ARTHRITIS & RHEUMATISMVol. 54, No. 4, April 2006, pp 1143–1150DOI 10.1002/art.21729© 2006, American College of Rheumatology

Evaluation of Classical Complement PathwayActivation in Rheumatoid Arthritis

Measurement of C1q–C4 Complexes as Novel Activation Products

Diana Wouters,1 Alexandre E. Voskuyl,2 Esmeralda T. H. Molenaar,2 Ben A. C. Dijkmans,2

and C. Erik Hack2

Objective. Novel activation products that are sta-ble and minimally susceptible to in vitro artefacts haverecently been described in the classical complementpathway. The present study assessed circulating levelsof these products, i.e., covalent complexes between therecognition molecule of the classical pathway (C1q) andactivated C4, in plasma samples from patients withrheumatoid arthritis (RA) to establish the relationshipbetween these levels and the clinical and immunologicparameters in these patients.

Methods. C1q–C4 levels were measured in plasmasamples from 41 patients with active RA and 43 patientswith inactive RA. These levels were related to othercomplement activation products and to disease activityaccording to the Disease Activity Score in 28 joints(DAS28), using Spearman’s rank correlations.

Results. C1q–C4 plasma levels were significantlyhigher in patients with active RA as compared withpatients with RA in clinical remission (median 3.3arbitrary units [AU], range 0.4–13.4 versus 1.7 AU,range 0.2–5.5; P � 0.0001), suggesting that activation ofthe classical complement pathway reflects disease activ-ity. This was supported by a significant correlationbetween C1q–C4 levels and the DAS28 (r � 0.398, P �0.0002). Levels of other complement activation prod-

ucts, such as activated C4 (C4b/c), were also signifi-cantly elevated in patients with active disease comparedwith patients with inactive disease (P � 0.03), and werecorrelated with C1q–C4 levels (r � 0.329, P � 0.002).Levels of C1q–C4 complexes were higher in synovialfluid samples than in plasma samples from the 4patients tested.

Conclusion. Systemic complement activation viathe classical pathway in patients with RA correlates withdisease activity. These results indicate that C1q–C4complexes may be used as a biomarker for RA.

Rheumatoid arthritis (RA) is an inflammatoryautoimmune disease characterized by chronic inflamma-tion of the joints, eventually leading to bone and carti-lage destruction. Although the etiology of RA is un-known, complement activation has been implicated inthe pathogenesis of the disease. Various studies identifycomplement activation as a main event in the inflamma-tory cascade in RA (1–3). Evidence of complementactivation in synovial fluid is abundant. For example,levels of complement proteins are depressed in thesynovial fluid of patients with RA, reflecting consump-tion of complement. Moreover, elevated levels of com-plement cleavage products, such as sC5b-9, C3a, Bb, andC1inh–C1s complexes, have been observed in synovialfluid (4–9).

Involvement of complement in the pathogenesisof RA was also confirmed in experimental studies.Collagen-induced arthritis (CIA), an experimental ani-mal model for human RA, was induced in C3- and factorB–deficient mice (10). In complement-deficient mice,arthritis was reduced or completely absent, whereasnormal mice were susceptible to CIA, indicating animportant role of complement in the induction of dis-

1Diana Wouters, MSc: Sanquin Research at CLB and Aca-demic Medical Centre, Amsterdam, The Netherlands; 2Alexandre E.Voskuyl, MD, PhD, Esmeralda T. H. Molenaar, MD, PhD, Ben A. C.Dijkmans, MD, PhD, C. Erik Hack, MD, PhD (current address:Crucell Holland BV, Leiden, The Netherlands): VU University Med-ical Centre, Amsterdam, The Netherlands.

Address correspondence and reprint requests to Diana Wout-ers, MD, Department of Immunopathology, Sanquin Research atCLB, Plesmanlaan 125, 1066 CX Amsterdam, The Netherlands.E-mail: d.wouters@sanquin.nl.

Submitted for publication June 2, 2005; accepted in revisedform December 19, 2005.

1143

ease. Consistent with this observation, systemic admin-istration of a monoclonal anti-C5 antibody preventedCIA in susceptible mice (11).

RA patients have increased levels of circulatingimmune complexes (12,13). Part of these complexescontains rheumatoid factors (RFs), which are autoanti-bodies against human IgG. RF-containing immune com-plexes are capable of activating complement via theclassical pathway (14–17), with IgM-RF being consider-ably more effective in complement activation thanIgG-RF (18). Thus, immune complexes, particularlythose in inflamed joints, are often assumed to be themain trigger for complement activation in RA. It is notclear to what extent circulating immune complexescontribute to complement activation in RA (19). How-ever, other activators of complement may also contrib-ute to complement activation in RA. Levels of theacute-phase protein, C-reactive protein (CRP), are ele-vated in the majority of patients with RA and areassociated with disease activity. CRP bound to a ligandcan activate complement, and there is evidence thatCRP-mediated activation of complement occurs in RA(20). Both immune complexes and ligand-bound CRPactivate complement via the classical pathway (18,21–23).

