Open Access
Add to BibSonomy
Abstract
Non-muscle-invasive bladder cancer (NMIBC) is a common urinary tumor and has a high recurrence rate due to improper or inadequate conservative treatment. The early and accurate prediction of its recurrence can be helpful to implement timely and rational treatment. In this study, we explored a preoperative serum surface-enhanced Raman spectroscopy based prognostic protocol to predict the postoperative prognosis for NMIBC patients at the time even before treatment. The biochemical analysis results suggested that biomolecules related to DNA/RNA, protein substances, trehalose and collagen are expected to be potential prognostic markers, which further compared with several routine clinically used immunohistochemistry expressions with prognostic values. In addition, high prognostic accuracies of 87.01% and 89.47% were achieved by using the proposed prognostic models to predict the future postoperative recurrence and recurrent type, respectively. Therefore, we believe that the proposed method has great potential in the early and accurate prediction of postoperative prognosis in patients with NMIBC, which is with important clinical significance to guide the treatment and further improve the recurrence rate and survival time.
© 2022 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Bladder cancer (BC) is a common malignant tumor in genitourinary system, which has become the ninth most frequently-diagnosed cancer worldwide [1]. In 2020, bladder cancer is diagnosed in about 570,000 patients worldwide, leading to approximately 210,000 deaths [2]. At first diagnosis, 75% of bladder cancer patients present as non-invasive cases [3]. Although non-muscle-invasive bladder cancer (NMIBC) is considered as less malignancy and non-life-threatening, it usually suffers from a high recurrence rate due to its improper or inadequate conservative treatment, in contrast to radical cystectomy for muscle-invasive bladder cancer (MIBC) [4]. It is reported that the recurrence rate for NMIBC patients after receiving transurethral resection can be as high as 50% to 70% and about 40% of those recurrent cases even further progress to MIBC [5]. If the recurrence and muscle infiltration of bladder cancer can be precisely assessed and predicted at an early stage, it can be helpful to establish timely and rational treatment plans for NMIBC patients and to alleviate the risk of potential toxic effects of adjuvant chemotherapy or immunotherapy [6], thus early and significative prognostic factors are expected to improve the recurrence rate and survival time for patients with NMIBC significantly.
In clinical practice, the commonly used prognostic factors for NMIBC include the number of tumors, their sizes and the prior recurrence rate [7–9]. However, only those prognostic factors are usually inadequate, and the accuracy of prognostic judgment relies heavily on the urologist’s experience. In contrast to those macro-level prognostic factors, tumor markers are a subset of molecules produced by tumor cells, which can reflect tumor diagnosis, prognosis and tumor response to treatment [10]. Tumor markers can improve the prognosis of patients with recurrent bladder cancer by improving the detection of occult symptoms [11]. Some studies [12–14] have shown that the overexpression of tumor suppressor p53 can be an independent predictor for bladder cancer recurrence. However, the abnormal p53 is more frequently associated with the prognosis of MIBC other than NMIBC [15], and some researchers have also found that it is not the only key prognostic factor associated with the bladder cancer relapse. Reinert et al. have demonstrated several panels of genes in urine (e.g. EOMES, HOXA9, POU4F2, TWIST1, VIM, and ZNF154) to be potential DNA methylation biomarkers for assessing surveillance of recurrent bladder cancer [16]. Nevertheless, bladder cancer prognosis based on those DNA methylation biomarkers is still in the early stage and often suffers from low specificity, which needs to be further scientifically and clinically investigated [17,18]. Urinary biomarkers, e.g. NMP22 BladderChek, NMP22, BTA STAT, BTA Trak, ImmunoCyt, UroVysion, Cxbladder monitor and Bladder cancer (UBC) test, have also been investigated in various studies because of their non-invasive and cost-effective characteristics, however, the clinical role of urinary biomarkers in the follow-up of bladder cancer patients remains undefined due to the unsatisfactory sensitivity and specificity [19]. Besides, optical imaging modalities, such as blue-light cystoscopy and narrow-band imaging cystoscopy, are often suggested as regular follow-up examination every 3 months to screen the bladder cancer recurrence [20], in which the detection sensitivity can be improved by imaging specific molecules or collecting additional spatial and spectral information [21,22]. However, the cystoscopy examination is invasive, untimely and relatively expensive, which puts a huge burden on using cystoscopy as a regular follow-up examination for all patients. Therefore, a non-invasive, timely and cost-effective method to predict future postoperative bladder cancer recurrence with high accuracy, especially for NMIBC, would be of great clinical significance.
Raman spectroscopy, a rapid and non-destructive spectroscopic technique based on Raman effect, can collect the information about molecular vibration and rotation by analyzing the frequency shifts of the acquired Raman signal [23]. Such technique can be applied on both in-vivo and in-vitro samples and has been widely investigated in many clinical applications, especially for cancer diagnosis [24]. Among those in-vitro samples, body fluids (e.g., blood and urine) are often served as ideal specimen for Raman acquisition, because they typically contain rich biochemical information about body microenviroment as well as can be collected and measured with high repeatability [25]. Considering the weak spontaneous Raman signal and strong fluorescence interference of those body fluid samples, surface-enhanced Raman spectroscopy (SERS) is usually employed to enhance the Raman signal by several orders of magnitude [26]. In recent years, SERS measurements on body fluids have been comprehensively investigated in bladder cancer diagnosis [27–29]. Furthermore, a most recent study showed that SERS analysis of pretreated plasma samples can accurately predict disease recurrence in MIBC patients with a high accuracy of 84.1% [30]. However, the postoperative prognosis of recurrence in NMIBC patients is still a huge challenge, especially for predicting the recurrent type. To the best of our knowledge, although SERS technique can provide rich biochemical information that might be closely related to bladder cancer prognosis, there has been no effort in the literature to predict the postoperative prognosis of NMIBC patients based on such technique.
