SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References

Urine is one of the diagnostically important bio fluids, as it has different metabolites in it, where many of them are native fluorophores. Native fluorescence characteristics of human urine samples were studied using excitation–emission matrices (EEMs) over a range of excitation and emission wavelengths, and emission spectra at 405 nm excitation, to discriminate patients with cancer from the normal subjects. The fluorescence spectra of urine samples of cancer patients exhibit considerable spectral differences in both EEMs and emission spectra with respect to normal subjects. Different ratios were calculated using the fluorescence intensity values of the emission spectra and they were used as input variables for a multiple linear discriminant analysis across different groups. The discriminant analysis classifies 94.7% of the original grouped cases and 94.1% of the cross-validated grouped cases correctly. Based on the fluorescence emission characteristics of urine and statistical analysis, it may be concluded that the fluorophores nicotinamide adenine dinucleotide (NADH) and flavins may be considered as metabolomic markers of cancer.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References

Fluorescence spectroscopy has been considered as one of the potential diagnostic tools for the detection of various cancers, as this technique is highly sensitive, minimally invasive, possible to make portable and has many complementary techniques. The photophysical properties of intrinsic fluorophores such as reduced nicotinamide adenine dinucleotide (NADH), flavin adenine dinucleotide (FAD), collagen, elastin and their cross-links, the aromatic amino acids including tryptophan, tyrosine and phenylalanine, and porphyrins have been considered as a tool to understand the alterations in the functional, morphological and microenvironmental changes in cells and tissues. Among these intrinsic fluorophores, collagen, elastin and more generally proteins are related to the structural arrangement of cells and tissues. The other fluorophores, pyridoxine derivatives, NADH (reduced), FAD and porphyrin are related to metabolic processes [1]. Although many had reported extensively on the use of native fluorescence spectroscopy in the discrimination of different pathological conditions of tissue from their normal counterpart, only limited reports are available to understand the reasons for the altered spectral signature between normal and abnormal cells and tissues [2-7]. Hence, there is a need to understand the microscopic origins of the spectral differences to develop a new strategy for improving the diagnostic potential of fluorescence spectroscopy in oncology. Besides the above, many also showed interest on the furthermore characterization of the altered spectral signature with respect to their biochemical and structural changes during the transformation of normal to malignant conditions [8, 9]. In this regard, although many reported on the photon propagation in turbid media, there is a need for improved model for optical molecular imaging under in vivo conditions [10, 11].

Most of the cancer patients report to the physician at the late stage of the disease, the condition where only palliative treatment is possible and this increases the morbidity of the disease. Under these circumstances, it is felt that there is a need for rapid screening techniques for the early cancer diagnosis to improve the therapeutic efficacy. Although many applications of native fluorescence spectroscopy of biomolecules are reported on the characterization of cellular metabolic pathways and the discrimination of malignant from normal conditions of tissues, only limited studies have been reported on the applications of native fluorescence spectroscopy of biofluids in diagnostic oncology. Only limited data are available on the optical spectroscopic characterization of body biofluids such as blood [1, 12-14], saliva [15, 16] and urine[17, 18]. Although Leiner et al. has discussed the possible shoulder emission spectra of various molecules at different excitation and emission ranges, they have not compared cancer samples with normal urine samples [17].

Among various biofluids, urine is also considered to be one of the important biological fluids as it has many metabolites including some key native fluorophores [19]. Studies reported that there is a considerable variation in the concentration and conformation of some of the metabolites due to food intake, body metabolism and in particular due to some altered metabolic and pathological conditions [20, 21]. Based on these, data reported that the fluorescence of urine has been considered in the clinical diagnosis of a number of health conditions, including proteinuria, renal disorder, nephritic syndrome, hepatopathy and neural ceroid lipofuscinosis [22]. For example; Rabinwitz reported that there is a considerable variation in the red and blue fluorescence emission intensity and their ratio between the urine of benign and malignant patients with respect to normal counterpart and highlighted the importance of porphyrin and its relation in the cancer genesis [23]. Dubayova et al. reported on the use of synchronous fluorescence spectroscopy for diagnostic monitoring of urine and subsequently by excitation–emission matrix [18, 19]. Anwer et al. had reported on the use of autofluorescence in discriminating the individuals with bacteriuria with the normal subjects and subsequently Sandeep Perinchery et al. reported the autofluorescence of urine in the diagnosis of human urinary tract infection[20, 22]. Recently, Yi-Qun Wan made use of synchronous fluorescence spectroscopy and reported that the variation in isoxantopterin levels is significant in the urine of stomach cancer patients when compared with the normal subjects [24].

Although these findings support the hypothesis that native fluorescence spectroscopy may provide a promising method of measuring various pathobiochemical changes in urine for disease diagnosis, research in the field of analysis of urine for disease diagnosis by native fluorescence spectroscopy is still in the initial phase. Furthermore, the reported studies do not suggest a comparison of the fluorescence emission characteristics of urine at different stages of malignancy and different etiologies with respect to healthy subjects [25]. Besides, the study on optical spectroscopy of urine in diagnostic oncology has not yet been fully optimized both experimentally and statistically. In this study, urine samples from patients with Head and Neck, Cervix and Breast cancer was considered because the incident rate of these cancers is high in India. The purpose of this study is to determine whether the fluorescence emission spectral characteristics of urine at 405 nm excitation provide statistical difference between patients with cancer and normal subjects.