A parameter that reliably reflects complementactivation in vivo might constitute a biomarker of RA.However, most, if not all, activation markers of thecomplement system are susceptible to in vitro artefacts,resulting in artificially high levels of activation productsin plasma samples. We recently described novel activa-tion products of complement that not only are specificfor activation of the classical pathway of complement,but also are remarkably stable in plasma (24). Thepresent study was performed to investigate plasma levelsof this new complement parameter, which consists ofcovalent complexes between the recognition molecule ofthe classical complement pathway, C1q, and activatedC4, in RA. We also sought to establish the relationshipbetween classical complement pathway activation andother immunologic parameters in patients with active orinactive RA, and we assessed the association of theseparameters with disease activity.

PATIENTS AND METHODS

Patients. We selected patients with active RA (n � 41)and patients with inactive RA (n � 43) from a cohort of 187patients investigated previously (20). All patients with RAfulfilled the American College of Rheumatology (ACR; for-merly, the American Rheumatism Association) 1987 criteriafor RA (25). Inactive RA was defined by the ACR criteria for

clinical remission (26). Disease activity was assessed by calcu-lating the modified Disease Activity Score in 28 joints(DAS28) (27). In addition, 10 patients treated with anti–tumornecrosis factor (anti-TNF) antibody (infliximab) were studied.All patients received 3 mg/kg infliximab at weeks 0, 2, 6, and 14and every 8 weeks thereafter. Evaluation of disease activityand collection of plasma samples were carried out in thesepatients at baseline and at weeks 22 and 52 after the start oftherapy. The protocol was approved by the local institutionalethics review committee. For control experiments, humanEDTA plasma was obtained from 21 healthy volunteers.

Collection of blood samples. Blood was collected in 10mM EDTA (final concentration) and centrifuged for 10 min-utes at 1,300g to remove the blood cells. Plasma was stored inaliquots at �70°C until used.

Proteins and antibodies. Monoclonal antibody (mAb)anti–C1q-85, directed against the globular heads of C1q, hasbeen described previously (28). The mAb anti–C4-4 wasobtained from a fusion of spleen cells from a mouse immu-nized with C4 and is directed against the C4d fragment,recognizing both native and activated C4. The use of thisantibody in immunoassays has been described previously (29).The mAb anti–C4-1, directed against an activation-dependentneoepitope, has also been described previously (30). C1q wasisolated from human plasma according to the method de-scribed by Tenner and colleagues (31).

Enzyme-linked immunosorbent assay (ELISA) for ac-tivated C4. Overall complement activation was measured byassessing the amount of C4b/bi/c (abbreviated further asC4b/c). For the quantification of these activation products, anELISA was used that has been described previously (29).Briefly, an mAb (anti–C4-1) recognizing a neoepitope onactivated C4 was used as catching antibody. Biotinylatedpolyclonal rabbit anti-human anti-C4 antibody was used fordetection. Since C4 can be hydrolyzed by surrounding watermolecules, C4b/c levels in plasma or serum may be artificiallyhigh due to temperature changes. Aged human serum, con-taining a known amount of activated C4, was used for thecalibration curve.

ELISA for C1q–C4 complexes. To measure C1q–C4complexes, an ELISA that was recently described by our group(24) was used. The catching antibody in this ELISA, anti–C4-4,recognizes both native and activated C4. To prevent saturationof the coating antibody, C1q–C4 complexes were separatedfrom unbound C4 by ammonium sulfate precipitation. To thisend, 1 volume of plasma was incubated with 1 volume of 66%saturated (final concentration 33%, or 1.29M) ammoniumsulfate (Merck, Darmstadt, Germany), dissolved in phosphatebuffered saline (PBS), pH 7.4, containing 10 mM EDTA.Mixtures were placed on ice for 1 hour and then centrifugedfor 30 minutes at 1,300g at 4°C. Precipitates were dissolved inELISA buffer (high-performance ELISA [HPE] [BusinessUnit Immune Reagents; Sanquin, Amsterdam, The Nether-lands] to which 0.5M NaCl and 10 mM EDTA were added toprevent aspecific binding of C1q and in vitro complementactivation, respectively).

