Histomorphologic spectrum of nodal marginal zone lymphoma as defined by its methylome

Abstract

Objectives

Primary nodal marginal B-cell lymphoma (NMZL) is rare and histologically very variable. Its large-cell presentation is difficult to distinguish from nodal diffuse large B-cell lymphoma (nDLBCL) due to the absence of specific markers for nodal marginal zone lymphomas in general.

Methods

Using a comprehensive cohort of NMZLs and a control cohort of nDLBCLs, we conducted a methylome analysis on subgroups of both.

Results

The methylomes were strikingly different between the cohorts but unexpectedly homogeneous within the NMZL cohort. This allowed us to describe the morphologic spectrum of NMZL in all its value ranges. The considerable overlap in growth pattern and cytology of NMZL with nDLBCL was explored morphometrically, leading to an operational tool for separating both by a simple measurement of cell size and nuclear size. This was integrated in a hierarchical approach, including a scoring system for the parameter growth pattern, follicular colonization, follicular dendritic network, IgD expression, and Ki-67 rate, and led to a proposal for a classifier that we present here.

Conclusions

This methylome-based study extends the morphological spectrum of NMZL towards large cell morphology and offers a conventional way to distinguish it from nDLBCL.

INTRODUCTION

Primary nodal marginal B-cell lymphoma (NMZL, International Classification of Diseases for Oncology [ICD-O] code: 9699/3) is rare,^1,2 and its patterns of lymph node involvement are as variable as its cytomorphology.³ In the case of large-cell cytology, the distinction between NMZL and nodal diffuse large B-cell lymphoma (nDLBCL,⁴ ICD-O code: 9680/3) is an unsolved problem in daily routine. Unlike mantle cell lymphoma⁵ with its cyclin D1 expression as a discriminative marker, even in cases of large-cell morphology, no such sensitive and specific marker has been found to date. Although several immune markers have been proposed, such as T-Bet,⁶ Irta-1,⁷ and MNDA,^8,9 they have not yet become established due to their lack of robustness and specificity.¹⁰

The distinction between nDLBCL and large-cell NMZL is not purely academic as, similar to what has been repeatedly communicated for extranodal¹¹ (eg, gastric marginal zone) lymphoma,¹² the large-cell variant of NMZL might have a far better prognosis.^13,14

Against this background, we set out to explore the nosological border between NMZL and nDLBCL. To this end, we (the German Reference Centers for Lymphoma Diagnosis) selected 89 cases for which the diagnosis of NMZL had been made in the respective center with its local diagnostic algorithm. As a reference group for nDLBCLs, we extracted 47 cases from 1 reference center.

The diagnoses of NMZL in our cohort were made by expert lymph node pathologists and given as such to the patients. We chose to further characterize this lymphoma by analyzing its methylome NMZL and comparing it with that of nDLBCL.

Here we show that the methylome of NMZL is surprisingly homogeneous and clearly distinct from that obtained for our nDLBCLs, which, as it turned out, is in concordance with published data.

Thus, having critically verified our diagnoses of NMZL, we explored the morphologic spectrum of patterns of spread within the lymph node, including its behavior toward follicles. Furthermore, we applied morphometrics on cell size and nuclear size to create a practical, simple numerical distinction between large-cell NMZL and nDLBCL. Cell dimension and morphology have been historical milestones for the study of other malignancies, such as primary mediastinal B-cell lymphoma.¹⁵

As a final diagnostic adjunct, we observed IgD expression in various quantities and intensities in half of the NMZL cases, while this was infrequent in our nDLBCL cohort.

MATERIALS AND METHODS

Lymph Node Selection

We collected 89 NMZLs TABLE 1 from 9 German Reference Centers for Lymphoma and 47 nDLBCLs (Supplemental Table 1; all supplementary material is available at American Journal of Clinical Pathology online) from our archives, together with 6 cases of lymph node reaction to toxoplasmosis (toxoplasmic lymphadenitis) as a nontumor reference for the cell and nuclear size of mantle cells, centroblasts, and monocytoid B cells. The diagnoses of NMZL and nDLBCL had been made by the experts of the reference centers applying their local diagnostic procedure, including immunohistochemistry, clonality analysis, fluorescence in situ hybridization, and mutation analysis. Although these diagnoses had been made by lymphoma experts and communicated as such to the treating clinicians, they are, scientifically speaking, still bona fide diagnoses based on current knowledge, which is fuzzy at the nosological border between NMZL and nDLBCL.

