Cancer Immunology, Immunotherapy

, Volume 62, Issue 6, pp 1107–1122

Monoclonal antibodies toward different Tn-amino acid backbones display distinct recognition patterns on human cancer cells. Implications for effective immuno-targeting of cancer


  • Daniel Mazal
    • Departamento de Anatomía Patológica y Citología del Hospital de la MujerCentro Hospitalario Pereira Rossell
  • Richard Lo-Man
    • Unité de Régulation Immunitaire et VaccinologieInstitut Pasteur
    • Institut National de la Santé et de la Recherche Médicale
  • Sylvie Bay
    • Unité de Chimie des BiomoléculesInstitut Pasteur
    • Centre National de la Recherche Scientifique UMR 3523
  • Otto Pritsch
    • Departamento de Inmunobiologia, Facultad de MedicinaUniversidad de la República
    • Unidad de Biofísica de ProteínasInstitut Pasteur de Montevideo
  • Edith Dériaud
    • Unité de Régulation Immunitaire et VaccinologieInstitut Pasteur
    • Institut National de la Santé et de la Recherche Médicale
  • Christelle Ganneau
    • Unité de Chimie des BiomoléculesInstitut Pasteur
    • Centre National de la Recherche Scientifique UMR 3523
  • Andrea Medeiros
    • Departamento de Bioquímica, Facultad de MedicinaUniversidad de la República
  • Luis Ubillos
    • Departamento de Inmunobiologia, Facultad de MedicinaUniversidad de la República
  • Gonzalo Obal
    • Departamento de Inmunobiologia, Facultad de MedicinaUniversidad de la República
    • Unidad de Biofísica de ProteínasInstitut Pasteur de Montevideo
  • Nora Berois
    • Laboratorio de Glicobiología e Inmunología TumoralInstitut Pasteur de Montevideo
  • Mariela Bollati-Fogolin
    • Unidad de Biología CelularInstitut Pasteur de Montevideo
  • Claude Leclerc
    • Unité de Régulation Immunitaire et VaccinologieInstitut Pasteur
    • Institut National de la Santé et de la Recherche Médicale
    • Departamento de Inmunobiologia, Facultad de MedicinaUniversidad de la República
    • Laboratorio de Glicobiología e Inmunología TumoralInstitut Pasteur de Montevideo
Original Article

DOI: 10.1007/s00262-013-1425-7

Cite this article as:
Mazal, D., Lo-Man, R., Bay, S. et al. Cancer Immunol Immunother (2013) 62: 1107. doi:10.1007/s00262-013-1425-7


The Tn antigen (GalNAcα-O-Ser/Thr) is a well-established tumor-associated marker which represents a good target for the design of anti-tumor vaccines. Several studies have established that the binding of some anti-Tn antibodies could be affected by the density of Tn determinant or/and by the amino acid residues neighboring O-glycosylation sites. In the present study, using synthetic Tn-based vaccines, we have generated a panel of anti-Tn monoclonal antibodies. Analysis of their binding to various synthetic glycopeptides, modifying the amino acid carrier of the GalNAc(*) (Ser* vs Thr*), showed subtle differences in their fine specificities. We found that the recognition of these glycopeptides by some of these MAbs was strongly affected by the Tn backbone, such as a S*S*S* specific MAb (15G9) which failed to recognize a S*T*T* or a T*T*T* structure. Different binding patterns of these antibodies were also observed in FACS and Western blot analysis using three human cancer cell lines (MCF-7, LS174T and Jurkat). Importantly, an immunohistochemical analysis of human tumors (72 breast cancer and 44 colon cancer) showed the existence of different recognition profiles among the five antibodies evaluated, demonstrating that the aglyconic part of the Tn structure (Ser vs Thr) plays a key role in the anti-Tn specificity for breast and colon cancer detection. This new structural feature of the Tn antigen could be of important clinical value, notably due to the increasing interest of this antigen in anticancer vaccine design as well as for the development of anti-Tn antibodies for in vivo diagnostic and therapeutic strategies.


Tn antigenO-glycosylationCancerImmunotherapyVaccineAntibody


Glycosylation, a complex form of post-translational modification affecting over 50 % of all proteins, constitutes a key regulator of many eukaryotic processes [1]. Many markers of pre-malignant and malignant cells correspond to the carbohydrate and the peptidic moieties of mucins [2, 3]. These structures greatly contribute to the phenotype and the biological behavior of cancer cells. It has been shown that cell-surface carbohydrates are involved in tumor metastasis, and the role of O-glycans depends on their structure [4]. In addition, O-glycosylation-dependent mechanisms are also implicated in tumor immune escape [5]. Mucin-type O-glycans are built up by a sequential, step-by-step, process in the Golgi apparatus, starting with the addition of a N-acetyl-galactosamine (GalNAc) to a serine or a threonine residue. This results in the core GalNAcα-O-Ser/Thr, also known as the Tn antigen. This initial key step of the O-glycosylation process is catalyzed by the UDP-GalNAc:polypeptide N-acetylgalactosaminyltransferases (GalNAc-T) (EC [6]. GalNAcα-O-Ser/Thr is then further elongated by other glycosyltransferases to generate complex O-glycans. For example, the Tn structure is the acceptor substrate for core 1 β3-galactosyltransferase (β3Gal-T) to generate the core 1 disaccharide O-glycan Galβ1-3GalNAcα-O-Ser/Thr, also known as Thomson–Friedenrich antigen. Therefore, in normal tissue O-glycoproteins, subsequent addition of sugar residues hides the Tn determinant, while in tumor mucins, incomplete glycosylation results in the exposure of this core antigen [7]. The mechanisms underlying the Tn antigen expression are still poorly understood, although elucidated in a few cases. It is well known that a low core 1 β3-galactosyltransferase activity correlates with Tn antigen expression. This enzyme requires for its activity the expression of a molecular chaperone, designated Cosmc [8]. Mutations in Cosmc are the molecular basis of truncated O-glycosylation in blood cells in the autoimmune Tn syndrome [9], as well as in the Tn expression in some cancer cells [10, 11]. Additionally, in breast cancer, we have found that the “non-mammary gland” apomucin MUC6 is a very good acceptor substrate for GalNAc-Ts and that the MUC6-Tn glycopeptide is a poor acceptor substrate for the core 1 β3Gal-T [12]. These results indicate that “non-mammary gland” apomucin expression could be, at least in part, responsible for Tn antigen expression in breast cancer cells.