The ELISA was then performed with plates coatedwith anti–C4-4 mAb at 5 �g/ml overnight at room temperature(RT), in 0.11M sodium acetate buffer, pH 5.5. Plates werewashed 5 times with PBS–0.02% (weight/volume) Tween priorto 1 hour of incubation with dissolved precipitates at RT. After

1144 WOUTERS ET AL

washing 5 times with PBS–0.02% Tween, plates were incu-bated for 1 hour at RT with biotinylated mAb against C1q(anti–C1q-85) diluted in HPE to detect C1q–C4 complexes.After 5 washes with PBS–0.02% Tween, plates were incubatedwith polymerized horseradish peroxidase (Business Unit Re-agents; Sanquin), diluted 1:10,000 in PBS–2% (v/v) cow milkfor 30 minutes at RT. After 5 washes, the ELISA wasdeveloped with 100 �g/ml tetramethylbenzidine in 0.11Msodium acetate, pH 5.5, containing 0.003% (v/v) H2O2. Sub-strate conversion was stopped by addition of 100 �l H2SO4.

Absorbance was measured at 450 nm with a Titertekmultiscan. Purified C1q, containing C1q–C4 complexes, wasused as the calibration curve for this assay. Levels of complexesin plasma were expressed as arbitrary units (AU), based on theamount of complexes in the purified C1q sample.

Absorption of RFs from plasma. One volume ofplasma was incubated for 30 minutes at RT with 4 volumes ofRF neutralization agent (human gamma globulin–coated mi-croparticles; Abbott Laboratories, Abbott Park, IL) and sub-sequently centrifuged for 10 minutes at 2,000 revolutions perminute. Supernatant was used for testing.

Determination of IgM-RFs. IgM-RFs were measuredin a regular ELISA using human IgG (25 �g/ml) as antigen. Apositive plasma sample containing 200 IU of IgM-RF was usedfor the calibration curve. IgM-RF levels above 12.5 IU/ml wereconsidered to be positive.

Statistical analysis. Levels of complement activationproducts above the upper limit of normal values were consid-ered to be increased. Comparisons between patients withactive disease and those with inactive disease were made usingthe Mann-Whitney test, the unpaired t-test, or the chi-squaretest, depending on whether the values were normally distrib-uted. Comparisons between samples before and after freezingand thawing were made using the paired t-test. Correlationsbetween the various complement activation products anddisease activity were analyzed using Spearman’s rank correla-tion coefficients. P values (2-tailed) less than 0.05 were con-sidered statistically significant.

RESULTS

Patient characteristics. The characteristics of the2 groups of patients (active versus inactive RA) areshown in Table 1. There was no significant difference inage, sex, or disease duration between the 2 groups ofpatients. The DAS28 and the frequency of treatmentwith disease-modifying antirheumatic drugs were, asexpected, significantly higher among the patients withactive disease. Correspondingly, CRP plasma levels andlevels of IgM-RF were also higher in the patients withactive disease.

Increase of C1q–C4 levels in RA. The medianC1q–C4 plasma level in the whole group of 84 RApatients was 2.15 AU (range 0.2–13.4), which was signif-icantly (P � 0.0001) higher than in the healthy controlgroup (median 0.83 AU, range 0.45–1.84) (Figure 1).Increased C1q–C4 plasma levels were found in 80% of

the 41 patients with active RA and in 58% of the 43patients with inactive RA. The median level of C1q–C4in patients with active RA (3.3 AU, range 0.4–13.4) wassignificantly higher than that in patients with inactiveRA (1.7 AU, range 0.2–5.5) (P � 0.0001, by Mann-Whitney test) (Figure 1). This indicates that the classicalpathway of complement is activated to a higher degreein active disease than in inactive disease.

No influence of RF in C1q–C4 ELISA. A largepercentage of RA patients have RFs that are mainly IgMautoantibodies against epitopes on the constant regionof IgG. Since the antibodies in the C1q–C4 ELISA areof the IgG subclass, RFs may cause false-positive signalsin the ELISA. Therefore, interference by RFs in the

Table 1. Characteristics of the patients with active RA and those withinactive RA*

Active RA(n � 41)

Inactive RA(n � 43) P

Age, years 61 (29–84) 59 (24–86) 0.06Female, no. (%) 35 (85) 30 (70) 0.09Disease duration, years 7 (0–50) 8 (2–32) 0.79DAS28 5.4 (0.9–8.1) 1.7 (0.1–3.6) �0.0001DMARD, no. (%) 40 (98) 25 (58) �0.0001CRP, mg/liter 14 (0.6–112) 3 (0.1–20) �0.0001C4b/c, nM† 36 (9.6–113) 21 (8.7–88) 0.03C1q–C4, AU‡ 3.3 (0.4–27.7) 1.7 (0.2–5.5) 0.0001

* Except where indicated otherwise, values are the median (range).RA � rheumatoid arthritis; DAS28 � Disease Activity Score in 28joints; DMARD � disease-modifying antirheumatic drug; CRP �C-reactive protein; AU � arbitrary units.† Normal value � 8 nM.‡ Normal value � 0.9 AU.

