Clearance mechanism of a mannosylated antibody–enzyme fusion protein used in experimental cancer therapy

MFECP1 is a mannosylated antibody–enzyme fusion protein used in antibody-directed enzyme prodrug therapy (ADEPT). The antibody selectively targets tumor cells and the targeted enzyme converts a prodrug into a toxic drug. MFECP1 is obtained from expression in the yeast Pichia pastoris and produced to clinical grade. The P. pastoris -derived mannosylation of the fusion protein aids rapid normal tissue clearance required for success-ful ADEPT. The work presented provides evidence that MFECP1 is cleared by the endocytic and phagocytic mannose receptor (MR), which is known to bind to mannose-terminating glycans. MR-transfected ﬁbroblast cells internalize MFECP1 as revealed by ﬂow cytometry and confocal microscopy. Immunoﬂuorescence microscopy shows that in vivo clearance in mice occurs predominantly by MR on liver sinusoidal endothelial cells, although MR is also expressed on adjacent Kupffer cells. In the spleen, MFECP1 is taken up by MR-expres-sing macrophages residing in the red pulp and not by dendritic cells which are found in the marginal zone and white pulp. Clearance can be inhibited in vivo by the MR inhibitor mannan as shown by increased enzyme activities in blood. The work improves understanding of interactions of MFECP1 with normal tissue, shows that glycosylation can be exploited in the design of recombinant anticancer therapeutics and opens the ways for opti-mizing pharmacokinetics of mannosylated recombinant therapeutics.

MFECP1 is a mannosylated antibody -enzyme fusion protein used in antibody-directed enzyme prodrug therapy (ADEPT). The antibody selectively targets tumor cells and the targeted enzyme converts a prodrug into a toxic drug. MFECP1 is obtained from expression in the yeast Pichia pastoris and produced to clinical grade. The P. pastoris-derived mannosylation of the fusion protein aids rapid normal tissue clearance required for successful ADEPT. The work presented provides evidence that MFECP1 is cleared by the endocytic and phagocytic mannose receptor (MR), which is known to bind to mannose-terminating glycans. MR-transfected fibroblast cells internalize MFECP1 as revealed by flow cytometry and confocal microscopy. Immunofluorescence microscopy shows that in vivo clearance in mice occurs predominantly by MR on liver sinusoidal endothelial cells, although MR is also expressed on adjacent Kupffer cells. In the spleen, MFECP1 is taken up by MR-expressing macrophages residing in the red pulp and not by dendritic cells which are found in the marginal zone and white pulp. Clearance can be inhibited in vivo by the MR inhibitor mannan as shown by increased enzyme activities in blood. The work improves understanding of interactions of MFECP1 with normal tissue, shows that glycosylation can be exploited in the design of recombinant anticancer therapeutics and opens the ways for opti-Introduction Antibody-directed enzyme prodrug therapy (ADEPT) is a twophase cancer treatment that uses a systemically administered antitumor antibody-enzyme conjugate to localize enzyme in tumors (Bagshawe 1987). After sufficient clearance from normal tissues, a prodrug is administered (Bagshawe et al. 1988;Bagshawe 1995) and is selectively converted into an active cytotoxic drug by enzyme at the tumor site. The first ADEPT system used in clinical trials employed a nonglycosylated chemical conjugate of the bacterial enzyme carboxypeptidase G2 (CP) from Pseudomonas sp. strain RS-16 and the F(ab) 2 fragment of A5B7, a monoclonal antibody to carcinoembryonic antigen (CEA) (Bagshawe 1995;Bagshawe et al. 1995;Napier et al. 2000). Persistence of this conjugate in blood necessitated the additional administration of a galactosylated antibody to CP in order to clear excess antibody-enzyme from circulation prior to administration of prodrug (Napier et al. 2000). Clearance of A5B7-CP was most likely via the galactose-specific hepatic asialoglycoprotein receptor (Sharma et al. 1994;Stockert 1995).
A subsequent and more direct approach for obtaining rapid clearance is to use a recombinant antibody-enzyme fusion protein produced in the methylotrophic yeast Pichia Pastoris, a eukaryotic microorganism which performs posttranslational glycosylation with mannose (Trimble et al. 1991;Bretthauer and Castellino 1999;Cereghino and Cregg 2000). A particular advantage of P. pastoris as opposed to Saccharomyces, which adds hypermannosylation to proteins, is the significantly shorter chain-length of the high-mannose structures. Furthermore, a number of P. pastoris products are already in clinical development (Herbst et al. 2002;Gerngross 2004;Malkin et al. 2005;Kobayashi 2006) demonstrating feasibility and safety of using these in man.
