In‐vivo quantitative assessment of the therapeutic response in a mouse model of collagen‐induced arthritis using ¹⁸F‐fluorodeoxyglucose positron emission tomography

Summary

Mouse collagen‐induced arthritis (CIA) is the most commonly used animal model to investigate underlying pathogenesis of autoimmune arthritis and to demonstrate the therapeutic efficacy of novel drugs in autoimmune arthritis. The conventional read‐outs of CIA are clinical score and histopathology, which have several limitations, including (i) subjected to observer bias; and (ii) longitudinal therapeutic efficacy of a new drug cannot be determined. Thus, a robust, non‐invasive, in‐vivo drug screening tool is currently an unmet need. Here we have assessed the utility of ¹⁸F‐fluorodeoxyglucose positron emission tomography (¹⁸F‐FDG) as an in‐vivo screening tool for anti‐inflammatory drugs using the mouse CIA model. The radiotracer ¹⁸F‐FDG and a PET scanner were employed to monitor CIA disease activity before and after murine anti‐tumour necrosis factor (TNF)‐α antibody (CNTO5048) therapy in the mouse CIA model. Radiotracer concentration was derived from PET images for individual limb joints and on a per‐limb basis, and Spearman's correlation coefficient (ρ) was determined with clinical score and histology of the affected limbs. CNTO5048 improved arthritis efficiently, as evidenced by clinical score and histopathology. PET showed an increased uptake of ¹⁸F‐FDG with the progression of the disease and a significant decrease in the post‐treatment group. ¹⁸F‐FDG uptake patterns showed a strong correlation with clinical score (ρ = 0·71, P < 0·05) and histopathology (ρ = 0·76, P < 0·05). This study demonstrates the potential of ¹⁸F‐FDG PET as a tool for in‐vivo drug screening for inflammatory arthritis and to monitor the therapeutic effects in a longitudinal setting.

Introduction

Collagen‐induced arthritis (CIA) is used widely to evaluate the therapeutic potential of novel anti‐inflammatory drugs. The conventional read‐outs of this model are clinical and histological scores, the latter being considered the gold standard 1, 2, 3. These read‐outs have many limitations, including: (i) longitudinal studies in the same mouse cannot be performed; (ii) clinical and histological scores are subjected to observer bias and are semi‐quantitative; and (iii) in‐vivo cellular events cannot be captured in its native environment. To alleviate these limitations, and to make the preclinical drug screening more robust, reproducible and efficient, a validated drug screening tool is currently an unmet need.

The application of ¹⁸F‐fluorodeoxyglucose positron emission tomography (¹⁸F‐FDG PET) in inflammatory arthritis is in its infancy, compared to the modality's advancement in oncology 4, 5. Recent studies have suggested the potential of PET for evaluating inflammatory arthritis in a longitudinal manner in the in‐vivo environment 5, 6, 7. In continuation of our previous work, to meet these unmet needs and to have a quantitative tool to directly measure the degree of inflammation, herein we have evaluated the utility of ¹⁸F‐FDG PET in the CIA model, which closely resembles human autoimmune arthritis, and shares pathological features of autoimmune arthritis such as rheumatoid arthritis 8, 9.

Materials and Methods

Induction of CIA

The study was approved by the University of California Davis Institutional Animal Care and Use Committee. Arthritis was induced (n = 30) using type II bovine collagen (CII) 5, 8, 10 in eight‐week‐old arthritis‐susceptible male/1J mice (Jackson Laboratory, ME, USA). Briefly, 100 μg bovine CII emulsified in 100 μL of complete Freund's adjuvant (CFA) was injected intradermally into the proximal tail of male DBA/1J mice, followed by a second intradermal injection on day 21 with emulsified incomplete Freund's adjuvant. Twenty‐five mice showed signs of arthritis and were included in this study.

