The potential tumor-recognizing capacity of B cells infiltrating human breast carcinoma is an important aspect of breast cancer biology. As an experimental system, we used human medullary breast carcinoma because of its heavy B lymphocytic infiltration paralleled to a relatively better prognosis. Ig-rearranged V region VH-JH, Vκ-Jκ, and Vλ-Jλ genes, amplified by RT-PCR of the infiltrating B cells, were cloned, sequenced, and subjected to a comparative DNA analysis. A combinatorial single-chain variable fragment Ab minilibrary was constructed out of randomly selected VH and Vκ clones and tested for binding activity. Our data analysis revealed that some of the VH-JH, Vκ-Jκ, and Vλ-Jλ region sequences were being assigned to clusters with oligoclonal predominance, while other characteristics of the Ab repertoire were defined also. A tumor-restricted binder clone could be selected out of the single-chain variable fragment κ minilibrary tested against membrane fractions of primary breast tumor cells and tumor cell lines, the VH of which proved to be the overexpressed VH3-1 cluster. The specific binding was confirmed by FACS analysis with primary breast carcinoma cells and MDA-MB 231 cell line. ELISA and thin layer chromatography dot-blot experiments showed this target Ag to be a ganglioside D3 (GD3). Our results are a proof of principle about the capacity of B cells infiltrating breast carcinomas to reveal key cancer-related Ags, such as the GD3. GD3-specific Abs may influence tumor cell progression and could be used for further development of diagnostic and/or therapeutic purposes.
Recombinant Abs and their fragments represent now a very high percentage of all biological proteins undergoing clinical trials for diagnosis and therapy (1, 2). Clinical research in the area of Ab-based tumor-targeted therapy has been driven for many years by the prospect of identifying cell surface Ags with sufficient restrictive tissue expression patterns to allow a selective and specific tumor tissue accumulation of the Ab (3). Technology for Ab design has taken enormous strides forward through new library display, and selection procedures that serve as perfect tools for searching for tumor-specific or any other Ags have been described (4, 5). Even with these new procedures, there are difficulties in identifying real and reliable tumor-specific structures with characteristics that enable them to be used further as potential targets for tumor diagnostics or therapeutics (6, 7). Among the strategies to obtain reliable tumor target molecules, one of the most challenging and perhaps insightful is the exploitation of the intelligence of the natural human immune response to tumor-restricted Ags (8, 9, 10, 11).
The presence of B lymphocytes that infiltrate cancerous tissues may reflect an ongoing immune response against transformed cells. In addition, these cells may provide an interesting source for obtaining Abs to tumor Ags, and therefore some insight into the natural anti-cancer immune response. Although some important earlier investigations in melanoma have been described (12, 13, 14), almost all efforts concerning infiltrating lymphocytes were focused on T cells (15, 16). The paucity of information concerning tumor-infiltrating B (TIL-B)3 cells has, in part, been due to the low amount of these cells found in various tumor tissues. Further studies on melanoma (17); neuroectodermal tumor (18); lung (19), ovarian (20), and colon cancer (21); and recently on breast cancer types (22, 23, 24) have shown that TIL-B cells are of potential interest. Immunocompetent B cells in some solid tumors were contributed to spontaneous tumor regression as well (25). Both naturally occurring and vaccine-induced Ab responses to some breast cancer Ags could be associated with improved survival in some cases (26).
We postulate that TIL-B cells have specific tumor-recognizing capacity, and could serve as a new source besides peripheral blood and lymph nodes to search for tumor-binder Abs. We postulate that the analysis of the Ab repertoire of these TIL-B cells could lead to the further understanding of the precise nature of natural Abs to tumor Ags.
