Biology and significance of alpha‐fetoprotein in hepatocellular carcinoma

Hepatocellular carcinoma (HCC) is one of the most common causes of cancer‐related deaths globally due, in part, to the majority of patients being diagnosed with intermediate or advanced stage disease. Our increased understanding of the heterogeneous molecular pathogenesis of HCC has led to significant developments in novel targeted therapies. Despite these advances, there remains a high unmet need for new treatment options. HCC is a complex disease with multiple pathogenic mechanisms caused by a variety of risk factors, making it difficult to characterize with a single biomarker. In fact, numerous biomarkers have been studied in HCC, but alpha‐fetoprotein (AFP) remains the most widely used and accepted serum marker since its discovery over 60 years ago. This review summarizes the most relevant studies associated with the regulation of AFP at the gene and protein levels; the pathophysiology of AFP as a pro‐proliferative protein; and the correlation of AFP with molecular HCC subclasses, the vascular endothelial growth factor pathway and angiogenesis. Also described are the historical and current uses of AFP for screening and surveillance, diagnosis, its utility as a prognostic and predictive biomarker and its role as a tumour antigen in HCC. Taken together, these data demonstrate the relevance of AFP for patients with HCC and identify several remaining questions that will benefit from future research.


| INTRODUC TI ON
Hepatocellular carcinoma (HCC), the leading primary malignancy of the liver, is one of the most common cancers globally and results in significant health-related problems, making it the third most frequent cause of cancer-related deaths. 1 their role in this disease remains uncertain. Despite these advances, the current prognosis for advanced disease remains dismal with median OS ranging from 7.3 to 13.6 months, median progression-free survival (PFS) from 3.1 to 7.4 months and objective response rates from 2% to 24%. 6 Current research efforts are aimed at identifying specific HCC patients who will benefit from new therapies, and several prognostic or predictive biomarkers are being used or studied to improve outcomes. The most commonly used biomarker for HCC is serum alpha-fetoprotein (AFP). 9 Historically, AFP has been used for screening and diagnosing HCC, predicting prognosis and monitoring response to treatment. However, its use in some settings has been controversial, particularly with respect to surveillance and diagnosis.

| B I OMARK ER S IN H CC
Although the search for other diagnostic, prognostic or predictive biomarkers for HCC has been extensive, AFP remains the most commonly used biomarker in HCC. For patients who are at risk of developing HCC, additional biomarkers could detect the cancer at an earlier, potentially curative stage. For those with a diagnosis of HCC, novel biomarkers could identify biochemical or clinical factors indicative of clinical outcomes and/or measure of disease burden ('prognostic'). Likewise, they could also be correlated with tumour response and clinical outcomes to specific therapies ('predictive').
There are several factors that determine the effectiveness of a biomarker, including those attributable to the type of tumour, its prevalence in the investigated population and the availability of an effective therapeutic regimen.
A growing body of literature describes potential predictive and prognostic biomarkers that may inform diagnosis and treatment of HCC. For early detection and/or diagnosis of HCC, oncofetal antigens, proteoglycans, enzymes and isoenzymes, such as the fucosylated fraction of AFP (AFP-L3), des-gamma carboxyprothrombin (DCP; also known as prothrombin induced by vitamin K absence or antagonism II), versican and glypican 3, have been evaluated. [10][11][12] Clinical features related to tumour stage and treatment have also been characterized, including Barcelona Clinic Liver Cancer (BCLC) the historical and current uses of AFP for screening and surveillance, diagnosis, its utility as a prognostic and predictive biomarker and its role as a tumour antigen in HCC. Taken together, these data demonstrate the relevance of AFP for patients with HCC and identify several remaining questions that will benefit from future research.