Materials and Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References

Samples

First voided morning urine samples were collected in a sterile container from cancerous patients of different etiologies and stages who are admitted for treatment in various hospitals in Chennai and from healthy volunteers. As a total of 80 normal subjects, both male (N = 32) and female (N = 48) in the age group of 18–65 years and 90 pathologically confirmed cancerous patients (of which 31 from Head and Neck cancer, 39 from Cervical cancer, 13 samples from breast cancer and 7 samples from others) both male (N = 18) and female (N = 72) in the age group of 18–75 years were collected. Both the cases are selected in such a way that they are free from any abnormalities like diabetes, jaundice and bacterial infections. The urine samples were stored in the refrigerator at 4°C and were examined as such within 48 h after collection.

Chemicals

Nicotinamide adenine dinucleotide and Flavin adenine dinucleotide were obtained from Sigma-Aldrich Co. (St. Louis, MO). Stock solutions of 0.5 mmol of both were prepared using deionised water and was used as standards.

Native fluorescent measurements

The excitation–emission matrices were obtained using a commercially available spectrofluorometer (Fluoromax-2 SPEX, Edison, NJ) in the following ranges, 250–450 nm for excitation and 270–750 nm for emission. In EEM measurements, the spectral band passes were kept as 5 nm in both excitation and emission and data were collected at a wavelength interval of 20 nm. The steady state fluorescence emission spectra of urine samples and the standard solutions were also measured using the same spectrofluorometer. The samples were excited at 405 nm and the native fluorescence emission spectra were recorded in the region 425–750 nm. The details about the excitation source, detectors and excitation, emission monochromators were already discussed [26]. Excitation and emission slit width were fixed as 5 nm. The acquisition interval and the integration time were maintained as 1 nm and 0.1 s, respectively.

Statistical analysis

Detailed statistical analysis of the native fluorescence emission spectral data at 405 nm excitation was carried out. This includes the following three primary steps: (1) Normalization of each fluorescence spectrum; (2) Identifying characteristic spectral features of each experimental group and introduction of different ratio parameters; and (3) Classification by receiver operator characteristic analysis and stepwise multiple linear discriminant analysis using SPSS/PC + 15 software [1, 27]. As this study is aimed to discriminate the malignant subjects from the normal subjects, the analysis was performed across 80 normal subjects and 90 patients with different cancers.

Normalization

Each fluorescence spectrum was normalized by dividing the fluorescence intensity at each emission wavelength by the peak emission intensity of the spectrum. Normalizing a fluorescence spectrum removes absolute intensity information, and the main advantage of utilizing normalized spectrum is that fluorescence intensity does not need to be recorded in calibrated units.

Introducing ratio variables

To quantify the results and to estimate the diagnostic potentiality of the present technique, 19 ratio variables were introduced. These variables were calculated using fluorescence intensities at those emission wavelengths, which represent characteristic spectral features of different groups of subjects studied.

Mean and standard error values of all the ratio variables were calculated for each group of experimental subjects. A two-sampled t-test was performed to determine the level of significance (P value) with which each ratio variable discriminates diseased subjects with respect to normal subjects.

Receiver operator characteristic analysis

Receiver operator characteristic (ROC) curve was plotted for each of the ratio parameters to determine the optimal cutoff value that would give the maximum sensitivity and specificity. The area under the curve (AUC) for each ratio variable was also calculated. The AUC must be closest to 1 for maximum discrimination efficiency [25].

Stepwise multiple linear discriminant analysis

In case of stepwise method, Wilk's lambda method was followed for entering or removing independent variables. Wilk's lambda method is a variable selection method for stepwise discriminant analysis that chooses variables for entry into the equation on the basis of how much they lower Wilk's lambda. At each step, the variable that minimizes the overall Wilk's lambda is entered. The Wilk's lambda is the most economical method, which selects independent variables that minimize Wilk's lambda. The next step is to specify the criteria for entry and removal (criteria or p criteria), or to take the defaults.

The discriminant analysis used a partial F-test (F to enter 3.84; F to remove 2.71) and a stepwise method (maximum number of steps = 38) to sequentially incorporate the set of 19 variables into a Fisher linear discriminant function. Stepwise discriminant analysis performed across n groups would generally result in the coefficients of n Fisher linear discriminant or classification functions (one for each group) and (− 1) canonical discriminant functions. The classification function of each group could discriminate only that group from the rest of the groups in the analysis. To check the reliability of our analysis, leave-one-out cross-validation was used. In this procedure, one particular sample case was eliminated and discriminant analysis was used to form a classification algorithm using the remaining samples. The resulting algorithm was then used to classify the excluded case. This process was repeated for each of the sample cases. Sensitivity and specificity values were determined by comparing the spectroscopic classification with histopathological examination.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References

Excitation–emission matrices of urine samples from normal human subjects and patients with cancer

