Fluorescence Liquid Biopsy for Cancer Detection Is Improved by Using Cationic Dendronized Hyperbranched Polymer

(1) Background: Biophysical techniques applied to serum samples characterization could promote the development of new diagnostic tools. Fluorescence spectroscopy has been previously applied to biological samples from cancer patients and differences from healthy individuals were observed. Dendronized hyperbranched polymers (DHP) based on bis(hydroxymethyl)propionic acid (bis-MPA) were developed in our group and their potential biomedical applications explored. (2) Methods: A total of 94 serum samples from diagnosed cancer patients and healthy individuals were studied (20 pancreatic ductal adenocarcinoma, 25 blood donor, 24 ovarian cancer, and 25 benign ovarian cyst samples). (3) Results: Fluorescence spectra of serum samples (fluorescence liquid biopsy, FLB) in the presence and the absence of DHP-bMPA were recorded and two parameters from the signal curves obtained. A secondary parameter, the fluorescence spectrum score (FSscore), was calculated, and the diagnostic model assessed. For pancreatic ductal adenocarcinoma (PDAC) and ovarian cancer, the classification performance was improved when including DHP-bMPA, achieving high values of statistical sensitivity and specificity (over 85% for both pathologies). (4) Conclusions: We have applied FLB as a quick, simple, and minimally invasive promising technique in cancer diagnosis. The classification performance of the diagnostic method was further improved by using DHP-bMPA, which interacted differentially with serum samples from healthy and diseased subjects. These preliminary results set the basis for a larger study and move FLB closer to its clinical application, providing useful information for the oncologist during patient diagnosis.


