Monocytes mediate shaving of B-cell-bound anti-CD20 antibodies

Abstract

Anti-CD20 monoclonal antibodies are promising for the treatment of B-cell malignancies such as chronic lymphocytic leukaemia and autoimmune diseases where auto-antibodies play an important role. Anti-CD20 such as rituximab (RTX) mediates B-cell depletion through mechanisms such as complement-mediated cytotoxicity and antibody-dependent cellular cytotoxicity. However, in haematological malignancies, such effector mechanisms can be saturated and result in release of malignant B cells with reduced levels of CD20. It has been hypothesized that this is the result of monocyte-mediated shaving of the CD20/RTX complex from the B-cell surface. Here, we confirm, that in vitro co-culture of human monocytes and RTX-labelled syngeneic B cells results in reduced expression of CD20/RTX complex on the B cell surface. This shaving mechanism was the result of active protease activity because EDTA and PMSF were able to mediate partial inhibition. Also, a series of alternative anti-CD20 antibodies representing both type I and type II antibodies were tested for their ability to induce the shaving reaction. These results demonstrate that a monocyte-mediated shaving reaction can lead to complete loss of most anti-CD20 antibodies from the surface of B cells even from healthy donors and this is an important obstacle for antibody-mediated immune therapy. The findings demonstrate the necessity of developing novel antibodies that maintain high effector functions without enabling activation of the shaving reaction.

Introduction

Monoclonal antibodies against tumour antigens or tissue-specific markers have become a key element in cancer immunotherapy.¹ Rituximab (RTX), which is specific for CD20 and therefore targets B cells, was the first antibody approved by the Food and Drug Administration and its effect on B-cell malignancies depends on immunological mechanisms such as complement-dependent cytotoxicity (CDC), antibody-dependent cellular cytotoxicity (ADCC) and phagocytosis.^2–5 In addition, direct induction of apoptosis in B cells also seems to be involved.⁶ Treatment with RTX is effective in autoimmune diseases where antibodies play an important role⁷ and also in several forms of B-cell lymphoma.⁸ However, in certain haematological malignancies such as chronic lymphocytic leukaemia, only a partial effect has been observed,⁹ and it is therefore pivotal to identify mechanisms that hinder the full effect of B-cell depletion strategies or that will optimize treatment strategies.

Monocytes/macrophages can, under certain conditions, remove cell-bound IgG without destroying the opsonized cell¹⁰ and this mechanism has recently been shown to account for a phenomenon called ‘shaving’, where monocytes can remove anti-CD20 antibodies together with CD20 from the surface of antibody-coated target cells through an endocytic reaction called trogocytosis that depends on Fcγ receptor I (FcγRI) expression on the acceptor cell.¹¹ Under these circumstances, the shaved target cells are viable but have reduced CD20 expression¹² and probably reflect a phenomenon previously known as antigenic modulation but is unrelated to internalization of antigen on the target cell. Such an effect is also seen in patients with chronic lymphocytic leukaemia who receive RTX treatment.¹³ Here, a rapid clearance of malignant B cells from the bloodstream is observed, but a small fraction of uncleared cells and cells that are later released from lymphoid tissues seems to obtain a reduction in CD20 expression because of shaving, which occurs, for example, by liver Kupffer cells when effector mechanisms such as CDC and ADCC have been saturated. As a result, a subsequent new bolus of RTX will have little effect on the remaining malignant B cells and so the shaving reaction has large clinical implications.

Effector function of anti-CD20 antibodies varies based on division into type I (RTX-like) and type II (tositumomab), where type II antibodies have increased B-cell depleting capacity in vivo.¹⁴ Until now, this difference between antibodies has not been explained in relation to affinity, opsonization, induction of phagocytosis, isotype or half life of the antibody, but they are known to have different abilities for redistributing CD20 in the plasma membrane. Hence, testing the effect on monocyte-mediated shaving would be important for a better understanding of anti-CD20 antibody function.

