Humanization of an Anti-CCR4 Antibody That Kills Cutaneous T-Cell Lymphoma Cells and Abrogates Suppression by T-Regulatory Cells

This article is featured in Highlights of This Issue, p. 2309

Cutaneous T-cell lymphoma (CTCL) is the second most common extranodal non–Hodgkin lymphoma in adults, characterized by primary accumulation of clonally derived malignant CD4⁺ T cells in the skin. There are 13 clinically and histologically distinct types of CTCL, and the majority falls into 3 classes: mycosis fungoides, the most common type of CTCL, accounts for almost 50% of all primary cutaneous lymphomas; primary cutaneous CD30⁺ lymphoproliferative disorders—more specifically primary cutaneous anaplastic large cell lymphoma (PC-ALCL)—the second most common CTCL, account for circa 30%; and Sézary syndrome, the most aggressive type of CTCL, makes up circa 5% (1, 2). Other forms of CTCL include adult T-cell leukemia/lymphoma (2%–5%) and peripheral T-cell lymphoma (PTCL; 2%). Although CTCL usually has indolent clinical behavior, in advanced stages, it can progress into an aggressive phenotype with poor prognosis and survival (1, 3, 4) with severe immunodeficiency characteristically developing during disease progression (5). Current hypotheses maintain that the malignant T cells drive this evolving immunodeficiency through the constitutive secretion of immunosuppressive cytokines, dysregulated expression of immunoregulatory proteins by the malignant T cells, and loss of T-cell receptor repertoire complexity due to competitive replacement of normal T cells by the clonally expanded malignant T cells. In addition, a subset of the malignant T cells may act as T-regulatory cells (Tregs) to suppress antitumor responses in some patients with CTCL (6–8). Current therapies for CTCLs do not prevent new lesions from emerging (9) and durable long-term remissions are rare (10). Therefore, disease-specific and more effective therapeutics that can decrease toxicity profiles and induce durable responses will greatly benefit patients with CTCL.

Normal skin-homing CD4⁺ T cells express cutaneous lymphocyte–associated antigen and CC chemokine receptors (CCR) CCR4, CCR6, CCR7, and CCR10 (11–14). Among these, only CCR4 is universally expressed at high levels on the malignant skin-homing T cells, and its surface expression is closely associated with the enhanced skin-homing characteristics of CTCL cells and unfavorable disease outcome (15, 16). Tregs also selectively express high levels of CCR4 compared with other T-cell subsets. The 2 specific ligands for CCR4, chemokines CCL17, and CCL22, produced by tumor cells and cells of the tumor microenvironment, attract CCR4⁺ Tregs to the tumor, where they suppress host immune responses against tumor and create a favorable environment for cancer cells to grow (17, 18). Thus, the high-level expression of CCR4 on CTCL cells and its preferential expression on Tregs (19) make CCR4 not only a potential ideal therapeutic target for CTCLs, but also for other type of cancers for which CCR4⁺ Tregs are involved in their immune evasion.

In this study, we characterized and humanized a mouse anti-CCR4 monoclonal antibody, mAb1567, that recognizes both the N-terminal (NT) and the extracellular domains of CCR4. The antibody exhibited potent antitumor effects in a CTCL mouse model and its mechanism(s) of action, including complement-dependent cytotoxicity (CDC), neutrophil,- and NK-mediated antibody-dependent cellular cytotoxicity (ADCC), were elucidated by a number of in vitro studies. In addition, mAb1567 could not only inhibit Tregs migration toward CCR4 ligand, CCL22, but also abrogate suppression by Tregs in the T-cell proliferation assay. Finally, after the affinity maturation of humanized mAb1567, the resulting mAb2-3 was further improved in affinity and showed stronger CDC and ADCC activities against CCR4⁺ tumor cells.