It is well documented that the urine is one of the potential diagnostic biofluids to monitor the physiological and pathological conditions, as urine has many metabolites in it [20, 22]. Among these, a number of them are native fluorophores [19]. It may not be possible to characterize all the fluorophores present in urine using conventional fluorescence emission spectroscopy. This is because, many native fluorophores exhibit similar emission characteristics and their spectra may be perturbed due to various photophysical phenomenons such as energy transfer process, self-quenching etc. Hence, it is worth to monitor the absorption and emission characteristics of multicomponents present in the urine using EEMs. This may help us in monitoring the entire spectral signature as a function of excitation and emission wavelength and to identify suitable spectral regions of excitation and emission wavelength for further analysis. In this context, a comparison of undiluted urine samples of 80 normal subjects and 90 cancer patients of different etiological condition using EEMs was carried out. Fig. 1a–c represents the three-dimensional (3D) and Fig. 2a–c represents the two-dimensional (2D) EEMs plots. The 3D plot clearly indicates that the EEMs of urine of normal subjects have an emission peak around 439 ± 10 nm when excited at 360 ± 10 nm. On the other hand, the diseased samples exhibit distinct emission peaks, such as 435 ± 10 nm, 513 ± 10 nm and 530 ± 10 nm for excitations at 360 ± 10 nm, 383 ± 10 nm and 449 ± 10 nm, respectively.

image

Figure 1. Averaged three-dimensional native fluorescence EEMs of undiluted urine samples of (a) normal, (b) cancer, in the wavelength region of 250–450 nm for excitation and 270–750 nm for emission. (c) Shows the difference between normal and cancer subjects. The excitation and emission slit width were kept as 5 nm.

Download figure to PowerPoint

image

Figure 2. Averaged two-dimensional native fluorescence EEMs of undiluted urine samples of (a) normal, (b) cancer, in the wavelength region of 250–450 nm for excitation and 270–750 nm for emission. (c) Shows the difference between normal and cancer subjects. The excitation and emission slit width were kept as 5 nm.

Download figure to PowerPoint

In the EEMs of human urine, it is practically impossible to attribute the fluorescence (emission) peak to a specific fluorophore because many fluorophores in urine possess similar spectral characteristics, some may be buried due to the intervening effects that affects the spectra. Therefore, the separation of the composite spectrum into the components corresponding to single fluorophores has not been attempted, and our discussion focuses on the overall features of the mixture of all fluorophores present in urine. To make tumor-associated spectral deviations more clearly discernible, we calculated a standard EEM by averaging the normalized EEMs of the healthy control group and from this, the averaged EEMs of malignant subjects are subtracted according to

  • display math

where D is the matrix of the difference EEM; Dij is the difference of the fluorescence intensities (percent of the maximum intensity) at a given excitation wavelength i and emission wavelength j; Nij is an element of the normal EEM; Nmax is the fluorescence intensity of the peak maximum of the normal subject; Cij is an element of the cancer-EEM to be subtracted; and Cmax is the fluorescence intensity of the peak maximum of the cancer-EEM. This kind of subtraction reveals differences in curve shape rather than differences in absolute intensities [28].

From the difference graph (Figs. 1c and 2c), it is evident that the normal and diseased samples show significant difference in the UV and visible regions. Irrespective of the excitation wavelength, the samples show considerable differences in the visible region. It shows two regions of positive deviations (characteristic maximum around 405/473 and 294/484 nm) and a region of negative deviations (characteristic minimum around 383/530 nm).

Fluorescence emission spectroscopy of urine of normal subjects and patients with cancer

Although EEMs reveal that there is a considerable emission in both UV–Visible excitation wavelengths, we limited our discussion for emission spectra at 405 nm excitation wavelength to verify the spectral signature of urine of normal and cancer patients at the selected wavelength excitation may be absorbed by the fluorophores, NADH, flavin and porphyrin. In the visible excitation wavelength regions, the emission spectrum at 405 nm excitation shows the emission signatures of many fluorophores viz., NADH, flavins and endogenous porphyrin. The averaged fluorescence emission spectra of urine of normal (n = 80) and cancer patients (n = 90) at 405 nm excitation are shown in Fig. 3a. From Fig. 3a, it is observed that the fluorescence emission spectrum of normal urine sample shows a primary peak around 478 nm, a secondary peak around 617 nm with weak hump around 580 nm. In the case of diseased conditions, the primary peak is considerably red shifted with a major peak around 520 nm. Besides these, the spectrum has humps around 617 nm. The averaged normalized fluorescence emission spectra of both the conditions of urine sample are also given in Fig. 3b. From the averaged normalized spectrum of diseased case, the weak band around 578 nm is absent with respect to the normal spectrum and also one can see that the peak around 617 nm is prominent in the case of normal subjects when compared with the diseased subjects.

image

Figure 3. Averaged fluorescence emission spectra of urine of 80 normal and 90 cancer patients at 405 nm excitation. (a) absolute spectra and (b) normalized spectra along with the difference between them and the emission spectrum of nicotinamide adenine dinucleotide (NADH) and flavin adenine dinucleotide (FAD) standard fluorophores as inset.

