Visualization of the composition of the urinary fluorescent metabolome. Why is it important to consider initial urine concentration?

Urine is a highly complex fluorescent system, the fluorescence of which can be affected by many factors, including the often-ignored initial urine concentration in comprehensive fluorescent urine analysis. In this study, a total urine fluorescent metabolome profile (uTFMP) was created as a three-dimensional fluorescence profile of serial synchronous spectra of urine diluted by geometric progression. uTFMP was generated using software designed for this purpose after recalculating the 3D data concerning the initial urine concentration. It can be presented as a contour map (top view) or as a more illustrative and straightforward simple curve, thus usable in various medicinal applications.


Introduction
Biological systems such as cells, tissues, and body fluids contain many native fluorophores exhibiting autofluorescence. Under physiological conditions, their composition and concentration in the system are relatively stable. In contrast, the pathological process may affect their concentration, mutual relation, and the change in the local microenvironment, which can be manifested by altered fluorescence. Autofluorescence of biological systems is a phenomenon with huge potential in non-invasive diagnostics of pathological states [1,2] 3D fluorescence analysis in various graphic realizations, mainly in the form of an excitation-emission matrix (EEM), allows the description and visualization of the composition of the fluorescent metabolome of a given biological system. Thus, comparing the analyzed sample with a defined standard clearly identifies differences caused by qualitative or quantitative changes in compositions in response to a pathological process. Although the first applications of 3D fluorescence analysis for the graphical characterization of biological systems were introduced four decades ago, the technique has not yet become established in routine diagnostics [3][4][5][6][7].
A special type of 3D fluorescence analysis represents a synchronous spectrum obtained by scanning the fluorescence during the synchronous (parallel) motion of the excitation and emission monochromator at the same rate but with a predefined constant wavelength difference Δλ (CW spectra). Lloyd [8] introduced these types of non-conventional fluorescence spectra to characterize multicomponent mixtures, and Vo-Dinh subsequently applied them to the study of biological systems [9].
However, despite the higher resolution, the frequent mismatch of Δλ with the Stokes shift complicates the interpretation of CW synchronous spectra. Characterizing the spectrum at the chosen difference in Δλ requires the experimental definition of the fluorophores forming the individual peaks. [10,11].
Due to the concentration stability of the microenvironment of cells, the in vitro and in vivo fluorescence analysis of tissues is analytically simpler than the fluorescence analysis of body fluids, the most complex of which is urine. Although easily accessible, intensely naturally fluorescent body fluid [23,38], urine forms an extremely demanding multi-fluorescence system. The presence of several fluorophores with very similar fluorescence characteristics, or changes in magnitude in concentrations of different fluorophores, high inter and intra variability in concentration as well as the influence of many exogenous factors (e.g. diet or drugs) on the final urine composition explains the reason why fluorescence analysis of urine rarely occurs in clinical laboratory diagnostics.
The problem of urine concentration variability was solved after introducing the concentration matrix model [25], which presented a method of solving urine concentration variability in fluorescence analyses for the first time. The urine sample was diluted via geometrical progression so that the linearity of fluorescence versus fluorophore concentration was identifiable over the entire range of excitation wavelengths and fluorophores. Subsequently, the 3D processing of a series of these spectra into the form of a contour map was named fluorescent concentration matrix (uFC-Matrix), graphically defining the quantitative and qualitative composition of the fluorescent urine metabolome. uFCMatrix provides an idea of the composition, the initial concentration of the urine sample, and the amount of water contained [26,27]. Kollarik et al [34] applied uFCMatrix to analyze urologic patients, and Birková et al to characterize the urine of patients with malignant melanoma [31]. The application of uFCMatrix also inspired Martinicky et al [33] in screening female patients with ovarian cancer.
uFCMatrix is comprehensive and extremely rich in information. For fluorescence analytical practice, it represents an irreplaceable aid in 'tracking' fluorophores and in the selection of the optimal urine dilution for further detailed fluorescence analysis of individual fluorophores. However, it can be difficult to 'read' uFCMatrix for general medical practice (except for assessing total urine concentration). Birková et al [30] proposed a simplified graphic representation of uFCMatrix in the form of a single curve as a frontal view of a series of synchronous spectra at different dilutions of urine. This curve represents the urine fluorescence profile (uFP).
The uFP is more illustrative and straightforward than the uFCMatrix but does not include the initial concentration of the urine sample, which shows considerable concentration variability. The presented paper addresses this issue by processing the uFCMatrix data into a urinary total fluorescent metabolome profile (uTFMP), where the initial urine concentration is considered.
Chemometric methods, artificial intelligence, and their application in accessing information from complicated analytical systems, such as body fluids, significantly advance diagnostic possibilities [11,16,19,30,40,41]. Information about the metabolome composition obtained by a suitable analytical technique must first be adequately processed so that the subsequent application of mathematical procedures is effective and provides unbiased conclusions.
This study focuses on the optimal model of data processing obtained by specific 3D fluorescence analysis of urine for visualization of the fluorescence metabolome of urine and subsequent data mining. It modifies the urine fluorescence profile by considering the initial urine concentration.

