Novel humanized anti-CD3 antibodies induce a predominantly immunoregulatory profile in human peripheral blood mononuclear cells

Immunology Letters 125 (2009) 129–136

Contents lists available at ScienceDirect

Immunology Letters journal homepage: www.elsevier.com/locate/

Novel humanized anti-CD3 antibodies induce a predominantly immunoregulatory profile in human peripheral blood mononuclear cells Hernandez M. Silva a,b,c , Pedro M.M.M. Vieira b,c , Patricia L.N. Costa a , Bárbara M.S. Pimentel a , Ana M. Moro c,d , Jorge Kalil b,c , Andrea Q. Maranhão a,c , Verônica Coelho b,c , Marcelo M. Brigido a,c,∗ a

Departamento de Biologia Celular, Universidade de Brasília, 70910-900 Brasília, DF, Brazil Immunology Laboratory, Heart Institute (InCor), University of São Paulo Medical School, CEP 05403-000 São Paulo, SP, Brazil Institute for Investigation in Immunology, Millennium Institute, São Paulo, Brazil d Laboratório Especial de Biofármacos em Célula Animal, Instituto Butantan, CEP 05503-900 São Paulo, SP, Brazil b c

a r t i c l e

i n f o

Article history: Received 23 April 2009 Received in revised form 15 June 2009 Accepted 22 June 2009 Available online 30 June 2009 Keywords: Antibody engineering Antibody humanization Anti-CD3 CD3 IL-10

a b s t r a c t Strategies to minimize the immunogenicity and toxicity of murine anti-CD3 antibodies (e.g. OKT3) are of special interest for organ transplantation and for the treatment of autoimmune diseases. In the present work, we have developed two humanized anti-CD3 antibodies. These molecules were shown to bind to human CD3, though less efficiently, and display less mitogenic activity than OKT3. These results prompted us to investigate whether this reduced mitogenic potential was associated with the development of anti-inflammatory properties. Indeed, in peripheral blood mononuclear cells (PBMCs), the humanized antibody versions induced a predominantly anti-inflammatory cytokine profile, in contrast with the proinflammatory profile induced by OKT3. Neither OKT3 nor the humanized versions induced the expression of IL-4, IL-2 or TGF-␤. Both humanized antibodies induced significantly lower production of IFN-␥ and IL-5 and slightly higher production of IL-10 than OKT3. This immunomodulatory profile was most evident by the 80-fold higher ratio of IL-10/IFN-␥ production in PBMCs cultured in the presence of the humanized antibodies, compared to those stimulated with OKT3. Furthermore, these humanized anti-CD3 antibodies induced a late FOXP3 gene expression while OKT3 led to a more transient expression of FOXP3. Taken our results, we suggest that these humanized anti-CD3 antibodies may promote the development of T cells with immunoregulatory activity. © 2009 Elsevier B.V. All rights reserved.

1. Introduction Despite the therapeutic progress achieved in the fields of organ transplantation and autoimmune disease, control of the immune response in such pathological conditions remains challenging. It is well-established that T cells play a central role in the recognition of allo- or autoantigens, governing the inflammatory immune response in transplantation and autoimmune reactions. Therefore, the development of new approaches to control the inflammatory immune response that target T cells is of special interest. During the past two decades, OKT3 (muromonab-CD3, Ortho Biotech), the first mouse monoclonal antibody specific for human CD3 to be used in clinical practice [1], was largely used to treat

Abbreviations: CDR, complementary determinant region; Fab, antigen binding fragment; PCR, polymerase chain reaction; FvFc, single chain Fv fragment fused to a human IgG1 Fc; VL, light chain variable region; VH, heavy chain variable region. ∗ Corresponding author at: Departamento de Biologia Celular, Universidade de Brasília, 70910-900 Brasília, DF, Brazil. Tel.: +55 61 33072423; fax: +55 61 33498411. E-mail address: [email protected] (M.M. Brigido). 0165-2478/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.imlet.2009.06.009

acute rejection episodes [2]. The suppressive activity was initially attributed mainly to the depletion of T cells. However, later reports showed that the effects of anti-CD3 monoclonal antibodies (mAb) may be more diverse and complex, even including immunoregulatory mechanisms [3]. Renewed interest in CD3 specific antibodies has emerged with the development of new antibody formulations involving immunoglobulin humanization [4–7]. When administrated in a short-term treatment, these antibodies are capable of eliciting an immunoregulatory profile of antigen-specific unresponsiveness that is sustained in the absence of chronic immunosuppression both in transplantation and in the context of autoimmune diseases [4,8–11]. Studies by several groups indicate that the tolerance induced by anti-CD3 humanized antibodies may be associated with the development of a subset of CD4 regulatory T cells [3]. OKT3 is the only anti-CD3 antibody approved by FDA for clinical use to date. Unfortunately, due to its murine origin OKT3 triggers an immunogenic reaction in vivo, which limits its widespread and prolonged use in transplantation, as well as the extension of its use to other clinical areas such as autoimmunity. Antibody humanization has been used to overcome the problems involved with heterolo-

