40 on B220+ cells from BSA- or protein l-injected mice 48 h after injection (Fig. 7B).
Bacterial endotoxin (LPS) is a polyclonal activator of murine B cells (Morrison and Ryan, 1979). To ensure that endotoxin in the protein L could not account for the observed effects on B cells, a batch of protein L with minimal endotoxin contamination was chosen for use (50 ng endotoxin mg−1 protein L). In addition, two types of control experiments were carried out to exclude the possibility that the increased surface expression of CD86 and MHC-II on B cells from mice injected with protein L resulted from endotoxin. First, mice were injected with the same amount of Escherichia coli endotoxin as that contained in the 100 µg of protein L used in the above experiments (5 ng) as well as with a 10-fold excess of this amount of endotoxin (50 ng). Mice injected with 5–25 µg of endotoxin were used as a positive control. No increase in CD86 expression was observed on B220+ splenocytes from mice injected 4 h earlier with amounts of endotoxin ranging between 5 ng and 50 ng (n = 11; data not shown). In contrast, B220+ splenocytes from mice given between 5 µg and 25 µg of endotoxin had a large upregulation of CD86 (n = 5; data not shown). In a second approach, CD86 and MHC-II expression on B cells from C3H/HeJ and C3H/HeN mice, which are LPS hyporesponsive or have a normal response to LPS, respectively (Poltorak et al., 1998), was analysed 4 h after administration of protein L. While B220+ splenocytes from C3H/HeJ (LPS hyporesponsive) mice had increased surface expression of CD86 and MHC-II after protein L injection, no alteration in the expression of these molecules was observed after injection of 5 µg of purified endotoxin (Fig. 8). However, C3H/HeN mice responded to both protein L and endotoxin injection by increasing surface expression of CD86 and MHC-II on B220+ cells (Fig. 8). Together, these data demonstrate that the minute quantity of endotoxin in the protein L used for these studies does not account for the observed increase in CD86 and MHC-II after injection of this protein.
Specific targeting of a protein to a defined cell population in vivo may be a means of modulating the host response to an administered agent. Such an approach may enhance the desired effect while minimizing adverse effects on non-targeted cells and tissues. An example of specific tissue and cell population targeting of an administered protein in vivo was shown in the present study by the Ig κ light chain-binding protein, protein L.
Whole-body autoradiographic analysis on sections from mice given [125I]-protein L demonstrated that protein L preferentially targeted secondary lymphoid organs, the spleen and lymph nodes. Furthermore, no accumulation in primary lymphoid organs, the bone marrow and thymus, or non-lymphoid tissue was apparent. Protein L principally targeted the spleen, and a much higher level of radiolabel was present in the white pulp compared with the red pulp. As the former is highly organized lymphoid tissue composed largely of T and B lymphocytes (Picker and Siegelman, 1999), this suggested that protein L preferentially targets the predicted cell population, that is, cells expressing surface Ig or cells that can otherwise interact with Ig by, for example, surface Fc receptors. Preferential binding of protein L to B cells in vivo, and a slight capacity to target Fc receptor-bearing cells, was directly demonstrated by flow cytometry analysis. The observed targeting of protein L to splenic white pulp and its binding to B cells and Fc receptor-expressing cells was a feature of the ability of protein L to bind murine Ig. This was supported by data showing that neither protein H, which binds human but not murine IgG (Åkesson et al., 1990; 1994; Gomi et al., 1990), nor BSA accumulated in the spleen or lymph nodes upon administration to mice.
Some radioactivity was apparent in the liver and kidney cortex after administration of [125I]-protein L. This suggested that complexes between Ig and protein L were formed that accumulated in the kidney and liver upon clearance from the circulation (Johansson et al., 1988; 1996). Indeed, analysis of plasma from mice injected with [125I]-protein L by gel filtration chromatography supported the formation of complexes between protein L and plasma Ig.