To address this hypothesis, a detailed Ig repertoire analysis of the expressed human medullary breast carcinoma (MBC) TIL-B cell Ig V regions was performed. Our model system for characterizing the expressed Ig repertoire (22, 27, 28) was established from high grade MBC because this has the highest B cell infiltration (29). A combinatorial single chain Fv κ (scFvκ) minilibrary from selected VH and Vκ region genes was generated and tested thereafter for breast tumor cell-binding capacity. The findings presented in this work provide a new strategy for identifying novel tumor-specific Ags, and serve as proof of principle for the potential tumor Ag-binding capacity of B lymphocytes infiltrating human breast carcinomas.
Materials and Methods
PCR amplification of TIL-B Ig V regions
Tumor tissue was obtained aseptically during surgery from a patient with MBC. The tissue sample was minimized for attached normal tissue. RNA was extracted according to the manufacturer’s instructions (RNeasy Mini kit; Qiagen), while cDNA was synthetized by the commercially available kit (Pharmacia Biotech). To amplify the human Ig VH-JH, Vκ-Jκ, and Vλ-Jλ encoding regions, specific primers were designed (30) and PCR was performed (35 cycles: 1 min, 94°C; 1 min, 60°C; and 1 min, 72°C; PerkinElmer/Cetus thermocycler).
Cloning and sequencing of VH-JH, Vκ-Jκ, and Vλ-Jλ TIL-B Ig regions
PCR products were purified and blunt end ligated into pUC18 (SmaI/BAP) plasmid vector (Pharmacia Biotech). Escherichia coli TG1 bacteria were transformed, and positive clones were selected by PCR, as described (22, 28). Sequencing of plasmid dsDNA (QIAprep Spin Miniprep kit; Qiagen) was performed partly by Sequence Version 2.0 DNA Sequencing kit (USB) and mainly by automatic sequencing (Dye Terminator Sequence Reaction Kit, DyeEx Spin kit (Qiagen; ABI PRISM Software, automatic sequencer of PerkinElmer). More than 60 VH, Vκ, and Vλ clones were sequenced, respectively. Further comparative DNA sequence analysis was performed by accessible software and databases available through the Internet: BIOEDIT 5.0.9 (31) was used for editing sequences, Clustal X 1.8 (32) for sequence alignment, and TREEVIEW 1.5.2 (33) for displaying sequence trees based on homology level. Sequence comparison was made to KABAT National Institutes of Health (〈http://immuno.bme.nwu.edu〉), IMGT (〈http://imgt.cines.fr.〉), GenBank, Embnet via National Center for Biotechnology Information Blast Engine (〈www.ncbi.nlm.nih.gov/BLAST/〉), and SRS (〈http://srs.hgmp.mrc.ac.uk〉). Our sequences were compared with germline sequences and related coding regions (V, (D), J) participating in the whole length of the V region according to the International ImMunoGeneTics database results (34) and referred to compiled germline. To find homologous sequences, a databank search via National Center for Biotechnology Information Blast server to GenBank/European Molecular Biology Laboratory Net databases was conducted and the generated data termed as Blastn result. More details about data analysis will be published elsewere (35).
Construction of combinatorial scFvκ minilibrary
The construction of the scFvκ minilibrary was conducted, as described (27, 28), by selecting VH and Vκ genes and amplified with specific primers based upon DNA sequence analysis. Assembly reactions of the selected rearranged Ig V region H (26VH) and L chain (32Vκ) genes were conducted by a three-step PCR amplification, using a linker peptide (Gly4Ser3) coding sequence. The purified and SfiI- and NotI-digested Vκ-Jκ fragments were ligated into pHEN1 phage vector, and E. coli TG1 bacteria (Stratagene) were infected according to methods (36) slightly modified. The scFvκ combinatorial minilibrary generated had 832 possible combinations and 5 × 106 members in size.
