2.1.

Animals and Arthritis Model

Animals were maintained in accordance with the Guide for the Care and Use of Laboratory Animals published by the US National Institute of Health (NIH publication No. 85–23, revised 1996), and examinations were approved by the responsible authority. Female Lewis rats (body weight 140 to 160 g) were purchased from Charles River Laboratories (Sulzfeld, Germany). The study was performed in 15 rats (30 joints) with collagen-induced arthritis (CIA) and seven control animals with healthy joints (14 joints).

For induction of CIA, bovine type II collagen (Collagen Acid Soluble, from Bovine Nasal Septum, Sigma–Aldrich Chemie GmbH, Steinheim, Germany) was dissolved in 0.01 M acetic acid. Subsequently, the solution was emulsified with an identical volume of incomplete Freund’s adjuvant (Sigma–Aldrich Chemie GmbH, Steinheim, Germany) for 30 min in a well-closed container placed in an ultrasonic bath (35 kHz) (Sonorex RK 31, Bandelin, Berlin, Germany) filled with ice water.

For collagen injection the rats were immobilized by inhalation anesthesia with 2% isoflurane (Forene, Abbott GmbH, Wiesbaden, Germany) and medical oxygen. A dorsal area at the base of the tail was shaved and the skin disinfected. In 15 rats, 0.5 ml (containing 0.5 mg collagen) of the emulsion was intracutaneously injected at the base of the tail at different sites (each about 0.05 ml). The seven animals serving as controls were injected with 0.5 ml physiologic salt solution under identical conditions. The procedure was repeated in all rats after seven days.

Arthritis develops 13 to 15 days after the first collagen injection, but is highly variable in severity. Animals may show different degrees of arthritis in the right and left ankle joint, or there may be animals without signs of clinical or histological changes in one or both joints.

2.2.

Clinical Evaluation (Score)

After the second collagen injection, the rats gait and the distal joints, especially the tibiotarsal joints, were inspected for swelling and reddening daily for describing the developing of arthritis. Before NIRF imaging each ankle joint was given a score from 0 to 3 ($0=\text{no clinical symptoms}$, $1=\text{mild swelling}$, $2=\text{moderate swelling}$, and $3=\text{severe swelling}$ with reduced mobility of the tibiotarsal joint).

2.3.

NIRF Imaging

NIRF imaging of ankle joints was performed at the onset of clinical symptoms of arthritis. For examination, the rats were anesthetized by a combination of inhalation and injection anesthesia. First they were immobilized with isoflurane followed by subcutaneous injection of $80\mathrm{mg}/\mathrm{kg}$ ketamine hydrochloride (Curamed Pharma GmbH, Karlsruhe, Germany) and $4\mathrm{mg}/\mathrm{kg}$ xylazine (Rompun 2%, Bayer Vital GmbH, Leverkusen, Germany) in a mixing syringe and $3.5\mathrm{mg}/\mathrm{kg}$ of diazepam (Diazepam ratiopharm 10 solution for injection, ratiopharm GmbH, Ulm, Germany) subcutaneously. An indwelling cannula (Vasocan Braunüle, $0.9\mathrm{mm}\text{diameter}\times 25\mathrm{mm}$, B.Braun AG, Melsungen, Germany) was placed in the tail vein (lateral coccygeal vein) and connected with an automatic injector (P4000, IVAC/Alaris Medical Systems Inc., San Diego, USA) for the dye injection. The rats were placed in prone position and the hind legs extended caudolaterally. Hair was not removed since the ankle region is nearly not covered by fur and any manipulation that might induce hyperemia should be avoided.

