Hepatocellular Carcinoma: Making Sense of Morphological Heterogeneity, Growth Patterns, and Subtypes

Abstract

Hepatocellular carcinomas are not a homogenous group of tumors, but have multiple layers of heterogeneity. This heterogeneity has been studied for many years with the goal to individualize care for patients and has led to the identification of numerous hepatocellular carcinoma subtypes, defined by morphology and or molecular methods. This article reviews both gross and histological levels of heterogeneity within hepatocellular carcinoma, with a focus on histological findings, reviewing how different levels of histological heterogeneity are used as building blocks to construct morphological hepatocellular carcinoma subtypes. The current best-practice for defining a morphological subtype is outlined. Then, the definition for thirteen distinct hepatocellular carcinoma subtypes is reviewed. For each of these subtypes, unresolved issues regarding their definitions are highlighted, including recommendations for these problematic areas. Finally, three methods for improving the research on hepatocellular carcinoma subtypes are proposed: 1. Use a systemic, rigorous approach for defining hepatocellular carcinomas subtypes (four point model); 2. Once a definitions for a subtype is established, it should be followed in research studies, as this common denominator enhances the ability to compare results between studies; 3. Studies of subtypes will be more effective when morphological and molecular results are used in synergistic and iterative study designs where the results of one approach are used to refine and sharpen the results of the other. These and related efforts to better understand heterogeneity within hepatocellular carcinoma are the most promising avenue for improving patient care by individualizing patient care.

1. Introduction

Historically, hepatocellular carcinomas were largely thought of as a single homogenous tumor, but over time many different layers of heterogeneity were recognized at the levels of etiology, clinical presentation, the degree of fibrosis in the background liver, radiology findings, H&E morphology, molecular studies, and outcomes [1]. This heterogeneity is important to understand because it potentially informs all key aspects of patient care, including diagnosis, treatment, and prognosis, as well providing insight into the pathogenesis of hepatocellular carcinoma, which may in turn unlock even better treatments. Despite decades of progress in many areas, particularly with identifying the most common etiologies for hepatocellular carcinoma and in improving diagnostic tools (laboratory testing, radiology, histology), the complexities within these layers of heterogeneity have been daunting to sort out in order to maximize individual patient care. This article focuses on heterogeneity from a histological perspective, reviewing the current state-of-the-art understandings and discussing how histological heterogeneity is used as building blocks to construct ever more meaningful hepatocellular carcinoma subtypes.

2. Macroscopic heterogeneity

Small hepatocellular cellular carcinomas are defined by a diameter less than 2 cm. They have been classified into two major subtypes based on whether the hepatocellular carcinoma is well circumscribed and stands out from the background liver (distinctly nodular) or does not (vaguely nodular) [2]. The vaguely nodular pattern is thought to develop from dysplastic nodules and frequently has a few residual portal tracts. Vaguely nodular hepatocellular carcinomas are almost always seen in livers with advanced fibrosis/cirrhosis (>90%) and they tend to be well differentiated (>90%). They typically do not have a well-developed pseudocapsule, are hypovascular on imaging, and only rarely have vascular invasion (5%). In contrast, the distinctly nodular form of small hepatocellular carcinoma is more likely to have a pseudocapsule (50%), is more often moderately differentiated (60%), is hypervascular on imaging, and is more likely to have vascular invasion (40%) [2].

Additional macroscopic growth patterns have been described for hepatocellular carcinomas. The nodular type and the massive type grow as distinct nodules, with or without smaller satellite nodules. The nodular type can be of any size, while the massive type replaces almost an entire lobe of the liver [2]. Pedunculated hepatocellular carcinomas grow as mass lesions protruding from the capsule surface [3]. Diffuse hepatocellular carcinoma grows as a proliferation of small tumor nodules that have a similar size and shape to cirrhotic nodules (synonym is cirrhotomimetic). With diffuse hepatocellular carcinoma, imaging and gross examination frequently underestimate the extent of the tumor, as the tumor nodules closely mimic ordinary cirrhotic nodules. Of the different macroscopic growth patterns, only the diffuse pattern is routinely classified as a distinct hepatocellular carcinoma subtype.

3. Histological heterogeneity

Histological heterogeneity is important at the diagnostic level, as pathologists need to recognize when different morphologies are consistent with hepatocellular carcinoma and when they are not. Because of this important diagnostic need, pathologists identified and reported a wide variety of histological variations in the morphology of hepatocellular carcinoma. They also developed robust immunostain approaches to confirm a diagnosis of hepatocellular carcinoma across a wide spectrum of morphologies [1].

Over time, many of the individual findings that make up histological heterogeneity have been used as building blocks to construct specific hepatocellular carcinoma subtypes; this latter usage will be emphasized in this review. These basic building blocks can be classified into broad categories of (1) cytological findings of individual tumor cells, (2) architectural growth patterns, or (3) changes within the tumor matrix, such as fibrosis or inflammation.

3.1. Cytological heterogeneity

The most common cytoplasmic findings are clear cell change, fatty change, bile production, pale bodies, Mallory Denk bodies, and hyaline bodies. Less common changes include lipofuscin pigment and giant cell transformation. Clear cell change and fatty change are key features in the clear cell and steatohepatitic variants of hepatocellular carcinoma respectively. Bile production is not specific for mutation status, but is more common in CTNNB1 mutated hepatocellular carcinomas [4, 5]. Likewise, pale bodies (Figure 1A) are more common in fibrolamellar carcinoma than conventional hepatocellular carcinoma but are not specific for fibrolamellar carcinoma [6]. Hyaline bodies (Figure 1B), but not Mallory Denk bodies (Figure 1C), have been linked to a worse prognosis [7].

