Selective antibody activation through protease-activated pro-antibodies that mask binding sites with inhibitory domains

Systemic injection of therapeutic antibodies may cause serious adverse effects due to on-target toxicity to the antigens expressed in normal tissues. To improve the targeting selectivity to the region of disease sites, we developed protease-activated pro-antibodies by masking the binding sites of antibodies with inhibitory domains that can be removed by proteases that are highly expressed at the disease sites. The latency-associated peptide (LAP), C2b or CBa of complement factor 2/B were linked, through a substrate peptide of matrix metalloproteinase-2 (MMP-2), to an anti-epidermal growth factor receptor (EGFR) antibody and an anti-tumor necrosis factor-α (TNF-α) antibody. Results showed that all the inhibitory domains could be removed by MMP-2 to restore the binding activities of the antibodies. LAP substantially reduced (53.8%) the binding activity of the anti-EGFR antibody on EGFR-expressing cells, whereas C2b and CBa were ineffective (21% and 9.3% reduction, respectively). Similarly, LAP also blocked 53.9% of the binding activity of the anti-TNF-α antibody. Finally, molecular dynamic simulation showed that the masking efficiency of LAP, C2b and CBa was 33.7%, 10.3% and −5.4%, respectively, over the binding sites of the antibodies. This strategy may aid in designing new protease-activated pro-antibodies that attain high therapeutic potency yet reduced systemic on-target toxicity.

blots, successively increased ( Fig. 2B-D). Cleaved antibodies migrated similarly as the native (unmodified) anti-EGFR antibodies in the reducing SDS-PAGE, indicating MMP-2 cleaved at the predicted position within the substrate peptide of the recombinant antibodies. These data suggested the inhibitory domains can be effectively removed by disease-associated proteases (MMP-2). As a control, unmodified anti-EGFR antibody was not digested by the MMP-2 in all the conditions tested ( Fig. 2A). Size-exclusion high-performance liquid chromatography analyses showed similar results between anti-EGFR antibodies and LAP-anti-EGFR antibodies, with a major peak detected in both HPLC and Coomassie staining, indicating monomeric conformation. A minor peak preceding the major peak may indicate slight aggregation in both of anti-EGFR and LAP-anti-EGFR antibodies in solution, but the aggregation of LAP-anti-EGFR antibodies were not more severe than anti-EGFR antibodies, indicating the LAP domain does not cause aggregation of pro-antibodies to a significant extent. (supplementary Figure 1). We further confirmed the clipped inhibitory domains dissociate from anti-EGFR antibodies by immunoprecipitation in supplementary Figure 2, which shows that LAP domain (with a molecular weight approximately 40 kDa) was not co-precipitated (thus not seen in the gel) with the antibodies treated by MMP-2. This data support the idea that clipped inhibitory domains would likely diffuse away from the anti-EGFR antibodies after digestion.