Here, we confirm, that in vitro co-culture of monocytes and RTX-labelled B cells results in reduced expression of RTX on the surface. We find that this reaction is dependent on the Fc part of RTX but is not the result of simple endocytosis. Instead, active protease activity is involved because EDTA and PMSF were able to partly inhibit the reaction. Also, we tested a series of alternative type I and type II anti-CD20 antibodies for their ability to induce the shaving reaction and here the murine type I antibody AT80 showed reduced ability to initiate the shaving reaction compared with a series of other type I and type II anti-CD20 antibodies. Our findings demonstrate that a general strategy for developing novel antibodies against haematological malignancies is necessary and has to address the inhibitory functions of the shaving reaction.

Materials and methods

Isolation of cells

Peripheral blood mononuclear cells were isolated from buffy coats obtained from healthy donors from the Department of Clinical Immunology, Rigshospitalet using Lymphoprep (Axis-Shield, Oslo, Norway). They were washed in RPMI-1640 containing Glutamax. Monocytes were than separated by positive selection with anti-CD14 conjugated to paramagnetic beads using a commercial kit from Miltenyi Biotech (Bergisch Gladbach, Germany). Similarly, syngeneic B cells were isolated by negative selection with a commercial kit from Miltenyi Biotech. In experiments where myeloid mouse CD11b⁺ cells were used as effector cells, spleen cells from conventional BALB/c mice (Taconic, Ry, Denmark) were prepared as a single-cell suspension and positive isolation for CD11b was performed with a commercial kit from Miltenyi Biotech. In vitro generation of monocyte-derived dendritic cells with granulocyte–macrophage colony-stimulating factor and interleukin-4 was performed as previously described.¹⁵

Induction of shaving reaction

B cells were then cultured as 1 × 10⁶ cells/well in a 24-well flat-bottom culture dish in X-Vivo 15 serum-free cell culture medium (Cambrex, Charles City, IA). To test the shaving reaction, RTX antibody (MabThera® from Roche, Basel, Switzerland) was added in a final concentration of 5 μg/ml. Syngeneic monocytes were added in titrated numbers up to 1 × 10⁶ cells/well. After 18 hr, cells were harvested and labelled with relevant antibodies for flow cytometry. Based on flow cytometry data, mean fluorescence intensity (MFI), % shaving was defined as (1 − [{MFI of monocyte containing RTX sample − MFI of monocyte containing isotype control}/{MFI of RTX sample − MFI of isotype control}]) × 100.

Flow cytometry

Cells from co-cultures were labelled with FITC-conjugated anti-RTX antibody (Clone MB2 A4 from AbD Serotec, Dusseldorf, Germany) or a relevant isotype control (IgG2a) and the effect of RTX was tested. Related antibody combinations where used when testing alternative anti-CD20 antibodies. After labelling and washing, cells were resuspended in PBS containing 1% BSA as running buffer and directly analysed on a FACSCalibur (San Jose, CA) using bd cell quest pro software (Becton Dickinson, Franklin Lane, NJ). Analyses of monocytes and B cells were separated using gates defined by forward and side scatter. In separate experiments, cell viability including the effect of CDC was tested with a commercial kit from BD (Becton Dickinson) using FITC-annexin V and propidium iodide.

Antibodies and other reagents

Human IgG was from Sigma (St Louis, MO), anti-CD14 and anti-CD64 was from BD. Type I and type II anti-CD20 antibodies were a kind gift from Mark S. Cragg and Claude H.T. Chan. In experiments, testing the effect of hyperosmolar sucrosis, 0·4 m sucrosis (Sigma) was added to the monocyte–B-cell co-culture. Similarly, 80% active human AB serum was used when testing CDC. In experiments with blockade of protease activity, 10 mm EDTA (Sigma) was used and 3 mm PMSF, 2·8 mg/ml aprotinin, 20 nm bestatin hydrochloride, 5 mm 1,10-phenanthroline monohydrate, 5 μm phosphoramidon disodium salt and 5 μg/ml α₂-macroglobulin (all from Sigma) were used for inhibition of specific classes of proteases.

