OUP user menu

B lymphocytes in the normal brain: contrasts with HIV‐associated lymphoid infiltrates and lymphomas

I. C. Anthony, D. H. Crawford, J. E. Bell
DOI: http://dx.doi.org/10.1093/brain/awg118 1058-1067 First published online: 1 May 2003


In recent years evidence has accumulated which suggests that the brain may not be the immunologically privileged site it was once considered to be. It is now widely accepted that T lymphocytes perform surveillance functions in normal brain parenchyma. However, as yet there are no reports of B lymphocytes entering brain parenchyma in the healthy state. This study aimed to determine first the prevalence of B lymphocytes in normal brain, and subsequently whether advancing HIV infection led to changes in the brain B lymphocyte population, which might contribute to the increased risk of lymphoma seen in AIDS. Our results show that B lymphocytes do enter all parts of the normal human brain in very low numbers and that the B lymphocytes within the brain parenchyma display an activated (CD23 positive) phenotype. In contrast, intravascular B lymphocytes have a much lower expression of activation markers. B lymphocytes were found in increased numbers in both the brain parenchyma and perivascular spaces of pre‐AIDS brains. However, brains from the majority of AIDS subjects, including those with primary CNS lymphoma (PCNSL) (outside the area of neoplastic involvement) contained fewer B lymphocytes than normal or pre‐symptomatic HIV‐infected brains. A subset of AIDS brains, previously shown to have pleomorphic lymphoid infiltrates in the perivascular spaces, had significantly increased numbers of B lymphocytes in both the brain parenchyma and perivascular spaces. Virtually all AIDS‐related PCNSL are known to be Epstein–Barr Virus (EBV) positive, in contrast to non‐HIV PCNSL and non‐CNS AIDS‐related lymphomas. We examined the EBV status of brain parenchymal B lymphocytes to investigate whether EBV‐positive B lymphocytes are more frequent in HIV‐infected brains than normal, thus explaining the propensity for CNS lymphomas in AIDS. In situ hybridization studies showed EBV positivity only in AIDS‐related PCNSL cases within the lymphoma deposits. PCR‐based studies detected high EBV copy numbers in PCNSL tumour tissue, and low copy numbers in AIDS cases with pleomorphic lymphoid infiltrates. As none of the B lymphocytes in this latter group were EBV positive on in situ hybridization, bearing in mind that this appears to be a prerequisite for PCNSL development, we find no evidence that pleomorphic infiltrates represent a pre‐malignant PCNSL state.

  • Keywords: B lymphocytes; brain; EBV; HIV; primary CNS lymphoma
  • Abbreviations: BBB = blood brain barrier; EBERs = EBV small mRNAs; EBV = Epstein–Barr virus; PCNSL = primary CNS lymphoma


Until recent years the brain has been considered an immunologically privileged site. This concept was based on early studies which showed that allografts survived better in the CNS than in peripheral sites (Woodruff, 1960). The view was prevalent that the CNS lacked both lymphatic drainage (Gell and Coombs, 1968) and antigen‐presenting lymphocytes. In addition, the blood brain barrier (BBB) was believed to prevent the entry of immune effector cells. However, evidence has accumulated in recent years to suggest that the CNS is not excluded from regular surveillance by the immune system. Not only does it contain cells such as astrocytes, microglia, endothelial cells and pericytes, which are capable of antigen presentation (Male et al., 1987; Fabry et al., 1994), but CD4 T lymphocytes also enter the CNS in an apparently random manner (Hickey et al., 1991). Only T lymphocytes that are activated are able to enter the normal CNS, and those that fail to encounter antigen leave within 1–2 days of entry. Cells capable of reacting with brain parenchymal antigens remain in the CNS, or cyclically re‐enter, and thereby initiate inflammation (Hickey et al., 1991).