Figure 1. Plasma levels of C1q–C4 complexes in patients with activerheumatoid arthritis (RA) (n � 41), patients with inactive RA (n �43), and normal healthy volunteers (n � 21). Bars show the median,boxes show the 25th and 75th percentiles, and bars outside the boxesshow the 10th and 90th percentiles (in arbitrary units [au]). � � P �0.0001; �� � P � 0.0002; ��� � P � 0.0001.

CLASSICAL COMPLEMENT PATHWAY ACTIVATION IN RA 1145

assay for C1q–C4 needs to be excluded. We did severalexperiments to address this issue. First, RA sampleswere tested in an ELISA with irrelevant mAb, whichyielded no response (results not shown). Second, weabsorbed RFs from RA plasma samples by incubationwith latex beads coated with human IgG. Using thismethod, more than 90% of IgM-RFs were removedfrom the plasma samples, as assessed in an ELISA forRF (all absorbed samples yielded levels below thenormal value of 12.5 IU/ml). However, C1q–C4 levelswere unchanged upon absorption of the RF. Thus, bothcontrol experiments indicated that RFs have no influ-ence on the C1q–C4 ELISA results.

Stability of C1q–C4. To assess the stability ofC1q–C4 complexes in RA plasma samples, we tested theeffect of repeated freezing and thawing on C1q–C4levels. We compared this effect with the effect ofrepeated freezing and thawing on C4b/c levels (30).Fourteen RA plasma samples, randomly chosen fromthe 84 available samples, were repeatedly frozen at�80°C and subsequently thawed. C4b/c and C1q–C4levels were measured in the plasma samples after 5cycles of freezing and thawing, and values were com-pared with those for fresh samples. The median C4b/cplasma level after 5 cycles of freezing and thawing (426.5nM, range 237.9–1,038.0) was more than 10-fold higherand significantly increased as compared with the medianlevel in fresh samples (41.07 nM, range 13.0–105.0; P �0.0001, by paired t-test). In contrast, the median C1q–C4plasma level in the fresh samples (2.8 AU, range 0.4–7.9)did not increase after 5 cycles of freezing and thawing(median 2.6 AU, range 0.3–13.9; P � 0.73, by pairedt-test), indicating that C1q–C4 levels are not sensitive toin vitro activation, consistent with observations in previ-ous studies (24).

Plasma levels of C4b/c in active and inactive RA.Plasma levels of C4b/c were increased in all 84 RApatients. The median level of C4b/c was significantlyhigher in patients with active RA than in patients withinactive RA (P � 0.03, by unpaired t-test) (Table 1).Levels of the various complement parameters correlatedwith each other. Using Spearman’s rank correlationcoefficient, C1q–C4 levels correlated significantly withC4b/c levels (r � 0.329, P � 0.002).

Correlation of C1q–C4 levels with IgM-RFplasma levels. Since IgM-RFs are able to activate theclassical pathway of complement (15,16), we assessedthe relationship between IgM-RF levels and C1q–C4complexes. In all 84 RA patients, levels of IgM-RFswere determined and assessed for correlations withcomplement activation products. The median level of

RF in all RA patients was 51.7 IU/ml, ranging from 1.2IU/ml to 412.3 IU/ml. Levels of IgM-RF were signifi-cantly higher (P � 0.0001) in patients with active RAthan in patients with inactive RA. Eighty-eight percentof the patients with active RA and 47% of the patientswith inactive RA were positive for IgM-RF (i.e., IgM-RFlevels were above the normal value of 12.5 IU/ml). No

Figure 2. Correlation between levels of IgM rheumatoid factor (RF)and C1q–C4 levels (A), the Disease Activity Score in 28 joints (DAS)and C1q–C4 levels (B), and the DAS and IgM-RF levels (C) in allrheumatoid arthritis patients (n � 84). Associations were determinedby Spearman’s rank correlation coefficient. au � arbitrary units.

1146 WOUTERS ET AL

significant correlation was found between positivity forIgM-RF and C4b/c levels (r � 0.071, P � 0.52). How-ever, a significant correlation between C1q–C4 levelsand the presence of IgM-RF was observed (r � 0.449, P� 0.0001) suggesting that the presence of IgM-RF isinvolved in classical complement pathway activation(Figure 2A).

