Expression of the ULBP ligands for NKG2D by B-NHL cells plays an important role in determining their susceptibility to rituximab-induced ADCC

Authors

  • Atsushi Inagaki,

    1. Department of Medical Oncology and Immunology, Nagoya City University Graduate School of Medical Sciences, Nagoya, Aichi, Japan
    Search for more papers by this author
  • Takashi Ishida,

    Corresponding author
    1. Department of Medical Oncology and Immunology, Nagoya City University Graduate School of Medical Sciences, Nagoya, Aichi, Japan
    • Department of Medical Oncology and Immunology, Nagoya City University Graduate School of Medical Sciences, 1 Kawasumi, Mizuho-chou, Mizuho-ku, Nagoya, Aichi, 467-8601, Japan
    Search for more papers by this author
    • Fax: +81-52-852-0849.

  • Hiroki Yano,

    1. Department of Medical Oncology and Immunology, Nagoya City University Graduate School of Medical Sciences, Nagoya, Aichi, Japan
    Search for more papers by this author
  • Toshihiko Ishii,

    1. Department of Medical Oncology and Immunology, Nagoya City University Graduate School of Medical Sciences, Nagoya, Aichi, Japan
    Search for more papers by this author
  • Shigeru Kusumoto,

    1. Department of Medical Oncology and Immunology, Nagoya City University Graduate School of Medical Sciences, Nagoya, Aichi, Japan
    Search for more papers by this author
  • Asahi Ito,

    1. Department of Medical Oncology and Immunology, Nagoya City University Graduate School of Medical Sciences, Nagoya, Aichi, Japan
    Search for more papers by this author
  • Masaki Ri,

    1. Department of Medical Oncology and Immunology, Nagoya City University Graduate School of Medical Sciences, Nagoya, Aichi, Japan
    Search for more papers by this author
  • Fumiko Mori,

    1. Department of Medical Oncology and Immunology, Nagoya City University Graduate School of Medical Sciences, Nagoya, Aichi, Japan
    Search for more papers by this author
  • Jianmin Ding,

    1. Department of Medical Oncology and Immunology, Nagoya City University Graduate School of Medical Sciences, Nagoya, Aichi, Japan
    Search for more papers by this author
  • Hirokazu Komatsu,

    1. Department of Medical Oncology and Immunology, Nagoya City University Graduate School of Medical Sciences, Nagoya, Aichi, Japan
    Search for more papers by this author
  • Shinsuke Iida,

    1. Department of Medical Oncology and Immunology, Nagoya City University Graduate School of Medical Sciences, Nagoya, Aichi, Japan
    Search for more papers by this author
  • Ryuzo Ueda

    1. Department of Medical Oncology and Immunology, Nagoya City University Graduate School of Medical Sciences, Nagoya, Aichi, Japan
    Search for more papers by this author

Abstract

Antibody-dependent cellular cytotoxicity (ADCC) is a major antitumor mechanism of action of therapeutic monoclonal antibodies (mAbs). The aim of this study was to identify tumor-associated factors which determine susceptibility to rituximab-induced ADCC. Thirty different CD20+ non-Hodgkin lymphoma cell lines were phenotyped for characteristics such as level of expression of NKG2D ligands, and the influence thereof on susceptibility to rituximab-induced ADCC was established. The present study demonstrated that tumor cell susceptibility to rituximab-induced ADCC was determined by 3 major tumor-associated factors: (i) the amount of the target molecule, CD20; (ii) the amount of the ligands for inhibitory killer Ig-like receptors, major histocompatibility complex class I; and (iii) the amounts of some of the NKG2D ligands, especially UL16-binding protein (ULBP) 1–3. The importance of the ULBPs was confirmed using antibody blockade. In conclusion, this is the first report to show the importance for rituximab-induced ADCC of ULBPs expressed on tumor cells. The ULBPs could be valuable diagnostic biological markers and significant targets for immunotherapy to improve efficacy not only of rituximab but also of other therapeutic mAbs. © 2009 UICC

