ARTHRITIS & RHEUMATISMVol. 52, No. 1, January 2005, pp 201–211DOI 10.1002/art.20745© 2005, American College of Rheumatology

Increase in Activated CD8� T Lymphocytes ExpressingPerforin and Granzyme B Correlates With Disease Activity in

Patients With Systemic Lupus Erythematosus

Patrick Blanco,1 Vincent Pitard,1 Jean-Francois Viallard,2 Jean-Luc Taupin,1

Jean-Luc Pellegrin,2 and Jean-Francois Moreau1

Objective. Cytotoxic T lymphocyte–mediated kill-ing using granzyme B has recently been proposed to bea preferential and selective source of autoantigens insystemic autoimmune diseases, including systemic lu-pus erythematosus (SLE), while other reports haveindicated that cytolytic activity in SLE patients wasdecreased. The aim of this study was to examine thephenotypic and functional status of the CD8� T cells inSLE patients.

Methods. Phenotype analysis of CD8� T cells wascarried out using flow cytometry. The cytotoxic potentialof CD8� T cells and its consequences were examined inredirected-killing experiments. SLE patients with qui-escent disease (n � 41) were compared with SLEpatients with active disease (n � 20), normal individu-als (n � 36), and control patients with vasculitis (n �14). Cytotoxic CD8� T cell differentiation was exam-ined by coculture with differentiated dendritic cells(DCs) in the presence of SLE patient sera.

Results. Patients with disease flares were charac-terized by higher proportions of perforin- and/or gran-zyme B–positive lymphocytes with a differentiated effec-tor phenotype (CCR7� and CD45RA�). The frequencyof these cells in peripheral blood correlated with clinicaldisease activity as assessed by the SLE Disease Activity

Index. These cells generated high amounts of solublenucleosomes as well as granzyme B–dependent uniqueautoantigen fragments. Finally, the activation of DCswith serum from a patient with active lupus inducedgranzyme B expression in CD8� T lymphocytes.

Conclusion. DCs generated in the presence ofsera from SLE patients with active disease could pro-mote the differentiation of CD8� effector T lymphocytesthat are fully functional and able to generate SLEautoantigens. Our data disclose a new and pivotal roleof activated CD8� T lymphocytes in SLE pathogenesis.

Systemic lupus erythematosus (SLE) is a systemicautoimmune disease with multiorgan involvement char-acterized by an immune response against nuclear com-ponents (1). SLE patients experience a waxing andwaning disease course and a wide array of clinicalmanifestations reflecting the systemic nature of thedisease. The skin, kidneys, joints, and central nervoussystem may become the target of SLE-induced inflam-mation at its onset or during the course of the disease.Environmental triggers such as viruses (2) may act in thecontext of susceptibility genes, including genes involvedin antigen/immune complex clearance, lymphoid signal-ing, and apoptosis among several others (3), explainingwhy the pathogenesis of this disease remains largelyunknown.

The autoimmune response in SLE patients wasrecently found to be driven by unabated activation ofmyeloid dendritic cells (DCs) through interferon-�(IFN�) produced by another subset of DCs (i.e., plas-macytoid DCs) (4). The professional antigen-presentingcells capture, process, and present autoantigens to Tcells, thereby initiating the full autoimmune response. Inthis respect, much of the attention is now focused on

1Patrick Blanco, MD, Vincent Pitard, MSc, Jean-Luc Taupin,PhD, Jean-Francois Moreau, MD, PhD: CNRS–UMR5164 andIFR66, Universite de Bordeaux 2, Bordeaux, France; 2Jean-FrancoisViallard, MD, PhD, Jean-Luc Pellegrin, MD, PhD: Hopital du Haut-Leveque, CHU de Bordeaux, Bordeaux, France.

Address correspondence and reprint requests to PatrickBlanco, MD, CHU de Bordeaux, Place Amelie Raba Leon, Bordeaux33076, France. E-mail: patrick.blanco@u-bordeaux2.fr.

Submitted for publication June 15, 2004; accepted in revisedform September 21, 2004.

201

how these autoantigens that drive the autoimmuneresponse are produced and/or selected in order tounderstand the pathophysiology of SLE (5–8).

Apoptotic and necrotic cells are strong candi-dates as a source of such autoantigens (9,10). Anoverload on the capacity to clear dying cells may allowfor an unrestricted availability of autoantigens for DCs(11). In normal mice, the intravenous administration ofhigh numbers of apoptotic thymocytes was shown todetermine, by itself, the generation of antinuclear andantiphospholipid antibodies (12). Moreover, a recentreport from Casciola-Rosen et al proposed cytotoxic Tlymphocyte–mediated killing to be a preferential andselective source of autoantigens (13). That group re-ported the finding from in vitro studies that an exclusiveproperty of autoantigens is their ability to be cleaved bygranzyme B, a serine protease released by cytotoxic Tlymphocytes (13). However, the question remainswhether cleavage by granzyme B could be operating invivo in SLE, since no studies showing this have yet beenreported. The results obtained in the present studyallowed us to better delineate a still-unknown role forCD8� cytotoxic T lymphocytes in SLE pathogenesis andperpetuation.

PATIENTS AND METHODS

Patients and controls. Sixty-one consecutive SLE pa-tients were included in the present study between September2001 and June 2003. Patients met at least 4 of the AmericanCollege of Rheumatology 1982 revised criteria for SLE (14).All clinically and biologically relevant information concerningthe patients is shown in Table 1.

Clinical disease activity was scored using the SLEDisease Activity Index (SLEDAI) (15,16). Two groups ofpatients were defined. The active-disease group included 20patients with a flare of disease, defined either as a minimal3-point increase in the SLEDAI score compared with the scoreat the previous examination or as a score �6 for patients atdiagnosis. The quiescent-disease group included 41 patientswith a SLEDAI score �6 and with no variations throughoutthe entire followup period. A third group comprised 14 controlpatients with vasculitis.