We previously created a fusion protein of CP with MFE-23, a single-chain Fv antibody to CEA, and exploited P. pastoris mannosylation to clear the MFE-CP fusion protein from blood (Medzihradszky et al. 2004;Sharma et al. 2005) prior to administration of prodrug. The P. pastoris expressed nonclinical grade fusion protein of the CEA-binding antibody (MFE-CP) was shown to be glycosylated with high-mannose structures of 5 -13 residues at two of the three potential N-glycosylation sites of each unit of the dimeric enzyme (Medzihradszky et al. 2004). Blood clearance of MFE-CP via liver was rapid leading to tumor-to-plasma ratios of over 1000:1 in human colon carcinoma xenograft models 6 h after administration (Medzihradszky et al. 2004;Sharma et al. 2005) and effective therapy in combination with a bisiodo-phenol mustard prodrug (Sharma et al. 2005). The first clinical grade MFE-CP, which is abbreviated to MFECP1, is made to GMP standards in P. pastoris and is currently in Phase I/II clinical trials where similar hepatic clearance is observed Mayer et al. 2004). We hypothesized that, in both mice and man, blood clearance of MFE-CP is mediated by the mannose receptor (MR) interacting with P. pastoris-derived high-mannose structures.
MR is a member of the C-type lectin family (Taylor et al. 1990), and is an endocytic and phagocytic pattern recognition receptor that binds in a calcium-dependant manner to mannose, fucose, and N-acetylglucosamine terminating glycans (Largent et al. 1984). In liver, it is found on sinusoidal endothelium and Kupffer cells (Takahashi et al. 1998), where it acts as a molecular scavenger by clearing pathogenic microorganisms and endogenous glycoproteins with high mannose structures (Taylor et al. 2005). MR also plays an important role in cell activation and antigen presentation (Linehan 2005;Taylor et al. 2005;Martinez-Pomares et al. 2006) and, therefore, could be involved in the immune response to glycosylated protein therapeutics.
In the present study, we investigate the interaction of MFECP1 with MR-expressing cells and provide in vitro and in vivo evidence that the antibody -enzyme fusion protein clears via MR. Furthermore, we show that the in vivo clearing cells are predominantly sinusoidal endothelial cells and not Kupffer cells. We also investigate uptake of MFECP1 in potential antigen presenting cells (APC) in the spleen and show that this occurs by MR-expressing macrophages and not dendritic cells. Finally, we demonstrate that pharmacokinetics of MFECP1 can be significantly modified by inhibition of MR in vivo.

Analysis of MFECP1
MFECP1 (clinical grade MFE-CP) was produced at a final yield of 80 mg. The purified material gave a characteristic profile for this protein when tested by size-exclusion chromatography ( Figure 1A). Functional activity of the CP moiety was confirmed by testing for enzyme activity which was found to be 124 U/mg, where 1 U is the amount of enzyme required to hydrolyze 1 mmol of methotrexate per min at 37 8C.

N-glycan analysis of MFECP1
The degree of mannosylation of MFECP1 was determined by DNA sequencer-assisted fluorophore-assisted carbohydrate electrophoresis (DSA-FACE) analysis ( Figure 1B). This analysis involves derivatisation of the PNGase F released carbohydrate chains with 8-amino-1,3,6-pyrenetrisulfonic acid and separation by polyacrylamide electrophoresis (Callewaert et al., 2001). Comparison of the profile of MFECP1 with that of RNase B shows that Man 8 GlcNAc 2 and Man 9 GlcNAc 2 are the predominant carbohydrate structures ( Figure 1B, Panels 2 and 3, respectively). In addition, higher mannosylated and lower mannosylated structures are also present, however, at a lower abundance than the Man 8 GlcNAc 2 and Man 9 GlcNAc 2 structures. This analysis agrees with previously published mass spectrometry data on MFE-CP (Medzihradszky et al. 2004). The liquid chromatography mass spectrometry (LC/MS) analysis of trypsin digested glycopeptides identified two glycosylation sites, Asn-442 containing Man 5 GlcNAc 2 to Man 13 GlcNAc 2 structures and Asn-492 containing Man 8 GlcNAc 2 to Man 10 GlcNAc 2 structures.