Mac-1 cell line was isolated from a patient with PC-ALCL, one of CTCLs (20), obtained from Dr. Thomas S. Kupper (Brigham and Women's Hospital, Harvard Medical School, Boston, MA) and cultured in 10% FBS RPMI-1640. Luciferase-expressed Mac-1 cells were stably transduced with a luciferase reporter retrovirus and authenticated by detecting luminescence. 293F cell line was purchased from Invitrogen. 293T (CRL-11268) and Cf2Th (CRL-1430) cell lines were purchased from American Type Culture Collection and incubated in 10% FBS Dulbecco's Modified Eagle's Medium. No additional authentication of these cell lines was conducted by the authors.

mAb1567 was purchased from R&D systems and the other 1567 variant antibodies were produced as described previously (21). Briefly, scFv-Fcs were constructed by cloning the single-chain variable region (scFv) into pcDNA3.1-Hinge vector in frame with human IgG1 Fc region. IgG1 was generated by cloning heavy-chain variable region (VH) and light-chain variable region (VL) into TCAE5.3 vector (22). Antibodies were produced in 293T or 293F cells and purified by proteinA-Sepharose (Amersham) affinity chromatography. For staining, Mac-1 was stained with anti-CCR4 antibodies, detected by fluorescein isothiocynate-conjugated goat–anti-human IgG or anti-mouse IgG antibodies (Sigma), and analyzed with FACSCalibur and CellQuest software.

Mac-1 cells (1 × 10⁶ cells/well) were placed in Transwell migration wells (Corning) with or without mAb1567 for 3 hours at 37°C. Migrated cells harvested from the bottom chamber containing 50 ng/mL human CCL17 or CCL22 (R&D Systems) were enumerated by fluorescence-activated cell sorting (FACS) analysis. Human CD4⁺ T cells were isolated by CD4⁺ T-Cell Isolation Kit (Miltenyi Biotech) and placed in Transwell migration assays with c1567IgG. Migrated cells (CD4⁺CD25^high) were enumerated as above in response to 100 ng/mL CCL22. Percentages of migrated cells were calculated by dividing the number of transmigrated Mac-1 or CD4⁺CD25^high cells by the number of input cells.

For lactate dehydrogenase (LDH) release assay, SCID/Beige mouse neutrophils, human peripheral blood mononucleated cells (PBMC), or human NK cells and neutrophils were used as effector cells and Mac-1, Cf2Th-CCR4, or Cf2Th were used as target cells. Target cells (1 × 10⁴ cells/well) were plated into 96-well plates and antibodies were added. After 1 hour, effector cells were added at an appropriate effector/target (E/T) ratio and incubated (PBMCs, NK, and neutrophils for 4, 16, and 6 hours, respectively). The supernatants were recovered by centrifugation at 300 × g and measured using nonradioactive cytotoxicity assay kits (Promega) at 490 nm. For ⁵¹Cr release assay, 1 × 10⁶ Mac-1 cells were labeled with 100 μCi (3.7 MBq) of Na⁵¹Cr (Amersham International), washed, and used as targets. ⁵¹Cr-labeled target cells (5,000 cells/well) were seeded into 96-well plates and the release of ⁵¹Cr into supernatants was determined. The cytotoxicity was calculated by the following formula:

where E is released LDH from E/T culture with antibody; S_E, spontaneous released LDH from effectors; S_T, spontaneous released LDH from targets; and M, the maximum released LDH from lysed targets.

The 5 × 10⁴ Mac-1 cells per well, resuspended with medium containing rabbit serum (Cedarlane Laboratories) or mouse serum (IMS-COMPL; Innovative Research), were plated in 96-well plates with anti-CCR4 antibodies. After 4-hour incubation, the supernatants were recovered and detected by LDH release assay, calculating as: |$\% \,\,{\rm Cytotoxicity} = {\rm 100} \times \left({{\rm E} - S_{\rm E} - S_{\rm T}} \right){\rm /}\left({{\rm M} - S_{\rm T}} \right)$|⁠.