130

H.M. Silva et al. / Immunology Letters 125 (2009) 129–136

gous antibodies in clinical applications [12,13] and is as a valuable tool for the development of a new generation of CD3 specific antibodies. In this work, we developed two humanized versions of OKT3. The development strategy was to use the closest human germline sequence to receive the murine CDRs, as was previously and successfully described by our group [14]. These humanized versions were shown to bind to human CD3, but had less mitogenic activity than OKT3. We also observed that the humanized antibodies induced a predominantly anti-inflammatory cytokine profile with late expression of FOXP3 when compared with OKT3, suggesting that humanized anti-CD3 antibodies might stimulate the development of T cells with immunomodulatory activity. 2. Materials and methods 2.1. Computational analysis Human VH and VL framework sequences were obtained from the closest germline sequence extracted from SwissProt [15] for the VH or GenBank (http://www.ncbi.nlm.nih.gov/igblast/), for the VL. Sequence search was performed with FASTA [16] or BLAST [17] algorithms. Clustal W [18] was used for multisequence alignment, running in BioEdit version 5.0.9 (http://www.mbio.ncsu.edu/bioedit/bioedit.html). Sterically constrained atoms were detected using similar tri-dimensional models in Protein Data Bank (www.pdb.org). Model visualization and atomto-atom distance calculations were performed in RASMOL version 2.6 [19]. Variable region numbering followed Kabat’s convention [20], except for the 1SYS pdb model analysis, where we preserved the original numbering for clarity. 2.2. Construction of the humanized FvFc sequences and expression vectors All plasmids were constructed using standard cloning methods [21]. Synthetic genes coding for the humanized VHs and VL were produced (GenScript Corp.). VH and VL coding sequences were cloned in a FvFc vector [22], generating the humanized antiCD3 FvFc sequences. The whole FvFc constructs were transferred to the pMIRES vector between the XmaI and EcoRI restriction sites. pMIRES is a mammalian dicistronic expression vector for the production of humanized FvFcs that contains a CMV promoter, an Ig leader sequence, an EMCV-IRES-NEO element from the pLXIN vector (Clontech, Palo Alto, CA), and a SV40 polyadenylation sequence.

cell lines were analyzed for antibody production by sandwich ELISA, as described below. The clones that produced the highest amount of recombinant antibodies were selected and grown in Ham-F12 medium supplemented with 1.25% Ultra-low IgG Fetal Bovine Serum (Invitrogen, USA), until the necessary amount of antibody was achieved for use in experiments. All proceedings were performed in sterile conditions to avoid contamination with endotoxin. 2.4. FvFc production measurement The recombinant FvFcs present in cell culture supernatant were quantified by sandwich ELISA. Ninety-six-well microtiter plates (Nunc, USA) were coated with a polyclonal goat anti-human IgG antibody (Pierce, USA), and then serial dilutions of culture supernatants from transfected cells were loaded into the plates. Bound humanized FvFc were detected using goat anti-human IgG (Fc specific) conjugated with alkaline phosphatase (Sigma, Germany). All samples were measured in duplicate. A standard curve was generated using known concentrations of human IgG. 2.5. Humanized FvFc purification About 1 l of culture medium was collected from the selected antibody-expressing CHO-K1 clones. Due to its human IgG1 Fc, the humanized FvFcs were purified on a 1 ml HiTrap Protein-A HP column (GE Healthcare, Sweden). Columns were equilibrated with 5 CV (column volumes) of washing buffer (20 mM sodium phosphate, pH 7.0, 0.2 ␮m filtered). The samples were filtered on 0.22 ␮m membranes and loaded onto the columns at a flow rate of 1 ml/min, and the columns were washed with 20 CV washing buffer. The antibodies were eluted with 100 mM citric acid (pH 3.5). One milliliter fractions were collected into tubes containing 50 ␮l of 2 M Tris–HCl (pH 9.0) for neutralization. Relevant fractions were pooled after analysis by 10% SDS-PAGE, and dialyzed against PBS. All solutions were sterile and the process was realized in germ-free conditions to avoid any contamination. Finally, antibody concentration was determined by ELISA, as described above. The quality and purity of the preparations were analyzed by reducing and non-reducing SDSPAGE with Coomassie blue staining and by Western blotting using anti-human IgG (Fc specific) conjugated with alkaline phosphatase (Sigma, Germany) where the corresponding antibodies fragments were visualized as monomeric 55 kDa and dimeric 110 kDa bands (Fig. 1 and data not shown). The purity observed was around 95% (Fig. 1). 2.6. Humanized FvFcs direct binding activity assay

2.3. Cell culture, transfection and selection of expressing clones CHO-K1 cells (ATCC number: CCL-61) were cultured in Ham-F12 (Hyclone, USA) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin (100 ␮g/ml). Cells were grown in six-well flat-bottom tissue culture plates (TPP, Switzerland) and were observed daily until 80–90% confluence was reached. Then, cells were transfected using JetPEI reagent (Polyplus Transfection, France) according to the manufacturers’ protocols. Untransfected cells served as background controls in every plate. Stable CHOK1 clones producing the humanized FvFc versions T and R were obtained by G418 selection (600 ␮g/ml) starting from 72 h after transfection. Mock-transfected CHO-K1 cells died after 7–10 days of selection. Clones were isolated by limiting dilution, dispensing approximately 0.5 cell per well into 96-well plates, and maintaining under continued selection (600 ␮g/ml G418). Isolated clones were then seeded in 75 cm2 flasks and maintained under continued selection for 14 days before switching to culture medium without G418. Culture supernatants from the individual clonal