With regard to the humoral arm of the immune response, there are as yet no reports of B lymphocytes entering the normal CNS. Several in vitro studies, using models of the BBB, have shown that while B lymphocytes may adhere to cerebral endothelium, it is CD4 T lymphocytes that migrate most effectively through the barrier (Pryce et al., 1994). However, this model replicates only some of the in vivo functions of the BBB. It is known that B lymphocytes are capable of responding to antigen within the CNS. For instance, Knopf and colleagues showed that in rats naïve to the antigen in question, infusion of antigen through an indwelling brain parenchymal catheter was followed by recruitment of B lymphocytes and production of antibody despite the presence of an intact BBB (Knopf et al., 1998). When the experiment was repeated in pre‐immunized rats, the B lymphocyte response was even stronger. This suggests that antigen can be detected behind the BBB and furthermore that antigen‐specific B lymphocytes can respond to antigen within the brain.

Infection with HIV is characterized by the early appearance of anti‐HIV antibodies in the serum. These antibodies generally persist for the lifetime of the individual, despite the steady decline in immune responses that occur in untreated individuals. HIV nucleic acid can be detected in brain tissue from many patients with HIV infection, but the level of the pro‐viral load is generally low unless florid productive HIV infection (HIV encephalitis) is present (Achim et al., 1994; Bell et al., 1998). The levels of proviral DNA detected in the CNS of pre‐symptomatic patients are very much lower than those found in HIV encephalitis (Bell et al., 1993). In view of the fact that antigen‐specific B lymphocytes are capable of responding to antigen in the CNS, and that anti‐HIV antibodies are produced for the duration of the infection, it seems logical to predict that B lymphocytes should enter the brain in the presence of HIV infection of the CNS. However, there have been no reported studies to date on CNS B lymphocyte trafficking in HIV infection. This is surprising given the extensive autopsy data, which suggests that before the introduction of HAART (highly active antiretroviral therapy) 5–10% of the AIDS population developed primary CNS lymphoma (PCNSL) (Auperin et al., 1994; Jellinger and Paulus, 1995). While the incidence of PCNSL appears to have declined significantly in HAART‐treated subjects, the increased life expectancy of such individuals may be threatened by a risk of malignancy, which remains higher than for the normal population (Kirk et al., 2001; Rabkin, 2001). AIDS‐associated PCNSLs are all of B lymphocyte origin, and these lymphoproliferations are almost universally Epstein–Barr virus (EBV)‐driven (MacMahon et al., 1992; Auperin et al., 1994; reviewed in Brooks et al., 1999). However, it is unclear under what circumstances EBV‐infected B lymphocytes enter the CNS.

The first aim of this study was to determine whether B lymphocytes enter the brain in normal healthy individuals when, presumably, no exogenous immunological challenge originates from CNS tissue. Our results show that B lymphocytes are present within normal brains in very low numbers, and that all the B lymphocytes within brain tissue display an activated phenotype.

The second aim was to investigate whether HIV‐associated lymphomas were heralded by any change in the B lymphocyte population within the CNS as compared with normal subjects. HIV infection may be uniquely interesting in this context because of the relative preservation of B lymphocyte responses in contrast to the progressive disablement of the T lymphocyte system. To this end, the B lymphocyte population was investigated in the brains of pre‐symptomatic HIV‐positive patients, in AIDS subjects with no significant CNS pathology, and in those with PCNSL. In the subjects with PCNSL, we concentrated on the parts of the brain not affected by PCNSL in an attempt to gauge the setting in which PCNSL might develop. We also investigated cases displaying pleomorphic lymphoid infiltrates within the brain since these are poorly understood at present, particularly with respect to their relationship to PCNSL. In addition, we sought to determine the EBV status of all the CNS lymphocyte populations examined in this study.


Five patient groups were used in the study. The cases were carefully selected to fit within the definition of the group as given, thus minimizing potential confounding variables. These cases were part of the Edinburgh HIV brain bank and use of this resource in research was approved by the Lothian Ethics of Research committee.

Normal brains

Group 1 (n = 7; five males; age range 16–31 years, mean age 24 years). All these cases died as a result of accidents and had no evidence of CNS disease at autopsy, nor history of illness relating to the CNS.

Pre‐symptomatic HIV‐infected patients

Group 2 (n = 6; six males; age range 28–40 years, mean age 32 years). These individuals were all HIV‐positive drug users infected by needle sharing and who died of drug‐related accidents before developing an AIDS‐defining illness. This was confirmed upon post mortem examination.

