An Update on the Role of Immunohistochemistry in the Evaluation of Pancreatic/Liver/Gastrointestinal Luminal Tract Disorders

Geisinger Medical Laboratories (Danville, Pennsylvania) performed tissue microarray sections and routine sections stained with Dako and Ventana Systems on normal pancreatic tissue of 40 cases. The data revealed that both normal pancreatic ductal and acinar epithelial cells are IHC positive for CK7, CAM5.2, MOC-31, Ber-EP4, prostate stem cell antigen (PSCA), ATRX/DAXX, and DPC4/SMAD4. Normal pancreatic ductal epithelial cells also express MUC6, CK19 (focal), claudin 18 (focal), annexin A8 (weak), claudin4 (weak), MUC1 (weak, on luminal side), and mesothelin (weak). Expression of S100A, monoclonal carcinoembryonic antigen (mCEA), and CA19-9 in normal pancreatic ductal epithelial cells is usually absent or weakly positive. The weak positivity of mCEA is on the luminal side of normal pancreatic duct epithelial cells. Normal pancreatic acinar cells express BCL10 and trypsin; they can weakly express CA19-9 as well. Normal and reactive pancreatic ducts are usually negative for CK20, CK17, maspin, IMP3 (KOC), trypsin, MUC2, MUC4, and MUC5AC. Approximately 10% of pancreatic ducts and acini are focally positive for CDX2.

Autoimmune pancreatitis (AIP) is categorized into types 1 and 2. The former is mediated by immunoglobulin (Ig) G4–positive plasma cells, whereas the latter is characterized by neutrophilic destruction of pancreatic ductal cells. Accordingly, type 1 AIP lesions show dense distribution of IgG4-positive plasma cells, measured by tissue IgG4 counts (>50/high-power field for resection specimens and >10/high-power field for biopsy specimens) or IgG4:IgG ratios (>40%).¹  Neutrophils present in most AIP type II lesions are almost entirely absent in type I AIP.²  Recently, ductal programmed death ligand-1 (PD-L1) reactivity has been reported to have potential to distinguish type 2 AIP from other forms of pancreatitis and pancreatic ductal adenocarcinoma (PDAC). The sensitivity and specificity of PD-L1 as a marker of type 2 AIP were 70% and 99%, respectively.³ 

Differentiation of PDAC cells from reactive pancreatic ductal epithelial cells can sometimes be very challenging. Many markers, such as mCEA, CA19-9, MUC1, MUC5AC, SMAD4 (loss), mesothelin, and p53, have been reported. In our experience, a panel consisting of pathophysiology von Hippel-Lindau (pVHL), maspin, S100P, and IMP3 works very well in the distinction of PDAC from normal/reactive pancreatic ducts. Negativity of pVHL and positivity of the remaining 3 markers support a diagnosis of PDAC, instead of a reactive process.^4–7  Of note, maspin is normally expressed in normal gastric mucosa and duodenal mucosa. When background staining for S100P is present, S100A6 can be a good substitute, although weak nuclear and cytoplasmic staining for S100A6 can be seen in normal/reactive pancreatic ducts.

Pancreatic acinar cell carcinoma (ACC) typically stains positive for CK AE1/3, BCL10, trypsin, chymotrypsin, and PDX-1.⁸  BCL10 is considered the most sensitive and specific marker to confirm the diagnosis of ACC⁹  (Figure 1, A through C). Other ACC markers include lipase, amylase, and carboxyl ester lipase.¹⁰  ACC tumor cells express α₁-antitrypsin, an acute-phase plasma protein that can be secreted from normal hepatocytes, small intestinal enterocytes, monocytes, macrophages, and bronchial epithelial cells. Therefore, it is not a useful marker for ACC. A recent multicenter study demonstrated that carboxypeptidase A1 (CPA1), a zinc metalloprotease produced by pancreatic acinar cells, was strongly expressed only in acinar cells of the pancreas^11,12  (Figure 1, A, B, and D). The sensitivity and specificity of CPA1 in diagnosis of ACC were 100% and 95%, respectively.

There are some pitfalls in diagnosis of ACC using special or IHC stains.¹³  About 20% of ACCs show both nuclear and membranous positivity of β-catenin, and 20% of ACCs display loss of expression of SMAD4/DPC4. Periodic acid–Schiff diastase stain is usually positive in ACC. Trypsin immunostain may give a strong background staining. Hepatocellular differentiation markers, including albumin mRNA in situ hybridization (ISH), may be positive in ACC. When more than 30% of tumor cells are positive for neuroendocrine markers, the tumor should be regarded as mixed acinar and neuroendocrine carcinoma.