The use of therapeutic monoclonal antibody (mAb) for the treatment of cancer has evolved into a promising approach over the last several years, as exemplified by the great success of mAb such as rituximab,1, 2 trastuzumab,3 bevacizumab4, 5 and cetuximab.6 Antibodies of the human IgG1 isotype are commonly used for therapeutic applications as they can mediate multiple effector functions including antibody-dependent cellular cytotoxicity (ADCC), complement-dependent cytotoxicity (CDC), and direct apoptosis induction.7–9 ADCC is triggered following binding of the antibody Fc region to the Fcγ-receptor (FcγR) on effector cells, and is believed to represent the major in vivo antitumor mechanism of rituximab, an anti-CD20 chimeric IgG1 mAb.10–12 In particular, Carton et al.,12 found that follicular lymphoma patients homozygous for FcγRIIIa-158VV, which has the highest affinity for the IgG1 Fc region, have the best clinical and molecular responses to rituximab. The clear role for FcγR-bearing effectors in mediating the response to rituximab in clinical settings further demonstrates the importance of ADCC. Therefore, a better understanding of ADCC will allow the development of novel, more effective treatment strategies using not only rituximab but also other therapeutic mAbs such as trastuzumab and cetuximab, and may help overcome the resistance which can develop against the effects of these therapeutic mAbs. However, although the clinical efficacy of rituximab in patients with B-cell non-Hodgkin lymphoma (B-NHL) is no longer in question,1, 2 its mechanism of action has yet to be fully elucidated. In particular, the tumor-associated factors which determine susceptibility to rituximab-induced ADCC have not been identified. Therefore, the aim of this study was to identify those tumor-associated factors. Because of the importance of natural killer (NK) cells for ADCC,13–16 we especially focused on the molecules expressed by the tumor cells that interact with NK cells.

Abbreviation

ADCC, antibody-dependent cellular cytotoxicity; BL, Burkitt lymphoma; CDC, complement-dependent cytotoxicity; CLL, chronic lymphocytic leukemia; DLBCL, diffuse large B-cell lymphoma; mAb, monoclonal antibody; MCL, mantle cell lymphoma; MFI, mean fluorescence intensity; MHC, major histocompatibility complex; MICA, major histocompatibility complex (MHC) class I chain-related genes A; MICB, major histocompatibility complex (MHC) class I chain-related genes B; NHL, non-Hodgkin lymphoma; ULBP, UL16-binding protein.

Material and methods

Human PBMC and serum

All donors provided written informed consent before sampling, according to the Declaration of Helsinki, and the present study using human samples was approved by the institutional review board of Nagoya City University Graduate School of Medical Sciences.

Human B-NHL cell lines

All cell lines analyzed in this study were CD20+ human B-NHL cell lines, as follows: diffuse large B-cell lymphoma (DLBCL) cell lines (OCI-Ly-1, -2, -7, -8, Farage, DB, SU-DHL-7, -10, VAL, WSU-DLCL2, NCU-L4, RL, Toledo, HT, Karpas-422, Pfeiffer and KIS-1); Burkitt lymphoma (BL) cell lines (NCU-L1, -L3, Raji, Daudi, BL-53, PA682, BL-60, BL-41 and P3HR1); chronic lymphocytic leukemia (CLL) cell lines (MEC-2, HANDMAN, Mo1043 and WAC-CD5+); and a single mantle cell lymphoma (MCL) cell line (Sp-49).

Antibodies

Mouse mAbs to human UL16-binding protein (ULBP)1 (clone: 170818, IgG2a), ULBP2 (165903, IgG2a), ULBP3 (166510, IgG2a), Allophycocyanin (APC)-conjugated antihuman major histocompatibility complex (MHC) class I chain-related genes A (MICA) (159227, IgG2b), MHC class I chain-related genes B (MICB) (236511, IgG2b), Phycoerythrin (PE)-conjugated antihuman NKG2D (149810, IgG1) and appropriate isotype-matched control mAbs were purchased from R&D Systems, (Minneapolis, MN). PE-conjugated goat F(ab′)2 antimouse IgG, and recombinant human NKG2D/Fc chimera were also purchased from R&D Systems, Inc. PE-conjugated antihuman CD20 (2H7, IgG2bκ), CD56 (B159, IgG1κ), CD48 (Tá145, IgMκ), fluorescein isothiocyanate (FITC)-conjugated antihuman human leukocyte antigen (HLA)-A, B, C (G46–2.6, IgG1κ), CD46 (E4.3, IgG2aκ), CD55 (IA10, IgG2aκ), CD59 (p282, IgG2aκ), CD69 (FN50, IgG1κ) and appropriate isotype-matched control mAbs were purchased from BD Biosciences (San Jose, CA). PE-conjugated antihuman HLA-G (MEM-G/9, IgG1) mAb was purchased from Acris Antibodies GmbH (Himmelreich, Germany). FITC-conjugated goat antimouse Ig-specific polyclonal antibody was also purchased from BD Biosciences. Anti-cathepsin B mAb (Ab-1, IgG2a) was purchased from CALBIOCHEM (San Diego, CA). Rituximab was kindly provided by Chugai Pharmaceutical Co. (Tokyo, Japan).

