Diffuse large B-cell lymphoma (DLBCL) is the most common subtype of non-Hodgkin's lymphoma in adults. It is generally treated by a combination of chemotherapy and CD20-specific mAbs, such as rituximab, which act, at least partially, by activating antibody-dependent cell-mediated cytotoxicity (ADCC). ADCC involves NK cells, particularly the CD56dim NK-cell subset expressing CD16, the low affinity Fcγ receptor. Here, we show that CD16 expression levels are decreased in a cohort of 36 newly diagnosed DLBCL patients compared with those in 20 healthy controls (HCs). CD137, a co-stimulatory molecule expressed on activated NK cells, was also expressed at lower levels in patients compared with controls. Cells sampled from our cohort also showed severely reduced degranulation activity when challenged with rituximab-coated tumor cells, which could not be corrected by stimulation with high doses of IL-2. These results suggest that rituximab-induced NK-cell ADCC could be defective in some DLBCL patients at diagnosis. These patients should be closely monitored and attempts made to improve their NK-cell function.
Among non-Hodgkin's lymphoma subtypes in adults, diffuse large B-cell lymphoma (DLBCL) is the most common. Treatment of DLBCL combines specific antibodies (e.g. rituximab) with chemotherapy (R-CHOP) . Rituximab is a humanized IgG1 mAb specific for CD20, which is expressed on healthy B cells and overexpressed on malignant B cells. Rituximab induces several mechanisms including direct signaling, complement-mediated cyto-toxicity, and antibody-dependent cell-mediated cytotoxicity (ADCC) . Rituximab-mediated ADCC mainly triggers innate effector cells, such as NK cells. Few studies have specifically focused on NK-cell activity in the context of B-cell lymphoma [3-5], even though many studies indicate that patients with enhanced antitumor immune responses may have better clinical outcomes [6-8]. In DLBCL patients, a lower absolute NK-cell count at diagnosis has been linked to poor survival . This may be related to cytokine imbalances as a phase I study  showed that in vivo IL-2 stimulation of NK cells enhances the efficacy of rituximab when treating B-cell lymphomas.
As a part of the innate immune system, NK cells monitor for and eliminate infected or abnormal cells. To do this, they express an array of activating and inhibitory receptors . Regulatory receptors control how malignant cells are eliminated, and several activating receptors, including NKp30, NKp46, NKG2D, DNAM-1, and the inducible co-stimulatory molecule CD137 (4–1BB), have been linked to antitumor immunity [12-14]. Efficient tumor cell elimination generally requires collaboration between these receptors, as described for myeloma or melanoma [12, 13].
In addition to regulatory receptors, NK cells also express CD16 (FcγRIIIA, gene FCGR3A), a low-affinity Fcγ receptor that can detect antibodies bound to tumor cells. CD16 is strongly and constitutively expressed on the highly cytotoxic CD56dim NK-cell subset, which represents about 90% of peripheral NK cells . The clinical response to rituximab is known to be affected by FCGR3A polymorphisms [16-18], which may be a reflection of the role of CD16 in NK-cell-mediated ADCC with this treatment. This study has directly assessed NK-cell phenotype and CD16 expression levels in DLBCL patients at diagnosis while also studying ADCC function and CD137 upregulation.
Results and discussion
Reduced CD16 surface expression on NK cells from DLBCL patients
Blood samples were taken from DLBCL patients at diagnosis. Flow cytometric analysis was used to assess levels of activating NK-cell receptors involved in antitumor function. The results for 36 DLBCL patients and 20 healthy donors (HCs) are shown in Figure 1 and Table 1. Of all the markers analyzed, only DNAM-1, CD137, and intracellular perforin were significantly reduced in the overall NK-cell population in DLBCL patients compared to HCs. The CD56dim/CD56bright ratio for PBMC from patients (median: 4.8 (SEM = 1.2; range: 0.4–35.5)) was comparable to that for HCs (7.5 (SEM = 1.5; range: 1.3–28.4), ns), ruling out an indirect effect on activation markers due to an imbalance of NK-cell subsets. In the CD56bright NK-cell subset, cells expressing NKp46 and DNAM-1 – both of which stimulate NK-cell functions – were significantly less abundant in patients (p < 0.0001 and p = 0.0085, respectively). This suggests that NK cells in DLBCL patients have a reduced activation potential. More importantly, expression of the FcγRIIIa receptor (CD16) on the CD56dimCD16+ subset was markedly reduced in patients at diagnosis (CD16 median fluorescence intensity (MFI): 4761) compared to HCs (MFI: 7957; p = 0.004; Fig. 1C). In patients with solid tumors, such as kidney carcinomas, no similar reduction in CD16 levels was observed (data not shown). The effects on receptor expression in DLBCL patients were not related to a soluble compound in the serum as incubating PBMCs from healthy donors with serum from DLBCL patients did not reduce CD16 expression, or cause any modifications to surface expression of DNAM-1, NKG2D, or CD137 (data not shown).