Anti‐tumour necrosis factor (TNF)‐α antibody (CNTO5048) treatment

Rat/mouse chimeric monoclonal antibody specific for mouse TNF‐α was a gift from Janssen Research and Development (Spring House, PA, USA). Each mouse received 300 μg of CNTO5048 dissolved in sterile phosphate‐buffered saline (PBS) intraperitoneally every 2 days (morning) for 10 days, starting on day 45 after the first dose of CII 11. Mice were allocated randomly to treatment (n = 10) and vehicle control groups (sterile PBS, untreated, n = 5). On day 45, clinical score, histological score and ¹⁸F‐FDG PET uptake were recorded before injecting CNTO5048 (pretreatment).

¹⁸F‐FDG PET scan

Mice were scanned using ¹⁸F‐FDG PET at days 0 (n = 20), 28 (n = 20), 45 (n = 16) and 56 (n = 12). Briefly, mice were anaesthetized followed by intravenous injection of 7·4 MBq ¹⁸F‐FDG and scanned sequentially starting at 30 min post‐injection on small‐animal PET (D‐PET; Siemens Healthcare, Knoxville, TN, USA). PET images were reconstructed using the ‘fastMAP’ method, as it provides higher recovery coefficients, a lower spillover fraction and optimized quantitative accuracy compared to previous studies 6, 7, 12. Regions of interest (ROI) were drawn at carpal and tarsal joints of each paw. The maximum ¹⁸F‐FDG uptake (standardized uptake value normalized by body weight) per ROI was determined for each major joint to generate a comprehensive PET score (PS). PS in the most affected joint, as determined by clinical score in each animal, was used for statistical analysis.

Clinical score (CS)

Before PET scan, the severity of the arthritis was scored using clinical score for each limb, as described in an earlier report 5. Due to the asymmetrical nature of the disease, CS of the most inflamed joint in each animal was considered for statistical analysis.

Histological score (HS)

After PET scan, mice were killed on days 0 (n = 6), 28 (n = 6), 45 (n = 4) and 56 (n = 4) for histopathological analyses of the affected limbs. Limbs were dissected and processed in 10% formalin followed by decalcification. Subsequently, tissue sections were made and stained with haematoxylin and eosin (H&E) to evaluate the severity of the inflammatory infiltrates, which was graded on a scale of 0–3, where 0 = normal, 1 = mild, 2 = moderate and 3 = severe 5. This finding is represented as histological score (HS) for each inflamed joint (ankle joint). The joint which clinically showed maximal inflammation was chosen for histopathological analysis, and statistical tests were performed with those data.

Statistical analysis

CS, HS and PS are presented as median with range. The non‐parametric Mann‐Whitney U‐test was used to determine the statistical differences between groups/time‐points. Spearman's rank correlation coefficients (ρ) between measures were computed in the r statistical analysis environment. A P‐value of < 0·05 was considered statistically significant.

Results

Induction of CIA in the DBA/1J mouse

CIA was induced effectively in male DBA/1J mice, as evidenced by clinical progression of the disease as confirmed by histology (Fig. 1). With the passage of time and compared to day 0 (CS = 0), the median CS of the most affected limb increased gradually to 2 (range= 1–2) at day 28 (P < 0·01), 2 (range = 2–3) at day 45 (P < 0·01) and 3 (range = 2–3) at day 56 (P < 0·01). Histologically, the disease became worse with the passage of time, as determined by severity of inflammatory infiltrates, i.e. histological score (HS). Compared to day 0 (HS = 0), the median HS on day 28 was 1 (range = 1–2), on day 45 was 3 (range = 2–3) and on day 56 was 3 (range = 2–3) (Fig. 2). These conventional read‐outs ascertained establishment of the mouse CIA model. In spite of a progressive increase of CS between days 45 and 56 (untreated), the difference in CS between these two time‐points was statistically insignificant (P = 0·10). The difference between HS at day 45 and those at day 56 (untreated) did not reach statistical significance (P = 0·10). The lack of statistically significant differences in CS and HS was attributed to the small sample size (n = 5) in the untreated group.

Figure 1.