Internalization in vitro of MFECP1 by an MR-transfected fibroblast cell line
To determine whether MR has the ability to internalize the mannosylated MFECP1, an MR-transfected rat fibroblast cell was incubated with the antibody-enzyme fusion protein at 37 8C for various time intervals and analyzed by flow cytometry ( Figure 2A, Table I). The cell line has previously been shown to mediate efficient endocytosis of mannose-BSA (Taylor et al. 1992;Taylor et al. 1990). Considerable fluorescence intensity and, hence, internalization was observed for MFECP1 after 30 min when detected with an anti-CP antibody followed by R-phycoerythrin (R-PE) labeled IgG and compared to the omission of the anti-CP control experiment. The median fluorescence intensity shows that the highest internalization of MFECP1 was observed after 45 min and internalization decreased at the 1-h and 5-h intervals. The fluorescence intensity represents only internal MFECP1 because no fluorescence above background levels was observed when the cells were incubated with the antibodies prior to fixation (data not shown) to reveal surface binding. MFECP1 is Mannosylated antibody-enzyme clearance by mannose receptor expected to be removed from the cell surface during cell detachment with EDTA-trypsin due to the Ca 2þ dependency of the carbohydrate -MR interaction. Addition of mannan at 37 8C to the fibroblast cell/MFECP1 dish prevented internalization of the glycoprotein. Mannan is a known inhibitor of carbohydrate -MR interactions (Taylor et al. 1992) and, therefore, indicates that internalization of MFECP1 is specifically mediated by MR.
When MFECP1 was incubated at 4 8C for 1 h, the antibodyfusion protein was also not internalized, the observed fluorescence was comparable to the negative control. Furthermore, MFECP1 was not internalized by rat-fibroblasts transfected with an empty vector (data not shown).
Confocal microscopy was used to visualize internal MFECP1 ( Figure 2B). MFECP1 was incubated at 37 8C for 1 h with MR-transfected fibroblast cells described under Internalization in vitro of MFECP1 by an MR-transfected fibroblast cell line. The cell was labeled with the cell tracer, carboxyfluorescein diacetate succinimidyl ester (CFSE). MFECP1 was clearly visible inside the cell when revealed with Alexa fluor594 labeled goat antimouse IgG that binds to anti-CP. The internal MFECP1 was visualized as red spots, indicating its location in internal vesicles. For the otherwise similar anti-CP omission control experiment, no MFECP1 was visible.  Figure 3A), indicating that MFECP1 is cleared also in vivo in liver by MR. Two cell types in the liver, sinusoidal endothelial cells and liver residence macrophages, Kupffer cells, express MR. In order to address whether there is a preference of clearance of MFECP1 by one of these cell types, liver tissue was stained for CD31, an antigen highly expressed on endothelial cells. Merging of the images of CD31 and MFECP1 staining showed considerable colocalization of these two molecules ( Figure 3B and D). Therefore, MFECP1 appears to be cleared by the endothelial cells. Staining of liver tissues with a pan macrophage antibody that recognizes the F4/80 antigen (expressed on almost all macrophages) shows distinct locations of Kupffer cells, and MFECP1 and very little colocalization ( Figure 3C). Hence, MFECP1 appears to be cleared in liver predominantly by endothelial cells and not by the Kupffer cells although both cell types express MR. To substantiate this finding, the immunofluorescence staining was repeated and Kupffer cells were visualized using anti-CD68, an independent macrophage marker that recognizes the late endosomal glycoprotein macrosialin. Results obtained by merging of the image of CD68 staining with that of MFECP1 showed that the majority of Kupffer cells do not colocalize with the enzyme-fusion protein ( Figure 3E). In conclusion, endothelial cells are the preferred cell type involved in clearance of MFECP1 in the liver.

Localization of MFECP1 in vivo in spleen
Mechanisms of clearance of MFECP1 were investigated in the spleen. Five minutes after i.v. injection of MFECP1, the  H. Kogelberg et al. fusion protein accumulated in the spleen in the red-pulp area and uptake by MR and macrophages was investigated ( Figure 4A). Merging of spleen images stained for MFECP1 and MR showed colocalization of the two proteins and that MR may be involved in clearance of MFECP1 in the spleen. Triple staining with the F4/80 recognizing pan macrophage antibody shows colocalization of MR, MFECP1, and macrophages, which accumulate in the redpulp area of the spleen. Therefore, it appears that MR expressed on macrophages clears the antibody -enzyme fusion protein in the spleen. The spleen sections were also triple stained for MFECP1, MR, and dendritic cells in order to investigate whether these cells would be involved in clearing the fusion protein ( Figure 4B). Dendritic cells are found in the marginal zone and in the white-pulp area of the spleen, in a different location to the MR-expressing macrophages. Dendritic cells in the spleen do not express MR and are not involved in clearing of MFECP1.

