Development of a new high‐affinity human antibody with antitumor activity against solid and blood malignancies

mAbs have emerged as a promising strategy for the treatment of cancer. However, in several malignancies, no effective antitumor mAbs are yet available. Identifying therapeutic mAbs that recognize common tumor antigens could render the treatment widely applicable. Here, a human single‐chain variable fragment (scFv) antibody library was sequentially affinity selected against a panel of human cancer cell lines and an antibody fragment (named MS5) that bound to solid and blood cancer cells was identified. The MS5 scFv was fused to the human IgG1 Fc domain to generate an antibody (MS5‐Fc fusion) that induced antibody‐dependent cellular cytotoxicity and phagocytosis of cancer cells by macrophages. In addition, the MS5‐Fc antibody bound to primary leukemia cells and induced antibody‐dependent cellular cytotoxicity. In the majority of analyzed cancer cells, the MS5‐Fc antibody induced cell surface redistribution of the receptor complexes, but not internalization, thus maximizing the accessibility of the IgG1 Fc domain to immune effector cells. In vitro stability studies showed that the MS5‐Fc antibody was stable after 6 d of incubation in human serum, retaining ~60% of its initial intact form. After intravenous injections, the antibody localized into tumor tissues and inhibited the growth of 3 different human tumor xenografts (breast, lymphoma, and leukemia). These antitumor effects were associated with tumor infiltration by macrophages and NK cells. In the Ramos B‐cell lymphoma xenograft model, the MS5‐Fc antibody exhibited a comparable antitumor effect as rituximab, a chimeric anti‐CD20 IgG1 mAb. These results indicate that human antibodies with pan‐cancer abilities can be generated from phage display libraries, and that the engineered MS5‐Fc antibody could be an attractive agent for further clinical investigation.—Sioud, M., Westby, P., Vasovic, V., Fløisand, Y., Peng, Q. Development of a new high‐affinity human antibody with antitumor activity against solid and blood malignancies. FASEB J. 32, 5063–5077 (2018). www.fasebj.org

In addition to surgery, chemotherapy, and radiotherapy, immunotherapy represents a therapeutic avenue that has recently shown great promise in cancer (1). mAbs represent an example of successful targeted immunotherapy and are one of the fastest growing areas in oncology (2). Most commonly used mAbs in cancer immunotherapy are of the IgG class due to their long half-life and stability in serum. These therapeutic mAbs kill cancer cells via several mechanisms, including antibody-dependent cellular cytotoxicity (ADCC), antibody-dependent cell-mediated phagocytosis (ADCP), and complement-dependent cytotoxicity. The best examples of antibody therapy come from the treatment of non-Hodgkin lymphoma (NHL) and HER2-positive breast cancer by rituximab (anti-CD20) or trastuzumab (anti-HER2), respectively (2,3). However, the emergence of unresponsive tumors due to low or absent targeted cell surface receptors highlights the need to develop additional targeted options (3,4). Moreover, in many cancers, no mAbs that target and kill tumor cells are yet available.
In addition to mAbs that specifically target cancer cells, a new type of mAbs called checkpoint inhibitors [e.g., ipilimumab (anti-cytotoxic T-lymphocyte-associated antigen 4)] and nivolumab (anti-programmed cell death-1) showed remarkable efficacy in skin and lung cancers by "releasing the brakes" on T cells (5). Unfortunately, these mAbs work best in tumors with high DNA mutations (neoantigens), such as melanoma, but are less successful in other tumors, such as prostate, breast, and blood cancers (6)(7)(8)(9). Even in melanoma, the therapeutic benefit of ipilimumab, for example, has been limited to a fraction of patients (5). Recent data suggest the involvement of preexisting microbial T cells in antitumor immunity of checkpoint inhibitors (10). Whatever the origin of T cells, however, most tumors escape traditional T-cell killing through alteration of their antigen-processing machinery and/or down-regulation of major histocompatibility complex class I expression, rendering the neoantigens undetectable by patient T cells (11). Thus, novel treatment options are still needed that recruit and activate more immune effector cells (e.g., NK cells, macrophages, neutrophils) to kill tumor cells independent of their mutational load and major histocompatibility complex expression status. In addition, if antibodies could be engineered to recognize a common target for most cancer patients, the same antibody can be used for these patients, making the treatment widely applicable and hopefully cheaper.
Phage display, which involves the cloning of the variable gene segments from heavy and light chains of IgGs in Escherichia coli and their display on filamentous bacteriophages, has been used for the selection and engineering of therapeutic mAbs with distinct functional properties (12). The goal of the present study was to investigate the possibility of selecting universal antitumor single-chain variable fragment (scFv) antibodies from a human scFv antibody library. By removing phage antibodies binding to common cell surface receptors expressed by normal cells and alternating cancer cells during the biopanning experiments, we have isolated cancer cell-binding scFv antibody fragments. Among the selected candidates, one variant (named MS5) recognized several types of cancer cells but did not, or weakly, recognized normal cells such as peripheral blood mononuclear cells (PBMCs). The MS5 scFv was fused to the human IgG1 Fc domain (hinge, CH2-CH3) resulting in the generation of an MS5-Fc antibody that efficiently induced ADCC and inhibited tumor growth in 3 different tumor xenografts, supporting its potential clinical use in cancer immunotherapy.

