A 55-year-old, previously-well man presented with a new-onset seizure characterized by right head tilt, visual flashes of red light and loss of consciousness. Aside from type 2 diabetes mellitus, hypertension and dyslipidemia, there was no relevant past medical history. He reported daily marijuana use. Neurological exam showed only frontal dysfunction (impaired serial sevens, bilateral Luria sequence difficulty, poor verbal fluency).

MRI revealed ill-defined, robustly enhancing lesions in the cortical-subcortical regions of the cerebral hemispheres bilaterally (Figure 1), with disproportionate surrounding T2-–fluid-attenuated inversion recovery hyperintensity at the larger left frontal lesion. The lesions demonstrated a T2 isointense signal relative to the cortex with diffusion restriction. Parts of the left frontal cortical lesion also exhibited bright T2 cortical signal with associated cortical thinning, reminiscent of focal scarring (a “burnt-out” appearance). Based on neuroimaging, the initial differential considerations were broad. Given the multifocal nature of the lesions, metastasis was considered; however, the imaging features were not entirely typical, and a subsequent screening body CT was negative for malignancy. Additional considerations included a granulomatous inflammatory process (e.g., sarcoidosis, granulomatosis with polyangiitis), lymphoma, histiocytosis and granulomatous infections (tuberculosis, fungal infections). Given the apparent cortical thinning in some areas and the multifocal but predominantly cortically based pattern, a high-grade glioma was less favored.

Figure 1. Contiguous slices of MRI axial T2-weighted sequence (A, B) with magnified representative images (A1, B1) showing an isointense cortical-subcortical lesion (arrow) in the left frontal lobe with surrounding T2 hyperintensity. There is an associated avidly bright T2 signal and mild cortical thinning in the adjacent cortex (dashed arrows). The lesion demonstrates diffusion restriction with diffusion weighted imaging bright signal (C) and low apparent diffusion coefficient (not shown) and robust enhancement on post-gadolinium T1-weighted image (D). Axial T2 (E) and post-gadolinium T1 (F) images show other smaller enhancing multifocal lesions in the right frontal operculum (arrow) and right temporal lobe (dashed arrows).

A left frontal craniotomy for resection of the left frontal lesion was performed. Preliminary intraoperative pathology from the frozen section suggested a high-grade glioma. The patient was discharged on dexamethasone (2 mg BID) following an uneventful postoperative course, with follow-up planned under neuro-oncology. Interval MRI demonstrated no residual left frontal lesion. Unusually, spontaneous regression of the right frontal and right temporal lesions was observed. This was atypical for glioblastoma (GBM) but was initially attributed to the effects of dexamethasone. ^{Reference Goh, See, Ang and Ng1}

Initial pathological examination (Figure 2) showed largely reactive astrocytes, with a sparse population of mitotically active atypical cells confined to a focal area, which, on immunohistochemistry, expressed GFAP accompanied by a high Ki-67 proliferation index. Therefore, the histology was interpreted as a high-grade glioma, which was additionally immunonegative for IDH1 R132H with retention of ATRX. Flow cytometry was not sent.

Figure 2. Hematoxylin-eosin reveals a “washout” section featuring atypical cells with distorted nuclear contour and presence of mitosis (arrow) (A); strong GFAP expression (B); high Ki-67 immunolabeling of atypical cells (C); CD68 immunolabels numerous histiocytes among the atypical cells (D); PAX5 immunolabels scattered atypical lymphoid cells displaying distorted nuclear contour with marked pleomorphism, lacking the classical angiocentric pattern (query effects of post-steroid treatment) (E); and CD20 immunolabels small aggregates of atypical lymphoid cells with profoundly distorted cellular contour but without the classical angiocentric pattern (F).

At the time of the initial neuro-oncology review, the next-generation sequencing (NGS) results, exact tumor grade and MGMT promoter methylation status were pending. A decision was made to proceed with radiation and temozolomide, given the presumptive diagnosis of GBM (isocitrate dehydrogenase–wildtype). Subsequently, NGS identified two different TP53 mutations (R273H with variant allele frequency (VAF) 13.1% and H193Y with VAF 7.3%) and an unmethylated MGMT promoter, but an absence of molecular alterations associated with GBM (TERT promoter mutation, EGFR amplification, +7/−10 chromosomal copy-number variations).

