Decreased green autofluorescence in cancerous tissues is a potential biomarker for diagnosis of renal cell carcinomas

Abstract

This study primarily focuses on the potential sources of autofluorescence, including keratins (KRT) encoded by KRT1, KRT7, and KRT8, to investigate their contributions to the differences in autofluorescence between cancerous tissues and adjacent non-tumor tissues, as well as their potential for real-time diagnosis of RCC. First, the autofluorescence of renal cell carcinoma (RCC) tissues under 488 nm laser excitation was observed and compared with the autofluorescence of neighboring non-tumor tissues. Then, the effect on the autofluorescence intensity was analyzed by knocking down the KRT1/KRT7 gene. In addition, autofluorescence data were collected from 174 pairs of tumor and adjacent non-tumor tissue samples (from 60 RCC patients). Diagnostic performance was evaluated using ROC analysis to determine the threshold value for tumor autofluorescence intensity. Under 488 nm laser excitation, the intensity of green autofluorescence in cancerous tissues of RCC patients was significantly lower than that in non-tumor tissues. Further analysis showed that KRT1 knockdown resulted in a 73% reduction in autofluorescence intensity, suggesting that KRT1 plays a key role in the reduced autofluorescence observed in tumor tissues. In addition, analysis of autofluorescence data from 174 tumor and adjacent non-tumor tissue samples showed an AUC of 0.880 for ROC analysis, a diagnostic sensitivity and specificity of 0.843 and 0.835, respectively, and a threshold value of 27.45 for using tumor autofluorescence intensity. KRT1 is a major contributor to the tumor autofluorescence observed in RCC. An autofluorescence-based diagnostic model facilitates real-time assessment of surgical margins during partial nephrectomy, thereby potentially improving surgical success rates.

Introduction

Renal cell carcinoma (RCC) originates from renal tubular epithelial cells and accounts for more than 90% of renal malignancies, with a mortality rate of approximately 30-40%^1,2. The major histological subtypes of RCC is clear cell renal cell carcinoma, which accounts for 2% of all malignancies³. RCC is mostly asymptomatic in its early stages, and patients are often in a progressive stage when they are diagnosed. Radical nephrectomy and partial nephrectomy (PN) with preservation of nephrons are effective treatments for RCC (especially limited RCC)⁴. However, determination of tumor boundaries during performing PN is a major challenge⁵. Therefore, exploring simple, real-time and accurate diagnostic methods for RCC is of great significance for improving the prognosis.

Autofluorescence technology was based on utilizing bio-macromolecules that can be excited to fluoresce in vivo. In the past decade, which was demonstrated to be a promising tool for the rapid diagnosis of advanced cancers, including gastric cancer, lung cancer, and basal cell carcinoma etc^{6, 7–8}. Our previous study found that green autofluorescence at a wavelength of 500–550 nm, excited by a 488 nm laser, exhibited significant decreased autofluorescence in lung cancer⁷. Later study revealed that the source responsible for autofluorescence mainly came from keratin 1 (KRT1)⁹. As an epithelial neoplasm, RCC is characterized by widespread cytokeratin expression, which suggested the possibility that KRT1 dominated green autofluorescence would be a tool in the differentiation between cancerous and non-neoplastic tissue. KRT1, encoded by KRT1, was also named as cytokeratin-1 in non-keratinized epithelial cells. In addition, cytokeratin-7 and cytokeratin-8 were indicated to be expressed in several types of RCC¹⁰. Therefore, this study mainly focused on potential source of autofluorescence, including cytokeratin-1, cytokeratin-7, and cytokeratin-8 to investigate their contributions to the autofluorescence differences between cancerous tissue and adjacent non-neoplastic tissues, aiming to identify key proteins responsible for this phenomenon and its potential in real-time diagnosis of RCC.

Materials and methods

Patients and ethical statement

In this study, we prospectively collected postoperative fresh tumor specimens from 60 patients with renal cell carcinoma (RCC) who underwent partial nephrectomy or radical nephrectomy, as well as from 4 patients each with bladder cancer, upper tract urothelial carcinoma (UTUC), gastric cancer, colon cancer, and rectal cancer who underwent radical surgery. The specimens were collected from February 1, 2023, to July 10, 2024. All specimens included in the study were confirmed by postoperative pathology. For tumor tissues collection, the location and extent of the tumor were first confirmed to avoid including excessive necrotic tissue or fat, which could affect subsequent experiments (Pathologically confirmed as RCC postoperatively). Normal peritumoral tissues was collected from renal parenchyma that appeared normal and was at least 3 cm away from the tumor margin. For each RCC patient, 1–3 pairs of tumor and adjacent normal tissues were collected (depending on the size of the specimen resected during surgery), resulting in a total of 174 pairs of tissue samples (i.e., 174 tumor tissue samples + 174 adjacent normal renal tissue samples). The study strictly adhered to the principles expressed in the Declaration of Helsinki, and procedures followed were in accordance with the ethical standards set by the Ethics Committee of the First Affiliated Hospital of Gannan Medical University and were approved by the Committee (Approval No. 2LSC-2024-359), with informed consent obtained from all patients. The patients with renal cancer, mainly clear cell renal cell carcinoma, were hospitalized in the Department of Urology of the First Affiliated Hospital of Gannan Medical University.

