ARTHRITIS & RHEUMATISMVol. 56, No. 12, December 2007, pp 4132–4141DOI 10.1002/art.23054© 2007, American College of Rheumatology

Hyperexpression of Cyclooxygenase 2 in the Lupus ImmuneSystem and Effect of Cyclooxygenase 2 Inhibitor Diet Therapy

in a Murine Model of Systemic Lupus Erythematosus

Li Zhang, Anne M. Bertucci, Kimberly A. Smith, Luting Xu, and Syamal K. Datta

Objective. To investigate the role of cyclooxygen-ase 2 (COX-2) in the functioning of different cell typesinvolved in the lupus autoimmune response, and toexamine the therapeutic effect of COX-2 inhibitors inmice prone to spontaneously develop systemic lupuserythematosus (SLE).

Methods. Lupus-prone (SWR � NZB)F1 micewere fed with a diet containing different doses of theCOX-2–specific inhibitor celecoxib or the nonspecificinhibitor aspirin, or a combination of both, and theeffects of the therapy on autoantibody production, de-velopment of lupus nephritis, and mortality were deter-mined. Expression of COX-2 by different cells of thelupus immune system and the effect of COX-2 inhibitorson the function of these cells in vitro and in vivo wereassessed.

Results. The immune cells of mice with SLEspontaneously hyperexpressed COX-2, and COX-2 in-hibitors could cause cell apoptosis. Treatment withCOX-2 inhibitors resulted in decreased autoantibodyproduction and inhibition of the T cell response to themajor lupus autoantigen, nucleosome, and its presenta-tion by antigen-presenting cells. Surprisingly, a signifi-cant increase in survival occurred only in mice receivingintermittent therapy with the lowest dose of celecoxib(500 parts per million), approximating <100 mg ofcelecoxib/day in humans. A continuous diet, but notintermittent feeding, with the combination of celecoxib

and aspirin delayed development of nephritis tempo-rarily, but failed to prolong survival. Indeed, treatmentwith aspirin alone increased mortality.

Conclusion. The contributions of the major play-ers in the pathogenic autoimmune response, namely, Tcells, B cells, dendritic cells, and macrophages that areabnormally hyperactive in lupus, depend on the in-creased expression and activity of COX-2, similar toinflammatory cells in target organs. Intermittent pulsetherapy with low doses of select COX-2 inhibitors wouldbe of value in the treatment of lupus.

In systemic lupus erythematosus (SLE), hyperac-tivity of the immune system leads to activation of certainautoimmune T helper cells, which drive autoimmune Bcells to produce somatically mutated IgG autoantibodiesagainst apoptotic nuclear antigens (1–6). IgG immunecomplexes containing autoantigenic DNA and RNAbind Fc� receptors and Toll-like receptors (TLRs),which results in stimulation of type I interferon (IFN)production by hyperactive dendritic cells (DCs); more-over, autoimmune B cells bearing the T helper cell–driven, high-affinity, somatically mutated B cell recep-tors are dually stimulated by TLRs (7–9). These eventsfurther amplify the ability of antigen-presenting cells(APCs) to present autoantigens to pathogenic T helpercells, which is an essential element of disease develop-ment (10–13).

Normally, autoreactive T and B cells are elimi-nated by functional inactivation (anergy) and byactivation-induced cell death (AICD; apoptosis), via Fassignaling (14). However, autoimmune T helper cells inhuman lupus resist AICD by up-regulating the expres-sion of cyclooxygenase 2 (COX-2) and the antiapoptoticmolecule cFLIP (15,16). Even after activation with fullcostimulation using anti-CD28 antibodies, anti-CD3 an-tibodies, and interleukin-2, normal T cells do not up-regulate COX-2 to the same extent as that by lupus T

Supported by the NIH (grants R37-AR-39157 and R01-AI-41985) and the Solovy Arthritis Research Society.

Li Zhang, PhD, Anne M. Bertucci, BS, Kimberly A. Smith,MS, Luting Xu, PhD, Syamal K. Datta, MD: Northwestern UniversityFeinberg School of Medicine, Chicago, Illinois.

Address correspondence and reprint requests to Syamal K.Datta, MD, Division of Rheumatology, Northwestern University Fein-berg School of Medicine, 240 East Huron Street, McGaw M300,Chicago, IL 60611. E-mail: skd257@northwestern.edu.

Submitted for publication March 7, 2007; accepted in revisedform August 20, 2007.

4132

cells stimulated with plastic-bound anti-CD3 alone(which normally induces anergy) (16). Surprisingly, onlycertain COX-2 inhibitors (celecoxib [Celebrex] beingone of them) could cause apoptosis of lupus T cells.Moreover, in vitro in lupus-prone (SWR � NZB)F1

(SNF1) mice, celecoxib could also markedly block theresponse of T cells to nucleosomes, the major lupusautoantigen, as well as block pathogenic autoantibodyproduction induced by the nuclear autoantigen–specificT helper cells (16).

To determine the mechanism of COX-2 inhibi-tion in vivo, we investigated the expression of COX-2 inother cells of the lupus immune system in an SNF1

mouse model of lupus. In addition, we examined theeffects of treatment with a COX-2 inhibitor diet onspontaneously developing SLE in SNF1 mice (17,18).

MATERIALS AND METHODS

Mice. NZB, SWR, and BALB/c mice were purchasedfrom Jackson Laboratory (Bar Harbor, ME). SNF1 hybridswere bred. Female mice were used in the present study, withthe approval of the Animal Care and Use Committee ofNorthwestern University.

