A Limited Immunohistochemical Panel Can Subtype Hepatocellular Adenomas for Routine Practice

Brent K. Larson, DO; Maha Guindi, MD


Am J Clin Pathol. 2017;147(6):557-570. 

In This Article


Since clonality studies first demonstrated that the telangiectatic variant of FNH was, in fact, the variant of HCA now recognized as IHCA,[25,26] successive molecular studies have revealed at least four subtypes of HCA.[5] Extensive morphologic and immunohistochemical studies have shown good correlation with molecular subtyping. H-HCA is characterized by macrovesicular steatosis, b-HCA by cytologic atypia and pseudoglandular formations, and IHCA by ductular reaction, sinusoidal dilatation, and inflammatory infiltrates.[6,8] But while there may be correlations with genotypic subtypes, these morphologic features are of variable utility in diagnostic practice. While steatosis is considered the hallmark of H-HCA, it is neither sensitive nor specific, with 60% to 70% of H-HCAs showing marked steatosis,[19,21,27] although other subtypes can also be steatotic.[24,27,28] Some suggest that no histologic factors are predictive of b-HCA.[27]

In the current study, when analyzed independently, only pseudoportal tracts, ductular reaction, and inflammation were significantly correlated with a subtype (IHCA), and none were more than 70% sensitive. After overall morphologic evaluation and subsequent immunohistochemical staining, more than 40% of adenomas were reclassified, demonstrating the need for ancillary studies.

β-Catenin and GS staining markedly increased the number of b-HCAs identified from three to 11 (four pure b-HCAs and seven b-IHCAs), indicating that rare cytologic or architectural atypia are not sensitive enough measures to identify these subtypes. Over one-third of adenomas morphologically classified as IHCAs were reclassified as b-HCAs or untypable HCAs after immunostaining. This again reinforces the importance of β-catenin and GS staining, as b-IHCAs frequently had inconspicuous atypia. As in some previous studies, morphology of IHCA was variable,[18,20,21] with a total of 10 adenomas reclassified from IHCA to untypable HCA (n = 7) or from untypable HCA to IHCA (n = 3) after SAA staining. While some individual morphologic features were significantly associated with IHCA, they were clearly not sensitive predictors of a final immunohistochemical diagnosis. In particular, the large number of HCAs initially classified as IHCAs morphologically that were reclassified as untypable after immunostaining frequently showed one or a few of the morphologic features associated with IHCAs without showing the full complement.

A plethora of immunostains can now be used to subclassify HCAs. Negative L-FABP immunohistochemistry is diagnostic of H-HCA.[6] Nuclear β-catenin staining and diffuse GS positivity correlate well with molecular disruptions of the β-catenin pathway.[4,6] SAA and CRP are useful for diagnosing IHCA.[7,8] Additional markers may be on the horizon, as well, such as a STAT3 immunostain.[29] When there is diagnostic difficulty in separating HCA from FNH and HCC, even more markers may be employed, including glypican 3, CD34, reticulin, CK7, CK19, Ki-67, and HSP70.[9,18,29,30]

While many ancillary techniques exist, relatively few are available or practical in routine clinical practice. Molecular studies are the definitive method for subtyping HCAs, but they are not widely available. Bioulac-Sage and coauthors[7] have suggested a diagnostic algorithm predominantly using immunohistochemistry and reserving molecular methods as a last resort. Several other groups have confirmed the usefulness of immunohistochemistry in subclassifying HCAs, with most studies focusing on a panel minimally including GS, L-FABP, β-catenin, SAA, and CRP.[7,9,13,18–22,27]