Cell lines

Selected from human renal cortical proximal tubule epithelial cells HK-2 and human renal clear cell adenocarcinoma cells 786-O (both purchased from Wuhan Pricella Biotechnology Co., Ltd, Wuhan, China). The cells were grown in DMEM Medium Modified (C11965500BT, Gibco) supplemented with 10% FBS (100–500, GEMINI), 100U/ml penicillin and 0.1 mg/ml streptomycin (15140122, Gibco) at 37℃ in humidified 5% CO2 atmosphere. The HK-2 cell lines with KRT1/KRT7 knockdown (KRT1/7-shRNA-HK-2) was constructed by transfecting HK-2 cells with a plasmid that interferes with keratin 1/ keratin 7 (Genechem, Shanghai, China) using Lipofectamine 2000 following the manufacturer’s instructions (GLPBIO, Montclair, California, USA).The sequences of the KRT1-shRNA and KRT7-shRNA were 5’GACTCAAATCAGTGAAACTAA3’ and 5’CCGCGAGGTCACCATTAACCA3’, respectively (Genechem, Shanghai, China).

Determinations of the autofluorescence of human subjects’ tumor tissues

Firstly, collect fresh postoperative cancerous and peritumoral normal renal tissue samples from RCC patients according to the standard procedure (each tissue sample should be around 0.5 cm in size, and avoid staining, fixation, or immersion in normal saline and formalin to prevent interference with fluorescence imaging). Place the samples into specimen tubes and label them with the date, name, hospital number, and sample type (renal cancerous tissue or normal renal tissue). After collecting the specimens, they should be sent for autofluorescence imaging as soon as possible (the time interval from obtaining the specimen to imaging should be less than 15 min, and the tissue specimens should be cut into regular thin slices as much as possible to facilitate fluorescence imaging). Use tweezers to remove the cancerous and peritumoral normal tissue samples from the specimen tubes and place them on a glass slide. Then, place the slide on the stage of the autofluorescence imaging device, focus the 488 nm excitation light on the sample surface, and then rotate the focusing knob until a clear image is obtained and save it (note that imaging should be conducted in a relatively dark environment, and the conditions for each imaging session should be as consistent as possible). A portable autofluorescence imaging equipment (Jiangsu Kunhui Biotechnology Co. Limited, Jiangsu, China) was used to determine the autofluorescence of cancerous tissues and non-neoplastic tissues of human subjects. The excitation wavelength and the emission wavelength were 488 nm and 500-550 nm, respectively. The equipment parameters used for capturing the autofluorescence images was as follows: microscopy type was set to wide-field mode, laser power in working state was set to 10µW, frame rate was set to100 frames/s, exposure time was set to 30000µs, image size was set to 1920 × 1200 pixels, imaging resolution was set to 1450 pixels/mm, physical size corresponded to 1.32mmx0.83 mm. The autofluorescence was quantified by the following method: Five spots with a size of approximately 65 × 65 µm² were randomly selected on the scanned image using Image J. The average autofluorescence intensities was defined as the autofluorescence intensity of each spot.

Determinations of the autofluorescence of cells

The autofluorescence of HK-2 cells, 786-O cells, and KRT1/7-shRNA-HK-2 cells was detected under a confocal laser scanning microscope (STELLARIS 5, Leica, Germany) with an excitation wavelength of 488 nm and an emission wavelength of 500–550 nm. The autofluorescence was quantified by the following method: six spots with a size of approximately 2.6 × 2.6 µm² were randomly selected on the scanned image using Image J. The average autofluorescence intensities was defined as the autofluorescence intensity of each spot.

Western blot assays

The tissue samples (20 µg) and cell samples was lysed with RIPA buffer, and the lysates were centrifuged at 12,000 g for 15 min at 4 °C. The protein concentrations of the samples were quantified using BCA Protein Assay Kit (Epizyme, Shanghai, China), then prepared with 6x loading buffer. Total proteins were then separated by 8-10% SDS-PAGE gel and subsequently transferred to a 0.45 μm PVDF membrane (Servicebio, Wuhan, China). Blots were incubated with monoclonal anti-Cytokeratin-1 (ab185628, Abcam, UK) (1:1000 dilution), polyclonal anti-cytokeratin 7/8 (15539-1-AP /17514-1-AP, Proteintec, USA) (1:2500 dilution) or β-actin/GAPDH (1:2500, 66009-1-Ig/60004-1-Ig, Proteintec, USA) with 0.5% BSA overnight at 4 °C, then incubated with HRP conjugated Goat Anti-Rabbit IgG (H + L) (1:2000, SA00001-2, Proteintec, USA) or HRP conjugated Goat Anti-Mouse IgG (1:2000, SA00001-1, Proteintec, USA). The ECL detection system (Amersham ImageQuant 800, Cytiva, USA) was used to detect protein signals. The intensity of the quantitative bands was determined by optical densitometry using Image J. All experiments were conducted in triplicate.