Figure 1. Cytoplasmic inclusions in tumor cells.

Panel A. Pale bodies are more common in fibrolamellar carcinoma than in conventional hepatocellular carcinoma but are not specific, as shown in this image of pale bodies in a conventional hepatocellular carcinoma. Several of the pale bodies are indicated by arrows. Panel B. Hyaline bodies are round to oval, densely pink inclusions. Several of the hyaline bodies are indicated by arrows. Panel C. Mallory Denk bodies are in tumor cells are similar to those seen in non-tumor cells, composed of ropy to chunky brightly eosinophilic material. Several Mallory Denk bodies are indicated by arrows.

The amount of cytoplasm is also used as one of the features of fibrolamellar carcinoma, where tumor cells have abundant eosinophilic cytoplasm. Prominent nucleoli are a common feature of fibrolamellar carcinoma but are nonspecific and are not used as a feature for other subtypes of hepatocellular carcinoma.

Cytological features are also used to grade hepatocellular carcinomas, based on the amount of cytoplasm, nuclear irregularities, and nucleolar changes [8, 9]. Tumor grade is another layer of tumor heterogeneity, as about 20% of hepatocellular carcinomas will have multiple grades present in the same tumor, a phenomenon called grade progression. When grade progression is present, then the higher grade tends to drive prognosis [10].

3.2. Inflammation and fibrosis

Many hepatocellular carcinomas have mild, patchy, and nonspecific inflammatory changes. When the inflammation is striking, however, this finding is used to identify lymphocyte rich hepatocellular carcinoma and granulocyte colony stimulating factor (GCSF) producing hepatocellular carcinoma. In addition to subclassification of hepatocellular carcinoma, there also currently is considerable interest in how the type and degree of inflammation within hepatocellular carcinoma is related to response to check point inhibitors.

Abundant intratumoral fibrosis is evident in both fibrolamellar carcinoma and in scirrhous hepatocellular carcinoma. Larger bands of fibrosis can also create distinctive nodularity within some tumors. Steatohepatitic hepatocellular carcinoma often have intratumoral pericellular fibrosis.

3.3. Microscopic growth patterns

There are four principal microscopic growth patterns (Figure 2A–D): trabecular (approximately 70% of tumors have this as the predominant growth pattern), solid (20%; synonym is compact), pseudoglandular (10%; synonym is pseudoacinar), and macrotrabecular (1%). All of these growth patterns are defined by their H&E appearance and most hepatocellular carcinomas show a mixture of these patterns. The frequency for these patterns varies considerably in different studies because so many tumors have mixed patterns and the patterns tend to run along a continuum, especially between solid and trabecular patterns. Any of these patterns can be seen with any of the different hepatocellular carcinoma subtypes.

Figure 2. Growth patterns in hepatocellular carcinomas.

These growth patterns can be seen in any subtype. Panel A. A solid growth pattern is shown. Panel B. A trabecular growth pattern is illustrated. Focal bile is also present. Panel C. This tumor has a pseudogland growth pattern. Panel D. A macrotrabecular growth pattern is seen.

The solid growth pattern (Figure 2A) has tumor cells growing in broad sheets without well-defined trabeculae readily visible at low and medium power (eg 4X, 10X, 20X). At higher power (eg 40X), however, areas of trabecular growth are often still discernable. The trabecular growth pattern (Figure 2B) has tumor cells growing in thin or moderately thick trabeculae, which should be readily evident at low and medium power. The trabecular and solid growth patterns often co-occur, where they can blend almost imperceptibly into each other. The pseudogland pattern (Figure 2C) often co-occurs with bile production and shows dilated spaces surrounded by tumor cells. The macrotrabecular growth pattern (Figure 2D) shows trabeculae that are at least 10 cells in thickness (other criteria are also used such as >6). The macrotrabecular growth pattern is used as the key feature for a proposed variant called macrotrabecular massive hepatocellular carcinoma [11], but is also found in wide range of other subtypes of hepatocellular carcinoma, leading to classification challenges that are discussed further below.

3.4. Other layers of heterogeneity

Some hepatocellular carcinomas grow as distinct single nodules of tumor, but are further subdivided into many smaller nodules by bands of fibrosis, a finding that has been called fibronodular hepatocellular carcinoma [12]. Distinctive radiological findings have also been reported for fibronodular hepatocellular carcinoma [12]. In some hepatocellular carcinomas, the tumor sinusoids can be dilated and peliotic [13]. Additional but uncommon findings include foamy macrophages within the tumor sinusoids [9] and focal myxoid changes [14].

3.5. Clonal progression

Clonal progression is defined as the presence of a distinct nodule of tumor that is morphological different than the rest of the tumor (Figure 3). When small, the nodule of clonally progressed tumor is typically clearly located within the lower grade tumor, but over time it can largely replace the original tumor clone, leaving behind only a thin discontinuous rim, or sometimes a smaller peripheral nodule, of the original tumor clone [2]. The nodules of clonally progressed tumor are typically less well differentiated and have a higher proliferative rate when compared to the background lower grade tumor morphology. In addition, a nodule of progressed hepatocellular carcinoma can subsequently develop additional levels of clonal progression, although more than 3 totals levels is uncommon. This aspect of hepatocellular carcinoma represents a major hurdle to accurate subtyping, as areas of clonally progressed carcinoma often have a different morphology, leading to challenges for tumor classification. Clonal progression should be distinguished from grade progression. Grade progression encompasses both clonal progression as well as tumor that are morphologically homogenous, other than for changes in tumor grade.

