Establishment of Real-Time Multispectral Imaging for the Detection of Bladder Cancer Using a Preclinical in Vivo Model

BBN-induced carcinogenesis

Male C57BL/6JRj mice were obtained from Janvier Labs (Janvier Labs, France). Mice received 0.1% BBN (Sigma Aldrich Biochemie GmbH, Hamburg, Germany. Cat.-Nr. B8061) ad libitum in drinking water for 3 (n = 6) or 5 (n = 24) months. The drinking water containing BBN was prepared once a week. The experiments were conducted with the approval of the Baden Württemberg animal ethics committee (animal experiment number: 1326). Mice were sacrificed, and bladders were prepared for ex vivo detection of tumours by rMSI. Hexaminolevulinat (HAL), CD90.2-FITC (Miltenyi Biotec B.V. & Co. KG, Bergisch Gladbach, Germany), CD47-FITC (BD Biosciences, San Jose, CA, USA), CD47-FITC/HAL combinations and CD90.2-FITC/HAL combinations were directly applied on the mouse tumours. Five non-treated control mice (no BBN-treatment, normal urothelium) and 30 BBN-treated mice were analyzed. For histopathological visualization 2μm thick whole slide section of each FFPE block were cut, mounted on conventional STARFROST glass slides and stained with Hematoxylin and Eosin (HE).

Preparation of murine bladders for imaging with the multispectral camera system

The bladders were incubated for 1 hour in 8 mM HAL (Sigma-Aldrich Biochemie GmbH, Hamburg, Germany) and different concentrations (1μg/ml –4μg/ml) of FITC-labelled antibodies against CD47 (BD Biosciences, San Jose, CA, USA), or CD90.2 (Miltenyi Biotec B.V. & Co. KG, Bergisch Gladbach, Germany) in DMEM (ThermoFisher, Paisley, Scotland) containing 15% FCS (Sigma Aldrich, Steinheim, Germany), 1% Penicillin/Streptomycin (PAN-Biotec GmbH, Aidenbach, Germany), 5% Glutamax (ThermoFisher, Grand Island, New York, USA), 1% non-essential aminoacids (ThermoFisher, Paisley, Scotland) and 1% sodium pyruvate (Life Technologies Europe BV, Bleiswijk, Netherlands). The bladders were fixed onto a cork, which was painted black with a permanent marker, glued into a 12-well plate (Sarstedt AG & Co. KG, Nürnbrecht, Germany) using a double-sided scotch tape. For imaging, the bladder was immersed into DPBS (Lonza Group Ltd, Basel, Switzerland).

Multispectral imaging of excised mouse bladders

The general principle of the multispectral camera system is described elsewhere [6]. Briefly, the camera system consists of a USB camera (Ximea GmbH, Münster, Germany) with a custom-designed filter (Chroma Technology Cooperation, Bellows Falls, VT, USA), a video adapter (R.Wolf, Knittlingen, Germany), a microcontroller unit, a multi-LED light source (Omicron-Laserage Laserprodukte GmbH, Rodgau, Germany), and a PC for data acquisition. For imaging of the specimens, a 4 mm 30° endoscopy optic and a liquid light guide were used (Karl Storz GmbH, Tuttlingen, Germany). The white light images and the fluorescence images of murine bladders were acquired concurrently and saved as uncompressed 16-bit tif-images with bayer pattern (raw-image format).

Image/video processing

The image processing was performed using Python and OpenCV. All images were background corrected, transformed to RGB, colour corrected and gamma corrected. Additionally, a simple intensity-based threshold was used for masking the data. The fluorescence data were unmixed into a PpIX fluorescence channel and a background fluorescence channel, which combined tissue autofluorescence and fluorescein fluorescence from labelled antibodies, using matrix multiplication. The ratio of PpIX and BGD fluorescence highlighted areas with high PpIX fluorescence. For quantification 100 sequential images were selected, and the median value of the PpIX and BGD ratios were calculated for each frame. The median value of the 100 frames was calculated yielding a single value per mouse bladder. The statistical analysis was performed using the python library scipy.stats (https://docs.scipy.org/doc/scipy/reference/stats.html).

