Tumoral Intraductal Neoplasms of the Bile Ducts Comprise Morphologically and Genetically Distinct Entities

With approval of the institutional review boards, the surgical pathology and consultation files of the authors between 1994 and 2018 were searched for cases of tumoral (grossly visible) intraductal neoplasms of the bile ducts with or without an associated invasive carcinoma. Tumoral intraductal neoplasms that were not entirely submitted for microscopic evaluation as well as dysplastic changes that did not form grossly visible lesions (nontumoral dysplasia, biliary intraepithelial neoplasia) were excluded.³⁵  Available gross photographs and descriptions as well as all histologic sections were re-evaluated to confirm the diagnoses and further characterize the spectrum of architecture (eg, papillary, tubulopapillary, and tubular), morphologic differentiation (eg, pancreatobiliary, gastric, intestinal, and oncocytic), and presence, pattern, and extent of invasive carcinoma by using the current WHO criteria for pancreatic and biliary intraductal neoplasms.^2,30,31,36  Available medical records, including imaging study reports, were reviewed to obtain clinical data regarding age, sex, tumor size, location (intrahepatic or extrahepatic), type of surgical intervention, margin status, and date of last follow-up. Contributing physicians were contacted to obtain this information for consultation cases.

Representative full-faced formalin-fixed, paraffin-embedded tissue sections from each case were immunolabeled for MUC1, MUC2, MUC5AC, MUC6, CK7, CK20, and CDX2 by using the standard avidin-biotin peroxidase method (Supplemental Table 1; see supplemental digital content at https://meridian.allenpress.com/aplm in the December 2023 table of contents). The controls used were as follows: Nonneoplastic tissues and carcinomas from the pancreas (MUC1 and CK7), colon (MUC2, CK20 and CDX2), and stomach (MUC5AC and MUC6) served as controls. The staining reaction was evaluated in tumor cell membranes and cytoplasm (MUC1, MUC2, MUC5AC, MUC6), cytoplasm (CK7 and CK20), and nucleus (CDX2). Moderate to strong staining of at least 10% of the lesional cells was considered to represent a positive result.

A representative formalin-fixed, paraffin-embedded block of neoplastic and normal control tissue from each case was selected for next-generation sequencing by the Memorial Sloan Kettering-Integrated Mutation Profiling of Actionable Cancer Targets (MSK-IMPACT) platform. For each case, twenty 10-μm sections of tissue were cut. Owing to small size of the invasive carcinoma component in most cases, only the intraductal component was needle microdissected. DNA was extracted from intraductal neoplasms and normal tissues. Deep coverage, targeted next-generation sequencing was performed for a panel of 468 genes known to undergo somatic genomic alterations in cancer (Supplemental Table 2).^16,20  Briefly, massively parallel sequencing libraries (Kapa Biosystems, New England Biolabs) with barcoded universal primers were generated from 115 to 250 ng of genomic DNA from each case. After amplification and DNA quantification, the barcoded libraries were subjected to solution-phase hybrid capture with synthetic biotinylated DNA probes (Nimblegen Seq-Cap) that target exons from all key genes and introns known to harbor recurrent translocation breakpoints. Each hybrid capture pool was sequenced with deep coverage in a single paired-end lane of an Illumina flow cell. The sequencing data were analyzed for single-nucleotide variants (SNVs), small insertion/deletions (indels), and copy number variations. Somatic SNVs and indels were called by using MuTect and the SomaticIndelDetector tools in the Genome Analysis Toolkit.^37,38  All candidate mutations and indels were reviewed manually with the Integrative Genomics Viewer.³⁹  The mean coverage was 400× for tumor DNA and 50× for normal DNA. Altered genes were categorized into molecular pathways, based on review of the GeneCards Human Gene integrative database (Supplemental Table 3).