AIDS patients with no significant CNS pathology

Group 3 [n = 5; three patients were drug users (one female), the remaining two were homosexual males; age range 32–45 years, mean age 37 years]. All patients died of non‐CNS‐related AIDS conditions. These patients displayed no evidence of PCNSL, CNS opportunistic infections or HIV encephalitis. Two cases in this group had evidence of systemic lymphoma without CNS involvement.

AIDS patients with non‐neoplastic lymphoid infiltrates in the CNS

Group 4 (n = 5). Four patients were adults, of whom three were drug users (all females) and one was a homosexual male (adult age range 23–38 years, mean age 32 years). One patient was an African child (age 6 months). The child died of systemic cytomegalovirus and pneumocystis infection. Despite relatively high T lymphocyte counts in two individuals in this group, all had symptoms of AIDS. At routine neuropathological examination these subjects displayed pleomorphic perivascular lymphoid infiltrates in all or many areas of the brain. These infiltrates contained occasional plasmacytoid cells and were not considered morphologically malignant. Careful examination of these brains had revealed no evidence of other significant CNS pathology, apart from one case that had a single focus of ependymal cytomegalovirus infection.

AIDS PCNSL patients

Group 5 (n = 6; five homosexual men, one intravenous drug user; age range 28–36 years, mean age 31 years). Five individuals had focal deposits of neoplastic B lymphoid lymphocytes aggregated in the perivascular compartment and also invading the brain tissue locally, while one individual had widespread disseminated CNS B‐cell lymphoma. Full autopsy in these patients failed to reveal evidence of lymphoma outside the CNS. None had HIV encephalitis but three displayed very slight evidence of focal cytomegalovirus infection in addition to lymphoma.



In all the subjects described above, the brain was removed at autopsy and fixed intact in formalin for 2–3 weeks. Blocks were removed for histology from the frontal, parietal and occipital lobes, central white matter, temporal hippocampus, basal ganglia, thalamus, midbrain, pons, medulla and cerebellum. Most blocks were available from every case and cases were not included unless a wide selection of blocks was available. All samples were paraffin embedded and 5 µm sections were cut from each block. Sections were attached to superfrost slides (BDH, VWR International Ltd, Poole, Dorset, UK) and stored at 37°C for 48 h. Sections were de‐waxed in xylene and rinsed in alcohol in preparation for immunocytochemistry. Procedures for immunohistochemistry, including the antigen pre‐treatments and antibody concentrations, are listed in Table 1.

View this table:
Table 1

Antibodies and immunohistochemical techniques

Antibody/sourceTarget cell typePre‐treatmentBlocking agentsPrimary antibody dilutionSecondary antibodyTertiary stepVisualizing agent
CD20/Dako (monoclonal) for single or double immunohisto‐chemistryB lymphocytesMicrowave in citric acid3% H2O2 (10 min) 20% rabbit serum (10 min)1:300 (30 min)RAMBO (Dako) 1/200 (30 min)ABC (Dako) (30 min)DAB (Vector) for single immuno‐histochemistry or VIP (Vector) for double immuno‐histochemistry
CD20/Dako (monoclonal) before re‐staining procedure (see text)B lymphocytesMicrowave in citric acid3% H2O2 (10 min) 20% rabbit serum (10 min)1:150 (30 min)RAMBO (Dako) 1/200 (30 min)ABC‐alkaline phosphotase (Vector) (30 min)New Fucsin Red(Dako)
CD79a/Dako (monoclonal)B lymphocytesMicrowave in EDTA3% H2O2 (10 min) 20% rabbit serum (10 min)1:75 (30 min)RAMBO (Dako) 1/200 (30 min)ABC (Dako) (30 min)DAB (Vector)
CD23/Novocastra (monoclonal)Activated B lymphocytesMicrowave in citric acid3% H2O2 (10 min) 20% rabbit serum (10 min)1:30 (30 min)RAMBO (Dako) 1/200 (30 min)ABC (Dako) (30 min) DAB (Vector)
CD3/Dako (polyclonal)T lymphocytes (CD4 and CD8)Microwave in EDTA3% H2O2 (10 min) 20% swine serum (10 min)1:100 (30 min)SARBO (Dako) 1/200 (30 min)ABC (Dako) (30 min) DAB (Vector)
Von Willebrand Factor/Dako (polyclonal)Endothelial cellsMicrowave in citric acid3% H2O2 (10 min) 20% swine serum (10 min)1:200 (30 min)SARBO (Dako) 1/200 (30 min)ABC (Dako) (30 min) DAB (Vector)
Ki‐67/Dako (monoclonal)Proliferating cellsMicrowave in EDTA3% H2O2 (10 min) 20% rabbit serum (10 min)1:40 (30 min)RAMBO (Dako) 1/200 (30 min)ABC (Dako) (30 min) DAB (Vector)
Ki‐67/Dako (polyclonal)Proliferating cellsMicrowave in EDTA3% H2O2 (10 min) 20% swine serum (10 min)1:40(30 min)SARBO (Dako) 1/200 (30 min)ABC (Dako) (30 min) DAB (Vector)