Human peripheral blood mononuclear cells (PBMCs) were isolated from heparinized blood of three healthy volunteers by Ficoll-Hypaque density gradient centrifugation. Antibodies used were: FITC conjugated goat anti-mouse IgG (Sigma, Germany; used to detect the direct binding of OKT3 to human T cell surface); goat anti-human IgG–FITC (Sigma, Germany; used to detect the direct binding of the humanized FvFcs to human T cell surface); phycoerythrin (PE) conjugated mouse anti-human CD4 and CD8 (BD Pharmigen, USA; both used simultaneously to select the T cell subpopulations); mouse anti-human CD3 (OKT3) FITC conjugated (Ebioscience, USA); mouse anti-human CD3 PE conjugated clone UCTH1 (Dako, Denmark) and Orthoclone OKT3 (Ortho biotech, USA). Four hundred thousand (4 × 105 ) PBMCs were incubated with the different antibodies for 30 min at 4 ◦ C, washed three times in FACS buffer (PBS, 2% FCS and 0.01% sodium azide), spun down and resuspended in 400 ␮l of FACS buffer. For the samples with the recombinant humanized FvFcs and the OKT3, a second incubation was performed with goat anti-human IgG-FITC or goat anti-mouse

H.M. Silva et al. / Immunology Letters 125 (2009) 129–136

IgG-FITC, respectively. All samples were acquired using a FACScalibur cytometer (Becton Dickinson, USA) and analyzed using CellQuest PRO (Becton Dickinson, USA) or FlowJo software version 7.2.5 (Treestar, USA). Results are expressed as the percentage of stained cells. 2.7. Blocking capacity of the humanized FvFc In order to analyze the binding specificity of the two humanized FvFcs, a blocking experiment was performed. The ability of the FvFc to block the binding of the original OKT3-FITC (Ebioscience, USA) or the related mouse anti-human CD3 PE-conjugated clone UCTH1 (Dako, Denmark) to surface CD3 molecules was tested. PBMCs were initially incubated with the two humanized FvFcs or with unconjugated OKT3 (Ortho biotech, USA), washed, and then incubated with sub-saturating concentrations of the OKT3-FITC or of the antiCD3-PE clone UCTH1. The median fluorescence intensity (MFI) was compared in samples with OKT3-FITC or anti-CD3-PE clone UCTH1 alone and in samples with the different humanized FvFc, or OKT3 plus OKT3-FITC or anti-CD3-PE clone UCTH1. The percent inhibition of binding was calculated by comparing the shift of the MFI in the blocking experiments to those observed upon the direct binding of conjugated antibodies.

131

7500 Sequence Detection System (Applied Biosystems, USA). Reactions consisted of an initial 10 min 95 ◦ C denaturation step followed by 40 cycles of 15 s at 95 ◦ C and 60 s at 60 ◦ C. All reported mRNA levels were normalized to the GAPDH mRNA, where GAPDH = 1. The FOXP3 and GAPDH primers sequences were as follows—Foxp3 primers: forward 5 GAGAAGGGCAGGGCACAAT 3 and reverse 5 GTGGGCCTGCATGGCAC 3 ; GAPDH primers: forward 5 TGGTCTCCTCTGACTTCAACA 3 and reverse 5 AGCCAAATTCGTTGTCATACC 3 . To eliminate amplifications from contaminating genomic DNA, these primers were designed to span an intron/exon boundary and thus to anneal specifically to cDNA. Normalized values for Foxp3 mRNA expression in each sample was calculated using the Pfaffl method [24]. All samples were run in triplicate. 2.11. Data analysis Data are presented as mean ± SEM data from independent experiments. The results were analyzed using unpaired T test between the comparable groups using GraphPad Prism software version 5 (GraphPad software). 3. Results 3.1. Construction and expression of humanized anti-CD3 FvFcs