Flow cytometry analysis of splenocytes from mice injected with protein L–FITC showed that the major cell population targeted was B220+ cells, B lymphocytes (Johnson et al., 1997). No significant binding of protein L to other splenocytes, such as cells expressing CD4 or CD8 (T cells and certain dendritic cell subsets; Ledbetter et al., 1980; Dialynas et al., 1983; Pulendran et al., 1997; Vremec et al., 2000), or NK1.1 (some T-cell subsets and natural killer cells; Yokoyama and Seaman, 1993; Vicari and Zlotnik, 1996) was apparent after administration of protein L–FITC. The kinetics of protein L interaction with B cells in vivo was rapid and transiently detected. FITC+B220+ splenocytes were apparent within 30 min of administration with few FITC+ cells remaining at 12 h. The rapid reduction in the number of FITC+B220+ cells detected may result from degradation of protein L–FITC internalized by B cell surface Ig (Lanzavecchia, 1990; Watts, 1997). In support of this, confocal microscopy showed that protein L was internalized into vacuoles of B220+ cells from mice injected with protein L–FITC. However, additional studies are needed to address more specifically the intracellular fate of protein L and whether fusion of antigens to protein L alters their degradation and/or antigen presentation by B cells. Alternately, FITC+ B cells could migrate out of the spleen or die within the spleen. Deletion of B cells is a potential outcome of an encounter between peripheral B cells and multivalent antigen (Finkelman et al., 1995; Lang and Nemazee, 2000). The occurrence of any of these mechanisms would no longer allow detection of FITC+ B cells by flow cytometry.
We predicted that the mechanism of protein L interaction with B cells in vivo was due to direct interaction with immunoglobulin on the surface of the these cells. The data presented here support this prediction. First, no FITC+ B cells were detected after injection of mice with BSA–FITC, ovalbumin–FITC (data not shown) or protein H–FITC (data not shown), proteins that do not have the capacity to bind murine Ig. In addition, no binding of protein L to T cells or natural killer cells, cell populations that do not express Ig on their surface, was apparent after injection of protein L–FITC. Together, these data suggest that the targeting of protein L to B cells in vivo is mediated by direct interaction of protein L to surface Ig on B cells.
Protein L showed a strong preference for targeting B cells in vivo. However, small populations of FITC+CD11c+ and FITC+F4/80+ splenocytes, dendritic cells and macrophages were also detected in mice given protein L–FITC. In contrast, no such cells were found after injection with BSA–FITC. Thus, it is unlikely that FITC+CD11c+ cells in mice given protein L–FITC arise from passive transfer of FITC into CD11c+ cells or endocytic uptake of free protein. Furthermore, FITC+CD11c+ cells were still apparent in FcRγ–/– mice injected with protein L. This suggests that FcγRI or FcγRIII-mediated uptake of Ig–protein L complexes does not account for the FITC staining of CD11c+ cells after administration of protein L. The ability of protein L to target CD11c+ and F4/80+ splenocytes may still, however, depend on the capacity of protein L to bind murine Ig. Protein L–Ig complexes internalized by macropinocytosis (Rodriguez et al., 1999) may account for the presence of FITC+ macrophages and dendritic cells after administration of protein L–FITC. Alternatively, FcγRII, which is still present on dendritic cells and macrophages, as well as on B cells, in FcRγ–/– mice (Takai et al., 1994; 1996), may be involved in internalization of protein L–Ig complexes. Experiments in FcγRII–/– mice were performed in attempt to address this possibility. However, protein L had a toxic effect when injected into FcγRII–/– mice. This may be caused by the capacity of protein L to exert an unregulated stimulatory effect on the B cells in these mice, as FcγRII–/– functions as a negative regulator of immune complex-triggered activation (Takai et al., 1996). Although in vitro studies have shown that protein L can trigger release of inflammatory cytokines by mast cells and basophils (Patella et al., 1990; Genovese et al., 2003), whether protein L has a pathogenic effect on B lymphocytes or other cells is not yet known. Considering the in vivo targeting of protein L described here, it is noteworthy that, although protein L is a surface-associated molecule, substantial amounts of the protein are released from the bacterial surface (Kastern et al., 1990). However, the pathogenic consequences of cellular targeting by protein L in vivo remain to be established.
