ARTHRITIS & RHEUMATISMVol. 56, No. 1, January 2007, pp 117–128DOI 10.1002/art.22303© 2007, American College of Rheumatology

Use of Molecular Imaging to Quantify Response toIKK-2 Inhibitor Treatment in Murine Arthritis

Elena S. Izmailova,1 Nancy Paz,1 Herlen Alencar,2 Miyoung Chun,1 Lisa Schopf,1

Michael Hepperle,1 Joan H. Lane,1 Geraldine Harriman,1 Yajun Xu,1 Timothy Ocain,1

Ralph Weissleder,2 Umar Mahmood,2 Aileen M. Healy,1 and Bruce Jaffee1

Objective. The NF-�B signaling pathway pro-motes the immune response in rheumatoid arthritis(RA) and in rodent models of RA. NF-�B activity isregulated by the IKK-2 kinase during inflammatoryresponses. To elucidate how IKK-2 inhibition sup-presses disease development, we used a combination ofin vivo imaging, transcription profiling, and histopa-thology technologies to study mice with antibody-induced arthritis.

Methods. ML120B, a potent, small molecule in-hibitor of IKK-2, was administered to arthritic animals,and disease activity was monitored. NF-�B activity indiseased joints was quantified by in vivo imaging.Quantitative reverse transcriptase–polymerase chainreaction was used to evaluate gene expression in joints.Protease-activated near-infrared fluorescence (NIRF)in vivo imaging was applied to assess the amounts ofactive proteases in the joints.

Results. Oral administration of ML120B sup-pressed both clinical and histopathologic manifesta-tions of disease. In vivo imaging demonstrated that

NF-�B activity in inflamed arthritic paws was inhibitedby ML120B, resulting in significant suppression ofmultiple genes in the NF-�B pathway, i.e., KC, epithelialneutrophil–activating peptide 78, JE, intercellular ad-hesion molecule 1, CD3, CD68, tumor necrosis factor �,interleukin-1�, interleukin-6, inducible nitric oxide syn-thase, cyclooxygenase 2, matrix metalloproteinase 3,cathepsin B, and cathepsin K. NIRF in vivo imagingdemonstrated that ML120B treatment dramatically re-duced the amount of active proteases in the joints.

Conclusion. Our data demonstrate that IKK-2inhibition in the murine model of antibody-inducedarthritis suppresses both inflammation and joint de-struction. In addition, this study highlights how geneexpression profiling can facilitate the identification ofsurrogate biomarkers of disease activity and treatmentresponse in an experimental model of arthritis.

NF-�B signaling is thought to play an importantrole in the inflammatory response underlying the patho-genesis of autoimmune diseases, including rheumatoidarthritis (RA). NF-�B activity is regulated by the IKKcomplex. This complex consists of 2 kinases, IKK-1 andIKK-2, and a regulatory subunit, IKK�/NEMO. Recentdata indicate that IKK-2, rather than IKK-1, participatesin the pathway by which proinflammatory stimuli induceNF-�B function (1).

NF-�B activity is implicated in promoting bothinflammation and tissue remodeling, by activating thetranscription of many key genes (2). Recent data fromboth clinical RA studies and studies in rodent models ofarthritis suggest that inflammation and destruction ofarticular structures occur independent of one another(3,4). It is well established that NF-�B is involved in theregulation of multiple proinflammatory mechanisms.Activation of NF-�B is necessary and sufficient fortranscriptional activation of intercellular adhesion mol-

1Elena S. Izmailova, PhD, Nancy Paz, MS (current address:Merrimack Pharmaceuticals, Cambridge, Massachusetts), MiyoungChun, PhD (current address: University of California, Santa Barbara),Lisa Schopf, PhD, Michael Hepperle, PhD (current address: PhenomixCorporation, San Diego, California), Joan H. Lane, DVM, DACVP,Geraldine Harriman, PhD, Yajun Xu, PhD, Timothy Ocain, PhD,Aileen M. Healy, PhD, Bruce Jaffee, PhD: Millennium Pharmaceuti-cals, Inc., Cambridge, Massachusetts; 2Herlen Alencar, MD, RalphWeissleder, MD, PhD, Umar Mahmood, MD, PhD: MassachusettsGeneral Hospital, Charlestown.

Drs. Izmailova, Schopf, Lane, Harriman, Ocain, and Jaffeehave or have had stock or stock options in Millennium Pharmaceuti-cals, Inc. Drs. Weissleder and Mahmood have stock or stock options inVisen Medical. Dr. Mahmood holds patents on activatable opticalagents.

Address correspondence and reprint requests to Elena S.Izmailova, PhD, Millennium Pharmaceuticals, Inc., 35 LansdowneStreet, Cambridge, MA 02139. E-mail: Elena.Izmailova@mpi.com.

Submitted for publication March 10, 2006; accepted in revisedform September 22, 2006.

117

ecule 1 (ICAM-1) and the chemokines monocyte che-motactic protein 1 and interleukin-8 (IL-8). These mol-ecules facilitate infiltration of inflammatory cells intothe diseased joint. NF-�B also regulates transcription oftumor necrosis factor � (TNF�), IL-1�, IL-6, induciblenitric oxide synthase (iNOS), and cyclooxygenase 2(COX-2) (5), all of which are required for the initiation,amplification, and maintenance of chronic inflamma-tion. The role of NF-�B in the mechanisms regulatingtissue remodeling is less well defined.

