How Do mAbs Make Use of Complement to Kill Cancer Cells? The Role of Ca2+

We examined the kinetics and mechanisms by which monoclonal antibodies (mAbs) utilize complement to rapidly kill targeted cancer cells. Based on results from flow cytometry, confocal microscopy and high-resolution digital imaging experiments, the general patterns which have emerged reveal cytotoxic activities mediated by substantial and lethal Ca2+ fluxes. The Ca2+ fluxes are common to the reported pathways that have been utilized by other toxins in killing nucleated cells. These reactions terminate in very high levels of cell killing, and based on these considerations, we suggest additional strategies to further enhance mAb-based targeting of cancer with complement.


Introduction
Complement was first described and characterized by Bordet more than 100 years ago [1]; he demonstrated it to be a heat-labile factor in serum that promoted destruction (lysis) of bacteria and/or hemolysis of erythrocytes, each opsonized with the antibodies in immune sera. Complement is an important "first responder" that orchestrates the rapid clearance and destruction of a variety of microbial invaders as well as damaged and dying cells. It is therefore quite reasonable to anticipate that it should also be capable of destroying antibody-opsonized tumor cells. The importance of complement (C) in health and disease is now very well recognized and several outstanding reviews that describe its pathways and biological actions are available [2][3][4].
The first figure in the review in this volume by Golay and Taylor succinctly summarizes the important steps and controls in C-mediated killing of malignant cells opsonized with specific mAbs [5]. The traditional view of the mechanism by which C mediates the killing of antibody-opsonized cells was based on classic experiments that focused on C-mediated lysis of non-nucleated sheep erythrocytes that were first opsonized with polyclonal rabbit antibodies before they were brought into contact with a source of C and then incubated for a considerable period of time at 37 • C to promote hemolysis [6][7][8].
The results of these studies led to the concept that insertion of the membrane attack complex (MAC) pore(s) into the erythrocyte cell membrane allowed for influx of water and ions into the cell, ultimately leading to swelling of the cells followed by osmotic lysis and killing of the cells [6][7][8][9][10]. This model system has of course proven to be invaluable for dissecting out and identifying virtually all of the key components of C, including pathways, activating factors and inhibitors.

