C1q restrains autoimmunity and viral infection by regulating CD8+ T cell metabolism

See allHide authors and affiliations

Science  04 May 2018:
Vol. 360, Issue 6388, pp. 558-563
DOI: 10.1126/science.aao4555

Complement is a CD8+ T cell metabolic rheostat

Systemic lupus erythematosus (SLE) is associated with deficiencies in the complement protein C1q. Although C1q plays a role in the clearance of apoptotic cells, there are several redundant clearance pathways. Disruption of one pathway does not lead to an autoimmune defect. In a chronic graft-versus-host disease model of SLE, Ling et al. show that C1q dampens CD8+ T cell responses to self-antigens. C1q modulates metabolism through the mitochondrial cell-surface protein p32/gC1qR. The lack of C1q during a viral infection also enhances CD8+ T cell responses. Thus, C1q plays a role as a “metabolic rheostat” for effector CD8+ T cells.

Science, this issue p. 558


Deficiency of C1q, the initiator of the complement classical pathway, is associated with the development of systemic lupus erythematosus (SLE). Explaining this association in terms of abnormalities in the classical pathway alone remains problematic because C3 deficiency does not predispose to SLE. Here, using a mouse model of SLE, we demonstrate that C1q, but not C3, restrains the response to self-antigens by modulating the mitochondrial metabolism of CD8+ T cells, which can themselves propagate autoimmunity. C1q deficiency also triggers an exuberant effector CD8+ T cell response to chronic viral infection leading to lethal immunopathology. These data establish a link between C1q and CD8+ T cell metabolism and may explain how C1q protects against lupus, with implications for the role of viral infections in the perpetuation of autoimmunity.

Systemic lupus erythematosus (SLE) is an autoimmune condition that develops as a result of complex genetic and environmental interactions. B and CD4+ T cell abnormalities are well known features of SLE (1), but the role of CD8+ T lymphocytes remains poorly understood. Transcriptomic data suggest that a CD8+ T cell signature can predict disease outcome (2, 3).

There is evidence for a strong association between SLE and complement C1q deficiency (4). Previous work has shown that C1q deficiency leads to the ineffective clearance of apoptotic cells and consequently enhanced exposure to self-antigens, which facilitate autoimmunity (5). However, there are multiple pathways, including those mediated by C3, through which apoptotic cell clearance occurs (6). This suggests that the contribution of C1q is redundant and the “waste disposal” hypothesis (5) is inadequate to fully explain why C1q deficiency, and not C3 deficiency, is associated with autoimmunity. Alternative, but not mutually exclusive, hypotheses have been proposed (7). However, an explanation for this strong association in terms of classical complement pathway abnormalities alone remains unsatisfactory. Given that there is evidence that C1q has multiple roles that are independent of complement activation (8), we searched for an alternative function that could explain why C1q is so critical for maintaining self-tolerance.

Chronic graft-versus-host-disease (cGvHD) is a well-established inducible model of SLE. We used the bm12-cGvHD model (9) and injected splenocytes from B6(C)-H2-Ab1bm12/KhEgJ (bm12) mice into coisogenic C57BL/6 (B6) recipients lacking C1q (C1qa−/−) or C3 (C3−/−). Lupus autoantibody levels were similar at disease onset (weeks 0–4), but increased at later time points only in the C1qa−/− mice (Fig. 1A). At week 10, C1qa−/− mice displayed more severe glomerulonephritis and increased glomerular deposition of immunoglobulin G (IgG) and C3 (Fig. 1B). They also had splenomegaly with higher percentages of germinal center B cells, follicular helper T cells (TFH), activated CD4+ and CD8+ T cells than the wild-type (WT) and C3−/− counterparts (fig. S1). During the course of the disease there were no differences in blood B and CD4+ T cell activation between experimental groups, but the proportion of activated CD44hiCD62LCD8+ T cells in the C1qa−/− mice was increased with a relative expansion of KLRG1+IL-7R short-lived effector cells (SLECs) and a corresponding reduction in KLRG1IL-7R+ memory precursor effector cells (MPECs) (Fig. 1C) (10). Consistent with the alterations in blood, cGvH-treated C1qa−/− mice had early (from week 1) splenic CD8+ T cell abnormalities, whereas the initial B and CD4+ T cell responses were similar to WT and C3−/− animals (fig. S1). Furthermore, the in vitro restimulation of C1qa−/− CD8+ T cells resulted in increased interferon-γ (IFN-γ) and granzyme B expression and fewer IL-2+ cells (fig. S2), indicating that the lack of C1q, but not of C3, resulted in CD8+ T cell responses skewed toward an effector phenotype. To determine whether bystander inflammation or self-antigen stimulation promoted CD8+ T cell activation during bm12-cGvH induction, naïve CD8+ T cells from B6.CD45.1+ and ovalbumin (OVA)–specific T cell receptor (TCR) transgenic (OT-I) mice were cotransferred into B6.CD45.2+ animals, which were challenged with bm12 splenocytes. Donor CD45.1+CD8+ T cells expanded and became activated like the host CD45.2+CD8+ T cells, whereas pentamer+ OT-I cells remained quiescent (fig. S3), suggesting that TCR engagement by self-antigen was required.