Fluorescence images were obtained in reflection geometry. NIR dye fluorescence was excited with an optical parametric oscillator (OPO) (GWU-Lasertechnik, Erftstadt, Germany) based on an optically nonlinear $\mathrm{Ba}{\mathrm{B}}_2{\mathrm{O}}_4$-crystal, pumped by the third harmonic ($\lambda =355\mathrm{nm}$,$E_{\text{pulse}}=80\mathrm{mJ}$) of a Q-switched $\mathrm{Nd}:\mathrm{YAG}$ laser (GCR–10, Spectra-Physics Inc., USA). The OPO provides laser radiation tuneable between 415 nm and 2.2 μm and was set to ${\lambda }_{\mathrm{ex}}=740\mathrm{nm}$ in our experiments to excite NIR fluorescence of the dye under investigation. Two long-pass interference filters (${\lambda }_{50\%}=800\mathrm{nm}$) and a long-pass edge filter (${\lambda }_{50\%}=780\mathrm{nm}$, 3 mm) were used to cut off scattered excitation light from emitted fluorescence light. The cut-off wavelength at $\lambda =860\mathrm{nm}$ results from a sensitivity drop of the detector. The fluorescence was imaged onto the photocathode of a water/Peltier-cooled, intensified CCD camera (Model ICCD-576, Princeton Instruments Inc, Trenton, USA) with the help of a standard object lens of 50 mm focal length ($f=1.4$). In order to suppress background signals caused by ambient light, the intensifier of the CCD camera was gated applying an electrical pulse ($-180\mathrm{V}$) of about 10 ns duration derived from a HV pulse generator. The HV pulse generator was synchronized with an advanced trigger pulse provided by the power supply of the laser system and appropriately delayed by means of a digital delay generator. The delay of the gating pulse was adjusted to image the fluorescence of the cyanine dye (decay time $\sim 300\mathrm{ps}$) and to suppress long-living autofluorescence components with typical decay times of 3 ns. Fluorescence was recorded using a short exposure time (0.2 s) and 4 times accumulation to avoid motion artifacts. For close monitoring of dye uptake immediately after injection, 180 images were recorded within 660 s (i.e., one image every 3.65 s), followed by time points at 30, 60, and 120 min post injection.

A stable fluorescent cube consisting of covalently bound cyanine dyes embedded in plastic material placed in the top right corner of the measuring field served as reference for fluorescence intensity.

2.4.

NIRF Dye

The nonspecific NIRF dye, TSC, is a tetrasulfonated carbocyanine dye based on an indotricarbocyanine chromophore. TSC is a low-molecular-weight dye ($836.9\mathrm{g}/\mathrm{mol}$) with both anionic and hydrophilic properties.^23,30 In phosphate-buffered saline TSC shows an absorption maximum at ${\lambda }_{\mathrm{abs}}=755\mathrm{nm}$ and a fluorescence emission maximum at ${\lambda }_{\mathrm{em}}=778\mathrm{nm}$.²³ TSC has a plasma binding of about 74% and in rats excretion is mostly (70%) hepatobiliary in feces.²³ The dye was dissolved in physiological saline and injected as an intravenous bolus into the tail vein at a dose of $1\mathrm{\mu mol}/\mathrm{kg}$ and a rate of $4\mathrm{ml}/\min$. The fluorescent contrast agent was administered at the start of the third frame.

2.5.

NIRF Imaging Data Analysis

The acquired data were transferred to a desktop computer and stored. The first two frames of the sequence served as background frames. Circular regions of interest (ROIs) with a constant diameter of 20 pixels were placed over the reference cube in the top right corner and the left and right ankle joints with the center of the ROI positioned over the external malleolus by one reader (BE) blinded to MRI results, histopathological and clinical grading.

Mean fluorescence intensities $I_F$ were the average of all fluorescence intensity values of pixels covered by the ROI of a particular region under investigation divided by the number of pixels. All fluorescence intensities were normalized ($I_{\mathrm{NF}}$) to the fluorescence of the reference mean $I_F(\text{reference})$ to obtain relative values using the following equation:

Contrast enhancement was determined by calculating $I_{\mathrm{NF}}$ at 5 and 10 min after dye administration and additionally after 30, 60, and 120 min.

2.6.

MR Imaging

The rats were examined 5 h after NIRF imaging on a 1.5 T whole-body MR scanner (Magnetom Sonata, Siemens, Erlangen, Germany) using an extremity coil (Siemens, Erlangen, Germany). The rats were placed in prone position with the hind legs extended caudolaterally on a Styrofoam support and positioned on a horizontal line in the center of the coil. MR images were acquired with a T1-weighted spin-echo (SE) sequence in coronal slice orientation using the following sequence parameters: echo time 13 ms, repetition time 350 ms, field of view $160\times 144{\mathrm{mm}}^2$, matrix $173\times 320$, slice thickness 2 mm, and number of slices 13. The total acquisition time was 4 min and 13 s.

Following acquisition of nonenhanced MR images, the gadolinium-based contrast agent gadodiamide (Omniscan, Amersham Buchler GmbH & CO. KG, Braunschweig, Germany) was injected as a bolus into the tail vein at a dose of $0.2\mathrm{mmol}\mathrm{Gd}/\mathrm{kg}$. Acquisition of contrast-enhanced images was started at the same slice position about 3 min after administration of contrast agent.

Due to the start of image acquisition about 3 min after contrast medium administration and the acquisition time, the MRI data values were identified as 5 min value.

2.7.