Evaluation of the signal intensity was done by visual inspection of the processed videos. The signal intensity for a bladder was rated 1.0 (strong), 0.5 (moderate), 0.25 (weak), 0.125 (very weak), and 0 (no signal, see also Table 1).

BBN-induced bladder cancer recapitulates human BC

We used the established N-butyl-N-(4-hydroxy butyl)-nitrosamine (BBN) bladder cancer (BC) model to induce tumours [15, 16] administrating 0.1% BBN ad libitum (Fig. 1A). BBN specifically induced tumours in the bladder, resulting in the development of invasive tumours by five months. Recent studies demonstrated that the tumours substantially reflect morphological and molecular characteristics of human BC [17–19].

OPEN IN VIEWER

Fig. 1

BBN-induced bladder cancer recapitulates human bladder cancer. (A) Schematic representation of the BBN-induced bladder cancer model. Male mice with C57BL/6 genetic background were obtained from Janvier Labs at the age of 6 weeks. Starting at the age of 8 weeks, they received 0.1% BBN ad libitum in drinking water, which was freshly prepared every week. After 3 or 5 months of treatment (at the age of ∼20 or ∼28 weeks, respectively) the mice were sacrificed, and the bladders were excised for microscopy. (B) Macro-view of mouse bladders. (Top left) no BBN treatment; normal bladder; pathological result: normal bladder. (Top right) BBN treatment for 5 months; enlarged bladder; pathological result: squamous CIS with focal invasion. (Bottom left) BBN treatment for 5 months; enlarged bladder; pathological result: squamous carcinoma. (Bottom right) BBN treatment for 5 months; enlarged bladder; high vascularization; pathological result: squamous CIS. (C) Histological evaluation of normal and tumour mouse bladder tissues. (Top row) Representative images of a BBN induced carcinoma of a murine bladder. Black arrows highlight the superficial urothelial lining. Carcinoma infiltrates show an undermining growth pattern (marked area). Higher magnifications (right panel) show poorly differentiated carcinoma cells with highly pleomorphic and atypical nuclei with prominent nucleoli as well as prominent intercellular bridges resembling a feature of squamous differentiation (representative atypical nuclei are marked with black arrows). Keratinization is not present. (Middle row) Urothelial lining of a BBN murine bladder with prominent squamous metaplasia with focal high-grade atypia (marked area). In the 40× magnification (right panel) the metaplastic cells show marked high-grade atypia with pleomorphic nuclei and disturbed epithelial architecture (representative atypical nuclei are marked with black arrows). (Bottom row) Normal urothelial lining without squamous metaplasia or dysplasia. Urothelial cells are marked with black arrows. (D) Graphical illustration of the histological evaluation of mice tissues. H&E staining, followed by pathologic analysis, were conducted with 30 BBN-induced bladder tumours. 5M and 3M stand for 5 months and 3 months, respectively. The detailed information on tissue histology is depicted in Table I. Please note that invasive cancer comprises all samples with invasion of the submucosal tissue layer or the muscularis propria, i.e. all carcinoma and CIS with focal invasion.

Histological evaluation of mice bladder tissue revealed that 50% (12 of 24) of the 5-months BBN-treated mice develop invasive BC (defined as cancer cells invading the submucosal tissue layer or the muscularis propria) (Figs. 1B–D). As reported earlier, the majority of these tumours are of basal origin and represent squamous cell carcinoma (SCC) (Table 1). Moreover, 17% (4/24) of the five months BBN-treated mice exhibit carcinoma in situ (CIS) with focal invasion, and 33% (8 of 24) developed CIS. On the other hand, none (0/6) of the three months BBN-treated mice developed macroscopic lesions (Figs. 1B–D). While CIS could be observed in 50% of these mice (3/6), the other 50% (3/6) showed focal metaplasia (Fig. 1C, D). As expected, none of the control mice (0/5) showed any signs of abnormality in the bladder tissue after 3 or 5 months (Figs. 1C and 1D).

