AFP is an oncofoetal glycoprotein, largely expressed in foetal yolk sac and liver during embryonic development. Its concentration gradually decreases after birth to < 10 ng/mL in 12–18 months. During pregnancy, AFP levels rise again in maternal serum. AFP levels may also increase in several pathological situations, such as hepatocellular carcinoma (HCC), as well as in other gastrointestinal, lung, and testicular carcinomas [70]. Furthermore, it has also been reported to increase in benign liver diseases, namely in acute and chronic viral hepatitis and cirrhosis [70–72].

Although it is not specifically produced in HCC, AFP has been historically used as a tumour marker for this disease [26, 73]. Nevertheless, AFP expression may vary significantly in HCC, in a range from normal to > 100,000 ng/mL [74]. Up to 30% of HCC patients show normal AFP values at the time of diagnosis, remaining low even in advanced disease [75], and independently of the tumour size at diagnosis [76]. This results in large numbers of false negatives and low sensitivity of the test for HCC diagnosis [76, 77]. On the other hand, elevated AFP levels may be found in liver benign diseases, as mentioned before, resulting in high numbers of false positives and low specificity [26, 76]. Together, these findings rule out the use of AFP for HCC diagnosis and for screening purposes [76].

To improve the sensitivity and the diagnostic power of AFP for HCC, its glycoforms have been investigated using lectin electrophoretic techniques. It was found that AFP presents a distinctive affinity for LCA and, accordingly, three glycoforms can be separated, namely AFP-L1, AFP-L2, and AFP-L3. AFP-L1 does not bind to LCA and is the major glycoform of AFP in the serum of chronic hepatitis and cirrhosis. AFP-L2 is mostly resultant from yolk sac tumours and is also produced during pregnancy, showing an intermediate affinity for LCA [78]. AFP-L3 is the LCA-bound fraction of AFP, being produced in HCC even at early stages [26, 29].

AFP-L3 presents α1,6 core fucosylation on the reducing terminus of N-acetylglucosamine of AFP molecule, and this altered glycosylation provides the reactivity towards LCA (Figure 1). Elevated AFP-L3 levels in HCC result from overexpression of fucosyltransferase 8 (Fut 8; responsible for core fucosylation of proteins in the liver) [79] and also from the increased release of AFP-L3 from hepatocytes into plasma in HCC [80].

AFP-L3 has shown to be a highly specific biomarker for HCC, allowing to differentiate HCC from cirrhosis, and also serves as a predictive marker for the development of HCC during the follow-up of patients with cirrhosis [28, 30, 31]. The AFP-L3 percentage of the total AFP (AFP-L3%), calculated as [(AFP-L3/total AFP) × 100], has been used as a marker for early diagnosis of HCC, for therapy evaluation and to predict the prognosis of HCC [26, 32, 33], being particularly useful in the “diagnostic dilemma” of cases with total AFP concentrations between 10 ng/mL and 200 ng/mL and to exclude HCC in benign conditions with elevated AFP [27, 34]. The clinical performance of the AFP-L3% biomarker versus total AFP for the diagnosis and monitoring of HCC is presented in Table 2.

Comparing the results obtained for AFP-L3% and total AFP quantification, when performed on the same samples, it is clear that AFP-L3% has superior behaviour, with improved sensitivity and specificity, demonstrating the benefits of quantifying this glycobiomarker instead of only measuring the total protein concentration.

Recent studies on fucosylated glycoforms of other liver-secreted proteins, namely haptoglobin, kininogen, α1-antitrypsin and Golgi protein 73 have shown the promising potential of these glycobiomarkers for differential diagnosis of HCC versus cirrhosis with improved sensitivity and specificity [85, 86]. In addition, the monosialylated AFP (msAFP) has also proved to be more specific than AFP, showing the potential to be a diagnostic marker for HCC in nondiagnostic AFP conditions [87].