2.8. Fluorescein labeling of PBMC cells and proliferation assays Labeling of cells with carboxyfluorescein diacetate 5,6 succinimidyl ester (CFSE) (Molecular Probes, USA) prior to assaying proliferation allows the quantification of cell division, as described previously [23]. Briefly, isolated PBMC cells were washed in PBS, counted, adjusted to 107 cells/ml of PBS, and incubated with 2.5 ␮M of CFSE (Sigma–Aldrich, USA). The cells were incubated at 37 ◦ C protected from light for 8 min under gentle agitation, and then five volumes of cold 10% FBS supplemented RPMI 1640 (Invitrogen, USA) were added to quench the reaction for 5 min on ice. The labeled cells were washed three times with the same medium, spun down and resuspended in culture medium. CFSE-labeled human PBMCs (1 × 106 cells/well) were incubated with 1 ␮g of the humanized FvFcs or OKT3 for 5 days at 37 ◦ C. On the fifth day, cells were harvested, transferred to 4 ml FACS tubes, washed three times, spun and resuspended in 400 ␮l of FACS buffer. Data were acquired using a FACScalibur cytometer as described above. 2.9. Cytokine production detection by cytometric bead array Human PBMCs (2 × 106 cells/well) were incubated with 1 or 5 ␮g of the purified humanized FvFcs or OKT3 at 37 ◦ C. Culture supernatants were collected at 72 or 192 h. The concentrations of cytokines in the supernatant fluids were measured by flow cytometry using the cytometric bead array (CBA) human Th1/Th2 cytokine kit (Becton Dickinson, USA), which detects the cytokines IL-2, IL4, IL-5, IL-10, TNF-␣, and IFN-␥. The lower limit of detection of cytokines in this assay was 2.5 pg/ml. 2.10. FOXP3 real time quantitative PCR Total cellular RNA was extracted using Trizol reagent (Invitrogen, USA) from 2 × 106 PBMCs previously incubated with 1 or 5 ␮g of the humanized FvFcs or OKT3 for 72 or 192 h at 37 ◦ C. Total RNA was reverse-transcribed using Superscript II reverse-transcriptase, with oligo(dT)12–18 primer and random primers (Invitrogen, USA) in a final volume of 20 ␮l. For the detection of FOXP3 mRNA, the reaction mixture (10 ␮l) contained 0.3 ␮M of forward and reverse FOXP3 primers and the SYBR-Green PCR Master Mix (Applied Biosystems, USA), according to the manufacturers’ recommendations. The mRNA levels were quantified using the Perkin-Elmer ABI Prism

OKT3 humanization was performed by the CDR grafting method [25], in which murine CDRs were grafted to the closest human germline framework sequence. The closest human sequence found for VH was H-1 HG3 (GenBank accession number P01743) [26], which shares a 71% identity and 83% sequence similarity with OKT3 VH. For the VL, we used the sequence L6 (GenBank accession CAB37836) [27], which shares 61% of identity and 75% of similarity with OKT3 VL. The JH and Jk used were JH4 and Jk2, respectively. Structural analyses were performed using the closest murine Fab crystal (1MRC from PDB which possess 82.20% of similarity with OKT3 in VH), looking for possible constraints on antibody structure for the CDR grafting. These analyses suggested that residue 87, located on framework 3, should be important for the spatial conformation of CDR 2, due to a putative hydrogen bond between Thr87 and Arg66 . Therefore, we proposed two constructions of humanized antibody for the VH domain, one relying on the human germline Arg residue at position 87 (named version R), and the other preserving the murine Thr residue (named version T) [28]. The proposed VH/VL were chemically synthesized for cloning as FvFc fragments (a single chain Fv fused to a human IgG1 Fc) [22] (Fig. 1) onto the mammalian expression vector, pMIRES, as described in Section 2. Both humanized constructions (R and T) were transfected into CHO-K1 cells. The antibody fragments were expressed stably in CHO-K1 cell lines, and clones expressing antibodies were obtained by positive selection. Humanized antibody fragments were purified by affinity chromatography from pooled cell culture supernatants, and the antibodies produced were quantified by sandwich ELISA (data not shown). 3.2. In vitro functional characterization of recombinant humanized antibodies To analyze the binding capacity of the humanized FvFcs to human CD3 on the T lymphocyte surface, we performed an assay of direct binding to peripheral blood mononuclear cells (PBMCs) by flow cytometry. The humanized versions R and T and OKT3 showed similar cell-binding capacities, showing 90%, 87% and 95% staining of gated cells, respectively (Fig. 2). Despite these results, we observed that the humanized versions exhibited significantly lower median fluorescence intensities (MFIs) (77.06 and 153.20, for T and R versions, respectively), compared to OKT3 (568.97) (Fig. 2). These

132

H.M. Silva et al. / Immunology Letters 125 (2009) 129–136

(data not shown). However, they were not able to completely block OKT3 binding to T lymphocytes. While OKT3 blocked up to 82% of OKT3-FITC binding (Table 1), R and T versions of the humanized antibodies blocked only 49% and 21%, respectively (Table 1). These data confirmed that the humanized versions are capable of binding to CD3 molecules on T lymphocyte surface, but with a lower ability to compete with OKT3. 3.3. Anti-CD3 humanized FvFcs display a lower mitogenic potential

Fig. 1. The FvFc molecule migrates as a 55 kDa monomer. (A) FvFc preparation was run under reducing condition in 10% SDS-PAGE and stained with Coomassie blue. Lane 1: Unstained Protein Molecular Weight Marker (Fermentas® ); lane 2: 1 ␮g of purified FvFc preparation; lane 3: 500 ng of human serum IgG (SIGMA® ). Arrow indicates the FvFc band. (B) Schematic representation of an FvFc molecule. FvFcs are immunoglobulin-like structures comprising a homodimeric pair of VH, VL, hinge, CH2 and CH3 domains. The variable regions are joined with a flexible polypeptide linker (Gly4 Ser)3 .

results suggest that the humanized proteins present an apparent reduction in binding to PBMCs. To confirm this hypothesis, we carried out blocking assays. We observed that the R and T versions were capable of completely blocking the binding of the related antiCD3-PE clone UCTH1, blocking 85% and 82% of binding, respectively