Figure 3. Clonal progression.

This case shows a background steatohepatitic hepatocellular carcinoma (right side of image) with a nodule of poorly differentiated carcinoma that has lost the typical findings of steatohepatitic hepatocellular carcinoma (left side of image).

4. Definition of a morphological subtype

The current definition of a fully established morphological subtype includes four key components [15] as detailed below. The terms subtype and variant are interchangeable.

1. Distinctive histological findings that are consistently present. This criterion is typically the first to be recognized by pathologists. The next 3 criteria improve specificity and add clinical meaning.

2. Immunohistochemistry or other confirmatory testing. This criterion significantly improves specificity, making the subtype more meaningful for clinical care and research studies.

3. Clinical correlates. This criterion makes the subtype clinically meaningful.

4. Unique molecular findings. This criterion provides insight into potential treatment targets and can feed back into criterion 2, serving as the basis for molecular or immunohistochemically based confirmatory tests.

One important practical matter is that these criteria will be developed over a period of time and it is not expected that every element will be present when a subtype is first described. Nonetheless, it is anticipated that true subtypes will develop each of these criteria over time. A natural corollary is that proposed histological subtypes should be abandoned if data clearly shows a criterion cannot be met.

5. Hepatocellular carcinoma subtypes

Overall, about 70% of all hepatocellular carcinomas can be subclassified into morphological subtypes. At this time, there are 13 fairly well-established hepatocellular carcinoma subtypes (Table 1). In the sections below, a brief morphological definition is provided for each subtype, along with ongoing challenges to their current definitions. These problematic areas await future studies for their resolution, but interim recommendations are made. The strength of the recommendations is classified as shown in Table 2. Key immunostain findings, clinical correlates, and molecular correlates are also provided.

Table 1.

Hepatocellular Carcinoma Subtypes by order of frequency.

Subtype	Frequency	Prognosis*	Unique histology	Confirmatory immuno-histochemistry or FISH	Unique clinical findings**	Unique molecular findings***
Steatohepatitic	20%	similar	Yes	No	Risk factor of fatty liver disease	No
Macrotrabecular massive	10%	Worse	Partial	No	Elevated serum AFP	No
Clear cell	7%	better	Yes	No	No	No
Scirrhous	5%	similar to better	Yes	No	No	No
Chromophobe	5%	Similar	Yes	Yes	No	Yes
Cirrhotomimetic	1%	worse	Partial	No	No	No
Fibrolamellar carcinoma	1%	Similar to better	Yes	Yes	Young age, no underlying liver disease	Yes
Granulocyte-colony-stimulating-factor producing	<1%	worse	Yes	Yes	Elevated serum white blood cell counts, IL6, CRP	No
Lymphocyte rich	<1%	better	Yes	No	No	No
Lymphoepithelioma hepatocellular carcinoma	<1%	Insufficient data	Yes	No	No	No
Combined tumors						No
hepatocellular-cholangiocarcinoma	1%	worse	Yes	Yes	Both AFP and CA19–9 can be elevated; Higher risk for LN metastases	No
hepatocellular and neuroendocrine	<1%	worse	Yes	Yes	No	No
Carcinosarcoma	<1%	worse	Yes	No	No	No

^*Compared to conventional hepatocellular carcinoma

^**Other than prognosis

^***Sufficiently unique to help define the subtype

Table 2.

Strength of Recommendation

Grade	Recommendation criteria
A	Well-established consensus diagnostic approach has been established
B	Well-established consensus is lacking, but at least one high quality study specifically addresses the point using objective criteria
C	Well-established consensus is lacking and recommendation is based on expert opinion only, with limited or no objective data
D	Insufficient published data, no recommendation.

5.1. Steatohepatitic hepatocellular carcinoma

Morphological definition: As originally defined, the steatohepatitic variant of hepatocellular carcinoma should have features of steatohepatitis (Figure 4A0, with macrovesicular steatosis, at least mild intratumoral inflammation (neutrophilic or lymphocytic), balloon cells, Mallory Denk bodies, and intratumoral pericellular fibrosis. The cut-off for the percent of the tumor with these findings needed to qualify for the steatohepatitic variant was originally defined as 5% [16], but later moved to 50 % [17].

Figure 4. Most common subtypes of hepatocellular carcinoma.

Panel A. Steatohepatitic hepatocellular carcinomas show fat, inflammation, and often fibrosis. Panel B. Clear cell hepatocellular carcinomas have abundant clear cytoplasm. Most cases are well to moderately differentiated. Panel C. . Scirrhous hepatocellular carcinomas have abundant intratumoral fibrosis. They can closely mimic peripheral or small duct type cholangiocarcinoma. Panel D Macrotrabecular massive hepatocellular carcinoma. This hepatocellular carcinoma has small basophilic cells and a macrotrabecular growth pattern.

Definitional challenges:

1. The minimum percent of tumor that needs to show steatosis/steatohepatitis in order to qualify as the steatohepatitic variant remains to be defined based on objective criteria. Current recommendation: at least 50%. Strength of recommendation: C

2. The minimum amount of steatosis within the areas of steatosis/steatohepatitis has also not been defined by data driven studies. The study that first identified the steatohepatitic subtype of hepatocellular carcinoma reported that 36% of their cases had minimal steatosis [16], while a subsequent study indicated that all of their cases had at least moderate macrovesicular steatosis (ie greater than 33%) [17]. Current recommendation: at least moderate macrovesicular steatosis. Strength of recommendation C:

3. Many cases show only steatosis but not steatohepatitis. Do these cases also qualify as the steatohepatitic hepatocellular carcinoma? Current recommendation: Given that steatosis and steatohepatitis are a continuum, it seems reasonable to include both steatosis and steatohepatitis, but there is insufficient data to make a recommendation. Strength of recommendation D.

