Hepatocellular adenoma subtype classification using molecular markers and immunohistochemistry


  • Potential conflict of interest: Nothing to report.


Hepatocellular adenomas (HCA) with activated β-catenin present a high risk of malignant transformation. To permit robust routine diagnosis to allow for HCA subtype classification, we searched new useful markers. We analyzed the expression of candidate genes by quantitative reverse transcription polymerase chain reaction QRT-PCR followed by immunohistochemistry to validate their specificity and sensitivity according to hepatocyte nuclear factor 1 alpha (HNF1α) and β-catenin mutations as well as inflammatory phenotype. Quantitative RT-PCR showed that FABP1 (liver fatty acid binding protein) and UGT2B7 were downregulated in HNF1α-inactivated HCA (P ≤ 0.0002); GLUL (glutamine synthetase) and GPR49 overexpression were associated with β-catenin–activating mutations (P ≤ 0.0005), and SAA2 (serum amyloid A2) and CRP (C-reactive protein) were upregulated in inflammatory HCA (P = 0.0001). Immunohistochemistry validation confirmed that the absence of liver-fatty acid binding protein (L-FABP) expression rightly indicated HNF1α mutation (100% sensitivity and specificity), the combination of glutamine synthetase overexpression and nuclear β-catenin staining were excellent predictors of β-catenin–activating mutation (85% sensitivity, 100% specificity), and SAA hepatocytic staining was ideal to classify inflammatory HCA (91% sensitivity and specificity). Finally, a series of 93 HCA was unambiguously classified using our 4 validated immunohistochemical markers. Importantly, new associations were revealed for inflammatory HCA defined by SAA staining with frequent hemorrhages (P = 0.003), telangiectatic phenotype (P < 0.001), high body mass index, and alcohol intake (P ≤ 0.04). Previously described associations were confirmed and in particular the significant association between β-catenin–activated HCA and hepatocellular carcinomas (HCC) at diagnosis or during follow-up (P < 10−5). Conclusion: We refined HCA classification and its phenotypic correlations, providing a routine test to classify hepatocellular adenomas using simple and robust immunohistochemistry. (HEPATOLOGY 2007.)

Hepatocellular adenomas (HCA) are rare benign liver tumors, most frequently occurring in women who are using oral contraception. Although HCA are mostly found as a single nodule, the presence of more than 10 in the liver indicates a specific nosological entity termed liver adenomatosis.1 Two genetic alterations, the biallelic inactivation of hepatocyte nuclear factor 1 alpha (HNF1α) and the activating mutation of β-catenin, have been described in HCA.2, 3 Recently, a comprehensive analysis of genetic, pathological, and clinical features in a series of 96 HCA enabled the identification of 4 HCA subtypes.4 Biallelic HNF1α mutations defined the first group of HCA, phenotypically characterized by marked steatosis, lack of cytological abnormalities, and inflammatory infiltrates. Presence of a β-catenin–activating mutation defined the second group of HCA representing 15% of the cases generally characterized by a higher risk of malignant transformation in hepatocellular carcinoma (HCC). The third group of HCA was defined by the presence of inflammatory infiltrates and showed more or less obvious additional features such as sinusoidal dilatation, dystrophic vessels, and ductular reaction. Finally, the fourth group of HCA included the lesions nonmutated for HNF1α or β-catenin and without inflammatory infiltrate. Using this classification, the lesions initially termed telangiectatic focal nodular hyperplasia5 and recently proposed as HCA subtype resembled inflammatory HCA.6, 7

Considering the high risk of malignant transformation related to β-catenin–activated HCA, our recent pathomolecular classification has clinical implications in patient management.4, 8 To simplify and optimize the use of this classification, we searched for new robust markers useful in routine diagnosis. For this purpose, we tested for the expression of candidate genes at the ribonucleic acid (RNA) level using quantitative reverse transcriptase polymerase chain reaction (QRT-PCR) and at the protein level using immunohistochemistry (IHC). We analyzed the expression level of these markers to test their specificity and sensitivity in HCA cases previously characterized as HNF1α-inactivated, β-catenin–activated, inflammatory including lesions with a telangiectatic phenotype. Four markers were validated using IHC, providing a valuable tool that was used to classify a whole monocentric series of 93 HCA samples.


