Cancer Diagnosis and Therapy
Optimal excitation-emission wavelengths for autofluorescence diagnosis of bladder tumors
Version of Record online: 15 JAN 2003
Copyright © 2003 Wiley-Liss, Inc.
International Journal of Cancer
Volume 104, Issue 4, pages 477–481, 20 April 2003
How to Cite
Zheng, W., Lau, W., Cheng, C., Soo, K. C. and Olivo, M. (2003), Optimal excitation-emission wavelengths for autofluorescence diagnosis of bladder tumors. Int. J. Cancer, 104: 477–481. doi: 10.1002/ijc.10959
- Issue online: 7 FEB 2003
- Version of Record online: 15 JAN 2003
- Manuscript Accepted: 19 NOV 2002
- Manuscript Revised: 24 OCT 2002
- Manuscript Received: 9 SEP 2002
- National Medical Research Council (Singapore)
- SingHealth Cluster Research Council (Singapore)
- Singapore Cancer Society
- bladder tumor;
- light-induced autofluorescence;
Tissue autofluorescence depends on endogenous fluorophores in the tissue, which undergo a change associated with malignant transformation. This change can be detected as an alteration in the spectral profile and intensity of autofluorescence. Our purpose was to determine the optimal excitation and emission wavelengths for autofluorescence diagnosis of bladder cancer. A total of 52 bladder tissue specimens were obtained from 25 patients undergoing mucosal biopsies or surgical resections of bladder tumors. Light-induced autofluorescence measurements were performed to study the spectroscopic differences between normal and malignant bladder tissue. Fluorescence excitation wavelengths varying from 220 to 500 nm were used to induce tissue autofluorescence, and emission spectra were measured in the 280–700 nm range. These spectra were then combined to construct 2-dimensional fluorescence excitation-emission matrices (EEMs). Significant changes in fluorescence intensity of EEMs were observed between normal and tumor bladder tissues, the most marked differences being at the excitation wavelengths of 280 and 330 nm. The diagnostic algorithm based on the combination of the fluorescence peak intensity ratios of I350/I470 at 280 nm excitation and I390/I470 at 330 nm excitation yielded a sensitivity of 100% [95% confidence interval (CI) 0.95–1.0] and specificity of 100% (95% CI 0.90–1.0). The results of the present fluorescence EEM study demonstrate that autofluorescence spectroscopy can distinguish malignant from normal bladder tissue and that excitation wavelengths of 280 and 330 nm are the most significant for differentiation between normal and malignant bladder mucosae with a high degree of diagnostic accuracy. © 2003 Wiley-Liss, Inc.
Bladder cancer is one of the most prevalent malignant diseases worldwide.1 In Singapore, the incidence of bladder cancer ranks ninth in men and is relatively less frequent in women.2 Early cancer detection and localization with effective treatment are crucial to increasing survival rates and have a significant impact on reducing the high recurrence rate (55–70%).3 Currently, conventional white-light cystoscopy combined with random biopsy is widely used for surveillance of patients who are at high risk of bladder malignancy. However, because early bladder neoplasms, such as dysplasia and carcinoma in situ (CIS), are only a few cell layers thick (0.2–1 mm), they can be very difficult to visually detect by conventional diagnostic methods. Thus, the new techniques for improving the detection of early urothelial neoplasms by guiding the routine biopsy are highly desirable.