Correlation of complement activation with dis-ease activity. C1q–C4 levels were higher in the patientswith active RA than in the patients with inactive disease,suggesting that classical complement pathway activationis related to disease activity. We therefore analyzed thecorrelation between C1q–C4 complexes and the DAS28,which is an index of disease activity (27). A significantcorrelation was found between C1q–C4 levels and theDAS28, when using Spearman’s rank correlation coeffi-cient (r � 0.398, P � 0.0002) (Figure 2B). Moreover,C4b/c levels correlated significantly with the DAS28 (r �0.361, P � 0.0008) (results not shown). As expected, theIgM-RF levels correlated significantly with the DAS28(r � 0.487, P � 0.0001) (Figure 2C).

C1q–C4 in synovial fluid of RA patients. Weassessed levels of the C1q–C4 complex in the synovialfluid and corresponding EDTA plasma of an additional4 RA patients. In 3 of the 4 patients, C1q–C4 levels insynovial fluid were much higher than the levels inplasma; in 1 patient (patient 4 in Table 2), we foundequal C1q–C4 concentrations in both plasma and syno-vial fluid. In general, the total protein concentration wassomewhat lower in synovial fluid than in plasma fromthese patients. Consistent with these observations, C1qlevels were lower in the synovial fluid of all 4 patients,and especially patient 4, whose C1q concentration insynovial fluid was very low. Therefore, after correctionfor the C1q concentration, all 4 patients had significantlyhigher C1q–C4 complex levels in synovial fluid than inplasma (Table 2 and Figure 3).

Course of C1q–C4 levels during anti-TNF treat-ment. Ten RA patients receiving anti-TNF therapy wereincluded in this study. Prior to each infusion, the patientswere clinically examined, and EDTA plasma was col-lected. At 22 and 52 weeks after the start of treatment,the patients showed clinical improvement comparedwith baseline, as reflected by a significantly decreasedDAS28 (P � 0.05) (Figure 4A). In parallel with the DASchanges, C1q–C4 plasma levels decreased after anti-TNF treatment (Figure 4B). Although this decrease wasnot statistically significant, we observed an obvious trendin the decreasing levels of C1q–C4. Furthermore, thedifference in C1q–C4 levels at 52 weeks after anti-TNFtreatment, as compared with baseline, was borderlinesignificant (P � 0.08).

DISCUSSION

Several studies (1,20,32), including the presentone, have demonstrated a relationship between diseaseactivity and complement activation in RA. Nevertheless,among the known biomarkers, plasma levels of comple-

Table 2. Complement parameters in plasma and synovial fluid (SF) samples from 4 patients withrheumatoid arthritis

Patient 1 Patient 2 Patient 3 Patient 4

Plasma SF Plasma SF Plasma SF Plasma SF

C1q–C4, AU* 4.5 12.6 3.2 13.8 2.1 17.3 2.9 3.0C1q, �g/ml 163 102 207 125 181 119 194 29Total protein, mg/ml 74 60 82 67 80 69 78 74Adjusted C1q–C4, AU† 2.8 12.3 1.5 11.0 1.2 14.6 1.5 10.4

* AU � arbitrary units.† Values are the ratio of C1q–C4 levels to C1q levels.

Figure 3. C1q–C4 complex levels in paired plasma and synovial fluid(SF) samples from 4 rheumatoid arthritis patients. Values are the ratioof C1q–C4 levels to C1q levels. Bold lines indicate the median value.� � P � 0.03.

CLASSICAL COMPLEMENT PATHWAY ACTIVATION IN RA 1147

ment activation products are rarely used as an inflam-mation marker in RA, because analysis of complementactivation in patients is hampered by in vitro artefacts.Recently, we described a new marker for complementactivation, i.e., C1q–C4 complexes, which are remark-ably stable in EDTA plasma. In the present study, wefound that C1q–C4 levels were elevated in RA patientsas compared with the levels in healthy controls, and that

these levels were higher in patients with active RA thanin patients with RA in clinical remission.

Increased complement activation in RA patientshas been described before (1,2,20,32) and is supportedin the present study by means of a novel technique toassess classical complement pathway activation. SinceC1q–C4 complexes are a specific parameter of classicalpathway activation, increased plasma levels of thesecomplexes in RA indicate involvement of the classicalpathway of complement in RA. The molecular size ofC1q–C4 complexes depends on whether the attached C4molecule is completely cleaved into the C4d fragment.Therefore, as long as their precise molecular composi-tion is not clear, calculation of the molar concentrationof C1q–C4 complexes will be difficult, if not impossible.For this reason, levels of C1q–C4 complexes were ex-pressed in arbitrary units in the present study, with 100AU being the amount of complexes in a purified C1qsample. Currently, we are investigating the compositionof C1q–C4 complexes in more detail, to improve quan-tification.