For patients who presented with a disease flare (n �20), the concomitant or the closest biologic variables measuredwithin 2 weeks of diagnosis were considered for statisticalanalyses. For patients with quiescent disease throughout theentire followup period (n � 41), the last biologic variables wereused for statistical analyses. Healthy individuals from our staff(30 women and 6 men) were studied as a control group. Allblood samples were obtained after the patients and controlsubjects had given their informed consent.

Flow cytometric analysis. In all cytometric analyses, atotal of at least 5,000 lymphocytes from SLE patients or controlsubjects were analyzed using a 4-color flow cytometer(FACSCalibur; Becton Dickinson, Mountain View, CA). Spe-cific antibodies directed at surface markers included anti-CD3,anti-CD8, anti-CD4, anti-CD45, anti-CD45RA, anti–HLA–DR, and anti-CCR7 (all from Becton Dickinson). Antibodieswere incubated in whole blood before red blood cells werelysed using fluorescence-activated cell sorting lysing solution(Becton Dickinson). For intracellular staining, cells were firstlabeled with anti-CD3, anti-CD8, and anti–HLA–DR for 30minutes and were then resuspended in Permeafix (BectonDickinson) for 30 minutes at room temperature before label-ing with anti–granzyme B (Tebu, Paris, France), antiperforin(Becton Dickinson), or an isotype control antibody (BectonDickinson).

Purification of CD8� T lymphocytes. CD8� T lym-phocytes were purified from peripheral blood mononuclearcells (PBMCs; obtained after Ficoll-Hypaque gradient centrif-ugation) by using a magnetic cell separation method (MACS;Miltenyi Biotec, Bergisch Gladbach, Germany). In all experi-ments, purity was subsequently checked and always exceeded97%.

Anti-CD3–redirected killing assay and nucleosomequantification assay. The assay was carried out as previouslydescribed (17). Briefly, Fc receptor–bearing K562 cells labeledwith 51Cr were seeded in the presence of either anti-CD3(OrthoClone OKT3; Janssen-Cilag, Boulogne-Billancourt,France) or an IgG2a isotype-matched control antibody (anti-CD19; Becton Dickinson) at a final concentration of 10 �g/ml.In a second step, purified CD8� T lymphocytes from patientswith active or quiescent disease were added in each well atvarious target:effector ratios. For nucleosome release, K562cells were incubated with purified CD8� T lymphocytes at atarget:effector ratio of 1:20, and supernatants were harvestedat 6, 12, and 24 hours. Soluble nucleosome quantification wasdone using a commercially available enzyme-linked immu-nosorbent assay (ELISA) kit (Cell Death Detection ELISAPlus; Roche Diagnostics, Mannheim, Germany).

Monocyte-derived DC generation and mixed lympho-cyte reaction. Plastic-adherent monocytes were cultured in6-well plates (106/well) for 4 days in RPMI 1640 supplementedwith 10% fetal calf serum and with 100 ng/ml of recombinanthuman granulocyte–macrophage colony-stimulating factor(GM-CSF) (Leucomax; Novartis Pharma, Rueil-Malmaison,France) along with 20 ng/ml recombinant human interleukin-4(IL-4) or supplemented with 25% fresh lupus patient sera. Onday 5, monocyte-derived DCs were cultured at 104/well with105 allogeneic lymphocytes for 6 days. At specified time points,proliferation was checked by standard 3H-thymidine incorpo-ration, and T lymphocyte phenotypes were analyzed by flowcytometry.

Cleavage of endogenous autoantigens after in vitroincubation of Fas-negative K562 cells with purified CD8� Tlymphocytes. K562 cells were incubated for 4 hourswith purified CD8� T lymphocytes as previously described(13) at a target:effector cell ratio of 1:5 in the absence orpresence of a 50 �M final concentration of Z-IETD-fluoromethylketone (Z-IETD-FMK; Calbiochem, FontenaySous Bois, France). Cell extracts were separated by 10%

202 BLANCO ET AL

Table 1. Summary of the demographic, clinical, and biologic characteristics and treatments of SLE patients in the study*

Patient/age/sexLymphocytes/ml

(% CD8�,HLA–DR�)Disease flare

manifestations† SLEDAI score‡ Treatment‡

With quiescent disease1/30/F 1,154 (1.40) A, Sk 1 Pred., HCQ2/27/F 3,411 (10.62) Bl, A 2 Pred.3/23/M 2,004 (13.52) Sk, Bl, P, A 4 Pred., AZA4/34/F 1,154 (1.40) A, Bl 4 None5/21/F 2,423 (2.96) A, Sk 3 Pred., HCQ6/28/F 1,396 (2.74) Sk, A 3 Pred.7/22/F 1,759 (3.23) P, A, Sk 0 None8/45/F 1,950 (3.3) P, A, Bl 2 Pred.9/52/F 2,045 (10.33) A, Sk 2 None10/33/M 1,231 (4.72) A, NP 2 Pred.11/45/F 1,344 (5.72) Bl, A 2 Pred., HCQ12/37/F 1,528 (3.66) A, Sk 3 Pred., AZA13/40/F 1,682 (8.96) A, Sk, Bl 2 Pred., HCQ14/32/F 1,161 (9.95) Sk, A 5 Pred., HCQ15/25/F 3,530 (14.01) K, Sk, A 0 None16/50/F 1,640 (8.57) A, P 4 None17/40/F 1,027 (4.74) Bl, APS, Sk 3 Pred., AZA18/24/F 3,784 (0.96) A, P 4 Pred., HCQ19/20/F 1,428 (4.08) K, A 0 Pred.20/60/F 1,196 (1.88) K, A 2 None21/17/F 2,194 (3.58) A, Sk 2 None22/23/F 2,538 (3.97) A, K, P 5 None23/18/F 2,006 (2.64) A, Sk 2 None24/56/F 2,063 (5.55) P, A 2 None25/36/F 1,383 (3.50) Bl, A, NP 5 Pred.26/16/F 886 (7.94) Sk, A, APS 2 Pred.27/43/F 434 (3.1) A, Sk, Bl 6 Pred., MMF28/30/F 988 (5.27) P, Bl 2 HCQ29/25/F 1,700 (3.34) APS, A, P 3 Pred.30/25/F 1,976 (2.47) A, P, Bl 2 None31/43/F 4,226 (9.9) APS, NP 5 HCQ32/38/F 314 (4.64) Sk, A, Bl 2 Pred.33/64/F 1,332 (11.18) Sk, Bl, APS 5 Pred.34/70/F 2,823 (12.00) Sk, A, Bl 3 Pred.35/16/F 1,577 (7.63) SK, A, Bl 4 None36/17/F 1,031 (6.19) Sk, A 3 HCQ, Pred.37/41/F 1,072 (7.87) Bl, A, P 5 Pred.38/22/F 1,722 (8.22) A, Sk 3 None39/32/F 2,124 (10.2) A, P, Bl, Sk 5 HCQ, Pred.40/25/F 1,590 (1.82) A, Sk 2 None41/38/F 1,833 (5.6) Bl, NP, A 5 Pred., HCQ