4. Most hepatocellular carcinomas with the steatohepatitic pattern are found in patients with the metabolic syndrome [17] or with alcohol use [18], where both the tumor and the background liver show fatty liver disease, but a small subgroup of cases do not have these risk factors [19]. Should these two be grouped together? Current recommendation: they should be considered as separate entities, suggested terms: classic steatohepatitic hepatocellular carcinoma / nonclassic steatohepatitic hepatocellular carcinoma. Strength of recommendation C:

Immunostain confirmatory tests: None have been developed. Trichrome stains can identify intratumoral fibrosis but are not necessary.

Key clinical correlations: Metabolic syndrome or alcohol use

Molecular correlations: In comparison to conventional hepatocellular carcinoma, there are less frequent CTNNB1, TP53, and TERT promoter mutations [11, 20], but more frequent activation of the IL6/JAK/STAT pathway [11]. These molecular observations are not yet strong enough to define this subtype.

5.2. Clear cell hepatocellular carcinoma

Morphological definition: The tumor cells have clear cell change that resembles the clear cell morphology seen in clear cell carcinomas of other organs (Figure 4B), such as clear cell renal carcinoma. The clear cell change should be definite and diffuse. The tumor grade is typically well differentiated or moderately differentiated.

Definitional challenges:

1. The minimum percent of tumor that needs to show clear cell change in order to qualify as the clear cell variant of hepatocellular carcinoma remains to be well defined based on objective criteria. Current recommendation: at least 50%, with a cut off of 90% giving a more pure group of tumors [9]. Strength of recommendation: C

2. Some cases show both clear cell change and macrovesicular steatosis. Current recommendation: classification should be based on the predominate pattern. Strength of recommendation: C

3. Some hepatocellular carcinomas with clear cell features have high grade cytology (Figure 4C); how should these be classified? At this point, there is no data on which to base a recommendation. Current recommendation: none, strength of recommendation: D

Immunostain confirmatory tests:None have been developed.

Key clinical correlations: Inconsistent, but general trend towards a better prognosis based on smaller size, better differentiation, and less vascular invasion [21–23].

Molecular correlations: Limited data. IDH1 mutations have been reported [24]. These molecular observations are not strong enough to define this subtype.

5.3. Scirrhous hepatocellular carcinoma

Morphological definition: The tumor should have abundant intratumoral fibrosis (Figure 4C). The fibrosis should be diffuse, with over 50% of the tumor showing dense intratumoral fibrosis [9, 25].

Definitional challenges:

1. Different studies have used different cut-offs for the percentage of the tumor that needs to have the scirrhous morphology. The best study on this question found an optimal cut-off of 50% [9, 25]. There are many cases with more focal (less than 50%) but sometimes striking intratumoral fibrosis; how should these be classified? Current recommendation: classify as per the predominant pattern and not as scirrhous hepatocellular carcinoma. Strength of recommendation: C

2. Some cases have a scirrhous morphology but show either clear cell change or macrovesicular steatosis, enough to qualify for clear cell hepatocellular carcinoma or steatohepatitic hepatocellular carcinoma respectively [25]. This challenging area represents a gap in knowledge that has not been addressed in the literature. Current recommendation: no recommendation, but it would not be unreasonable to classify by the predominant pattern until there is further clarification. Strength of recommendation: D

Immunostain confirmatory tests: None have been developed.

Key clinical correlations: none

Molecular correlations: TSC1/TSC2 mutations [11]. These molecular observations are not strong enough to define this subtype.

5.4. Macrotrabecular-massive hepatocellular carcinoma

Morphological definition: The tumor cells are growing in trabeculae at >6 cells in thickness (Figure 4D). There is no requirement for the tumor to be massive in overall size, despite the name.

Definitional challenges:

1. There have been multiple cut offs used to define macrotrabeculae growth pattern but the most common are >6 [26] and ≥10 in thickness [9], while earlier definitions also used ≥20 [27]. There is no objective data to favor one over the other. The lower threshold is likely to capture more cases but could also be less discriminating. Recommendation: Either the 6 or 10 cutoff is reasonable, with no data to favor one of the other. Strength of recommendation: C.

2. The distinction between the macrotrabecular growth pattern and macro-trabecular massive subtype is not well delineated. This is the most important challenge for this potential subtype. Should the definition be based on the percent of the tumor with a macrotrabecular growth pattern alone, or should the definition include cytological findings? The earliest proposals for this subtype, presented at meetings [28, 29] and published in book chapters [9, 30], advocated for a combined definition using a macrotrabecular growth pattern on H&E, plus tumor cells having a small and basophilic morphology, as this composite pattern was strongly associated with elevated serum AFP levels and extensive angiolymphatic invasion. Subsequent papers, however, used cut-offs that range from 30 to 50% for the macrotrabecular growth pattern, with no requirements for tumor cytology [31, 32]. Part of the definitional challenge is that the macrotrabecular growth pattern itself is an indicator of poor prognosis [33], making it hard to anchor the definition for the macrotrabecular massive subtype primarily to clinical outcomes. Based solely on morphological observations, it seems likely that criteria based only on architecture will lead to a less specific subtype, since the macrotrabecular growth pattern can be associated with a wide range of other morphological findings. Recommendation: No recommendation. Intuitively it seems reasonable to incorporate uniform cytological findings into the definition, but unfortunately this critical question remains unresolved, with no peer reviewed published data to serve as a guide. Strength of recommendation: D.

