ARTHRITIS & RHEUMATISMVol. 52, No. 12, December 2005, pp 3880–3889DOI 10.1002/art.21488© 2005, American College of Rheumatology

Intraarticular Glucocorticoid Treatment Reduces Inflammationin Synovial Cell Infiltrations More Efficiently

Than in Synovial Blood Vessels

Erik af Klint, Cecilia Grundtman, Marianne Engstrom, Anca Irinel Catrina,Dimitrios Makrygiannakis, Lars Klareskog, Ulf Andersson, and Ann-Kristin Ulfgren

Objective. To investigate whether intraarticular(IA) glucocorticoid (GC) therapy diminishes synovialcell infiltration, vascularity, expression of proinflamma-tory cytokines, and adhesion molecule levels in patientswith chronic arthritides.

Methods. Thirty-one patients with chronic ar-thritides received a single IA injection of triamcinolonehexacetonide to treat active large-joint inflammation.Synovial biopsy specimens were obtained with arthro-scopic guidance before and 9–15 days after injection.The presence of T lymphocytes, macrophages, intercell-ular adhesion molecule 1 (ICAM-1), vascular endothe-lial growth factor (VEGF), the pan-endothelial markerCD31, and the proinflammatory cytokines interleukin-1� (IL-1�), IL-1�, tumor necrosis factor (TNF), andhigh mobility group box chromosomal protein 1(HMGB-1) was studied by immunohistochemistry andreal-time reverse transcriptase–polymerase chain reac-tion.

Results. IA GC treatment resulted in good clinicalresponse in 29 of 31 joints. After therapeutic interven-tion, the number of synovial T lymphocytes declined,whereas the number of macrophages remained un-changed. Overall synovial protein expression of TNF,IL-1�, extranuclear HMGB-1, VEGF, and ICAM-1 wasreduced at followup tissue sampling, while no signifi-cant effects were observed regarding vascularity. Incontrast, expression of IL-1�, VEGF, and cytoplasmicHMGB-1 protein in vascular endothelial cells was notaffected. GC therapy down-regulated levels of messen-ger RNA encoding IL-1� and IL-1�, but not TNF orHMGB-1.

Conclusion. Synovial cell infiltration and pro-inflammatory cytokine expression were affected in amultifaceted manner by IA GC treatment. Markedreduction of synovial T lymphocytes, TNF, IL-1�, ex-tranuclear HMGB-1, ICAM-1, and VEGF occurred inassociation with beneficial clinical effects. Unexpect-edly, macrophage infiltration and proinflammatory en-dothelial cytokine expression remained unchanged.These findings may reflect mechanisms controlling thetransiency of clinical improvement frequently observedafter IA GC injection.

The discovery more than 50 years ago of glu-cocorticoids (GCs) and their potent therapeutic effectsin rheumatoid arthritis (RA) represents a major break-through for ameliorating inflammatory diseases. How-ever, dose-related toxicity mediated by GCs constrainstheir systemic therapeutic usefulness and encourageslocal administration whenever feasible. Hence, intraar-ticular (IA) GC injections are widely used in the man-agement of arthritic conditions. Their clinical efficacy isunpredictable, however, and varies from long-term re-mission to minimal benefit. The reasons for the diverse

Supported by The Swedish Rheumatism Association, KingGustaf V’s 80-year-Foundation, the Borje Dahlin’s Foundation, theFreemason Lodge in Stockholm, the Swedish Medical ResearchCouncil, and the Åke Wiberg Foundation.

Erik af Klint, MD, Cecilia Grundtman, MSc, MarianneEngstrom, BSc, Anca Irinel Catrina, MD, Dimitrios Makrygiannakis,MD, Lars Klareskog, MD, Ulf Andersson, MD (current address:Astrid Lindgren Children’s Hospital, Stockholm, Sweden), Ann-Kristin Ulfgren, PhD: Karolinska Institutet, Stockholm, Sweden.

Dr. af Klint and Ms Grundtman contributed equally to thiswork.

Dr. Klareskog is a consultant for and/or has received researchgrants from Wyeth, Schering-Plough, Abbott, Astra-Zeneca, BristolMyers-Squibb, and Centocor. Dr. Andersson has received researchgrants (more than $10,000 per year) from Critical Therapeutics andowns stock in Critical Therapeutics.

Address correspondence and reprint requests to Ulf Anders-son, MD, Department of Woman and Child Health, Astrid LindgrenChildren’s Hospital, Q2:02, S-17176 Stockholm, Sweden. E-mail:ulf@mbox313.swipnet.se.

Submitted for publication October 18, 2004; accepted inrevised form August 31, 2005.

3880

clinical responses have not been fully elucidated, partlybecause of a paucity of published data on in vivo effectsof IA GCs on chronic human synovitis. In vitro studiesindicate that important antiinflammatory effects of GCsare mediated by suppression of transcription and releaseof multiple cytokines and chemokines, by inhibition ofthe influx of effector cells to sites of inflammation, andby redirection of T lymphocyte responses from a pro- toan antiinflammatory influence (for review, see refs. 1and 2).

In the present study we utilized sequential arthro-scopic-guided biopsy and ex vivo techniques to examinesynovial tissue changes in chronic arthritis reflecting cellmigration and proinflammatory cellular functions afterIA GC injection. A previous study demonstrated that IAGC therapy reduces the synovial membrane volume ininflamed joints (3). We analyzed whether this therapeu-tic effect could be linked to any decrease in T lympho-cytes and/or macrophages, expression of intercellularadhesion molecule 1 (ICAM-1) in the synovial mem-brane, or the size of the vascular compartment. Arecently reported study demonstrated a marked reduc-tion of synovial macrophage numbers after 2-week sys-temic GC therapy for RA (4).

