PATIENTS

Patients with RA according to the American College of Rheumatology criteria21 and with active disease, defined by a disease activity score (DAS)22 >3.2, who were enrolled in a double blind, multicentre clinical trial with anti-TNFα monoclonal antibody (D2E7) at our centre, were studied. Disease modifying antirheumatic drug treatment was withheld and non-steroidal anti-inflammatory drugs or low dose oral steroids (<10 mg/day), or both, were kept constant in a three week wash out period before and during the study. Patients were randomly assigned to receive 0.5, 1, 3, 5, or 10 mg/kg of D2E7 and each dose group included two patients treated with placebo, who received active drug at six weeks if they still fulfilled the entry criteria.11 ,13

D2E7 or placebo was given as a slow intravenous infusion over three to five minutes. The preparation consists of a non-pyrogenic solution of 25 mg/ml D2E7 mAb in 1.2% mannitol, 0.12% citric acid, 0.02% sodium citrate. The placebo consists of the same ingredients, except that D2E7 is excluded.

Results of the multicentre study including 120 patients in three centres show that clinical response is already apparent within 24 hours and maximal between one and two weeks after D2E7 administration.11 ,13 The clinical effect is maximal at a dose of 1 mg/kg and shows a plateau in the dose-response curve thereafter.11 ,13 Therefore, for the aims of this study, we focused on the short term effects seen during the first two weeks after infusion of D2E7/placebo and subdivided the patients into three dose groups consisting of placebo (n=11), 0.5 mg/kg (n=8), and 1–10 mg/kg D2E7 (n=31).

CONCENTRATIONS AND GENE EXPRESSION OF CYTOKINES IN PERIPHERAL BLOOD

Blood samples were drawn at 8–9 am on days 0, 1, and 14 after infusion in endotoxin-free Vacutainer tubes with 15% EDTA-K₃ (Becton and Dickinson, Rutherford, NJ). Samples were directly centrifuged (2250 g, 10 minutes and 15 000 g, 5 minutes) to assess circulating cytokine concentrations in platelet-free plasma. Aliquots were stored at −20°C until assay.

IL1β was measured using a high sensitivity enzyme linked immunosorbent assay (ELISA; sensitivity 0.1 pg/ml; Quantikine HS, R&D, Minneapolis, USA) according to the manufacturer's instructions. Interleukin 1 receptor antagonist (IL1ra) was measured by radioimmunoassay (sensitivity 40 pg/ml; interassay and intra-assay variation <10%, respectively). IL6 and both soluble TNF receptors (sTNFR) were measured by ELISA (sensitivity 8 pg/ml and 0.1 ng/ml respectively) as previously described.23 ,24 Total TNFα was measured by ELISA (sensitivity 10 pg/ml). The latter had been validated in spiking experiments showing adequate TNFα recovery and no hampering by therapeutic concentrations of D2E7 (range 25–250 mg/l).11

Aliquots of 500 μl blood for RNA isolation were mixed with guanidinium isothiocyanate at a ratio of 1:1 and stored at −70°C. Isolation of whole blood mRNA and reverse transcriptase polymerase chain reaction (RT-PCR) analysis were performed as previously described by Netea et al.25 Briefly, aliquots of 0.5 μg total RNA were diluted in 20 μl RT buffer (50 mM Tris-HCl pH 8.3, 75 mM KCl, 3 mM MgCl₂) containing 10 mM dithiothreitol, 5 μM random hexamers, 250 μM dNTPs, 20 U RNAsin, and 200 M MLV RT. RT reaction was performed for 10 minutes at 20°C, 45 minutes at 42°C, and 10 minutes at 95°C in a Mastercycler 5330 (Eppendorf, Hamburg, Germany). PCR reaction mixtures consisted of 3 μl cDNA in 50 μl PCR buffer (20 mM Tris-HCl pH 8.4, 50 mM KCl, 1.5 mM MgCl₂, 0.001% gelatin) containing 100 μM dNTPs, 1.25 U Taq polymerase, and 0.3 μM of each primer.25 A total of 29 PCR cycles (denaturation 30 seconds at 92°C, annealing 30 seconds at 55°C, and extension 90 seconds at 72°C) were performed for IL1β mRNA and TNFα mRNA, and 24 cycles were performed for β₂ microglobulin. PCR products underwent electrophoresis on 2% agarose gel and were stained with ethidium bromide. Gels were scanned on a densitometer and analysed using the Molecular Analyst software. Results are expressed as a ratio of IL1β or TNFα mRNA to the housekeeping gene.