All sections were counterstained with haematoxylin.

ABC = streptavidin biotinylated horseradish peroxidase; DAB = diaminobenzidine; EDTA = ethlyenediaminetetraacetic acid; RAMBO = rabbit anti‐mouse biotinylated; SARBO = swine anti‐mouse biotinylated.


The antibodies used to detect B lymphocytes were anti‐CD20 and anti‐CD79a. In order to verify that B lymphocytes were truly extravascular and not contained within small calibre capillaries, double immunocytochemistry was performed, applying first an antibody to Von Willebrand factor (an endothelial cell marker). Before applying the second antibody (anti‐CD20 antibody), sections were re‐immersed in 3% hydrogen peroxide (H2O2). Following incubation with blocking serum, free biotin sites on the first biotinylated secondary antibody were blocked using an avidin/biotin blocking kit (Vector). The full immunocytochemical procedure was then repeated for the B lymphocyte antibody, with substitution of the peroxidase substrate Vector VIP for diaminobenzidine as the final visualization agent, followed by haemtoxylin counterstaining.

In order to investigate the activation status of B lymphocytes, sections previously stained with anti‐CD20 antibody were de‐stained after assessment for the presence of B lymphocytes and photographic capture, and subsequently stained using an anti‐CD23 antibody. CD20 was detected using an alkaline phosphatase endpoint and sections were photographed. The cover slips were then removed by immersion of mounted slides in 50% ethanol and the sections placed in an antibody stripping solution (1 vol. 5% sulphuric acid, 1 vol. 2.5% potassium permanganate, 50 vol. distilled water) for 14 min with shaking. Sections were de‐stained in 0.5% sodium metabisulphate for 1 min with shaking, followed by three rinses in distilled water. The regular immunocytochemical procedure was then used to re‐stain the same sections with an antibody to CD23, which was visualized with diaminobenzidine. Sections were finally re‐photographed and the photographs compared. Control sections were included; these were stained for CD20‐positive cells, then immersed in the stripping solution. Subsequently, secondary anti‐mouse antibody, but no primary antibody, was applied. These controls demonstrated that all of the original staining and antibody had been removed from the section. This technique was found to work well in further characterizing individual cells within a single section since B lymphocyte presence proved to be a rare event within the brain tissue of most of the subjects in this study. This procedure was employed rather than double staining because both the primary antibodies (anti‐CD20 and anti‐CD23) were mouse monoclonal antibodies and double staining could have led to a cross reaction between the anti‐mouse secondary antibodies.

A proliferation marker, Ki‐67, was used to detect potentially dividing cells amongst the lymphocytic infiltrates. Subsequently, antibodies specific for CD3 and CD20 were used to identify Ki‐67‐positive cells as T or B lymphocytes.

