Variable sensitivity to complement-dependent cytotoxicity in murine models of neuromyelitis optica

To investigate the effect of HC on murine CNS cells, we assessed the sensitivity of mouse neurons, oligodendrocytes (OLs), oligodendrocyte progenitor cells (OPCs), and astrocytes to human serum in monocultures. Monocultures were exposed to either PBS as a control or 2–5% (vol/vol) HC. Cell death was measured by the addition of DRAQ7 to the media to label dead/dying cells, followed by IncuCyte live cell imaging. After 4 h of incubation with 2% HC, 90.0 ± 1.9% of the mouse neuronal population was DRAQ7+ (Fig. 1a–b). OLs and OPCs were also sensitive to HC exposure with 70.8 ± 2.8% and 42.2 ± 5.6% cell death, respectively, after 4 h incubation (Fig. 1c–f and Additional file 2: Movie 1 and Additional file 3: Movie 2). In contrast, HC did not induce astrocyte cell death as determined by LDH release assay (Fig. 1g) or IncuCyte live cell imaging (data not shown). The degree of cell death in these murine CNS monocultures in the presence of HC is summarized in Table 1.

In the CNS, the interaction of neurons and glia may alter their responses to environmental stressors. We investigated how mouse neurons and glial cells responded to HC in mixed cell cultures prepared from PLP-eGFP [20] mouse pups in which both OPCs and differentiated OLs were labeled with eGFP. Cell type specific marker staining showed that the neuroglial mixed cultures consisted of 70% astrocytes (GFAP and AQP4), 10–15% oligodendroglial cells (Olig1), and 10–15% neurons (βIII-Tubulin). Most oligodendroglial cells were OPCs, but occasionally maturing OLs (O4+) were noted (Additional file 1: Figure S1).

Cell death in mixed cultures was monitored by IncuCyte live imaging. In the mixed cultures, HC was not toxic to astrocytes (Fig. 2a, arrow heads). Neurons remained sensitive to HC in the mixed cultures (Fig. 2a, arrows); however, the magnitude of loss was significantly reduced. The addition of 5% HC caused 31.4 ± 4.0% neuronal death in the mixed culture after 4 h incubation (Fig. 2b and Additional file 4: Movie 3); whereas, HC resulted in greater than 90% neuronal death in the monoculture (Fig. 1b). Neuronal death was confirmed by co-labeling with the dead cell dye propidium iodide (PI) and neuronal marker NeuN (Fig. 2c).

In the neuroglial mixed cultures, the majority of eGFP+ oligodendrocyte lineage cells were bipolar and tripolar progenitors, while a minority of the cells displayed a highly complex network of processes consistent with differentiated OLs [26]. OLs died rapidly in the presence of HC alone with 56.0 ± 1.4% of OLs lost by 4 h (Fig. 2d). OL cell death was reduced, however, in mixed cultures when compared to monocultures (70.8 ± 2.8%, Fig. 1e). In contrast to their behavior in monoculture (Fig. 1d–e), OPCs were not sensitive to HC in the neuroglial mixed cultures, even after an extended incubation time of 3 days (Fig. 2e and Additional file 4: Movie 3). As observed in the monoculture, astrocytes remained insensitive to HC in the neuroglial mixed culture (Fig. 2a and Additional file 4: Movie 3). The percent of CNS cell death in mixed cultures is summarized in Table 1.

We hypothesized that astrocytes in mixed cultures provided protection against HC to adjacent oligodendrocytes and neurons. To test this, we used a recombinant monoclonal antibody specific for AQP4 (rAb #53) [12] to target astrocytes via complement-mediated lysis (CDC) [13]. Live staining of rAb #53 of astrocytes at 4 or 37 °C displayed a robust membrane staining and dim cytoplasmic staining, indicating only a small portion of AQP4 being internalized from the cell surface (Fig. 3a). In combination with HC, rAb #53 induced significant astrocyte death, resulting in a marked increase of LDH in the culture medium (Fig. 3b).