We further assessed the expression of synovialproinflammatory cytokines including tumor necrosis fac-tor (TNF) and interleukin-1 (IL-1), since these cytokinesare key mediators in chronic synovitis (5). The regula-tion of cytokine production and release is highly com-plex, and we consequently studied therapy-inducedchanges in both cytokine protein and cytokine messen-ger RNA (mRNA) levels. In addition, we investigatedchanges in the expression of high mobility group boxchromosomal protein 1 (HMGB-1). HMGB-1 repre-sents a proinflammatory molecule that is abundantly andaberrantly expressed in the synovial tissue in RA andother chronic arthritides (for review, see refs. 6 and 7).The intranuclear architectural protein HMGB-1 hasrecently been identified as a potent cytokine whenpresent extracellularly (8). The molecule can be activelysecreted by stimulated macrophages or monocytes in aprocess that requires HMGB-1 translocation from thenucleus to secretory lysosomes (9). Furthermore,HMGB-1 has been demonstrated to be a nuclear dangersignal that is passively released by necrotic, as opposedto apoptotic, cells that will induce inflammation (10).Extracellular HMGB-1 acts as a cytokine by signaling viathe receptor for advanced glycation end products andpossibly via members of the Toll-like receptor family(11,12).

PATIENTS AND METHODS

Patients. Thirty-one patients with chronic arthritis (21women and 10 men; mean age 48 years, range 20–83) who hadactive inflammation with effusion in at least 1 large joint (30 of31 studied joints were knee joints), were included in this study.Clinical and demographic characteristics of the patients arepresented in Table 1. Sixteen patients fulfilled the 1987American College of Rheumatology criteria for RA (13) (10patients were seropositive), 7 patients had unclassified oligo-arthritis, 3 patients had unclassified polyarthritis, 2 patientshad spondylarthritis, 2 patients had psoriatic arthritis, and 1patient had juvenile idiopathic arthritis. Disease durationvaried from 6 weeks to 20 years (mean 5.5 years). Theinvestigated joints had exhibited active arthritis for 7 days to 3years. Patients were recruited from the outpatient clinic of therheumatology unit at Karolinska University Hospital, Solna,Sweden. All clinical examinations and sequential arthroscopicsampling were performed by the same physician (EaK). Allpatients had given informed consent, and the study wasapproved by the Karolinska University Hospital ethics com-mittee.

Arthroscopic biopsies. Synovial membrane sampleswere obtained by an arthroscopic technique as previouslydescribed (14) and were taken before and 9–15 days after IAinjection of 40 mg triamcinolone hexacetonide (Lederspan;Wyeth Lederle, Solna, Sweden). Synovial biopsy sites wereselected to be from areas exhibiting maximal signs of inflam-mation and were documented by photography and mapped inclose proximity to the primary biopsy sites to avoid samplingvariation at the second arthroscopy biopsy sampling.

Tissue processing and immunohistochemical studies.The synovial biopsy specimens were snap frozen within 30seconds in liquid isopentane, stored at �70°C, and thenembedded in OCT compound (TissueTek; Sakura Finetek,Zoeterwoude, The Netherlands) when sectioning was per-formed. Cryostat sections from the biopsy samples (6–8 �m,cryostat setting 7 �m) were placed on SuperFrost Plus slides(Menzel-Glaser, Braunschweig, Germany) and air dried for 30minutes. Sections for cytokine staining were initially fixed for20 minutes with freshly prepared 2% (volume/volume) form-aldehyde (Sigma, St. Louis, MO), dissolved in phosphatebuffered saline (PBS; pH 7.4) at 4°C, washed in PBS, and thenleft to air-dry before storage at �70°C. Sections for phenotypiccellular characterization were fixed with acetone, then left toair-dry at room temperature before storage at �70°C. Thestaining procedures have been described in detail previously(14).

Primary antibodies. All primary antibodies used, apartfrom the anti–HMGB-1 and anti–vascular endothelial growthfactor (anti-VEGF) antibodies, were mouse IgG1 monoclonalantibodies. Anti-CD3 antibody (clone SK7) was obtained fromBecton Dickinson (San Jose, CA), anti-CD68 (clone PG-M1)and anti-CD163 (clone Ber-MAC3) from Dakopatts(Glostrup, Denmark), anti-CD54 (anti–ICAM-1) (clone84H10) from Serotec Scandinavia (Oslo, Norway), anti-CD31(human CD31) (clone EN4) from Sanbio Bio-Zac (Uden, TheNetherlands), anti-TNF (clone 2C8) from Biodesign (Saco,ME), anti–IL-1� (clone 1277-89-7) and anti–IL-1� (clone 2D8combined with 1437-96-5) from Immunokontakt (Bioggo,Switzerland), control mouse IgG1 from Dakopatts, anti–

EFFECT OF IA GLUCOCORTICOIDS ON SYNOVIAL CELL INFLAMMATION 3881

HMGB-1 peptide affinity-purified polyclonal rabbit IgG anti-body from PharMingen (San Diego, CA), anti-VEGF poly-clonal rabbit IgG antibody (A-20 [sc-152]) from Santa CruzBiotechnology (Santa Cruz, CA), and control polyclonal rabbitIgG from Dakopatts.

Secondary antibodies. Biotinylated F(ab�)2-fragmented goat-anti-mouse IgG1 was purchased from Caltag(Burlingame, CA). Biotinylated horse anti-mouse IgG andbiotinylated goat anti-rabbit IgG were from Vector (Burlin-game, CA).

Serum analysis. Serum levels of VEGF were assessedby standard sandwich enzyme-linked immunosorbent assay(ELISA) (Quantikine; R&D Systems Europe, Abingdon, UK).Recombinant human VEGF165 was used as a standard in theassay. All serologic analyses were performed in duplicate.Optical densities were determined using a microplate reader(Emax; Molecular Devices, Menlo Park, CA) at 450 nm with aconnected software program (SoftMax version 2.01; SoftMax,Menlo Park, CA).