Recovery of binding activities of the pro-anti-EGFR Abs after MMP-2 digestion. We next tested
whether the binding activities of the recombinant antibodies were inhibited in the presence of the inhibitory domains and whether MMP-2 digestion was capable of restoring the activities of the antibodies. The culture supernatants from cells expressing CBa-anti-EGFR, C2b-anti-EGFR or LAP-anti-EGFR antibodies were digested by MMP-2 for different amounts of time and then tested by a cell-based ELISA where EGFR-expressing MDA-MB-468 cells were used as an antigen source. Prior to MMP-2 digestion, CBa, C2b and LAP domains reduced 9.3%, 21% and 53.8%, respectively, binding activities of the recombinant antibodies when compared to binding activity of the unmodified anti-EGFR antibodies. Removal of the inhibitory domains restored binding activities of the antibodies (Fig. 3). In general, longer incubation with MMP-2 better recovered antibody binding activity, in line with more abundant mature antibodies seen in western blot analyses in Fig. 2. CBa (Fig. 3A) and C2b (Fig. 3B) masked antibody binding sites ineffectively, enhancement of antibody binding after MMP-2 treatment was not as evident as LAP-anti-EGFR antibody, which rose progressively from 46.2% to 100% after treatment with MMP-2 for 1 hour (Fig. 3C). These data suggest that LAP is most effective in masking the binding sites of antibodies and that masking can be fully removed by MMP-2. To confirm the masking effect of pro-antibodies is dependent on inhibitory domains, rather than the MMP-2 substrate peptide (GPLGVR), we produced and purified recombinant MMP-2 substrate peptide-linked anti-EGFR antibodies (GPLGVR-anti-EGFR Abs). The binding activity of GPLGVR-anti-EGFR antibodies was indistinguishable with or without MMP-2 treatments  3D). We also confirmed that the GPLGVR peptide was successfully removed from the heavy chains of anti-EGFR antibodies by MMP-2 (Fig. 3E). Thus, masking of the binding activity of pro-antibodies is dependent on the inhibitory domain (LAP), not the MMP-2 substrate peptide (GPLGVR).
The EC 50 and binding specificities of LAP-anti-EGFR antibodies before and after MMP-2 digestion. We next tested how binding kinetics of anti-EGFR antibodies were altered by LAP domains. Purified anti-EGFR antibodies and LAP-anti-EGFR antibodies were incubated with MMP-2 at 37 °C for 1 h and then tested by cell-based ELISA with the indicated dilution of the antibodies. Figure 4A shows that before MMP-2 treatment, the binding kinetic curve of LAP-anti-EGFR antibodies shifted to the right, whereas after incubation with MMP-2, the binding kinetic curve of LAP-anti-EGFR antibodies shifted to the left, to the levels similar to the unmodified anti-EGFR antibodies. The EC 50 values of anti-EGFR antibody, LAP-anti-EGFR antibody and MMP-2-activated LAP-anti-EGFR antibody in EGFR-overexpressing MDA-MB-468 cells were 2.4 nM, 7.5 nM and 2.7 nM, respectively. In addition, we compared the binding activity of anti-EGFR and LAP-anti-EGFR antibodies using HaCaT keratinocyte cells, which express a moderate level of EGFR on cell surface (supplementary Figure 3). The binding kinetics of anti-EGFR antibodies were decreased by LAP domains in moderate EGFR-expressing HaCaT cells (EC 50 7.5-fold higher, Fig. 4B). The difference of EC 50 in HaCaT cells was higher than the data obtained in the high EGFR-expressing MDA-MB-468 cells. This result is consistent with Carrido G. et al. who also showed that binding activity of an intermediate affinity anti-EGFR antibody (nimotuzumab) decreased when EGFR expression is decreased 26 . In conclusion, our results show that the LAP reduces the binding activity of the antibody in both high and moderate EGFR-expressing cells.
We further confirmed the specificities of anti-EGFR antibody and LAP-anti-EGFR antibody for binding to EGFR overexpressed on MDA-MB-468 cells. We used a HA-tagged anti-EGFR antibody as the readout for cell-based competition ELISA. Graded concentrations of anti-EGFR antibodies or LAP-anti-EGFR antibodies (with or without MMP-2 digestion) were mixed with a fixed amount of HA-tagged anti-EGFR antibody prior to addition to EGFR-overexpressing MDA-MB-468 cells, then we detected bound HA-tagged anti-EGFR antibody by a HA tag-specific antibody. As shown in Fig. 4C, increasing amounts of anti-EGFR antibodies indeed competed with the HA-tagged anti-EGFR antibody for binding sites on MDA-MB-468 cells as demonstrated by the proportionally decreased absorbance at 405 nm. However, untreated LAP-anti-EGFR antibodies did not significantly decrease the signal, indicating that LAP-anti-EGFR antibodies could not efficiently compete with HA-tagged anti-EGFR antibody. On the contrary, MMP-2-pretreated LAP-anti-EGFR antibodies proportionally decreased the absorbance, similar to the extent of the anti-EGFR antibodies. These results suggest that MMP-2-treated LAP-anti-EGFR antibodies can specific bind to the EGFR antigen expressed on the cells.
The serum stability of LAP-anti-EGFR antibodies. We examined the serum stability of LAP-anti-EGFR antibodies in vitro. The anti-EGFR antibodies or LAP-anti-EGFR antibodies were incubated with 20% fetal bovine serum (FBS) at 4 °C for indicated time, then analyzed protein stability by western blotting, and binding activity by cell-based ELISA. Figure 5A shows that both of anti-EGFR antibodies and LAP-anti-EGFR antibodies are stable in serum. In addition, the binding activity were not affected after serum incubation throughout the indicated time (Fig. 5B). These results indicate that the LAP-anti-EGFR antibodies are proteolytically stable in serum.