Inhibition of clearance of MFECP1 by mannan in vivo
Having established that MR is involved in the clearance of MFECP1 in the liver and spleen in vivo in mice, it was investigated whether prior addition of the known MR inhibitor mannan would inhibit clearance. Enzyme activity levels were measured in blood between 5 min and 24 h after MFECP1 Mannosylated antibody-enzyme clearance by mannose receptor injection with or without prior intraperitoneal injection of mannan ( Figure 5) using two cohorts of mice as described in the Materials and methods section. MFECP1 clearance in blood is very rapid and undetectable after 5 h without addition of mannan. At 5 min and 15 min after MFECP1 injection, the enzyme levels were comparable with or without addition of mannan, mean values (of three mice) of 11.9 + 1.0 with and 11.36 + 0.7 without and 9.3 + 0.6 with and 7.9 + 0.5 without mannan, respectively. The later time points showed considerable increases in enzyme levels with prior addition of the MR inhibitor mannan, 2-fold after 30 min (7.1 + 2.4 with and 3.5 + 1.0 without mannan), 27fold after 1 h (6.7 + 0.3 with and 0.25 + 0.02 without mannan), 75-fold after 3 h (4.64 + 0.7 with and 0.062 + 0.07 without mannan) and 206-fold after 5 h (4.1 + 0.5 with and 0.02 without mannan). After 24 h, enzyme levels in blood fall below detection levels, also when mannan is added prior to MFECP1. Inhibition of blood clearance of MFECP1 by mannan provides further evidence of the specificity of clearance by MR and, in addition, shows that pharmokinetics of MFECP1 can be changed when MR is inhibited.

Discussion
This study investigates the clearance mechanism of a recombinant glycoprotein therapeutic used in experimental cancer therapy. The glycoprotein, MFECP1, is a fusion of a bacterial enzyme and an scFv antibody and would not be glycosylated in native form. However, expression in P. pastoris X33 strain has led to addition of high-mannose type oligosaccharides, which were found to be advantageous in achieving rapid clearance of the recombinant protein from noncancerous tissues in vivo. This is a prerequisite for prodrug treatment in ADEPT in which enzyme activity remaining in the blood can activate prodrug and cause systemic toxicity. It was hypothesized that clearance was mediated via MR in mice and man. Our results demonstrate that MFECP1 is specifically internalized by human MR-transfected fibroblast cells in vitro and that internalization is blocked by the known MR inhibitor mannan. In vivo, we show by immunofluorescence that MFECP1 is specifically internalized by MR-expressing cells in the liver and spleen in mice. We also demonstrate that this uptake is blocked by pre-injection with mannan, since approximate 200-fold higher enzyme levels were measured in blood 5 h post administration of MFECP1 in the presence of the inhibitor compared to administration of MFECP1 alone.
Mannose is a preferred ligand for MR (Kery et al. 1992) and MFECP1 homo-dimer contains four N-linked highmannose structures which were shown by tandem mass spectrometry to extend from 5 to 13 units in length in nonclinical grade MFE-CP (Herbst et al. 2002). In our current study, we demonstrate by DSA-FACE experiments that the same mannosylation pattern is present on the clinical product, MFECP1. MR uptake of proteins with similar mannosylation to MFECP1 has been reported by several other workers and in some cases has been shown to depend on monosaccharide density because efficient uptake was obtained by addition of at least 16 mannose residues. These studies have shown uptake of mannosylated albumins by MR in isolated perfused rat livers (Jansen et al. 1991), hepatic sinusoidal cells (Taylor et al. 1987) and in vivo in mice (Opanasopit et al. 2001). Other glycoproteins, including RNase B have been shown to be rapidly cleared by aleveolar macrophages, mediated by their terminating GlcNAc and/or mannose residues (Stahl et al. 1978). RNase B contains one N-linked glycosylation site of high-mannose type with 5 -9 mannose residues (Joao and Dwek 1993).