Quantification of staining results. All stained sectionswere examined with a Polyvar II microscope (Reichert-Jung,Vienna, Austria) equipped with a Leica 300F digital color

video camera that digitized the microscope images to beprocessed in a Quantimet 550S image analyzer (Leica Cam-bridge, Cambridge, UK) linked to a PC computer. The sectionswere assessed on 3 different occasions by 2 independentobservers who were blinded to the identity of the specimens,with concordant results. The area of specific immunostainingwas expressed as a percentage of the total counterstained tissuearea evaluated. Analysis of an entire tissue section typicallyinvolved 22–253 microscopic fields at a magnification of � 250.

RNA extraction and real-time reverse transcriptase–polymerase chain reaction (PCR). Total RNA was extractedfrom snap-frozen synovial biopsy samples from 12 patientsbefore and after IA GC injection, using an RNeasy Minikit(Qiagen, Valencia, CA). Before RNA elution, residualgenomic DNA was digested using RNase-Free DNase I (Qia-gen). The integrity and quality of the total RNA were assessedwith an Agilent 2100 Bioanalyzer (Agilent Technologies,Waldbronn, Germany). Total RNA (18–748 ng/�l) wasreverse-transcribed to complementary DNA (cDNA) usingrandom hexamer (0.1 �g/�l; Gibco Life Technologies, GrandIsland, NY) and SuperScript reverse transcriptase (200 units/�l; Gibco Life Technologies). Amplification was performed

Table 1. Demographic and clinical characteristics of the patients at baseline*

Patient Diagnosis RF Age/sexDuration of

disease, monthsDuration of activearthritis, months Therapy

1 RA � 32/F 3 0.75 NSAID2 RA � 40/F 180 0.75 Pred. 2.5 mg/day, NSAID3 RA � 42/F 3 3 NSAID4 RA � 46/F 3 3 05 RA � 53/M 84 0.75 Pred. 7.5 mg/day, NSAID6 RA � 57/F 72 6 Etan. 50 mg/week, AZA 100 mg/day, NSAID7 RA � 58/F 23 0.75 08 RA � 63/M 3 1 NSAID9 RA � 65/F 36 2 MTX 10 mg/week, NSAID

10 RA � 68/F 12 1.75 NSAID11 RA � 35/F 240 0.5 MTX 15 mg/week, NSAID12 RA � 38/M 36 36 NSAID13 RA � 44/F 60 1.5 MTX 17.5 mg/week14 RA � 60/M 3 2 015 RA � 70/F 60 3 MTX 20 mg/week16 RA � 83/M 18 0.5 MTX 15 mg/week, pred. 10 mg/day17 Oligoarthritis � 21/M 4 4 018 Oligoarthritis � 28/M 48 2.5 HCQ 400 mg/day19 Oligoarthritis � 39/F 168 7 020 Oligoarthritis � 53/M 120 2 021 Oligoarthritis � 56/F 8 3 NSAID22 Oligoarthritis � 61/F 12 6 023 Oligoarthritis � 66/F 1.5 1.5 Pred. 5 mg/day, NSAID24 Polyarthritis � 20/F 8 8 NSAID25 Polyarthritis � 43/F 144 1.5 Pred. 2.5 mg/day, NSAID26 Polyarthritis � 73/F 24 0.5 MTX 12.5 mg/week27 Spondylarthritis � 20/M 6 1.5 NSAID28 Spondylarthritis � 36/F 72 2.5 SSZ 2 gm/day, NSAID29 PsA � 53/F 12 3 MTX 10 mg/week, NSAID30 PsA � 52/F 36 0.25 MTX 20 mg/week, NSAID31 JIA � 33/M 204 0.25 SSZ 3 gm/day, NSAID

* RF � rheumatoid factor; RA � rheumatoid arthritis; NSAID � nonsteroidal antiinflammatory drug; pred. � prednisolone; etan. � etanercept;AZA � azathioprine; MTX � methotrexate; HCQ � hydroxychloroquine; SSZ � sulfasalazine; PsA � psoriatic arthritis; JIA � juvenile idiopathicarthritis.

3882 AF KLINT ET AL

with an ABI Prism 7700 Sequence Detection System (PerkinElmer, Norwalk, CT) using the 5� nuclease method (TaqMan)with a 2-step PCR protocol (95°C for 10 minutes, followed by

40 cycles of 95°C for 15 seconds and 60°C for 1 minute) forIL-1�. Amplification of IL-1� was performed using the humanPCR cytokine kit (Applied Biosystems, Foster City, CA).GAPDH was used as a housekeeping gene control. Forquantification of mRNA for IL-1�, IL-1�, HMGB-1, and TNF,a QuantiTec SYBR Green kit was used according to theinstructions of the manufacturer (Qiagen). Amplification wasperformed using the same Sequence Detection System asabove but with a 3-step PCR protocol (95°C for 10 minutes,followed by 40 cycles of 95°C for 15 seconds, 60°C for 30seconds, and 72°C for 30 seconds; then 95°C for 15 seconds and60°C for 20 seconds and ramping for 20 minutes to 95°C for 15seconds).

Relative quantification of mRNA levels was performedusing the standard curve method. The standard curve thatconsisted of a pool of all patient samples was created using 5different dilutions (1, 1:5, 1:25, 1:125, and 1:625). The amountof mRNA in each sample and the amount of endogenouscontrol GAPDH could be deduced from target and GAPDHmRNA standard curves, respectively. Standard curves wereused for different targets; thus, the relative amounts cannot becompared between targets. The amount of each sample wascalculated as the ratio between the relative amount of targetand the relative amount of the corresponding endogenouscontrol (GAPDH mRNA). The samples were run in duplicate.Samples without cDNA were used as negative controls.