Administration of celecoxib and aspirin. Celecoxib wasprovided by Pfizer (Groton, CT). Aspirin was purchased fromSigma (St. Louis, MO). Groups of 12-week-old SNF1 micewere given a regular diet of rodent chow (Teklad no. 7912)either by itself or supplemented (19) with 500, 1,000, or 1,500parts per million (ppm) (or mg per kg of diet) of celecoxib(Research Diets, New Brunswick, NJ), which was administeredintermittently in a cycle of 3 weeks of celecoxib diet followedby 4 weeks of drug-free diet. The regimen was continuedthrough the remaining lifespan of the animals. Another 4groups of mice were administered continuously a diet of either500 ppm of celecoxib, 400 ppm of aspirin, or 500 ppm ofcelecoxib plus 200 ppm aspirin (continuous combination treat-ment), or received the regular (drug-free) diet as control.Finally, additional groups of SNF1 mice received a combina-tion diet of 500 ppm celecoxib plus 200 ppm aspirin adminis-tered in an intermittent regimen of 3 weeks on the drugs and4 weeks off (intermittent combination treatment), or receivedthe regular (drug-free) diet as control.

All mice were monitored weekly for the developmentof proteinuria, by testing with Albustix (VWR Scientific,Chicago, IL), determining diet intake (weight of chow left oncage top), and tracking body weight. Treatments lasted untilthe mice were moribund.

To study early immunologic changes after treatmentwith celecoxib or aspirin, an additional batch of 12-week-oldSNF1 mice (5 mice/group) was treated with the same regimensas described above, and killed after 6 weeks. Celecoxib levels inthe plasma of mice receiving different doses of the drug weremeasured by high-performance liquid chromatography(HPLC) (19) conducted in the Clinical Pharmacology CoreFacility at Northwestern University.

Quantitation of autoantibodies. IgG-class autoanti-bodies to double-stranded DNA (dsDNA), single-strandedDNA (ssDNA), nucleosomes, and histones in the sera andculture supernatants were estimated by enzyme-linked immu-nosorbent assay (ELISA) (13,20).

Measurement of intracellular COX-2 and analysis ofsurface marker staining by flow cytometry. Total spleen cellsfrom 6-week-old unmanipulated SNF1, SWR, and BALB/cmice were stained with rat or hamster monoclonal antibodies(mAb) to mouse CD4 (for T cells), mouse CD19 and CD86(for activated B cells), mouse CD11c (for DCs), and mouseCD11b (for macrophages), conjugated with peridinin chloro-phyll A protein (PerCP) or phycoerythrin (BD PharMingen,San Jose, CA or eBioscience, San Diego, CA, respectively) at4°C for 30 minutes. After washing and fixation, cells werepermeabilized and stained with mouse anti–COX-2 mAb or itsisotype control (BD PharMingen) at room temperature (RT)for 30 minutes. After extensive washing, cells were incubatedwith second-step goat anti-mouse fluorescein isothiocyanate(FITC)–conjugated IgG (Molecular Probes, Carlsbad, CA) atRT in the dark for 30 minutes. Usually, 200,000 cell eventswere collected after live-cell gating with FACSCalibur, withthe results analyzed by CellQuest (BD PharMingen) or FlowJo(TreeStar; FlowJo, Ashland, OR) software. Isotype-matchedcontrols were used for marking positive and negative cellpopulations.

Western blot for COX-2 expression. Whole cell lysateswere prepared from fractionated spleen cell subsets or wholesplenocytes. CD4 T cells (macrophages plus DCs) were puri-fied from splenocytes using a technique as previously described(21,22). Western blotting was then carried out (15,16), andblots were probed with primary goat anti–COX-2 antibody(sc-1745; Santa Cruz Biotechnology, Santa Cruz, CA) over-night at 4°C in 5% nonfat dry milk in TBST (60 mmoles/literTris base, 120 mmoles/liter NaCl, 0.5% Tween 20), and thenincubated with a horseradish peroxidase–conjugated second-ary antibody (sc-2020; Santa Cruz Biotechnology) in 5% drymilk in TBST for 1 hour at RT. Bound antibodies weredetected with enhanced chemiluminescence detection reagents(Amersham Pharmacia Biotech, Buckinghamshire, UK).

Induction of apoptosis. Spleen cells from 6-month-oldSWR or SNF1 mice were cocultured with the various concen-trations of celecoxib for 24 hours. Apoptotic cells were de-tected by staining with annexin V and propidium iodide (BDPharMingen), and at the same time, whole splenocytes werestained with antibodies to CD4-PerCP, B220-allophycocyanin,CD19-PerCP, CD11c-FITC, and CD11b-FITC to analyze ap-optosis of CD4 T cells, B cells, DCs, and macrophages,respectively. Apoptosis in the specific cell subpopulationsgated by flow cytometry was calculated as (% of experimentalapoptosis � % of spontaneous apoptosis)/(100 � % of spon-taneous apoptosis) (16).

Enzyme-linked immunospot (ELISpot) assay. Amouse IFN� ELISpot assay was used to detect the T cellresponse to nucleosomes, as described previously (21).

Helper assays for IgG autoantibody production.Splenic T cells (2.5 � 105/well) from mice treated continuouslyfor 6 weeks or splenic T cells from control SNF1 mice werecocultured with treated or control SNF1 splenic B cells (2.5 �105/well) in 96-well plates for 7 days in the presence of 5 �g/mlnucleosomes, as described previously (13,20). Culture super-

COX-2 AND LUPUS AUTOIMMUNITY 4133

natants were collected, freeze-thawed, and assayed by ELISAfor IgG autoantibodies.

Histopathologic analysis. Kidneys were fixed in 10%buffered formalin and paraffin embedded. Tissue sectionswere then stained with hematoxylin and eosin and periodicacid–Schiff. The sections were graded in a blinded manner forpathologic changes on a scale of 0–4�, as described previously(21,23).

Statistical analysis. Log rank tests and Student’s2-tailed t-tests were used for statistical analyses. Results areexpressed as the mean � SEM.