K E Y W O R D S
alpha-fetoprotein, biomarkers, hepatocellular carcinoma

Key points
• Alpha-fetoprotein (AFP) is the most widely accepted serum biomarker used in the management of hepatocellular carcinoma (HCC).
• The regulation and pathophysiology of AFP informs the basis for its clinical relevance in the context of HCC.
• Several first-line and second-line trials show the prognostic effect of AFP, and now ramucirumab studies with enriched patient populations have demonstrated its utility as a predictive marker for antiangiogenic treatment.
• Growing evidence suggests pretransplant AFP is a useful prognostic marker for selecting liver transplant candidates and assessing risk of recurrence.
• Future clinical studies will strengthen our understanding of this important biomarker for patients with HCC. stage and macroscopic vascular invasion. Several retrospective studies evaluating laboratory and clinical findings during treatment have also identified potential predictive markers of response or resistance to therapy. During treatment with sorafenib, hand-foot skin reaction, hypertension or diarrhoea have been associated with clinical benefits. 13 Likewise, low neutrophil to lymphocyte ratio, aetiology, extrahepatic spread, high s-c-KIT, low baseline hepatocyte growth factor concentration, FGF3/FGF4 amplification and rare vascular endothelial growth factor-A (VEGF-A) amplification have also been associated with improved response to sorafenib. [14][15][16][17] Conversely, elevated expression of VEGF-A, K19, CD133, epithelial cell adhesion molecule (EpCAM) or serglycin as well as single nucleotide polymorphisms and angiopoietin-2 levels in patients receiving sorafenib have been associated with poor prognosis. 12,18,19 An exploratory analysis of the RESORCE trial suggests that decreased expression of lectin-like oxidized LDL receptor 1, Ang1, cystatin-B, latency-associated peptide TGF-β1 or macrophage inflammatory protein 1α may be predictive of the OS and time to progression treatment benefit observed from regorafenib, but these findings require prospective studies to confirm. 20 In a large, phase 3, randomized, double-blind, placebo-controlled study, tivantinib did not improve OS in patients with high MET expression on tumour cells and advanced HCC that had progressed after sorafenib-based therapy, despite the anticipated predictive value of MET identified in a phase 2 study. 21 In the large, phase 3, randomized trial REFLECT, lenvatinib demonstrated noninferiority to sorafenib in OS and improved response rates in patients with unresectable HCC in the first-line setting. 22 Final biomarker analysis of the REFLECT trial suggested that higher baseline levels of VEGF, ANG2 and FGF21 were associated with worse OS in both treatment arms. 23 However, longer OS for lenvatinib vs sorafenib was observed in patients with high baseline levels of FGF21, suggesting FGF21 could be predictive for reduced OS with sorafenib compared to lenvatinib. 23 Despite the identification of candidate biomarkers, AFP is still the most widely accepted and used serum marker in HCC. 10 The aim of this review is to understand the significance and clinical relevance of AFP as a biomarker and tumour antigen in HCC.