Nucleated Cells Are More Complicated: Important Questions
However, a considerable body of evidence, based on a series of studies by Shin's group on the lysis of nucleated Ehrlich ascites cells (EACs) opsonized with rabbit polyclonal antibodies, suggested that the osmotic lysis concept could not explain how these nucleated cells were killed. Instead, The influx of Ca 2+ mediated by MAC pores appeared to be the predominant lethal event [10][11][12][13][14][15]. Table 1. Observed consecutive steps in monoclonal antibody (mAb)-mediated complement-dependent cytotoxicity (CDC).
Step Time Frame * Method Figure  References mAb binding 30 s Flow cytometry 8 [16] C1q binding and colocalization with mAb 60 s Flow cytometry 2, 8 [16,17] C3b deposition and mAb colocalization 90 s Flow cytometry, high-resolution digital imaging (HRDI) 3, 4, 10 [16,18,19] C9 binding and colocalization of C9/C3b/mAb 60-120 s HRDI, confocal 11, 12 [20] Ca 2+ influx, transition-state intermediate  s Confocal, flow cytometry 9, 10, 15, 16 [16,21] tunneling nanotubule (TNT) generation 120 s Spinning disk 4, 7 [19,22] mitochondrial poisoning 180 s 60-180 s HRDI, confocal 9-11 [16,20] decay of transition-state intermediate 225 s Confocal 9, 10, 15, 16 [16,21] MAC formation and cell death 60-240 s Confocal 8-10, 15 [16,20,21]   It is well established that during C activation, innocent bystander opsonization or lysis of nearby non-targeted cells is negligible [4,[39][40][41]. Therefore, implicit in the specificity of the lytic C cascade is the presumption that there is both concentration and localization of activated C components at the nexus of C activation, which, for the classical pathway, would be the cell-bound mAbs at the plasma membrane target site (e.g., CD20). We have demonstrated this phenomenon of colocalization of reactive proteins at multiple steps in the pathway, starting with C1q binding [17]. First, we observed a high level of colocalization of Al647 OFA with Al488 C1q on Daudi cells quantitated by HRDI measurements (Figure 2 and Table 2). However, when we performed the experiment under identical conditions with Al647 RTX-opsonized Daudi cells, we found that although the amount of cell-bound RTX was comparable to the amount of cell-bound OFA, there was a very modest level of C1q binding ( Table 2) and considerably less colocalization of cell-bound C1q with RTX. Based on analyses of their CD20 epitope specificities, it is now well established that OFA binds much closer to the cell membrane and with a slower off rate than RTX [19,31,[42][43][44], and this has also been reflected "downstream" in that OFA is able to mediate CDC of B cells much more effectively than does RTX. This is particularly apparent in the case of ARH77 cells and CLL B cells. These cells are more resistant to mAb-induced CDC than most B cell lines because they express low levels of CD20 but high levels of C control proteins CD55 and/or CD59, and therefore the differences in the CDC efficacy of OFA versus RTX are readily demonstrated [19,31,43,44]. These findings also speak to the issue of thresholds for C activation [9,45,46] and cell killing by the MAC. Binding of RTX to B cells does indeed allow for some C1q binding, C activation, and subsequent C3b deposition and colocalization of the deposited cell-bound C3b with cell-bound RTX ( Figure 3) [18,19]. However, we found that on reaction in NHS, the amount of C3b deposited on OFA- Daudi cells were incubated with 10 µg/mL Al647 mAb and 1.6 µg/mL Al488 C1q in medium for 60 min at 37 • C. Samples were washed, fixed, concentrated, and then analyzed by multispectral high-resolution digital imaging (HRDI). Fluorescence signals for Al647 OFA-opsonized samples were analyzed for colocalization with bound Al488 C1q. Light scatter, fluorescence and bright field images of representative doubly positive cells. Results are representative of three similar experiments. Figures 1 and 2  reacted CLL cells was 5-10-fold greater than the amount of C3b deposited on RTX-opsonized CLL cells, quantitated with flow cytometry measurements [31]. Thus, although there is comparable binding of these CD20 mAbs to the CLL cells and there is enough C1q bound to RTX-opsonized cells to activate C, less C3b is deposited on the cells compared to the amount of C3b deposition mediated by OFA [17,31]. In other words, the C3b deposition threshold needed to achieve adequate generation of the MAC to enable cell killing is not reached for most RTX-opsonized CLL cells. It is therefore understandable why OFA is considerably more effective than RTX in promoting CDC of CLL cells. Representative images from samples opsonized as indicated and then analyzed by HRDI. Similarity bright detail score (SBDS) values given below the images are the mean ± SD (n) for values obtained on 'n' different replicate samples (at least 10,000 cells/sample) prepared and analyzed over a period of 18 months. The AF488 maleimide binds to the free SH group on C3b. Note how it is colocalized with mAb 7C12, specific for C3b. Reprinted from Journal of Immunological Methods, Vol. 317, 2006, Beum, P.V. et al., Quantitative analysis of protein colocalization on B cells opsonized with rituximab and complement using the ImageStream multispectral imaging flow cytometer, pp. 90-99, with permission from Elsevier, Amsterdam, The Netherlands [18].
However, RTX can promote very rapid C3b deposition on B cells in the circulation of non-human primates. We found that when RTX is infused intravenously into cynomolgus monkeys, it rapidly binds to circulating B cells and this is followed, within 2 min, by deposition of C3 fragments in close juxtaposition with B cell-bound RTX [36]. Moreover, we also obtained blood samples from CLL patients treated with either RTX or OFA, and we observed C3 fragments deposited on their B cells in close juxtaposition with cell-bound RTX/OFA on samples taken within an hour of the start of the mAb infusions [31,47,48]. We note that multiple lines of evidence indicate that mechanisms mediated by receptors for the Fc region of human IgG (FcγR) expressed on macrophages are principally responsible for the in vivo efficacy of RTX [49][50][51].

C3b Deposition Kinetics, a Key Intermediate Step in CDC; the "Discovery" of Streamers
We next asked whether the kinetics of C3b deposition and killing of B cell lines or of CLL cells opsonized with these CD20 mAbs would also reflect differences between OFA and RTX. We conducted real-time spinning disk confocal fluorescence microscopy experiments in which Alexalabeled mAbs specific for CD20 were reacted with B cells and then incubated in NHS as a C source supplemented with Alexa-labeled mAb 3E7 as a marker to reveal C3b/iC3b deposition. Importantly, mAb 3E7 does not cross-react with C3 and therefore can report C3b deposition in situ. [19]. We confirmed that under these conditions, both RTX and OFA promoted rapid C activation (~2 min), and that substantial colocalization of deposited C3b with cell-bound RTX was easily demonstrable on Daudi cells (Figure 4, panels A-C). However, an unexpected and initially very puzzling observation set the stage for more detailed investigations that have ultimately allowed us to carefully However, RTX can promote very rapid C3b deposition on B cells in the circulation of non-human primates. We found that when RTX is infused intravenously into cynomolgus monkeys, it rapidly binds to circulating B cells and this is followed, within 2 min, by deposition of C3 fragments in close juxtaposition with B cell-bound RTX [36]. Moreover, we also obtained blood samples from CLL patients treated with either RTX or OFA, and we observed C3 fragments deposited on their B cells in close juxtaposition with cell-bound RTX/OFA on samples taken within an hour of the start of the mAb infusions [31,47,48]. We note that multiple lines of evidence indicate that mechanisms mediated by receptors for the Fc region of human IgG (FcγR) expressed on macrophages are principally responsible for the in vivo efficacy of RTX [49][50][51].