Since the humanized FvFcs have an human IgG1 Fc fragment, we investigated their mitogenic potential. To evaluate this activity, we performed a T cell proliferation assay using flow cytometric measurement of CFSE dye dilution. The proliferation pattern of cells after stimulation with OKT3 showed that this antibody induced up to 62% of total cells to proliferate (Fig. 3). However, when we analyzed the mitogenic potency of the humanized versions, no significant T cell proliferation was induced. We observed proliferation of only 8% and 13% of total cells for T and R versions, respectively (Fig. 3). These results reveal that the mitogenic potentials of the humanized versions T and R were much weaker than that induced by OKT3. 3.4. Cytokine profiles induced by anti-CD3 antibodies To further investigate the biological properties of the humanized versions we cultivated PBMC in the presence of the anti-CD3 antibodies at different concentrations and analyzed the cytokine profiles. We observed that the humanized antibodies induced a significantly lower production of IFN-␥ and IL-5 compared with OKT3, both 72 and 192 h after stimulation with 5 or 1 ␮g/ml of antibody (Fig. 4A). In contrast, IL-10 production was slightly higher, but not

Fig. 2. Anti-CD3 antibodies direct binding to the human lymphocyte cell surface. The anti-CD3 humanized versions T (D) and R (E) displayed similar binding capacities to OKT3 (C), staining 87%, 90% and 95% of gated cells, respectively. Lymphocytes were gated in a forward versus side scatter dot plot. To avoid the detection of human immunoglobulins on the B cells surface, a second gate was performed on cells stained with phycoerythrin (PE)-labeled mAb for CD4 and CD8. CD3 specific antibodies were detected with FITC-labeled anti-human or anti-mouse IgG, which was plotted as a histogram. (A) and (B) represent cells incubated only with the anti-mouse IgG FITC and cells incubated only with the anti-human IgG FITC, respectively.

H.M. Silva et al. / Immunology Letters 125 (2009) 129–136

133

Table 1 Inhibition of OKT3-FITC binding to human PBMCs surface by humanized anti-CD3 versions. Blocking antibody

FITC median fluorescence intensity

% of OKT3-FITC binding inhibition

Antibody quantity added (ng)

OKT3 R T OKT3-FITC

24.14 71.69 109.41 139.49

82 49 21 0

100 300 300 –

Human PBMCs were incubated with different concentrations of recombinant humanized FvFc or unconjugated OKT3, washed, and then incubated with sub-saturating concentrations of the mAb OKT3-FITC. The percentage of OKT3-FITC binding inhibition reflects the decrease in the MFI of OKT3-FITC when the cells were cultured with anti-CD3 antibodies.

statistically different, in cells treated with humanized antibodies 72 and 192 h after stimulation (Fig. 4A). Moreover, we did not observe any significant difference in TNF-␣ production induced by OKT3 and by the humanized versions (Fig. 4A). We also analyzed IL-2 and IL-4 production, but the values were almost under the detection limit of 2.5 pg/ml of the CBA standard curve (data not shown). Unexpectedly, when we compared the relative ratios of IL-10 to IFN-␥, we observed at least a 80-fold higher relative production of IL-10 in cells cultured with the T and R versions than those treated with OKT3 (Fig. 4B). In addition, we assessed TGF-␤1 production by ELISA, but no significant differences from control condition were observed (data not shown).

tive PCR. Although the results were not significantly different, we observed a slight transient increase in FOXP3 expression at 72 h and a drop to baseline levels at 192 h in cells stimulated with 1 ␮g/ml of OKT3 (Fig. 5). However, when cells were stimulated with 5 ␮g/ml of OKT3, we observed the same increase in FOXP3 expression at 72 and a sustained increase in expression at 192 h (Fig. 5). In contrast, when cells were stimulated with 1 or 5 ␮g/ml of humanized T version or with 1 ␮g/ml of the R version, we observed no increase in early expression of FOXP3. Instead, the humanized versions induced a later expression of FOXP3 at 192 h (Fig. 5).

4. Discussion 3.5. Humanized anti-CD3 antibodies induce late FOXP3 expression We next evaluated whether the cytokine profile induced by stimulation with humanized anti-CD3 antibodies was related to the expression of the transcription factor FOXP3 by real time quantita-

Anti-CD3 antibodies have been successfully used in the clinic for the last 20 years, mostly for the prevention and treatment of organ allograft rejection. Aiming to minimize the immunogenicity and toxicity induced by murine anti-CD3 antibodies, humanization of these antibodies has been pursued by several groups [6,7,29,30].

Fig. 3. Lymphocytes proliferation induced by anti-CD3 antibodies. Human CFSE-labeled PBMCs at 1 × 106 cells/well were incubated for 5 days with 1 ␮g/ml of OKT3 (B), humanized version T (C), and humanized version R (D). Cells were then harvested and cell proliferation was assessed by flow cytometric measurement of CFSE dilution. Part A shows the CFSE-labeled pattern of control non-proliferating cells in culture without any stimulus at the gate of 98.72%. The numbers presented in the two determined gates represent the percentage of cells in relation to the total cell number.

