Diagnostic Value of Immunohistochemical Markers in Four-grade Histological Classication of Hepatocellular Carcinoma

Backgrounds: Accurate differential diagnosis regarding histological grades of hepatocellular carcinoma (HCC) is critical for targeted treatment and prognostic evaluation. However, the currently used descriptive diagnosis of histological grading have to observe tedious item of neovascularization, stromal invasion, cellular and structural atypia, which have drawbacks of subjectivity and error-prone. Immunohistochemical (IHC) markers-based diagnosis only needs to determine whether there is staining in the cell membrane, cytoplasm and / or nucleus. Using IHC markers targeted heat shock protein (HSP)70, glypican (GPC)3, glutamine synthetase (GS), and organic anion transport peptide (OATP), we sought to establish an easy method for the diagnosis of the histopathological grades and further explore the best ecacy by their combined or independent application. Methods: We retrospectively conducted a study of 157 patients with 200 histologically conrmed HCCs, which were classied into early (n= 45), well (n=31), moderately (n=68), and poorly (n=56) differentiated (diff.) HCC. The sensitivity, specicity, accuracy of HSP70, GPC3, GS and OATP on each histological grade were examined. Results: HSP70 and GPC3 showed difference in most histological grades of HCC ( P < 0.05). GS distinguished early HCC from three other histological grades (p < 0.01). OATP8 only differentiated early HCC from poorly diff. HCC (P=0.019). When any two of the three indicators (HSP, GPC3, and GS) were negatively expressed, the diagnostic ecacy of early HCC was the highest, with an area under the curve (AUC) of 0.802 and accuracy of 80.5%. The optimal ecacy for poorly diff. HCC detection was obtained when both GPC3 and HSP70 were positively expressed (74.4% accuracy; AUC = 0.703). However, for well and moderately diff. HCC, a relative satisfactory AUC value (>0.60) failed to yield either


Background
In the diagnosis of hepatocellular carcinoma (HCC), proper histopathology-based classi cation is bene cial to make a appropriate therapy plan and predict prognosis [1], mainly because different histological grades of HCC show difference in many aspects of diagnosis, treatment and prognosis. For example, patients with early HCC (eHCC) may have a higher surgical cure rate, a lower recurrence rate, and higher short-and longterm survival rates than those with advanced HCC (adHCC) [2][3] [4]. EHCC can be completely cured [5], but poorly differentiated (diff.) HCC exhibits relatively worse recurrence-free and overall survival rates [6]. Compared with patients histologically diagnosed as well diff. HCC, those who diagnosed as moderately diff. HCC tend to have advanced clinical stage of HCC [7].
There is a critical problem of currently used histological grading method for HCC. No matter two-(eHCC and adHCC) [7] [8], three-(well, moderately, and poorly) [9][10], four-grade (grade I/II/III/IV) [11] [12] classi cation methods and our newly developed a four-grade histological classi cation method, which classi ed HCC as "early/well/moderately/poorly differentiated" [13], consistently adopt diagnostic method to observe the morphological and structural criteria (cellular atypia and structural atypia) of both the lesions and surrounding tissues under the microscope, and thus make a comprehensive diagnosis. In details, hematoxylin-eosin (HE) staining was needed to observe cell density, nuclear / cytoplasmic ratio, cell atypia, nuclear atypia, especially the arrangement of trabeculae; silver staining to observe whether reticular bers were clear or sparse; CD34 to observe neovascularization; Victoria blue (VB) to observe stromal invasion; and cytokeratin (CK)7 staining to distinguish ductular reaction [14].
Obviously, these observation items are tedious and time-consuming. The identi cation of trabeculae and neovascularization require a certain pathological knowledge of the readers. In addition, there are more descriptive indicators rather than quantitative indicators in the basis of diagnosis. Therefore, the gold standard of histological grading method is a little subjective and not easy to be popularized in clinical practice.
Even worse, identi cation of stromal invasion of tumoral hepatocytes and tumor boundaries, which are considered as of great importance in histological grading diagnosis of HCC, are sometimes di cult to recognize in specimen obtained by ne-needle aspiration because of its limited amount [15]. All these factors contribute to the challenge and even dilemma for achieving a quick and con rmed histological grading diagnosis.
The novel IHC methods make diagnosis according to whether the tumor cell membrane, nucleus and cytoplasm are stained or not. Readers only need to distinguish the cell membrane, nucleus and cytoplasm, and there is no need to observe the tedious structures such as portal vein area, cytoskeleton, and trabeculae. Therefore, it is easier to observe, faster to obtain results, and require fewer expertise pathological technology of operator, so it is more objective. Most importantly, the reported positive results of heat shock protein 70 (HSP70), glypican 3(GPC3), glutamine synthetase (GS), and organic anion transport peptide 8 (OATP8) [16][17][18] [19], provided a promising insight in feasibility of IHC marker-based HCC diagnosis. Nevertheless, there is very few data available regarding IHC marker-based histological diagnosis of HCC, not to mention the combined use of IHC markers to achieve best effectiveness for histological diagnosis.
In this regard, taking our four-grade histological classi cation as a golden standard, we conducted a retrospective research to explore the diagnostic ability of the IHC markers. By introducing an alternative diagnostic criteria by IHC indicator, this study might be bene cial for optimize and popularize the histologcial grading diagnosis of HCC in clinical practise.

