Brain (2000), 123, 2321–2337

The peripheral benzodiazepine binding site in thebrain in multiple sclerosisQuantitative in vivo imaging of microglia as a measure ofdisease activity

R. B. Banati,1,10 J. Newcombe,4 R. N. Gunn,1 A. Cagnin,1 F. Turkheimer,1 F. Heppner,3,11 G. Price,7

F. Wegner,9 G. Giovannoni,5 D. H. Miller,5 G. D. Perkin,3 T. Smith,4,6 A. K. Hewson,4,8 G. Bydder,2

G. W. Kreutzberg,10 T. Jones,1 M. L. Cuzner4 and R. Myers1

1MRC Cyclotron Unit and 2Robert Steiner Magnetic Correspondence to: R. B. Banati, MRC Cyclotron Unit,Resonance Imaging Unit, Imperial College School of Imperial College School of Medicine, HammersmithMedicine, Hammersmith Hospital, 3Department of Hospital, Ducane Road, London W12 0NN, UKNeuroscience, Imperial College School of Medicine,Charing Cross Hospital, 4Neuroinflammation Group and5NMR Research Unit, Institute of Neurology, UniversityCollege London, 6Eisai London Research Laboratories,University College London, London, 7Department ofNeuroscience Research, SmithKline-BeechamPharmaceuticals, Harlow, 8Department of Physiology,University of Cambridge, Cambridge, UK, 9Paul FlechsigInstitute, Leipzig, 10Department of Neuromorphology,Max-Planck-Institute of Neurobiology, Munich, Germanyand 11Department of Neuropathology, University of Zurich,Switzerland

SummaryThis study identifies by microautoradiography activatedmicroglia/macrophages as the main cell type expressingthe peripheral benzodiazepine binding site (PBBS) at sitesof active CNS pathology. Quantitative measurements ofPBBS expression in vivo obtained by PET and [11C](R)-PK11195 are shown to correspond to animal experimentaland human post-mortem data on the distribution patternof activated microglia in inflammatory brain disease. Filmautoradiography with [3H](R)-PK11195, a specific ligandfor the PBBS, showed minimal binding in normal controlCNS, whereas maximal binding to mononuclear cells wasfound in multiple sclerosis plaques. However, there wasalso significantly increased [3H](R)-PK11195 binding onactivated microglia outside the histopathologically definedborders of multiple sclerosis plaques and in areas, suchas the cerebral central grey matter, that are not normallyreported as sites of pathology in multiple sclerosis. Asimilar pattern of [3H](R)-PK11195 binding in areascontaining activated microglia was seen in the CNS

Keywords: PK11195; microglia; peripheral benzodiazepine receptor; mitochondrial benzodiazepine receptor; PET; multiplesclerosis; EAE; inflammation

© Oxford University Press 2000

of animals with experimental allergic encephalomyelitis(EAE). In areas without identifiable focal pathology,immunocytochemical staining combined with high-resolution emulsion autoradiography demonstrated thatthe cellular source of [3H](R)-PK11195 binding isactivated microglia, which frequently retains a ramifiedmorphology. Furthermore, in vitro radioligand bindingstudies confirmed that microglial activation leads to arise in the number of PBBS and not a change in bindingaffinity. Quantitative [11C](R)-PK11195 PET in multiplesclerosis patients demonstrated increased PBBS expres-sion in areas of focal pathology identified by T1- and T2-weighted MRI and, importantly, also in normal-appearinganatomical structures, including cerebral central greymatter. The additional binding frequently delineatedneuronal projection areas, such as the lateral geniculatebodies in patients with a history of optic neuritis. Insummary, [11C](R)-PK11195 PET provides a cellularmarker of disease activity in vivo in the human brain.

2322 R. B. Banati et al.

Abbreviations: BP � binding potential; CR3 � complement receptor 3; EAE � experimental allergic encephalitis; EDSS �Expanded Disability Status Scale; GFAP � glial fibrillary acid protein; PBBS � peripheral benzodiazepine binding site;PMA � phorbol myristate acetate; PP � primary progressive; RR � relapsing–remitting; SP � secondary progressive;SRCC � Spearman’s rank correlation coefficient; TAC � time–activity curve(s)

IntroductionIn brain lesions without direct damage to the blood–brainbarrier, the predominant cell type expressing binding sitesfor the isoquinoline PK11195 is activated microglia, thebrain’s intrinsic macrophage (Banati et al., 1997). Origin-ally discovered as an additional binding site of certainbenzodiazepines, such as diazepam, the binding site forPK11195 is abundantly expressed in peripheral organs andhaematogenous cells and is often referred to as the ‘peripheralbenzodiazepine binding site’ (PBBS) (for review, see Hertz,1993), although it is structurally and functionally unrelatedto the central benzodiazepine receptor associated with GABA-regulated channels.

Labelled with carbon-11, PK11195 has previously beenused in a small number of patients to image focal accumula-tions of brain macrophages by PET (Benavides et al., 1988;Ramsay et al., 1992; Banati et al., 1999). There is increasingawareness that microglia are early sensors of brain pathology(Kreutzberg, 1996), and [11C](R)-PK11195 PET offers theprospect of measuring microglial activation in the livinghuman brain in diseases such as multiple sclerosis, in whichactivated microglia may contribute to tissue destruction anddisease progression (Cuzner, 1997; Sriram and Rodriguez,1997). However, a number of fundamental methodologicalissues have so far limited the interpretation of [11C](R)-PK11195 imaging data in brain inflammation. Whereas thereis evidence that areas with increased PK11195 binding containactivated microglia (Dubois et al., 1988; Myers et al., 1991a;Stephenson et al., 1995; Conway et al., 1998), the cell typeitself has not yet been shown directly (i.e. by double-labellingin situ) at the single-cell level to express the [11C](R)-PK11195 binding site in inflammatory tissue. It is also notyet established whether an increase in microglial PK11195binding would reflect a rise in the number of binding sitesor a change in their affinity. Furthermore, unlike the situationwith other ligands used in PET, the near absence of significantPK11195 binding in the normal brain makes it difficult toestablish a typical or normal pattern of PK11195 binding andpredict its change in CNS disease. To this end, in vitrodata from experimental allergic encephalomyelitis (EAE), ananimal model of multiple sclerosis, and from post-mortemCNS tissue of multiple sclerosis patients are used to interpretin vivo data of [11C](R)-PK11195 binding in multiple sclerosispatients. The kinetic behaviour of [11C](R)-PK11195 indisease tissue is quantitatively assessed by a method thataccounts for, for example, regional blood flow-dependentchanges in ligand delivery, as they may be expected ininflammatory pathology, and is relevant to the issues thatarise from the unstable stable behaviour of the ligand in