Fig. 1 Autoimmune features and CD8+ T cell response in C1qa−/− mice after bm12-cGvH induction.

(A) Autoantibody levels after bm12 injection (arrows) (n = 5 mice per group). (B) IgG, C3, and periodic acid–Schiff (PAS) staining of kidney sections at week 10. Quantification of glomerular IgG and C3 deposition expressed as arbitrary fluorescence units (AFU). ND, not detectable. Glomerulonephritis score: 0 to 4, bars indicate the median; *P < 0.05; Kruskal-Wallis H test. (C) Flow cytometric analysis of blood cells after cGvH induction (arrows) (n = 8 to 10 mice per group). (D and E) C1qa−/− mice were administrated phosphate-buffered saline (PBS), antibody to CD8α (anti-CD8α), or isotype-matched IgG2b antibody (n = 4 to 7 mice per group). (D) Autoantibody and IgG levels after cGvH induction (arrows). (E) Images and quantification of glomerular IgG and C3 deposition at week 10. *P < 0.05, **P < 0.01, ****P < 0.001; one-way analysis of variance (ANOVA) [(B) and (E)]; *P < 0.05, **P < 0.01, ***P < 0.005, ****P < 0.0001 (WT versus C1qa−/−), #P < 0.05, ##P < 0.01, ###P < 0.005, ####P < 0.0001 (C1qa−/− versus C3−/−) two-way ANOVA [(A) and (C)]; *P < 0.05, **P < 0.01, ***P < 0.005 (isotype versus anti-CD8α), ##P < 0.01, ###P < 0.005, ####P < 0.0001 (PBS versus anti-CD8α) two-way ANOVA (D). Data are mean ± SEM unless indicated otherwise; pooled results of two experiments (C); representative of two [(D) and (E)] or three [(A) and (B)] experiments. Scale bars, 100 μm [(B) and (E)].

CD8α+ dendritic cells (DCs) cross-present apoptotic cell-associated antigens to CD8+ T cells (11). However, the cross-priming by CD8α+ DCs in C1qa−/− animals was not impaired (fig. S4, A to C). Furthermore, after cGvH induction, the number and phenotype of CD8α+ DCs was unaffected by C1q deficiency (fig. S4, D and E). We then depleted CD8+ T cells to demonstrate their direct contribution to the autoimmune response in cGvHD. Although similar autoantibody levels were initially detected in all groups (Fig. 1D), from week 4 onwards, CD8+ T cell-depleted C1qa−/− mice displayed a progressive decline in lupus-associated autoantibodies, whereas total IgG levels remained unaffected (Fig. 1D). CD8+ T cell-depleted C1qa−/− mice also showed reduced glomerular deposition of IgG and C3 compared with nondepleted mice (Fig. 1E). Thus, these data suggest that CD8+ T cells are responsible for perpetuating the lupus-like disease observed in cGvH-treated C1qa−/− mice.