Analysis of MR Images

MRI always has a morphologic background; therefore, it is necessary to include the precontrast signal intensity in analyzing MRI contrast medium enhancement. The MR images were analyzed by one reader (DP) blinded to time point of acquisition, NIRF imaging findings, histopathological, and clinical scores. The reader determined relative signal intensities (SI) [signal-to-noise ratio (SNR)] in the ankle joints before and after administration of contrast agent using the software ImageJ (National Institutes of Health, Maryland, USA). SI of the left and right ankle joints were measured in circular ROIs with a constant diameter and the center positioned in the same way as for analysis of NIRF images. SI of the background was measured outside the animal. SNRs of the ankle joints before (${\mathrm{SNR}}_{\mathrm{pre}}$) and after contrast enhancement (${\mathrm{SNR}}_{\text{post}}$) were calculated as ratios of SI in the joints to background intensities. The percent signal intensity increase ($\mathrm{\Delta }\mathrm{SI}\%$) in ankle joints after contrast enhancement was then calculated according to the following formula:

2.8.

Histology

The animals were euthanized in deep anesthesia by intravenous injection of T61 ad us. vet. (Intervet Deutschland GmbH, Unterschleißheim, Germany) at a dose of $2\mathrm{ml}/\text{animal}$ immediately after MRI.

For histological workup, the hind legs of the animals were removed and fixed in 4% buffered formaldehyde solution (Mallinckrodt Baker, Deventer, Holland). Subsequently, the specimens were placed in EDTA decalcifying solution (Herbeta Arzneimittel, Berlin, Germany) for 5 weeks at 60°C. The solution was changed every week. The decalcified hind legs were embedded in paraffin, thereafter about 4-μm thick histological sections were cut and stained with hematoxylin and eosin (H&E).

The histopathological specimens (Fig. 1) were assessed for signs of arthritis using the following criteria: synovitis, periarthritis, tenosynovitis, periostitis, and cartilage/bone destruction. Each criterion was graded semiquantitatively with predetermined parameter values by one reader (IG) blinded to NIRF and MRI values as well as clinical score on the basis of a standardized human synovitis scale,³¹ which distinguishes four degrees of severity: 0 (absence of criterion), 1 (mild), 2 (moderate), and 3 (severe). The sum score was graded as follows: 0, no arthritis; 1 to 5, low-grade arthritis; 6 to 10, moderate arthritis; and 11 to 15 high-grade arthritis.

2.9.

Statistical Analysis

To compare different groups the medians (50th percentile) were calculated and tests of descriptive statistics were used.

For analysis of NIRF and MRI data, the whole group of animals with arthritis and the control group were tested for significant differences in $I_{\mathrm{NF}}$, signal intensity, and signal increase at different time points given in minutes (NIRF: 5, 10, 30, 60, 120; MRI: 5). Additionally, the animals with arthritis were classified into three different groups (low-grade, moderate, and high-grade) based on histopathological scores. SI and signal increases, respectively, of both imaging modalities were compared between these subgroups and with controls at time points given above. Differences between any two CIA groups or controls were tested for significance using the nonparametric Mann-Whitney U-test. Significance was assumed at $p<0.05$. Differences in $I_{\mathrm{NF}}$ at different time points and differences in signal increases between all CIA groups and controls were tested for significance using the nonparametric Kruskal-Wallis analysis. A very highly significant difference was assumed at $p<0.001$, a highly significant difference at $p<0.01$.

Associations between clinical parameters, histology, and the two imaging modalities were identified by calculating Spearman correlation coefficients ($r_s$). Significance was assumed at $p<0.05$. Statistical tests were performed using SPSS version 13.0 (SPSS Inc., Chicago, USA).

Sensitivity and specificity for the detection of arthritic joints were calculated. Sensitivity gives the percentage of correctly identified arthritic joints (number of arthritic joints with NIRF value higher than discriminator related to number of all arthritic joints). Specificity measures the proportion of healthy joints correctly identified by NIRF imaging (number of control joints with NIRF value lower than discriminator related to number of all control joints).

ROC curve analysis was used to calculate the cut-off value with the highest number of samples correctly classified (true positives plus true negatives).

$I_{\mathrm{NF}}$ at the different time points were graphically represented in box-and-whisker plots. Outliers are indicated by open circles, extremes by asterisks.

3.1.

NIRF Imaging

In pre-injection NIRF images, both control joints and arthritic joints showed very low $I_{\mathrm{NF}}$ (0.045 and 0.05) without differences between the groups.

Five minutes after dye administration, all $I_{\mathrm{NF}}$ values were significantly higher than the corresponding intensities before administration ($p<0.05$), regardless of the degree of arthritis.