Immunostain confirmatory tests: CD34 immunostain are helpful to identify the macrotrabecular growth pattern, especially in areas that transition to a more solid morphology, as the macrotrabeculae are brought out by the CD34 stain [30]. Other stains, such as ESM1 [34] can do the same thing, but CD34 is more readily available.

Key clinical correlations: Elevated serum AFP levels. The strength of this correlation depends on the strictness of the definition.

Molecular correlations: Higher rates of TP53 mutations and FGFR3 amplification. These molecular observations are not strong enough to define this subtype.

5.5. Chromophobe hepatocellular carcinoma

Morphological definition: The tumor cells have moderately abundant to abundant amphophilic to eosinophilic cytoplasm with bland nuclei throughout most of the tumor, but punctuated by tumor cells with greater nuclear anaplasia (Figure 5A), often scattered in small discrete foci [15]. The tumor also frequently has scattered cystic spaces (Figure 5B) filled with a thin serum like substance, occasionally with hemorrhage. Chromophobe hepatocellular carcinoma are positive for alternative lengthening of telomeres (ALT) [109], which is a mechanism for tumors to maintain their telomeres without TERT promoter mutations or other TERT gene rearrangements. The mechanism appears to rely on homologous recombination of the telomeres. ALT can be detected by FISH or CISH on paraffin embedded tissues [35, 36].

Figure 5. Less common subtypes of hepatocellular carcinoma.

Panel A. Fibrolamellar carcinoma. Intatumoral bands of fibrosis can be seen at low power. Panel B. Fibrolamellar carcinoma cells have abundant eosinophilic cytoplasm and prominent nucleoli. In the center top of the image, a tumor cell has abortive bile canaliculi located within the cytoplasm, a finding that is common but no specific for fibrolamellar carcinoma. Panel C. Chromophobe hepatocellular carcinoma. Scattered larger anaplastic nuclei are seen scattered in background of generally low grade nuclei—the sharp distinction of the two cell populations is captured in the term sudden anaplasia Panel D. Chromophobe hepatocellular carcinoma. Large pseudocysts are seen at low power.

Definitional challenges:

1. Clear cell carcinomas can enter the differential. The cytoplasmic changes are distinct when viewing cases side by side, but can be challenging when viewed in isolation. Clear cell carcinomas lack the scattered foci of anaplasia, generally do not have areas of cyst formation, and are ALT negative. Strength of recommendation: C

Immunostain confirmatory tests: None

Key clinical correlations: None

Molecular correlations: Tumor are positive for ALT by either FISH or other molecular methods.

5.6. Fibrolamellar carcinoma

Morphological definition: The tumor cells have abundant cytoplasm and prominent nucleoli (Figure 5C). There is striking intratumoral fibrosis and often pseudoglands, pale bodies, and hyaline bodies. The fibrosis can be variable in its density, but the cytological findings are homogenous within the tumor [6].

Definitional challenges:

1. Some studies have proposed mixed tumors, tumors that are part fibrolamellar carcinoma and part hepatocellular carcinoma, or part fibrolamellar carcinoma with areas of neuroendocrine differentiation. These findings have not been substantiated in subsequent studies and recent studies have shown that at least some “mixed hepatocellular carcinomas—fibrolamellar carcinomas” are conventional hepatocellular carcinomas with BAP1 mutations [37]. Current recommendation: at this point, the best definition for fibrolamellar carcinoma is as above. At this point, there is no compelling evidence for a combined hepatocellular carcinoma-fibrolamellar carcinoma or for a combined fibrolamellar carcinoma--neuroendocrine carcinoma. Strength of recommendation: A

Immunostain confirmatory tests: Fibrolamellar carcinomas are positive for CK7 and CD68 [38]. FISH testing or PCR is useful to identify the DNAJB1-PRKACA fusion gene [39].

Key clinical correlations: Fibrolamellar carcinomas occur in younger patients (average age of 26 years) [6] with no underlying liver disease and patients have more frequent metastatic disease to hilar lymph nodes than conventional hepatocellular carcinoma.

Molecular correlations: Protein kinase A is activated, in most cases (>99%) by a fusion gene DNAJB1-PRKACA created by a microdeletion on chromosome 19 [40]. Less common mechanisms include biallelic mutations in PRKAR1A [41]. The molecular findings are strong enough to define this tumor, when used in association with morphology that is compatible with fibrolamellar carcinoma. The same fusion is found in other tumors [42], so the diagnosis should not be based solely on molecular findings.

5.7. Diffuse hepatocellular carcinoma

Morphological definition: The tumor cells grow in small nodules that are about the same size and shape as that of the cirrhotic liver, making them difficult to identify by radiology and sometimes by gross examination. The definition also allows that the small nodules in some cases may coalesce to form a larger tumor mass that is visible on imaging and gross examination, but, by definition, there should be substantial tumor burden with the cirrhosis-like growth pattern. Additional terms include cirrhotomimmetic and cirrhosis-like hepatocellular carcinoma.