Cancer cell lines and blood cells
Cancer cell lines used in this study ( Table 1) were obtained from the American Type Culture Collection (Rockville, MD, USA) and cultured under recommended conditions. Human PBMCs were isolated from buffy coats by Lymphoprep density gradient centrifugation. Monocytes were enriched from PBMCs by using plastic adherence, and purification was verified by phenotypic analysis of the surface marker CD14 + . To generate M2 macrophages, monocytes were cultured for 6 d in RPMI-1640 medium supplemented with 20 ng/ml of M-CSF. For full M2 polarization, IL-4 (20 ng/ml) was added during the last 48 h of culture (13). NK cells were purified from PBMCs by using an NK cell isolation kit and auto MAC Pro Separator according to the manufacturer's instructions (Miltenyi Biotec, Bergisch Gladbach, Germany). Purification was verified by phenotypic analysis of the surface marker CD56. T cells and B cells were purified from PBMCs by using either the Dynal CD4 or CD19 Positive Isolation Kit, respectively, as described in the manufacturer's instructions (Invitrogen Dynal As, Oslo, Norway). PBMCs from patients with leukemia were isolated as noted earlier. Cells were cultured in RPMI-1640 or DMEM supplemented with 10% heat-inactivated fetal calf serum (FCS) and antibiotics. Cell stimulation with the MS5-Fc antibody or Fc control was performed in X-VIVO 15 medium (Lonza, Basel, Switzerland). CD34 + cells were isolated from bone marrow (BM) mononuclear cell preparations as previously described (14). Approval for obtaining blood samples from patients with leukemia was granted by the Regional Committees for Medical and Health Research Ethics (REK 2017/1596).

Selection of scFv antibodies
The synthetic human scFv library (Griffin-1 library) used in this study was a generous gift from Dr. Greg Winter (University of Cambridge, Cambridge, United Kingdom). The library was constructed by recloning synthetic heavy-and light-chain variable regions from the lox library vector (15) into the phagemid vector pHEN2. Before affinity selection, the library (1 3 10 10 phage particles) was preabsorbed on normal human mammary epithelial cells (HMECs) (2 3 10 7 ) and blood leukocytes (2 3 10 7 cells) to remove most phages that bind to common receptors expressed by normal cells. The preabsorbed library was then incubated with the breast cancer cell line MDA-MB-453 (1 3 10 7 cells) in PBS supplemented with 5% FCS for 1 h at room temperature with gentle mixing. Nonbound phages were removed by washing 10 times with 5 ml PBS (pH 7.5) containing 1% FCS and twice with 5 ml PBS (pH 6.5). Cell-bound phages were eluted by adding 500 ml 0.1 M Tris glycine pH 2.2 and then neutralized with 40 ml of 2 M Tris-base and then titrated. To amplify the recovered phages, 40 ml of exponentially growing E. coli TG1 cells were infected with 250 ml of the eluted phages, plated on 2xTY plates containing 100 mg/ml ampicillin and 1% glucose, and incubated overnight at 30°C. Bacterial cells were scraped from the plates, and phages were amplified, PEG purified, and titrated. For the second round of affinity selection, MDA-MB-453 amplified phages (1 3 10 10 phages) were preabsorbed on normal cells and then affinity selected on prostate cancer cell line PC3 (1 3 10 7 cells) as described earlier.

Phage amplification
In brief, 40 ml of exponentially growing E. coli strain TG1 cells were infected with the eluted phages and grown in 2xTY containing 100 mg/ml ampicillin and incubated for ;2 h at 37°C with shaking. The cells were infected with helper phage M13KO7 in a ratio of at least 20:1 (phage:bacteria) for 30 min at 37°C. The infected bacteria were centrifuged and resuspended in 100 ml 2xTY containing 100 mg/ml ampicillin and 25 mg/ml kanamycin and incubated in a shaker at 30°C overnight. After the culture was centrifuged at 9000 g for 10 min, the phage particles were precipitated by adding 1/5 volume of PEG 6000/NaCl and then incubated for 2 h before centrifugation at 11,000 g for 30 min. Phage pellets were resuspended in PBS buffer and then titrated as indicated in the following section. Phages were also prepared from individual ampicillin-resistant colonies, PEG precipitated, and titrated; their binding to cancer cells was investigated by using flow cytometry.

Phage titration
Serial dilutions of eluted or amplified phages (10 ml/sample) were added to exponentially growing TG1 cells along with 3 ml top agar and then plated on agar plates. After incubation at room temperature for 1 h, plates were incubated overnight at 37°C, and the number of plaques was determined for each dilution. control Fc were affinity purified on a protein G column. Positive fractions and protein purity were determined by using 10% SDS-PAGE analysis with Imperial Protein Stain (Thermo Fisher Scientific) to visualize proteins. Positive fractions were collected, pH adjusted to 7.5, and then stored at 280°C until use. For conjugation to 5(6)-carboxyfluorescein N-hydroxysuccinimide or 5(6)-carboxytetramethylrhodamine ester, antibody purification and elution were performed in phosphate buffer, and pH was adjusted to 8.00. The final antibody concentrations were evaluated from OD 280 nm (NanoDrop, Saveen & Werner AB, Limhamn, Sweden) or by the Bradford protein assay.

Serum stability
The MS5-Fc antibody (50 mg in PBS) was incubated in 40% human serum at 37°C in a humidified tissue culture incubator. The final volume was 350 ml. Aliquots (30 ml/each) were collected at various days and immediately stored at 220°C until analysis by Western blots using an anti-His tag mAb. Densitometric quantification of the reacting full-length protein was calculated by using the Image Lab TM 4.1 Imaging System (Bio-Rad, Hercules, CA, USA).

