Reliable cell and tissue morphology-based diagnosis of endemic Burkitt lymphoma in resource-constrained settings in Ghana

The study protocol was approved by the institutional review boards of Kwame Nkrumah University of Science and Technology (CHRPE/AP/175/17) and Korle-Bu Teaching Hospital (KBTH; IRB/00080/2016) and by the Heads of Department of the pathology units of Komfo Anokye Teaching Hospital (KATH) and KBTH. Written informed consent was obtained from the parents/guardians on admission before inclusion of patients in the study.

Patients (N = 50) with head, neck, or abdominal masses suggestive of eBL were recruited into the study in 2018 according to the recommendation of the attending physicians (Table 1). Tumour FNAs were obtained at admission for preparation of smears for diagnostic purposes. Additional smears were stored at 4 °C for retrospective and independent assessment of cell morphology (Fig. 1a) and cytogenetics (Fig. 2).

Archival FFPE tumour samples (N = 111) from patients diagnosed with eBL or other childhood lymphomas between 2010 and 2017 were retrieved (Table 1), and new sections (3–4 μm thickness) were cut. The tissue sections were mounted on pre-coated slides (Leica BOND™ Plus) and stored at − 20 °C for later haematoxylin-eosin staining (Fig. 1b) and additional tests.

Only a limited number of specimen slides were available for retrospective analysis, and complete morphological, immunohistochemical, and FISH characterization was therefore not possible.

Retrospective morphological assessment of tumour FNA smears and FFPE tumour sections was performed by two experienced morphologists (CS, PHN) blinded to the original diagnosis. For FNAs, the diagnosis of eBL was based on finding of monotonous, medium-sized, blastoid cells with basophilic cytoplasm usually containing lipid vacuoles and round nuclei with finely clumped chromatin containing multiple medium-sized basophilic nucleoli. For the FFPE sections, the eBL diagnosis was based on finding of a solid tumour composed of medium-sized, blastoid cells with diffuse monotonous growth pattern, cohesive with squared-off borders of retracted basophilic cytoplasm (usually with visible lipid vacuoles), and round nuclei with finely clumped chromatin containing multiple medium-sized basophilic nucleoli. The eBL tumours also contained many tingible body macrophages, giving rise to a characteristic “starry sky” appearance. Mitotic figures were common due to the high proliferation rate of the tumour cells. Only vital and sufficiently fixated tumour tissue was evaluated, and samples containing solely necrotic tissue and/or autolyzed tumour tissue were discarded.

The frozen sections were brought to room temperature and fixed to the slides (58 °C, 45 min). Paraffin was removed in three changes of Tissue-Clear (Sakura Finetek Europe B.V.; 10 dips in the first two changes and incubation (15 min) in the third), followed by three changes of 100% ethanol (10 dips in the first two and incubation (1 min) in the last change). The sections were subsequently washed in two changes of 96% ethanol followed by two changes of 70% ethanol (10 dips in the first, and incubation (1 min) in the second change of each). The sections were then rinsed by incubation in two changes (3 min each) of wash buffer (Tris-HCl buffer, pH 7.6), followed by pre-treatment buffer (MES-buffer, pH 6.55) at boiling temperature (10 min) in a microwave oven (Whirlpool, CRISP) and cooling (15 min) in the pre-treatment buffer. The sections were then rinsed in two changes of wash buffer as above, blotted with tissue paper and digested with pepsin (ZytoVision, GmbH, Germany; 8 min, room temp), rinsed again as above, dehydrated in two changes of 70, 96, and 100% ethanol (10 dips in first change of each, and incubation (2 min) in the second). To further enhance dehydration, sections were subsequently air-dried (room temp, 15 min). c-myc/igh fusion probes (ZytoVision, Germany, z-2105-200) and c-myc dual split probes (DAKO, Agilent, USA) were added (1.5-10 μL, depending on the size of the section), the sections covered with coverslips (12 × 12 or 22 × 22 mm²) and sealed with Fixogum (ZytoVision, E-4005-126). Following incubation (85 °C, 5 min followed by 37 °C, overnight) of the slides on a hybridiser (ThermoBrite, Statspin, Abbot Molecular), the seals and coverslips were removed, and the sections rinsed (1 min) in Stringent buffer (SSC Buffer 20x concentrate (Sigma), Triton X-100 (Sigma)). The sections were transferred to a Coplin jar filled two-thirds with Stringent buffer and incubated (64 °C, 10 min) in a water-bath to remove unspecific binding of the probes. Finally, the sections were rinsed in two changes of wash buffer, dehydrated, and mounted in fluorescence mounting medium (10 μL) containing Dapi stain (Veatashield®, Vector Lab. Inc., Burlingame, CA 94010). The above procedures were done in batches of ten sections plus known positive (BL tissue) and negative (reactive lymph node) controls.