the vascular (plasma) compartment and the absence of ananatomically definable reference tissue in multiple sclerosis.

Material and methodsHuman post-mortem issuePost-mortem brain and spinal cord tissue from 11 patientswith clinically and histopathologically confirmed multiplesclerosis (mean age 57 years, range 35–63 years; meandisease duration 18 years, range 1–28 years; mean intervalfrom death to snap-freezing 41 h, range 21–61 h) and ninecontrol subjects without neurological disease (mean age55 years, range 37–72 years; mean interval from death tosnap-freezing 38 h, range 14–82 h) were obtained from theMultiple Sclerosis Society Tissue Bank at the Institute ofNeurology. Immunostaining, histological grading and filmautoradiography were performed on a total of 52 tissue blocksfrom different areas of brain and spinal cord.

Adoptive transfer of EAEAdult female Lewis rats (190–220 g) were purchased fromCharles River (Margate, Kent, UK) and housed in pairs in astandard animal facility for at least 1 week beforeexperimentation. EAE was adoptively transferred to naı̈veanimals by intraperitoneal injection of 4 � 107 splenocytescultured for 3 days with myelin basic protein (MBP) takenfrom animals 11 days after sensitization with MBP andcomplete Freund’s adjuvant (Hewson et al., 1995).

Histology and immunocytochemistryHistopathological evaluation of snap-frozen multiple sclerosistissue samples was carried out using cresyl violet, haematoxy-lin–eosin and oil red O. For immunocytochemistry, 10 µmcryostat sections of human CNS and rat brain and rat lumbarspinal cord (between levels L3 and L5) were processed andstained with a panel of antibodies using the avidin–biotin–horseradish peroxidase method as described (Smith et al.,1996). For human tissue the following antibodies wereused at appropriate dilutions: EBM11 (1 : 100) (Dako, HighWycombe, UK) and HLA class II DQ (1 : 500) (a gift fromProfessor F. Vartdal and Dr S. Funderud, Oslo, Norway) asmarkers of microglia/macrophages; CD2 for lymphocytes(1 : 100) (Dako); polyclonal anti-GFAP (glial fibrillary acidicprotein) for astrocytes (1 : 400) (Dako); and 14E for oligoden-drocytes and a reactive astrocyte subpopulation (1 : 10 super-natant) (Newcombe and Cuzner, 1988). Rat CNS sectionswere stained with markers for microglia [OX-42 recognizing

Peripheral benzodiazepine binding in CNS pathology 2323

CR3 complement receptor (1 : 3000), Dako], macrophages[ED1 (1 : 500), Dako] and astrocytes [polyclonal anti-GFAP(1 : 400)]. Controls included omission of the primary antibodyand replacement with a control IgG1 (Dianova, Hamburg,Germany).

AutoradiographyFilm autoradiography and photoemulsion microauto-radiography (Banati et al., 1997) were performed on unfixedcryostat sections of snap-frozen CNS tissue from humanpost-mortem material and animals with EAE, using custom-synthesized single enantiomer [3H]R-PK11195 (Amersham,UK), which in previous studies has been shown to possesshigher affinity for the PBBS than the commonly used racemateof PK11195 (Shah et al., 1994). Tritium standards (AmershamInternational, Little Chalfont) co-exposed on each film wereused to quantify the autoradiographically measured binding.If not specified otherwise, mean binding values were derivedfrom at least 10 sampled areas. The cell type binding [3H](R)-PK11195 was identified by further immunohistochemistrycarried out immediately after the development of the microau-toradiographic label using markers for microglia and macro-phages (for human brain, EBM11; for rat brain, OX-42) andastrocytes (GFAP).

In vitro radioligand bindingTo assess whether increased ligand binding reflects anincrease in the number of binding sites or a change in theaffinity of the binding site, additional measurements of in vitroradioligand binding with [3H](R)-PK11195 (1 nM, specificactivity 84 Ci/mmol) in the presence of unlabelled PK11195(concentration range 0.3 nM to 1 µM) were carried out onstandard membrane preparations (15 mg/ml) of animmortalized cell line with properties of activated microglialcells (Bocchini et al., 1992), applying established bindingprotocols (Itzhak et al., 1993). Maximal binding (Bmax) andaffinity (pKi) were measured before and after maximalstimulation with 1 mM PMA (phorbol myristate acetate).