Flow cytometry analysis
Binding of polyclonal and monoclonal scFv phage antibodies to human cells was determined by using flow cytometry. In brief, cells were gently detached from the culture dish by scrapping, washed twice with PBS, and then aliquots of 10 5 cells were plated in a conical 96-well microplate in 100 ml staining buffer (PBS buffer + 2% FCS) and then incubated with phage antibodies (10 8 TU) for 30 min on ice. Cells were washed twice with 400 ml staining buffer and then incubated with biotin-conjugated anti-M13 mAb for 30 min at 4°C. After washing, cells were incubated with phycoerythrin (PE)-conjugated streptavidin, washed, and analyzed by using flow cytometry (Canto II; BD Biosciences, San Jose, CA, USA). Similarly, binding of MS5-Fc antibody and Fc control to cancer cells or normal cells was determined by using flow cytometry. Single-cell suspensions (10 5 cells) were incubated with the test molecules (5-10 mg/ml) for 30 min at 4°C in PBS buffer containing 2% human serum (staining buffer). After washing with staining buffer, cells were incubated with FITC-conjugated anti-human Fc IgG for 30 min on ice. Washing was repeated, and cells were resuspended in 300 ml staining buffer before being analyzed with the use of flow cytometry. In some experiments, cells were stained with biotin-labeled MS5-Fc antibody or Fc control followed by PE-conjugated streptavidin. All flow data were analyzed by using FlowJo software (FlowJo, Ashland, OR, USA).

Affinity measurements
Affinity measurements by ELISA for cell surface antigens were performed as described by Bator and Reading (16). Briefly, the antibody at various concentrations was mixed with cell suspensions incubated for 2 h on ice with occasional mixing before cells were pelleted by centrifugation. Supernatants were removed and retained for the quantification of unbound MS5-Fc antibody molecules using the human IgG ELISA Quantification Kit (Bethyl Laboratories, Montgomery, TX, USA). Bound MS5-Fc antibody concentrations were obtained by subtraction of unbound molecules from the initial antibody concentrations, and these values are then used to construct Scatchard plots.

Fluorescence microscopy analysis
Cancer cells were cultured in Lab-Tek chamber slides (Nalge Nunc International, Naperville, IL, USA) for 24 h in RPMI-1640 or DMEM medium supplemented with 10% FCS. The medium was replaced with X-VIVO 15 medium supplemented with 2% human serum, and the cells were incubated with 5(6)carboxyfluorescein-conjugated MS5-Fc antibody or Fc control (10 mg/ml) for 60 min at 4°C. Subsequently, the cells were incubated with Hoechst 33342 (1 mg/ml) for 5 min, washed, and then fixed with 4% paraformaldehyde for 30 min at 4°C. Slides were covered with Dako Cytomation fluorescent mounting medium before images were taken by using confocal microscopy (LSM 510; Carl Zeiss, Olympus, Tokyo, Japan). In some experiments, stained cells were incubated at 37°C for 3-6 h to allow receptor internalization and then processed as described earlier.

NK cell degranulation
To measure surface expression of CD107a/LAMP-1, a surrogate marker for NK cell degranulation, NK cells (5 3 10 4 in 200 ml X-VIVO medium) were incubated with the cancer cell-coated MS5-Fc antibody or Fc control for 5 h at 37°C. During stimulation, PE-Cy-conjugated anti-CD107a (2 ml/well) and monensin (0.2 ml/well) were added to the cell cultures. After incubation, the cells were harvested, washed, and stained with FITC-conjugated anti-CD56 before being analyzed with the use of flow cytometry. The data were analyzed by using FlowJo software.

ADCC assay
ADCC assay was conducted by using the Lactate Dehydrogenase Cytotoxicity Detection kit (Promega) in accordance with the manufacturer's instructions. Briefly, cancer cells (target cells) pretreated with either the MS5-Fc antibody or Fc control (10 mg/ml) were incubated with NK cells (effector cells) at various effector-to-target ratios for 18 h at 37°C. NK cells were prepared from human PBMCs. After incubation, plates were centrifuged at 300 g for 5 min, and 50 ml of supernatant from each sample was transferred to a 96-well plate to determine the amount of LDH released. The percentage of specific cell lysis was calculated as follows: (experimental releasebackground release/maximum releasebackground release) 3 100. The effector-to-target ratio (25:1) was determined from pilot experiments. In parallel experiments, culture supernatants were collected, and cytokine contents were measured by using ELISAs.

ADCP assay
To quantify ADCP, tumor cells were harvested by enzymatic dissociation, labeled with carboxyfluorescein succinimidyl ester (CFSE), washed with serum-free medium, and plated at a density of 5 3 10 4 cells per well in 100 ml X-VIVO 15 medium in a 96-well plate. After incubation at 37°C for 4 h, tumor cells were opsonized by addition of the MS5-Fc antibody or Fc control for 30 min at room temperature. Human M2 macrophages were harvested by cell scraping and subsequently pelleted, washed, and added to opsonized tumor cells at a density of 5 3 10 4 cells per well in 50 ml medium. Cells were incubated at 37°C for 10 h, pelleted, washed with X-VIVO 15 medium, and stained with anti-CD163 mAb. Samples were washed and then analyzed by using flow cytometry.