Methylation Profiling

DNA was extracted from formalin-fixed, paraffin-embedded (FFPE) material with the GeneRead DNA FFPE Kit (Qiagen) and quantified with the Qubit 1X dsDNA HS Assay Kits (Thermo Fisher Scientific). Due to the size of the biopsy specimens and the DNA quality, we ended up with a sufficient amount of DNA (500 ng) from 35 NMZLs and 17 nDLBCLs. The bisulfite conversion was performed with the EZ DNA Methylation kit (Zymo Research), and the DNA was then restored with the Illumina Restoration Kit. The hybridization on the MethylationEPIC Beadchip (Illumina) was done following the manufacturer’s instructions. Chips were scanned with the NextSeq 550Dx instrument. Raw data were preprocessed with GenomeStudio software (Illumina). The analysis was performed on Qlucore Omics Explorer software as follows: first, a feature selection was applied to filter the loci that are significantly differentially methylated across the 2 groups (variance filter σ ≥ .4 and 2-group comparison q < 0.05); then, an unsupervised clustering was carried out to organize each sample in a hierarchy tree based on similarity (method applied: Euclidian distance, Qlucore Omics Explorer software).

Histomorphologic Analysis

The NMZL cohort was systematically analyzed as follows:

For the growth pattern, we adopted the definitions of Salama et al¹⁶ as they were given and depicted in their seminal article, “Immunoarchitectural Patterns in Nodal Marginal Zone B-Cell Lymphoma”:

• Perifollicular: annular distribution of the neoplastic cells around uninvolved normal secondary follicles

• Interfollicular: neoplastic cells in the interfollicular areas that surround residual germinal centers

• Nodular/follicular: distinct nodules that are well demarcated from the uninvolved interfollicular areas

• Diffuse: diffuse sheets of neoplastic cells with effacement of nodal architecture

The cell morphology as characterized by the pathologist was addressed, and 3 categories for cell size were set up: (1) “small,” reminiscent of normal marginal zone B cells; (2) “large,” within the range of normal centroblasts or monocytoid B cells; and (3) “abnormally large,” cells exceeding the size of centroblasts and monocytoid B cells. Lymph follicles were characterized morphologically and by means of CD21 (Dako, IR608) immunostaining, highlighting the follicular dendritic cells (FDCs). The follicles themselves were characterized as present or absent, and the FDC network was scored as normal, expanded, disrupted, compressed, or absent. The IgD expression (BioSB, BSB2961) status was scored as negative, weakly positive, partially weakly positive, partially strongly positive, or strongly positive. The Ki-67 proliferation rate was determined by estimating the percentage of Ki-67–positive (Dako, M7240) neoplastic cells and roughly categorized as “low,” ranging from less than 1% to 5%; “intermediate,” ranging from 5% to 39%; and “high,” being higher than 40%.

The same procedure was followed for the nDLBCL control cohort. The immunohistochemical stains were performed with the Dako-Omnis automated platform (Dako).

Morphometrics

A high-resolution representative picture was acquired from the H&E staining of each lymphoma sample. For each picture, the diameter of 30 cells and their nuclear diameter was measured (Supplemental Table 2) with ImageJ software (National Institutes of Health) and the data analyzed with GraphPad (v. 8, Dotmatics; GraphPad Software). We defined 1 and 2 standard deviations in the cell size and nuclear size of centroblasts as the largest reactive B cell. With this as a reference, we defined the labels as “small” (cell size, <10.5 µm; nuclear size, <7.7 µm), “large” (cell size, <14 µm; nuclear size, <9.4 µm), and “abnormally large” (cell size, >14 µm; nuclear size, >9.4 µm) for both measurements. Moreover, we calculated the cell/nucleus ratio and the width of the cytoplasmic rim with the subtraction of nuclear size from cell size. We intentionally used the terms small, large, and abnormally large in both the histomorphologic description and the morphometric categorization. The dual characterization was largely consistent in itself but revealed discrepancies that were minimal.

RESULTS

Methylation

A methylation profile was obtained for 35 NMZLs and 17 nDLBCLs (array data available: GSE260667). The comparison between the NMZL and nDLBCL groups highlighted a separation of the cases into distinct clusters with just a few cases falling out of the respective group FIGURE 1. Forty-three significantly differentially methylated CpG loci emerged from the analysis (Supplemental Table 4), with 53.5% of them belonging to promoter regions. Unexpectedly, no internal separation of the NMZL cohort profiles was correlated with any histomorphologic features of this lymphoma, and the same was true for the parameters we applied to describe them (Supplemental Figure 4).

Histomorphology

The marginal zone of the lymph nodes was selectively occupied in a minority of cases, mostly by small cell variants, the pattern being perifollicular or interfollicular. Expansion of this pattern led to either follicular compression or follicular colonization, ultimately leading to the disruption of the FDC network. Both processes were highlighted by the CD21-stained FDCs FIGURE 2G, H. When the follicular dendritic network was eventually completely destroyed and effaced, the resulting growth pattern was diffuse.