¹⁸F‐fluorodeoxyglucose positron emission tomography (¹⁸F‐FDG PET) imaging, a quantitative in‐vivo pre‐clinical drug screening tool. Mouse collagen‐induced arthritis (CIA) was induced in male/1J mice using type II bovine collagen and treated with mouse anti‐tumour necrosis factor (TNF)‐α (CNTO5048) antibody on day 45 for 10 days. In the untreated group, the disease was allowed to progress until day 56. A mosaic of representative photographs of clinical presentation, histology (×10) and maximum‐intensity projection from PET. The pseudo‐colour in PET indicates higher glucose metabolism, hence increased cellular metabolic activity. The yellow arrows in histology represent sites of inflammatory infiltrates.

Figure 2.

¹⁸F‐fluorodeoxyglucose positron emission tomography (¹⁸F‐FDG PET) measured the therapeutic response of mouse anti‐tumour necrosis factor (TNF)‐α in the collagen‐induced arthritis (CIA) model. In the mouse CIA model, anti‐TNF‐α (CNTO5048) therapy was initiated on day 45 and given every 2 days for next 10 days. Clinical score (CS), histological score (HS) and ¹⁸F‐FDG PET scan (PET) were performed on day 45 before starting the therapy (Pre‐Rx). The therapeutic response was determined by CS followed by PET. Histopathological analysis was performed to confirm the findings. Box‐and‐whisker plot representing the disease progression and therapeutic response as determined by CS, HS and PET. The maximum and minimum of each data set is represented by ends of the whiskers. The Mann–Whitney U‐test was used to determine statistical significance. Rx= CNTO5048 treatment, PET score (PS) = ¹⁸F‐FDG uptake.

¹⁸F‐FDG PET correlates with clinical and histological assessment of CIA progression

¹⁸F‐FDG uptake (PS) in the most affected joint as determined by CS were considered for analysis. PS was determined to have a median value of 1·02 (range = 0·85–1·05) on day 0. The PS increased to median values of 1·52 (range = 1·14–2·01) on day 28 (P < 0·05), 3·51 (range = 1·23–4·76) on day 45 (P < 0·05) and 1·76 (range = 1·45–2·78) on day 56 (P < 0·05), compared to PS on day 0. Analogous to the CS, the difference between the PS on days 45 and 56 (untreated) was statistically insignificant (P = 0·073), and was attributed to the small sample size (n = 5) in the untreated group. We assessed the temporal correlation between the PS, CS and HS using data collected at the four time‐points (days 0, 28, 45 and 56). PS had a strong positive correlation with CS (ρ = 0·71, P < 0·05) and with HS (ρ = 0·76, P < 0·05). This result was in concordance with our earlier study 5, albeit that study was performed in a small number of animals.

¹⁸F‐FDG PET assessment of therapeutic response of anti TNF‐α in mouse CIA model

For this analysis, CS, HS and PS on day 45 served as pretreatment values. On day 56 mice were scored clinically, followed by PET scan and subsequent histology. As per expectation, CNTO5048 treatment reduced significantly the median CS to 1 (range = 0–2, P < 0·01) and HS to 1 (range = 0–2, P < 0·01) compared to their respective pretreatment levels (CS = 2, HS = 3). Post‐treatment PS on day 56 was significantly lower (median = 0.97, range = 0·62–2·12, P < 0·01) compared to pretreatment level on day 45 (median = 3·51, range = 1·23–4·76) (Fig. 2). The median CS and HS in the vehicle‐treated mice were 3 (range = 2–3) and 3 (range = 2–3), respectively (Fig. 2), suggesting that the decrease in the severity of inflammation was a consequence of CNTO5048, and not a spontaneous regression of the disease. The difference in the PS at day 56 for the animals that received treatment and for animals at day 0 (before induction of CIA) was statistically insignificant (P = 0·20), indicating that a baseline level of PS was attained due to treatment.