Analysis of the tumor-associated MS5-Fc antibody by fluorescence microscopy
Two weeks after tumor cell inoculation, 5(6)-carboxyfluoresceinlabeled MS5-Fc antibody or Fc control were injected intravenously into separate animals using 200 mg of the conjugate in 100 ml of physiologic saline. Mice were euthanized 20 h after injection, and tumors and other organs (lungs, kidneys, and heart) were removed and incubated overnight in 4% paraformaldehyde solution and then embedded in optimal cutting temperature medium. Tumors and tissue sections (10 mm) were incubated with Hoechst 33342 for nuclei staining. Thereafter, the slides were covered with Dako Cytomation fluorescent mounting medium before being examined by using an epifluorescence microscope (Leica DM RHC; Leica Microscopy As, Oslo, Norway). Samples were also examined by confocal microscopy (LSM 510; Carl Zeiss).

Xenograft animal models
Female BALB/c nude mice were produced at the animal core facility (Oslo University Hospital) and housed in microisolation cages during the course of the experiments. Mice were inoculated subcutaneously with cancer cells and then randomized into different groups (5-7 mice/group). Test molecules were administrated intravenously on various days after tumor cell inoculation (100 mg/mouse/injection). Animals were assessed daily for clinical symptoms and adverse effects. Tumor size was measured by using a caliper twice a week, and tumor volumes were calculated by using the formula (L 3 W 2 )/2, with L and W representing length and width, respectively. All animal studies were approved by the Institutional Animal Care Use Committee at Oslo University Hospital. Randomization was used in all animal experiments.

Analysis of tumor-infiltrating macrophages and NK cells
Detection of infiltrating immune cells in tumor tissues was performed according to standard immune-histochemical methods using 10 mm cryostat sections of tumors or organ tissues fixed with 4% paraformaldehyde and embedded in optimal cutting temperature medium. For detection of macrophages and NK cells, PE-conjugated anti-mouse F4/80 and PE-conjugated anti-NKp46 were used, respectively. Before incubation with antibodies, sections were blocked with 5% rat serum in PBS. Stained sections were analyzed with an epifluorescence microscope (Leica DM RHC; Leica Microscopy As).

Statistical analysis
Results are reported as means 6 SD. Statistical significance of differences was assessed by using Student's t test. The antitumor effects of the MS5-Fc antibody vs. the Fc control were assessed by using a 2-tailed test. The level of significance was set at a value of P , 0.05.

Isolation of cancer cell-binding scFv antibody fragments
To determine the feasibility of selecting antibodies that recognize common cell surface receptors expressed by cancer cells, a semisynthetic human antibody library was sequentially screened on different types of cancer cells. The library was first preabsorbed on PBMCs and normal HMECs to remove most of the phages that bind to receptors expressed by normal cells. Second, the preabsorbed library was sequentially affinity selected on breast MDA-MB-453, prostate PC3, lung SW900, glioblastoma U87MG, and lymphoma Ramos cancer cell lines (Fig. 1A). By alternating cancer cells during the selection process, the selected phage antibody fragments may bind to common receptors or carbohydrate/lipid structures expressed preferentially or exclusively by cancer cells. The results of phage enrichment during the affinity selection protocol are shown in Table 1. After selection on prostate cancer cell line PC3, the number of eluted phages increased from 2.1 3 10 3 to 8.5 3 10 4 , a 40-fold enrichment. After selection on Ramos cells, the number of eluted phages also considerably increased from 3.5 3 10 5 (selection on U87MG cells) to 4.5 3 10 6 . However, there was a moderate enrichment after selection on SW900 and U87MG cells, suggesting that these cancer cell lines may express fewer binding receptors. The overall enrichment was ;2080-fold. We next examined the binding of Ramos-selected phages to the used cancer cell lines. The screening protocol resulted in the enrichment for phages binding to all 5 cancer cell lines relative to the unselected original library (Fig. 1B). No significant binding to BM CD34 + hematopoietic stem cells and normal HMECs was detected. Thus, specific scFv antibody fragments were enriched by the phage-display selection protocol used in this study. Analysis of individual random phage clones confirmed the strong binding of the selected phages (Fig. 1C). Most of the selected single phage clones did not bind to blood lymphocytes or BM CD34 + hematopoietic stem cells. Analysis of the DNA inserts of 50 single phage clones that recognized cancer cells revealed the presence of 3 dominant phage clones ( Table 2). Importantly, 70% of all phage clones sequenced were found to be clone MS5, indicating that this clone was highly enriched. MS5 bound to most tested cancer cell lines, but weakly or not to peripheral blood lymphocytes and normal HMECs.
To investigate the impact of the selection protocol on the enrichment of MS5 and MS10 phage clones, 150 phage clones were randomly picked from the eluted phages (30 phages/cell type) and sequenced, yielding 125 intact antibody clones. Analysis of the CDR3 region sequences of the heavy chains indicated a clear enrichment of MS5 and MS10 phages (Supplemental Table 1). The CDR3 sequences of the MDA-MB-453-binding phages were different (24 unique cell-binding scFv antibody fragments, including scFv MS5 and MS10). The frequency of MS5 and MS10 phages increased after affinity selection on prostate cancer cell line PC3. There was a modest enrichment of these 2 phage clones after affinity selection on SW900 and U87MG cancer cell lines. However, a high proportion (84%) of the eluted phages after affinity selection on Ramos cells was found to be either MS5 or MS10 phages. Indeed, 13 and 8 of the 25 intact scFv antibodies displayed the heavy chain CDR3 sequence of scFv MS5 or MS10, respectively. Some uncontrollable factors, such as receptor density on the different cell lines, may facilitate the enrichment of these 2 phage clones. In contrast to all sequenced CDR3 regions,  the scFv MS5 has a longer synthetic CDR3 sequence, which is similar to natural antibodies.