In this study, a preoperative serum SERS method was proposed to predict the postoperative NMIBC prognosis, in which the preoperative serum samples were collected from NMIBC patients approximately 1 day before implementing partial cystectomy or radical cystectomy. To the best of our knowledge, this is for the first time that the postoperative serum SERS technique is employed to predict the postoperative prognosis of NMIBC patients at the time even before treatment. Through the biochemical analysis on SERS measurements of those preoperative serum samples, significant differences were found in biomolecules, including amino acids, adenine and nucleic acids, which are expected to be used as potential markers for predicting NMIBC recurrence and its recurrent type, and further compared with several routine clinically used immunohistochemistry expressions with prognostic values. In addition, prognostic models based on partial least squares analysis and linear discriminant analysis (PLS-LDA) and preoperative serum SERS measurements were established to predict whether NMIBC will be recurrent or not and whether the short-term recurrence will be invasive or non-invasive with high prediction accuracy. Therefore, this method has demonstrated great potential for accurately predicting the postoperative prognosis of NMIBC patients and further guiding the subsequent treatment to improve the recurrence rate and survival time.
2. Materials and methods
2.1 Experimental design and implementation of this prospective study
The experimental design and implementation of this prospective study is illustrated in Fig. 1. A total of 154 NMIBC patients were recruited, who were histologically diagnosed as NMIBC from 2016 to 2020 at Department of Urology in First Hospital of China Medical University in China. The exclusion criterion of this study was defined as follows: patients with urinary tract infection or severely impaired liver and kidney function; patients having undergone radiotherapy, chemotherapy or biological targeting therapy; patients having recently used hormones or immune inhibitors; patients with tumors in other sites or familial hereditary malignancies; patients with severe heart, brain and lung diseases. Ethical approval of this study (ID: 2017-37) was obtained from the institutional review board of the Medical Ethics Committee of the First Hospital of China Medical University, and written informed consent was signed by each patient.
The preoperative blood samples were obtained from the above patients in the morning (7:00 to 8:00 A.M.) after an overnight fast. More specifically, the blood samples were collected approximately 1 day before implementing partial cystectomy or radical cystectomy on each patient. Those blood samples were clot for 90 minutes at 4°C and then centrifuged at 4000 rpm for 10 minutes at room temperature. The serum samples were separated from the blood samples and immediately stored at -80°C. At the same period of collecting preoperative serum samples, biopsy and immunohistochemistry examination were also carried out on each patient to evaluate the expression levels of Ki67, CK20, P53, CK, GATA and CK7 following the protocol as previously described in [31], for comparison and validation purposes. More specifically, the primary antibodies specific for Ki67, CK20, P53, CK, GATA and CK7 (Fuzhou Maxim Company, China, 1:200), goat-anti rabbit secondary antibody (Fuzhou Maxim Company, China) and a DAB kit (Fuzhou Maxim Company, China) were used according to the manufactures’ guidelines; After imaging under an Olympus IX71 microscope, the image-pro Plus 6.0. software was used to analyze those expression levels by calculating the integrated optical density per stained area. Among those immunohistochemistry expressions, Ki-67 is widely accepted as a cellular proliferation marker closely associated with tumor stage, grade and bladder cancer recurrence [32,33]; p53 is commonly involved in genomic stability, cell cycle and apoptosis [34]; CK7, CK20 and GATA are assumed to be effective markers of luminal differentiation [35,36].
After undergoing routine treatment (i.e., partial cystectomy or radical cystectomy), those patients further received regular follow-up examinations every 3 months, including urine test, chest X-rays, abdominal and pelvic ultrasonography, cystoscopy, and cytologic examination. During the follow-up period, 59 patients were found to suffer from bladder cancer recurrences and served as recurrence group, while the other 95 patients were served as non-recurrence group. Besides, the short term recurrences (< 6 months) are commonly assumed as more life-threatening, and the early and accurate prediction on its recurrent type (MIBC or NMIBC) can guide the timely and rational treatment; thus, 19 patients with short term recurrences were particularly picked up from the recurrence group, which can be further divided into 13 NMIBC cases and 6 MIBC cases; the 13 NMIBC cases were served as non-invasive recurrence group while the 6 MIBC cases were served as invasive recurrence group. The overall characteristics of the patients are listed in Table 1 and the detailed characteristics of each patient can be found in Table S1 in the supplementary material.
Table 1. The overall characteristics of the patients
Before SERS acquisition, 1 µl concentrated silver colloids was mixed into 1 µl preoperative serum sample from each NMIBC patient with 1:1 ratio, then was shaken for 3 minutes, and 0.7 µl of the mixture was finally transferred onto an aluminum foil covered slide. The recipe and the procedure of silver colloid fabrication have been illustrated in Fig. 1. The silver colloids were prepared according to a published method [37], which can be simply and rapidly fabricated with a high success rate. More specifically, immediately after mixing 9 ml of sodium hydroxide solution (0.1 M) and 10 ml of hydroxylamine hydrochloride solution (0.06 M), the resulting mixture was added into 180 ml of silver nitrate solution (1.11×10−3 M) and then stirred for 5 minutes to obtain a homogenous silver colloid solution; thereafter, 4 ml of the silver colloid solution was picked up and centrifuged at 10,000 rpm for 10 minutes, discarding 3.8 ml of the supernatant to finally obtain the concentrated silver colloids. In order to characterize the silver colloids, the transmission electron microscope (TEM) micrograph and UV/visible absorption spectrum of the silver colloids were measured. According to Fig. 2, the silver nanoparticles are spherical with an average diameter of approximately 60 nm, and the maximum absorption locates at around 430 nm. The SERS measurements were acquired from the above mixtures by a confocal Raman microscope (HR Evolution, Horiba JY, France) and the exactly same configurations were set for every preoperative serum sample. More specifically, the excitation wavelength was 785nm with the laser power at the sample of approximately 4.3 mw; the objective lens was 20× with a numerical aperture of 0.4; the integration time was 10 seconds; the measured wavenumber range was from 400 cm−1 to 1800 cm−1 with a spectral resolution of 1 cm−1. In order to alleviate the fluctuations of SERS measurements caused by coffee-ring effect and manual operation, five different positions on each serum sample were chosen to take five SERS spectra, including both the center and the edge, and the average spectrum of those five SERS spectra was finally used to represent each preoperative serum sample for subsequent spectral analysis.