With active disease101/28/F 1,144 (34.76) K, A, Sk 12 Pred., AZA102/28/F 668 (26.34) A, P, K 7 None103/32/F 946 (21.53) Sk, A, P 9 Pred., HCQ104/66/F 216 (19.27) Sk, A, P, Bl 8 Pred.105/40/F 3,073 (29.23) Sk, A, Bl, P 9 None106/33/F 2,053 (28.96) A, Sk 7 Pred., HCQ107/70/M 1,292 (26.33) A, K, Sk 15 None108/26/F 1,254 (28.29) A, K 8 Pred., MMF109/47/F 1,291 (16.70) P, A 8 Pred.110/67/F 1,005 (39.45) P, A, Sk 13 None111/28/F 1,507 (22.06) A, Sk, P, Bl 8 Pred.112/36/F 382 (26.09) A, Sk 12 Pred., AZA113/46/F 655 (26.84) A, Sk, K 12 None114/37/F 1,599 (18.31) Bl, NP, P 8 Pred.115/38/F 2,693 (12.81) P, A, K 12 None116/70/F 1,460 (22.30) A, Sk, NP 8 None117/13/F 511 (25.72) A, Sk, K 12 None118/28/F 932 (22.4) K, P, Bl, NP 10 None119/16/F 1,224 (25.2) Bl, A, NP 8 HCQ120/36/F 542 (19.00) A, Sk 7 None

* SLE � systemic lupus erythematosus; ACR � American College of Rheumatology; SLEDAI � SLE Disease Activity Index; A � musculoskeletalsystem; Sk � mucocutaneous lesions; Pred. � prednisone; HCQ � hydroxychloroquine; Bl � hematologic abnormality; P � pericarditis; AZA �azathioprine; NP � neuropsychiatric disorders; K � renal disease; APS � antiphospholipid syndrome; MMF � mycophenolate mofetil.† Organs or organ systems affected or syndromes occurring during a disease flare in patients who had previously been diagnosed as having SLEaccording to the ACR criteria.‡ At time of blood sampling.

ACTIVATED CD8� T LYMPHOCYTES IN SLE PATHOGENESIS 203

sodium dodecyl sulfate–polyacrylamide gel electrophoresisunder reducing conditions and then electroblotted ontonitrocellulose membranes. To ensure that equal amounts ofproteins were loaded per lane, protein concentrations foreach condition were determined using the bicinchoninicacid/copper sulfate method following the instructions of themanufacturer (Sigma, Pont-de-Claix, France). Incubationwith monospecific SLE patient anti–U1–70 kd serum at afinal dilution of 1:2,000 was used to identify autoantigenfragments. We observed that incubation of either K562 cellsor CD8� T lymphocytes alone did not affect the autoanti-gens analyzed (data not shown).

Statistical analysis. All T cell subpopulation percent-ages and absolute cell counts between groups were comparedusing the nonparametric Mann-Whitney U test, with a level ofsignificance of P � 0.05. We used the Spearman test todetermine the correlation between percentages of CD8�perforin- or granzyme B–positive T cells and the SLEDAIscore. The tests were carried out using Statistica statisticalsoftware (StatSoft, Tucson, AZ).

RESULTS

Activation of a high proportion of CD8� Tlymphocytes in SLE patients. Following our preliminaryobservations, we confirmed that activated T cells in SLEpatients are confined to the CD8� compartment andincrease during disease flare (18). The phenotypes of Tlymphocytes in 61 consecutive patients followed up fromDecember 2000 to June 2003 were evaluated. As men-tioned in Patients and Methods, we defined a group ofpatients with active disease (n � 20) and a group withquiescent disease (n � 41). In addition, those 2 groupswere compared with a group of age- and sex-matchedhealthy control volunteers (n � 36) who were free of anyautoimmune disease and/or infection, as well as with agroup of control patients with vasculitis (n � 14).

As we reported recently (18), and as shown inFigure 1A, patients with disease flares were character-ized by a statistically significant increase in the percent-age of CD8� T lymphocytes expressing HLA–DR (P �10�6) compared with the other groups, whereas thepercentages of total CD3�, CD3�,CD4�, orCD3�,CD8� T lymphocytes were not found to bestatistically different among all 4 groups. This activatedphenotype was virtually restricted to the sole CD3�,CD8�compartment and mildly affected CD3�,CD4� T cells(P � 0.01), which therefore barely accounted for theincrease in CD3�,HLA–DR� cells seen in the course ofdisease acceleration. It also suggested that CD8� cellsbearing an activated T cell phenotype might have ac-quired a cytotoxic phenotype. We therefore examinedtheir intracellular expression of perforin and gran-

zyme B by flow cytometry (Figure 1B). We noted an�3-fold increase in the median percentages of perforin-or granzyme B–positive CD8� T cells in the active-disease group compared with the quiescent-diseasegroup (P � 10�6 for both comparisons, by Mann-Whitney U test). The intracellular expression of perforinand granzyme B in the quiescent-disease group wassimilar to that in the healthy controls (P � 0.3 and P �0.06, respectively).