Samples
Spectral data from morning urine samples analyzed at our Department in the past were used in the study. To demonstrate the concentration-related changes in the profiles, the urines of patients with various diagnoses were selected: figure 4: both samples -suspicion of nephrotic syndrome; figure 5: patient 1 -colon adenocarcinoma, patient 2 -malignant prostate and bladder cancer, patient 3 -bladder cancer in remission, patient 4 -invasive urothelial carcinoma, patient 5bladder cancer in remission + administration of folic acid, patient 6-bladder cancer in remission, atypical urothelial cells, proteinuria. The urines of healthy persons with a negative semi-quantitative strip analysis were used for comparison with the patients' urines.

Measurements and data processing
Before fluorescent measurements, urine samples were processed according to Kusnir et al [25]. Briefly, 5 ml of fresh urine was centrifuged (10 min, 1100 rpm), after which the supernatant was transferred into a new test tube. The set of urine concentrations was prepared by deionized water dilution of the supernatant via geometric progression as follows: undiluted, 1/4; 1/ 16; 1/64.; 1/ 256; 1/1024. The samples were measured using quartz cuvettes with a path length of 1 cm and a volume of 3.5 ml. The simple synchronous spectra were scanned with a constantly set wavelength difference Δλ = 30 nm using the Luminescence Spectrofluorimeter Perkin Elmer LS 55 (USA) and visualized as the dependence of the fluorescence intensity (y-axis) on the excitation wavelength (x-axis) in the range 250-550 nm in the FLWinlab.
The concentration matrix (uFCMatrix) was generated in the FLWinlab program from the synchronous spectra of a geometrically diluted urine sample from the undiluted sample (pf = 0) to the 1000-fold waterdiluted sample (pf = 3) (modification of [25]). uFC-Matrix in the 3D arrangement shows the dependence of the fluorescence intensity on the excitation wavelength and the negative logarithm of the volume fraction of urine in a given dilution (F = f (λex; pf)).
The author's software generated a total urinary fluorescent metabolome profile (uTFMP) as a profile of simple synchronous spectra of individual urine sample dilutions (pf) multiplied by the appropriate factor k n for a given dilution n = 2-5 or n = 2-6 only for extremely concentrated urine. n = 1 for undiluted urine, n = 6 for 1000 times diluted urine. 2.3. Authors' software for creating a total fluorescent metabolome profile (uTFMP) Data processing for the construction of the model was done with an authors' software created especially for this purpose. The data processing and visualization program was written in Python 3. The program processed data from .sp files (in ASCII format) obtained by FLWinlab software. The data were processed by common Python libraries, NumPy and SciPy. The Mathplotlib library was used for data visualization. PyCharm by JetBrains was chosen as an integrated development environment (IDE). The program is available through a Git repository that guarantees the current version of the program among all team members. The results shown in figures 4 and 5 were achieved with the authors' program.