FNA smears were fixed (3.7% formalin, 5 min) and rinsed in two changes of wash buffer (3 min each). The smears were further dehydrated as described above and divided into two halves. Probes as above (10 μL each) were added to one half each. The slides were covered, sealed, hybridized and processed further as described above.

Each FISH slide was independently evaluated for c-myc break (Fig. 2a) and c-myg/igh fusion (Fig. 2b) by three pathology microscopists blinded to the original diagnosis of the patient, using an Olympus BX61 fluorescence microscope equipped with an attached Zeiss camera. For the split probe, the sample was scored positive if > 10% of the red and green signals were clearly separated (Fig. 2a). For the fusion probes, the sample was scored positive if > 10% of the red and green signals were immediately next to each other or overlapping (yellow signal) (Fig. 2b).

Expression of C-MYC was evaluated according to the guidelines for routine diagnostic work-up at the Department of Pathology, Herlev Hospital [11]. FFPE tissue sections were pre-treated (pH = 9.0) on a PT Link pre-treatment module (Dako), including paraffin removal, rehydration, and epitope retrieval, and subsequently stained in a Dako Autostainer Link 48, using EnVision FLEX+ visualization kits (Dako) and monoclonal C-MYC antibody Epitomics; clone y69/EP121, 1:100 dilution EnVision Flex Antibody Diluent). All steps were completed according to the manufacturer’s instructions. The stained sections (Fig. 1c) were evaluated and scored by two experienced haemato-pathologists using a double-headed microscope (Olympus BX51, equipped with a colour view camera and analySIS getIT 5.0 software (Soft Imaging Systems Munster, Germany)). MYC expression was evaluated on full slide sections in hot spot areas and all staining intensities were included as previously described [12].

Sensitivity and specificity, and their confidence intervals, were calculated as described in detail elsewhere [13], using the MedCalc online calculator (https://​www.​medcalc.​org/​calc/​diagnostic_​test.​php).

Blinded Giemsa-stained FNA tumour smears (Fig. 1a) from 28 of the patients (17 with an original eBL diagnosis) from 2018 (Table 1) were available for retrospective morphological review. The results were in good agreement with the original diagnoses made (Fig. 3a and Additional file 1: Table S1).

We next evaluated freshly stained haematoxylin-eosin sections (Fig. 1b) from the archival tissue blocks (Table 1). A reliable retrospective assessment was only possible for 85 (including 42 from patients with an original eBL diagnosis), as the remaining samples had to be discarded because of poor specimen quality (autolysis or no viable tumour tissue due to tumour necrosis) (Additional file 1: Table S2). The overall sensitivity, specificity, and accuracy estimates were somewhat lower than for the FNA smears (Fig. 4a), probably related to the overall lower quality of the archival compared to the fresh samples. The ages of the samples were not significantly related to rejection rate, suggesting that inadequate sample preparation rather than inadequate storage was the main cause of low sample quality.

To further assess the accuracy of the original cell/tissue morphology-based diagnosis, we also analysed FNA smears from 20 of the 2018-patients (including 13 patients originally diagnosed as eBL) for FISH evidence of breakage of the c-myc oncogene (Fig. 2a). The sensitivity, specificity, and accuracy estimates (Fig. 3b and Additional file 1: Table S3) were lower than above. We did not find evidence of c-myc breakage in one of the patients originally diagnosed as eBL, whereas c-myc breakage was detected in four patients originally diagnosed as non-eBL. While c-myc breakage is not specific for eBL, lack of c-myc breakage is rare in eBL, and this analysis thus suggests some over-diagnosis originally. However, cryptic rearrangements in some cases cannot be formally ruled out [14]. Examination of slides for FISH evidence of c-myc translocation into the igh locus (Fig. 2b) yielded similar results (Fig. 3c and Additional file 1: Table S4).

Corresponding FISH analysis of archival FFPE sections showed high sensitivity but low specificity (Fig. 4b-c and Additional file 1: Tables S5-S6). The low specificity was due to the fact that we did not find evidence of c-myc breakage in half (7/14) of the patients originally diagnosed as eBL. As lack of c-myc breakage is rare in eBL, this finding supports the above indication of a degree of over-diagnosis when supportive molecular evidence of eBL is not available. This conclusion is further supported by examination of biopsy sections for immuno-histochemical evidence of C-MYC expression (Fig. 1c). Breakage and translocation of c-myc is often associated with high expression of C-MYC protein [15], and immunohistochemistry is generally less sensitive to sample deterioration than FISH [16, 17]. This analysis (Fig. 4d and Additional file 1: Table S7) showed low specificity due to absence of C-MYC expression in biopsies from eight of 34 patients originally diagnosed as eBL.