According to global cancer statistics in 2020, there are estimated 905,677 new cases and 830,180 deaths from liver cancer, ranking seventh and second for cancer incidence and death, respectively, and most of these cases occur in East Asia (1). Early diagnosis and curative treatment are crucial for improving the prognosis of patients with primary hepatic carcinoma (PHC) (2).

Aptamers are artificial nucleic acid ligands of biological molecules that are selected via the systematic evolution of ligands by exponential enrichment (SELEX) method (3, 4). Aptamers are similar to antibodies in function but superior to antibodies in application and thus promising in medical diagnosis (5). By using aptamer-based fluorescence spectroscopy, Wu et al. (6) developed a highly sensitive and specific method for the detection of non-small cell lung cancer cells in human serum and pleural effusion, with an area under the receiver operator characteristic curve (AUROC) of 0.974 for the early diagnosis of lung cancer.

Serum, the largest reservoir of disease biomarkers, provides rich information on disease occurrence and development (7) and therefore is a valuable complex target for selecting aptamers against disease biomarkers. Gold et al. (8) selected a group of aptamers against serum and then developed an aptamer chip that could detect 813 serum proteins with high throughput, and they found 12 proteins related to lung cancer from them and established a valuable discriminant model for lung cancer diagnosis, including early lung cancer (9).

Previously, we developed a universal method for the selection of serum aptamers and generated a group of aptamers against PHC serum (10). Briefly, the sera from patients with hepatocellular carcinoma (HCC) or intrahepatic cholangiocarcinoma (ICC) were pooled and used as targets for positive selection, and sera from normal controls were pooled and used as targets for negative selection. The pooled serum was incubated with the random single-stranded oligonucleotide library (artificially synthesized) and then the free oligonucleotide sequences (in negative selection) or bound oligonucleotide sequences (in positive selection) were separated and amplified for the next round of selection. The procedure (incubation, separation and amplification) was iterated to select and enrich the oligonucleotide sequences capable of binding to specific molecules (biomarkers) of PHC serum. After three rounds of negative selection followed by nine rounds of positive selection, the bound sequences of the last round of selection were cloned and sequenced, and finally we obtained a group of aptamers that were able to specifically bind to the serum biomarkers of PHC and to be used in the detection of PHC.

Based on analyzing the binding bands of aptamers to serum targets via polyacrylamide gel electrophoresis (PAGE), we found that some of these aptamers exhibited high diagnostic performance for PHC, with AUROCs superior to that of AFP (10). However, the PAGE-based procedure for the diagnostic evaluation of aptamers is low-throughput, laborious, and time-consuming, although it has the advantages of visibility and low cost. Fluorescence intensity is a frequent and simple method for the detection of aptamer-target binding (11, 12). A label-free fluorescence-based aptamer sensor could detect low abundance of exosomes derived from cancer cells in human blood with high sensitivity and specificity (13). Therefore, a fluorescence-based method to detect the binding of aptamers to serum targets may be an ideal alternative for the PAGE-based method.

Serum autofluorescence and cell-free DNA (cfDNA)-related fluorescence intensities are correlated with tumors (14, 15). In our previous work, by using a conventional real-time PCR system as a fluorimeter, we conveniently detected the autofluorescence and cfDNA-related fluorescence intensities of serum specimens, and by combining them, we developed a simple and high-throughput diagnostic method for PHC, but the method was not robust in differentiating PHC from liver cirrhosis (LC) (16). We speculated that using the same method to additionally detect the serum fluorescence intensity of the aptamer against PHC serum and combining it with the autofluorescence and cfDNA-related fluorescence intensities (together named aptamer-based triple serum fluorescence intensities, ATSFI) could yield robust diagnostic performance in differentiating PHC from LC.

Therefore, in the present study, we used a real-time PCR system to sequentially detect ATSFI in patients with PHC or LC in one tube and combined the indicators of ATSFI to develop a novel method with robustness, convenience and clinical potential for the diagnosis of PHC.

In the present study, we successfully developed a novel and simple but robust method for the diagnosis of PHC. We sequentially measured the ATSFIs of PHC and LC in one tube by using a conventional real-time PCR system, evaluated the diagnostic value of ATSFIs for PHC, and analyzed the associations of ATSFIs with clinical and pathological characteristics. We found that some ATSFI indicators alone or in combination were capable of differentiating PHC (including early-stage HCC, small PHC and AFP-negative PHC) from LC with excellent performance, independence of but complementary to AFP, and good clinical utility and that aptamer-related fluorescence intensities were generally independent of the clinicopathological characteristics of PHC but correlated with multiple laboratory characteristics of PHC serum.