Results and discussion
3.1. Fluorescent metabolome of urine Urine in its entirety is characterized by the blue-green fluorescence affected mainly by metabolites of tryptophan and catecholamines present in urine in a concentration about ten times higher than other fluorophores. Significant urinary fluorophores and their fluorescence characteristics can be seen in figure 1.
For the visualization of the fluorescent metabolome composition, synchronous fluorescence spectrum Δλ = 30 nm has been administered, enabling the distinction of the basic groups of fluorophores and allowing a sensitive response to changes in the composition of the metabolome of urine. However, the interpretation of the spectrum is not simple. Although this spectrum captures most of the fluorophores of the urine metabolome, except for porphyrins, λex does not always coincide with the real value of excitation maxima. Therefore, fluorophores' position in the synchronous excitation spectrum must be determined experimentally after adding the given metabolite into the urine. The experimentally assigned excitations to the individual metabolites are listed in table 1.

Visualization of the fluorescent metabolome of urine
Visualizing the composition of the urinary fluorescent metabolome based on the scan of the synchronous excitation spectrum made it possible to introduce a fourth important concentration dimension into the fluorescence analysis. Synchronous excitation spectra Δλ = 30 nm of the geometric dilution progression of the urine sample were arranged in series, forming a 3D fluorescence body, which can be displayed as a contour map -top view, as a dependence of fluorescence

Urine fluorescent profile (uFP)
The front projection of the 3D fluorescent body represents the fluorescence profile (uFP), which is plotted as the dependence of the maximum fluorescence intensity on the excitation wavelength of the synchronous spectrum [30]. The fluorescence maximum represents the peak of the linearity of the  dependence of the fluorescence intensity on the concentration. The fluorescence profile is an illustrative view of the composition of the urinary fluorescent metabolome. However, the fact that it only partially includes the initial concentration of the urine sample and its effect on the fluorescence intensity can be considered a certain handicap.

Common approaches to urinary fluorescent metabolome analysis
Comparing the fluorescence of urines with various initial concentrations can lead to misinterpretation in the composition of the fluorescent metabolome, making a serious problem in the diagnostic process, where changes caused by different urine concentrations can be attributed to the pathological process. Therefore, the processing of spectra related to the initial concentration of urine to compare different urines and identify pathological changes in the composition of the fluorescent urinary metabolome can represent an analytical and interpretative advantage.
Briefly, to compare different approaches in the visualization of the fluorescent metabolome of urine and the possible misinterpretation of their changes, two samples with different initial urine concentrations and similar fluorescent metabolome composition were selected (figure 4): urine A with low pf values, indicating low initial urine concentration, and in contrast, urine B with higher values corresponding to more concentrated urine.

Fluorescence analysis of constantly diluted urine
This approach was the first when we started to analyze urine fluorescence [24]. The discrepancy between high urinary protein concentrations in patients with kidney disease and zero fluorescence intensity in the protein fluorescence region was the impetus for addressing the problem of analyzing urinary fluorescence associated with its concentration and/or an order of magnitude different concentrations of metabolites.
The misinterpretation of the results using this approach is shown in figure 4 (top).
Synchronous fluorescence spectra Δλ = 30 nm of equally diluted urine (1:50; gray colored area of spectrum in both cases A and B) differ significantly in their spectra (figure 4, top in the middle). The position of the spectrum in the uFCMatrix represents a gray line. The synchronous spectrum of urine A is in the linear region of group I fluorophores and partially overlaps the spectrum of group II and III fluorophores. The synchronous spectrum of urine B no longer shows group I fluorescence correctly (fluorescence quenching). In the linear region, it displays the group II and III fluorophores and, in addition, also indicates the presence of fluorophores of the IVth group. Assessment of the differences between the samples based on evaluating such spectra would distort the conclusions due to the concentration difference of the initial urine concentration.