AB, antibody; CI, confidence interval; CRP, C-reactive protein; FNH, focal nodular hyperplasia; GPC3, glypican 3; GS, glutamine synthetase; HCA, hepatocellular adenoma; HCC, hepatocellular carcinoma; HNF1α, hepatic nuclear factor 1 alpha; IHC, immunohistochemistry; L-FABP, liver-fatty acid binding protein; QRT-PCR, quantitative reverse transcription polymerase chain reaction; RNA, ribonucleic acid; SAA, serum amyloid A2.

Patients and Methods

Patients and Samples.

We used a whole series of 98 resected HCA collected prospectively at Bordeaux Hospital (France) from 1986 to 2006. We excluded 5 cases because they were extensively necrotic with no sufficient tumor tissue available. Finally, a total of 93 HCA provided a satisfactory sampling of fixed liver tissues after hepatectomy. Among them, 34 cases were previously reported as HCA4 or telangiectatic focal nodular hyperplasia in 13 cases.6 In 70 of 93 cases, a representative part of the principal nodule, as well as of the nontumor liver, was immediately frozen in liquid nitrogen and stored at −80°C until used for molecular studies. Clinical data for oral contraception, individual and familial medical history, circumstances of diagnosis, alcohol intake, smoking, body mass index, and associated diseases were systematically collected. All patients were recruited in accordance with French law and institutional ethical guidelines. The overall design of the study was approved by the ethical committee of Hospital Saint-Louis (Paris, France). Clinical and histological tumors characteristics are summarized in Table 1. Pathological reviewing was performed and quoted for the following features: steatosis, inflammatory infiltrate, dystrophic arteries, cytological abnormalities, and macroscopic and microscopic hemorrhage. In consequence, HCA cases were classified according to the pathomolecular classification taking into account the HNF1α or β-catenin mutations and the presence of inflammatory infiltrates as previously described.4 For comparison, 20 typical focal nodular hyperplasias (FNH) were also analyzed by QRT-PCR and IHC.

Table 1. Correlation Between Immunohistochemical Classification and Clinical Features
Patient Features (n = 93 cases)L-FABP Negative HCA (n = 31)β-Catenin Activated (n = 18)SAA Positive HCA and β-Catenin Negative (n = 37)Unclassified (n = 7)
  1. NOTE. P value obtained from Fisher exact test are indicated, NS: not significant.

Sex F (77), M (16)0.04 (F)0.007 (M)NSNS
Oral contraception Yes (67), no (7)NSNSNSNS
Age Median = 38 yNSNSNSNS
BMI >25 Yes (29), no (58)NSNS0.04 (yes)NS
Alcohol >40g/day Yes (8), no (65)0.03 (no)NS0.03 (yes)NS
Tobacco Yes (42), no (29)NSNSNSNS
Children Yes (49), no (13)NSNSNSNS
Abnormal GGT Up (38), normal (28)0.003 (N)NS0.03 (U)NS
Number of adenomas Unique (44), multiple (45)0.001 (M)NS0.04 (U)NS
Associated HCC Yes (6), no (87)<0.001 (no)<.0010.04 (no)NS
Cytological abnormalities Yes (27), no (66)<0.001 (no)<.001 (yes)NSNS
Telangiectasia Yes (36), no (57)<0.001 (no)NS<0.001 (yes)0.03 (no)
Macroscopic hemorrhage Yes (41), no (51)<0.001 (no)NS0.003 (yes)NS
Microscopic hemorrhage Yes (50), no (42)<0.001 (no)NS<0.001 (yes)NS
Steatotic HCA Yes (58), no (35)<0.0001 (yes)NS0.005 (no)NS

QRT-PCR Validation.