Subjects of imaging studyTwelve patients with multiple sclerosis according to thecriteria of Poser and colleagues (Poser et al., 1983) (eightwomen, four men; age range 28–66 years, average age46.3 years; for diagnosis, disease duration, disability scoresand treatment at the time of investigation, see Table 1) andeight healthy normal volunteers (three women, five men; agerange 34–70 years, average age 51.75 years) were studied.Eight patients were considered to have relapsing–remitting(RR) multiple sclerosis, i.e. they had stable disability betweenrelapses. One patient had secondary progressive (SP) multiplesclerosis with increased disability over 6 months withoutsuperimposed relapses. Three patients had progressive diseasewithout intermittent remission and were classified as having

primary progressive (PP) multiple sclerosis (Lublin andReingold, 1996). Except for one patient (Patient 5), whoreceived interferon β-1b treatment, none of the patientswas under any concurrent anti-inflammatory or immuno-suppressive therapy. Informed written consent was obtainedfrom all subjects. Ethical approval and permission was givenby the Hammersmith Hospital ethics committee and theAdministration of Radioactive Substances AdvisoryCommittee of the Department of Health (ARSAC), UK.

MRIClose to the time of the PET study (i.e. within 7 days), eachsubject underwent a three-dimensional T1-weighted MRI scan[voxel size 1 � 1 � 1.3 mm; 128 contiguous slices; repetitiontime (TR) 35 ms, echo time (TE) 6 ms, flip angle 35°]. Thesewere obtained from a 1.0 T Picker HPQ MRI scanner forthe purpose of co-registration with the PET image (Studholmeet al., 1997) and exclusion of incidental pathology. In the 12multiple sclerosis patients, additional 2D proton density spin-echo (TR 2500 ms, TE 20 ms) and T2-weighted spin echoimages (TR 2500/TE 80) were acquired. The T1-weighted3D spin echo images were repeated after contrastenhancement with intravenous dimeglumine gadopentate(Gd-DPTA, 0.1 mmol/kg).

PETData acquisitionThe PET study was performed on a CTI/SiemensECAT 953B PET scanner operated in 3D acquisitionmode. [11C](R)-PK11195 was injected as a bolus 30 s afterthe acquisition scan started. The mean tracer dose was360 � 30 MBq with a specific activity of 37 � 1 GBq/mmol.Dynamic data were collected over 60 min as 18 temporalframes. Attenuation correction factors were determined usinga 15 min transmission scan acquired before the dynamic scan.Scatter correction was achieved using a dual-energy windowmethod (Grootoonk et al., 1996). Data were reconstructed witha ramp filter at Nyquist cut-off, producing an image resolutionof 5.8 mm (full-width at half maximum) at the centre of thefield of view.

Kinetic modelling and analysisAs the kinetic behaviour of [11C](R)-PK11195 in the plasmacompartment was found to be unstable, probably because ofhigh and variable plasma protein binding, the regional uptakeof [11C](R)-PK11195, extraction of which into the brain issimilar in areas with and without blood–brain barrier (Priceet al., 1990; Cremer et al., 1992), was assessed by using abasis function implementation of a simplified reference tissuemodel (Lammertma and Hume, 1996; Gunn et al., 1997;Banati et al., 1999) and calculating the following rateconstants: RI, the ratio of delivery of radioligand between

2324 R. B. Banati et al.

Table 1 Patient data

Patient Age Sex Diagnosis Clinical course EDSS Subscores

Disease Last relapse Total Pyramidal Cerebellar Brainstem Sensory Bowel/ Visual Cerebralduration bladder(years)

1 38 F RRMS 10 1 month 2 0 0 1 2 0 0 02 55 F RRMS 5 1 month 4.5 3–4 2 1 4 1 1–2 13 28 F RRMS 6 3 months 2.5 1 2 1 1 2 0 14 51 M RRMS 6 3 weeks 3.5 3 3 1 0 2 0 15 51 F RRMS 26 21 months* 4 3–4 0 1 4 0 1 06 20 F RRMS 5 5 months† 2 2 1 1 2 0 0 27 53 F RRMS 6 6 months 3 1 2 1 3 0 0 08 34 F RRMS 2 5 months† 3 2 2 1 2 2 1 09 53 M SPMS 9 5 months 7 5 5 4 0 3–4 2 2

10 46 M PPMS 10 – 5.5 3 2 1 2 1 0 011 47 F PPMS 2 – 3.5 3 1 1 3 0 0 012 66 M PPMS 8 – 6.5 3 1 1 1 1 0 0

*Interferon β-1b treatment for 20 months; †in relapse at time of scan. RRMS� relapsing–remitting multiple sclerosis; SPMS � secondaryprogressive multiple sclerosis; PPMS � primary progressive multiple sclerosis.

the target and reference tissues; k2, the efflux rate constantfrom the target tissue; and the binding potential (BP) whichis a measure of ligand binding. In the reference tissuemodel, the reference input kinetic is usually derived from ananatomical structure that is devoid of specific ligand binding,such as the cerebellum. However, in multiple sclerosis,the often widespread tissue pathology may not allow theanatomical definition of unequivocally normal referencetissue. Therefore, cluster analysis (Ashburner et al., 1996;Gunn et al., 1998) was employed as an alternative, ‘data-led’ approach for the extraction of a normal ligand kineticto serve as the reference input function. In brief, voxels inthe raw dynamic data were segmented into 10 clustersdistinguished by the shape of their concentration time–activity curves (TACs), thereby associating each dynamicimage voxel with one of the cluster curves according to thelikelihood with which the TAC of the voxel belongs to agiven cluster (Gunn et al., 1998). In normal brain, themajority (~90%) of the voxels segregated into two clusters,one representing the TAC mainly from the skull and scalpand one representing the TAC of voxels mostly located inthe cerebral cortex (Fig. 1). The latter cluster was comparedwith a normalized mean TAC (population input kinetic)previously created from the normal ligand kinetics identifiedby cluster analysis in 14 healthy subjects (eight males andsix females; age range 32–80 years, mean age 57.3 years).Whether a TAC extracted by cluster analysis from the rawdynamic data of a patient was suitable to serve as the patient’snormal ligand kinetic was assessed by testing for dissimilaritywith the previously established normal population inputkinetic (χ2 test, P � 0.05) (Fig. 1).