134

H.M. Silva et al. / Immunology Letters 125 (2009) 129–136

Fig. 4. Cytokines produced by human PBMCs in response to OKT3 or humanized versions R and T. (A) Human PBMC were incubated for 72 or 192 h in the presence of soluble OKT3 or humanized versions R or T at concentrations of 5 ␮g/ml (OKT3 5, R5, T5) or 1 ␮g/ml (OKT3 1, R1, T1). Culture supernatants were collected after and cytokine concentrations were measured using CBA beads (BD Biosciences) by flow cytometry. The control condition shows the cytokines measured in cell culture supernatant without any antibody stimulus. All data were represented as the mean ± SEM of three independent experiments. (B) Relative IL-10/IFN-␥ molar ratio induced by the anti-CD3 antibodies. The IL-10/IFN-␥ molar ratio was calculated using the cytokine production induced by OKT3 or humanized versions R and T. (*) represents p < 0.05.

We, here, present two humanized anti-CD3 antibodies developed by our group as FvFc fragments that display differential binding and immunologic properties in relation to the murine OKT3 antibody. We chose this antibody format because it displays several advantages, as described in the literature [31]. This fragment preserves the targeting specificity of the whole mAb, it is minimally immunogenic and has superior biodistribution and blood clearance properties [31–35]. To perform the antibody humanization we identified the human immunoglobulin germline sequence [25] most similar to OKT3, to graft the murine CDRs. Then, we carried out a three-dimensional

structural analysis of the closest murine antibody crystal structure of OKT3 (1MRC) reported in the Protein Data Bank [28]. Among the differences between the murine and the humanized sequences, we identified a residue in the murine VH, Thr87 , situated between the CDR2 and framework 3 that could have an impact on the conformation of CDR2. Therefore, we constructed two versions of humanized VHs, one bearing the murine Thr87 (named version T), and the other preserving the human germline Arg87 residue (version R). Both humanized versions of anti-CD3 antibodies were able to bind to the human CD3 molecule on T lymphocytes. However, they could not completely block OKT3 binding to T lymphocytes in competition

H.M. Silva et al. / Immunology Letters 125 (2009) 129–136

Fig. 5. Expression of Foxp3 by human PBMCs in response to stimulation by OKT3 or by the humanized versions R and T. Human PBMC were incubated for 72 or 192 h in the presence of soluble OKT3 or humanized versions R or T at 5 ␮g/ml (OKT3 5, R5, T5) or 1 ␮g/ml (OKT3 1, R1, T1, respectively). cDNA samples prepared from treated PBMC were subjected to real time quantitative PCR analyses using primers specific for Foxp3 and GAPDH. The relative quantity of Foxp3 in each sample was normalized to the relative quantity of GAPDH using the Pfaffl method [24]. All data were represented as the mean ± SEM of three independent experiments.

assays, especially the T version. Although we have not directly measured the affinity of our humanized antibodies, these data suggest a loss of affinity for CD3 in the humanized. More specific experiments involving surface plasmon resonance (SPR) will be performed to evaluate the antibody–antigen binding characteristics and to test if this hypothesis is correct. The potential to induce proliferation and cytokine production of the humanized antibodies was evaluated. This is especially relevant for our antibodies due to the presence of a human IgG1 Fc. The proliferation induced by both the R and T humanized versions was significantly lower than that observed for OKT3, indicating a less potent mitogenic capacity (Fig. 3). The possible low antigen affinity of these antibodies may be implicated in impairing the coating of T lymphocytes and the recognition of the antibodies at the cell surface by the low affinity Fc␥ receptors on phagocytes. Considering their low capacity to induce proliferation, we asked whether our antibodies displayed a differential cytokine profile. In contrast with OKT3, which induced high amounts of IFN-␥ and significant production of TNF-␣, IL-5 and IL-10 (Fig. 4A), the humanized versions induced a slightly higher production of IL-10, similar production of TNF-␣ and a significantly lower production of IFN␥ and IL-5 than OKT3 (Fig. 4A). Despite the similar amounts of TNF-␣ induced, in vitro, by both OKT3 and our humanized antibodies, the IL-10/IFN-␥ ratio was markedly higher for the humanized antibodies (Fig. 4B). It has consistently been reported in the literature that the presence of high levels of IFN-␥ can inhibit the suppressive effects of IL-10 and alter the pattern of IL-10-induced genes [36]. However, when IFN-␥ levels drop below those of IL-10, the anti-inflammatory signaling pathway predominates, inactivating effector cells and resolving inflammation [36]. Therefore, based on our results, we suggest that the humanized antibodies have the potential to induce an anti-inflammatory microenvironment, probably mediated by the high IL-10/IFN-␥ ratio, favoring the development of immunoregulatory conditions through the activation of IL-10 suppressor mechanisms. Due to difficulties in the recombinant protein expression and the limited amount of humanized antibodies, at the moment, we were unable to test whether our humanized antibodies induced the generation of Tregs. Nevertheless, we evaluated whether our anti-CD3 antibodies induced the expression of the FOXP3 gene, a forkhead family transcriptional factor involved in regulatory T cell suppressive activity [37]. We observed a differential pattern of