Results

Shaving of the RTX/CD20 complex

Monocyte-mediated shaving of RTX/CD20 complexes from the surface of CD20⁺ cells has recently been reported as a major obstacle for RTX treatment in haematological malignancies.¹³ Our results here confirm the shaving reaction and demonstrated monocyte-mediated shaving of RTX antibody from the surface of CD19⁺ B cells. This phenomenon was monocyte number-dependent, not evident with 1 × 10⁴ monocytes, increased at 1 × 10⁵ and almost complete after the addition of 4 × 10⁵ to 5 × 10⁵ monocytes (Fig. 1). After addition of 1 × 10⁶ monocytes for 18 hr, the percentage of monocyte-mediated RTX shaving was typically 70–80%. The following experiments were performed after monocyte incubation for 18–24 hr, but total shaving could be observed as early as 1–2 hr after monocyte–B-cell co-culture (data not shown). Interestingly, in-vitro-generated monocyte-derived dendritic cells also induced RTX shaving (Fig. 1g). B cells were also viable after 24 hr of co-culture, but when testing for CDC, addition of activated autologous serum to co-cultures resulted in some induction of B-cell apoptosis, which seemed to vary between donors (data not shown). Hence, as complement-mediated killing does not seem to be the only effector function of RTX, monocyte-mediated shaving could be an important problem both in leukaemic and non-leukaemic applications, as it renders target cells less sensitive for natural killer (NK) cell-mediated killing.

Figure 1.

Human monocytes mediate shaving of B-cell bound rituximab (RTX): 1 × 10⁶ B cells were co-cultured with 1 × 10⁴ to 1 × 10⁶ syngeneic CD14⁺ monocytes from healthy donors together with RTX antibody. Remaining RTX on the B-cell surface was analysed after 18 hr using flow cytometry. Data are presented as original histograms from the experiment (a–f) and as % shaving ± SD (g). Also, the effect of in-vitro-generated monocyte-derived dendritic cells was tested (g). Data presented are representative of at least three independent experiments with similar findings.

Mechanisms of monocyte-mediated RTX/CD20 shaving

We next investigated the mechanisms resulting in monocyte-mediated shaving. We used a modified RTX where the Fc part was deleted and demonstrated that interaction with the Fc part of the antibody was pivotal for monocyte-mediated shaving (Fig. 2). However, using another approach to test for Fc dependency, addition of pooled human IgG or anti-CD64 antibody, to block Fc receptors, only resulted in a minor inhibition of RTX shaving, which could reflect the relatively long co-culture period used. We then tested whether the mechanisms for cleavage of the RTX complex could be the result of simple endocytosis, but as addition of hyperosmolar sucrose did not inhibit RTX shaving this does not seem likely (Fig. 3).

Figure 2.

Monocyte-mediated shaving is dependent on the Fc part of rituximab (RTX): 1 × 10⁶ B cells were co-cultured with 1 × 10⁶ syngeneic CD14⁺ monocytes from healthy donors together with RTX antibody or an Fc-truncated analogue. Remaining RTX on the B-cell surface was analysed after 18 hr using flow cytometry. Data are presented as % shaving, which indicates % loss of RTX. Data presented are representative of at least two independent experiments with similar findings.

Figure 3.

Monocyte-mediated shaving is independent of endocytosis: 1 × 10⁶ B cells were co-cultured with 1 × 10⁶ syngeneic CD14⁺ monocytes from healthy donors together with rituximab (RTX) antibody and 0·4 m hyperosmolar sucrosis. Remaining RTX on the B-cell surface was analysed after 18 hr using flow cytometry. Data are presented as % shaving, which indicates % loss of RTX. Data presented are representative of at least two independent experiments with similar findings.

Protease activity in the shaving reaction

To investigate the involvement of proteases in the shaving reaction, 10 mm EDTA was added to the B-cell–monocyte co-culture and this led to a partial inhibition of the shaving reaction (Fig. 4). Protease inhibitors can be divided into aspartic protease inhibitors, cysteine protease inhibitors, metalloproteinase inhibitors and serine protease inhibitors. Here, the serine protease inhibitor PMSF caused a partial decline in shaving activity (Fig. 5), whereas aprotinin did not. Also, the metalloproteinase inhibitors bestatin hydrochloride 1,10-phenanthroline monohydrate and phosphoramidon disodium salt did not have any effect (data not shown). The endoprotease inhibitor α₂-macroglobulin, which also acts as a cysteine protease inhibitor, serine protease inhibitor, metalloproteinase inhibitor and aspartic protease inhibitor, also did not have any effect. PMSF does also have cysteine protease inhibitor activity and phosphoramidon disodium salt has metalloproteinase inhibitor activity.