Definition of volumes of interestTo allow anatomical localization of the regional [11C](R)-PK11195 binding, the binding potential images were co-

registered and overlaid onto the individuals’ own MRIs(Studholme et al., 1997). For the calculation of regionalmean BP values, the following volumes of interest (VOIs)were defined anatomically on the individuals’ volumetricMRI and applied to their parametric BP image: right and leftthalamus; brainstem (between the upper and lower bordersof the pons); and right and left hemisphere excluding thethalamus. The cerebellum, seen within the restricted field ofview (10.65 cm) of the PET camera only to a varying extent,was excluded from formal analysis.

Additional VOIs were determined for the multiple sclerosispatients according to the volume of pathological tissuerevealed by proton-density, T2-weighted and T1-weightedMRI. Analyze™ software (Robb and Hanson, 1991) wasused to outline MRI lesions (Fig. 6E and F) by a localthreshold technique and for the calculation of volume (mm3).The definition of the MRI-defined lesions followedestablished criteria (Filippi et al., 1998). The MRI lesionmask for the lesions seen in the T1-weighted image volumeincluded all hypointense voxels, except those with signalintensity close to that of cerebrospinal fluid. The latter wereincluded in a mask of ‘black holes’ (Truyen et al., 1996)(Fig. 6B). The mean percentage difference (i.e. inter-ratervariability of the MRI lesion load calculation) between threeraters, who were blinded to the clinical data, was �6% forthe T2-weighted lesion loads and �5% for the T1-weightedlesion loads. Comparison of the intensity distribution of thevoxels included in the group of black holes against theremaining hypointensities of the T1-weighted MRI showedthat both lesion types were separated accurately, i.e. theywere significantly different with essentially no overlap(P � 0.0001).

By analogy with the MRI lesion loads, hemispheric[11C](R)-PK11195 lesion loads for both hemispheres werecalculated as described previously (Banati et al., 1999) by

Peripheral benzodiazepine binding in CNS pathology 2325

Fig. 1 The majority of the time–activity curves extracted by cluster analysis from the dynamic [11C](R)-PK11195 PET data of a normal brain fell into two clusters, one localizing to extracerebal structures (A) andthe other to healthy brain tissue (B). The latter represents the normal reference kinetic that would have beenobtained from an anatomically defined mask similar to that seen in B. The population input kinetic (solidline, with standard deviations) was used to decide whether an individual dynamic data set (here fromPatient 9) contained a time–activity curve suitable to serve as the individual’s reference input function(dotted line and arrow).

including only those voxels in the CNS that had a valueabove a threshold BPT (BP � BPT � 2 SD of background).

While the MRI lesion loads are reported as absolutevolumes (cm3) to allow comparison with published values,the hemispheric [11C](R)-PK11195 lesion loads are reportedas relative volumes of pathology [i.e. as a percentage ofthe sampled hemispheric volume, excluding thalamus andbrainstem, which were measured separately (see above)],thus accounting for the different field of view of the PETcamera compared with the MRI, and hence the differentabsolute sampling volume.

StatisticsSpearman’s rank correlation coefficient (SRCC) was usedto calculate correlations between disability scores, diseaseduration and total [11C](R)-PK11195 lesion load (mm3) orthe [11C](R)-PK11195 lesion load (mm3) within the VOIsdefined as being abnormal by MRI (referred to as ‘[11C](R)-PK11195 overlap’). Student’s t-test was used to determinethe significance of the differences in mean BP values betweennormal control subjects and patients.

Z-values were calculated to determine the significance ofincreased [11C](R)-PK11195 binding in brains of individual

patients compared with the group of normal control brains.As [11C](R)-PK11195 binding shows a unidirectional change,i.e. it only increases, a Z-value of 1.6 in a one-tailed Z-testrepresents a level of significance of P � 0.05 (Snedecor andCochran, 1980).

ResultsHuman post-mortem studies: [3H](R)-PK11195binding in normal control and multiple sclerosistissueThe correlation of histopathological lesion grading withmeasured [3H](R)-PK11195 binding values and the expressionof immunocytochemical markers for macrophages andactivated microglia is summarized in Fig. 2. Briefly, [3H](R)-PK11195 binding was slightly higher in normal controlcortical and central grey matter than in normal white matter.Maximal [3H](R)-PK11195 binding, of more than four timesabove normal control white matter, was reached within thoseareas of multiple sclerosis plaques that contained oil red O-positive macrophages. Even within or around oil red O-negative but EBM11-positive areas, binding was more thanthree times above the normal level. Regions of increased[3H](R)-PK11195 binding were also seen in histologically

2326 R. B. Banati et al.

Peripheral benzodiazepine binding in CNS pathology 2327

normal-appearing multiple sclerosis white matter and centralgrey matter where the increased [3H](R)-PK11195 bindingco-localized with EBM11-positive parenchymal microglia(Fig. 3K and L). Apart from some binding to infiltrating cellsin perivascular cuffs, no obvious binding to any other celltype or structure was found.

Animal studies: [3H](R)-PK11195 binding inEAEIn longitudinally cut spinal cords (Fig. 4) from rats with thetypical disease course of EAE, autoradiographs showednormal [3H](R)-PK11195 binding along the spinal grey andwhite matter columns up to day 4. At the peak of clinicaldisease, [3H](R)-PK11195 binding increased in white andgrey matter and reached maximal values in small foci.[3H](R)-PK11195 binding, similar to that seen in the spinalcord, was also found in the midbrain (Fig. 4C). After clinicalrecovery, binding returned to normal in white matter butremained increased in grey matter (Fig. 4D). OX-42immunohistochemistry demonstrated that the elevatedregional binding co-localized with perivascular infiltrates orthe cytoplasmic processes of non-phagocytic and still ramifiedactivated microglia along normal-appearing white mattertracts. At this level of single-cell resolution, there was noindication of [3H](R)-PK11195 binding to any other glialcell type.