MUC1 is a heavily glycosylated transmembrane protein (with 50% to 90% of its molecular mass corresponding to carbohydrates), that belongs to the MUC family. It is physiologically expressed in breast tissue, specifically on the glandular or luminal epithelial cells of the mammary gland, as well as in the apical surfaces of simple secretory epithelia of gastrointestinal, respiratory, urinary, and reproductive tracts [88], protecting the underlying epithelia.

In most human epithelial cancers, MUC1 is overexpressed and exhibits aberrant glycosylation [89–92]. In breast cancer cells there is increased sialylation of MUC1, which results in truncation of sugar branches that would be otherwise elongated. Therefore, MUC1 presents mainly core 1 O-glycans (namely T antigen) in opposition to the extensively branched core 2 O-glycans carried by normal MUC1, and the Tn antigen is also overexpressed, as well as the respective sialylated derivatives (ST and STn antigens) [91, 93–95]. There is also an increased expression of high-mannose-type N-glycans [96]. In cancers located in other tissues, different aberrant glycosylation of MUC1 occurs [55, 92, 97].

The CA15-3 assay detects the soluble forms of MUC1 (the peptide backbone), and it is used for therapy monitoring in patients with metastatic disease [98–100]. Since the referred forms are also secreted from cells in other tumours and diseases, the assay lacks the required specificity for breast cancer. Besides, the assay also presents low sensitivity for in situ or low stage disease, as the levels of the soluble forms of MUC1 rarely increase at the early stages of breast cancer. Therefore, it cannot be used for screening or early detection purposes [91, 98]. A related assay, known as CA27.29, also detects the soluble forms of MUC1 but uses a different antibody and a different detection strategy. Both assays originate similar but not identical results [101]. Although it seems that the CA27.29 assay has greater sensitivity than the CA15-3 assay [102], its routine use is yet not recommended due to insufficient data [100, 103].

Aberrant glycoforms of MUC1 have been evaluated in order to be used as breast cancer biomarkers with improved clinical performance over the conventional CA15-3 and CA27.29 assays. An antibody-lectin sandwich assay was designed to detect the aberrant glycosylation of MUC1 in sera from breast cancer patients, using an antibody to capture the protein and the lectin Con A to detect high mannose N-glycans present in MUC1 [35]. The results revealed that mannosylated N-glycans of cancer-associated MUC1 were the most effective targets for the antibody-lectin sandwich assay. In addition, it was shown that the glycosylation level increased with the progression of the disease and the assay efficiently discriminated breast cancer stage I (sensitivity: 63%, specificity: 69%), IIA (sensitivity: 77%, specificity: 75%), IIB (sensitivity: 69%, specificity: 86%) and III (sensitivity: 80%, specificity: 65%) from benign breast disease. There were no differences in the mannosylated N-glycan level of MUC1 between the normal and the benign groups. Immunodetection of the MUC1 protein revealed no differences between the groups in a sandwich enzyme-linked immunosorbent assay (ELISA) using two different anti-MUC1 antibodies, confirming that the increased response of cancer samples in the developed assay was due to the increased N-glycosylation of MUC1 rather than to an increased concentration of MUC1. The enhanced sensitivity of the developed assay versus the current ELISA CA15-3 assay was justified by the heavy glycosylation of MUC1 that resulted in an amplified analytical signal. Although a larger sample size is needed to validate the reported findings, the study demonstrated the advantages of detecting the aberrant glycosylation of the biomarker instead of the total protein concentration.

MUC16 is a large membrane glycoprotein that belongs to the MUC family. It is highly O-glycosylated, with a carbohydrate content accounting for approximately 77% of the total weight of the MUC [104]. CA125 is an epitope present on MUC16, which is detected by specific monoclonal antibodies, such as OC125 and M11, used in CA125 immunoassays [105]. MUC16 is physiologically expressed by secretory epithelial cells of the eye, oral cavity, heart, lung, gastric tract, gall bladder, cervix, and uterus, providing lubrification and protection to the epithelial surfaces [106]. In healthy populations, CA125 serum levels are < 35 U/mL [107], and the values tend to decline with aging and menopause. On the other hand, elevated CA125 levels can be found in healthy women, during the follicular phase of the menstrual cycle [108] and during the first trimester of pregnancy [109], and in non-malignant diseases such as cirrhosis, hepatitis, endometriosis, and ovarian cysts [110–112].