CD4⁺CD25^high and CD4⁺CD25⁻ T cells were sorted by Beckman CoulterMoFlo sorter using mouse-anti-human CD4-PE-Cy5 (RPA-T4) and antihuman CD25-PE (M-A251) antibodies (BD Pharmigen). CD4⁺CD25⁻ Teffs (2,500) were cultured with or without CD4⁺CD25^high Tregs (1,250) in 96-well plates with 25,000 irradiated (3,000 rad) CD3-depleted PBMCs. Cells were stimulated with 0.05 μg/mL plate-bound anti-CD3 (UCHT1) and 1 μg/mL soluble anti-CD28 (CD28.2) antibodies (BD Pharmigen). Anti-CCR4 antibodies were included in the appropriate cultures. The cultures were pulsed on day 5 after culture initiation with 1 μCi ³H-labeled thymidine/well (Perkin Elmer). Proliferation of cultures was measured in terms of incorporation of ³H-thymidine by reading counts in a scintillation counter (Perkin Elmer).

2 × 10⁶ luciferase-Mac-1 or 1 × 10⁷ Mac-1 cells were injected subcutaneously into the dorsolateral flank in 6-week SCID/Beige mice (Charles River). After 24 hours of injection, mice were randomly assigned into different groups and treated with 3 mg/kg of mAb1567 and mouse IgG2b (twice a week for 3 weeks) or 5 mg/kg of control-scFv-Fc, c1567-scFv-Fc, and h1567-scFv-Fc (twice a week for 4 weeks) by intraperitoneal injection. Body weight and tumor size were measured using digital caliper and Xenogen imaging. Tumor volumes were calculated as length × (width)² × 0.52. Animal care was carried out in accordance with the guidelines of Animal Care and Use Committee of Dana-Farber Cancer Institute (Boston, MA).

Data were analyzed using 2-sided unpaired Student t test. *, **, and *** indicate P < 0.05, 0.01, and 0.001, respectively. All values and bars are represented as mean ± SD.

CCR4 has 4 regions exposed at the cell surface: the NT (∼30–50 aa) and 3 extracellular domains loops (ECL, each of ∼10–30 aa), which are important for ligand binding, intracellular signaling, and other biologic functions. In this study, 2 commercially available murine anti-CCR4 mAbs, mAb1567 (R&D Systems), and 1G1 (BD Pharmingen), both generated by immunizing mouse with full-length human CCR4 (hCCR4)-expressing cells (13, 23), were initially selected for evaluation. Both mAb1567 and 1G1 showed specific binding activity in FACS analysis to hCCR4-expressing Cf2Th-CCR4 cells but not to the parental Cf2Th cells. In comparison, mAb1567 had relatively higher affinity than 1G1 under the same antibody concentrations tested (data not shown). Therefore, we selected only mAb1567 for further characterization.

Binding of mAb1567 was further tested using the CCR4⁺ Mac-1 cells by FACS analysis and the half maximal effective binding concentration (EC₅₀) is about 0.45 nmol/L (Fig. 1A). Chemotaxis inhibition assay showed that mAb1567 effectively inhibited chemotaxis of Mac-1 cells in a dose-dependent manner toward both CCR4 ligands, CCL17 and CCL22 (Fig. 1B). We next examined the epitope recognized by mAb1567, in particular, whether it recognizes solely the NT or a nonlinear conformational-dependent epitope comprising of both NT and ECL, by using hCCR4 and hCCR8 NT swapping chimeras that either contained CCR8-NT/CCR4-ECLs (Chi#1) or CCR4-NT/CCR8-ECLs (Chi#2; ref. 24). As shown in Supplementary Fig. S1A, all constructs encoding wild-type or chimeras CCR4 and CCR8 expressed similar levels on the cell surface as validated by antibody staining against the hemagglutinin tag. mAb1567 specifically recognized cell surface full-length hCCR4 but not hCCR8. It bound to Chi#1 and Chi#2 in a similar level as WT CCR4, indicating that the epitope of mAb1567 is not solely a linear epitope on NT of CCR4, rather both ECLs and NT contribute to the binding of mAb1567 with CCR4. However, the CCR4-Nt alone is also sufficient for some degree of mAb1567 binding to CCR4-Nt-Fc as determined in ELISA studies to plate-bound mAb1567 (Supplementary Fig. S1B–S1D). Moreover, mAb1567 showed high specificity for CCR4 rather than CCR5, the most similar CCR molecule to CCR4 (Supplementary Fig. S1E).