C3b Deposition Kinetics, a Key Intermediate Step in CDC; the "Discovery" of Streamers
We next asked whether the kinetics of C3b deposition and killing of B cell lines or of CLL cells opsonized with these CD20 mAbs would also reflect differences between OFA and RTX. We conducted real-time spinning disk confocal fluorescence microscopy experiments in which Alexa-labeled mAbs specific for CD20 were reacted with B cells and then incubated in NHS as a C source supplemented with Alexa-labeled mAb 3E7 as a marker to reveal C3b/iC3b deposition. Importantly, mAb 3E7 does not cross-react with C3 and therefore can report C3b deposition in situ. [19]. We confirmed that under these conditions, both RTX and OFA promoted rapid C activation (~2 min), and that substantial colocalization of deposited C3b with cell-bound RTX was easily demonstrable on Daudi cells (Figure 4, panels A-C). However, an unexpected and initially very puzzling observation set the stage for more detailed investigations that have ultimately allowed us to carefully decipher the intricacies of the CDC killing mechanism for nucleated cells. We found that very soon after the C3b deposition reaction could be detected, long very thin fragments of cell membrane extended from the Daudi cells before they were killed, and it was possible to detect both membrane-bound mAb (RTX or OFA) along with colocalized C3b on these fragments. Control experiments in the absence of mAb 3E7 clearly demonstrated that these structures were not an experimental artifact (Figure 4, panel D). We initially called these membrane fragments "streamers", but additional experiments revealed that we were studying the formation of tunneling nanotubules (TNTs), in a reaction that is mediated by rapid entry of large amounts of Ca 2+ into a cell [22,52,53].
Antibodies 2020, 9, x FOR PEER REVIEW 8 of 26 bound mAb (RTX or OFA) along with colocalized C3b on these fragments. Control experiments in the absence of mAb 3E7 clearly demonstrated that these structures were not an experimental artifact ( Figure 4, panel D). We initially called these membrane fragments "streamers", but additional experiments revealed that we were studying the formation of tunneling nanotubules (TNTs), in a reaction that is mediated by rapid entry of large amounts of Ca 2+ into a cell [22,52,53]. The magnification in (A-C) was ×40, but, in panel (D) and all other figures derived from spinning disc confocal microscopy experiments, magnification at ×63 was used. [19].
We performed a similar experiment substituting ARH77 cells for Daudi cells. As noted previously, OFA, but not RTX, can mediate CDC of ARH77 cells. We observed colocalization of OFA or of RTX with C3b on ARH77 cells when the cells activated C in the presence of NHS. However, TNTs/streamers were released only by OFA-reacted ARH77 cells, but not by ARH77 cells reacted with RTX ( Figure 5, panels A-D) [19]. This finding strongly suggests that C activation on ARH77 cells induced by RTX was not adequate to promote entry of Ca 2+ into these cells. That is, in view of the large differentials in C3b deposition and cell killing for OFA versus RTX-opsonized CLL cells, these The magnification in (A-C) was 40×, but, in panel (D) and all other figures derived from spinning disc confocal microscopy experiments, magnification at 63× was used [19].
We performed a similar experiment substituting ARH77 cells for Daudi cells. As noted previously, OFA, but not RTX, can mediate CDC of ARH77 cells. We observed colocalization of OFA or of RTX with C3b on ARH77 cells when the cells activated C in the presence of NHS. However, TNTs/streamers were released only by OFA-reacted ARH77 cells, but not by ARH77 cells reacted with RTX ( Figure 5, panels A-D) [19]. This finding strongly suggests that C activation on ARH77 cells induced by RTX was not adequate to promote entry of Ca 2+ into these cells. That is, in view of the large differentials in C3b deposition and cell killing for OFA versus RTX-opsonized CLL cells, these findings again support the idea that the threshold for C3b deposition required for generation downstream of sufficient amounts of the MAC to effectively permeabilize ARH77 cells and promote Ca 2+ entry is not reached for RTX-reacted ARH77 cells. The patterns we have described (colocalization of mAb and C3b, formation of TNTs) are not unique to CD20 mAbs. Certain mAbs activate C very effectively on binding to target cells because they bind at very high levels (>80,000 mAbs per cell). These include HB28 (anti-β2 microglobulin, mouse (IgG2b), alemtuzumab (anti-CD52), and W6/32 (anti-HLA), and all of these mAbs have also been demonstrated to produce streamers/TNTs on binding to target cells in the presence of C [19,22,33,54].
Antibodies 2020, 9, x FOR PEER REVIEW 9 of 26 effectively on binding to target cells because they bind at very high levels (>80,000 mAbs per cell). These include HB28 (anti-β2 microglobulin, mouse (IgG2b), alemtuzumab (anti-CD52), and W6/32 (anti-HLA), and all of these mAbs have also been demonstrated to produce streamers/TNTs on binding to target cells in the presence of C [19,22,33,54].