A representative formalin-fixed, paraffin-embedded block of neoplastic tissue from IOPNs with enough material (n = 6) was selected for the MSK-Fusion assay. The MSK-Fusion assay is a custom targeted, RNA-based panel that uses Archer Anchored Multiplex PCR (polymerase chain reaction) technology and next-generation sequencing to detect gene fusions in 62 genes (including DNAJB1, PRKACA, and ATP1B1) known to be involved in chromosomal rearrangements.^40–42  These custom assays have been validated and approved for clinical use at Memorial Sloan Kettering Cancer Center (New York, New York) by the New York State Department of Health Clinical Laboratory Evaluation Program.

Statistical comparisons of binomial data were performed by using the Fisher exact test. Statistical comparisons of continuous data were made with the Student t test. Survival comparisons were made by log-rank test of Kaplan-Meier survival curves. Unsupervised hierarchical clustering of the cases, based on molecular pathways, was performed by the Ward method using Euclidean distance measurements.

The clinicopathologic features of the 41 cases analyzed are summarized in Table 1.

The study group included 18 females and 23 males who ranged from 42 to 81 years in age (mean, 69 years). Presenting symptoms included jaundice, dark urine, pruritus, bloating, abdominal pain, loss of appetite, nausea, vomiting, and weight loss. One patient had primary sclerosing cholangitis. Three patients had a history of a prior malignancy, including one each with breast carcinoma, colonic adenocarcinoma, and prostatic carcinoma.

Tumors were relatively evenly distributed throughout the biliary tract: Of the 41 tumors, 44% (n = 18) were intrahepatic, 27% (n = 11) were hilar, and 29% (n = 12) were located within the distal bile duct. All patients were treated by surgical resection, and none received neoadjuvant chemotherapy. Nine of 11 patients with disease recurrence or metastases received chemotherapy and/or radiation therapy.

The tumors ranged from 1.2 to 14 cm (mean, 4.6 cm) in greatest dimension. The intraductal nature of the tumor was specifically documented in 39 of 41 cases (95%) (Figure 1, A and B). Most tumors were described as friable/papillary (n = 19, 46%) or polypoid (n = 11, 27%) masses. Nine (22%) were described as predominantly solid and 2 as cystic (5%). None of the tumors were accompanied by grossly visible mucin in the lumen.

Most tumors (n = 29, 71%) displayed a predominantly papillary or tubulopapillary growth pattern, and the remainder (n = 12, 29%) displayed a tubular growth pattern. Pancreatobiliary differentiation was most common (n = 14, 34%), whereas intestinal differentiation was least common (n = 5, 12%) (Figure 2, A through F). Seven of 41 cases (17%) showed the characteristic features of an IOPN and were composed of arborizing papillae punctuated by intraepithelial lumina and lined by 1 to 2 cell layers of oncocytic cells with large, uniform nuclei and single, prominent nucleoli resembling pancreatic IOPN (Figure 3, A and B).^15,17,18  Six of 41 cases (15%) had a predominantly tubular growth pattern characterized by back-to-back small glands lined by cuboidal cells with minimal to modest amounts of cytoplasm, lacking obvious intracellular mucin, and round to ovoid nuclei reminiscent of pancreatic ITPN (Figure 3, C and D).²⁰ 

Most tumors (n = 32, 78%) contained areas of high-grade dysplasia and 28 of 41 (68%) harbored an associated invasive carcinoma component. Of note, tumors with gastric differentiation were predominantly noninvasive (67% versus 22% in other subtypes, P = .02) and tumors with pancreatobiliary differentiation were predominantly invasive (93% versus 56% in other subtypes, P = .03). The invasive component was of tubular type in 26 of 28 cases (93%) and colloid type in 2 cases (7%), both with intestinal differentiation. One biliary IOPN showed stromal mucin accumulation around the invasive component, resembling a colloid carcinoma. Of the known 30 cases, lymphovascular invasion and perineural invasion was detected in 7 (23%) and 8 (27%) cases, respectively. A lymph node dissection was performed in 26 cases; only 1 case was associated with lymph node metastases.