Flow cytometric analysis

After blocking surface Fc receptors with heat-inactivated human serum, flow cytometric analyses were performed as described previously.17, 18 For determining the total level of expression of all NKG2D ligands on the NHL cell lines in their entirety, direct binding of NKG2D/Fc chimeric molecules was assessed using a PE-conjugated anti-NKG2D mAb.19 After staining, the cells were analyzed on a FACScalibur (Becton Dickinson, San Jose, CA). The level of expression of each molecule is presented as the mean fluorescence intensity (MFI) ratio, i.e., the MFI with test mAb divided by that of its isotype control mAb.

ADCC assay

To standardize effector cell function in repeat assays and minimize differences caused by varying degrees of HLA incompatibility, PBMC obtained from 4 healthy individuals were aliquotted and stored frozen at −80°C until use, so that the same batches could be employed in every experiment. Standard 4-hr 51Cr release assays were performed with or without rituximab at a final concentration of 10 μg/ml, and at the effector: target ratio of 50:1, as described previously.17 Percent target cell lysis mediated by each individual PBMC is presented as the average of triplicate determinations. The cytotoxicity against B-NHL cell lines in the presence (rituximab-induced ADCC) or absence of rituximab is then expressed as the mean percent lysis by the 4 individual PBMC. In some experiments, the target B-NHL cell lines were pre-incubated with antibodies to NKG2D ligands, ULBP1-3, to block receptor-ligand interactions, or with isotype control mAb for 1 hr at 37°C. In this experiment using mAb blockade, effector PBMC were freshly isolated from a healthy individual, and percent ADCC-mediated lysis presented as mean ± standard deviation (SD) of triplicates.

CDC assay

CDC assays were performed using pooled human serum as a complement source, with rituximab at a final concentration of 10 μg/ml, as described previously.17

Analysis of NK cell activation

Isolation of NK cells from PBMC using NK Cell Isolation Kits (Miltenyi Biotec, Bergisch Gladbach, Germany) was as previously described.20 The isolated NK cells and an equal number of target cells were dispensed into round-bottom 96-well plates. Rituximab (final concentration, 10 μg/ml) was added to each well and incubated at 37°C, 5% CO2 for 24 hr, and double-staining with anti-CD56 and anti-CD69 mAbs was performed. Phorbol 12-myristate 13-acetate (PMA) and ionomycin were purchased from CALBIOCHEM. CD69 expression is a reliable activation marker for NK cells21; hence, CD69-positive NK cells are defined here as activated NK cells. The percent activated NK cells was calculated according to the following formula: the percent activated NK cells = 100 × (CD69 and CD56 double-positive cells)/(CD56-positive cells).

Statistical analysis

Correlations between levels of expression of each molecule (CD20, MHC-class I, total NKG2D ligands, ULBP1-3, MICA, MICB, CD48, HLA-G, cathepsin B, CD46, CD55 and CD59) in B-NHL cell lines and the percent of cell lysis mediated by rituximab-induced ADCC/CDC or that by PBMC alone were assessed using the Spearman rank correlation coefficient. The significance of changes in the percent activated NK cells in the absence or presence of rituximab or target cells was determined using the Wilcoxon signed rank test. Data were analyzed with the aid of StatView software (version 5.0, SAS Institute, Cary, NC). In this study, p < 0.05 was considered significant.

Results

Correlations between CD20 expression and lysis by rituximab-induced ADCC/CDC in B-NHL cell lines

Flow cytometric analyses of CD20 expression in the 30 B-NHL cell lines tested in the present study are shown in Table I. The original disease of each patient from whom the cell line was derived, the CD20 MFI ratio, and rituximab-induced ADCC/CDC data are also shown in Table I. There was a significant correlation between the level of CD20 expression (determined as the CD20 MFI ratio) and rituximab-induced ADCC-mediated cell lysis, but not rituximab-induced CDC (coefficients = 0.511, p = 0.0060, Fig. 1, upper panel; and 0.232, p = 0.2106, Fig. 1, lower left panel, respectively).

Figure 1.

Correlations between CD20, CD55 or CD59 expression and lysis of B-NHL cell lines by rituximab-induced ADCC/CDC. ADCC (upper panel) or CDC (lower 3 panels) in the presence of 10 μg/ml rituximab against 30 different CD20+ B-NHL cell lines determined by 51Cr release assays. Effector PBMC for ADCC assays were obtained from 4 healthy volunteers and stored frozen at −80°C until use. Percent lysis was calculated as described in Materials and Methods. Y-axis: % cell lysis by ADCC (upper panel) and CDC (lower panels). X-axis: MFI ratio of CD20 (upper and lower left panels), CD55 (lower middle panel) and CD59 (lower right panel). Each dot plot in each panel represents 1 cell line, and ○ indicates diffuse large B-cell lymphoma; ● Burkitt lymphoma; Δ CLL; ▴ Mantle cell lymphoma. The coefficients and P values assessed by Spearman rank correlation coefficient testing are indicated in each panel.