Median plasma levels of the measured comple-ment activation products (C4b/c and C1q–C4) wereelevated in patients with active RA compared with thosewith inactive RA. Although the median values weresignificantly different, there was overlap between pa-tients with active RA and patients with RA in clinicalremission. Some patients whose disease was in remissionhad levels of complement activation products that werein the range of those in patients with active disease. Thisindicates that at least a subgroup of patients with diseasein clinical remission have ongoing inflammation. At themoment we can only speculate whether the clinicalcourse in these patients differs from that in patients inremission who have low levels of complement activation.

C1q–C4 plasma levels were higher in active RAthan in inactive RA, suggesting that classical comple-ment pathway activation is correlated with disease activ-ity. Supporting this assumption, C1q–C4 levels werefound to correlate significantly with parameters of dis-ease activity (the DAS28). We also measured C1q–C4plasma levels in serial samples from RA patients whowere receiving infliximab (anti-TNF therapy). In thesepatients, plasma levels of C1q–C4 decreased from base-line to week 52, in parallel with the DAS (Figure 4),further supporting the notion that these levels reflectdisease activity in RA.

Both CRP and circulating immune complexes arepotential activators of the classical complement pathway(21,22,33), and increased levels of both have beendescribed previously in RA. CRP-mediated complement

Figure 4. Disease Activity Score in 28 joints (DAS28) (A) andC1q–C4 plasma levels (B) in anti–tumor necrosis factor–treated pa-tients with rheumatoid arthritis (n � 10) at baseline, 22 weeks, and 52weeks after start of therapy. Bars show the median, boxes show the25th and 75th percentiles, and bars outside the boxes show the 10thand 90th percentiles. Statistical differences between the different timepoints were compared using Wilcoxon’s matched pairs test. TheDAS28 at 22 and 52 weeks after start of treatment was signficantlydecreased compared with baseline. Levels of C1q–C4 complexes (inarbitrary units [au]) were not significantly decreased, although thedifference between 52 weeks and baseline reached borderline signifi-cance.

1148 WOUTERS ET AL

activation may explain at least part of the classicalpathway activation in RA (20), but some of the activityis independent of the effects of CRP and may resultfrom interaction of complement with immune com-plexes. Circulating immune complexes in RA are pooractivators of complement (19), but this does not rule outa role for immune complexes in the observed activationof the classical pathway. It has been described thatIgM-RF–containing immune complexes are able to ac-tivate complement (17,18,34–36), particularly in RApatients with vasculitis (37). In the present study, asignificant correlation was found between IgM-RF andC1q–C4 complex levels, which supports the role ofimmune complexes in the observed classical comple-ment pathway activation.

In general, levels of complement activation prod-ucts are higher in synovial fluid than in plasma, and ithas been suggested that excess production from thejoints, or spillover, contributes to plasma levels (8). Wemeasured C1q–C4 complexes in synovial fluid and cor-responding EDTA plasma samples from 4 RA patients.Following correction for the C1q concentration, wefound significantly higher levels of C1q–C4 complexes insynovial fluid compared with plasma. Taking into ac-count the molecular size of the C1q–C4 complexes andthe ratio between synovial fluid concentration andplasma concentration, it is most likely that C1q–C4complexes are produced locally in the joint, and that theplasma levels measured in RA are the result of spillover.This would explain the higher C1q–C4 plasma levels inpatients with active disease, since complement activationin the joint is higher in these patients (5,6,38).

Studies on complement activation in human dis-ease are easily blurred because of in vitro activation ofthe system during processing of samples. Indeed, amajor disadvantage of measuring C4b/c is the highsensitivity for in vitro artefacts, due to temperaturechanges or inaccurate handling of samples. Spontaneoushydrolysis of the intramolecular thio–ester bond maylead to artificially high levels of C4b/c, since the neo-epitope recognized by mAb anti–C4-1 is dependent onthe integrity of this bond. In the present study, weobserved a �10-fold increase in C4b/c after 5 cycles offreezing and thawing of RA samples. In contrast, levelsof C1q–C4 complexes did not change upon freezing andthawing, demonstrating that C1q–C4 complexes arehighly stable in plasma, as has been observed before inplasma samples from normal donors (24). In our opin-ion, the fact that C1q–C4 complexes are hardly, if at all,generated in vitro in EDTA-containing plasma samplesis an important advantage over conventional methods to

assess complement activation. In addition, the assay forC1q–C4 complexes was not influenced by RFs. Becausein vitro artefacts play a much lesser role in this newassay, and because C1q–C4 complexes specifically re-flect classical complement pathway activation, we con-clude that measurement of these complexes is of extravalue in studying complement activation in general, andin RA in particular.

Thus, we have shown that significantly elevatedplasma levels of C1q–C4 complexes correlate with dis-ease activity in RA patients. This new complementparameter is remarkably stable in EDTA plasma sam-ples and is hardly, if at all, influenced by in vitroartefacts. We therefore suggest that C1q–C4 complexesmay be helpful for use as a biomarker in assessinginflammatory status and disease activity in RA patients.