Over the past years, the Tn antigen has attracted much interest in cancer biology and clinical oncology for several reasons. First, it may be a useful diagnostic marker since it is rarely found in normal tissues, whereas it is widely expressed in a variety of adenocarcinomas [7, 13] and in some malignant hematopoietic cells [14]. Second, it was reported that Tn could be an early biomarker of cancer, both in humans [15] and in animal models [16]. Third, Tn has been implicated in the metastasis of tumor cells, and a direct correlation has been shown between carcinoma aggressiveness and the density of this antigen, for example, extent of tissue spread and vessel invasion [17]. Fourth, the Tn antigen was shown to induce efficient therapeutic anti-tumor responses in an experimental model [18]. Consequently, this antigen represents an important component for the design of therapeutic glycoconjugate anti-tumor vaccines, for selective eradication of tumors [19, 20].

The Tn antigen was characterized using both specific monoclonal antibodies (MAb) and lectins, but these approaches lead to highly variable recognition of tumor cells [21]. For example, three anti-Tn antibodies, raised against partially deglycosylated intestinal mucin peptides, have shown different tissue reactivity [22]. In addition, using the isolectin B4 from Vicia villosa seeds (VVLB4), the Tn antigen was shown to be expressed in normal pancreatic acinar cells [23, 24], whereas in contrast, two anti-Tn monoclonal antibodies (TEC-02 and 12A8-C7-F5) failed to recognize these normal tissues [25]. In a histochemical study performed on 322 cases of breast ductal carcinomas, VVLB4-binding was shown to correlate with tumor stage, lymphatic invasion, and lymph node metastasis, whereas the reactivity of the anti-Tn monoclonal antibody HB-Tn1 was not clearly related to breast cancer aggressiveness [26].

Although the chemical structure of the Tn determinant is known to be GalNAcα-O-Ser/Thr, its immunological definition is more complex and some Tn-binding proteins clearly require more complex epitopes than a single Tn residue. Indeed, the characterization of the fine specificity of some anti-Tn antibodies showed that they require the involvement of additional amino acids in the antigenic determinant [27]. Moreover, although plant lectins (VVLB4 and Salvia sclarea) appear to recognize single Tn epitopes [28, 29], some anti-Tn monoclonal antibodies require at least two consecutive Tn residues for binding [29, 30]. Indeed, when an unglycosylated amino acid is introduced between two Tn residues, both MLS128 and 83D4 MAb lose their reactivity [29]. The MAb PMH1 can react with single or multiple Tn on a specific MUC2 apomucin peptidic chain but recognize neither the unglycosylated peptide nor the asialo-ovine submaxillary mucin (aOSM) [31]. These results show that, in addition to the role played by the Tn density, the O-glycosylated amino acid residues (Ser vs Thr) as well as the neighboring peptide backbone could be important factors that modulate the structure of the Tn epitope.

The fact that some anti-Tn monoclonal antibodies binding require several consecutive Tn residues is consistent with the fact that this antigen is known to be displayed as clusters on native structures. This provided the basis for the rational design of glycopeptide-based anticancer vaccines containing glycotopes organized in clusters. Kuduk et al. [32] observed that clusters of three consecutive Tn residues covalently linked to the carrier protein KLH were able to induce in mice high IgM and IgG anti-Tn antibody titers, which were strongly reactive with the Tn-positive human colon cancer cell line LSC, but not with the Tn negative colon cancer cell line LSB. We have developed a fully synthetic Tn immunogen that does not require a protein carrier, called multiple antigenic glycopeptide (MAG), based on a dendrimeric lysine core with four arms containing a CD4+ T cell epitope linked to Tn residues [33]. A therapeutic immunization performed with this immunogen induced a strong IgG anti-Tn antibody response, associated with an increase survival of tumor-bearing mice [34]. Moreover, the MAG carrying the tri-Tn glycotope was much more efficient than the mono-Tn analogue to promote the survival of mice grafted with the mammary adenocarcinoma TA3/Ha cell line in immunotherapeutic experimental settings [35].

Taking into account the heterogeneous immune response that may be obtained using different sources of Tn antigen, it is very important to analyze the variation of the Tn antigen expression pattern in clinical samples, in order to improve the design of Tn-based vaccines. In the present study, we performed an immunohistochemical analysis of the Tn antigen expression in breast and colon cancers, using a panel of well-characterized anti-Tn monoclonal antibodies with subtle differences in their fine specificities. We generated this panel of antibodies using MAG carrying different types of tri-Tn glycotopes, and we compared their reactivity with the anti-Tn MAb 83D4 [29], produced from a mouse immunized with a human breast cancer extract [36]. We demonstrated, for the first time, that the amino acid carrier of the GalNAc (Ser vs Thr) plays a key role in the immunohistochemical profile of Tn antigen expression on breast and colon cancer cells. These results are of critical importance for the design of new immunotherapy based on the Tn antigen.