SYNOVIAL BIOPSIES AND IMMUNOHISTOCHEMISTRY

Percutaneous synovial biopsy specimens of the knee were obtained with a Parker Pearson needle at baseline and 14 days after the first dose of D2E7/placebo in all consenting patients. Biopsy was preceded by knee joint examination for tenderness (0=none, 1=response on questioning, 2=spontaneous response, 3=withdrawal) and swelling (0=none, 1=thickening without loss of bony contours; bulge sign, 2=loss of bony contours; palpable but not tightly distending effusions, 3=bulging synovial proliferation; tightly distending effusions). Macroscopic knee joint scores were calculated by adding up the tenderness and swelling scores. An average of 30 biopsy specimens was obtained at each occasion and immediately fixed in 10% formalin and embedded in paraffin. Serial 7 μM microtome sections were mounted on superfrost slides and either stained with haematoxylin and eosin (H&E) or used for histochemical staining as previously described.26 ,27 For each marker, all sections were stained in the same run to minimise interassay variations. Slides were incubated with anti-TNFα (IgG1, Monosan, Uden, The Netherlands) or anti-IL1β mAb (IgM, 12E9, Oncogene Science, Manhasset, NY, USA). This primary step was followed by incubation with normal horse serum and with biotinylated horse antimurine IgG. Slices were stained with avidin peroxidase (Elite kit, Vector, Burlingame, CA), developed with diaminobenzidine and counterstained with haematoxylin for three minutes. Controls consisted of (a) irrelevant primary isotype-specific IgG1 and IgM antibodies obtained from normal horse serum, and (b) omission of the secondary antibodies.

All areas of each section were randomly analysed by two “blinded” observers using semiquantitative five point scales. H&E sections were scored for the presence of lymphocytes, plasma cells, and polymorphonuclear cells with a scale ranging from 0 (no or minimal infiltration) to 4 (abundant inflammatory infiltrate), the lining hyperplasia was also scored (0=1–2 layers, 1=3–4 layers, 2=5–6 layers, and 3=more than 6 layers). An “inflammation score” was calculated by adding these four components (range 0–15), as previously described by Tak et al. 14 The staining for IL1 and TNF in lining, sublining, and vessels was also scored semiquantitatively on a five point scale (0 to 4). Differences in readings of one point were taken as the average, differences exceeding one point were resolved by mutual agreement.

STATISTICAL ANALYSIS

Analysis was performed using the SAS statistical package (SAS 6.04 PC version). Data are expressed as means (SD) or as median (25th (p25)–75th (p75) centile) if appropriate. Within-group comparisons were analysed by paired Student'st or Mann Whitney rank sum test. Baseline comparisons between groups were performed using one way analysis of variance or Kruskal-Wallis tests. Correlations are expressed using Spearman's rank correlation coefficient.

CLINICAL RESULTS, ACUTE PHASE REACTION, AND LEUCOCYTE SUBSETS

Table 1 summarises the baseline characteristics of the patients included in the present study. Patients were subdivided into groups receiving placebo, 0.5 mg/kg D2E7, and 1–10 mg/kg D2E7 because, as previously mentioned, the dose-response curve in the multicentre study (n=120) showed a plateau at doses of 1 mg/kg D2E7.11 ,13

Administration of D2E7, but not placebo, rapidly reduced disease activity as measured by the DAS, a composite disease activity score22 (fig 1A), and its individual components, including swollen and tender joint counts, patient wellbeing (data not shown), and erythrocyte sedimentation rate (ESR; fig 1B). In the group treated with 1–10 mg/kg D2E7, the decrease in DAS was already significant within 24 hours after the infusion (mean (SD) 4.87 (0.79)v 5.25 (0.95); p<0.002) and reached a nadir at week 2 (4.06 (0.85), p<0.0001). The 0.5 mg/kg group showed a drop in DAS values significant at week 2 (p<0.01). At this time point, 20/31 (65%) and 3/8 (38%) of the patients treated with 1–10 and 0.5 mg/kg D2E7, respectively, fulfilled the EULAR criteria for clinical response28 in contrast with only 1/11 (9%) patients receiving placebo. Five (16%) patients in the 1–10 mg/kg group showed clinical response already 24 hours after infusion.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

(A) Disease activity score (DAS), (B) erythrocyte sedimentation rate (ESR), and (C) C reactive protein (CRP) in patients receiving a single dose of placebo or 0.5 mg/kg or 1–10 mg/kg anti-tumour necrosis factor monoclonal antibody. Boxes represent the 25th, 50th, and 75th centiles; vertical lines indicate the 5th and 95th centiles. Measurements on days 0, 1, and 14 after infusion (x axis). Comparisons versus baseline shown as *p<0.05, **p<0.005, and ***p<0.0005.