The 2019 WHO classification of neuroendocrine neoplasms categorizes pancreatic neuroendocrine neoplasms into pancreatic neuroendocrine tumor (PanNET) and pancreatic neuroendocrine carcinoma (PanNEC).^14,15  The former is separated into 3 grades (G1 to G3) based on mitotic rate or Ki-67 index, and the latter is further divided into small cell neuroendocrine carcinoma and large cell neuroendocrine carcinoma based on histomorphology. Both PanNETs and PanNECs usually stain positive for cytokeratin, synaptophysin, chromogranin, and insulinoma-associated protein 1 (INSM1).

According to the 2019 WHO classification of tumors of the digestive system,¹⁵ Ki-67 is no longer used as a marker to differentiate PanNET G3 from PanNEC.¹⁶  A panel consisting of SST2, ATRX, DAXX, p53, and RB1 is usually helpful in differentiating PanNET G3 from PanNEC. The vast majority of NETs show wild-type pattern of p53 expression, intact expression of SST2, and loss of α-thalassemia/mental retardation X-linked (ATRX)/death domain–associated protein (DAXX), whereas most PanNECs display SST2 negativity, mutation pattern of p53, and loss of RB1.¹⁷  The molecular signature of PanNET G3 is MEN1, DAXX/ATRX, and mTOR mutations, whereas PanNEC typically shows mutations in TP53, RB1, KRAS, and SMAD4 genes.¹⁸  In difficult cases, testing for pancreatic hormones may help, as most PanNETs produce hormones whereas PanNECs do not. Table 1 lists the IHC markers useful in differentiating PanNET G3 from PanNEC.

Recent studies demonstrate that status of alternative lengthening of telomeres and loss of ATRX/DAXX are associated not only with adverse prognostic pathologic features but also with shortened 5-year relapse-free survival in nonfunctional PanNETs.^19,20 

These 2 populations of cells have different immunoprofiles. Hepatocellular differentiation is determined on 2 levels: (1) At the cytologic level, cells with hepatocytic differentiation express arginase-1, HepPar-1, and keratin 8/18 and show positivity on albumin RNA ISH.^21–23  (2) At the architectural level, hepatocytic differentiation is evidenced by presence of canalicular structure, which can be highlighted by polyp CEA, villin, CD10, and bile salt export pump (BSEP) antibodies.²⁴  TTF1 antibody clone 8G7G3/1 recognizes hyperplastic cytoplasm.²⁵  Malignant hepatocytic cells also express α-fetoprotein, glypican-3, glutamine synthetase, and heat shock protein 70 (HSP70).²⁶  Cholangiocytes have no specific markers. They express keratins 7 and 19 and display positivity on albumin RNA ISH.

In the diagnosis of liver tumors, the distinction between hepatocellular carcinomas (HCCs) and intrahepatic cholangiocarcinomas (iCCAs) by histomorphology sometimes can be challenging.^27,28  A recent study reported that C-reactive protein (CRP) showed a sensitivity of 75.7% and a specificity of 91.1% in the diagnosis of iCCA. Another new marker, hepatocyte nuclear factor 1β (HNF-1β), exhibited improved performance, with a sensitivity of 100% and a specificity of 92.31% (Figure 2, A through I).²⁹  Please note that this application is limited to workup on primary liver tumors, as HNF-1β expression is shown in a variety of other tumors, such as ovarian clear cell carcinomas, endometrial carcinomas, colorectal carcinomas (CRCs), clear cell renal cell carcinomas, yolk sac tumors, PDACs, and urothelial carcinomas. HNF-1β is a more sensitive and specific marker than CRP for the diagnosis of iCCA and to identify the iCCA component in combined hepatocellular-cholangiocarcinoma. Lack of HNF-1β expression may be used to exclude iCCA from consideration in cases of adenocarcinomas of unknown primary.^29,30 Table 2 lists the hepatocellular and cholangiocytic markers.