| HIS TORY AND PHYS I OLOGY OF ALPHA-FE TOPROTEIN
Alpha-fetoprotein is a 70 kD glycoprotein that is produced by the fetal liver and yolk sac during the first trimester of pregnancy. In 1956, it was identified from a protein fraction detected in human fetal serum that was not detected in adult serum (Figure 1). 24 The isolated protein was subsequently termed AFP, and acts as the fetal equivalent of serum albumin. In normal physiology, AFP declines rapidly after birth and remains at low levels over the entire lifespan ( Figure 2). A study conducted in the late 1980s concluded that the absolute size of the fetus as well as gestational age might play a significant role in determining maternal and fetal AFP concentrations, and that there is a significant correlation between maternal, cord arterial and venous AFP. 25 Likewise, low levels of maternal serum AFP during the second trimester were subsequently shown to be associated with a very low risk of preterm birth, pre-eclampsia and placental complications, and vice versa. 26 Owing to the variability in absolute AFP concentrations in healthy newborns, the kinetics of AFP declining during the neonatal period is commonly followed in clinical practice. 27 The AFP gene is one of the four members of the albumin gene family localized in a tandem arrangement to form a multigene cluster, 28 and there are three major isoforms defined by their affinity for the lectin Lens culinaris agglutinin (AFP-L1, AFP-L2 and AFP-L3) that F I G U R E 1 Fetal and maternal serum proteins fractions. Electrophoretic separation of serum proteins stained with Ponceau red from fetal (top) and maternal (bottom) blood samples. Fetal samples present an additional fraction between the albumin and α1 protein bands. Based on this position, the fraction was termed alpha-fetoprotein. From Bergstrand and Czar. 24 Copyright © Medisinsk Fysiologisk Forenings Forlag (MFFF), reprinted by permission of Taylor & Francis Ltd, http://www.tandf online.com, on behalf of MFFF F I G U R E 2 Relative levels of alpha-fetoprotein in the fetal liver, a healthy adult, or in a patient with hepatocellular carcinoma are found in varying amounts in different physiological or pathological conditions. 4 AFP expression is primarily regulated at the transcriptional level (Figure 3), and the gene has an upstream regulatory region that consists of a tissue-specific promoter, three independent enhancers and two silencer regions, which may be involved in the decreased AFP gene expression in adult livers. 29 Following fetal liver development, the AFP enhancers are normally blocked from the gene promoter and instead act to maintain albumin gene transcription into adulthood. 30 Alpha-fetoprotein was recognized as the first oncofetal biomarker after it was identified in the serum of patients with HCC and undifferentiated teratoma. The test for AFP was the first serologic assay used for the detection and clinical follow-up of patients with HCC. 31

| Mechanisms of AFP overexpression
Although expressed in 60% to 80% of HCC, genetic regulation of AFP is complex and has not been fully characterized. 33 Based largely on preclinical studies, AFP expression appears to be suppressed at the promoter and at two enhancers by corepressors and methylated histones in adult cells. 34 Repression is mediated, in part, by zinc-fingers and homeoboxes 2 (ZHX2) and BTB domain containing 20 (ZBTB20). 35,36 In HCC, hypermethylation and resultant silencing of ZHX2 was found to be a potential mechanism of AFP overexpression, and overexpression of ZHX2 inhibited the AFP synthesis and secretion. 35 Further, deregulation of pathways impacting ZBTB20 expression in the liver, which may be through the microRNA122 pathway, was also shown to contribute to AFP overexpression and HCC tumour aggressiveness. 37 Specifically, microRNA122 regulated the levels of ZBTB20 via a complex pathway involving the Cut homeobox 1 (CUX1), a protein that regulates cell motility and invasion. 37 Despite some understanding of the regulation of AFP expression based on preclinical work, the limited clinical observations in patients with HCC have been less clear. A small study in Peruvian patients with HCC demonstrated a contrasting relationship between the two repressors in which ZBTB20 was downregulated whereas ZHX2 was enhanced. However, both repressors appeared to contribute to AFP overexpression. 38 To add complexity, other studies have identified additional effects of cytokine signalling, such as TGF-β, on AFP expression, 39 which is only further complicated by the well-known fact that TGF-β also has distinct roles in tumour inhibition vs tumour progression depending on the stage of disease. 40 Despite progress to date, additional research is clearly needed to better characterize the regulation of AFP in HCC.