Glycopeptide synthesis

The syntheses of the glycopeptides (tetravalent MAG and linear structures) and of the control peptide were performed as previously described by solid phase peptide methodology using Fmoc chemistry [18, 37]. The MAG has a C-ter tetravalent lysine core displaying four N-ter tri-Tn clusters linked to the peptide KLFAVWKITYKDT (poliovirus 103–115 or PV). The tri-Tn structure is either S*T*T* or S*S*S* [* = α-D-GalNAc] leading to, respectively, MAG:Tn3-PV [MAG(S*T*T*)] and MAG:Tn(S)3-PV [MAG(S*S*S*)] [18]. The linear glycopeptides are composed of an N-ter tri- or di-Tn cluster (S*T*T*, S*S*S*, T*T*T*, or S*T*T or ST*T*) linked to a C-ter eight amino acids sequence (Gly)6LysGly (G6 KG). The corresponding α-mannosylated MAG and unglycosylated peptides were also synthesized with the STT backbone. When present, the biotin is linked to the ε-amino group of lysine [18, 37]. All the derivatives were purified by reverse phase high-performance liquid chromatography (RP-HPLC) using a Perkin–Elmer pump system with a UV detector and a Nucleosil C18 column (5 μ, 100 or 300 Å, 250 × 10 mm). They were analyzed by RP-HPLC (C18 column 150 × 4.6 mm), and the gradient was performed with water (0.1 %TFA)/acetonitrile over 20 min. The compounds were characterized by amino acid analysis (AAA) and electrospray mass spectrometry (ESI–MS) analyses. The analyses of Tn3-G6 K(biot)G [S*T*T*-G6 K(biot)G], STT-G6 K(biot)G, Tn3-G6 KG [S*T*T*-G6 KG], and STT-G6 KG have been previously reported [37].

MAG:Tn3-PV or MAG(S*T*T*)

HPLC: gradient 20–50 %, retention time 10.4 min; ESI–MS: 10447.92 (calcd 10448.63); AAA: Ala 3.87 (4), Asp 4.38 (4), Ile 4.10 (4), Leu 4.21 (4), Lys 16.3 (15), Phe 4.0 (4), Ser 3.60 (4), Thr 16.63 (16), Tyr 4.31 (4), Val 4.11 (4).

MAG:Tn(S)3-PV or MAG(S*S*S*)

HPLC: gradient 10–60 %, retention time 11.9 min; ESI–MS: 10336.05 (calcd 10336.57); AAA: Ala 3.82 (4), Asp 4.45 (4), Ile 4.12 (4), Leu 4.08 (4), Lys 16.1 (15), Phe 4.0 (4), Ser 11.02 (12), Thr 7.42 (8), Tyr 4.35 (4), Val 4.05 (4).


HPLC: gradient 10–40 %, retention time 9.2 min; ESI–MS: 9956.82 (calcd 9956.01); AAA: Ala 3.91 (4), Asp 4.48 (4), Ile 4.05 (4), Leu 4.13 (4), Lys 16.1 (15), Phe 4.0 (4), Ser 3.62 (4), Thr 15.13 (16), Tyr 4.29 (4), Val 4.13 (4).

S*S*S*-G6 K(biot)G

HPLC: gradient 5–25 %, retention time 8.1 min; ESI–MS: 1642.43 (calcd 1642.67); AAA: Gly 7.0 (7), Lys 1.03 (1), Ser 2.71 (3).

T*T*T*-G6 K(biot)G

HPLC: gradient 10–17 %, retention time 6.9 min; ESI–MS: 1684.98 (calcd 1684.73); AAA: Gly 7.0 (7), Lys 1.10 (1), Thr 2.80 (3).

Production of anti-Tn monoclonal antibodies

Animals’ experimentation was done in accordance with institutional guidelines. BALB/c mice were intraperitoneally immunized every 3 weeks with alum together with 10 μg of MAG(S*T*T*) or MAG(S*S*S*) based on a tri-Tn build on a STT and a SSS backbone, respectively [18]. Throughout the immunization protocol, mice were bled and their sera tested for the presence of anti-Tn antibodies by enzyme-linked immunosorbent assay. The mice splenocytes, collected after 4–5 immunizations, were fused with P3X63 myeloma cells (ATCC) using poly-ethylene glycol (Sigma). The generated hybridomas were screened for the production of Tn-specific MAbs by ELISA for the recognition of S*T*T*-G6 K(biot)G, T*T*T*-G6 K(biot)G, and S*S*S*-G6 K(biot)G and for the lack of reactivity against STT-G6 K(biot)G coated on streptavidin plates. The MAbs were purified from ascitic fluids or from supernatants using T-GelTM purification kit (Pierce) according to the manufacturer’s instructions. The antibody concentration was measured with Bio-Rad protein assay. The previously characterized MAb 83D4 (IgM), which binds specifically to the Tn antigen [29], was precipitated from ascitic fluids by dialysis against demineralized water at 4 °C, dissolved in a small volume of 0.5 M NaCl in PBS, and purified by gel-filtration chromatography on Sephacryl S-200.

Analysis of monoclonal antibody specificity by ELISA

MAbs were tested for Tn recognition by ELISA using non-biotinylated synthetic Tn cluster glycopeptides adsorbed on plates, the corresponding unglycosylated peptides, and aOSM (a standard natural source of Tn antigen). The glycopeptides consisted of (1) various linear peptide sequences with di- or tri-Tn build on a S/T backbone to define fine specificity and (2) tetravalent MAG bearing either α-D-GalNAc or α-D-Man on a STT backbone. Goat anti-mouse IgG or goat anti-human IgG peroxidase conjugate (Sigma) was used.