Primers and probe for IL-1� were obtained fromApplied Biosystems and have been described previously (15).Primers for IL-1� and HMGB-1 (from CyberGene, NovumResearch Park, Huddinge, Sweden) were as follows: IL-1�

Figure 1. Quantitative results of digital image studies of the immu-nostained tissue area occupied by T lymphocytes (CD3) or macro-phages (CD68, CD163) before (white boxes and circles) and after(gray boxes and circles) intraarticular glucocorticoid therapy in 31patients with chronic arthritis. Data are presented as box plots, wherethe boxes represent the 25th to 75th percentiles, the lines within theboxes represent the median, and the lines outside the boxes representthe 10th and 90th percentiles. Circles indicate outliers. Corticosteroidinjection down-regulated the number of T lymphocytes (� � P � 0.05),but not the number of macrophages.

Figure 2. A, Cytokine protein expression before (white boxes and circles) and after (gray boxes and circles) intraarticular (IA) glucocorticoid (GC)therapy in biopsy specimens from 31 patients with chronic arthritis. Immunostaining for interleukin-1� (IL-1�), IL-1�, tumor necrosis factor (TNF),and high mobility group box chromosomal protein 1 (HMGB-1) was evaluated by digital image analysis. The overall protein expression of IL-1�,TNF, and HMGB-1 was significantly reduced after therapy (� � P � 0.05). B, Expression of mRNA for IL-1-1�, IL-1�, TNF, and HMGB-1 before(white boxes and circles) and after (gray boxes and circles) IA GC therapy in paired biopsy specimens from 12 patients with chronic arthritis,analyzed by real-time reverse transcriptase–polymerase chain reaction. IL-1� and IL-1� mRNA levels were significantly reduced after a singleinjection (� � P � 0.05). Data are presented as box plots, where the boxes represent the 25th to 75th percentiles, the lines within the boxes representthe median, and the lines outside the boxes represent the 10th and 90th percentiles. Circles indicate outliers.

EFFECT OF IA GLUCOCORTICOIDS ON SYNOVIAL CELL INFLAMMATION 3883

forward 5�-CTG-ATG-GCC-CTA-AAC-AGA-TGA-G, re-verse 5�-GGT-CGG-AGA-TTC-GTA-GCT-GGA-T (16,17);HMGB-1 forward 5�-GCG-GAC-AAG-GCC-CGT-TA, re-verse 5�-AGA-GGA-AGA-AGG-CCG-AAG-GA (18). Prim-ers for TNF, designed with Primer Express software (PerkinElmer) and obtained from CyberGene, were as follows: for-ward 5�-CCA-GGG-ACC-TCT-CTC-TAA-TCA-GC, reverse5�-CTC-AGC-TTG-AGG-GTT-TGC-TAC-A (19). Primersand probe for GAPDH were designed with Primer Expresssoftware (primers obtained from CyberGene, probe obtainedfrom Scandinavian Gene Synthesis [Koping, Sweden]) andwere as follows: forward probe 5�-TAT-TGT-TGC-CAT-CAA-TGA-CCC-CTT-CAT-TGA, forward primer 5�-AGG-GCT-GCT-TTT-AAC-TCT-GGT-AAA, reverse primer: 5�-CAT-ATT-GGA-ACA-TGT-AAA-CCA-TGT-AGT-TG (19).

Statistical analysis. Statistical analyses were per-formed using SPSS 11.5 for Windows (SPSS, Chicago, IL).Wilcoxon’s signed rank test was used to compare parametersbefore and after IA GC treatment. Adjustment for multiplecomparisons was made by the Bonferroni correction method. Pvalues less than or equal to 0.05 were considered significant.

RESULTS

Improvement in clinical signs of inflammationafter IA GC injection. At the initial arthroscopy, allinvestigated joints had clinical signs of inflammation,with swelling and pain accompanied by IA effusion. Atthe time of followup arthroscopy 9–15 days after GCinjection, the treated joints exhibited marked clinicalimprovement in the majority of the patients. In 2injected joints, there were minimal or no clinical effects.In the remaining 29 joints, swelling and pain werereduced and effusion was undetectable. However, thepostinjection arthroscopy still showed signs of persistingsynovitis, including villus formation and increased vas-cularity in all patients.

Reduced numbers of synovial tissue T lympho-cytes, but not of macrophages, after IA GC injection.Before treatment, all joints expressed a considerable

Figure 3. IL-1� (A and B) and TNF (C and D) expression with brown (diaminobenzidine) immunoperoxidasestaining in synovial sections from 1 GC-injected joint in a patient with rheumatoid arthritis. Cell nuclei werevisualized with blue hematoxylin counterstaining. IL-1� was mainly observed in the vascular compartments, withno observable change of staining intensity after (B) versus before (A) IA GC injection. TNF was readilydetectable before GC injection (C), intracellularly as well as extracellularly in inflammatory cell infiltrates (thickarrow in C), and its expression was substantially reduced 2 weeks after GC injection (D). Thin arrows indicatelarge vessels expressing IL-1� or TNF. See Figure 2 for definitions. (Original magnification � 250.)

3884 AF KLINT ET AL

number of T lymphocytes in the synovium, mainly in theform of lymphoid aggregates and as scattered cells in thesublining tissue. The number of T lymphocytes wasmarkedly reduced in the great majority of treated joints(Figure 1). Similarly, the number of lymphoid aggregateswas diminished in almost every biopsy sample at fol-lowup. In contrast, no statistically significant effect onmacrophage infiltration, as indicated by CD68 or CD163phenotypic staining, was observed at the time of thesecond biopsy (Figure 1).

Diverse modulation of synovial cytokine proteinexpression by IA GC. The expression of TNF and IL-1�protein detected by immunohistochemical staining inthe synovial membrane was significantly diminished, butnot abrogated, after IA GC injection (Figures 2A and 3Cand D). TNF was mainly observed in sublining

macrophage-like cells and in the lining layer, while IL-1�was demonstrated in macrophages and fibroblasts, asjudged by morphology. In contrast, there were no dif-ferences in the appearance of IL-1� in tissues ob-tained before and after GC injections (Figures 2A and3A and B). IL-1� was mainly observed in vascularendothelial cells present in the majority of bloodvessels, and only occasionally in cells outside thevascular compartment.

