Brief history of tumor biomarkers development

Biomarkers are designated as “a biological molecule found in blood, other body fluids, or tissues that is a sign of a normal or abnormal process, or a condition or disease. A biomarker may be used to see how well the body responds to a treatment for a disease or condition” according to the National Cancer Institute ( Tumor biomarkers exist in tumor tissues or body fluids such as blood, urine, stool, saliva, and are produced by the tumor or the body’s response to the tumor.1 The goal of the tumor biomarker field is to develop sensitive, specific, reliable, cost-effective, reproducible, and powerful detection and monitoring strategies for tumor risk indication, tumor monitoring, and tumor classification so that patients can receive the most appropriate treatment and doctors can monitor the progress, regression, and recurrence of the tumors.2 Since the discovery of Bence-Jones protein (BJP), the first tumor biomarker, in 1846, this field has been through many stages and has made significant and substantial progress with the joint efforts of researchers, clinical staff, and patients.

Discovery and exploration stage (1847–1962)

In 1847, Henry Bence-Jones described the findings of a large number of immunoglobulin light chains from the urine of a patient with multiple myeloma, and named it BJP, the first biochemical tumor biomarker described in diagnostic laboratory medicine3 (Fig. 1). The monitoring of BJP in urine has become one of the parameters related to the diagnosis and prognosis of multiple myeloma.4 The discovery of BJP marks the beginning of research on tumor biomarkers. Subsequently, hormones, isozymes, and other tumor biomarkers that displayed abnormalities during the occurrence and development of tumors were discovered. In 1927, Selmar Ashheim and Bernhard Zondek found a gonadal stimulating substance—human chorionic gonadotropin (HCG) from the blood and urine of pregnant women.5 Later on, HCG was identified as a tumor biomarker, which is frequently associated with gestational trophoblastic disease and testicular germ cell tumor.6 In 1959, lactate dehydrogenase (LDH), the first “isoenzyme”, was discovered in the bovine heart by Clement L Markert at Johns Hopkins University,7 and numerous clinical evidence subsequently demonstrated that LDH was an essential prognostic factor for different tumors.8 In 1962, Meador found that some tumors spontaneously produced adrenocorticotropic hormone-like substances, which hindered the normal secretion mechanism of adrenocorticotropic hormone and induced metabolic abnormalities dominated by hypokalemia.9 Despite the aforementioned breakthroughs in knowledge about tumor biomarkers, these biomarkers did not translate from bench to bedside for cancer diagnosis or monitoring.

Fig. 1
figure 1

Timeline of the history of tumor biomarker

Clinical application stage (1963–1978)

The next significant advances came from GI Abelev who is well known for his 1963 discovery that mice inoculated with liver cancer cells can synthesize alpha-fetoprotein (AFP)10 (Fig. 1). AFP has been used as a biomarker in clinical screening, diagnosis, prediction, and treatment evaluation of hepatocellular carcinoma (HCC).11 At around the same time (in 1965), Goldenberg and Freeman found that carcinoembryonic antigen (CEA) in fetal colon mucosa12 contributed a crucial part in tumor diagnosis and prognosis evaluation of lung cancer,13 breast cancer,14 ovarian cancer,15 colorectal cancer (CRC),16 etc. The discovery of AFP and CEA has promoted the clinical application of tumor biomarkers. However, the application of CEA as a tumor biomarker was later challenged by Paul Lo Gerfo and his colleagues in 1971. They measured the level of CEA in the serum or plasma of 674 hospitalized patients by radioimmunoassay (RIA). It was found that CEA expression level was elevated in the serum of patients with multitudinous diseases, but not cancer-specific, which hampered the potential absolute benefit of CEA assessment independent of other surveillance tools.

Exploration stage (1979–2004)

James Watson and Francis Crick’s discovery of the DNA double helix structure in 1953 ushered in a new era in tumor biomarker research.17 After this discovery, modern molecular biology significantly promoted the research on tumor biomarkers, with a large number of genes involved in tumor occurrence and progression being discovered. In 1979, p53 was found by David Lane and confirmed by other independent groups.18,19 Although being considered as a cell tumor antigen at the beginning, p53 was defined as a cancer suppressor gene in 198920 and more than 50% of p53 was mutant in cancer patients.21 In 1981, Robert Weinberg and Geoffrey Cooper discovered the small fragments of DNA in transgenic experiments in which the transformation of mouse NIH/3T3 fibroblast cells transfected with DNA extracted from human tumors cell lines was successfully induced and soon followed the isolation of homologous oncogenes HRAS and KRAS from human tumors.22,23 This discovery paved the way for the development of the first KRASG12C inhibitor, sotorasib, which was approved by the United States Food and Drug Administration (FDA) in 2021 for non-small cell lung cancer (NSCLC) treatment.24 The first human oncogene, retinoblastoma (Rb) gene, was successfully cloned in 198625 (Fig. 1). After that, large numbers of proto-oncogenes, oncogenes, tumor suppressors, receptors, and kinases were discovered, and some of them were successfully used as tumor diagnostic, prognostic, and therapeutic biomarkers.

Innovation and development stage (2005-)

The rapid development of science and technology is driving the field of tumor biomarkers to an innovation and development stage. An increasing number of methods and technologies have been developed and applied in tumor biomarker discovery and detection. The molecular biological technologies, such as genomics, transcriptomics, proteomics,26 metabolomics,27,28 the clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated protein 9 (Cas9) gene-editing technology,29,30 and high-throughput sequencing,31 make it possible to obtain large-scale information on the diversity of tumor biomarkers. In 2005, the first high-throughput sequencer was launched, which contributed to the sequencing technology. In recent years, nucleic acid-based liquid biopsy for monitoring cancer has attracted much attention.32 For example, cell-free DNA (cfDNA) in the plasma of cancer patients contains tumor-derived DNA sequences, which can be used as biomarkers for the early detection of cancer, guiding treatment, and monitoring drug resistance.33 In 2016, ExoDx Lung (ALK), the world’s first exosomal oncology diagnostic test, was launched. It can be used for the diagnosis of NSCLC and the screening of NSCLC patients for targeted therapy with anaplastic lymphoma kinase (ALK) inhibitors. Subsequently, the epidermal growth factor receptor (EGFR) gene mutation detection kit in plasma DNA samples was launched in 2018 and used to screen patients for EGFR-targeted drug therapy. China’s self-developed “CellRich” circulating tumor cell detection system was approved by the National Medical Products Administration in 2018 and used to capture tumor circulating cells. Furthermore, the emergence of “precision medicine” has pushed the field of tumor biomarkers to a new stage, which requires the discovery of more effective tumor biomarkers and the integration of multiple tumor biomarkers to support personalized medicine.34

Clinical application of tumor biomarkers

Early screening of tumors

Early screening of tumors is the most powerful public health tool that enables early detection, thus reducing annual cancer incidence, providing a higher chance of treatment, improving patient response to medical interventions, and prolonging patient survival, especially for those cancers with high mortality such as CRC.35 In addition to invasive or expensive screening methods such as endoscopy, low-dose computed tomography, and tissue biopsy, noninvasive and cost-effective screening based on biomarkers from body fluids including blood, stool, saliva, and urine has been gaining extensive attention. To date, thousands of these biomarkers including proteins, cytokines, metabolites, hormones, microRNA, and circulating DNA have been explored, and several of them have been successfully developed and used in the early screening of cancers.2,34 For example, AFP was the first blood biomarker used for the screening of HCC in the populations since 1964.36 After that, other biomarkers called “classical” tumor markers, such as CEA, carbohydrate antigen 19-9 (CA19-9),37 carbohydrate antigen 125 (CA125),38 prostate-specific antigen (PSA),39 and LDH,40 have been used in the clinical screening of various kinds of cancers. In addition to the “classical” tumor marker, a broad range of novel biomarkers have been explored in recent years, which include microRNA41 and other RNAs,42 microbial proteins,43 circulating nucleosomes,44 circulating tumor DNA,45 and circulating tumor cells.46 Albeit currently undergoing clinical trials or preclinical studies and unavailable in the market, they have great potential for clinical screening. As some biomarkers from body fluids may be difficult to detect because of fundamental biological barriers such as short circulation times and very low density, synthetic biomarkers including small-molecule, DNA-based, mammalian cell-based, and bacterial cell-based sensors have been developed to amplify tumor signals, thus enhancing the sensitivity and efficiency of early-stage tumor detection.47 New screening tests based on these novel techniques can be used in the clinic in the near future. Collectively, for the detection of early-stage cancer, the noninvasive or minimally invasive test is ideal, and developing such techniques is desirable in clinical applications.

Tumor auxiliary diagnosis

Due to the risk of false positive or negative results, relying solely on one biomarker level is not an accurate and reliable strategy for tumor diagnosis. Instead, the combination of tumor biomarkers with other methods such as tissue biopsy and endoscopy is a promising alternative to improve the effectiveness of screening.48,49 For example, the combined detection of AFP with cfDNA can improve the specificity of HCC diagnosis to 94.4%, which was superior to that of AFP alone in terms of higher sensitivity and better clinical correlation.50 The advantages of biomarker panels have been confirmed as compared with a single biomarker, especially a panel of biomarkers reflecting changes in independent pathways. The combination of periodin (POSTN) with CA15-3 and CEA for the diagnosis of breast cancer can improve the diagnostic performance of CA15-3 and CEA.14 For CRC, the detection of hemoglobin using fecal immunochemical testing in combination with transferrin in stool improves the diagnostic accuracy for CRC.51 The combination of various diagnosis strategies with biomarkers could result in an easier, faster, more accurate, and more specific diagnosis of cancer.

Prediction of tumor prognosis and curative effect

Precision stratification of cancer patients based on prognosis and therapeutic decision biomarkers has enabled the selection of treatment strategies and more effective treatments for individual cancer patients. One successful example is to distinguish the type of breast cancer by the expression of human epidermal growth factor receptor 2 (HER2), estrogen receptor (ER), and progesterone receptor (PR) in breast cancer tissues. These biomarkers help to identify the triple-negative breast cancer (TNBC) lacking the expression of ER, PR, and HER2 which is the most aggressive type of breast cancer associated with poor prognosis and limited treatment options, thus improving the management and treatment options with the ultimate goal of improving clinical outcomes.52 Moreover, identifying curative predictive biomarkers to distinguish patients who are most likely to respond to anticancer therapy from all cancer patients enhances therapeutic efficiency, decreases treatment costs, and avoids adverse events. For example, the implementation of patient selection prior to programmed cell death 1 (PD-1)/programmed cell death ligand 1 (PD-L1) inhibitors therapy by the combination of biomarkers reflecting tumor immune microenvironment and tumor cell-intrinsic features, such as PD-L1, tumor-infiltrating lymphocyte, tumor mutational burden, mismatch-repair deficiency, and gut microbiota, could enhance the treatment effect of anti-PD-1/anti-PD-L1 therapy in clinical practice.53

Tumor recurrence monitoring

The level of tumor biomarkers is valuable for indicating the disease recurrence of tumor patients. Some classic biomarkers for tumor diagnosis and prognosis, such as PSA, CEA, CA19-9, and CA72-4, are used for indicating the recurrence of cancers including prostate cancer, gastric cancer, breast cancer, and liver cancer.54,55 The CEA is increased in most liver recurrence cases of gastric cancer (90%), while the increase of CA19-9 after surgery in patients with gastric cancer could predict peritoneal recurrence more accurately (78.9%).56 In recent years, extensive molecular and genetic characterization of disseminated tumor cells and blood-based biomarkers have contributed significantly to monitoring cancer recurrence. Postoperative methylated septin 9 in plasma may represent a potential noninvasive biomarker for CRC recurrence monitoring in addition to CRC diagnosis and prognosis compared with CEA and CA19-9.57 The circulating tumor DNA (ctDNA) minimal residual disease (MRD) following treatment in solid tumors predicts relapse and highlights the application of this potentially transformative biomarker.58

Collectively, tumor biomarkers play an active role in all aspects of clinical application, such as early screening, diagnosis, prognosis, and relapse monitoring, and are of great value in helping patients prolong their survival and improve their quality of life. To date, excellent progress has been made in the discovery and application of biomarkers. Besides classical biomarkers used in clinical practice, recent advances in molecular biology technologies have significantly improved the discovery of new candidates for cancer management, but most of them are still in the early stage of development and validation. Great effort could be made to find new biomarkers with the right degree of specificity, sensitivity, and reliability, so as to provide evidence for individualized decision-making during the overall management of cancer patients. In this review, we summarize the current progress that has been made in cancer biomarker development and discuss the promise, limitations, and further challenges in biomarker development.