Analysis of monoclonal antibody specificity by surface plasmon resonance

Interactions between anti-Tn MAbs and glycopeptides were analyzed by performing surface plasmon resonance experiments on a BIAcore 3000 instrument (GE Healthcare, Sweden). Biotinylated-peptides were bound on sensor surfaces with immobilized streptavidin (SA sensorchip, GE Healthcare, Sweden). The first flow cell (Fc) was used as a reference surface containing 17 RU of the non-glycosylated peptide STT-G6 K(biot)G, the second Fc contained 27 RU of glycopeptide S*T*T*-G6 K(biot)G, the third 26 RU of glycopeptide T*T*T*-G6 K(biot)G, and the fourth 26 RU of glycopeptide S*S*S*-G6 K(biot)G. The MAbs were diluted in HBS-EP buffer (10 mM HEPES, 150 mM NaCl, 3 mM EDTA, and 0.005 % Surfactant P20, pH 7.4) and were passed over the sensorchip at different concentrations (2A10, 9A7 and 6E11 at 10, 25, 50, 100 and 500 nM; 15G9 at 0.67, 2.2, 6.7, 12, 25 nM; 83D4 at 0.25, 2.5, 6.25, 12.5, 25 and 125 nM). All experiments were run in duplicates at a 30 μL/min flow rate, a contact time of 180 s and a dissociation time of 360 s, with the biosensor instrument thermostated at 25 °C. After dissociation, the sensorchip was regenerated by injecting 10 mM glycine–HCl (pH 2.5) at the end of each experiment. All data processing was carried out using the BIAevaluation 4.1 software provided by BIAcore. Binding responses were first double referenced by subtracting signals corresponding to both reference flow cell and from the average of blank (buffer) injections. According to the software manual, our experimental data were fitted to the bivalent interaction model. Nonlinear fitting methods were used to simultaneously obtain association and dissociation parameters.

Analysis of monoclonal antibody recognition on cancer cell lines by flow cytometry

MAbs were tested by flow cytometry on Tn-expressing human Jurkat (ATCC), MCF-7 (ATCC), and LS174T (ATCC) cell lines. Binding of antibodies to Jurkat cells was revealed with goat anti-mouse IgG antibody conjugated to phycoerythrin, and PFA fixed cells were analyzed by FACS. MCF-7 and LS174T cells were fixed, permeabilized (4 % PFA, 0.1 % Tween 20), and incubated with the different monoclonal antibodies. The specific binding was developed with an anti-mouse antibody conjugated with FITC (Dako, Carpinteria, CA) and further analyzed on a CyAn™ ADP Flow Cytometer (Dako). All data obtained were analyzed using Summit v4.3 software.

Patients and immunohistochemical analysis

The immunohistochemical evaluation of each anti-Tn monoclonal antibody was performed on tissues from 116 randomly chosen patients between 2002 and 2004: 72 patients who had undergone surgical treatment for a breast lump at Pereira Rossell Hospital (Montevideo, Uruguay) with histopathological diagnosis of breast cancer, and 44 patients who had undergone surgical treatment for colon cancer at Maciel Hospital (Montevideo, Uruguay). The study was examined and approved by the ethical review board of the School of Medicine, Montevideo, Uruguay. Normal breast tissues were obtained from reduction mammoplasties (n = 5), and normal colon tissues were obtained from the proximal resection border of the surgical sample (n = 10). All specimens were formalin-fixed (4 %) paraffin embedded tissues, and histological sections were subjected to hematoxylin and eosin staining (standard procedure) as well as immunohistochemical evaluation for Tn antigen expression. All breast malignant tumors were conventionally classified by histological type and grading (Scarff-Bloom-Richardson system with Nottingham modification), and all colon malignant tumors were classified by the Duke’s procedure. In order to perform a more accurately comparison in the immunohistochemical pattern among the 5 anti-Tn MAbs, the reactivity of each was analyzed by using tissue microarrays (TMAs). H&E stained slides of all specimens were analyzed before construction of the TMA to identify representative tumor areas. TMAs were constructed as described previously [38], using 2 tissue cores (diameter 4 mm) each containing one sample from a different region of the tumor. Clinico-pathologic parameters of patients included in the TMA are summarized in Table 1. For antibody immunostaining, the quenching of endogenous peroxidase activity was performed with 3 % H2O2 in PBS for 20 min and blocked with normal goat serum for 20 min (to decrease background staining). Anti-Tn MAbs were incubated overnight at 4 °C: 83D4 (10 μg/ml), 2A10 (20 μg/ml), 6E11 (20 μg/ml), 9A7 (20 μg/ml), and 15G9 (10 μg/ml). Sections were thereafter treated with peroxidase-conjugated goat anti-mouse polyvalent antibody (Sigma) (60 min at room temperature). Reactions were revealed with 3,3′-diaminobenzidine (Sigma), washed in water, counterstained in Mayer`s hematoxylin, dehydrated in ethanol and xylene, and mounted. Between each step, sections were washed in PBS. For every assay, negative controls using PBS without primary antibody were included. To release the Tn antigen, samples were incubated in the presence of 2 U/ml α-N-acetylgalactosaminidase from chicken liver (Prozyme, Hayward, CA) in 100 mM sodium citrate/phosphate buffer, pH 4, for 20 h at 37 °C in a humidified chamber. Control samples were incubated similarly in the absence of the enzyme. Slides were then washed three times in PBS and stained with the anti-Tn antibodies as described above. An immunohistochemical score was used which includes an assessment of both the intensity of staining and the percentage of stained cells. For the intensity, a score index of 0, 1, 2, and 3 corresponding to negative, weak, moderate, and strong staining intensity was used and the percentage of positive cells at each intensity was estimated subjectively. A final score of 0–300 is the product of both the intensity and the percentage. Each tumor core was scored individually, and then the mean of the 2 readings was calculated. If one core was uninformative, either lost or contained no tumor tissues, the overall score applied was that of the remaining core [39]. Scores were established jointly by two observers (DM and EO). Clinico-pathological information was masked to the observers.
Table 1

Characteristics of patients and tumors

Breast cancer

 Age in years, median (range)