Generation and characterization of the MS5-Fc antibody
Based on the flow data (Fig. 1C), we chose to focus on scFv MS5 for further study. Although soluble scFv antibody fragments have a wide range of applications in research, diagnostics, and therapy, the scFv-Fc format offers several advantages over candidate scFvs, including bivalent binding, longer half-life, and Fc-mediated effector functions (17). We therefore genetically fused the VH and VL gene segments of scFv MS5 to human IgG1 Fc domain (hinge, CH2-CH3) in frame with human IL-2 leader sequence (for secretion). The IgG1 domain contains the triple mutations DEL (S239D/I332E/A330L), known to enhance ADCC and ADCP via higher binding to Fcg RIIIa (18). Because of the hinge region, the scFv-Fc (MS5-Fc) fusion protein is expected to form an S-S linker dimer in solution.
Recombinant proteins were produced in HEK293T cells and purified by using protein G-agarose chromatography.
As a control, we purified the Fc domain of human IgG1 carrying the DEL mutations. The integrity of the MS5-Fc antibody and Fc control was confirmed by Coomassie Brilliant Blue staining of an SDS-PAGE gel ( Fig. 2A). Under reducing conditions, the fusion protein migrated as ;55 kDa monomers, whereas under nonreducing conditions, it migrated as a band of ;100 kDa, corresponding to the dimeric structure (12,19). Moreover, gel filtration showed a single peak eluted at the position around the size of the dimer (90-100 kDa), indicating that MS5-Fc antibody does not form aggregates larger than the dimer in solution (Supplemental Fig. 1A-D). We next assessed the binding of the MS5-Fc antibody to Ramos, MDA-MB-453, SW900, PC3, U87MG, and HL60 cancer cell lines. Consistent with the phage clone, the MS5-Fc antibody bound to the tested cancer cells (Fig. 2B). Its predicted binding affinity to MDA-MB-453, Ramos, and PC3 cells is high (5-10 nM). Table 3 shows the binding potency of the MS5-Fc antibody to a large panel of human cancer cell lines derived from solid and blood cancers. With respect to hematopoietic cells, the MS5-Fc antibody did not bind to hematopoietic stem/progenitor KG1a leukemic cells (20) or to Nalm-6, a pre-B acute lymphocytic leukemia. In line with this observation, the MS5-Fc antibody did not bind to normal BM CD34 + hematopoietic stem/progenitor cells. In contrast to either malignant or normal early progenitor cells, the MS5-Fc antibody bound to lymphoma and leukemia cell lines that are derived from intermediate and late myeloid or lymphoid precursors, respectively. Notably, the MS5-Fc antibody exhibited no binding to mature blood T cells and only a weak binding to B cells. Moreover, it did not bind to the myeloma U266 cell line, a cancer of mature B cells. Collectively, these cellbinding profiles suggest that the receptor for the MS5-Fc antibody is expressed on most malignant leukemia and lymphoma cells but not, or only weakly, on early progenitor cells and mature blood leukocytes.

The origin of the Fc domain does not affect the antibody binding profile
We next investigated whether MS5-Fc antibody fragment maintains its specificity when fused to a different Fc domain. Similarly, we genetically fused the scFv MS5 sequence with the sequence encoding the Fc domain of mouse IgG2a to generate MS5-mIgG2a fusion protein. We then produced the recombinant protein in HEK293T cells, followed by purification on protein G chromatography and characterization by Western blotting. Mouse IgG2a  was used as a control. Data evaluating cell populations in whole PBMCs revealed no binding of MS5-mIgG2a fusion protein to either lymphocyte (R1 gate) or monocyte (R2 gate) populations (Fig. 3A). Consistent with the data obtained with the human IgG1-Fc fusion protein, MS5-mIgG2a did not bind to purified blood T cells (Fig. 3B). A weak binding was also observed with purified blood B cells (Fig. 3C). Again, the hematopoietic progenitor leukemia cell line KG1a failed to bind to MS5-mIgG2a (Fig. 3D), whereas lymphoma and leukemia cell lines did (Fig. 3E, F).
Comparable results were obtained with a third construct containing the human IgG2 Fc domain (data not shown). In summary, the binding specificity of the MS5-Fc antibody was not affected by the nature of its Fc domain.

Redistribution and internalization of the MS5-Fc antibody receptor complexes
To gain insight into the interactions of the MS5-Fc antibody with cancer cells, confocal immunofluorescence microscopic analysis was used to characterize the binding. When the staining was conducted at 4°C, the MS5-Fc antibody receptor complexes were distributed homogeneously on the cell surface (Fig. 4A). After incubation at 37°C, the complexes clustered into specific membrane sites (Fig. 4B). Thus, the binding of the MS5-Fc antibody to cancer cells seems to induce the receptor redistribution into separate membrane compartments. Whereas the MS5-Fc antibody receptor complexes massively clustered at several sites of the cell membrane, those found in Ramos lymphoma cells concentrated in big spots, likely representing large assemblies of coalesced lipid rafts. Depending on the cell type, the MS5-Fc antibody may induce redistribution of the receptor complexes into separate membrane compartments. In the majority of analyzed cancer cells, the antibody binding did not induce receptor internalization. A representative example of receptor internalization upon MS5-Fc antibody binding to the prostate cancer cell line PC3 is shown in Fig.  4B. The Fc control showed no binding. Similarly, the MS5-Fc antibody exhibited no significant binding to normal blood T cells and B cells, myeloid precursor leukemia KG1a, and B-cell precursor leukemia Nalm-6 ( Fig. 4C).