Figure 1C shows the T cell phenotype of 4representative individuals from 3 of the 4 groups. Not allHLA–DR� T cells expressed also perforin or granzymeB, and while a majority of cells displayed this phenotype,some CD8� cells were found to be positive for onlyHLA–DR or perforin or granzyme B, reflecting thedifferential kinetics of the expression of these markers atthe cell surface and/or of their recirculation capacity atthe periphery. However, among the 3 groups, the moststringent difference by far concerned the percentage ofdouble-positive CD8� T cells (HLA–DR and perforinor granzyme B positive), suggesting a true increase inrecently cytotoxic and activated effector T cells only inpatients with disease flares. Because it has recently beenfound that subpopulations of effector cells in the restingmemory T cell compartment can be distinguished basedon their expression of CCR7/CD45RA (19), we exam-ined the expression of these 2 cell markers at the surfaceof the blood CD8� T cells (Figure 1D). Comparedwith the normal control or quiescent-disease groups, amarked decrease in the percentages ofCCR7�,CD45RA�,CD8� (naive) T cells, associatedwith a reciprocal increase in CCR7�,CD45RA�,CD8�and CCR7�,CD45RA�, CD8� effector T cells, wasnoted in the SLE patients with disease flares.

Taken together, these data—the percentages aswell as the absolute cell counts (data not shown)—demonstrated that SLE patients with disease flares arespecifically characterized by an altered differentiation oftheir CD8� T cells toward a cytotoxic T lymphocytephenotype. These data disclose a hitherto ignored butpotentially important role for the cytotoxic CD8� Tcells in the pathogenesis of SLE.

Correlation between SLE activity and an in-creased proportion of perforin- or granzyme B–expressingCD8� T lymphocytes in the blood. In order to formallydemonstrate the relationship between the increasedproportion of cytotoxic CD8� T cells and the fluctua-tions of clinical manifestations associated with SLE,we plotted the percentages of peripheral blood

204 BLANCO ET AL

perforin-positive (Figure 2A) or granzyme B–positive (Fig-ure 2B) CD8� T cells against the SLEDAI scores for eachSLE patient (n � 61) at the time of flow cytometricevaluation of that patient’s cells. The percentages ofperforin- or granzyme B–positive CD8� T cells correlatedstrongly with the SLEDAI scores. The Spearman statistical

test yielded high values for the correlation coefficients,similar in both instances (R � 0.731 and R � 0.733), witha very high degree of significance (P � 10�6). Thisdemonstrated that the increase in circulating perforin- orgranzyme B–positive CD8� T cells thoroughly reflectedthe activity of the disease.

Figure 1. Skewing of CD8� T cells toward a cytotoxic effector phenotype in systemic lupus erythematosus (SLE) patients with disease flares. A,Comparisons of percentages of CD3�,HLA–DR�, CD3�,CD4�,HLA–DR�, and CD3�,CD8�,HLA–DR� lymphocytes among total lympho-cytes from healthy controls, SLE patients with active disease, SLE patients with quiescent disease, and control patients with vasculitis. Squares,triangles, and diamonds inside boxes represent the median value for each group. Boxes include the 25th–75th percentiles; bars outside boxesrepresent the 10th and 90th percentiles; open circles represent values beyond the 10th and 90th percentiles. P values were determined byMann-Whitney U test. B, Comparisons of intracellular expression of perforin and granzyme B in peripheral blood CD8� T cells from healthycontrols, SLE patients with active disease, and SLE patients with quiescent disease. Squares and triangles inside boxes represent the median valuefor each group. Percentiles and outlying values are as described in A. P values were determined by Mann-Whitney U test. C, Flow cytometric analysisshowing the T cell phenotype of 4 representative individuals from 3 of the 4 groups. Active 1 and active 2 represent 2 patients from the group withactive disease. Percentages on the y-axis of positive cells among total peripheral blood lymphocytes, as defined by size, cell content, and CD45expression, are indicated in the double-positive dot-plot quadrant. D, Flow cytometric analysis of the expression of CCR7 and CD45RA on thesurfaces of peripheral blood CD8� T cells from healthy controls, SLE patients with active disease, and SLE patients with quiescent disease.Percentages on the y-axis are expressed among CD8� T cells. Squares, triangles, diamonds, and circles inside boxes represent the median value foreach group. Percentiles and outlying values are as described in A. P values were determined by Mann-Whitney U test. ISO � isotype controlantibody.

ACTIVATED CD8� T LYMPHOCYTES IN SLE PATHOGENESIS 205

Peripheral blood CD8� T cells from SLE pa-tients acting as functional cytotoxic effectors in vitrogenerating soluble nucleosomes. In the absence of nom-inal antigens, we carried out a nonspecific, anti-CD3–dependent, redirected killing to test whether circulatingCD8� T lymphocytes from SLE patients were func-tional cytotoxic effectors. This was accomplished byusing Fas-negative K562 erythroleukemic human cells,which abundantly express Fc� receptors at their cellsurfaces, as target cells. In this 4-hour chromium releaseassay, target cells bound anti-CD3 through its Fc do-main, leaving the F(ab�)2 domain free to activate theeffector T cells nonspecifically.