RESULTS

Significantly prolonged survival of lupus-proneSNF1 mice following intermittent treatment with low-dose COX-2 inhibitor diet. Because the COX-2 inhibitorcelecoxib can cause apoptosis of human lupus T cellsand also blocks autoantigen recognition and autoanti-body production in vitro in the cells of SNF1 mice (16),we first treated SNF1 mice with serologically active lupususing intermittent pulses of celecoxib, to determine theeffects of this therapy on lupus in vivo. We treated12-week-old SNF1 mice with a diet containing high(1,500 ppm), moderate (1,000 ppm), or low (500 ppm)doses of celecoxib in the following cycle: 3 weeks offeeding with the celecoxib diet followed by 4 weeks withregular diet.

As determined by HPLC, the mean concentra-tions of celecoxib in the plasma of mice receiving 500ppm, 1,000 ppm, and 1,500 ppm celecoxib were 0.75 �M,1.3 �M, and 2.0 �M, respectively (0.27 �g/ml, 0.50�g/ml, and 0.76 �g/ml, respectively). The maximumplasma concentration (Cmax) of celecoxib (Celebrex)reaches 0.705 �g/ml in human subjects, as has beendetermined in patients receiving a single 200-mg dose(see Physician’s Desk Reference for the pharmacokinet-ics of Celebrex); this Cmax value in humans is 2.6-foldhigher than that reached in the mice receiving thelow-dose (500 ppm) celecoxib diet.

Although differences in the incidence of severeproteinuria between the treatment groups were notmarked, intermittent therapy with the lowest dose ofcelecoxib had the most beneficial effect in prolongingsurvival up to 24 months of age (P � 0.05 by log ranktest), as shown in Figure 1A. The low-dose intermittentcelecoxib therapy also was the most effective in delayingthe onset of severe proteinuria (Figure 1B), but cumu-latively, the between-group differences (by log rank test)did not reach significance.

Significant delay in onset of severe proteinuria,but failure to prolong mouse survival time, followingcontinuous treatment with low-dose COX-2 inhibitordiet. To examine whether continuous therapy with cele-coxib would be more beneficial than intermittent ther-apy, and also whether adding aspirin would counteractany cardiovascular side effects, SNF1 mice were startedon the COX-2 inhibitor diet at 12 weeks of age, at a timewhen high levels of serum autoantibodies were presentbut proteinuria was �2� (�100 mg/dl). A total of 40mice was randomly separated into 4 treatment groups(10 mice per group). Three treatment groups were fedcontinuously with a diet containing 500 ppm of cele-coxib, 400 ppm of aspirin, or a combination of 200 ppmaspirin and 500 ppm celecoxib (continuous combination

Figure 1. Effect of intermittent therapy with 500, 1,000, or 1,500 partsper million (ppm) dietary doses of celecoxib (CC) in the (SWR �NZB)F1 (SNF1) mouse model of spontaneous systemic lupus erythem-atosus. A, Percent survival and B, incidence of severe lupus nephritisamong the treatment groups versus untreated controls. Starting at age3 months, female SNF1 mice (10 mice/group) received cycles of aregular rodent chow diet containing celecoxib for 3 weeks, followed bythe regular diet alone for 4 weeks, and the treatment cycles werecontinued throughout the observation period. The control mice re-ceived the same regular diet without the drug.

4134 ZHANG ET AL

treatment). Another treatment group received a dietcontaining the combination of 200 ppm aspirin and 500ppm celecoxib, but this was administered intermittently,comprising 3 weeks on combination therapy and 4 weeksoff the therapy (intermittent combination treatment).The control group of mice received the same diet butwithout the drugs.

Figure 2B shows that continuous treatment withthe combination of COX-2 inhibitors significantly de-layed the onset of severe proteinuria as compared withthat in the control group (P � 0.05 by log rank test). Atage 26 weeks, 70% of control mice had developedpersistent proteinuria of �2� that rapidly progressed to4�, whereas in the continuous combination treatmentgroup the incidence of severe proteinuria was only 30%.

However, in mice receiving 500 ppm celecoxib, 400 ppmaspirin, or intermittent combination treatment, therewas no significant difference in the incidence of severeproteinuria, as compared with controls, at age 26 weeks.Moreover, by 9.5 months of age, the incidence of severeproteinuria, even in the continuous combination treat-ment group, quickly reached 100% (as in the controls).

Survival in all of the treated groups was notprolonged as compared with that in the control group(P � 0.05 by log rank test) (Figure 2A). Indeed, thegroup receiving aspirin alone had accelerated mortality,although the dietary intake and body weight of theanimals were similar to those in the other groups.

When the 2 sets of results were compared,namely, those in mice treated intermittently with cele-coxib (500 ppm) (Figure 1A) versus those in micetreated intermittently with the celecoxib plus aspirincombination therapy (Figure 2A), clearly the intermit-tent therapy with low-dose celecoxib alone was better inprolonging overall survival (survival up to age 24 monthsversus survival up to age 13.5 months; P � 0.05 by logrank test) and was more effective in decreasing theincidence of nephritis, especially after 9 months of age(Figures 1B and 2B). At age 38 weeks, 100% of the micein the intermittent combination therapy group had de-veloped severe nephritis, as compared with only 65% ofthose treated intermittently with celecoxib alone (at9.5–15 months of age; P � 0.05 by log rank test). Theonly difference between the 2 treatment groups was thepresence of aspirin (200 ppm) in the diet of the inter-mittent combination treatment group, which appearedto result in a worse outcome.

Significant reductions in serum antinuclearautoantibody levels following treatment with low-doseCOX-2 inhibitor diet. To determine the effect of in vivoinhibition of COX-2, we measured serum levels of IgGantinuclear autoantibodies after short-term treatment ofSNF1 mice for 6 weeks. All 3 low-dose treatment groups(500 ppm celecoxib, 400 ppm aspirin, and continuouscombination treatment) showed significantly reducedserum concentrations of IgG anti-dsDNA, anti-ssDNA,antinucleosome, and antihistone autoantibodies as com-pared with the control group (P � 0.01 to P � 0.05)(Figure 3). In the SNF1 mouse model of lupus, it isknown that nephritogenic autoantibodies have high af-finity for ssDNA and nucleosomes (9,24).