Introduction
In recent years, cancer has emerged as the second leading cause of death worldwide, accounting for almost 10 million deaths per year, according to estimates from World Health Organization (WHO) [1]. Early detection and diagnosis of cancer is the best means for reducing the associated mortality and morbidity associated with it because long term survival rates drop significantly once metastasis has occurred [1]. Developing novel noninvasive (or minimally invasive) strategies for early-stage cancer diagnosis with sensitivity may contribute considerably to the success of cancer therapy and the improvement of patients' survival rate.
In the last decade, liquid biopsy has arisen as a promising non-invasive strategy for cancer diagnosis to overcome the limitations of traditional techniques. Liquid biopsy refers to the analysis of tumor biomarkers isolated from biological fluids of cancer patients [2]. The peripheral blood of a cancer patient contains a pool of cells and cell products derived from the tumor, including circulating tumor cells (CTCs), platelets, circulating free (cf) DNA or RNA, exosomes, proteins and metabolites [3][4][5][6]. All this information makes liquid biopsy an interesting high-information-content technique for its application in early detection of cancer, in risk estimation for metastatic relapse, and cancer evolution and treatment response monitoring [7], avoiding invasive methods of sample collection, such as a tissue biopsy.
Mass spectrometry [8,9], calorimetry [10][11][12][13][14], and fluorescence spectroscopy [15] are some of the biophysical techniques that have been used for cancer patient diagnosis by analyzing blood serum. In the case of fluorescence, it is a useful technique traditionally employed for characterizing the structure of biomolecules and its interactions. Detection by fluorescence spectroscopy requires excitation of the biological sample with a radiation beam at a certain wavelength, which promotes fluorescence emission from the different components of the biological sample (mainly macromolecules present in the sample) at larger wavelengths [16]. When there is a pathological state, biochemical, physicochemical, and histological changes occur, and they are reflected in tissue or serum composition. Therefore, it could be reasonable to think that an altered fluorescence emission spectrum would be detected. On this basis, several studies have been published showing that fluorescence spectroscopy can be applied to the diagnosis of tumor diseases and reported that when comparing samples from individuals with and without pathology, the fluorescence properties are different [17,18]. This phenomenon has been studied in both solid tissue [19,20] and serum samples. Breast, cervix, colon, and oropharyngeal cancer [21][22][23][24] were some of the pathologies studied.
Advances in nanotechnology and nanoscience have led to the creation of novel horizons for the use of nanomaterials in cancer diagnosis. Among these, dendrimers (threedimensional hyperbranched polymers) have gathered increasing attention because of their unique structural properties, including monodispersity, internal molecular cavities, tunable size, and numerous peripheral functional groups [25,26]. The use of dendrimers as nanocarriers for drug and contrast agent delivery for cancer treatment and diagnostic imaging has been intensively investigated by several research groups [27][28][29]. More recently, the multivalency of dendrimers has attracted research interest for the fabrication of dendrimerbased biosensors for biomarker detection. The multiple functional groups on the surface of dendrimers are used to immobilize antibodies or aptamers combined with highly sensitive electrochemical or fluorescence methods for specific tumor biomarker detection [30][31][32][33][34][35][36]. However, the use of dendrimers for biomarker detection purposes can also be figured out through the direct interaction of dendrimers with proteins and biomolecules in a biological fluid, thus approaching the principles of liquid biopsy mentioned above. Indeed, it is well studied that dendrimers, as other nanomaterials, interact with proteins and other biomolecules in a biological fluid, and this provokes the formation of the so-called protein corona around them. Although this phenomenon can bring limitations to drug delivery applications [37], interest has been reported for exploiting it for a variety of medical applications [38] and, particularly, for diagnosis purposes [39,40].
The interaction between dendrimers and proteins in blood depends on the peripheral features of the dendrimer, such as positive or negative charge [41], and also the characteristics of protein contents of a given blood sample, and this can be detected by fluorescence spectroscopy [42][43][44]. Typically, tryptophan residues give rise to fluorescent emission upon excitation at 330 nm, and protein modifications around them cause fluorescence changes (either changes in emission intensity or characteristic emission wavelength) that can be detected [45]. This was taken as the basis of this study, in which we assessed a novel and simple approach for cancer detection using a dendronized hyperbranched polymer based on bis(hydroxymethyl)propionic acid (DHP-bMPA) (Figure 1a). DHP-bMPA is based on a fourth generation hyperbranched polymer of 2,2-bis(hydroxymethyl)propionic acid, bis-MPA [46] conjugated at its periphery with third generation bis-MPA dendrons bearing eight terminal glycine moieties. Previous studies conducted by our group demonstrated that some cationic DHP-bMPA derivatives were suitable for DNA transfection [47] and drug delivery [48]. However, their potential as a diagnostic tool remains to be explored [49]. The hypothesis underlying our work is: (1) pathology alters the composition of blood serum (protein content by down-and up-regulation of certain disease-associated proteins, protein modification, and protein-associated metabolites); (2) dendrimers interact differentially with those proteins exhibiting an altered pattern due to the disease; and (3) fluorescence spectroscopy provides discriminating information for these phenomena (fluorescence liquid biopsy, FLB). Here, DHP-bMPA was used in combination with fluorescence spectroscopy in order to improve the classification between cancer and healthy subjects. In this pilot study, 20 pancreatic ductal adenocarcinoma (PDAC) samples and 24 ovary cancer samples were analyzed and compared with their healthy counterparts. Fluorescence spectroscopy was performed in the presence or absence of DHP-bMPA and a biparametric analysis of the fluorescence emission curves was carried out in order to obtain a single fluorescence spectroscopy score (FS score ) to assess blood serum alterations and predict risk of disease (PDAC or ovarian cancer). The addition of DHP-bMPA resulted in an improved classification of subjects, with an overall improvement in statistical specificity and sensitivity of the classification model in both cancer patient cohorts, when compared with the model elaborated in the absence of dendrimer. Despite the low number of samples considered in this study, this novel strategy exhibited excellent potential for improving clinical cancer diagnostic in a quick, minimally-invasive, and cost-effective manner.