QRT-PCR was performed using TaqMan pre-designed assays and the ABI Prism 7900HT System (Applied Biosystems) as previously described.9 Briefly, all expression results of a gene were normalized to internal control ribosomal 18S. Expression results of a gene in a sample are expressed relative to the mean expression level of the corresponding gene in nontumor samples.

Immunohistochemistry Analyses.

All 93 cases included in this study were analyzed by IHC, including the 70 cases also studied by molecular biology. IHC was performed on paraffin section of 10% fixed tumor and nontumor tissue with the following 4 antibodies (AB): anti-liver-fatty acid binding protein (L-FABP) (polyclonal AB, Abcam, 1/50 dilution), anti–β-catenin (monoclonal mouse AB, BD Biosciences, 1/200 dilution), anti-glutamine synthetase (GS) (monoclonal mouse AB, BD Biosciences, 1/400 dilution), and anti-serum amyloid A (SAA) (monoclonal mouse AB, Dako, 1/50 dilution). For each immunohistochemical procedure, antigen retrieval was performed in citrate buffer, and detection was amplified by the Dako Envision system.

For L-FABP, the use of a commercial antibody was optimized for IHC; the staining was considered normal when the protein was expressed in tumor hepatocytes, as in normal hepatocytes of the surrounding tissue; otherwise, it was quoted negative, even if a few stained hepatocytes were scattered in the tumor. β-Catenin–positive immunostaining corresponded to nuclear and cytoplasmic expression, whatever the number of tumor-stained hepatocytes, always associated with a positive cytoplasmic overexpression of GS, either homogeneous or heterogeneous but on more than 50% of the tumor surface, as previously described10; this abnormal overexpression of GS contrasted with normal liver, in which the staining is always restricted to pericentral hepatocytes. Adenoma was considered positive for SAA when tumor hepatocytes expressed SAA in their cytoplasm, whatever the intensity of the staining. The positive staining was often rather strong and homogeneously distributed; occasionally it was patchy and/or faint. In most cases, a clearcut difference was seen between the periphery of the SAA-positive adenoma and the surrounding negative nontumoral parenchyma. Furthermore, the SAA staining often appeared diffuse in the cytoplasm of hepatocytes, but sometimes granular. By contrast, SAA staining is always negative on the other cellular types of the liver.

Finally, we use the polyclonal anti CD3 and monoclonal anti CD20 AB (both from Dako, 1/100 dilution) for characterization of T and B lymphocytes, respectively, in the inflammatory infiltrate of this adenoma type.

Statistical Analysis.

Statistical analysis was carried out using Stata 8.0 software (Stata Corp, College Station, TX). Qualitative and categorized quantitative variables were compared with each other in contingency tables using a chi-square statistic or Fisher's exact test. For quantitative variables, data were expressed as mean and its standard error. The differences between quantitative variables were evaluated with a t test or analysis of variance when the variances were similar, or with the Kruskall-Wallis test when a nonparametric test was requested. All reported P values were 2-tailed; a P value of less than 0.05 was considered statistically significant.


Validation of Classifying Markers Using QRT-PCR.

A series of 40 HCA with high-quality RNA was used to validate expression of markers relatively to the pathomolecular classification. Among these cases, 9 and 11 were mutated for HNF1α and β-catenin, respectively; 23 cases demonstrated an inflammatory infiltrate with a β-catenin mutation (6 cases) or without (17 cases), and 3 HCAs were nonmutated without inflammatory infiltrate. We selected 2 genes, FABP1 and UGT2B7 coding for L-FABP and the uridine diphosphate glycosyltransferase 2B7, respectively. Expression of these genes are known to be positively regulated by HNF1α and highly expressed in normal liver tissues.11, 12 As expected, we found a low expression of these 2 genes using QRT-PCR in HNF1α-inactivated HCA when compared with HNF1α nonmutated subtypes (Fig. 1A) with a mean decrease of 87-fold and 260-fold for FABP1 and UGT2B7, respectively (P < 0.0001). The diagnostic value of FABP1 and UGT2B7 expression for the presence of HNF1α mutations was completely specific and sensitive as assessed by area under the receiver operating characteristics curve [standard error; confidence interval (CI) = 1 (0.00; 95% CI = 1)].