Inhibition of early glomerulonephritis by low-dose COX-2 inhibitor treatment. Persistent, severe pro-teinuria (2� to 4�, or 100 mg/dl to �2,000 mg/dl) inSNF1 mice is always associated with a 3� to 4� grade ofglomerulonephritis, as determined by histopathology

Figure 2. Effect of continuous therapy with low doses of celecoxib(CC), aspirin (ASP), or a combination of celecoxib and aspirin(Combo), or intermittent therapy with the combination diet (ComboIntermittent) in the (SWR � NZB)F1 (SNF1) mouse model ofspontaneous systemic lupus erythematosus. A, Percent survival and B,incidence of severe lupus nephritis among the treatment groups versusuntreated controls. Three-month-old female SNF1 mice (10 mice/group) received a diet containing celecoxib (500 parts per million[ppm]) or aspirin (400 ppm) or received continuous combination orintermittent combination therapy (combination diet containing 500ppm celecoxib and 200 ppm aspirin), or the regular diet as control.

COX-2 AND LUPUS AUTOIMMUNITY 4135

(21,25), Therefore, we wanted to assess renal pathologicfeatures at the earliest ages of the mice. Kidney sectionsfrom control mice and from age-matched, low-doseCOX-2 inhibitor–treated mice were examined at the ageof 18 weeks, at which time the mice had been treated for6 weeks. In contrast to the treated mice, the controlgroup had markedly higher histopathology scores (P �0.005), with typical lesions of severe lupus glomerulone-phritis, including glomerular enlargement, hypercellu-larity, crescents, mesangial thickening, and glomerulo-sclerosis (Figures 3B and C). In addition, perivascularand interstitial infiltration with mononuclear cells wasevident in the control mice (Figures 3B and C).

Hyperexpression of intracellular COX-2 in Tcells, B cells, DCs, and macrophages of lupus-proneSNF1 mice. Human lupus T cells resist AICD by up-regulating COX-2 expression, and high doses (50 �M) ofcelecoxib in vitro can cause apoptosis of T cells (16).However, intermittent therapy with the lowest dose ofcelecoxib had the most beneficial effect in vivo in thelupus-prone SNF1 mice (Figure 1), without diminishing

the expression of intracellular COX-2 (results notshown). Therefore, celecoxib might be affecting otherCOX-2–expressing cells that are involved in lupuspathogenesis, by inhibiting the enzymatic function ofCOX-2.

As indicated in Figures 4A–C, intracellularCOX-2 levels in CD4 T cells, as well as in DCs,macrophages, and spontaneously activated(CD19�,CD86high) B cells, of SNF1 mice were a mean2–3.5-fold higher than those in control SWR or BALB/cmice (P � 0.05), indicating that COX-2 is important inthe functioning of various lupus cells. Although Westernblotting involved fractionation of cell subsets, whichmight have caused some cellular activation during invitro manipulations, the results for COX-2 expressionwere nevertheless consistent with those obtained by flowcytometry of gated splenocyte subsets.

Induction of apoptosis of T cells, B cells, DCs,and macrophages in vitro following treatment withCOX-2 inhibitors. Because SNF1 mouse T cells, acti-vated B cells, DCs, and macrophages have higher basal

Figure 3. Effects of low doses of celecoxib (CC), aspirin (ASP), orboth (Combo) given continuously on serum levels of IgG-class auto-antibodies (autoAb) and development of glomerulonephritis. A,Twelve-week-old mice (5 mice/group) were continuously treated withthe celecoxib diet (500 parts per million [ppm]), aspirin (400 ppm),combination diet, or control (Ctrl) diet for 6 weeks. Blood sampleswere assessed for IgG autoantibodies to double-stranded DNA(dsDNA), single-stranded DNA (ssDNA), nucleosomes, and histone. �

� P � 0.05; �� � P � 0.01, versus controls. B, Another group of micewas treated with regimens identical to those in A and evaluated, aftertermination, for renal pathologic features of lupus nephritis, withscoring of 5 mice per group on 2 sections of each kidney. �� � P �0.005 versus controls. Results in A and B are the mean and SEM. C,Representative examples of kidney sections from the animals de-scribed in B (stained with hematoxylin and eosin; original magnifica-tion � 200). In contrast to the kidney from an animal treated with thecelecoxib (500 ppm) diet (right), the kidney from an animal receivingthe control diet (left) shows marked enlargement and hypercellularityof glomeruli with crescent formation, hyalinization, sclerosis, basementmembrane and mesangial thickening, and interstitial infiltrates ofmononuclear cells.

4136 ZHANG ET AL

Figure 4. Hyperexpression of cyclooxygenase 2 (COX-2) in cells of autoimmune (SWR � NZB)F1 (SNF1) mice, and apoptosis of CD4 T cells, Bcells, macrophages (M�), and dendritic cells (DCs) in vitro following treatment with celecoxib. A, Intracellular COX-2 expression in gated CD4 Tcells, spontaneously activated (CD86high) B cells, macrophages, and DCs, after staining of whole splenocytes from 6-week-old unmanipulated SNF1,SWR, or BALB/c mice and analysis by flow cytometry. � � P � 0.05 versus SNF1 strain. B, Representative histograms (determined using FlowJosoftware; Y-axis scale indicates the percent maximum), showing expression of COX-2high cells (thick lines) within a marker, versus isotype control(shaded gray areas). Percentage values in A are the mean and SEM of 5 mice per group from 5 experiments. Values shown over the horizontal barsin B are the percentage values within the range (� SEM) of the mean values shown in A. C, Western blotting for COX-2 expression in fractionatedsplenocytes (upper panel) or whole splenocytes (lower panel) from SWR, NZB, SNF1, and BALB/c mice. Results are representative of 2 separateexperiments. D, Frequency of specific apoptosis induced in vitro by celecoxib in splenocyte subsets of SNF1 mice. Splenocytes from 6-month-oldSNF1 mice with overt nephritis or age-matched SWR mice were cultured with celecoxib in the indicated concentrations for 24 hours, and then stainedfor analysis by flow cytometry. Apoptotic (annexin V–positive, propidium iodide–negative) cells were analyzed in gated cell subsets (n � 5 per strain).Values are the mean and SEM. � � P � 0.05; �� � P � 0.01, versus the same cell subsets in SWR mice.