On the Importance of Ca 2+
Based on these observations, we next focused on investigating the possible role of Ca 2+ in the cell-killing phase of the CDC reaction [22]. Upstream deposition of C3b occurs in a process which requires Ca 2+ , but the downstream terminal steps in the C cascade, in particular, generation of the MAC, do not directly require Ca 2+ . Therefore, we briefly reacted Daudi cells with OFA in C5-depleted serum to deposit active C3b but not permit subsequent MAC formation. The cells were then washed and incubated in NHS-EDTA (to chelate Ca 2+ ) or in NHS. Under both conditions, The MAC is then generated and the cells are killed ( Figure 6); however, TNTs are not produced when the cells are killed in NHS-EDTA, providing strong evidence that in NHS it is entry of Ca 2+ into the cells that promotes TNT formation (Figure 7) [22]. The degree of killing for C3b-opsonized cells reacted in NHS-EDTA was somewhat lower and slower than in in NHS. We suggest that under these conditions, where Ca 2+ entry into cells is precluded, we are instead studying "death by drowning" of the cells due to influx of large amounts of water and loss of cellular constituents upon permeabilization of the cell membrane. However, it is our working hypothesis that when the cells are killed under normal physiological conditions for CDC, it is the influx of lethal amounts of Ca 2+ that provides the most immediate fatal blow. A similar finding of slower cell killing was reported by Papadimitriou et al., who examined MAC killing of EACs in the presence of EGTA in which Ca 2+ was chelated, thus precluding its rapid entry into the cells [14,15].

On the Importance of Ca 2+
Based on these observations, we next focused on investigating the possible role of Ca 2+ in the cell-killing phase of the CDC reaction [22]. Upstream deposition of C3b occurs in a process which requires Ca 2+ , but the downstream terminal steps in the C cascade, in particular, generation of the MAC, do not directly require Ca 2+ . Therefore, we briefly reacted Daudi cells with OFA in C5-depleted serum to deposit active C3b but not permit subsequent MAC formation. The cells were then washed and incubated in NHS-EDTA (to chelate Ca 2+ ) or in NHS. Under both conditions, the MAC is then generated and the cells are killed ( Figure 6); however, TNTs are not produced when the cells are killed in NHS-EDTA, providing strong evidence that in NHS it is entry of Ca 2+ into the cells that promotes TNT formation (Figure 7) [22]. The degree of killing for C3b-opsonized cells reacted in NHS-EDTA was somewhat lower and slower than in in NHS. We suggest that under these conditions, where Ca 2+ entry into cells is precluded, we are instead studying "death by drowning" of the cells due to influx of large amounts of water and loss of cellular constituents upon permeabilization of the cell membrane. However, it is our working hypothesis that when the cells are killed under normal physiological conditions for CDC, it is the influx of lethal amounts of Ca 2+ that provides the most immediate fatal blow. A similar finding of slower cell killing was reported by Papadimitriou et al., who examined MAC killing of EACs in the presence of EGTA in which Ca 2+ was chelated, thus precluding its rapid entry into the cells [14,15].  At this point, our research direction was strongly influenced by a general theory as to how toxins kill cells, developed more than 40 years ago by Schanne et al. [55]. They noted that in a first step, toxin could compromise the integrity of cell membranes by a variety of mechanisms that were usually independent of Ca 2+ . However, they suggested that "the second step of toxin induced killing most At this point, our research direction was strongly influenced by a general theory as to how toxins kill cells, developed more than 40 years ago by Schanne et al. [55]. They noted that in a first step, toxin could compromise the integrity of cell membranes by a variety of mechanisms that were usually independent of Ca 2+ . However, they suggested that "the second step of toxin induced killing most likely represents an influx of Ca 2+ across the damaged plasma membrane . . . and represents a final common pathway by which the cells are killed." Indeed, under normal physiological conditions, The external extracellular Ca 2+ concentration in blood and interstitial fluid is in the millimolar range, but the Ca 2+ concentration in most nucleated cells is approximately 0.1 uM, and levels above 2 uM are usually lethal [9,[56][57][58]. Therefore, it is quite reasonable to expect that when the plasma membrane of a cell is effectively permeabilized by the MAC "toxin", The cell will then be killed as a consequence of influx of Ca 2+ and poisoning of many of its major metabolic pathways due to Ca 2+ -mediated excessive activation of a variety of proteases, endonucleases, and phospholipases [12,14,15].