Ex vivo real-time multispectral imaging of BC in BBN-induced mouse model

We used the ex vivo organ culture to evaluate the suitability of the new real-time multispectral imaging in the preclinical mouse model, based on the report for the ex vivo detection of human bladder tumour tissue by the CD47 antibody [12]. Excised mouse bladders were prepared as described in the Materials and Methods section and incubated with Hexaminolevulinat (HAL, Hexvix®), CD47-FITC, CD90.2-FITC or a combination of CD90.2-FITC/ CD47-FITC and HAL, along with bladders from mock-treated control mice. The selection of the chemicals was based on the application in human BC imaging (HAL) or their reported expression in human BC [12–14]. Fig. 2A shows images of murine bladders recorded with the multispectral imaging system, which allows simultaneous imaging of the white light reflectance and the fluorescence of multiple fluorophores. The bladders were treated with HAL, a substrate of the heme metabolism that is metabolized to protoporphyrine IX (PpIX) and accumulates in cancer cells [20–22]. The PpIX fluorescence is excited at 405 nm, and it fluoresces in the red region of the visible spectrum (peak of the fluorescence emission at 635 nm). The red fluorescence can be detected predominantly in the bladder samples of the five months BBN-treated mice (Fig. 2A, middle column, indicated by arrows). Of note, the white light reflectance of these tissues indicates strong proliferation/hyperplasia compared to the mock-treated controls or the 3 months BBN-treated mice (Fig. 2A, left column), in line with tumour detection of human BC by white light as indicated above. This observation is also supported by the histological evaluation, which shows invasive carcinoma or CIS for these samples (Table 1).

OPEN IN VIEWER

Fig. 2

Ex vivo real-time multispectral imaging of BC in BBN-induced mouse model. (A) Endoscopic images of excised murine bladders under BBN treatment acquired by the multispectral camera system. Image modes: WL: white light, Fluo: fluorescence, PpIX: unmixed PpIX fluorescence channel and ratio PpIX/BGD:The ratio between the unmixed PpIX signal and the unmixed autofluorescence/FITC (BGD) signal. (Colormap: Blue = low values, Orange/Red = high values). Arrows indicate the PpIX signal. (B) Category plot of the ratio between the unmixed PpIX signal and the unmixed autofluorescence/FITC signal per murine bladder. The ratio describes a coarse measure of the overall presence of PpIX in the investigated murine bladders. Labels show the pathological staging of the murine bladders. Mann-Whitney-U-Test was used for statistical analysis (*p < 0.05, **p < 0.005). (C) Bar plot showing the percentage of bladders with positive PpIX signals on visual inspection of the processed videos. The different treatment categories are indicated. 5M = 5 months, and 3M = 3 months, respectively. (D) Subjective rating of PpIX signal intensity per bladder (= relative PpIX signal intensity). The signal intensity of a bladder was rated with 1.0 for a strong signal, with 0.5 for a moderate signal, 0.25 for a weak signal and with 0.125 for very weak signal and 0 for no signal (see also Table I). Bars show mean +/–standard deviation. The sum of all ratings per category was divided by total mouse number in each category and multiplied by 100 to calculate the arbitrary values. The different treatment categories are indicated. 5M = 5 months, and 3M = 3 months, respectively. (E) Bar plot showing the percentage of bladders with positive PpIX signal on visual inspection of the processed videos. Histological results shown in B were pooled as follows: Carcinoma (Squamous Carcinoma + CIS, Squamous Carcinoma, Urothelial Carcinoma), CIS (Squamous CIS, Squamous CIS with focal invasion), Metaplasia (Squamous Metaplasia), Normal Urothelium. –HAL bladders were excluded. (F) Subjective rating of PpIX signal intensity per bladder (= relative PpIX signal intensity). The signal intensity of a bladder was rated with 1.0 for a strong signal, with 0.5 for a moderate signal, 0.25 for a weak signal and with 0.125 for very weak signal and 0 for no signal (see also Table 1). Bars show mean +/–standard deviation. The sum of all ratings per category was divided by total mouse number in each category and multiplied by 100 to calculate the arbitrary values. The different treatment categories are indicated. 5M = 5 months, and 3M = 3 months, respectively.