COX-2 AND LUPUS AUTOIMMUNITY 4137

levels of COX-2 proteins as compared with those innonautoimmune SWR or BALB/c strains, we checkedwhether inhibition of COX-2 enzymatic activity couldcause the apoptosis of these cells. Figure 4D shows thattreatment with celecoxib at 25 �M induced the apoptosisof T cells, B cells, DCs, and macrophages of SNF1 mice,after 24 hours’ incubation, to a greater extent than thatin the cells from SWR mice. Since the plasma concen-tration of celecoxib in mice that received the low-dose(500 ppm) celecoxib diet was 0.75 �M, it was interestingthat this concentration of celecoxib could induce splenicDCs, but not other cells, to undergo some apoptosis inthis short-term assay. Indeed, SNF1 and human lupus Tcells are similar in terms of requiring high levels ofCOX-2 to undergo apoptosis in vitro (16).

Suppression of the T cell response to nucleo-somes and inhibition of autoantigen presentation byAPCs in vivo following treatment with low-dose COX-2inhibitors. SNF1 mice at 12 weeks of age were started onthe lowest dose of COX-2 inhibitor–containing diet for a10-week period. T cells from treated mice were cocul-tured with APCs from age-matched (22-week-old) un-treated control SNF1 mice or, conversely, APCs fromtreated mice were cocultured with T cells from un-treated control SNF1 mice in the presence of nucleo-somes. The IFN� production response by T cells to theautoantigen was measured. T cells from the 3 treatedgroups, especially the low-dose (500 ppm) celecoxib andcontinuous combination treatment groups, showed asignificantly reduced autoantigen presentation and Tcell response (P � 0.01), even at the higher doses ofautoantigen (Figures 5A–C).

Suppression of autoantibody-inducing T helpercell function and reduced capacity of autoimmune Bcells to receive help from T cells following low-doseCOX-2 inhibitor therapy. To determine whether thereduction in serum autoantibodies by COX-2 inhibitionin vivo was due to suppression of autoimmune T cells orB cells, or both, we tested T cells and B cells from bothcontrol and COX-2 inhibitor–treated mouse spleensafter 6 weeks of the COX-2 inhibitor diet. Crisscrosshelper assays were done in the presence of nucleosomes,and autoantibody levels were measured in the superna-tants of cell cultures.

We found that the levels of IgG-class anti-dsDNA, anti-ssDNA, antinucleosome, and antihistoneautoantibodies in culture supernatants of T cells fromthe 3 different treatment groups of SNF1 mice cocul-tured with B cells of untreated control SNF1 mice weresignificantly reduced in comparison with the levels inculture supernatants of T cells from untreated control

Figure 5. Effect of low-dose cyclooxygenase 2 inhibitor diet therapy onthe nucleosome-induced interferon-� (IFN�) response in lupus T cellsand the autoantigen presentation function of lupus antigen-presentingcells (APCs). Three-month-old (SWR � NZB)F1 (SNF1) mice weretreated with the lowest doses of A, celecoxib (CC) or B, aspirin (ASP), orC, with a low-dose combination of both (Combo), or given a regular dietwithout drugs as control (Ctrl), continuously for 10 weeks. Splenic T cellsand APCs were then separated and crisscross cocultures were done asfollows: untreated T cells (Ctrl-T) � untreated APCs (Ctrl-APC) (servingas the untreated control in A–C), untreated T cells (Ctrl-T) � treatedAPCs (CC-APC, ASP-APC, or Combo-APC), treated T cells (CC-T,ASP-T, or Combo-T) � treated APCs (CC-APC, ASP-APC, or Combo-APC), or treated T cells (CC-T, ASP-T, or Combo-T) � untreated APCs(Ctrl-APC), all in the presence of various doses of nucleosomes. IFN�responses on enzyme-linked immunospot assay are expressed as the meanand SEM number of IFN�-positive spots per 106 T cells (n � 4 pergroup). Values are the mean and SEM. � � P � 0.01 versus untreatedcontrol. The mean � SEM baseline number of IFN� spots in SNF1 T cellscultured with APCs without nucleosomes was 6 � 2 spots per 106 T cells.

4138 ZHANG ET AL

animals cocultured with B cells of untreated controlanimals (P � 0.001) (Figure 6). A similar extent ofsuppression could be seen in cocultures of B cells fromtreated animals and T cells from untreated animals whencompared with cocultures of B and T cells from un-treated animals. These results show that COX-2 inhibi-tors can suppress both the function of T helper cells andthe ability of autoimmune B cells to receive autoantigen-specific help for autoantibody production.