RT-qPCR assays

Total RNA was extracted using TransZol UP Plus RNA Kit (TransGen, Beijing, China). RNA purity and concentration were determined by NanoDrop 2000 (Thermo Fisher Scientific). Reverse transcription reactions were performed using EasyScript One-Step gDNA Removal and cDNA Synthesis SuperMix (TransGen, Beijing, China), following the manufacturer’s instructions to generate cDNA. Primers were obtained from Shanghai Generay Biotechnology (KRT1 F: 5’AAGTCCTCTGGTGGCAGTTC3’ and R: 5’TTTTCTCCGGTAAGGCTGGGG3’, KRT7 F:5’CTGCCTACATGAGCAAGGTG3’ and R: 5’ CAGCTCTGTCAACTCCGTCT 3’, KRT8 F:5’ GCAGCAACTTTCGCGGT 3’ and R:5’ GTCTCCAGCATCTTGTTCTGC 3’, GAPDH F: 5’GAAAGCCTGCCGGTGA CTAA3’ and R: 5’GCATCACCCGGAGGAGAAAT3’), qPCR reactions were performed using PerfectStart Green qPCR SuperMix (TransGen, Beijing, China) following the manufacturer’s instructions. Data were collected using the StepOne Real-Time PCR System (Applied Biosystems) and statistically analyzed by GraphPad Prism 9.3.

Statistical analyses

All data are presented as Mean ± SEM, except where noted. In this study, we used the Shapiro-Wilk test to assess the normality of the data. For the comparison of two independent samples, we used the independent samples t-test if the data met the normality assumption, and the non-parametric test (Mann-Whitney U test) if the data did not meet the normality assumption. For the comparison of paired samples, we used the paired samples t-test if the data met the normality assumption, and otherwise we used the non-parametric test (Wilcoxon signed-rank test). R software (version 4.4.2) and Graphpad prism 9.3 was used for ROC analyses and for determinations on if the data were in normal distribution. P values < 0.05 were considered statistically significant.

Results

The green autofluorescence intensity of cancerous tissues in RCC patients is significantly lower than non-neoplastic tissues

As demonstrated in our previous findings, green autofluorescence at a wavelength of 500–550 nm, excited by a 488 nm laser, exhibited significant decreased autofluorescence in lung cancer, indicating its potential diagnostic value in real-time diagnosis of tumors⁷. To elucidate its potential clinical significance, this study employed a 488 nm laser to examine various tumor tissues and their corresponding adjacent non-neoplastic tissues, including colon cancer, gastric cancer, rectal cancer, bladder cancer, upper urinary tract tumors, and renal cell carcinoma (RCC). While varying degrees of autofluorescence differences were observed across all tumor types (Figure S1A), RCC tissues exhibited the most pronounced autofluorescence differences between cancerous and adjacent non-neoplastic tissues (Figure S1B), highlighting a stronger potential for clinical application.

To this end, we obtained specimens from 37 RCC patients. As illustrated in Fig. 1A, the autofluorescence intensity of cancerous tissues was significantly diminished compared to non-neoplastic tissues. Statistical analysis comparing autofluorescence intensities between tumor tissues and adjacent non-neoplastic tissues from these 37 patients yielded consistent results (Fig. 1B). These findings prompted us to investigate the molecular mechanisms underlying this disparity in autofluorescence intensity. Consequently, we evaluated autofluorescence intensity in both normal renal tubular epithelial cells (HK-2) and primary clear cell renal carcinoma cells (786-O). The results demonstrated a notable reduction in autofluorescence intensity in the 786-O cell lines (Fig. 1C and D).

[See PDF for image]

Fig. 1

Carcinoma tissues showed decreased green autofluorescence intensity of in RCC patients. A. Representative images of carcinomas and counterpart non-neoplastic tissues under white light and 488 nm laser. Scale bar: white light 0.2 cm/ fluorescence 10 μm. B. Quantitative analysis of autofluorescence images showed that the green autofluorescence intensity of carcinoma tissues of RCC patients was significantly lower than that of non-neoplastic tissues (N = 37 Patients, ***P < 0.001, Wilcoxon signed-rank test). C. The green autofluorescence of 786-O cells was significantly lower than that of HK-2 cells. The excitation wavelength and the emission wavelength were 488 nm and 500-550 nm, respectively. Scale bar = 100 μm. D. Quantitative analysis of autofluorescence images showed that the green autofluorescence intensity of 786-O cells was significantly lower than that of HK-2 cells (**P < 0.01, independent samples t-test).