PCR screen and fingerprint analysis of scFv inserts
The scFvκ combinatorial minilibrary was plated on selective culture medium (100 μg/ml ampicillin 1% glucose 2×TY) at 30°C. Ninety clones were randomly picked, and PCR screening for the scFv insert was performed using LMB2 (forward) and LMB3 (backward) primers and standard conditions (30 cycles at 94°C for 1 min, 55°C for 1 min, and 72°C for 1.5 min, followed by an extension at 72°C for 10 min). A masterplate was set up in 2×TY ampicillin/glucose culture medium containing 15% glycerol and kept frozen until further usage. In addition, the amplified PCR samples were digested with 5 U of the enzyme HaeIII for 1 h at 37°C, and the restriction pattern of fingerprint analysis was visualized by gel electrophoresis on 3% agarose.
Cell membrane preparation of fresh cultured tumor cells and cell lines
Cell lines of different histological origin (MDA-MB 231 and MDA-ZR75-1 breast cancer cell lines, 293 kidney epithelial cell line, LS174T colorectal carcinoma line, SK MEL-28 melanoma cell line) and the control COS7 cell lines were obtained from American Type Culture Collection. They were cultured in RPMI 1640 (Invitrogen Life Technologies) supplemented with 10% (v/v) heat-inactivated FCS (Invitrogen Life Technologies), 50 μg/ml penicillin (Invitrogen Life Technologies), 100 μg/ml streptomycin, and 2 mM l-glutamine (Invitrogen Life Technologies). All cells were cultured at 37°C in a 5% CO2 atmosphere. Growth medium was changed every 3 days, and cells were subcultured according to their growth rate. Small pieces of tumor tissue (0.6 × 0.6 × 0.6 cm), aseptically obtained after surgery from two patients with breast carcinomas, were minced under medium supplemented with 20% culture supernatant from the 293 cell line in growing phase and 0.01% sodium pyruvate and kept under suitable conditions. Cultured cells (1 × 107 each) were processed further to obtain membrane fractions, as described (37).
Soluble scFv ELISA on membrane preparation
Ninety-six-well Maxisorp (Nunc) microtiter plates were coated (16 h, 4°C) with 1–10 μg of membrane preparations of breast carcinoma cells and control cell lines. Plates were washed five times with PBS and blocked with 200 μl of 2% BSA (Sigma-Aldrich) in PBS. Soluble scFv production was induced from individual bacterial clones by standard isopropyl β-d-thiogalactoside (IPTG) induction procedure. In ELISA, when testing soluble scFv fractions against tumor cell membrane preparations, anti-c-myc mAb (Sigma-Aldrich), HRP-conjugated anti-mouse Ab (Amersham), and the ABTS (Boehringer Mannheim) were used and analyzed by an ELISA reader (415 nm) (MWG Biotec). In further blocking experiments, selected hybridoma supernatants with known specificity, alkaline phosphatase (AP)-conjugated anti-c-myc Ab (Sigma-Aldrich), and p-nitrophenyl phosphate (Sigma-Aldrich) subtrate system were used, and the reaction was evaluated at 405 nm.
Dot-blot analysis of gangliosides
A total of 5 μg of seven different gangliosides (GD1a, GD1b, GD2, GD3, GM1, GM2, and GM3; Calbiochem) was placed on silica plates (2–20 μm porus size; Sigma-Aldrich), dried, and blocked with PBS/2% milk for 2 h. Soluble fraction (10 μl) produced by the tumor-binder TIL-B scFvκ bacterial clone was added onto the dots containing the targeted gangliosides and dried. After washing and drying the plates, AP-conjugated anti-c-myc Ab in PBS-BSA (1%) and bromochloroindolyl phosphate/NBT substrate (Sigma-Aldrich) were used for analyzing the binding capacity. Negative controls were set up with culture supernatants of pHEN1 vector/no insert containing E. coli TG1 bacteria. As a positive control, we used supernatant from the hybridoma, HCB-C3 with known GD3 specificity (M. Glassy, unpublished observations). To evaluate the binding potential of the G2scFv Ab derivate (dAb) fragment to purified gangliosides and different relevant cell lines, further dot-blot experiments were set up with membrane preparations of breast tumor cells (TU1, TU2, MDA-MB231) and irrelevant cell lines (COS7).