In recent years, laser-induced autofluorescence (LIAF) has shown promise for detecting bladder cancers.4, 5, 6, 7, 8, 9 This is because subtle changes in tissue architecture and the biochemical constituents associated with malignant transformations can be probed as an alteration in the spectral shape and intensity of tissue autofluorescence.9 Koenig et al.6 used a 337 nm nitrogen laser light to study autofluorescence spectra of bladder cancer and the fluorescence intensity ratio I385/I455 to yield a sensitivity and specificity of 97% and 98%, respectively, for differentiating malignant from nonmalignant bladder lesions. Anidjar et al.4 used laser wavelengths of 308, 337 and 480 nm to obtain autofluorescence spectra of bladder tissue in vivo and demonstrated the high diagnostic accuracy of LIAF in distinguishing tumors from normal bladder, including CIS. They found that among the 3 excitation wavelengths used, 308 nm was most suitable for tissue excitation as it provided the greatest amount of information related to endogenous fluorophores and allowed more accurate diagnosis of bladder tumors.4 Different excitation wavelengths may induce different types of tissue fluorophore to fluoresce, resulting in different spectral profiles and intensities of tissue spectra with different diagnostic abilities for cancer detection. So far, most autofluorescence studies on bladder tissues have been carried out using only a limited number of laser wavelengths.4, 7, 8, 9 The optimal excitation wavelengths in the entire spectrum of light between 220 and 500 nm for spectroscopic diagnosis of bladder cancer are still largely underexplored. Our purpose was to use the entire fluorescence excitation-emission matrices (EEMs), rather than a few excitation wavelengths, to study the spectroscopic properties of bladder tissues in vitro and to determine the optimal excitation and emission wavelengths for distinguishing malignant from normal bladder tissue. The fluorescence peak intensity ratios of I350/I470 at 280 nm excitation and I390/I470 at 330 nm excitation were used as diagnostic algorithms to evaluate the diagnostic accuracy for bladder tissue classification.
MATERIAL AND METHODS
Bladder tissue samples were obtained from 25 patients (20 men and 5 women, median age 67.3 years) with known or suspected bladder malignant lesions undergoing diagnostic biopsies or resections of bladder tumors. Preoperatively, all patients signed an informed consent permitting the investigative use of tissues, and our study was approved by the Ethics Committee of the National Cancer Centre, Singapore. After cystoscopic biopsy or surgical resection, tissue specimens were rinsed, placed in bottles with physiologic saline solution (pH = 7.4) and delivered for spectroscopic measurements immediately. Following spectral measurements, tissue samples were fixed in 10% neutral buffered formalin and then submitted for histopathologic examination.
Autofluorescence spectra of bladder tissue samples were measured using a spectrofluorometer (RF-5301PC; Shimadzu, Kyoto, Japan). The instrument consists of a 150 W white light source (xenon lamp), an excitation monochromator, a tissue chamber and an emission monochromator equipped with a photomultiplier tube detector (R955; Hamamatsu, Shizuoka, Japan). A personal computer controlled spectral data acquisition. For autofluorescence measurements, each tissue specimen was placed and fixed in a sample holder with saline in the tissue chamber. A fiberoptic probe (Ocean Optics, Dunedin, FL) was used to deliver the filtered light from the excitation monochromator, to illuminate tissue at normal incidence and pick up the emitted fluorescence, and then coupled to the emission monochromator for spectral analysis. Tissue autofluorescence spectra were acquired at 29 excitation wavelengths varying from 220 to 500 nm in 10 nm increments. Correspondingly, emission wavelengths varied from 280 to 700 nm in 5 nm increments. Spectral resolution was 5 nm for both the excitation and emission monochromators. The spot size of the excitation light on the tissue sample was approximately 3 mm, and all samples used were large enough to cover the spot size of excitation light. All measured spectra were corrected for source-intensity variations and spectral response of the system, and all spectral data were also normalized to the tissue fluorescence peak intensity at 470 nm at an excitation wavelength of 360 nm, to correct for the influence of variations in probe positioning and enable comparison of the relative intensities between different tissue samples. These normalized fluorescence spectra were then combined to generate fluorescence EEMs for each tissue sample. The total measurement time for each EEM was approximately 10 min. At the end of each EEM measurement, the fluorescence spectrum at the first excitation wavelength (220 nm) was measured again to evaluate the effect of photobleaching, and no obvious decrease in fluorescence peak intensity at 350 nm was observed.
For the assessment of diagnostic sensitivity and specificity, histopathologic results were regarded as the gold standard. Normal mucosa and inflammation were classified as benign or normal, whereas CIS and stage TxGx tumors were classified as malignant. Unpaired Student's t-test was used to test the difference between the fluorescence peak intensity ratios at 350 to 470 nm with 280 nm excitation and 390 to 470 nm with 330 nm excitation for both normal and malignant bladder tissues.