Table I. CD20 Expression in the B-NHL Cell Lines and the Percent Lysed Cells by Rituximab Induced ADCC/CDC
Cell lineCD20- MFI+/−The percentlysed cells by ADCC/CDC
Diffuse large B-cell lymphoma (DLBCL)
Ly-2chemical structure image42.5/75.5
Ly-1chemical structure image38.0/24.7
Ly-7chemical structure image38.0/83.5
Ly-8chemical structure image36.3/8.0
Faragechemical structure image36.1/19.8
DBchemical structure image32.6/5.8
SU-DHL-10chemical structure image31.8/49.3
VALchemical structure image31.2/29.0
WSU-DLCL2chemical structure image28.7/74.8
NCU-L4chemical structure image26.2/75.8
RLchemical structure image25.3/37.7
SU-DHL-7chemical structure image22.3/21.5
Toledochemical structure image20.4/0.4
HTchemical structure image18.6/13.8
Karpas-422chemical structure image17.5/42.7
Pfeltterchemical structure image15.1/69.0
KIS-1chemical structure image14.6/1.5
Burkitt lymphoma (BL)
NCU-L3chemical structure image34.4/13.7
Rajichemical structure image29.6/3.8
Daudlchemical structure image24.3/82.1
NCU-L1chemical structure image22.9/41.3
BL-53chemical structure image22.7/77.1
PA682chemical structure image16.3/34.8
BL-60chemical structure image12.9/34.9
BL-41chemical structure image11.8/4.3
Chronic lymphocytic leukemia (CLL)
MEC-2chemical structure image25.8/20.9
HANDMANchemical structure image22.5/9.9
Mo1043chemical structure image20.8/5.1
WAC-CD5+chemical structure image19.8/9.1
Mantle cell lymphoma (MCL)
Sp-49chemical structure image20.1/66.3

Correlations between expression of the complement inhibitors CD46, CD55 or CD59 and cell lysis by rituximab-induced ADCC/CDC in B-NHL cell lines

There were significant inverse correlations between CD55 and CD59 expression levels and rituximab-induced CDC in the B-NHL cell lines [coefficient = −0.639, p = 0.0006 (Fig. 1, lower middle panel), coefficient = −0.552, p = 0.0030 (Fig. 1, lower right panel), respectively]. However, there were no significant correlations between CD46 expression and rituximab-induced CDC (coefficient = −0.214, p = 0.2490) (data not shown). As expected, there were no significant correlations between rituximab-induced ADCC and CD55 (coefficient = −0.071, p = 0.6996), CD59 (coefficient = 0.331, p = 0.0746), or CD46 (coefficient = −0.343, p =0.0648) (data not shown) levels in these B-NHL cell lines.

Correlations between MHC class I expression and rituximab-induced ADCC/CDC

There was a significant inverse correlation between MHC class I expression levels and percent lysis by rituximab-induced ADCC (coefficient = −0.565, p = 0.0023) (Fig. 2, upper left panel), and that by PBMC alone, which is almost entirely due to NK cell-mediated lysis,13, 22 (coefficient = −0.473, p = 0.0109) (data not shown) in these B-NHL cell lines. As expected, there was no significant correlation between MHC class I expression level and rituximab-induced CDC (coefficient = −0.337, p = 0.0693) (Fig. 2, lower left panel).

Figure 2.

Correlations between the expression of MHC class I or total NKG2D ligand and rituximab-induced ADCC in B-NHL cell lines. ADCC (upper panels) or CDC (lower panels) in the presence of 10 μg/ml rituximab against 30 different CD20+ B-NHL cell lines determined by 51Cr release assays. Y-axis: % lysis. X-axis: MFI ratio of MHC class I (left panels) and total NKG2D ligand (right panels). Each dot plot in each panel represents 1 cell line in the same manner as in Figure 1. The coefficients and P values assessed by Spearman rank correlation coefficient testing are indicated in each panel.

Correlations between the level of expression of all NKG2D ligands and rituximab-induced ADCC/CDC

Significant correlations were found between total NKG2D ligand expression levels and the degree of rituximab-induced ADCC (coefficient = 0.459, p = 0.0134) (Fig. 2, upper right panel), and that of lysis by PBMC alone, namely NK cell-mediated lysis (coefficient = 0.469, p = 0.0115) (data not shown) in B-NHL cell lines. Again, as expected, there were no significant correlations between rituximab-induced CDC lysis and the level of NKG2D ligand expression (coefficient = 0.086, p = 0.6420) (Fig. 2, lower right panel).