Technologies used in the detection of tumor biomarkers

Multiple technologies have been developed for the detection of tumor biomarkers as follows (Fig. 2). In the past decades, various immunoassay methods have played crucial roles in the discovery of tumor biomarkers. Meanwhile, molecular hybridization technology and gene amplification detection technology further broaden the horizon of the application of tumor biomarkers in clinical practice. Immunohistochemistry (IHC) brings about the original distribution of biomarkers in fixed tissue. Furthermore, rapidly developed DNA sequencing and gene-editing technologies accelerate the speed and numbers of digging out prognostic and predictive tumor biomarkers. Other technologies, such as liquid biopsy and different microscopy technologies, as well as single-cell sequencing analysis,59 also provide tremendous convenience in cancer therapy.

Fig. 2
figure 2

Technologies for the detection of tumor biomarkers

RIA technology

RIA technology is an analytical method proposed in the late 1950s by the United States chemist Solomon A. Berson and Rosalyn S. Yalow.60 It integrates immunologic and radiolabeling techniques to quantitate minute amounts of biological substances based upon the competition between labeled and unlabeled antigens for specific antibody sites, forming antigen–antibody complexes.60,61 RIA usually uses radionuclide 125I as a tracer, which has been widely used for its advantages of highly radioactive.62,63,64 In addition, RIA is advantageous in measuring a variety of immunoreactive substances for its high sensitivity, and specificity.65,66 For example, RIA is utilized in the detection of early-stage tumors, and is an effective method in combination with clinical pathological assay to provide comprehensive evaluations of tumors.67 However, the shortcomings of RIA are also prominent, such as isotope contamination due to the radioactive wastes, the requirement of specific safety equipment, and the excessive radiation exposure of workers induced by the long incubation time. which limits its wide use.65,68 RIA tends to be eliminated with the rapid update of other immunoassay methods, such as enzyme-linked immunosorbent assay and fluorescent immunoassay (FIA), which use other substances such as fluorescent dye instead of radioactive isotopes to label antigens.

FIA technology

FIA, combining the specificity of the immunological response with the sensitivity of fluorescent technology, is a popular and fast-growing nonisotopic immunoassay technology. As a new immunoassay technology using fluorescein-labeled antibodies or antigens as tracers, the principle of FIA is similar to enzyme-linked immunosorbent assay. The fluorescein is chemically bound to the antibody (or antigen) molecule, and after that, the latter is combined with the matching antigen (or antibody). The fluorescence is observed, or the fluorescence intensity is measured by a fluorescence detector which determines the presence, distribution, and content of antigens (or antibodies) in samples. FIA has the advantages of high specificity, high sensitivity, and good practicability with cheap, stable, and safe reagents.69 Moreover, FIA avoids the risk of handling radioactive materials. Thus, FIA is widely used in the biomedical field in the measurement of drugs, hormones, and proteins; the identification of antibodies, and the quantification of antigens.70

The development of various fluorescent probes and instruments also contributes to the continuing evolution of FIA.2,71,72 Multiple FIA-related technologies with high detection sensitivities and various measurable properties have been developed, including fluorescent excitation transfer immunoassay, fluorescence polarization immunoassay, and time-resolved fluorescence immunoassays.73,74,75,76,77 For example, the multicolor quantum dots based on fluorescence polarization immunoassay have been applied in the detection of tumor biomarkers such as α-AFP and CEA.78

Molecular hybridization technology

Molecular hybridization technology is an important method for the investigation of gene expression and genome function by assessing chromosomal aberration using a fluorescent probe.79 The principle of molecular hybridization technology is to form stable double-stranded hybrid molecules between DNA or RNA from different species, thereby detecting complementary sequences or recognizing binding sites of transcription factors. Common molecular hybridization techniques include fluorescence in situ hybridization (FISH) and in situ hybridization.79,80 In situ hybridization uses labeled complementary DNA or RNA strands to localize specific DNA or RNA sequences on chromosomes or tissue sections fixed on slides (in situ), and the FISH technique helps to localize genes to different chromosomal locations. They are all molecular tools in cancer diagnosis, treatment, and prognosis.79,81 With the advantages of easy manipulation, fast hybridization process, possible automation of process and scoring, the FISH technique is wildly utilized to detect the tumor biomarkers in diagnosis and metastasis prognosis, such as the analysis of circulating tumor cells (CTCs) obtained from the patient blood sample82 in various of cancers, including lung cancer, glioma, breast cancer, ovarian cancer, and soft tissue sarcomas. Extensive prognostic and predictive biomarkers, such as ALK, mesenchymal-epithelial transition factor (c-MET), and ROS1, are identified by FISH.82 Significantly, the timeless and costless FISH remains a gold standard in ALK-rearrangements NSCLC.83 In 2011, a novel anticancer drug crizotinib and its companion, the ALK FISH probe detection kit, were simultaneously approved by the FDA, which highlights the crucial position of the FISH assay in guiding ALK-targeted therapy.84

In short, FISH is an increasingly demanded tool for biomarker research and personalized medicine despite the fact that the process of FISH may be time-consuming and costly when performed with standard chemicals and the retention of the fluorescence is limited.82

Gene amplification detection technology

Polymerase chain reaction (PCR), a molecular biological technology used to amplify specific DNA fragments, is an invaluable tool for the assessment of nucleic acids in tissues and body fluids. It can synthesize and amplify specific DNA into billions of copies in a few hours by separating the DNA into two strands and incubating it with oligonucleotide primers and DNA polymerase in vitro.85 PCR technology has developed to the third generation since the invention of Kary Mullis in 1985,86 and holds a pivotal position in biological research. The PCR technology includes three major steps: denaturation of double-stranded (ds) DNA template, annealing of primers (forward and reverse primers), and extension/elongation of dsDNA molecules.85 The quantitative real-time PCR (qPCR)-based assay is considered to be the gold standard for prognostic and predictive biomarker analysis for the quantitative advantage.85

The application of PCR in diagnostic gene mutation analysis, such as the B-raf proto-oncogene (BRAF), EGFR, Kirsten rat sarcoma viral oncogene homolog (KRAS), neuroblastoma RAS viral oncogene homolog (NRAS), and phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha (PIK3CA) genes from the blood, is meaningful in initial cancer stratification and the monitoring of cancer progression. Moreover, several PCR assays approved by the FDA are used for the diagnosis of KRAS mutation status in formalin-fixed paraffin-embedded tissue, thereby guiding anti-EGFR antibody treatment for metastatic CRC.87 Similarly, qPCR assays are effective in the detection of MRD in leukemia, such as the quantification of BCR-ABL-positive cells post-induction chemotherapy/transplantation in acute lymphoblastic leukemia (ALL).85 PCR technology is also widely used to detect abnormal genes and abnormal mRNA amplification in tumors, such as MYCN amplification in neuroblastoma.88 Ligand-targeted PCR is essential for the detection of folate receptor-positive circulating tumor cells as a potential diagnostic biomarker in pancreatic cancer.89

PCR methods are of great advantages in the detection of nucleic acid biomarkers, including relatively simple manipulation, providing rapid inexpensive diagnosis with good sensitivity, valuable for clinical molecular pathology.87 Nevertheless, several intrinsic drawbacks of PCR that restrain its application have room for improvement, such as the requirement for instruments, experienced operators, laboratory setting, and sophisticated operations.90 Collectively, PCR is a valuable tool in tumor biomarker detection, while novel PCR-based methods remain to be explored to meet the needs of patient monitoring in clinic.87

DNA sequencing technology

DNA sequencing technology is a commonly used technology in molecular biology research, which is used to analyze the arrangement of the base sequence of specific DNA fragments. The world’s first method of DNA sequencing was invented by British biochemist Frederick Sanger, who performed the first complete DNA genome sequencing, bacteriophage ϕX174 in 1977.91,92 Since that, DNA sequencing technology has witnessed rapid development, which is now in its fourth generation of DNA sequencing technology.93 It not only opens up new perspectives in traditional biology, medical research, and other fields, but also promotes the further development of bioinformatics, molecular genetics,94 genomics,95 precision medicine,96 and other disciplines, which advances the progress of life science research.

DNA sequencing technology is not only the gold standard for microbial identification but can also be used to detect the existence of tumor biomarkers which indicate the occurrence and development of tumors. At present, next-generation sequencing technology (NGS) is the most widely used among four DNA sequencing technologies in clinical practice, which can detect multiple genomic alterations including nucleotide substitutions, small insertions, deletions, copy number variations, and chromosomal rearrangements. NGS promotes the identification of somatic mutations associated with acute myeloid leukemia (AML), melanoma, mesothelioma, small cell lung cancer (SCLC), breast cancer, and prostate cancer.97 For example, NGS was applied to detect mutations of many cancer-related genes, such as TP53,98 phosphatase and tensin homolog (PTEN),99 KRAS,100 and breast cancer type 1/2 susceptibility protein (BRCA1/2).101 These detections are valuable for assessing the family history and the risk of tumorigenesis, and improving clinical diagnostics.97 Except for providing a high sensitivity in gene mutations, NGS is dramatically cost-effective and less time cost compared with current PCR-based tests. For example, the cost of using PCR to detect RAS mutations is as high as several thousand dollars, while NGS only costs one-third for detecting the same mutations. Notably, NGS simultaneously sequences the remaining gene samples in the same pathway or multiple samples in a single sequencing run with high speed and accuracy, which avoids incurring additional operating costs.97

Moreover, RNA sequencing has been utilized in multifarious aspects of cancer management, including prognostic and predictive biomarker identification, the characterization of cancer heterogeneity, and the monitoring of drug resistance. Some special genomic biomarkers, including miRNA, lncRNA, and circRNA, have been discovered by RNA sequencing.102 For example, isocitrate dehydrogenase (IDH) mutation which is a good prognostic biomarker for glioma, and nuclear cyclooxygenase 2 combined with HER2 which serve as potential biomarkers for the diagnosis and prognosis of CRC, are identified by RNA sequencing.102

In conclusion, the advent of sequencing technology sequences individual cancer genomes, which opens a new chapter in precision cancer therapy. Novel sequencing technologies have the potential to decode massive amounts of cancer genomes rapidly and cheaply to benefit cancer precision therapy.

IHC technology

IHC, a technology used to detect the distribution of antigens (or antibodies) on formalin-fixed paraffin-embedded tissue sections,103 identifies targets through antigen–antibody interactions, and the antibody binding site is identified by direct labeling or secondary labeling method.

IHC is a gold standard and ubiquitously applied technology in cancer identification and diagnosis, especially in assessing biomarkers used for characterizing tumor subtypes, confirming tissue origin, distinguishing metastasis from primary tumor, providing prognosis information, stratifying patients for treatment selection, and predicting therapy response in various cancers.104,105,106,107,108,109 The American Society of Clinical Oncologists and College of American Pathologists (ASCO/CAP) provides the HER2 scoring guidelines to determine breast cancer pathological classification and clinical stage by using IHC-based staining intensity and the percentage of HER2+ cells in cancer tissues.110 IHC is used to detect p53 in cancer tissues.111,112 The detection of the excision repair cross-complementation group 1 protein by IHC has been approved for predicting the response of NSCLC patients to chemotherapy.105

As an indispensable technology, IHC holds the unique advantage of correlating the presence of protein with its location in tissues or cells compared with other protein detection methods, which is essential for illustrating protein function in normal and pathological tissues.113 Moreover, IHC can be operated by easy preparation and automated manipulation.82 However, several limitations still exist in IHC, especially a lack of reproducibility. Conflicting results often occur when different antibodies are used. The variables of the protocol affect the reliability of IHC, including the fixation time of tissues, the absolute level of the antigen, the affinity and concentration of antibody, and the sensitivity of the detection system.111 Thus, high-quality control of regents, standardized protocols, automated IHC, or combined IHC with transcriptomics will improve the accuracy, reproducibility, and reliability of IHC and accelerate its application in the discovery and validation of biomarkers.114

Liquid biopsy technology

Liquid biopsy, a minimally invasive methodology, is used to obtain tumor-derived information from body fluids so as to facilitate cancer diagnosis.115 Currently, liquid biopsy is used to detect cfDNA, cell-free RNA, CTCs, extracellular vesicles,116,117 ctDNA,118 circulating RNA, and exosomes119,120 in blood or other body fluids.116,121 Liquid biopsy can enhance patient overall survival (OS) by improving early cancer detection and monitoring treatment response continuously. Thus, liquid biopsies are widely used in the clinical biomarker screening of tumors, such as endometrial cancer,122 lung cancer,123 pancreatic cancer,124 CRC,125 melanoma,126 renal cell carcinoma (RCC),127 breast cancer, ovarian cancer, cervical cancer, and bladder cancer.128 In addition, liquid biopsy technology is also utilized to detect and monitor KRAS, BRAF, and EGFR mutations in patients with lung cancer, CRC, and breast cancer.128

Liquid biopsy can reduce the risk of biopsy by noninvasive sampling,128 and it has the advantages of convenient sampling and easy operation. Moreover, liquid biopsies have the potential to better detect heterogeneity across regions of the tumor.115 Although there are still some challenges to overcome in terms of assay sensitivity and specificity, liquid biopsy technology provides new opportunities for personalized cancer treatment and has the potential to revolutionize the field of oncology.