During cancer development, MUC16 is aberrantly expressed, both in its amount and glycosylation [46, 113, 114]. Aberrant MUC16 has been reported to be produced in diverse carcinomas such as ovarian, lung, breast, gastrointestinal, kidney, cervical, uterine, and endometrial [48, 49]. Cancer-associated CA125 antigen presents both N- and O-glycosylation, with N-linked glycans being mainly composed of bi-antennary complex-type and high mannose-type oligosaccharide chains, and O-glycans are, predominantly, mono- and disialylated core type 1 (Galβ1,3-GalNAc) and core type 2 glycans [Galβ1,3-(GlcNAcβ1,6)-GalNAc]. The majority of O-glycans are sialylated and/or fucosylated, and carry a maximum of two fucoses or two sialic acids, or two fucoses and two sialic acids. The Le^x epitope is also present, both in O- and N-glycans [47].

Since MUC16 can be overexpressed in a variety of non-ovarian cancers and benign diseases, the CA125 assay shows poor specificity for ovarian cancer diagnosis and frequently presents false positives. Additionally, the assay has also low sensitivity for ovarian cancer, as around 50% of patients with early stages of ovarian cancer do not express elevated CA125 values, resulting in false negatives [115]. Therefore, the CA125 assay is not recommended for screening ovarian cancer in asymptomatic women. However, it is recommended for screening women with a family history of ovarian cancer, in conjunction with other imaging tests. Furthermore, CA125 testing is recommended as an adjunct to discriminate between benign and malignant suspicious pelvic masses, particularly in postmenopausal women. In spite of its limited sensitivity and specificity for ovarian cancer, CA125 concentration correlates with disease stage, tumour size, and prognosis, so its measurement is recommended and useful for monitoring the response to therapeutic treatment and for evaluation of disease status and detection of recurrence. As CA125 concentration may vary due to differences in calibration, assay design, and reagent specificities, it is recommended that the same assay method should be used throughout for a single patient [99].

Other serum biomarkers have been proposed for ovarian cancer detection, namely human epididymis protein 4 (HE4), alone or in combination with CA125, but the available data are insufficient to conclude a significant improvement in the clinical performance of the assays versus the CA125 assay alone [116]. Multiple-marker based algorithms, namely ROMA and OVA1, have also been developed and approved to assess ovarian cancer risk in women with a pelvic mass, showing increased benefit over the traditional CA125 assay. However, they are not recommended for ovarian cancer screening or diagnosis and the results must be interpreted together with an independent clinical and radiological evaluation [117, 118].

The glycosylation profile of CA125 antigen from human amniotic fluid and from a human ovarian carcinoma cell line 3 (OVCAR-3) was compared, using lectin-affinity chromatography [46]. The results showed a higher abundance of O-glycans in the CA125 antigen from the ovarian cancer cells, with higher expression of GalNAc and lower expression of Galβ1,3-GalNAc, compared to the amniotic fluid samples. Also, the terminal GalNAcβ1,4-GlcNAc structure was found to be predominantly present in ovarian cancer cells. High-mannose-type and bisecting GlcNAc N-glycans were also found to be more abundant in ovarian cancer cells. Although this was a preliminary study, it demonstrated the differential glycosylation pattern between CA125 antigens from normal and cancer origins, and the potential use of a lectin-based assay to discriminate CA125 glycoforms and to complement the total CA125 concentration test, improving its diagnostic performance towards the detection of ovarian cancer. In another study, a sandwich ELISA was developed to measure the level of STn antigen expressed on MUC16, which allowed to differentiate between patients with endometriosis and ovarian cancer [50]. An anti-MUC16 monoclonal antibody was used for capture and an anti-STn monoclonal antibody was used for detection. It was shown that STn antigen was noticeably detectable in the MUC16-enriched fractions from OVCAR-3 cells but negligible in those from peritoneal fluid of patients with endometriosis, and that the level of STn-MUC16 was significantly higher in the samples from patients with ovarian cancer, compared with the samples from patients with endometriosis. These results were consistent with immunohistochemical analysis. Furthermore, the levels of MUC16 and STn-MUC16 were compared for different clinical stages and cytological grades of ovarian cancer, and it was found that an elevated level of STn-MUC16 was more closely associated with an advanced clinical stage and cytological grade than that of MUC16 concentration alone. The level of STn-MUC16 discriminated patients with endometriosis and with ovarian cancer, with 44% sensitivity and 100% specificity.

