Acta Neuropathologica

, Volume 115, Issue 3, pp 345–356

CNS T-cell lymphoma: an under-recognized entity?

Authors

    • Department of PathologyStanford University Medical Center
  • Christopher Y. Park
    • Department of PathologyStanford University Medical Center
  • William D. Howell
    • Department of PathologyStanford University Medical Center
  • Lawrence T. Smyth
    • Department of PathologyKaiser Permanente Medical Center
  • Mayuri Desai
    • Department of PathologyKaiser Permanente Medical Center
  • Diane M. Carter
    • Department of PathologyKaiser Permanente Medical Center
  • Hannes Vogel
    • Department of PathologyStanford University Medical Center
Original Paper

DOI: 10.1007/s00401-007-0338-y

Cite this article as:
Dulai, M.S., Park, C.Y., Howell, W.D. et al. Acta Neuropathol (2008) 115: 345. doi:10.1007/s00401-007-0338-y

Abstract

The incidence of CNS lymphoma has increased significantly in the past 30 years, primarily in the elderly and immunocompromised. While T-cell lymphomas comprise 15–20% of systemic lymphomas, they comprise less than 4% of primary CNS lymphomas, suggesting that they may be under-recognized compared to their systemic counterparts. To investigate this, we studied brain biopsies from three patients who were diagnosed with T-cell lymphoma confined to the brain. They had enhancing lesions by MRI, arising in the cerebellum and brainstem in one and temporal lobe in two. We compared these to biopsies from three patients who had reactive lymphoid infiltrates and who had clinical signs/symptoms and radiographic findings that were indistinguishable from the lymphoma group. Biopsies from both the lymphoma group and reactive group showed considerable cytomorphologic heterogeneity. Although one lymphoma case contained large atypical cells, the other two contained small, mature lymphocytes within a heterogeneous infiltrate of neoplastic and reactive inflammatory cells. Surface marker aberrancies were present in two lymphoma cases, but this alone could not reliably diagnose T-cell lymphoma. The proliferation index was not useful for differentiating lymphoma from reactive infiltrates. In five of the six cases the diagnosis was most influenced by clonality studies for T-cell receptor-gamma gene rearrangements. We conclude that because of the high degree of overlap in cytomorphologic and immunophenotypic features between T-cell lymphoma and reactive infiltrates, T-cell lymphoma may not be recognized unless studies for T-cell receptor gene rearrangements are performed for CNS lesions composed of a polymorphous but predominantly T-cell infiltrate.

Keywords

T-cell lymphomaPrimary CNS lymphomaT-cell receptor rearrangementMonoclonal T-cellReactive lymphoid infiltrate

Introduction

The incidence of primary central nervous system (CNS) lymphomas increased by more than tenfold between 1973 and 1992 [22, 23, 33]. This coincided with an increased number of immunocompromised individuals as HIV infection became a major public health concern [4, 7, 34, 35]. The incidence of immunocompetent individuals afflicted with CNS lymphomas has also increased, especially in the elderly who have become one of the fastest growing segments of the population due to increased life expectancy [7, 10, 22, 26, 33, 34]. In the last 10–15 years, however, the rate of increase in CNS lymphoma of the elderly has slowed significantly [23].

Currently, lymphomas account for 3–6% of all primary brain neoplasms [6, 22, 35]. Of these, the vast majority are B-cell lymphomas, and only approximately 2–3% of CNS lymphomas are thought to be derived from T-cells [6, 19, 24, 35, 41, 42]. This contrasts significantly with the fact that T-cell lymphomas comprise 15–20% of all systemic non-Hodgkin lymphomas [24]. T-cell lymphomas are reportedly more common in eastern countries, and the proportion of T-cell lymphomas of the CNS (T-CNSL) is as high as 16.7% in Korea and 8–14% in Japan [6, 8, 19, 26, 34, 35, 41, 42]. The reasons for these discrepancies remain largely unknown. Interestingly, a majority of T-CNSL arise in immunocompetent individuals [35].

Establishing a diagnosis of T-CNSL can be a particularly daunting task [6, 34]. Radiographic features of T-CNSL are difficult to separate from reactive processes or even other types of neoplasms [6,26]. Often brain biopsies contain very limited amounts of tissue, making histopathologic evaluation difficult. Because lymphomas encompass a diverse group of disorders, establishing specific criteria to help the pathologist make a diagnosis of lymphoma is challenging. Because of this, T-CNSL may currently be under-diagnosed on brain biopsies and may be mistaken for non-neoplastic lymphoid infiltrates [7].