Immunohistochemical staining results did not consistently distinguish between different morphologic subtypes (Table 2). All morphologic subtypes showed staining for CK7 and MUC1. Although 2 of 5 intestinal-type IPNs (40%) showed MUC2 staining, this marker was also positive in 3 of 13 pancreatobiliary-type IPNs (23%) and 3 of 6 IOPNs (50%). Similarly, MUC5AC and MUC6 were also expressed in most subtypes. Only CK20 showed higher rates of expression in intestinal-type IPNs than other subtypes (80% versus 9%; P < .001). CDX2 was not sensitive for any of the subtypes. Of note, all gastric-type IPNs as well as a vast majority (4 of 6, 67%) of ITPNs and IOPNs showed consistent CK7 and MUC6 coexpression, which was less common among other subtypes (81% versus 33%; P = .004). ITPNs were also consistently negative for MUC5AC.

The results of molecular studies are summarized in Tables 3 and 4, and details regarding each of the individual cases are described in Supplemental Tables 4 and 5.

No germline alterations were identified in any cases. The overall number of alterations, including both somatic mutations and copy number variations, per case ranged from 0 to 64 (median, 9; Figure 4) with the highest number of alterations occurring in a mismatch repair–deficient tumor. The most altered gene was ELF3, which encodes a transcription factor; 13 of 41 cases (32%) harbored either mutations or copy number variations. Of 41 cases in our cohort, other frequently altered genes included TP53 (n = 9, 22%), APC (n = 9, 22%), KRAS (n = 6, 15%), AXIN1 (n = 6, 15%), SMAD4 (n = 6, 15%), CDKN2A/B (n = 6, 15%), SOX9 (n = 6, 15%), ARID2 (n = 5, 12%), ATM (n = 5, 12%), and HIST2H3C/D (n = 5, 12%).

Molecular abnormalities varied somewhat depending on morphologic subtypes (Tables 3 and 4). Pancreatobiliary-type IPNs (n = 14) more commonly harbored alterations in TP53 (n = 6, 43%), ARID2 (n = 5, 36%), and CDKN2A/B (n = 5, 36%) than other subtypes (P = .04, P = .002, and P = .01, respectively). On the other hand, gastric-type IPNs (n = 9) showed higher rates of APC (n = 5, 56%) and KRAS (n = 4, 44%) mutations than other subtypes (P = .01 for both), and intestinal-type IPNs (n = 5) were more likely to harbor mutations in SMAD4 (n = 3, 60%) than other subtypes (P = .02). Two of 7 biliary IOPNs (29%) harbored mutations in BLM, which were not seen in any other tumor type (P = .03). One of 7 other biliary IOPNs (14%) was found to have a mutation in ASXL1 similar to that previously reported in pancreatic IOPNs.¹⁶  Moreover, 3 of 6 IOPNs (50%) tested revealed an ATP1B1-PRKACB fusion, which was previously reported in both pancreatic and biliary IOPNs.^19,28,29  The ITPNs (n = 6) exhibited some recurrent alterations in SWI/SNF complex genes, FGF/FGFR, and PI3K, but none of these alterations were preferential for this subtype (P = .15). Although alterations in the KMT (also known as MLL) lysine methyltransferase family are detected in pancreatic ITPNs,²⁰  they were not found in biliary ITPNs in this series. Rather, 6 IPNs of other subtypes had alterations in genes of the KMT family.

When genes altered were categorized into the predominant pathways (Table 5), the 2 categories most affected were Wnt signaling (n = 21) and DNA damage repair (n = 21), followed in frequency by RAS-MAPK (n = 20) and PI3K (n = 18). Alterations affecting RB1 (n = 15), TP53 (n = 14), SWI/SNF (n = 13), and NOTCH (n = 11) were less common. All gastric-type IPNs harbored at least 1 Wnt pathway gene alteration, which was less common in other subtypes (P = .001). Although MAPK pathway–related alterations were relatively more common in gastric-type IPNs (n = 7, 78%), there was no statistically significant difference among subtypes (P = .07). Tumors with an intestinal phenotype showed disproportionate TGF-β alterations (n = 3, 60%, P = .04) and SWI/SNF alterations were common among tumors with a pancreatobiliary phenotype (n = 8, 57%; P = .02). However, there was no statistically significant difference between intraheaptic and extrahepatic IPNs (Table 6).