Notably, NGS identified an unusual finding: a MYD88 L265P mutation (VAF 14.1%), a gene involved in innate immune mechanisms via the NF-κB signaling pathway. Recurrent somatic mutations at this hotspot (L265P) are considered one of the genetic hallmarks of primary CNS lymphoma (PCNSL), occurring in ∼ 70–85% of the cases. ^{Reference Yamaguchi, Ohka and Kitano2,Reference Nayyar, White and Gill3} The MYD88 L265P mutation is also detectable in nearly all cases of lymphoplasmacytic lymphoma/Waldenström’s macroglobulinemia (involving the CNS as Bing–Neel syndrome). ^{Reference Varettoni, Arcaini and Zibellini4}

The significance of this mutation in the context of GBM was uncertain, prompting further work-up for a possible lymphoproliferative disorder. Work-up for Waldenström’s macroglobulinemia was negative. Additional immunohistochemistry markers (e.g., CD3, CD20, PAX5, CD138, CD68, CD163, CD79a) did not clearly reveal a lymphoproliferative disorder. The specimens were sent for DNA methylation profiling at the National Institutes of Health (NIH) (Bethesda), which did not classify the tumor into a specific entity but suggested that it was not compatible with glioma. Instead, the methylation signature was more suggestive of a hematologic process. The stained sections were reviewed by a hematopathologist at NIH and a hematopathologist internally, who both found no evidence for lymphoma. As the classical angiocentric pattern of PCNSL was lacking, the case was consulted internally with a second neuropathologist, who felt that malignant glioma was still a possibility.

This prompted a repeat multidisciplinary review with renewed attention to the preoperative steroid exposure. The patient had received dexamethasone prior to the MRI (single 10 mg intravenous dose and four 4 mg oral doses) and continued on this regimen until the surgery was performed two days later. A re-examination of the pathology slides was performed internally by two neuropathologists and a hematopathologist, which identified scattered aggregates of PAX5- and CD20-positive atypical B lymphocytes within a background of reactive gliosis and lymphohistiocytic infiltrate, suggestive of diffuse large B-cell lymphoma. Repeat NGS testing confirmed the previously detected mutations.

The final diagnosis was revised to PCNSL. He demonstrated a good response to temozolomide and radiation therapy (no residual disease on MRI head; negative bone marrow examination, negative positron emission tomography scan). He tested negative for HIV. Three-month follow-up MRI showed an out-of-radiation-field recurrence in the right occipital lobe. For ongoing management, the patient was started on ibrutinib as a possible bridge to autologous transplantation, an approach chosen to reduce the risk of combined radiation and methotrexate leukoencephalopathy.

Non-diagnostic biopsy in PCNSL due to preoperative steroid exposure is a well-recognized phenomenon, attributed to the lymphocytotoxic effects of corticosteroids, with systematic meta-analysis indicating a relative risk of approximately 3, regardless of steroid duration, dose or preoperative taper. ^{Reference Tosefsky, Rebchuk, Martin, Chen, Yip and Makarenko5} Prior steroid administration can significantly complicate histopathological interpretation, increasing diagnostic ambiguity or misinterpretation in up to 50% of cases in some series. ^{Reference Rice, Renowden, Urankar, Love and Scolding6} Cases of steroid-exposed PCNSL mimicking inflammatory demyelination on pathology are known. ^{Reference Barrantes-Freer, Engel and Rodríguez-Villagra7} However, other studies report no significant histopathological or radiographic alterations from steroid exposure in PCNSL, especially following a brief course, and do not support the need for re-biopsy. ^{Reference Bullis, Maldonado-Perez and Bowden8,Reference Porter, Giannini and Kaufmann9}

The left frontal lesion demonstrated an area of cortical thinning that may have resulted from tumor regression due to steroid exposure. The tumor may have further regressed during the two-day interval between MRI and surgery. A closer timed MRI to the surgery may have been more beneficial. A few factors led to increased suspicion for lymphoma (atypical MRI features for GBM, steroid-responsiveness, MYD88 mutation and lack of GBM molecular features), and close re-evaluation of pathology led to the revised diagnosis. This exemplifies the usefulness of molecular testing in raising a diagnostic possibility that was not initially suspected.

In summary, our case highlights how even a short duration of steroid exposure can complicate histopathological interpretation in PCNSL, and an accurate diagnosis requires attentiveness to all aspects of the diagnostic pathway, including molecular genetics testing.