A total of 52 bladder tissue samples were used for fluorescence EEM studies. Pathologic evaluations showed that there were 14 normal and inflammatory specimens, 3 with CIS, 22 with a stage pTa tumor, 6 with a stage pT1 tumor and 7 with a muscle-invasive pT2 tumor (Table I).
|Histology||Number of lesions|
|Transitional cell carcinoma (stage/grade)|
Figure 1 shows the average fluorescence EEMs of normal and malignant tumor bladder tissues. EEMs of bladder mucosas exhibited characteristic contours centered at 350 nm (220–230 and 270–290 nm excitation), 390 nm (320–340 nm excitation), 470 nm (270–280 and 330–370 nm excitation) and 520 nm (410 and 440 nm excitation), which are most likely due to emission from tryptophan, collagen, NADH and oxidized riboflavins (flavin adenine dinucleotide, FAD), respectively, in bladder tissue.10, 11, 12, 13 An additional fluorescence peak at 635 nm (410 nm excitation) was also observed in some tumor tissues. This may be a result of the increased fluorescence of endogenous porphyrins accumulated in tumors, which had also been found in other organs.14, 15, 16 However, diagnosis based on this red fluorescence peak intensity may be inaccurate as the red fluorescence may be due to porphyrins produced by certain bacteria, and other tissue cells of epithelial origin may also concentrate porphyrins.14
Although the excitation-emission fluorescence peak positions were not significantly altered in tumors, significant changes in fluorescence intensities existed between normal and tumor tissue, indicating that these excitation-emission wavelength pairs observed in fluorescence EEMs may be diagnostically useful for bladder cancer detection. Among the excitation wavelengths below 400 nm, some (e.g., at 230 and 360 nm) can induce a strong autofluorescence emission band (e.g., at 350 and 470 nm). Based on absolute intensity measurements, the intensity decrease or increase could allow differentiation between normal mucosa and tumor tissue. However, some UV excitation light wavelengths can effectively excite several fluorophores of bladder tissue to fluoresce, resulting in more than 1 fluorescence band with different peak intensities between normal and tumor tissue (Fig. 1). For instance, at 280 nm excitation, 2 fluorescence bands appeared in tissue autofluorescence and the intensity of the fluorescence peak at 350 nm was consistently higher in tumor than in normal tissue, while the intensity of the fluorescence peak at 470 nm was lower in bladder tumors. At 330 nm excitation, fluorescence peak intensities at 390 and 470 nm for tumor tissues were less than those for normal tissues. These observations suggested that the autofluorescence peaks induced by 280 nm excitation, at 350 and 470 nm emissions, and that induced by 330 nm excitation, at 390 and 470 nm emissions, would maximize the ability of bladder tumor discrimination. Absolute intensity determinations were not required in this situation since a definite diagnosis could be established based on the fluorescence peak intensity ratios at 350 nm to 470 nm and 390 nm to 470 nm. Therefore, 280 and 330 nm were selected as optimal excitation wavelengths for bladder cancer diagnosis. Compared to the autofluorescence results obtained with excitation wavelengths below 400 nm, autofluorescence signals of bladder tissue using excitation wavelengths above 400 nm were weak with considerable background noise; thus, excitation wavelengths above 400 nm were not suggested for autofluorescence diagnosis of bladder tumors.
Figure 2 shows the ratios of the fluorescence peak intensities at 350 nm to 470 nm and 390 nm to 470 nm, correlating with their histologic findings. At 280 nm excitation, the mean ratio value of I350/I470 in normal tissue was 2.81 ± 0.44, which is significantly different from the mean value of 4.56 ± 0.65 in malignant tissue (unpaired Student's t-test, p < 0.001). Using I350/I470 of 3.50 as a cut-off, this algorithm yields a sensitivity and specificity of 97% [95% confidence interval (CI) 0.90–1.0] and 93% (95% CI 0.88–0.98)17 for differentiating malignant from normal tissue. Only 1 of 38 malignant tissues was misclassified as normal, and 1 of 14 normal tissues was misdiagnosed as tumor. Similarly, with the excitation wavelength of 330 nm, the mean ratio value of I390/I470 for normal tissue was 1.09 ± 0.08, which is also significantly different from the mean value of 1.30 ± 0.09 for malignant tissue (unpaired Student's t-test, p < 0.001). By selecting I390/I470 of 1.18 as a cut-off for tissue classification, sensitivity of 95% (95% CI 0.90–0.98) and specificity of 92% (95% CI 0.85–0.98) were achieved.