The $I_{\mathrm{NF}}$ of control joints and joints from CIA rats were compared. The group of arthritic joints showed statistically significant higher $I_{\mathrm{NF}}$ values than the control joints 5, 10, 30, and 60 min after dye administration ($p<0.05$). The difference in $I_{\mathrm{NF}}$ between the group of arthritic joints and controls was no longer significant after 120 min (Figs. 2 and 3).

Fig. 1

H&E-stained histological sections of a control joint (left, 1 and 2) and an arthritic joint (right, 3 and 4). 1 and 3: (1) Overview of a tibiotarsal joint in a control rat and (3) in an arthritic joint showing (A) tibia, (B) talus, (C) calcaneus, and (D) tarsal bone. 2: Detail of tibiotarsal joint in a control animal showing (E) the joint space with synovial membrane and (F) the articular cartilage. 4: Detail of tibiotarsal joint of an arthritic joint showing (E) proliferating synovial membrane filling the joint space and covering the destroyed cartilage.

Fig. 2

Box-and-whisker plots of normalized fluorescence intensities $I_{\mathrm{NF}}$ for all groups of joints investigated at different time points. The $I_{\mathrm{NF}}$ values of the joints with moderate and high-grade arthritis are markedly above the values of the control joints between 5 and 60 min after dye injection ($p<0.05$). The $I_{\mathrm{NF}}$ values of joints with low-grade arthritis are only slightly above those of the controls ($p>0.05$). The difference in $I_{\mathrm{NF}}$ values between the three grades of arthritis is especially pronounced 5 min after dye administration. The differences between all groups were very highly significant between 5 and 30 min after dye administration ($p<0.001$), highly significant at 60 min ($p=0.001$), and nonsignificant at 120 min and before dye administration. Outliers are denoted by open circles, extremes by asterisks.

Fig. 3

Fluorescence images in false colors (range 0 to 100,000 counts for all color bars) before and after administration of $1\mathrm{\mu mol}/\mathrm{kg}$ TSC. The time course is shown in a rat with high-grade CIA on the left and a healthy control rat on the right. The fluorescent cube used as reference is seen in the upper right corner of the images. Red color (high fluorescence intensity) indicates areas of inflammation. The high signal intensities in the back region are caused by shaving and skin inflammation due to the intradermal injections. The hot spots in the ankle joints of the CIA rat (left) 1 and 10 min after injection clearly indicate arthritis, while the control joints (right) appear homogeneous. The fluorescence intensities decrease with the time elapsed since dye injection. Two hours after injection there is only very weak relative fluorescence intensity visible.

In all three arthritis subgroups, the highest $I_{\mathrm{NF}}$ was found 5 and 10 min after dye administration. The highest ratio of median $I_{\mathrm{NF}}$ of all arthritic joints to all control joints was obtained 10 min after dye administration [$\text{median}{I}_{\mathrm{NF}}(\text{arthritis})/\text{median}{I}_{\mathrm{NF}}(\text{control})=4.76/2.58=1.84$]. The relative fluorescence intensity showed a marked decrease for all groups at 30 min and a further decrease at 120 min after dye administration (Fig. 2).

In more detail, the median $I_{\mathrm{NF}}$ of the animals with low-grade arthritis was above those of the controls at all time points (Fig. 2), but the difference was not statistically significant. Furthermore, there is no significant difference between moderate and high-grade arthritis to all time points.

In contrast, the median $I_{\mathrm{NF}}$ of animals with moderate arthritis and the ones with high-grade arthritis differ significantly from those of animals with no or with low-grade arthritis for all time points except for 120 min (Fig. 2). The highest ratio (2.14) between the median $I_{\mathrm{NF}}$ of moderate and high-grade arthritis joints versus the median $I_{\mathrm{NF}}$ of controls was reached 10 min after dye administration.

Kruskal-Wallis analysis was used to test differences between the three groups (low-grade, moderate, and high-grade) and controls at different time points after dye administration. Very highly significant differences ($p<0.001$) were seen 5, 10, and 30 min after dye administration; no significant differences were found before and 120 min after dye administration.

Using the relative fluorescence intensities of all groups determined at 5 min and 10 min after dye administration in ROC curve analysis we identified 3.07 at 5 min and 3.15 at 10 min as the cut-off values with the highest numbers of samples correctly classified (85%). Using these values, we achieved a sensitivity of 84.6% and specificity of 85.7% at both time points.

Using the discriminator of 3.5 at 5 and 10 min after dye administration, the calculated sensitivity was 77% for both time points, while specificity was 93% and 100%, respectively, for classifying a joint as arthritic regardless of arthritis grades. Using the same discriminator for joints with moderate and high-grade arthritis, sensitivity was 89% with specificity of 81% and 85%, respectively.

