Main

Classical Hodgkin lymphoma and anaplastic large cell lymphoma are both characterized by the presence of CD30-positive tumor cells.1, 2 Furthermore, anaplastic large cell lymphoma is divided into ALK+ cases, presenting with a translocation affecting the ALK locus,3, 4, 5 and ALK cases, in which translocations involving DUSP22 have been described in a fraction of cases.6 Whereas classical Hodgkin lymphoma often presents in young adults,2 anaplastic large cell lymphoma arises at various ages. ALK+ anaplastic large cell lymphoma is more frequently found in children and adolescents, whereas ALK anaplastic large cell lymphoma is often observed in older patients.7, 8 Classical Hodgkin lymphoma can be cured in about 90% of patients with currently applied treatment approaches.9 However, ALK anaplastic large cell lymphoma often shows poor outcome,10 particularly in the elderly. Therefore, a correct diagnostic classification of both diseases is mandatory for an appropriate therapeutic strategy.

Although classical Hodgkin lymphoma is a B-cell neoplasm, the tumor cells, Hodgkin, and Reed-Sternberg cells—have largely lost their B-cell phenotype.11, 12 Anaplastic large cell lymphoma is a T-cell neoplasm, but the tumor cells often present with a null phenotype and can be negative for all T-cell markers.13, 14 In ALK+ anaplastic large cell lymphoma immunohistochemical detection of ALK expression is a decisive marker for the differential diagnosis, but this marker cannot be applied for ALK anaplastic large cell lymphoma. In some cases of anaplastic large cell lymphoma cytotoxic molecules like TIA1, perforin, and granzyme B are expressed.15 However, there are rare Hodgkin lymphoma cases in which expression of these cytotoxic molecules can also be found in Hodgkin and Reed-Sternberg cells.15, 16 In the past years, the B-cell transcription factor PAX5 has been discovered to be still expressed at low level in Hodgkin and Reed-Sternberg cells of most classical Hodgkin lymphoma cases.11, 17, 18 Classical Hodgkin lymphoma can therefore in most cases reliably be distinguished from ALK anaplastic large cell lymphoma. However, some PAX5-negative classical Hodgkin lymphoma cases18 are difficult to be distinguished from ALK anaplastic large cell lymphoma. Ancillary molecular studies showing rearrangements of B- or T-cell receptors can be helpful in these cases. However, owing to a low tumor cell content, mostly observed in Hodgkin lymphoma and sometimes also found in anaplastic large cell lymphoma, they do not always give indicative results.

Therefore, the aim of the present study was to identify a set of characteristic markers, which can successfully be applied in the differential diagnosis of classical Hodgkin lymphoma and ALK anaplastic large cell lymphoma.

Materials and methods

Microarray Data Analysis and Selection of Target Genes

Gene expression data from the studies by Brune et al19 and Eckerle et al20 were reevaluated. A supervised comparison between ALK anaplastic large cell lymphoma cases and all Hodgkin lymphoma cases from both studies was performed. After the global filtering, a two-sample t-test (assuming equal variance in both groups) was applied to identify genes that were differentially expressed between the two groups. To account for multiple testing, the false discovery rate was used, as described by Benjamini and Hochberg.21 For visualization, a heat map of all significantly upregulated or downregulated transcripts was created (P<0.05, false discovery rate <0.1). Genes to be immunohistochemically validated on the protein level were chosen with regard to the strongest upregulation, the most homogenous expression patterns in the lymphoma samples, and the commercial availability of antibodies.

Tissue Samples and Immunohistochemical Stainings

Thirty classical Hodgkin lymphoma cases (10 nodular sclerosing subtype, 10 mixed cellularity, and 10 Hodgkin lymphoma cases with expression of granzyme B—9 mixed cellularity and 1 lymphocyte depleted), 20 nodal ALK anaplastic large cell lymphoma as well as 11 primary cutaneous ALK anaplastic large cell lymphoma, diagnosed according to the World Health Organization 2008 classification,2 were selected from the archives of the Dr Senckenberg Department of Pathology in Frankfurt. We specifically selected granzyme B expressing Hodgkin lymphoma cases, because in classical Hodgkin lymphoma with PAX5,+ Hodgkin and Reed-Sternberg cells sometimes show coexpression of cytotoxic molecules, in particular granzyme B, thus sharing even more features with anaplastic large cell lymphoma. A validation cohort of 30 additional classical Hodgkin lymphoma cases, 13 ALK and 11 ALK+ anaplastic large cell lymphoma cases was selected from the archives of the Department of Experimental, Diagnostic and Specialty Medicine, Haematopathology Section, S. Orsola-Malpighi Hospital, University of Bologna, Italy and stained on tissue microarray format. Stainings were considered positive if at least 50% of the tumor cells were positive. The local ethics committees of the Universities in Frankfurt and Bologna approved the study. Detailed information on the immunohistochemical profiles of the cases are found in Table 1. Whole tissue sections were stained for CD83, MDC/CCL2, TUBB2B, STAT3, and CAPN2 using the Peroxidase-FLEX EnVision Kit (Dako) as described previously.22, 23 The antibodies used, dilutions, and providers are listed in Table 2. Antigen unmasking was performed for 3 min in a pressure cooker or microwave in Tris-EDTA at pH 8.0 or citrate buffer at pH 6.0. All cases had previously been stained for CD20, CD30, CD3, PAX5, TIA1, granzyme B, and LMP1. Hierarchical cluster analysis was conducted on the four immunohistochemical stainings included in the classifier using R and Bioconductor.