Hierarchical clustering based on the pathways revealed 2 major tumor groups: gastric-type, noninvasive IPNs located within the intrahepatic bile ducts and pancreatobiliary-type IPNs with invasive carcinoma located within the distal bile ducts. Pancreatobiliary-type IPNs had more genetic alterations than other groups (15 versus 4; P = .002) affecting PI3K, histone modifier, SWI/SNF, p53, p16/RB, and TGF-β pathways (Figures 5 and 6).

Clinical follow-up information was available for 38 of 41 patients (93%) with a median follow-up of 58.5 months (range, 4–143 months). Fifteen of 38 patients (39.5%) died of disease; sites of dissemination included lung (n = 4), and/or brain (n = 2), and/or liver (n = 3). Six of 38 patients (16%) were alive with disease with a median follow-up of 31 months (range, 4–76 months). Six of 38 patients (16%) were alive with no evidence of disease with a median follow-up of 83.5 months (range, 15–129 months).

Overall, the 5-year survival was 80% for patients who had a noninvasive tumoral intraductal neoplasm of the bile ducts, compared with 60% for patients who had a tumoral intraductal neoplasm of the bile ducts associated with invasive carcinoma. Although the ratio of disease-related deaths was higher for patients who had a tumoral intraductal neoplasm of the bile ducts associated with invasive carcinoma (56% versus 10%), since the number of disease-related deaths in the noninvasive group was low, the difference was not found to be statistically significant by Kaplan-Meier log-rank test (P = .1). There was also no statistically significant difference in survival rates between intrahepatic and extrahepatic IPNs (Table 7). Similarly, age, sex, the presence of invasive carcinoma, tumor size, lymphovascular invasion, perineural invasion, incomplete tumor resection, and lymph node metastases were not predictive of overall survival. Moreover, there was no correlation among genetic alterations or pathways affected and clinical outcome.

In this study, we explored the molecular characteristics of 41 tumoral intraductal neoplasms of the bile ducts, one of the largest series analyzed to date, by high-depth targeted next-generation sequencing for a large panel of key cancer-associated genes (n = 468), along with a clinicopathologic correlative analysis. Various previously unrecognized characteristics of these tumors and their subtypes were elucidated.

Our results confirm that tumoral intraductal neoplasms of the bile ducts are indeed highly analogous to their pancreatic counterparts in their display of a range of histomorphologic differentiation with distinct characteristics. However, IOPNs (n = 7, 17%) and ITPNs (n = 6, 15%), which are not recognized as distinct entities by the current (2019) WHO classification, seem to be represented at a higher percentage among tumoral intraductal neoplasms of the bile ducts than in the pancreas.^30,31  Also, intestinal-type IPNs are relatively rare in our cohort (n = 5, 12%) but this could be population related, as previous studies analyzing tumoral intraductal neoplasms of the bile ducts in Eastern cohorts reported a higher proportion (up to 50%) of intestinal-type IPNs.^43–46  Of note, while other subtypes are found in both intrahepatic and extrahepatic locations, all our intestinal-type IPNs were extrahepatic. Immunohistochemical studies revealed significant overlaps between subtypes (such as CK7 and MUC1 expression is very common regardless of subtype, including 60% of the intestinal-type IPNs). However, there were some staining trends. For example, CK20 and CDX2 expression was identified only in the intestinal-type IPNs, even though the sensitivity was poor (80% and 20%, respectively), whereas MUC6 was not detected in any of the intestinal-type IPNs. Moreover, gastric-type IPNs, biliary IOPNs, and biliary ITPNs were characterized with CK7 and MUC6 coexpression. Gastric-type IPNs may also express MUC5AC but biliary ITPNs consistently lacked MUC5AC, resembling pancreatic intraductal neoplasms (Table 2).^18,21  Previous studies analyzing smaller series have also reported similar results.^1,5  As such, current data suggest that immunohistochemical staining may not be helpful in subtyping these neoplasms.