2.2 Analysis of SERS measurements
Each SERS spectrum was preprocessed to first eliminate the noise by a Savitsky-Golay smoothing algorithm [38], then remove the fluorescence interference by a fifth-order polynomial fitting algorithm [39], and finally be normalized by dividing the SERS intensities of each wavenumber by the maximum SERS intensity. Those preprocessed SERS spectra were served as the inputs of further biochemical analysis and prognostic protocol establishment.
For biochemical analysis, two-sample t-test was applied on the intensities of obvious preoperative serum SERS peaks between recurrence group versus non-recurrence group as well as between invasive recurrence group versus non-invasive recurrence group. Those characteristic preoperative serum SERS peaks with statistically significant differences (p < 0.05) were tentatively assigned to specific biomolecules to investigate the potential prognostic markers for predicting the recurrence and recurrent type of NMIBC patients. In addition, the Pearson product-moment correlation coefficients [40,41] among characteristic preoperative serum SERS peaks, immunohistochemistry expressions, linear discriminant score and bladder cancer recurrence were calculated and investigated to further validate those potential prognostic markers from the clinical point of view. It should be noted that the linear discriminant score derived from the following PLS-LDA based prognostic model were essentially the integration of intensities at certain wavenumbers with certain weights, thus the linear discriminant score was actually used to represent the joint effect of those characteristic preoperative serum SERS peaks in this study.
For establishing the prognostic protocol, multivariate analysis [42] was employed to build two prognostic models: the first prognostic model was used to predict whether there would be recurrence or not (i.e., prediction between recurrence group versus non-recurrence group), and the second prognostic model was used to predict whether the short-term recurrence would be invasive or non-invasive (i.e., prediction between invasive recurrence group versus non-invasive recurrence group). In the above prognostic models, partial least square analysis (PLS) algorithm was first employed to compress each preoperative serum SERS spectrum into 15 PLS scores; then, only the PLS scores with statistically significant difference (p < 0.05) was extracted by two-sample t-test [43] and served as the input of subsequent linear discriminant analysis (LDA) to train each prognostic model. Both prognostic models were tested in a leave-one-out cross-validation manner [44], in which the classification accuracy, sensitivity, specificity, receiver operating characteristic (ROC) curves and the area under ROC (AUC) were used as the criteria to evaluate their performances.
3. Results and discussions
3.1 Enhancement effect of silver colloids
To evaluate the enhancement effect of silver colloids, the SERS spectrum and the spontaneous Raman spectrum of the same preoperative serum sample with and without silver colloids were measured, as shown in Fig. 3. It can be observed that the spontaneous Raman signal of the preoperative serum sample was very weak, in contrast to significantly enhanced signals in many major vibration bands of the SERS spectrum. In addition, the silver colloids did not generate obvious Raman signal and were assumed to contribute negligible interference on the SERS measurements. Such large enhancement effect and negligible interference have proved the proper choice of the above silver colloids.
Fig. 3. Comparison of the SERS spectrum and the spontaneous Raman spectrum of the same preoperative serum sample as well as the Raman spectrum of the silver colloids.
3.2 Biochemical analysis of preoperative serum SERS spectra
After preprocessing, the average preoperative serum SERS spectra of recurrence versus non-recurrence groups and invasive recurrence versus non-invasive recurrence groups are shown in Fig. 4. Major SERS peaks of preoperative serum are obviously found at 494, 589, 639, 725, 812, 887, 857, 1004, 1073, 1135, 1206, 1330, 1581 and 1654 cm-1, which can be tentatively assigned to certain vibrational modes and specific biomolecules, as listed in Table 2.
Fig. 4. The average preoperative serum SERS spectra (a) between recurrence versus non-recurrence groups, and (b) between non-invasive recurrence versus invasive recurrence groups, in which the shaded area represents the standard deviations within each group.
Table 2. The tentative assignments of the major SERS peaks of preoperative serum
3.2.1 Recurrence versus non-recurrence
By applying two-sample t-test on those major preoperative serum SERS peak intensities between recurrence and non-recurrence groups, statistically significant differences were found at 725 cm-1 (p < 0.05), 1073 cm-1 (p < 0.05), 1330 cm-1 (p < 0.01), 1581 cm-1 (p < 0.01) and 1654 cm-1 (p < 0.01), as shown in Fig. 5. More specifically, compared with the non-recurrence group, the recurrence group exhibited larger preoperative serum SERS signals at both 725 cm-1 and 1330 cm-1, which corresponded to nucleic acid bases. This should be attributed to the fact of abnormal DNA/RNA metabolism in the recurrence group, i.e., high proliferation of tumor cells in the recurrence group typically required larger consumption of nucleic acid bases to synthesis DNA/RNA [45]. This was consistent with the findings in a published study [46], in which the concentration of cell-free DNA in the serum of bladder cancer patients is higher than that of healthy individuals. Another study [47] has also suggested cell-free DNA as a potential biomarker for bladder cancer prognosis. Besides, larger preoperative serum SERS signals at 1581 cm-1 and 1654 cm-1 were also observed in the recurrence group, which can be assigned to amino acid (phenylalanine) and its basic chemical compound (amide-I). Such increment of protein substances in the recurrence group might be caused by the following two aspects: on the one hand, tumor cells synthesize a large number of proteins during the process of reproduction, thus required more amino acid reserves; on the other hand, the apoptosis or necrosis of tumor cells can cause protein degradation to produce a large number of free amino acids [48]. Therefore, the free amino acids in preoperative serum might also be a potential factor related to bladder cancer recurrence. As far as we know, collagen is assumed as an active participant in tumor progression, which can change the shape of cells, participate in cell movement and play an important role in tumor cell migration and reproduction [49]. This complies with the findings of stronger preoperative serum SERS signal at 1073 cm-1 (corresponding to collagen) in the recurrence group, thus we speculate collagen as an ‘accomplice’ role in the bladder cancer recurrence.