Applicability of LAP inhibitory domain in an anti-TNF-α antibody.
The strategy of masking antibody binding sites seemed to be most successful when using LAP. We therefore tested whether LAP can be used as a universal blocker for antibodies. To this end, we linked LAP to an anti-TNF-α antibody through the same MMP-2 substrate peptide. Similar to LAP-anti-EGFR antibody, the LAP domain can be effectively removed from the heavy chain of anti-TNF-α antibody by MMP-2 treatment in an incubation time-dependent fashion (Fig. 6A). The LAP reduced antibody binding to TNF-α by 53.9%, as compared to the unmodified anti-TNF-α antibody (Fig. 6B). Addition of MMP-2 to the recombinant antibody restored its binding activity to 100%, indicating MMP-2 can effectively remove the LAP inhibitory domain to revive the antibody function. These data indicate that linking LAP to a therapeutic antibody through a protease substrate peptide may be a useful strategy to render antibody function latent until it is activated by disease-associated proteases.

Illustration of the masking efficiency of the inhibitory domains by computer molecular dynamic simulation.
We used computer molecular dynamic simulation to explore the possible mechanisms of the masking efficiency of inhibitory domains over the binding sites of antibodies (complementarity-determining regions, CDRs) (Fig. 3). We set the anti-EGFR Fab protein structure as main motif, then linked different inhibitory domains to the N-terminus of the heavy chain of an anti-EGFR Fab through a substrate peptide of MMP-2, then simulated the pro-anti-EGFR Fab by computer dynamic simulation. To explain the masking efficiency of inhibitory domains, we analyzed the spaces between CDRs and inhibitory domains. The program HOLLOW 27 was used to calculate the channel spaces and interior spaces of protein structures by setting a sphere inside the structure then filled the sphere space with dummy atoms, and then to calculate the total numbers of atoms to obtain the sphere volume. We set a hollow sphere between the binding sites of the antibodies (CDRs) and the inhibitory domains (LAP/C2b/CBa), and then analyzed the masking efficiency of the inhibitory domains by calculating the volume of the hollow sphere (filled with oxygen atoms). A smaller sphere volume would indicate a greater masking efficiency of the inhibitory domain. Figure 7 shows the simulation of protein structures of anti-EGFR Fab (Fig. 7A), LAP-Fab (Fig. 7B), C2b-Fab (Fig. 7C) and CBa-Fab ( Fig. 7D) with hollow sphere. The sphere volume of the LAP-Fab was smallest (filled with 3980 oxygen atoms), followed by the C2b-Fab (5382 oxygen atoms) and finally the CBa-Fab (6322 oxygen atoms). According to these calculations (Table 1), the masking efficiency of the LAP domain, C2b domain and CBa domain were 33.7%, 10.3% and −5.4%, respectively. These results suggested that the LAP domain covered the CDR most extensively among the three different inhibitory domains, which is in line with our data showing reduced binding activities of pro-anti-EGFR antibodies imposed  by the inhibitory domains (Fig. 3). Thus the computer simulation can be used to explain the masking efficiency of inhibitory domains over the binding sites of antibodies.