KRT1 was the primary contributor of green autofluorescence

Our previous study suggested that the green autofluorescence might be derived from keratins^7,9. To elucidate the role of cytokeratin-1 (KRT1) in RCC, we examined its expression levels in both cancerous and non-neoplastic tissues from four RCC patients. RT-qPCR results demonstrated a significant downregulation of KRT1 expression in cancerous tissues relative to non-neoplastic tissues (Fig. 2A). Western blot reveled similar phenomenon which corroborated this finding, confirming markedly reduced cytokeratin-1 expression in cancerous tissues (Fig. 2B and C). The 786-O and HK-2 cell lines demonstrated a trend consistent with the observations in the tissue samples (Fig. 2D and F). Given the role of KRT1 in biological autofluorescence, we hypothesized that KRT1 might be closely associated with the reduction in tumor autofluorescence. Consequently, we knocked down KRT1 expression in HK-2 cells using a KRT1-shRNA. RT-qPCR and Western blot analyses confirmed stable knockdown efficiency (Fig. 2G and I). Interestingly, we observed that the intensity of autofluorescence decreased 73%, which was concomitantly with the reduction in KRT1 expression (Fig. 2J and K).

Given that various cytokeratins can contribute to autofluorescence, we also investigated the potential impact of KRT7 and KRT8 on RCC autofluorescence. Consistent with our findings for KRT1, KRT7 expression was significantly lower in tumor tissues compared to adjacent non-neoplastic tissues in RCC patients (Figures S2A–S2C) and exhibited a similar differential expression pattern in cell lines (Figures S2D–S2F). However, when we knocked down KRT7 expression in HK-2 cells using the same approach (Figures S2G–S2I), it was observed no significant reduction in the intensity of autofluorescence (Figures S2J–S2K). Furthermore, when analyzing KRT8, it was found that its expression was higher in tumor tissues and reduced in adjacent non-neoplastic tissues, both at the tissue level (Figures S3A–S3C) and the cellular level (Figures S3D–S3F).

[See PDF for image]

Fig. 2

KRT1 was the primary contributor of green autofluorescence. Compared with normal renal tissues, the mRNA level (A) and protein level of the KRT1 (B-C) in cancerous tissues were significantly decreased (N = 4, *P < 0.05, **P < 0.01, Paired Samples t-test). Compared with HK-2 cells, the mRNA level (D) and protein level of the KRT1 (E-F) in 786-O cells were significantly decreased (*P < 0.05, **P < 0.01, Independent Samples t-test). A KRT1-silenced HK-2 cell lines (shRNA-KRT1-HK-2) was constructed. Compared to the control group, the mRNA (G) and protein levels (H-I) of KRT1 in shRNA-KRT1-HK-2 cells were significantly reduced (**P < 0.01, ***P < 0.001, Paired Samples t-test). (J/K) Compared with the control group, the green autofluorescence intensity of shRNA-KRT1-HK-2 cells was significantly decreased (**P < 0.01, Paired Samples t-test). Scale bar = 100 μm.

Green autofluorescence is a novel biomarker for RCC diagnosis

To assess the diagnostic value of KRT1-dominant green autofluorescence in RCC and determine its potential for real-time diagnosis, we prospectively collected a total of 174 pairs of tissue samples (174 cancerous tissue samples + 174 adjacent normal renal tissue samples) from 60 RCC patients (Table S1). As shown in Fig. 3A, in 10% (6/60) of RCC patients, the tumor tissues exhibited higher autofluorescence intensity compared with the adjacent normal renal tissues. Overall, the autofluorescence intensity of tumor tissues was 47.5% lower than that of adjacent normal renal tissues (Fig. 3B). Then, all the tissue samples were divided into training set (121 pairs of tissue samples from 42 RCC patients) and validation set (53 pairs of tissue samples from 18 RCC patients) randomly. ROC analysis revealed an area under the curve (AUC) of 0.880 (Fig. 3C) by using the training set. When the optimal cut-off value was determined to be 27.45 using the Youden index method, the sensitivity and specificity were 0.843 and 0.835, respectively.

Based on the validation set data comprising 53 pairs of tissue samples, we assessed the performance of the optimal cutoff value (27.45) for distinguishing cancerous from non-cancerous tissue in patients with RCC using R software. The results demonstrated that this cutoff value achieved a true positive rate (TPR) of 0.755 (40/53), indicating that 75.5% of actual cancerous tissues were correctly identified. The true negative rate (TNR) was 0.943 (50/53), showing that 94.3% of non-tumor tissues were accurately classified. Additionally, the overall accuracy of the model was 0.849 (90/106), meaning that 84.9% of all samples were correctly classified. The precision rate was 0.930 (40/43), indicating that 93.0% of predicted cancerous samples were indeed cancerous. The recall rate, equivalent to the TPR, was also 0.755 (40/53), reflecting the model’s ability to identify cancerous tissues. Finally, the F1 score of 0.833, which is the harmonic mean of precision and recall, provided a balanced measure of the model’s performance. Collectively, these metrics indicated that the cutoff value of 27.45 performed well in differentiating cancerous from non-cancerous tissue in RCC patients (Fig. 3D).