The proinflammatory mediator HMGB-1 wasaberrantly expressed outside the cell nucleus inmacrophage-like cells, in numerous synoviocytes, and inendothelial cells in a minority of blood vessels (Figure4). Before GC injection, HMGB-1 was detectable extra-cellularly, with a staining pattern encompassing inflam-matory cells that also expressed cytoplasmic HMGB-1

Figure 4. HMGB-1 expression in inflammatory cells (A), lining layer (B), and endothelial cells (C and D) inbiopsy specimens obtained before (A and C) and after (B and D) IA GC injection in the knee joint of a patientwith rheumatoid arthritis. Massive HMGB-1 staining (brown) was evident both intracellularly and extracellularlyin inflammatory infiltrates (thick arrow) before GC injection (A). This staining pattern was substantially changedafter GC injection, when the extracellular HMGB-1 staining was almost abolished and the intracellular HMGB-1was mainly confined to the lining layer cells (thick arrow) and was not seen in the sublining cell infiltrates (B).The prominent cytoplasmic HMGB-1 staining in vascular endothelial cells appeared unchanged after (D) versusbefore (C) GC injection (arrows). Thin arrow in A shows a large vessel with no HMGB-1 staining. Thin arrowin B shows HMGB-1 staining in a capillary. See Figure 2 for definitions. (Original magnification � 250 in A, C,and D; � 100 in B.)

EFFECT OF IA GLUCOCORTICOIDS ON SYNOVIAL CELL INFLAMMATION 3885

(Figure 4A). Overall HMGB-1 staining in the synovialtissue sampled postinjection was strongly diminished(Figures 2A and 4A and B), but in a selective manner.Extracellular HMGB-1 staining was almost eliminatedat followup, and cytoplasmic HMGB-1 staining in syno-viocytes and macrophage-like cells was considerablyreduced. In contrast, the cytoplasmic HMGB-1 compo-nent in vascular endothelial cells and in macrophage-likecells in the lining layer did not decrease (Figures 4B andD). Nuclear HMGB-1 expression was not down-regulated, and even appeared with increased intensity inmany synovial cells (Figure 4B).

Selective effects of IA GC injections on synovialcytokine mRNA levels. IA GC injection affected cyto-kine mRNA expression similarly among all 12 patientsfrom whom paired synovial samples were technicallysuitable for mRNA extraction. However, results ob-tained in the cytokine mRNA analysis only partiallyparalleled those observed in cytokine protein assess-ments (Figure 2B). Levels of mRNA for both IL-1� andIL-1� were strongly down-regulated by IA GC. Incontrast, TNF and HMGB-1 mRNA levels were notsignificantly altered.

Effects of IA GC on the synovial vascular com-partment. Immunohistologic studies did not reveal anyreduction in the vascularity of the tissue as a conse-

quence of GC therapy. We used the pan-endothelialmarker CD31 for digital image analysis assessing tissuearea (Figure 5A), and manual counting of blood vesselsby microscopy (data not shown), and found no indica-tions of significantly reduced vascularity.

The adhesion molecule ICAM-1 was demon-strated by immunostaining to be present in blood vesselsand in nonvascular tissue, particularly in the synoviallining layer. IA GC injection consistently and signifi-cantly down-regulated ICAM-1 expression in all studiedtissue compartments in 30 of the 31 patients (Figure 5A).

The expression of vascular growth factors, such asVEGF, was also significantly reduced in the synovialtissue sections after GC injection (Figure 5A). However,this decrease was seen only within the inflammatory cellinfiltrates, and not within the vascular compartments.These immunohistochemical findings were further sup-ported by results of ELISA study of peripheral bloodsamples (Figure 5B). In healthy controls, the mean �SD serum VEGF level was 331.4 � 33.3 pg/ml (range25.2–1,315.1). In patients, the serum VEGF level was923.7 � 105.1 pg/ml (range 366.8–2,198.8) before and620.2 � 50.3 pg/ml (range 321.2–1,234.9) 9–15 days afterGC injection. Although the serum VEGF levels inpatients declined markedly (P � 0.05) after therapy,

Figure 5. A, Quantitative results of immunohistochemistry and digital image analysis of the expression of molecules associated with the vascularsystem, i.e., as the pan-endothelial marker CD31, intracellular adhesion molecule 1 (ICAM-1), and vascular endothelial growth factor (VEGF),before (white boxes and circles) and after (gray boxes and circles) intraarticular (IA) glucocorticoid (GC) therapy in 31 patients with chronic arthritis.Expression of ICAM-1 and VEGF was significantly reduced after therapy (� � P � 0.05). Vascularity did not change during the study period, asjudged by CD31 staining. B, Serum VEGF levels were analyzed in blood samples obtained before (white box and circles) and 9–15 days after (graybox and circles) IA GC injection in 20 patients with chronic arthritis. The same analysis was also performed in 56 healthy age-matched blood donors,in a single blood test per person (stippled box and circles). VEGF levels were significantly higher in patients before versus after treatment, and bothbefore and after treatment, patients exhibited higher serum VEGF levels than controls (� � P � 0.05). Data are presented as box plots, where theboxes represent the 25th to 75th percentiles, the lines within the boxes represent the median, and the lines outside the boxes represent the 10th and90th percentiles. Circles indicate outliers.

3886 AF KLINT ET AL

they remained elevated when compared with those inhealthy controls.

DISCUSSION

To our knowledge, this is the first study toexamine therapeutic effects of IA GC injection on thetarget organ of chronic arthritides, the synovial mem-brane. However, arthroscopy and ex vivo studies, whichform the basis for the observations described in thisreport, have inherent limitations. Although, it is desir-able from a scientific standpoint to perform repeatedarthroscopies during followup, ethical considerationspreclude this strategy. Arthroscopy allows visualizationmainly of the surface of synovium and not of pathologicprocesses in the adjacent bone tissue. Likewise, withsynovial biopsy, only regions of superficial synovium oflimited size, which may not necessarily reflect the globalIA inflammatory activity, are sampled. However, onewould expect that the strongest therapeutic effectswould be on superficial surfaces, where injected GCreaches maximal levels. In the present study pretreat-ment biopsies were all obtained from areas expressingmaximal signs of inflammation, and careful mapping wasundertaken to direct the followup sampling to the sameregions to enable evaluation of therapy-inducedchanges.