Table 1 Expression of immunohistochemical markers in different subsets of classical Hodgkin lymphoma and anaplastic large cell lymphoma
Table 2 Antibodies and dilutions applied for immunohistochemistry

Results

Gene Expression Analysis

In a reevaluation of gene expression analyses performed by Brune et al19 and Eckerle et al20 a supervised comparison of the gene expression profiles of microdissected tumor cells of classical Hodgkin lymphoma and ALK anaplastic large cell lymphoma was performed. Eighteen transcripts were found to be overexpressed in Hodgkin and Reed-Sternberg cells of classical Hodgkin lymphoma compared with the tumor cells of ALK anaplastic large cell lymphoma and 15 transcripts were overexpressed in the tumor cells of ALK anaplastic large cell lymphoma (fold change<−1.8 or >1.8, P-value<0.05, false discovery rate<0.1, Table 3, Figure 1). Based on significant differential expression, macrophage-derived chemokine/chemokine ligand 22 (MDC/CCL22) (39.8-fold upregulated in Hodgkin and Reed-Sternberg cells), tubulin beta 2B (TUBB2B) (9.9-fold upregulated in Hodgkin and Reed-Sternberg cells), CD83 (8.4-fold upregulated in Hodgkin and Reed-Sternberg cells), signal transducer and activator of transcription 3 (STAT3) (4.2-fold upregulated in Hodgkin and Reed-Sternberg cells), and calpain 2 (CAPN2) (5.0-fold upregulated in the tumor cells of ALK anaplastic large cell lymphoma) were selected for immunohistochemical studies (Figure 1).

Table 3 Transcripts differentially upregulated and downregulated in Hodgkin and Reed-Sternberg cells of classical Hodgkin lymphoma compared with tumor cells of ALK anaplastic large cell lymphomaa
Figure 1
figure 1

Heat map of the genes differentially expressed in the microdissected tumor cells of classical Hodgkin lymphoma and ALK anaplastic large cell lymphoma from the studies by Brune et al19 and Eckerle et al.20 NA, not annotated; red color: high expression; black color: intermediate expression; green color: low expression. Genes selected for immunohistochemical studies are highlighted by arrows.

Immunohistochemical Investigation of Differentially Expressed Genes on Protein Level

In majority of the 30 classical Hodgkin lymphoma cases of the first series, stained on whole tissue sections, Hodgkin and Reed-Sternberg cells were positive for TUBB2B (80%) as well as STAT3, CD83, and MDC/CCL22 (83% each, Table 1 and Figure 2). MDC/CCL22 and CD83 were strongly expressed in the cytoplasm and Golgi field of Hodgkin and Reed-Sternberg cells, whereas TUBB2B showed a homogeneous expression in the cytoplasm. An enhanced STAT3 expression was observed in the nucleus and cytoplasm of Hodgkin and Reed-Sternberg cells. A weak expression of STAT3 was also observed in histiocytes and the tumor cells of some anaplastic large cell lymphoma cases (Figures 2e and f). Only 10% of the 31 anaplastic large cell lymphoma in the first series presented with a strong nuclear and cytoplasmic STAT3 expression and cytoplasmic TUBB2B expression. Like the Hodgkin and Reed-Sternberg cells, a subgroup of macrophages/dendritic cells stained positive for CD83 and MDC/CCL22 in some cases, whereas anaplastic large cell lymphoma tumor cells were consistently negative in this first series. Interestingly, among 10 classical Hodgkin lymphoma cases with expression of granzyme B, Hodgkin and Reed-Sternberg cells were less frequently positive for the Hodgkin classifier markers, when compared with the typical classical Hodgkin lymphoma cases being negative for cytotoxic molecules. In these cases, MDC/CCL22 was positive in only 5/10 cases (50%), CD83 was positive in 7/10 cases (70%), TUBB2B and STAT3 were each positive in 8/10 cases (80%).