Sections stained with each of these antibodies were assessed from the 11 named areas of the brain in each case. The numbers of parenchymal and perivascular B lymphocytes were quantified in each slide. Positive cells proved to be sufficiently rare that subjective counting was deemed more accurate than automated image analysis. In the PCNSL cases, areas of lymphoma were deliberately avoided so that assessment of the parenchymal and perivascular B lymphocyte presence was restricted to non‐lymphomatous areas. The number of cells per cm2 in each 5 µm section was calculated from the total cell count within a section of known area. Statistical analysis was made between groups for parenchymal and perivascular B lymphocyte counts using the Student’s t‐test.

In situ hybridization

In situ hybridization to detect EBV small mRNAs (EBERs) was performed on de‐waxed paraffin sections using a Dako (Ely, Cambridgeshire, UK) peptide nucleic acid EBER probe, with a Dako peptide nucleic acid detection kit, in accordance with the manufacturer’s instructions.

PCR analysis

In those cases for which frozen CNS tissue samples were available, PCR was undertaken to detect EBV presence. Semi‐quantitative PCR to detect the BamW region of the EBV genome was undertaken as described previously (Hopwood et al., 2002). DNA was extracted from frozen tissue using proteinase K (10 mg/ml) for digestion of tissue and phenol/chloroform for separation of nucleic acids, followed by ethanol precipitation. Control DNA was extracted from the EBV‐positive Burkitts lymphoma cell line Namalwa, and the EBV‐negative lymphoma cell line BJAB.


Examination of the brain in seven normal young subjects (Group 1) showed that B lymphocytes were present in the parenchyma in very small numbers, but consistently so in all cases (Fig. 1 and Table 2). Double staining with the anti‐Von Willebrand antibody confirmed that these occasional parenchymal B lymphocytes were not intravascular nor were they perivascular in distribution. B lymphocytes were found in the parenchyma of all areas of the normal brain examined and no evidence was found to suggest that they preferentially ‘home’ to specific areas of the brain (Table 3 shows the data from the cases in Group 1).

Fig. 1 A single CD20‐positive B lymphocyte (stained brown) in the parenchyma of a normal brain.

View this table:
Table 2

Quantitation of parenchymal and perivascular B lymphocytes in the CNS in five study groups

Age (years)/genderHIV statusCNS pathologyCD4/CD8 blood countsAverage number of CNS parenchymal B cells/cm2Average number of CNS perivascular B cells/cm2EBV viral load in brain
Group 1: normal subjects
Group 2: pre‐symptomatic HIV‐infected subjects
28MPositiveMild lymphocytic meningitis90/10100.5180.183NA
31MPositiveMild lymphocytic meningitis211/12670.1380.025NA
40MPositiveMild lymphocytic meningitisNA1.3840.629
31MPositiveMild lymphocytic meningitis425/13160.2530.864NA
30MPositiveMild lymphocytic meningitisNA0.1351.012
33MPositiveMildlymphocytic meningitis312/7010.0730NA
Group 3: AIDS patients with no significant CNS pathology
32MAIDSNil (systemic lymphoma)40/26000.023
36FAIDSNil (systemic lymphoma)23/740.0390NA
Group 4: AIDS patients with pleomorphic lymphoid infiltrates in the CNS
29FAIDSPleomorphic infiltrate248/11540.30810.474NA
38FAIDSPleomorphic infiltrate85/30204.109+
23FAIDSPleomorphic infiltrate246/5100.13854.027+
37MAIDSPleomorphic infiltrate (minimal CMV)22/6620.2104.770NA
0.5FAIDSPleomorphic InfiltrateNA0.2140.499NA
Group 5: AIDS patients with PCNSL
51MAIDSCNS lymphoma18/27600.061NA
28MAIDSCNS lymphoma and CMV4/38400NA
30MAIDS (drug user)CNS lymphoma and CMV4/u00NA
36MAIDSCNS lymphoma 1/330.0160NA
37MAIDSCNS lymphoma and CMV8/5700+++
34MAIDSDisseminated CNS lymphomaNA00.047+++

NA = not available; EBV = Epstein–Barr virus; CMV = cytomegalovirus; M = male; F = female.