[See PDF for image]

Fig. 3

KRT1-mediated green autofluorescence may become a novel biomarker for RCC diagnosis. A. Among 60 RCC patients, 6 RCC patients had cancerous tissues with higher autofluorescence intensity than adjacent normal renal tissues. B. Compared with non-neoplastic tissues, the average autofluorescence intensity in carcinoma tissues decreased by more than 47.5% (N = 60 Patients, ***P < 0.001, Wilcoxon signed-rank test). C. Taking the autofluorescence intensity as a variable, the ROC analysis based on the training set showed that the area under the curve (AUC) for differentiating cancerous tissues from non-neoplastic tissues in RCC patients was 0.880. When the optimal cut-off value was determined to be 27.45 by the Youden index method, the sensitivity and specificity were 0.843 and 0.835, respectively (N = 121 pairs of samples). D. Based on the validation set, the accuracy, precision, recall, F1-score, TPR, and TNR of this optimal cut-off value (27.45) for distinguishing between cancerous tissues and non-neoplastic tissues in RCC patients were 0.849, 0.930, 0.755, 0.833, 0.755, and 0.943, respectively (N = 53 pairs of samples).

Discussion

This study found that autofluorescence was significantly reduced in cancerous tissues of RCC patients (predominantly clear cell RCC, with the remainder being papillary RCC and chromophobe RCC) compared to normal renal tissues (Table S1). This phenomenon was also observed in the renal cancerous cell line (786-O cells). Further validation in KRT1/7-knockdown HK-2 cell models confirmed that KRT1 was the primary contributor to autofluorescence differences under 488 nm excitation light, while KRT7 and KRT8 made minimal contributions. This conclusion is supported by the following: after constructing shRNA-KRT1/7-transfected HK-2 cells (Fig. 2/S2), The autofluorescence was detected using a confocal laser scanning microscope. The results showed that compared with the control group, the green autofluorescence in KRT1-knockdown HK-2 cells was significantly reduced. Quantitative analysis indicated that the fluorescence intensity was decreased by 73% (p < 0.01), whereas KRT7 knockdown caused no significant change in autofluorescence intensity (p > 0.05, Figure S2). Additionally, the overexpression of KRT8 in renal cancerous tissues and 786-O cancerous cells was inconsistent with the trend of decreased autofluorescence intensity in 786-O cells and cancerous tissues (Figure S3). These findings indicate that KRT1 may be the main contributor to green autofluorescence under 488 nm excitation light. Thus, the reduction in autofluorescence intensity in RCC is likely primarily attributed to the lower expression of KRT1 in cancerous tissues compared to normal renal tissues. However, the green autofluorescence was not completely abolished, which may be attributed to the efficiency of KRT1 knockdown, the involvement of other cytokeratins, and other minor autofluorescent molecules in the same emission range etc. It is worth noting that substances capable of producing autofluorescence include keratin, flavin (FAD), elastin, lipofuscin, bilirubin and its derivatives¹¹. Among these, substances emitting in the 500–550 nm range include keratins, FAD, and bilirubin (not expressed in the kidney). Relevant studies have shown that the optimal excitation light range for FAD fluorescence spectra is 350–450 nm^12,13. Although FAD can also absorb 488 nm laser light and be excited, the absorption is reduced by approximately 50% compared to that at 450 nm¹⁴. We believe that FAD may contribute to the reduction of tumor autofluorescence, but it is definitely not the primary factor. Furthermore, after knocking down KRT1 expression in HK-2 cells, the autofluorescence intensity decreased by approximately 73%. This data has been validated through multiple experiments and at least indicates that under our experimental conditions (488 nm excitation light), KRT1 is the primary source of tumour autofluorescence.