The ATSFI method has the advantages of high throughput, rapidity, convenience, and low cost. The original serum was directly used to measure autofluorescence followed by the measurements of cfDNA-related fluorescence intensity after adding EvaGreen and aptamer-related fluorescence intensity after adding aptamer and incubation. The ATSFIs could be measured sequentially in one tube within one hour, which not only made the measurement convenient, rapid and economical but also minimized measurement error. The real-time PCR system was used as a fluorimeter to measure the ATSFIs, which provided the advantages of high throughput in sample detection, precise control in detection temperatures, convenience in instrument accessibility, and “zero cost” in fluorescence detectors.

The ATSFI method exhibited high diagnostic performance for PHC. The cfDNA-related fluorescence indicators (SE-S8 and SE-S37) and aptamer-related fluorescence indicators (SEA-SE8 and SEA-SE37) were valuable in the differentiation of PHC from LC, particularly the indicator SEA-SE37, which exhibited a higher AUROC than AFP and was comparable to our previous PAGE-based method (10). The diagnostic model that incorporated the six original ATSFI indicators exhibited excellent diagnostic performances for PHC, including AFP-negative PHC, early-stage HCC and small PHC, with AUROCs greater than 0.93 and accuracies greater than 85%. Furthermore, the diagnostic value of the ATSFI in combination with AFP was further improved, even in AFP-negative PHC, with AUROCs up to approximately 0.95 and accuracies up to approximately 90%, which indicates that the ATSFIs are independent of but complementary to AFP in the diagnosis of PHC and therefore have good feasibility in clinical practice.

The ATSFI method is superior to our previous methods. Our previous diagnostic models that combined serum autofluorescence and cfDNA-related fluorescence intensities with conventional blood tests exhibited excellent diagnostic performances for the diagnosis of PHC (16, 21), comparable to the ATSFI method. However, the ATSFI model was developed with the six original fluorescence indicators from one test and without the combination of conventional blood tests, which does not need the collection of conventional blood test results and avoids the blood test results incomparable between medical centers due to difference of methods or reagents. Furthermore, the ATSFI method overcome the disadvantages of the low-throughput, laborious and time-consuming nature of our previous PAGE-based method that combined the grey indicators of two aptamers with AFP and exhibited diagnostic performance comparable to the ATSFI method (10). Therefore, the ATSFI method shows an obvious improvement in convenience and clinical feasibility compared with previous methods.

Decision curve analysis (DCA) has been widely utilized to evaluate the practical value of an indicator or predictive model in clinical decision-making (19, 20), which compares the net benefit of selective intervention according to the indicator or model with the net benefit of nonselective intervention (towards all or none), and greater net benefit than the all or none indicates that the indicator or model has better clinical utility. In the present study, DCA demonstrated that patients at a risk of PHC could obtain a much greater net benefit from diagnostic intervention to be decided according to the ATSFI compared with AFP, especially in combination ( Figure 6 ). These results suggest that the ATSFI is valuable in clinical application.

Serum autofluorescence is produced by native fluorescent substances in the blood and has been reported to have potential in the diagnosis of liver cancer (22), but in the present study, serum autofluorescence indicators (S8 and S37) were not valuable for the diagnosis of PHC, although they were significantly lower in PHC than in LC. We observed a strong correlation of autofluorescence intensities with serum levels of both total and direct bilirubin in correlation analysis, which indicates that the autofluorescence is mainly from serum bilirubin. Since serum bilirubin is just an indicator for the impairment of liver function, it is rational for the serum autofluorescence indicators to be not significant in differentiating PHC from LC.

When the nucleic acid dye EvaGreen was added, the serum fluorescence intensity increased in both PHC and LC, but the corresponding fluorescence indicators (SE8 and SE37) showed weak diagnostic value in differentiating PHC and LC. However, the changes in fluorescence intensity (SE-S8 and SE-S37) exhibited significant diagnostic value due to the different increases between PHC and LC, indicating that compared with LC, PHC has higher levels of circulating cfDNA that bind to EvaGreen and emit stronger fluorescence. Circulating cfDNA is the fragment of DNA derived from the apoptosis and necrosis of cells, especially in patients with tumor (23). Circulating cfDNA has potential diagnostic value in liver cancer (24), but it is not robust enough to be used independently (25). Here, an interesting issue is whether HBV-DNA affects cfDNA levels and their fluorescence intensities? The patients in the present study were treating with nucleic acid analogues according to the guideline (26), and thus only 137 patients had measurable HBV-DNA levels from 2.52×10² to 9.01×10⁷ IU/mL. We performed a bivariate correlation analysis between HBV-DNA levels and cfDNA fluorescence intensities in these patients, and no correlation was found (r=0.015, P=0.863 for HBV-DNA and SE8; r=0.015, P=0.866 for HBV-DNA and SE37). This may be related to the fact that HBV-DNA in the blood is not free but present in the Dane particles (27).