MS5-Fc antibody induces NK-cell activation and ADCC
The CD107a, which is a component of cytotoxic granules, accumulates on the surface of activated NK cells and cytotoxic T cells upon granulation/activation (21,22). We therefore analyzed its surface display in response to the antibody treatment. MDA-MB-453, Ramos, and PC3 cells were used as target cells. As shown in The effects were antigen-specific, as KG1a cells (which more likely lack the MS5 scFv receptor) were not affected. In the context of antibody treatment, NK cells are unique in that they express only the low-affinity activating FcgR CD16 (FcgRIIIA) and no inhibitory receptors, underscoring a significant role in ADCC (23). The ability of MS5-Fc antibody to induce ADCC was assessed by using the CytoTox Non-radioactive Cytotoxicity Assay (Promega) based on LDH release (24). MDA-MB-453, Ramos cells, and HL60 cells were used as target cells and freshly isolated human NK cells as effector cells. Fig. 5B presents a significant increase in cytolysis by 5 mg/ml of the MS5-Fc antibody at an effector-to-target cell ratio of 25:1, revealing 25 6 5% cytotoxicity of MDA-MB-453, 30 6 5% of Ramos, and 20 6 5% of HL60 cells. The ADCC effect was antigen-specific, as KG1a cells were not affected. Similarly, autologous CD4 + T cells were not killed. In accordance with ADCC activity, cancer cell-coated MS5-Fc antibody induced IFN-g and TNF-a production by NK cells (Fig. 5C).

MS5-Fc antibody induces ADCP
In addition to ADCC, we investigated the capacity of MS5-Fc antibody to recruit macrophages and induce Normal mammary epithelial cells -Binding of the MS5-Fc antibody to the indicated cells was determined by using flow cytometry. Biotin conjugated MS5-Fc antibody was used in these experiments. Of note, a MS5-mouse IgG2a Fc fusion protein exhibited the same binding profile as the human MS5-Fc antibody. -, no significant binding; 6, very weak binding; +, weak binding; ++, medium binding; +++, strong binding; ++++, very strong binding.
ADCP in vitro. Target cancer cells were fluorescently labeled with CFSE and then incubated for 30 min with either MS5-Fc antibody or Fc control before adding monocyte-derived M2 macrophages and further incubation at 37°C for 8 h. Thereafter, the cells were stained with APC-conjugated anti-CD163, a marker for M2 macrophages (25), before analysis by flow cytometry (Fig. 6A, a representative example). Dual-labeled macrophages (APC/CFSE) are considered to represent phagocytosis of cancer cells by macrophages. The MS5-Fc antibody displayed a phagocytic activity against different types of cancer cells, whereas no significant effect was found on receptor-negative KG1a cells (Fig. 6B). The Fc control showed no significant effect.

MS5 antibody binds to primary tumor cells and induces ADCC
The MS5-Fc antibody was further characterized by using primary malignant cells of patients with leukemia in a functional assay using allogenic human NK cells. Figure 7A shows the binding of the MS5-Fc antibody to blood cells from 1 patient with chronic myeloid leukemia (CML). In agreement with the data obtained with cancer cell lines, the MS5-Fc antibody bound to primary leukemic cells (R1 gate) but not to the normal lymphocyte population (R2 gate) from the same patient. Similar results were obtained with 4 additional PBMC preparations from 2 patients with CML and 2 patients with acute myeloid leukemia (AML). We next used the samples from patients with CML and AML and allogenic NK cells to analyze the ability of the MS5-Fc antibody to induce ADCC. Leukemia cell lysis was significantly enhanced in the presence of the MS5-Fc antibody compared with that obtained with the Fc control (P , 0.02) (Fig. 7B).

MS5-Fc antibody localizes into tumor tissues
We next investigated whether the MS5-Fc antibody could target tumor cells in vivo. Nude mice bearing subcutaneous MDA-MB-453 or Ramos tumors were injected intravenously with 5(6)-carboxyfluorescein-labeled MS5-Fc antibody. After 20 h, animals were euthanized to assess tumor tissues and normal organs for fluorescence (Supplemental Fig. 2). As opposed to the Fc control, the data show that the MS5-Fc antibody can target tumor cells in vivo. Confocal microscopic analysis of tumor sections showed the binding of the MS5-Fc antibody to tumor cells. Thus, the engineered antibody penetrated into tumor tissues. No significant fluorescence was detected in normal tissues such as the lungs. A nonspecific uptake of antibody-free fluorescence dye into the tumor is unlikely because different MS5-Fc-fluorochrome conjugates produced similar results (data not shown).