To explore whether C1q also modulates CD8+ T cell immunity during infections, we used lymphocytic choriomeningitis virus (LCMV) models. C1qa−/− mice, subjected to acute LCVM-Armstrong (Arm) infection, had an aberrant effector LCMV-specific CD8+ T cell response at day 8 (Fig. 2A and fig. S5A), but did not show markedly different memory and recall responses (fig. S5, B to D). We next used the chronic LCMV-clone 13 (Cl13) model where an exaggerated effector immune response can cause lethal lung immunopathology (12). When compared with WT mice, Cl13-infected C1qa−/− mice experienced greater body weight loss and had to be culled at day 11 (Fig. 2B). Examination of C1qa−/− lung tissue showed edema that was absent in the controls (Fig. 2C and fig. S6A). Consistent with a vigorous CD8+ T cell response, C1qa−/− mice showed increased LCMV-specific gp33+ and gp276+CD8+ T cell populations (Fig. 2D and fig. S6B). On day 8, when C1qa−/− mice still had numbers of LCMV-specific CD8+ T cells comparable to WT animals, virus-specific C1qa−/− CD8+ T cells were functionally overreactive, with enhanced degranulation and cytokine production (Fig. 2, E and F). Consistent with an enhanced CD8+ T cell response, serum viral loads were lower in Cl13-infected C1qa−/− mice compared with WT mice (Fig. 2G). Moreover, the up-regulation of PD-1 expression was similar in WT and C1qa−/− LCMV–specific CD8+ T cells, indicating that C1q deficiency did not impair the PD-1 signaling pathway (Fig. 2H). These findings demonstrate that C1q plays a pivotal role in regulating effector CD8+ T cell responses in both autoimmunity and viral infection.

Fig. 2 Essential role for C1q in chronic LCMV infection.

(A) Numbers of splenic np396+, gp33+, and gp276+ CD8+ T cells and proportions of SLECs and MPECs among LCMV-specific CD8+ T cells in LCMV-Arm–infected WT and C1qa−/− mice at day 8 (n = 5 to 8 mice per group). (B to G) Analysis of WT and C1qa−/− mice infected with LCMV-Cl13. (B) Percentage of body weight loss (n = 5 mice per group). (C) Representative lung histology on day 11. Scale bars, 100 μm. (D) Numbers of splenic LCMV-specific CD8+ T cells at day 11 (n = 5 mice per group). [(E) and (F)] Percentages of CD8+ T cells positive for CD107a and the proportion of LCMV-specific CD8+ T cells producing IFN-γ, tumor necrosis factor–α (TNF-α), and interleukin-2 (IL-2) after incubation with LCMV gp33 peptide (E) or gp276 peptide (F) at day 8 (n = 6 mice per group). (G) Serum viral load measured using quantitative polymerase chain reaction (PCR). (H) PD-1 expression on LCMV-specific CD8+ T cells at day 8 after LCMV-Cl13 and LCMV-Arm. NS, not significant; *P < 0.05, **P < 0.01, ****P < 0.0001; unpaired Student’s t test [(A) and (D) to (F)]; two-way ANOVA (B). Data are mean ± SEM and representative of two experiments.

Complement can mediate its cellular effects via both extracellular and intracellular pathways (13). To explore how C1q affect CD8+ T cells, we cotransferred naïve CD8+ T cells, isolated from B6.CD45.1+ and C1qa−/−.CD45.2+ mice into B6.CD45.1+.CD45.2+ mice, which were challenged with bm12 CD4+ T cells one day later. C1q-sufficient and C1q-deficient donor CD8+ T cells showed similar expansion and activation, suggesting a cell-extrinsic effect of C1q (fig. S7). We corroborated this using the lymphopenia-induced proliferation model by cotransferring carboxyfluorescein diacetate succinimidyl ester (CFSE)–labeled CD8+ and CD4+ T cells from B6.CD45.1+ mice into irradiated B6.CD45.2+ and C1qa−/−.CD45.2+ mice. Fourteen days later, donor B6.CD45.1+CD8+ T cells showed greater proliferation and activation in C1qa−/−.CD45.2+ recipients, whereas the cotransferred B6.CD45.1+CD4+ T cells were unaffected (Fig. 3A and fig. S8A). As in the bm12-cGvHD model, C1q operated independently of C3 (fig. S8B). Lymphopenia-induced T cell expansion is triggered by low-affinity interactions (14). Analysis of OT-I proliferation with OVA peptides of different affinities showed that C1q had an inhibitory effect only in response to partial (T4) and weak (G4) agonists but not to a strong (N4) ligand (fig. S8C). Similarly, C1q inhibited human CD8+ T cell activation, proliferation, and cytotoxic functions under suboptimal stimulation (fig. S9). C1q was detected mainly on activated CD8+ T cells (mouse and human) (Fig. 3B and fig. S10A) and almost exclusively on MPECs (Fig. 3C). Preincubation with the globular C1q region, but not with the collagen tail, inhibited C1q binding in a dose-dependent manner (fig. S10, B and C), indicating that C1q recognizes activated CD8+ T cells through its globular domain. Correspondingly, expression of the globular C1q receptor (p32/gC1qR) (15), a mitochondrial molecule present on the surface of several immune cells (fig. S11A), was increased on activated mouse and human CD8+ T cells (fig. S11, B and C). Consistent with the preferential binding of C1q to MPECs (Fig. 3C), cGvH-treated C1qa−/− MPECs expressed lower levels of the anti-apoptotic factor Bcl-2 and higher levels of Blimp-1, a repressor that promotes cytotoxic T functions (16), than WT MPECs (Fig. 3D and fig. S12, A and B). Furthermore, the proportion of C1qa−/− MPECs, but not SLECs, secreting granzyme B was higher compared with WT cells (fig. S12C). Abnormal Bcl-2 expression in C1qa−/− mice suggests that C1q may influence MPEC viability. In cGvH-treated C1qa−/− mice, MPECs, but not SLECs, displayed a higher rate of bromodeoxyuridine (BrdU) decay compared with WT animals (Fig. 3E), indicating a more rapid turnover of this subpopulation. Moreover, the percentage of C1qa−/− MPECs expressing active caspase 3/7 was higher (Fig. 3F and fig. S12D). Altogether, these findings suggest that C1q controls the programming and survival of MPECs through its globular domain.