Several lines of evidence indicate that heightenedprotease activity is a contributing factor in the tissueremodeling associated with RA (6,7). Proteases play arole in pannus growth and degradation of the articularstructure. Both matrix metalloproteinases (MMPs) andcysteine proteases are implicated in the development ofRA (8–11). NF-�B regulates the production of someMMPs, such as MMP-3 and MMP-13 (12), but its effecton cysteine proteases remains unclear.

One of the major challenges in assessing theefficacy of antirheumatic drugs is the inability to esti-mate protease activity in the joints directly and non-invasively in vivo. Radiographic assessment of the jointdoes not generally take place until late in the course ofthe disease, which limits its usefulness. The recentdevelopment of near-infrared fluorescence (NIRF) im-aging probes allows for the localization, visualization,and quantification of signal in areas with increasedproteolytic activity in vivo (13). Such molecular report-ers may serve as sensitive indicators for use in objectiveassessment of treatment response.

ML120B is a potent and selective small moleculeinhibitor of IKK-2 (14) that suppresses the clinical andhistologic disease manifestations of antibody-inducedarthritis in mice. IKK-2 inhibition also results in signif-icant suppression of multiple genes in the NF-�B path-way that are involved in the development of murineantibody-induced arthritis.

In the present study, we combined molecularimaging with gene expression profiling and histopatho-logic examination to examine the effects of IKK-2inhibition in murine antibody-induced arthritis. In vivoimaging demonstrated direct inhibition of NF-�B activ-ity by ML120B in inflamed paws. We found that theIKK-2 inhibitor suppressed expression, in the joints, ofseveral proinflammatory genes, including the adhesionmolecule ICAM-1, the chemokines JE, KC, and epithe-lial neutrophil–activating peptide 78 (ENA-78), the in-flammation mediators TNF�, IL-1�, IL-6, COX-2, andiNOS, as well as 2 markers of mononuclear cell infiltra-

tion, CD3 and CD68. Of note, ML120B treatment alsoinhibited the expression of the tissue remodeling genesMMP-3 and the cysteine proteases cathepsin B andcathepsin K. Furthermore, IKK-2 suppression dramati-cally reduced the level of destructive enzymes in dis-eased joints as demonstrated by noninvasive NIRFimaging, and these findings were concordant with histo-logic evidence of cartilage damage and bone resorption.Thus, our data demonstrate that IKK-2 inhibition ofantibody-induced arthritis in mice is likely the resultof suppression of both inflammation and degradation ofbone and cartilage.

MATERIALS AND METHODS

Induction of arthritis in mice. Female BALB/c micewere purchased from Charles River (Wilmington, MA). Allmice were housed at Millennium Pharmaceuticals with freeaccess to standard rodent chow diet and water and werestudied at 8–9 weeks of age. Animal studies were performedaccording to Institutional Animal Care and Use Committeestandards. For induction of arthritis, mice were injected intra-venously with 2 mg anti–type II collagen (anti-CII) antibody,according to the protocol recommended by the manufacturer(Chemicon, Temecula, CA). Three days later, animals receivedan intraperitoneal injection of 12.5 �g lipopolysaccharide(LPS) (Escherichia coli O111:B4).

Administration of the IKK-2 inhibitor ML120B. TheIKK-2 inhibitor ML120B was administered orally at 10 mg/kg,30 mg/kg, or 60 mg/kg in 0.5% hydroxypropyl-methylcellulose/0.2% Tween twice daily starting on day 0 after anti-CIIantibody injection. Untreated animals and animals injectedwith LPS only were used as nonarthritic controls.

Clinical disease scoring. The severity of arthritis wasgraded visually by assessing the level of swelling in each paw,including the tarsus (ankle) and carpus (wrist) joints. Scoreswere assessed consistently 1 hour after morning dosing, at thefollowing time points: day 0 (before administration of anti-CIIantibody), day 3 (before disease boost with the LPS injection),day 5 or 6 (when clinical symptoms became clearly visible), anddays 7 and 9 (peak of disease activity). The following scoringsystem was used: 0.5 � slight redness, 1.0 � 1 or more digitsswollen, 1.5 � 1 or more digits swollen and red/swollen tarsus,2.0 � moderate swelling, 2.5 � moderate swelling and swollendigits, 3.0 � severe swelling, 3.5 � severe swelling andmoderate ankylosis, and 4.0 � severe swelling and completeankylosis (maximum possible total score per animal 16).

Histopathology. Hind paws from one side of eachmouse were fixed for 4 days in 10% buffered formalin,decalcified for 2 weeks in 5% formic acid, and processed forparaffin embedding. Eight-micrometer sections were stainedwith toluidine blue and scored, under blinded conditions, by aveterinary pathologist (at BoulderPath, Boulder, CO) forinflammation, pannus formation/cartilage loss, and bone ero-sion. Severity was scored on a 0–5 scale.

In vivo NF-�B imaging. For NF-�B imaging studies(15), antibody-induced arthritis was introduced into transgenic

118 IZMAILOVA ET AL

female BALB/c mice as described above. ML120B was admin-istered orally at 60 mg/kg twice daily. Imaging experimentswere performed 1 hour after compound administration (thetime of maximum concentration in peripheral blood) on days0, 3, 7, and 9. For the imaging procedure, mice were anesthe-tized with 2% isoflurane in oxygen, and luciferin (150 mg/kg)was injected intraperitoneally. Ten minutes after luciferininjection, the animals were imaged in an IVIS200 system(Xenogen, Alameda, CA), with a 1-minute bioluminescenceexposure. Signal quantification was based on region of interestanalysis.