Figure 3 presents the scatter plot of the intensity ratio of I390/I470 at 330 nm excitation vs. the intensity ratio of I350/I470 at 280 nm excitation for different pathologic types. A linear decision line separates bladder tissue samples with cancer from normal tissue, with sensitivity of 100% (95% CI 0.95–1.0) and specificity of 100% (95% CI 0.90–1.0). The results presented above show that the nondimensional intensity ratios of I350/I470 at 280 nm excitation and I390/I470 at 330 nm excitation can be used as discriminating criteria with a high degree of accuracy for bladder cancer diagnosis.
Measurements of in vitro autofluorescence of fresh tissue samples are usually the first step in the process of developing clinical devices for fluorescence spectroscopy and imaging of diseased tissues. In vitro examination has the advantage of allowing analysis of fluorescence spectra over a complete range of excitation wavelengths. We measured the fluorescence EEMs of bladder tissue samples using excitation wavelengths of 220–500 nm and determined the excitation wavelengths with the greatest discrepancy between normal and tumor bladder tissue. Our results demonstrate increased fluorescence at 350 and 635 nm emission and a decrease at 390, 470 and 520 nm emission associated with bladder tumors. These fluorescence peaks observed in fluorescence EEMs are close to the emission spectra of pure tryptophan (emission peak approx. 340 nm), collagen (emission peak approx. 385 nm), NADH (emission peak approx. 460 nm), FAD (emission peak approx. 510–530 nm) and porphyrins (emission peak 635 nm);10, 11, 12, 13, 14 we thus tentatively identify these biochemical compounds as major endogenous fluorophores of bladder tissue. The changes in fluorescence intensity of bladder tumors may involve the influence of factors such as tissue architectures, endogenous fluorophores and light penetration depth in tissue. The significant decrease of collagen (390 nm), NADH (470 nm) and FAD (520 nm) fluorescence in malignant tumors may be due to changes in tissue morphology (e.g., thickening of the urothelial layer) and the decrease in endogenous fluorophore (e.g., collagen, NADH and flavin) quantum yields.18, 19 The normal urothelium thickness of the bladder is estimated to be <150 μm,20 and the 300–400 nm light is able to excite bladder tissue layers on the order of 500 μm. As a result, tissue autofluorescence signals from bladder tissue include contributions not only from the urothelium but also from deeper tissue layers, such as the collagen located in the lamina propria and the NADH in muscle layers. In contrast to the lamina propria and muscle, there is no collagen and only a small amount of NADH compounds in the urothelium.6, 9 The increased thickness of the urothelium in malignant lesions thus led to the decrease of the collagen (390 nm) and NADH (470 nm) signals by filtering the autofluorescence from deeper tissue layers. The decrease in the contribution of collagen and NADH fluorescence from healthy to neoplastic tissue has also been observed in other tissue sites (e.g., colon, esophagus and bronchus).21, 22, 23
Another important observation is the increase in the fluorescence of tryptophan residues (emission peak at 350 nm) in bladder tumors when using excitation wavelengths below 300 nm. This increase was probably due to hyperactivity or urothelial hyperproliferation in tumor. Brancaleon et al.24 also observed increased fluorescence of tryptophan in skin tumor. The increased proliferative and decreased apoptotic activity of malignant cells could cause an increased number and/or larger size of differentiated cells of bladder tumor lesions, resulting in higher concentrations of proteins in cells and increased thickness of the urothelium (therefore larger tryptophan fluorescence signals).5