Baseline acute phase reaction parameters, such as the ESR, C reactive protein (CRP), and platelet counts tended to be higher in the 1–10 mg/kg group (table 1) and were also significantly decreased two weeks after administration of anti-TNF mAb but unchanged or increased in the placebo group (figs 1B and C). The 0.5 mg/kg group, which included fewer patients (n=8), also showed a clear decrease in ESR and CRP after two weeks.

Baseline white blood cell and leucocyte subset counts were similar in all groups (table 1). Total white blood cell counts did not change during the study (data not shown), but a rise in day 1 lymphocyte counts was seen in all groups, being more marked in patients receiving 1–10 mg/kg D2E7 (median 14.5% before infusionv 22% at day 1; p<0.01). In this group only, the lymphocyte counts were still higher than baseline at day 14. Conversely, there was a drop in day 1 polymorphonuclear cell counts in all groups, which only reached significance in the 1–10 mg/kg group at day 1 (median 77% at baseline v 66.5% at day 1; p=0.02).

SYSTEMIC CYTOKINE MEASUREMENTS

Baseline IL1β and TNFα levels were below or around the detection limit of the radioimmunoassays currently used in our laboratory and similar to those found in healthy subjects (data not shown). A high sensitivity ELISA (Quantikine HS, R&D, Minneapolis, USA) showed that IL1β protein levels were low and did not change significantly after the first dose of D2E7 or placebo (median levels at baseline, day 1, and day 14 were 0.55, 0.4, and 0.5 pg/ml after 1–10 mg/kg D2E7 and remained at 0.6 pg/ml before and after placebo administration).

Mean baseline TNFα levels (free and bound) measured by ELISA did not change after infusion of placebo but increased after D2E7 administration (41 pg/ml at baseline v 76 and 63 pg/ml at days 1 (p<0.0001) and 14 (p<0.01), respectively), which may reflect the formation of circulating TNF-anti-TNF complexes.

To complement the data at the protein level, systemic IL1β and TNFα mRNA levels in whole blood were measured in the placebo and 1–10 mg/kg D2E7 group by RT-PCR. Baseline IL1β/β₂ microglobulin ratios in patients with RA were raised compared with those in healthy controls (median (range) 1.45 (0.57–2.95) v0.35 (<0.10–0.69)). As shown in fig 2, IL1β mRNA expression decreased within 24 hours after D2E7 administration (p=0.002) and remained lower than baseline at day 14 (p=0.007), whereas no significant changes occurred after placebo. Systemic TNFα mRNA levels were lower than those of IL1β, overlapped largely with those found in healthy controls (median (range) 0.13 (<0.10–0.85) in RAv <0.10 (<0.10–0.30) in controls), and remained unchanged during the study. Systemic levels of IL1β and TNFα mRNA were not correlated (r _s (95%CI)=0.13 (−0.18 to 0.43)).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Circulating concentrations of (A) interleukin 1 receptor antagonist (IL1ra), (B) IL6, (C) 55th centile (p55), and (D) 75th centile (p75) soluble tumour necrosis factor receptors (sTNFR). (E) IL1β and (F) TNFα mRNA in whole blood normalised for the presence of β₂ microglobulin (β₂M). Boxes represent 25th, 50th, and 75th centiles; vertical lines indicate the 5th and 95th centiles. Measurements on days 0, 1, and 14 after infusion (x axis). Comparisons versus baseline shown as *p<0.5, **p<0.005, and ***p<0.0005.

The concentrations of IL1ra and IL6 decreased sharply after D2E7 administration (p<0.0001 and <0.005 at days 1 and 14, respectively, in the 1–10 mg/kg group). Both sTNFR types decreased also after treatment. In patients receiving 1–10 mg/kg D2E7, the p75 levels dropped from median baseline values of 5 ng/ml to 4.3 (p<0.005) and 3.7 ng/ml (p=0.0001) after 1 and 14 days respectively and the p55 sTNFR levels fell from median baseline values of 3.1 ng/ml to 2.5 ng/ml at day 14 (p<0.002) (fig 2).

Positive correlations between acute phase reactants and the endogenous antagonists of IL1 and TNF were found: CRP measurements correlated with the levels of IL1ra (r _s=0.47; p<0.001), p55 sTNFR (r _s=0.66; p<0.0001) and IL6 (r _s=0.65; p<0.0001), whereas ESR levels correlated with p75 sTNFR (r _s=0.51; p<0.002). In contrast, neither TNFα nor IL1β, at protein or mRNA level, were related to acute phase parameters.