59.1 (26–82)

 UICC stage


20 (27.7 %)


21 (29.1 %)


27 (37.5 %)


4 (5.5 %)

 Histological type


66 (91,66 %)


4 (5.4 %)


1 (1.35 %)


1 (1.35 %)

 Histological grade

  Grade 1

16 (22.2 %)

  Grade 2

32 (44.4 %)

  Grade 3

24 (33.3 %)

Colon cancer

 Age in years, median (range)

65 (29–90)

 UICC stage


4 (9.1 %)


12 (27.3 %)


21 (47.7 %)


7 (15.9 %)

 Histological grade


10 (22.7 %)


26 (59.1 %)


8 (18.2 %)


Analysis of MAb fine specificity using synthetic glycopeptides

We previously demonstrated that a synthetic Tn immunogen displaying a tri-Tn cluster based on a STT amino acid backbone [namely MAG(S*T*T*)] was particularly efficient in inducing an anti-Tn-specific antibody response, leading to the therapeutic control of the TA3/Ha murine breast carcinoma [35]. To investigate the diversity of the Tn-specific antibody response induced by this synthetic Tn glycopeptide, we first generated MAbs from MAG(S*T*T*) immunized mice. These MAbs were then screened for their capacity to recognize the Tn antigen as a cluster of three Tn corresponding to α-D-GalNAc (*) residues displayed on a STT, TTT, or SSS backbone at the N-ter of an irrelevant poly-Gly peptide. As all MAbs generated in these conditions were positive for the recognition of Tn, whatever the amino acid carrier backbone, we selected arbitrarily eight of these MAbs for further analysis (fine specificities are detailed for 4 out of eight MAbs in Table 2). We also generated MAbs from MAG(S*S*S*) immunized mice and isolated a S*S*S* specific MAb (15G9). This Tn-specific MAb was highly sensitive to the Tn backbone, as it failed to recognize a S*T*T* or a T*T*T* structure (Table 2). Intriguingly, none of the MAb recognized aOSM, produced by chemical desialylation of OSM, and generally used as a natural form of Tn. In addition, all MAbs produced using the synthetic MAG immunogens failed to recognize blood group A erythrocytes (data not shown), confirming the specificity of these antibodies for Tn antigen. To confirm that the reactivity is truly through the Tn clusters, we compared the antibody binding of the MAG:Tn3-PV versus both the α-mannosylated analogue and the unglycosylated peptide counterpart. As showed in the Supplemental Figure 1, the MAbs were only reactive on the α-GalNAc-bearing glycopeptide.
Table 2

Detailed characteristics of anti-Tn MAbs














Tumor cell





GalNAc carriera








Tumor cell recognitionc





aDetermination by ELISA using synthetic glycopeptides

bRecognition of aOSM by MAbs was tested by ELISA

cRecognition of Tn-positive Jurkat cells by flow cytometry

The MAbs binding specificities and affinities were further analyzed by surface plasmon resonance, using a BIAcore system. The biotinylated glycopeptides were immobilized onto flow cell surfaces of streptavidin-coupled SA sensorchips, and binding of the MAbs was measured by running them across this surface. As shown in Fig. 1, MAb 83D4 recognized all three glycopeptides S*T*T* (KD = 1.0 × 10−9 M), T*T*T* (KD = 9.9 × 10−9 M), and S*S*S* (KD = 2.0 × 10−9 M) (Fig. 1a), whereas MAb 2A10 displayed the best specificity for S*T*T* (KD = 1.1 × 10−7 M) and S*S*S* (KD = 2.6 × 10−7 M), but no binding was detected for T*T*T* (Fig. 1b). All three glycopeptides S*T*T*, T*T*T*, and S*S*S* were also recognized by MAbs 6E11 (KD = 1.3 × 10−7, 2.8 × 10−7, and 6.8 × 10−8 M, respectively; Fig. 1c) and 9A7 (KD = 6.4 × 10−8, 1.4 × 10−7, and 1.5 × 10−7 M, respectively; Fig. 1d), whereas MAb 15G9 only recognizes S*S*S* (KD = 1.1 × 10−8 M; Fig. 1e). Overall, these results were in agreement with those obtained by ELISA. Collectively, these data show that the Tn-based MAG synthetic immunogen generates a variety of fine and accurate specificities for the Tn antigen.
Fig. 1

Analysis of MAb specificities by surface plasmon resonance measurements. Binding analysis of different anti-Tn MAbs: a 83D4 (25 nM), b 2A10 (100 nM), c 6E11 (100 nM), d 9A7 (100 nM), and e 15G9 (12 nM) was performed on biosensor surfaces with immobilized biotinylated glycopeptides. These glycopeptides display different tri-Tn clusters (S*T*T*, T*T*T*, or S*S*S*) linked to an irrelevant poly-glycine peptide different from the immunogen PV peptide (see “Methods”). The biotinylated unglycosylated peptide (STT) was used as the negative control. RU resonance units
Fig. 2

Anti-Tn MAbs differentially recognize native forms of Tn on tumor cell lines. Anti-Tn MAbs were tested for the binding to Tn-expressing Jurkat, MCF-7, and LS74-T cells, by flow cytometry, using PE conjugated anti-mouse IgG antibodies, except for 83D4 MAbs for which anti-mouse IgM was used. Filled gray histograms correspond to the secondary reagent alone and black lines to indicated MAb

Analysis by FACS of MAb recognition on different cancer cell lines

We next evaluated the capacity of these MAbs to recognize the Tn antigen on human Jurkat cells. All MAbs, but one (15G9), can bind to Jurkat cells (Fig. 2a; Table 2). Taking into account that the Tn antigen is a useful marker for breast and colon tumors and that it is being evaluated as an immunotherapeutic target in patients, we then compared the immunoreactivity of these anti-Tn MAbs on these cancer cell lines. Percent of positive cells for each anti-Tn MAbs was evaluated by flow cytometry on MCF-7 breast and LS174T colon cancer cell lines as shown on Table 3 and Fig. 2b, c. The highest percentage of cell staining was obtained with the 83D4 and 6E11 MAbs, followed by 9A7 and 2A10, while very few cells were stained by the 15G9 MAb.
Table 3