Figure 1.

Validation of markers using QRT-PCR. (A) Level of expression of FABP1 and UGT2B7 in HNF1α-mutated HCA (n = 9) compared with HNF1α-nonmutated HCA (n = 31). (B) Level of expression of GLUL and GPR49 in β-catenin mutated HCA (n = 11) compared with β-catenin nonmutated HCA (n = 29). (C) Level of expression of SAA2 and CRP in inflammatory HCA (n = 23) compared with noninflammatory HCA (n = 17). All results were normalized with the level of expression measured in 6 nontumor liver tissues.

Using the same series of 40 HCA, we analyzed 2 well-known genes targeted by β-catenin: GLUL coding for GS and GPR49, an orphan nuclear receptor10, 13, 14; their level of messenger RNA expression was for each transcript correlated with the presence of a β-catenin activating mutation (P = 0.0004 and P = 0.0001, respectively). The mean increase in β-catenin–mutated HCA was 19-fold and 34-fold when compared with non–β-catenin–mutated HCA, respectively (Fig. 1B). The diagnostic value of GLUL and GPR49 expression for the presence of β-catenin mutation was assessed by area under the receiver operating characteristics curve (standard error; CI = 0.87; 0.07; 95% CI = 0.72–1) and 0.92 (0.04; 95% CI = 0.84–1), respectively.

Finally, we tested the expression of 2 genes coding for major proteins of the acute-phase inflammatory response: SAA2 coding for the serum amyloid A2 and CRP coding for the C-reactive protein. Both transcripts were 9- and 22-fold upregulated in inflammatory HCA when compared with noninflammatory HCA, respectively (P < 0.0001, Fig. 1C). The diagnostic value of SAA2 and CRP expression for the presence of inflammatory infiltrates was assessed by area under the receiver operating characteristics curve (standard error; CI = 0.9; 0.05; 95% CI = 0.8–0.99) and 0.98 (0.02; 95% CI = 0.94–1), respectively.

QRT-PCR Expression Level Correlates With Protein Expression.

Among the 6 markers validated by QRT-PCR, we selected 3 of them specific for each subtype of HCA to test their expression at the protein level (L-FABP, GS, SAA). We also assayed for the expression of β-catenin that is accumulated in cytoplasm and nucleus when it is activated by mutation. Protein validation was performed using IHC in a series of 70 classified HCA that includes the 40 cases tested in QRT-PCR experiments. Among these 70 HCA, 25 were HNF1α-mutated, 13 showed an activating β-catenin mutation, 26 were non-mutated and inflammatory, and 6 were nonmutated and noninflammatory. No tumors harbored both HNF1α and β-catenin mutation.

L-FABP immunostaining analysis resulted in a constant cytoplasmic staining with variable intensity of all 70 tested nontumor livers. A complete absence of staining was observed in all 25 HNF1α-mutated HCA cases, contrasting with the nontumor surrounding tissue that was homogeneously immunostained (Fig. 2A, B) although sometimes faintly. The lack of L-FABP was obvious in nearly all tumoral hepatocytes, steatotic or not (Fig. 2C); only rarely few scattered isolated tumoral hepatocytes were stained, or corresponding sometimes to a few entrapped normal hepatocytes at the periphery (Fig. 2C). For the other HCA subtypes, L-FABP staining was comparable to the corresponding nontumor tissues (100% specificity and sensitivity).

Figure 2.