DISCUSSION

Based on our studies on lupus T cells in vitro inhumans (16), we designed the present study to test the

effects of deletion of autoimmune T helper cells in vivoby intermittent pulse therapy with celecoxib in mice withlupus. Even when the therapy was administered inter-mittently, the beneficial effects of celecoxib were evidentin mice that had clinically overt lupus autoimmunity atthe start of treatment. Surprisingly, the lowest dose ofcelecoxib diet (500 ppm) was the most effective indelaying the onset of severe nephritis and in increasingsurvival. Plasma levels of celecoxib were, on average,0.27 �g/ml (0.72 �M) at the lowest dose, which is2.6-fold less than that achieved by intake of 200 mg ofcelecoxib/day in humans. Although such low doses areless likely to cause adverse cardiovascular events (26), toavoid complications, we fed mice with lupus continu-ously with a combination of low doses of celecoxib andaspirin (27–29). This regimen, when given continuouslybut not intermittently, did delay lupus nephritis tempo-rarily, but failed to prolong survival, probably due tosome long-term adverse effects of continuous therapy invivo. Indeed, addition of aspirin led to a worsenedoutcome following intermittent celecoxib therapy, andaspirin given by itself accelerated mortality in the micewith lupus, although aspirin is an approved drug forpatients with lupus and is not toxic at the low dosageused in this study in other strains of mice (28).

In the early weeks of therapy, the combinationregimen with both celecoxib and aspirin was effective inreducing serum levels of potentially pathogenic autoan-tibodies and in ameliorating the histopathologic changesof lupus glomerulonephritis (Figure 3). Moreover, ther-apy with the COX-2 inhibitors in the short term, even atthe lowest doses, could also suppress the functions ofautoimmune T helper cells, B cells, and APCs, reducingautoantibody production in response to the major lupusautoantigen, nucleosome, and also its presentation (Fig-ures 5 and 6). Although the low doses of celecoxib didnot cause apoptosis of T or B cells in the short term invitro, these autoimmune cells could have been inhibitedby other mechanisms, such as the blocking of NF-kBactivation (30). Indeed, activation markers, such asCD86 and CD40, were decreased overall in the B cellsand DCs of SNF1 mice after low-dose celecoxib therapy(results not shown).

Nevertheless, due to some adverse effects in vivoover the long term, there was no overall benefit, in termsof survival, with continuous administration of celecoxiband/or aspirin. It could be that generation of regulatoryT cells, which are important in controlling lupus(21,31,32), was impaired by continuous inhibition ofCOX-2 (33). Thus, for this or additional reasons, inter-mittent, but not continuous, therapy with low doses of

Figure 6. Effect of low-dose cyclooxygenase 2 (COX-2) inhibitor diettherapy in vivo on autoantigen-specific T helper cell and B cellfunction, assessed as IgG autoantibody (autoAb) production. Twelve-week-old (SWR � NZB)F1 (SNF1) mice received the lowest doses ofCOX-2 inhibitor–containing diet for 6 weeks prior to termination. T orB cells from mice fed continuously with 500 parts per million (ppm)celecoxib (CC T or CC B), 400 ppm aspirin (ASP T or ASP B), or alow-dose combination of both drugs (Combo T or Combo B) wereused. IgG autoantibody production was determined in T cells fromtreated mice cocultured for 7 days with B cells from control (Ctrl)SNF1 mice in the presence of 5 �g/ml of nucleosomes (top) or in Bcells from treated mice cocultured for 7 days with T cells from controlSNF1 mice in the presence of nucleosomes (bottom). Levels of IgGautoantibodies produced in the culture supernatants are the mean andSEM from 5 experiments. �� � P � 0.001 versus control. Baselinelevels of IgG autoantibodies produced by B cells cultured alone wereas follows: for anti–double-stranded DNA (anti-dsDNA), 0.004 �0.001 mg/dl, for anti–single-stranded DNA (anti-ssDNA), 0.003 �0.0006 mg/dl, for antinucleosome, 0.003 � 0.001 mg/dl, and forantihistone, 0.003 � 0.0005 mg/dl.

COX-2 AND LUPUS AUTOIMMUNITY 4139

celecoxib alone is better in prolonging survival in theselupus-prone mice.

It is interesting that celecoxib in the lowest doserange could cause modest apoptosis of SNF1 DCs in ashort-term in vitro assay, but low-dose celecoxib diettherapy significantly suppressed autoantigen presenta-tion by lupus APCs. Up-regulation of COX-2, andconsequently of prostaglandin E2, is needed for survival,maturation, and activation of human DCs (34,35), andhyperactivity of DCs in lupus leads to immunogenicpresentation of autoantigens (7,36). Indeed, not only Tcells, but also DCs and macrophages of SNF1 miceconstitutively hyperexpressed COX-2 (Figure 4), andNF-�B activation, which is needed for functioning ofthese cells of the lupus immune system, could have beeninhibited by celecoxib (30).

The hyperactive B cells of lupus not only producepathogenic autoantibodies, but also costimulate andpresent autoantigens to pathogenic T cells, and maystimulate themselves (1,37–39). We have found thatspontaneously activated lupus B cells constitutively ex-press high levels of COX-2, as do mitogenically stimu-lated, normal human B cells (40,41). Interestingly,CD86high (activated) B cells from SNF1 mice expressedmore COX-2 than did the same subset from nonauto-immune strains, indicating a lupus-intrinsic defect (Fig-ure 4). Indeed, COX-2 inhibitor therapy could suppressboth autoantigen-specific T helper function and the abilityof autoimmune B cells to receive autoantigen-specific helpfor autoantibody production (Figures 5 and 6).

Finally, COX-2 is also up-regulated in inflamma-tory macrophages and glomerular mesangial cells of thekidney in lupus (42,43). Thus, major players in thepathogenic autoimmune response that are abnormallyhyperactive and resistant to apoptosis in lupus dependon increased activity of COX-2, similar to inflammatorycells in the target organs. In addition, our observationsshow that intermittent therapy with low doses of aCOX-2 inhibitor increases survival irrespective of theseverity of nephritis, by an additional mechanism(s).