Fig. 5. The comparisons between recurrence versus non-recurrence groups: (a) the subtracted preoperative serum SERS spectrum and (b) the box-plots of preoperative serum SERS intensities at 725, 1073, 1330, 1581, 1654 cm−1.
3.2.2 Invasive recurrence versus non-invasive recurrence
By applying two-sample t-test on those major preoperative serum SERS peak intensities between invasive recurrence versus non-invasive recurrence groups, statistically significant differences were found at 494 cm-1 (p < 0.05), 589 cm-1 (p < 0.05), 639 cm-1 (p < 0.05), 812 cm-1 (p < 0.05), 1073 cm-1 (p < 0.01) and 1330 cm-1 (p < 0.05), as shown in Fig. 6. More specifically, compared with non-invasive recurrence group, the invasive recurrence group exhibited smaller preoperative serum SERS signal at 1330 cm-1 (corresponding to trehalose), because trehalose was found to behave inhibitory effect on migration and invasion of tumor cells [50]. In contrast, the preoperative serum SERS signals at 494 cm-1, 589 cm-1, 639 cm-1 and 812 cm-1, corresponding to protein substances such as L-arginine, amide-VI, L-tyrosine and L-serine, were found to be larger in the invasive recurrence group than that in the non-invasive recurrence group. This might be attributed to the fact that high malignancy of invasive recurrence group can lead to higher proliferation of tumor cells and need to use up more amino acids to synthesize a large number of proteins [48]. Meanwhile, the consumption of massive nutrients and energy during cell proliferation may also lead to the disorder of amino acid metabolism [51]. The invasive recurrence group showed larger SERS signal at 1073 cm-1 corresponding to collagen because of its participation in tumor progression, in which similar trend was also found between the recurrence group versus non-recurrence group. This further confirms the great potential of collagen as a key indicator for bladder cancer prognosis. Interestingly, it is worth to mention that the above biochemical analysis results between invasive recurrence group versus non-invasive recurrence group are mostly consistent with our previous study of using serum SERS measurements to diagnose MIBC and NMIBC [29]. Although the serum samples in invasive recurrence group were preoperatively obtained at the first diagnosis of NMIBC, those serum samples still had some characteristics similar to MIBC. Therefore, we suspect that NMIBC patients with preoperative serum existing such characteristics at the first diagnosis are more likely to suffer from invasive recurrence.
Fig. 6. The comparison between invasive recurrence versus non-invasive recurrence groups: (a) the subtracted preoperative serum spectrum and (b) the box-plots of preoperative serum SERS intensities at 494, 589, 639, 812, 1073, 1330 cm−1.
The correlations among characteristic preoperative serum SERS peaks, immunohistochemistry expressions, linear discriminate score and bladder cancer recurrence were tested using Pearson correlation, as listed in Table 3. The correlations between those five characteristic preoperative serum SERS peaks and bladder cancer recurrence were found to be higher than almost all immunohistochemistry expressions only except Ki-67. Interestingly, the largest correlation between characteristic preoperative serum SERS peaks and immunohistochemistry expressions was observed between preoperative serum SERS peak at 725 cm-1 and Ki-67. This is reasonable because both the preoperative serum SERS peak at 725 cm-1 corresponding to nucleic acid bases and Ki-67 are closely associated with the proliferation of tumor cells [52,53]. Although both the preoperative serum SERS peak at 725 cm-1 show slight degradation on correlation to bladder cancer recurrence compared to Ki-67, the serum SERS measurements can be acquired at the point of care whereas immunohistochemistry examination usually takes about one week due to its complicated operations. Therefore, the preoperative serum SERS technique based bladder cancer prognosis has significant advantage of rapid detection. Furthermore, it can be found that the linear discriminant score (i.e., the joint effect of those characteristic SERS peaks) shows the largest correlation (R = 0.74) to bladder cancer recurrence, thus the joint usage of those characteristic preoperative serum SERS peaks is supposed to provide more promising prognosis of bladder cancer recurrence. This is another significant advantage of the preoperative serum SERS technique based bladder cancer prognosis and can also be easily implemented, because the full serum SERS spectrum with all characteristic SERS peaks can be obtained from a single preoperative serum sample in a single acquisition. In contrast, different immunohistochemistry examinations and different tissue sections are always required to detect different immunohistochemistry expressions in clinical practice, thus the joint usage of different immunohistochemistry expressions for bladder cancer prognosis would be much more complicated and time-comsuming.
Table 3. Pearson product-moment correlation coefficients between each pair of characteristic preoperative serum SERS peaks, immunohistochemistry expressions, linear discriminate score and bladder cancer recurrence
3.3 Multivariate analysis of preoperative serum SERS spectra
By applying PLS-LDA algorithm, a prognostic protocol was established, consisting of two prognostic models to predict between recurrence group versus non-recurrence group and between invasive recurrence group versus non-invasive recurrence group, respectively. In the first prognostic model, the first 15 PLS scores were derived by PLS algorithm from the pre-processed preoperative serum SERS spectra of recurrence group and non-recurrence group, and only the PLS scores with statistically significant differences were served as the input to train and test the first prognostic model based on LDA algorithm, including the 1st (p < 0.01), 2nd (p < 0.01), 4th (p < 0.01), 5th (p < 0.05), 6th (p < 0.05), 8th (p < 0.05), 9th (p < 0.01), 12th (p < 0.05), 13th (p < 0.05), 14th (p < 0.05) and 15th (p < 0.05) PLS scores. In the second prognostic model, similar analysis was implemented on the pre-processed preoperative serum SERS spectra of invasive recurrence group versus non-invasive recurrence group, and statistically significant differences were only found in the 1st (p < 0.05), 2nd (p < 0.05) and 3rd (p < 0.01) PLS scores. The PLS loadings corresponding to those PLS scores are shown in Fig. 7 (a) and (b), and it can be found that most of the peaks in those PLS loadings are in good agreement with the preoperative serum SERS peaks showing statistically significant differences in Fig. 4, especially for the first few PLS loadings. Therefore, PLS algorithm not only compresses the data to avoid over-fitting problem and save computation, but also indeed retains the most critical spectral information about predicting bladder cancer prognosis. In addition, the scatter plots of the linear discriminant scores are shown in Fig. 8 (a) and (b), and excellent separation between recurrence group versus non-recurrence group as well as between invasive recurrence group versus non-invasive recurrence group can be clearly observed.