Validation of classifying markers using immunohistochemistry (x: original magnification). HNF-1α inactivated HCA: HES (A) and L-FABP immunostaining (B to D). (A) Typical aspect of severe steatotic adenoma (right) (original magnification, 40×); (B) absence of L-FABP expression in HCA (right), contrasting to the non-tumor liver (left) (original magnification, 40×); (C) Limit between nontumor liver (below) and HCA (top): L-FABP staining is lacking in steatotic as well as in nonsteatotic tumoral hepatocytes (original magnification, 100×); (D) Numerous microadenomas (stars) in the whole liver of a germline HNF1α-mutated adenomatosis, lacking L-FABP (original magnification, 20×); β-cateninactivated HCA. (E) Homogeneous GS staining of HCA (left), contrasting with nontumor liver (right), where only pericentral hepatocytes are stained (arrow) (original magnification, 40×); (F) Cytoplasmic and nuclear overexpression of β-catenin in HCA (left), whereas normal hepatocytes exhibited normal membranous staining (right) (original magnification, 200×). Telangiectatic/inflammatory HCA: HES (G) and SAA immunostaining (H). (G) Typical aspect of telangiectatic HCA (left) (original magnification, 40×); (H) SAA is strongly overexpressed in tumoral hepatocytes (left) (original magnification, 40×); whereas there is no SAA expression in inflammatory cells of the HCA (inset; original magnification, 200×); note the sharp limit between the stained HCA and the adjacent nontumoral liver (right).

Immunostaining analysis using antibodies directed against β-catenin and GS showed the following: Abnormal cytoplasmic and nuclear staining of β-catenin was observed in 11 of the 13 β-catenin–mutated cases (100% specificity, 85% sensibility, Fig. 2F). However, β-catenin staining was sometimes difficult to interpret because nuclear staining was only observed in few isolated hepatocytes (3 cases), and in the other 8 cases, positive nuclei were grouped in more or less large areas. In these cases, interpretation of the β-catenin staining was greatly facilitated using GS-positive immunostaining, a marker less specific but more sensitive (89% specificity, 100% sensibility). Indeed, a strong homogeneous (Fig. 2E) or heterogeneous cytoplasmic overexpression of GS was observed in all β-catenin–mutated cases and helped to focus the search for nuclear β-catenin immunostaining in the corresponding area. For nonactivated β-catenin cases, the normal membranous staining of hepatocytes was heterogeneous but similar in the tumor and nontumor liver. GS staining was generally absent in the adenoma, contrasting with the strong perivenular staining restricted to 1 or 2 plates of hepatocytes in the nontumor liver. However, irregular and less intense staining was observed around some venular structures in the adenoma as well as at its periphery.

Among the 70 HCA cases tested by molecular biology, 33 demonstrated inflammatory infiltrates, most frequently focal and limited to the HCA lesion. The inflammatory infiltrate was more or less important, most of the time located around arteries; it was polymorphous, usually composed of mixed T (CD3 positive) and B (CD20 positive) lymphocytes and histiocytes. Among these 33 inflammatory adenomas, 7 were mutated for β-catenin, and none were HNF1α-mutated. All the telangiectatic HCAs were inflammatory, and 6 were β-catenin mutated. In case of inflammatory adenomas, an obvious SAA immunostaining was observed in the tumor hepatocytes without any staining of the nonhepatocytic cells, that is, Kupffer cells or inflammatory cells inside the HCA (Fig. 2H). Furthermore, no reinforcement of the staining was observed in the tumor hepatocytes at proximity to the inflammatory infiltrates. The boundaries between the stained tumor and the nonstained surrounding nontumor liver were clearcut and corresponded to the limit of the adenoma. Although in 5 cases areas of hepatocytes were found positive in nontumor liver, SAA was obviously overexpressed in the corresponding HCA. Overall, SAA immunostaining was of 94% specificity and sensitivity when compared with the histopathological definition of an inflammatory adenoma. In noninflammatory cases, SAA staining was absent in the adenoma. Occasionally a faint granular staining was observed in small areas of nontumor liver.

Correlation Analysis Between Immunohistochemistry HCA Classification and Clinical and Pathological Features.