MS5-Fc antibody inhibits growth of human tumor xenografts in vivo
We next investigated the capacity of the MS5-Fc antibody to inhibit tumor growth in vivo. First, the antitumor activity was evaluated in a human breast cancer model in which MDA-MB-543 cells (1 3 10 7 cells/mouse) were inoculated subcutaneously into BALB/c nude mice. These mice bear monocytes/macrophages and NK cells capable of mediating ADCC and ADCP (26). The mice were treated intravenously either with MS5-Fc antibody or Fc control (100 mg/injection) on d 3, 7, and 10. Tumor growth was monitored, and tumors were collected and weighed at the end of the experiments (d 20). As shown in Fig. 8A, MS5-Fc antibody treatment inhibited tumor growth. At d 20, tumors in MS5-Fc antibody-treated mice (n = 6) reached a mean volume of 1641 mm 3 , whereas the control mice (n = 7) developed tumors with a mean volume of 2754 mm 3 (P , 0.004). Consistent with tumor volume, the treatment reduced the mean tumor weight by 48% compared with the Fc control-treated mice.
An additional series of experiments following the same design were performed by using Ramos lymphoma cells. The cells (1 3 10 7 cells/mouse) were subcutaneously transplanted into nude mice (n = 5/group), and treatment was given intravenously on d 4, 8, 12, and 18. Mice treated with the MS5-Fc antibody displayed a significant reduction in tumor growth rate compared with those treated with the Fc control (Fig. 8B). At d 22, tumors in MS5-Fc antibody-treated mice had reached a mean volume of 518 mm 3 , whereas those treated with Fc control reached 1200 mm 3 (P , 0.002). In line with tumor growth curves, MS5-Fc antibody treatment reduced the mean tumor weight by 65% compared with Fc control-treated mice (d 22).
To extend the therapeutic use of the MS5-Fc antibody, we also evaluated its antitumor effect on leukemia cells. Nude mice (n = 4/group) were subcutaneously challenged with HL60 cells (1 3 10 7 cells/mouse) and then treated with either MS5-Fc antibody or Fc control on d 3, 7, and 11. Here, we decided to terminate the experiment at d 33 when tumors in the control group reached a mean size of 1200 mm 3 . Figure 8C shows tumor weights at d 33. Fc control-treated mice developed tumors with an average weight of 0.925 g, whereas those treated with MS5-Fc antibody developed smaller tumors with an average weight of 0.263 g (P , 0.001). One mouse in the MS5-Fc antibody-treated group did not develop a tumor.
Together, the data confirm the potential versatility of the MS5-Fc antibody in blocking tumor growth.
Given the importance of innate immune cells in antibody antitumor effects, we examined whether MS5-Fc antibody treatment enhances their infiltration into tumor tissues. Figure 8D displays representative tumor sections stained with monoclonal antibodies specific for NK cells (NKp46) or macrophages (F4/80). The MS5-Fc antibody treatment increased the infiltration of both NK cells and macrophages into tumor tissues compared with those treated with the Fc control. A similar increase in the number of infiltrating NKp46 + and F4/80 + macrophages was also seen in MS5-Fc antibody-treated HL60 tumors (data not shown).

Comparison of the MS5-Fc antibody antitumor effect vs. that of rituximab
The CD20 mAb rituximab is incorporated into standard care for B-cell NHL (27). In the next experiments, the Notably, the level of CD20 expression on Ramos cells is in general higher than that for the MS5 binding receptor (Fig. 9B). Given the challenges involved in overcoming resistance to rituximab treatment (e.g., loss of CD20 from the cell surface) (4), the engineered MS5-Fc antibody could be an attractive means for treating patients with lymphoma.

MS5-Fc antibody showed an enhanced stability in human serum
The stability in human serum is an important factor affecting the therapeutic efficacy of mAbs. Therefore, we investigated the stability of MS5-Fc antibody in 40% human serum. The antibody was incubated at 37°C for various time points, and the stability was assessed with Western blot experiments using an anti-His tag mAb to detect the full-length protein (Fig.  10A). Densitometric quantification of the signals indicates that ;60% of the antibody remained intact after 6 d of incubation (Fig. 10B), indicating that the engineered MS5-Fc antibody is not highly susceptible to serum proteases.