The purpose of this study was to review the literature for diagnostic criteria for T-CNSL and to evaluate their usefulness in making a diagnosis of lymphoma versus a reactive process. We also compared the clinical, radiographic and pathologic characteristics of T-cell lymphoma cases with reactive lymphoid infiltrates in the brain.

Materials and methods

We searched the archives of the Stanford Medical Center Laboratory of Neuropathology for brain biopsies in which the diagnosis of T-cell lymphoma had been made between the years 1996 and 2006. Three patients with this diagnosis were identified (cases 1, 2 and 3). We then identified three patients who had reactive lymphoid infiltrates in brain biopsies and in whom no diagnosis of lymphoma was ever made (cases 4, 5 and 6).

Brain biopsy specimens from all six patients had been fixed in 10% buffered formalin and submitted for routine processing and paraffin embedding. Hematoxylin and eosin (H&E) as well as reticulin stains were performed by routine methods.

For cases 3 and 6, the following immunohistochemical stains with their respective antibody dilutions were performed: CD3 (1:200, Cell Marque, Hot Springs, AR, USA), CD4 (1:20, Vision BioSystems’ Novocastra, Norwell, MA, USA), CD5 (1:50, Vision BioSystems’ Novocastra), CD7 (1:50, Vision BioSystems’ Novocastra), CD8 (1:200, Dako, Carpinteria, CA, USA), CD20 (1:1000, Dako), CD30 (1:40, Dako), ALK-1 (1:100, Dako) and Ki-67 (1:200, Dako). Immunohistochemical stains were performed on a Dako Autostainer using standard protocols.

For cases 1, 2, 4 and 5, the following immunohistochemical stains with their respective dilutions were performed: CD3 (1:50, Vector Laboratories, Burlingame, CA, USA), CD4 (1:5, Vector Laboratories), CD5 (1:200, Vector Laboratories), CD7 (1:25, Vector Laboratories), CD8 (1:200, Dako), CD20 (1:1200, Dako) and Ki-67 (1:100, Dako). These immunohistochemical stains were performed on a Ventana Benchmark XT platform using standard protocols. Additionally, these cases were stained with CD30 (1:40, Dako) and ALK-1 (1:100, Dako) which were performed on a Dako Autostainer.

In situ hybridization (ISH) for Epstein-Barr virus (EBV) was also performed on all six cases using a cocktail of oligonucleotide probes labeled with alkaline phosphatase (Ventana Medical systems, Tucson, AZ, USA).

Molecular assays for B-cell and T-cell clonality utilized commercially available kits to detect clonal (IgH) and T-cell receptor-gamma gene (TCR-γ) rearrangements (InVivoScribe Technologies, San Diego, CA, USA). Both kits employed primers and conditions validated in the BIOMED-2 multicenter European study [3, 29, 48]. An additional PCR tube was included in each kit as an amplification positive control. Genomic DNA was isolated from paraffin-embedded tissue as previously described [46]. The multiplex PCR was performed as specified in the kit protocols. The PCR products were resolved on an Applied Biosystems 3100 capillary electrophoresis instrument and analyzed with Genescan software (Applied Biosystems).

Cases exhibiting a Gaussian distribution of peaks within the specified molecular weight range for a given PCR primer pair were interpreted as a polyclonal background T-cell population and, thus, negative for clonal TCR-γ gene rearrangements [3, 29, 46, 48]. A clone was defined as one or two peaks of at least twice the height of the background, if present, of polyclonal peaks. The presence of three or more distinct peaks of similar magnitude was determined to represent oligoclonality. Cell line DNA for positive, negative (polyclonal) and sensitivity (2% clonal DNA in a background of 98% polyclonal DNA) controls were purchased from InVivoScribe Technologies. PCR (reagent) grade water was used as a blank control (Teknova).

Case histories

Cases 1–3: T-cell lymphomas

Case 1

A 56-year-old male presented with facial and hand numbness of 13 months’ duration. His symptoms progressed rapidly to increased somnolence, severe headaches, difficulty walking, nausea and decreased cognition. An MRI of the head showed enlarging enhancing lesions in bilateral cerebellar hemispheres, smaller areas of enhancement in brainstem and a small enhancing lesion in the wall of left lateral ventricle. The radiographic differential diagnosis included lymphoma, metastasis, multicentric glioma and an inflammatory process. The pathologic diagnosis was, “atypical lymphohistiocytic proliferation consistent with T-cell lymphoma.” The patient had an excellent initial response to chemotherapy, but he was lost to follow-up and died 6 months following his biopsy, the exact cause of which is unknown.