Despite the fact that the IA GC injections led togood clinical recovery of almost all treated joints duringthe study period, both arthroscopy and microscopyexaminations demonstrated that many signs of inflam-mation in the synovial tissue prevailed. IA GC injectionmarkedly reduced the number of synovial T lymphocytesbut not of macrophages, which differs from resultsobserved after 2 weeks of treatment with systemicallyadministered GC (4). One of the most striking findingsof the latter study was that prednisolone therapy in RAwas associated with a marked reduction in macrophageinfiltration in the synovial tissue. Separate biologic ef-fects exerted by local versus systemic GC administrationpossibly explain the discrepant results concerning mac-rophage accumulation. High-dose systemic, but not lo-cal, GC therapy will induce a profound monocytopenia(20), impeding the access of monocytes to the inflamedsynovial tissue. Reduced numbers of macrophages aswell as functionally pacified macrophages both correlatewith good clinical efficacy of any therapy administeredfor synovitis.

We cannot presently explain the mechanisms forthe marked reduction of synovial T lymphocytes after IAGC injection in our study, which did not occur to thesame extent after systemic GC administration (4). It has

been demonstrated that activated human T lymphocytesare susceptible to induction of apoptosis in response tohigh doses of GC (21), which may be achieved onlylocally after IA GC injection. However, when we usedthe TUNEL assay staining technique in paired biopsyspecimens from 20 treated joints, we could not demon-strate signs of increased apoptosis 9–15 days after GCinjection (results not shown).

We observed that IA GC injections affected theexpression of cytokine protein and mRNA levels with, atfirst glance, contradictory outcomes. GC treatment didnot diminish IL-1� protein expression during the studyperiod, despite the fact that IL-1� mRNA was reduced.IL-1� protein as well as mRNA were down-regulated.Both TNF and HMGB-1 protein levels were decreased,while neither TNF mRNA nor HMGB-1 mRNA wassignificantly reduced.

Transcription of IL-1�, as well as IL-1�, is underNF-�B control (for review, see ref. 22), so the observeddown-regulation of IL-1 mRNA levels by GC treatmentis a predictable result. IL-1� protein lacks a knownsecretory pathway, precluding a hypothetical influenceof GC on IL-1� release. Undiminished IL-1� proteinexpression, despite suppressed mRNA levels, could pos-sibly be explained by a long half-life of IL-1� proteinwithin the endothelial cells. IL-1� transcription is di-rectly suppressed by a regulatory region containing anegative GC response element present in the IL-1�promoter, which would contribute to the IL-1� down-regulation (23).

TNF production is known to be negatively regu-lated by GCs, both transcriptionally by decreasing thesteady-state level of TNF mRNA and posttranscription-ally by reducing the rate of secretion (24). Interferon-�(IFN�) could possibly be a factor explaining the lack ofeffect of GCs on TNF mRNA in our study. IFN�strongly counteracts the effects of GCs on TNF at themRNA level but not at posttranscriptional sites of action(24).

HMGB-1 transcription is up-regulated by estro-gen (25), but there are no reports of mediation of itsregulation by GCs. The reduced extracellular HMGB-1protein levels observed in our study are likely due toposttranscriptional therapeutic effects. GC-induced li-pocortins suppress phospholipase A2 (PLA2) formation(26), a crucial factor for HMGB-1 protein secretionfrom activated macrophages (9,27). Several therapeuticantagonists of the NF-�B pathway have been shown toinhibit extracellular HMGB-1 release (28), suggestingthat this may also be a mechanism by which GCs may actto inhibit inflammatory HMGB-1 expression.

EFFECT OF IA GLUCOCORTICOIDS ON SYNOVIAL CELL INFLAMMATION 3887

There are few corroborating reports of studiesinvestigating local effects of drug therapy on the synovialmembrane in chronic arthritis. Two-week treatment ofrheumatoid synovitis with high-dose oral prednisoloneresulted in a clear trend toward reduction in levels ofTNF and IL-1�, but not IL-1� (4). A study of intrave-nous pulse methylprednisolone therapy with arthrosco-pically directed synovial biopsy before and 24 hours afterinfusion revealed a strong decrease of TNF and IL-8levels in RA (29). Moreover, disease relapse was asso-ciated with recurrence of synovial TNF expression.However, no effects on IL-1� or IL-1 receptor antago-nist levels were observed. Young and coworkers in-vestigated therapeutic effects of IA GC injection inosteoarthritis, using serial arthroscopy sampling andimmunostaining (30). They found a modest reduction ofmacrophage numbers in the lining layer, but no changesin the sublining tissue 1 month after injection. Cytokineassessment in a study of combination treatment withsystemic GCs and gold showed no effect on synovialIL-1� expression 2 weeks after initiation of therapy,while a substantial reduction was noted 10 weeks later(31).

To date, only 2 types of RA treatment have beenshown to work within a time frame measured in days:GC and TNF-blocking therapy. We have previouslypublished a study on anti-TNF (infliximab) treatment ofpatients with active RA, utilizing assessment methodssimilar to those in the present study (32). Synovial tissueTNF production was distinctly reduced 14 days after asingle intravenous infliximab infusion, whereas no suchchanges in IL-1� and IL-1� expression in the synoviumwere recorded.