Download figure to PowerPoint

To know whether there is any difference between different regions of cancer, the averaged fluorescence emission spectra of urine from 80 normal subjects and patients with head and Neck cancer (n = 31), cancer of cervix (n = 39) and breast cancer (n = 13) were also compared (Fig. 4a). The overall emission intensity in the wavelength region 425–700 nm is 50% less for normal with respect to urine of cancer patients. Among the three types of cancers, cervical cancer exhibits more emission than the other two in the order of cervical cancer > breast cancer > head and neck cancer > normal. The normal spectrum has a peak centered at 478 nm with small humps around 580 and 617 nm, respectively. However, all the abnormal urine samples exhibit a major peak around 520 nm. There is a 40 nm red shift with respect to normal urine emission peak. Besides these red shifts, abnormal samples exhibit distinct hump around 478 nm, which is absent in normal urine and very weak hump with respect to normal urine is seen at 617 nm. Similarly, the hump at 580 nm which is seen in normal is absent in cancer patients. From the difference spectrum (Fig. 4b), it is also observed that the difference is positive at wavelength lesser than the isosbestic point (518 nm) and negative at wavelengths greater than the isosbestic point. This may be due to the increased emission intensity with respect to normal subjects at wavelength greater than the isosbestic point and it is lesser at wavelength lesser than the isosbestic point. This increase and decrease in cancer patients may be due to variation in the distribution and contribution of protein-free and -bound flavin and porphyrin derivatives and NADH. This clearly indicates that there is a considerable variation between normal and samples with abnormality.

image

Figure 4. Averaged fluorescence emission spectra of urine of 80 normal and 90 cancer patients of different origins at 405 nm excitation. (a) absolute spectra and (b) normalized spectra along with the difference between them.

Download figure to PowerPoint

Results of statistical analysis

To quantify and characterize the general features of the fluorescence emission spectral differences between normal subjects and patients with cancer, 19 ratio variables were introduced and calculated for each fluorescence emission spectrum at 405 nm excitation. The mean (±) standard error values of these ratio variables for normal and diseased group of samples studied are given in Table.1 along with the P values. The P values of all the 19 ratio variables are <0.05, indicating a very high statistical significance. The results of the ROC and stepwise multiple linear discriminant analysis performed are given below.

Table 1. Mean values (± SD) of the ratio parameters used for statistical analysis along with their P values
Wavelength of excitationParameterNormalCancerSignificance
405 nmI500/I525 (V1)1.17 ± 0.090.89 ± 0.160.000
I550/I525 (V2)0.69 ± 0.030.79 ± 0.070.000
I625/I475 (V3)0.18 ± 0.050.42 ± 0.470.002
I525/I575 (V4)2.13 ± 0.251.94 ± 0.270.000
I470/I475 (V5)0.99 ± 0.160.96 ± 0.050.000
I470/I520 (V6)1.14 ± 0.160.65 ± 0.300.000
I470/I525 (V7)1.21 ± 0.180.67 ± 0.320.000
I470/I580 (V8)2.79 ± 0.731.51 ± 0.880.000
I470/I617 (V9)4.58 ± 1.462.85 ± 1.510.000
I475/I520 (V10)1.15 ± 0.140.67 ± 0.300.000
I475/I525 (V11)1.22 ± 0.170.68 ± 0.310.000
I475/I580 (V12)2.81 ± 0.701.55 ± 0.880.000
I475/I617 (V13)4.57 ± 1.432.92 ± 1.490.000
I520/I525 (V14)1.05 ± 0.011.02 ± 0.020.000
I520/I580 (V15)2.41 ± 0.342.17 ± 0.360.000
I580/I617 (V16)1.63 ± 0.291.99 ± 0.340.000
I480/I520 (V17)1.15 ± 0.140.68 ± 0.290.000
I480/I525 (V18)1.15 ± 0.140.67 ± 0.290.000
I520/I575 (V19)2.25 ± 0.291.98 ± 0.310.000

Discrimination by receiver operator characteristic curve

Figure 5 shows the ROC plot for the 19 ratio variables. The cutoff values for each ratio variable and their respective sensitivity and specificity values are given in Table 2. From the table, it is observed that the AUC for the variables V1, V6, V10, V11, V17 and V18 are closer to 1 (≈ above 0.9). Among these, V11 provides 91.30% sensitivity and 83.30% specificity.

image

Figure 5. ROC curves for the ratio variables used in the discriminant analysis. The optimum cutoff point was defined as the closest point on the ROC curve to the point (X, Y) = (0, 1), where X = 1 – specificity and Y = sensitivity.

Download figure to PowerPoint

Table 2. Cutoff value, AUC ± SE, sensitivity and specificity of the ratio variables
ParametersCutoff valueAUC ± SESensitivity (%)Specificity (%)
I500/I525 (V1)1.0550.939 ± 0.0290.0084.40
I550/I525 (V2)0.7250.090 ± 0.0220.0015.60
I625/I475 (V3)0.1850.209 ± 0.0345.0022.20
I525/I575 (V4)1.9350.703 ± 0.0480.0051.10
I470/I475 (V5)0.9840.752 ± 0.0475.0066.70
I470/I520 (V6)0.9390.930 ± 0.0290.0080.00
I470/I525 (V7)0.9830.929 ± 0.0290.0081.10
I470/I580 (V8)2.1890.865 ± 0.0381.3077.80
I470/I617 (V9)3.5780.780 ± 0.0475.0065.60
I475/I520 (V10)0.9520.936 ± 0.0291.3081.10
I475/I525 (V11)0.9880.938 ± 0.0291.3083.30
I475/I580 (V12)2.2530.867 ± 0.0380.0078.90
I475/I617 (V13)3.5820.777 ± 0.0476.3063.3
I520/I525 (V14)1.0380.919 ± 0.0287.5081.10
I520/I580 (V15)2.2440.686 ± 0.0470.0060.00
I580/I617 (V16)1.7280.209 ± 0.0340.0022.20
I480/I520 (V17)0.9660.936 ± 0.0290.0083.30
I480/I525 (V18)0.9620.925 ± 0.0291.0380.00
I520/I575 (V19)2.1010.745 ± 0.0471.3065.60