Discussion
Systemic injection of antibodies may cause considerable adverse effects. We used an inhibitory domain to mask the binding sites of antibodies to develop MMP-2-activated pro-antibodies as a strategy to reduce systemic targeting and adverse effects of the antibodies. This strategy takes advantage of the fact that many proteases such as MMP-2 are selectively upregulated in disease sites 28 . Using this strategy, inactive pro-antibodies may be selectively converted into active antibodies at the disease sites by MMP-2, while leaving the target molecules in normal tissues untouched. We demonstrated that N-terminally linked LAP can indeed mask the binding sites of the recombinant LAP-anti-EGFR antibody and LAP-anti-TNF-α antibody. Cleavage of LAP domain from the recombinant antibodies revived the binding activities comparable to that of their unmodified forms. Our finding that LAP can mask the binding sites of two antibodies suggests that LAP may be applicable to other antibodies to generate disease site-selective pro-antibodies. Thus LAP-masked pro-antibody is a promising strategy for the development of personalized protease targeted therapies.
With regard to the relationship between the reduction of binding activity and the reduction of the on-target toxicity of antibody, Desnoyers and coworkers showed that a peptide-masked EGFR probody displaying a higher EC 50 (than the parental antibody; 48-fold difference) could improve the safety of therapeutic antibodies in nonhuman primates 29 ; Yang and coworkers also demonstrated that a probody with 20-fold weaker binding activity (than that of parental antibody) displayed a lower targeting in adjacent normal tissues 30 ; Donaldson and coworkers also proved a peptide-masked scFv with lower binding affinity (eight-fold lower than unmasked scFv) displayed a lower targeting to keratinocyte-derived HaCaT cells 31 . Finally, Crombet and coworkers pointed out that lowering the affinity between a monoclonal anti-EGFR antibody (nimotuzumab) and EGFR reduced on-target toxicity to normal cells. Their mathematical models predict that the binding affinity (K D ) for anti-EGFR antibodies should be in the range of 10 −8 M-10 −9 M to maximize tumor cell targeting while minimizing normal cell toxicity 32 . It is therefore concluded that lowering binding activity in terms of EC 50 or K D may reduce on-target toxicity to normal    tissues. In our study, the EC 50 of LAP-anti-EGFR antibody is 3.1-fold higher than the parental anti-EGFR antibody (Fig. 4A), thus we expect LAP may reduce the on-target toxicity of antibodies in normal tissue. Several strategies have been attempted to design pro-antibodies; however, technical challenges remain. For example phage display technology was used to select peptide capable of masking the binding sites of anti-EGFR antibodies 29,33,34 . These phage display-selected peptides are likely non-endogenous and may provoke neutralizing antibodies once injected to humans, thus limiting their long term use in humans. In addition to phage display technology, other investigators used epitope sequences recognized by the antibody (such as anti-EGFR antibody) as masking peptides 31 . However, the epitope peptides bind to the antibody with high affinity (i.e., the nature of antibody-antigen binding), such that it may remain bound after proteases have cleaved the linker between the antibody and the peptides. Furthermore, the affinity-based approach requires search and selection of a unique phage display-selected peptide/epitope peptide for each antibody, making generalized application of a selected peptide to other antibodies difficult.
To address these technical challenges, we used the concept of steric hindrance to select inhibitory domains from endogenous proteins. Thus LAP-masked pro-antibodies are expected to have low immunogenicity and are safer for use in humans. On the other hand, the masking peptide LAP can dissociate from the antibodies once the intervening linker is cleaved by the protease. This notion is supported by our data that binding activity of our pro-antibodies can be revived upon treatment with MMP-2. Considering that all antibodies are composed of light chains and heavy chains and that different antibodies share a very similar three dimensional structure, grafting the LAP domain to other antibodies may impose similar steric hindrance to their binding sites. Thus the LAP domain may represent an efficient and rapid way to generate protease-activated pro-antibodies.
Proteases play an important role in the progression of many diseases 35 . Proteases cleave many different targets (extracellular matrix, cytokines, growth factors and cytokine/growth factor receptors that modulate pathophysiology in diseases 28,36,37 . High expression of proteases is a characteristic of many diseases. Given their differential expression patterns at disease sites and in normal tissues 38 , proteases are attractive molecular targets for the design of new therapies or diagnoses that act preferentially at disease sites while leaving normal tissues untouched. Indeed, several reports have engineered molecular probes 39,40 or protease-activated prodrugs suited for this aim 29,41 . For example Desnoyers et. al. have developed an EGFR-specific probody that is activated by the urokinase-type plasminogen activator (uPA) and membrane-type serine protease 1 (MT-SP1) 29 . We choose to target MMP-2 in our proof-of-concept test of protease-activated pro-antibodies. This concept, however, is not limited to MMP-2 but may be further expanded for other protease-related diseases by replacing the substrate linkers between the inhibitory domains and targeting antibodies. For example, cysteine protease cathepsins 36 and legumain 37 are well-known proteases that are overexpressed in some tumors or arthritis sites. Specific linker substrates can be tailor-made to suit patients' protease profiling at the disease sites to achieve personalized medicine.
Our computer molecular dynamic simulation shows that the LAP inhibitory domain masked the binding sites of antibodies most extensively. This result is consistent with the inhibition of binding activity before MMP-2 activation (Fig. 3). However, many of the selected inhibitory domains failed in our test possibly due to distorted conformations when linked to the antibodies. We consider that the linker between inhibitory domains and antibodies was too flexible to lose direction, so it may cause ineffective masking of the antibodies. Further study to modify the linker sequences between the inhibitory domain and antibody binding sites via simulation of molecular dynamics may reveal the optimized linker for each antibody. Though computer-simulation may not reveal real protein structures, it can provide insights to rational design and improve developmental efficiency of protease-activated pro-antibodies. In the future, crystallography may be required to solve the definitive structures of pro-antibodies to fully explain the mechanism on how LAP masks the binding activities of the antibodies.
In summary, LAP domain-masked, MMP-2-activated, pro-antibodies possess the following advantages: (1) pro-antibodies can increase disease site selectivity, thus reducing systemic adverse effects, by targeted activation of the pro-antibodies by proteases over-expressed at disease sites; (2) pro-antibodies have low immunogenicity due to the endogenous origin of the inhibitory domain; (3) steric hindrance of the inhibitory domain can be easily removed from the pro-antibodies once the linker is cleaved by the protease; (4) the LAP inhibitory domain can be adopted for other antibodies; and (5) the linker substrate can be changed to suit other disease-associated proteases. According to these advantages, the LAP domain may be useful for the development of novel pro-antibodies. This strategy may improve the safety of the antibodies and likely increase patient compliance to achieve a better therapeutic index. Plasmid construction of pro-antibody. A schematic representation of the pro-antibody expression constructs is shown in Fig. 1B. To construct the expression vectors encoding the full-length anti-EGFR antibody and anti-TNF-α antibody, a variable fragment of human anti-EGFR antibody was cloned based on the h528Fv DNA sequence 42 and a variable fragment of human anti-TNF-α antibody was cloned based on the Remicade sequence (from US Patent 7,070,775) by assembly PCR. The VH-CH1 domain of the antibody, assembled with the human IgG1 Fc domain via assembly PCR, was cloned into the pLHCX mammalian expression vector to acquire the human Igκ leader sequence (LS) at the N terminus of the antibody heavy chain expression sequence. The VL-Cκ Scientific RepoRts | 7: 11587 | DOI:10.1038/s41598-017-11886-7 domain of antibody with a leader sequence was inserted at the N terminus and a furin-2A (F2A)-based bicistronic sequence 43 was cloned into the pLHCX-LS-anti-EGFR/TNF-α heavy chain expression vector to form an anti-EGFR/TNF-α antibody expression cassette. The coding sequences of complement factor 2, complement factor B and latency-associated peptide genes were purchased from TransOMIC Technologies (Huntsville, AL, USA). The LAP and C2b of C2 and CBa of CB fragments-containing MMP-2 substrate peptide 44 were cloned and amplified by PCR, and then inserted into the antibody expression cassette, between the F2A and the heavy chain of the antibody. The MMP-2 substrate peptide (GPLGVR)-linked anti-EGFR antibodies heavy chain were cloned and amplified by PCR, and then replaced the heavy chain to GPLGVR-heavy chain in the antibody expression cassette.