Cell-Bound Hexamer-Forming mAbs Bind C1q
In a very productive collaboration, we made use of hexamer-forming mAbs that were developed by Prof. Paul Parren and his colleagues at Genmab [16,20,[59][60][61][62]. We used these mAbs to further study the role of Ca 2+ entry into cells in the mAb-mediated CDC reaction. These mAbs are modified in the Fc region, which substantially enhances their potential to form hexamers upon binding to cells. The modifications (e.g., E430G) promote much more effective and rapid binding of the classical pathway initiating factor, hexameric C1q, to the mAb-opsonized cells, and thereby increase the CDC potential of a wide range of mAbs. We also note that numerous other strategies are under investigation for increasing the ability of mAbs to promote CDC of tumor cells [26,27,41,[63][64][65][66][67][68][69][70][71][72]. It will be interesting to compare the efficacy of these strategies if they progress to clinical trials.
We first compared 7D8 (a close analogue of OFA [73]), RTX and the hexamer-forming variants, 7D8-Hx and RTX-Hx. Although approximately comparable amounts of the different CD20 mAbs bind to a given cell, 7D8-Hx and RTX-Hx are considerably more effective at promoting C1q binding and rapid C-mediated killing of the target cells ( Figure 8) [16]. We also found that all four mAbs bound to the cells within 1-2 min.

Four-Color Confocal Microscopy Movies
We next performed a series of kinetic experiments to study CDC-mediated by 7D8-Hx based on four-color confocal microscopy movies. Raji B cells or CLL cells were internally labeled with the green fluorescent Ca 2+ indicator, Fluo-4, to monitor Ca 2+ influx. Viable mitochondria were visualized with TMRME (red). The cells were dispersed in NHS containing Alexa-405-labeled anti-C3b/iC3b mAb 3E7 (blue) to follow C3b deposition and TO-PRO-3 (purple) was added to the medium as a live/dead indicator. The screenshots (Figure 9) [16] from the movies of Raji cells reacted with 7D8-Hx demonstrate that many of the cells are opsonized with C3b (binding of blue anti-C3b mAb 3E7) after approximately 90 s. Soon thereafter, enough MAC must have penetrated the cells, because many of them are bright green (indicating influx of Ca 2+ ), but some are still alive, and their mitochondria appear to be alive and intact, based on the viable red TMRME signal (135 and 180 s). We have identified these bright green cells as "transition-state intermediates". They are alive, but are doomed, because lethal amounts of Ca 2+ had entered the cells. Soon, by 180-270 s, many of the cells are dying or are dead, and this is evident based on three separate criteria: first, most of the Fluo-4 has leaked out of the cells; second, The mitochondria have been poisoned and are no longer stained by TMRME; and third, TO-PRO-3 has entered the cells. Careful inspection of the movies suggested that there was a slight lag (5-10 s) between the coincident leakage of Fluo-4/loss of the TMRME signal, and the final entry of TOPRO-3 into the dead cells. More details on these phenomena can be found in the Figure 9 caption.  or are dead, and this is evident based on three separate criteria: first, most of the Fluo-4 has leaked out of the cells; second, the mitochondria have been poisoned and are no longer stained by TMRME; and third, TO-PRO-3 has entered the cells. Careful inspection of the movies suggested that there was a slight lag (5-10 s) between the coincident leakage of Fluo-4/loss of the TMRME signal, and the final entry of TOPRO-3 into the dead cells. More details on these phenomena can be found in the Figure 9 caption.