Discrimination by stepwise multiple linear discriminant analysis

The stepwise multiple linear discriminant analysis performed across the whole set of 90 cancerous patients and 80 normal subjects resulted in the following expression for a canonical discriminant function (DF):

  • display math

It is seen from the above expression for canonical discriminant function, of 19 ratio variables are introduced, that only five ratio variables turned out to be significant and were included in the linear discriminant function. Fig. 5 shows the scatter plot of the discriminant score i.e. the value of the discriminant function for normal and cancerous individuals. The above canonical discriminant function provides 95% sensitivity and 94.4% specificity.

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References

Although there is a considerable improvement in the treatment and reconstructive modalities, cancer remains a major threat to the public. In this context, early diagnosis is considered to be one of the clinical importances, as it may reduce patient morbidity and mortality. Based on these, many showed interest to probe the possibility of using native fluorescence spectroscopy as one of the methods in early diagnosis of cancer. Several studies had been reported on the native fluorescence spectroscopy of various tissues [2, 3, 29-31]. As any metabolic changes in the body may reflect in the body fluids, attempts have also been made on the characterization of body biofluids, viz., blood in diagnosis of liver diseases and cancer [1, 14]. It has also been reported that urine is a diagnostically important biofluids, which has many metabolites. Furthermore, it was reported that many of the urine metabolites are native fluorophores [19]. It is well established that riboflavin absorbed through the intestine is synthesized to flavin mononucleotide (FMN) and FAD, coenzymes of flavin enzymes, in tissues. On the other hand, free riboflavin is being considered to be the dominant flavin compound excreted into human urine, and the examination on flavin metabolites in human urine has been scarcely done [32]. Data reported on the use of native fluorescence spectroscopy of urine in clinical oncology. For example, Rabinwitz had reported that the ratio of red fluorescence and blue fluorescence from urine may be considered for studying the tumor growth and its altered mechanism [23]. Subsequently, Masilamani et al. reported the detection of cancer by native fluorescence [25]. However, the native fluorescence spectroscopic characterization of urine in diagnostic oncology is still in the primitive stage. In this context, attempts were made to characterize the native fluorescence spectroscopy of urine by the excitation–emission matrices and fluorescence emission spectral measurement at 405 nm excitation. The fluorescence emission spectral data were also subjected for statistical analysis to verify whether there exists any diagnostic possibility.

As urine has native fluorophores as metabolites, EEMs were measured to identify the location of the different native fluorescence peaks and their relative fluorescence emission contributions. Figure 1a,b compares the averaged three-dimensional presentation of EEMs of human urine of normal subjects and cancer patients. From the figure, it is observed that there is a considerable difference in the EEMs topograms between them. The 3D EEMs of normal subject show that there is a maximum emission around 440 ± 10 nm at excitation 360 ± 10 nm. However, it is observed that 3D EEMs of cancer subjects have three peaks at 439 ± 10 nm, 513 ± 10 nm and 530 ± 10 nm at excitations 360 ± 10 nm, 383 ± 10 nm and 449 ± 10 nm, respectively. The emission around 440 ± 10 nm may be ascribed to the presence of NADH or/and NAD(P)H. The additional peaks observed at longer excitation–emission wavelengths viz., 513 ± 10 and 530 ± 10 nm may be attributed to flavin. Similar observations were also made from the averaged 2D contour plots of normal and cancerous subjects Fig. 2a,b.

To further identify special emission features, attempts were also made to compute the difference EEMs between normal and cancer group by both 3D and 2D plot (Figs. 1c & 2c). From the difference graph, it is evident that the two regions of positive deviations (characteristic maximum around 405/473 and 294/484 nm) may be due to NAD(P)H and a regions of negative deviations (characteristic minimum around 383/530 nm) may be due to flavins.

On the basis of the observations made from EEMs, we have also carried out the measurements of fluorescence emission spectra at different excitation wavelengths. Among various excitation wavelengths, the emission spectra for excitation at 405 nm are considered to be an optimized wavelength for both spectral and statistical analysis as it provides emission contribution of NADH, flavin and endogenous porphyrin.