Case 2

A 42-year-old female presented with seizures beginning 9 months prior to diagnosis. She also developed left superior quadrantanopsia. An MRI of the head showed enhancement of cortical surfaces of the right temporal lobe. Subsequently, a biopsy was performed which was suggestive of cerebral infarction. Four months later, a new small, enhancing lesion was noted in the posterior right temporal lobe. The radiologic differential diagnosis included infarction, neoplasm, encephalitis and infection. The pathologic diagnosis was, “findings most consistent with T-cell lymphoma.” After receiving radiation and chemotherapy, the patient showed no residual/recurrent disease 26 months after diagnosis.

Case 3

A 54-year-old immunocompetent male presented with seizure activity. He also had progressively worsening bilateral hand tremors, right facial twitch and right-sided weakness. He was treated for epilepsy due to negative neuroimaging. His CSF was negative for atypical or malignant cells. After 5 months, a repeat MRI showed volume loss with mild enhancement in the right superior temporal gyrus (Fig. 1a). The radiologic differential diagnosis included an inflammatory process, a post-infectious process and a demyelinating process. The pathologic diagnosis was, “large cell lymphoma, of probable T-cell lineage.” He was treated with high dose methotrexate, procarbazine and vincristine, however, his neurologic status worsened after two cycles of chemotherapy. The patient then received one course of whole brain radiation which he tolerated well. However, he died of a suspected pulmonary embolism 3 months after diagnosis, but no autopsy was performed.
https://static-content.springer.com/image/art%3A10.1007%2Fs00401-007-0338-y/MediaObjects/401_2007_338_Fig1_HTML.jpg
Fig. 1

Axial and coronal MRI images with contrast demonstrate enhancing lesions in a the right superior temporal gyrus in case 3 of the T-CNSL group and b multiple areas including the left thalamus, internal capsule, periventricular regions and left temporal lobe in case 6 of the reactive group

Cases 4–6: reactive infiltrates

Case 4

A 37-year-old male presented with nonspecific motor abnormalities. An MRI of the head showed enhancing lesions in the cerebellum and brainstem with mass effect. The primary radiologic diagnosis was lymphoma. The pathologic diagnosis was, “atypical lymphoid infiltrate, favor reactive.” The patient then transferred care, but an enlarged cervical lymph node was excised several months later, which showed noncaseating granulomata. This was thought to be either tuberculosis or, more likely, sarcoidosis. No additional information was obtainable regarding the lymph node biopsy or the follow-up of the patient.

Case 5

A 56-year-old female presented with dizziness, ataxia, dysarthria and double vision worsening over several months and rapid onset of dementia evolving over 10 weeks. An MRI of the head showed bilateral deep white matter confluent areas of increased signal intensity with enhancement at the interface of the gray and white matter. The radiologic differential diagnosis included lymphoma, demyelinating process, chronic microvascular ischemia and gliomatosis cerebri. Due to worsening of her symptoms, the patient was given intravenous steroids prior to obtaining a biopsy. The pathologic diagnosis was “chronic inflammatory infiltrate, favor reactive.” Subsequent to her biopsy, the patient’s symptoms and radiographic findings improved with steroids. However, she then had a relapse of symptoms 7 months later, and an MRI was interpreted as progression of demyelinating lesions. One year after her biopsy, the patient is still receiving steroids. She has recovered intellectual function, but has diplopia and is ataxic, requiring a walker. The clinical suspicion for multiple sclerosis is low at this point.

Case 6

A 61-year-old male presented with mental status changes. An MRI of the head showed multiple enhancing lesions in the left thalamus, internal capsule, left periventricular region, left temporal lobe, left frontal lobe and corpus callosum (Fig. 1b). The radiologic differential diagnosis included lymphoma, sarcoidosis and gliomatosis cerebri. His symptoms and radiographic findings waxed and waned for almost 4 years before a biopsy was performed. The pathologic diagnosis was, “polyclonal perivascular and parenchymal small T-cell infiltrate.” He episodically received steroid treatments, and 20 months after biopsy, the patient’s symptoms and radiographic findings completely resolved. He was thereafter lost to follow-up.