Viral load scale: – = no EBV detected by PCR; + = 10–1000 copies of EBV genome detected/µg DNA; ++ = 1000–20 000 copies; +++ = 20 000–106 copies.

View this table:
Table 3

Data from Group 1 (normal) demonstrating the presence of parenchymal CD20‐ positive cells in all regions of the brain

Area of brainTotal number of CD20+ cells in all sections examinedNumber of sections with CD20+ cells present/total number of cases examined Mean area of sections examined [cm2 (range)]
Frontal22/64.87 (4.19–5.5)
Parietal44/74.15 (2.56–7.5)
Occipital21/34.87 (4.44–5.19)
Basal ganglia33/55.69 (5.19–6.13)
Temporal hippocampus22/63.97 (3.25–4.56)
Cerebellum55/55.08 (4.44–6.44)
Mid‐brain11/73.34 (2.5–4.13)
Pons41/34.35 (3.63–5.31)
Medulla11/31.81 (1.56–1.94)

CD20‐positive cells were detected only in the brain parenchyma, no cells were found in the perivascular space. In all areas of the brain it is apparent that very few sections contained more than one CD20+ parenchymal cell. *Only one section examined.

When sections in which B lymphocytes were identified were re‐stained with an antibody to the activation marker CD23, it was shown that intravascular B lymphocytes were a mix of activated (CD20+ CD23+) and non‐activated (CD20+ CD23–) lymphocytes, whereas those outside the vascular compartment were all of the activated phenotype (Fig. 2A and B). This latter finding proved consistent for all the study groups.

Fig. 2 (A) A single CD20‐positive B lymphocyte (stained red) in the parenchyma of a normal brain. (B) The same section as in A after the antibody and stain have been removed and the section has been re‐stained using an anti‐CD23 antibody (stained brown).

On routine staining, pre‐symptomatic HIV‐infected patients (Group 2) displayed mild focal lymphocytic infiltrates in the leptomeninges and around some white matter vessels. We have shown previously that these are largely made up of T lymphocytes (Bell et al., 1993). This study shows that in addition to CD3‐positive T lymphocytes, B lymphocytes appeared in the perivascular infiltrates (P < 0.02 when compared with controls) (Fig. 3) and were also found in increased numbers in the parenchymal compartment, although this trend was not statistically significant (Table 2). In contrast, once patients progressed to AIDS (Group 3), with a corresponding fall in the CD4 and CD8 blood lymphocyte counts (Table 2), there was a decline in the parenchymal B lymphocytes to virtually zero, both in the parenchymal and perivascular compartments. This decline in parenchymal cells compared with control brains was significant (P < 0.009). Surprisingly, when the brains were examined from patients who had developed PCNSL (Group 5), the areas of the brain outside the neoplastic foci also showed a virtual absence of parenchymal and perivascular B lymphocytes (Table 2). As with the non‐lymphoma AIDS subjects (Group 3), this reduction of parenchymal lymphocytes in non‐lymphomatous areas of PCNSL brains was significantly different from controls (P < 0.002) and the reduction of perivascular B lymphocytes compared with pre‐symptomatic patients was also significant (P < 0.026).

Fig. 3 CD20‐positive B lymphocytes (stained purple, blue arrow) in the perivascular space of a blood vessel in a pre‐symptomatic, HIV‐positive individual. The vascular endothelium is stained with Von Willebrand antibody (stained brown, black arrow).

Group 4 included AIDS subjects in whom the brain displayed quite widespread pleomorphic lymphoid infiltrates that did not have the characteristics of lymphoma and which were accompanied by similar lymphoid infiltrates in non‐CNS organs. These patients had somewhat higher CD4 and CD8 lymphocyte counts than the other AIDS groups in this study (Table 2). Quantitation of B lymphocytes in multiple areas from these brains showed a widespread increase in parenchymal B lymphocytes compared with other AIDS subjects (P < 0.037), with a very significant increase in the perivascular compartment (P < 0.002).

Figures 4 and 5 compare the means for parenchymal and perivascular B lymphocytes between the patient groups.

Fig. 4 Quantitation of parenchymal B lymphocytes in different patient groups.