From the steady-state native fluorescence measurements, it is observed that the urine of different types of cancer patients exhibits significant spectral signatures with respect to normal subjects. Figure 3a compares the averaged fluorescence emission spectrum of cancer subjects with that of normal subjects. It is observed that the averaged emission spectrum of urine from normal subjects exhibits a broad emission maxima in the region 460–525 nm. The major peak at 478 nm may be attributed to enzyme bound NAD(P)H [1] and the small humps centered around 510 and 617 nm may be due to flavin and porphyrins [2]. On the other hand, cancerous samples exhibit a primary emission peak around 520 nm, which may be attributed to flavin with a small hump at 478 nm. Furthermore, there is a considerable emission around 617 nm in normal subjects, which is not observed in the case of cancer subjects. This may be due to the overwhelming emission contribution of flavin over NADH in cancer urine than that of normal subjects. In general, protein-bound flavin such as FAD gives less fluorescence over FMN. Cancer cells utilize more FAD for energy metabolism [33]. During the unfolding of flavoprotein, the structure of the protein may be disrupted and interactions with flavin break down, usually leading to the increase in flavin in blood plasma. The fluorophores NADH and flavin that is not bound to proteins in the plasma are filtered by glomerulus and excreted in urine [34]. This is one of the reasons for more fluorescence in urine as it has more free riboflavin. The other possible reason for more fluorescence in the urine of cancer patients is due to the presence of more FMN than FAD as the fluorescence quantum yield(Qf) of FMN (Qf = 0.27) is more than FAD (Qf = 0.032) [35]. Furthermore, the interlocking coordination may be affected in the defective cell during the transformation of normal into cancer. This may influence the distribution of ATP, ADP and NADH in defective cells [36]. In this context, free and/or protein-bound flavin and NADH may be considered as metabolomic markers of cancer, which are present in urine. However, the biochemical pathways relating riboflavin with cancer is highly complex. This is because certain patients with cancer excrete less riboflavin than do normal individuals [37]. In this context, it is essential to understand the details of the uptake, binding and removal of riboflavin and its derivatives from the normal and neoplastic tissues and also the excretory mechanisms. Similarly, different opinion has been reflected about the difference in emission of porphyrins in human urine. It has been reported that the level of porphyrin in blood plasma of abnormal cases is elevated when compared with the normal subjects without any increase in urinary porphyrins due to the insufficient supply of iron to the bone marrow [1]. In this regard, Masilamani et al. have reported that the porphyrin peak is distinct in the abnormal case when compared with the normal case, which is contrary to our results. Furthermore, based on the report on the native fluorescence of tissues and biofluids, the peak around 478 nm is attributed to unbound, free NADH [25], and 617 nm is due to endogenous porphyrins [2]. From Fig. 4a, the emission maxima around 520 nm from urine of cancerous patients is differing in the order of cervical cancer > breast cancer > head and neck cancer > normal subjects. This variation in flavin emission with different types of cancer may be due to variation in flavin excretion. The fluorescence maxima of cancer subjects is red shifted with respect to the normal subjects may be due to the acidic microenvironment [3]. The prominent maxima of head and neck cancer, breast cancer and cervical cancer exhibit a 14, 13 and 9 nm blue shift, respectively, instead of 531 nm suggesting an accumulation of positive ions causing the flavins to emit at shorter wavelengths. Our current results supports the earlier published results of Alfano et al., [2]. From Fig. 3a, it is also observed that the averaged fluorescence spectrum of normal subjects shows a small hump around 580 nm. The origin for this emission is not known and it is almost absent in the case of cancer subjects.

By considering the difference in spectral signature between cancer and normal subjects, ROC and stepwise multiple linear discriminant analysis were carried out to distinguish cancer patients from the normal subjects. To quantify these observed spectral differences among normal and cancerous subjects, 19 ratio variables were computed from fluorescence intensities at different emission wavelengths and were introduced in the analysis. Each emission wavelength used in these ratio variables represents a specific characteristic spectral feature of one or more groups. For instance, I470, I475 and I480 have been chosen to represent the NADH. I500, I520, I525 and I550 represent the contribution of flavins. I617 has been chosen to represent the characteristic emission due to endogenous porphyrin. Although the origins of I575 and I580 are not known, they have also been included in the statistical analysis for giving good discrimination. Based on the ROC analysis, the samples are discriminated with 90–91% sensitivity and 80–84% specificity. Among many ratio variables, V11 provides 91.3% sensitivity and 83.3% specificity. On the other hand, stepwise multiple linear discriminant analysis was performed across normal and cancerous samples shows that of 80 normal subjects, 77 subjects are correctly discriminated and four subjects are misclassified as cancer patients yielding a sensitivity of 95%, whereas of 90 cancer subjects, 85 are correctly classified and five cases are misclassified as normal subjects yielding a specificity to detect cancer subject is 94.4% (Table 3, Fig. 6). In these discriminant analyses, 94.70% of the original grouped cases and 94.10% of the cross-validated grouped cases were correctly classified.

image

Figure 6. Scatter plot showing the distribution of DF for normal (○) and cancer (●) subjects. The algorithm shows a sensitivity and specificity of 95% and 94.4%, respectively, in discriminating the normal subjects from the cancer subjects.

Download figure to PowerPoint

Table 3. Classification results of the discriminant analyses
CasesActual groupPredicted group membershipTotal% of correctly classified cases
NormalCancer
  1. * Bold figures represent the specificity or sensitivity value of the corresponding group.