Figure 4.

Monocyte-mediated shaving is dependent on protease activity: 1 × 10⁶ B cells were co-cultured with 1 × 10⁶ syngeneic CD14⁺ monocytes from healthy donors together with rituximab (RTX) antibody and 10 mm EDTA. Remaining RTX on the B-cell surface was analysed after 18 hr using flow cytometry. Data are presented as % shaving, which indicates % loss of RTX. Data presented are representative of at least three independent experiments with similar findings. *P < 0·05.

Figure 5.

Monocyte-mediated shaving is dependent on protease activity: 1 × 10⁶ B cells were co-cultured with 1 × 10⁶ syngeneic CD14⁺ monocytes from healthy donors together with rituximab antibody and 10 mm EDTA. Remaining RTX on the B-cell surface was analysed after 18 hr using flow cytometry. Data are presented as % shaving, which indicates % loss of RTX. Data presented are representative of at least two independent experiments with similar findings. *P < 0·05.

Shaving reaction with alternative anti-CD20 antibodies

Next, we tested a panel of alternative type I and type II anti-CD20 antibodies to identify possible anti-CD20 antibodies with reduced effect on monocyte-mediated shaving. First, a series of mouse antibodies was tested. Here, antibodies such as mouse CD20-2, mouse CD20-6 (type II) and Ritm2a (type I) induced similar levels of monocyte-mediated shaving whereas type I antibody AT80 resulted in less shaving both when human monocytes were used (Fig. 6) and when mouse CD11b⁺ spleen cells were used as effector cells (data not shown). When tested in different donors, the % shaving observed with mouse AT80 was typically between 20 and 47%, whereas other mouse antibodies induced shaving at 60–90%. We then tested related human or chimeric antibodies BHH2, CD20-2, CD20-6, CD20-G and chimeric AT80 (chAT80). However, here we observed 67–84% shaving, which was comparable to the level observed with RTX (Fig. 7).

Figure 6.

AT80 mouse anti-human CD20 antibody induces shaving at lower levels: 1 × 10⁶ B cells were co-cultured with 1 × 10⁶ syngeneic CD14⁺ monocytes from healthy donors together with different mouse anti-human CD20 antibodies. Remaining antibody on the B-cell surface was analysed after 18 hr using flow cytometry with fluochrome-conjugated polyclonal rabbit anti-mouse antibody. Data are presented as % shaving, which indicates % loss of anti-CD20 antibody. Data presented are representative of at least three independent experiments with similar findings.

Figure 7.

Human rituximab (RTX) analogues have similar shaving capacity: 1 × 10⁶ B cells were co-cultured with 1 × 10⁶ syngeneic CD14⁺ monocytes from healthy donors together with various human anti-CD20 antibodies. Remaining antibody on the B-cell surface was analysed using fluorochrome-conjugated anti-human Fc antibody SB2H2 after 18 hr using flow cytometry. Data are presented as % shaving, which indicates % loss of RTX. Data presented are representative of at least two independent experiments with similar findings.

Discussion

Recently, it was reported that monocytes have an inhibitory effect on ADCC because they can remove antibody such as RTX from the surface of target B cells and in this way cause a reduced ability of NK cells to bind RTX via the FcγRIII.^11,12 Hence, monocytes seem to compromise RTX treatment, in particular in haematological malignancies with a large B-cell load.¹³ Here, we confirm these observations and demonstrate that the shaving mechanism is independent of endocytosis but relies on protease activity after monocyte binding to the Fc part of RTX. Also, we have screened a series of alternative type I and II anti-CD20 antibodies to identify antibodies with a reduced effect on monocyte-mediated shaving.