FACS analysis with soluble scFv fractions
From two patients with invasive ductal carcinoma, two separate breast cancer cell lines were established from the surgically removed tissues. Cells from log-phase growing cultures with >95% viability were gently scraped off, and 3 × 105 cells per test tube were used. Soluble scFv fraction of the binder-positive clone was prepared by standard IPTG induction method in large scale. Tumor cells were incubated with soluble scFv Ig fraction (diluted 1/1 in FCS RPMI 1640) at 37°C for 1 h in pretreated (1% BSA) plastic tubes. After three washes with FCS RPMI 1640, anti-c-myc 9E10 mAb (Sigma-Aldrich) was added for another 1-h incubation. Three washes with 1% BSA PBS were followed by administration of FITC anti-mouse IgG F(ab′)2 (Sigma-Aldrich) for 1 h at 4°C. Cells were washed three times with 1% BSA PBS and PBS, and fixed in 1% Formalin-PBS. Ten thousand cells were counted in a FACSCalibur (BD Biosciences) and analyzed by CellQuest.
Revealing the V gene usage, mutational pattern, and other characteristics of B cells infiltrating MBC by comparative DNA sequence analysis
A significant lymphocyte cell infiltration was found in human MBC, as shown in Fig. 1.
VH-JH, Vκ-Jκ, and Vλ-Jλ V regions were cloned, and the V gene usage was analyzed. Of the clones with the expected insert size, 67 VH (460 bp), 72 Vκ (440 bp), and 63 Vλ (430 bp) clones were selected for sequencing. The VH, Vκ, and Vλ sequences could be grouped into different families based on searching for the original germline sequences by IMGT database. Expressed Ig V regions were determined using the KABAT database, and characteristics were defined using commercially available softwares. Fifty-one percent (24 of 47) of the VH region sequences belong to the VH3 family, and the others are members of the VH4 (23.4%), VH5 (21.2%), and VH1 (4.3%) families. We found the Vκ1 as being the most represented family (19 of 38 (50.0%)), followed by Vκ4 (29.0%), Vκ2 (18.4%), and Vκ3 (2.6%). Based on the highest homology level (90–100%) to IMGT-defined germline sequences, a family ranking has been made. TIL-B VH and VL sequences with the same compiled germline sequence were grouped into one cluster. In most of the families, clusters with high internal homology (94.5–99.8%) could be distinguished. Some clusters defined were overrepresented, while others contained just a few members (Fig. 2). The VH3 family sequences (72%) could be grouped into the VH3/1 (10 of 22 (45%)) and VH3/4 (5 of 22 (23%)) main clusters. Of the eight clusters, 42% (8 of 19) of the Vκ1 sequences belong to the overrepresented one. The Vλ V region sequences showed less variability as the three families were represented by approximately equal members, that is 26.9% (7 of 26) Vλ1, 38.5% (10 of 26) Vλ2, and 34.6% (9 of 26) Vλ3.
Sequences belonging to the defined clusters were grouped into a tree structure. The overrepresented VH3/1 cluster showed a high difference (>12%) in comparison with all other VH3 sequences. While compiling the nearest germline sequences to our defined clusters, some genes seem to participate in more cases. Some combinations are more frequent (VH3a/D4a/J4a) than others, suggesting more efficiency in the representation. By contrast, when the same VH segments were combined with the above different D or J segments (VH3a/D4b/J4b), they resulted in underrepresentation of the given Ig. The tendency in how the nearest VH3 sequences build groups is depicted in Fig. 3,A. Contrary to that, the divergence among sequences of VH5 family is very low (0.2–0.4%), and these sequences share the closest germline sequence found (Fig. 3 B).