OriginalNormal 95.0 5.094.7
Cancer5.6 94.4
Cross-validatedNormal 93.8 6.294.1
Cancer5.6 94.4

Conclusion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References

In conclusion, the photophysical characteristics of urine of normal and cancer subjects were carried out by EEMs and fluorescence emission spectroscopic technique. As flavin is being considered to be one of the metabolites in urine, fluorescence emission characteristics of normal and cancerous subjects at 405 nm have alone be considered for spectral comparison and for detailed statistical analysis. The discrimination reveals that fluorescence emission spectra excited at 405 nm discriminates the cancer patients from normal subjects with a sensitivity and specificity of 95% and 94.4%, respectively, with an overall accuracy of 94%. On the basis of these observations, it is obvious that the present technique may be used as a complementary technique to the existing conventional methods of disease diagnosis. In this preliminary report, we have grouped cancer samples from different etiology as cancer for statistical analysis with normal subjects. Further studies are to be carried out to collect more data on various cancer patients and to analyze the possibility of discriminating the different etiologies of cancer. To elucidate the possible reason for the altered spectral signatures due to the native fluorescence from flavin and its derivatives and the relation connecting to riboflavin metabolism, further combined studies by the biochemist, oncologist and physicians are required.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References

This study was supported by Board of Research in Nuclear Sciences, Department of Atomic Energy, Government of India, Project no. 2009/34/38/BRNS/3206. We thank Dr.M.Bagavandas, School of Public Health, SRM University, Kattankulathur, for the support in statistical analysis.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  • 1
    Madhuri, S., N. Vengadesan, P. Aruna, D. Koteeswaran, P. Venkatesan and S. Ganesan (2003) Native fluorescence spectroscopy of blood plasma in the characterization of oral malignancy. Photochem. Photobiol. 78, 197202.
  • 2
    Alfano, R. R., D. B. Tata, J. Cordero, P. Tomashefsky, F. W. Longo and M. A. Alfano (1984) Laser induced fluorescence spectroscopy from native cancerous and normal tissue. IEEE J. Quant. Electron. 20, 15071511.
  • 3
    Alfano, R. R., G. C. Tang, A. Pradhan, W. Lam, D. S. J. Choy and E. Opher (1987) Fluorescence spectra from cancerous and normal human breast and lung tissues. IEEE J. Quant. Electron. 23, 18061811.
  • 4
    Heintzelman, D. L., U. Utzinger, H. Fuchs, A. Zuluaga, K. Gossage, A. M. Gillenwater, R. Jacob, B. Kemp and R. R. Richards-Kortum (2000) Optimal excitation wavelengths for in vivo detection of oral neoplasia using fluorescence spectroscopy. Photochem. Photobiol. 72, 103113.
  • 5
    Glassman, W. S., M. Steinberg and R. R. Alfano (1994) Time resolved and steady state fluorescence spectroscopy from normal and malignant cultured human breast cell lines. Lasers Life Sci. 6, 9198.
  • 6
    Schantz, S. R., H. E. Savage, P. Sacks and R. R. Alfano (1997) Native cellular fluorescence and its application to cancer prevention. Environ. Health Perspect. 105, 941944.
  • 7
    Ganesan, S., P. G. Sacks, Y. Yang, A. Katz, M. Al-Rawi, H. E. Savage, S. P. Schantz and R. R. Alfano (1998) Native fluorescence spectroscopy of normal and malignant epithelial cells. Cancer Biochem. Biophys. 16, 365373.
  • 8
    Drezek, R., C. Brookner, I. Pavlova, L. Boiko, A. Malpica, R. Lotan, M. Follen and R. Richards-Kortum (2001) Autofluorescence microscopy of fresh cervical-tissue sections reveals alterations in tissue biochemistry with dysplasia. Photochem. Photobiol. 73, 636641.
  • 9
    Muller, M. G., T. A. Valdez, I. Georgakoudi, V. Backman, C. Fuentes, S. Kabani, N. Laver, Z. Wang, C. W. Boone, R. R. Dasari, S. M. Shapshay and M. S. Feld (2003) Spectroscopic detection and evaluation of morphologic and biochemical changes in early human oral carcinoma. Cancer 97, 16811692.
  • 10
    Cai, W., X. Ni, S. K. Gayen and R. R. Alfano (2006) Analytical cumulant solution of the vector radiative transfer equation investigates backscattering of circularly polarized light from turbid media. Phy.Rev. E. 74, 056605.
  • 11
    Cong, W., H. Shen, A. Cong, Y. Wang and G. Wang (2007) Modeling photon propagation in biological tissues using a generalized Delta-Eddington phase function. Phy.Rev. E. 76, 051913.
  • 12
    Madhuri, S., S. Suchitra, T. G. Srinivasan, P. Aruna and S. Ganesan (1999) Native fluorescence characteristics of blood plasma of normal and liver diseased subjects. Med. Sci. Res. 27, 635639.