| Role as pro-proliferative protein
AFP may regulate the growth of neoplastic and normal cells by several mechanisms that include apoptotic regulation and cytoplasmic signalling modulation. Although increasing evidence suggests that AFP may regulate the growth of tumour cells, the specific mechanism for its growth-promoting activity is unclear. In HL-60 cells and HepG2 cells, AFP was shown to protect against apoptosis induced by various factors. 41,42 Since tumour proliferation by AFP was found to be dependent on the cyclic AMP-protein kinase A pathway and the initiation of Ca2+ influx, a possible explanation could be that increases in intracellular Ca2+ from AFP-induced Ca2+ influx results in increased DNA synthesis and tumour proliferation mediated F I G U R E 3 Concordant dysregulated expression of alpha-fetoprotein and other genes in the fetal liver, a healthy adult, or in a patient with hepatocellular carcinoma by intracellular cAMP and protease A activity. 43 In vitro, AFP also affects the function of caspase-3 via indirect interaction with the Xlinked inhibitor of the apoptosis protein, XIAP, and the cellular inhibitor of apoptosis protein, cIAP-2. 44 As a result, AFP may play a critical role in the inhibition of the apoptotic signal transduction mediated by caspase-3. 45 Several studies also suggest that tumour growth results from AFP-mediated suppression of the antitumour immune response. 46 AFP was shown to interact with macrophages and to decrease their phagocytic activity and Ia antigen expression. In addition, AFP inhibits the activity of natural killer cells, reduces proliferation of T-lymphocytes and promotes the activity of T-suppressor cells (reviewed in Terentiev et al 31 ).
Detailed mechanisms regarding the upregulation of cell proliferation and tumour growth by AFP remain unknown. However, studies have shown that AFP binds to specific receptors located on the surface of normal and tumour cells, and the presence of AFP and its receptor in human placenta suggests a possible receptor-mediated mechanism for placental transport of AFP between the fetal and maternal circulations. 31 AFP may influence the delivery of fatty acids to proliferating cells that require increased energy supply and intermediate products of β-oxidation of fatty acids. In reaction to lymphocytic blast transformation, AFP and the heptapeptide AFP [14][15][16][17][18][19][20] might also cause a moderate stimulation of proliferation of lymphocytes and inhibition of proliferation of PHA-activated lymphocytes at concentrations of 10 7 to 10 9 M. 31 Since the heptapeptide AFP [14][15][16][17][18][19][20] inhibits proliferation of lymphocytes in a dose-dependent manner from patients with acute and chronic lymphocytic leukaemia with low sensitivity to antitumour agents, AFP may be a biologically active ligand on certain human cells. Taken together, these data suggest that AFP may have dual regulatory effects on cell proliferation and tissue growth through both stimulatory and inhibitory effects. 41,[47][48][49]

| Correlation with molecular HCC classes
Hepatocellular carcinoma is a heterogeneous tumour on the macroscopic, histopathological and molecular level. Molecular aberrations have been identified that allow the subclassification of HCC by molecular and clinical characteristics. 50 In addition, different transcriptomic subclasses have been described and linked to histological subtypes. 51 Other predominant determinants included genetic and epigenetic alterations such as chromosome instability, CTNNB1 and TP53 mutations, and parental imprinting. 51 A meta-analysis of gene expression profiles from 603 HCC patient samples identified three robust HCC subclasses, S1-S3, associated with various parameters that included tumour size, degree of cellular differentiation and serum AFP levels. 52 The G1, G2 and G3 subclasses, which are characterized by high proliferation and chromosomal instability and the S2 subclass, which is linked to large tumours, have been shown to be associated with high AFP serum levels (G1-G3: AFP > 100 ng/mL; P < .001, and S2: median AFP = 171 ng/mL; P < .001). 52,53 Furthermore, AFP and EpCAM expression have been used to divide HCC patients into prognostic molecular subclasses. 3,54,55 These results suggest AFP may be useful for detecting a molecular subclass of HCC. However, translating this knowledge into clinical practice will require further research.

| AFP, VEGF and angiogenesis
Hepatocellular carcinomas are highly vascular, and their growth is dependent on angiogenesis. 56 The therapeutic effect of transarterial (chemo-) embolization and agents with antiangiogenic properties such as sorafenib support the vascular nature of these tumours. 57 High serum and tissue VEGF levels are associated with poor disease-free and OS in HCC. 58 In addition, VEGF has been reported as a prognostic biomarker in advanced HCC. 17 AFP and EpCAM, a hepatic stem cell expression marker, have also been implicated in the prognosis of HCC patients, 55 and HCC patients with high AFP serum levels (>300 ng/mL) and positive staining for EpCAM have been shown to have significantly higher VEGF tissue expression and microvessel density. 59 AFP has also been specifically evaluated as a predictor of efficacy to antiangiogenic therapy in HCC. There is a paucity of literature demonstrating a mechanistic link between AFP and angiogenesis. However, emerging data suggest crosstalk of AFP and VEGF signalling cascades. Genetic mechanisms do exist that may concordantly dysregulate the expression of AFP and several other known and unknown genes involved in angiogenesis during HCC progression. 59 Likewise, silencing of AFP has been shown to inhibit VEGF production in HCC cells in vitro. 60 Taken together, these findings are consistent with AFP expression being associated with potentially more angiogenic tumours and, as described above, may denote particular subclasses of HCC.