Divergences among different Vκ1 sequences may be significant. Based on a very low divergence level (0.2%) and one mutual germline sequence, eight Vκ sequences could be classified into the overrepresented cluster (Vκ1/1). All of the other Vκ1 sequences differed >5% to that one (Fig. 3 C). Representative data about the Vλ3 L chain family showed an overrepresented cluster (Vλ 3/1) with 90% of the sequences with the same germline origin (data not shown).
In the case of selected representative VH and VL families, we investigated the number of the members belonging to one defined cluster, the length of sequences, and the number of replacement and silent (R/S) mutations in the framework, as well as in the CDRs. The length of the CDR3 region; the deletion and/or insertion mutations in the V, D, and J regions; and the percentage of identity in CDR3 were also analyzed. The VH, Vκ, and Vλ lengths defined were 351 and 369 bp in the overrepresented VH3 and VH5 families, 324 and 342 bp in the representative Vκ families, and 333 and 321 bp in the selected Vλ cases. A high number of R as well as S mutations was found in the framework regions of the overexpressed VH3/1 cluster that was not observed in the VH5 family (Table I). A relatively high R/S ratio in CDR3 was characteristic only for the VH3/1 cluster. Deletion mutations caused the low homology level of CDR3, in comparison with the germline region. A very high internal homology level and a low closest germline homology were characteristic of the VH3/1 cluster. The Vκ4/1 cluster showed a 3.3 R/S mutation ratio, whereas mutations in the Vλ families were difficult to detect. CDR1, CDR2, and CDR3 region sequences of the most relevant clusters were compared with each other and the blast query result. DNA mismatches were defined to each other and to the germline sequences, and the gaps representative for the given CDR3 were determined as well (Table I).
|Cluster (members no.) .||Length .||R/S in FR Mutation .||R/S in CDR 1, 2 Mutation .||CDR3 Length/Deletion (d) Insertion (i)/Identity % .||Identity % .||.||.|
|.||.||.||.||.||Internal .||Germline .||Blastn .|
|Vλ2 (10)||333||1/0.1||0.4/0||35/Vd4, Jd2/83||99.0||97.6||98.2|
|Cluster (members no.) .||Length .||R/S in FR Mutation .||R/S in CDR 1, 2 Mutation .||CDR3 Length/Deletion (d) Insertion (i)/Identity % .||Identity % .||.||.|
|.||.||.||.||.||Internal .||Germline .||Blastn .|
|Vλ2 (10)||333||1/0.1||0.4/0||35/Vd4, Jd2/83||99.0||97.6||98.2|
Total lengths of the cloned sequences, ratio of R/S mutations in framework regions (FR) 1, 2, 3, and 4, CDR1 and CDR2 regions, and the average mutations per sequence are indicated. The CDR3 length includes the R/S region gaps. The percent of identity refers to the germline sequence (IMGT database was used).
Characterization of the scFvκ combinatorial minilibrary and selection of tumor-binder clone
Ninety-seven percent of the clones of our scFvκ library (5 × 106 members) contained an insert. Of these, 45% contained a scFv insert (∼800 bp), and 54% of clones had a 600-bp truncated scFv (Fig. 4). Thirty-three percent of the clones with an insert of 600 bp showed identical fingerprint patterns (data not shown). DNA sequencing showed that these truncated clones were comprised of a VH, the linker, and a part of the framework region 1 of Vκ.