As shown in Figure 3A, the fresh and enriched(�97% pure) CD8� T lymphocyte fraction (sorted

following positive anti-CD8–coated bead selection)from the blood of patients with active disease (n � 3)displayed 3-fold higher cytotoxic activity than the puri-fied CD8� T lymphocyte fraction from the blood ofpatients with quiescent disease (n � 3). All 3 patientswith disease flares had percentages of CD3�,CD8�,HLA–DR�,granzyme B–positive cells ranging from

Figure 2. Correlation between percentages of cytotoxic effector Tcells circulating in the blood and Systemic Lupus ErythematosusDisease Activity Index (SLEDAI) scores in SLE patients. A, Percent-ages of CD8�,perforin-positive T lymphocytes among totalCD3�,CD8� lymphocytes plotted against SLEDAI scores. B, Per-centages of CD8�,granzyme B–positive T lymphocytes among totalCD3�,CD8� lymphocytes plotted against SLEDAI scores. In bothinstances, correlation coefficients and P values were obtained using theSpearman test (see Patients and Methods).

Figure 3. Cytotoxicity of, and induction of high levels of solublenucleosomes by, CD8� T cells from systemic lupus erythematosus(SLE) patients with disease flares. A, Redirected killing of Fas-negative K562 cells by freshly purified CD8� T lymphocytes fromSLE patients with quiescent (triangles) or active (squares) disease inthe absence (open triangles or squares) or presence (solid triangles orsquares) of 10 �g/ml anti-CD3. Values are the mean and SD specific51Cr release (target:effector ratios of 1:1, 1:5, 1:10, and 1:20) obtainedfrom 3 separate 51Cr release assays performed on cells from 3 SLEpatients with quiescent disease and 3 SLE patients with active disease.B, Soluble nucleosome generation following anti-CD3–redirected kill-ing of K562 cells by CD8� T lymphocytes (target:effector ratio 1:20)from SLE patients with quiescent (squares) or active (diamonds)disease in the absence (open squares or diamonds) or presence (solidsquares or diamonds) of 10 �g/ml anti-CD3. Values are the meanoptical densities (ODs) obtained at 6-hour, 12-hour, and 24-hourcoincubation times from 1 experiment representative of 3.

206 BLANCO ET AL

42% to 60% of total lymphocytes, whereas the corre-sponding percentages among the 3 patients with quies-cent disease were �15%. Because CD8� sorted lym-phocytes may contain natural killer (NK) cells, andbecause K562 cells are exquisitely sensitive to killing byNK cells, control experiments were carried out in theabsence of anti-CD3 antibody. Under these conditions,the killing remained at the baseline level, indicating thatall the cytotoxic effects on K562 cells revealed in thepresence of anti-CD3 were attributable to CD3�,CD8�T lymphocytes, the percentage of which increased dra-matically in patients with disease flares.

Soluble nucleosomes are found in high amountsin sera from SLE patients (20,21) and are thought toplay a key role in the autoimmune reaction againstnuclear components. We therefore asked whetherCD8� T lymphocytes from patients had the capacity togenerate nucleosomes when mixed with target cells invitro. Again, freshly purified CD8� cells from SLEpatients were added to K562 cells in the presence orabsence of anti-CD3, and a fraction of the cell superna-tant was harvested at 6, 12, and 24 hours of coincubationtime. The concentrations of soluble nucleosomespresent in the supernatants were then determined usinga commercially available semiquantitative ELISA. Datafrom a representative experiment are shown in Figure3B and demonstrate that supernatants from culturescontaining CD8� T lymphocytes from patients withactive disease generated soluble nucleosomes earlierand in higher amounts than CD8� T lymphocytes frompatients with quiescent disease. Together, these dataconfirmed the presence of a greater number of func-tional cytotoxic effector cells among CD8� T cells fromSLE patients with disease flares.

Cleavage by cytotoxic effector T cells of autoan-tigens recognized by autoantibodies found in SLE pa-tients. Casciola-Rosen et al recently demonstrated thatgranzyme B cleavage in target cells during cytotoxiclymphocyte granule–induced cell death led to the pro-duction of unique peptide fragments exclusively fromautoantigens in patients with systemic autoimmune dis-eases including SLE. To ascertain whether similar auto-antigen fragments could be preferentially generated byCD8� T lymphocytes from SLE patients, we coculturedK562 cells with purified CD8� T lymphocytes from SLEpatients with active or quiescent disease and analyzedthe cleavage of autoantigens through immunoblottingusing patient sera. In order to avoid confusing theinterpretation of the data, we used a monospecific SLE

patient serum directed against U1–70-kd autoantigen todetect its fragments potentially produced by cytotoxic Tcells.

As shown in Figure 4, purified CD8� T lympho-cytes were able to spontaneously generate granzymeB–specific fragments of U1–70-kd autoantigen (smallsolid arrow). In contrast, the production of the uniquegranzyme B fragment was abolished by the addition of50 �M Z-IETD-FMK, a known inhibitor of granzyme B,to the reaction wells (lane 3), and the granzyme Bfragment was undetectable using CD8� T cells purifiedfrom the blood of patients with quiescent disease (lane4). K562 cells alone (lane 1) or purified CD8� T cellsalone (lanes 5 and 6), both in the presence of anti-CD3,did not lead to any granzyme B–specific U1–70-kd band.Taken together, these data demonstrated that purifiedCD8� T lymphocytes from SLE patients with activedisease have the capacity to generate unique granzymeB fragments that may subsequently be the target of theautoimmune response.

Figure 4. Granzyme B–specific autoantigens generated by freshlypurified CD8� T lymphocytes from systemic lupus erythematosus(SLE) patients with active disease. K562 cells were coincubated (lanes2–4) or not (lane 1) with freshly purified CD8� T lymphocytes(target:effector ratio 1:5) from SLE patients with quiescent disease(lane 4) or from SLE patients with active disease in the presence(lane 3) or absence (lane 2) of granzyme B inhibitor (Z-IETD-fluoromethylketone). Effectors were also incubated in the absence ofK562 cells (lanes 5 and 6) as controls. After a 4-hour incubation, cellswere lysed in loading sodium dodecyl sulfate–polyacrylamide gelelectrophoresis buffer before being electrophoresed. U1–70-kd au-toantigen was detected by Western blotting using a monospecific SLEpatient serum. Large solid arrow indicates intact antigen; small solidand open arrows indicate granzyme B–specific and caspase-specificautoantigen fragments, respectively.