| AFP for defining patients at risk of HCC development: screening and surveillance
Early detection of HCC may improve outcomes, and persistently elevated AFP levels have been identified as a risk factor for development of HCC and could potentially help define at-risk populations. 61,62 The use of AFP alone to screen the general population for HCC has proven controversial since elevated AFP levels may occur in other benign liver conditions. [63][64][65] Because the prevalence of HCC in the general population is low and the positive predictive value (PPV) of AFP is poor (at 5% prevalence, its PPV has been calculated to be 25.1% using a cut-off of 20 ng/mL), AFP alone is not recommended as a screening tool to identify individuals with an increased HCC risk in the general population. 66 For patients with risk factors for developing HCC, surveillance with AFP is suboptimal as AFP levels are normal (≤20 ng/mL) in about 30% to 40% of patients with HCC and may also be elevated because of nontumour-related causes, such as chronic viral hepatitis, leading to reported sensitivities of 58% to 68% and specificities of 80% to 94% (20 ng/mL cut-off). 11,63,[66][67][68][69][70] Moreover, AFP is particularly insensitive for the detection of small HCC, which limits the usefulness of screening in this important setting. Of note, it has been shown that antiviral therapy can reduce AFP baseline levels, improving the diagnostic accuracy of AFP in patients with chronic hepatitis B. 71,72 In turn, delayed AFP response to antiviral therapy may serve as an indicator of increased HCC risk. 73 It has been debated whether applying a different AFP cutoff value might turn AFP into a suitable surveillance parameter.
However, owing to the inverse relationship of sensitivity and specificity, a meta-analysis of studies with varying AFP thresholds has clarified that a different AFP cut-off value results in either unacceptably high false-positives or false-negatives. 67  Given AFP's performance as a screening parameter in HCC surveillance, AFP alone is not sufficient for the diagnosis of HCC and, therefore, contemporary imaging techniques (for the most part, CT or MRI) are used for screening. 76 However, AFP does have a role as a supplemental test in the clinical setting when the imaging findings are inconclusive or cannot be clearly differentiated from other kinds of liver cancer such as cholangiocarcinoma. A large, randomized, controlled trial that assessed the effect of screening on HCC mortality in at-risk patients determined that ultrasound and serum AFP testing every 6 months reduced HCC mortality by 37% despite poor compliance with the program. 77 Likewise, a surveillance program study that enrolled patients with Child A/B cirrhosis found that ultrasound screening combined with AFP significantly increased the sensitivity of ultrasound screens from 43.9% to 90.2%. 78 Consequently, practice guidelines currently recommend liver ultrasound with or without AFP levels as a screening method for patients at risk for HCC. 9,79,80 Guidelines from Japan recommend inclusion of AFP levels in surveillance programs. 81,82 The biomarker-combined Japan Integrated Staging (bm-JIS) score includes AFP, AFP-L3 and DCP specific for HCC to stratify patients and predict prognosis. 83 In the United States, guidelines suggest that serum AFP levels are optional for use in conjunction with ultrasound for screening at-risk populations. 80 European guidelines have recently reported that adding AFP measurements to ultrasound as the predominant surveillance tool leads to a modest increase of 6% to 8% in detected HCCs. 9