Our data also show that the interaction between protein L and B cells in vivo results in B-cell activation. Splenic B cells showed upregulation of MHC-II and CD86, surface molecules important in interacting with helper T cells during an immune response, after administration of protein L. No such activation was observed in mice given BSA. This suggests that the upregulation of these molecules required interaction of the administered protein with B-cell surface Ig. The protein L used in this study contains four tandem Ig-binding domains, a feature that may enhance the capacity to cross-link surface Ig on B cells. Alternatively, the ability of a single Ig-binding domain of protein L to interact with the light chain of two Ig molecules in a sandwich fashion (Graille et al., 2001) may be involved in cross-linking Ig when protein L interacts with a B cell. Other reagents that cross-link surface Ig on B cells are known to activate the cells and upregulate surface expression of co-stimulatory molecules such as CD86 (Kosco-Vilbois et al., 1993; Ho et al., 1994; Lenschow et al., 1994; Constant et al., 1995). Despite the interaction of protein L with B cells in vivo, however, injection of protein L did not result in increased total serum Ig. That is, the total serum level of IgM and IgG1 analysed 4, 7 or 14 days after administration of 100 µg of protein L was similar to that in mice injected with either PBS or BSA (data not shown). Likewise, the total level of IgM and IgG1 in protein L-treated mice did not differ from that in serum taken before protein L administration.
Together, our data suggest that protein L can be used as an agent specifically to target a linked protein or compound to secondary lymphoid tissues and to B cells in vivo. In the case of a fused protein, the interaction between protein L and B cells, which also activates the B cells, may facilitate the initiation of an immune response by enhancing T-cell co-stimulation and thus T-cell activation (Cassell and Schwartz, 1994; Ho et al., 1994). Interestingly, a fusion protein between a non-toxic form of cholera toxin and two domains from Staphylococcus aureus protein A that confer binding to Ig heavy chains administered to mice also targets B cells in vivo (Ågren et al., 1997; 2000). Thus, the approach of targeting administered agents to a defined cell population in vivo holds promise for new prophylactic and therapeutic disease treatments.
Unless otherwise stated, C57BL/6 mice were used. Mice used in the whole-body autoradiographic studies were kept at the animal facilities at Active Biotech AB (Lund, Sweden), whereas those used in other experiments were kept in the facilities at Lund University. Mice were purchased from M&B (Ry, Denmark). C3H/HeN and C3H/HeJ mice were bred and kept at Lund University. Mice deficient in the FcR common γ chain, FcRγ–/– mice (Takai et al., 1994), were generously provided by Nils Lycke (Department of Clinical Immunology, Göteborg University, Göteborg, Sweden). Mice were used at 6–10 weeks of age.
Phycoerythrin (PE)-conjugated mAbs against mouse CD45R/B220 (RA3-6B2), CD40 (3/23), CD86 (GL1), CD4 (L3T4), CD69 (H1.2F3), CD11c (HL3), NK1.1 (PK136), I-Ab (AF6-120.1), mouse IgG2aκ (G155-178, isotype control) and rat IgG2aκ (R35-95, isotype control) were purchased from Pharmingen. Biotinylated anti-mouse CD80 (RMMP-1) and PE-conjugated anti-mouse CD8α (CT-CD8α) were purchased from Caltag Laboratories. mAbs from hybridoma 2.4.G2 (anti-FcγRII/III; Unkeless, 1979), RA3.6B2 (anti-CD45R/B220; Coffman, 1982), K9.178 (anti-Kb; Hämmerling et al., 1982), F4/80 (Austyn and Gordon, 1981) and 14.4.4S (anti-I-Ek; Ozato et al., 1980) were purified from culture supernatants using a γ-bind plus column (Amersham Pharmacia Biotech) and labelled with FITC (Sigma Chemical) or biotin (Sigma Chemical). Streptavidin–FITC (Vector Laboratories) and streptavidin–PE (Vector Laboratories) were used as second-step reagents.