Synthesis and Morphological Characterization of DHP-bMPA
The dendronized hyperbranched polymer, DHP-bMPA, used for this study was synthesized from a tert-butyloxycarbonyl (t-Boc) protected DHP-bMPA precursor bearing an average of 424 t-BocNH groups, which was previously prepared and reported by our group [45]. The amino groups were deprotected with trifluoraoacetic acid. The t-Boc cleavage reaction was followed by Fourier transform infrared (FTIR) spectroscopy ( Figure 1b). The chemical structure of the trifluoroacetate product, DHP-bMPA ( Figure 1a) was confirmed by nuclear magnetic resonance (NMR) and FTIR spectroscopy (Materials and Methods).
The morphological characterization of DHP-bMPA was carried out by transmission electron microscopy (TEM). The sample was prepared by depositing a 0.5 mg mL −1 solution of DHP-bMPA in PBS (phosphate-buffered saline) on a copper grid. DHP-bMPA appeared as rounded objects with average diameter of 15.9 ± 2.2 nm calculated from TEM images ( Figure 1c). This size is consistent with the expected formation of unimolecular micelles well dispersed in aqueous solution, and in reasonable agreement with dynamic light scattering determinations (Supplementary Table S1). No aggregation, turbidity, or color changes were observed upon mixing DHP-bMPA with serum.

Preliminary Fluorescence Studies
At 330 nm wavelength, tryptophan residues in proteins are excited and emission fluorescence curves recorded from 400 to 700 nm compile the information of the proteins' aromatic residues and their environment (neighboring residues exposure to solvent molecules). All serum proteins and their potential interactions with other serum proteins and metabolites would be reflected in this fluorescence spectrum [45]. In order to set up experimental conditions, preliminary fluorescence studies were first carried out with a set of blood donors (BD) samples (three different samples taken as example from the BD group), both in the absence ( This could reflect that a different interaction between DHP-bMPA and the protein content of BD samples occurred as they were extracted from different subjects, and this leads to an inherent interindividual variability. Also, the normalization of the fluorescence spectra highlighted a signal shift to lower wavelength values in the case of BD/DHP-bMPA (Inset in Figure 2a). In an attempt to study further the origin of these differences, the next step consisted in comparing the experimental fluorescence signal of BD/DHP-bMPA and the theoretical fluorescence signal obtained as the sum of both fluorescence spectra, BD and DHP-bMPA, recorded separately. Both types of spectra are plotted in Figure 2b. It can be observed that the experimental spectra recorded for the three BD/DHP-bMPA samples cannot be explained by the sum of the individual fluorescent spectra of DHP-bMPA and BD, and this is consistent with the existence of interaction of DHP-bMPA with the protein content in BD samples. Furthermore, there is a displacement at 525-550 nm between the experimental and theoretical fluorescence maxima (Figure 2b, Inset) that cannot be explained by the DHP-bMPA signal contribution. This shift could be related to BD protein residues specifically contributing to the fluorescence of the sample in that wavelength range and that can be altered upon interacting or binding to DHP-bMPA.
These preliminary results revealed that some serum components interact with DHP-bMPA, resulting in changes in the environment of aromatic residues of serum proteins, and these changes could be detected using fluorescence spectroscopy. Accordingly, we planned to study if this interaction measured by fluorescence might be used for detecting serum protein content differences triggered by the disease and this could lead to distinguish between healthy controls and a certain pathological status based on serum samples analysis.

Fluorescence Studies of Patient Serum Samples
Two different pathologies, pancreatic ductal adenocarcinoma cancer (PDAC) and ovarian cancer (OV), were studied to explore the applicability of the interaction with DHP-bMPA to detect a pathological status based on fluorescence spectra.    When comparing mean curves signal values (Figure 6a,b), it is apparent that the presence of DHP-bMPA promoted changes in the shape (around 525 nm) and the maximum value of fluorescence signals (about 20% increase) only in the OV group ( Figure 6b). OC samples spectra did not change with DHP-bMPA incubation. Again, as it occurred in PDAC disease samples, DHP-bMPA promoted different changes in OC and OV, thus supporting the objective of using fluorescence spectroscopy of serum samples to assess disease-induced blood serum alterations and assign a probability of cancer from serum samples.

Evaluation of the Potential Diagnostic Capability of DHP-bMPA
According to the results described above, DHP-bMPA seems to interact with proteins from serum samples producing different responses (changes in fluorescence emission) in PDAC and OV samples compared to their respective BD and OC control groups. These observations fostered the statistical evaluation of fluorescence spectra in order to confirm the possibility of using the incubation treatment with DHP-bMPA to discriminate between healthy and disease status. This would impart DHP-bMPA with diagnostic capability.