We then tested the antitumor effect of mAb1567 in vivo in a CTCL model using immunodeficient SCID/Beige mice that lack T and B cells and have defective NK function. SCID/Beige mice implanted with Mac-1 cells can efficiently form subcutaneous tumors (25). As shown in Fig. 1C, the tumor size in the mAb1567-treated group was 3- to 4-fold smaller than seen in the control group. None of the mice showed mAb1567 treatment–related toxicity.

To further understand the mechanism underlying the antitumor effect of mAb1567 seen in the SCID/Beige mice, we tested whether mAb1567 can mediate CDC and/or neutrophil-mediated ADCC effects against CCR4⁺ tumor cells in vitro. mAb1567 induced a significant lysis of Mac-1 cells in a dose-dependent manner in the presence of mouse complement as compared with the mouse IgG2b isotype control antibody (Fig. 2A). Rabbit complement was also tested and mAb1567 mediated a much more potent CDC activity and reached 80% of the target cell lysis (Fig. 2B). Next, neutrophils isolated from the SCID/Beige mice were tested in an in vitro ADCC assay. As shown in Fig. 2C, mAb1567 specifically mediated approximately 20% lysis via mouse neutrophils as compared with control at E/T (neutrophils/Mac-1) ratio of 80:1. These results show that mAb1567 can directly mediate not only CDC but also mouse neutrophil-induced ADCC activities.

To humanize mAb1567 for further preclinical studies, the cDNAs encoding the VH and VL genes from the hybridoma cell line were individually recovered by RT-PCR using primers specific for mouse antibody variable genes. The VH and VL of mAb1567 belong to mouse V_H1 (IGHV1S56*01) and V_K8 (IGKV8-27*01) families and were rearranged with the J_H1 (IGHJ1*01) and J_K2 (IGKJ2*01) segments, respectively. The cloned and rearranged VH and VL genes were then assembled as a single-chain antibody variable region fragment (scFv) using a (G₄S)₃ linker. Binding of the recombinant mAb1567 to CCR4 was verified in both scFv-Fc IgG1 minibody (c1567-scFv-Fc; Fig. 3A) and full-length chimeric IgG1 form (c1567-IgG; data not shown).

As NK cell–mediated ADCC is one of the most important mechanisms of action for immunotherapy with human IgG1 antibodies, we further tested whether recombinant mAb1567 can mediate ADCC via NK cells. Chimeric 1567 in both scFv-Fc or IgG1 forms were highly effective in killing Mac-1 cells in an in vitro ADCC assay using human PBMCs (Fig. 3B) or purified NK (CD56⁺CD16⁺) cells (Fig. 3C) from healthy donors as effector cells at different E/T ratios.

Next, the structure-guided complementarity-determining region (CDR) grafting approach was used to humanize mAb1567. Homology 3-dimensional modeling of the VH and VL chains of mAb1567 using Web Antibody Modeling program (26) was generated to known antibody structures in the PDB database. For selecting the human acceptor framework template for CDR-grafting, the VH and VL amino acid sequences of mAb1567 were separately compared with human Ab sequences in the IGBLAST database to identify the most similar human Ab and Ig germline VH and VL sequences (Fig 3D). The human VH (McAb Ctm01, PDB:1ae6H) and VL (GenBank #ABG38372) share 82% and 84% amino acid sequence homology to the VH and VL of mAb1567, respectively; the best matched human Ig germline V sequences are IGHV1-3*01 (67% homology to mAb1567-VH) and IGKV4-1*01 (83% homology to mAb1567-VL). The framework residues of mAb1567 were manually changed to the selected human framework residues to generate the humanized mAb1567 (h1567). GROMOS force field energy minimization parameter was then applied to homology model h1567 using DeepView program (27). Examination of this energy minimized homology model of h1567 was carried out to ensure that no residues had distorted geometry or steric clashes with other residues.