The protein L used throughout these studies was a recombinant form produced in E. coli. The recombinant protein L consists of the four tandem B1–B4 Ig-binding domains (Kastern et al., 1992), i.e. the A, B5, C1 and C2 repeats as well as the cell wall-spanning region are removed (see Kastern et al., 1992 for a schematic diagram of the domain structure of protein L that illustrates the domains of protein L used in these studies). Recombinant protein L was obtained as lyophilized rProtein L™ from ACTIgen. The BSA used was IgG-free, low-endotoxin fraction V (Sigma Chemical). Both protein L and BSA were reconstituted in endotoxin-free phosphate-buffered saline (PBS) to a concentration of 1 mg ml−1. The proteins were assayed for endotoxin content using Limulus amoebocyte lysate (LAL) Pyrotell®-T (The Associates of Cape Cod) with E. coli endotoxin as a standard according to the manufacture's protocol. Various lots of protein L were screened for endotoxin content, and a lot that contained < 50 ng of endotoxin mg−1 protein L was chosen and used for all studies. The endotoxin-free nature of the BSA used throughout the studies was confirmed using the LAL assay. Protein H, a bacterial Ig-binding protein with affinity for human and rabbit IgG but not for murine IgG (Gomi et al., 1990; Åkesson et al., 1994), was used as a control protein that does not bind murine Ig in the animal section studies. The recombinant protein H used lacked the membrane-spanning region and was expressed and purified as described previously (Åkesson et al., 1990). The protein concentration of protein H was assessed optically by determining the absorbance at 280 nm (Nilson et al., 1995). FITC-labelled proteins were prepared using standard methods for experiments requiring directly labelled proteins.
Radioactive labelling of proteins
Fifty micrograms of either protein L or protein H was labelled with N-succinimidyl-4-(125I) iodobenzoate to a specific activity of 5.5 and 17 µCi µg−1 respectively. The proteins were dialysed against PBS (pH 7.4) before labelling. The iodine-containing preparations were purified by preparative chromatography using a PD-10 desalting column (Amersham Pharmacia Biotech), and fractions containing the highest amount of labelled protein were pooled. The final injection solution was prepared by dilution to a concentration of 5.0 µg ml−1 and stored at −70°C until use.
Analysis of labelled proteins and plasma samples
The radioactivity of the labelled proteins and plasma samples was analysed by γ-counting (Packard Cobra Auto-gamma 5003), and data were corrected for the detector efficiency and the decay of 125I. The radioactive proteins and plasma samples were also analysed by high-performance liquid chromatography (HPLC) to determine the total radioactivity. For this analysis, 50 µl of a radiolabelled protein diluted 1:40 or 50 µl of undiluted plasma was injected into a Waters 717 Auto Sampler chromatographic system. The protein L or protein H samples were injected on a Superdex 200 column (Amersham Pharmacia Biotech) while the plasma samples were injected on a TSK G 3000 SW column (7.5 × 600 mm; Tosohaas) with a 2 µm prefilter (Upchurch Scientific) and a TSK G 3000 SW guard column (7.5 × 75 mm; Tosohaas). The samples were eluted with PBS (pH 7.4) at a flow of 0.75 ml min−1 (protein L or protein H) or 1 ml min−1 (plasma samples) at ambient temperature. Detection was performed with a UV detector (A280 nm, Waters 486) and a radioactivity detector (Flo-One, A-515-AX; Packard Instruments) coupled in series. Data were collected for 45–60 min depending on the sample. Fractions containing complex-bound proteins, as well as dimers or larger aggregates, were observed as high-molecular-weight peaks appearing earlier than monomers. Fractions containing iodinated fragments, N-succinimidyl-4-(125I) iodobenzoate reagent or free iodine were observed as low-molecular-weight peaks appearing later in the chromatogram. The radioactivity chromatograms were integrated, and the area of the radiolabelled test compound peak was expressed as a percentage of the total area.