Figure 2
figure 2figure 2

Immunhistochemical stainings differentiating classical Hodgkin lymphoma and anaplastic large cell lymphoma. (a and b) CD83; (c and d) MDC/CCL22; (e and f) STAT3; (g and h) TUBB2B; (i and j) CAPN2. (a, c, e and g) Hodgkin and Reed-Sternberg cells of classical Hodgkin lymphoma show expression of the respective immunohistochemical markers. (b, d, f and h) Tumor cells of ALK-negative anaplastic large cell lymphoma are negative for the respective markers. (i) Hodgkin and Reed-Sternberg cells are negative for CAPN2, whereas a subset of lymphocytes is positive. Inset: classical Hodgkin lymphoma case with CAPN2-positive Hodgkin and Reed-Sternberg cells. (j) Tumor cells of ALK anaplastic large cell lymphoma show expression of CAPN2.

CAPN2 was expressed in the tumor cells of both classical Hodgkin lymphoma and ALK anaplastic large cell lymphoma cases (23 and 45%, respectively) and was therefore not further considered.

In the training set, stained on tissue microarray, Hodgkin and Reed-Sternberg cells were also positive in majority of the 30 classical Hodgkin lymphoma cases for TUBB2B and MDC/CCL22 (77% each), STAT3 (93%), and CD83 (90%). In contrast, tumor cells of the 13 ALK anaplastic large cell lymphoma cases were only infrequently positive for TUBB2B (8%), STAT3 (15%), and CD83 (31%). They were always negative for MDC/CCL22.

Establishment of a Classifier to Efficiently Distinguish Classical Hodgkin Lymphoma and ALK Anaplastic Large Cell Lymphoma

A scoring system (0–4) was created reflecting the number of positive stainings (MDC/CCL22, CD83, STAT3, and TUBB2B) per case (0,no positive stainings; 4,all four stainings positive). The first case series, stained on whole tissue sections, was evaluated applying this scoring system. A threshold of ≥2-positive stainings proved to be optimal for the classification as Hodgkin lymphoma, whereas cases with a score 0–1 were regarded as anaplastic large cell lymphoma. Applying this scoring system to the second series of cases stained on tissue microarray, all 30 classical Hodgkin lymphoma cases were correctly classified and only one out of 13 ALK anaplastic large cell lymphoma cases was misleadingly classified as Hodgkin lymphoma (42/43, 97%, Table 4). Applying the classifier to all cases, two out of 60 Hodgkin lymphoma cases and two out of 34 nodal ALK anaplastic large cell lymphoma were misclassified, yielding an accuracy of 96% (90/94). The misclassified Hodgkin lymphoma cases were two cases with granzyme B expression. In an unsupervised hierarchical clustering by the classifier, two large branches were observed (Figure 3).

Table 4 Scores obtained in different subsets of classical Hodgkin lymphoma and anaplastic large cell lymphoma applying the scoring system including the immunohistochemical markers TUBB2B, STAT3, MDC, and CD83
Figure 3
figure 3

Unsupervised hierarchical clustering by the classifier of four immunohistochemical stainings for MDC/CCL22, CD83, STAT3, and TUBB2B. Red color: tumor cells positive, green color: tumor cells negative. Color code: yellow: classical Hodgkin lymphoma, green classical Hodgkin lymphoma with granzyme B expression, red: ALK anaplastic large cell lymphoma, orange: cutaneous anaplastic large cell lymphoma, blue: ALK+ anaplastic large cell lymphoma.

Discussion

Several attempts have been made to establish classifiers differentiating between ALK anaplastic large cell lymphoma and peripheral T-cell lymphoma not otherwise specified,24, 25, 26 but to our knowledge this has not been achieved for differential diagnosis of classical Hodgkin lymphoma and ALK anaplastic large cell lymphoma. The aim of the present study was to find new immunohistochemical markers which can be applied in the differential diagnosis of ALK anaplastic large cell lymphoma and classical Hodgkin lymphoma in daily routine practice. Our classifier was established by the reevaluation of gene expression studies of microdissected tumor cells and shows a high specificity and sensitivity.