REFERENCES

1. Makinde VA, Senaldi G, Jawad AS, Berry H, Vergani D. Reflec-tion of disease activity in rheumatoid arthritis by indices ofactivation of the classical complement pathway. Ann Rheum Dis1989;48:302–6.

2. Petersen NE, Elmgreen J, Teisner B, Svehag SE. Activation ofclassical pathway complement in chronic inflammation: elevatedlevels of circulating C3d and C4d split products in rheumatoidarthritis and Crohn’s disease. Acta Med Scand 1988;223:557–60.

3. Hietala MA, Nandakumar KS, Persson L, Fahlen S, Holmdahl R,Pekna M. Complement activation by both classical and alternativepathways is critical for the effector phase of arthritis. Eur J Im-munol 2004;34:1208–16.

4. Brodeur JP, Ruddy S, Schwartz LB, Moxley G. Synovial fluidlevels of complement sC5b-9 and fragment Bb are elevated inpatients with rheumatoid arthritis. Arthritis Rheum 1991;34:1531–7.

5. Mollnes TE, Paus A. Complement activation in synovial fluid andtissue from patients with juvenile rheumatoid arthritis. ArthritisRheum 1986;29:1359–64.

6. Moxley G, Ruddy S. Elevated C3 anaphylatoxin levels in synovialfluids from patients with rheumatoid arthritis. Arthritis Rheum1985;28:1089–95.

7. Inman RD, Harpel PC. C1 inactivator-C1s complexes in inflam-matory joint disease. Clin Exp Immunol 1983;53:521–8.

8. Swaak AJ, van Rooyen A, Planten O, Han H, Hattink O, Hack E.An analysis of the levels of complement components in thesynovial fluid in rheumatic diseases. Clin Rheumatol 1987;6:350–7.

9. Shingu M, Watanabe Y, Tomooka K, Yoshioka K, Ohtsuka E,Nobunaga M. Complement degradation products in rheumatoidarthritis synovial fluid. Br J Rheumatol 1994;33:299–300.

10. Hietala MA, Jonsson IM, Tarkowski A, Kleinau S, Pekna M.Complement deficiency ameliorates collagen-induced arthritis inmice. J Immunol 2002;169:454–9.

11. Wang Y, Rollins SA, Madri JA, Matis LA. Anti-C5 monoclonalantibody therapy prevents collagen-induced arthritis and amelio-rates established disease. Proc Natl Acad Sci U S A 1995;92:8955–9.

12. Zubler RH, Nydegger U, Perrin LH, Fehr K, McCormick J,Lambert PH, et al. Circulating and intra-articular immune com-plexes in patients with rheumatoid arthritis: correlation of 125I-Clq binding activity with clinical and biological features of thedisease. J Clin Invest 1976;57:1308–19.

CLASSICAL COMPLEMENT PATHWAY ACTIVATION IN RA 1149

13. Hay FC, Nineham LJ, Perumal R, Roitt IM. Intra-articular andcirculating immune complexes and antiglobulins (IgG and IgM) inrheumatoid arthritis: correlation with clinical features. AnnRheum Dis 1979;38:1–7.

14. Tanimoto K, Cooper NR, Johnson JS, Vaughan JH. Complementfixation by rheumatoid factor. J Clin Invest 1975;55:437–45.

15. Sato Y, Sato R, Watanabe H, Kogure A, Watanabe K, NishimakiT, et al. Complement activating properties of monoreactive andpolyreactive IgM rheumatoid factors. Ann Rheum Dis 1993;52:795–800.

16. Sato Y, Watanabe H, Kogure A, Miyata M, Watanabe K, Nishi-maki T, et al. Complement-activating properties of IgM rheuma-toid factors reacting with IgG subclasses. Clin Rheumatol 1995;14:425–8.

17. Robbins DL, Fiegal DW Jr, Leek JC, Shapiro R, Wiesner K.Complement activation by 19S IgM rheumatoid factor: relation-ship to disease activity in rheumatoid arthritis. J Rheumatol1986;13:33–8.

18. Sabharwal UK, Vaughan JH, Fong S, Bennett PH, Carson DA,Curd JG. Activation of the classical pathway of complement byrheumatoid factors: assessment by radioimmunoassay for C4.Arthritis Rheum 1982;25:161–7.

19. Hack CE, Eerenberg-Belmer AJ, Lim UG, Haverman J, AalberseRC. Lack of activation of C1, despite circulating immune com-plexes detected by two C1q methods, in patients with rheumatoidarthritis. Arthritis Rheum 1984;27:40–8.