Adopting the definitions of the growth patterns of NMZL published by Salama et al,¹⁶ our cohort consisted of 12 cases with a perifollicular, 34 cases with an interfollicular, 23 cases with a nodular, and 19 cases with a diffuse growth pattern.

Next, operationally subdividing the lymphoma cell size into “small” (reminiscent of normal marginal zone cells), “large” (reminiscent of normal centroblasts) FIGURE 3C and monocytoid B cells FIGURE 3D-G, and “abnormally large” (cells exceeding the size of normal reactive B-blasts) FIGURE 3H, our cohort had the following characteristics (detailed in TABLE 1): 40 cases belonged to the small-cell category FIGURE 3A, B, 38 cases had a large-cell morphology FIGURE 3C-G, and 10 featured abnormally large cells.

There was no stringent correlation between growth pattern and cell size, although the perifollicular pattern was almost exclusively found in small-cell NMZLs. The diffuse growth pattern was prevalent in large and abnormally large-cell lymphomas. Unaffected follicles were more frequent in small-cell lymphomas. Follicular colonization was found equally distributed over the 3 size groups, leading to the disruption of the FDC networks prevalent in large and abnormally large cell lymphomas and, not surprisingly, the complete effacement of the follicle, leading to the nodular or diffuse growth patterns in these lymphomas. Compression of the FDC network was observed mostly in the small-cell group, then in the large-cell group, and only exceptionally in the abnormally large category and appeared as such as a very peculiar feature of NMZL.

Sclerosis, which was infrequent, was present exclusively in large and abnormally large-cell NMZLs.

IgD Expression

An unexpected finding that had not been observed by others was the expression of IgD in 51% of our NMZL cohort. However, the expression pattern of IgD was very heterogeneous FIGURE 2; details given in TABLE 1. IgD at a very low expression level was found in 22 cases, most of them being in the small-cell category FIGURE 2A-C. Some lymphomas were partially IgD positive FIGURE 2D with both low (7 cases) and high (2 cases) expression levels. Consistent IgD expression at high levels was found in 8 lymphomas, 5 of them belonging to the small-cell category FIGURE 2F. Sixteen cases of the large-cell category expressed IgD, and only 1 of 6 cases of the abnormally large-cell category was IgD positive.

Against this background, we stained reactive lymph nodes and, albeit inconsistently, found IgD expression in marginal zone B cells, mostly at lower expression levels as compared to mantle zone B cells FIGURE 4.

Ki-67 Proliferation Index

The NMZLs had a wide range of proliferation rates (as measured by Ki-67 staining) ranging from less than 1% to 80%. Most cases fell into an intermediate proliferation range (Ki-67 <39% positivity). There was an interesting correlation between the Ki-67 index and cell size: cases with low proliferation (Ki-67 <6%) never occurred in the abnormally large size category. Highly proliferating NMZLs are more likely to have a disrupted FDC network, while follicles without colonization are more typical of intermediate and low proliferation (Supplemental FIGURE 1A, B).

Morphometrics

The measurements of cell size and nuclear size of both NMZL and nDLBCL gave us a clear view of the heterogeneity of appearance between these 2 cohorts. Both formed continuous clouds that considerably overlapped in the center. We decided to use centroblasts from toxoplasmic lymphadenitis cases as a reference for defining the dimensions of the paradigmatic large cell and its nucleus. Using 2-fold standard deviation of centroblasts as a reference, we divided the cases into 3 categories—“small,” “large,” and “abnormally large”—both for the cell size and for the nuclear size FIGURE 5A. Interestingly, with 1 exception, only nDLBCL cases fell into the category “abnormally large” for both measurements simultaneously. Thus, morphometrics yield data largely congruous but not identical to the cytomorphologic assessment. While 10 cases were assigned “abnormally large” by the pathologist, only 4 cases fell into the morphometrical category “abnormally large cell size.” Operationally speaking, when the case has abnormally large cells and nuclei (above 14 µm and 9.4 µm, respectively), it is most likely to be an nDLBCL, and at the opposite extreme of the spectrum, a lymphoma can be identified as an NMZL if the cell size and nuclear size are smaller than 10.5 µm and 7.7 µm, respectively FIGURE 5B. The cell/nucleus ratio of NMZL was significantly higher than nDLBCL FIGURE 5C, meaning that in nDLBCL, large and abnormally large cells also have a large nucleus, whereas in NMZL, the large cells can have small nuclei, resulting in a wide, often clear, cytoplasmatic rim.