RNA extraction and polymerase chain reaction (PCR)analysis. Total RNA was prepared from each animal sepa-rately using left rear paw whole-joint homogenate. RNA wasextracted by the single-step method using RNA Stat-60 (Tel-Test, Friendswood, TX). After treatment with DNase I (Qia-gen, Valencia, CA), complementary DNA (cDNA) was syn-thesized using the MultiScribe Reverse Transcription kit(Applied Biosystems, Foster City, CA). Gene expression wasmeasured by TaqMan real-time PCR according to the protocolrecommended by the manufacturer (Applied Biosystems).

Target gene probes were labeled with 6-FAM, and theinternal reference probe, rodent GAPDH, was labeled withVIC. PCRs were performed with the forward and reverseprimers (200 nM) and the probe (100 nM) for GAPDH and theforward and reverse primers (600 nM) and probe (200 nM) forthe gene of interest. The experiments were performed on anABI Prism 7700 Sequence Detection System (Applied Biosys-tems) under the following conditions: 2 minutes at 50°C and 10minutes at 95°C, followed by 2-step PCR for 40 cycles of 95°Cfor 15 seconds followed by 60°C for 1 minute. The number ofPCR cycles needed for FAM and VIC fluorescence to cross athreshold of a statistically significant increase in fluorescence(threshold cycle [Ct]) was measured using Applied Biosystemssoftware. Relative target gene expression was determinedusing the following formula: relative expression � 2-��Ct,where ��Ct � (Ct target gene – Ct reference gene in experimentalcDNA sample) � (Ct target gene – Ct reference gene in mockreverse-transcribed RNA sample).

In vivo protease optical imaging and image analysis.Fluorescence reflectance imaging was performed using anepifluorescence system (bonSAI; Siemens, Erlangen, Ger-many) that is capable of near simultaneous data acquisition inmultiple channels, including a broad-spectrum visible whitelight channel providing anatomic detail and an NIR channelproviding molecular imaging information. The system consistsof a 150W halogen excitation light source connected to theacquisition box through an optical waveguide. The built-infilter wheel is set to deliver light at 400–745 nm for white lightimages and at 645–675 nm for NIR images. On the detectionside, a second filter wheel uses a 4-step neutral optical densityfilter for white light and a 720–750–nm filter for activatedProsense (Visen Medical, Woburn, MA) NIR fluorescencedetection. A charge coupled device camera with a matrix sizeof 1,360 � 1,024 pixels and a resolution of 0.116 mm/pixel wasused for image acquisition. Anesthesia was maintained bymask inhalation of isoflurane vaporized at concentrations ofup to 4% in the induction phase, and 1.5% during imaging.The isoflurane was delivered along with 100% oxygen at a flow

rate of 2 liters/minute. During imaging, body temperature waskept constant at 37°C.

Previous experiments aimed at optimizing imagingtime indicated that statistically significant differences in signalintensity between control and diseased animals can be de-tected as early as day 6. Fluorescence signal intensity reachesits maximum at the peak of disease activity, between days 7 and10. We performed imaging on days 6 and 9 after anti-CIIantibody injection, to detect early disease response to treat-ment and to assess the effect of the added compound whendisease activity is maximal. The imaging procedure was carriedout 24 hours after intravenous injection of the protease-activated probe Prosense, which contained a total of 2 nmolesof quenched NIRF dye. This dose and timing were based ondoses that had previously been used in tumor model systems.Signal intensity of this class of activated probe is linear withrespect to enzyme concentration. Exposure time was 0.3seconds for all fluorescence images, and the data were storedin DICOM (Digital Imaging and Communications in Medi-cine) format for subsequent analysis. System control and datastorage were performed with a PC using Syngo software(Siemens). Using custom software (CMIRImage; Center forMolecular Imaging Research, Massachusetts General Hospi-tal), signal intensities from all fluorescence images were deter-mined using a circular region of interest placed over anklejoints, skin (snout), and a reference standard. The fluorescencesignal from the mouse paws, expressed as relative fluorescenceintensity, was calibrated to the standard, subtracting the skinvalue as background. Immediately after the last imaging ses-sion, animals were killed and paws were obtained for RNAexpression analysis and histologic study.

Statistical analysis. Summary statistics are reported asthe mean � SEM for each treatment group. Data were testedfor normal distribution. One-way analysis of variance withDunnett’s multiple comparison test was used to identify signif-icant differences between experimental and vehicle-treatedcontrol groups. Pearson’s correlation coefficient was used toexpress the correlation between fluorescence intensity and themeans of total clinical scores. All statistics were generatedusing GraphPad Prism software (GraphPad, San Diego, CA).P values less than 0.05 were considered significant.

RESULTS

Increase in NF-�B activity in vivo in murinearthritis. To assess NF-�B activity during arthritis, weinduced antibody-induced arthritis in transgenic miceexpressing the luciferase gene, under control of theNF-�B–inducible promoter (15). We visualized NF-�Bactivity in diseased joints by in vivo bioluminescenceimaging, captured images, and quantified the lumines-cence signal intensity in nonarthritic controls and ar-thritic paws during disease progression. On day 7 afterinjection of anti-CII antibodies, signal intensity was6.5-fold higher in arthritic animals than in the control

RESPONSE TO IKK-2 INHIBITION IN ARTHRITIC MICE 119

group. On day 9, signal intensity was 9.4-fold higher inthe arthritic animals compared with controls (Figure 1).