Electron microscopy technology

Electron microscopy (EM), a powerful imaging technique used to visualize the ultrastructure of cells and tissues with high resolution, is applied based on a special type of microscope, electron microscope.129,130 The first electron microscope was built in 1931 by a German engineer and academic professor Ernst Ruska.131 The electron microscope uses signals obtained from the interaction between an electron beam and the sample to achieve information about sample structure, morphology, and composition.132

EM has been widely used in investigating tumorigenesis-related cellular and subcellular change133 and observing the ultrastructure changes in cancer cells,134 as well as in clinical applications for cancer diagnosis and treatment.134,135,136 The ultrastructural features of tumor cells by EM can provide vital clues such as evidence or biomarkers of cytodifferentiation for correct diagnosis, which is difficult for diagnosis of light microscopy.137,138 Especially, the ultrastructural examination provided by EM is necessary for the precise categorization of biomarkers in apparently undifferentiated carcinoma.138 Thus, EM is useful in the differential diagnosis of tumors, particularly in small-cell “undifferentiated” tumors, such as neuroblastoma, rhabdomyosarcoma, Ewing’s sarcoma, undifferentiated squamous cell carcinoma (SCC) of the lymphoepithelioma type, and malignant lymphoma, amelanotic melanoma, and spindle-cell carcinoma.137 Scanning electron microscopy has been used as an alternative to examine the morphology of exosomes which is a diagnostic biomarker usually detected by liquid biopsy.139 In conclusion, EM is a valuable complementary tool for tumor diagnosis, especially providing valuable information on tumor differentiation which is difficult to define by light microscopies.134,140

CRISPR/Cas9 technology

The CRISPR/Cas9 technology is a gene-editing tool that is based on the bacterial immune system. The basic principle of CRISPR/Cas9 is to use a guide RNA molecule to direct a nuclease, Cas9, to a specific target gene. The nuclease then cleaves the DNA at the target site, allowing for precise modifications of genome.141,142

By using CRISPR/Cas9 technology to precisely edit cancer-related genes, researchers have created highly specific molecular probes for the detection of cancer biomarkers in body fluids, such as blood, urine, and saliva. CRISPR/Cas9 system is extensively used for different kinds of cancer biomarkers including virus nucleic acids, ctDNAs (i.e., EGFR mutation), miRNAs (i.e., miR-17, miR-31), proteins (i.e., TGF-β1, CEA, PSA, AFP), and extracellular vesicles.90 CRISPR/Cas9 can combine with other assays for tumor biomarker identification, such as qPCR, FISH, and nanotechnology, providing an efficient way for tumor biomarker discovery. Moreover, CRISPR/Cas9 has exerted significant effects in the treatment of cancers, such as pancreatic cancer, prostate cancer, breast cancer, ovarian cancer, liver cancer, and CRC.143

The CRISPR/Cas9 system enjoys some advantages, including low cost, high efficiency, low application complexity, easy-to-operate, and time-saving.30 The exquisite specificity is also a character of the CRISPR/Cas9 system which could distinguish single base mismatch in target nucleic acid.90 Moreover, CRISPR/Cas9-based nucleic acid amplification strategies exhibit high detection sensitivity comparable with PCR. However, several aspects of CRISPR/Cas9-based diagnosis still need to be improved. CRISPR/Cas9-based analysis requires the fluorescence spectrophotometer and electrochemical workstation which is inconvenient for detection. Thus, the portable and quantitative detection strategy should be further explored to monitor cancer biomarkers. Cancer progression is influenced by the level of multiple biomarkers such as various miRNAs, ctDNAs, and proteins, which makes the design of the high-throughput CRISPR/Cas9-based strategy for cancer biomarkers detection promising and significant.90 In conclusion, CRISPR/Cas9 technology is a powerful gene-editing tool that holds great promise and opportunities for the development of personalized cancer management.144

Classification of tumor biomarkers

Tumor biomarkers are diverse and can be classified by different standards. Here, we divide tumor biomarkers by tissue origin: tumor biomarkers derived from blood, tumor tissues, and other biofluids such as feces, urine, and saliva (Fig. 3).

Fig. 3
figure 3

Overview of human tumor biomarkers

Tumor biomarkers derived from blood

Tumor biomarkers in the blood are highly significant for tumor diagnosis and treatment. They have vital reference values for early tumor diagnosis, tumor stage assessment, anticancer strategy selection, treatment response monitoring, and prognosis.2,145 Here, we summarize the common tumor markers in blood and their roles in cancers.

Embryonic antigen tumor biomarkers

The 1960s saw the discovery of AFP and CEA, two tumor biomarkers that are still widely employed as tumor biomarkers. AFP and CEA are embryonic antigen substances which are proteins that only appear in the fetal period and gradually decline and disappear in adulthood.146,147,148,149 The reemergence of these embryonic antigens in cancer patients may be related to the activation of certain genes that have been turned off in adulthood when malignant cells transform, and these genes make embryonic antigens. While there aren’t many embryonic antigen tumor biomarkers, the ones that exist are crucial biomarkers for cancer care in clinical practice.


AFP, first discovered in 1956 by Bergstrand Czar,146 is a 3–5% carbohydrate-containing single-chain glycoprotein.150 Encoded by the AFP gene located in the q arm of chromosome 4 (4q25), AFP is a member of the albuminoid gene superfamily.151 As the amino acid sequences of AFP and albumin are very similar and highly homologous, AFP is considered as an analog of serum albumin in the fetal period and is the main protein in fetal circulation. At 18 months after birth, albumin synthesis gradually increases, and AFP concentration gradually decreases. The concentration of AFP in healthy adult serum is less than 10 μg/L.147 AFP is currently the most widely used tumor biomarker for HCC and has been used for more than 60 years. Elevated AFP can be seen in ~80% of HCC patients.152,153 Thus, AFP is currently applied for HCC screening, especially in China, Japan, Africa, and Alaska. The international academic community recommends limiting the reference value of AFP to 20 μg/L. Moreover, early-stage HCC is frequently detected by AFP detection combined with ultrasound.154 Tumor prognosis and treatment monitoring are additional applications for AFP. In patients with HCC, a sharp increase in AFP indicates tumor recurrence or metastasis. AFP >200 μg/L after surgery indicates incomplete removal or metastasis of HCC.155 Nonetheless, AFP levels are not the perfect diagnostic criteria for HCC. Approximately 40% of patients with early-stage HCC express normal or acceptable AFP levels. The elevation of AFP levels is observed in patients with chronic liver diseases, including ~20% of patients with hepatitis and 40% of patients with cirrhosis.156


CEA was first extracted from human CRC tissues and embryonic tissues in 1965, hence it was named for CEA.148 CEA belongs to a family of glycoproteins on the cell surface, and its gene is located on chromosome 19q.157 The production of CEA in the digestive tract starts at the early fetal stage (week 9–13). In addition to normal adult tissues such as the colon, stomach, cervix, sweat glands, and prostate, CEA is highly expressed in various tumors.149

As a broad-spectrum tumor biomarker, CEA is elevated in 70% of CRC, 55% of pancreatic cancer, 50% of gastric cancer, 45% of lung cancer, 40% of breast cancer, 40% of urethral carcinoma, and 25% of ovarian cancer patients.149,158,159 Serum CEA levels are proportional to tumor burden. Accordingly, CEA is applied to aid the diagnosis, determine the prognosis, monitor recurrence, and evaluate the therapeutic efficacy in cancer patients.160 In patients with breast cancer, CEA is one of the most frequently used biomarkers in the diagnosis, prognosis, and prediction of survival for different breast cancer molecular subtypes.161 In CRC patients, CEA level is a meaningful prognostic and diagnostic biomarker. The levels of CEA predict the 5-year survival rate of patients: 69% of patients have a CEA level below 5 ng/mL, 44% have a level of 5–200 ng/mL, and only 7% have a level equal or greater than 200 ng/mL.162 The elevated CEA level also has a bearing on poor prognosis and progression of lung adenocarcinoma patients with mutant EGFRs, and gastric cancer patients with lymph node metastasis.163 Additionally, CEA is also used for efficacy evaluation and recurrence detection after tumor treatment.164

Nevertheless, CEA lacks good sensitivity and specificity, which renders CEA inappropriate for tumor screening. A combination of CEA with other biomarkers could improve its actual significance in clinical practice.149,164


Squamous cell carcinoma antigen (SCCA), a tumor-specific antigen, was first isolated from cervical SCC tissue by Kato and Torigoe in the 1970s.165 Initially, SCCA was used as a tumor biomarker for cervical cancer, and it has a high independent diagnostic value in cervical cancer.166 The serum level of SCCA correlates with the stage, the degree of invasion, recurrence, and the progression of cervical SCC.159 Cervical cancer patients with a high-level of pretreatment serum SCCA exhibit a higher risk for death than patients with low serum SCCA. Pretreatment SCCA cutoff ranging from 1.1 to 40.0 ng/mL is related to recurrence and death.166,167 Subsequent research has revealed that SCCA exists in tumors in the mouth, pharynx, esophagus, lung, and other tissues. In particular, high levels of SCCA have been found in multiple SCCs including lung cancer, esophageal cancer, and genitourinary system cancer in addition to cervical cancer, suggesting its essential role in the diagnosis and prognosis of the above cancers.168,169 Furthermore, elevated serum SCCA is associated with the therapeutic effect of postoperative chemotherapy in esophageal squamous cell carcinoma (ESCC),170 and with tumor-node-metastasis stage in head and neck squamous cell carcinoma (HNSCC).171 Peripheral SCCA has also been extensively utilized as one of the tumor biomarkers for monitoring NSCLC and predicting patients’ response to platinum combination chemotherapy, and serum SCCA level accurately reflects the survival status of patients.169 Despite its limited sensitivity in routine tests, SCCA is still a valuable diagnostic and prognostic biomarker in cancers.


Tissue polypeptide-specific antigen (TPS) is an M3 antigen determinant on the 18 fragments of cell keratin.172 TPS is synthesized in the S and G2 phases of the cell cycle, and the level of TPS in serum specifically indicates the proliferative activity of cells.173 The levels of TPS mostly depend on the number of cells in the proliferative phase instead of the total number of tumor cells, which is different from other tumor biomarkers.173 The serum levels of TPS are noticeably increased in multiple tumors, such as endometrial cancer, bladder cancer, NSCLC, skin cancer, carcinoma of male urethra, prostate cancer, pancreatic cancer, CRC, gastric cancer, esophageal cancer, neuroblastoma, and nephroblastoma.174,175,176,177 Thus, TPS has been employed as a serum tumor biomarker. Due to its lack of sensitivity and organ specificity, the prime application of TPS is monitoring treatment efficiency, and predicting tumor progression and recurrence, rather than diagnostic utility. In breast cancer patients, elevated serum levels of TPS could predict distant metastasis after treatment,178 and are recognized as an independent prognostic factor for disease-free survival (DFS) and OS of patients.179 In gastric cancer, TPS is applied in monitoring the palliative treatment response of patients with a 75% detection rate.180 The potential clinical role of TPS in RCC prognosis has also been demonstrated.181

Additionally, it is worth noting that TPS levels can alter in response to some pathological and physiological conditions, such as chronic pancreatitis, liver cirrhosis, ovulation, and menopausal status. Thereby, TPS in combination with other prognostic factors is necessary to improve the clinical use of serum TPS levels in predicting patient prognosis and facilitating the individualization of therapy for cancer patients.182 Further clinical studies are required to fully determine the utility of TPS alone or in combination.182,183 In conclusion, TPS has a unique value in the prediction of recurrence and metastasis, treatment monitoring, and prognostic evaluation in cancer patients.177


PSA, a serine protein kinase-releasing enzyme specifically secreted by the epithelial cells of prostate,184 is encoded by the prostate-specific gene kallikrein 3 which is a member of the tissue kallikrein family.185 PSA was first identified in the late 1970s.186 The elevated serum PSA levels represent prostate pathologies including prostatitis, benign prostatic hyperplasia, and prostate cancer.187,188 For the early diagnosis of prostate cancer, the positive cut-off value of serum PSA is greater than 10 ng/mL. In 1986, PSA was approved by the FDA as an adjunctive test for the detection of prostate cancer in men over the age of 50.185,189 Subsequently, in 1994, PSA was approved by the FDA as a diagnostic biomarker.189 Later on, PSA became popular in prostate cancer detection and patient management including screening, risk stratification for recurrence, surveillance following diagnosis, and monitoring therapy.186,188 Total PSA essentially consists of free PSA and bound PSA, and the higher percentage of the free PSA is connected to the lower the cancer risk. Studies have shown that a free PSA percentage >25% indicates the cancer risk is <10%, but a free PSA percentage <10% means the cancer risk is ~50%.187

However, PSA holds a poor specificity of 20–40% in prostate cancer diagnosis. Some noncancerous pathologies such as inflammation, trauma, or benign prostatic hyperplasia may also elevate the PSA level, which leads to a high rate of false positives. Besides, PSA is unable to differentiate between indolent and aggressive forms of prostate cancer, which may ignore aggressive prostate cancer with low initial serum PSA levels.187,190 All the aforementioned factors make prostate cancer now an “overdiagnosed” and “overtreated” cancer.185 To sum up, PSA level is a promising biomarker in prostate cancer diagnosis and prediction.