More importantly, this study for the first time revealed the key role of KRT1 in the differential green autofluorescence intensity in RCC, providing new insights for the real-time diagnosis of RCC. Through analysis of autofluorescence differences between tumor and adjacent non-neoplastic tissues under 488 nm excitation light and 500–550 nm emission wavelength, we found that the green autofluorescence intensity in cancerous tissues was significantly lower than in non-neoplastic tissues. ROC curve analysis showed that using green autofluorescence intensity as a variable, RCC tumor tissues could be differentiated from non-neoplastic tissues with high sensitivity (0.843) and specificity (0.835), suggesting strong clinical diagnostic potential. Its clinical applications include tumor lymph node dissection and determination of positive surgical margins. Currently, poor intraoperative tumor localisation and a high rate of positive margins are important factors contributing to the poor prognosis obtained in patients undergoing PN¹⁵. Intra-operative frozen (IFS) is a reliable method for pathological assessment of diseased tissue during surgery¹⁶. It can be prepared and pathologically diagnosed in approximately half an hour, providing an relatively immediate assessment of the surgical margins with about 90% accuracy, thus reducing the risk of postoperative recurrence¹⁷. In addition, immunofluorescence (IF) uses fluorescently labeled antibodies to bind with target antigens and observes signals via fluorescence microscopy to achieve antigen localization, identification, and quantification¹⁸. This is achieved by using the expression patterns of specific protein markers to aid in pathological diagnosis¹⁹. CK AE1/AE3 is a broad-spectrum cytokeratin antibody mixture composed of two monoclonal antibodies, AE1 and AE3²⁰. In the differential diagnosis of RCC, it helps distinguish renal epithelial tumors (such as RCC) from other non-epithelial tumors²¹. Thus, IF based on CK AE1/AE3 can improve the diagnostic accuracy of some renal cancers. It should be noted that CK AE1/AE3 is not a specific marker for all RCCs, as some RCCs may not express or only weakly express cytokeratin²². For example, Akgul et al. evaluated the IHC expression profile of TFE3 gene fusion-related RCC (TFE3-RCC) and its correlation with fluorescence in situ hybridization detection. The results showed that CK7 was negative in the majority of TFE3-RCC cases (59 out of 74 cases), while CK AE1/AE3 was positive in 47 out of 66 cases (71.2% of TFE3-RCC cases), with diffuse expression in 42.4% of the positive cases²³. Therefore, in practical applications, CK AE1/AE3 should be interpreted in conjunction with other clinical and pathological features for comprehensive assessment. Additionally, both IFS and IF are relatively time-consuming, involve technically complex procedures, and require a certain amount of tissue (invasive sampling). Moreover, the interpretation of IF fluorescence images demands specialized training, whereas most pathology departments are more familiar with the chromogenic patterns of IHC. On the contrary, this real-time and convenient diagnostic method based on autofluorescence may serve as an important complement to the currently routinely used invasive tests, such as IFS and IF.

It is worth noting that, based on the training dataset, our ROC analysis revealed that when the optimal cutoff value was set at 27.45, the sensitivity and specificity were 0.843 and 0.835, respectively. This may suggest that the classifier’s overall accuracy in distinguishing cancerous tissue from adjacent normal renal tissue in these RCC patients is not extremely excellent (AUC = 0.880 < 0.9). In addition, we also observed heterogeneity and partial overlap in the distribution of average autofluorescence intensity values between cancerous tissue and adjacent normal renal tissue in 60 RCC patients (Fig. 3A shows that among the 60 RCC patients, 6 patients (2 with pRCC, 1 with chRCC, and 3 with ccRCC) exhibited higher autofluorescence in cancerous tissues compared to normal renal tissues. Based on more detailed patient data, these RCC patients were found to have a more severe inflammatory phenotype, which is consistent with our previous findings that tissue inflammation can increase autofluorescence⁹). Meanwhile, we further evaluated the impact of the intra-group and inter-group variability (coefficient of variation (CV)) of autofluorescence intensity values between cancerous and adjacent normal renal tissues in these RCC patients on the research outcomes. The results showed that the average CV of autofluorescence intensity values in adjacent normal renal tissues within RCC patients (intra-group) was 11.27%, while the average CV of autofluorescence intensity values in cancerous tissues was 9.42%. This indicates that there is a certain difference in autofluorescence intensity values between cancerous tissues and adjacent normal renal tissues within RCC patients. However, the overall CV is relatively low, suggesting that the intra-patient variability is relatively small and generally controllable. The high inter-group variability (CV_T = 65.80% /CV_N = 40.84%) suggests that there is a large variability in autofluorescence intensity values between cancerous tissues and adjacent normal renal tissues among different RCC patients. If the inter-group variability is high, even if there are significant differences in the mean values (Fig. 3B), the classifier may still have difficulty distinguishing between the two groups. This partial overlap may lead to the classifier producing more false positives and false negatives.

Therefore, we plan to increase the sample size in future studies to enhance the overall classification accuracy of the diagnostic model. Additionally, the diagnostic model used in this study relies solely on the parameter of autofluorescence intensity, which still has certain limitations (AUC = 0.880). In recent years, endoscopic systems utilizing multispectral imaging technology have been validated as effective tools for real-time assessment of a range of malignancies^24,25. This highlights the advantages and potential of autofluorescence-based multispectral imaging technology in instantaneous diagnosis. Thus, Future studies will combine patients’clinical data with hyperspectral emission profiles to improve autofluorescence diagnostics’ sensitivity and specificity for clinical applications.