Materials and Methods
Transient expression and purification of pro-antibody. 293 T cells (1.5 × 10 6 cells/well) were seeded in 6-well plate contained in DMEM supplemented with 10% cosmic calf serum and placed in an incubator at 37 °C overnight. The cells were transfected with 2.5 μg pLHCX-pro-anti-EGFR or pLHCX-anti-EGFR plasmids by TransIT-LT1 transfection reagent (Mirus Bio Corporation, Madison, WI, USA) and cultured in DMEM serum-free medium for 48 hours. The cells and media were collected for subsequent experiments. Alternatively, 6 × 10 7 Expi293 cells in suspension were transfected with 30 μg pLHCX-LAP-anti-TNF-α or pLHCX-anti-TNF-α plasmids by the Expifectamine reagent (Life Technologies, Carlsbad, CA, USA). Cells were cultured in Expi293 media with enhancers as specified by the manufacturer. The media were harvested for subsequent experiments 5 days post-transfection. The anti-EGFR antibodies and LAP-anti-EGFR antibodies were purified by protein A sepharose 4 Fast Flow chromatography (GE Healthcare, Little Chalfont, UK). The purity of purified anti-EGFR antibodies and LAP-anti-EGFR antibodies were analyzed by SDS-PAGE in reducing and nonreducing conditions (supplementary Figure 4).