Neuron-specific enolase (NSE), a member of the enolase gene superfamily in glycolysis, was originally identified by Moore and McGregor in 1965 as an enzyme enriched in neurons and peripheral neuroendocrine cells.191,192 NSE consists of five dimeric isoenzymes with three different subunits, α, β, and γ, and is a sign of mature neural differentiation.193 Cell proliferation accelerates in response to oncogenic transformation in either central or peripheral neurons, accompanied by enhanced glycolysis and elevated NSE expression. Consequently, NSE plays pivotal roles in diagnosis, prognosis, and treatment efficacy evaluation in cancers originating from neural and neuroendocrine.194,195 Moreover, elevated NSE is also observed in SCLC which is with neuroendocrine properties. Serum NSE is currently believed to be a clinically potential biomarker for staging, monitoring treatment, and predicting relapse of SCLC.196,197 Interestingly, NSE also exerts a significant function in NSCLC. An analysis of 363 patients with advanced and metastatic NSCLC showed that patients with high NSE level (≥26.1 ng/mL) have significantly shorter progression-free survival (PFS) (5.69 vs 8.09 months) and OS than patients with low NSE level (11.41 vs 24.31 months).191 Besides, increased serum NSE levels are found in 30–69% of patients with NSCLC,198,199 which is in accordance with a study of 621 NSCLC patients which shows high NSE level (>12.5 ng/mL) is a prognosticate of poor outcome.200 Thus, serum NSE level is a predictive biomarker of cancer treatment response and an independent prognostic factor.191


α-l-Fucosidase (AFU), consisting of two isoforms, AFU1 and AFU2, which are encoded by FUCA1 and FUCA2 genes, respectively, is a lysosomal enzyme that clears the terminal α-l-fucose residues from glycoproteins.201 AFU is involved in the metabolism of glycoproteins, glycolipids, and oligosaccharides, and is widely distributed in human tissues and blood. The serum AFU level remains low under normal circumstances. While the serum AFU level increases rapidly as long as tumors attack the body, its level is closely related to the tumor stage and size.202 Multiple studies have shown that AFU is one of the most valuable biomarkers for HCC detection, with 85% sensitivity and 91% specificity.203,204 85% of patients with HCC can be diagnosed with AFU detection six months prior to the ultrasonography detection.205 Patients with a preoperative AFU >35 U/L have a lower recurrence-free survival (RFS) rate and OS rate than those with AFU ≤35 U/L, and they tend to form macrovascular invasion. Therefore, serum AFU is of great significance for judging the treatment effect, prognosis, and recurrence of HCC.205,206 Besides, the low AFU levels are significantly associated with longer OS in ESCC, which indicates that AFU is a potential prognostic biomarker for long-term survival in patients with early-stage ESCC.207 However, the serum levels of AFU are also mildly elevated in certain nonneoplastic conditions such as cirrhosis, chronic hepatitis, and gastrointestinal bleeding.203,208 Presently, the combination of AFU and AFP biomarkers is used in the diagnosis of HCC, which enhances the diagnostic specificity, and makes the diagnosis more stable and reliable for high-risk groups such as hepatitis and cirrhosis.155


LDH, an enzyme that catalyzes the reversible transfer of pyruvate to lactate and NADH to NAD + , consists of two different isoforms, lactate dehydrogenase A (LDHA) and LDHB.209,210 The two isoforms can form five homotetramers or heterotetramers with different functions.210 In the reverse reaction, LDHB is more effective at converting lactic acid back to pyruvate than LDHA is at converting pyruvate to lactic acid.211,212 Multiple factors, such as the oncogene c-Myc and hypoxia-inducible factor (HIF-1α), stimulate the transcription of LDHA,213,214 which results in the overexpression of LDHA in most tumor tissues.215

High expression of LDHA provides cancer cells with many benefits, and multiple studies have proved that high levels of serum LDH are associated with the proliferation of cancer-initiating cells, enhanced aggressiveness and metastasis, the poor prognosis of cancers, as well as radiation and chemotherapy resistance.216,217,218 The serum LDH level is considered to be a primary predictor of prognosis in patients with adverse prognosis and distant metastases in melanoma, RCC, and CRC.216 Accordingly, an analysis of 76 studies comprising 22,882 patients with solid tumors reveals that high serum LDH levels are linked to poor survival in patients with solid tumors, in particular in melanoma, prostate cancer, and RCC, and is a valuable and affordable prognostic biomarker in metastatic cancers.40 Serum LDH levels are closely correlated with OS in an analysis of 2507 cancer patients with brain metastasis216 and are a poor prognosticator for OS and DFS in nasopharyngeal carcinoma (NPC) patients. Furthermore, the elevated serum LDH levels could be used to develop individualized treatment strategies.219 A study of a total of 68 studies including 31,857 patients illustrates that LDH overexpression is a predictor to guide individual therapy in solid tumors,220 such as testicular cancer,221 SCLC,219 and gastrointestinal cancer.222,223 In conclusion, LDH is a valuable indicator of cancer diagnosis, efficacy evaluation, and recurrence and metastasis.


Carbohydrate antigen 72-4 (CA72-4) is a mucin carcinoid embryonic antigen found in liver metastases of breast cancer in 1981 and is highly expressed in human adenocarcinoma.224 Enhanced serum CA72-4 levels are effective indicators for the diagnosis of cancers, including gastric cancer, pancreatic cancer, CRC, breast cancer, ovarian cancer, lung cancer, cervical cancer, and endometrial cancer.225,226 Notably, CA72-4 exerts diagnostic value in patients with digestive system tumors, especially gastric cancer, with superior sensitivity and specificity.227 Studies have demonstrated that the sensitivity and specificity of CA72-4 applied in the diagnosis of gastric cancer alone are 49 and 96%, respectively, which outperforms other tumor biomarkers such as CEA (sensitivity 41%, specificity 93%), CA19-9 (sensitivity 44%, specificity 92%), and CA242 (sensitivity 38%, specificity 97%).228 The serum level of CA72-4 is also correlated with the malignant grade of gastric cancer. Thus, CA72-4 is used as the best serum marker for gastric cancer diagnosis in China.229 However, CA72-4 also has limitations. It has been uncovered that CA72-4 is highly expressed in normal tissues in addition to tumor tissues such as the endometrium and the colonic transitional mucosa, which results in false positives in patients with atrophic gastritis.230 The sensitivity of CA72-4 in the diagnosis of gastric cancer is far from satisfactory.231 The combination with other biomarkers may gain increased sensitivity and specificity of CA72-4 in tumor applications. In conclusion, serum CA72-4 is a unique biomarker of gastric cancer for screening, diagnosis, the prediction of metastasis and recurrence, and the evaluation of treatment efficiency.229


CA125, a highly glycosylated mucin, is originally discovered in a monoclonal antibody OC125 screening against the ovarian cancer cell line OVCA433.232,233 Thus, CA125 has become one of the most important biomarkers for monitoring epithelial ovarian cancer, and its sensitivity in the diagnosis of epithelial ovarian cancer reaches ~70%.234 The key role of CA125 in the prognosis of ovarian cancer patients has also been recognized. The Gynecologic Cancer Group (GCIG) has shown that the serum level of CA125 is associated with the progression and recurrence of ovarian cancer. According to the criteria of GCIG, patients with serum CA125 levels within the reference range (<35 U/mL) after surgery or chemotherapy are considered fully effective. While the CA125 level increased to twice of the minimum value (≥70 U/mL), the progression or recurrence is considered.235 Moreover, CA125 is also a diagnostic and prognostic biomarker for other nonovarian tumors, such as cervical cancer, endometrial carcinoma,236 and gastric cancer.237 Of note, ~1% of healthy people and 5% of patients with menstrual or benign diseases such as endometriosis and coronary artery disease have varying degrees of elevated serum CA125 levels.238,239,240


Carbohydrate antigen 242 (CA242), a sugar chain antigen containing sialic acid, is obtained after the immunization of mice with a human CRC cell line COLO 205.241 An analysis of serum CA242 levels from 34,680 patients with 27 clinically defined diseases suggests that patients with pancreatic cancer, cervical cancer, and lymphoma have the highest level of serum CA242, which are followed by esophagus cancer, CRC, ovarian cancer, and breast cancer.242 Hence, the primary application of CA242 is as a biomarker for CRC and pancreatic cancer.243 Serum CA242 has a normal reference value of less than 17 U/mL. The sensitivity for diagnosing pancreatic cancer and CRC is ~70 and 45%, respectively, and the specificity is ~95 and 83%, respectively.244,245,246 As CA242 exhibits a lower sensitivity for diagnosing pancreatic cancer, the combination of CA242 with CEA is a promising strategy for improving diagnosis sensitivity in pancreatic cancer.247 In addition, CA242 is also used as a clinical indicator of progression or recurrence during chemotherapy for pancreatic cancer.241,242


Carbohydrate antigen 15-3 (CA15-3, also known as mucin 1) is a large transmembrane glycoprotein derived from the MUC1 gene.248 It is expressed in normal tissues including the breast, esophagus, stomach, duodenum, pancreas, uterus, prostate, and lung.248,249 Notably, CA15-3 is overexpressed in the majority of human cancers, and is thought to be a key biomarker for cancers, especially for indicating cancer metastasis.250 The reference value for normal serum CA15-3 levels is less than 28 U/mL.243 In breast cancer, serum CA15-3 is used as an auxiliary diagnostic index with a diagnosis sensitivity of 61.5–70% which is higher than that of CEA.243 Thus, CA15-3 in combination with CEA is the most popular method for breast cancer diagnosis.251 Meanwhile, CA15-3 is also a crucial indicator for the evaluation of postoperative recovery, recurrence, and metastasis of breast cancer.248 It is noteworthy that serum CA15-3 is also elevated to varying degrees in benign diseases of the breast, liver, gastrointestinal tract, lung, and other organs, but the positive rate is low.250


Similar to CA15-3, carbohydrate antigen 27-29 (CA27-29) is a critical epitope for the MUC1 protein.243 With a sensitivity of 84% for breast cancer detection, CA27-29 is primarily utilized in breast cancer patients for diagnosis, and efficacy evaluation.243 Additionally, it also be used in combination with other markers to increase the specificity of breast cancer diagnosis.252 The elevated CA27-29 is also observed in other cancers including CRC, stomach cancer, pancreatic cancer, ovarian cancer, and benign diseases of the breast and liver.253


Carbohydrate antigen 50 (CA50) was initially identified as a cancer-specific antigen screened from monoclonal antibodies against CRC cell line COLO 205 in 1983.254 It is generally absent in normal tissues, but elevated in multifarious cancers. Patients suffering from pancreatic cancer, lung cancer, and colon cancer exhibit the highest levels of serum CA50. Serum CA50 is quite effective in the diagnosis of pancreatic cancer, with a sensitivity of more than 84%.255 Meanwhile, patients suffering from gastric cancer and rectum cancer reveal comparable serum CA50 levels.256,257 Similar to other carbohydrate antigens, serum CA50 is also increased in patients with non-neoplasm diseases such as chronic pancreatitis, colitis, cholecystitis, and pneumonia.256