Recognition of breast (MCF-7) and colon (LS174T) cancer cell lines by anti-Tn monoclonal antibodies

Monoclonal antibody


 % Positive cells


 % Positive cells
















Immunohistochemical evaluation on breast and colon tumors

Once established the optimal conditions for immunostaining using each MAb, the tissues of 45 patients were examined in whole tissue sections (15 primary breast tumors, 15 primary colon tumors, 5 specimens from normal breast tissues, and 10 specimens from normal colon tissues). All MAbs showed constantly a predominating cytoplasmatic staining pattern of the tissues from cancer patients but did not stained normal breast and colon samples (Figs. 3a, 4). In this preliminary evaluation, few tumors were positive for MAb 15G9. Moreover, we observed that some tumor samples displayed different patterns of reactivity with these antibodies (Fig. 3b). Binding of mAbs was abolished when tissue samples were treated with α-N-acetylgalactosaminidase, which cleaves GalNAc from the protein backbone, confirming the specificity of the interaction (Supplemental figure 2).
Fig. 3

Immunohistochemical evaluation of anti-Tn MAbs in breast tumors. a Anti-Tn MAbs 83D4, 2A10, 6E11, 9A7, and 15G9 were probed against a primary breast tumor (magnification ×40). The immunohistochemical staining (score 0–300) was assessed as the product of the intensity (0, 1, 2, or 3) and the percentage of positive cells. Staining intensity and immunohistochemical score assigned to each sample were 83D4 (3/300), 2A10 (2/160), 6E11 (1/30), 9A7 (1/70), and 15G9 (0/0). b Representative picture of positive (6E11) or negative (9A7) staining of a primary breast tumor. Magnification ×100. Staining intensity and immunohistochemical scores were 6E11 (3/300) and 9A7 (0/0)
Fig. 4

Immunohistochemical evaluation of anti-Tn MAbs in colon tumors. Anti-Tn MAbs 83D4, 2A10, 6E11, 9A7, and 15G9 were probed against a primary colon tumor or normal tissue. Magnification ×400. Staining intensity and immunohistochemical score assigned to each sample were 83D4 (3/294), 2A10 (2/196), 6E11 (2/190), 9A7 (1/95), and 15G9 (1/90)

To compare more accurately the immunostaining pattern of the anti-Tn MAbs in clinical samples, we prepared tissue microarrays using 72 breast and 44 colon tumors. The Table 4 shows the percent of positive tumors for each antibody. In breast cancer samples (Fig. 5), the highest sensitivity was observed for 6E11 and 83D4 (98 and 95.8 %, respectively), followed by 2A10 (88.8 %) and 9A7 (83.3 %). Interestingly, the MAb 15G9, specific for a cluster of S*S*S*, recognized only 9.7 % of tumors. Remarkably, the simultaneous analysis using 6E11 and 83D4 was able to detect all breast tumors (data not shown). In colon cancer samples, the highest sensitivity was observed for MAb 83D4 (81.8 %), followed by 6E11 (61.3 %), 2A10 (50 %), 9A7 (47.7 %), and 15G9 (13.6 %). Simultaneous analysis using 83D4 or 6E11 associated with 2A10 or 9A7 did not improve this percentage for colon cancer detection. In breast cancer, we observed a similar immunostaining pattern among all antibodies, whereas in colon cancer, we found differences in tissue staining between MAb 83D4 compared with the other four antibodies. MAb 83D4 was the only one showing reactivity with the secreted material (Fig. 6). The staining by anti-Tn MAbs (Table 4) expressed as the immunohistochemical score (intensity of staining + the percentage of stained cells) did not significantly correlated with any of the clinico-pathological parameters examined (stage and tumor differentiation). In a few tumors, MAb 15G9 had a score 2 or 3 for staining intensity. As a whole, the results of the immunohistochemical assessment of 116 human tumors showed the existence of different recognition profiles among the five antibodies evaluated, demonstrating that the aglyconic part of the Tn structure (Ser vs Thr) plays a key role in the anti-Tn specificity for breast and colon cancer detection.
Table 4

Anti-Tn antibodies reactivity in colon and breast cancers related to clinico-pathological features of patients



n (%)



n (%)



n (%)



n (%)



n (%)


Colon cancer (n = 44)

Tumor stage

I (n = 4)

4 (100)


3 (75)


4 (100)


2 (50)


1 (25)


II (n = 12)

10 (83)

6 (50)

7 (58)

7 (58)

3 (25)

III (n = 21)

16 (76)

10 (48)

13 (62)

8 (38)

2 (10)

IV (n = 7)

6 (86)

3 (43)

3 (43)

4 (57)


Histological differentiation

Well (n = 10)

8 (80)


4 (40)


6 (60)


3 (30)


1 (10)


Moderate (n = 26)

21 (81)

13 (50)

14 (54)

14 (54)

3 (11)

Poor (n = 8)

7 (87)

5 (62)

7 (87)

4 (50)

2 (25)


36 (82)

22 (50)

27 (61)

21 (48)

6 (14)

Breast cancer (n = 72)

Tumor stage

I (n = 20)

18 (90)


17 (85)


19 (95)


17 (85)


2 (10)


IIa (n = 21)

20 (95)

19 (90)

21 (100)

18 (86)

3 (14)

IIb (n = 27)

27 (100)

25 (93)

27 (100)

23 (85)

2 (7)

III (n = 4)

4 (100)