In our studies, MFECP1 clearance occurred rapidly in liver in mice. MR is expressed in liver on two cell types, Fig. 4. MFECP1 colocalizes in the spleen with the mannose receptor (MR) and macrophages, which express MR (A) but not with dendritic cells, which do not express MR (B). Mice were injected i.v. with 0.3 mL MFECP1 (approximately 25 U) and tissue was removed after 5 min. The spleen sections were incubated with rabbit anti-CP followed by Alexa Fluor w 546 labeled goat antirabbit IgG, with Alexa Fluor w 488 labeled rat antimouse CD206 (MR) and with rat antimouse Pan macrophage antibody followed by goat antirat Alexa Fluor w 488 antibody (A). The spleen sections were also incubated with rabbit anti-CP followed by Alexa Fluor w 546 labeled goat antirabbit IgG, with Alexa Fluor w 488 labeled rat antimouse CD206 and with rat antimouse dendritic cell antibody and secondary Alexa Fluor w 647 labeled goat antirat antibody (B). Blue reveals nuclear staining by 4 0 ,6-diamidino-2-phenylindol. Images were revealed by fluorescence microscopy.
Mannosylated antibody-enzyme clearance by mannose receptor sinusoidal endothelial cells and Kupffer cells as derived from the localization of internalized mannosylated ligands (Hubbard et al. 1979) and revealed by immunofluorescence with anti-MR and antimacrophage antibodies (Takahashi et al. 1998). However, our results showed that MFECP1 is predominately cleared by the endothelial cells and not the Kupffer cells. Although some reports have stated the opposite, our findings are consistent with most published work on clearance of mannosylated glycoproteins and neoglycoproteins. For example, in a pioneering study using iodinated mannosylated ligands, it was shown by electron microscope autoradiography that the radioactivity was predominantly associated with the endothelial cells and not the Kupffer cells (Hubbard et al. 1979). In an attempt to specifically target liver macrophages with mannose-terminal glucocerebrosidase in the lysosomal storage disorder, Gaucher's disease, it was concluded that the major site of delivery were the endothelial cells and not, as attempted, the macrophages (Bijsterbosch et al. 1996). It might be concluded that addition of appropriate mannosylation is a good strategy if delivery to sinusoidal endothelial cells is a desired outcome. For example, the prevention of an immune response where the indigested foreign peptides are presented on major histocompatibility complex (MHC) molecules on Kupffer cells and activated T cells. Mannosylating proteins might thus be a good strategy for preventing an immune response by re-directing clearance away from Kupffer cells. Evasion of immune responses is a desired outcome for MFECP1 in ADEPT, especially for multi-cycle treatment.
A further site of MFECP1 uptake in vivo was the spleen. Here, two APC, macrophages, and dendritic cells, might be expected to be involved. Dendritic cells have been shown in vitro to use MR for internalization of mannosylated proteins and uptake by this route may be problematic for protein therapeutics as this can lead to increased immunogenicity. For example, P. pastoris-expressed mannosylated ovalbumin (OVA) with either N-linked glycans, an attached S/T-rich O-linked region from crytococcal MP98, or both N-and O-linked glycans colocalizes with MR in dendritic cells and can induce OVA-specific CD4 þ T-cell proliferation (Lam et al. 2005). Furthermore, the approach has been successfully used to generate an antibody and T-cell response against MUC1 when the protein was linked to the MR ligand mannan and injected in patients with adenocarcinoma (Apostolopoulos and Mc Kenzie 2001). Tumor rejection, mediated by CD8þ T cells, was achieved with hydrophobized mannan nanoparticles that contained a fragment of the HER2 oncoprotein (Shiku et al. 2000).
Our in vivo results of spleen uptake showed that only macrophages express MR, and MFECP1 was only detected in macrophages. Providing more evidence, that uptake was MR mediated. The macrophage and dendritic cell distributions in the spleen were consistent with other reports where MRexpressing macrophages have been detected in the red pulp and not in the white pulp and marginal zone, whereas dendritic cells localize in the marginal zone and in the white pulp (Linehan et al. 1999). Because MFEC1 is not found in the dendritic cell residing marginal zone area, this might lead to a dampened immune response for this mannosylated antibody -enzyme fusion protein. This is consistent with its use in the clinic which showed no increase in immune response when compared to the nonmannosylated nonclinical grade MFE-CP .