A microarray-based platform has also been developed for profiling specific aberrant glycoforms including T, Tn, and STn antigens, present on CA125 (MUC16) and CA15-3 (MUC1) [51]. A blinded cohort study of patients with elevated CA125 levels (30–500 kU/L) and a pelvic mass, from the UK Ovarian Cancer Population Study (UKOPS), was performed by analysing a combined glycoform profile that comprised STn-CA125, ST-CA125, and STn-CA15-3 (the three best-performing glycoforms). Glycoprofiling analysis was able to distinguish benign ovarian neoplasms from invasive epithelial ovarian/tubule cancer with a specificity of 61.1% at 90% sensitivity, whereas the CA125 specificity for the commercial assay was only 41.3% at the same sensitivity. The developed microarray assay thus showed its potential to improve the differential diagnosis of ovarian cancer and to significantly reduce the number of patients elected for further testing. A refinement of this study was achieved through the use of proximity-ligation assay (PLA), in which aberrant glycoforms (Tn, STn, and T) of MUC16 and MUC1 in ovarian cancer tissue were detected [119]. PLA reactions for Tn-MUC16, STn-MUC16, Tn-MUC1, and STn-MUC1 were negative in benign lesions but often positive in borderline and malignant lesions. The developed assay allowed to improve the analytical performance of the previous microarray assay, increasing the sensitivity to 72% and 83% in two series of samples, with 100% specificity. The described examples reinforce the usefulness of employing the analysis of MUC16 aberrant glycoforms to improve the clinical performance of MUC16 as a biomarker for ovarian cancer.

The CA19-9 antibody recognizes a glycan epitope present in MUC-type glycoproteins in serum, and in glycolipids (monosialogangliosides) in tissues. The immunodeterminant structure is SLe^a [Siaα2,3-Galβ1,3-(Fucα1,4)-GlcNAc], a sialylated glycoform of the Le^a blood group antigen [120, 121]. The CA19-9 antibody reacts with some normal tissues, namely the pancreatic, gall bladder, and gastric epithelia, showing that CA19-9 can be expressed under physiological conditions [120]. Nevertheless, CA19-9 is also expressed in cancer, for example, carried by MUCs which are known to play a role in cancer progression and metastization [121].

CA19-9 has been extensively reported to be a useful diagnostic aid for pancreatic cancer and it is the gold standard biomarker for this disease, being used to discriminate malignant from benign disorders of the pancreas [42, 43]. Elevated levels of CA19-9 in pancreatic cancer patients have also been associated with disease recurrence [44] and poor prognosis [39]. CA19-9 can also be elevated in other gastrointestinal, ovarian, and hepatocellular cancers, although its routine measurement in such cases is not recommended [40, 99, 121]. However, elevated CA19-9 levels can also be found in benign diseases (and even in healthy individuals, depending on their genetic variation for the expression of CA19-9 and related epitopes) [122], and in about 20% of the patients with pancreatic adenocarcinoma its level may be normal [40, 121].

Due to its insufficient specificity and sensitivity, CA19-9 is not recommended for screening, diagnosis, staging, surveillance, or treatment monitoring of patients with pancreatic or colorectal cancer (CRC) [40, 103]. However, it can be used to monitor the response to therapy for locally advanced metastatic pancreatic cancer and the increase in CA19-9 levels may indicate disease progression [40].