Analyses of MMP-2 digestion of pro-Abs by western blotting. The pro-antibodies and antibodies
with or without MMP-2 treatment were mixed with 6× SDS reducing loading dye and boiled for 10 min. Samples were separated by 10% SDS-PAGE (sodium dodecyl sulfate-polyacrylamide gel electrophoresis) and then transferred to nitrocellulose (NC) membranes (Millipore, Billerica, MA, USA). After blocking with PBS containing 5% milk at 4 °C overnight, the membranes were incubated with HRP-conjugated goat anti-human IgG Fcγ antibodies (Jackson ImmunoResearch Laboratories, West Grove, PA, USA) at room temperature for 1 hour. After washing in PBST (PBS containing 0.05% Tween 20) three times and in PBS once, the blots were developed by ECL reagent (Millipore, Billerica, MA, USA) and chemiluminescence was detected on autoradiography films.

Analyses the protein aggregation of anti-EGFR antibodies and LAP-anti-EGFR antibodies by size-exclusion high-performance liquid chromatography. Anti-EGFR antibodies and
LAP-anti-EGFR antibodies (100 μg/sample) were injected into an Aligent Bio SEC-5 column (300 × 7.8 mm, 300 Å) and separated at 1 ml/min in 50 mM sodium phosphate buffer, pH7.0. The protein peaks were detected at 280 nm.