Correlations between the levels of the individual NKG2D ligands ULBP1–3, MICA or MICB and rituximab-induced ADCC/CDC

Significant correlations were found between ULBP1, 2 or 3 expression levels and the degree of rituximab-induced ADCC [coefficient = 0.443, p = 0.0170 (Fig. 3, upper left panel), coefficient = 0.413, p = 0.0263 (Fig. 3, upper middle panel), or coefficient = 0.391, p = 0.0353 (Fig. 3, upper right panel), respectively], but not between MICA and MICB expression and rituximab-induced ADCC of these targets [coefficient = −0.051, p = 0.7856 (Fig. 2, lower left panel), coefficient = −0.144, p = 0.4395 (Fig. 2, lower right panel), respectively]. Significant correlations were also found between ULBP1 and 2 expression levels and the degree of NK cell-mediated lysis (coefficient = 0.384, p = 0.0385, coefficient = 0.530, p = 0.0043, respectively; data not shown), but not between ULBP3, MICA and MICB expression and the NK cell-mediated lysis of these targets (coefficient = 0.278, p = 0.1348, coefficient = −0.060, p = 0.7481, coefficient = −0.084, p = 0.6522, respectively; data not shown). Once more, there were no significant correlations between rituximab-induced CDC lysis and the level of expression of ULBP1 (coefficient = 0.202, p = 0.2761), ULBP2 (coefficient = −0.270, p = 0.1455), ULBP3 (coefficient = −0.009, p = 0.9608), MICA (coefficient = −0.001, p = 0.9952) or MICB (coefficient = −0.047, p = 0.7994) (data not shown). We also investigated correlations between the NKG2D ligands in their entirety, and each NKG2D ligand separately. Total NKG2D ligand expression was significantly correlated with the level of expression of ULBP1 (coefficient = 0.773, p < 0.0001), ULBP2 (coefficient = 0.463, p = 0.0126), ULBP3 (coefficient = 0.513, p = 0.0057), but not MICA (coefficient = 0.087, p = 0.6394) or MICB (coefficient = −0.084, p = 0.6505) (data not shown).

Figure 3.

Correlations between the expression of the NKG2D ligand, ULBP1–3, or MICA/B and rituximab-induced ADCC in B-NHL cell lines. ADCC in the presence of 10 μg/ml rituximab against 30 different CD20+ B-NHL cell lines determined by 51Cr release assays. Y-axis: % lysis. X-axis: MFI ratio of ULBP1 (upper left panel), ULBP2 (upper middle panel), ULBP3 (upper right panel), MICA (lower left panel), and MICB (lower right panel). Each dot plot in each panel represents 1 cell line in the same manner as in Figure 1. The coefficients and p values assessed by Spearman rank correlation coefficient testing are indicated in each panel.

Correlations between the expression of the 2B4 (CD244) ligand CD48 and rituximab-induced ADCC/CDC

No significant correlations were found between the level of expression of CD48 and the percent of cells lysed by rituximab-induced ADCC (coefficient = 0.319, p = 0.0855), and that of NK cell-mediated lysis (coefficient = 0.308, p = 0.0970) in the B-NHL cell lines. Unexpectedly, however, there was a significant inverse correlation between CD48 levels and rituximab-induced CDC (coefficient = −0.481, p = 0.0096) (data not shown). Like CD55 and CD59, CD48 is also a glycosylphosphatidylinositol (GPI)-linked membrane protein. In the present study, significant correlations were also found between levels of expression of CD48 and CD55, and between CD48 and CD59 in these B-NHL cell lines (coefficients = 0.454, p = 0.0146; and 0.743, p < 0.0001, respectively; data not shown).

Correlations between HLA-G or cathepsin B expression and rituximab-induced ADCC/CDC

No significant correlations were found between HLA-G or cathepsin B expression and the percent of cells lysed by rituximab-induced ADCC (coefficient = 0.019, p = 0.9189, and coefficient = 0.071, p = 0.7022, respectively) or CDC (coefficient = 0.033, p = 0.8574, and coefficient = 0.016, p = 0.9303, respectively) in these B-NHL cell lines (data not shown).