Liver sinusoidal endothelial cells (LSECs) differ from classic vascular endothelial cells in multiple aspects:^21,24,31  (1) they lack basement membranes and tight junctions; (2) they are architecturally noncontinuous and fenestrated; and (3) they act mainly as scavengers, eliminating macromolecules and small particulates from the blood, and as immune gatekeepers through their striking endocytotic activity.³²  Not surprisingly, LSECs present a unique immunoprofile: They strongly express macrophage markers (CD4, CD14, CD16, CD31, CDw32 [FcRII], ICAM-1) but rarely express CD34 or factor VIII. They stain positive for ERG but negative for vimentin and D2-40.^33–35 

After vascular remodeling as seen in benign and malignant liver conditions, LSECs undergo capillarization and can switch immunophenotypically and functionally to classic vascular endothelial cells, which can aid in diagnosis.³⁶ 

Bile ductular reactions are commonly seen in cholestatic liver diseases. Using a panel of markers such as CD56, EMA, and CD10 may help to differentiate probable etiologies.³⁷  In fulminant hepatic failure and cirrhosis, hepatic ductules stain positive for CD10, CD56, and EMA. In an autoimmune bile duct injury, such as primary biliary cholangitis and primary sclerosing cholangitis, hepatic ductules express CD56 but not EMA or CD10. When the biliary system is obstructed, hepatic ductules express EMA but not CD10 or CD56.³⁸ 

The 2019 WHO classification of tumors of the digestive system¹⁵  classifies hepatocellular adenoma (HCA) into 4 types: HNF1A mutated, inflammatory, β-catenin activated, and unclassified.^39,40  Recent substantial accumulation of histomorphologic and molecular data enables a more accurate subclassification of HCA (Table 3).^41–43  Please note that in HCC, expression of LFABP can be lost in 16% to 25% of cases.⁴⁴  Interpretation of serum amyloid A (SAA) and CRP immunostain results requires more caution; particularly, they should not be used as a reliable feature to imply a specific type of adenoma as potential precursor for HCC, as their expression can be detected in HCC.⁴⁵  The expression of CRP and SAA can also be seen in focal nodular hyperplasia. This is sometimes a more important diagnostic challenge because focal nodular hyperplasia is often the most important morphologic mimicker of inflammatory hepatic adenoma when dealing with small biopsies. In addition, SAA and CRP can be expressed in cirrhotic nodules. As to the sonic hedgehog–activated HCA markers, ASS1 generally is more sensitive than PTGDS.^45–47 

Histologic diagnosis of Barrett esophagus relies on identification of goblet cells, that is, intestinal metaplasia, in the tubular esophagus mucosa that is 1 cm above the gastroesophageal junction (GEJ). Recognition of goblet cells can occasionally be challenging.⁴⁸  Alcian blue stain at pH 2.5 highlights the acidic mucin in the metaplastic goblet cells. The utility of immunostaining markers for Barrett mucosa is debatable in general, and their use is not recommended for clinical practice. The specialized goblet cells have been reported to stain positive for CK7, villin, SOX-9, HepPar-1, Das1, and CDX2.^49–52  It is worth noting that gastric cardiac mucosa can sometimes stain positive for villin.

There are currently no sensitive and specific markers for diagnosing dysplasia in Barrett esophagus.^53,54  Cytologic and architectural alterations are still the best criteria for this purpose. Some authors believe aberrant p53 expression, P504S positivity, and IMP-3 (KOC) overexpression can help with the distinction.^55,56  However, investigations exploring the utility of biomarkers in difficult cases are still waiting for randomized controlled prospective clinical trial data. Nevertheless, the commonly used markers determining cytologic dysplasia include p53 immunostain and the TissueCypher (Castle Biosciences, Pittsburgh, Pennsylvania) assay.⁵⁷  The latter chooses 9 protein markers for immunofluorescent stain, captures images by an automatic imaging system, and puts the 15 best-performing quantitative image analysis features, such as nuclear morphology and tissue architecture, into an algorithm to generate a risk score (0–10) to classify the risk for progression to high-grade dysplasia/esophageal adenocarcinoma within 5 years into a 3-tier (low, intermediate, or high) or a 2-tier (low, high) risk stratification in patients.^58–60  Both assays have low sensitivity, with p53 immunostain of 31% to 62% and TissueCypher of 29% to 30.4% (for 3-tier risk classification) and 67.7% (for 2-tier risk classification).