Owing to overexpression of several genes in the Hodgkin and Reed-Sternberg cells of classical Hodgkin lymphoma we were able to distinguish between ALK anaplastic large cell lymphoma and classical Hodgkin lymphoma. The one gene overexpressed in ALK anaplastic large cell lymphoma did not prove to be diagnostically powerful. One of the genes upregulated in classical Hodgkin lymphoma was TUBB2B, a member of the beta-tubulin family, involved in the formation of microtubuli and therefore has a role in several important cellular processes, including mitosis. TUBB2B has so far not been described to be upregulated in Hodgkin and Reed-Sternberg cells, whereas another member of the beta-tubulin family, TUBB3, was observed to be expressed in both the tumor cells of classical Hodgkin lymphoma and anaplastic large cell lymphoma.27 Recently, refusion of Hodgkin and Reed-Sternberg cells with a persisting microtubule bond between daughter cells could be demonstrated to be an important feature of giant cell formation.28 However, the reason of a selective tubulin overexpression in the Hodgkin and Reed-Sternberg cells as demonstrated in this study so far remains obscure.

The remaining significantly overexpressed genes are related to the interactions between Hodgkin and Reed-Sternberg cells and their microenvironment. CD83 is usually expressed in mature dendritic cells29 and triggers CD4+ T-cell maturation, but is also expressed by activated B cells and involved in the modulation of B-cell receptor signaling.30 CD83 has previously been described to be upregulated in Hodgkin and Reed-Sternberg cells of Hodgkin lymphoma.31 Transgenic overexpression of CD83 in B cells resulted in an upregulation of the major histocompatibility complex II (MHC II) and interleukin 10.30 However, MHC II is shown to be downregulated in a subset of classical Hodgkin lymphoma cases,32 partly owing to translocations involving the MHC II transactivator (CIITA).33 Decreased MHC class II expression has been linked to reduced tumor cell immunogenicity. A strong expression of STAT3 was observed in Hodgkin and Reed-Sternberg cells in the present and previous studies.34, 35, 36 STAT3 is mainly expressed in macrophages and dendritic cells and can exhibit—depending on the stimulating cytokine—both pro- and antiinflammatory properties.37 However, STAT3 is also found to be expressed in ALK+ anaplastic large cell lymphoma.38 In contrast, pediatric cases of ALK anaplastic large cell lymphoma were negative for phosphorylated STAT3.39 Our results fit well with the observations made by Piva et al,26 when considering the number of ALK anaplastic large cell lymphoma strongly expressing phosphorylated STAT3 in most if not all neoplastic cells. This can also be observed on RNA level in the gene expression arrays (Figure 1). However, it is important to point out, that in the present study only the amount of STAT3 protein was assessed, whereas phosphorylation and activation status were not evaluated. MDC/CCL22 is usually also expressed by antigen-presenting cells and is upregulated by TH2-type cytokines.40 It was shown to be expressed in Hodgkin and Reed-Sternberg cells of classical Hodgkin lymphoma41 and age-related Epstein–Barr virus-associated B-cell lymphoproliferative disorders.42 MDC/CCL22 and TARC/CCL17 share the receptor CCR4, by which TH2 and regulatory T cells can be attracted by Hodgkin and Reed-Sternberg cells.43, 44 However, it could recently be shown that the majority of T cells in classical Hodgkin lymphoma microenvironment represents TH1 cells.45

In conclusion, three of four genes overexpressed in the Hodgkin and Reed-Sternberg cells in comparison with the tumor cells of ALK anaplastic large cell lymphoma are also expressed in antigen-presenting cells and are related to the interactions between Hodgkin and Reed-Sternberg cells and surrounding T cells. Moreover, Hodgkin and Reed-Sternberg cells are known to express several additional markers usually expressed by antigen-presenting cells.46, 47, 48Although Hodgkin and Reed-Sternberg cells have usually lost their B-cell phenotype, they have maintained the features for an interaction with T cells, which is not observed in the tumor cells of ALK anaplastic large cell lymphoma. Interestingly, Hodgkin lymphoma cases with granzyme B expression were less frequently positive for MDC/CCL22 and CD83 than conventional classical Hodgkin lymphoma cases. These cases may therefore differently interact with their microenvironment and may be closely related to anaplastic large cell lymphoma.

In summary, the genes identified in the present study reflect a different pathogenesis of classical Hodgkin lymphoma and ALK anaplastic large cell lymphoma. By closely interacting with surrounding T cells, Hodgkin and Reed-Sternberg cells have maintained parts of their B-cell identity and show a fundamentally different interaction with their microenvironment than tumor cells in ALK anaplastic large cell lymphoma. These features can be practically applied in the diagnostic distinction between classical Hodgkin lymphoma and ALK anaplastic large cell lymphoma. However, owing to the limited case number investigated in the present study, our newly established classifier should be validated in a large patient series with clinical data.