Patient enrollment
Between January 2014 and August 2018, we retrospectively enrolled a total of 200 newly discovered and untreated hepatic lesions obtained from 157 consecutive patients. These lesions were de nitively diagnosed as HCC and histologically graded after ultrasound-guided biopsy and subsequent IHC staining. Clinical information (gender, age, associated liver disease, Child-Pugh classi cation), radiological images, and histopathology reports were retrospectively collected from a review of the electronic medical record system, radiology database, and pathology records, respectively, of our hospital. The institutional review board of the Medical Center of Yokohama City University approved this study (approval no. B180200054), and written, informed consent was obtained from all patients. Patients with the following characteristics were excluded from this study: (1)Child Pugh score system classi ed as grade C; (2)Insu cient specimen from the target lesion, making it impossible to obtain a de nitive pathological diagnosis; (3)Only one of the four IHC staining procedures (OATP, GPC3, CS and HSP) was conducted. Such cases were excluded to ensure that the IHC diagnostic value was compared between the same lesions as much as possible to maintain homogeneity of analysis.

Current used descriptive diagnosis
Based on the evaluation of the location and needle path of the target lesions via preoperative grayscale ultrasonography, a 21-gauge ne needle (SONOPSY; Hakko, Tokyo, Japan) aspiration biopsy under ultrasonography or contrast enhanced ultrasound guidance was performed. At least two samples from each lesion and one sample from the adjacent liver tissue (negative control) were obtained. Formalin-xed, para n-embedded tissue sections were consecutively cut and immunostained with antibodies directed against HE, VB, silver staining, CD34, and CK7 antigens.
We classi ed HCC into early, well, moderately, and poorly differentiated (diff.). The parameters evaluated in this study included cell atypia, structural atypia, reticular ber patterns, neovascularization, stromal invasion, and ductular reaction (Table 1) [13]. The density of cancerous cells, nuclear size and morphology, nuclear/cytoplasmic ratio, and trabecular structure were observed via HE staining. Decrease of reticular ber were evaluated using silver staining. By assistance of VB staining, intratumoral stromal invasion (tumor cell invasion into the intratumoral portal tracts [20]) for the diagnosis of early HCCs was determined [21]. We used the CK7 to stain the bile duct, which is a suggestive of change in bile metabolism in HCC [22]. CD34 was used to stain the endothelial cells to detect neovascularization [23]. Observation of CK7 staining was focus on the periportal bile duct of HCC lesions, which is re ection of dysfunction of bile-producing metabolic pathways [24,25]. CD34 staining targeted at the cytoplasm/membrane of vascular endothelial cells, which reveal sinusoidal capillarization ( Fig. 1-4). variable ages (with normal distribution) are presented as means ± standard deviation (SD), and mean age was compared between histological grades using one-way analysis of variance (ANOVA). Continuous variable sizes (skewness distribution, homogeneity of variance) are described as medians (inter quartile range) and compared between different histology grades using the nonparametric Kruskal-Wallis test. The rest of the baseline data and imaging indicators were tested among the four histological groups using the chi-squared test and Fisher's exact test. Two-tailed Fisher's exact and chi-squared tests were used to compare the frequencies of all categorical variables for inter-group differences in histology grades. The area under the curve (AUC) of a receiver operating characteristic (ROC) curve was used to determine the diagnostic usefulness of indicators for each diff. HCC. The level of signi cance was set at P < 0.05. The cut-off value of continuous data (lesion size) was obtained from the ROC curve analysis.The size value yielding the maximum sum of sensitivity and speci city is determined as the cutoff value. The above statistical analyses were performed using SPSS version 26.0 for Windows (SPSS Inc., Chicago, IL, USA). The sensitivity, speci city, and accuracy of each indicator were calculated manually.