Fluorescence analysis of undiluted urine
This is a widespread approach to urine analysis. The importance of the initial concentration of the two urines mentioned above, A and B, was again manifested by significant differences in the spectrum's shape (figure 4; yellow colored area of spectrum in both cases A and B). The position of the spectrum in the uFCMatrix represents the yellow line. In both samples, fluorescently silent fluorophores of the Ist and IInd groups can be found. Urine A fully captures the fluorescence of IIIrd and IVth groups of fluorophores. Urine B displays dominant fluorescence of the IVth group of fluorophores; the IIIrd group shows as little an indication of fluorescence due to quenching at high concentrations. Neither of these images reflected similarity in the composition of the metabolome, the resulting images of both urines were diametrically different.
A well-grounded analysis of undiluted urine is in those cases where of interest are the fluorophores of a low concentration in urine (e.g., group IV and porphyrins) (table 1).
In the presented cases, the essential role of the fourth dimension -the concentration -can be seen, as well as the importance of the initial visualization of the composition of the fluorescent urine metabolome using uFCMatrix.
3.4. An approach to urinary fluorescent metabolome analysis considering the initial concentration 3.4.1. Urine fluorescent profile (uFP) Various urinary fluorophores, which differ by order of magnitude in their concentrations, require a series of dilutions and subsequent processing of the scanned spectra into unity. uFP reveals all fluorophore groups regardless of their location in the uFCMatrix. Based on this image, urine A resembles the profile of urine B, getting it closer to reality, but both vary in fluorescence intensity at a certain wavelength range. The intensity of group I in both urines is the same in this profile ( figure 4 and table 2), despite their localization in uFCMatrix and different pf. The uFP only partially covers the initial concentration of the urine sample. It is suitable for qualitative visualization of the urine metabolome.

Total fluorescent metabolome profile of urine (uTFMP)
An accurate profile of the quantitative composition of the urine metabolome is created after consideration of the dilution factor and multiplication of the individual spectra of uFCMatrix by a specific dilution value. The dominant peak observed in the wavelength range of 250-300 nm results from the intense fluorescence of tryptophan and its metabolites, catecholamines, and indole derivatives. Due to the orders of magnitude different concentrations of individual groups of fluorophores (I-IV), the potential creation of a real fluorescence profile (consider the absolute numbers of the fluorescence intensity at a high dilution of urine, e.g., 250-or 1000-fold diluted) would be unsuitable for a graphical comparison of the complex The display of the urine composition by applying uTFMP visualizes all groups of fluorophores (of high and low concentrations) and partially deals with the differences in the initial urine concentrations simultaneously. After the inclusion of the dilution factor, the profiles of both urines (A and B) are very similar and better reflect the concentration dependence Ist to 3rd groups fluorophores ( figure 4).
Comparative analysis of both types of profiles (uFP and uTFMP) reveals significant differences in the intensities of zones I. and III. ( figure 4 and table 2). While at uFP, the intensity of the I. center is the same in both urines, urine B characterizes a higher fluorescence intensity in uTFMP that can be evident from the pf position in uFCMatrix. The maximum intensity of zone I for urine A is at pf = 1.8 (urine diluted approximately 60 times), and for urine B, it is at pf = 2.4 Table 2. Comparison of two urines A and B with different initial concentration and their visualization results using two different approaches uFP and uTFMP. Similar results are highlighted in the grey box and the ratio close to 1 is highlighted in bold. In the table, the zone area is used, not the maximum fluorescence intensity.  (urine diluted approximately 250 times). For zone III, the fluorescence center of urine A is at pf = 0, i.e., undiluted urine, but for urine B at pf = 1.2, the urine is diluted 16 times.
In contrast, the intensity of zone III is very similar in both urines in the case of uTFMP but much higher in urine A compared to urine B in the case of uFP. The position of the center of zone III in the uFCMatrix is again different in the two urines. Urine B is more concentrated than urine A, which should be considered.
For zone IV, both profiles (uFP and uTFMP) are the same, as in both urines (A and B), the zone's center has the maximum intensity in undiluted urine (pf = 0). The fluorescence of uFP is equal to that of uTFMP in only zone IV when the maximum intensity is at pf = 0.
A detailed analysis of this comparison is given in table 2, where zone area, not maximum fluorescence intensity, was used for the quantitative comparison of uFP and uTFMP. Similar areas of zones are highlighted in the grey box.
The analysis above of individual profiles confirms that uTFMP represents the most informative compromise applied to the visualization of the fluorescent urinary metabolome.