After the aptamer was further added, the fluorescence intensity ascended continuously. Similarly, the corresponding original fluorescence indicators (SEA8 and SEA37) presented weak diagnostic value for PHC, but due to the fluorescence ascending less in PHC than LC, the related indicators SEA-SE8 and SEA-SE37 exhibited strong diagnostic value and were more valuable than cfDNA-related indicators. The change in fluorescence intensity after adding aptamers depends on the serum levels of free aptamers; that is, when more aptamers bind to targets, fewer free aptamers bind to the dyes, which results in weaker fluorescence intensity (28, 29). Thus, the smaller increase in fluorescence intensity in PHC than in LC suggests more specific targets of the aptamer in PHC serum than in LC serum.

The ATSFIs were measured at two temperature points, 8°C and 37°C. The serum fluorescence intensity was found to be stronger at 8°C than at 37°C, but the diagnostic performance was better at 37°C than at 8°C, and SEA-SE37 exhibited the greatest AUROC among all single fluorescence indicators. These results indicate that nonspecific binding at higher temperatures is reduced, which is a well-known phenomenon in biological binding reactions. Interestingly, the fluorescence indicators at the two temperature points were all independent variables in the diagnostic model, suggesting that they differently contribute to the diagnosis of PHC.

No significant differences were found between the aptamer-related fluorescence intensity indicator SEA-SE37 (similar in SEA-SE8) and the PHC subgroups of age, sex, Child-Pugh grade, tumor type, tumor size, BCLC stage, and TNM stage, suggesting that aptamer-related fluorescence intensity is independent of these clinicopathological characteristics of PHC. However, SEA-SE37 was significantly different between AFP-positive and AFP-negative PHCs, but its correlation with AFP was weak. These features give the indicator SEA-SE37 high diagnostic performance in various subtypes of PHC, including AFP-negative PHC, small PHC, and early-stage HCC, important for the translational application of the aptamer in clinical practice. The mechanism that SEA-SE was irrespective of PHC clinicopathological characteristics is unclear. We speculate that this may be related to a “combination detection” effect. The aptamer was generated using pooled serum and thus may target multiple types of molecules due to the extremely complex of serum components. Therefore, in the reaction between aptamer and serum, the aptamer binds to multiple targets simultaneously, similar to the combined detection of multiple indicators that yields better performance than a single indicator. Additionally, the targets of the aptamer may be irrespective of tumor size and stage of PHC. However, these speculations have yet to be confirmed.

We previously reported that the combination of serum autofluorescence and cfDNA fluorescence intensities had moderate diagnostic value for differentiating PHC from LC (21). In the present study, we combined aptamer-related fluorescence intensity with autofluorescence and cfDNA fluorescence intensities (“triple” instead of “dual”) and obtained robust diagnostic performance. The ATSFIs reflect the characteristics of PHC serum in three dimensions: autofluorescence, cfDNA-related fluorescence and aptamer-related fluorescence. Serum autofluorescence intensity has a certain correlation with malignant tumors (14). Serum cfDNA is well known to be associated with tumors and their levels can be analyzed by fluorescence (15). The aptamer specifically binds to targets, and the bound status between aptamers and targets can be analyzed by fluorescence (30). Therefore, the combination of the three dimensions in a logistic diagnostic model was able to provide powerful diagnostic ability in the present study. This novel concept could be transformed to other tumor diagnoses if a specific aptamer is available.

Although the aptamer-related fluorescence intensity (SEA-SE37) exhibited high diagnostic performance for PHC, even better than AFP, we did not know the target molecules of the aptamer. Therefore, we tried to realize the underlying mechanism of the aptamer for the diagnosis of PHC from clinical aspects. The SEA-SE37 was significantly correlated with many blood laboratory indicators with low or moderate correlation coefficients (absolute r=0.219-0.420) ( Table 3 ), while some conventional laboratory blood tests have been proven to be valuable for the diagnosis of PHC in our studies (21, 31, 32) and others (33, 34), particularly in combination. Furthermore, SEA-SE37 was independent of the clinicopathological characteristics of PHC ( Figure 7 ). The weak or no associations of aptamer-related fluorescence intensity with conventional laboratory blood tests and clinicopathological indicators may indicate that the aptamer is generally independent of single clinical and pathological factors but dependent on comprehensive characteristics of PHC serum, which makes SEA-SE37 present better diagnostic performance than AFP.