Table 1. NK-cell phenotype in healthy donors and DLBCL patients at diagnosis
Defective ADCC activity in NK cells from DLBCL patients
Our phenotypic observations suggest that the cytotoxic function of NK cells from DLBCL patients at diagnosis is reduced compared to healthy subjects. To analyze this, we investigated NK-cell degranulation either as a part of natural cytotoxicity upon stimulation by the NK-sensitive K562 cell line, or in rituximab-mediated ADCC directed against the CD20+ lymphoma cell line Raji. During cytotoxic degranulation, CD107a is transferred from cytolytic granules onto the cell surface, surface CD107a labeling can therefore be used as a surrogate marker for cytotoxic activity . We assessed CD107a labeling on NK cells from 11 patients and 15 HCs, NK cells from patients had a strongly reduced capacity to delocalize CD107a after stimulation both with K562 (median experimental/control ratios: 2.0 for patients versus 4.4 for HCs; p = 0.0003) and with rituximab + Raji (median experimental/control ratios: 2.5 for patients versus 6.3 for HCs; p = 0.0005; Fig. 2A). Stimulating DLBCL NK cells with IL-2 during cytotoxicity assays increased their ADCC capacity slightly (median of experimental/control ratios: 1.7 in medium versus 2.4 with IL-2; p = 0.043) but did not fully restore ADCC to the levels measured for HCs (Fig. 2B). The CD107a experimental/control ratio in patients was also found to positively correlate with CD16 expression after stimulation by rituximab + Raji (Rho = 0.7; p = 0.020; Fig. 2C), but not after exposure to K562 (Rho = 0.2; p = 0.509). In our cohort, CD16 expression levels did not significantly associate with FCGR3A polymorphism (data not shown). Adding IL-2 to culture medium increased CD16 expression on NK cells from HCs, but this effect was very weak or negligible on DLBCL NK cells (data not shown). Taken together, these results confirmed that the reduced CD16 expression on patients’ NK cells was, at least partially, responsible for the impaired rituximab-mediated ADCC measured.
CD16-mediated activation induces lower CD137 levels on DLBCL NK cells
Recently, Levy and colleagues  used a xenotransplanted human lymphoma model to show that the antilymphoma activity of rituximab could be enhanced by administering an anti-CD137 agonist mAb. To test whether ADCC-related NK-cell activation led to upregulated CD137 expression in patient samples, we analyzed CD137 expression in 13 DLBCL and 16 HC samples at the end of co-incubation. In vitro unstimulated NK cells express very low levels of CD137 (median in HCs: 0.35% and in patients: 0.9%; Fig. 2D). For both HCs and patients, the frequency of CD137+ cells was moderately increased by stimulation with K562. This increase was dramatic after stimulation with rituximab-coated tumor B cells, while cells coated with an irrelevant antibody induced no increase. However, with rituximab the increase in CD137 expression was significantly greater for HC samples (median CD137+ cells: 47.4%) than for DLBCL samples (median CD137+ cells: 27.1%, p = 0.001). After stimulation by rituximab-coated Raji, the frequency of CD137+ NK cells was positively correlated with CD16 expression in HCs, a similar trend was noted for DLBCL patients (Fig. 2E).
