Density and size of lymphoid follicles are useful clues in differentiating primary intestinal follicular lymphoma from intestinal reactive lymphoid hyperplasia

Whole slide imaging (WSI), the scanning of conventional glass slides as digital files, has been increasingly employed by pathologists for assisting diagnosis, education, and research purposes. Recently several studies have validated the accuracy of WSI in pathological diagnosis, suggesting that WSI can be used for routine practice, which has been realized in a Dutch academic pathology laboratory where the pathological diagnosis is fully digitalized [1‐4]. The concordance rates in diagnosis between using conventional light microscope and WSI range from 87 to 96%, depending on various subspecialties [1, 5‐9]. However, most of previously published studies mainly targeted small specimens and excluded cases requiring immunohistochemistry, such as lymphoma cases. So far there is only one published report investigating the concordance rate between digital pathology and conventional microscopic histopathology and the researchers found that WSI was a reliable technology for lymphoma diagnosis [10].

Follicular lymphoma (FL) is one of the most common low-grade B cell lymphomas. FL is a heterogeneous disease, and may be further divided into many different subtypes in terms of age of onset, involved organs and genetic abnormality. In the revised fourth edition of World Health Organization (WHO) classification of tumors of hematopoietic and lymphoid tissues, primary intestinal FL (PI-FL) is listed as a rare and specific variant of FL with distinct clinical and biological features [11]. PI-FL predominantly occurs in the second portion of duodenum as multiple whitish polypoid nodules [12, 13]. Unlike patients with systemic nodal FL, most patients with PI-FL have localized diseases and excellent prognosis even without treatment [11‐13]. However, long-term follow-up is recommended as rare cases with histological transformation to high-grade lymphoma have been reported [14, 15]. Considering the uniqueness of PI-FL, thorough clinical, laboratory, endoscopic and radiologic workups to differentiate PI-FL from systemic nodal FL with secondary gastrointestinal (GI) tract involvement have been suggested [16].

PI-FL shares the same histopathology, immunophenotype, and the hallmark t(14;18) translocation with systemic nodal FL [12, 17]. On the other hand, expression of certain markers, such as memory B-cell marker CD27, molecules related to B-cell homing such as chemokine (C-C motif) ligand 20 (CCL20) and mucosal vascular addressin cell adhesion molecule 1 (MAdCAM-1), lack of activation-induced cytidine deaminase (AID), and variable immunoglobulin heavy chain gene (VH) usage suggest a genetic similarity of PI-FL with gastric mucosa-associated lymphoid tissue lymphoma, but different from systemic nodal FL [13, 17‐20]. In addition, follicular dendritic cells (FDC) residing in the periphery of neoplastic follicles is more frequent in PI-FL cases as compared to nodal FLs, in which the FDC usually form dense meshworks. Takata K. et al. thus suggested that recognizing different FDC patterns is useful for differentiating PI-FL from secondary GI involvement by systemic FL [13].

It is important to make an accurate diagnosis by endoscopic biopsy specimens to facilitate appropriate management. In this study we retrospectively analyzed PI-FL cases in an attempt to identify morphological features of PI-FL by measuring various parameters on WSI. These morphological clues might be useful in differentiating reactive lymphoid hyperplasia from PI-FL in endoscopic biopsy specimens.

Although rare, PI-FL has been characterized recently, particularly the duodenal-type FL. Surgical specimens are easier for diagnosis; however, with increasing use of endoscopy, accurate pathologic diagnosis on small biopsy specimens remains challenging. In this current study of PI-FL with WSI, we found that the average density and diameter of neoplastic follicles were larger than those of reactive lymphoid hyperplasia in the GI tract. Interestingly, in contrast to the study by Crocker et al in 1983, in which they used image analyzer and found that reactive lymph nodes had larger mean follicle areas than nodal FL [26]. The discrepancy between their and our studies may come from the different types of specimens/organs used for analysis, that is, large nodal tissue vs. endoscopic intestinal biopsy specimens. In addition, the neoplastic follicles in our PI-FL samples tended to be located more superficially, although not statistically significant. Occasional ulcerations or erosions of the surface mucosa were noted as well. Thus, we suggest that when encountering endoscopic biopsies with the presence of superficially located large follicles and a high density of follicles, immunohistochemistry is mandatory to exclude the possibility of PI-FL. Further WSI studies might include cases of systemic nodal FL with secondary involvement of GI tract, to compare the histologic parameters of neoplastic follicles between primary and secondary FL.