We monitored the kinetics of neuron, oligodendrocyte, and astrocyte death in the mixed cultures following targeted rAb #53-mediated astrocyte damage with IncuCyte live cell imaging. No cell death was noted with rAb #53 or isotype control (Iso) without the presence of HC. Treatment with rAb #53+ 5% HC induced rapid cell death, notably among astrocytes (Fig. 4a–b). Dead neurons were identified as small round DRAQ7+ particles with high fluorescence intensity (Fig. 4a, arrows), whereas dead astrocytes were distinguished as larger but less intense DRAQ7+ cells (Fig. 4a, arrow head). Dead astrocytes floated into the culture supernatant, resulting in areas of astrocyte depletion (Fig. 4a, circled area). After the peak of astrocyte death (4 h), viable astrocytes repopulated these areas (Additional file 5: Movie 4). Neuronal death increased in the presence of rAb #53-mediated astrocyte destruction, with approximately 70% of neurons dying within 2 h. While some neuronal death may result from their sensitivity to HC alone, only 30% neuronal death was observed with Iso+HC (Fig. 4c) or HC alone (Fig. 2b–c). Astrocyte loss induced by rAb #53 plus HC also increased the death of OLs (Fig. 4d). In contrast, rAb #53-mediated astrocyte destruction caused no loss of OPCs (Fig. 4e). The percent cell death is summarized in Table 1.

The sensitivity of neurons, OL, and astrocytes to HC was not dependent on the source of human sera. Both pooled and single donor samples of HC from different vendors were tested for cytotoxic activity on mixed neuroglial cultures and produced identical cytotoxic effects (Additional file 1: Figure S2 and Table S1). Consistent with previous reports [18, 19], we did not detect any cytotoxic effects of mouse serum on neurons or OLs in mixed culture in the presence or absence of AQP4-specific rAb #53 (Additional file 1: Figure S2 and Table S1).

In order to determine the sensitivity of CNS cells to HC in a complex tissue network, cell death was examined in mouse organotypic cerebellar slices. Slices treated with HC alone had no effect on astrocyte morphology and tissue structure (Fig. 5a). A low level of neuronal death was detected at 48 h following HC treatment: 17.4 ± 0.7% of calbindin-positive neurons were PI-positive (Fig. 5b–c). Neuronal death, however, was not limited to Purkinje cells, as coincident NeuN and PI staining showed a general loss of neurons throughout the slice, including the granule cell layer (Additional file 1: Figure S3).

In cerebellar slice cultures, all PLP-eGFP positive cells differentiated into mature oligodendrocytes and co-localized with CC1 (data not shown), a mature oligodendrocyte marker [27]; OPCs no longer expressed eGFP [20]. HC caused a significant loss of PLP-eGFP+ cell bodies from 24 h treatment. At 48 h, the number of eGFP+ oligodendrocyte cell bodies was reduced by 25% with some PLP-eGFP+ OLs displaying disorganized processes (Fig. 5d–e). OPCs were visualized with NG2 and Olig2 co-staining and did not express PLP-eGFP (Fig. 5f, arrows). No loss of OPCs was observed in cerebellar slices treated with HC alone for 48 h (Fig. 5f–g).

To further investigate the role of astrocyte injury on cell-specific survival within organotypic cultures, we treated cerebellar slices with NMO rAb #53+HC. In rAb #53+HC-treated slices, astrocytes became swollen, and the network of astrocytic processes was disrupted within 8 h. At 48 h, massive destruction of the astrocyte network was apparent (Fig. 6c–d). In contrast, no significant morphological changes in astrocytes were detected using an isotype control rAb+HC or rAb #53 alone (Fig. 6a–b). At 48 h, neuronal death in the slices treated with rAb #53+HC increased about threefold compared to HC alone: 52.3 ± 2.7% calbindin-positive neurons were PI-positive following rAb #53+HC (Fig. 7a–b), as compared with 17.4 ± 0.7% with HC alone (Fig. 5b–c). With increasing astrocyte destruction, the loss of OLs rapidly escalated: approximately 30% of OLs were lost at 12 h, whereas, at 48 h, about 50% of OLs were lost (Fig. 7c–d). The loss of OLs at 48 h was increased approximately 1.6-fold compared to HC alone (Fig. 5d–e). Surprisingly, approximately 50% of NG2+ Olig2+ OPCs were lost in slices by 24 h after treatment with rAb #53+HC (Fig. 7e–f, arrows), despite their lack of sensitivity to HC in neuroglial mixed cultures following astrocyte loss (Fig. 4e and Additional file 5: Movie 4) and their resistance to HC alone in organotypic cultures. The results are summarized in Table 1.

