Differentiating between HGDN and WDHCC represents a challenge even to experienced hepatic clinicians, radiologists and hepatopathologists, and the pathological differentiation of pre-neoplastic lesions, particularly HGDN and small WDHCC, is always questionable [20, 21, 26, 27]. Although several immunohistochemical markers such as GPC3, HSP70, GS, and EZH2 have been reported to play roles in the diagnosis of HCC, some limitations remain [10, 13, 28]; e.g., the sensitivity and specificity of GPC3 for the diagnosis of small HCC were 77% and 96% respectively in resected cases , and 61.4% and 92% respectively in needle biopsies . Based on our experience in EHBH, the immunohistochemical sensitivity of GPC3 in 3,232 cases of HCC (from August 2010 to July 2011) was only 63.1%, while those of HSP70 and GS were not as high as expected (data not shown). Such limitations may result in confusion between small WDHCC and HGDN.
In the present study, we used ACY1 and SQSTM1, which were initially identified by screening in our laboratory , and a ‘star molecule’ GPC3 to establish diagnostic panels to differentiate between HGDN and WDHCC using logistic regression analyses. The models were then further validated in an independent set of WDHCC and HGDN samples. ACY1, SQSTM1, and GPC3 expression differed significantly between WDHCC and HGDN (Additional file 1: Table S4). In addition, there were no differences in expression levels of ACY1, SQSTM1 or GPC3 in HCCs <2 cm or 2–3 cm in diameter (Additional file 1: Figure S2).
Moreover, the sensitivity and specificity of ACY1 + SQSTM1 + GPC3 were higher than those of any single marker or any two-marker combination, with a sensitivity and specificity of 93.8% and 95.2%, respectively, for this new diagnostic model of ACY1 + SQSTM1 + GPC3 combination, constructed by logistic regression. The immunostaining scores for ACY1, SQSTM1, and GPC3 can be input into Model 4 during routine daily practice. The model can be easily set up and processed using a workstation. An output value ≤0.6366 is considered highly predictive of HGDN, while an output >0.6366 predicts WDHCC. This three-marker combination (−/+++/+++) demonstrated the highest sensitivity and specificity in terms of diagnostic value for diagnosing HCC, especially early highly-differentiated HCC.
Tommaso et al. recently observed that the use of an additional marker (clathrin heavy chain) improved the performance (sensitivity) of the immunomarker panel GPC3 + HSP70 + GS . We aim to investigate the use of additional markers, including those mentioned above, together with our previous proteomics results, to further improve the sensitivity and specificity of the marker panels.
We demonstrated that ACY1 was expressed at low levels in WDHCC, while SQSTM1 was expressed at high levels in WDHCC tissues, compared with LGDN and HGDN. ACY1 is a cytosolic, homodimeric, zinc-binding enzyme that catalyzes the hydrolysis of acylated L-amino acids to L-amino acids and acyl groups . SQSTM1 is an adapter protein that binds ubiquitin and may regulate signaling cascades through ubiquitination. It may regulate the activation of nuclear factor-κB by tumor necrosis factor-α, nerve growth factor and interleukin-1 [32–34]. The present study demonstrated a gradual decrease in ACY1 expression and a gradual increase in SQSTM1 and GPC3 expression from LGDN to MDHCC, which were confirmed by Spearman correlations and were in accordance with the stepwise progression of hepatocarcinogenesis. Although ACY1 and SQSTM1 showed no prognostic values in this present study, they presented significant diagnostic values and raised the sensitivity of GPC3 for the detection of WDHCC.
GPC3 is a member of the glypican family of glycosyl-phosphatidylinositol-anchored cell surface heparan sulfate proteoglycans . It is expressed in embryonic mesodermal tissues and plays an important role in embryonal growth [36, 37]. In addition to HCC, GPC3 displays loss-of-function mutations in Simpson-Golabi-Behmel syndrome [36, 37], and changes in GPC3 expression levels have been detected in lung squamous cell carcinomas . In the present study, TNM stage and serum AFP were independent prognostic factors for OS and TTR, in agreement with previous reports [24, 39, 40]. Kaplan-Meier and multivariate survival analyses revealed that lower GPC3 expression was significantly linked to both poor OS and increased risk of recurrence after surgical resection in HCC patients. However, apart from studies on GPC3 staining in HCC tissues, few studies have reported any association between high GPC3 expression and poor outcome in HCC patients [16–19]. This discrepancy might be partly related to the following factors. The above studies were based on relatively small sample sizes (n = 61, 86, 107 and 185, respectively), and the use of different GPC3 scoring systems may lead to contradictory results for predicting long-term prognoses . There may also have been differences between studies in terms of factors such as antibody sources and maximum follow-up time (Additional file 1: Table S5). In addition, age, HBsAg, serum AFP, TNM, and tumor differentiation differed significantly between GPC3-low and GPC3-high patients (Additional file 1: Table S6), and these results were similar to those from previous reports [16–18]. To the best of our knowledge, the present study evaluated GPC3 prognostic values using the largest sample size (n = 500) with the longest follow-up time (up to 12 years).
To date, few and limited data have been reported regarding the use of both serological and immunohistochemical biomarkers to predict postoperative prognosis in patients with HCC. As shown by Kaplan-Meier analysis, although either serum AFP or GPC3 staining alone had prognostic values, OS and TTR were lower in patients with both positive serum AFP and low GPC3 expression. In addition, TNM staging and serum AFP combined with GPC3 staining were adopted from Cox multivariate regression analyses, indicating that TNM and serum AFP/GPC3 staining may be a promising prognostic parameter in HCC patients undergoing surgical resection.