Inhibition of rituximab-induced ADCC in B-NHL cell lines by mAbs to ULBP1–3

The significant correlations found between the expression of ULBPs and percent of B-NHL cells lysed in rituximab-induced ADCC in the present study prompted us to assess the role of these molecules using mAbs to ULBP1–3 to block receptor-ligand interactions. Flow cytometric analyses of ULBP1–3 on the B-NHL cell lines OCI-Ly-1, -8, Farage, and NCU-L3 are shown in Figure 4a. Expression of MHC class I by these cell lines is also shown. In the presence of mAb to ULBP1, 2 or 3, a decrease of rituximab-induced ADCC was observed. A combination of mAbs to ULBP1 and 2 reduced ADCC against OCI-Ly-1 most effectively (percent lysis in the presence of both mAbs together compared to isotype control was 25.0 ± 1.7% compared to 58.4 ± 3.0%). This effect was also seen with OCI-Ly-8 (41.8 ± 2.2% and 67.3 ± 4.6%, respectively), as well as NCU-L3 (24.1 ± 0.3% and 44.3 ± 1.4%, respectively) (Fig. 4b), which express both ULBP1 and 2 molecules. In the presence of mAb to ULBP1, no decrease of rituximab-induced ADCC was observed with the Farage line, which does not express ULBP1. However, employing mAb to either ULBP2 or 3 did result in a decrease of lysis mediated by rituximab-induced ADCC against this line (percent lysis in the presence of isotype control mAb, anti-ULBP2 and 3 mAbs was 64.3 ± 7.9%, 45.5 ± 0.3% and 53.3 ± 3.7%, respectively) (Fig. 4b).

Figure 4.

Inhibition of rituximab-induced ADCC with mAbs to ULBP1–3 in B-NHL cell lines. (a) Expression of MHC class I and ULBP1–3 on the CD20+ NHL cell lines OCI-Ly-1, -8, Farage, and NCU-L3. Each cell line was stained with fluorescence-conjugated mAb (blank histogram) or isotype control mAb (filled histograms). (b) ADCC against each cell line was measured by a standard 4 hr 51Cr release assay in the presence of effector PBMC obtained from a healthy volunteer. The blocking mAbs against ULBP1 & 2 (final concentration: 25 μg/ml, each) or ULBP1 (50 μg/ml), ULBP2 (50 μg/ml) or ULBP3 (50 μg/ml), or isotype control mAb (50 μg/ml) were added to the ADCC assay. Percent cell lysis is presented as the mean ± SD of triplicates. Each result represents 3 independent experiments. N.E., not evaluated.

NK cell activation in the presence of rituximab

We next examined NK cell activation during rituximab-induced ADCC. The target cell line in this experiment was the Burkitt line P3HR1, expressing CD20, MHC class I and ULBP1–3 (Fig. 5a). Flow cytometric analysis of NK cell activation from1 of 7 healthy individuals (donor 1) is shown in Figure 5b. The mean percent activated NK cells from all 7 donors without any stimulation was 6.0 ± 4.3% (mean ± SD), which increased in a statistically significant manner to 8.5 ± 3.5% in the presence of rituximab only (Figs. 5b and 5c, p = 0.0180). After incubation together with target cells, the fraction of activated NK cells was also significantly increased compared to NK cells alone (17.6 ± 10.4%; p = 0.0180) (Figs. 5b and 5c). However, when NK cells were mixed with the target cells in the presence of rituximab, much more dramatic activation (72.8 ± 21.3%) was observed (Figs. 5b and 5c). This value was nearly as great as the positive controls, at 90.2 ± 8.9% (data not shown) [purified NK cells stimulated by PMA (final concentration 100 ng/ml) and ionomycin (final concentration 1.0 μg/ml) for 24 hr to confirm the upregulation of CD69].

Figure 5.

NK cell activation on mixing with target cells in the presence of rituximab. (a) Burkitt lymphoma cell line, P3HR1, express both CD20 and MHC class I. (b) NK cell activation determined by CD56 and CD69 expression. CD69 levels in NK cells mixed with or not mixed with the target cell (P3HR1) for 24 hr in the presence or absence of 10 μg/ml rituximab. Control: NK cells stimulated by PMA and ionomycin for 24 hr to confirm the up-regulation of CD69. The percent activated NK cells is indicated in the upper-right of each lower panel. This result represents NK cell activation obtained from donor 1. (c) A comparison of the percent activated NK cells from 7 healthy individuals incubated with or without target cells (P3HR1) in the presence or absence of 10 μg/ml rituximab for 24 hr. Significant differences between the groups are indicated in the graph by # (p < 0.05). The means ± SD value of each group are indicated below.

Discussion

The present study demonstrated that tumor cell susceptibility to rituximab-induced ADCC was determined by 3 major tumor-associated factors: first, the amount of the target molecule CD20 on the cell surface; second, the amount of ligand for inhibitory killer Ig-like receptors (KIRs), i.e., MHC class I; and, finally, the amounts of some of the ligands of the activating receptor NKG2D, ULBP1–3, especially ULBP1 & 2. Of these 3 factors, the observation that ULBP expression by B-NHL cells determines their susceptibility to rituximab-induced ADCC is a significant novel finding.