CA19-9 is a carbohydrate and, thus, the aberrant glycosylation associated with cancer is already capitalized in this biomarker. Still, the clinical performance of CA19-9 in the early diagnosis of pancreatic cancer can be further extended if other cancer-associated aberrant glycans and proteins are detected in combination with CA19-9. In fact, a wide range of glycosylation alterations have been associated with pancreatic cancer such as increased expression of SLe^x antigen, increased expression of truncated O‑glycans (Tn and STn), increased branched and fucosylated N‑glycans, upregulation of specific proteoglycans and galectins, and increased O‑GlcNAcylation (Figure 1) [123]. For example, it was shown that CA19-9 expression in MUC5AC and MUC1 proteins may be increased, independently of protein elevation, and that the combined measurement of MUC5AC and CA19-9 presents better performance and improved specificity to differentiate pancreatic cancer subjects from controls [124, 125]. At a threshold of 96% specificity, 78% and 83% of pancreatic cancer patients showed increased expression of CA19-9 in MUC5AC and MUC1, respectively [124]. The complementary relationship between protein and glycan elevations resulted in better sensitivities achieved when both glycan and protein were measured, versus measuring just the protein. In comparison, the median sensitivity of CA19-9 for the diagnosis of pancreatic cancer calculated from pooled data is 79% (70–90%) with a median specificity of 82% (68–91%) [43]. Furthermore, the combined detection of CA19-9 and other cancer-associated aberrant glycostructures, namely wheat germ agglutinin (WGA)-binding glycans in serum MUCs [126], CA242 [127], CA72-4 [128, 129] or SLe^x [130, 131] enabled to increase the sensitivity and specificity of the assays, compared to the measurement of CA19-9 alone. The advantageous combination of CA19-9 with other cancer-associated proteins or glycostructures has also been demonstrated for the detection of gastric cancer [132–134].

CEA is a highly glycosylated (with approximately 60% carbohydrate content) complex glycoprotein belonging to the immunoglobulin gene superfamily [135]. It is an oncofoetal antigen, produced by the intestinal tissue of a normal foetus, but expressed in normal adults only in trace amounts. Being a cell surface immunoglobulin, CEA functions are related with cell adhesion, intercellular recognition, and attachment, exerting also immunoregulatory roles [52, 136]. CEA has been detected in the glycocalyx of surface epithelial cells of normal colon mucosa [137], and more than 95% of the normal population expresses CEA in levels below 5 ng/mL. Smoking habits somewhat increase CEA normal serum values, and CEA tends also to slightly increase in older people [138, 139]. CEA can also be increased in non-malignant diseases, such as liver, gastrointestinal, renal, and other chronic diseases [138, 140, 141].

CEA is also expressed in diverse carcinomas including breast [142], gastric [143], pancreatic [144], renal [145], and lung [146]. In colon carcinogenesis, CEA overexpression contributes to disturb the normal histological architecture of colonic epithelium [147] and facilitates cellular migration and metastization [148].

Both in benign and malignant diseases there is no progressive increase in CEA levels, with values higher than 10 ng/mL being rarely observed [138]. In addition, although CEA is expressed in the great majority of CRCs, its levels are usually normal at the time of diagnosis due to the first-pass effect in CEA metabolization [149]. In spite of its low specificity and sensitivity for CRC diagnosis, CEA has been extensively used as the main biomarker for CRC, as well as for tumours of the gastrointestinal tract. Although it is not recommended for CRC screening, it can be used for staging (as increased serum levels correlate with advanced stages) and for surgical treatment planning. Elevated preoperative CEA may correlate with poorer prognosis and it is recommended that postoperative serum CEA testing may be performed every 2–3 months in patients with stage II or III disease or for ≥ 2 years after diagnosis, as an increase in postoperative CEA level may indicate recurrence. CEA is the marker of choice for monitoring metastatic CRC during systemic therapy [40, 52, 99, 150].