In relation to human lupus, this study shows theimportance of COX-2 as a target for lupus therapy, usingmore benign COX-2 inhibitors that lack adverse cardio-vascular effects. It would be interesting to explorewhether intermittent pulse therapy with low doses of aselective COX-2 inhibitor in combination with cardio-protective statins, which also have beneficial effects inlupus models (44), would provide added benefit. More-over, with a currently permissible COX-2–specific inhib-itor such as celecoxib, a low-dose intermittent adminis-tration might have a more favorable outcome.

ACKNOWLEDGMENTS

We thank Dr. Hee-Kap Kang and Michael Liu for theircritical comments and advice regarding the ELISpot assay andELISA. We also thank Dr. David Pisetsky (Duke University)for suggesting the use of aspirin as part of the therapy.

AUTHOR CONTRIBUTIONS

Dr. Datta had full access to all of the data in the study andtakes responsibility for the integrity of the data and the accuracy of thedata analysis.Study design. Zhang, Datta.Acquisition of data. Zhang, Bertucci, Smith, Xu.Analysis and interpretation of data. Zhang, Bertucci, Xu, Datta.Manuscript preparation. Zhang, Bertucci, Xu, Datta.Statistical analysis. Zhang.

REFERENCES

1. Shlomchik MJ, Craft JE, Mamula MJ. From T to B and backagain: positive feedback in systemic autoimmune disease. Nat RevImmunol 2001;1:147–53.

2. Krishnan S, Farber DL, Tsokos GC. T cell rewiring in differenti-ation and disease. J Immunol 2003;171:3325–31.

3. Datta SK, Zhang L, Xu L. T-helper cell intrinsic defects in lupusthat break peripheral tolerance to nuclear autoantigens. J MolMed 2005;83:267–78.

4. Mohan C, Adams S, Stanik V, Datta SK. Nucleosome: a majorimmunogen for the pathogenic autoantibody-inducing T cells oflupus. J Exp Med 1993;177:1367–81.

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. Greidinger EL, Gazitt T, Jaimes KF, Hoffman RW. Human T cellclones specific for heterogeneous nuclear ribonucleoprotein A2autoantigen from connective tissue disease patients assist in auto-antibody production. Arthritis Rheum 2004;50:2216–22.

7. 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.

8. Leadbetter EA, Rifkin IR, Hohlbaum AM, Beaudette BC, Shlom-chik MJ, Marshak-Rothstein A. Chromatin-IgG complexes acti-vate B cells by dual engagement of IgM and Toll-like receptors.Nature 2002;416:595–8.

9. Berland R, Fernandez L, Kari E, Han JH, Lomakin I, Akira S, etal. Toll-like receptor 7-dependent loss of B cell tolerance inpathogenic autoantibody knockin mice. Immunity 2006;25:429–40.

10. Daikh DI, Finck BK, Linsley PS, Hollenbaugh D, Wofsy D.Long-term inhibition of murine lupus by brief simultaneous block-ade of the B7/CD28 and CD40/gp39 costimulation pathways.J Immunol 1997;159:3104–8.

11. Peng SL, McNiff JM, Madaio MP, Ma J, Owen MJ, Flavell RA, etal. �� T cell regulation and CD40 ligand dependence in murinesystemic autoimmunity. J Immunol 1997;158:2464–70.

12. Datta SK, Patel H, Berry D. Induction of a cationic shift in IgGanti-DNA autoantibodies: role of T helper cells with classical andnovel phenotypes in three murine models of lupus nephritis. J ExpMed 1987;165:1252–68.

13. Mohan C, Shi Y, Laman JD, Datta SK. Interaction between CD40and its ligand gp39 in the development of murine lupus nephritis.J Immunol 1995;154:1470–80.

14. Refaeli Y, Van Parijs L, London CA, Tschopp J, Abbas AK.Biochemical mechanisms of IL-2–regulated Fas-mediated T cellapoptosis. Immunity 1998;8:615–23.

4140 ZHANG ET AL

15. Yi Y, McNerney M, Datta SK. Regulatory defects in Cbl andmitogen-activated protein kinase (extracellular signal-related ki-nase) pathways cause persistent hyperexpression of CD40 ligand inhuman lupus T cells. J Immunol 2000;165:6627–34.

16. Xu L, Zhang L, Yi Y, Kang HK, Datta SK. Human lupus T cellsresist inactivation and escape death by upregulating COX-2. NatMed 2004;10:411–5.

17. Datta SK, Schwartz RS. Genetics of expression of xenotropic virusand autoimmunity in NZB mice. Nature 1976;263:412–5.

18. Datta SK, McConahey PJ, Manny N, Theofilopoulos AN, DixonFJ, Schwartz RS. Genetic studies of autoimmunity and retrovirusexpression in crosses of NZB mice. II. The viral envelope glyco-protein gp 70. J Exp Med 1978;147:872–81.

19. Zeytin HE, Patel AC, Rogers CJ, Canter D, Hursting SD, SchlomJ, et al. Combination of a poxvirus-based vaccine with a cycloox-ygenase-2 inhibitor (celecoxib) elicits antitumor immunity andlong-term survival in CEA.Tg/MIN mice. Cancer Res 2004;64:3668–78.

20. Kaliyaperumal A, Michaels MA, Datta SK. Naturally processedchromatin peptides reveal a major autoepitope that primes patho-genic T and B cells of lupus. J Immunol 2002;168:2530–7.

21. Kang HK, Michaels MA, Berner BR, Datta SK. Very low-dosetolerance with nucleosomal peptides controls lupus and inducespotent regulatory T cell subsets. J Immunol 2005;174:3247–55.

22. Kang HK, Liu M, Datta SK. Low-dose peptide tolerance therapyof lupus generates plasmacytoid dendritic cells that cause expan-sion of autoantigen-specific regulatory T cells and contraction ofinflammatory Th17 cells. J Immunol 2007;178:7849–58.