The h1567 sequence shown in Fig. 3D has 21 and 11 amino acid differences in the framework regions compared with the mouse VH and VL, respectively. The humanized VH and VL gene were de novo synthesized and codon-optimized for mammalian cell expression. The binding affinity of h1567 and c1567 scFv-Fcs to CCR4 was then compared by FACS with Mac-1 cells. The h1567 had approximately 2-fold decrease in binding as compared with c1567 but both are in the nanomolar range, with EC₅₀ of 2.2 and 1 nmol/L, respectively (Fig. 3A). The humanized h1567 scFv-Fc maintained potent NK-mediated ADCC killing of Mac-1 cells compared with c1567 (Fig 3C).

To test the in vivo antitumor effect of h1567, the luciferase-expressing Mac-1 cells were subcutaneously implanted into the dorsolateral flank of SCID/Beige mice, and mice were intraperitoneally treated with 5 mg/kg of control-scFv-Fc, c1567-scFv-Fc, h1567-scFv-Fc, or equivalent volumes of saline. Tumor growth in mice was monitored for luciferase intensity by IVIS imaging. All mice were sacrificed on day 28 and tumors were excised for photographing and measuring tumor weight. As shown in Fig. 4, tumors were significantly reduced in size at day 21 in c1567- and h1567-treated groups but not in control-scFv-Fc or PBS-treated groups as measured by IVIS imaging (Fig. 4A, top, and C), size of the excised tumors (Fig. 4A, bottom), tumor volume (Fig. 4B), and tumor weight (Fig. 4D).

Although the h1567 exhibited similar biologic activity as its murine counterpart in both in vitro and in vivo, the relative apparent binding affinity of h1567 is 2-fold lower than c1567 (Fig. 3A). To further affinity mature the h1567, we conducted VL-chain shuffling and alanine scanning to identify key residues in CDRs, followed by selection and screening of phage display library constructed by random mutagenesis of key residues in the CDRs (see Supplementary Method and SupplementaryFig. S2 for details).

The 2 affinity-improved h1567 variants, mAbs 1-44 and 2-3, that showed higher binding affinity to Mac-1 cells than parental h1567 with EC₅₀ of 1.47 and 1.39 nmol/L, respectively (Supplementary Fig. S2F) were further evaluated for their capacity to mediate ADCC activity using human NK cells. The result showed that improvement in ADCC activity of the h1567 variants is correlated with their binding affinity, 2-3-scFv-Fc exhibited the best human NK-mediated ADCC activity for both Mac-1 cells (Fig. 5A) and Cf2Th-CCR4 cells but not to negative control Cf2Th (Supplementary Fig. S3). Moreover, because parental mAb1567 could induce mouse neutrophil-mediated ADCC, h1567 and 2-3 were tested for human neutrophil-mediated ADCC assay and mAb2-3 showed enhanced cytotoxic activity (Fig. 5B) compared with h1567. Furthermore, slightly improved CDC activity against Mac-1 cells was also observed for both 1-44 and 2-3 variants, but more for the 2-3-scFv-Fc (Supplementary Fig. S4).

Fc engineering was also conducted on mutant antibodies 1-44 and 2-3 to further enhance ADCC activity by mutating 3 residues (S239D, A330L, and I332E) in CH2 domain, which have been shown to increase ADCC effect of human IgG1 antibody (28). As shown in Fig. 5C, ADCC-mediated by 1–44- and 2-3-scFv-mFcs was significantly enhanced as compared with their WT Fc counterparts or the WT h1567. However, as the A330L mutation in the Fc domain can ablate CDC function (29), we also tested and confirmed that CDC activity for the scFv-mFcs forms of 1-44 and 2-3 scFv-mFcs was completely abolished (Fig. 5D). Taken together, the affinity optimized variants of humanized 1567, in particular the 2-3 variant, showed improved ADCC and CDC effector functions, and NK cell–mediated ADCC activity can be further enhanced through Fc engineering.