ACTIVATED CD8� T LYMPHOCYTES IN SLE PATHOGENESIS 207

Unique ability of DCs generated in the presenceof SLE serum to induce the expression of granzyme B inCD8� T lymphocytes. In order to demonstrate whichfactors could account for such an activation of theCD8� T cell compartment, we first incubated allogeneicPBMCs (obtained by Ficoll-Hypaque gradient centrifu-gation) with SLE serum from patients with active andquiescent disease. We did not observe any difference interms of intracellular perforin and granzyme B expres-sion after 10 days of culture under these conditions, astested by flow cytometry (data not shown), suggestingthat soluble factors present in SLE patient sera (thera-peutic agents, cytokines, etc.) could not be the directcausative agent for the altered differentiation of CD8�T cells.

Since investigators in our group and others haverecently demonstrated that unabated induction of DCsby IFN� may drive the autoimmune response in SLE(4,8,22), we addressed the question of whether DCsgenerated in the presence of sera from SLE patientswith active or quiescent disease (hereafter referred to asactive SLE DCs or quiescent SLE DCs, respectively) hadthe capacity to specifically induce the expression ofintracellular perforin and granzyme B in allogeneicnaive CD8� T cells. To this end, we cultured normalmonocytes as described elsewhere (4) with SLE serafrom patients with active or quiescent disease for 4 days.Alternatively, DCs differentiated in the presence ofGM-CSF and IL-4 were used as a control. In this firststep, we confirmed that monocytes cultured in thepresence of sera from SLE patients with active diseasedifferentiated into DCs, whereas sera from patients withquiescent disease or from normal individuals were inef-fective. In the second step, allogeneic peripheral bloodlymphocytes were added to the original washed culturesand monitored for their proliferation and cell markerexpression over time. As expected, DCs differentiated inthe presence of GM-CSF and IL-4 or active SLE DCsinduced a robust proliferation of lymphocytes, whereasquiescent SLE DCs were unable to do so (ref. 4 and datanot shown).

The intracellular staining for granzyme B ofCD8� T cells from cocultures was analyzed on days 0, 3,and 6 (Figure 5A). Clearly, active SLE DCs had theunique ability to induce the intracellular expression ofgranzyme B in CD8� T cells. Indeed, after 6 days ofcoculture with active SLE DCs, �40% of allogeneicCD8� T lymphocytes expressed intracellular granzymeB, as compared with 20% with quiescent GM�IL-4 DCs(Figure 5A). Moreover, this effect could be inhibited

by adding 10 �g/ml of blocking anti-IFN� mono-clonal antibody during the first step of active SLE DC

Figure 5. Unique ability of dendritic cells (DCs) generated in thepresence of serum from patients with systemic lupus erythematosus(SLE) to drive the differentiation of allogeneic T lymphocytes towardfunctional cytotoxic effector cells. A, Induction of intracellular gran-zyme B expression in CD8� allogeneic T lymphocytes. Monocyteswere either cultured in the presence of recombinant humangranulocyte–macrophage colony-stimulating factor (GM-CSF) andinterleukin-4 (IL-4) or were cultured with sera from SLE patients withactive disease in the absence or presence of blocking anti–interferon-�(anti-IFN�) monoclonal antibody. On day 4, differentiated DCs wereharvested and cocultured with allogeneic lymphocytes from normalindividuals. On days 0 (open bars), 2 (striped bars), and 6 (solid bars),CD8� T lymphocytes were labeled for intracellular expression ofgranzyme B. Shown are mean percentages of granzyme B–positivecells among CD8� T lymphocytes from 1 experiment representative of3. B, Freshly isolated CD8� T lymphocytes from SLE patients withactive disease act in a manner similar to that of allogeneic CD8� Tlymphocytes primed in vitro by SLE DCs in a cytotoxic functionalassay. Shown is redirected killing of Fas-negative K562 cells by freshlypurified CD8� T lymphocytes from SLE patients with active disease(squares) or by in vitro SLE DC–primed allogeneic CD8� T lympho-cytes (triangles) in the absence (open squares or triangles) or presence(solid squares or triangles) of 10 �g/ml anti-CD3. Values are the meanand SD specific 51Cr release (target:effector ratios of 1:1, 1:5, and 1:10)obtained from triplicates from 1 representative experiment of 3.

208 BLANCO ET AL

generation. Intracellular perforin staining gave similarresults (data not shown). As a control, normal lympho-cytes cultured with quiescent SLE DCs expressed nei-ther intracellular perforin nor granzyme B (data notshown).

In addition, we compared the cytotoxic potentialof these in vitro–primed CD8� T lymphocytes with thatof their counterparts found in the blood of SLE patientswith active disease (Figure 5B). The CD3-dependentredirected killing against K562 cells did not disclose anydifference between the 2 types of T lymphocytes, con-firming that CD8� T lymphocytes not only acquiredcytotoxic phenotypes when cultured in the presence ofSLE DCs, but also became fully functional and indistin-guishable from freshly purified blood CD8� T lympho-cytes from SLE patients with active disease. These datademonstrate that activation of DCs by IFN� in SLEpatients with active disease is sufficient to drive thedifferentiation of CD8� T cells toward fully activecytotoxic effector T lymphocytes.