This work demonstrated that monocytes are able to remove B-cell-bound RTX at monocyte : B-cell ratios of 1 : 2 in vitro and that this is dependent on the Fc part of RTX. Recent work has shown that the high-affinity receptor for IgG, FcγRI, is responsible for this and expression of this receptor on monocytes provides a competitive advantage to hinder NK-cell-mediated ADCC through FcγRIII with lower affinity.¹² This group also demonstrated that addition of human IgG could restore NK-cell-mediated ADCC in these co-cultures. However, in our assay, the addition of human IgG or anti-CD64 only had a minor effect on monocyte-mediated shaving. This could reflect that the addition of IgG in their assay had a direct effect on the NK cells, which also have an ability to perform shaving of target cells. Hence, monocytes could either be dependent on cross-linking of even low numbers of free FcγRI to induce shaving or be activated in alternative ways. Interestingly, we also observed that monocyte-derived dendritic cells can mediate the shaving reaction, and this could represent an additional mechanism whereby dendritic cells in the tumour microenvironment act as a ‘black hole’, hindering effective anti-tumour immune responses.

Hyperosmolar sucrosis is an inhibitor of endocytosis. In our assay, hyperosmolar sucrosis did not lead to inhibition of the shaving reaction and this indicates that this phenomenon is not the result of B-cell-mediated endocytosis of the CD20/RTX complex or of simple endocytosis by monocytes. This observation is in line with detailed analysis from Beum et al.¹¹ who recently demonstrated that the shaving reaction is similar to a processing mechanism originally described by Griffin et al.,¹⁰ which is now named trogocytosis. Here phagocytic cells can remove immune complexes from lymphocytes without destroying them, which is in contrast to phagocytosis, where the phagocytic cell surrounds and engulfs the target cell, or to endocytosis or macropinocytosis, which involve internalization of the particulate or soluble immunocomplexes by the acceptor cells. Hence, the shaving reaction seemed to be dependent not only on cytochalasin D¹² but also on protease activity as a protease inhibitor mixture could inhibit the effect of THP-1-cell-mediated shaving.¹⁶ In our study, we confirmed that protease activity is also involved in the shaving reaction performed by conventional monocytes as EDTA leads to a partial inhibition. Further investigation of protease reactivity revealed that serine proteases are likely to be involved because PMSF resulted in some inhibition of the shaving reaction.

Recently, Beum et al.¹¹ demonstrated that monocyte-mediated shaving of therapeutic antibodies is a general phenomenon that can be extended to, for example, cetuximab, used for treatment of colorectal cancers and other tumors, and trastuzumab, used for treatment of breast cancer. This demonstrates that trogocytosis or shaving of therapeutic antibodies is likely to occur against most therapeutic antibodies used in the clinic and underscores the importance of identifying novel antibodies that bypass this reaction, in particular in cancer therapy where the target cell load is high and therefore more likely to result in competition between monocyte-mediated shaving and NK-cell-mediated ADCC. We therefore screened a series of mouse and human anti-CD20 antibodies to identify candidate antibodies with reduced capacity for the shaving reaction. Here, human anti-CD20 antibodies BHH2, CD20-2, CD20-6, CD20-G and CD20-8 all induced monocyte-mediated shaving at a similar level as RTX. When we tested mouse anti-human CD20 antibodies, most antibodies such as mouse CD20-1, CD20-2, mouse CD20-6, Ritm2a, HI47, NK1-B20 2b, NK1 B20 1, IF5, LT20 and NK1 B20 2a (representing different type I and II antibodies) also induced shaving at a similar level both when human monocytes and mouse spleen CD11b⁺ cells were used as acceptor cells. However, mouse AT80 induced shaving at a lower level, indicating that antibody-specific differences can be found. Unfortunately, the chimeric antibody chAT80 that expresses a human Fc again induced shaving at a level comparable to RTX.

In conclusion, we demonstrated that monocyte-mediated shaving of RTX on the surface of B cells is a general phenomenon and leads to complete loss of RTX from the B-cell surface. This mechanism is independent of simple endocytosis and involves serine protease activity and a functional Fc part of the opsonizing antibody. The shaving reaction seems to be a general phenomenon for most antibodies tested, but our results demonstrate that candidate antibodies with altered and reduced ability for shaving can be identified. For future immunotherapy with therapeutic antibodies, research into factors regulating antibody shaving ability will be important for increased efficacy of depleting antibodies in relation to their ADCC induction and clinical response, in particular in cancer.

Disclosures

The authors declare to have no financial conflicts or interest.