Histomorphologic Comparison of NMZL With nDLBCL

Having presented the descriptive parameters of NMZL in detail, we compared the value ranges of these parameters for nDLBCL. The parameters of nDLBCL were homogeneous with regard to the cell of origin. In comparison with NMZL, several interesting parameters came up. As its name indicates, nDLBCL cases very often had a diffuse growth pattern with only few exceptional nodular and interfollicular lymphoma cases. Cases of NMZL, on the other hand, showed a mostly interfollicular growth (Supplemental Figure 2A). In nDLBCL, there was a very strong prevalence of complete effacement of the follicle and its FDC network; colonized follicles were found only in a small minority of cases; conversely, in NMZL, there was a prevalence of colonized follicles, followed by intact follicles and only a small group without follicular remnants. The compression of the FDC network was exclusive to NMZL (Supplemental Figure 2C, D). The 51% IgD positivity rate in NMZL as opposed to the 17% rate in nDLBCL may also serve as an ancillary diagnostic indicator (Supplemental Figure 2B), and this difference is statistically significant (χ² with Yates’s correction test, P = .0005, Supplemental Table 3). Proliferation rate distribution also differed between NMZL and nDLBCL, the latter showing the highest proliferative rate (Ki-67 >40%) in all cases except 1, while NMZL covered the whole range of proliferation (from <1% to 80% of Ki-67–positive cells). In general, nDLBC had significantly larger cells and nuclei (Supplemental Figure 2E, F). The overlapping region of the morphometric analysis FIGURE 5B is the core problem. Interestingly, all the differences observed for the whole cohorts were likewise found in the overlapping subgroup (Supplemental Figure 3A-D).

DISCUSSION

In this study, we present the first genome-wide methylation analysis on a cohort of NMZLs, demonstrating the possibility of separating them from nDLBCLs. The basis of this data set should nevertheless be interpreted with caution, especially when it comes to the genes differentially methylated in both lymphoma types (cf. Supplemental Table 4) since it is a long way from the methylation state to the emergence of a protein. Furthermore, there is another limitation in terms of practicability: although we collected a large cohort considering the rarity of NMZL, the requirement of a high amount of high-quality DNA for the methylation profile assay is very stringent, limiting the number of profiles (35 of 89 cases) we were able to obtain.

The difficulties that pertained to this differential diagnosis are in part due to the lack of recognizability of the large-cell variant of NMZL. Up to now, the histologic transformation of NMZL associated with clinical progression is considered an “aggressive entity” and a DLBCL.^17-19 Since we have large-cell NMZL in our cohort, this redefinition of an NMZL becoming a DLBCL is nosologically misleading based on published data on gastric small- and large-cell marginal zone lymphoma.^12,20-23 The article by Spina et al,²⁴ who profiled 35 cases of NMZL, used the 2008 World Health Organization definition of this lymphoma as a “small-cell lymphoma,” and consistently, the highest proliferation rate they report was in the range of 45%. Thus, it is most likely that the large-cell variants that we are dealing with in our cohort were missing. The same restriction to small-cell NMZL cases is present in the work of Pillonel et al²⁵ dealing with sequencing NMZL, Granai et al²⁶ on IGHV mutational status, and the meta-analysis by Vela et al²⁷ on the mutational landscape of NMZL. Consequently, these molecular studies should be repeated on the basis of our cohort.

Future work will also concentrate on addressing whether or not the large-cell variant of NMZL corresponds to the BN2 cluster of DLBCL, as characterized by Schmitz et al²⁸ and validated by Wright et al.²⁹

To overcome this dilemma, we generated an operational tool with techniques already available in routine pathology practice to help recognize large-cell NMZL and exclude nDLBCL. As detailed in FIGURE 6 the procedure is composed of a decision tree (cyclin D1, CD10, morphometrics) followed by an evaluation of ordinal or continuous scale parameters (growth pattern, follicular colonization, FDC network, IgD, Ki-67) and a final majority vote if these parameters fall under the nondecisive value ranges. These aspects allow us to make a statement about the likelihood of the aggressiveness of NMZL; it is true that some of our NMZL cases show features of aggressive malignancy, such as a high proliferation rate, large-cell morphology, and a diffuse growth pattern with destruction of the lymph node architecture. However, it is very rare that all these features appear combined in NMZL, in contrast to the consistency that was observable within the nDLBCL group. Our findings led us to propose this new diagnostic procedure as a helpful tool for pathologists in the differential diagnosis of NMZL. Morphometrics is particularly interesting also in the light of future developments when digital pathology and artificial intelligence will enter routine practice.

Acknowledgments

The authors thank Dr Martin L. Hansmann, Dr Markus Tiemann, Dr Hans-Heinrich Wacker and Dr Harald Stein for the cases contributed; Dr Alexander Grunenberg for his help in the collection of cases; Dr Hans Kestler from the institute of Medical Systems Biology of Ulm University for his contribution to the methylation omics data analysis; and Marlene Schuster and Juliane Nell for technical help.

Conflict of interest disclosure: The authors have nothing to disclose.

REFERENCES