The present observation that the level of expression of the target molecule, CD20, in B-NHL cell lines determined their susceptibility to rituximab-induced ADCC is not unexpected, and is consistent with a previous report using mouse EL4 cells compared to clones expressing different levels of exogenous human CD20.20 With respect to CDC, the lysis by rituximab-induced CDC was found to correlate with the level of expression of CD55 and CD59, but not CD20. These results are consistent with the study reported by Golay et al.,23 suggesting that susceptibility to rituximab-induced CDC is determined by complement inhibitors such as CD55 and CD59, rather than the amount of the target antigen.

It is generally accepted that cells are resistant to NK cell-mediated lysis due to expression of MHC class I molecules, which act as ligands of NK inhibitory receptors.24 For this reason, we examined MHC class I expression on NHL cell lines, and confirmed a significant inverse correlation between the amount of MHC class I expressed and the degree of rituximab-induced ADCC. The present observations support the proposed enhanced ADCC approach with blocking NK cell inhibitory receptors by Binyamin et al.25 However, positive triggering of NK cells depends largely on the activating receptor NKG2D which recognizes ULBPs as well as MICA and MICB nonclassical MHC molecules in humans.26 In addition, the 2B4 molecule represents another type of activating NK cell receptor, the ligand of which, CD48, is widely expressed on hematopoietic cells.27 Because NKG2D and 2B4 are thought to play important roles in immune surveillance against tumors, we quantified the level of expression of all of the NKG2D ligands and of CD48 by these NHL cell lines. As expected, we confirmed significant correlations between NKG2D ligand expression and rituximab-induced ADCC, but not between CD48 expression and ADCC, against the B-NHL cell lines. The present observations are consistent with the report by Fischer et al.,14 that the major fraction of NK cells acting as ADCC effectors in response to rituximab are NKG2D-positive. Therefore, we next focused on individual NKG2D ligands, and confirmed significant correlations between ULBP1–3 expression and rituximab-induced ADCC, but not between MICA or MICB expression and ADCC against the B-NHL cell lines. Frequent expression of ULBPs on B-NHL cell lines in the present study is consistent with the observations that ULBPs were highly expressed on B cells but were not detectable on T and NK cells among healthy hematopoietic cells.28 In the present study, the correlation between ULBP3 expression and rituximab-induced ADCC was weaker than for ULBP1 or 2 expression. In addition, significant correlations were observed between ULBP1 and 2 expression levels and the NK cell-mediated lysis, but not between ULBP3 and the NK cell-mediated lysis against these targets. These findings are consistent with reports that ULBP3 binds much more weakly to human NKG2D than does ULBP1 and ULBP2,29 and that in cytotoxicity assays ULBP3-transfected human target cells deliver a much weaker activating signal to human NK cells than ULBP1 and ULBP2.30 Regarding the other NKG2D ligands, MICA/B, it has been reported that MIC molecules are expressed mainly on solid tumors rather than leukemia/lymphoma.31, 32 In fact, in the present study, only a small number of B-NHL cell lines demonstrated significant expression of MICA/B; for this reason, neither MICA nor MICB expression correlated significantly with the total level of NKG2D ligand expression on the 30 B-NHL cell lines. This might be one explanation why MICA/B expression did not correlate significantly with rituximab-induced ADCC.

Unexpectedly, the present study documented significant inverse correlations between CD48 expression and rituximab-induced CDC in these B-NHL cell lines, although CD48 has no complement inhibitory functions. The close correlations between the expression of CD48 and CD55, and between CD48 and CD59 observed in the present study suggest that CD48 might have an indirect inverse correlation with CDC.

HLA-G is a nonclassical MHC class I antigen with very little sequence variability. Because HLA-G exerts multiple immune-regulatory functions such as inhibition of NK cell cytolysis, its expression on tumor cells may favor their escape from antitumor immune responses, thus allowing tumor progression.33, 34 These findings regarding HLA-G prompted us to examine its expression in NHL cell lines, but no significant correlations with rituximab-induced ADCC in the B-NHL cell lines were observed. Perforin, exclusively expressed by NK cells and cytotoxic T lymphocytes, is essential for inducing tumor cell apoptosis.35, 36 Cathepsin B, which readily cleaves perforin, can associate with some tumor cell surfaces and maintain proteolytic activity in the extracellular environment.37, 38 In addition, surface cathepsin B provides a self-protection mechanism for cytotoxic lymphocytes.39 These findings prompted us to examine surface cathepsin B expression levels in NHL cell lines, but again, no significant correlations with ADCC were found.