This is the first report to address the therapeuticeffects of GCs on the proinflammatory cytokineHMGB-1 in human arthritis. This protein is of majorinterest for several reasons: 1) it has been demonstratedto be present extracellularly in considerable amounts ininflamed, but not normal, joints; 2) it causes destructivearthritis when injected IA in animal studies; and 3)neutralizing anti–HMGB-1 therapy is beneficial in ex-perimental arthritis (for review, see ref. 6). HMGB-1 isubiquitously present as a nuclear factor in all mamma-lian cells, although we did not demonstrate the nuclearexpression in all cells in our specimens, probably due toinsensitivity of the staining technique. Nevertheless, it isthe extranuclear expression of HMGB-1 that is of maininterest for inflammation research.

HMGB-1 may be passively released intracellu-larly by necrotic, but not apoptotic, cells, or it can beactively secreted by specialized cells including activatedmacrophages (9). It is thus biologically significant that in

the present study, the extracellular HMGB-1 staininghad decreased considerably in the postinjection speci-mens. Activation of macrophages results in an accumu-lation of HMGB-1 in cytoplasmic secretory lysosomes.The lysosomes then are exocytosed when triggered bylysophosphatidylcholine (9), a lipid generated by PLA2at sites of inflammation. Since GCs are potent inhibitorsof the synthesis of PLA2 (26), it is conceivable that IAGC injections caused reduced extracellular HMGB-1secretion via a down-regulation of lysophosphatidyl-choline formation required for HMGB-1 release. Arecent study on systemic GC therapy in dermatomyositisand polymyositis likewise demonstrated a potent sup-pression of aberrant HMGB-1 expression in and adja-cent to mononuclear inflammatory cells (33).

A novel and thus unexpected result of this studywas the finding of a relative insensitivity of the vascularcompartment to the antiinflammatory effects of GCs.We did not observe a reduction in the area or numbersof blood vessels or a reduction of the proinflammatorycytokines IL-1� and HMGB-1 in the endothelial cells.The short followup period (maximum 15 days) compli-cates the interpretation of these findings. However,when patients with chronic myositis received systemicGC therapy for 3–6 months, inflammatory, extranuclearHMGB-1 expression prevailed selectively in the vascularendothelial cells (33), while vascular IL-1� expressiondeclined but did not normalize (34). Furthermore, anotherstudy using cell cultures demonstrated strong inhibitoryeffects of dexamethasone on lipopolysaccharide-inducedIL-1 formation in human monocytes, but not in endo-thelial cells (35). These findings together with the resultsof our study might have some relevance with regard tocardiovascular function. There are many inflammatorydisorders that are successfully treated with systemicGCs, yet the same patients are at increased risk fordeveloping cardiovascular disease.

Other parameters related to vascular inflamma-tion and activation, i.e., expression of ICAM-1 andproduction of VEGF, were susceptible to GC-inducedsuppression. The down-regulation of VEGF revealed byimmunostaining of the synovial specimens was con-firmed when serum VEGF levels were analyzed beforeand after the IA injections.

In summary, the results of our study emphasizethe central role of TNF and IL-1�, and possibly alsoHMGB-1, in the pathogenesis of chronic arthritis. These3 mediators are involved in a positive feedback system(5–8) and could thus be anticipated to be affected in thesame direction by this therapy. The observed suppres-sion of aberrant HMGB-1 expression after IA GCtherapy represents a novel finding. Future studies are

3888 AF KLINT ET AL

needed to determine whether specific HMGB-1–blocking therapy will be beneficial for patients withchronic arthritis.

REFERENCES

1. Almawi WY, Beyhum HN, Rahme AA, Rieder MJ. Regulation ofcytokine and cytokine receptor expression by glucocorticoids.J Leukoc Biol 1996;60:563–72.

2. Moreland LW, O�Dell JR. Glucocorticoids and rheumatoid arthri-tis: back to the future? [review]. Arthritis Rheum 2002;46:2553–63.

3. Ostergaard M, Stoltenberg M, Gideon P, Henriksen O, LorenzenI. Changes in synovial membrane and joint effusion volumes afterintraarticular methylprednisolone: quantitative assessment of in-flammatory and destructive changes in arthritis by MRI. J Rheu-matol 1996;23:1151–61.

4. Gerlag DM, Haringman JJ, Smeets TJ, Zwinderman AH, KraanMC, Laud PJ, et al. Effects of oral prednisolone on biomarkers insynovial tissue and clinical improvement in rheumatoid arthritis.Arthritis Rheum 2004;50:3783–91.

5. Feldmann M, Brennan FM, Foxwell BM, Maini RN. The role ofTNF � and IL-1 in rheumatoid arthritis. Curr Dir Autoimmun2001;3:188–99.

6. Andersson U, Erlandsson-Harris H. HMGB1 is a potent trigger ofarthritis. J Intern Med 2004;255:344–50.

7. Andersson U, Tracey KJ. HMGB1 as a mediator of necrosis-induced inflammation and a therapeutic target in arthritis. RheumDis Clin North Am 2004;30:627–37.

8. Andersson U, Wang H, Aveberger AC, Janson A, Palmblad K,Yang H, et al. High mobility group-1 (HMG-1) stimulates proin-flammatory cytokine synthesis in human monocytes. J Exp Med2000;192:565–70.

9. Gardella S, Andrei C, Ferrera D, Lotti LV, Torrisi MR, BianchiME, et al. The nuclear protein HMGB1 is secreted by monocytesvia a non-classical, vesicle-mediated secretory pathway. EMBORep 2002;3:995–1001.

10. Scaffidi P, Misteli T, Bianchi ME. Release of chromatin proteinHMGB1 by necrotic cells triggers inflammation. Nature 2002;418:191–5.

11. Huttunen HJ, Kuja-Panula J, Sorci G, Agneletti AL, Donato R,Rauvala H. Coregulation of neurite outgrowth and cell survival byamphoterin and S100 proteins through receptor for advancedglycation end products (RAGE) activation. J Biol Chem 2000;275:40096–105.

12. Park JS, Svetkauskaite D, He Q, Kim JY, Strassheim D, IshizakaA, et al. Involvement of toll-like receptors 2 and 4 in cellularactivation by high mobility group box 1 protein. J Biol Chem2004;279:7370–7.