IMMUNOHISTOLOGICAL FINDINGS

A total of 25 patients underwent synovial biopsies at baseline and 14 days after infusion of placebo (n=8) or 1–10 mg/kg D2E7 (n=17). At this time in the treated group, the DAS had dropped from 5.0 (1.0) to 4.0 (1.1) (p<0.0005), and 11 patients fulfilled the EULAR criteria for clinical response.28 Conversely, DAS scores remained unchanged or increased (4.8 (1.3) at baselinev 5.1 (1.6) after 14 days) in the placebo group, where only one patient fulfilled the EULAR response criteria28 (fig 3A). The macroscopic joint scores for swelling and pain at the biopsied joint were unchanged two weeks after placebo but decreased in treated patients (p=0.02; fig 3B). The microscopic scores for inflammation14 showed a more scattered pattern and no significant changes in either group (fig 3C). Baseline microscopic inflammation scores were weakly correlated with the macroscopic scores for pain and swelling (r_s =0.3; p<0.05).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

(A) Disease activity score (DAS), (B) pain and swelling scores, and (C) histological scores for inflammation in 25 patients undergoing serial biopsies. Time in weeks and placebo (n=8) or anti-tumour necrosis factor (anti-TNF) treatment (n=17) shown in the x axis. Scores for more than one patient (× exact number) shown as bold lines.  

TNFα and IL1β staining were seen in all layers of most biopsy specimens. The scores for both cytokines in the lining, sublining, and especially in vessels were mutually correlated (r_s =0.43; p=0.002 for the lining and r_s =0.47 and 0.58; p<0.0005 for the sublining and vessels, respectively). Nevertheless, some differences were found in the staining patterns of both cytokines: TNF showed a more dense and intense staining in sublining layers, lymphocyte aggregates, and vessels than IL1, which tended to be more diffuse (fig 4). The microscopic scores for inflammation were positively correlated with the scores for TNF and IL1 in the sublining (r_s =0.65 and 0.54 respectively; p<0.0001) and to a lesser extent with the TNF staining in the lining (r_s =0.38; p=0.005).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Immunohistochemical staining in the same patient at baseline ((A) tumour necrosis factor α (TNFα), (B) interleukin 1β (IL1β)) and 14 days after ((C) TNFα, (D) IL1β) the first dose of anti-TNFα. Note the unchanged expression of cytokines in the synovial lining (s, arrows), sublining, and blood vessels (bv). No IL1β or TNFα staining was seen in controls with (E) an irrelevant primary antibody and (F) when the secondary antibody was omitted. Original magnification ×200.

Figure 5 shows the individual synovial scores for IL1 and TNF after administration of anti-TNF mAb or placebo. As shown, there was a marked interindividual and intra-individual variation in the synovial staining for these cytokines. Decreases greater than one point in the scores for IL1β or TNFα were rare among placebo treated patients. On the other hand, both increased and decreased scores for IL1β and TNFα were detected in the lining and sublining of patients treated with anti-TNF. With the exception of a trend towards a decrease in the IL1β staining in vessels among the treated patients (p<0.06), no significant changes were seen.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

Semiquantitative scores for interleukin 1β (IL1β; left columns) and tumour necrosis factor α (TNFα; right columns). Staining in (A) lining, (B) sublining, and (C) vessels in biopsy specimens taken at baseline and at day 14 after administration of placebo (n=8) or anti-TNF treatment (n=17). Scores for more than one patient (× exact number) shown as bold lines.

Additionally, analysis was performed according to the clinical response achieved at day 14 after infusion. At this time point, a total of 12 patients (all but one in the group treated with anti-TNF) were responders according to the EULAR criteria28 (DAS decreased from 4.9 (1.1) to 3.4 (0.6); p<0.0001), whereas 13 patients, including seven who had received placebo, were non-responders (DAS 5.1 (1.2) v 5.2 (1.2)). Macroscopic knee joint score for swelling and pain dropped from 2 (2.3) to 0.8 (1.1) (p<0.05) among responders but remained unchanged (2.2 (2.0)v 2.2 (2.0)) in the non-responders.

In contrast with the clinical improvement observed in responders, the microscopic inflammation scores (H&E) and the staining for IL1β or TNFα in lining, sublining, and vessels were not significantly changed from baseline compared with day 14 (data not shown). Subgroup analysis taking into account the individual dose of D2E7 given gave the same results.