Inhibition of the development of antibody-induced arthritis by ML120B treatment. Clinical dis-ease symptoms. To determine if inhibition of theNF-�B pathway affects antibody-induced arthritis devel-opment, we used ML120B, a potent and selective in-hibitor of IKK-2 kinase (14). The antibody-inducedarthritis model was initiated by injection of anti-CIIantibodies, and animals were treated orally twice dailywith various concentrations of ML120B or vehiclealone, as described above. Vehicle-treated animals de-veloped severe arthritis, with clinical disease symptoms(redness, swelling, ankylosis) appearing on day 5 and

peaking between days 7 and 9 (Figure 2A). Clinicaldisease symptoms were almost completely inhibitedwith 60 mg/kg of ML120B. A small number of animals(2 of 6) developed scores of 0.5 in 1 paw on days 7 and10 with ML120B at this dose level. Mice treated withML120B exhibited a dose-dependent decrease in clinicalsymptoms compared with vehicle-treated animals (Fig-ure 2A). Based on these data, further imaging studiescharacterizing the effects of IKK-2 inhibition were per-formed using the maximally efficacious dose evaluated,i.e., 60 mg/kg.

Histopathologic assessment. We compared nonar-thritic, vehicle-treated, and ML120B-treated animals ondays 6 and 9 after disease initiation. Histologic staining

Figure 1. Up-regulation of NF-�B activity in vivo in the murine arthritis model. The model was generated in transgenic mice expressing theluciferase gene under the control of the NF-�B–inducible promoter, as described in Materials and Methods. A, In vivo bioluminescence imaging.Shown are overlays of photographic and color-coded bioluminescence images of the front and hind paws of nonarthritic control mice (day 0) andarthritic mice on days 3, 7, and 9 after the injection of anti–type II collagen antibodies. B, Quantitative analysis of total bioluminescence signalintensity in front and hind paws. Values are the mean � SEM signal intensity in nonarthritic controls (day 0) and in arthritic mice on days 3, 7, and9, and are representative of 3 independent experiments. Each group of mice consisted of 6–8 animals. � � P � 0.05 versus controls.

120 IZMAILOVA ET AL

Figure 2. Suppression of disease development in the murine antibody-induced arthritis model by administration of the IKK-2 inhibitor ML120B.A, Mean � SEM clinical scores on days 0, 3, 5, 7, and 10 in mice administered anti–type II collagen (anti-CII) antibodies and subsequently treatedwith vehicle or with ML120B twice daily at the doses shown. Each group of mice consisted of 6–8 animals. Values shown are representative of 5independent experiments. �� � P � 0.01 versus vehicle-treated animals. B, Histopathologic assessment, on days 6 and 9 after administration ofanti-CII, of joint sections from representative mice treated with vehicle or with ML120B at 60 mg/kg (toluidine blue–stained; originalmagnification � 10). C, Total pathology scores in the paws of nonarthritic control, vehicle-treated, and ML120B-treated mice on days 6 and 9 afteradministration of anti-CII. Values are the mean and SEM.

RESPONSE TO IKK-2 INHIBITION IN ARTHRITIC MICE 121

of paw sections (Figure 2B) indicated that as early as day6, joints from the vehicle-treated group had moderatecellular infiltration, edema, and minimal growth of pan-nus into the cartilage and subchondral zone. Cellularinfiltrates were composed predominantly of neutrophils,accompanied by fibroblast-like cells and smaller num-bers of lymphocytes and macrophages. The joints dis-played minimal-to-mild chondrocyte loss or collagendisruption, while a few animals (2 of 8) had small areasof bone resorption and rare osteoclasts in affected jointsat this time point. On day 9, animals in the vehicle-treated group showed marked cellular infiltrates andedema, pronounced pannus formation, mild-to-markedchondrocyte and collagen loss, larger areas of boneresorption, and more frequent osteoclasts comparedwith day 6. In contrast, joints from ML120B-treatedanimals were mostly disease free on both day 6 and day

9. A small number of animals (2 of 8 on day 6) showedsome cell infiltration, mild edema, minimal pannusformation, and cartilage loss (Figure 2C). Thus, animalstreated with 60 mg/kg of ML120B twice daily exhibitedsuppressed disease development as judged by his-topathologic as well as clinical features.

Inhibition of NF-�B activity in arthritic paws byML120B treatment. To determine whether ML120Btreatment directly inhibits NF-�B activity in vivo, wemeasured NF-�B activity by quantifying the lumines-cence signal intensity in paws of compound-treatedanimals and compared the findings with those in thevehicle-treated group. Luminescence signal intensityin ML120B-treated animals was significantly lower thanthat in the vehicle-treated animals and similar to thelevels detected in nonarthritic controls (Figure 3A).Moreover, luminescence signal intensity correlated

Figure 3. Suppression of NF-�B activity in vivo in arthritic mice treated with the IKK-2 inhibitor ML120B. A, Overlays of photographic andcolor-coded bioluminescence images of the front and hind paws of nonarthritic control mice, vehicle-treated arthritic mice, and 60 mg/kgML120B–treated mice on days 7 and 9. B, Quantitative analysis of total bioluminescence signal intensity. ��� � P � 0.001 versus vehicle-treatedanimals. Embedded graph shows the total clinical scores of the animals used in the imaging experiments. Values are the mean � SEM.