Fluorescence Curves Parametrization and Statistical Analysis Applied to PDAC Samples Fluorescence Signals Parametrization
The fluorescence signal was analyzed by calculating two parameters: G 1 and WL max (according to Equations (1) and (2) in Section 3). The skewness, G 1 , describes the asymmetry of the spectrum, and WL max is the wavelength for the maximal intensity in the spectrum. The methodology we developed was protein concentration-independent (avoiding the problems derived from protein concentration determinations). A complete descriptive analysis of the parameters can be found in Supplementary Table S2.

Statistical Analysis of Parameters: Univariate Analysis
Thus, DHP-bMPA promoted changes in G 1 values of BD, with an increase of 35%, and PDAC, with an increase of 20%. WL max values decreased 23 nm in BD and 5.5 nm in PDAC, and this change seemed to be the reason for the loss of significance between BD and PDAC previously observed. Based on these data, the effect of DHP-bMPA in presence of serum samples could be more remarkable in BD group.
The statistical analysis of G 1 and WL max parameters from fluorescence spectra of BD and PDAC serum samples was performed. Despite the apparent similarity of the fluorescence spectra of the BD group and PDAC patients in the absence of DHP-bMPA (Figures 3a and 4a), there were statistically significant differences (p < 0.050) in fluorescence curves-associated parameters (G 1 and WL max ) when comparing both BD and PDAC (Supplementary Table S3). When DHP-bMPA was added to serum samples (Figure 3b,c and Figure 4a), fluorescence curves-associated parameters changed (Supplementary Table S3): G 1 maintained its significant difference between groups (p < 0.050), but WL max increased its p-value to 0.379 (losing statistical significance). If we have a deep look to the parameter values, WL max mean values without DHP-bMPA were 18 nm higher in BD, and G 1 was 0.04 lower in PDAC patients. Accordingly, the presence of DHP-bMPA enhanced the difference in the mean values of G 1 (∆G 1 = 0.12) and brought closer WL max values between BD and PDAC, until they were almost the same (∆WL max = 1.50 nm) (Figure 7). Thus, DHP-bMPA promoted changes in G 1 values for BD, with an increase by 35%, and PDAC, with an increase by 20%. WL max values decreased by 23 nm in BD and by 5.5 nm in PDAC, and this change seemed to be the reason for the loss of statistical significance between BD and PDAC previously observed. Based on these data, the effect of DHP-bMPA on serum samples could be more remarkable in the BD group.
Statistical Analysis of Parameters: Multivariate Analysis As described above, monovariant analysis of the individual parameters (G 1 and WL max ) obtained from the fluorescence spectra showed that they were statistically different (p-value < 0.050). We constructed a classification model with these two parameters and calculated a single classification index, called FS score (Fluorescence Spectrum Score), for classifying samples from each group (BD vs. PDAC), following a methodology previously described in our group [12].
We elaborated two scores calculated on the basis of the parameters G 1 and WL max , FS score and FS score -DHP-bMPA, (that is, FS score for samples untreated and treated with DHP-bMPA, respectively) and evaluated their capability to classify subjects and, therefore, predict risk of PDAC. The results, gathered in Table 1, indicated that G 1 and WL max parameters included in FS score -DHP-bMPA were statistically significantly different (p < 0.050), while in FS score only the WL max parameter was statistically significant. Therefore, when using DHP-bMPA, the model was improved. The values of the FS score comprises a range between 0 and 1. Typically, a 0.5 threshold is used. FS score values > 0.5 indicate that the model is classifying a serum sample as altered (that is, reflecting PDAC disease), whereas FS score values < 0.5 indicate that the model was classifying a serum sample as non-altered (that is, not reflecting any disease, closer to the BD group).
A good performance of the model corresponds to maximizing the number of true positives and negatives and minimizing the number of false positives and negatives. The results indicated that when using DHP-bMPA, the model performed considerably better. That is, there were less false positives and negatives. The number of true negatives (FS score -DHP-bMPA < 0.5 in BD group) and true positives (FS score -DHP-bMPA > 0.5 in PDAC patients) increased in comparison to the FS score -based model. Similarly, the number of false positives (FS score -DHP-bMPA > 0.5 in BD group) and false negatives (FS score -DHP-bMPA < 0.5 in PDAC patients) decreased in comparison to the FS score -based model ( Table 2). ↑ means that the value increases when using DHP-bMPA; ↓ means that the value decreases when using DHP-bMPA.
Summarizing the statistical performance, FS score exhibited good performance indexes, according to the success rate (75.6%) and area under the Receiver Operating Characteristic (ROC) curve (0.78 AUC), with better specificity (88.0%) than sensitivity (60.0%) (Figure 8, ROC curve, Table 3). When using DHP-bMPA, FS score -DHP-bMPA improved all indexes, with better success rate (91.1%), larger AUC (0.89), and higher specificity (96.0%) and sensitivity (85.0%). In addition, according to the Net Reclassification Index (NRI) of Pencina [50], 33.0% of patients were more correctly reclassified using FS score -DHP-bMPA compared to FS score . This reclassification involved mainly the PDAC group (25.0%), not the BD group (8.0%).  The statistical analysis of the FS score values showed significant differences between BD and PDAC group (p = 0.001). Differences were higher for FS score -DHP-bMPA (p < 0.001) (Supplementary Table S4, Figure 9). The improvement in the predictive capability of FS score when DHP-bMPA was employed can be observed in Figure 10. There were three BD subjects with FS score > 0.5 (12.0% false positive rate) and eight PDAC patients with FS score < 0.5 (40.0% false negative rate). Data were improved by using DHP-bMPA: 1 BD subject with FS score -DHP-bMPA > 0.5 (4.0% false positive rate) and three PDAC patients with FS score -DHP-bMPA < 0.5 (6.7% false negative rate). Figure 10. Distribution of the FS score (left) and FS score -DHP-bMPA (right) within BD (green) and PDAC (red) groups. The lines represent an equivalent Gaussian distribution. The FSscore threshold for discriminating between an unaltered and altered fluorescence spectrum is 0.5 (dotted horizontal line).