  • 13
    Masilamani, V., R. Kalaivani, O. AI-Daghri, H. Raja, S. E. Sivanandam, L. Anand, S. Ganesan, A. R. Diab, M. AI-SaIhi, C. Mohan, R. Thirunanlyan and K. Vijayasarathi (2005) Optical diagnosis of cancer from blood components. Egypt. J. Biophys. 1, 530.
  • 14
    Lualdi, M., A. Colombo, E. Leo, D. Morelli, A. Vannelli, L. Battaglia, E. Poiasina and R. Marchesin (2007) Natural fluorescence spectroscopy of human blood plasma in the diagnosis of colorectal cancer: feasibility study and preliminary results. Tumori 93, 567571.
  • 15
    Bigler, L. R. and C. F. Streckfus (2009) Salivary biomarkers for the detection of malignant tumors that are remote from the oral cavity. J. Lab. Clin. Med. 29, 7185.
  • 16
    Lima, D. P., D. G. Diniz, S. A. S. Moimaz, D. H. Sumida and A. C. Okamoto (2010) Saliva: reflection of the body. Int. J. Infect. Dis. 14, e184e188.
  • 17
    Leiner, M. J. P., M. R. Hubmann and O. S. Wolfbeis (1987) The total fluorescence of human urine. Anal. Chim. Acta 198, 1323.
  • 18
    Dubayova, K., J. Kusnır and L. Podracka (2003) Diagnostic monitoring of urine by means of synchronous fluorescence spectrum. J. Biochem. Biophys. Methods 55, 111119.
  • 19
    Kusnir, J., K. Dubayova, L. Leskova and M. Lajtar (2005) Concentration matrices - solutions for fluorescence definition of urine. Anal. Lett. 38, 15591567.
  • 20
    Perinchery, S. M., U. Kuzhiumparambil, S. Vemulpad and E. M. Goldys (2010) The potential of autofluorescence spectroscopy to detect human urinary tract infection. Talanta 82, 912917.
  • 21
    Saude, E. J., D. Adamko, B. H. Rowe, T. Marrie and B. D. Sykes (2007) Variation of metabolites in normal human urine. Metabolomics 3, 439451.
  • 22
    Anwer, A. G., P. M. Sandeep, E. M. Goldys and S. Vemulpad (2009) Distinctive autofluorescence of urine samples from individuals with bacteriuria compared with normals. Clin. Chim. Acta 401, 7375.
  • 23
    Rabinowitz, H. M. (1949) A Correlation of fluorescence of human urine with benign and malignant growth. Cancer Res. 9, 672676.
  • 24
    Wan, Y. Q., L. J. Tang and T. Tan (2010) Simultaneous determination of xanthopterin and isoxanthopterin in human urine by synchronous fluorescence spectroscopy. J. Fluoresc. 20, 11911198.
  • 25
    Masilamani, V., T. Vijmasi, M. A. Salhi, K. Govindaraj, A. P. Vijaya-Raghavan and B. Antonisamy (2010) Cancer detection by native fluorescence of urine. J. Biomed. Opt. 15, 057003.
  • 26
    Ebenezar, J., P. Aruna and S. Ganesan (2010) Synchronous fluorescence spectroscopy for the detection and characterization of cervical cancers in vitro. Photochem. Photobiol. 86, 7786.
  • 27
    Sivabalan, S., C. P. Vedeswari, S. Jayachandran, D. Koteeswaran, C. Pravda, P. Aruna and S. Ganesan (2010) In vivo native fluorescence spectroscopy and nicotinamide adinine dinucleotide/flavin adenine dinucleotide reduction and oxidation states of oral submucous fibrosis for chemopreventive drug monitoring. J. Biomed. Opt. 15, 017010.
  • 28
    Leiner, M. J. P., R. J. Schaur, G. Desoye and O. S. Wolfbeis (1986) Fluorescence topography in biology. III: Characteristic deviations of tryptophan fluorescence in sera of patients with gynaecological tumours. Clin. Chem. 32, 19741978.
  • 29
    Richards - Kortum, R. and E. Sevick – Muraca (1996) Quantitative optical spectroscopy for tissue diagnosis. Annu. Rev. Phys. Chem. 47, 555606.
  • 30
    Ramanujam, N. (2000) Fluorescence spectroscopy of neoplastic and non-neoplastic tissues. Neoplasia. 2, 89117.
  • 31
    Wagnieres, G. A., W. M. Star and B. C. Wilson (1998) In vivo fluorescence spectroscopy and imaging for oncological applications. Photochem. Photobiol. 68, 603632.
  • 32
    Ohkawa, H., N. Ohishi and K. Yagi (1983) New metabolites of riboflavin appear in human urine. J. Biol. Chem. 258, 56235628.
  • 33
    Powers, H. J. (2003) Riboflavin (vitamin B-2) and health. Am. J. Clin. Nutr. 77, 13521360.
  • 34
    Gropper, S. S., J. L. Smith and J. L. Groff (2009) Water- soluble vitamins. In Advance Nutrition and Human Metabolism (Edited by P. Adams, A. Lustig, E. Feldman and E. Downs), pp. 329333. Wadsworth, Cengage Learning, Belmont, CA.
  • 35
    Munro, A. W. and M. A. Noble. (1999) Fluorescence analysis of flavoproteins. In Methods in Molecular Biology, Flavoprotein Protocols, Vol. 131 (Edited by S. K. Chapman and G. A. Reid), pp. 2548. Humana Press Inc., Totowa, NJ.
  • 36
    Mikirova, N., H. D. Riordon and P. Rillema (2003) Detection of energy metabolism level in cancer patients by fluorescence emission from serum. J. Ortho. Med. 18, 924.
  • 37
    Rivlin, R. S. (1973) Riboflavin and cancer: A Review. Cancer Res. 33, 19771986.