122 IZMAILOVA ET AL

Figure 4. Suppressed expression of disease-mediating genes in arthritic mice treated with the IKK-2 inhibitor ML120B. Gene expression in the pawsof nonarthritic control mice, vehicle-treated arthritic mice, and 60 mg/kg ML120B–treated mice on days 6 and 9 was measured by quantitative reversetranscriptase–polymerase chain reaction. Values are the mean � SEM relative expression (RE) of each gene. � � P � 0.05; �� � P � 0.01; ��� �P � 0.001, versus vehicle-treated animals. ICAM-1 � intercellular adhesion molecule 1; ENA-78 � epithelial neutrophil–activating peptide 78;TNFa � tumor necrosis factor �; IL-1b � interleukin-1�; iNOS � inducible nitric oxide synthase; COX-2 � cyclooxygenase 2; MMP-3 � matrixmetalloproteinase 3.

RESPONSE TO IKK-2 INHIBITION IN ARTHRITIC MICE 123

with clinical scores during disease progression (r � 0.6,P � 0.001). Thus, these results demonstrate a cor-relation between the inhibition of NF-�B activity inarthritic paws in vivo and clinical disease activity(Figure 3B).

Suppression of the expression of disease-mediating genes by ML120B treatment. The devel-opment of RA is associated with chronic active in-flammation in the synovial tissue and damage toarticular surfaces. Antibody-induced arthritis exhibitsboth periarticular inflammation and cartilage destruc-tion in a subacute to chronic-active setting. To gaininsight into the pathogenic processes blocked byML120B, we examined gene expression in whole-jointhomogenates from nonarthritic control, vehicle-treated,and 60 mg/kg ML120B–treated mice, by quantitativereverse transcriptase–PCR.

First, we examined molecules involved in thedevelopment of inflammation. We found an 11–13-foldincreased induction of ICAM-1 on days 6 and 9 in thepaws of vehicle-treated animals compared with thecontrol group. A similar expression profile was ob-served for ENA-78, KC, and JE, neutrophil and mono-cyte chemoattractants induced by NF-�B (Figure 4).Compared with vehicle-treated animals, ML120B-treated mice exhibited significantly suppressed ICAM-1levels on days 6 (61%) and 9 (86%). Moreover, onboth day 6 and day 9, ML120B-treated animals dis-played significantly inhibited gene expression for chemo-kines KC (88% and 110%), ENA-78 (98% and 99%),and JE (77% and 58%) compared with vehicle-treatedanimals.

We estimated modulation of the composition ofcellular infiltrate in joints via examination of CD3 andCD68 expression, for T cells and macrophages, respec-tively. The expression of CD3 was increased 3.3-fold andCD68 expression was increased 7.2-fold on day 9 in thevehicle-treated animals compared with the controlgroup, indicating increased infiltration of both T cellsand macrophages. Levels of CD3 and CD68 expressionin the joints of ML120B-treated animals were compara-ble with levels in the joints of the nonarthritic controls;this was corroborated histologically by observation ofdecreased cellular infiltrate.

With regard to expression of the inflammationmediators TNF�, IL-1�, IL-6, iNOS, and COX-2, wedetected a modest but significant increase in expressionof TNF� (3-fold; P � 0.008) and COX-2 (2.5-fold; P �0.0004) in the vehicle-treated mice, and these animalsexhibited a dramatic increase in levels of messenger

RNA (mRNA) for IL-1� (11-fold), IL-6 (76-fold), andiNOS (10-fold). ML120B treatment caused a profoundinhibition of all proinflammatory markers on both day 6and day 9 (TNF� 48% and 96%, respectively, IL-1� 91%and 98%, respectively, IL-6 94% and 100%, respectively,iNOS 80% and 66%, respectively, and COX-2 128% and116%, respectively) (Figure 4).

We next investigated whether IKK-2 inhibitionameliorates joint damage via matrix remodeling, byquantifying the expression of matrix-degrading enzymes.Expression of MMP-3 (stromelysin) was elevated up to7.3-fold in the vehicle-treated group. ML120B adminis-tration suppressed MMP-3 expression to levels compa-rable with those in control mice. We also measuredexpression of genes for cysteine proteases and foundthat cathepsin B expression was elevated 3.2-fold onday 6 and 6-fold on day 9 in vehicle-treated mice.Expression of cathepsin K was significantly increasedonly on day 9 (12-fold). ML120B administration signif-icantly inhibited the expression of cathepsin B on bothday 6 and day 9 (73% and 66%, respectively). CathepsinK expression was inhibited by 70% on day 9 (Figure 4).The gene expression profiling results are summarized inTable 1.

Reduction of levels of destructive proteases in thejoint by ML120B treatment. Since we detected increasedmRNA levels for cathepsins B and K in arthritic joints,which were inhibited in the ML120B-treated group, weselected an optical probe activated by cathepsin pro-teases to monitor protease activity in vivo. The Prosense“smart” probe is an enzyme-cleavable fluorescent probespecific for a defined subset of proteases, includingcathepsins B and K. Probe fluorescence remainsquenched when the probe is administered intravenously,until activation by enzyme cleavage in the target tissue(16). The Prosense probe was injected into control,vehicle-treated, and 60 mg/kg ML120B–treated animals,and an in vivo image was captured and fluorescenceintensity in paws quantified (Figure 5A). The fluores-cence signal was increased 4.2-fold in the paws ofvehicle-treated mice on day 6 after the injection ofanti-CII antibodies and 8-fold on day 9 (Figure 5B),which indicates augmented probe enzymatic cleavage inarthritic paws, resulting in fluorescence activation. Theincreased fluorescence signal during disease progressioncorrelated with clinical scores (r � 0.95, P � 0.001)(Figure 5C). The signal intensity was significantly lowerin the animals receiving ML120B and was similar to thelevels found in the control group (Figures 5A and B).