3 (75)

4 (100)

2 (50)


Histological grade

I (n = 16)

16 (100)


14 (87)


16 (100)


13 (81)


3 (18)


II (n = 32)

31 (97)

28 (87)

32 (100)

27 (84)

2 (6)

III (n = 24)

22 (92)

22 (92)

23 (96)

20 (83)

2 (8)


69 (96)

64 (89)

71 (98)

60 (83)

7 (10)
Fig. 5

Human tissue microarray analysis of Tn antigen expression in breast tumors. A representative case stained by MAbs 83D4, 2A10, 6E11, and 9A7 and negative for MAb 15G9. Magnification ×40. Staining intensity and immunohistochemical score assigned to each sample were 83D4 (3/300), 2A10 (2/180), 6E11 (2/200), 9A7 (1/40), and 15G9 (0/0)
Fig. 6

Human tissue microarray analysis of Tn antigen expression in colon tumors. A representative case showing staining both in cancer cell and in the secretory material (MAb 83D4), staining only in cancer cells (MAbs 2A10, 6E11, and 9A7), and absence of staining (MAb 15G9). Magnification ×100. Staining intensity and immunohistochemical score assigned to each sample were 83D4 (3/300), 2A10 (2/120), 6E11 (2/140), 9A7 (1/30), and 15G9 (0/0)


Synthetic vaccines based on Tn epitopes are currently being evaluated as potential immunotherapeutics in the treatment of cancer. However, there exists conflicting results regarding the immunohistochemical expression of this antigen in human tumors. This controversy could be linked to the different specificities of the antibodies used in these studies [21, 22, 40]. This is a key issue considering that the expression of the Tn antigen by a tumor is a fundamental requirement for the efficiency of Tn-based vaccines. Taking into account that most of the synthetic Tn vaccines are based on GalNAcα-O-Ser/Thr residues organized in clusters (mainly tri-Tn), it is crucial to determine whether variants of Tn clusters could exist in clinical samples, in order to improve vaccine design as well as to select the patients eligible for such immunotherapy. In the present work, using synthetic MAG immunogens, we have generated a panel of 8 anti-Tn monoclonal antibodies exhibiting specific binding against Tn clusters bearing different glyco-amino acid motifs. Three out of 8 MAbs required at least a minimal dimeric Tn structure (i.e., S*T* or T*T*), although they preferentially bound to tri-Tn as a S*T*T* backbone. Among the 5 other MAbs, 3 also required 2 Tn residues (data not shown). However, di-Tn-specific MAbs were able to recognize efficiently consecutive di-Tn (XS*T*X) but not a “non-consecutive” Tn dimer (XS*XT*X). The 2 remaining MAbs required 3 consecutive Tn as a minimal structure. One of these antibodies displayed a very restricted specificity, such as the MAb 15G9 which is specific for tri-Tn linked to GalNAc-Ser residues. This antibody is highly sensitive to the peptide backbone and failed to recognize S*T*T* or T*T*T* structures. It is the first antibody characterized to be specific for trimeric GalNAcα-O-Ser (S*S*S*). The MAb B230.9 was previously generated following immunization of mice with Tn-serine trimers conjugated to KLH [41]. This antibody displays preferential reactivity toward Tn-serine dimeric and trimeric structures. However, it was not reported if the aglyconic part of the Tn antigen (Ser vs Thr) could play a role in MAb B230.9 binding. It has been demonstrated that GalNAcα-O-Ser and GalNAcα-O-Thr adopt completely different three-dimensional orientations [42]. The GalNAc moiety in Thr is almost perpendicular to the peptide backbone, whereas in Ser residues, the sugar moiety adopts a parallel disposition. Our present results are in good agreement with the findings that the peptide sequence used for clustering Tn antigens plays a significant role on both their flexibility and their global shape and that the sequence of Ser or Thr residues has a crucial influence on the orientation of the carbohydrate moiety [43]. These differences most probably play a significant role in the biological functions of O-glycoproteins. For example, it was demonstrated that when a Ser residue replaces a Thr residue in a O-glycosylated sequence of an anti-freeze protein, the resulting structure loses its anti-freeze activity [44]. These distinct spatial arrangements of GalNAcα-O-Ser and GalNAcα-O-Thr residues may have important implications in the presentation of the epitope, as well for the immunogenicity of Tn bearing glycopeptides.

Using binding analysis by surface plasmon resonance, we also demonstrated that the MAb 2A10 displayed the highest specificity for S*T*T* and S*S*S*, but not for T*T*T*, and that MAbs 83D4, 6E11, and 9A7 recognize all three glycopeptides. Taken together, these results confirm that the five MAbs have selective specificities for GalNAc O-linked amino acidic sequences organized in a cluster of Tn residues. Our results also confirm the role played by the serine and threonine residues in the conformation of Tn antigen identified by MAbs 15G9 and 2A10 (both IgG1). These observations are in agreement with the recent findings reported by Blix et al. [45], who generated seven different anti-Tn MAbs raised against Jurkat T cells. The authors found that IgM MAbs bound the terminal GalNAc residue of the Tn antigen irrespective of the peptide context or with low selectivity to the glycoproteins, while IgG MAbs recognized the Tn antigen in the context of a specific peptide motif. It is important to remark that the different binding patterns displayed by our antibodies toward synthetic Tn glycopeptides were also evidenced in FACS (Fig. 2) and Western blot (data not shown) analysis using cancer cell lines, confirming that the synthetic immunogens were able to induce antibodies capable of recognizing native cell glycoproteins. Several data suggest that although the structural and sequence requirements for O-glycosylation of serine and threonine residues are similar, serine sites are glycosylated less effectively than threonine in vitro [46]. Differences at the molecular level for Thr versus Ser recognition have also been shown for polypeptide N-α-acetylgalactosaminyltransferases (GalNAc-Ts), the key enzymes which initiate O-glycan biosynthesis by the transfer of GalNAc from UDP-GalNAc to the hydroxyl groups of serine and threonine residues in a polypeptide. The higher in vitro catalytic efficiency toward threonine versus serine is the result of enhanced binding as well as increased reaction velocity, both effects being the result of steric interactions between the active site of the enzyme and the methyl group of threonine [47].