Conclusion

In summary, Our study shows that the autofluorescence of renal cancerous tissues under 488 nm excitation light is primarily derived from KRT1, and this autofluorescence has the potential to serve as a biomarker for real-time diagnosis of RCC. The diagnostic efficacy of the single-variable model used in this study needs further optimization through a multi-factorial diagnostic model that includes patient age, sex, tumor type, stage, size, and spectral characteristics. We plan to collect large sample data in the future to construct a multi-factorial diagnostic model aimed at achieving a more satisfactory diagnostic outcome.

Acknowledgements

We gratefully acknowledge the Ethics Committee of the First Affiliated Hospital of Gannan Medical University and the individuals who have contributed to our research.

Author contributions

X.Z. designed the concept. W.W., J.Z. and T.X. conducted the experiments and drafted the manuscript. W.W., Y.L., Y.W., and G.W. provided a critical revision of the manuscript for important intellectual content. T.X. and X.Z. acquired funds, H.L., L.Z., and B.J. performed the statistical analysis and W.Y. provided study supervision.All authors reviewed the manuscript.

Funding

This study was supported by the Jiangxi Provincial Key R&D Program (No. 20212BBG71013), Ganzhou Science and Technology Innovation Talent Project (No. 2022CXRC9621), Jiangxi Department of Education Science and Technology Project (No. GJJ190819), and Ganzhou Science and Technology Guidance Program (No. GZ2023ZSF097).

Data availability

The datasets used and/or analysed during the current study available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

References

1. Ricketts, CJ et al. The Cancer genome atlas comprehensive molecular characterization of renal cell carcinoma. Cell. Rep.; 2024; 43, 4 113063.1:CAS:528:DC%2BB3sXhslCnsbbI [DOI: https://dx.doi.org/10.1016/j.celrep.2023.113063] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/38578829]

2. Hsieh, JJ et al. Renal cell carcinoma. Nat. Reviews Disease Primers; 2017; 3, 17009. [DOI: https://dx.doi.org/10.1038/nrdp.2017.9] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/28276433]

3. Nezami, B. G. & Maclennan, G. T. Clear cell renal cell carcinoma: A comprehensive review of its histopathology, genetics, and differential diagnosis. Int. J. Surg. Pathol.10668969241256111https://doi.org/10.1177/10668969241256111 (2024).

4. Liu, Z et al. Partial versus radical nephrectomy for the treatment of pT3aN0M0 renal cell carcinoma: A propensity score analysis. Asian J. Surg.; 2023; 46, 9 pp. 3607-3613. [DOI: https://dx.doi.org/10.1016/j.asjsur.2023.04.058] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/37147255]

5. Egen, L et al. Biophotonics-Intraoperative guidance during partial nephrectomy: A systematic review and Meta-analysis. Eur. Urol. Focus; 2024; 10, 2 pp. 248-258. [DOI: https://dx.doi.org/10.1016/j.euf.2024.01.005] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/38278713]

6. Matsui, R; Inaki, N; Tsuji, T. Diagnosis of advanced gastric cancer using image enhancement and autofluorescence imaging systems. Asian J. Endoscopic Surg.; 2021; 14, 4 pp. 700-706. [DOI: https://dx.doi.org/10.1111/ases.12925]

7. Zhang, M et al. Decreased green autofluorescence of lung parenchyma is a biomarker for lung cancer tissues. J. Biophotonics; 2022; 15, 8 e202200072.1:CAS:528:DC%2BB38XhvVGnsbfF [DOI: https://dx.doi.org/10.1002/jbio.202200072] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/35429151]

8. Kwiatek, S et al. Can fluorescence and autofluorescence imaging be useful in diagnosis of basal cell cancer? Proposition of algorithms. Photodiagn. Photodyn. Ther.; 2020; 30, 101697.1:CAS:528:DC%2BB3cXksFGksrs%3D [DOI: https://dx.doi.org/10.1016/j.pdpdt.2020.101697]

9. Zhang, M. et al. Keratin-Based epidermal green autofluorescence is a common biomarker of organ injury. J. BioRxiv. 564112https://doi.org/10.1101/564112 (2020).

10. Skinnider, BF et al. Distribution of cytokeratins and vimentin in adult renal neoplasms and normal renal tissue: potential utility of a cytokeratin antibody panel in the differential diagnosis of renal tumors. Am. J. Surg. Pathol.; 2005; 29, 6 pp. 747-754. [DOI: https://dx.doi.org/10.1097/01.pas.0000163362.78475.63] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/15897741]

11. Hegyi, J et al. New developments in fluorescence diagnostics. J. Dtsch. Dermatol. Ges; 2011; 9, 5 pp. 368-372. [DOI: https://dx.doi.org/10.1111/j.1610-0387.2011.07582.x] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/21251227]