124 IZMAILOVA ET AL

DISCUSSION

Our results demonstrate that the combination ofmolecular imaging methods, gene expression profiling,and histopathologic analysis is a powerful approach tounderstanding the mechanism of action of small mole-cule inhibitors in complex models of disease. We as-sessed the effects of the IKK-2 inhibitor ML120B in amurine arthritis model and demonstrated that, at theefficacious dose, this compound suppressed NF-�B ac-tivity and inhibited mediators of inflammation and jointdestruction.

There are some differences between the patho-genesis of murine antibody-induced arthritis and that ofRA in humans, i.e., the antibody-induced arthritis issubacute while RA is generally chronic with periodicflares. Nevertheless, the known shared features of thecharacteristic inflammatory response to injury renderthe model informative with respect to the pathways ofinterest for potential therapeutic intervention.

Several studies have assessed in vitro NF-�B–DNA binding activity in synovial tissue both from hu-mans and from animals with experimental disease. NF-�B–DNA binding was significantly greater in human RAtissue compared with that from patients with osteo-arthritis, and moreover, was localized to the sites ofmaximum tissue destruction (17). Increased NF-�B–

DNA binding activity has been demonstrated in synovialextracts from rats and mice following disease develop-ment (18,19). The inhibition of NF-�B binding afterIKK-2 inhibitor treatment has also been demonstratedin the mouse collagen-induced arthritis model (19) andthe rat adjuvant-induced arthritis model (18). In vivoimaging enabled us to visualize and quantify targetactivity in the inflamed joints of mice. Our data extendprevious observations by providing in vivo evidence ofNF-�B activity during disease development and its sup-pression in arthritic joints via the administration of theIKK-2 inhibitor ML120B.

NF-�B regulates the expression of multiplegenes involved in arthritis pathogenesis. This regulationmay be direct, via activation of gene transcription, orindirect, via the secondary downstream effects ofNF-�B–regulated gene products. For example, TNF�production is regulated by NF-�B directly via theTNF� promoter, which contains an NF-�B binding site(20,21). NF-�B also regulates the expression of celladhesion molecules and inflammatory cell chemo-attractants (5), and therefore, indirectly regulates cellmigration. Our mRNA expression analyses provide evi-dence of decreased cell infiltration into the joints ofML120B-treated animals, which was confirmed by his-topathologic analyses. One potential mechanism ex-

Table 1. Regulation of gene expression in the antibody-induced arthritis model and suppression by the IKK-2 inhibitorML120B*

Gene

Day 6 Day 9

Mean fold inductionover nonarthritic

controls

Mean %inhibition with

ML120B P†

Mean fold inductionover nonarthritic

controls

Mean %inhibition with

ML120B P†

CD3 2.2 75 �0.05 3.3 89 NSCD68 3 40 �0.01 7.2 87 �0.001ICAM-1 13 61 �0.01 11 86 �0.05JE 40 77 �0.01 29 58 NSKC 8.8 88 �0.001 5.2 110 �0.001ENA-78 22 98 �0.001 37 99 �0.001TNF� 3 48 �0.05 3.4 96 �0.05IL-6 76 94 �0.001 72 100 �0.001IL-1� 11 91 �0.001 10 98 �0.001iNOS 10 80 �0.001 8.4 66 �0.05COX-2 2.5 128 �0.001 3.2 116 �0.001MMP-3 7.3 97 �0.001 3.5 119 �0.05Cathepsin B 3.2 73 �0.001 6 66 �0.01Cathepsin K 1.7 �14 NS 12 70 �0.01

* NS � not significant; ICAM-1 � intercellular adhesion molecule 1; ENA-78 � epithelial neutrophil–activating peptide 78;TNF� � tumor necrosis factor �; IL-6 � interleukin-6; iNOS � inducible nitric oxide synthase; COX-2 � cyclooxygenase 2;MMP-3 � matrix metalloproteinase 3.† Significance of the percent inhibition with ML120B.

RESPONSE TO IKK-2 INHIBITION IN ARTHRITIC MICE 125

Figure 5. Reduced levels of destructive proteases in the joints of arthritic mice treated with the IKK-2 inhibitor ML120B. Protease-activated probewas injected intravenously 24 hours prior to the imaging experiment. A, Superimposition of white light and near-infrared fluorescence images of thepaws of nonarthritic control, vehicle-treated, and 60 mg/kg ML120B–treated mice on days 0, 6, and 9. B, Time course of fluorescence intensity.Embedded graph represents the means of total clinical scores of the animals used in the imaging experiments. Values are the mean � SEM.��� � P � 0.001 versus vehicle-treated animals. C, Correlation between mean total clinical scores and relative fluorescence intensity in arthritic hindpaws 24 hours after injection of Prosense probe injection. Fluorescence intensities in arthritic animals correlated strongly with clinical scores(r � 0.95, P � 0.001).