Although one recent study showed a promising role for p53 immunostain in identifying higher risk of progression, regardless of histomorphologic presentation, technical variation in evaluation of p53 expression by this method, such as antibody used, staining method, and definition and interpretation of positivity, also makes results noncomparable between laboratories. Similarly, with the TissueCypher system, the effects of preanalytic and analytic factors on performance of TissueCypher are not fully understood yet. In addition, the data from a large cohort (10 000 patients) study conducted by GMC investigators showed that at an acceptability threshold of $100 000/quality-adjusted life year, TissueCypher is marginally cost-effective in identifying risk of progression within 5 years after diagnosis of Barrett esophagus. The 2022 American Gastroenterological Association guideline does not recommend either. Attempts with p17 and Ki-67 immunostains show some positive results but lack wide validation.

The Cancer Genome Atlas project classifies gastric adenocarcinoma into 4 major genomic subtypes: Epstein-Barr virus–infected tumors, tumors with microsatellite instability, genomically stable tumors, and chromosomally unstable tumors.⁶¹  This classification greatly helps determine a patient’s treatment plan. The current National Comprehensive Cancer Network guideline (Gastric Carcinoma, v2.2022) lists HER2 overexpression/amplification, PD-L1 expression, microsatellite instability, high tumor mutation burden (≥10 mutations/megabase), Epstein-Barr virus positivity, and NTRK gene fusion as targetable prognostic markers.⁶² 

About 20% of gastric and GEJ cancers have HER2/neu overexpression and/or amplification. Trastuzumab provides clear survival benefit in patients of advanced stage, as demonstrated in the Trastuzumab for Gastric Cancer trial.⁶³  Trastuzumab in combination with chemotherapy is now the standard of care when treating HER2-positive metastatic gastric and GEJ cancers. Of note, response to the tyrosine kinase inhibitor afatinib can be associated with EGFR and HER2 gene amplification.^64,65 

Claudin 18.2, a tight junction protein, is another targetable molecule in gastric cancer. It is highly expressed in cancers of the uterine cervix, pancreas, ovary, lung, and esophagus.⁶⁶  The monoclonal claudin 18.2 antibody zolbetuximab is associated with significantly longer survival in advanced gastric cancer.⁶⁷ 

Although not currently targetable, loss of E-cadherin protein is prognostic in gastric cancer.⁶⁸  Hereditary diffuse gastric cancer (HDGC) can be caused by germline CDH-1 gene mutation and presents with signet cell features, which confer worse prognosis, but loss of E-cadherin protein is neither a sensitive nor a specific marker for screening for HDGC. Patients with germline detrimental mutation of CDH-1 have up to an 80% risk of developing gastric cancer in both sexes and up to a 60% risk for lobular breast cancer in women. Genetic counseling and testing should be recommended for suspected HDGC cases.

Benign glandular proliferative lesions involving the appendix include those arising from the appendix, such as reactive hyperplasia (secondary to appendicitis and diverticulum), and conditions with other tissue origins, such as endometriosis, endosalpingiosis, and mesothelial cyst.⁶⁹  In reactive hyperplasia, appendiceal epithelial cells express CK20, CDX2, SATB2, and variable CK7. When Müllerian-origin glandular hyperplasia is present, the epithelial cells express ER and PAX8, as seen in endometriosis and endosalpingiosis. The stromal cells in endometriosis stain positive for CD10. The epithelial cells in endometriosis can undergo mucinous metaplasia, which may display CK20 positivity on immunostain. Endosalpingiosis may demonstrate variable expression of Wilms tumor 1 (WT-1) in the epithelial cells, but calretinin should be negative.⁷⁰  Benign mesothelial cells can form cystic structures, especially in the mesoappendix. The lining cells of mesothelial cysts stain positive for both WT-1 and calretinin.