12. Ramanujam, N. Fluorescence spectroscopy of neoplastic and non-neoplastic tissues. Neoplasia; 2000; 2, 1–2 pp. 89-117.1:STN:280:DC%2BD3cvhvFKqtA%3D%3D [DOI: https://dx.doi.org/10.1038/sj.neo.7900077] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/10933071][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1531869]

13. Kolenc, OI; Quinn, KP. Evaluating cell metabolism through autofluorescence imaging of NAD(P)H and FAD. Antioxid. Redox Signal.; 2019; 30, 6 pp. 875-889.1:CAS:528:DC%2BC1MXptlarug%3D%3D [DOI: https://dx.doi.org/10.1089/ars.2017.7451] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/29268621][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6352511]

14. Gaitan, B et al. Development of an endoscopic Auto-Fluorescent sensing device to aid in the detection of breast Cancer and inform photodynamic therapy. Metabolites; 2022; 12, 11 1097.1:CAS:528:DC%2BB38XjtFahurbJ [DOI: https://dx.doi.org/10.3390/metabo12111097] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/36422237][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9697641]

15. García-Perdomo, HA et al. Effect of positive surgical margins in patients who undergo a partial nephrectomy regarding recurrence, overall survival, Recurrence/Progression-Free survival, and Metastasis-Free survival. A systematic review and Meta-Analysis. Clin. Genitourin. Cancer; 2022; 20, 5 pp. 459-472. [DOI: https://dx.doi.org/10.1016/j.clgc.2022.05.011] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/35840533]

16. Algaba, F et al. Intraoperative frozen section diagnosis in urological oncology. Eur. Urol.; 2005; 47, 2 pp. 129-136.1:STN:280:DC%2BD2M%2FjsVegsg%3D%3D [DOI: https://dx.doi.org/10.1016/j.eururo.2004.08.010] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/15661406]

17. Garcia, MT et al. Accuracy of frozen section in intraoperative margin assessment for breast-conserving surgery: A systematic review and meta-analysis. PloS One; 2021; 16, 3 e0248768.1:CAS:528:DC%2BB3MXnt1Gnt7Y%3D [DOI: https://dx.doi.org/10.1371/journal.pone.0248768] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/33735315][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7971883]

18. Im, K et al. An introduction to performing Immunofluorescence staining. Methods Mol. Biol.; 2019; 1897, pp. 299-311.1:CAS:528:DC%2BC1MXitVCqu7%2FF [DOI: https://dx.doi.org/10.1007/978-1-4939-8935-5_26] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/30539454][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6918834]

19. Truong, LD; Shen, SS. Immunohistochemical diagnosis of renal neoplasms. Arch. Pathol. Lab. Med.; 2011; 135, 1 pp. 92-109. [DOI: https://dx.doi.org/10.5858/2010-0478-RAR.1] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/21204715]

20. Ordóñez, NG. Broad-spectrum immunohistochemical epithelial markers: a review. Hum. Pathol.; 2013; 44, 7 pp. 1195-1215.1:CAS:528:DC%2BC3sXislems7Y%3D [DOI: https://dx.doi.org/10.1016/j.humpath.2012.11.016] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/23427873]

21. Kalinin, Y et al. Lingual metastasis as the first presentation of clear cell renal cell carcinoma: report of a rare case clinically mimicking a benign lesion. Oral Oncol.; 2023; 137, 106293.1:CAS:528:DC%2BB3sXit1Oiug%3D%3D [DOI: https://dx.doi.org/10.1016/j.oraloncology.2022.106293] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/36565486]

22. Theparee, T; Umetsu, SE; Chan, E. Pancreatic serous neoplasm and metastatic clear cell renal cell carcinoma: diagnostic pitfalls resolvable by a panel of immunohistochemical stains to include PAX8 and CK7 but not CAIX. Am. J. Surg. Pathol.; 2025; 49, 4 pp. 394-402. [DOI: https://dx.doi.org/10.1097/PAS.0000000000002357] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/40096283]

23. Akgul, M et al. Morphologic and immunohistochemical characteristics of fluorescent in situ hybridization confirmed TFE3-Gene fusion associated renal cell carcinoma: A single institutional cohort. Am. J. Surg. Pathol.; 2020; 44, 11 pp. 1450-1458. [DOI: https://dx.doi.org/10.1097/PAS.0000000000001541] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/32701515]

24. Kriegmair, MC et al. Multiparametric cystoscopy for detection of bladder Cancer using Real-time multispectral imaging. Eur. Urol.; 2020; 77, 2 pp. 251-259. [DOI: https://dx.doi.org/10.1016/j.eururo.2019.08.024] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/31563499]

25. Grune, B et al. Ex vivo validation of a real-time multispectral endoscopic system for the detection and biopsy of bladder tumors. Transl Androl. Urol.; 2021; 10, 6 pp. 2373-2383. [DOI: https://dx.doi.org/10.21037/tau-20-137] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/34295724][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8261439]