DISCUSSION
In the present study, we found that scFv antibody libraries provide a rich source of antibody fragments that specifically recognize receptors expressed by diverse tumor cells. One of the selected scFv antibody fragments was turned into a human IgG1 antibody that killed cancer cells in vitro and inhibited tumor growth in 3 different human xenograft models, suggesting that it could have widespread therapeutic use for a variety of cancers.
Consistent with its binding profile, the MS5-Fc antibody induced ADCC and ADCP against various cancer cell lines. The use of KG1a cells and blood T cells, more likely not expressing the MS5-Fc receptor, as targets revealed no significant effect of the MS5-Fc antibody on NK cell activation and induction of ADCC, supporting the assumption that the antibody effects are receptor dependent. From the flow cytometry analyses, we can conclude that the receptor for MS5 scFv is expressed in lymphoma and leukemia cells but not in CD34 + hematopoietic stem cells, early hematopoietic progenitor cells, or mature peripheral blood leukocytes; a weak binding to B cells was detected, however. This binding profile is interesting because most, if not all, current therapeutic antibodies against lymphoma and leukemia cells also target normal blood leukocytes (2,3). For instance, rituximab, which is part of the standard treatment regimen for B-cell lymphoma, kills both normal and malignant B cells, resulting in long-term profound deletion in circulating B cells. Thus, targeting proteins expressed in normal blood cells requires that the patient can do without the normal cells from which the cancer cell originates. With respect to therapy, antibody internalization is an important issue to consider during antibody development projects. Tumor-specific antibodies that internalize efficiently provide a means for delivering cytotoxic drugs into target cells (28). However, for cancer immunotherapy relying on ADCC, ADCP, and/or complement-mediated cytotoxicity, noninternalizing antibodies are preferred. This approach would maximize the accessibility of the Fc domain to immune effector cells such as NK cells, macrophages, and neutrophils. Notably, antigen downregulation and rapid internalization have been some of the main reasons for the limited therapeutic efficiency of several mAbs, including gemtuzumab (anti-CD33) and rituximab (4,(29)(30)(31)(32). Under our experimental conditions, the confocal microscopic analysis showed that the MS5-Fc antibody is not internalized by the majority of examined cancer cell lines. Thus, it should be suitable for the activation of innate immune cells against tumor cells. Previous studies have shown that antibody internalization often depends on the target antigen and the epitope recognized by the antibody (31). Our data suggest that internalization may also depend on the cell types. Indeed, the prostate cancer cell line PC3 internalized the MS5-Fc-receptor complexes, but the breast cancer cell line MDA-MB-453 did not (Fig. 3).
To evaluate the antitumor effects of MS5-Fc antibody, 3 aggressive xenograft models were used. The MDA-MB-453 breast cancer cells do not respond to hormone therapy and are especially aggressive due to their high metastatic potential. The human Ramos Burkitt lymphoma is a type of high-grade NHL, and it is frequently used to assess the therapeutic performance of antibodies targeting lymphoma (33). The HL60 cells, perhaps the most widely used, best-known myeloid cell line, originated from a female patient with AML (34). Adults with AML have some of the  highest unmet needs of all cancer patients (35). In these 3 different human tumor xenograft models, the MS5-Fc antibody exhibited significant antitumor activity. Indeed, only 3-4 intravenous injections of small amounts (100 mg/ injection) of the antibody were sufficient to inhibit tumor growth, measured by using tumor size and/or tumor mass, compared with Fc control-treated mice. Moreover, the treatment was accompanied by infiltration of innate immune cells into tumor tissues. Previous preclinical and clinical studies support the role of effector cells, especially NK cells and macrophages, in the antitumor activity of several therapeutic antibodies such as rituximab, trastuzumab, and cetuximab (3,36,37). The in vivo experiments also revealed that the MS5-Fc antibody and rituximab have comparable antitumor effects. Although comparable until d 18 post-cancer cell inoculation, the effect of rituximab was superior to that of the MS5-Fc antibody at d 22 (7 d after the last treatment). This enhanced activity could be due to rituximab's higher serum half-life in mice and/or the high expression of the CD20 receptor on Ramos B-cell lymphoma. A half-life of 9.5 d was found for the full-length human IgG1, whereas that of scFv-IgG1-Fc fusion was ;4.5 d (38). Despite being effective in the treatment of B-cell lymphoma patients, ;50% of patients with relapsed/refractory CD20 + follicular lymphoma do not respond to initial therapy with rituximab, and close to 60% of previously rituximab-responsive patients will no longer benefit with retreatment with rituximab (27). Similarly, the clinical outcome of high fraction (40%) of diffuse large B-cell lymphoma remains unsatisfactory after rituximab treatment (39). There are currently no effective therapies for these patients, giving rise to a significant unmet medical need. With its antitumor effect, the MS5-Fc antibody could lead to novel therapeutic treatment for millions of patients with lymphoma or leukemia. This notion is further supported by the finding that MS5-Fc antibody can bind and kill primary leukemia cells in vitro. Moreover, it showed a significant serum half-life in vitro and exhibited no significant aggregation in solution. As discussed by Roberts et al. (40), protein aggregation presents a key challenge in the development of therapeutic proteins.
Screening of phage libraries on intact cells preserves the original conformation of cell surface proteins and protein-protein interactions that could be relevant in vivo. Although a number of selected antibodies and peptides from phage display libraries have been used for tumor imaging and/or drug delivery without knowledge of the cellular receptors, their clinical use will be further facilitated by the characterization of the binding receptors. Immunoprecipitation experiments followed by spectrometric identification of captured protein species identified HSPs, including HSP70 and GRP78, as potential MS5 interacting partners (data not shown). HSP70 and GRP78 are known to be expressed on the cell surface of most cancer cells but not normal cells (41). These expression patterns fit well with the binding profile of the MS5-Fc antibody to cancer cells. However, when we investigated the binding of the MS5-Fc antibody to recombinant HSP70 and GRP78 proteins in Western blot experiments, no reactivity was detected (data not shown). We believe that the receptor is formed by the interaction between at least 2 protein partners expressed on the cell surface of cancer cells. Hence, the binding may arise from the formation of a native conformational epitope involving the interaction of these protein partners. Such interaction is more likely to be lost upon the preparation of membrane proteins for affinity purification. Therefore, a combination of several cellular and molecular biology techniques will be needed to tackle this challenging question that is under investigation.
In summary, we demonstrate here that sequential affinity selection of an antibody library againts a panel of human cancer cell lines can lead to the isolation of antibodies with pan-cancer binding abilities. The engineered MS5-Fc antibody showed in vitro and in vivo efficacy against cancer cell lines of different origins. Given the emergence of unresponsive tumors to the current immunotherapies and the fact that many cancer types still have no immunotherapy options available, the developed MS5-Fc antibody represents an attractive candidate for further development toward clinical studies. Moreover, the identification of antibodies that recognize common cell surface antigens/structures specifically or preferentially expressed by cancer cells may open the avenue to truly cheaper cancer immunotherapy.