Variability of uTFMP
The results show that the uTFMP of patients may vary significantly. uTFMP reveals the differences between the individual samples and appears to be an optimal tool for visualizing the fluorescent metabolome of urine. Figure 5 displays the diversity of the fluorescent metabolome and highlights the urinary concentration variability.
Including the initial concentration of the urine sample allows a clear and straightforward comparison of fluorescence of differently concentrated urines. Visual comparison of patient urine with healthy standards immediately informs about changes (qualitative and quantitative) in urine metabolome.

Fluorescent zonation as a tool for processing and sorting profiles
Since the concentration matrices/urine profiles of patients with different diagnoses, in addition to the four primary fluorescent centers (I.-IV.), have significant peaks different from healthy individuals, complex urine fluorescence visualizing tools [26,30] and uTFMP were divided into eight zones ( figure 6). Zonation is important in the identification of metabolites as possible markers of various diseases. The classification of fluorophores into individual zones is given in table 1.
A change in the uTFMP reflects the change in metabolic composition. Identification of the change in specific zones compared to the standard (average of the concentration fluorescence profiles of healthy individuals) allows the identification of fluorophores and subsequently indicate the metabolic process associated with a given state or disease.
3.7. Software for mathematical data processing and generation of visualization of metabolome composition uTFMP is generated usually from five spectra after the spectrum of a given dilution is multiplied by the appropriate factor k n . The sixth urine dilution (1000x) is only included for uFCMatrix formation assembling. A high dilution of a urine sample with a low initial concentration would distort the real profile by multiplying the fluorescence since the peak of zone I at Δλ = 30 nm may interfere with the Raman peak of water. In urine with a very low initial concentration ( figure 5, patient 2), there is only one contour in the area of pf 2.4-3 (left corner) in uFCMatrix. In highly concentrated urines, if the uFCMatrix in area I shows a fluorescence with a tendency to exceed the y-axis range ( figure 5 patient 3), the sixth dilution of the sample (1000x) is also taken into account, and the spectrum is multiplied by the appropriate factor.
Multiplying the spectra by accurate dilution creates a data set for the real fluorescence profile (uFPreal), presenting thus the expected fluorescence for a given dilution -the so-called real/valid values. The fluorescence of uTFMP is a function of excitation wavelength and a maximum of fluorescence multiplied by the factor: F uTFMP = f(λ EX ; max(F n .k n ).
The output for the user is uFCMatrix and uTFMP, graphically defining the composition of the fluorescent metabolome ( figure 6). Deviations from the standard composition of the fluorescent metabolome are presented in the form of a table indicating the change in the relevant zone.
The software processes accurate profile data and creates uTFMP in parallel. These data provide the core for machine learning using artificial neural networks and other chemometric techniques to sort urine samples into defined diagnostic groups.

Conclusion
The potential of urinary fluorescence analysis has not yet been fully exploited in the diagnostic process. Consideration of initial urine concentration and easy comparison of fluorescent signals could bring a shift in this area. The generation of uTFMP brings several benefits: it represents a low-cost technology with the potential to be a part of non-invasive urinalysis. It meets the requirement of simple visualization of the composition of the fluorescent metabolome, simplifies sample comparisons, and provides data for machine learning. Visualization of the system composition via uTFMP does not require detailed knowledge of its composition. The system is considered as a whole. The visual record in the form of uTFMP reflects the quantitative, and qualitative representation of individual components and their mutual interactions. If the composition of the standard system is known, graphical changes can be either directly concretized or specifically targeted to the selected compound/s in further analyses.
The application of uTFMP for monitoring patients with various diagnoses and 'feeding' the software with spectral data and their derivations considering initial urine concentration in connection with a specific diagnosis could bring insight into the composition of the fluorescent metabolome of urine of such patients and may also reveal unexpected changes in the urine, either the presence of unexpected metabolites or a significantly altered concentration of normally present metabolites.

Data availability statement
The data that support the findings of this study are available upon reasonable request from the authors.

Disclosures
The authors declare no conflicts of interest.

Ethical compliance
All samples and procedures performed in this study respond to anonymous waste material and were analyzed in accordance with Helsinki Declaration and its later amendments or comparable ethical standards.