Finally, as the majority (94%) of freshly isolated CD4⁺CD25^high Tregs from peripheral blood express high level of surface CCR4 (30) and they have been reported to migrate to tumors secreting CCL22 (31), we investigated whether Ab1567 could have an antitumor role by modulating the chemotactic recruitment and suppressive activity of human CD4⁺ Tregs. First, we confirmed that CD4⁺CD25^high Tregs migrated toward CCL22 much more effectively than CD4⁺CD25⁻ T cells (Fig. 6A). Next, using peripheral blood CD4⁺ T cells in combination with examining the Treg phenotype of the migrated cells, we confirmed that c1567 completely inhibited the migration of CD4⁺CD25^high Tregs in a transwell chemotaxis assay at concentrations greater than 2 μg/mL (Fig. 6B).

In addition, as we are unaware of any published data on the role of CCR4 in Treg function, we also examined whether 1567 engagement of CCR4 could result in modulation of Treg suppression activity in an in vitro Treg suppression assay. As shown in Fig. 6C, the proliferation of CD4⁺ T-effector cells (Teffs alone, lane 1) was inhibited by highly purified CD4⁺CD25^high Tregs (1:2 ratio) by 78% (lane 7), which is our typical Treg suppression effect on Teffs (32). Surprisingly, in the presence of c1567IgG or h1567scFv-Fc, the proliferation of Teff was stimulated to 183% and 207%, respectively (lanes 3 and 5), but there was no stimulatory effect on Treg (lanes 4 and 6). In the Treg/Teff coculture (1:2 ratio), the proliferation of Teff was inhibited directly (lane 7) and there was no reversal of this inhibition by control mAb (lane 8). Moreover, T-cell proliferation was restored with a net positive response to 258% and 221% (lanes 9 and 10) in the presence of c1567IgG or h156scFv-Fc, respectively.

The increase in CD4⁺CD25⁻ T-cell proliferation in the presence of anti-CCR4 antibodies was further observed by flow cytometry. We detected CD4⁺CD25⁻ T-cell proliferation on the basis of carboxyfluorescein succinimidyl ester fluorescence intensity of labeled CD4⁺CD25⁻ T cells in the Treg/Teff coculture. The analysis revealed that over the 7-day study, CD4⁺CD25⁻ T-cell proliferation responded to stimulation of anti-CCR4 antibodies in a time-dependent manner and was approximately 50% higher than with control antibody–treated cells that showed no proliferation (Supplementary Fig. S5). Importantly, Treg-mediated suppression of Teff proliferation occurred even in the presence of anti-CD3/CD28 costimulation but could be reversed in the presence of anti-CCR4 antibody resulting in an increased proliferative capacity of CD4⁺CD25⁻ T cells.

In this study, we humanized a mouse anti-CCR4 antibody, mAb1567, which recognizes both the NT and the ECLs of CCR4 with high affinity and inhibits migration of CCR4⁺ tumor cells toward its 2 ligands, CCL22 and CCL17. The antibody exhibited potent antitumor effect in a CTCL mouse model with the mechanisms of CDC and neutrophil-mediated ADCC likely involved (Figs. 1 and 2). The chimeric or humanized mAb1567 also showed potent CDC and human NK cell–mediated ADCC activity in vitro and therapeutic effects in vivo (Figs. 3 and 4). In addition, this antibody effectively inhibited the chemotaxis of CD4⁺CD25^high Tregs to CCL22. Interestingly, it also stimulated CD4⁺CD25⁻ cell proliferation and inhibited Tregs' immune-suppressive activity (Fig. 6 and Supplementary Fig. S5). The affinity of h1567 was further improved by employing a targeted mutagenesis strategy in combination with phage-display library selection and the affinity optimized h1567 variant mAb2-3 showed better CDC and ADCC activities against CCR4⁺ tumor cells in vitro (Fig. 5). These studies support that the affinity selected mAb2-3 may provide a novel immunotherapy option not only to directly kill the CCR4⁺ tumor cells, but also may have a role in other cancers by suppressing Treg trafficking and overcoming the suppressive effect of CCR4⁺ Tregs to enhance host antitumor immune responses.