Definitional challenges:

1. At times, this subtype has inappropriately been applied to cases where a small number of tumor satellite nodules were not evident on imaging, but identified by microscopic examination. The definition, however, requires that substantial numbers of tumor nodules were not evident by imaging. Current recommendation: No definite cut-off has been established, but in most cases there are 30 or more small tumor nodules that were not present on imaging. Strength of recommendation: C

2. An important question is whether this represents a true hepatocellular carcinoma subtype versus a growth pattern that can rarely be seen with any subtype of hepatocellular carcinoma. The cytological findings are not consistent, with most cases showing a not-otherwise specific morphology, but some cases showing a steatohepatitic or clear cell morphology [43]. This observation of inconsistent morphology argues against a true subtype, yet the growth pattern is so distinctive and so uncommon that some sort of distinguishing terminology seems warranted. When enough cases are put together for an in-depth analysis, it may be that a morphologically homogenous group emerges. If studies show this is not the case, then this subtype can be reclassified as a gross pattern only. For the time being, it seems reasonable to keep it as a subtype. Strength of recommendation: C

Immunostain confirmatory tests: None

Key clinical correlations: None

Molecular correlations: None

5.8. Lymphocyte rich hepatocellular carcinoma

Morphological definition: The tumor should have abundant intratumoral lymphocytes. The lymphocytosis should be diffuse (Figure 6A), involving over 50% of the tumor. Within the areas of lymphocytosis, the number of lymphocytes should exceed the number of tumor cells on H&E. Older terms no longer in use include hepatocellular carcinoma with lymphoid rich stroma and inflammatory hepatocellular carcinoma.

Figure 6. Rare subtypes of hepatocellular carcinoma.

Panel A. Diffuse hepatocellular carcinoma. The tumor nodules are small and round, about the same size as an ordinary cirrhotic nodule. Panel B. Lymphocyte rich hepatocellular carcinoma. There is striking and diffuse lymphocytosis in the tumor sinusoids Panel C. GCSF producing hepatocellular carcinoma. The tumor sinusoids are markedly distended by neutrophils.

Definitional challenges:

1. Some hepatocellular carcinomas have noticeable lymphocytosis, but the findings do not reach the above criteria, either because the inflammation is intense but localized, or because it is mild but diffuse. Current recommendation: classify as per the predominant pattern, independently of the lymphocytic inflammation, for those cases where the findings to not reach the level of lymphocyte rich hepatocellular carcinoma. Strength of recommendation: C

2. Should lymphocyte rich hepatocellular carcinoma be separated from lymphoepithelioma like hepatocellular carcinoma? Current recommendation: Yes, they should be classified as separate tumors. Currently, lymphoepithelioma like hepatocellular carcinoma and lymphocyte rich hepatocellular carcinoma are often used as synonyms. There is a difference, however, in the morphology. Lymphoepithelioma like hepatocellular carcinoma is defined as a poorly differentiated carcinoma growing in ill-defined sheets that closely resembles lymphoepithelioma like carcinomas from other organs. In contrast, lymphocyte rich hepatocellular carcinomas are usually well to moderately differentiated. Strength of recommendation: C

Immunostain confirmatory tests: None have been developed.

Key clinical correlations: None

Molecular correlations: Tumors are EBV negative; in contrast to some lymphocyte rich carcinomas from other organs [44]. There is no evidence for microsatellite instability [44]. Amplifications of chromosome 11q13.3 have been reported [45]. These molecular observations are not strong enough to define this subtype.

5.9. Lymphoepithelioma like hepatocellular carcinoma

Morphological definition: Lymphoepithelioma like hepatocellular carcinoma is a poorly differentiated carcinoma growing in ill-defined sheets in a background of striking lymphoplasmacytic inflammation. The H&E findings should be indistinguishable from lymphoepithelioma like carcinomas arising in other organs, so metastatic disease should be excluded.

1. Definitional challenges: It remains an unresolved question whether these tumors represent de-differentiated lymphocyte rich hepatocellular carcinomas. Current recommendation: There is no convincing published evidence that lymphocyte rich hepatocellular carcinomas progress to lymphoepithelioma like hepatocellular carcinomas, but both tumors are rare and data is limited. Given the lack of evidence that they are related, the current recommendation is that they should be classified separately. Strength of recommendation: C

Immunostain confirmatory tests: None have been developed.

Key clinical correlations: None

Molecular correlations: None to date. Tumors are EBV negative; in contrast to some lymphoepithelioma like carcinomas from other organs.

5.10. Granulocyte colony stimulating factor (GCSF) producing hepatocellular carcinoma

Morphological definition: The tumor should have abundant intratumoral neutrophils (Figure 6B). Most are poorly differentiated hepatocellular carcinomas, often with sarcomatoid areas. Other organs also have GCSF producing carcinomas, which can metastasize to the liver. In addition, rare cholangiocarcinomas produce GCSF. For these reasons, hepatic differentiation should be demonstrated by immunohistochemistry. Correlation with imaging findings is important to confirm a liver primary.

Definitional challenges:

1. Is this a definite subtype of hepatocellular carcinoma, versus a heterogeneous group of poorly differentiated hepatocellular carcinomas that aberrantly produce GCSF? Current recommendation: Morphological and molecular studies are limited largely to case reports, so it remains unclear if this a homogenous or heterogeneous group of tumors. Based on the unique clinical and molecular findings, it seems reasonable to study them as a separate hepatocellular carcinoma variant. Strength of recommendation: C

Immunostain confirmatory tests: GCSF immunohistochemistry can be positive in tumor cells [46, 47].

Key clinical correlations: Patients have elevated peripheral white blood cell counts, serum IL-6 levels, and often serum CRP [47–49].

Molecular correlations: Tumor cells produce GCSF.