DISCUSSION

Recent studies have focused on the role of apo-ptosis and the antigen-presenting function of DCs in thepathophysiology of human SLE. The role of cytotoxicCD8� T lymphocytes in the defense against viral agentsor organ-specific autoimmune diseases is well docu-mented, but it is still unexplored in SLE. The presentstudy reveals a quantitative and functional increase inCD8� cytotoxic T lymphocytes that is highly correlatedwith SLE disease activity and that may be responsible forthe increased production of autoantigens. In addition,following stimulation with IFN� derived from sera fromSLE patients with active disease, monocyte-derived DCsacquired the unique ability to induce the differentiationof naive CD8� T lymphocytes toward a functionalcytotoxic phenotype identical to that observed in vivo.Thus, our data imply a previously ignored role of CD8�T lymphocytes in the generation of high amounts ofnuclear autoantigens which, as a consequence, mayoverwhelm the physiologic clearance pathway.

Owing to their central role in the humoral re-sponse against autoantigens, CD4� cells are thought tobe the primary T lymphocyte subpopulation involved inlupus autoimmune response (23). Although some re-ports of studies in humans suggested that cell cytotoxi-city is impaired in SLE (24,25), studies in several rodentmodels indicate that CD8� T lymphocytes may also

contribute to this response, either directly, as a noxiouselement of the cellular response, or indirectly, by pro-viding supplies for overcoming mechanisms of toleranceto autoantigens. It is significant that in a rat model ofautoimmune glomerulonephritis (i.e., Goodpasture’ssyndrome), in which anti–glomerular basement mem-brane antibodies are pathogenic, CD8� cell depletioncan prevent or treat the renal disease without affectingserum levels of autoantibodies (26). Mice deprived ofCD8� T cells following deficiency in major histocom-patibility complex class I antigen expression are resistantto experimental SLE (27). NZB mice deficient in �2-microglobulin had a lower incidence and a delayed onsetof antierythrocyte autoantibody production comparedwith that seen in normal NZB mice (28). More recently,NZB mice deficient in type I IFN receptor were shownto have a significant decrease in splenic CD8� cells anda reduced lupus-like disease (29).

In vitro studies reported by Casciola-Rosen et alindicate that apoptosis involving granzyme B may beimportant for autoantigen generation (13), but no directlinks between cytotoxic T lymphocytes and autoantigengeneration have ever been demonstrated in SLE pa-tients. In those studies, the effective generation ofgranzyme B autoantigen fragments was dependent onthe relative exogenous inhibition of the caspase pathway.That is not the case in our study, since CD8� Tlymphocytes freshly isolated from the peripheral bloodof SLE patients with active disease had the intrinsiccapacity to generate nontolerized granzyme B autoanti-gen fragments without any inhibition of the caspasepathway (U1–70 kd and topoisomerase I), suggesting apeculiar status for these cells in patients with active SLE(Figure 4 and data not shown). Among several possibil-ities, this discrepancy may rely on the fact that we aredealing with in vivo–activated T lymphocytes, whereasAndrade et al dealt with lymphokine-activated killercells cultured for 4 days (30).

In this view, the circumstances of T lymphocyteactivation by DCs in vivo may be of paramount impor-tance but difficult to investigate in humans. However,our in vitro experiments suggest that normal T lympho-cytes are converted to an activated and functional pheno-type only when cocultured with monocyte-derived DCsin the presence of sera from patients with active SLE.Inhibition of IFN� in SLE sera by blocking antibody ledto the abrogation of this process, a finding that empha-sizes and allows us to better understand the involvementof IFN� in the pathogenesis of SLE.

ACTIVATED CD8� T LYMPHOCYTES IN SLE PATHOGENESIS 209

The mechanism(s) leading to the wide activationof CD8� T lymphocytes is still elusive, but we suggestthat it is a consequence of the DC system activationfound in SLE (4). In this regard, the increased level ofimmune complexes leading preferentially to cross-presentation (31) and/or a direct presentation of viralantigens (2) may considerably increase the proportion ofactivated T cells.

In summary, the present data support the hypothe-sis of an existing vicious circle initiating and perpetuat-ing SLE disease, in which IFN�-activated DCs stronglyalter the differentiation of CD8� T lymphocytes, gen-eration of nontolerized autoantigens, and efficient “an-tigenic feeding” of DCs. IFN� is indispensable in thisprocess by acting on DCs, and it represents a key targetfor SLE therapy.

ACKNOWLEDGMENTS

We are indebted to N. Berrie, J. C. Carron, M. Garcie,and F. Saussais, and to all members of the Laboratory ofClinical Immunology at the Centre Hospitalier Regional deBordeaux who skillfully contributed to this study. We thankDrs. F. Halary and J. Dechanet-Merville for critically readingthe manuscript and for their helpful suggestions.

REFERENCES

1. Ruiz-Irastorza G, Khamashta MA, Castellino G, Hughes GR.Systemic lupus erythematosus. Lancet 2001;357:1027–32.

2. James JA, Harley JB, Scofield RH. Role of viruses in systemiclupus erythematosus and Sjogren syndrome. Curr Opin Rheuma-tol 2001;13:370–6.

3. Wakeland EK, Liu K, Graham RR, Behrens TW. Delineating thegenetic basis of systemic lupus erythematosus. Immunity 2001;15:397–408.

4. Blanco P, Palucka AK, Gill M, Pascual V, Banchereau J. Inductionof dendritic cell differentiation by IFN-� in systemic lupus ery-thematosus. Science 2001;294:1540–3.

5. Casciola-Rosen LA, Anhalt G, Rosen A. Autoantigens targeted insystemic lupus erythematosus are clustered in two populations ofsurface structures on apoptotic keratinocytes. J Exp Med 1994;179:1317–30.

6. Bijl M, Limburg PC, Kallenberg CG. New insights into thepathogenesis of systemic lupus erythematosus (SLE): the role ofapoptosis. Neth J Med 2001;59:66–75.

7. Pollard KM. Cell death, autoantigen cleavage, and autoimmunity.Arthritis Rheum 2002;46:1699–702.

8. Bennett L, Palucka AK, Arce E, Cantrell V, Borvak J, BanchereauJ, et al. Interferon and granulopoiesis signatures in systemic lupuserythematosus blood. J Exp Med 2003;197:711–23.