Because we obtained high specificity and sensitivity of the markers used in immunohistochemistry when compared with the pathomolecular classification, we analyzed an additional series of 23 HCA for which only formalin-fixed tissue was available, using L-FABP, SAA, GS, and β-catenin antibodies according to the criteria described previously. Therefore, the whole series of 93 HCA were classified as HNF1α-inactivated (L-FABP negative, 31 cases), β-catenin activated (nuclear β-catenin staining and glutamine synthetase overexpression, 18 cases), or inflammatory (SAA positive without β-catenin activation, 37 cases). Among these cases, 10 HCA were both β-catenin activated and inflammatory. The remaining 7 cases did not exhibit any immunohistochemical signature because they were L-FABP positive and β-catenin, GS, and SAA negative.

Using this immunohistochemical classification, we then searched for correlations with the clinical and histopathological features of the patients among the 93 immunohistochemically classified (Table 1). As previously described with the molecular classification,4 L-FABP–negative adenomas were severely steatotic (P < 10−6), without cytological abnormalities (P < 10−4) and inflammatory infiltrates (P < 10−10). The lack of steatosis is significantly associated with SAA-positive HCA when compared with all others (P = 0.005, Table 1); however, some of these inflammatory HCA exhibited steatosis, usually mildly intense. A β-catenin activation was associated with cytological abnormalities (P < 10−8); the occurrence of HCC (P < 10−5) and HCA developed in males were more frequently β-catenin activated (P = 0.007, Table 2). We also found new correlations: macroscopic and microscopic hemorrhages were more frequent in inflammatory HCA (P ≤ 0.003) and exceptional but severe in HNF1α- inactivated cases (P < 0.001); telangiectatic phenotype was highly related to SAA staining (P < 0.001) with or without β-catenin activation; high body mass index, alcohol intake, and elevated gamma glutamyltransferase in sera were also significantly associated with inflammatory HCA (P ≤ 0.04); adenoma nodules were more numerous in L-FABP–negative (P = 0.001). In this series, because L-FABP–negative HCA occurred more frequently in otherwise normal liver, the presence of steatosis in the nontumor liver (observed in 21 cases) or fibrosis (observed in 7 cases) was not related to a specific HCA subtype.

Table 2. Clinical Pathological Data of β-Catenin Adenoma Patients with HCC
IDSexHistoryCharacteristics of HCCOutcome
Age (HCC)Number of NodulesMax size (cm)SurgeryEdmondsonPathology
1F● Oral contraception (3 y)37 y>10 nodules (multicentric HCC)12Liver biopsyGrade 2Trabeculo-glandularDeath
● 24 y: Unique adenoma (17 cm), SAA+, incomplete resection
2F● 35 y: Cushing's syndrome47 y112Liver biopsyGrade 2Trabeculo-glandularDeath
● 39 y: liver adenoma 1.5 cm (liver biopsy)
● 40 y: bilateral adrenalectomy
● 42 y: liver adenoma 15 cm (right hepatectomy)
469F● Oral contraception (18 years)32 y118Right hepatectomyPart HCA Part HCC: Grade 3-4Macrotrabecular Pleiomorph.Rapid Multifocal recurrence Death
361F● Malgache origin66 y16SegmentectomyPart HCA Part HCC: Grade 2MacrotrabecularRecurrence 3 y later: surgery Alive (> 3 y)
● Contraception (intra muscular during 18 y)
● 40 y: uterine cancer (surgery & chemotherapy)
● Followed during 20 y for an hemangioma
● Type 2 diabetes (BMI: 29)
5M● Glycogenosis type 137 y10 HCA one HCC24 Tumorectomies (HCA and HCC)Grade 2Trabeculo-glandularAlive (> 4 y)
● adenomatosis (<10 HCA, largest 7.5 cm)
786M● Alcohol 60 g/ day60 y115Right hepatectomyGrade 1Trabecular (borderline tumor: HA/ very wd HCC)Alive (> 3 y)
● Tabacco. 45 pack year
● BMI: 30
● Followed during 12 y for hemangioma

Our series included 23 cases of adenomatosis (defined as more than 10 nodules identified in the liver by imaging). The principal analyzed adenoma was L-FABP negative or SAA positive (without β-catenin activation) in 15 and 4 cases, respectively. In these cases, when several adenomas were analyzed, they were of similar subtype in a defined patient. In contrast, the index analyzed nodule in the remaining 4 adenomatosis cases, which were β-catenin activated, was more heterogeneous. Two of them were patients with glycogenosis type 1; 1 was associated with an HCC; in these 2 cases, some nodules exhibited an abnormal nuclear immunostaining with β-catenin, others not. Finally, in 2 additional cases the principal nodule was both β-catenin activated and SAA positive, the other analyzed nodules demonstrated a variable pattern.