In vitro radioligand bindingThe in vitro experiments in which [3H](R)-PK11195 wasdisplaced by unlabelled PK11195 (concentration range 0.3 nMto 1 µM) before and after maximal stimulation with PMA(1 mM) of a murine brain macrophage/microglia cell linedemonstrated a significant (P � 0.001) increase in Bmax

[unstimulated, 5.54 pM/µM (SEM 1.27, n � 3); stimulated,9.21 pM/µM (SEM 0.53, n � 3) and no change in affinity[unstimulated, pKi � 8.6 � 0.12 (SD); stimulated, pKi �8.5 � 0.12 (SD)] (Fig. 5).

Imaging studiesNormal brainBinding of [11C](R)-PK11195 in the hemispheres (excludingthe thalamus) of normal brains was minimal, with poorcontrast among the grey matter, white matter and ventricles

Fig. 2 [3H](R)-PK11195 binding in normal controls and multiple sclerosis. Tissue types: CeG � tissue containing central grey matter;CoG � cortical grey matter; W � white matter; G � grey matter; Q � multiple sclerosis plaque. Areas: L � lobe; O � occipital;P � parietal; SC � spinal cord; T � temporal; V � ventricular. Histology is scored on a scale of 0–4; the first number is the score for oil redO-positive macrophages showing the degree of ongoing demyelination, the second is the score for perivenular inflammatory cuffing. Thehistology sections were evaluated by two independent observers. [3H](R)-PK11195 binding in Bq/mg tissue was calculated using [3H]standards co-exposed on the same autoradiographic films as the tissue sections. Binding values were derived from a minimum of 10 samplingareas (multiple sclerosis plaque, area with oil red O-positive macrophages, n � 11; multiple sclerosis plaques, area without oil red O-positivemacrophages, n � 23; multiple sclerosis white matter containing activated microglia, n � 23; normal grey and white matter and background,n � 10 each). *All spinal cord sections were cut transversely except sample 3, which was cut longitudinally.

[mean background value 0.19 � 0.02 (SD)]. In normal brains,an average 5.2 � 3.1% (SD) of all voxels (sampled from thetotal volume of both hemispheres without thalamus) hadvalues above the threshold BPT. Some constitutive bindingwas seen in normal brainstem [mean BP 0.38 � 0.05 (SD)]and in the normal thalami [mean BP 0.40 � 0.05 (SD)].

Multiple sclerosis brainHighly focal binding and regionally circumscribed increasesin [11C](R)-PK11195 binding were observed in the brains ofmultiple sclerosis patients (data are summarized in Tables 2and 3). In multiple sclerosis brain, the mean BP of brainstemwas 0.49 � 0.07 (SD) while the mean BP of the thalamiwas 0.56 � 0.25 (SD), which is significantly higher than inthe normal control brains (P � 0.001 for brainstem, P � 0.05for thalami). Unlike in normal brain, [11C](R)-PK11195binding to multiple sclerosis thalami was frequentlyasymmetrical. In most multiple sclerosis patients (exceptPatients 4 and 8), at least one thalamus and/or the brainstemhad significantly increased signal compared with controls.This also applied to Patients 11 and 12, who had PP multiplesclerosis and normal MRI of the brain. Patient 11 hadan additional, clinically silent area of increased [11C](R)-PK11195 binding in the right frontal lobe.

The average hemispheric [11C](R)-PK11195 lesion load inthe group of multiple sclerosis brains was approximatelytwice that of normal brain [10 � 5.8% (SD)], ranging from4 to 30%. Due to the highly regional binding pattern and thelarge sampling volume, i.e. a full hemisphere, the [11C](R)-PK11195 lesion load per hemisphere is a greatly dilutedmeasure and was significantly increased in only four patients(Patients 5, 7, 8 and 10).

In the multiple sclerosis brains, the additionally sampledareas of MRI-defined pathology (data are summarized inTable 3) showed varying amounts of overlap with the areasof increased [11C](R)-PK11195 binding: the highest overlapwas seen with gadolinium-enhancing lesions (present inPatients 6, 8 and 9), in whom 30% of the gadolinium-enhancing volume had significantly increased [11C](R)-PK11195 binding (Fig. 6G). For the group of patients withRR multiple sclerosis who were not in relapse, the followingvalues for percentage overlap between MRI-definedpathology and increased [11C](R)-PK11195 binding werecalculated: hyperintense areas in T2-weighted MRI,10 � 3.5% (SD); hypointense areas in T1-weighted MRI

2328 R. B. Banati et al.

Fig. 3 [3H](R)-PK11195 binding in multiple sclerosis: combined high-resolution autoradiography and immunohistochemistry. (A)Autoradiograph taken from a section of a normal control tissue (sample 12 in Fig. 2) [binding in fmol/mg tissue, 18.8 � 2.8 (SD) in greymatter, 11.66 � 2.3 (SD) in white matter]. The square frame delineates the area from which the photomicrographs shown in D, G and J aretaken. (B) (Upper section) [3H](R)-PK11195 autoradiograph showing increased binding [46.28 � 4.16 (SD) fmol/mg tissue] within the areaof a multiple sclerosis plaque (arrows) in the central grey matter (sample 6 in Table 1). The accentuated edges (upper arrow) of the multiplesclerosis plaque co-localize with the presence of oil red O-positive macrophages, where binding is maximal [56.51 � 0.82 (SD) fmol/mgtissue]. (Lower section) Parietal ventricular plaque and white matter (sample 1 in Fig. 2). The square frame is positioned over an area withactivated microglia but no oil red O-positive macrophages or perivascular infiltrates; photomicrographs of this area are shown in E, H, K andL. (C) [3H](R)-PK11195 autoradiograph of a multiple sclerosis tissue from cervical spinal cord (sample 3 in Fig. 2). F and I arephotomicrographs of the framed area. (D–I) Staining of resting (D, G) and activated (E, F, H, I) parenchymal microglia with EBM11. F and Ishow activated EBM11-positive microglia following a myelinated tract. (J–L) High-resolution photoemulsion [3H](R)-PK11195autoradiography combined with EBM11 immunohistochemistry. In areas of activated microglia specific co-localization of silver grains withthe EBM11-positive cytoplasm of parenchymal microglia is demonstrated (arrows in K and L). Sections were lightly counterstained withhaematoxylin. OG � occipital cortical grey matter; CeG � central grey matter; PV � parietal ventricular; SC � spinal cord. Scale bar �10.7 mm for A and B, 7.2 mm for C, 1 mm for D–F, 300 µm for G–J and 180 µm for K–L.