To discriminate tumor from normal bladder tissue, 2 nondimensional intensity ratios, I350/I470 at 280 nm excitation and I390/I470 at 330 nm excitation, were employed as diagnostic algorithms. As the ratio of I350/I470 at 280 nm excitation was related to tissue fluorescence of tryptophan and NADH and I390/I470 at 330 nm excitation was related to fluorescence of collagen and NADH, a variation in one or several fluorophore emission intensities will change the ratio values, thereby revealing changes of the spectral distribution of autofluorescence in tumor tissue. The increased value of the fluorescence ratios of I350/I470 and I390/I470 for different stages and grades of bladder tumors (including CIS) compared to normal tissue reflects an increase in the tryptophan fluorescence contribution and a decrease in the collagen and NADH fluorescence contribution. Statistical analysis showed that both intensity ratios yielded high diagnostic sensitivity and specificity (>90%) (Fig. 2). Combining the 2 ratio diagnostic algorithms can further improve the diagnostic yield for bladder cancer detection (Fig. 3). In contrast to absolute intensity measurements for tumor differentiation, the intensity ratios I350/I470 and I390/I470 are independent of measurement conditions such as excitation light power fluctuation or probe positioning variation; therefore, they can be used for comparative analysis of spectra collected from different tissue sites and/or subjects. Koenig et al.6 utilized the intensity ratio I385/I455 at 337 nm laser excitation to give sensitivity and specificity of 97% and 98% for differentiating malignant from nonmalignant bladder lesions, and similar diagnostic results were reported by Anidjar et al.,4 who used the intensity ratio I360/I440 at 308 nm laser excitation for bladder tumor diagnosis. All of these findings support the use of light-induced autofluorescence spectroscopy as a new diagnostic technique for occult urothelial tumors.
However, despite very promising rates in determining bladder cancer using autofluorescence spectroscopy, bladder lesions cannot be visualized and the bladder surface must be scanned point by point with the tip of the fiber probe to assess the entire bladder mucosa. This point-by-point measurement is time-consuming, especially for examination of large tissue areas. Thus, the autofluorescence spectroscopy and imaging system is required for detecting and localizing subtle changes in bladder lesions not distinguishable by white light cystoscopy.18 The results of our fluorescence EEM study could be particularly useful to design a simplified autofluorescence spectroscopy and imaging device for providing more accurate detection of occult urothelial lesions. A combination of full emission spectra from 2 excitation wavelengths (280 and 330 nm) with autofluorescence ratio imaging (I350/I470 and I390/I470) may significantly enhance the contrast between normal and malignant bladder tissue and allow regional observation of urothelial carcinoma of the bladder, overcoming the limitations of point-by-point measurements based on spectroscopy alone.
In conclusion, our fluorescence EEM study demonstrates the ability of autofluorescence spectroscopy to distinguish malignant from normal bladder mucosa in vitro. Our results show that tissue autofluorescence of bladder tumor is characterized by an increase in the emission of tryptophan and porphyrins and a decrease in the emission of collagen, NADH and FAD compared to normal tissue. The finding of increased fluorescence of tryptophan residues related to bladder cancer could be particularly useful for detecting the urothelial hyperproliferation of early bladder lesions (e.g., dysplasia and CIS) through the combination of the increase of tryptophan fluorescence and the decrease of collagen. Furthermore, with the most significant excitation wavelengths at 280 and 330 nm, the diagnostic algorithm based on the combination of fluorescence intensity ratios of I350/I470 and I390/I470, which contains the clinically valuable fluorescence information on all 3 fluorophores (i.e., tryptophan, collagen and NADH), can provide good differentiation between normal and malignant bladder tissue. If the UV doublet (280 and 330 nm excitation) induced autofluorescence spectroscopy and imaging system were successfully developed for in vivo applications, this technique may also be useful as an adjunct to conventional cystoscopy for optically guided biopsies for the histopathologic investigation of early bladder lesions.
- 2Trends in cancer incidence in Singapore (1993–1997). Singapore Cancer Registry Rep 2000; 5: 110., , , .
- 10Protein fluorescence. In: Principles of fluorescence spectroscopy. New York: Plenum, 1983. 341–81..
- 12Fluorescence of organic natural products. In: SchulmanSG, ed. Molecular luminescence spectroscopy, vol. 1. New York: John Wiley & Sons, 1993. 167–370..
- 17Statistical methods for rates and proportions, 2nd ed. New York: John Wiley & Sons, 1981..
- 20Urinary bladder and ureter. In: SternbergSS, ed. Histology for pathologists. New York: Raven Press, 1992. 709–720..