The clinical profiles of the PHC and LC groups differed significantly ( Table 1 ). To realize whether the diagnostic performance of ATSFIs was related to the difference in clinical profiles, we performed propensity score matching (PSM) for PHC and LC using age, sex, etiology, Child-Pugh grade, and liver function indicators, and 150 pairs were successfully matched ( Supplemental Table S1 ). In the matched set, SEA-SE37 remained the single indicator with the best diagnostic value and had an AUROC comparable to that before matching (0.858 vs. 0.879) ( Supplemental Figures S1A, B ), and the AUROC of the ATSFI model was also comparable to that before matching (0.917 vs. 0.937) ( Supplemental Figures S1C, D ). In addition, the pre-match AFSFI model exhibited an AUROC of 0.915 in the matched set ( Supplemental Figures S1E ). These results suggest that diagnostic performance is similar before and after PSM and virtually unaffected by the clinical profiles.

In the differences of the clinical profile, higher ALB levels and lower TBIL levels were found in the PHC group compared to the LC group (contradicting the tendency of PHC to occur in advanced fibrosis) ( Table 1 ). This was due to the different liver function status between PHC and LC in the present study. PHC patients were hospitalized mainly for surgery and other antitumor treatments, which usually require liver function to be compensated. In contrast, LC patients were hospitalized mainly because of deteriorating or decompensated liver function. Therefore, the proportion of Child-Pugh grade A was significantly greater in the PHC group than in the LC group, while the opposite was for grades B and C ( Table 1 ).

In the present study, we did not collect non-LC patients as controls, including chronic hepatitis and normal controls, due to the following considerations. Firstly, our new method is necessary to compare with the best current biomarker AFP, and serum AFP levels are usually elevated in the patients with chronic hepatitis hospitalized at our hospital due to active or severe disease, which may therefore lead to bias in this comparison. Secondly, from a clinical point of view, the differential diagnosis between liver cancer and cirrhosis is more important than that between liver cancer and hepatitis, because majority of PHC occurred in the background of cirrhosis rather than hepatitis. Thirdly, our previous study found that the grayscale value of aptamer-serum binding bands between cirrhosis and chronic hepatitis were very close (10), suggesting that cirrhosis may be “representative” of chronic hepatitis to some extent. As for the normal control, this is a “case-control” diagnostic study that tends to exaggerate diagnostic effects according to the STARD guidelines (35). However, the above-mentioned control groups should still be considered in the future for systematic evaluation of our new diagnostic method.

Additionally, further studies in other aspects will help us to better understand the ATSFI assay for the diagnosis of PHC. Validation studies in external patient cohorts can further evaluate the diagnostic performances of the ATSFIs for PHC. The etiology of patients in the present study was predominantly hepatitis B virus; thus, the diagnostic performance of the ATSFI method should be evaluated in PHC patients caused by other factors. Studies on the target molecules of the aptamer are helpful to elucidate the mechanisms underlying aptamer-based diagnosis.

In conclusion, a novel, feasible and robust diagnostic method for PHC has been successfully developed and internally validated. This method was established on ATSFIs that were sequentially measured in one tube by a conventional real-time PCR system, with the advantages of high throughput, rapidity, convenience, and low cost. ATSFIs were excellent in the diagnosis of PHC, including AFP-negative, early-stage and small PHCs, independent of and complementary to AFP. The excellent diagnostic performance of the ATSFI method is based on the comprehensive integration of triple serum fluorescence characteristics, especially aptamer-based fluorescence, and the independence of the clinicopathological characteristics of PHC. However, further studies that evaluate the diagnostic performance in external patient cohorts, analyze the exact targets of the aptamer, and explore the underlying diagnostic mechanisms will provide more information to understand the ATSFI method.

The datasets generated for this study are available on request to the corresponding authors.

The studies involving human participants were reviewed and approved by The Ethics Committee on Medical Research of the First Affiliated Hospital of Nanchang University. The ethics committee waived the requirement of written informed consent for participation.

KZ and TW conceived and designed the study. JZ, FH, and HZ participated in the experiments. JZ, TW, and KZ analyzed and interpreted the data. SY and ZL collected clinical data. JZ, KZ, and TW prepared and revised the manuscript. All authors have read and approved the content of the final manuscript.

This work was supported by the Natural Science Foundation of China (82160444, 81702107, 81760536) and the Science and Technology Project of Jiangxi Province, China (20171ACB21055 and 20192BBG70048).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.