Because of the intrinsically weak activity of mouse complement [18] and the presence of complement inhibitor(s) in the mouse serum [19], complement sourced from human serum is required to trigger complement activation in NMO models. Our results demonstrate that murine neurons and oligodendroglia are sensitive to HC in monocultures and mixed cell cultures. Human neurons, however, did not demonstrate a similar sensitivity to HC (Additional file 1: Figure S4). Similarly, rat OPCs and OLs showed no sensitivity to rat serum (Additional file 1: Figure S4). These data suggest that human serum may contain cross-reactive antibodies that are mediating murine neuron and oligodendrocyte cell death. Using heat inactivation, which eliminated complement protein activity but maintained human antibody function, we observed binding of HC-derived human IgG to murine neurons, OLs, OPCs, and astrocytes (Additional file 1: Figure S5). Furthermore, IgG-depleted HC was not toxic to mouse neurons, OPCs or OLs in the monocultures or in mixed cell culture, and retained its ability to induce CDC on cultured astrocytes in the presence of rAb #53 (Fig. 8a–b and e–f). To determine if the complement pathway is required for HC-mediated cell death of murine cells, we heat inactivated or depleted HC of complement components (C1q, C3, C5, or C9) and applied it to murine monocultures. Inactivation of the complement pathway prevented HC-mediated cell death of murine cells (Fig. 8c–f). Taken together, these results suggest cross-reactive antibodies in human sera bind to the surface of mouse cells and may activate the complement cascade.

Complete activation of the complement cascade leads to the formation of the terminal membrane attack complex (MAC) and cell lysis. To determine which cells are targeted for MAC formation, we used selective depletion of proteins along the classical and common pathways to prevent cell death and localize the deposition of activated complement components. Using C5-depleted HC, we arrested complement activation following C3 proteolytic cleavage resulting in C3d deposition on complement-targeted cells (Fig. 9a). In neuronal monocultures treated with C5-depleted HC, C3d was detected on the cell surface. When neuronal monocultures were treated with C4-depleted human complement (Fig. 9b), there was no C3d deposition, indicating that activation of the classical pathway was essential for neuronal injury with HC alone. In neuroglial mixed cultures treated with C5-depleted HC, C3d was detected on the surface of neurons and OLs (arrows and arrow heads, respectively), but not OPCs or surrounding astrocytes (Fig. 9c). Since HC-derived IgG bound to murine neurons, OLs, OPCs, and astrocytes (Additional file 1: Figure S5), the resistance of OPCs and astrocytes to CDC in mixed culture was either due to the local expression of complement inhibitors or the inability of surface-bound human IgGs to reach a sufficient threshold to activate the complement cascade. When mixed cultures were treated with NMO rAb #53 and C5-depleted HC, C3d deposits were found on the surface of neurons (arrows), OLs (arrow heads), and surrounding astrocytes (stars), but not on OPCs (Fig. 9c). In organotypic cerebellar slices treated with C5-depleted or complete HC, deposits of C3d and MAC were detected along fibers containing neuro-filaments (Fig. 9d, arrows) and surrounding oligodendrocyte cell bodies (Fig. 9d, arrowheads). No detectable C3d or MAC deposits were noted in slices exposed to medium alone (data not shown). These results demonstrate that complement in human serum is activated by human anti-murine cross-reactive antibodies in primary cell monocultures, mixed cultures, and cerebellar slice cultures.