Major Histocompatibility Complex Class I Shedding and Programmed Cell Death Stimulated Through the Proinflammatory P2X₇ Receptor: A Candidate Susceptibility Gene for NOD Diabetes

NOD mice carrying an H2-E transgene (NOD.E), as previously described (23), were bred in the Central Biological Services Unit at Hammersmith Hospital and were a generous gift from Prof. E. Simpson. Institute guidelines for the care of laboratory animals were followed. C57BL/10 mice were purchased from Harlan Olac (Bicester, U.K.). All mice used were age 6–14 weeks. All reagents used were obtained from Sigma (Poole, U.K.), unless otherwise stated.

Mesenteric lymph node cells were prepared from adult mice. Cell suspensions (10⁷/ml) in phenol red–free Dulbecco’s modified Eagle’s medium (DMEM) were stained with the indicated combination of CD4^APC, CD8^APC, CD8^PE, CD62L^FITC (Becton Dickinson, Sunnyvale, CA), or anti–D^b-FITC (clone B220.249; gift from Dr. M. Millrain) antibodies to permit separation of different cell types during subsequent analysis, washed, and resuspended in DMEM. Cells were then equilibrated with propidium iodide together with annexin V^FITC or annexin V^CY5 (Becton Dickinson), as appropriate, for 4 min and analyzed by flow cytometry on a FACScalibur machine. Data were analyzed using CellQuest (Becton Dickinson) or FlowJo (Treestar) software. Baseline fluorescence was established for ∼1 min before benzoyl ATP (BzATP) was added. Debris was excluded from analysis based on forward and side scatter.

Stimulation of the P2X₇ receptor by ATP, or more potently by its analog BzATP, results in translocation of PS to the outer leaflet of the plasma membrane (12) and rapid shedding of the lymphocyte homing receptor CD62L (13). These responses are absent in lymphocytes from mice lacking P2X₇ (11). After brief stimulation of the P2X₇ receptor, exposure of PS is reversible and does not initiate programmed cell death (12). To compare the sensitivities of NOD background and C57BL/10 lymphocytes to P2X₇-dependent stimulation, we developed real-time flow cytometric assays in which the responses of the two populations could be compared in a single tube. To avoid potential influences resulting from disease, lymphocytes from NOD mice expressing a type 1 diabetes–protective H2-E transgene (NOD.E) (23) were used. Lymphocytes from NOD.E and C57BL/10 mice were labeled with PE- and CYCHROME-conjugated anti-CD4 antibodies, respectively, to allow cells from different mice to be distinguished and then mixed. Cells were stained with fluorescein isothiocyanate (FITC)-conjugated anti-CD62L antibody to monitor cell surface CD62L and with CY5-conjugated annexin V to monitor cell surface exposure of PS. After stimulation with BzATP, NOD.E lymphocytes exhibited markedly greater rates of PS translocation and CD62L shedding than C57BL/10 lymphocytes (Fig. 1).

It has been reported that MHC class I expression is low in NOD mice (24) and autoimmune patients (25–27), although these findings are controversial (28–30). Given the increased P2X₇ receptor–induced shedding of CD62L in NOD.E mice, we tested the hypothesis that P2X₇ stimulation can result in the loss of other cell surface molecules. In particular, we questioned whether low levels of cell surface MHC class I molecules (if apparent) might reflect P2X₇-dependent shedding and not defective constitutive expression. Consistent with this hypothesis, P2X₇ stimulation resulted in significant shedding of MHC class I (Fig. 2, Table 1) and CD4 and CD8 (data not shown). In each case, although the extent of shedding was less than that observed for CD62L, loss after P2X₇ stimulation was greater from NOD.E lymphocytes than from C57BL/10 lymphocytes. Importantly, although NOD.E and C57BL/10 lymphocytes were mixed and stained with anti–MHC class I (anti-D^b) antibody in a single tube to minimize experimental artifacts, initial cell surface expression of MHC class I (D^b) was not significantly different. Given the suggestion that decreased levels of MHC class I might reflect the smaller size of NOD lymphocytes (31), cell volume was also compared. There was little or no initial size difference between lymphocytes from the two strains, although P2X₇-stimulation resulted in the marked shrinkage (albeit followed by size recovery) of NOD.E, but not C57BL/10, lymphocytes (Fig. 2).

NOD lymphocytes have been reported to be resistant to a range of stimuli that induce programmed cell death (32–35). Because prolonged (though not brief) P2X₇ stimulation results in aponecrosis and NOD lymphocytes are relatively sensitive to P2X₇ stimulation, NOD lymphocytes might be expected to be sensitive, rather than resistant, to P2X₇-dependent cell death. Indeed, NOD lymphocytes (and thymocytes [data not shown]) were found to be significantly more susceptible than those from C57BL/10 mice to P2X₇-induced aponecrosis, as indicated by terminal cell membrane rupture and consequent uptake of propidium iodide (Fig. 3). Thus, the relative resistance or sensitivity of NOD lymphocytes to programmed cell death is stimulus dependent.

In rodents and humans, neonatal pancreatic β-cells undergo a wave of physiological cell death in a process known as tissue remodeling (5–7,36,37). Although usually benign, this event, at least in the NOD mouse strain, may trigger leukocytic infiltration and, eventually, autoimmune destruction and diabetes (5–7). Similarly, the diabetogenic activity of both coxsackie B4 virus and streptozotocin result from induction of damage to β-cells (8,9). These observations suggest that the early stage of disease may reflect an aberrant response to immune “danger” signals. Two such signals released at sites of tissue damage are ATP and NAD, both of which activate the proinflammatory P2X₇ receptor (11,12,38). Indeed, given the central role of the P2X₇ receptor in IL-1β secretion and shedding of the lymphocyte homing receptor, CD62L, both of which are involved in the pathogenesis of NOD diabetes (14,15,18,20,39), any factors affecting the expression, regulation, or innate sensitivity of this receptor may affect disease susceptibility. This raises the possibility that the gene encoding the NOD P2X₇ receptor might be a candidate susceptibility gene in type 1 diabetes.