CA19-9 is initially found in human CRC cell line SW1116 and belongs to the mucin glycoprotein antigen.258 It is extensively distributed on the cell membrane of Lewis antigen-positive epithelial cells such as the pancreatic duct, gallbladder, and gastrointestinal tract. CA19-9 is currently the most commonly used and gold-standard biomarker for pancreatic cancer,259,260 and holds a median sensitivity of 79% for diagnosis of pancreatic cancer.261,262 In addition, CA19-9 has also been used as a biomarker for other cancers, particularly digestive tract cancers.263 Other diseases such as liver damage, bile duct obstruction and inflammation, pancreatitis, acute diarrhea, stomach ulcer, and pulmonary fibrosis have also been linked to increased CA19-9 levels.264,265,266 Notably, CA 19-9 is not expressed in cells from patients with Lewis allele deficiencies, and it is necessary to ascertain the patient’s Lewis gene type information when applying CA19-9 as a diagnostic biomarker.267,268


Human epididymal protein 4 (HE4), an orotic acid protein, is first identified in distal epididymal epithelial cells.269 HE4 is widely expressed in the trachea, salivary gland, lung tissue, etc., and is highly expressed in ovarian cancer, endometrial cancer, and lung cancer. Meanwhile, age and menopausal status are also momentous factors affecting HE4 levels.270,271 At present, serum HE4 is primarily used for the diagnosis and recurrence monitoring of ovarian cancer with a sensitivity of 67%. HE4 is also used to evaluate the treatment effect of ovarian cancer.270,272 In addition, HE4 is also overexpressed in other non-gynecologic malignancies, including NSCLC, pancreatic cancer, and transitional cell carcinoma.273


Ferritin is the leading protein that is essential for iron storage and detoxification.274 Ferritin is present in numerous normal tissues such as liver, spleen, bone marrow, and body fluids.274 Serum ferritin levels are linked to a broad range of conditions. The low serum ferritin concentration indicates iron deficiency, e.g., anemia and diarrhea,275 and the high serum ferritin concentration indicates iron overload, e.g., hemochromatosis and hemolytic anemia, or infection and liver disease.276 Moreover, ferritin is overexpressed in various cancers, such as HCC, lung cancer, lymphoma, melanoma, and CRC.277,278 As indicated by its potential to promote tumor proliferation, angiogenesis, immunosuppression, and tumor drug resistance,276 ferritin is valuable in evaluating the progression and prognosis of cancer patients. Nevertheless, a number of factors influence ferritin levels, and ferritin’s limited specificity for tumor detection means that it is not an ideal diagnostic marker for cancers.279


Prostate-specific antigen precursor (p2PSA) is a precursor that is first secreted in the prostate gland ducts during the production of PSA.280 p2PSA is a relatively stable pro-PSA and has certain clinical value in the diagnosis of early prostate cancer. The prostate health index, which forecasts the diagnosis of prostate cancer, is calculated by PSA and p2PSA. Currently, the prostate health index is the strongest predictor of diagnosis at initial biopsy when total PSA levels are between 2.0 and 10 ng/mL in prostate cancer patients, and the prostate health index has been approved by the FDA for early diagnosis and risk grading of prostate cancer.280,281


HCG is a polypeptide hormone composed of two noncovalently linked subunits (α and β). The smaller α subunit is the part of follicle-stimulating hormone and luteinizing hormone, while the larger β subunit is unique to HCG.282,283 Serum levels of HCG in non-pregnant and menopause women maintain at a low level of 5–10 U/L, and increase dramatically during pregnancy.284 Increased serum HCG levels are observed in trophoblastic tumors, ovarian cancer, testicular cancer, breast cancer, lung cancer, HCC, CRC, and kidney cancer. Although HCG level could be employed for monitoring the disease progression, it is too low to be regarded as a diagnostic marker.282,285


CAM17.1 is a mucin with high specificity for the digestive system, such as the pancreas, colon, small intestine, and biliary tract.286 Several studies revealed that CAM17.1 is particularly overexpressed in pancreatic cancer with a serum cut-off value is 39 U/L.287 CAM17.1 has a sensitivity of 86% for the diagnosis of pancreatic cancer, and a higher sensitivity of 89% in patients without jaundice.286 These findings suggested that CAM17.1 is a potential biomarker for pancreatic cancer diagnosis, which triggers the need for further study.


Protein induced by vitamin K absence or antagonist-II (PIVKA-II) is an abnormal prothrombin elevated in the conditions of vitamin K reduction or the presence of vitamin K antagonists.288 PIVKA-II is primarily used for the early detection of HCC, with a sensitivity and specificity of 97.5 and 90%, respectively.289,290 In other tumors such as gastric cancer and pancreatic cancer, PIVKA-II is also elevated at varying degrees.291 In addition to being able to differentiate between other non-malignant conditions such cirrhosis or chronic hepatitis, serum PIVKA-II is more accurate than AFP in the diagnosis of early-stage HCC.292,293 It is noteworthy that certain patients with vitamin K deficiency also exhibit elevated PIVKA-II levels.288


Gastrin-releasing peptide (GRP), first isolated from gastric nerve fibers by McDonald in 1978, is a gastrointestinal hormone that exits in the normal bronchial epithelial cells, pulmonary fibroblast, central nervous system cells, and neuroendocrine cells.294 Significantly, GRP is overexpressed in multiple cancers, including 62% of CRC patients, 59% of pancreatic cancer patients, 60% of prostate cancer patients, 39% of breast cancer patients, and 74% of SCLC patients.294 Since GRP has a short half-life and is unstable, it is more appropriate to detect its precursor, pro-GRP.294 With a sensitivity of 47 to 86%, serum pro-GRP detection is mainly utilized for the diagnosis, efficacy, and prognosis analysis of SCLC, outperforming NSE.294,295 The combined application of pro-GRP and NSE increases the sensitivity of SCLC detection.296 In addition, pro-GRP is also elevated in a few other diseases, such as gastritis and acute hepatitis, but the positive rate is generally low.297

Tumor biomarkers derived from tumor tissues

Since the six hallmarks of cancer were proposed in 2000, tumor characteristics are considered to be a set of functional capabilities acquired by human cells during the transition from a normal to a tumor growth state.298 To date, tumors have possessed fourteen major characteristics, including sustaining proliferative signaling, evading growth suppressors, enabling replicative immortality, inducing angiogenesis, resisting cell death, activating invasion and metastasis, genome instability, and mutation, tumor-promoting inflammation, deregulating cellular metabolism, avoiding immune destruction, unlocking phenotypic plasticity, nonmutational epigenetic reprogramming, and polymorphic microbiomes, and senescent cells.298,299,300 Herein, we summarize the tumor biomarkers from tumor tissues divided by cancer hallmarks (Fig. 4).

Fig. 4
figure 4

The 14 cancer hallmarks-based biomarkers. Fourteen major characteristics of tumor cells have been proven so far, which have been divided into acquired hallmarks including sustaining proliferative signaling, evading growth suppressors, resisting cell death, enabling replicative immortality, inducing angiogenesis, activating invasion and metastasis, enabling hallmarks including genome instability and mutation, tumor-promoting inflammation, nonmutational epigenetic reprogramming, and polymorphic microbiomes, and emerging hallmarks including deregulating cellular metabolism, avoiding immune destruction, unlocking phenotypic plasticity, and senescent cells. Each of the cancer hallmarks is involved in numerous essential biomarkers that play vital roles in tumor progression

Sustaining proliferative signaling

Cancer cells are capable of multiple approaches to acquire the ability to sustain proliferation. Stimulated by growth factors and other proliferative signals, proliferation-related signaling pathways, such as the RAS, the phosphoinositide 3-kinase (PI3K)-protein kinase B (AKT)-mammalian target of rapamycin (mTOR) pathway, and the RAF-mitogen-activated protein kinase (MAPK) kinase (MEK)-extracellular signal-related kinase (ERK) pathway, are activated in tumor cells, which subsequently regulate tumor cell proliferation, migration and invasion, gene transcription, cellular metabolic reprogramming, and tumor microenvironment (TME) remodeling.301,302,303


RAS genes, named after the rat sarcoma,304 were identified as the transformative factor in the Harvey and Kirsten strains of rat sarcoma viruses305 and were identified in the human genome in 1982.304,305 RAS proteins belong to a superfamily of GTPases, and three RAS genes (HRAS, NRAS, and KRAS) encode four highly homologous RAS proteins: HRAS, NRAS, KRAS4A, and KRAS4B, with the latter two KRAS isoforms arising from alternative splicing.306,307

RAS proteins couple cell surface receptors to intracellular effector pathways through binding to GTP or GDP, followed by a cycle between the GDP-bound inactive state (RAS-GDP) and the GTP-bound active state (RAS-GTP). Under physiological conditions, RAS proteins retain an inactive state, and are incapable of interacting with downstream effectors. When activated by upstream receptors, RAS is activated by guanine nucleotide exchange factors (GEFs) which promote GDP to GTP exchange, thereby recruiting diverse downstream effectors such as the RAF-MEK-ERK pathway and the PI3K-AKT-mTOR pathway.301,308 RAS activation has been linked to multiple tumor phenotypes, including cell cycle progression, proliferation, metastasis, and apoptosis resistance.301 Furthermore, RAS is involved in diverse metabolic processes such as aerobic glycolysis, glutaminolysis, redox homeostasis, and lipid metabolism in tumor cells to support tumor growth.309 Importantly, RAS activation remodels the TME,301 including the initiation and maintenance of proangiogenesis,310 the production of proinflammatory factors,311 and immune escape.301

RAS mutation is a prominent factor that plays a vital role in tumorigenesis and progression.312,313 Approximately 21% of all malignancies have RAS mutations,308 which include CRC,314 pancreatic ductal adenocarcinoma (PDAC),315 lung adenocarcinoma,316 and melanoma.317

Although the function of RAS in the physiological or pathological states has been thoroughly elucidated in the past decades, numerous unresolved concerns still need to be investigated. For instance, the regulatory relationship between RAS and downstream effectors other than PI3K and MAPK.318 To sum up, RAS is a crucial biomarker for tumor diagnosis, prognosis, and treatment.


KRAS is by far the most frequently amplified and mutated RAS isoform among the three RAS genes, accounting for 85% of all RAS mutations.319 KRAS mutations were first identified in 1984 in patients with squamous cell lung cancer.320 Notably, KRAS mutations are present in 88% of pancreatic cancer, 50% of CRC, and 32% of lung cancer.319,321 The most common mutations in KRAS are G12D, G12V, G12C, G13D, and Q61R, which account for 70% of RAS mutations in cancer patients.321 KRASG12C mutation is the most frequent,321 and the G12C mutation alters KRAS conformation and shape by forming binding pockets, leading to increased affinity for GTP and sustained activation of KRAS, ultimately triggering the transduction of downstream oncogenic signaling.319,321

KRAS mutations have emerged as biomarkers for the prognosis, diagnosis, and treatment of some tumors, including PDAC,322 CRC,323 and lung cancer.324 A study in a pooled analysis has found that KRAS mutations are independently associated with shorter time to recurrence, survival after relapse, and OS in patients with microsatellite-stable resected stage III CRC.313 Patients with the KRASG12C mutation are related to inferior PFS and OS compared with patients with non-mutated tumors, according to a prognosis analysis in 1239 patients with metastatic CRC.323 Moreover, KRAS mutations link to the poor prognosis of patients with PDAC, and KRAS mutation assays provide significant predictive information on tumor progression and recurrence, which are of great value in the diagnosis, prognosis, and treatment of PDAC.325 Consistently, PDAC patients with KRASG12D mutation have shorter survival than all other PDAC patients.326 In lung adenocarcinoma patients, the patients with KRASG12C mutation have worse DFS than patients with nonG12C mutation KRAS or wild-type KRAS.324

Mechanistically, KRAS drives tumor development and progression through various signaling pathways. For example, the extensive metabolic reprogramming induced by KRAS mutations, such as glycolysis, glutamine metabolism, lipid metabolism, and nucleotide biosynthesis to facilitate tumorigenesis, has attracted much attention in recent years.327,328 KRAS-mutant cells exhibit the upregulation of glucose transporters329 and metabolic enzymes involved in the glycolysis,330,331 resulting in increased glucose flux in the glycolytic pathway.329 KRASG12D stimulates hexosamine biosynthesis and the pentose phosphate pathway to regulate glucose metabolism in PDAC.332 KRAS-mutant cells produce nicotinamide adenine dinucleotide phosphate (NADPH) by promoting glutamine catabolism,329 and intracellular fatty acid uptake and oxidation.333 Furthermore, KRAS leads to the transcriptional upregulation of MYC and the nonoxidative pentose phosphate pathway gene RPIA through activating MAPK, thereby enhancing nucleotide biosynthesis in PDAC cells.327

In summary, KRAS mutations are among the most prevalent drivers of tumorigenesis, and their activation is correlated with tumor progression and poor prognosis.334,335 The evidence presented above strongly suggests that KRAS is a crucial tumor biomarker.