23. Kalled SL, Cuttler AH, Datta SK, Thomas DW. Anti-CD40 ligandantibody treatment of SNF1 mice with established nephritis:preservation of kidney function. J Immunol 1998;160:2158–65.

24. Gavalchin J, Datta SK. The NZB � SWR model of lupusnephritis. II. Autoantibodies deposited in renal lesions show arestricted idiotypic diversity. J Immunol 1987;138:138–48.

25. Kaliyaperumal A, Michaels MA, Datta SK. Antigen-specific ther-apy of murine lupus nephritis using nucleosomal peptides: toler-ance spreading impairs pathogenic function of autoimmune T andB cells. J Immunol 1999;162:5775–83.

26. Solomon DH, Avorn J, Sturmer T, Glynn RJ, Mogun H, Schnee-weiss S. Cardiovascular outcomes in new users of coxibs andnonsteroidal antiinflammatory drugs: high-risk subgroups andtime course of risk. Arthritis Rheum 2006;54:1378–89.

27. Wilner KD, Rushing M, Walden C, Adler R, Eskra J, Noveck R,et al. Celecoxib does not affect the antiplatelet activity of aspirin inhealthy volunteers. J Clin Pharmacol 2002;42:1027–30.

28. Sansom OJ, Stark LA, Dunlop MG, Clarke AR. Suppression ofintestinal and mammary neoplasia by lifetime administration ofaspirin in ApcMin/� and ApcMin/�, Msh2�/� mice. Cancer Res2001;61:7060–4.

29. Cheng Y, Wang M, Yu Y, Lawson J, Funk CD, FitzGerald GA.Cyclooxygenases, microsomal prostaglandin E synthase-1, andcardiovascular function. J Clin Invest 2006;116:1391–9.

30. Shishodia S, Koul D, Aggarwal BB. Cyclooxygenase (COX)-2inhibitor celecoxib abrogates TNF-induced NF-�B activationthrough inhibition of activation of I�B� kinase and Akt in humannon-small cell lung carcinoma: correlation with suppression ofCOX-2 synthesis. J Immunol 2004;173:2011–22.

31. Singh RR, Ebling FM, Albuquerque DA, Saxena V, Kumar V,Giannini EH, et al. Induction of autoantibody production islimited in nonautoimmune mice [published erratum appears inJ Immunol 2002;169:1647]. J Immunol 2002;169:587–94.

32. Zheng SG, Wang JH, Koss MN, Quismorio F Jr, Gray JD, HorwitzDA. CD4� and CD8� regulatory T cells generated ex vivo withIL-2 and TGF-� suppress a stimulatory graft-versus-host diseasewith a lupus-like syndrome. J Immunol 2004;172:1531–9.

33. Mahic M, Yaqub S, Johansson CC, Tasken K, Aandahl EM.FOXP3�CD4�CD25� adaptive regulatory T cells express cyclo-oxygenase-2 and suppress effector T cells by a prostaglandinE2-dependent mechanism. J Immunol 2006;177:246–54.

34. Baratelli F, Krysan K, Heuze-Vourc’h N, Zhu L, Escuadro B,Sharma S, et al. PGE2 confers survivin-dependent apoptosisresistance in human monocyte-derived dendritic cells. J LeukocBiol 2005;78:555–64.

35. Fogel-Petrovic M, Long JA, Knight DA, Thompson PJ, UphamJW. Activated human dendritic cells express inducible cyclo-oxygenase and synthesize prostaglandin E2 but not prostaglandinD2. Immunol Cell Biol 2004;82:47–54.

36. Kalled SL, Cutler AH, Burkly LC. Apoptosis and altered dendriticcell homeostasis in lupus nephritis are limited by anti-CD154treatment. J Immunol 2001;167:1740–7.

37. Mamula MJ, Fatenejad S, Craft J. B cells process and presentlupus autoantigens that initiate autoimmune T cell responses.J Immunol 1994;152:1453–61.

38. Sfikakis PP, Boletis JN, Lionaki S, Vigklis V, Fragiadaki KG,Iniotaki A, et al. Remission of proliferative lupus nephritis follow-ing B cell depletion therapy is preceded by down-regulation of theT cell costimulatory molecule CD40 ligand: an open-label trial.Arthritis Rheum 2005;52:501–13.

39. Desai-Mehta A, Lu L, Ramsey-Goldman R, Datta SK. Hyperex-pression of CD40 ligand by B and T cells in human lupus and itsrole in pathogenic autoantibody production. J Clin Invest 1996;97:2063–73.

40. Ryan EP, Pollock SJ, Murant TI, Bernstein SH, Felgar RE, PhippsRP. Activated human B lymphocytes express cyclooxygenase-2 andcyclooxygenase inhibitors attenuate antibody production. J Immu-nol 2005;174:2619–26.

41. Mongini PK, Inman JK, Han H, Fattah RJ, Abramson SB, AtturM. APRIL and BAFF promote increased viability of replicatinghuman B2 cells via mechanism involving cyclooxygenase 2. J Im-munol 2006;176:6736–51.

42. Tomasoni S, Noris M, Zappella S, Gotti E, Casiraghi F, BonazzolaS, et al. Upregulation of renal and systemic cyclooxygenase-2 inpatients with active lupus nephritis. J Am Soc Nephrol 1998;9:1202–12.

43. Zoja C, Benigni A, Noris M, Corna D, Casiraghi F, Pagnoncelli M,et al. Mycophenolate mofetil combined with a cyclooxygenase-2inhibitor ameliorates murine lupus nephritis. Kidney Int 2001;60:653–63.

44. Lawman S, Mauri C, Jury EC, Cook HT, Ehrenstein MR. Ator-vastatin inhibits autoreactive B cell activation and delays lupusdevelopment in New Zealand black/white F1 mice. J Immunol2004;173:7641–6.

COX-2 AND LUPUS AUTOIMMUNITY 4141