Discussion

This proof‐of‐concept study sought to assess the utility of ¹⁸F‐FDG PET as a quantitative preclinical in‐vivo drug screening tool using the commonly used mouse CIA model. Among several models of murine inflammatory arthritis (K/BxN/TNF‐α transgenic mice), the CIA model was chosen in this study due to its close resemblance to human autoimmune arthritis, such as rheumatoid arthritis 8, 9. This model has been employed extensively to study the pathogenesis of autoimmune arthritis and to test the therapeutic potential of novel drugs. Previous studies have shown that being an inflammatory response to type II bovine collagen, the infiltrates in the joints of CIA are predominantly leucocytes such as neutrophils, monocytes and lymphocytes, including various subpopulations of T cells such as effector memory CD4⁺ T cells, T helper type 17 (Th17) cells and regulatory T cells (T_regs) 13, 14, 15, 16, 17.

In this study, to determine the disease severity histologically, we assessed the severity of inflammatory infiltrates in the affected joint and represent that as histological score (HS). ¹⁸F‐FDG is a glucose analogue taken up by metabolically active cells, and the fluorescence is captured by PET. In an inflammatory milieu, cells become metabolically more active, resulting in increased uptake of ¹⁸F‐FDG. Thus, the uptake of ¹⁸F‐FDG PET corroborates directly with the extent of inflammatory infiltrates (HS) in the affected joint. Herein, we observed that the magnitude of ¹⁸F‐FDG PET uptake in both pre‐ and post‐treatment scenarios are in line with corresponding clinical and histological scores which, in turn, validates the rationale for using ¹⁸F‐FDG PET as an in‐vivo drug screening tool for anti‐inflammatory drugs. This technique could potentially eliminate intraspecies biological variation, as the same animal can be studied at different stages of the disease without the need for killing the animal. The eventual benefits of this model are (i) a quantitative tool to measure the degree of inflammation and (ii) increased study power and a reduction in the number of animals required for a preclinical study.

Non‐invasive in‐vivo molecular imaging has the potential to offer unparalleled insight into the underlying disease process in the native environment of the human or animal body. The use of molecular imaging in autoimmune diseases is still in its infancy, but is gaining interest rapidly. There are few imaging techniques which have been used for in‐vivo assessment of the preclinical model of autoimmune arthritis, which have focused on quantifying: (i) structural changes such as bone destruction [X‐ray computed tomography (CT)] 18, magnetic resonance imaging (MRI) 19, 20 and ultrasound 21; and (ii) molecular or functional changes (PET) 5, 6, single photon emission computed tomography (SPECT) 22 and optical imaging 23. For a detailed review of these techniques, please refer to Mountz et al. 24 and Put et al. 25. PET allows for the production of detailed, three‐dimensional maps indicating nanomolar‐level quantification of radiotracer concentration. An adjunctive information provided by an ¹⁸F‐FDG scan in our study is its ability to quantify degree of inflammation in an in‐vivo system and thus disease severity or therapeutic response can be determined in longitudinal studies in the same mouse.

A study by Irmler et al. used ¹⁸F‐FDG PET to monitor the therapeutic response in the G6PI model of inflammatory arthritis 6. Another study by Terry et al. employed another nuclear imaging technique (SPECT) using ¹¹¹InCl₃ to measure the therapeutic response of etanercept in the mouse CIA model 26. The advantage of PET imaging over SPECT is that the former has a much higher photon sensitivity (by approximately two to three orders of magnitude) 27, which is especially advantageous for quantification of radiotracer uptake in small joints, such as those of the paws in a mouse model of arthritis. To date, there are no published data regarding the use of ¹⁸F‐FDG PET in determining therapeutic efficacy using the mouse CIA model.

In summary, we have shown herein that ¹⁸F‐FDG PET is an in‐vivo preclinical drug screening tool for anti‐inflammatory therapies using the most widely used mouse model of autoimmune arthritis (CIA model). The use of this imaging modality will allow longitudinal quantitative evaluation of the degree of inflammation, and thus has the potential to accelerate drug development and reduce overall therapy assessment cost.

Disclosure

The authors declare no disclosure.