Results
Clinical characteristics  Inter-group comparison of the IHC markers for four histological grades In general, when each diagnostic indicator was analyzed between any two histological grades, differences in GPC3, GS, HSP, and OATP8 expression were statistically signi cant between some, but not all histological grades (

Analysis of the diagnostic e cacy of the IHC markers
In general, some of the IHC indicators we studied performed well in the diagnosis of early HCC, followed by poorly diff. HCC (Table 4). Diagnostic e cacy for well and moderately diff. HCC was not very high (data not listed). In detail, optimal e cacy for early HCC detection was obtained when any two of the three IHC indicators GPC3, HSP70, and GS were negatively expressed. The sensitivity, speci city, accuracy, and AUC were 79.5%, 80.8%, 80.5%, and 80.2, respectively. For poorly diff. HCC, the combination of positive GPC3 and HSP70 expression yielded a relatively acceptable accuracy (74.4%) and AUC (0.703). Although the negative expression of OATP8 was signi cant in the diagnosis of poorly diff. HCC (accuracy = 60.0%, AUC = 0.566), the combination of OATP8 in any form failed to improve the AUC and accuracy (accuracy = 67.3%, AUC = 0.703, data was not shown in tables). The highest e cacy for diagnosing well diff. HCC was the combination of positive HSP70 and GPC3 expression, while for moderately diff. HCC, the combination was of any two indicators of negative OATP expression and positive HSP and GPC3 expression.
Unfortunately, their accuracy and AUC were all lower than 0.60, inferior to those of early and poorly diff. HCC. Our research showed that the combination of some IHC indicators yielded good diagnostic accuracy for early and poorly diff. HCC. Early HCC was much easier to be detected than the other histological grades. Unfortunately, the diagnostic accuracies of each IHC indicator for well and moderately diff. HCC were not as good as those for early and poorly diff. HCC.
GPC3 is a member of the glypican family of heparinsulfate proteoglycans, which are cell proliferation inhibitors and apoptosis inducers [29].
GPC3 is negatively expressed in both normal liver tissue and benign hepatocellular nodules. Although GPC3 is not a speci c hepatocellular marker, the reported sensitivity (91.7%) and speci city (100%) for small HCC detection are higher than those of alpha-fetoprotein (AFP). Thus, GPC3 is considered a reliable and early HCC biomarker [30]. Studies have reported that GPC3 overexpression increases with the progression of hepatocarcinogenesis [17,31,32].
HSP70 is a member of the HSP family, which is involved in the regulation of cell cycle progression and apoptosis. In early HCC, HSP70 synthesis may be stimulated by nutritional depletion and hypoxia resulting from insu cient blood supply [33]. Chuma et al. [33] reported the upregulation and signi cant overexpression of HSP 70 in early and advanced HCC, respectively.
Consistently with the promising ndings from previous studies stated above, our study showed that GPC3 and HSP70 expression increased with the advancement of histological grade. Moreover, negative and positive GPC3 and HSP70 expression displayed good values for early and poorly diff. HCC, respectively. Nevertheless, some reports have suggested that it is di cult to diagnose poorly diff. HCC using IHC markers, because poorly diff. HCC may lose its immunoreactivity to some IHC markers [34]. In contrast, our study revealed that all the IHC markers we investigated were valuable for distinguishing poorly diff. HCC from other histological grades. Notably, GPC3 was positive in 87.5% (49 of 56) of poorly diff.
HCC, similar to previous studies (85-89%) [35]. Furthermore, it was exciting to nd that the combination of GPC3 and HSP70 in our study yielded a relatively fair diagnostic value for poorly diff. HCC (AUC = 0.703).
GS is an enzyme which supplies tumors (not speci c to HCC) with energy by synthesizing glutamine. It has been speculated that as HCC progresses, the proliferation rate of tumor cells increases, and the expression of GS, which plays a key role in cell proliferation, increases accordingly [32]. Speci cally, the degree (no, weak, moderate, or high) of GS expression has been reported as a sensitive marker of active βcatenin signaling in human HCC [36]. Nuclear expression of β-catenin might be more prone to occur in tumor cells of poorly diff. HCC, while βcatenin mutations are mainly found in non-HBV (hepatitis B virus), well diff., and chromosome-stable HCC [36]. Unfortunately, whether GS expression in relation to the histological de-differentiation of HCC is unclear, as published results are contradictory [31][32] [37]. In our study, a positive GS outcome indicated that the GS expression level could distinguish early HCC from other histological grades. However, there was no of early HCC was extremely low (25%). However, this does not mean that GS is useless for HCC grading differentiation. Surprisingly, GS speci city (97.4%) was much higher than that of GPC3 (66.5%) and HSP70 (55.6%), suggesting that negative GS expression can almost rule out the possibility of advanced HCC. Taking advantage of the potential complementarity between GPC3 and HSP70, the combined application of these IHC indicators hint at the best e cacy for diagnosing early HCC (AUC = 0.802, accuracy = 80.5%).