The present observations led us to focus on ULBPs, and we confirmed here their significant role in modulating susceptibility of B-NHL cells to rituximab-induced ADCC, using blocking mAbs in 51Cr release assays. As expected, a reduction of rituximab-induced ADCC in the presence of mAb against these ULBP was observed. These results provide evidence of a possible mechanism whereby ULBPs expressed on tumor cells activate effector NK cells via increased binding to NKG2D, leading to enhanced ADCC. In addition, the present findings with the OCI-Ly-8, Farage and NCU-L3 lines, which express MHC class I molecules, are consistent with the report that ULBPs function to transduce a dominant stimulatory signal to NK cells, overcoming an inhibitory signal generated by MHC class I engagement of KIRs.29 Among the 3 major tumor-associated factors determining their susceptibility to rituximab-induced ADCC determined here, 2 factors, MHC class I and ULBPs, are molecules that interact with NK cells, and in fact, with the exception of ULBP3, their expression levels were also significantly associated with the degree of NK cell-mediated lysis. These present observations indicating the actual importance of NK cell-associated molecules for rituximab-induced ADCC led us to investigate the manner of NK cell activation during ADCC. We have demonstrated here that FcγR stimulation of NK cells via interactions solely with the Fc portion of rituximab did activate them, and although this was relatively weak, it was statistically significant. In addition, combining this FcγR stimulation with the activation of NK cells via interactions with target tumor cells extremely potentiated NK cell activation. These findings are consistent with the recent report by Bryceson et al.,40 that resting NK cell cytotoxicity was induced by FcγR stimulation alone, and that NK cell activation mediated by FcγR stimulation synergized with stimulation mediated by activation via NK cell receptors such as NKG2D and 2B4. Collectively, the present study demonstrated the key mechanism of ADCC, which is the target-specific cytotoxicity mediated by highly activated NK cells. The target specificity of ADCC is surely dependent on antibody-antigen interaction. Highly NK cell activation is due to the combination of FcγR stimulation via the Fc portion of the therapeutic antibody, together with stimulation of activating NK cell receptors via their ligands expressed on the tumor cells, such as ULBPs.

Several approaches to stimulate effector cell function have been suggested to argument ADCC, including the defucosylation of therapeutic mAb and combination treatment with immunomodulatory cytokines, which we have been investigating.17, 22, 41–44 The present study emphasizes the importance of tumor cell expression of NKG2D ligands, especially ULBPs, for efficient ADCC. This finding will allow the rational design of more effective immunotherapy, including antibody-based therapy. For instance, all-trans-retinoic acid upregulates surface expression of ULBP3 and MICA on B-CLL cells,45 and the histone deacetylase inhibitor, valproate, can increase cell surface expression of MICB in hepatoma cells.46 This upregulation of NKG2D ligands on tumor cells led to enhancement of tumor cell killing by cytotoxic lymphocytes via NKG2D. Treatment with therapeutic mAb which induce ADCC together with agents that upregulate NKG2D ligands offers a novel combined treatment approach worthy of pursuit. Indeed, von Strandmann et al.,47 have already recognized the importance of ULBP2 and designed a novel recombinant bispecific protein (ULBP2-BB4) for therapeutic use. ULBP2 binds NKG2D on NK cells, and the BB4 moiety binds to CD138. The ULBP2-BB4 construct enhanced NK-mediated lysis of CD138+ multiple myeloma cells. Our present findings also indicate that this approach targeting ULBP2 should be fruitful.

In conclusion, this is the first report to show that expression of ULBP on tumor cells is important for rituximab-induced ADCC. NKG2D ligands, especially ULBPs, could be valuable targets for immune therapy to improve the efficacy not only of rituximab but also other therapeutic antibodies. In addition, analysis of ULBP expression by B-NHL tumor cells would provide us with important clinical information predicting the response to rituximab therapy, and determining a suitable treatment strategy for patients with CD20-positive B-NHL.

Acknowledgements

The authors thank Ms. Chiori Fukuyama for her excellent technical assistance. Grant support: Grants-in-Aid for General Scientific Research (No. 19390266, R. Ueda, No. 80405183, T. Ishida), and for Scientific Research on Priority Areas (No. 17016065 and 16062101, R. Ueda) from the Ministry of Education, Culture, Science, Sports, and Technology, Japan; Grants-in-Aid for Cancer Research from the Ministry of Health, Labor, and Welfare, Japan (No. 17S-1, and 17-16, S. Iida, No. 19-8, T. Ishida).

Ancillary