126 IZMAILOVA ET AL

plaining this observation may be inhibition of chemo-tactic chemokines and/or leukocyte adhesion moleculesin the joint.

Previous reports on the effects of IKK-2 inhibi-tors in animal models were limited to genes directlyregulated by NF-�B (19,22). We investigated the im-pact of IKK-2 inhibition on genes that are both directlyand indirectly regulated by NF-�B in the developmentof experimental arthritis. As expected, we observed theinhibition of several genes that are directly regulatedby NF-�B and involved in cell migration, inflammation,and matrix degradation, including ICAM-1, TNF�,IL-1�, IL-6, KC, ENA-78, JE, iNOS, COX-2, andMMP-3, in the diseased paws of animals treated with theIKK-2 inhibitor. Reports concerning the role of NF-�Bin cathepsin expression are controversial, and it is notclear whether IKK-2 inhibition has a direct or indirecteffect (23,24). However, in vivo IKK-2 inhibition byML120B suppressed the transcription of the cathepsin Band cathepsin K genes. Overall, our data indicate thatIKK-2 inhibition has effects on multiple genes involvedin the inflammatory and destructive components ofdisease.

We observed increased expression of mRNA forseveral proteases, including cathepsins B and K, inarthritic animals. Since reduced expression of thesegenes was found in the ML120B-treated group, wehypothesized that ML120B therapy decreases the levelsof destructive enzymes in joints. To test this, we usedNIRF imaging with a protease-specific probe. We se-lected a probe that is activated by the cysteine proteasecathepsin, because the results of previous studies dem-onstrated that cysteine proteases are up-regulated in RAsynovial tissue and fluid (25). Increased cysteine pro-tease expression is predominantly restricted to the syno-vium at sites of joint damage and can be detected asearly as 2 weeks after the onset of disease symptoms(26). These enzymes directly degrade cartilage and bonematrices. In addition to their destructive activities, theycan activate MMPs (27). Consistent with these findings,we observed a dramatic increase in fluorescence signalintensity in the joints of vehicle-treated animals, whichindicated increased amounts of active proteases. Incontrast, signal intensity in the ML120B-treated groupwas similar to that in the nonarthritic control animals.

The reduction of protease-activated fluorescencesignal in response to methotrexate treatment has beenobserved in the murine collagen-induced arthritis model(16). However, that study did not include analysis of thecorrelation between the decrease in protease-activated

fluorescence signal and clinical disease indices such asredness, swelling, and ankylosis. In the present study wedemonstrated that protease-activated NIRF imagingprobes can be used as sensitive biomarkers of subacuteto chronic active joint disease activity and treatmentefficacy in a murine antibody-induced arthritis model.We observed complete disease suppression with theIKK-2 inhibitor at the efficacious concentration used inthis study. Current efforts are aimed at testing theprognostic utility of the Prosense probe for predictingclinical and histopathologic disease amelioration.

In conclusion, we have demonstrated, using invivo imaging and gene expression profiling, that theML120B compound offers protection from inflamma-tion and joint destruction in subacute to chronic activemurine antibody-induced arthritis. Moreover, weshowed a direct correlation between the drug’s disease-modifying effects and biochemical target activity. Thesestudies highlight how gene expression profiling can beimplemented to identify surrogate biomarkers of diseaseactivity and treatment response in experimental modelsof arthritis.

ACKNOWLEDGMENTS

The authors would like to thank K. Anderson and E.Grant for technical assistance, and A. Parker and C. Fraser forcritical review of the manuscript.

AUTHOR CONTRIBUTIONS

Dr. Izmailova had full access to all of the data in the study andtakes responsibility for the integrity of the data and the accuracy of thedata analysis.Study design. Drs. Izmailova, Alencar, Chun, Schopf, Xu, Ocain,Weissleder, Mahmood, Healy, and Jaffee.Acquisition of data. Dr. Izmailova, Ms Paz, and Drs. Alencar, Mah-mood, and Jaffee.Analysis and interpretation of data. Drs. Izmailova, Alencar, Schopf,Lane, Xu, Ocain, Mahmood, Healy, and Jaffee.Manuscript preparation. Dr. Izmailova, Ms Paz, and Drs. Alencar,Schopf, Lane, Ocain, Mahmood, Healy, and Jaffee.Statistical analysis. Dr. Izmailova, Ms Paz, and Dr. Jaffee.Medicinal chemistry. Dr. Hepperle.Invention of chemical matter, progression of key compound. Dr.Harriman.Project leadership. Dr. Xu.Manuscript editing. Dr. Weissleder.

REFERENCES

1. Ghosh S, Karin M. Missing pieces in the NF-�B puzzle. Cell2002;109 Suppl:S81–96.

2. Tak PP, Firestein GS. NF-�B: a key role in inflammatory diseases.J Clin Invest 2001;107:7–11.

3. Van den Berg WB, van Riel PL. Uncoupling of inflammation and

RESPONSE TO IKK-2 INHIBITION IN ARTHRITIC MICE 127

destruction in rheumatoid arthritis: myth or reality? [editorial].Arthritis Rheum 2005;52:995–9.

4. Smolen JS, Han C, Bala M, Maini RN, Kalden JR, van der HeijdeD, et al, for the ATTRACT Study Group. Evidence of radio-graphic benefit of treatment with infliximab plus methotrexate inrheumatoid arthritis patients who had no clinical improvement: adetailed subanalysis of data from the Anti–Tumor Necrosis FactorTrial in Rheumatoid Arthritis with Concomitant Therapy study.Arthritis Rheum 2005;52:1020–30.