In conclusion and an advantage to ADEPT, the current study shows that MFECP1 is rapidly cleared in liver and spleen by MR and in both organs it is the cell type with the less immunological response, predominantly sinusoidal endothelial cells in liver and macrophages in spleen, that is involved in the clearance. To our knowledge, MFECP1 is the first recombinant therapeutic with N-linked high-mannose structures to be tested in man. Elucidating its mechanism of clearance is of relevance for clinical development of this and other mannosylated compounds. The finding that clearance can be inhibited by mannan leads the way to optimize pharmacokinetics of mannosylated recombinant therapeutics.

Materials and methods
Protein production and analysis MFECP1 was produced in P. pastoris (strain X33) by fermentation interfaced with expanded bed adsorption immobilized metal affinity chromatography (EBA-IMAC) according to our standard protocols (Tolner et al. 2006a(Tolner et al. , 2006b. The 200-mM imidazole eluant from EBA-IMAC was concentrated 10-fold, dialyzed into PBS and purified by size exclusion chromatography using a Superdex 200 column (GE Healthcare, Amersham, UK). The final product was applied to an endotoxin removal gel (Pierce Biotechnology, Rockford, IL), filter sterilized (0.2 mm, Nalgene), dispensed glass vials (GE Healthcare) and stored at 280 8C.
N-glycan analysis of MFECP1 N-glycan analysis of MFECP1 was performed by DSA-FACE analysis according to the procedure described previously (Callewaert et al. 2001).

Enzyme activity
The enzyme activity was expressed in units (U), where one unit is the amount of enzyme required to hydrolyze 1 mmol of MTX per minute at 37 8C. For concentrations of H. Kogelberg et al. 2.5 U/mL or above, MFECP1 enzyme activity was quantified spectrophotometrically by measuring a change in absorbance at 320 nm due to CPG2 cleavage of methotrexate (MTX) (McCullough et al. 1971;Sharma et al. 2005). The rate of absorbance change was measured using a Hitachi U-2001 spectrophotometer with UV-Solutions software. Enzyme concentrations below 0.27 U/mL were measured using the sensitive high performance liquid chromatography (HPLC) assay using methotrexate as a substrate as described previously (Sharma et al. 2005).

Flow cytometric analysis of internalization of MFECP1
Aliquots of 1 Â 10 6 cells were preincubated at 4 8C in 35-mm 2 dishes with 10 mg MFECP1 in 500 mL DMEM medium containing 25 mM HEPES ( pH 7.3) for 1 h with or without 0.5 mg of mannan (Sigma-Aldrich, Dorset, UK). Cells were transferred to a 37 8C humidified CO 2 atmosphere for 5, 30, 45 60, 180, or 300 min. After one wash with 1 mL PBS, cells were detached with 0.3 mL trypsin/EDTA (Cambrex) and added to 3 mL DMEM medium. Aliquots of 5 Â 10 5 cells were subsequently washed with 2 mL PBS/ BSA and fixed with 100 mL fixation buffer (IntraStain fixation and permeabilisation kit, DakoCytomation, Dako, Glostrup, Denmark), washed with 2 mL PBS/BSA and treated with 100 ml permeabilization buffer in the presence of 1 mg mouse anti-CP, washed with 2 mL PBS/BSA and treated with 100 mL permeabilization buffer in the presence of 1 mg R-PE-conjugated goat antimouse IgG (BD Pharmingen, San Diego, CA) for flow cytometry analysis. In control experiments, the mouse anti-CP was omitted. Cells were maintained on ice prior to analysis by flow cytometry on an FACSCalibur TM cytometer with CelQuest software (Becton Dickonson, Cowley, UK) within no more than 12 h.
Flow cytometric analysis of surface binding of MFECP1 Cells were incubated with MFECP1 for 60 min and detached from the dish as described under Flow cytometric analysis of internalization of MFECP1. They were then incubated with 1 mg mouse anti-CP in 100 mL PBS at 4 8C for 60 min, washed with 2 mL PBS/BSA and treated with 1 mg R-PEconjugated goat antimouse IgG in 100 mL PBS/BSA at 4 8C for 60 min and fixed with 100 mL fixation buffer. Analysis by flow cytometry is as described under Flow cytometric analysis of internalization of MFECP1.