The PI3K-AKT-mTOR pathway plays valuable roles in various cellular processes, such as cell proliferation, angiogenesis, protein translation, and metabolic reprogramming.302

In normal cells, growth factor-stimulated PI3K activation leads to the conversion of phosphatidylinositol-3,4-bisphosphate (PIP2) to phosphatidylinositol-3,4,5-trisphosphate (PIP3), followed by the recruitment of AKT and 3-phosphoinositide-dependent kinase 1 to the plasma membrane. Following that, 3-phosphoinositide-dependent kinase 1 phosphorylates and activates AKT, thus phosphorylating the downstream mTOR complex, contributing to cell survival and proliferation.302,336,337 The atypical serine/ threonine kinase mTOR consists of rapamycin-sensitive mTOR complex 1 (mTORC1) and rapamycin-insensitive mTORC2.338 AKT drives mTORC1 activation either directly by phosphorylating mTORC1 at Ser2448 or indirectly by inhibiting TSC1/TSC2.302 mTOR supports cell growth and proliferation by promoting cell cycle,302 sensing nutrient signaling339,340 by phosphorylating its downstream effectors such as S6K and 4EBP1.302,341 The tumor suppressor PTEN is a critical negative regulator of the PI3K signaling pathway.302 PETN rapidly metabolizes PIP3 by removing the 3’-phosphate of PIP3, which in turn terminates PI3K signaling.342

In cancer cells, the PI3K-AKT-mTOR signaling pathway is abnormally activated via the stimulation of tyrosine kinase growth factor receptors,343 the loss of PTEN functions, and the mutations of PIK3CA, thereby promoting tumorigenesis in a wide variety of human cancers.302,342 The PI3K-AKT-mTOR pathway exerts significant impacts on multiple cancers including lung cancer,344,345 ovarian cancer,302 breast cancer,346 and NPC.347,348 The PI3K-AKT-mTOR has been proven to be crucial in ovarian tumorigenesis and drug resistance.302 The level of pAKT is a diagnostic biomarker for the treatment of SCLC involving the combination of clinically approved inhibitors against isoform-specific PI3K and mTOR.345 In addition, the class I isoform of PI3K, the most well-known PI3K protein, contains four distinct isoforms of catalytic structural domain: p110α (PIK3CA), p110β (PIK3CB), p110γ (PIK3CG), and p110δ (PIK3CD).343 pIK3CA and PTEN aberrations lead to the activation of the PI3K-AKT-mTOR pathway.349,350 The TCGA database has shown that PIK3CA gene mutations occur in a variety of cancers, including 53% of endometrial cancer, 35% of breast cancer, 23% of cervical cancer, 21% of gastric cancer, 20% of head and neck cancer, 20% of CRC, 15% of lung cancer, and 10% of glioblastoma.343 PIK3CA mutation, PTEN loss, and pAKT activation are predictive biomarkers for the efficacy of tumor treatment.350,351 Moreover, PIK3CA mutations act as diagnostic biomarkers for HR+ and HER2- metastatic breast cancer.352 In summary, the PI3K-AKT-mTOR pathway is an essential biomarker pathway for tumor diagnosis, prognosis, and treatment.


The RAS-RAF-MEK-ERK pathway participates in the regulation of key processes such as cell proliferation, differentiation, migration, and apoptosis,303 which can be activated by growth factors, cytokines, integrins, and chemokine receptors.303,353 Active RAS binds to RAF kinase, which results in RAF dimerization and activation.354,355 RAF proteins possess three isoforms including BRAF, CRAF, and ARAF which share conserved regions of the regulatory domain and kinase domain,356 and among them, ARAF exhibits the lowest kinase activities.357 The active RAF dimer phosphorylates and activates MEK1/2 which subsequently phosphorylates and activates ERK1/2, followed by the phosphorylation and activation of downstream effectors and the proliferation of cells.336 Various proteins, such as Hsp90,358 p50CDC37,359 and KSR,360 are engaged in the regulation of RAF activation.

Abnormalities epically mutations in RAF-MEK-ERK signaling lead to the aberrations of cell proliferation.361 A mutation analysis of more than 3000 samples from 12 tumor types has shown that the mutations of RAF-MERK-ERK signaling occur in ~50% of cancers.362 In particular, BRAF mutations are widely investigated in cancers.362 Studies have revealed that the hyperactivity of the BRAF-MEK-ERK pathway is correlated with worse survival in patients with ER-negative or progesterone receptor-negative breast cancers,363 suggesting that the alterations of the RAS-RAF-MEK-ERK pathway could serve as predictive and prognostic biomarkers for breast cancer.303 Meanwhile, the aberrations of the RAS-RAF-MEK-ERK pathway can be predictive biomarkers of drug sensitivity in cancer therapies.303 In conclusion, the RAF-MEK-ERK signaling cascade functions as a significant biomarker in tumor progression.


PTEN was discovered in 1997 as a tumor suppressor,364 and it was the first phosphatase proven to have tumor suppressive effects.365,366 As a phosphoinositide 3-phosphatase, PTEN negatively regulates the PI3K-AKT-mTOR pathway by converting PIP3 to PIP2, thereby hindering the proliferation and survival of tumor cells.365,367 Furthermore, PTEN exerts both enzymatic and nonenzymatic effects in cellular epithelial-mesenchymal transition (EMT), migration and invasion, glucose and lipid metabolism, cell cycle, DNA repair, genomic stability, and gene transcription.365,368

PTEN function and expression are frequently altered in a variety of cancers.369 Accordingly, PTEN acts as a prognostic and predictive biomarker in various cancers including prostate cancer, RCC, PDAC, CRC, breast cancer, endometrial cancer, brain cancers, skin cancers, and hematological malignancies.370 Aberration of PTEN is associated with the mutations, downregulation or deletion of the PTEN gene, and the abnormal subcellular localization of PTEN protein.371,372 PTEN deletion modulates the downstream effector of mTORC1 by regulating 4EBP1 and p70S6 kinase to increase protein synthesis.372 Significantly, PTEN deletion is strongly linked to a shorter OS and DFS of cancer patients.370 Taken together, PTEN is a significant biomarker for tumor prognosis. The mechanism studies of PTEN activation will be beneficial for the development of antitumor strategies based on the recovery of PTEN function.

Evading growth suppressors

In addition to inducing and maintaining growth stimulus signals, tumor cells also eschew powerful programs to evade growth restriction and blockade, which mainly rely on the action of tumor suppressor genes.


Rb, the first tumor suppressor gene to be identified, was originally discovered in retinoblastoma children.373,374 The alteration of the Rb gene or inactivation of the Rb protein is one of the most common events in cancers.375 Rb primarily restricts the transcription of cell cycle genes by regulating E2F transcription factors.376 Rb proteins are phosphorylated by cyclin-dependent kinases (CDKs),377 which lead to Rb functional inactivation, followed by E2F transcriptional activation and cell cycle progression.378 Inactivation of Rb causes abnormalities in cell division, defects in cell cycle withdrawal, impaired induction of cellular senescence, and impaired cell cycle checkpoint control.379 The function of Rb in tumor cells is disrupted in various ways including Rb gene mutation, Rb protein hydrolysis, Rb-E2F interaction elimination, and the overactivation of CDK.375 Consequently, Rb dysregulation acts as a prognostic biomarker in cancers.380


TP53, often referred to as the “guardian of the genome”, is a gene encoding the p53 tumor suppressor protein.381 Numerous studies have shown that p53 plays an integral role in biological processes such as cell cycle arrest, aging, DNA repair, and apoptosis.382,383,384

In human tumors, TP53 is the most commonly mutated gene with ~50% of tumors carrying the mutations or deletion of TP53.385,386 In addition to mutations or deletion, tumors may lose the function of p53 due to other mechanisms. For example, overexpression of viral oncoproteins or MDM2 leads to the degradation of p53 protein.387 The expression and function loss or gain function of TP53 are associated with poor prognosis, immune escape, and anticancer drug resistance. Thus, TP53 can serve as an effective predictive biomarker to evaluate prognosis and monitor therapeutic responses in various cancers.388 An analysis of over 29,000 cases from the International Agency for Research on Cancer database revealed that TP53 mutations are potential prognostic biomarkers, and can be used to bolster the predictive accuracy of the OS and DFS of cancer patients.389

Enabling replicative immortality

Unlimited proliferation is a critical characteristic of tumor cells.299 Normal cells undergo senescence due to their recurrent division cycle, whereas tumor cells are capable of unlimited replication, a phenomenon known as immortalization. The protection of chromosome ends by telomeres is crucial for tumor immortalization.299


Telomere is a repetitive DNA–protein complex located at the end of chromosome.390 Telomeres in normal cells gradually shorten with continuous cell division and eventually fail to protect the end of chromosomal DNA, thus triggering DNA damage, cellular senescence, and apoptosis. Therefore, the length of telomeres is closely related to the cellular lifespan.299,391

Telomerase is a DNA polymerase that maintains telomere length by adding telomeric repeat fragments to telomeric DNA ends, thus compensating for the attrition of chromosomal ends in continuous cell division.390,392 Telomerase is encoded by the human telomerase reverse transcriptase (hTERT) gene which is the catalytic subunit of telomerase holoenzyme.390,393 hTERT is silenced in almost all somatic cells and is significantly re-expressed in ~90% of human cancers by various approaches.390 Thus, the large majority of normal human somatic cells lack the telomerase-maintenance mechanism, while a tremendous proportion of cancer cells have a highly active telomerase-maintenance mechanism.392 The activation of the telomerase-maintenance mechanism is observed in numerous human cancers, such as breast cancer, CRC, kidney cancer, cervical cancer, liver cancer, lung cancer, pancreatic cancer, prostate cancer, thyroid cancer, and bladder cancer,394 which ensure the replicative immortality of cancer cells. In clinical practice, cancer patients with high hTERT levels are along with worse survival than those with low hTERT levels. Moreover, cancer patients with high hTERT levels have a higher risk of disease recurrence and death.395 Therefore, telomerase is an independent prognostic biomarker of OS in cancer patients.395 Besides, TERT promoter mutations increase the expression of telomerase directly, which contributes to tumorigenesis and is associated with poor OS of cancer patients, suggesting that TERT promoter mutations are prognostic biomarkers for cancers.396 Moreover, the nonenzymatic functions of telomeres promote cancer cell proliferation and the resistance of apoptosis,299 regulate chromatin structure,397 impair DNA damage repair,397 and increase antioxidant protein expression,393 although the detailed mechanism remains to be elucidated.299

Inducing angiogenesis

In tumor initiation and progression, the new vascular system can transport nutrients and oxygen, and excrete metabolic waste, which is critical for tumor growth.299,398 The transition from prevascular hyperplasia to highly vascularized and progressively outgrowing tumors is known as the “angiogenic switch”. In the early stage of tumor development, the angiogenic switch is highly activated, which in turn sustains the continuous generation of new blood vessels, and causes the transition from dormant hyperplasia to outgrowing vascularized tumor, ultimately promoting rapid proliferation of cancer cells.299,398,399 The angiogenic switch, which favors a proangiogenic outcome during tumor angiogenesis, is controlled by the balance between proangiogenic and antiangiogenic factors secreted by tumor cells or TME cells.398 Studies have ascertained that angiogenesis significantly contributes to the development of various cancers, including CRC, breast cancer, bladder cancer, RCC, and NSCLC.400,401 A large number of angiogenic factors such as vascular endothelial growth factors (VEGFs) have been found to induce the proliferation and differentiation of endothelial cells directly or indirectly.