Fluorescence Curves Parametrization and Statistical Analysis Applied to OV Samples Fluorescence Signals Parametrization
As in PDAC disease, the fluorescence spectra were analyzed by calculating the same two parameters, G 1 and WL max . A complete descriptive analysis of the parameters can be found in Supplementary Table S5.

Statistical Analysis of Parameters: Univariate Analysis
The statistical analysis of G 1 and WL max from fluorescence spectra of OC and OV samples was performed (Supplementary Table S6) and they did not exhibit significant differences (p > 0.050). However, the presence of DHP-bMPA promoted changes in G 1 and WL max , rendering them statistically different between the two groups, OC and OV (p < 0.001) for both parameters. If we analyze the parameters, WL max values increased by 3 nm in OC and by 20.5 nm in OV group when DHP-bMPA was employed. The same effect was observed in the case of G 1 , which increased by 0.1 in OC and by 0.4 in cancer patients when DHP-bMPA was used. The boxplots ( Figure 11) of G 1 and WL max comparing both groups, OC and OV samples, reflected the effect of the presence of DHP-bMPA. Based on these results, it could be said that the effect of DHP-bMPA on the fluorescence spectrum of serum samples was more remarkable in OV than in the OC group. Having in mind the previous results regarding PDAC, for which BD spectra were more affected, DHP-bMPA seemed to interact in a different way with serum proteins depending on the disease. This is consistent with a different serum protein composition in different diseased and control subjects.