Results

All six patients evaluated in this study were immunocompetent individuals. The mean age of the patients in the lymphoma group was 51 years. Likewise, the mean age of patients in the reactive group was 51 years. Both groups included two males and one female. Both groups also included two patients with only supratentorial disease and one with infratentorial disease.

All six brain biopsies demonstrated diffuse infiltration into the brain parenchyma by the lymphocytic infiltrates. In addition to diffuse infiltration, concentric perivascular aggregates of lymphocytes were also seen in cases 2–6. However, this was not a prominent feature in case 1 (Fig. 2a–f).
https://static-content.springer.com/image/art%3A10.1007%2Fs00401-007-0338-y/MediaObjects/401_2007_338_Fig2_HTML.jpg
Fig. 2

All six cases demonstrate varying degrees of diffuse lymphocytic infiltration. a No perivascular infiltrates are observed in case 1. b, c However, this is a prominent feature of the other two cases from the lymphoma group, cases 2 and 3. df Cases 4–6 comprising the reactive group also demonstrate prominent perivascular lymphocytic infiltrates (H&E, original magnifications ×100)

Cases 1, 2, 4, 5 and 6 all contained infiltrates composed of small mature-appearing lymphocytes with round to slightly angulated nuclei. Occasional larger cells were identified admixed with the smaller lymphocytes (Fig. 3a–f). In case 1, many of the larger cells had abundant foamy cytoplasm and were identified as macrophages. An immunohistochemical stain for CD163 confirmed this.
https://static-content.springer.com/image/art%3A10.1007%2Fs00401-007-0338-y/MediaObjects/401_2007_338_Fig3_HTML.jpg
Fig. 3

a The infiltrates in case 1 are predominantly composed of small lymphocytes with occasional larger cells intermixed. Many of the larger cells represent macrophages. b A majority of the lymphoma cells in case 2 are also relatively small. c Case 3 demonstrates large lymphoma cells with coarse chromatin and increased mitotic activity. df Cases 4–6 show reactive lymphocytic infiltrates composed predominantly of small cells with occasional larger cells intermixed. The morphologic features are virtually indistinguishable from cases 1 and 2 of the lymphoma group. (H&E, original magnifications ×400)

Case 3 was composed primarily of large, atypical cells with irregular nuclear contours and scant cytoplasm (Fig. 3c). Some cells were multinucleated. Significant mitotic activity, frequent apoptotic bodies and patchy necrosis were also present. These features of malignancy were not seen in any of the other five cases.

The immunohistochemical phenotypes of the six cases are presented in Table 1. Cases 1 and 2 had numerous B-cells admixed with the lymphocytic infiltrate (Fig. 4a, b). Case 5 had approximately equal numbers of CD4 and CD8 positive T-cells, whereas case 6 had more CD8 than CD4 positive T-cells (Fig. 4c, d). A reticulin stain highlighted perivascular reticulin deposition in all cases, including the ones with only reactive infiltrates (Fig. 4e). Ki-67 staining of lymphoid cells was felt to be quite variable from case to case as well as within each biopsy. The proliferative fraction was notably high in case 4 (Fig. 4f) compared to the other cases with only reactive infiltrates. In situ hybridization (ISH) for EBV was positive in only case 3.
Table 1

Immunohistochemical/molecular results

Case

Diagnosis

CD3

CD5

CD7

CD4

CD8

CD20

CD30

ALK-1

Ki-67

EBV-ISHa

T-Cell Clonality

1

Lymphoma

+

+

Weak

− in neoplastic cells,

+ in few reactive cells

+

+ in many reactive

B-cells

Overall: 30%

Range: 10–50%

+

2

Lymphoma

+

+

+

− in neoplastic cells,

+ in few reactive cells

+

+ in many reactive

B-cells

Overall: 30%

Range: 5–50%

+

3

Lymphoma

+

+

+ in Neoplastic cells

Overall: 5%

Range: 0–20%

+

No result due to specimen degradation

4

Reactive

+

+

+

+ in majority of lymphs

+ in few lymphs

+ in many reactive

B-cells

Overall: 5%

Range: 1–30%

Polyclonal

5

Reactive

+

+

+

+ in half of lymphs

+ in half of lymphs

+ in few reactive

B-cells

Overall: 5%

Range: 1–15%

Polyclonal

6

Reactive

+

+

+

+ in few lymphs

+ in majority of lymphs

+ in few reactive

B-cells

Overall: 1%

Range: 0–1%

Oligoclonal

aEpstein-Barr virus in situ hybridization

https://static-content.springer.com/image/art%3A10.1007%2Fs00401-007-0338-y/MediaObjects/401_2007_338_Fig4_HTML.jpg
Fig. 4