Of 80 soluble scFv-producing bacterial clones (exhibiting either the 800- or the 600-bp insert) tested by ELISA, 1 clone (G2/clone 15) showed a binding capacity to membrane fractions of fresh cultivated breast cancer cells of an invasive breast carcinoma patient, and of the MDA MB-231 breast cancer cell line (Fig. 5). No binding could be detected to the LS174T colorectal cancer, 293 kidney epithelial, and the control COS cell lines. The result shows that by this strategy already, without previous panning procedures, just after setting up a master plate from the scFvκ library in one 96-well ELISA plate, we were able to obtain a tumor-binder clone. After sequencing the positive clone (G2/clone 15) and making comparative data analysis with BLAST, a 98% identity in the VH region to a GD2/GD3-binder Ab was found. CDR1, CDR2, and CDR3 VH sequences showed high identity level between the G2 clone and that of the GD2/GD3-binder VH sequence (Fig. 6). The truncated Vκ L chain belongs to the Vκ4 family. There was one nucleotide difference in the linker sequence of this scFv dAb G2 clone. Interestingly, the G2 clone had only expressed a 25-nt-long region of a Vκ chain. Sequence analysis of nine randomly picked negative clones from our masterplate showed that three of them had a full-length DNA as scFv, and two others were truncated scFv-s (with a truncated VH in framework region 1 at the 5′ end). The four other clones were of smaller size (truncated scFv-s), with complete (3) or shortened (1) VH regions fused either to a complete (3 case) or shortened (1 case) linker sequence and a truncated Vκ.
Characterization of the binder TIL-B scFv clone shows evidence of GD3 ganglioside binding
ELISA-blocking experiments were performed with human or mouse hybridoma supernatants having known ganglioside-binding capacity. Membrane fractions of the MDA MB-231 breast carcinoma cell line and of fresh cultivated tumor cells were coated to polystyrene plate and incubated with soluble scFv dAb G2 clone. In other experiments, the GD3-reactive Ab (HCB-C3) was added before the soluble scFv dAb G2. The direct binding of soluble scFv dAb G2, detected using an AP-labeled anti-c-myc Ab, was inhibited by the anti-GD3 mAb HCB-C3 (up to 75%) (Fig. 7). When another ganglioside-specific Ab (HCB-C3) and indifferent control Ig were added, there was no blocking effect. In addition, the scFv dAb G2 showed reactivity to a dried cell membrane ghost preparation of the GD3 ganglioside-positive SK MEL-28 (human melanoma cell line), whereas there was no binding to LS174 colon carcinoma cell ghost, known to be ganglioside negative. Finally, the soluble fraction of scFv dAb G2 bound to GD3 ganglioside on silica gel-coated membrane in blotting experiments. There was only a marginal or no staining to the other six ganglioside types or the negative control sample. Binding capacity of our G2 scFv dAb fragment to gangliosides and membrane preparations of some relevant breast cancer and other cell lines loaded on TLC plates showed evidence of GD3 specificity (Fig. 8).
Soluble scFv Ab fragments out of the TIL-B minilibrary show binding capacity to invasive breast cancer cells by ELISA and FACS analysis
Two adherent robust cell lines, TU1 and TU2, were established out of the separated and cultured cancerous tissues of two patients with invasive ductal breast cancer (Fig. 9). Although a detailed characterization of these lines will be published elsewhere, some features concerning ganglioside expression are explained and presented in Fig. 10. The reaction pattern with HCB-C3 ganglioside D3-specific Ab and culture supernatants containing soluble scFv anti-GD3 (G2 clone, truncated scFv, B2 clone, whole scFv) fragments on ganglioside-expressing cells showed further evidence concerning GD3 binding of the selected clones. Mid-log phase cells showed strong HLA Ag positivity with the W6/32 mAb, and MUC-1 positivity of breast tumor TU2 cells was found (Fig. 10). The surface-attached cultured breast cancer and melanoma cell lines suddenly detached from the chamber slide after the reaction with anti-ganglioside Abs, giving a further indirect evidence about the targeted structures, as that is a characteristic feature of gangliosides. In the immunofluorescence experiments, the mouse mAb 9410 to anti-c-myc was used as a control, and it was negative. Culture supernatants containing soluble scFv anti-GD3 (G2 clone, truncated scFv) showed positive binding to both the TU1 and TU2 breast cancer cell lines. As the binding intensity was similarly strong, only data with the TU2 invasive ductal breast cancer cells are shown (Fig. 11,A). Other selected clone supernatants from the phage library with the same overexpressed VH3/1 VH DNA sequence, but representing a whole scFv fragment (B2 scFv), also gave positive fluorescence with the given cells (Fig. 11,B). Similarly, incubation with soluble scFv fragments, obtained after panning against the MDA-MB 231 cell line, resulted in significant positive tumor cell binding (Fig. 11 C). The SK Mel-28 human melanoma cell line, being strongly positive for the ganglioside GD3, was the target for the hybridoma HCB-C3. The immunofluorescence labeling reaction with HCB-C3 anti-GD3 Ab against the TU1 and TU2 breast cell lines was significant. However, binding of the G2 scFv dAb and the immunofluorescence positivity against breast tumor cells was even stronger, while no remarkable labeling to normal cells was found (B. Kotlan, J. Toth, J.-L. Teillaud, M. McKnight, and M. Glassy, manuscript in preparation).