With use of a higher discriminator, specificity increases while sensitivity decreases. For instance, with a discriminator of 5, moderate or high-grade arthritis is detected with a specificity of 100% at 5 and 10 min while sensitivity is only 67%.

3.2.

MR Imaging

In pre-injection MR images, there were significant differences in SI between control joints and the group of all arthritic joints ($p<0.05$) but no significant differences in SI between low-grade, moderate, and high-grade arthritis.

Five minutes after administration of the contrast agent, MRI SNRs were significantly higher than the corresponding intensities before administration ($p<0.05$) regardless of the degree of arthritis.

The group of all arthritic ankle joints had higher SNR on post-contrast images than the control group (Fig. 4), and the percentage signal increase was also significantly higher.

Fig. 4

T1-weighted spin echo MR images of the ankle joints in a rat with bilateral high-grade arthritis (left) and a control (right). Upper row: before administration of contrast agent; lower row: after administration of contrast agent. There is marked enhancement in the area of the ankle joints in the rat with high-grade arthritis (arrows) while there is no signal increase after administration of contrast agent in the control rat.

The post-contrast SNR and the percentage signal increase in rats with low-grade arthritis did not differ significantly from controls. Moderate and high-grade arthritic joints differed significantly from low-grade arthritic joints and controls (Fig. 5).

Fig. 5

Increase in $\mathrm{\Delta }\mathrm{SI}\%$ in controls and the three arthritis groups. Kruskal-Wallis analysis yielded very highly significant differences ($p<0.001$) between all groups. The $\mathrm{\Delta }\mathrm{SI}\%$ values in moderate and high-grade arthritis are markedly higher than in low-grade arthritis and control animals. The values $\mathrm{\Delta }\mathrm{SI}\%$ of the rats with low-grade arthritis are only slightly higher than the values of the control animals. The open circle indicates an outlier.

Kruskal-Wallis analysis yielded a very highly significant difference ($p<0.001$) in $\mathrm{\Delta }\mathrm{SI}\%$ between the three arthritis groups and controls.

The $\mathrm{\Delta }\mathrm{SI}\%$ of 59 is the cut-off value with the highest number of samples correctly classified (87.5%). This cut-off yielded 88.5% sensitivity and 85.7% specificity.

Using the discriminator of $\mathrm{\Delta }\mathrm{SI}\%$ 80 at 5 min after administration, we calculated a sensitivity of 77% and a specificity of 100%, respectively, for classifying a joint as arthritic regardless of arthritis grade.

It is worth noting that there is no significant difference between moderate and high-grade arthritis in terms of post-contrast SNR or percent signal increases. The sensitivity and specificity for the groups of high-grade and moderate arthritis are 100% and 91%, respectively, when using $\mathrm{\Delta }\mathrm{SI}\%=80$ as discriminator between controls and inflamed joints.

3.3.

Histology

For comparative analysis of $I_{\mathrm{NF}}$ (NIRF) and SNR (MRI), histology was performed for joints with collagen injection (30 joints in 15 animals) and for controls (14 joints in 7 animals).

None of the 14 control joints showed histologic evidence of arthritis.

In the group with collagen injection (30 joints), four joints in four different animals did not develop arthritis and were excluded from statistical analysis.

In the remaining 26 joints, which developed histologic changes after collagen-injection, the changes were classified as low-grade arthritis in eight joints, moderate arthritis in eleven joints, and high-grade arthritis in seven joints.

3.4.

Clinical Findings

There were no clinical signs of arthritis in any of the 7 control animals (14 joints).

In the animals with collagen injection, we observed initial clinical symptoms in the form of unilateral or bilateral swelling of the ankle joints on days 13 to 15 after the first collagen injection. Out of 30 joints from 15 animals, 7 joints showed no clinical symptoms, 4 joints showed mild swelling, 6 joints moderate swelling, and 13 joints severe swelling.

3.5.

Correlation

Relative fluorescence intensity $I_{\mathrm{NF}}$ 5 min after dye administration showed high correlation with both the histopathological scores ($r_s=0.742$) (Fig. 6) and the clinical scores ($r_s=0.741$) of the individual joints. The percentage signal increase on MRI also correlated well with the histopathological and clinical scores ($r_s=0.816$, $r_s=0.761$). Correlations were also high for the two imaging modalities, NIRF and MRI ($r_s=0.685$), as well as for clinical and histological assessment of arthritis ($r_s=0.874$).

Fig. 6

Scatter diagram of the correlation between normalized fluorescence intensity of the ankle joint 5 min after injection of TSC versus the histopathological score ($r_s=0.74$, $p<0.001$).