Statistical Analysis of Parameters: Multivariate Analysis
As in the case of PDAC, we constructed two scores calculated on the basis of the two parameters, G 1 and WL max , FS score and FS score -DHP-bMPA, (that is, FS score for samples untreated and treated with DHP-bMPA, respectively) and evaluated their capability to predict risk of ovarian cancer. The results, gathered in Table 4, indicated that G 1 parameter included in FS score -DHP-bMPA was statistically significantly different (p < 0.050). Therefore, when using DHP-bMPA, the model was improved. Again, a 0.5 threshold was set (FS score > 0.5 classified a sample as diseased status, and FS score < 0.5 classified a sample as non-diseased status) and the results indicated that the model performed better when using DHP-bMPA. There were more true positives and negatives and less false positives and negatives when using the dendrimer in the experimental protocol; the number of true negatives (FS score -DHP-bMPA < 0.5 in OC group) and true positives (FS score -DHP-bMPA > 0.5 in OV group) increased compared to FS score model. Similarly, the number of false positives (FS score -DHP-bMPA > 0.5 in OC group) and false negatives (FS score -DHP-bMPA < 0.5 in OV group) decreased compared to FS score model (Table 5). FS score showed poor performance indexes, according to success rate (57.1%), area under curve (0.61 AUC), and statistical specificity and sensitivity around 45.8% and 68.0%, respectively ( Figure 12, Table 6). When using DHP-bMPA, FS score -DHP-bMPA showed much better performance indexes: success rate (91.8%), area under curve (0.98), specificity (92.0%), and sensitivity (91.7%). Also, according to Net Reclassification Index (NRI) of Pencina [50], approximately 70.0% of patients were correctly reclassified using FS score -DHP-bMPA. This reclassification involved mainly the OV group (45.8%), and to a lesser extent the OC group (24.0%). Delog test was used to compare areas under curves (AUC), and they were statistically significantly different (p < 0.001) in their AUC.  The statistical analysis of the FS score values indicated there were no significant differences between OC and OV groups (p = 0.174), but remarkably significant differences for FS score -DHP-bMPA (p-value < 0.001) (Supplementary Table S7, Figure 13). Figure 13. Distribution of the probability score (FS score and FS score -DHP-bMPA) within OC (green) and OV (red) groups. The p-value (Wilcoxon test) indicates there is a statistically significant difference between OC and OV groups (p-value < 0.001) in FS score -DHP-bMPA.
The excellent improvement in the predictive value of FS score when DHP-bMPA was included in the experimental protocol can be observed in Figure 14. There were 8 OC samples with FS score > 0.5 (32.0% false positive rate) and 13 OV samples with FS score < 0.5 (54.2% false negative rate). The classification was improved by using DHP-bMPA: There were only two OC samples with FS score -DHP-bMPA > 0.5 (8.0% false positive rate) and two OV samples with FS score -DHP-bMPA < 0.5 (8.3% false negative rate). Figure 14. Distribution of the FS score (left) and the FS score -DHP-bMPA (right) within OC (green) and OV (red) groups. Notes: The lines represent an equivalent Gaussian distribution. The FS score threshold for discriminating between an unaltered and altered fluorescence spectrum is 0.5 (dotted horizontal line).

Synthesis and Characterization of the Dendronized Hyperbranched Polymer DHP-bMPA
The t-Boc-protected dendronized hyperbranched polymer (DHP-bMPA) was synthesized as previously reported by us [47]. Terminal amines were deprotected as depicted in Scheme 1, by dissolving t-Boc protected DHP-bMPA (1.00 eq.) in a mixture of chloroform and trifluoroacetic acid (5 mL, 1:1 volume/v). The reaction mixture was stirred at room temperature for 3 h. The solvent and the excess of trifluoroacetic acid were removed under vacuum. Then, the product was dissolved in methanol and precipitated in cold ether. After decanting, the remaining solvent traces were removed under reduced pressure and the product was obtained as a white solid. 1

Dendrimer Characterization Techniques
1 H and 13 C Nuclear Magnetic Resonance (NMR) experiments were performed using an AV-400 ( 1 H: 400 MHz, 13 C: 100 MHz) spectrometer (Bruker, Billerica, MA, USA) employing deuterated methanol (CD 3 OD) or deuterated dimethyl sulfoxide ((CD 3 ) 2 SO) as solvents. The chemical shifts are given relative to tetramethylsilane (TMS) in ppm and the coupling constants in Hz; as internal standard, the solvent residual peak was used for spectrum calibration. Infrared spectra were recorded between 4000 and 600 cm -1 on a Vertex 70 spectrophotometer (Bruker, Billerica, MA, USA), which worked in attenuated total reflection (ATR) mode.
Transmission electron microscopy (TEM) images were recorded on a TECNAI T20 system (FEITM, Hillsboro, OR, USA) with a beam power of 200 kV. A droplet of a solution of the sample at 0.5 mg mL −1 in PBS was deposited on a holey carbon film 300 mesh coppered grids (Electron Microscopy Sciences, Hatfield, PA, USA) and a 3% w/v aqueous solution of phosphotungstic acid was used as negative stain.