The data presented in this paper show that NK-cell-mediated ADCC may be impaired in DLBCL patients at diagnosis. How this affects individual responses to R-CHOP therapy remains to be determined. It will be necessary to monitor the NK-cell compartment in large patient cohorts to determine how NK-cell phenotype and function are affected. Nevertheless, it seems likely that improving NK-cell function in these patients could enhance the efficiency of lymphoma-specific antibodies . Given the correlation between ADCC activity and CD137 expression, CD16 and CD137 might be potential candidates to optimize mAb-mediated therapies. Next generation anti-CD20 antibodies, such as GA101 [22-25] and LFB-R603 , have already been engineered to optimize CD16-mediated ADCC. Furthermore, the advantages of combining a tumor-specific mAb with an anti-CD137 agonistic mAb have recently been described in xenotransplant models of lymphoma  and breast cancer . However, given the reduced CD137 expression on patient's NK cells before and after acti-vation shown here, a similar strategy might be unable to enhance NK-cell–mediated ADCC in DLBCL patients.
Materials and methods
Patients were diagnosed with DLBCL without circulating tumor cells in the Hematology department of the Saint-Louis Hospital, Paris, France, between 2007 and 2009 (Supporting Information Table 1) . The study cohort included 49 newly diagnosed DLBCL patients and 31 HCs. All participants gave their informed consent to participate in this study, which was approved by the Saint-Louis Hospital review board. Samples from 36 patients and 20 HCs were used for phenotype analysis experiments; ADCC was studied in 16 patients and 19 HCs (of which three patients and eight HCs were also used for NK-cell phenotyping).
PBMCs from each individual was isolated by density gradient centrifugation. Cells were stored in liquid nitrogen before analysis. All phenotype and functional analyses were performed on thawed PBMCs for which viability was greater than 50%.
Flow cytometry analysis
NK-cell phenotypes were determined as previously described  using the following mAbs from Becton Dickinson, Beckman Coulter or R&D Systems: FITC–DNAM-1 (DX11), FITC–HLA-DR (L243), FITC–CD94 (HP-3D9), FITC-Perforin (δG9), PE–NKp30 (Z25), PE–NKp46 (BAB281), PE–NKG2C (134591), allophycocyanin–NKG2D (ON72), allophycocyanin–CD137 (4B4–1), allophycocyanin–NKG2A (Z199), allophycocyanin/H7–CD16 (3G8), V450–CD3 (UCHT1), and PE/Alexa750–CD56 (B159). Flow cytometry data were acquired on a BD Biosciences Canto II (Becton Dickinson) and analyzed using FACSDiva software (Becton Dickinson). To ensure consistency across flow cytometry acquisitions, patient samples were always paired with HC samples. For CD16 expression levels, the data shown are the mean of four individual MFI quantifications per sample. MFI for negative controls were identical for patients and HCs.
CD107a degranulation assay
PBMCs were cultured overnight in one of the following four conditions: alone; with the chronic myelogenous leukemia cell line K562; with rituximab-coated (5 μg/mL) CD20+ Burkitt's lymphoma cell line Raji; or with irrelevant IgG1 isotype control-coated Raji cells. A PE–CD107a-specific mAb (H4A3, Becton Dickinson) was added to cell cultures. Where indicated, IL-2 (1000 UI/mL) was added to the cell cultures the day before the experiment. After overnight culture, cells were labeled with allophycocyanin/H7–CD16, allophycocyanin–CD137, PerCP–CD3 and PE/cyanin7–CD56, and analyzed as described above. To ensure consistency across experiments, patient samples were always paired with HC samples.
Statistical analyses were performed using Statview software (v5.0). The statistical tests used were: Mann–Whitney test, signed Wilcoxon rank test, and Spearman correlation test. In all cases, p ≤ 0.05 was taken as the significance threshold.
This work was supported by research grants from the Association pour la Recherche sur le Cancer (ZK, #DOC20100600956), Institut National du Cancer (grant #R09081HH), and Assistance Publique–Hopitaux de Paris (Translational Research grant in Biology 2010, #RTB10002). The authors would like to thank Mrs. J. Berrou for her help in ana-lyzing data in kidney cancer patients and Dr. P. Haas for his efforts in critically reviewing this manuscript.
Conflict of interest
The authors declare no financial or commercial conflict of interest.