The nature of the human Ig repertoire in cancer patients is of interest and could provide important insight into aspects of tumor biology. The TIL-B cells that accumulate in solid tumor tissues could be a potential source for identifying novel tumor Ags. The data presented in this work on comparing TIL-B cell Ig V region sequences at the DNA level confirm our earlier observations on the oligoclonal nature of Ig repertoire in MBC (22) and suggest that the TIL-B Ig repertoire may be Ag driven. In the defined four VH, four Vκ, and three Vλ families, some overrepresented clusters could be distinguished according to the tree analysis. The VH5 and selected Vλ families showed internal similarity, whereas VH3 and Vκ1 families were moderately diversified. Some characteristic features of early B cells have been observed, such as a rather low mutation rate distributed in Ig framework regions and a significant internal sequence homology. Interestingly, a higher than 3 R/S ratio in CDR regions was found only in the selected overrepresented clusters. Our earlier and present results together with the recently published data about ductal breast carcinoma and MBC (22, 24, 35, 38) suggest that at least part of the B cell repertoire of TIL-B cells is tumor-associated Ag driven (39, 40, 41). The focus of our study was to obtain experimental proof of this Ag-driven response by identifying TIL-B-specific tumor-binder cancer-restricted clone(s).
Our investigations suggest that a tumor-restricted response of clonally selected TIL-B cells in MBC does exist. We selected a positive tumor-binder scFv (dAb VH) clone from our scFvκ minilibrary that showed strong restricted binding to invasive ductal breast carcinoma cells. Although this positive clone was first thought to be a truncated scFv with only a functional VH domain, it showed a tumor-restricted pattern and significant binding potential to invasive ductal breast carcinoma cells. Such VH-binding ability has been previously detected in melanoma (42) and autoimmune cases (43). DNA sequence analysis showed that our VH is a member of the overrepresented VH3/1 cluster. The VH3/1 cluster showed very high homology to the VH of a GD2/GD3 ganglioside-binder Ab. In addition, one isolated clone from the Vκ4/1 cluster showed a high identity to a ganglioside (GM3)-binder Ig VL chain (data not shown).
Gangliosides have been extensively investigated for their structural and functional properties (44, 45), as well as their role in tumor cell transformation and potential capacity as suitable tumor targets for diagnostics or therapeutics (46, 47). In addition to the well-defined GD2 overexpression in neuroblastomas and melanomas (48, 49), GD3 may be a useful marker in breast carcinomas. As GD3 gangliosides are poorly immunogenic, it has been difficult to obtain human Abs to them. Therefore, a new source and method of obtaining Ab fragments that react with the ganglioside D3 on breast cancer may be significant. Because GD3 is associated with invasive ductal carcinomas (50, 51), it is of interest to discover whether our TIL-B-derived Ab fragments recognize this Ag. Because no anti-GD3 Ab fragments from TIL-B cells have been reported to date, our data serve as first proof and potential for obtaining these human Abs out of MBCs. The importance and capacity of gangliosides and ganglioside-specific Abs as markers of tumor cells have been investigated (52, 53, 54, 55), and therapeutic trials have been put on the way (56, 57, 58).