Immunoprecipitation of the MMP-2-treated LAP-anti-EGFR antibodies. Purified
GPLGVR-anti-EGFR antibodies (MMP-2 substrate peptide-linked anti-EGFR Abs) and LAP-anti-EGFR antibodies were incubated with 5 μg/ml MMP-2 or buffer only (DMEM) at 37 °C for 1 h. The MMP-2 reaction were stopped with 15 mM EDTA and then immunoprecipitated by protein A sepharose beads with rotation at room temperature for 30 minutes. The beads were pelleted from the mixture by centrifugation and washed six times by PBS. The beads were mixed with 6× sodium dodecyl sulfate (SDS) sample loading buffer for SDS-PAGE and then the gel stained with Coomassie brilliant blue staining solution.

Analyses of the binding activities of the MMP-2-treated pro-anti-EGFR Abs by cell-based ELISA. EGFR-overexpressing MDA-MB-468 cells and moderate EGFR-expressing HaCaT cells (10 5 cells/well)
were grown overnight in 96-well microtiter plates pre-coated with 40 μg/mL of poly-L-lysine (50 μL/well). After removing the original culture medium from the 96-well plates, the cells were washed with DMEM (200 μL/well) three times, the indicated dilution of anti-EGFR antibodies or MMP-2 pretreated or untreated pro-anti-EGFR antibodies were added to the cells (50 μL/well) at room temperature for 2 hours. After the cells were washed with DMEM (200 μL/well) three times to remove unbound antibodies, 1 μg/ml of HRP-conjugated goat anti-human IgG Fcγ antibodies (Jackson ImmunoResearch Laboratories) were added to the cells (50 μL/well) at room temperature for 1 hour. Finally, after washing with DMEM three times and PBS one time, ABTS substrates were added for 30 min before optical absorbance values at 405 nm were measured using an ELISA reader (Molecular Devices, Menlo Park, CA, USA). The EC 50 for antibodies were analyzed by nonlinear regression curve fitting using the computer program GraphPad Prism 6.0 (GraphPad Software, San Diego, CA, USA). Data are presented as the mean of three independent experiments.
Scientific RepoRts | 7: 11587 | DOI:10.1038/s41598-017-11886-7 Examination of the expression level of EGFR on the cell surface by flow cytometry. EGFR expression were measured by staining the cells with 1 μg/mL Erbitux in PBS containing 0.05% bovine serum albumin (BSA) for 50 minutes at 4 °C, followed by 1 μg/mL FITC-conjugated goat anti-human IgG Fc (Jackson ImmunoResearch Laboratories, West Grove, PA) for 50 minutes at 4 °C. After extensive wash in ice-cold PBS containing 0.05% BSA, the surface immunofluorescence of viable cells was measured with a FACScan flow cytometer (Beckman Coulter FC500 Cytometer) and fluorescence intensities were analyzed with CXP software (Beckman Coulter).

Examination of the specificity of anti-EGFR antibodies and LAP-anti-EGFR antibodies by
competition cell-based ELISA. EGFR overexpressing MDA-MB-468 cells (10 5 cells/well) were seeded in 96-well microtiter plates as above. Serially diluted anti-EGFR antibodies, untreated or MMP-2 pretreated LAP-anti-EGFR antibodies were prepared and mixed 1:1 (v/v) with 50 nM of HA-tagged anti-EGFR antibody, and then the mixture were added to the cells at room temperature for 1 h. After proper washing in PBS, cells were sequentially incubated with HRP-conjugated anti-HA antibodies and ABTS at room temperature. The optical absorbance were measured at 405 nm using an ELISA reader.

Examination of the serum stability of LAP-anti-EGFR antibodies in vitro.
The anti-EGFR antibodies or LAP-anti-EGFR antibodies were incubated with 20% fetal bovine serum (FBS, Biological Industries) at 4 °C for 0 h, 12 h, 24 h or 48 h, then a proportion of the incubations were mixed with 6× sodium dodecyl sulfate (SDS) reducing loading dye for gel electrophoresis and western blotting analyses. The remaining incubations were serially diluted and tested for binding to EGFR by cell-based ELISA as described above.