| AFP as a prognostic factor in HCC
Elevated AFP serum levels are associated with a poorer prognosis in HCC patients, 14,54 and serum AFP concentrations ≥400 ng/mL consistently denote poorer prognosis in different clinical settings. 84 AFP has been proven to be a valuable addition to models used for identifying the best candidates for liver transplantation (also in the living donor liver transplantation setting); additionally, high AFP serum levels have been shown to predict the risk of tumour recurrence after hepatic resection, the risk of drop-out while on the waiting list for liver transplantation and the risk of tumour recurrence after liver transplantation (Table 1).  Recently, the combination of pretransplant AFP (with a cut-off of 200 ng/mL) and 18 F-fluorodeoxyglucose positron emission tomography has been reported to be superior in predicting 5-year disease-free survival rates in comparison with the traditional Milan criteria. 107 The rate of increase of AFP before liver transplantation (increase in AFP >50 ng/mL or >15 ng/mL per month) 91,102 has been shown to be a predictor of HCC recurrence and to be associated with worse survival after transplantation. 91 Interestingly, successful downstaging of AFP to levels ≤400 ng/mL before transplantation resulted in significantly better survival, compared with patients who failed to have a reduction in AFP to ≤400 ng/mL (P ≤ .001), and an equivalent survival compared with patients whose AFP had always been ≤400 ng/mL. 92 A similar decrease in recurrence after transplantation has been reported in patients who were bridged with locoregional therapy and whose pretransplant AFP was lowered to ≤13 ng/mL. 108 Therefore, it has been suggested to include AFP levels in the selection of patients for inclusion on liver transplantation waiting lists. 97 In the nonsurgical setting, pre-intervention AFP levels have been shown to predict survival prognosis with locoregional therapies as well as with sorafenib, lenvatinib, regorafenib, cabozantinib and ramucirumab. 5,22,[109][110][111][112][113][114][115][116] Two studies evaluating radiofrequency ablation (RFA) for first-line treatment of HCC in patients with Child-Pugh A/B cirrhosis demonstrated 5-year OS rates of 68% to 76%. 111,115 Univariate and multivariate analysis of prognostic factors associated with survival showed high baseline AFP levels was associated with HCC recurrence and poor prognosis. Transcatheter arterial chemoembolization (TACE) with lipiodol used as first-line treatment for 4966 Japanese patients with HCC demonstrated a 34% 5-year survival rate, and AFP was among several predictive factors identified in multivariate analysis. 116 In patients with advanced HCC, the prognostic value of elevated baseline serum AFP levels has been reinforced in several recently completed phase 3 trials. The association of elevated AFP concentrations with poor prognosis has been consistently demonstrated in these trials including SHARP, 5 REFLECT, 22 RESORCE,112 CELESTIAL 113,117,118 and REACH. 114 Owing to the clear prognostic significance of AFP in the advanced setting, baseline AFP concentration has been increasingly used as a stratification factor in phase TA B L E 1 AFP in liver transplantation: prognosis and risk of recurrence TA B L E 1 (Continued) 3 clinical trial design in advanced HCC. 112 Since these treatment modalities are commonly repeated or administered over time in the same patient, in contrast to surgery, it makes sense to consider the dynamic in AFP levels during a treatment course. Generally, the response to locoregional or systemic treatment is assessed by radiological imaging techniques using the RECIST version 1.1 or mRECIST criteria. However, serial serum AFP measurements during the course of HCC treatment are common in clinical practice based on the hypothesis that AFP reflects tumour activity and burden, and can be assessed much more frequently. 119 There have been a number of studies validating the utility of serial AFP measurements to monitor treatment response to locoregional therapies, and a reduction in the serial trend of AFP has been shown to be a marker of response to locoregional therapies including RFA, TACE or SIRT. 120,121 However, the advantage of using AFP over radiographic assessments is still unclear.
For systemic therapy, initial data from patients undergoing cytotoxic chemotherapy demonstrated that AFP response was a favourable prognosticator. 122 Case studies have suggested that downtrending AFP levels were associated with radiological response; however, validation in more robust clinical studies is required. 127 Overall, there is evidence that the AFP dynamic reflects response to treatment well, and response assessment by serial AFP measurements should be included in future clinical trials to better characterize the role of AFP monitoring during treatment.