Kinetics of CDC Monitored by Multicolor HRDI
We also made use of HRDI technology to visualize cells during the CDC process. Figure 10, top panel identifies three distinct populations of cells at the 40 s mark that are either: alive-Fluo-4 very bright (transition-state intermediate) and TMRME positive; and finally dead-Fluo-4 weak, TMRME negative, C3b positive and TOPRO-3 positive [16]. Verification that the residual Fluo-4 stain is in the mitochondria (identified with Mitotracker Red) of the cells is demonstrated in Figure 10, bottom panel. It is also noteworthy that there is no noticeable swelling of the dead cells ( Figure 10, top panel) at early times, soon after they are killed. This again emphasizes that there is no evidence for an early osmotic burst reaction when the nucleated cells are first killed, in agreement with Papadimitriou et al. [15]. negative, C3b positive and TOPRO-3 positive [16]. Verification that the residual Fluo-4 stain is in the mitochondria (identified with Mitotracker Red) of the cells is demonstrated in Figure 10, bottom panel. It is also noteworthy that there is no noticeable swelling of the dead cells ( Figure 10, top panel) at early times, soon after they are killed. This again emphasizes that there is no evidence for an early osmotic burst reaction when the nucleated cells are first killed, in agreement with Papadimitriou et al. [15]. We also used HRDI to follow the CDC reaction for Z138 cells. In these experiments, we monitored C9 binding instead of Ca 2+ influx. We were able to verify that binding of C9 to the Z138 cells followed C3b deposition, and that live cells containing bound C9 could be identified (Figure 11), but we know that these are also "doomed" cells that will soon experience substantial fluxes of Ca 2+ [20]. In agreement with earlier studies, there is considerable colocalization of cell-bound mAb with deposited C3b, and in addition deposited C3b clearly serves as a "landing site" for binding of C5b-9, based on the colocalization of C3b and C9 (Figure 12). The identification of the transition-state intermediate (very bright homogeneous Fluo-4 signal) raised an interesting question: could we better validate its existence and stabilize it by slowing leakage of Fluo-4 out of the cell? The coincident question raised in these studies centered on the role of C9; is C9 essential to kill the cells as part of the MAC?
We also used HRDI to follow the CDC reaction for Z138 cells. In these experiments, we monitored C9 binding instead of Ca 2+ influx. We were able to verify that binding of C9 to the Z138 cells followed C3b deposition, and that live cells containing bound C9 could be identified (Figure 11), but we know that these are also "doomed" cells that will soon experience substantial fluxes of Ca 2+ [20]. In agreement with earlier studies, there is considerable colocalization of cell-bound mAb with deposited C3b, and in addition deposited C3b clearly serves as a "landing site" for binding of C5b-9, based on the colocalization of C3b and C9 (Figure 12). The identification of the transition-state intermediate (very bright homogeneous Fluo-4 signal) raised an interesting question: could we better validate its existence and stabilize it by slowing leakage of Fluo-4 out of the cell? The coincident question raised in these studies centered on the role of C9; is C9 essential to kill the cells as part of the MAC?

On the Role of C9
We identified small populations of cells that were killed by CDC but did not appear to be stained by C9 [20] (Figure 11, panel E), and we also found that CLL cells and Z138 cells reacted with 7D8-Hx could be killed by CDC in C9-depleted serum. In fact, we reported that the cells could also be killed in sera depleted of Factor B and Factor D ( Figure 13). Other mAbs that promote high levels of CDC, Data were obtained based on multispectral high-resolution fluorescence imaging by flow cytometry. CLL cells were reacted with Al546 Hx-7D8 in NHS for 8 min and after two washes were probed with both FITC mAb aE11 and with Al594 3E7 (for C3b/iC3b). Images representative of triple-positive cells are shown. Bright detail similarity score (BDSS) values for the merged images of the individual cells are given. A very high degree of colocalization of Hx-7D8 with C3b is evident, and colocalization of the mAb with C9 is also observed [20].