CEA exhibits extensive glycosylation on asparagine residues with up to 28 complex N-linked multi-antennary glycan chains with Le^x and SLe^x structural motifs [57], with the glycofraction consisting mainly of N-acetylglucosamine, mannose, galactose, fucose and sialic acid [150]. O-glycosylation is quantitatively less important than N-glycosylation [55]. The N-glycosylation profile of CEA purified from human colon carcinoma cells and from human liver metastases of CRC cells was analyzed by MS, showing distinctive features versus normal CEA N-glycosylation pattern, namely simultaneously increased bisection and branching, incomplete galactosylation or poly-LacNAc elongation on highly branched structures, moderate levels of sialylation and high levels of fucosylation (Figure 1) [53]. Overexpression of bi-antennary, tri-antennary, and tetra-antennary structures carrying sialic acid and fucose residues, as well as terminal SLe^x structure, were also reported in another study, where CEA from samples of human CRC tissue were analyzed (Figure 1) [54]. In this study, approximately 50% of CEA glycans presented core fucosylation. The glycosylation of CEA from normal versus tumour colon tissue was also characterized using a different approach, based on lectin binding and ELISA assays [55]. The results revealed that tumour-associated CEA contained high levels of Le^x and Le^y antigens (proved by the strong binding of LTA and UEA, respectively), as well as increased expression of mannose (shown by increased binding of NPA) and β1,6-branched N-glycans (demonstrated by strong binding of PHA-L). MAA binding was higher for the cancer samples, evidencing increased expression of sialic acid (Figure 1). The study demonstrated the feasibility of differential detection of tumour-associated CEA using lectin-based ELISA assays and a panel of lectins. A high-density lectin microarray comprising 56 lectins was also developed for the analysis of the glycosylation pattern of CEA from tumour tissues of CRC patients [56]. The results showed different CEA glycosylation profiles between normal and cancer tissues, with CEA from tumour tissues presenting the higher expression of fucose (both terminal and core), mannose and T antigen, and lower expression of GalNAc, N-acetylglucosamine, galactose, and branched and bisecting N-glycans. Moreover, a combinatorial assessment of nine lectins was found to be sufficient to distinguish CRC tumour tissues from tumour-adjacent normal tissues with 83% sensitivity and around 90% specificity. The reported clinical performance surpassed the average performance of the total CEA concentration assay (sensitivity of ~80% and specificity of ~70%) [151]. Additionally, the developed assay showed that alterations in the glycosylation pattern of CEA correlated well with CRC tumorigenesis and progression and that specific glycans were differently expressed on CEA in a stage-dependent manner. In spite of the scarce quantitative studies regarding the improved performance of CEA cancer-associated glycoforms over the conventional total CEA concentration assay, the above-referred studies corroborate the importance of glycosylation analysis of CEA to enhance the clinical performance of this CRC biomarker.

PSA is a glycoprotein primarily expressed in prostate tissue [152]. It is produced almost exclusively by the luminal epithelial cells of prostatic glandular tissue in benign and malignant situations, which results in its low specificity in discriminating both. Most frequently, elevated values of PSA are indicative of benign prostatic hyperplasia (BPH), being also elevated in cases of urinary retention, prostatitis, and ejaculation [153, 154].

Total PSA serum concentration remains the most commonly used cancer biomarker for early detection of prostate cancer, and to monitor therapeutic response, aggressiveness, and recurrence [155, 156]. Nonetheless, its use is limited since there is no single cut-off value that simultaneously yields the high sensitivity and specificity required for a screening test [157–159]. In fact, there is a continuum of risk associated with PSA level and no PSA cut-off point can be clearly defined at which biopsy is recommended [160]. Furthermore, total PSA concentration cannot discriminate between potentially aggressive and indolent prostate cancers [161]. Due to its poor specificity, total PSA quantification is often associated with the risk of overdiagnosis [155, 162] which leads to overtreatment and the associated side effects such as sexual dysfunction, urinary incontinence, bowel problems, together with increased healthcare costs [163].