Subjects and Samples
Experimental blood samples were obtained from four different sets, grouped two by two with the appropriate controls (diseased vs. controls): pancreatic cancer patients (PDAC) vs. blood donors (BD); ovarian cancer patients (OV) vs. benign ovarian cyst patients (OC). All subjects gave their informed consent for inclusion before they participated in the study. The study was conducted in accordance with the Declaration of Helsinki, and the protocol was approved by the Ethics Committee of CEICA (PI16/0228).

Pancreatic Cancer (PDAC): Patients (PDAC Diseased Group) and Blood Donor Subjects (BD Control Group) Cohort Sample Description
Samples and data from blood donor subjects as control group patients included in this study were obtained from blood donors at Blood and Tissue Bank of Aragon and provided by the Biobank of the Aragon Health System, integrated in the Spanish National Biobanks Network (PT20/00112). They were processed following standard operating procedures with the appropriate approval of the Ethics and Scientific Committees. The cohort consisted of 25 serum samples from Spanish Caucasian subjects, apparently diseased-free.
A total of 20 serum samples from patients diagnosed with pancreatic cancer at the oncology service at Hospital Universitario Miguel Servet in Zaragoza (Spain) were used in this study. Samples were collected after histopathological confirmation and prior to treatment. They were handled, stored, and provided by Biobank of the Aragon Health System.
The age from PDAC group is statiscally higher than the age from BD group (p < 0.001, Wilcoxon test, Table 7). In the BD group, the age from males was statistically higher than the age from females (p = 0.036, t-test, Table 8). In the PDAC group, there were no statistically significant differences in age depending on gender (p = 0.703, Wilcoxon test, Table 8). Samples from ovarian cancer (24) and benign cysts (25) patients diagnosed at the gynecology service at Hospital Virgen del Camino in Pamplona (Spain) were included in this study. Serum samples were collected before surgery, handled, stored, and provided by the Biobank Navarrabiomed integrated in the Spanish National Biobanks Network (PT17/0015/0007). They were processed following standard operating procedures with the appropriate approval of the Ethics and Scientific Committees.
We recruited 49 samples, 24/49 (49.0%) were from female patients diagnosed of ovarian cancer (OV), and 25/49 (51.0%) were from female patients with initially suspected ovarian cancer, but finally diagnosed as benign (OC). There was only one missing age data in the OV group.
The age from OC group is statistically higher than age from OC group (p = 0.011, Wilcoxon test, Table 9).

Blood Sample Processing
Peripheral blood (10 mL) from each subject was collected and transferred into serum separator tubes for subsequent analysis. Tubes were gently mixed immediately after blood collection and then centrifuged at 3200× g rpm for 10 min. Separated serum was carefully aspirated to avoid hemolysis and contamination of the separated blood phases, and immediately stored in 1 mL aliquots at −80 • C until analysis.

Fluorescence Spectroscopy
Serum samples were diluted 1:25 in PBS and, when appropriate, mixed with DHP-bMPA nanoparticles to a final concentration of 500 µg mL −1 prior data acquisition. A total volume of 80 µL was transferred into 96-well microplates (96-well PCR plate, non-skirted, from 4titude, Surrey, UK). Fluorescence measurements were performed in a CLARIOstar plate reader (BMG Labtech, Germany) selecting an excitation wavelength of 330 nm, with excitation and emission bandwidths of 10 nm, and recording the fluorescence emission spectra in the range 400-700 nm. The excitation wavelength was slightly higher than optimal for tryptophan excitation (295 nm), but was a constraint from the instrument. Multimode plate readers with fluorescence capabilities very seldom offer the possibility to use lower excitation wavelengths.
All the fluorescence spectra, F(λ), where F is the fluorescence emission intensity and λ is the wavelength, were normalized to an area equal to 1, and two parameters (G 1 and WL max ) were calculated according to the following equations: where <λ> is the average wavelength (or first moment) considering the spectrum as a probability density function: m 2 and m 3 are the second and third moments of that density distribution, G 1 is the parameter that provides a measure of the asymmetry or skewness of the spectrum. In addition, the wavelength corresponding to the maximal fluorescence intensity of the spectrum (mode of the probability density function), WL max , is a central measure providing information about the location of the main region of the spectrum.