A number of studies have been reported on a humanized anti-CCR4 antibody, KM-0761 (33, 34). These studies have not only showed KM-0761′s effect in CTCL tumor animal models and its mechanism of action, but also convincing support that CCR4 is an important antibody-based immunotherapy target for relapsed CCR4⁺ ATL or PTCL (35, 36). The KM-0761 antibody was originally raised in mouse by immunizing with the CCR4′s NT_(12–29) peptide (34, 37), it works in killing tumor cells by ADCC mainly through NK and/or macrophages (33, 38, 39), and augmentation of FcγR engagement and ADCC killing is achieved by defucosylation (34). In contrast, mAb1567 was generated with full-length CCR4, it recognizes an epitope comprising of both NT and ECLs of CCR4, and ADCC-enhanced killing is achieved through both affinity maturation and Fc engineering (29). In addition, mAb1567 showed both CDC and ADCC activity. The disparity in CDC activities between these 2 IgG1 antibodies is likely due to different recognition of the CCR4 epitope(s). CDC is often dependent on the distance between the plasma membrane and the constant region of the sensitizing antibody that mediates effector functions (40). It is possible that the CDC activity mediated by mAb1567 is related to the position and orientation of the recognized epitope, when bound, the mAb1567 can position its Fc region more proximal to the membrane to recruit complement efficiently. Another possibility is that binding to the different epitope may promote more efficient cross-linking of CCR4, and thus, increasing the binding avidity (41).

Antitumor effect mediated by cancer cell–directed antibodies can generally be attributed to ADCC, CDC, or direct antiproliferation. We found that mAb1567 did not show any inhibition of cell proliferation (data not shown). Effector cells that can mediate ADCC are NK cells, neutrophils, and monocytes/macrophages. The SCID/Biege mice used in our in vivo studies not only lack T- and B-lymphocytes but are also NK cell defective. Thus, we surmised that the antitumor activity of mAb1567 seen in SCID/Beige mice might be due to effector cells other than NK cell–mediated ADCC and/or to CDC. In vitro CDC experiments indeed showed that mAb1567 had CDC activity as compared with control Ab, although the level of lysis target Mac-1 cells was low, upto approxiamtely 20%. Most mouse strains (including the mouse stains tested here) have exceptionally low complement activity, much lower than that of humans or other animals, including rabbits, guinea pigs, and hamsters (42). CDC assay with rabbit complement indeed showed dramatically improved lysis of target cells by mAb1567 (Fig. 2B) upto approximately 80%. Furthermore, mAb1567 could mediate ADCC activity (approximately 25%) via neutrophils isolated from SCID/Beige mice. Because of limited number of neutrophils available, only a single dose of mAb1567 and isotype control mAb at 50 μg/mL was tested (Fig. 2C). These results suggest that the antitumor activity of mAb1567 in the mouse model is likely due to a combination of both CDC and through neutrophil-mediated ADCC, although it remains to be tested if mouse monocytes/macrophages may also play a role.

Many therapies for cancer are accompanied by adverse side effects and dose-dependent toxicities. The development of more effective cancer immunotherapy with better discrimination between tumor cells and normal cells is the most important goal of current anticancer research. CCR4 is universally expressed at high levels among most CTCL cells; however, a subset of normal T cells also can express CCR4. It was reported that the CCR4 is selectively expressed on T-helper 2 (T_H2) cells, but not T_H1 cells, which are the precursors of memory T cells (43). Of particular importance to the in vivo expression of CCR4 on T_H2 cells would be a concern of bystander cytotoxicity in response to anti-CCR4 antibody treatment. However, potential bystander effects can be managed as shown in the recent clinical trial reports on the anti-CCR4 mAb, KW-0761, for treatment of adult T-cell leukemia/lymphoma and PTCL, where good clinical activity was seen without serious side effects (35, 36). This may be due, at least in part, to the CCR4-expressing CTCL patient cells showing a significantly 4-fold higher expression level than T cells from healthy donors (44). This high CCR4 expression in CTCL patients may result in most anti-CCR4 antibody binding to tumor T cells but not normal T cells. We are also aware of reports that CCR4 is expressed on platelets (45, 46) but again in the KW-0761 clinical trial (35), grade 3 or 4 thrombocytopenia was only seen in 5 of 28 patients (18%) and not associated with bleeding.

