In colorectal carcinomas (CRCs), research has highlighted that dedifferentiated tumor cells at the advancing edges have important prognostic significance (12345). The dedifferentiated tumor cells take the form of either single cells or clusters and do not have a well-formed glandular lumen. According to the number of cells, a cluster is defined as a “tumor bud (TB)” when consisting of < 5 cancer cells or as a “poorly differentiated cluster (PDC)” when consisting of ≥ 5 cells (67). PDCs are therefore more easily recognizable by hematoxylin and eosin (H & E) staining than TBs, and they are less likely to be mistaken for peritumoral desmoplastic tissue or inflammatory cells (8). However, it is clear that PDCs show a histological similarity to TBs regarding the loss of gland formation, and PDCs and TB are occasionally observed synchronously in the same tumor (Fig. 1). These morphological similarities and simultaneous occurrence suggest that PDCs and TBs could represent sequential steps in tumor growth and that epithelial-mesenchymal transition (EMT) may play a role in the development of both PDCs and TBs. β-catenin, located at the cell membrane, forms a complex with E-cadherin and is crucial for maintaining cell-cell adhesion and epithelial cell polarity (9). Mutations in the APC gene lead to the nuclear translocation of β-catenin, where it binds to T-cell factor/lymphoid enhancer-binding factor (TCF/LEF) family members, functioning as an oncogenic transcription factor (9). Therefore, the preservation of membranous β-catenin and E-cadherin is indicative of an epithelial phenotype, whereas the nuclear expression of β-catenin and loss of membranous E-cadherin are considered hallmarks of EMT (9).

According to the images and definition in the original study, PDCs with artifactual retraction clefts also show morphological similarities to micropapillary clusters; although, PDCs are observed as more variable-sized clusters without any glandular differentiation in a background of desmoplastic stroma (10). By contrast, lumens within classic micropapillary clusters are not uncommon, and more prominent cystic lacunar spaces are displayed. In micropapillary clusters, the reversed cell polarity of tumor cells may contribute to the development of cystic lacunar spaces, which is demonstrated by the inside-out (I/O) pattern of MUC1/epithelial membrane antigen (EMA) immunohistochemical staining. MUC1/EMA is a glycoprotein typically expressed in the apical cell surface (i.e., the central part of the cell membrane) of normal glandular epithelium, plays an important role in glandular lumen formation, and has an inhibitory role in cancer-stromal interactions and E-cadherin-mediated intercellular interactions. Due to the reversal of cell orientation in micropapillary clusters, MUC1/EMA is located in the membrane toward the stroma (i.e., the peripheral part of the cell membrane) and disrupts cell adhesion to the stroma, resulting in characteristic lacunar spaces (1112).

Apart from a morphological resemblance, it is well documented that PDCs, TBs, and micropapillary patterns (MPPs) are all significantly correlated with aggressive tumor behavior, such as a lower survival rate and frequent lymph node metastasis and lymphovascular tumor emboli (LVE) (1213). Based on these morphological continuity and behavioral similarities, it is reasonable to assume that there is a pathogenetic overlap among PDCs, TBs, and MPPs and that they are sequential manifestations in the process of the same pathogenetic events associated with infiltrative growth. However, there are only a few studies focused on this topic (14).

In the present study, to ascertain the biological nature of PDCs, we examined ancillary immunohistochemical markers associated with the morphogenesis of TBs and MPPs, such as β-catenin, Ki-67, E-cadherin, epithelial cell adhesion molecule (EpCAM), MUC1, and EMA, and compared their expression in TBs and PDCs, as well as in differentiated neoplastic glands from the tumor center (TC).

In 2008, Ueno et al. (7) defined dedifferentiated tumor clusters at advancing edges according to the number of cells, as TBs or PDCs, and various grading systems based on the number of cells have been verified as reproducible (21623). Because PDCs are more easily recognized than TBs in routine H & E sections without using additional immunohistochemical staining, objectivity, and simplicity are the advantages of the PDC-based grading system. Additionally, it has been reported that the PDC-based grading system strongly indicates poor prognosis (11024). Consistent with results from previous reports, our data showed that PDC-based grading was a robust prognostic indicator and was significantly correlated with clinicopathological parameters, including primary tumor (T), regional lymph nodes (N), and distant metastasis (M) stages, LVE, and PNI. However, the grading system has not been universally used for several reasons. First, because the grading system is based on quantitative methodology, there may be resistance to the increased workload in routine diagnosis. Second, PDC is a newly introduced concept, and there is no conclusive evidence about its pathogenesis. Finally, the morphological continuity and similarity of PDCs and other histological findings such as TBs and MPPs can confuse pathologists. Ultimately, if PDCs, TBs and MPPs are histological patterns that are associated with the same carcinogenetic event, then the grading systems based on the counts of these components could be integrated. Thus, we conducted this study to demonstrate the biological characteristics of PDCs.