5. Pahl HL. Activators and target genes of Rel/NF-�B transcriptionfactors. Oncogene 1999;18:6853–66.

6. Bresnihan B. Pathogenesis of joint damage in rheumatoid arthritis.J Rheumatol 1999;26:717–9.

7. Tak PP, Bresnihan B. The pathogenesis and prevention of jointdamage in rheumatoid arthritis: advances from synovial biopsy andtissue analysis [review]. Arthritis Rheum 2000;43:2619–33.

8. Trabandt A, Gay RE, Fassbender HG, Gay S. Cathepsin B insynovial cells at the site of joint destruction in rheumatoid arthritis.Arthritis Rheum 1991;34:1444–51.

9. Murphy G, Knauper V, Atkinson S, Butler G, English W, HuttonM, et al. Matrix metalloproteinases in arthritic disease. ArthritisRes 2002;4 Suppl 3:S39–49.

10. Mengshol JA, Mix KS, Brinckerhoff CE. Matrix metalloprotein-ases as therapeutic targets in arthritic diseases: bull’s-eye ormissing the mark? [review]. Arthritis Rheum 2002;46:13–20.

11. Hansen T, Petrow PK, Gaumann A, Keyszer GM, Eysel P, EckardtA, et al. Cathepsin B and its endogenous inhibitor cystatin C inrheumatoid arthritis synovium. J Rheumatol 2000;27:859–65.

12. Han Z, Boyle DL, Manning AM, Firestein GS. AP-1 and NF-�Bregulation in rheumatoid arthritis and murine collagen-inducedarthritis. Autoimmunity 1998;28:197–208.

13. Rudin M, Weissleder R. Molecular imaging in drug discovery anddevelopment. Nat Rev Drug Discov 2003;2:123–31.

14. Nagashima K, Sasseville VG, Wen D, Bielecki A, Yang H,Simpson C, et al. Rapid TNFR1 dependent lymphocyte depletionin vivo with a selective chemical inhibitor of IKK�. Blood 2006;107:4266–73.

15. Carlsen H, Moskaug JO, Fromm SH, Blomhoff R. In vivo imagingof NF-�B activity. J Immunol 2002;168:1441–6.

16. Wunder A, Tung CH, Muller-Ladner U, Weissleder R, MahmoodU. In vivo imaging of protease activity in arthritis: a novelapproach for monitoring treatment response. Arthritis Rheum2004;50:2459–65.

17. Benito MJ, Murphy E, Murphy EP, van den Berg WB, FitzGeraldO, Bresnihan B. Increased synovial tissue NF-�B1 expression atsites adjacent to the cartilage–pannus junction in rheumatoidarthritis. Arthritis Rheum 2004;50:1781–87.

18. Schopf L, Savinainen S, Anderson K, Kujawa J, DuPont M, SilvaM, et al. IKK� inhibition protects against bone and cartilagedestruction in a rat model of rheumatoid arthritis. ArthritisRheum 2006;54:3163–73.

19. Podolin PL, Callahan JF, Bolognese BJ, Li YH, Carlson K, DavisTG, et al. Attenuation of murine collagen-induced arthritis by anovel, potent, selective small molecule inhibitor of I�B kinase 2,TPCA-1 (2-[(aminocarbonyl)amino]-5-(4-fluorophenyl)-3-thiophene-carboxamide), occurs via reduction of proinflammatory cytokinesand antigen-induced T cell proliferation. J Pharmacol Exp Ther2005;312:373–81.

20. Ghosh S, May MJ, Kopp EB. NF-� B and Rel proteins: evolution-arily conserved mediators of immune responses. Annu Rev Immu-nol 1998;16:225–60.

21. Baldwin AS Jr. The NF-� B and I � B proteins: new discoveriesand insights. Annu Rev Immunol 1996;14:649–83.

22. McIntyre KW, Shuster DJ, Gillooly KM, Warrier RR, Connaugh-ton SE, Hall LB, et al. Reduced incidence and severity ofcollagen-induced arthritis in interleukin-12-deficient mice. EurJ Immunol 1996;26:2933–8.

23. Li Q, Bever CT Jr. Effect of protein kinase modulators on theregulation of cathepsin B activity in THP-1 human monocyticleukemia cells. Oncol Rep 1998;5:227–33.

24. Liuzzo JP, Petanceska SS, Moscatelli D, Devi LA. Inflammatorymediators regulate cathepsin S in macrophages and microglia: arole in attenuating heparan sulfate interactions. Mol Med 1999;5:320–33.

25. Maciewicz RA, Wotton SF. Degradation of cartilage matrix com-ponents by the cysteine proteinases, cathepsins B and L. BiomedBiochim Acta 1991;50:561–4.

26. Cunnane G, FitzGerald O, Hummel KM, Youssef PP, Gay RE,Gay S, et al. Synovial tissue protease gene expression and jointerosions in early rheumatoid arthritis. Arthritis Rheum 2001;44:1744–53.

27. Eeckhout Y, Vaes G. Further studies on the activation of procolla-genase, the latent precursor of bone collagenase: effects of lyso-somal cathepsin B, plasmin and kallikrein, and spontaneousactivation. Biochem J 1977;166:21–31.

128 IZMAILOVA ET AL