Many negative results were also noted for OATP8. As a member of the OATP family, which strongly mediates the transportation of a variety of endogenous and exogenous substrates through the cellular membrane, OATP8 (synonymous with OATP1B3, gene symbol SLC21A8) is speci cally expressed at the basolateral hepatocyte membrane. OATPs are of considerable importance in HCC diagnosis and treatment, as they are closely related to the occurrence, recurrence, and prognosis of HCC [38]. However, only a few studies have explored the relationship between OATP expression and the histological grading of HCC. It is commonly acknowledged that, with increasing histological grading of HCC, bile formation decreases in the hepatocytes [11]. As OATP8 promotes hepatocyte recovery of bile acids from portal vein blood, OATP expression might theoretically be related to the histological classi cation of HCC. By semi-quantitatively evaluating the effect of OATP8 expression on the tumor cellular membrane relative to that on surrounding non-neoplastic hepatocytes, Kitao et al. reported that OATP8 expression signi cantly decreased from well to moderately to poorly diff. HCC [19], but the mechanism of this correlation remains unclear. However, the authors speculated that it may be attributable to the increased expression of hepatocyte nuclear factor (HNF)-3β, an essential transcription factor for hepatocyte differentiation [19]. Tsuboyama et al. also found that as histological grade advanced from well diff. HCC to poorly diff. HCC, the dominant OATP distribution changed from grade 2 to grade 0 [28]. Unfortunately, the number of cases in this study was too small, and the authors did not make further statistical analyses or explain the changing trend in OATP. The method we used to detect OATP was exactly the same as that used by Kitao et al. [19], but a signi cant difference was only detected between early and poorly diff. HCC. It is unclear what makes our results different.
Since there are few studies on the relationship between OATP and HCC histological grading, and more importantly, because whether used alone or in combination with other IHC markers, OATP contributed little to improve the diagnostic e cacy, we recommend OATP as an alternative supplementary rather than a preferred method for grading diagnosis.
In our study, we observed signi cant differences in lesion sizes between early HCC and other histological groups, suggesting that size may play a role in early HCC diagnosis. Some studies have reported that the histological grade of HCC progressed with the increase in tumor diameter [39] [40]. In view of this, we conducted a subgroup analysis of the most e cient diagnostic indicators for early HCC (combination of GPC3, HSP70, accuracy of > 79% and AUC of > 78%. The statistical results were much closer to those not grouped by size (AUC = 0.802, accuracy = 0.5%). These results suggest that size is a negligible factor in terms of the IHC indicator outcomes.
This study has a few limitations due to the specimen collection method. Compared with surgical resection, there might be a higher possibility of inadequate amount and heterogeneous staining of specimens obtained via ne needle aspiration biopsy [41], which may lead to errors in histopathological diagnosis. For example, once focal staining areas are undetectable in fractioned, tiny, and less representative tissue, they may be misdiagnosed as negative expression of immunohistochemical markers [31].

Conclusion
In conclusion, we comparatively analyzed IHC methods for distinguishing the histological stages of HCC. OATP, HSP70, GS, and GPC3, especially in certain combinations, produced considerate differential effects between different histological grades and further demonstrated good diagnostic ability in early and poorly diff. HCCs. Therefore, we recommended IHC markers-based diagnostics method as a promising option instead of currently used descriptive method mainly based on cell and structural atypia.

Consent for publication
Not applicable.

Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.