VEGF, originally known as vascular permeability factor, was discovered as a tumor secretory factor in 1983 by Senger et al.402 In 1989, Ferrara isolated VEGF and renamed it vascular endothelial growth factor.403 VEGFs are heparin-binding homodimeric glycoproteins whose family includes VEGF-A (commonly referred to as VEGF), VEGF-B, VEGF-C, VEGF-D, VEGF-E, VEGF-F, and placental growth factor (PlGF).401,404 VEGF has been demonstrated to be a potent inducer of angiogenesis,399 and is widely expressed in normal adult organs along with two other related endothelial growth factors VEGF-B and VEGF-C, suggesting their necessary roles in tissue angiogenesis and homeostasis.399 VEGFs in tumor tissues are extracted not only from tumor cells but also from host cells,405 and high levels of VEGFs are found in diverse tumor cells. Interestingly, tumor cells are able to produce VEGFs, but are unable to respond to them due to the absence of VEGF receptors (VEGFRs) on the cell surface.406 Multiple factors such as genetic and epigenetic regulation influence the VEGF levels in tumor cells. Among them, epigenetic factors include hypoxia-inducible transcription factors 1α and 2α, low pH, inflammatory cytokines, growth factors, androgens, estrogens, and chemokines.406,407 Genetic factors include the activation of oncogenes such as RAS, EGFR, HER2, and the deletion and mutational inactivation of oncogenes such as p53, PTEN, and VHL.406,408

VEGFs bind respectively to the three tyrosine kinase receptors (RTKs) VEGFR1–3 with different specificities and affinities and VEGFR2 mediates the main VEGFR signals.409 VEGFR-1, the first RTK to be identified as a VEGF receptor,410 binds VEGF with a binding affinity ten times higher than that of VEGFR-2, although its ability of signal transduction is quite weak.406 VEGFR-1 serves as a decoy receptor to chelate or trap VEGF under some circumstances, thus negatively regulating VEGF activity by preventing VEGF binding to VEGFR-2.411,412,413,414 The specific mechanism of VEGFR-1 in VEGF-mediated angiogenesis needs to be explored in further detail.406,415 In contrast, VEGFR-2 is expressed in almost all endothelial cells, and exerts function through the activation of VEGF, VEGF-B, C, or D.416 The binding of VEGF to VEGFR-2 causes receptor dimerization and subsequently activates the intracellular signaling cascades, such as the PI3K-AKT and the RAF-MEK-ERK pathways, which generates neovascular branches required for tumor growth, and ultimately promotes rapid tumor cell proliferation and migration.406 VEGFR-3 has similar functions to VEGFR-2, but its action site is mainly in the lymphatic blood vessels.401,415 VEGFR-3 is expressed in the lymphatic endothelial cells415 and mainly binds to VEGF-C and VEGF-D to induce lymphangiogenesis.417,418 In addition, VEGFs interact with the neuropilin receptor family.415

VEGF levels are associated with the aggressiveness of tumors.419 The plasma VEGF levels in various cancer patients are elevated and negatively correlated with tumor prognosis.420 Moreover, VEGF levels are used to predict the efficacy of oral tyrosine kinase inhibitors (TKIs) in cancer patients.401 For example, VEGFR inhibitor sorafenib has displayed better therapeutic efficacy against advanced clear cell renal cell carcinoma (ccRCC) patients with high levels of VEGF.401 In summary, circulating VEGF and VEGFR-2 have been used as crucial biomarkers for the prediction of prognosis and antiangiogenic drug efficacy.405,421


Fibroblast growth factor (FGF) is a secreted glycoprotein422 that engages in the regulation of organogenesis,423 angiogenesis, and wound repair.422,424 FGF binds to the transmembrane FGF receptor (FGFR) on the surface of target cells with high affinity.422,425 The mammalian FGFR family consists of four highly conserved transmembrane RTK FGFR1-4, and FGFR5 which has no intracellular tyrosine kinase structural domain but has FGF binding capacity.422 FGFR is widely expressed in a broad range of cells, especially endothelial cells.426

In tumors, FGF is essential for vascular endothelial integrity, angiogenesis, tumor proliferation, survival, and metastasis.425,426 Notably, abnormal FGF signaling accelerates tumor proliferation by promoting tumor angiogenesis.422 For example, the elevated level of FGF2 in prostate cancer induces neovascularization to boost tumor growth.427 The increased angiogenesis induced by FGF1 amplification in high-grade serous ovarian cancer leads to reduced OS in patients, suggesting that FGF1 is a prognostic biomarker for ovarian cancer.428

Furthermore, FGFR is strongly associated with the development of various tumors,422,429 such as prostate cancer,430,431 lung cancer,432 breast cancer,433 and pancreatic cancer.434 In particular, studies have revealed that FGFR with mutations or amplification functions as driving oncogene to aberrantly activate downstream pathways, resulting in mitogenic, mesenchymal, and antiapoptotic responses in cells.422 Somatic mutations of FGFR3 have been observed in more than 30% of bladder cancers.435,436 The somatic mutations of FGFR2 have been discovered in 12% of endometrial cancers, and mutant endometrial cancer cell lines are highly sensitive to FGFR TKIs.437 Besides, FGFR amplification is also tightly linked to the progression of numerous cancers.438,439 Approximately 10% of gastric cancers have shown FGFR2 amplification, which is associated with the poor prognosis of gastric cancer patients.440 Amplification of FGFR1 occurs in approximately 10% of breast cancers, especially ER+ type.441,442 In brief, FGF and FGFR are vital biomarkers of tumor prognosis and treatment.


Platelet-derived growth factors (PDGFs), an α-granule component secreted in an autocrine manner during platelet activation,443 are critical proangiogenic factors for tumor angiogenesis.443,444 The PDGF family contains four different monomeric polypeptide chains: PDGF-A, -B, -C, and -D, which form four homodimers through disulfide bonds (PDGF-AA, -BB, -CC, -DD) and a PDGF-AB heterodimer.443,445 The PDGF receptor (PDGFR) consists of RTKs PDGFRα and PDGFRβ. PDGF isoforms trigger different receptor dimerization and phosphorylation by binding to the corresponding PDGFRs, thus activating multiple downstream growth signaling pathways, such as PI3K, MAPK, and JAK/STAT pathways, to promote cancer cell proliferation, migration and invasion, angiogenesis, and drug resistance.443,446,447

PDGFs and their receptors are extensively expressed in a number of cancers, such as oral squamous cell carcinoma (OSCC),448 skin SCC,449 soft tissue sarcomas,450 ccRCC,451 dermatofibrosarcoma protuberans, gastrointestinal stromal tumors (GIST),452 CRC,453 breast cancer,447 pancreatic cancer,454 gastric cancer,455 neuroendocrine tumors,456 NSCLC, ovarian cancer, and HCC.443 High PDGF-A levels correlated independently and inversely with the risk of metastatic relapse in cancer patients.450 The level of PDGF-D is associated with advanced tumor stages and the development of bone metastasis.457,458 High expression of PDGFR-β is independently linked to prostate cancer recurrence.445 In conclusion, PDGFs and PDGFRs are meaningful diagnostic biomarkers.

Resisting cell death

Resisting cell death is a significant tumor hallmark that contributes to tumor progression and therapeutic resistance.299 Apoptosis that leads to programmed cell death hinders tumorigenesis, and the apoptotic program is considerably reduced in highly aggressive and therapy-resistant tumor cells.299 Increasing autophagy activation might inhibit tumorigenesis in parallel with or in concert with apoptosis.459,460 Moreover, necrosis also significantly contributes to tumor cell death.461 The identification of biomarkers in these processes is useful for tumor diagnosis or prognosis.


Sydney Brenner, Robert Horvitz, and John Sulston shared the 2002 Nobel Prize in Physiology or Medicine for their contributions to the discovery of apoptosis procedure.462 Cellular stress, DNA damage, and immune surveillance systems frequently cause apoptosis, a type of cell death that is initiated by the proteolytic cleavage of numerous proteins and the regulation of caspase protease activity.463 Apoptosis can be triggered through the intrinsic or mitochondrial pathway and the extrinsic pathway.464 The intrinsic pathway is controlled by the B-cell leukemia or lymphoma gene number 2 (BCL-2) family. BCL-2 induces mitochondrial outer membrane permeabilization and the release of multiple proapoptotic factors, followed by the release of cytochrome c from mitochondria to the cytoplasm. Subsequently, the apoptotic peptidase activating factor 1 interacts with cytochrome c, and form the apoptosome that induces the activation of the initiator caspase pro-caspase 9. Later on, the caspase 9 binds to the apoptosome and is cleaved and activated, which subsequently stimulates the activation of initiator caspase 3.465,466 This process in which cytochrome c is released from the mitochondria is negatively regulated by antiapoptotic BCL-2 family members such as BCL-2, B-cell lymphoma-extra large (BCL-XL), BCL-W, BCL-2-A1, and MCL1.463 The membrane permeabilization and the release of cytochrome c into cytoplasm are key processes for triggering apoptosis.463,467

The extrinsic apoptotic pathway is initiated through the proapoptotic death receptors which include Fas, the tumor necrosis factor receptor (TNFR) family such as TNFR1, TNFR2, and theTRAIL receptors DR4 and DR5. The proapoptotic death receptors bind to ligands and then trimerize and aggregate within the cell membrane, subsequently recruiting adapter proteins such as FADD, caspase 8 and/or caspase 10 to form the death-inducing signaling complex, which activates the initiator caspase 8, which in turn induces the activation of the effector caspases such as caspase 3, 6, and 7, and apoptosis.463,467 Consequently, the potential strategy for cancer therapy is targeting the proapoptotic and antiapoptotic proteins to induce apoptosis.468


The BCL-2 family proteins have four conserved BCL-2 homology (BH) structural domains (BH1, 2, 3, and 4) which can be divided into three subfamilies based on the homology and function of proteins: the antiapoptotic BCL-2 family members (such as BCL-2 and BCL-XL), the multi-BH-domain proapoptotic members, such as the BCL-2-associated X protein (BAX) and the BCL-2 antagonist/killer (BAK), and the proapoptotic “BH3-only” proteins, such as the BCL-2 interacting mediator of cell death (BIM), and PUMA.469

BCL-2 was the first identified apoptosis regulator, which was activated by chromosome translocation in human follicular lymphoma oncoprotein.470 The BCL-X gene was cloned in 1993,471 and the BCL-XL protein, which is localized in the mitochondrion, is the first protein whose three-dimensional structure has been identified in the BCL-2 protein family.472,473 The “BH3-only” proteins can be divided into activators and sensitizers.474,475 Activators of BH3 proteins, such as BIM, BID, initiate apoptosis by directly inducing BAX and BAK oligomerization and cytochrome c release. However, sensitizer BH3 proteins, such as BAD, and BIK, exert proapoptotic functions by binding to antiapoptotic BCL-2 family members, rather than directly activating BAX or BAK.475,476,477 The interaction of one protein’s BH3 α-helix with a sizable hydrophobic pocket on binding partners regulates the activity of BH3-only proteins,475 which initiates apoptosis by activating proapoptotic proteins or by inhibiting antiapoptotic proteins.478

BCL-2 can drive oncogenic transformation, hinder apoptosis, and increase tumor cell survival.479,480 The high expression of BCL-XL is involved in tumor cell invasion, the maintenance of tumor stem cell phenotype, angiogenesis, and metastasis through inducing apoptosis resistance.480 The overexpression of BCL-2 and/or BCL-XL may contribute to tumor progression and the resistance of chemotherapeutic agents in various tumors,467,475 including pancreatic cancer,481 ovarian cancer,481 lung cancer,481 prostate cancer,481 breast cancer,482 neuroblastoma,483 CRC,484 gastric cancer,485 HCC,469 chronic lymphocytic leukemia (CLL),469 lymphoma,481 and multiple myeloma.481 Furthermore, BCL-XL can be used as an independent biomarker for the prognosis prediction of CRC patients.484 BCL-2 is a prognostic biomarker in TNBC patients. Lower BCL-2 expression level is associated with better outcomes of TNBC patients treated with both adjuvant and neoadjuvant chemotherapy.486 In summary, BCL-2 and BCL-XL are essential biomarkers in tumor prognosis and treatment.