In keeping with other series of PI-FL [27, 28], the majority of our patients were middle-aged or elderly. One-third of our patients were asymptomatic and the remaining patients presented with ambiguous abdominal discomfort or bloody stools. In the literature, most of the cases were detected by endoscopy with typical polypoid lesions in the duodenum (65%), followed by jejunum and ileum (20%) [13]. Our cases showed a similar trend with duodenum as the most common site (11/21; 52%), followed by ileocecum and jejunum (8/21; 38%). In the largest series of PI-FL so far in the literature, Schmatz et al reported that in 11 (17%) of 63 patients another segment of the small bowel (i.e., jejunum and/or ileum) was involved [12]. In our study, in most of the cases (19/21; 91%), the tumors were located only in a single anatomic site. Primary FL at different GI sites has been reported to possess different endoscopic features as visualized by more advanced and detailed endoscopic examination. For example, the most frequent endoscope finding in the stomach was ulcerated submucosal tumor while small granular lesions were the predominant findings in the duodenum [29‐31]. All of our cases showed low clinical stage without mortality or high-grade transformation during the follow-up period. Application of the FLIPI developed for nodal FL seemed irrelevant to the prognosis of our patients. However, data from more cases should be collected to investigate the feasibility of using FLIPI in PI-FL.

Immunophenotypically, the neoplastic follicular center cells expressed germinal center cell markers CD10 and BCL-6, with aberrant expression of BCL-2 due to t(14;18)/IGH-BCL2 translocation. In our current study, all PI-FL cases demonstrated BCL-2 protein expression and all cases expressed CD10 and BCL-6. The characteristic duodenal pattern of FDC meshworks was observed in all of our duodenal examples of PI-FL, while those in the non-duodenal sites had either nodal or duodenal patterns. The phenomenon was also observed in a previous study [22]. Interestingly, one of our cases had FDC meshworks in a mixed nodal and duodenal pattern. To our knowledge, to date there are no data suggesting PI-FLs at different GI sites might have different prognosis. The significance of FDC patterns has not yet been fully investigated either.

Using FISH, we confirmed IGH/BCL2 translocation in most (94%) of the successfully assayed cases, while the neoplastic cells in all cases expressed BCL-2 protein. The hallmark translocation t(14;18)(q32;q21) was detected in around 85% of PI-FL by FISH [12, 22], similar to that of nodal FL. Although in our previous study we found a relatively lower frequency (63%) of t(14;18) in low-grade FL in Taiwan [23]; in the current study we found that PI-FL in Taiwan carried a similar rate of t(14;18) as that of the Western nodal FL and Japanese PI-FL. The prognostic significance of t(14;18) in patients with PI-FL has not been delineated yet. Studies on larger numbers of cases are warranted.

The incidence of PI-FL might be underestimated as the patients might be asymptomatic or might have ambiguous symptoms, as exemplified in our study that eight (38%) patients were discovered at health examination with endoscopy. Although the case number is limited, our study, so-far the largest from Taiwan, showed histologic and clinical features characteristic of PI-FL. Most importantly, our study is the first to utilize digital imaging technique to assess the histologic parameters of PI-FL vs. intestinal reactive lymphoid hyperplasia. Our results indicate that superficially located large follicles and a high density of follicles in endoscopic biopsy specimens are surrogates for PI-FL and immunohistochemistry is mandatory to confirm the diagnosis. Larger series of studies are required to better characterize and to verify the parameters measured from WSI and their usefulness as adjuncts for pathological diagnosis.