We also tested for specificity of our immunohistochemical markers regarding the most frequent type of benign hepatocellular tumors, that is, focal nodular hyperplasia (FNH). No cases of typical FNH (20 tested samples) exhibited a lack of L-FABP, a hepatocytic SAA staining, or a nuclear staining of β-catenin. SAA expression was negative in FNH even when the latter was associated in the same liver with 1 or several SAA-positive inflammatory HCA. Finally, all FNH demonstrated a heterogeneous staining of GS with a very characteristic pattern reinforced around hepatic veins without any nuclear or cytoplasmic β-catenin overexpression. This pattern of GS overexpression in FNH was completely different, however, from GS-positive staining found in β-catenin–mutated HCA. Finally, we have to note that in 4 cases among the 120 screened nontumor liver tissues, we found diffuse SAA overexpression.


In this study, we have identified 4 robust markers with expression at the RNA and protein level closely related to the genetic alterations and the pathological features previously identified in HCA.4 This panel of markers accurately classifies the different HCA subtypes with high specificity and sensitivity and can be easily assayed by either QRT-PCR experiments or using IHC analysis. Furthermore, the histochemical panel of 4 markers provides a significant refinement of the HCA pathomolecular classification with direct clinical relevance. Indeed, our findings seal the correlation between β-catenin activation and HCC risk of transformation. Here, we found that malignant transformation can occur and be detected at the time of the HCA diagnosis and also during follow-up, which echoes previous cases described.15–17 Our key finding concerns the now possible early and robust detection of these β-catenin–activated adenomas, which can be very difficult to differentiate from an already well-differentiated grade 1 HCC, as previously demonstrated.4 However, we have to stress the well-known difficulty of making the distinction between dysplastic HCA, exhibiting some cytonuclear abnormalities (corresponding mainly in our study to β-catenin–activated HCA) and well-differentiated HCC. This point is particularly important for β-catenin–activated HCA without overt associated HCC. Glypican 3 (GPC3) expression was higher in small HCCs than in cirrhosis and other benign lesions, including high-grade dysplastic nodules and HCA, indicating that the transition from premalignant lesions to small HCC is associated with a sharp increase of GPC3 expression in most cases.18–21 In the current series, 10 cases of β-catenin–activated HCA without HCC were negative for GPC3 staining (data not shown). However, considering that only 64% of HCC developed on normal liver were previously shown to be positively stained by GPC3,22 we need to identify new markers of malignant transformation of the β-catenin–activated adenomas and therefore precisely understand the malignant transformation of this type of HCA.

Furthermore, we found that SAA overexpression in inflammatory adenomas included all cases of lesions previously classified as telangiectatic FNH. This new observation demonstrates a common molecular link between telangiectatic FNH and inflammatory HCA. Therefore, in addition to the common phenotypic features previously described, it establishes that so-called telangiectatic FNH is in fact part of the inflammatory subtype of HCA with telangiectatic features. In these inflammatory tumors, we also identified a strong overexpression of the CRP transcript, suggesting that the observed acute inflammatory response could be aberrantly activated. Moreover, IHC analysis of SAA expression showed that the protein only accumulates in the tumor hepatocytes without reinforcement of the staining in inflammatory cells, nor in hepatocytes around the infiltrates. This observation suggests that deregulation of the acute phase pathway identified in HCA with inflammatory infiltrates occurs specifically in adenomatous hepatocytes, probably related to their own mechanism of tumorigenesis, thus suggesting that infiltrates by inflammatory cells might be a secondary event. Consistent with the observed acute-phase induction of SAA and C-reactive protein (CRP) proteins in the tumors, several patients harboring an inflammatory HCA demonstrated a systemic biological inflammatory syndrome that in some cases have regressed after curative hepatectomy.23 In this series, we did not measure systematically CRP or SAA in blood. In our recent experience, however, we found a raised CRP level in several patients with inflammatory adenomas. Consequently, a raised CRP level in a young woman with a benign liver nodule, if not explained by other reasons, is probably a good indicator of an inflammatory adenoma. Finally, inflammatory HCA also can be β-catenin activated. In our series, 2 β-catenin–activated HCA exhibiting an overexpression of SAA were associated with an HCC (concomitant or occurring 12 years later in 32-year-old and 37-year-old women, respectively). Larger studies are needed to determine whether the β-catenin–activated HCA with or without an inflammatory pattern have the same potential of malignant transformation.