a, b Numerous CD20-positive B-cells are intermingled with the neoplastic T-cells in cases 1 and 2. c In case 6, relatively few CD4 positive reactive T-cells are seen. d In contrast, a higher proportion of the T-cells are CD8 positive in case 6. e Prominent perivascular reticulin deposition is seen in a region of reactive lymphocytic infiltrates in case 5. f Case 4 shows increased Ki-67 positivity within reactive lymphocytic infiltrates. (H&E, original magnifications ×200)

The results of T-cell receptor-gamma gene rearrangement studies by PCR are also shown in Table 1. Clonal peaks were identified in the TCR-γ variable region Vγ9 in case 1 and Vγ1-8 in case 2 (Fig. 5a, b). The presence of a scant polyclonal T-cell background was observed in case 5 in the context of adequate control gene amplification (Fig. 5c). An oligoclonal pattern of peaks is present in case 6, consistent with a reactive process (Fig. 5d). No PCR result was obtained from case 3 due to excessive specimen degradation. See supplemental material for complete electropherograms from patients 1, 2, 5 and 6.
https://static-content.springer.com/image/art%3A10.1007%2Fs00401-007-0338-y/MediaObjects/401_2007_338_Fig5_HTML.gif
Fig. 5

a, b Cases exhibiting clonal TCR-γ gene rearrangements. Two strong peaks are seen in the Vγ9 region in case 1. In case 2, two peaks are present in the Vγ1–8 region. c A sparse, polyclonal (reactive) T-cell background is seen in case 5. d An oligoclonal pattern of peaks of roughly equal magnitude is seen in case 6

Cases 1, 2 and 3 were ultimately diagnosed as T-cell lymphoma. Cases 4, 5 and 6 had reactive lymphoid infiltrates, and no diagnosis of lymphoma was otherwise made in these patients. None of the six cases showed features of B-cell lymphoma either by morphology, immunohistochemistry or molecular studies, and no additional brain biopsies have been performed to date.

Discussion

The diagnosis of T-CNSL is often difficult, and radiologic examination does not lend itself well to establishing a diagnosis. MRI, in particular, has been shown to markedly underestimate the extent of disease as microscopic infiltration into the brain parenchyma can have completely normal radiographic findings [28]. Multiple lesions are most often noted in AIDS patients, while immunocompetent patients more often present with only a single lesion [4, 23, 26]. Cystic degeneration with necrosis and hemorrhage are also not typical of T-CNSL in immunocompetent patients, being seen more often in AIDS patients [4, 26]. Reports from western countries have shown an infratentorial predilection, while eastern countries have shown a strong supratentorial predilection [19, 26, 35, 42]. Two of the cases of T-CNSL we studied went against western convention and had only supratentorial lesions.

Kim et al. studied a group of immunocompetent patients who had primary peripheral T-cell lymphoma of the brain, all of whom had supratentorial lesions. Although all patients had contrast-enhancing lesions, ring-enhancement was present in five of the seven patients, and only two had severe peritumoral edema. High-signal intensity on T-1 weighted images, suggesting hemorrhage, was also present in three patients [26]. Compared to B-cell lymphomas in which there is usually homogeneous enhancement, T-CNSL often has heterogeneously enhancing lesions, which further complicates the differential diagnosis [27]. Heterogeneous enhancement with necrosis and hemorrhage on MRI may lead to the radiologic diagnosis of a high-grade glioma [26]. All six patients in our series, including those with reactive infiltrates, had enhancing lesions by MRI, however, none showed ring-enhancement. Lymphoma was specifically mentioned in the radiographic differential diagnosis of only one of the T-CNSL cases we studied. Other considerations by MRI in these T-CNSL cases included metastasis, glioma, demyelinating lesions, infection or other non-neoplastic inflammatory processes. Ironically, all three cases of reactive infiltrates listed lymphoma as a leading consideration in the MRI reports.