BAX, a cytosolic membrane protein that works as a critical regulator of the apoptotic process, was identified by immunoprecipitation and yeast two-hybrid screening.480,487 BAX protein has BH1, BH2, and BH3 structural domains,480,488 which is highly homologous with BCL-2.480 BAX stimulates apoptosis either by inhibiting BCL-2 and BCL-XL or by directly triggering the apoptotic pathway.480 BAX moves from cytoplasm to mitochondria during apoptosis, followed by oligomerization and the formation of pores in the outer mitochondrial membrane, thus facilitating the release of cytochrome c which activates the downstream effector caspases and leads to cell death.469,489,490 Downregulation and mutations of BAX are essential for apoptosis resistance,491 and BAX acts as a potential prognostic and predictive biomarker in various cancers including gastric cancer, esophageal cancer, and CRC.492 The somatic frameshift mutations of the BAX gene highly occur in CRC with the microsatellite mutator phenotype.493 BAX mutations are found in ~21% of human hematopoietic malignancies such as ALL.494 Reduced BAX expression is a major factor in cisplatin resistance of ovarian cancer cells,495 5-FU resistance of CRC cells,496 and zoledronate resistance of lung cancer cells.497 The decreased BAX/BCL-2 ratio can be induced by BAX abnormalities, which affects the temozolomide-induced resistance in U87MG cells and paclitaxel-resistant breast cancer cells. Thus, the activation of BAX could be used to promote apoptotic cell death and overcome resistance.498

Furthermore, the high BAK expression is correlated with improved OS and PFS in patients with advanced gastric cancer. BAK is a predictive and prognostic biomarker for the therapeutic effect of docetaxel in patients with advanced gastric cancer.499 BAX-BAK heterodimer is also used as a pharmacodynamic biomarker of on-target drug action of MCL1 inhibitors.500


In 1955, Christian de Duve discovered the lysosome,501 a key organelle for intracellular degradation, and subsequently introduced the term “autophagy” at the CIBA Foundation Symposium on Lysosomes in 1963.502 In 2016, Yoshinori Ohsumi was awarded the 2016 Nobel Prize for Medicine or Physiology for elucidating the mechanism of autophagy, which led to increasing attention to autophagy in health and disease.460,503 To date, there are three main types of autophagy: macroautophagy, microautophagy, and chaperone-mediated autophagy. The general term “autophagy” usually means macroautophagy.504

Autophagy is a multistep, highly conserved degradation process: the initiation and nucleation of the autophagosome, the expansion, and elongation of the autophagosome membrane, the closure and fusion with the lysosome, and the degradation of products.460,505 Briefly, autophagy is triggered by a variety of factors, including nutrient or growth factor deprivation, energy status, hypoxia, ROS, and other stress inducers.506 Subsequently, a flat membrane named the phagophore or isolation membrane sequesters cytoplasmic constituents. The elongating phagophore results in complete sequestration and the formation of a double-membraned organelle autophagosome. Then, the autophagosome fuses with the lysosome, and the inner membrane of the autophagosome and the cytoplasm-derived materials it contains are subsequently degraded by the lysosome, resulting in the production of amino acids and lipids which are exported to the cytoplasm for recycling.504,507,508

Mechanistically, autophagy-associated (ATG) proteins, a group of evolutionarily conserved proteins, are responsible for this process.509 Autophagy begins with the activation of unc-51 like autophagy activating kinase 1 (ULK1) (also known as ATG1) complex which includes ULK1, ULK2, ATG13, FIP200, and ATG101. The ULK1 complex subsequently activates class III PI3K complex which includes VPS15, VPS34, ATG14, Beclin1, UVRAG, and AMBRA1, which mediates vesicle nucleation.460 Then, the ATG5-ATG12 complex binds to ATG16 to extend the autophagosomal membrane, and members of the LC3 and GABARAP protein families conjugate with lipid phosphatidylethanolamine (PE) and recruit PE to the membrane. ATG4B binds to ATG7 and then couples with LC3-I and PE to form LC3-II. Eventually, autophagosomes fuse with lysosomes to degrade macromolecules and reuse them. The adapter protein sequestosome-1 (also known as p62) targets autophagosome-specific substrates and LC3-II which are simultaneously degraded.460,510

Autophagy is a double-edged sword in cancer. The enhanced autophagic flow in tumor cells accelerates tumor cell growth, while the induction of autophagy can prevent the development of cancer.459,460 Therefore, autophagy inhibition and promotion are both promising strategies for cancer therapy, and their application depends on the actual situation.511

Beclin 1

Beclin 1 was identified as a BCL-2 interaction factor in the yeast two-hybrid screen in 1988.512,513 Human Beclin 1, the mammalian orthologue of yeast Atg6, consists of a BCL-2-homology 3 structural domain,514 a flexible helical domain,515 a coiled coil domain,516 and an evolutionarily conserved domain.514 Moreover, Beclin 1 contains a leucine-rich nuclear export signal that is essential for its autophagic and tumor suppressor functions.517 Beclin 1 is phosphorylated by ULK1 and acts as an integral component of the PI3K complex to localize autophagy proteins to the phagosome. Furthermore, Beclin 1 interacts with and is inhibited by BCL-2/BCL-XL in the BH3 structural domain, which blocks the formation of the Beclin 1-VPS34 complex and inhibits Beclin 1 interacting with UVRAG, thereby inhibiting autophagy. The bind of AMBRA1 to Beclin 1 stabilizes the Beclin 1-VPS34 complex, thus promoting autophagosome formation.518,519

Studies have found that Beclin 1 is a prognosis biomarker for various cancers. Reduced expression of Beclin 1 has been observed in brain tumors and cervical cell carcinomas.520 The absence of BECN1 has been found in 40 to 75% of sporadic breast cancer and ovarian cancer,521 and 40% of prostate cancer.522 Low expression of Beclin 1 is associated with the malignant phenotype and poor prognosis of gastric cancer.523 Beclin 1 inhibits the proliferation of human breast cancer cells MCF7 in vitro and in vivo through regulating autophagy.524 On the contrary, the elevated Beclin 1 expression is related to distant metastasis and poor prognosis in CRC patients, and reduced survival in CRC patients with 5-FU treatment.525 Taken together, Beclin 1 may serve as a valid prognostic indicator and therapeutic target for cancers although further research is needed to determine its specific mechanism in different cancers.526,527


The microtubule-associated protein 1 light chain 3B (LC3B or MAP1LC3B) is a classical autophagy marker, is cleaved by protease at the C-terminus to form free LC3B-I, and LC3B-I binds to PE to form membrane LC3B-II in autophagy occurrence. The process in which LC3B-I converts to LC3B-II is essential for phagophore expansion and the formation of autophagosomes.528,529 Thus, LC3B is a marker for the detection of multiple autophagic fluxes.530 Accordingly, LC3B-II is one of the most commonly used biomarkers to detect the number of autophagosomes and autophagosome-related structures.518

The high LC3B expression is closely associated with the aggressive progression, and poor prognosis of multiple tumors, including gastric cancer,531 CRC,532 TNBC,533 melanoma,534 astrocytoma,535 esophageal cancer,536 and OSCC.537 Studies have found that LC3B has the highest expression in TNBC cells in different molecular subtypes of breast cancer,538 and its high expression is related to the progression and poor prognosis of TNBC patients.533 Moreover, LC3B is closely connected with the vascular invasion and lymph node metastasis of HCC, and is a potential therapeutic target for HCC.539 Collectively, LC3B is a meaningful prognostic biomarker in cancer management.527,534


ULK1, a conserved Ser/Thr kinase, plays a pivotal role in autophagy induction.540 High expression of ULK-1 is associated with poor prognosis in various tumors, including esophageal SCC,541 HCC,542 NPC,543 prostate cancer,544 and CRC.545 Studies have found that HCC patients with ULK1 and LC3B overexpression have larger tumors and a higher frequency of lymph node metastasis. The combination of ULK1 and LC3B is an independent predictor of OS and PFS in HCC patients.546 After androgen deprivation therapy, prostate cancer patients with high levels of ULK1 have shorter PFS and OS.544 In addition, elevated expression of ULK1 has been connected to lymph node metastasis547 and functions as a prognostic biomarker in patients with CRC.545 Interestingly, low expression of ULK1 is associated with operable breast cancer progression and is a poor prognostic biomarker for patient survival.548 In human NPC, ULK1 is also a promising biomarker for the prediction of poor prognosis and treatment response.543 Furthermore, ULK2 has been found to be expressed at higher levels in prostate cancer tissues compared with that in normal tissues.549 To better determine the prognostic value of ULK1 and ULK2 in different cancer types, comprehensive studies in prospective cohorts are necessary.530


p62 (also known as sequestosome-1, SQSTM 1) was originally identified as an atypical protein kinase C (aPKC) interacting protein.550 p62 consists of several structural domains, including the N-terminal PB1 domain, the ZZ-type zinc finger (ZZ) domain, the tumor necrosis factor receptor-associated factor 6 (TRAF6) binding (TB) domain, the LIR domain, the Kelchlike ECH-associated protein 1 (Keap1)-interacting region (KIR), and the C-terminal UBA domain.551,552 Each structural domain of p62 has a different function. The PB1 domain is essential for the formation of homodimeric aggregates that regulate autophagic degradation. Moreover, p62 can interact with other proteins containing the PB1 domain, such as MAPK.551 ZZ structural domain is involved in the activation of NF-kB signaling pathway,553 and the TB structural domain can interact with TRAF6 which induces protein polyubiquitination.554 The LIR structural domain affects autophagosome formation and autophagic degradation by mediating LC3-p62 interactions,551,555 and the KIR structural domain activates Nrf2 by binding with Keap1.556,557 UBA structural domain is involved in autophagic lysosomal degradation558 and apoptosis signaling pathways.551

As a marker for autophagic flow detection, p62 accumulation usually represents the inhibition of autophagy.460,559 Upregulation or reduced degradation of p62 is associated with tumor progression and anticancer drug resistance.552 p62 expression is increased in 60% of lung adenocarcinomas and 90% of lung SCCs.550 Numerous studies have shown that high p62 expression is correlated with the aggressiveness and poor prognosis of cancers, including endometrial cancer,560 OSCC,537 epithelial ovarian cancer,561 and NSCLC.562 In addition, elevated p62 expression is also correlated with the high-grade, distant metastasis and reduced 5-year survival of breast cancer patients,563 especially in patients with TNBC cancer.564 In short, p62 is a meaningful prognostic biomarker and a potential target for cancer therapy.551,552


Necrosis is derived from the Greek “nekros” for corpse.461 Necroptosis is a programmed necrotic cell death type in a caspase-independent f manner, induced by TNFR superfamily and mediated by receptor-interacting protein kinase 1 (RIPK1, also known as RIP1), RIPK3 (also known as RIP3), and mixed lineage kinase domain-like (MLKL).565,566 Necrosis is caused by numerous stimuli such as cytokines, viral infection, pathogen-associated molecular, T-cell receptors, interferon receptors, Toll-like receptors, cellular metabolism, genotoxic stress, and various anticancer compounds.565,567 Common morphological features of necrotic cells include moderate chromatin condensation, cytoplasmic organelle swelling, and the rupture of plasma membrane.568 The biochemical characters of necrotic cells include a drop in ATP level, the activation of RIP1, RIP3, and MLKL, the release of damage-associated molecular pattern molecules (e.g., HMGB1), the hyperactivation of poly(ADP-ribose) polymerase 1 (PARP1).568 The basic feature that distinguishes necrosis from apoptosis is the rapid loss of cell membrane potential. Cellular energy depletion, membrane lipid damage, and the impairment of steady-state ion pump function lead to loss of membrane potential which in turn leads to cytoplasmic swelling, plasma membrane rupture, and cell lysis, thus promoting necrotic cell death.461

Studies have identified that necrosis is an essential predictor for prognosis and treatment response in various tumors, including pancreatic cancer,569 RCC,570 breast cancer, lung cancer, CRC,571 and soft tissue sarcoma.571,572 Tumor necrosis is closely associated with cancer-specific survival, OS, RFS, and PFS in patients with RCC, and it can be a prognostic biomarker of patients in clinical practice.570 Therefore, the discovery of biomarkers that identify necrosis and molecular mechanisms of necrosis enables the development of necrosis-based antitumor therapies.569


The serine/threonine kinase RIPK1 is a key regulator of necrosis, and RIPK3 is a downstream regulator of RIPK1.573,574 The RIPK1-RIPK3-MLKL complex, also known as the “necrosome”, mediates upstream cell death receptors and downstream signaling.565 Necrosome is a multiprotein complex that contributes to TNF-induced cell death.575,576 Necrotic cells trigger caspase 8 inactivation and activate RIPK1 and RIPK3, followed by autophosphorylation and cross-phosphorylation between RIPK1 and RIPK3 to form necrosome. Then, MLKL is phosphorylated, followed by oligomerizing and translocating to the plasma membrane and stimulating the necroptosis.577,578

The RIPK3 expression is significantly reduced in AML patients,579,580 which is consistent with the high methylation level near the transcriptional start site of RIPK3.581 RIPK3 deficiency promotes leukemogenesis by enhancing the accumulation of leukemia-initiating cells, and hinders myeloid differentiation through reducing cell death and IL-1β production.579,