CCR4 is also highly expressed on the majority (94%) of CD4⁺CD25^highFoxP3⁺ Tregs (30) that are considered the most potent inhibitors of antitumor immunity and the greatest barrier to successful immunotherapy (31, 47). CCR4 also plays an important role in Treg recruitment to the site of action. In some solid tumors including breast cancer (17, 48), ovarian cancer (18), and oral squamous cell carcinoma (49), increased numbers of recruited Tregs that are chemoattracted through the CCR4–CCL22 axis likely play a significant role in the suppression of host antitumor immunity. This trafficking pattern makes CCR4 an even more attractive therapeutic target for a broader range of tumors. mAb1567 was tested for inhibition of Treg migration and showed complete inhibition of chemotaxis of CD4⁺CD25^high Tregs (Fig. 6A). Interestingly, it appeared that mAb1567 can also abrogate the immune suppressive function of CD4⁺CD25^high Tregs while stimulating the proliferation of CD4⁺CD25⁻ Teffs (Fig. 6C and Supplementary Fig. S5). It has been proposed that Tregs exert their function by multiple suppressive mechanisms including cell–cell contact, including competitive consumption of IL-2, production of the immunosuppressive cytokines IL-10 and TGF-β, cytolysis, metabolic disruption, and modulation of the function of antigen-presenting cells (5, 47). However, CCR4 is not known to play a role in CD4⁺CD25⁻ cell proliferation and although a previous study described that CCR4^−/− Tregs are defective in suppressive activity in a model of colitis, this was due to their lack of chemotactic recruitment to the mesenteric lymph nodes because their in vitro suppressive function was intact (50). It will be important to conduct additional studies to further confirm these findings and elucidate the mechanism(s) by which anti-CCR4 mAb1567 could act through CCR4 on these cells to exercise different functions—proliferative effect on CD4⁺CD25⁻ Teffs and partial to complete abrogation of Tregs' suppression.

In summary, we have humanized and affinity maturated murine anti-CCR4 antibody 1567 and showed its in vitro and in vivo antitumor activity against CTCL cells. The mechanisms of tumor cell killing by mAb1567 seem to be multiple including CDC as well as neutrophil- and NK cell–mediated ADCC. The affinity-matured mAb2-3 promoted more potent CTCL lysis by augmentation of both NK cell–mediated ADCC and CDC activities. Remarkably, the activity of mAb1567 also extended to the normal T-cell compartment with demonstrable proliferative effect on CD4⁺CD25⁻ Teffs and abrogation of Treg suppression. These findings support further evaluation of our anti-CCR4 antibody as an immunotherapy for CTCL and other cancers where augmentation of Teff functions against the tumor cells are likely to have a therapeutic benefit.

No potential conflicts of interest were disclosed.

Conception and design: J. Sui, M. Bai, R. C. Fuhlbrigge, Q. Zhu, T. S. Kupper, W. A. Marasco

Development of methodology: D.-K. Chang, J. Sui, S. Geng, A. Muvaffak, M. Bai, R. C. Fuhlbrigge, A. S. Lo, Q. Zhu, W. A. Marasco

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Geng, M. Bai, A. Yammanuru, L. Hubbard, J. Sheehan, J. J. Campbell, Q. Zhu, T. S. Kupper

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D.-K. Chang, J. Sui, S. Geng, M. Bai, A. Yammanuru, Q. Zhu, T. S. Kupper, W. A. Marasco

Writing, review, and/or revision of the manuscript: D.-K. Chang, J. Sui, R. C. Fuhlbrigge, J. Sheehan, Q. Zhu, T. S. Kupper, W. A. Marasco

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D.-K. Chang, J. Sheehan

Study supervision: J. Sui, Q. Zhu, W. A. Marasco

The authors thank Maryam Ali, Hong Tao, and Erin M. McConocha for technical support, and National Foundation of Cancer Research (CFCR) for contribution of equipment used in this study.

This work was funded by Skin Cancer Score project 2P50CA093683 to T. S. Kupper, J. Campbell, and W. A. Marasco and NIH AI058804 to Q. Zhu.

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.