Akin to IPMNs, the gastric-type of IPN was most frequently associated with noninvasive disease (n = 6, 67%) and the pancreatobiliary-type had the highest rate of invasion (n = 13, 93%) and the invasion was of tubular type.^47,48  Colloid carcinoma was relatively rare but when it occurred, it arose in association with intestinal-type, as is the case in the pancreas. Moreover, biliary IOPNs may also exhibit stromal mucinous changes and mimic colloid carcinoma.

Despite the presence of invasion in the majority of cases (n = 28, 68%), tumoral intraductal neoplasms of the bile ducts followed a protracted clinical course, akin to their pancreatic counterparts.^{16,20,30,31,36}  In a median follow-up of 58.5 months, the 5-year survival was 80% for patients who had a noninvasive tumoral intraductal neoplasm of the bile ducts and 60% for patients who had an invasive tumoral intraductal neoplasm of the bile ducts. These outcomes are similar to prior reports.^3,4,8  However, there was no statistical difference in survival between our noninvasive and invasive tumoral intraductal neoplasms of the bile ducts, which is rather unexpected. This might be, in part, due to the relatively low number of noninvasive tumoral intraductal neoplasms of the bile ducts. Initially, we thought the small size of invasive component in invasive tumoral intraductal neoplasms of the bile ducts might be a factor as well (there are 9 cases in our cohort that exhibit minimal [ie, <10% of the tumor] invasion). However, even when these cases were included in the noninvasive cohort, the status of invasion still did not affect survival, raising the possibility of inaccuracies in identifying and quantifying the extent of invasion. In fact, because of their tendency to grow in lobulated configurations, introducing a degree of subjectivity in distinguishing between expansile (pushing-border invasion) versus true intraductal (noninvasive) growth, assessing invasion can be very challenging in these tumors. Of note, there was also 1 patient with gastric-type IPN without identifiable invasive carcinoma who developed a biopsy-proven metastasis in the lung and died of disease. Multifocality and field effect phenomenon⁴⁹  may be factors in such cases with unexpected progression and emphasize the importance of extensive sampling and thorough examination along with long-term follow-up. While earlier studies showed that a positive surgical margin was also correlated with worse outcomes, only 3 tumors in our cohort had a positive margin, which limits the analysis of this feature.^3,46,50 

Our results also illustrate that while tumoral intraductal neoplasms of the bile ducts show some genetic similarities to other pancreatobiliary neoplasms, including aberrations in KRAS, CDKN2A, TP53, and SMAD4, which are commonly seen in pancreatic ductal adenocarcinoma, IPMNs, and conventional bile duct carcinoma,⁵¹  the genetic alterations and their corresponding prevalence are not always congruent. For example, ELF3, which encodes a transcription factor with possible interactions with ErbB as well as Notch signaling pathways, was altered in about one-third of tumoral intraductal neoplasms of the bile ducts; however, it has been rarely reported in pancreas (1%) and ampullary (2%–12%) carcinomas.^52,53  Similarly, APC mutations were more common in tumoral intraductal neoplasms of the bile ducts (22%) than in conventional bile duct adenocarcinomas (3%–7%).^33,54  Emerging studies analyzing smaller numbers of cases and/or a limited panel of genes have also reported similar results.^32,33,54  For example, in one of the first studies analyzing the genetic landscape of IPNs, Fujikura et al³²  performed whole exome sequencing on 7 cases, followed by validation targeted sequencing on 7 additional cases, and identified mutations in APC or CTNNB1 in 6 cases (43%). Additionally, in contrast with pancreatic IPMNs, especially intestinal-type IPMNs, which show relatively high frequency of RNF43 and GNAS mutations,²⁷ RNF43 was mutated only in 3 cases in our cohort (1 intestinal-type IPN, 1 gastric-type IPN, and 1 IOPN) and another case of intestinal-type IPN harbored GNAS mutation. However, previous studies analyzing RNF43 and/or GNAS mutations in Eastern cohorts that have many more (50%) intestinal-type IPNs reported higher frequencies for both mutations, and the mutations were significantly more common in intestinal-type IPNs.^25,44,45,55  Therefore, it is possible that the difference between those studies and our results is due to the relatively low number (n = 5, 12%) of intestinal-type IPNs in our cohort.