Fig. 7. The loadings of those PLS scores with statistically significant differences to predict (a) between recurrence versus non-recurrence groups, and (b) between invasive recurrence versus non-invasive recurrence groups.
Fig. 8. Scatter plots of the linear discriminant scores: (a) between recurrence group versus non-recurrence group and (b) between non-invasive recurrence group versus invasive recurrence group, in which the detailed information of each patient (including the patient number, age, gender, tumor stage and metastasis) can be found in Table S1 and Table S2 in the supplementary material.
The performance of above prognostic models was evaluated in the leave-one-out cross-validation manner with the criteria, including classification accuracy, sensitivity, specificity, receiver operating characteristic (ROC) curves and the area under ROC (AUC). More specifically, for the first prognostic model to predict recurrence group versus non-recurrence group, the recurrence was defined as positive and the non-recurrence group was defined as negative; for the second prognostic model to predict invasive recurrence group versus non-invasive recurrence group, the invasive recurrence was defined as positive and the non-invasive recurrence was defined as negative. As listed in Table 4, the overall classification accuracy, sensitivity and specificity for predicting recurrence or non-recurrence are 87.01%, 93.68% and 94.91% respectively, and those for predicting invasive recurrence and non-invasive recurrence are 89.47%, 92.30% and 83.33% respectively. Besides, according to the ROC curves of those two prognostic models in Fig. 9, large AUC values of 0.9609 and 0.9872 are achieved, which further proves the excellent performance of such prognostic protocol for NMIBC patients at the time even before treatment.
Fig. 9. The ROC curves of the prognostic model for predicting recurrence group versus non-recurrence group and the prognostic model for predicting invasive recurrence group versus non-invasive recurrence group.
Table 4. Overall classification results based on preoperative serum SERS measurements and PLS-LDA method after leave-one-out cross-validation
Compared to urinary biomarkers, the proposed preoperative serum SERS technique has great potential of being a more promising way for early, accurate and highly repeatable prediction of postoperative prognosis in NMIBC patients. Although urinary biomarkers have been widely investigated for the detection of recurrence and/or progression in the follow-up of NMIBC patients, most urinary biomarkers suffer from either low sensitivity or low specificity, and large fluctuations of sensitivity and specificity were also observed among different urinary biomarkers and even in the same urinary biomarker among different studies. For example, the sensitivities and specificities of 8 available biomarkers (i.e., NMP22 BladderChek, NMP22, BTA STAT, BTA Trak, ImmunoCyt, UroVysion, Cxbladder monitor and Bladder cancer (UBC) test) for NMIBC surveillance sparsely range from 11% to 100% and from 29% to 100%, respectively [19]. This might be attributed to the fact that urine can easily be interfered by other factors, such as other genitourinary diseases and hematuria [54]. In contrast, the proposed preoperative serum SERS technique achieved high sensitivity and specificity of 93.68% and 94.91% for predicting the bladder cancer recurrence, and the serum based examinations were assumed to be with higher repeatability [38,55]. In addition, both preoperative and postoperative urine samples are typically required to detect recurrence by using urinary biomarkers based methods [56]. In contrast, only preoperative serum samples is necessary by using the proposed preoperative serum SERS technique, thus the postoperative prognosis can be predicted at the time even before treatment.
To the best of our knowledge, this study is for the first time that preoperative serum SERS technique is employed to predict the prognosis of NMIBC patients, in which the the future status (i.e., postoperative recurrence and postoperative recurrent type) of NMIBC patients can be predicted at the time even before treatment. Compared to the serum SERS based bladder cancer diagnosis in our previous study [29] might easily mistaken for similar to this study, the preoperative serum SERS technique based bladder cancer prognosis in this study is very different and much more challenging, because of the following aspects: 1. This study aims to predict the future status (i.e., to predict postoperative recurrence and postoperative recurrent type before treatment) of NMIBC patients for guiding the subsequent treatment to inhibit recurrence, thus is with very different clinical significance, whereas our previous study aims to diagnose the existing status (i.e., to differentiate NMIBC and MIBC patients) with less difficulty; 2. This study only recruits NMIBC patients and then predicts their future recurrence and recurrent type, in which both smaller differences within only NMIBC patients and the prediction of future statuses make it much more challenging, whereas our previous study recruits both NMIBC and MIBC patients and diagnoses their existing statuses at the time of blood collection; 3. This study requires additional postoperative follow-up examinations with some patients even lost to follow-up, thus such prospective study is much more complicated and challenging, whereas our previous study only requires one-time preoperative examination. Therefore, we believe that this study should be essentially different from our previous study.
4. Conclusion
In this study, we explored a preoperative serum SERS based prognostic protocol to predict the postoperative recurrence and the type of recurrent bladder cancer for NMIBC patients. The biochemical analysis results suggest that the biomolecules related to DNA/RNA, protein substances and collagen are expected as potential indicators to predict the recurrence of bladder cancer, whereas the biomolecules related to protein substances, trehalose and collagen can further be used to predict whether the short-term recurrent bladder cancer is invasive or non-invasive, which were further compared with several routine clinically used immunohistochemistry expressions with prognostic values. In addition, combining with PLS-LDA algorithm, a prognostic protocol consisting of two prognostic models was established, and achieved high classification accuracies, sensitivities and specificities to predict whether bladder cancer will be recurrent or not and whether the short-term recurrence will be invasive or non-invasive for NMIBC patients at the time even before treatment. Therefore, the proposed method has great potential in the early and accurate prediction of postoperative prognosis, which can provide timely and rational guidance for clinical treatment and further improve the recurrence rate and survival time of NMIBC patients.
Funding
National Natural Science Foundation of China (61605025); Fundamental Research Funds for the Central Universities (N2119002); Program for the Introduction of High-End Foreign Experts (GXL20200218001); K. C. Wong Magna Fund in Ningbo University (501100020738).