The elevated PpIX is also seen in the image showing the ratio between the unmixed PpIX signal and the autofluorescence/background signal (BGD signal) (Fig. 2A, right column). The high PpIX signal of the 5 months BBN-treated murine bladders is depicted in Fig. 2B, which shows the median ratio between the median PpIX signal and the median background signal. Bladders of 5 months BBN-treated mice (n = 20) show a significantly higher PpIX signal/BGD ratio compared to all other groups (Fig. 2B, first group: +BBN, 5M + HAL). Histological analysis of the mouse bladders shown in Fig. 1C revealed invasive carcinoma and CIS, which is in good agreement with the clear positive PpIX fluorescence signal (Table 1).

As expected, all 5 months BBN-treated bladders (n = 4), which were not stained with HAL were negative for the PpIX signal (Fig. 2A–D: +BBN/–HAL), indicating the specificity of HAL-signal for tumour tissue. The weak PpIX/BGD signals in these –HAL control mice (Fig. 2A, bottom row) can be discriminated from the true PpIX-positive signal observed in +BBN(5M)/+HAL and +BBN(3M)/+HAL mice in the white light image and fluorescence image. The quantitative evaluation of these signals in the +BBN/–HAL-group clearly indicates them as false-positive when compared to the +BBN (5M)/+HAL group (Fig. 2C, D).

Notably, the 3 months BBN-treated mouse bladder (+BBN(3M)/+HAL) presented in Fig. 2A shows a positive PpIX signal in the peripheral region in the fluorescence image (Flou) as well as in the ratio (PpIX/BGD). At the same time, the bladder tissue of the BBN-untreated control mouse is PpIX-negative (Fig. 2A, second row, indicated by an arrow). Local positive PpIX signals were observed in 4 out of 6 mice in this group (histologically defined as CIS or metaplasia), but a weak, non-specific PpIX-signal was observed only in one of the five –BBN/+HAL mice and in none of the mock-treated control mice (Fig. 2C, 2D and Table 1). Please note that the ratio between PpIX and background of the 3 months BBN-treated group is not significantly different from the control groups, probably due to the overall negative signals, damping the specific signal values in the total calculation (Fig. 2B). These results suggest that local, premalignant lesions can be detected with the new imaging system combined with HAL. In fact, the evaluation of the results in relation to histology reveals that bladders of all mice with carcinoma (SCC and UCB) or CIS (CIS with or without local invasion) show a PpIX signal in the fluorescence image (Fig. 2E). Moreover, the relative PpIX signal intensity of the bladders with carcinoma and CIS is elevated compared to the bladders with metaplasia or normal urothelium (Fig. 2F). In fact, also the bladders with metaplasia exhibit slightly higher signal intensities compared to healthy bladders (∼8 versus ∼2,5, Fig. 2F), though the signal intensities vary significantly among individual samples.

In summary, distinct PpIX signals were detected in the bladder tissue of all (n = 20) of the 5 months BBN-treated mice incubated with HAL (Fig. 2C, D). Moreover, the relative PpIX signal intensity was the highest (52.5) in this group (Fig. 2D, and Table 1), whereas only one of the five control mouse bladders incubated with HAL (–BBN/+HAL) showed weak, false-positive PpIX fluorescence (relative PpIX signal intensity 2.5). Importantly, although the average signal intensities of the 3 months BBN-treated mice (3 months BBN/+HAL) were comparable to the control groups (–BBN/+HAL and +BBN/–HAL), they showed specific local signals, in line with early lesions observed in this group (CIS or metaplasia). Interestingly, PpIX signal intensity shows some degree of variation within the same group (i.e. histology) (Table 1), potentially due to the localization of the degenerated tissue.

Fluorescence imaging shows that the staining with FITC-conjugated antibodies against surface proteins CD47 and CD90.2 is inconspicuous. A FITC-signal could not be detected with bladders labelled with FITC-conjugated antibodies against CD47. Some of the CD90-labelled bladders showed a FITC signal in the peripheral areas, especially the one incubated with higher antibody concentrations. Similarly, strong green autofluorescence was observed in all bladders, comparable to signals observed with CD90.2-FITC and CD47-FITC antibodies, indicating that these antibodies had limited ability to detect BC in BBN-treated mice.