Fig. 3 C1q selectively regulates MPEC programming and survival.

(A) Analysis of CFSE+CD45.1+CD8+ and CD45.1+CD4+T cells cotransferred in sublethally irradiated CD45.2+ WT or C1qa−/− hosts (n = 5 mice per group). Percentages of activated and fast proliferating donor T cell subsets in spleen on day 14. (B) Flow cytometric gating of C1q staining on blood CD8+ T cells at day 14 after cGvHD. (C) KLRG1 and IL-7R expression in WT CD44+CD8+ T cells (left); histogram of C1q staining (middle) and quantification (right) on MPECs and SLECs. Dotted line indicates mean fluorescence intensity (MFI) of isotype control. Each symbol represents a mouse. (D) Expression of transcription factors in splenic SLECs and MPECs from WT and C1qa−/− mice 2 weeks after cGvH induction (n = 3 to 5 mice per group). (E) Decay of Brdu+ SLECs and Brdu+ MPECs over 6 days (from day 11 after cGvH induction) in WT and C1qa−/− mice. Half-life times (t1/2) of the decay and the r2 value of the linear regressions are indicated (n = 6 mice per group). (F) Fractions of splenic SLECs and MPECs caspase 3/7+ at week 3 after cGvHD (n = 6 mice per group). NS, not significant; *P < 0.05, **P < 0.01; unpaired Student’s t test [(A), (C), (D), and (F)]; two-way ANOVA (E). Data are mean ± SEM and representative of three experiments.

CD8+ T cells undergo major metabolic changes upon activation (17). CD44+CD62L+CD8+ (MPECs) and CD44+CD62LCD8+ (SLECs) T cells from cGvH-treated C1qa−/− animals exhibited similar extracellular acidification rate (ECAR) and basal oxygen consumption rate (OCR) when compared with WT cells. However, C1qa−/− MPECs, but not SLECs, had impaired mitochondrial spare respiratory capacity (SRC) (Fig. 4, A to C). SRC has been shown to correlate with mitochondrial mass (17), and MitoTracker staining showed reduced mitochondrial content in C1qa−/− MPECs compared with WT MPECs (Fig. 4D). The addition of C1q increased the MitoTracker staining (Fig. 4E) and up-regulated the expression of mitochondrial biogenesis genes, such as Tfam and Ppargc1b, in IL-15–differentiated memory-committed OT-I cells but not in IL-2–differentiated effector-like cells (Fig. 4F). Consistent with a C1q-dependent pathway regulating MPEC mitochondrial biogenesis, in vitro metabolic conditions favoring a MPEC molecular profile (18) promoted p32/gC1qR surface expression on activated CD8+ T cells (Fig. 4G and fig. S13). The internalization of surface-bound C1q occurred via an endocytic pathway (fig. S14), and C1q colocalized with p32/gC1qR in the mitochondria (Fig. 4H).

Fig. 4 C1q regulates mitochondrial metabolism in MPECs.