5.11. Mixed tumors

There are 3 types of mixed tumors: combined hepatocellular carcinoma—cholangiocarcinoma, carcinosarcoma, and combined hepatocellular carcinoma—neuroendocrine carcinoma. In each subtype, the hepatocellular component should look and stain like hepatocellular carcinoma, while the additional component of malignancy should likewise look and stain appropriately. For combined hepatocellular carcinoma-cholangiocarcinoma and for carcinosarcoma, the two morphologically distinct components of the malignancy are typically in separate nodules by light microscopy, often with a narrow “transition” zone at their interface where the two morphologies are more intermixed. Widely intermixed morphologies are rare and the diagnosed of mixed tumor in this setting should be carefully considered. Aberrant expression of immunostain markers in an otherwise homogenous tumor does not support a diagnosis of a mixed tumor. The two components should be within a single tumor, in contrast to cases with two spatially separate and independent primaries.

5.11.1. Combined hepatocellular carcinoma-cholangiocarcinoma

Morphological definition: A single tumor has two cell populations that are distinct on H&E, one of hepatocellular carcinoma and one of cholangiocarcinoma (Figure 7A). Each component should be confirmed by immunostains, with the hepatocellular carcinoma component but not the cholangiocarcinoma positive for hepatocellular markers such as HepPar1 or arginase, and vice versa for biliary markers.

Figure 7. Mixed subtypes of hepatocellular carcinoma.

Panel A. Combined hepatocellular carcinoma-cholangiocarcinoma, with two distinct population of cells. Panel B. Carcinosarcoma, with two distinct populations of cells. Panel C. Mixed hepatocellular carcinoma-neuroendocrine carcinoma, with two distinct populations of cells. The small cell carcinoma is growing in small clusters in the tumor sinusoids, highlighted by arrows.

Definitional challenges:

1. The best definition for many years was not clear. Thus, in some papers, hepatocellular carcinomas with a homogenous morphology that showed aberrant expression of CK19 or CK7 or other keratins were classified as mixed tumors. Less commonly, morphologically homogenous cholangiocarcinomas with aberrant expression of hepatocellular markers such as HepPar1 were sometimes mistaken as mixed tumors. As this subtype matured over the years, a uniform definition developed, one that is now agreed upon by expert review articles [50, 51], authoritative text books [9, 52], and a consensus article [53]. Current recommendation: the best definition for combined hepatocellular carcinoma-cholangiocarcinoma is as above. Strength of recommendation: A

Immunostain confirmatory tests: Hepatocellular markers and biliary markers are used to confirm the morphology, as noted above. CK19, CK7, CKAE1/3 are not specific for biliary differentiation but are helpful when they are positive only in the areas of morphological cholangiocarcinoma.

Key clinical correlations: Both serum AFP and serum CA19–9 levels are elevated in about 1/3 of patients [54]. Metastatic disease to hilar lymph nodes (13%) is more common than conventional hepatocellular carcinoma (2%) and less common than conventional cholangiocarcinoma (21%) [55]. The overall prognosis is in between that of conventional hepatocellular carcinoma and cholangiocarcinoma [56, 57].

• Molecular correlations: None to date. Overall, the genetic mutations in both components are more similar to the genetic changes seen in hepatocellular carcinoma than cholangiocarcinoma [58, 59].

5.11.2. Carcinosarcoma

Morphological definition: Within the same tumor, a carcinosarcoma has areas of hepatocellular carcinoma and areas of sarcoma (Figure 7B). The identification of a mesenchymal component should be based on both morphology and immunohistochemistry. The hepatocellular carcinoma should stain appropriately with keratins and markers of hepatocellular differentiation, while the sarcoma component should not. The sarcoma can be more specifically classified, with reports of leiomyosarcoma, rhabdomyosarcoma, chondrosarcoma, fibrosarcoma, or osteosarcoma [48, 60, 61], but often is an undifferentiated spindle cell sarcoma.

Definitional challenges:

1. In some cases, the sarcoma component produces GCSF. Current recommendation: If the histological findings are consistent with carcinosarcoma, the tumor should be classified as such regardless of GCSF production. Strength of recommendation: C

2. Separating a carcinosarcoma from sarcomatoid carcinoma occasionally can be challenging. In carcinosarcomas, the sarcomatoid component is keratin negative and has immunostain evidence for mesenchymal differentiation. In some cases, however, this distinction remains problematic because there can be very focal or equivocal staining for keratins within the sarcomatoid areas, with no other morphological or immunostain evidence for more specific mesenchymal differentiation. In these cases, especially on biopsy, it is reasonable to provide a differential that includes both entities. For resection specimens, absent or equivocal keratin is best classified as carcinosarcoma. Other soft signs include the amount of the mesenchymal component: a small microscopic focus is more common in sarcomatoid carcinoma, while a large bulky component favors carcinosarcoma. The hepatocellular carcinoma component is most commonly moderately differentiated for both entities and does not provide specific clues to separate these two entities. Strength of these recommendations: C.

3. Immunostain confirmatory tests: Stains for hepatocellular differentiation in the epithelial component are important to rule out metastatic disease from carcinosarcomas arising in other organs.

Key clinical correlations: none

Molecular correlations: none

5.11.3. Combined hepatocellular carcinoma—neuroendocrine carcinoma

Morphological definition: Within the same tumor, there is an area of hepatocellular carcinoma and an area of neuroendocrine carcinoma, usually small cell carcinoma (Figure 7C). The hepatocellular carcinoma component is typically substantially larger than the neuroendocrine component, which often is located within tumor sinusoids.

Definitional challenges:

1. Some neuroendocrine tumors/carcinomas can have a hepatoid morphology and produce AFP [62]. These tumors, however, are histologically homogenous, lacking the biphasic morphology of the combined hepatocellular carcinoma—neuroendocrine carcinoma. A biphasic morphology that stains appropriately is required for the definition. Strength of recommendation: A

Immunostain confirmatory tests: Stains are positive for hepatocellular differentiation in the epithelial component and negative in the neuroendocrine carcinoma component, with stains for neuroendocrine differentiation showing the opposite.