We chose our limited panel with the goal of identifying the two HCA subtypes that could benefit most from direct surgical intervention and with a focus on ancillary studies that could practically be adopted by a laboratory with a small or moderate volume of liver specimens. β-Catenin is already a widely used immunostain, showing nuclear positivity in a variety of neoplasms and fibromatoses. GS has additional use in assessing other liver nodules, aiding in the diagnosis of FNH[7,8,29] and the differentiation between benign and malignant nodules in cirrhosis.[31,32] The MC1 clone we used for immunostaining against SAA has cross-reactivity with amyloid A, a protein product of SAA metabolism that may aggregate in systemic amyloidosis. Therefore, not only can intracellular staining be used to subtype HCAs, but extracellular staining also can be used to subtype amyloid fibrils in other organs and clinical situations. CRP immunostaining is less specific for IHCA than SAA, particularly when FNH is in the differential.[8] L-FABP immunostaining was developed as a marker of H-HCA and remains, to our knowledge, essentially restricted to this specific clinical situation. Therefore, we felt β-catenin, GS, and SAA to be an appropriate starting point.

GS is an enzyme involved in ammonia metabolism and glutamine synthesis in the liver. It is concentrated in zone 3 of the normal hepatic parenchyma, with immunostaining effectively limited to a few hepatic plates around central veins. However, in hepatic neoplasia, GS gene expression may be upregulated by the β-catenin pathway.[4,33] In this study, nuclear β-catenin positivity was patchy, and all b-HCAs or b-IHCAs with nuclear β-catenin positivity (one and three, respectively) also showed diffuse GS positivity, making GS 64% more specific for this subtype, similar to previous reports.[18,21,27] This suggests that our diagnostic algorithm could be further abridged by foregoing β-catenin staining and using only GS to detect b-HCA/b-IHCA.

Other authors have expressed difficulty interpreting GS.[8,9,22] In our series, eight cases showed less than overwhelming positivity. Three showed diffuse weak positivity, and two showed strong patchy positivity (see Image 5B). Three additional cases with atypical GS staining showed patchy strong staining in small areas of residual HCA compressed by adjacent HCC. Strong, diffuse positivity for GS and nuclear positivity for β-catenin are most strongly associated with mutations in exon 3 of CTNNB1, the gene coding for β-catenin. However, b-HCAs and b-IHCAs with newly recognized CTNNB1 mutations in exons 7 and 8 and other rare loci often show patchy and/or weak GS positivity and lack nuclear β-catenin positivity.[4,34] While these non–exon 3 mutations appear to be of lesser malignant potential, risk still exists and warrants caution until better characterized. The criterion of more than 50% glutamine synthetase staining in our algorithm is designed to capture both the classical strong diffuse pattern and the weaker, patchier staining associated with rarer mutations while excluding patterns with minimal staining not associated with β-catenin pathway upregulation.

β-Catenin–activated inflammatory HCA is not typically considered a standalone subtype, although its identification is also important due to CTNNB1 mutations and malignant potential.[4] Because b-IHCA is primarily a molecular and immunophenotypic diagnosis without specific morphologic features, morphologic features of b-IHCA were not specifically assessed. Rather, considering b-IHCA as a subtype showing secondary changes superimposed on preexisting IHCA or b-HCA, it was included in analyses for morphologic features of IHCA and b-HCA. When included in these categories, b-IHCA accounted for 28% of all IHCAs and 64% of all HCAs with β-catenin activation, similar to distributions in the literature.[4,6,18,19,21]

IHCAs have been associated with overweight/obesity and other proinflammatory conditions.[14,15,19,20,35] Specifically, IHCAs have been associated with elevated serum CRP and gamma-glutamyl transpeptidase (GGT).[15,19,29] Recording preoperative serum CRP and GGT values was considered for this study but proved impractical due to the rarity of HCAs and the difficulty of collecting clinical information retrospectively. While patients with IHCA in our series were overweight compared with those with other HCAs, this difference did not reach statistical significance (P = .2519). When more features possibly indicating inflammation, such as components of the metabolic syndrome, were included, there was still no significant correlation with the presence of IHCA (P = .4495). This indicates that the idealized clinical profile may not be sufficiently sensitive or specific to influence the diagnosis of IHCA and reinforces the importance of ancillary pathologic studies.