(A and B) ECAR and OCR under basal conditions and after mitochondrial inhibitors in sorted splenic CD8+ T cells from WT and C1qa−/− mice 2 weeks after cGvHD. Curve shows mean ± SEM of four technical replicates of pooled samples from four animals. Data are representative of three experiments. (C) SRC of the CD8+ T cell subsets, as in (A) and (B). Each symbol represents a biological replicate. (D) MTDR staining in SLECs and MPECs 2 weeks after cGvHD (n = 3 mice per group). (E) MTDR staining of in vitro IL-2 (TE) and IL-15 (TM)–differentiated OT-I cells with and without hC1q (n = 6 mice per group). (F) Mitochondrial gene expression in TE and TM cells with and without hC1q (n = 5 mice per group). (G) Percentages of activated CD8+ T cells expressing p32/gC1qR under different metabolic conditions (n = 4 mice per group). (H) Confocal images of TM cells cultured with hC1q, stained with MitoTracker (red), anti-C1q (green), anti-p32/gC1qR (cyan), and 4′,6-diamidino-2-phenylindole (DAPI) (blue). Scale bar, 5 μm. NS, not significant; *P < 0.05, **P < 0.01, ***P < 0.005, ****P < 0.0001; two-way ANOVA [(A) and (B)], unpaired Student’s t test [(C) to (G)]. Data are mean ± SEM and representative of three experiments [(D) to (G)]. FCCP, fluorocarbonylcyanide phenylhydrazone; ΔMFI = MFI − FMO (fluorescence minus one); MTDR, MitoTracker deep red; 2DG, 2-deoxyglucose.

Altogether, these data link C1q to the metabolic reprogramming and regulation of activated CD8+ T cells and lead us to propose a new paradigm for the protective role of C1q in SLE: C1q limits tissue damage and autoimmunity by acting as a “metabolic rheostat” for effector CD8+ T cells that are capable of propagating autoimmunity through the generation of unique autoantigen fragments by granzyme B (19, 20) (fig. S15). The role of CD8+ T cells in SLE has been largely overlooked and remains poorly characterized, with conflicting findings in human and animal studies (2024), perhaps reflecting a changing role of these cells at different stages of the disease. By uncovering the role of effector CD8+ T cells in a lupus-like disease associated with C1q deficiency, our data demonstrate that an aberrant effector CD8+ T cell response to viral infection may auto-amplify the breakdown of self-tolerance. This is in addition to molecular mimicry and the bystander activation of autoreactive T cells (25). Furthermore, very little is known about the metabolic profile of CD8+ T cells in SLE. Because a CD8+ T cell transcriptional signature can predict the clinical outcome (2, 3), it is conceivable that metabolic abnormalities in these cells play a key role. Our study showing that C1q, a key lupus susceptibility gene in humans, can influence the mitochondrial metabolism of CD8+ T cells demonstrates this link. As p32/gC1qR is ubiquitously present in mitochondria and is indispensable for mitochondrial function (26), we hypothesize that the surface expression of p32/gC1qR coupled with another receptor may determine the specificity of the cellular effect(s) mediated by C1q. Thus, our findings describe a new paradigm to explain how C1q may prevent lupus flares and highlight the importance of the interplay between complement and immunometabolism in autoimmunity.

Supplementary Materials

Materials and Methods

Figs. S1 to S15

References (2738)

References and Notes

Acknowledgments: We thank the staff of the Imperial Central Biomedical Services for the care of the animals. We thank L. Lawrence for histological processing of the samples, H. T. Cook for histological analysis, N. Shaikh for technical support, D. Carling and G. Chennell for Seahorse analysis, C. Reis e Sousa for providing the OVA mouse embryonic fibroblasts, and E. Simpson for critical reading of the paper. Funding: This work was supported by the Wellcome Trust (grant reference number 108008/Z/15/Z) (to M.B.). J.A.H. is a recipient of a Wellcome Trust and Royal Society Sir Henry Dale Fellowship (101372/Z/13/Z). We also acknowledge a contribution from the National Institute for Health Research (NIHR) Biomedical Research Centre based at Imperial College Healthcare NHS Trust and Imperial College London. The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR, or the Department of Health. Author contributions: G.S.L. performed the experiments with assistance from G.C., I.Bar., Z.B., K.T., and S.R.; N.M.T., I.Bal., and P.G.A-R. provided key reagents; J.A.H. and J.S. assisted with data analysis and interpretation; M.B. supervised and conceived the project; and M.B. and G.S.L. wrote the paper. All authors commented on the manuscript. Competing interests: Authors declare no competing interests. Data and materials availability: All data are available in the main text or the supplementary materials.

Stay Connected to Science

Navigate This Article