Disclosures
The authors declare no conflicts of interest.
Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request.
Supplemental document
See Supplement 1 for supporting content.
References
1. S. Antoni, J. Ferlay, I. Ferlay, A. Znaor, A. Jemal, and F. Bray, “Bladder cancer incidence and mortality: a global overview and recent trends,” Eur. Urol. 71(1), 96–108 (2017). [CrossRef]
2. H. Sung, J. Ferlay, R. L. Siegel, M. Laversanne, I. Soerjomataram, A. Jemal, and F. Bray, “Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries,” CA: A Cancer J. Clin. 71(3), 209–249 (2021). [CrossRef]
3. B. W. G. Van Rhijn, M. Burger, Y. Lotan, E. Solsona, C.G. Stief, R.J. Sylvester, J.A. Witjes, and A.R. Zloatta, “Recurrence and progression of disease in non-muscle-invasive bladder cancer: from epidemiology to treatment strategy,” Eur. Urol. 56(3), 430–442 (2009). [CrossRef]
4. R. J. Sylvester, A. P. M. Van Der Meijden, W. Oosterlinck, J. A. Witjes, C. Bouffioux, L. Denis, D. W. W. Newling, and K. Kurth, “Predicting recurrence and progression in individual patients with stage Ta T1 bladder cancer using EORTC risk tables: a combined analysis of 2596 patients from seven EORTC trials,” Eur. Urol. 49(3), 466–477 (2006). [CrossRef]
5. B. P. Schrier, M. P. Hollander, B. W. G. van Rhijn, L. A. L. M. Kiemeney, and J. A. Witjes, “Prognosis of muscle-invasive bladder cancer: difference between primary and progressive tumours and implications for therapy,” Eur. Urol. 45(3), 292–296 (2004). [CrossRef]
6. K. H. Kurth and R. J. Sylvester, “Prognostic factors in non–muscle-invasive bladder tumors: i. clinical prognostic factors: a review of the experience of the eortc genito-urinary group II. Biologic prognostic markers,” Eur. Urol., Suppl. 6(14), 789–799 (2007). [CrossRef]
7. J. Fernandez-Gomez, E. Solsona, M. Unda and, et al., “Prognostic factors in patients with non-muscle-invasive bladder cancer treated with bacillus Calmette-Guérin: multivariate analysis of data from four randomized CUETO trials,” Eur. Urol. 53(5), 992–1002 (2008). [CrossRef]
8. F. Millan-Rodriguez, G. Chechile-Toniolo, J. Salvador-Bayarri, J. Palou, and J. Vicente-Rodriguez, “Multivariate analysis of the prognostic factors of primary superficial bladder cancer,” J. Urol. (N. Y., NY, U. S.) 163(1), 73–78 (2000). [CrossRef]
9. R. J. Sylvester, “Natural history, recurrence, and progression in superficial bladder cancer,” Sci. World J. 6, 2617–2625 (2006). [CrossRef]
10. T. Habuchi, M. Marberger, M. J. Droller, G. P. Hemstreet III, H. B. Grossman, J. A. Schalken, B. J. Schmitz-Dräger, W. M. Murphy, A. V. Bono, P. Goebell, R. H. Getzenberg, S. H. Hautmann, E. Messing, Y. Fradet, and V. B. Lokeshwar, “Prognostic markers for bladder cancer: International Consensus Panel on bladder tumor markers,” Urology 66(6), 64–74 (2005). [CrossRef]
11. V. B. Lokeshwar, T. Habuchi, H. B. Grossman, W. M. Murphy, S. H. Hautmann, G. P. Hemstreet III, A. V. Bono, R. H. Getzenberg, P. Goebell, B. J. Schmitz-Dräger, J. A. Schalken, Y. Fradet, M. Marberger, E. Messing, and M. J. Droller, “Bladder tumor markers beyond cytology: International Consensus Panel on bladder tumor markers,” Urology 66(6), 35–63 (2005). [CrossRef]
12. B. J. Schmitz-Dräger, P. J. Goebell, T. Ebert, and Y. Fradet, “p53 immunohistochemistry as a prognostic marker in bladder cancer,” Eur. Urol. 38(6), 691–700 (2000). [CrossRef]
13. A. Rodríguez-Alonso, S. Pita-Fernández, J. González-Carreró, and J. L. Nogueira-March, “p53 and ki67 expression as prognostic factors for cancer-related survival in stage T1 transitional cell bladder carcinoma,” Eur. Urol. 41(2), 182–189 (2002). [CrossRef]
14. F. Saint, M. A. L. F. Belda, R. Quintela, A. Hoznek, J. Patard, J. Bellot, Z. Popov, E. S. Zafrani, C. C. Abbou, D. K. Chopin, and S. G. D. D. Medina, “Pretreatment p53 nuclear overexpression as a prognostic marker in superficial bladder cancer treated with bacillus Calmette–Guérin (BCG),” Eur. Urol. 45(4), 475–482 (2004). [CrossRef]
15. N. Malats, A. Bustos, C. M. Nascimento, F. Fernandez, M. Rivas, D. Puente, P. M. Kogevinas, and F. X. Real, “P53 as a prognostic marker for bladder cancer: a meta-analysis and review,” Lancet Oncol. 6(9), 678–686 (2005). [CrossRef]
16. T. Reinert, M. Borre, A. Christiansen, G. G. Hermann, T. F. Orntoft, and L. Dyrskjot, “Diagnosis of bladder cancer recurrence based on urinary levels of EOMES, HOXA9, POU4F2, TWIST1, VIM, and ZNF154 hypermethylation,” PLoS One 7(10), e46297 (2012). [CrossRef]
17. A. M. Kamat, N. M. Hahn, J. A. Efstathiou, S. P. Lerner, P. Malmström, W. Choi, C. C. Guo, Y. Lotan, and W. Kassouf, “Bladder cancer,” Lancet 388(10061), 2796–2810 (2016). [CrossRef]