The algorithm in this study is meant to identify those subtypes prompting surgical management. The obvious shortcoming of this algorithm is its inability to subtype H-HCAs and UHCAs. Steatosis shows good correlation with H-HCA and may raise suspicions for this subtype, but as mentioned above, it is by no means specific. Although our algorithm cannot confirm, through absence of L-FABP staining, that a steatotic adenoma is an H-HCA, the algorithm is unlikely to overcall steatotic HCAs as H-HCAs/untypable HCAs, as positive stains will appropriately diagnose steatotic IHCA and b-HCA. While rare examples exist of patients with concomitant H-HCAs and other HCA subtypes, HNF1A mutations are mutually exclusive of mutations in the JAK/STAT and β-catenin pathways.[4,28,36] Thus, positive staining for β-catenin, GS, or SAA in our algorithm virtually rules out H-HCA without staining for L-FABP. Not staining for L-FABP will also make it impossible to diagnose incidental microscopic steatotic foci as microadenomas and more difficult to definitively diagnose adenomatosis, both of which are usually associated with H-HCAs.[16,17] But in this study, only one patient had more than 10 masses by imaging, and the single mass examined histologically was successfully classified as an IHCA.

Although they do not typically require surgical resection, a diagnosis of H-HCA does raise clinical issues such as monitoring for adenomatosis, screening for diabetes mellitus, and avoiding oral contraceptive pills. Should H-HCA be strongly suspected, one can go beyond the algorithm to send out for L-FABP staining. If questions still remain or if GS immunostaining is difficult to interpret, molecular studies can be performed, although this requires routinely freezing and holding fresh tumor tissue.

Without molecular techniques available for comparison, this study relied on immunohistochemistry as the diagnostic gold standard. To control for this imperfect gold standard, we examined HCA subtype distribution in the literature. The distribution in the literature varies widely, and that of the current study was significantly different from five other immunohistochemical studies. One reason for these differences may lie in differing patient populations and practice settings. Some studies were performed at tertiary centers with large referral practices, while others specifically examined patients in a community hospital setting. Definitions of each subtype also vary in the literature. The difference between b-HCA and well-differentiated HCC can also be nuanced and somewhat subjective and likely explains the differences in rates of b-HCAs in the literature.[22] Even with these potential confounders, we feel confident in our algorithm, as the distribution of subtypes in this study closely mirrors that in Bioulac-Sage and coauthors' landmark study[6] using molecular subtyping as the diagnostic gold standard.

All specimens studied were from resections to maximize the number of cases available for review; no biopsy specimens were included. Although archival cases available for study were almost exclusively resection specimens, we concur with the increasing calls for biopsy of suspected HCAs for confirmation and subtyping.[37,38] While we propose an algorithm to be performed on biopsy specimens using a study set of resection specimens, we feel that the results are easily extrapolated to biopsy samples, as prior literature bears out the use of each of these immunostains on biopsy material.[7,8,13,31,32]

The primary risks of HCAs are hemorrhage and malignant transformation. HCAs larger than 5 cm and those found in males are at significant risk of malignant transformation. While these features are known before tissue diagnosis, the pathologist can still significantly contribute to patient management by subtyping HCAs. b-HCAs, b-IHCAs, and IHCAs have an elevated risk of malignant transformation, and IHCAs may cause systemic inflammation. They benefit from excision, and so finding a feasible way to identify them in clinical practice is imperative.

This study is limited by its retrospective nature and by exclusive reliance on resection specimens, and further validation in prospective studies on biopsy material is needed. While a limited panel may not be feasible in all situations, as when the diagnosis of HCA vs FNH or HCC remains unclear, a practical three-immunomarker panel of GS, β-catenin, and SAA clearly has the potential for use in routine practice to guide management of HCAs.