Fig. 5 Quantitation of perivascular B lymphocytes in different patient groups.

In addition to using anti‐CD20 antibody, as described above, we also repeated the study using an anti‐CD79a antibody (B lymphocyte marker). Results using this antibody were not significantly different from those obtained using anti‐CD20 antibody and therefore are not presented here [P = 0.163 for Group 1 (normal); P = 0.738 for Group 2 (pre‐symptomatic HIV positive); P = 0.437 for Groups 3, 4 and 5 (AIDS)].

Semi‐quantitative assessment of the presence of T lymphocytes in the brains in these five groups confirmed previous findings of low‐level T lymphocytes in normal brains, increased numbers in the pre‐symptomatic HIV‐positive patients, and a decline in most AIDS subjects (Bell et al., 1993, 1996; Tomlinson et al., 1999). Pleomorphic infiltrates included a high proportion of CD3‐positive T cells. In subjects with PCNSL, the lymphomatous areas often had significant numbers of CD8 lymphocytes but these were not a feature of the non‐lymphomatous areas. Proliferative activity, as confirmed by Ki‐67 positivity, was seen in both the pleomorphic infiltrates and in the lymphomas but was confined to T cells in the former group, whereas in the latter both T and B lymphocytes were Ki‐67 positive, although proliferating B cells greatly outnumbered T cells (data not shown).

In situ hybridization for EBERs gave negative results in all groups except Group 5 (PCNSL). In individuals with PCNSL, EBV positivity was found within the lymphomatous areas but not in the rest of the brain. Estimation by PCR of the numbers of copies of EBV within frozen brain tissue confirmed the high level within lymphoma brains (up to 106 copies of EBV per 1 µg of DNA) (Table 2). Brain tissue from two subjects with pleomorphic lymphoid infiltrates (Group 4) displayed low copy number positivity on PCR (Table 2). PCR was negative for EBV in pre‐symptomatic HIV positive patients (Group 2, n = 2), in AIDS patients with no CNS pathology (Group 3, n = 4), and in HIV‐negative normal controls (n = 2: these were different patients from those included in Group 1 for whom no frozen tissue was available).


Our results show for the first time that B lymphocytes enter the parenchyma of normal human brain. Furthermore, this B lymphocyte entry occurs in all areas of the brain. This finding indicates the potential for humoral as well as cellular immune responses within the CNS and questions further the concept of the brain as an immunologically privileged site. Although it is not possible from immunopathological studies alone to assess the trafficking pathways of lymphocytes to and from the CNS, B lymphocytes do not appear to accumulate in the perivascular space in the normal brain. Since no obvious stimulus for immune excitation was detected in the normal brains, the presence of B lymphocytes presumably represents normal trafficking. However, the functional role of B lymphocytes in the CNS under normal conditions remains unclear, and it is not known whether they leave the CNS via the circulation or whether they are destined to die there. While the presence of any B lymphocytes within the brain indicates an absence of complete immune privilege, it does not necessarily mean that B lymphocytes have free movement into the brain, and in fact they may be restricted in their movement into this organ. In order to assess whether there is restricted movement of B lymphocytes into the brain, or whether the numbers found merely reflect the number of circulating B lymphocytes in the blood, a study of non‐CNS organs would have to be undertaken. We are currently addressing this question.