Other molecular forms of PSA and their ratios, namely free PSA and percent free PSA, proenzyme PSA (proPSA), its derivatives, and the prostate health index, as well as PSA derived-parameters and kinetics (PSA-density, PSA-velocity and PSA-doubling time), have been developed and tested, aiming to improve the specificity of the assays for the diagnosis of prostate cancer and for the differentiation between indolent and aggressive prostate cancers. Although these novel biomarkers have shown better clinical performance than total PSA quantification, allowing to reduce unnecessary biopsies while maintaining high specificity for prostate cancer [164–167], they also present some technical and clinical limitations [168, 169].

The PSA glycosylation profile has been investigated in order to detect cancer-specific alterations that can be employed as efficient prostate cancer glycobiomarkers. Several differences in glycosylation have been found between PSA from normal and from tumour origins (Figure 1), namely in the outer ends of bi-antennary complex structures. PSA produced in cancer cells shows a higher fucose content (core fucose linked α1,6 to GlcNAc) and the oligosaccharides are neutral (do not contain sialic acid), in contrast to normal PSA glycans, which are sialylated. Fucose is also linked α1,2 to galactose, forming the H2 epitope, absent in PSA from normal cells. This is related with the expression of α1,2-fucosyltransferase in tumour cells. There is also an increase in GalNAc content in PSA from tumour cells (from 25% to 65%) [58]. Furthermore, a sialic acid-α2,3-galactose linkage was found as an additional terminal carbohydrate structure, which is detected by MAA [60]. The main glycosylation changes found in PSA from cancer origin and the advantages of detecting the aberrant glycoforms over the conventional detection of PSA protein concentration are presented in Table 3.

From Table 3, we can conclude that the detection of PSA glycoforms shows superior clinical performance and significant advantages over PSA tests, in terms of increased sensitivity and specificity, particularly when PSA concentration is in the diagnostic grey zone, being useful to distinguish between aggressive and indolent prostate cancers and to avoid unnecessary biopsies [180].

Tg is a glycoprotein (with around 10% of its molecular weight comprised of carbohydrates) produced by thyroid cells, being the most abundant glycoprotein in the thyroid gland. Tg plays an essential role in the synthesis of thyroid hormones since it is the initial block to which iodine is incorporated to form triiodothyronine (T3) and thyroxine (T4) [181]. Tg is predominantly N-glycosylated (although it also shows O-glycosylation), containing 20 putative N-glycosylation sites [182]. The main types of glycans in human Tg are high-mannose and bi-antennary complex-type structures [183, 184], and the majority of carbohydrate chains are sialylated [185]. O-glycans in Tg are still not fully identified [186]. The glycan components of Tg are important for the physiological functions of this glycoprotein, including Tg transport, hormone synthesis, and Tg antigenicity [187–189]. For example, sialylation was found to significantly affect the immunoreactivity and solubility of Tg, being also important for the binding to its transmembrane transporter [186, 190].

Measurement of Tg serum concentration is the routine procedure in the follow-up of patients with thyroid carcinoma, since elevated levels of serum Tg after total thyroidectomy indicate residual or recurrent disease [191]. It can also be used for the diagnosis of several thyroid benign conditions [186]. Serum Tg quantification is performed by immunometric methods, which include immunoradiometric and immunochemiluminometric assays [192, 193]. Although most recent methods are very sensitive, detecting Tg concentrations down to 0.1 ng/mL, they present several limitations [194]. First, as they are based on monoclonal anti-Tg antibodies, and assays from different providers use diverse antibodies that recognize different Tg epitopes, a significant variability in results is observed when diverse assays are used [193, 195]. Therefore, it is recommended that serum Tg should be assessed in the same laboratory and using the same assay during the follow-up of an individual patient [191]. In addition, monoclonal anti-Tg antibodies are produced from normal thyroid tissue-derived Tg, which present epitopes that differ from the plasma-derived Tg produced by thyroid tumours, and this can result in false negatives in the presence of thyroid carcinoma [196]. A second major limitation relates with the fact that 25–30% of patients with differentiated thyroid cancer have endogenous circulating anti-Tg antibodies (a three-fold greater prevalence than that found in the general population) [193], which can interfere with the assays, usually producing low or undetectable Tg levels, impairing the use of this biomarker for follow-up purposes. On the contrary, the presence of heterophile antibodies (such as human anti-mouse antibodies) may result in falsely high Tg levels [186]. Finally, total Tg serum concentration levels cannot differentiate between benign and malignant thyroid disease [197], therefore more specific biomarkers are needed for the diagnosis of thyroid cancer.