Different mouse strains express one of two common alleles of the P2X₇ receptor (21), with that being expressed by NOD mice having a relatively low activation threshold. Because the expression of a single allelic difference (e.g., P2X₇-P) can be a poor predictor of complex physiological events, we compared the sensitivity of NOD and C57BL/10 lymphocytes to P2X₇-induced shedding of CD62L and MHC class I, externalization of PS, and aponecrosis. All were greater in NOD than in C57BL/10 lymphocytes. Thus, although the resistance of NOD lymphocytes to a range of stimuli that induce programmed cell death has been implicated in disease pathogenesis (32–35), our data indicate that sensitivity and resistance to programmed cell death is stimulus dependent; both properties may promote disease. It is important to note that we have used programmed cell death as a nonmechanistic term, defined as any one of a range of active processes stimulated through a physiological receptor by a relevant agonist or through the removal of a stimulus required for cell viability. However, there is no generally accepted terminology. Although it is often assumed that apoptosis implies caspase-dependent cell death, many studies on caspase-independent apoptosis exist. Furthermore, whereas apoptosis and programmed cell death are frequently used as synonyms, apoptosis has been used more recently to describe one end of a continuum of active mechanisms called programmed cell death (40). Nevertheless, it is clear that caspase involvement is not a good indicator of the physiological importance, or “programming,” of a cell death pathway (40). Indeed, programmed cell death can occur in the complete absence of caspase activity; even within programmed cell death in response to a given stimulus, caspase dependence can be variable, perhaps dependent on cell type or differentiation state. ATP-induced cell death has, for example, been found independent of caspase-1 (41), except after lipopolysaccharide-dependent potentiation of macrophages (42). It is necessary, therefore, to determine to which forms of the spectrum of death pathways known as programmed cell death NOD lymphocytes are resistant or sensitive. Finally, because ATP is released from dying cells, P2X₇ receptor–induced cell death may be an important confounding factor in some studies of NOD apoptosis (34,43).

Our data clarify a long-standing controversy over the level of MHC class I molecules on lymphocytes in diabetes. According to one hypothesis, low expression of MHC class I in cells from NOD mice and patients with a variety of autoimmune conditions predisposes the subject toward disease (24–27). By contrast, other reports suggest that NOD MHC class I expression is only slightly reduced (28) or simply reflects a size difference with control cells (29) and that MHC class I levels are normal in patients with type 1 diabetes (30). Consequently, whether or why autoimmune disease is associated with reduced expression of MHC class I molecules is hotly disputed (26,44). We showed here that lymphocytes from type 1 diabetes–resistant NOD mice expressing an H2-E transgene (to minimize experimental results that are the consequence, not cause, of disease) express normal levels of MHC class I molecules, but that they are rapidly shed after P2X₇-stimulation. Loss of MHC class I molecules was moderate (15–20%), corresponding with the relative underexpression of MHC class I reported by others (28). Hence, although our data support the notion that lymphocyte MHC class I expression may be decreased in type 1 diabetes, they suggest that this phenotype is the result, not the cause, of disease. This interpretation is also consistent with observations of increased serum levels of CD62L, CD4, CD8, and MHC class I during inflammation (45–47).

One test of this hypothesis that the P2X₇-P allele predisposes toward NOD diabetes is whether the region encompassing P2X₇ (the distal end of chromosome 5) appears as a susceptibility locus in a backcross between NOD mice and any strain expressing the alternative, high-threshold P2X₇ allele. On this point, the data are inconclusive. Thus, although this region (identified by D5Mit43) was associated (χ² = 4.4) with disease in the progeny of a backcross of (B10.H2^g7 x NOD)F₁ with NOD mice (48), at this relatively low level of significance, the association may have been a false-positive result. Perhaps equally likely, however, is the possibility that the backcross analysis used underestimated the potential role of the P2X₇-P allele. Thus, given that P2X₇-P and -L alleles differ in their stimulation threshold, then depending on the level of extracellular ATP within the inflamed NOD pancreas, a single copy of P2X₇-P may have been sufficient to confer sensitivity to activation. If this were the case, P2X₇ would not have been detected as a susceptibility gene in the backcross used.

We have shown that, compared with normal lymphocytes, NOD lymphocytes are highly sensitive to P2X₇ receptor–stimulated programmed cell death and the shedding of CD62L and other cell surface markers such as MHC class I. These and other data indicate that P2X₇ is a good candidate susceptibility gene in murine autoimmune diabetes. It is attractive to speculate that defective clearance of neonatal apoptotic pancreatic β-cells by NOD mice (49) results in sufficient ATP release to stimulate the low-stimulation threshold allele of P2X₇ (P2X-P) and consequent secretion of IL-1β from macrophages and CD62L shedding and extravasation of T-cells. Subsequent T-cell–mediated killing of pancreatic β-cells and consequent release of ATP might then exacerbate the P2X₇-dependent inflammatory cascade. Although the P2X₇ receptor is encoded by a gene in a region weakly (and thus potentially falsely) associated with susceptibility to diabetes in a backcross analysis, its effects may be underestimated by such studies (48). Establishing a congenic NOD strain bearing the distal region of C57BL chromosome 5 will help determine the role that P2X₇ allelic variation plays in type 1 diabetes.