On the Role of C9
We identified small populations of cells that were killed by CDC but did not appear to be stained by C9 [20] (Figure 11, panel E), and we also found that CLL cells and Z138 cells reacted with 7D8-Hx could be killed by CDC in C9-depleted serum. In fact, we reported that the cells could also be killed in sera depleted of Factor B and Factor D ( Figure 13). Other mAbs that promote high levels of CDC, including alemtuzumab and W6/32 (Mouse IgG2b, anti-HLA) also were able to mediate CDC in C9-depleted serum, but in all cases mAb-mediated CDC had an absolute requirement for C1q. These findings suggest that the alternative pathway of C (APC) does not appear to be needed to promote effective CDC mediated by the mAbs under investigation. It is generally believed that the APC, based on its inherent exponential amplification loop is responsible for most of the efficacy of C [74,75]. Our findings would suggest this might not be the case for CDC of tumor cells mediated by mAbs, where it appears that the classical pathway (C1q requirement) is key.  The results of the experiments with C9-depleted serum are intriguing, but not definitive because trace amounts of C9 could still be present in the depleted serum. However, The smaller C5b-8 pores that penetrate cells are approximately 3.5 nm in size (pores formed with C5b-9 are 10-11.5 nm) and in principle the C5b-8 pores should be adequate to allow for Ca 2+ entry and killing of the cell [76]. Therefore, in collaboration with Drs. Paul Morgan and Masashi Mizuno, we used flow cytometry to investigate whether CLL cells could be killed by Ca 2+ fluxes in serum genetically deficient in C9 [21]. Although we had limited amounts of the C9-deficient serum, we were able to demonstrate that mAb 7D8-Hx promoted substantial CDC of the CLL cells from six different patients ( Figure 14). Compared to the reaction in NHS, The CDC kinetics were only slightly slower in the C9-deficient serum. We also asked whether the smaller C5b-8 pores would stabilize the transition-state intermediate (bright green, Fluo-4 very positive) by slowing down exit of Fluo-4 from the cell. Indeed, at the 150 s point, approximately 70% of the cells had been killed, but the net Fluo-4 signal was maximized at this point, indicating that even though most of the cells had been killed due to Ca 2+ poisoning, The Fluo-4 had not yet leaked out of the cells (Figure 15). The bright green "dead intermediate" was therefore stabilized over a period of approximately 200 s (t = 100 s to 300 s). Due to very limited amounts of C9-deficient serum, we were not able to investigate the reaction with confocal microscopy movies. However, we suggest that further studies with serum from donors genetically deficient in C9 [77] would allow for more comprehensive investigation of these phenomena in the future. For example, it could be quite informative and useful to identify and differentiate tumor cells that are/are not susceptible to CDC in sera lacking C9.
, 9, x FOR PEER REVIEW 4. C9-deficient serum can mediate high levels of CDC of CLL cells opsonized w . C9-deficient (C9-D) serum lacks any detectable C9, but can mediate high levels o ls reacted with mAb Hx-7D8. The ability of mAb Hx-7D8 to promote CDC in CLL c rent patients was evaluated in either: C9-D serum (25%), C9-D serum (25%) supp μg/mL C9, or NHS (25%). Controls included Hx-7D8 reacted with cells in media (4 ples) as well as cells reacted with RTX (2 CLL samples) or an isotype control (Isotype samples) reacted with CLL cells in all three conditions. Each color represents a pn); all determinations were in duplicate or triplicate and means and SD are provid d SD: 58 ± 17% for CLL cells reacted in C9-D serum; 67 ± 20% for CLL cells reacted Figure 14. C9-deficient serum can mediate high levels of CDC of CLL cells opsonized with mAb Hx-7D8. C9-deficient (C9-D) serum lacks any detectable C9, but can mediate high levels of lysis of CLL cells reacted with mAb Hx-7D8. The ability of mAb Hx-7D8 to promote CDC in CLL cells from six different patients was evaluated in either: C9-D serum (25%), C9-D serum (25%) supplemented with 18 µg/mL C9, or NHS (25%). Controls included Hx-7D8 reacted with cells in media (4 of 6 CLL cell samples) as well as cells reacted with RTX (2 CLL samples) or an isotype control (Isotype, Hx-b12, 2-4 CLL samples) reacted with CLL cells in all three conditions. Each color represents a different patient (pn); all determinations were in duplicate or triplicate and means and SD are provided. Mean % CDC and SD: 58 ± 17% for CLL cells reacted in C9-D serum; 67 ± 20% for CLL cells reacted in C9-D serum + C9; 91 ± 7% for CLL cells reacted in NHS. * p < 0.05; ** p < 0.01 [21]. with 18 μg/mL C9, or NHS (25%). Controls included Hx-7D8 reacted with cells in media (4 of 6 CLL cell samples) as well as cells reacted with RTX (2 CLL samples) or an isotype control (Isotype, Hx-b12, 2-4 CLL samples) reacted with CLL cells in all three conditions. Each color represents a different patient (pn); all determinations were in duplicate or triplicate and means and SD are provided. Mean % CDC and SD: 58 ± 17% for CLL cells reacted in C9-D serum; 67 ± 20% for CLL cells reacted in C9-D serum + C9; 91 ± 7% for CLL cells reacted in NHS. * p < 0.05; ** p < 0.01 [21].