The vast majority of CNS lymphomas are diffuse large B-cell lymphomas [35]. In contrast, most T-cell lymphomas do not exhibit high-grade morphology and frequently do not show other atypical cytologic features. In the CNS, T-cell lymphomas often have small nuclei with bland morphology [7, 34, 35]. Only one of the T-CNSL cases we studied showed atypical morphologic features (case 3) not present in either of the other two T-CNSL cases in our series. Two of the lymphoma cases also had an even higher proportion of B-cells than in cases 5 and 6 which were reactive in nature. This mixed infiltrate, which is common in T-cell lymphoproliferative processes, can obscure the lymphoma amongst a heterogeneous population of cells. Slightly larger cells could also be identified in the three reactive cases as well as the lymphoma cases, and as in case 1, may represent macrophages. This macrophage infiltrate may lead to an erroneous diagnosis of a reactive process, especially since macrophages can be abundant after treatment with steroids.

Although there is no specific immunohistochemical marker for neoplastic T-cells, surface marker expression can be altered when T-cells undergo neoplastic transformation. Unfortunately, this phenomenon is not limited to neoplastic cells and can also occur in reactive T-cell infiltrates. The most common alteration in antigen expression among neoplastic or reactive T-cells is loss of CD7 [39]. Decreased CD7 expression, compared to presumably reactive T-cells within the same biopsy, was noted in case 1, while case 3 showed complete loss of another T-cell marker, CD5. In case 3, the neoplastic T-cells also showed aberrant expression of CD20. This rare and unusual phenotype has been discussed by Mohrmann et al. and Warnke et al. among others, in extranodal sites including a cervical lymph node and the skin [38, 49]. Although CD20 has traditionally been thought of as a B-cell specific antigen, recent studies indicate that up to 12% of normal peripheral blood T-cells may express CD20, and these cells are predominantly CD8-positive cytotoxic T-cells [38, 52]. These cells, as well as CD20 positive T-cell lymphoma cells, lack expression of other B-cell markers such as CD79a [38, 45, 51].

A commonplace initial approach to brain biopsies with atypical lymphoid infiltrates involves immunohistochemical stains for CD3 and CD20. Caution must be taken when using a limited screening panel that includes CD5 (instead of CD3) and CD20. This limited panel cannot differentiate between T-cell lymphomas and CD5+ B-cell lymphomas [45]. Had this been done as the initial workup in case 3, one might have mistaken the phenotype of CD5−/CD20+ cells as a large B-cell lymphoma. However, these cells showed strong expression of other T-cell markers, including CD3, which confirmed the true lineage.

The expression of the T-cell markers CD4 and CD8 is of considerable interest in the attempt to differentiate various T-lymphocyte proliferative disorders, both reactive and neoplastic. Studies of acute experimental autoimmune encephalitis in rats, non-neoplastic T-cells in B-cell lymphomas of the CNS, and various reactive conditions outside the CNS, such as in inflammatory skin disease and connective tissue disease, have shown that the majority of reactive T-cells are of the CD4 helper phenotype [5, 21, 31]. It has been suggested that a shift in the CD4:CD8 ratio may provide support for a clonal process, especially when the vast majority of cells have a CD8 positive cytotoxic T-cell immunophenotype. However, this is by no means diagnostic, and exclusive positivity for either CD4 or CD8 is only sometimes associated with monoclonality [51]. It should be noted that the proportion of CD4 positive helper T-cells to CD8 positive cytotoxic T-cells may fluctuate over the course of a pathologic process, further confounding their use in establishing neoplasia. There are also many conditions in which CD8 positive T-cells become activated and predominate, including polymyositis, inclusion body myositis and various viral infections [11, 13]. Some studies have shown a predominance of CD8 positive T-cells in some types of viral encephalomyelitis and in demyelinating brain lesions associated with multiple sclerosis [37]. In addition, natural killer (NK) cells share some functions and markers with cytotoxic T-cells and are usually CD8 positive. Normal γδ T-cells are typically negative for both CD4 and CD8 [16, 24]. Thus, the presence of NK or γδ T-cells can alter the “normal” CD4:CD8 ratio in reactive infiltrates. Reactive in nature, case 5 had a ratio of approximately 1:1, and case 6 had more CD8 positive T-cells than CD4 positive T-cells and a reversal of this ratio.