Peripheral benzodiazepine binding in CNS pathology 2329

Table 2 Quantification of [11C](R)PK11195 binding

Patient Binding potential Hemispheres: Lesion load (%)

Thalamus Brainstem Right Left

Right Left

1 0.4 (n.s.) 0.36 (n.s.) 0.48 (2*) 9.2 9.52 1.19 (15.8‡) 1.22 (16.4‡) 0.60 (4.4‡) 5.6 6.33 0.90 (10‡) 0.95 (11‡) 0.50 (2.4†) 5.9 6.24 0.34 (n.s.) 0.33 (n.s.) 0.4 (n.s.) 10 9.45 0.42 (n.s.) 0.38 (n.s.) 0.50 (2.4†) 12.6 (2.4†) 12 (2.2†)6 0.4 (n.s.) 0.38 (n.s.) 0.4 (n.s.) 4.6 6.57 0.56 (3.2‡) 0.48 (1.6*) 0.64 (5.2‡) 16.6 (3.7‡) 14 (2.8†)8 0.41 (n.s.) 0.36 (n.s.) 0.4 (n.s.) 13 (2.5†) 16 (3.5‡)9 0.62 (4.4‡) 0.64 (4.8‡) 0.54 (3.2‡) 8.3 6.5

10 0.61 (4.6‡) 0.55 (3.6‡) 0.51 (2.6†) 30 (8‡) 21.2(5.2‡)11 0.49 (1.8*) 0.33 (n.s.) 0.50 (2.4†) 7 412 0.57 (3.4‡) 0.61 (4.2‡) 0.58 (4‡) 6.6 5.3

Numbers in parentheses are Z-values for regional [11C](R)-PK11195 binding or lesion loads significantlyhigher than in normal controls: *P � 0.05; †P � 0.01; ‡P � 0.001. n.s. � not significant.

Table 3 MRI lesion load and overlap with [11C](R)PK11195binding

Patient MRI lesion load (cm3) MRI–PK11195 overlap (%)

T2 T1 Black holes T2 T1 Black holes

1 – 4 0.1 – 9.3 1.32 8.2 2.6 1.1 7.3 12.4 13 6.5 8 0.3 6.1 3 5.24 2.2 3 0.4 13.8 11 115 20.5 7.4 3.6 9.4 9.2 96 2 5.6 1.4 27.8 8 17.47 13.7 9 0.9 8.8 12 6.28 9.1 2.9 0.6 31 15 219 52 13.1 34 11.5 40 5

10 4.2 3.4 7.5 14.5 17 10.211* – – – – – –

12* – – – – – –

*Patients 11 and 12 had a normal MRI.

(excluding black holes), 12 � 3% (SD); black holes, 6 � 4%(SD). For patients in acute relapse (Patients 6 and 8), theoverlap with T2-weighted hyperintense areas increased to~30% and to ~20%, respectively, in the black holes of theT1-weighted image, which in the remaining patients containedthe lowest amount of significant binding. The patient withsecondary progressive multiple sclerosis (Patient 9) showedhigh [11C](R)-PK11195 binding co-localizing with thehypointensities of the T1-weighted MRI (excluding blackholes), reaching 40% overlap.

Clear and well circumscribed signals (BP � 0.9 � 2) wereseen in anatomical structures that had a normal appearanceon each of the three MRI sequences used in this study.[11C](R)-PK11195 PET signals frequently followed neuronalfibre tracts, for example tracking through pathways in thebrainstem (Fig. 6A).

In patients who suffered from or had a recently reported

visual dysfunction, signals were observed in anatomicallocations along the pathway of the neuronal networkcontrolling visual processing or eye movement. Thus, inPatient 9, who had internuclear ophthalmoplegia, a clinicalcondition caused by a lesion to the medial longitudinalfasciculus (MLF), a circumscribed PET signal could be seenco-localizing with the MLF at the level of the mesencephalon,i.e. the predicted site of pathology (Fig. 7A and B). Similarly,patients with a history of optic neuritis (Fig. 7C and D)showed signals in the lateral geniculate bodies, to which theoptic nerve projects. Patients who had never reported anyvisual dysfunction showed no [11C](R)-PK11195 bindingassociated with any part of the brain’s visual pathway.

In Patient 6 (Fig. 7E–H), who, at the time of the scans,presented with transient speech dyspraxia (reduced fluencyof spontaneous speech with only mild language disturbance)(Mohr et al., 1978), [11C](R)-PK11195 PET revealedincreased binding in the left frontal operculum close to thearea of Broca, which contains the superior longitudinalfascicle.

A significant positive correlation (SRCC � 0.76; P �0.012) existed between the [11C](R)-PK11195 overlap withthe T1-weighted MRI hypointensities (excluding the blackholes) and the total Expanded Disability Status Scale (EDSS)score. This correlation was maintained at a lower level ofsignificance (SRCC � 0.66; P � 0.05) after exclusion ofPatient 9 (SP multiple sclerosis), who had a particularlyhigh [11C](R)-PK11195 overlap with the T1-weighted MRIlesion load.