135

FOXP3 expression induced by the humanized antibodies and OKT3, which was both time and concentration dependent. The stimulation of PBMC with both concentrations of the humanized T version or with 1 ␮g/ml of the R version did not induce an early expression of FOXP3. Instead, these humanized versions mainly induced a late FOXP3 expression, at 192 h (Fig. 5). These results, together with the high IL-10/IFN-␥ production ratio and the low mitogenic activity are compatible with an immunoregulatory profile for our humanized antibodies. It should be pointed that FOXP3 has been reported to be transiently expressed in activated effector T cells bearing no suppressive activity [38], while stable and high expression of FOXP3 correlates with the induction of suppressive activity in Teff cells [38]. Therefore, it is possible that the expression of FOXP3 induced by the humanized antibodies is not merely due to T cell activation but may involve the generation of T cells with regulatory activity. This hypothesis will be tested in the future. In contrast with this potential immunoregulatory profile, OKT3 seems to show more pro-inflammatory characteristics, as described in the literature [3,39]. Though OKT3 also induced FOXP3, the pattern of expression seems to be more related to T cell activation where a transient increase of it expression is observed (Fig. 5). However, there is a puzzling result because when PBMCs were stimulated at a higher concentration of OKT3 (5 ␮g/ml), we observed a sustained increase in FOXP3 expression. This sustained expression over 192 h could be related to the high antibody concentration, which, due to its high affinity for the antigen, would favor the continuous activation of newly generated cells. However, the generation of Tregs induced by OKT3 cannot be excluded. In summary, both the humanized anti-CD3 antibody versions, T and R, were able to bind to the human lymphocyte CD3, but showed reduced binding and less mitogenic activity than the murine OKT3 mAb. This probably lower affinity may be partly involved in the predominantly anti-inflammatory cytokine profile and the induction of late sustained gene expression of FOXP3. Further studies are necessary to determine whether these humanized anti-CD3 antibodies induce T cells with suppressive activity, and to explore the molecular pathways involved in these conditions. Taken together, these data point to the potential of using humanized antibody fragments in the clinical context when immunomodulatory activity is required, especially for the treatment of autoimmune diseases and allotransplantation. Acknowledgements We thank Sandra Maria Monteiro for help in the FACS analyses and CNPq for the provided studentship. This work was supported by grants from MCT/BID (Brazilian Ministry of Science and Technology). References [1] Cosimi AB, Burton RC, Colvin RB, Goldstein G, Delmonico FL, LaQuaglia MP, et al. Treatment of acute renal allograft rejection with OKT3 monoclonal antibody. Transplantation 1981;32:535–9. [2] Group OMTS. A randomized clinical trial of OKT3 monoclonal antibody for acute rejection of cadaveric renal transplants. Ortho Multicenter Transplant Study Group. N Engl J Med 1985;313:337–42. [3] Chatenoud L. CD3-specific antibody-induced active tolerance: from bench to bedside. Nat Rev Immunol 2003;3:123–32. [4] You S, Leforban B, Garcia C, Bach JF, Bluestone JA, Chatenoud L. Adaptive TGF-␤-dependent regulatory T cells control autoimmune diabetes and are a privileged target of anti-CD3 antibody treatment. Proc Natl Acad Sci USA 2007;104:6335–40. [5] Belghith M, Bluestone JA, Barriot S, Megret J, Bach JF, Chatenoud L. TGF␤-dependent mechanisms mediate restoration of self-tolerance induced by antibodies to CD3 in overt autoimmune diabetes. Nat Med 2003;9:1202–8. [6] Friend PJ, Hale G, Chatenoud L, Rebello P, Bradley J, Thiru S, et al. Phase I study of an engineered aglycosylated humanized CD3 antibody in renal transplant rejection. Transplantation 1999;68:1632–7.