Animals and experimental design of whole-section studies
Eight female C57BL/6 mice were injected intravenously with a single dose of 50 µg kg−1 protein L, and eight additional mice were injected with 50 µg kg−1 protein H. Six animals from each administration group were used for autoradiography. For both protein L and protein H, mice were killed 1, 4, 16, 24 or 48 h after protein administration. An additional mouse was sacrificed 15 min after administration of protein L or 8 h after injection of protein H. After sacrifice, mice were immediately frozen in liquid nitrogen and processed as described below. The two other animals in each administration group were used for plasma analysis by HPLC. A mouse from each group was sacrificed 1 h after injection and another was sacrificed 24 h after injection; blood was collected by heart puncture into heparinized syringes and transferred into plastic tubes. The tubes were immediately placed in an ice–water bath and centrifuged at 1300 g for 15 min at 4°C. Plasma was separated and stored at −70°C for bioanalysis.
After killing, frozen carcasses were embedded in a gel of aqueous carboxymethyl cellulose (CMC), frozen in hexane, cooled with dry ice (−80°C) and sectioned sagittally at −20°C using a cyromicrotome (CM3600; Leica) as described previously (Ullberg, 1954; 1977). Briefly, 20 sections with a thickness of 10 µm or 20 µm were cut from each animal at different levels and caught on tape (no. 810; Minnesota Mining and Manufacturing). After freeze drying at −20°C for at least 24 h, sections were put on imaging plates (Fuji Photo Film) and exposed for 3 days. The imaging plates were analysed using a bioimaging analysis system (Bas 2000, Fuji Photo Film) with aida software (Raytest Isotopenmessgeraete).
Preparation of 125I-radio standard
A 10% (w/w) gelatin solution was prepared at a temperature of ≈ 50°C. A volume of 350 µl of the 125I-labelled dose solution was mixed with 9.7 ml of the gelatin solution. Eight 1:2 dilutions were then prepared at 50°C in a similar fashion. These solutions were frozen rapidly in small gelatin capsules and mounted vertically in a prefrozen CMC block at −20°C. From this block, 20-µm-thick sections were cut at different levels and caught on tape in the same way as that described for the animals and used as 125I-standards. Pieces from seven capsules were counted in triplicate in a gamma counter (Packard Cobra Auto-gamma 5003) to determine the radioactivity of the 125I-standards (Günther et al., 2000). The data obtained were also corrected for the detector efficiency and the decay of 125I.
The qualitative autoradiographic findings were complemented with quantitative data as measured by radioluminography (Günther et al., 2000). Imaging plates were exposed to the radioactive tissue sections and 125I-standards and analysed as described above. The obtained section image was displayed on a monitor, and regions of interest were delineated for calculation of the average radioactivity concentration within each selected area. The imaging results were transformed into concentration in c.p.m. mm−3 using the 125I-standards. The obtained values were approximated and set to c.p.m. mg−1 tissue.
Immunohistochemistry on whole-body sections
Ten micrometre serial whole-body sections (including the spleen) from animals injected with [125I]-protein L were prepared and transferred to glass slides using macro tape transfer technology (Instrumedics). Sections were fixed at 4°C for 10 min in acetone, quickly air dried and stained with mAbs against CD19, CD3, F4/80 and CD11c using a conventional avidin–biotin immunoperoxidase technique as recommended by BD Biosciences Pharmingen. Briefly, sections were incubated in avidin–biotin for 15 min and washed twice in PBS containing 20% fetal bovine serum (FBS) to block endogenous avidin binding sites. Primary antibodies against CD19 (clone 1D3; BD Biosciences Pharmingen), CD3 (clone 145-2C11; BD Biosciences Pharmingen), F4/80 (clone A3-1; Serotec) or CD11c (clone N418; Serotec) were added, and sections were incubated at room temperature for 1 h. Primary antibodies were detected with biotinylated goat anti-Syrian hamster IgG (Jackson ImmunoResearch Laboratories) for CD3 and N418 and with biotinylated goat anti-rat Ig (Jackson ImmunoResearch Laboratories) for CD19 and F4/80. After incubation for 30 min, sections were washed with PBS containing 20% FBS, avidin-biotinylated horseradish peroxidase complex (Vector Laboratories) was added, and sections were incubated for another 30 min. The staining was visualized using diaminobenzidine (DAB; Saveen) and counterstained with haematoxylin (Merck). Sections were assessed microscopically using a Leitz Aristoplam microscope and Leica q win software. Final images were produced using Adobe photoshop 5.0.