No other statistically significant correlations were found,for example between the global hemispheric [11C](R)-PK11195 lesion load and disability (total EDSS and individualsubscores), disease duration or the interval since the lastrelapse or between the amount of [11C](R)-PK11195 overlapwith T2-weighted MRI lesion volume and either disability ordisease duration. There was, however, a trend for an inverse

2330 R. B. Banati et al.

correlation between the interval since the last relapse and the[11C](R)-PK11195 overlap with the T2-weighted MRI lesionvolume, indicating that the amount of binding within the T2-weighted MRI lesions gradually declines during the relapse-free time (SRCC � 0.63; P � 0.09). No such trend was seenwith respect to the amount of [3H](R)-PK11195 bindingwithin the T1-weighted lesions.

DiscussionPattern of PK11195 binding in vitro in multiplesclerosis and EAE (histology)High-resolution microautoradiography in multiple sclerosisand EAE tissue combined with immunohistochemical cellidentification in the same tissue section demonstrated thatthe PBBS is expressed on invading blood-borne cells at sites

Peripheral benzodiazepine binding in CNS pathology 2331

of focal inflammation and on activated microglia remotefrom any obvious inflammatory pathology.

The present study extends earlier observations (Benavideset al., 1988) of focal binding in multiple sclerosis andcorroborates indirect (Dubois et al., 1988; Myers et al.,1991a; Stephenson et al., 1995; Conway et al., 1998) anddirect (Banati et al., 1997) evidence that activated microgliaare the primary source of PK11195 binding in vivo. Inkeeping with reported data (Myers et al., 1991a, b; Stephensonet al., 1995; Conway et al., 1998) is the absence of significantin vivo binding of PK11195 to astrocytes, although bindingof PK11195 to astrocytes in cell culture is well described(Hertz, 1993; Itzhak et al., 1996). Recently, it has also beenreported that, in a neurotoxic lesion model, immunoreactivityprimarily in and around the nucleus of reactive hippocampalastrocytes can be detected by a polyclonal antibody against theperipheral benzodiazepine receptor (Kuhlmann and Guilarte,

Fig. 5 In vitro displacement of [3H](R)-PK11195 (1 nM) withvarious concentrations of unlabelled PK11195 (concentration range0.3 nM to 1 µM) before and after maximal stimulation with PMA(1 mM) of a murine brain macrophage/microglia cell line,demonstrating a significant increase in maximal binding, i.e.number of binding sites, but no significant change in affinity. Errorbars denote standard error of the mean of three independentexperiments. Open circles � unstimulated cells; closed circles �stimulated cells.

Fig. 4 Cellular localization of [3H](R)-PK11195 binding on activated microglia in EAE. The schematic drawing shows how spinal cords werecut longitudinally to allow viewing of the [3H](R)-PK11195 binding pattern along the spinal columns. (A–E) [3H](R)-PK11195autoradiographs. (A) Normal spinal cord and (B) preclinical spinal cord, i.e. 4 days after induction of EAE, showing a normal binding pattern[grey matter 21.3 � 1.01 (SD) fmol/mg tissue, white matter 7.85 � 1.07 (SD) fmol/mg tissue]. At the peak of clinical disease (day 7) (C),binding increased in white [17.48 � 0.93 (SD) fmol/mg tissue] and grey [46.76 � 8.48 (SD) fmol/mg] matter, with areas of maximum focalbinding [64.48 � 3.02 (SD) fmol/ mg tissue]. Similar increases are seen in the midbrain of the diseased animal (C, left). epend. � ependymallayer showing constitutive [3H](R)-PK11195 binding. After recovery from clinical symptoms (day 10) (D), binding is decreased to normallevels in white matter but is still slightly elevated in grey matter [33.79 � 3.09 (SD) fmol/mg tissue]. (F–M) High-resolution photoemulsion[3H](R)-PK11195 autoradiography combined with OX-42 immunohistochemistry in EAE (7 days). The square frame in E delineates thetransition zone between white and grey matter, which is shown at higher magnification in F and G. A perivascular infiltrate (left arrows) andan activated microglial cell (right arrow) stained with OX-42 are shown in F. When the focus is on the photoemulsion layer (G), silver grainsindicating [3H](R)-PK11195 binding can be seen within the perivascular infiltrate and on the activated microglial cell in the parenchyma. Thelower-magnification micrographs in H and K reveal that the focal binding of [3H](R)-PK11195 is predominantly to the perivascular cuffs,while the increased [3H](R)-PK11195 binding (arrowheads in L, arrows in M) in the white matter outside the perivascular lesion is foundexclusively in OX-42-positive microglia (arrowheads in I, arrows in J). Sections were lightly counterstained with haematoxylin.

2000). In our study, however, microautoradiography andimmunocytochemistry performed on the same section showedthat [3H](R)-PK11195 binding was restricted to GFAP-negative and CR3-positive cells. This included cells with alarger soma and short, broad processes, typical of the fullymatured brain macrophages/microglia that are usually foundat later stages of brain damage, which, according to theirmorphology, may resemble reactive astrocytes (Kuhlmannand Guilarte, 2000). The failure to find complete overlap ofthe reported immunocytochemical stain for the peripheralbenzodiazepine receptor and the cellular distribution of themicroautoradiographic PK11195 label may reflect sub-unitdependent variations in either the sensitivity or the specificityof both labels for the PBBS at its variously reportedmitochondrial, cytosolic and nuclear sites (Anholt et al., 1986;Hertz, 1993; Hardwick et al., 1999). A further potentiallyimportant difference from previous reports is the use of theR-enantiomer of PK11195, which has a higher affinity forthe PK11195-binding site, rather than the commonly usedracemate (Shah et al., 1994). Our data on (R)-PK11195binding in animal and human tissue sections show that in vivoastrocytes are not a main contributor to regionally increasedbinding, an interpretation that is supported by the lack ofsignificantly increased [11C](R)-PK11195 binding inastrocyte-rich tissue, such as in patients with hippocampalsclerosis (but infrequent seizures) (Banati et al., 1999).