Unfortunately, owing to the small size of the invasive carcinoma component in most of our cases, we could not sequence the invasive carcinoma component separately. Therefore, we are unable to compare the spectrum of alterations in invasive tumoral intraductal neoplasms of the bile ducts with conventional bile duct adenocarcinomas.

The molecular findings in this study also support the separation of biliary IOPNs and ITPNs from the other tumoral intraductal neoplasms of the bile ducts subtypes. To the best of our knowledge, our study has the largest cohort to date analyzing molecular features of biliary IOPNs. Similar to their pancreatic counterparts, all 7 biliary IOPNs analyzed in this study lack the mutations commonly found in other subtypes of pancreatic and biliary intraductal neoplasms. Also, of several recurrently altered genes in pancreatic IOPNs,¹⁶ ASXL1 is found to be mutated in a case of biliary IOPN. Moreover, 50% of the biliary IOPNs revealed an ATP1B1-PRKACB fusion, which was previously reported in both pancreatic and biliary IOPNs.^19,28,29  Of note, compared to pancreatic IOPNs, although the frequency of the fusion is lower in our current study, it is most likely due to the old nature of the negative cases, which were resected before 2005 (ie, suboptimal tissue quality for fusion assay). Similarly, this study confirms that biliary ITPNs do not have KRAS mutations, nor do they have mutations in Wnt pathway genes. Previous studies on biliary ITPNs also revealed a very low prevalence of alterations in common oncogenic signaling pathways.^5,34,56  Instead, alterations in chromatin remodelling, SWI/SNF complex, PI3K pathway, and FGFR pathway genes seem to be significantly associated with biliary ITPNs as is the case for the pancreatic ITPNs.²⁰  Overall, these findings demonstrate that biliary IOPN and ITPN are different from the other subtypes of biliary intraductal neoplasms not only based on their distinctive histopathologic features, but also based on the genetic pathways of tumor development.

Some of these specific genes or pathways are of significance because there are targeted therapies being developed, some of which are already in clinical trials. For example, alterations found in several receptor tyrosine kinases in the ERBB, NTRK, FGFR, and VEGFR families, which were discovered in these intraductal tumors, have available therapeutic inhibitors.^57–60  Similarly, PI3K and mTOR inhibitors may be of value in the tumors that reveal alterations in this pathway. Other potentially actionable targets found in our cohort include MAP kinases, SWI/SNF, MDM2, JAK/STAT, and DNA methyltransferases.

Finally, the groups based on our hierarchical clustering appear to have some similarities with the classification system proposed by the Japan-Korea Cooperative Study Group.⁶¹  Gastric-type, noninvasive tumoral intraductal neoplasms that are located within the intrahepatic bile ducts, which are likely to be classified as Type 1 by the Japan-Korea Classification, were significantly more common in our Group 1. In contrast, pancreatobiliary-type, tumoral intraductal neoplasms associated with invasive carcinoma that are located within the distal bile ducts, which are likely to be classified as Type 2, were significantly more common in our Group 2. However, none of our cases in Group 1 had grossly visible mucin in the lumen.

In summary, our comprehensive analysis confirms that tumoral intraductal neoplasms of the bile ducts, similar to their pancreatic counterparts, are diverse tumors, not only morphologically but also genetically. They consist of 5 distinct groups: pancreatobiliary-type IPN, gastric-type IPN, intestinal-type IPN, biliary IOPN, and biliary ITPN. Even the patients with invasive carcinoma appear to have a relatively protracted clinical course. The most common genetic alterations occur in the Wnt, DNA repair, MAP kinase, and PI3 kinase pathways, as well as many receptor tyrosine kinases. Some of these alterations are potentially targetable. Continued investigation of the clinical, histologic, and molecular features of these tumors will likely shed new light on better risk stratification and customized therapy to optimize patient care and outcomes.

The authors thank Marina Asher for her assistance with immunohistochemical stains and Bruce Crilly for his assistance with the figures.