These results establish a theoretical and experimental basis of our original hypothesis about the targeted tumor cell-binding capacity of the overexpressed Ig V regions of TIL-B cells. A direct selection of overrepresented clusters based on Ig repertoire analysis for building a scFv minilibrary can simplify and shorten the process for searching for tumor target and tumor-reactive binder Abs. Our results provide a potential new strategy for identifying novel tumor Ags and can serve as a proof of principle for demonstrating the potential tumor Ag-binding capacity of B lymphocytes infiltrating human breast carcinomas and other solid tumors. Furthermore, our results suggest that it may be worthwhile to investigate other types of solid tumors with heavy or moderate lymphocytic infiltration.
In addition, the identified tumor Ags might be used in the development of new tumor diagnostic and immunotherapeutic drug discovery programs (59, 60, 61). TIL-B lymphocytes could provide a new source for the production of novel Ab fragments with improved properties for applications such as tumor targeting in vivo. In well-differentiated ductal carcinoma in situ and invasive ductal carcinomas (grade 1), numerous cells exhibited cell surface labeling, which was reported to be absent in moderately and poorly differentiated tumors of the same types. The fact that in malignant lesions an abnormal distribution pattern of O-acetylated disialogangliosides (GD3) might be defined in comparison with a benign proliferation renders the ganglioside a putative prognostic and diagnostic marker (62). Our GD3-specific Ab fragment labels the more aggressive MDA-MB 231 cells that lack the estrogen receptors and are not responsive to estrogen and anti-estrogens (tamoxifen and benzothiophene). Ganglioside expression varies significantly among different cell lines, and both GM3 and GD3 are suggested to be involved in regulation of growth factor functions and tumor cell proliferation (51). Because gangliosides are overexpressed and cancer-restricted molecules, they are promising targets in the field of cancer research (52, 56, 59, 60, 63).
With the advent of rDNA technology, it has become feasible to manipulate and shuffle Ab genes (64, 65, 66) and to create recombinant Abs (and their conjugates) of desired specificity, effector functions, and reduced immunogenicity in humans (67, 68, 69, 70).
Our findings prove for the first time that tumor-specific Ig V region gene fragments obtained directly of B cells infiltrating breast cancer reveal key tumor-related Ags, such as the GD3 ganglioside. A careful analysis of the VH and VL Ig sequences of TIL-B through the construction of Ab fragment libraries appears to be a novel approach to obtain specific tumor-binder Ab fragments and may reveal new target Ags in a direct way from breast carcinomas. From these fragments, the Ig V genes of cell infiltrates from any solid tumors may be engineered into fully human Ab constructs for tumor diagnostic and therapeutic use.
The authors have no financial conflict of interest.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
This project was supported by international (NATO CLG 978639) and Hungarian grants (Eotvos XLI/34/2000), Orszagos Tudomanyos es Kutatasi Alap (National Scientific and Research Fund) (T030380), and recently by the Rajko Medenica Research Foundation. Some earlier basic works were performed through scholarships granted to B.K. from International Agency for Research on Cancer, Societe du Cancer, France, bourses of French Government, and Centre International des Etudes Scientifiques, and benefited from Institut Curie, Association pour la Recherche sur le Cancer, Association pour la Recherche sur le Cancer program (C01-010), and Institut National de la Santé et de la Recherche Médicale.
Abbreviations used in this paper: TIL-B, tumor-infiltrating B; AP, alkaline phosphatase; dAb, Ab derivate; IPTG, isopropyl β-d-thiogalactoside; MBC, medullary breast carcinoma; R, replacement; S, silent; scFv, single-chain variable fragment.