Analysis of the binding activities of the MMP-2-treated LAP-anti-TNF-α antibodies by
ELISA. ELISA microplates (96-well, Maxisorp, Nunc, Roskilde, Denmark) were coated with 1 μg/mL of TNF-α (R&D Systems, Minneapolis, MN, USA) in 0.1 M NaHCO 3 (pH8.3) coating solution (50 μL/well) for 2 hours at 37 °C and then blocked with PBS containing 5% skim milk (200 μL/well) for 2 hours at 37 °C. The anti-TNF-α antibody or MMP-2 pretreated or untreated pro-anti-TNF-α antibodies were added to the wells (50 μL/well) at room temperature for 2 hours. After washing with PBST (PBS containing 0.05% Tween 20) three times and PBS once to remove unbound antibodies, 1 μg/ml of HRP-conjugated goat anti-human IgG Fcγ antibodies were added to the wells (50 μL/well) at room temperature for 1 hour. After extensive washing as above, ABTS substrate was added for 30 min before optical absorbance at 405 nm was measured using an ELISA reader.

Computer molecular dynamic simulation of pro-antibodies. The anti-EGFR Fab were built by
Homology modeling 46 , using the antibody against foot-and-mouth disease virus (Protein Data Bank (PDB) ID entry 1QGC) as a template. The LAP, C2b and CBa domain were extracted from pro-TGF-β1 (PDB ID 3RJR), complement factor 2 (PDB ID 3ERB) and complement factor B (PDB ID 2OK5) by Swiss-PDB viewer 47,48 , respectively. Then, the MMP-2 substrate linker (GGGGS-GPLGVR) were inserted into the C terminus of inhibitory domains by edit protein sequence using Discovery Studio 3.0 software. After insertion of the MMP-2 substrate linker into the C terminus of inhibitory domains, the inhibitory domain-linker was ligated to the N terminus of the heavy chain of anti-EGFR Fab by Swiss-PDB viewer. The structure of the pro-anti-EGFR Fab was built by homology modeling and dynamically simulated with the program Nanoscale Molecular Dynamics (NAMD) 49 , then illustrated by Pymol. The anti-EGFR Fab complementarity-determining regions (CDRs) of the light chain and heavy chain were defined by using the Kabat amino acid numbering scheme 50 .
Calculation of masking efficiency of the inhibitory domain of the pro-antibody. The masking efficiency of the inhibitory domains over the binding sites of antibodies were calculated with the program HOLLOW 27 . In this program, a hollow sphere above the binding sites of antibodies (CDRs) was set as a tool to calculate the space between CDRs and the inhibitory domain. The highest amino acid (57S, located at the heavy chain CDR2) of anti-EGFR Fab was selected as the center of the hollow sphere, and then a sphere was drawn with a 12 Å radius, which is the cutoff distance of non-bonded interactions (e.g. electrostatic interactions or wan der Waals forces); distance greater than this range the non-bonded interactions are assumed to zero 51 . A smaller sphere volume indicates a smaller space between CDRs and the inhibitory domain, showing the masking efficiency of the inhibitory domain was effective. To calculate the sphere volume of pro-anti-EGFR Fab (LAP-Fab/ C2b-Fab/CBa-Fab), the space of the sphere was filled with oxygen atoms. The number of oxygen atoms (N) was multiplied by the density of the atoms (V) to derive sphere volume (S) by the following formula: The density of oxygen atoms was calculated by volume of sphere formula (V = πr 3 × 4/3). The radius of oxygen atom is 1.52 Å. The sphere volume of anti-EGFR Fab (S Fab ) was set as 100%. Results were expressed as a percentage of sphere volume of pro-Fab (LAP-Fab/C2b-Fab/CBa-Fab) (S pro-Fab ) compared to the S Fab by the following formula: The percentage of sphere volume (% S) = 100 × (S pro-Fab /S Fab ). The masking efficiency was calculated by the following formula: The percentage of masking efficiency (% M) = % S Fab − % S pro-Fab .