Akin to many reactive processes, a majority of T-cell lymphomas are CD4 positive [7, 34, 39, 51]. It is noteworthy, however, that two of the T-CNSL cases we studied were positive for CD8 and not CD4, reminiscent of a few T-cell lymphomas/leukemias, which typically occur outside the CNS that are usually CD8 positive, such as subcutaneous panniculitis-like T-cell lymphoma, large granular lymphocyte leukemia and angioimmunoblastic T-cell lymphoma [24, 39]. However, CD4 positive non-neoplastic T-cells were also present in these biopsies. Choi et al. reported similar findings. Five of the seven T-CNSL cases they studied were positive for CD8, one of which had coexpression of CD4. The remaining two cases were negative for both CD4 and CD8 [7]. Thus, previous reports and our experience indicate that immunophenotyping for CD4 and CD8 is of limited diagnostic utility in distinguishing T-CNSL from reactive infiltrates.

Concentric perivascular infiltrates are a prominent feature of CNS lymphoma and were conspicuous in two of our T-CNSL cases [4, 27, 35]. Of note, all three cases in the reactive group also showed prominent perivascular lymphoid aggregates identical to those seen in the T-CNSL cases, including the concentric perivascular deposition of reticulin fibers associated with CNS lymphoma [4, 35]. We, therefore, find this feature unreliable in distinguishing neoplastic from reactive perivascular lymphocytic infiltrates.

The proliferative fraction of lymphoid cells in a normal lymph node and extranodal sites shows wide variability. Doisne et al. found that activated HIV-specific CD8 positive T-cells showed a high proliferative rate of 51%, suggesting that activation of these cells drives them to proliferate [13]. Liu et al. reported only a few Ki-67 positive cells in their case report of T-CNSL [34]. Although two lymphoma cases in our series had mean proliferation rates significantly higher than the reactive cases, there was wide variability in the proliferation indices within each biopsy. One case from the reactive group also showed a significantly high proliferation rate, emphasizing the fact that high proliferation rates in non-neoplastic T-cells may reflect activation. Sen et al. reported that Ki-67 staining might have overestimated the number of neoplastic cells in subcutaneous panniculitis-like T-cell lymphoma due to the many reactive lymphocytes that showed immunopositivity [40]. In another study of peripheral T-cell lymphoma, “high” Ki-67 staining was present in only 11% of the cases [51]. Therefore, measuring the proliferative fraction as a tool for differentiating reactive from neoplastic lymphoid infiltrates should be done with caution.

Although EBV genetic material is more commonly present in B-cell lymphomas, the most common exceptions among T-cell lymphoproliferative disorders include aggressive NK-cell leukemia, nasal-type NK/T-cell lymphoma, and in a minority of cases, enteropathy-type T-cell lymphoma [24, 34]. The presence of EBV in one of the T-CNSL cases we report is of uncertain significance. However, this may have erroneously supported a B-cell origin of the neoplastic cells if multiple antibodies to T-cell markers had not been evaluated.

Since low-grade T-cell lymphomas can be indistinguishable from reactive conditions (especially in the skin), even with the aid of immunohistochemistry, it is important to evaluate these lesions using molecular assays for T-cell receptor (TCR) gene rearrangements [6, 12, 20]. Most laboratories have resorted to testing only the TCR-γ gene since the variable region is less complex than that of the TCR-β variable region. Additionally, TCR-γ is also rearranged earlier in T-cell development, and some have suggested that TCR-γ gene rearrangements are present in almost all T-cell neoplasms [2, 12, 47]. Of note, neoplasms derived from natural killer cells do not undergo T-cell receptor gene rearrangements [39]. T-cell monoclonality is not diagnostic of lymphoma however [12, 20, 39, 47]. Monoclonal populations of T-cells have been described in peripheral blood specimens of elderly individuals who carry no diagnosis of lymphoma [12]. Clonal populations of T-cells have also been recognized in several inflammatory skin disorders including lichen planus, pityriasis lichenoides and lichen sclerosis et atrophicus. [1, 20]. However, Griesser et al. postulated that some of these clonal populations may represent an early stage of lymphoma development [20]. Oligoclonal peaks in T-cell populations have been seen in inclusion body myositis, normal endometrium and in brains of children with AIDS [11, 14, 32]. Therefore, the results of PCR for TCR gene rearrangements should be analyzed within the proper clinical and pathologic context.