Ca 2+ Appears to be the Key: Implications
Our findings, as well as those of the Shin group [10][11][12][13][14][15], provide compelling evidence that it is the influx of Ca 2+ into nucleated cells that is the primary cause of cell death in the CDC reaction mediated by anti-tumor mAbs. On this basis, there are several additional strategies that could be employed to bring even more efficacy to this process, based on increasing the flux of lethal amounts of Ca 2+ into the targeted cell. One possibility is also to stimulate the target cell with other mAbs and/or ligands that promote uptake of Ca 2+ into the cell. For example, Fifadara reported that mast cells costimulated at the high-affinity receptor for the Fc region of IgE (FcεRI) and chemokine receptor 1 (both of which promote Ca 2+ fluxes) were not killed, but produced TNTs [78]. Apparently, these cells were "on the edge" of being over-stimulated and killed by Ca 2+ , but they survived. If these mast cells were reacted with a mAb that only modestly activated C and the Ca 2+ flux was inadequate, then perhaps synergy in killing could be achieved by coincident treatment with agents that stimulate FcεRI and/or chemokine receptor 1. This proof-of-concept experiment could be extended to target other stimulatory sites on tumor cells that are known to mediate Ca 2+ entry. Another approach would be to make use of a mAb-drug conjugate. If a Ca 2+ ionophore (A23187 or ionomycin) [23] could be stably coupled to a C-fixing mAb with no damage to the potential of the mAb to mediate C activation, then the ionophore could independently further increase Ca 2+ entry into the targeted cell to increase its cell-killing potential. A similar strategy could be based on coupling pore-forming agents such as melittin or perforin to C-fixing anti-tumor mAbs [23]. Indeed, immunoconjugates of melittin have been investigated for cancer immunotherapy [79]. Finally, as we suggested recently, a third strategy would be to develop a C-fixing antibody drug conjugate that blocks the machinery that pumps Ca 2+ out of cells, thus synergizing with the CDC action of the anti-tumor mAbs [16].

The Future
As new and effective C-fixing mAbs are developed for cancer immunotherapy, it is very likely that they will closely follow the patterns we have described here (Table 1). Indeed, in collaboration with investigators at Genmab and Drs. Clive Zent and Richard Burack at the University of Rochester, we recently reported on the properties of CD37 mAbs that contain the E430G hexamer-forming modification [32]. We found that these mAbs promote very high levels (>98%) of CDC of CLL cells taken from newly diagnosed patients. The very high levels of killing are likely due to the fact that most CLL cells express approximately twice as many CD37 epitopes as CD20 epitopes [32,80]. Moreover, based on four-color confocal microscopy real-time movies, we found that upon binding of the CD37-Hx mAb to CLL cells in the presence of NHS, C3b is rapidly colocalized with cell-bound CD37-Hx mAb, and the same pattern of rapid cell killing mediated by Ca 2+ influx is evident [80]. This includes generation of transition-state intermediates followed by poisoning of mitochondria, leakage of Fluo-4 out of the cell, and finally cell death, all within just a few minutes. We suggest that this cytotoxicity pattern can serve as a "litmus test" for evaluation of the potential of future mAbs intended to be used for cancer immunotherapy based on C-mediated killing.

Summary
As illustrated in Figure 16, we have examined and characterized, from the "point of view" of the cell, many of the key steps in the classical complement pathway that are activated when highly effective C-activating mAbs bind to tumor cells. The entire reaction is quite rapid, and both cell lines and primary CLL cells can be killed within just a few minutes due to influx of lethal amounts of Ca 2+ , clearly not allowing much time for any complex signaling pathways to be activated. The patterns we have described are likely to be quite general, and it may be possible to make use of the lessons we have learned in the development of additional innovative strategies that employ C and mAbs in the treatment of cancer and other diseases as well.  Figure 16. Simplified schematic illustrating the key steps in the CDC reaction that are described in this review. The reaction takes place at 37 • C in 50% NHS, and the mAb is at a concentration of 10 µg/mL. The target cell is internally labeled with Fluo-4. The pale green color of the cell illustrates the low, physiological levels of cytoplasmic Ca +2 . (1). Multiple mAbs bind to the cell in proximity, ideally forming hexamers. (2). C1q then binds to the mAb hexamers on the cell. (3). Downstream, C3b is deposited on the cell, closely associated with bound mAb. (4). After sufficient C3b has deposited on the cell, C5b is generated, and C5b-8 and/or C5b-9 assemble close to the sites of C3b deposition. (5). The pores formed in the plasma membrane by C5b-8 and/or C5b-9 allow the influx of large amounts of Ca 2+ and the cell turns bright green (the live, transition-state intermediate). (6). Soon thereafter, The Ca 2 in the cell penetrates and poisons the mitochondria. (7). This is rapidly followed by cell death. Coincident with cell death, most of the Fluo-4 leaks out of the cell. Not shown are the TNTs that are generated in the cells soon after the initial influx of Ca 2+ , during steps (5) and (6). Approximate times for these events are provided in Table 1, and all of the reactions are fast. The hexamer-forming mAbs and C1q are bound to the cell in less than 1 min, and the cells are killed in less than 2 min.

Patents
R.P.T. and M.A.L. are listed as co-inventors on a patent that describes the use of hexamerforming antibodies.