13. 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.

14. Ulfgren AK, Grondal L, Lindblad S, Johnell O, Klareskog L,Andersson U. Interindividual and intra-articular variation ofproinflammatory cytokines in patients with rheumatoid arthritis:potential implications for treatment. Ann Rheum Dis 2000;59:439–47.

15. Seddighzadeh M, Larsson P, Ulfgren AC, Onelov E, Berggren P,Tribukait B, et al. Low IL-1� expression in bladder cancer tissueand survival. Eur Urol 2003;43:362–8.

16. Brink N, Szamel M, Young AR, Wittern KP, Bergemann J. Com-parative quantification of IL-1�, IL-10, IL-10r, TNF� and IL-7mRNA levels in UV-irradiated human skin in vivo. Inflamm Res2000;49:290–6.

17. Rad R, Dossumbekova A, Neu B, Lang R, Bauer S, Saur D, et al.Cytokine gene polymorphisms influence mucosal cytokine expres-sion, gastric inflammation, and host specific colonisation duringHelicobacter pylori infection. Gut 2004;53:1082–9.

18. Pachot A, Monneret G, Voirin N, Leissner P, Venet F, Bohe J, etal. Longitudinal study of cytokine and transcription factor mRNAexpression in septic shock. Clin Immunol 2005;114:61–9.

19. Khademi M, Illes Z, Gielen AW, Marta M, Takazawa N, Baecher-Allan C, et al. T cell Ig- and mucin-domain-containing molecule-3(TIM-3) and TIM-1 molecules are differentially expressed on humanTh1 and Th2 cells and in cerebrospinal fluid-derived mononuclearcells in multiple sclerosis. J Immunol 2004;172:7169–76.

20. Fauci AS, Dale DC, Balow JE. Glucocorticosteroid therapy:mechanisms of action and clinical considerations. Ann Intern Med1976;84:304–15.

21. Kirsch AH, Mahmood AA, Endres J, Bohra L, Bonish B, WeberK, et al. Apoptosis of human T-cells: induction by glucocorticoidsor surface receptor ligation and ex vivo. J Biol Regul HomeostAgents 1999;13:80–9.

22. Auron PE, Webb AC. Interleukin-1: a gene expression systemregulated at multiple levels. Eur Cytokine Netw 1994;5:573–92.

23. Zhang G, Zhang L, Duff GW. A negative regulatory regioncontaining a glucocorticosteroid response element (nGRE) in thehuman IL-1� gene. DNA Cell Biol 1997;16:145–52.

24. Luedke CE, Cerami A. Interferon-� overcomes glucocorticoidsuppression of cachectin/tumor necrosis factor biosynthesis bymurine macrophages. J Clin Invest 1990;86:1234–40.

25. Borrman L, Kim I, Schultheiss D, Rogalla P, Bullerdiek J.Regulation of the expression of HMGB1, a co-activator of theestrogen receptor. Anticancer Res 2001;21:301–5.

26. Touqui L, Alaoui-El-Azher M. Mammalian secreted phospho-lipases A2 and their pathophysiological significance in inflamma-tory diseases. Curr Mol Med 2001;1:739–54.

27. Muller S, Ronfani L, Bianchi ME. Regulated expression andsubcellular localization of HMGB1, a chromatin protein with acytokine function. J Intern Med 2004;255:332–43.

28. Wang H, Liao H, Ochani M, Justiniani M, Lin X, Yang L, et al.Cholinergic agonists inhibit HMGB1 release and improve survivalin experimental sepsis. Nat Med 2004;10:1216–21.

29. Youssef PP, Haynes DR, Triantafillou S, Parker A, Gamble JR,Roberts-Thomson PJ, et al. Effects of pulse methylprednisoloneon inflammatory mediators in peripheral blood, synovial fluid, andsynovial membrane in rheumatoid arthritis. Arthritis Rheum 1997;40:1400–8.

30. Young L, Katrib A, Cuello C, Vollmer-Conna U, Bertouch JV,Roberts-Thomson PJ, et al. Effects of intraarticular glucocorti-coids on macrophage infiltration and mediators of joint damage inosteoarthritis synovial membranes: findings in a double-blind,placebo-controlled study. Arthritis Rheum 2001;44:343–50.

31. Kirkham BW, Navarro FJ, Corkill MM, Panayi GS. In vivo analysisof disease modifying drug therapy activity in rheumatoid arthritisby sequential immunohistological analysis of synovial membraneinterleukin 1�. J Rheumatol 1994;21:1615–9.

32. Ulfgren AK, Andersson U, Engstrom M, Klareskog L, Maini RN,Taylor PC. Systemic anti–tumor necrosis factor � therapy inrheumatoid arthritis down-regulates synovial tumor necrosis factor� synthesis. Arthritis Rheum 2000;43:2391–6.

33. Ulfgren AK, Grundtman C, Borg K, Alexandersson H, AnderssonU, Harris HE, et al. Down-regulation of the aberrant expression ofthe inflammation mediator high mobility group box chromosomalprotein 1 in muscle tissue of patients with polymyositis anddermatomyositis treated with corticosteroids. Arthritis Rheum2004;50:1586–94.

34. Lundberg I, Kratz AK, Alexandersson H, Patarroyo M. Decreasedexpression of interleukin-1�, interleukin-1�, and cell adhesionmolecules in muscle tissue following corticosteroid treatment inpatients with polymyositis and dermatomyositis. Arthritis Rheum2000;43:336–48.

35. Zuckerman SH, Shelhaas JS, Butler LD. Differential regulation oflipopolysaccharide-induced interleukin 1 and tumor necrosis fac-tor synthesis: effects of endogenous and exogenous glucocorticoidsand the role of the pituitary-adrenal axis. Eur J Immunol 1989;19:301–5.

EFFECT OF IA GLUCOCORTICOIDS ON SYNOVIAL CELL INFLAMMATION 3889