18. E. Xylinas, L. A. Kluth, M. Rieken, P. I. Karakiewicz, Y. Lotan, and S. F. Shariat, “Urine markers for detection and surveillance of bladder cancer,” Urol. Oncol.: Semin. Orig. Investig 32(3), 222–229 (2014). [CrossRef]
19. F. Soria, M. J. Droller, Y. Lotan, P. Gontero, D. D’Andrea, K. M. Gust, M. Rouprêt, M. Babjuk, J. Palou, and Shahrokh F. Shariat, “An up-to-date catalog of available urinary biomarkers for the surveillance of non-muscle invasive bladder cancer,” World J. Urol. 36(12), 1981–1995 (2018). [CrossRef]
20. D. S. Kaufman, W. U. Shipley, and A. S. Feldman, “Bladder cancer,” Lancet 374(9685), 239–249 (2009). [CrossRef]
21. C. Zheng, Y. Lv, Q. Zhong, R. Wang, and Q. Jiang, “Narrow band imaging diagnosis of bladder cancer: systematic review and meta-analysis,” BJU Int. 110(11b), E680–E687 (2012). [CrossRef]
22. S. Daneshmand, A. K. Schuckman, B. H. Bochner M, S. Cookson, T. M. Downs, L. G. Gomella, H. B. Grossman, A. M. Kamat, B. R. Konety, C. T. Lee, K. S. Pohar, R. S. Pruthi, M. J. Resnick, N. D. Smith, J. A. Witjes, M. P. Schoenberg, and G. D. Steinberg, “Hexaminolevulinate blue-light cystoscopy in non-muscle-invasive bladder cancer: review of the clinical evidence and consensus statement on appropriate use in the USA,” Nat. Rev. Urol. 11(10), 589–596 (2014). [CrossRef]
23. Q. Tu and C. Chang, “Diagnostic applications of Raman spectroscopy,” Nanomedicine 8(5), 545–558 (2012). [CrossRef]
24. K. Kong, C. Kendall, N. Stone, and I. Notingher, “Raman spectroscopy for medical diagnostics: from in-vitro biofluid assays to in-vivo cancer detection,” Adv. Drug Delivery Rev. 89, 121–134 (2015). [CrossRef]
25. L. A. Austin, S. Osseiran, and C. L. Evans, “Raman technologies in cancer diagnostics,” Analyst 141(2), 476–503 (2016). [CrossRef]
26. K. C. Bantz, A. F. Meyer, N. J. Wittenberg, H. Im, Ö. Kurtuluş, S. H. Lee, N. C. Lindquist, S. Oh, and C. L. Haynes, “Recent progress in SERS biosensing,” Phys. Chem. Chem. Phys. 13(24), 11551–11567 (2011). [CrossRef]
27. X. Cui, T. Liu, X. Xu, Z. Zhao, Y. Tian, Y. Zhao, S. Chen, Z. Wang, Y. Wang, D. Hu, S. Fu, G. Shan, J. Sun, K. Song, and Y. Zeng, “Label-free detection of multiple genitourinary cancers from urine by surface-enhanced Raman spectroscopy,” Spectrochim. Acta, Part A 240, 118543 (2020). [CrossRef]
28. Y Zhang, X Lai, Q Zeng, L. Li, L. Lin, S. Li, Z. Liu, C. Si, M. Qi, and Z. Guo, “Classifying low-grade and high-grade bladder cancer using label-free serum surface-enhanced Raman spectroscopy and support vector machine,” Laser Phys. 28(3), 035603 (2018). [CrossRef]
29. S. Chen, S. Zhu, X. Cui, W. Xu, C. Kong, Z. Zhang, and W. Qian, “Identifying non-muscle-invasive and muscle-invasive bladder cancer based on blood serum surface-enhanced Raman spectroscopy,” Biomed. Opt. Express 10(7), 3533–3544 (2019). [CrossRef]
30. H. Qian, Y. Wang, Z. Ma, L. Qian, X. Shao, D. Jin, M. Cao, S. Liu, H. Chen, J. Pan, and W. Xue, “Surface-Enhanced Raman spectroscopy of pretreated plasma samples predicts disease recurrence in muscle-invasive bladder cancer patients undergoing neoadjuvant chemotherapy and radical cystectomy,” Int. J. Nanomed. 17, 1635–1646 (2022). [CrossRef]
31. Z. Zhang, G. Zhang, C. Kong, B. Zhan, X. Dong, and X. Man, “METTL13 is downregulated in bladder carcinoma and suppresses cell proliferation, migration and invasion,” Sci. Rep. 6(1), 1–10 (2016). [CrossRef]
32. Y. Tian, Z. Ma, Z. Chen, M. Li, Z. Wu, M. Hong, H. Wang, R. Svatek, R. Rodriguez, and Z. Wang, “Clinicopathological and prognostic value of Ki-67 expression in bladder cancer: a systematic review and meta-analysis,” PLoS One 11(7), e0158891 (2016). [CrossRef]
33. V. Margulis, Y. Lotan, P. I. Karakiewicz, Y. Fradet, R. Ashfaq, U. Capitanio, F. Montorsi, P. J. Bastian, M. E. Nielsen, S. C. Müller, J. Rigaud, L. C. Heukamp, G. Netto, S. P. Lerner, A. I. Sagalowsky, and S. F. Shariat, “Multi-institutional validation of the predictive value of Ki-67 labeling index in patients with urinary bladder cancer,” J. Natl. Cancer Inst. 101(2), 114–119 (2009). [CrossRef]
34. K. Fukasawa, F. Wiener, G. F. V. Woude, and S. Mai, “Genomic instability and apoptosis are frequent in p53 deficient young mice,” Oncogene 15(11), 1295–1302 (1997). [CrossRef]
35. F. Michel, F. Cancrini, G. Cancel-Tassin and, et al., “A study of the immunohistochemical profile of bladder cancer in neuro-urological patients by the French Association of Urology,” World J. Urol. 1, 1–9 (2022). [CrossRef]
36. X. H. Pei, F. Bai, M. D. Smith, J. Usary, C. Fan, S. Pai, I. Ho, C. M. Perou, and Y. Xiong, “CDK inhibitor p18INK4c is a downstream target of GATA3 and restrains mammary luminal progenitor cell proliferation and tumorigenesis,” Cancer Cell 15(5), 389–401 (2009). [CrossRef]