Key clinical correlations: none

Molecular correlations: none

6. Additional morpho-molecular patterns

6.1. Beta catenin mutated hepatocellular carcinomas

Many hepatocellular carcinomas with CTNNB1 mutations are well differentiated tumors composed of eosinophilic tumor cells growing in thin trabeculae, often with pseudoglands [5, 63]. While there is a strong statistical correlation between CTNNB1 mutations and this morphology, there are many exceptions in both directions, such as tumors having a compatible morphology but lacking CTNNB1 mutations and tumors with CTNNB1 mutations that lack the typical morphology. It seems possible that a morphological subtype can be delineated in time, but the current findings are insufficient.

6.2. Sarcomatoid hepatocellular carcinoma

Sarcomatoid changes are defined as areas of poorly differentiated spindle cell growth within a hepatocellular carcinoma. The area of sarcomatoid change is keratin positive, although staining can be focal and broad-spectrum keratin stains may be needed. The sarcomatoid area can never be further subclassified by morphology or by immunostains into a more specific type of sarcoma, in contrast to some cases of carcinosarcoma.

6.3. BAP1 mutated and PKA activated hepatocellular carcinomas

A subgroup of hepatocellular carcinomas in older patients, often women without underlying liver disease, have histological findings that at least focally resemble fibrolamellar carcinoma. This group of tumors commonly show loss of BAP1 and PKA activation by gains of PRKACA (PKA catalytic subunit) and loss of PRKAR2A (PKA inhibitory subunit) [37]. With additional studies, it seems likely this group of tumors could represent a specific subtype.

7. Best practices going forward

Currently, morphological subtypes of hepatocellular carcinoma are important primarily to pathologists, as knowing about them improves the accuracy of histological diagnosis. With the exception of fibrolamellar carcinoma, combined hepatocellular carcinoma-cholangiocarcinoma, and combined mixed hepatocellular carcinoma-neuroendocrine carcinoma, however, none of these subtypes currently impact patients: they do not change clinical management. Below are three proposals to consider for enhancing progress in subtyping hepatocellular carcinomas with the goal of making the subtypes more clinically relevant and, by so doing, improve patient care.

1. Adopt a formal, systematic approach to establishing a hepatocellular carcinoma subtype. There will be greater progress in understanding hepatocellular carcinoma subtypes and in improving patient care when a shared definition of a subtype is used. Without the adoption of shared definitions, if authors use whatever criteria to define a subtype that they deem best for the purposes of their study, the field gets muddy very quickly. Current best practice recommendations are outlined above for determining a morphological-based subtype of hepatocellular carcinoma (4 point model: unique findings in histology, immunohistochemistry, clinical, and molecular). This 4 point model should be followed until a better one emerges.

2. Insist that standard definitions for each recognized subtype be used in academic studies. Each study should follow the criteria in standard definitions, such as those published in the AFIP Liver Tumor fascicle or the WHO blue book, which are further discussed above. In so doing, studies will have more value, allowing the results of one study to be more accurately compared to others. No definition is perfect and it is anticipated that authors will find ways to improve definitions. In the process, the performance of any newly proposed set of criteria should be directly compared in the same study to those used by the standard definition in order to demonstrate superiority.

3. More rigorous morphology, molecular correlative studies. The morphological findings in hepatocellular carcinoma subtypes have been exhaustively studied by light microscopy. Likewise, by now there have been thousands of hepatocellular carcinomas studied by detailed molecular methods, including gene expression analysis, methylation analysis, copy number variation, chromosomal changes, and point mutations. Over time, returns naturally diminishing for any scientific approach applied to the same question, including the methods of histology and molecular studies applied to the problem of hepatocellular carcinoma subclassification. Thus, it seems reasonable to postulate that sequencing another several thousand hepatocellular carcinomas (or gene expression analysis, or CNV analysis, etc) is unlikely to reveal substantially more insights than is currently available. In the same fashion, additional morphological studies of known variants are more likely than not to be largely confirmatory in nature and the literature benefits little from them if they don’t focus specifically on correcting a gap-in-knowledge, one that is actually of clinical or diagnostic relevance. Yet, it seems likely that considerable progress can be made in understanding hepatocellular carcinoma subtypes by using the synergies that open up in studies that effective use both tools.

One important potential path forward are studies that better integrate morphology and molecular findings, where the findings in one discipline are used to refine the classification based on the other discipline, and vice versa. To date, there have been many fine papers that have examined data sets for morphology-molecular correlations and they have made substantial contributions. Despite these efforts, a reproducible and clinically relevant subtyping approach remains elusive for hepatocellular carcinomas, when clinical relevance is defined as influencing clinical management. In addition, most of the studies on morphology-molecular correlations, are either primarily morphological based studies with post hoc analysis of molecular changes, or primarily molecular studies with post hoc analysis of morphological findings. Because of the complexity of both the morphology and the molecular findings, better integrated and iterative studies are needed to fully take advantage of potential synergies between morphology and molecular patterns.

In conclusion, defining clinically relevant subtypes is at the forefront of providing personalized care for patients with hepatocellular carcinoma. Morphological subtypes are based on a constellation of unique histological findings, confirmatory immunostains or molecular tests, unique clinical findings, and unique molecular findings. The best-practices for studies of hepatocellular carcinoma subtypes includes using standard definitions. Subtypes will be more clinically meaningful as molecular results are more effectively integrated into subtype definitions.