Immunofluorescence confocal microscopy of internalization of MFECP1 Internalization of MFECP1 was also analyzed by confocal laser scanning microscopy. Cells were incubated with MFECP1 as described under Flow cytometric analysis of internalization of MFECP1 for 45 min at 37 8C in a humified CO 2 atmosphere. This was followed by labeling of cells in PBS with 25 mM cell trace CFSE cell proliferation kit (Molecular Probes, Invitrogen) and incubation for 15 min at 37 8C, followed by incubation for 30 min at 37 8C in DMEM. Cells are detached with trypsin/EDTA as described under Flow cytometric analysis of internalization of MFECP1, fixed, permeablized in the presence of 1 mg mouse anti-CP and then permeabilized in the presence of 1 mg Alexa fluor594 goat antimouse IgG (Molecular Probes, Invitrogen). Control experiments omitted the anti-CP antibody. The nuclei were stained with Hoechst 33258. Cytospins were prepared and mounted to glass slides with ProLong Gold antifade reagent (Molecular Probe, Invitrogen) and visualized using an Olympus w confocal scanning microscope (Olympus, London, UK).

MFECP1 localization in mice
Two hundred micrograms (equivalent to 25 U of CP), each of MFECP1 was injected "intravenously" via the tail vein in BALB/c mice. Five minutes post injection, the liver and spleen tissues were removed, frozen, subsequently stained, and examined by immunofluorescence microscopy.
Immunofluorescence triple staining of MFECP1, MR, and dendritic cells in spleen. Frozen cryostat sections were cut at 10 mm and fixed in acetone for 10 min. Sections were washed in PBS, blocked in 3% normal goat serum for 20 min. Staining was carried out using rabbit anti-CP antibody at 1/500, and rat antimouse dendritic cell antibody (Hycult biotechnology, Uden, The Netherlands) at 20 mg/ mL, both for 1 h at RT. MFECP1 was detected using an Alexa fluor w 546 goat antirabbit secondary antibody (Invitrogen) at 1/200 and dendritic cells with Alexa fluor w 647 goat antirat antibody (Invitrogen, 1/200), for 1 h at RT. Sections were stained for MR using rat antimouse CD206 conjugated to Alexa fluor w 488 (Serotec, Oxford, UK) at 10 mg/mL for 1 h. Sections were counterstained for nuclei with 4 0 ,6-diamidino-2-phenylindole (DAPI) mountant (Vector Laboratories, Burlingame, CA).
Immunofluorescence triple staining of MFECP1, MR, and macrophages in the liver and spleen. The methods described under Immunofluorescence triple staining of MFECP1, MR, and dendritic cells in spleen were used for MFECP1 and MR staining. Macrophages were stained with rat antimouse pan macrophage antibody (recognizing F4/80 antigen) at 20 mg/mL (Hycult biotechnology) and secondary detection was with an Alexa fluor w 647 goat antirat antibody. Sections were counterstained for nuclei with DAPI as mentioned.
Double staining of MFECP1, CD31, and MFECP1, macrophages in the liver. The methods described under Immunofluorescence triple staining of MFECP1, MR, and macrophages in liver and spleen were used for MFECP1 and macrophage staining. The liver sections, in addition, were treated with rat antimouse CD31 at 1/2 (a gift from Professor A. Mantovani) followed by a goat antirat Alexa fluor w 488 secondary antibody. Sections were counterstained for nuclei with DAPI as under Immunofluorescence triple staining of MFECP1, MR, and dendritic cells in spleen.
Triple staining of MFECP1, CD31 and CD68 in liver. The methods described under Double staining of MFECP1, CD31, and MFECP1, macrophages in the liver were used for staining of MFECP1 and endothelial cells with anti-CD31. The liver section was also stained with rat antimouse CD68 (a late endosomal macrophage marker) conjugated to Alexa fluor w 647 (Invitrogen) diluted 10 mg/mL. MFECP1 blood clearance with and without addition of mannan in mice Six mice were injected i.v. into the tail vein with MFECP1 at a concentration of 200 mg in 0.3 mL (equivalent to 25 U of CP) per mouse. Six further mice were treated in the same manner but given 10 mg (0.3 mL, i.p.) each of mannan, 1 min prior to the addition of MFECP1. The mice were then divided into two cohorts of six mice. Each cohort contained three mice that had received MFECP1 alone and three mice which had received mannan prior to the injection of MFECP1. Cohort 1 was bled at 5 min, 30 min, 3 h, and 24 h, and cohort 2 was bled at 15 min, 1 h, 5 h, and 24 h. Twenty five microliters blood was collected at each instance. MFECP1 concentrations were measured as CP enzyme activity in plasma.