Statistical Analysis
Fluorescence raw data were processed using Origin software (OriginLab, Northampton, MA). Two parameters summarizing the essential geometric features of each fluorescence spectrum were directly calculated, G 1 and WL max , and were used for calculating the FS score to compare fluorescence spectra and build the classification model by using a generalized linear model (GLM). FS score provides the probability of an individual fluorescence spectrum for reflecting serum alterations. Such FS score is a single value employed for classifying a given subject as control or diseased. As any probability, the FS score ranges between 0 and 1: the model classifies the subjects with FS score values > 0.5 as individuals with high probability of having serum alterations (i.e., diseased status), and the subjects with FS score values < 0.5 as individuals with low probability of having serum alterations (i.e., healthy status). The performance of the diagnostic test was evaluated by calculating common statistical performance indexes (e.g., sensitivity, specificity) and the receiver operating characteristic (ROC) curve. Net Reclassification Index (NRI) of Pencina and/or Delong test were performed to compare area under curve (AUC) ROC.
The Kolmogorov-Smirnov test was performed to assess the normal distribution of the variables. Averages between two independent groups were compared with the t-test in normal distributions. Medians between two independent groups were compared with the Wilcoxon test in non-normal distributions. Comparisons between groups for qualitative variables were made using a chi-squared test.

Conclusions
There is an urgent necessity for an efficient, reliable, quick, non-invasive diagnostic test for cancer patient management, applicable not only for early diagnosis, but also for patient stratification and prioritization, as well as personalized patient surveillance and monitoring. New diagnostic methods can be created by applying newly developed experimental techniques or by improving traditional ones through new technological or analytical developments. In this work, we present a novel diagnostic test consisting of using the fluorescence emission signal obtained from serum samples of patients with cancer (PDAC or ovary) by adding a specific type of dendritic polymer (a cationic dendronized hyperbranched polymer, DHP-bMPA). The differences observed in the fluorescence emission of patients' serum samples compared to the controls were parametrized in two spectra features: the skewness (G 1 ) and the wavelength at the maximum signal value (WL max ). The statistical analysis of these parameters led us to define two scores, FS score and FS score -DHP-bMPA (without and with DHP-bMPA, respectively). Accordingly, we were able to increase the performance of the test for classifying individuals into healthy and cancer groups with very good indicators (accuracy, specificity and sensitivity). It must be emphasized that this is a quick, minimally invasive, risk-free test that can be applied routinely, and it only requires a sample obtained from a blood draw and a fluorescence plate reader, a non-expensive equipment available at many clinical laboratories. The findings herein reported set the basis for an easy and versatile approach to liquid biopsy that combines fluorescence spectroscopy and nanomaterials. These are preliminary results that open a larger study to move fluorescence spectroscopy closer to the clinical application, providing useful information to the oncologist during patient diagnosis.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/ijms22126501/s1, Table S1: Dynamic light scattering data, Table S2: Main descriptive indexes for the two fluorescence curves-associated parameters, in each clinical group (BD and PDAC), in the presence and absence of DHP-bMPA, Table S3: Univariate analysis of fluorescence curves-associated parameters of BD group and PDAC group in presence and absence of DHP-bMPA, Table S4: FS score and FS score -DHP-bMPA values comparison between BD and PDAC, Table S5: Main descriptive indexes for the two fluorescence curves-associated parameters, in each clinical group, OC and OV, without and with DHP-bMPA, Table S6. Univariate analysis of parameters of OC and OV groups, Table S7. FS score and FS score -DHP-bMPA values comparison between OC and OV.

Institutional Review Board Statement:
The study was conducted in accordance with the Declaration of Helsinki, and the protocol was approved by the Ethics Committee of CEICA (PI16/0228).
Informed Consent Statement: All subjects gave their informed consent for inclusion before they participated in the study.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.