37. N, Leopold and B Lendl, “A new method for fast preparation of highly surface-enhanced Raman scattering (SERS) active silver colloids at room temperature by reduction of silver nitrate with hydroxylamine hydrochloride,” J. Phys. Chem. B 107(24), 5723–5727 (2003). [CrossRef]
38. L. Xia, J. Lu, Z. Chen, X. Cui, S. Chen, and D. Pei, “Identifying benign and malignant thyroid nodules based on blood serum surface-enhanced Raman spectroscopy,” Nanomedicine (N. Y., NY, U. S.) 32, 102328 (2021). [CrossRef]
39. Z. Huang, H. Lui, D. I. McLean, et al., “Raman spectroscopy in combination with background near-infrared autofluorescence enhances the in vivo assessment of malignant tissues,” Photochem. Photobiol. 81(5), 1219–1226 (2005). [CrossRef]
40. A Kamińska, M Roman, A Wróbel, M. Korbelik, and H. Zeng, “Raman spectroscopy of urinary extracellular vesicles to stratify patients with chronic kidney disease in type 2 diabetes,” Nanomedicine (N. Y., NY, U. S.) 39, 102468 (2022). [CrossRef]
41. H. Ito, H. Inoue, K. Hasegawa, Y. Hasegawa, T. Shimizu, S. Kimura, M. Onimaru, H. Ikeda, and S. Kudo, “Use of surface-enhanced Raman scattering for detection of cancer-related serum-constituents in gastrointestinal cancer patients,” Nanomedicine (N. Y., NY, U. S.) 10(3), 599–608 (2014). [CrossRef]
42. S. Chen, H. Lin, H. Zhang, F. Guo, S. Zhu, X. Cui, and Z. Zhang, “Identifying functioning and nonfunctioning adrenal tumors based on blood serum surface-enhanced Raman spectroscopy,” Anal. Bioanal. Chem. 413(16), 4289–4299 (2021). [CrossRef]
43. Q. Wu, S. Qiu, Y. Yu, W. Chen, H. Lin, D. Lin, S. Feng, and R. Chen, “Assessment of the radiotherapy effect for nasopharyngeal cancer using plasma surface-enhanced Raman spectroscopy technology,” Biomed. Opt. Express 9(7), 3413–3423 (2018). [CrossRef]
44. S. Zhu, X. Cui, W. Xu, S. Chen, and W. Qian, “Weighted spectral reconstruction method for discrimination of bacterial species with low signal-to-noise ratio Raman measurements,” RSC Adv. 9(17), 9500–9508 (2019). [CrossRef]
45. J. De Gelder, K. De Gussem, P. Vandenabeele, and L. Moens, “Reference database of Raman spectra of biological molecules,” J. Raman Spectrosc. 38(9), 1133–1147 (2007). [CrossRef]
46. C. Bettegowda, M. Sausen, R. J. Leary and, et al., “Detection of circulating tumor DNA in early-and late-stage human malignancies,” Sci. Transl. Med. 6(224), 224ra24 (2014). [CrossRef]
47. R. Kandimalla, A. A. Van Tilborg, and E. C. Zwarthoff, “DNA methylation-based biomarkers in bladder cancer,” Nat. Rev. Urol. 10(6), 327–335 (2013). [CrossRef]
48. J. M. Argilés and J. Azcón-Bieto, “The metabolic environment of cancer,” Mol. Cell. Biochem. 81(1), 3–17 (1988). [CrossRef]
49. S. Xu, H. Xu, W. Wang, S. Li, H. Li, T. Li, W. Zhang, X. Yu, and L. Liu, “The role of collagen in cancer: from bench to bedside,” J. Transl. Med. 17(1), 1–22 (2019). [CrossRef]
50. N. S. N. Chaitanya, A. Devi, S. Sahu, and P. Alugoju, “Molecular mechanisms of action of Trehalose in cancer: A comprehensive review,” Life Sci. 269, 118968 (2021). [CrossRef]
51. B. J. Altman and C. V. Dang, “Normal and cancer cell metabolism: lymphocytes and lymphoma,” FEBS J. 279(15), 2598–2609 (2012). [CrossRef]
52. T. Scholzen and J. Gerdes, “The Ki-67 protein: from the known and the unknown,” J. Cell. Physiol. 182(3), 311–322 (2000). [CrossRef]
53. D. G. Booth, M. Takagi, L. Sanchez-Pulido, E. Petfalski, G. Vargiu, K. Samejima, N. Imamoto, C. P. Ponting, D. Tollervey, W. C. Earnshaw, and P. Vagnarelli, “Ki-67 is a PP1-interacting protein that organises the mitotic chromosome periphery,” ELife 3, e01641 (2014). [CrossRef]
54. D. Tilki, M. Burger, G. Dalbagni, H. B. Grossman, O. W. Hakenberg, J. Palou, O. Reich, M. Rouprêt, S. F. Shariat, and A. R. Zlotta, “Urine markers for detection and surveillance of non-muscle-invasive bladder cancer,” Eur. Urol. 60(3), 484–492 (2011). [CrossRef]
55. I. Sanders, S. Holdenrieder, G. Walgenbach-Brünagel, A. V. Ruecker, G. Kristiansen, S. C. Müller, and J. Ellinger, “Evaluation of reference genes for the analysis of serum miRNA in patients with prostate cancer, bladder cancer and renal cell carcinoma,” Int. J. Urol. 19(11), 1017–1025 (2012). [CrossRef]
56. A. M. Kamat, J. Briggman, D. L. Urbauer, R. Svatek, G. M. N. González, R. Anderson, H. B. Grossman, F. Prat, and C. P. Dinney, “Cytokine panel for response to intravesical therapy (CyPRIT): nomogram of changes in urinary cytokine levels predicts patient response to Bacillus Calmette-Guérin,” Eur. Urol. 69(2), 197–200 (2016). [CrossRef]