In the current work, we found that inflammatory HCAs were more frequently developed in patients with a high body mass index and excessive alcohol consumption. These results suggest that alcohol consumption and obesity could have a direct role in the initiation of tumorigenesis of inflammatory HCA. Conversely, in our series, development of HCA in women is associated with oral contraception whatever the subtype. One hypothesis is that obesity and alcohol intake may be additional risk factors specifically for the development of inflammatory HCA and contribute to maintain the number of HCA diagnosed each year in Bordeaux, despite new oral contraceptives containing less estrogen and progesterone. Large multicentric and international studies are in progress to determine this point together with an evaluation of new genetic alterations such as CYP1B1 mutations predisposing to HCA.24

We also found 16 HCAs in men (17% of the cases), and these cases were frequently associated with specific known risk factors: glycogenosis in 2 cases,25, 26 familial polyposis coli,27, 28 androgen therapy for Fanconi's anemia,29 portacaval shunt,30 and HNF1α germline mutation.2, 31 Among the remaining 10 cases, 7 patients showed an inflammatory HCA with excessive alcohol consumption and excess body weight. However, steatosis was found in nontumor liver of 8 male patients, one demonstrated nonalcoholic steatohepatitis, and a case of adenomatosis in nonalcoholic steatohepatitis also has been reported.32 Altogether, HCA in men have a less frequent, specific association with rare diseases and metabolic disorders.

With our analysis, we have now refined the tools to better characterize HCA subtypes. Particularly, in HNF1α-mutated adenomatosis, L-FABP immunostaining highlighted the numerous, nearly coalescent microadenomas (Fig. 2D). Whether our panel of markers will also be accurate on biopsy samples remains to be determined, because the main issue remains the identification of β-catenin activated adenomas, considering their high risk of malignant transformation. However, immunostaining on a tiny biopsy has to be carefully and, most importantly, accurately interpreted, taking into account possible focal staining in tumor lesions. From a clinical perspective, we can anticipate the future potential identification of the 2 major HCA subtypes (HNF1α-inactivated and inflammatory) using a combination of clinical history, peripheral biology (gamma glutamyltransferase; inflammatory markers), and imaging (steatosis). Interestingly, when different nodules have been tested in the same patient presenting multiple nodules, they were of the same subtype in case of L-FABP–negative or inflammatory HCA. However, the occurrence of β-catenin activation was not found in all nodules. This observation suggests that β-catenin activation could be a secondary alteration occurring during adenoma tumorigenesis that strongly promotes malignant transformation. In the case of multiple lesions, these results enlightened the need to analyze the largest number of nodules to identify those with a β-catenin activation.

In conclusion, our immunohistochemically based classification of HCA closely correlates and refines the established pathomolecular classification.4, 8 Our panel of 4 antibodies should be a useful complement to routine histology used to establish HCA diagnosis and prognosis. An extended prospective study may be warranted to validate our data further in a routine clinical setting.


The authors thank Emmanuelle Jeannot and Lucille Mellottee for their help in mutation screening and excellent technical assistance. We also thank Philippe Bois for critical reading of the manuscript.