9. Utz PJ, Anderson P. Posttranslational protein modifications,apoptosis, and the bypass of tolerance to autoantigens. ArthritisRheum 1998;41:1152–60.

10. Rosen A, Casciola-Rosen L. Autoantigens as substrates for apo-ptotic proteases: implications for the pathogenesis of systemicautoimmune disease. Cell Death Differ 1999;6:6–12.

11. Rosen A, Casciola-Rosen L. Clearing the way to mechanisms ofautoimmunity. Nat Med 2001;7:664–5.

12. Mevorach D, Zhou JL, Song X, Elkon KB. Systemic exposure toirradiated apoptotic cells induces autoantibody production. J ExpMed 1998;188:387–92.

13. Casciola-Rosen L, Andrade F, Ulanet D, Wong WB, Rosen A.Cleavage by granzyme B is strongly predictive of autoantigenstatus: implications for initiation of autoimmunity. J Exp Med1999;190:815–26.

14. Tan EM, Cohen AS, Fries JF, Masi AT, McShane DJ, RothfieldNF, et al. The 1982 revised criteria for the classification of systemiclupus erythematosus. Arthritis Rheum 1982;25:1271–7.

15. Bombardier C, Gladman DD, Urowitz MB, Caron D, Chang CH,and The Committee on Prognosis Studies in SLE. Derivation ofthe SLEDAI: a disease activity index for lupus patients. ArthritisRheum 1992;35:630–40.

16. Gladman DD, Goldsmith CH, Urowitz MB, Bacon P, BombardierC, Isenberg D, et al. Crosscultural validation and reliability of 3disease activity indices in systemic lupus erythematosus. J Rheu-matol 1992;19:608–11.

17. Maziarz RT, Groh V, Prendergast M, Fabbi M, Strominger JL,Burakoff SJ. Non-MHC-restricted target-cell lysis by a CD4-CD8-TCR � � T-cell line, as well as by TCR � � T-cell lines, results fromlymphokine-activated killing. Int J Cancer 1991;48:142–7.

18. Viallard JF, Bloch-Michel C, Neau-Cransac M, Taupin JL, Garri-gue S, Miossec V, et al. HLA-DR expression on lymphocytesubsets as a marker of disease activity in patients with systemiclupus erythematosus. Clin Exp Immunol 2001;125:485–91.

19. Sallusto F, Lenig D, Forster R, Lipp M, Lanzavecchia A. Twosubsets of memory T lymphocytes with distinct homing potentialsand effector functions. Nature 1999;401:708–12.

20. Amoura Z, Piette JC, Chabre H, Cacoub P, Papo T, Wechsler B,et al. Circulating plasma levels of nucleosomes in patients withsystemic lupus erythematosus: correlation with serum antinucleo-some antibody titers and absence of clear association with diseaseactivity. Arthritis Rheum 1997;40:2217–25.

21. Amoura Z, Piette JC, Bach JF, Koutouzov S. The key role ofnucleosomes in lupus. Arthritis Rheum 1999;42:833–43.

22. Baechler EC, Batliwalla FM, Karypis G, Gaffney PM, OrtmannWA, Espe KJ, et al. Interferon-inducible gene expression signa-ture in peripheral blood cells of patients with severe lupus. ProcNatl Acad Sci U S A 2003;100:2610–5.

23. Lu L, Kaliyaperumal A, Boumpas DT, Datta SK. Major peptideautoepitopes for nucleosome-specific T cells of human lupus.J Clin Invest 1999;104:345–55.

24. Stohl W, Elliott JE, Li L, Podack ER, Lynch DH, Jacob CO.Impaired nonrestricted cytolytic activity in systemic lupus ery-thematosus: involvement of a pathway independent of Fas, tumornecrosis factor, and extracellular ATP that is associated with littledetectable perforin. Arthritis Rheum 1997;40:1130–7.

25. Kaplan M, Lu Q, Wu A, Attwood J, Richardson B. Demethyl-ation of promoter regulatory elements contributes to perforinoverexpression in CD4� lupus T cells. J Immunol 2004;172:3652–61.

26. Reynolds J, Norgan VA, Bhambra U, Smith J, Cook HT, PuseyCD. Anti-CD8 monoclonal antibody therapy is effective in thep-revention and treatment of experimental autoimmune glomerulo-nephritis. J Am Soc Nephrol 2002;13:359–69.

27. Mozes E, Kohn LD, Hakim F, Singer DS. Resistance of MHCclass I-deficient mice to experimental systemic lupus erythemato-sus. Science 1993;261:91–3.

28. Chen SY, Takeoka Y, Pike-Nobile L, Ansari AA, Boyd R,Gershwin ME. Autoantibody production and cytokine profilesof MHC class I (�2-microglobulin) gene deleted New Zealandblack (NZB) mice. Clin Immunol Immunopathol 1997;84:318–27.

210 BLANCO ET AL

29. Santiago-Raber ML, Baccala R, Haraldsson KM, Choubey D,Stewart TA, Kono DH, et al. Type-I interferon receptor deficiencyreduces lupus-like disease in NZB mice. J Exp Med 2003;197:777–88.

30. Andrade F, Roy S, Nicholson D, Thornberry N, Rosen A, Cas-ciola-Rosen L. Granzyme B directly and efficiently cleaves several

downstream caspase substrates: implications for CTL-inducedapoptosis. Immunity 1998;8:451–60.

31. Dhodapkar KM, Krasovsky J, Williamson B, Dhodapkar MV.Antitumor monoclonal antibodies enhance cross-presentation ofcellular antigens and the generation of myeloma-specific killer Tcells by dendritic cells. J Exp Med 2002;195:125–33.

ACTIVATED CD8� T LYMPHOCYTES IN SLE PATHOGENESIS 211