One of the most interesting observations of this study concerns the different recognition of breast and colon cancer tissues by our panel of Tn-specific MAbs. Two antibodies (6E11 and 83D4) showed a similar profile, especially in breast cancer, and both detected the majority of tumors. These two antibodies also showed a good specificity for three GalNAc on a STT backbone, which is precisely the epitope that we included in the MAG(S*T*T*) immunogen. This vaccine has proven to be very efficient in eradicating Tn-expressing tumors in mice [35] and is presently under preclinical development. Several authors found that cancer patients have antibodies to Tn glycopeptides and that variations in their serum levels correlate with the progression of cancer [48, 49]. Although the fine specificity of these antibodies has not been yet determined, differences in anti-Tn serum antibodies between individuals have been recently found using a carbohydrate antigen arrays [50]. It could be very interesting to determine whether the analysis of serum anti-Tn antibodies specific for a unique type of Tn antigen cluster (such as S*T*T*, T*T*T*, S*S*S*) could improve the diagnosis in patient follow-up.

The observation that MAb 15G9 was unreactive with the Jurkat lymphoma cell line (rich in Tn residues) and that this antibody displayed reactivity with only few tumors suggests a very restricted expression of the Tn antigen as trimeric GalNAcα-O-Ser (S*S*S*) clusters. That could be explained by the very low frequency of tri-serine amino acidic motifs in the proteome and the requirement of three consecutive GalNAc-serine motifs for 15G9 binding. Regarding mucin gene expression in breast cancer, MUC1 is frequently overexpressed and underglycosylated, bearing Tn residues identified by anti-Tn antibodies [7]. However, in MUC1, there is no sequence containing three consecutive Ser residues (GenBank accession number AAA60019.1), suggesting that 15G9 cannot bind Tn residues expressed in the MUC1 mucin. In contrast, tri-serine amino acidic motifs are present in other apomucins, such MUC2 (GenBank accession number AAA59164.1), MUC5AC (GenBank accession number CAA04737.1), MUC5B (GenBank accession number CAA96577.1), and MUC6 (GenBank accession number NP_005952.2). It is important to note that all these four apomucins, normally expressed in other tissues and usually undetectable in normal breast, have been characterized as breast cancer-associated mucins [5154]. Therefore, we hypothesize that the MAb 15G9 reactivity in breast cancer could be associated with underglycosylated “non-mammary” mucin expression. Regarding mucin expression in colon, MUC2 is present in normal cells and generally decreased in colorectal adenocarcinoma. In contrast, MUC1 expression is increased in colon cancers, and MUC5AC mucin, a product of normal gastric mucosa, is absent from normal colon, but frequently present in colon cancers [55]. As stated above, both MUC2 and MUC5AC contain tri-serine amino acidic motifs. It remains to be determined if the sub-group of tumors expressing trimeric GalNAcα-O-Ser (S*S*S*) exhibits a particular immunobiological behavior. In accordance with the results obtained with MAb 15G9, we showed previously that a cluster of tri-Tn displayed on homoserine residues is not very efficient to induce anti-Tn antibodies that recognize tumor cells [18].

The Tn antigen represents also an attractive target for antibody-mediated passive immunotherapy. Tumor regressions have been observed in experimental models using both mouse [56] and chimeric human/mouse [57, 58] anti-Tn monoclonal antibodies. The mechanism of the in vivo effect on diminishing tumor growth mediated by anti-Tn antibodies may be explained by: (1) antibody-dependent cellular cytotoxicity (ADCC) [5759]; (2) inhibition of cancer cell adhesion to lymphatic endothelium [60]; and (3) direct blocking of receptor signaling, such as epidermal growth factor receptor and insulin-like growth factor I receptor [61]. These different mechanisms may be strongly dependent not only of the antibody class but also of the fine specificity of the antibody toward the Tn antigen type. MAbs with various Tn fine specificity, such as those generated in this work, may be valuable tools to clarify the mechanisms of passive anti-tumor immunotherapy mediated by anti-Tn antibodies.

In conclusion, our results highlight a new structural feature of the tumoral Tn antigen with a strong potential clinical value. It was previously found that the density of the Tn determinant as well as the amino acid residues neighboring the O-glycosylation sites is critical for its recognition by specific antibodies elicited against different forms of this antigen. Here, we demonstrate that the aglyconic part of the Tn structure (Ser vs Thr) plays a key role in anti-Tn specificity for breast and colon cancer detection. Given the increasing interest of Tn antigen as a molecular cancer target, our findings can be very useful to design efficient Tn-based immunogens as well to select antibodies for in vivo diagnostic and therapeutic strategies.


This work was supported by grants from Programmes Transversaux de Recherche (PTR, Institut Pasteur, Paris, France) and ECOS France-Uruguay Program to Eduardo Osinaga, Sylvie Bay and Claude Leclerc and from the Ligue Nationale Contre le Cancer (Equipe Labellisée 2011) and Banque Privée Européenne to Claude Leclerc, and Programa Grupos de Investigación (CSIC, Universidad de la República, Uruguay) to Eduardo Osinaga and Otto Pritsch.

Conflict of interest

The authors declare that they have no conflict of interest.

Supplementary material

262_2013_1425_MOESM1_ESM.pdf (567 kb)
Supplementary material 1 (PDF 567 kb)

Copyright information

© Springer-Verlag Berlin Heidelberg 2013