In several chronic thyroid diseases, such as cancer and autoimmunity, alterations in Tg glycosylation have been observed [64, 198]. In thyroid carcinogenesis, changes in sialylation and fucosylation, increased β1,6 branching of N-glycans, alterations in the content and structure of poly-LacNAc chains, and changes in O-GlcNAcylation have been reported (Figure 1) [63–66]. Plasma N-glycomes from samples of thyroid cancer, benign thyroid nodules, and matched healthy controls were analyzed by MALDI time-of-flight MS (MALDI-TOF MS) and they were found to differ between the three groups in several aspects, namely complexity, galactosylation, fucosylation and sialylation [199]. Histochemical analysis of tissue specimens from human thyroid carcinomas showed increased sialylation (especially α2,3-linked sialic acid) on follicular epithelial cells [65]. Nonetheless, Tg displayed decreased sialylation in cancer transformation (Figure 1) [67]. Furthermore, differences in the lectin binding profile between normal Tg and Tg from thyroid carcinoma have been described. Tg from thyroid carcinoma showed a lower affinity for Con A and a higher affinity for WGA, in contrast with Tg from normal thyroid tissue [200]. Asialo-glycan chains with truncated structures were also found in malignant Tg, and carbohydrate chains of higher molecular mass, which have repeating Gal-GlcNAc disaccharides and peripheral α-fucosyl residues were present in metastatic papillary carcinoma [68]. In papillary thyroid carcinoma, phosphorylated high-mannose type and hybrid-type oligosaccharides each containing one phosphate group in a diester linkage were also found to be present instead of sialylated oligosaccharides [69].

Additionally, Tg from patients with metastatic thyroid cancer showed increased binding to LCA, which has an affinity for mannose and glucose residues, allowing to discriminate between benign thyroid disease and thyroid cancer, with the LCA-binding Tg fraction being significantly lower in patients with thyroid cancer [201]. In the referred study, the immunocaptured Tg was incubated with LCA and both LCA-bound and unbound fractions were measured. The LCA-reactive Tg ratio (%) was calculated as [(total Tg concentration – LCA-nonreactive Tg concentration)/total Tg concentration × 100]. For total Tg concentration < 50 ng/mL, the LCA-reactive Tg ratio could not distinguish between healthy volunteers, patients with benign tumours, and patients with thyroid carcinoma without metastasis but, for total Tg concentration > 200 ng/mL, the calculated parameter could discriminate between benign thyroid tumours and thyroid carcinoma without metastasis or with lymph node metastasis. For total Tg concentrations between 51 ng/mL and 200 ng/mL, the LCA-reactive Tg ratio did not differ between patients with benign tumours and those with thyroid carcinoma without metastasis, but significant differences in it were observed between patients with thyroid carcinoma with lymph node metastasis and those with benign thyroid tumour or thyroid carcinoma without metastasis. Therefore, the LCA-reactive Tg ratio showed to be a valuable marker for distinguishing between thyroid carcinoma and benign thyroid tumours. The same method was applied to another set of samples, and the results confirmed that it could be used as a screening method for primary thyroid carcinoma [62]. For an LCA-reactive Tg ratio cut-off of 81.8%, with patients below this level defined as positive, the rate of positivity was 11.3% in the benign group and 61.4% in the malignant group. Total Tg serum concentration measured in the same samples did not significantly differ between benign and malignant groups, which demonstrates the enhanced performance of the LCA-reactive Tg ratio over the total Tg concentration as a sensitive and specific biomarker for thyroid cancer.