136

H.M. Silva et al. / Immunology Letters 125 (2009) 129–136

[7] Woodle ES, Xu D, Zivin RA, Auger J, Charette J, O’Laughlin R, et al. Phase I trial of a humanized, Fc receptor nonbinding OKT3 antibody, huOKT3␥1(Ala-Ala) in the treatment of acute renal allograft rejection. Transplantation 1999;68:608–16. [8] Kohm AP, Williams JS, Bickford AL, McMahon JS, Chatenoud L, Bach JF, et al. Treatment with nonmitogenic anti-CD3 monoclonal antibody induces CD4+ T cell unresponsiveness and functional reversal of established experimental autoimmune encephalomyelitis. J Immunol 2005;174:4525–34. [9] Herold KC, Gitelman SE, Masharani U, Hagopian W, Bisikirska B, Donaldson D, et al. A single course of anti-CD3 monoclonal antibody hOKT3␥1(Ala-Ala) results in improvement in C-peptide responses and clinical parameters for at least 2 years after onset of type 1 diabetes. Diabetes 2005;54:1763–9. [10] Utset TO, Auger JA, Peace D, Zivin RA, Xu D, Jolliffe L, et al. Modified antiCD3 therapy in psoriatic arthritis: a phase I/II clinical trial. J Rheumatol 2002;29:1907–13. [11] Nicolls MR, Aversa GG, Pearce NW, Spinelli A, Berger MF, Gurley KE, et al. Induction of long-term specific tolerance to allografts in rats by therapy with an anti-CD3-like monoclonal antibody. Transplantation 1993;55:459–68. [12] Jones PT, Dear PH, Foote J, Neuberger MS, Winter G. Replacing the complementarity-determining regions in a human antibody with those from a mouse. Nature 1986;321:522–5. [13] Riechmann L, Foote J, Winter G. Expression of an antibody Fv fragment in myeloma cells. J Mol Biol 1988;203:825–8. [14] Caldas C, Coelho V, Kalil J, Moro AM, Maranhao AQ, Brigido MM. Humanization of the anti-CD18 antibody 6.7: an unexpected effect of a framework residue in binding to antigen. Mol Immunol 2003;39:941–52. [15] Bairoch A, Apweiler R. The SWISS-PROT protein sequence database and its supplement TrEMBL in 2000. Nucleic Acids Res 2000;28:45–8. [16] Pearson WR. Flexible sequence similarity searching with the FASTA3 program package. Methods Mol Biol 2000;132:185–219. [17] Altschul SF, Madden TL, Schaffer AA, Zhang J, Zhang Z, Miller W, et al. Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Res 1997;25:3389–402. [18] Higgins DG, Thompson JD, Gibson TJ. Using CLUSTAL for multiple sequence alignments. Methods Enzymol 1996;266:383–402. [19] Bernstein HJ. Recent changes to RasMol, recombining the variants. Trends Biochem Sci 2000;25:453–5. [20] Kabat EA, Wu TT, Perry HM, Gottesmann KS, Foeller C. Sequences of proteins of immunological interest. Bethesda, MD: Public Health Service, National Institutes of Health; 1991. [21] Sambrook J, Russell DW. Molecular cloning: a laboratory manual. Cold Spring Harbor, N.Y.: Cold Spring Harbor Laboratory; 2001. [22] Andrade EV, Albuquerque FC, Moraes LM, Brigido MM, Santos-Silva MA. Singlechain Fv with Fc fragment of the human IgG1 tag: construction, Pichia pastoris expression and antigen binding characterization. J Biochem 2000;128:891–5.

[23] Lyons AB, Parish CR. Determination of lymphocyte division by flow cytometry. J Immunol Methods 1994;171:131–7. [24] Pfaffl MW. A new mathematical model for relative quantification in real-time RT-PCR. Nucleic Acids Res 2001;29:e45. [25] Caldas C, Coelho VP, Rigden DJ, Neschich G, Moro AM, Brigido MM. Design and synthesis of germline-based hemi-humanized single-chain Fv against the CD18 surface antigen. Protein Eng 2000;13:353–60. [26] Rechavi G, Ram D, Glazer L, Zakut R, Givol D. Evolutionary aspects of immunoglobulin heavy chain variable region (VH) gene subgroups. Proc Natl Acad Sci USA 1983;80:855–9. [27] Huber C, Schable KF, Huber E, Klein R, Meindl A, Thiebe R, et al. The V␬ genes of the L regions and the repertoire of V␬ gene sequences in the human germ line. Eur J Immunol 1993;23:2868–75. [28] Costa PLN, Fonseca AS, Souza DP, Caldas CA, Coelho V, Moro AM, et al. Humanization and binding activity of humanized/murine hybrid anti-CD3 scFvs. Mol Cell Proteomics 2003;2:760. [29] Carpenter PA, Appelbaum FR, Corey L, Deeg HJ, Doney K, Gooley T, et al. A humanized non-FcR-binding anti-CD3 antibody, visilizumab, for treatment of steroid-refractory acute graft-versus-host disease. Blood 2002;99:2712–9. [30] Li B, Wang H, Dai J, Ji J, Qian W, Zhang D, et al. Construction and characterization of a humanized anti-human CD3 monoclonal antibody 12F6 with effective immunoregulation functions. Immunology 2005;116:487–98. [31] Holliger P, Hudson PJ. Engineered antibody fragments and the rise of single domains. Nat Biotechnol 2005;23:1126–36. [32] Hudson PJ, Souriau C. Engineered antibodies. Nat Med 2003;9:129–34. [33] Pavlinkova G, Colcher D, Booth BJ, Goel A, Wittel UA, Batra SK. Effects of humanization and gene shuffling on immunogenicity and antigen binding of anti-TAG-72 single-chain Fvs. Int J Cancer 2001;94:717–26. [34] Schwarz M, Rottgen P, Takada Y, Le Gall F, Knackmuss S, Bassler N, et al. Singlechain antibodies for the conformation-specific blockade of activated platelet integrin ␣II␤3 designed by subtractive selection from naive human phage libraries. FASEB J 2004;18:1704–6. [35] Zheng D. Antibody gene therapy: an attractive approach for the treatment of cancers and other chronic diseases. Cell Res 2007;17:303–6. [36] Herrero C, Hu X, Li WP, Samuels S, Sharif MN, Kotenko S, et al. Reprogramming of IL-10 activity and signaling by IFN-␥. J Immunol 2003;171:5034–41. [37] Hori S, Nomura T, Sakaguchi S. Control of regulatory T cell development by the transcription factor Foxp3. Science 2003;299:1057–61. [38] Wang J, Ioan-Facsinay A, van der Voort EI, Huizinga TW, Toes RE. Transient expression of FOXP3 in human activated nonregulatory CD4+ T cells. Eur J Immunol 2007;37:129–38. [39] Chatenoud L, Bluestone JA. CD3-specific antibodies: a portal to the treatment of autoimmunity. Nat Rev Immunol 2007;7:622–32.