Although there is no distinct marker for T-cell lymphomas, flow cytometry is an effective way of analyzing T-cells for expression of multiple surface markers. An aberrant phenotype on a T-cell population may provide some support for clonality, but many T-cell lymphomas do not demonstrate immunophenotypic abnormalities, as was the situation in case 2. Surface marker aberrancies may also be seen in reactive conditions and are not diagnostic of lymphoma [16, 39]. Recently, a flow cytometry antibody panel has become commercially available which covers approximately 70% of the normal human TCR-Vβ repertoire. This can be useful in identifying T-cell clones but is not without limitations. This panel will only identify T-cells expressing the αβ receptor, and the selected target T-cells must express CD3 since the intensity of this surface marker parallels the expression of the T-cell receptor [30]. Therefore, in situations where tissue is limited, one must carefully consider whether flow cytometry or studies performed on paraffin-embedded tissue will yield more diagnostic information.

Three types of T-CNSL have been reported in the literature more often than others and include adult T-cell lymphoma/leukemia (ATLL), anaplastic T-cell lymphoma (ATCL) and peripheral T-cell lymphoma [9, 15, 17, 36, 42, 43]. The prevalence of ATLL corresponds with that of human T-cell leukemia virus type 1 (HTLV-1), and the lymphoma cells are large with cerebriform nuclei [15, 16, 24, 36]. Due to the rarity of this disease in the United States and lack of characteristic findings, none of the patients we studied were evaluated for HTLV-1. In addition to anaplastic morphology, the cells in ATCL are typically positive for CD30 and ALK-1 by immunohistochemistry, and molecular analysis shows t(2;5) [9, 17, 24]. Only case 3 exhibited sufficient atypia to raise concern for ATCL, however, ALK-1 staining was negative and the morphology was not entirely characteristic. Other even more exceptional types of T-CNSL include post-transplant T-CNSL such as that reported by Su et al. occurring in a patient 4 years after a kidney transplant and an extranodal natural killer cell primary CNS lymphoma reported by Kaluza et al. [23, 25, 44]. Primary peripheral T-cell lymphoma of the CNS has been reported in the literature by several authors [43]. Rather than one distinct entity, this group represents extranodal T-cell lymphomas, which do not fit well into any other category as defined by the World Health Organization [24]. Two of the lymphoma cases we studied fall into this category.

A diagnostic pitfall in the evaluation of brain biopsies for lymphoma occurs when the patient has received steroids prior to the biopsy. Steroids induce apoptosis of lymphoid cells and can deplete the brain tissue of neoplastic infiltrates within 24–48 h, leaving behind a milieu of gliosis and relatively few reactive lymphocytes and/or macrophages [4, 8, 18, 33]. With time, however, the lymphoma becomes resistant to steroid treatment, most likely through selective growth of clones exhibiting steroid receptor deficiency or through a post-receptor defect in apoptotic pathways [18, 50]. Geppert et al. reported a relatively short time interval of a few days after treatment with steroids before lymphoma cells repopulated the brain parenchyma [18]. Ideally, biopsies should be obtained prior to initiating therapy with steroids. However, Choi et al. demonstrated that PCR for B-cell rearrangements still showed monoclonal bands even when patients received steroids prior to obtaining brain biopsies. They proposed that nuclear fragments may still retain genetic alterations, and it is possible that this may hold true for T-cell lymphomas as well [8]. Only case 5 had steroids prior to the biopsy, and her clinical course, thus far, has shown no evidence of lymphoma.

Because of the rarity of the diagnosis, we conclude that no typical or common pattern for T-CNSL emerges, either clinically, radiologically or pathologically. TCR gene rearrangement studies were necessary to establish the diagnoses in the cases we studied, the clinical outcomes of which may have validated the diagnoses. Because the morphologic features of T-CNSL are heterogeneous and share features with reactive processes, T-CNSL may be an under-recognized entity, and pathologists should have a low threshold for performing T-cell receptor gene rearrangement studies on brain biopsies containing mixed lymphoid infiltrates.

Supplementary material

401_2007_338_MOESM1_ESM.tif (558 kb)
(TIF 557 kb)
401_2007_338_MOESM2_ESM.tif (560 kb)
(TIF 560 kb)
401_2007_338_MOESM3_ESM.tif (549 kb)
(TIF 548 kb)
401_2007_338_MOESM4_ESM.tif (538 kb)
(TIF 537 kb)

Copyright information

© Springer-Verlag 2008