The 2019 WHO classification of tumors of the digestive system¹⁵  replaces the old terms goblet cell carcinoid and adenocarcinoma ex-goblet cell carcinoid with appendiceal goblet cell adenocarcinoma (GCA), low grade or high grade. The GCA can also be graded by a 3-tiered system (grades 1, 2, and 3) based on the proportion of tumor that consists of low-grade (tubular and clustered growth) and high-grade (loss of tubular or clustered growth) patterns.⁷¹  GCA should be differentiated from classic neuroendocrine tumor or conventional adenocarcinoma with tubular and/or signet ring cell features by morphologic features and IHC studies. Molecularly, GCA bears little resemblance to colorectal adenocarcinoma, with rare KRAS, BRAF, APC, and TP53 gene mutations detected.⁷² 

GCA is composed of tubules or clusters of amphicrine neoplasm containing gobletlike mucinous cells, variable numbers of endocrine cells, and Paneth-like cells. The tumor cells can be IHC biphenotypic: they strongly express intestinal epithelial cell markers, including CEA, CK7, CK20, SATBS, MUC2, and CDX2, and also variably express neuroendocrine markers such as INSM1, CD56, synaptophysin, and chromogranin. Appendiceal neuroendocrine tumor cells are the opposite: they express variable amounts of the intestinal epithelial cell markers mentioned above but with consistent strong expression of neuroendocrine markers.⁷³  Conventional appendiceal adenocarcinoma expresses intestine epithelial cell markers but not neuroendocrine markers.

Mucinous tumors in appendix can be de novo mesothelial hyperplasia/neoplasms or may arise from endometriosis. Rarely, ovarian mucinous tumor can spread to the appendix, although we now know that tumor spread with the opposite direction between these 2 organs dominates. On IHC, appendiceal mucinous adenocarcinoma stains positive for intestinal markers including CK7, CK20, CDX2, SATB2, β-catenin (nuclear and membranous), and MUC2.^74–76  Cadherin 17 (CDH17), a transmembrane protein involved in cellular adhesion, has distinct membranous and cytoplasmic expression in GI glands.^77,78  CDH17 is expressed in colorectal (including appendiceal), stomach, and pancreatic cancers and cholangiocarcinomas. Ovarian mucinous carcinoma tumor cells have varied immunoreactivity to intestinal cell marker proteins, but with no expression of SATB2, MUC2, or CDH17.⁷⁹  CK7 is usually diffusely positive in ovarian mucinous tumors, whereas CK20 and CDX2 are usually focally or patchily positive. In contrast, both CDX2 and CK20 are diffusely positive in most appendiceal carcinomas, with variable CK7 positivity.⁸⁰  On the other hand, they express PAX8 (20%–30%) and WT-1, which can further help with the distinction.^81,82  They are almost always negative for nuclear β-catenin expression, suggesting a different molecular pathogenesis than appendiceal mucinous adenocarcinoma.⁸³ 

A small proportion of CRCs overexpress the HER2 oncogene, and the effective targeting of this pathway in other malignancies such as breast and gastric cancer has led to efforts to determine if it can also be exploited as a target in CRC.^84,85  Activation of the HER2 pathway as a bypass signaling pathway has been identified as a mechanism of resistance for anti-EGFR therapy in both the first-line and salvage settings. It has also been shown that RAS and BRAF wild-type metastatic CRC enriches for the presence of HER2 amplification. In contrast to gastric and breast cancers, the diagnostic criteria for HER2 positivity in CRC have not been fully standardized and have varied across studies.^86–88  The HER2 Amplification for Colorectal Cancer Enhanced Stratification (HERACLES) trial was a multicenter, open-label phase II clinical trial in patients with CRC resistant to chemotherapy and anti-EGFR therapy. This trial enrolled metastatic CRC patients with wild-type KRAS and HER2 overexpression, who were then treated with a combination of trastuzumab and lapatinib, with an objective response rate of 30%.⁸⁹  They defined 3 conditions of HER2 positivity as follows (HERACLES criteria): (1) a HER2 IHC 3+ score in 50% or more of CRC cells; (2) a HER2 IHC 3+ score in 10% to 50% of the CRC cells, and a fluorescence ISH HER2/CEP17 2.0 or higher in 50% or more of CRC cells; (3) more than 50% of CRC cells with a HER2 IHC 2+ score and a fluorescence ISH HER2/CEP17 2.0 or higher. However, to date, HER2 diagnostic criteria for gastroesophageal adenocarcinoma⁸⁸  or other independent HER2 IHC scoring systems^90,91  have also been used in CRC HER2 studies. Moreover, the prognostic role of HER2 in CRC remains controversial. Some studies have shown that HER2 overexpression/amplification, as an adverse prognostic factor, is closely correlated with the tumor stage and survival in CRC patients.⁹²  However, other studies have shown no association between HER2 expression and patient survival.^88,93  These controversial results suggest that the role of HER2 in CRC needs to be further explored.