| Role of AFP in prognostic scoring systems
Among the more comprehensive HCC staging systems, five (three European and two Asian) have been broadly tested, some of which include biomarkers (AFP, AFP-L3 and DCP) as parameters to refine the staging system ( Table 2) Two scores that include AFP as a parameter have been proposed. 83,134 The BALAD score, which was intended as a staging system that is entirely based on serological markers and includes bilirubin, albumin, AFP-L3 and DCP as well as AFP, has been shown to stratify HCC patients according to survival. 134 Biomarkers (AFP, AFP-L3 and DCP) were also used to refine the JIS system (bm-JIS), resulting in superior stratification ability and better survival prediction in comparison with the conventional JIS score. 83 The BALAD system has been validated in non-Japanese populations. 141,142 Despite these considerable achievements, both scores have not yet been widely implemented and await confirmation in multicenter, international studies.  The effect of immunotherapy relies on the recognition of antigens expressed on cancer cells by the patient's immune system, which subsequently attacks and eliminates the malignant cells. It has long been proposed that AFP as an oncofetal antigen can become a target for immunotherapy because it features potentially immunogenic epitopes and is not expressed in healthy individuals after birth. 150 In addition, AFP promotes the proliferation of liver cancer, which makes it an even more worthy immunotherapeutic target.

| AFP A S A TUMOUR ANTI G EN IN H CC
Naturally, the immune system is tolerant against AFP being a self-protein, and only low immunity is mounted against the protein in HCC patients despite high plasma levels. 150 To overcome this tolerance, several AFP-based immune interventions have been tested in the past, which have, however, been mainly limited to animal models. [150][151][152][153] Further studies are needed to demonstrate a benefit of AFP-based immunotherapies in HCC patients. In addition to the previously mentioned clinical trial (NCT03349255), another similar clinical trial in progress (NCT03132792) includes patients with HCC who have progressed on or were intolerant to prior therapy and have either tumour AFP levels ≥400 ng/mL or AFP expression of ≥1+ in ≥20% of tumour cells by immunohistochemistry, and noncancerous liver tissue with ≤5% cells that stain positive by immunohistochemistry for AFP. Included patients will be treated with autologous genetically modified AFP c332 T cells that will specifically target the patient's own AFP-expressing HCC tumour cells.

| CON CLUS I ON S AND FUTURE DIREC TIONS
Hepatocellular carcinoma is a complex disease with multiple pathogenic mechanisms caused by a variety of risk factors, making it difficult to characterize HCC with a single biomarker. Since its discovery more than 60 years ago, the use of AFP in clinical practice has evolved, and the knowledge of its role in HCC has expanded.
Although AFP's performance as a screening, diagnostic and prognostic marker for HCC is not ideal, it is the most frequently used biomarker in the management of HCC (summarized in Figure 4). Despite Since the use of different cut-off values in past studies has prevented the use of widely accepted AFP thresholds in the clinic, it seems advisable for scientists to apply values that are common practice (eg 20 ng/mL for HCC screening and diagnosis) or have accumulated some evidence (eg 200 and/or 400 ng/mL for treatment stratification and HCC prognosis). In addition, serial measurements F I G U R E 4 The role of alphafetoprotein in the management of hepatocellular carcinoma of AFP over time characterize the dynamic development of this marker, which may add to clinical judgement. In other areas, the future of AFP depends on broader developments, such as whether patient stratification according to molecular subclasses will mature into clinical practice or whether donor liver allocation algorithms will adopt one of the recently proposed criteria.
Like all great masters, AFP is challenged by the next generation.
While AFP may be in its prime, its end is not in sight and it may remain HCC's most important biomarker for years to come.

ACK N OWLED G EM ENTS
We thank Dr Carolin Czauderna (University Medical Center Mainz) for the contribution of Table 2 to this article. Eli Lilly and Company contracted with Syneos Health for writing support from Andrea Humphries, PhD, and editing support from Teri Tucker and Antonia Baldo.