20. Molenaar ET, Voskuyl AE, Familian A, van Mierlo GJ, DijkmansBA, Hack CE. Complement activation in patients with rheumatoidarthritis mediated in part by C-reactive protein. Arthritis Rheum2001;44:997–1002.

21. Volanakis JE, Kaplan MH. Interaction of C-reactive proteincomplexes with the complement system. II. Consumption ofguinea pig complement by CRP complexes: requirement forhuman C1q. J Immunol 1974;113:9–17.

22. Romero IR, Morris C, Rodriguez M, Du Clos TW, Mold C.Inflammatory potential of C-reactive protein complexes comparedto immune complexes. Clin Immunol Immunopathol 1998;87:155–62.

23. Antes U, Heinz HP, Schultz D, Brackertz D, Loos M. C1q-bearingimmune complexes detected by a monoclonal antibody to humanC1q in rheumatoid arthritis sera and synovial fluids. RheumatolInt 1991;10:245–50.

24. Wouters D, Wiessenberg HD, Hart M, Bruins P, Voskuyl A, DahaMR, et al. Complexes between C1q and C3 or C4: novel andspecific markers for classical complement pathway activation.J Immunol Methods 2005;298:35–45.

25. Arnett FC, Edworthy SM, Bloch DA, McShane DJ, Fries JF,Cooper NS, et al. The American Rheumatism Association 1987revised criteria for the classification of rheumatoid arthritis.Arthritis Rheum 1988;31:315–24.

26. Pinals RS, Masi AT, Larsen RA, and the Subcommittee forCriteria of Remission in Rheumatoid Arthritis of the AmericanRheumatism Association Diagnostic and Therapeutic CriteriaCommittee. Preliminary criteria for clinical remission in rheuma-toid arthritis. Arthritis Rheum 1981;24:1308–15.

27. Prevoo ML, van ‘t Hof MA, Kuper HH, van Leeuwen MA, van dePutte LB, van Riel PL. Modified disease activity scores thatinclude twenty-eight–joint counts: development and validation in aprospective longitudinal study of patients with rheumatoid arthri-tis. Arthritis Rheum 1995;38:44–8.

28. Hoekzema R, Martens M, Brouwer MC, Hack CE. The distortivemechanism for the activation of complement component C1supported by studies with a monoclonal antibody against the“arms” of C1q. Mol Immunol 1988;25:485–94.

29. Wolbink GJ, Brouwer MC, Buysmann S, ten Berge IJ, Hack CE.CRP-mediated activation of complement in vivo: assessment bymeasuring circulating complement-C-reactive protein complexes.J Immunol 1996;157:473–9.

30. Wolbink GJ, Bollen J, Baars JW, ten Berge RJ, Swaak AJ,Paardekooper J, et al. Application of a monoclonal antibodyagainst a neoepitope on activated C4 in an ELISA for thequantification of complement activation via the classical pathway.J Immunol Methods 1993;163:67–76.

31. Tenner AJ, Lesavre PH, Cooper NR. Purification and radiolabel-ing of human C1q. J Immunol 1981;127:648–53.

32. Nydegger UE, Zubler RH, Gabay R, Joliat G, Karagevrekis CH,Lambert PH, et al. Circulating complement breakdown productsin patients with rheumatoid arthritis: correlation between plasmaC3d, circulating immune complexes, and clinical activity. J ClinInvest 1977;59:862–8.

33. Borsos T. Immune complex mediated activation of the classicalcomplement pathway. Behring Inst Mitt 1989;84:93–101.

34. Westedt ML, Daha MR, de Vries E, Valentijn RM, Cats A. IgAcontaining immune complexes in rheumatoid vasculitis and inactive rheumatoid disease. J Rheumatol 1985;12:449–55.

35. Pope RM, Yoshinoya S, McDuffy SJ. Detection of immunecomplexes and their relationship to rheumatoid factor in a varietyof autoimmune disorders. Clin Exp Immunol 1981;46:259–67.

36. Scott DG, Bacon PA, Allen C, Elson CJ, Wallington T. IgGrheumatoid factor, complement and immune complexes in rheu-matoid synovitis and vasculitis: comparative and serial studiesduring cytotoxic therapy. Clin Exp Immunol 1981;43:54–63.

37. Elson CJ, Scott DG, Blake DR, Bacon PA, Holt PD. Complement-activating rheumatoid-factor-containing complexes in patientswith rheumatoid vasculitis. Ann Rheum Dis 1983;42:147–50.

38. Abbink JJ, Kamp AM, Nuijens JH, Erenberg AJ, Swaak AJ, HackCE. Relative contribution of contact and complement activation toinflammatory reactions in arthritic joints. Ann Rheum Dis 1992;51:1123–8.

1150 WOUTERS ET AL