Administration of proteins for ex vivo flow cytometry analysis
C57BL/6 mice were injected intraperitoneally or intravenously with 100 µg of either protein L or BSA in 200 µl of PBS. Similarly, mice were injected with 200 µl of the PBS used as the diluent as negative controls. After administration, mice were sacrificed at the times specified for the various experiments, and spleens were harvested aseptically. Single-cell suspensions were prepared, and red blood cells were lysed by resuspending the splenocytes in ice-cold water for 10 s. The cells were then washed in fluorescence-activated cell sorting (FACS) buffer and stained with mAbs for flow cytometry. FACS buffer is Hanks' balanced salt solution (HBSS; Life Technologies) supplemented with 3% FCS, 0.01% sodium azide, 5 mM EDTA and 10 mM Hepes. mAb dilutions and washes were all done in FACS buffer. Before staining, cells were incubated on ice for 20 min in 50 µl of mAb 2.4.G2 (10 µg ml−1) to block FcγRII/III receptors.
To determine the specific cell populations bound by protein L in vivo, spleen cells from mice injected with FITC-conjugated protein L (protein L–FITC) or BSA (BSA–FITC) were stained with PE-conjugated mAbs against B220, CD8, CD4, CD11c or NK1.1 or with biotinylated anti-F4/80 in 50 µl of FACS buffer for 20 min on ice. Flow cytometry was performed on a Becton Dickinson FACSort flow cytometer (BD Biosciences Pharmingen), and data were collected on 500 000 gated lymphocytes. Dead cells were excluded by staining with 7-amino-actinomycin D (7AAD) at 1 µg ml−1 (Sigma Chemical). To ascertain the activation state of B cells after protein L injection, cells were stained with FITC-conjugated anti-B220 in combination with PE-conjugated anti-CD80, CD86, CD40, MHC-I or MHC-II- for 20 min on ice before analysis. Flow cytometry was performed as above, and data were collected on 100 000 gated B220+ cells.
For confocal microscopy studies, mice were injected intraperitoneally with 100 µg of protein L–FITC or BSA–FITC. Forty-eight hours later, splenocytes were harvested and stained on ice with biotinylated anti-B220 followed by streptavidin–Alexa Fluor 594 (Molecular Probes). Cells were kept on ice at all possible steps, and cold buffers and reagents were used. The cells were fixed in 1% paraformaldehyde on ice for 30 min and then at room temperature for another 30 min before being added to polylysine-treated slides and mounted with ProLong Antifade (Molecular Probes). The cells were inspected by epifluorescence and laser scanning confocal microscopy using Bio-Rad MRC-1024 confocal equipment attached to a Nikon Eclipse E800 upright microscope using a 522 ± 32 nm bandpass filter to detect green fluorescence and a 598 ± 40 nm bandpass filter to detect red fluorescence.
The authors are grateful to Anders Håkansson for help with confocal microscopy, Ulf Sjöbring for helpful discussions, Inga-Maria Frick for protein H, and Nils Lycke for FcRγ–/– and FcγRII–/– mice. Technical assistance with whole-body autoradiography by Karin Hallbeck, labelling of protein L by Jonas Tuvesson, and immunohistochemistry by Ann-Sofi Pålsson are also gratefully acknowledged. This study was supported by funding from the Swedish Research Council (projects 621-2001-1720 and K2003-06Z-07480-18B).