In our study, similar results regarding the expression of β-catenin and E-cadherin in PDCs and TBs clearly indicate that the morphogenesis of PDCs is also associated with EMT, considering the well-described interaction between β-catenin and E-cadherin in cell-cell adhesion. However, other factors must also play a role in the pathogenesis of PDCs because, although up to 90% of CRCs display dysregulated Wnt signaling and 60% harbor APC mutations, PDCs are only seen in a minority of cases (9). Regarding the staining patterns, it has been reported that the loss of cytoplasmic/membranous expression of β-catenin is associated with a poor survival rate (19). Consistent with this observation, we found that tumors from the control group (PDC G1) did not show cytoplasmic/membranous loss of β-catenin whereas tumors from patients in the study group (PDC G2 and G3) showed cytoplasmic/membranous loss (32/117, 27.4%). Previously, the loss or aberrant expression of E-cadherin was correlated with dedifferentiation (25). We observed the loss or aberrant expression of E-cadherin in differentiated tumor cells from the TC of the minor cases (8.5% in the study group and 6.3% in the control group), suggesting that the “budding immunophenotype” may not be restricted to dedifferentiated tumor cells (26).

EpCAM mediates the cell-cell adhesion of epithelial cells and is an additional marker for cancer-initiating stem cells (27). Gosens et al. (22) reported that the loss of membranous EpCAM is associated with TBs and nuclear β-catenin localization. Interestingly, we showed that the frequency of EpCAM expression was sequentially decreased in the TC, PDC, and TB samples, respectively. Moreover, we found that PDCs have a lower rate of proliferation than the neoplastic glands of TC and that TBs have the lowest proliferative activity. Therefore, it might be proposed that PDCs are an intergrade between differentiated neoplastic glands and dedifferentiated TBs. Our results also support previous results suggesting that the loss of the epithelial phenotype in tumor cells is accompanied by a loss of proliferative capacity, where the driving force for tumor infiltration can be attributed to factors such as stromal remodeling rather than proliferation (171828).

According to the definition reported by the World Health Organization, micropapillary CRC has small clusters of tumor cells within stromal spaces mimicking vascular channels (3). Despite their name, micropapillary clusters are devoid of fibrovascular cores, and PDCs are surrounded by clear, artifactual retraction spaces that display a morphological overlap with MPPs (16). As such, with only moderate interobserver reproducibility, even among expert pathologists, it has been a diagnostic challenge to differentiate nonclassical micropapillary clusters from extensive retraction artifacts (29). Moreover, a retraction artifact is not always an artifact that can be ignored, considering that extensive retraction artifacts are also strongly associated with poor outcomes (30). Recently, Barresi et al. (14) reported that the reversed pattern of MUC1 expression and loss of E-cadherin expression were also observed in all PDCs and TBs in their study cohort and that reversed cell polarity may reflect EMT. They proposed that the term “micropapillary” be replaced by PDC because PDC indicates the underlying biological phenomenon and is linked with a more objective grading system. In our study, a reversed pattern of MUC1 and EMA expression was not always identified, but was present in 48.7% and 45.3% of the PDC samples, respectively, while it was absent in the TC samples. Thus, we agree that PDCs have biological properties similar to micropapillary clusters. Compared with the study by Barresi et al. (14), the lower frequency of reversed expression patterns may have arisen from differences between the study cohorts (i.e., PDC G3 in Barresi et al. (14) vs. PDC G2 and G3 in the present study) and tissue blocks used for immunohistochemical staining (whole-tissue block in Barresi et al. (14) vs. TMA block in the present study).

The limitation of our study is that immunohistochemistry was performed on TMAs. As immunohistochemical staining can show heterogeneity, evaluation on TMAs may miss some positive cases. We evaluated four cores with a relatively large size (3 mm) for each case and tried to overcome the limitation of TMA usage. Another limitation is that we evaluated immunoreactivity in only PDCs and TBs. Future studies are needed to compare immunohistochemical results in micropapillary clusters.

We demonstrated that PDCs, TBs, and MPPs share immunohistochemically common results with morphological and behavioral similarities. They may represent sequential or similar growth patterns as found for other outgrowths of common biological processes such as dedifferentiation and alteration in the interaction between tumor cells and stroma. Therefore, we propose that the terminology and concepts concerning the PDCs, TBs, and MPPs with retraction artifacts should be carefully reviewed and clearly reestablished based on biological characteristics. Ultimately, an integration of grading systems based on the counts of these components should be considered.