Our data further indicate that, as for T lymphocytes (Hickey et al., 1991), only B lymphocytes that display an activated phenotype, defined by expression of CD23, are able to cross the BBB and enter the CNS. If activation is a prerequisite for B lymphocyte entry into the CNS, then facilitation by T lymphocytes would generally be required as an initial step. However, immunodysregulation seen in HIV infection includes a chronic, T lymphocyte‐independent B lymphocyte activation (Lane et al., 1983). We therefore sought to establish whether HIV infection might alter the normal migration patterns of B lymphocytes within the CNS, leading to increased accumulation. We have shown that this is indeed the case in pre‐AIDS patients when compared with controls, and this finding is again in line with the increased numbers of T lymphocytes in pre‐symptomatic brains (Tomlinson et al., 1999). However, brains from the majority of AIDS subjects, including those with PCNSL (outside the site of malignancy) contained fewer B lymphocytes than normal brains. A subset of AIDS brains, previously identified during routine neuropathological examination as having heavy pleomorphic lymphoid infiltrates in the perivascular spaces (Group 4), had significantly increased numbers of B lymphocytes. Analysis of these pleomorphic lymphoid infiltrates strengthens our impression that they do not represent merely an extension of the lymphoid infiltrate seen in pre‐symptomatic patients, but constitute a separate pathology. In the former group, the number of perivascular lymphocytes is significantly higher than in the latter despite declining lymphocyte counts. While the pre‐symptomatic infiltrates consisted of small T lymphocytes (Tomlinson et al., 1999), the pleomorphic infiltrates contained an admixture of larger lymphocytes, some of plasmacytoid appearance. The finding that the pleomorphic lymphocytic infiltrates contained substantial numbers of B lymphocytes raised the possibility that they represented a pre‐malignant PCNSL state. However, in PCNSL the neoplastic infiltrates are generally focal, whereas in the non‐PCNSL brains they were diffuse. Furthermore, the brains of PCNSL cases, outside the areas of lymphoma, contained virtually no B lymphocytes, suggesting that the appearance of neoplastic foci was not superimposed on a background akin to the pleomorphic infiltrates.

Since almost all AIDS‐related PCNSL are EBV‐driven lymphoproliferations (MacMahon et al., 1992; Auperin et al., 1994), we were interested to determine whether the increased number of B lymphocytes detected in the CNS in pre‐AIDS and/or AIDS pleomorphic infiltrates contained EBV‐infected cells. We hypothesized two mechanisms that could predispose to increased numbers of EBV‐infected B lymphocytes entering the brains of HIV‐infected individuals. First, the number of circulating latently infected B lymphocytes, which is between 5 and 500 per 107 B lymphocytes in a healthy carrier (Miyashita et al., 1995), is significantly increased in HIV infection (Birx et al., 1986). Secondly, like HIV, EBV is a potent T lymphocyte‐independent polyclonal activator of B lymphocytes. Although EBV gene expression is tightly controlled in healthy carriers (Miyashita et al., 1997), expression of the B lymphocyte activating genes EBV nuclear antigen 2 and latent membrane protein 1 has been detected in the immunocompromised host (Hopwood et al., 2002). We therefore sought evidence of EBV infection initially by screening for EBV DNA by PCR and then by in situ hybridization to detect EBERs in tissue sections. We found all the PCNSL tissue tested to be EBV positive in both assays, and low level positivity for EBV DNA in the pleomorphic infiltrates. No EBERs‐positive cells were detected in these infiltrates, nor in the perivascular lymphocytic accumulations in the pre‐AIDS brains. We conclude that the PCR test used in this study (sensitivity of 1–10 EBV‐positive cells in 106 EBV‐negative cells) detected the occasional EBV‐infected B lymphocyte either in the brain tissue or blood vessels, but that EBV infection is not the primary cause of pleomorphic B lymphocyte infiltrates.

In summary, we have demonstrated for the first time that B lymphocytes infiltrate the normal brain. This finding suggests that there is no impediment to the movement of B lymphocytes from the perivascular to the parenchymal compartment of the brain. Despite the lack of detectable EBV‐positive cells, the increased presence of B lymphocytes in the CNS before the onset of AIDS may have implications for patients maintained long term on HAART. However, the lack of B lymphocytes found in most AIDS brains, in particular in PCNSL brains outside the areas of malignancy, suggests that increased B lymphocyte infiltration of the CNS is not a predisposing factor in the short term for primary PCNSL development. Although in some pre‐AIDS and AIDS subjects there is an unusual accumulation of B lymphocytes in the perivascular space, these cells are not infected with EBV, and probably do not represent a pre‐malignant condition.


We wish to thank Ms F. Brannan for technical assistance, and Dr R. Elton for guidance on statistical analysis. This work was supported by MRC grant 9808080, NIH grant RO1‐13840 and a University of Edinburgh studentship awarded to I.C.A.


View Abstract