An important finding of our study was the extent and theclear demarcation of PBBS expression along anatomicaltracts by microglia that were still ramified, were not co-localized with reactive astrocytes and showed few signs ofactivation. Our data suggest that during autoimmune diseasethere is widespread, under-appreciated activation of microgliain the white matter that is unrelated to autoreactive T-cellswhich precedes obvious demyelination (Cuzner, 1997). It ispossible that regional microglia that have not yet developedinto full-blown macrophages but are already activated andreleasing locally active cytotoxic factors (Banati et al.,1993; Li et al., 1996) are crucial for the process of earlydemyelination leading to a ‘dying back gliopathy’ (Ludwinand Johnson, 1981; Rodriguez and Scheithauer, 1994).

The presence of PBBS-expressing (i.e. activated) microglia

2332 R. B. Banati et al.

in normal-appearing white matter may indicate preferentialmigration along white matter tracts that are ready fordemyelination. Alternatively, in situ proliferation rather thanmigration may be the primary cause of the regionallyincreased number of microglia in areas of subtle or imminentbrain damage. For example, in the facial nerve axotomymodels without blood–brain barrier damage or neuronal celldeath, rapid activation and proliferation of microglia occursalong the entire pathway of the lesioned nerve and

simultaneously in the affected facial nucleus (Kreutzberg,1996). Similarly, a peripheral lesion of the sciatic nerveinduces remote expression of microglial PBBS in thebrainstem in the nucleus gracilis via ipsilaterally ascendingnerve fibres (Banati et al., 1997). Likewise, in animals withEAE, the inflammatory damage in the spinal cord may haveremotely induced the increased PBBS expression in themidbrain through the affected ascending spinothalamic tracts(Fig. 4C).

Peripheral benzodiazepine binding in CNS pathology 2333

In vivo imaging of the glial immunopathologyin multiple sclerosisIncreased [11C](R)-PK11195-PET signals in the brainstemand the thalamic nuclei secondary to pathology elsewherehave been observed in patients with cortical damage dueto cerebrovascular stroke (Pappata et al., 2000) and inexperimental models of cortical damage associated with asecondary microglial reaction in the ipsilateral thalamus(Myers et al., 1991b; Sorensen et al., 1996). It is likely thatthe phenomenon of a ‘projected neuroinflammatory’ responsein the wake of primary inflammatory lesions elsewhere alonga neural pathway also occurs in multiple sclerosis patients.For example, in Patient 11, the increased thalamic bindingis likely to be the consequence of a cervicospinal cord lesionseen in the MRI at level C3/C4, whereas in multiple sclerosispatients with a recent history of optic neuritis the increased[11C](R)-PK11195 signal in the lateral geniculate bodies isthe result of damage to the optic nerve. A recent longitudinalMRI study of five patients with isolated neurologicalsyndromes suggestive of multiple sclerosis found adistribution pattern of MRI-defined lesions that was spatiallydissociated from the primary inflammatory or demyelinatingfoci but followed neural pathways in a fashion that ischaracteristic of Wallerian degeneration (Simon et al., 2000).It is unclear whether the local microglial activation in thelateral geniculate bodies reflects synaptic terminaldegeneration or whether the transient axonal irritation ofthe inflamed optic nerve is a sufficient stimulus.

As expected, areas where the blood–brain barrier isdisrupted, i.e. gadolinium-enhancing lesions in T1-weightedMRI, which are likely to contain infiltrating mononuclearcells (Nesbit et al., 1991), had the highest overlap withincreased [11C](R)-PK11195 binding, whereas old structurallesions, represented by black holes (Truyen et al., 1996), hadlittle binding. However, during relapse the binding in blackholes more than doubled, indicating that during relapse thereis renewed disease activity in existing areas of severe tissuedestruction. No relapse-associated change was observed inthe other hypointense areas of the T1-weighted MRI excludingblack holes. In contrast, the hyperintense areas of the T2-weighted MRI showed an increase of more than two-fold inoverlap with the [11C](R)-PK11195 signal during relapse, and

Fig. 6 MRI and [11C](R)-PK11195 PET. All images follow the radiological convention, i.e. the left side of the image corresponds to subject’sright side. (A) Three orthogonal views of [11C](R)-PK11195 images co-registered and overlaid on the MRI of Patient 9, showingspinothalamic tract-associated [11C](R)-PK11195 signals extending through the brainstem and pons into the thalamus. (B–D) T1-weighted (B)and T2-weighted (C) MRI and [11C](R)-PK11195 PET (overlaid onto T1-weighted MRI) (D) of Patient 9 show lesions in all different spin-echo MRI sequences that partially overlap with areas of significantly increased [11C](R)-PK11195 binding (red arrow).The white arrow pointsto a ‘black hole’ in an area that appears strongly hypointense in the T1-weighted MRI and has little binding of [11C](R)-PK11195. Note,however, that a similar black hole (yellow arrowhead) adjacent to the right occipital horn of the lateral ventricle shows significant [11C](R)-PK11195 binding. (E–F) Demonstration of the definition of the MRI lesion load masks in Patient 9 (purple, T1-weighted MRI lesionsexcluding black holes; blue, black hole only; green, gadolinium-enhancing areas; dark grey (in F), T2-weighted MRI lesions; red, areas ofoverlap between significantly increased [11C](R)-PK11195 binding and MRI-defined areas of pathology); yellow, areas of increased [11C](R)-PK11195 binding and no overlap with any MRI-defined pathology. (G) Average percentage volume of the MRI-defined lesions overlappingwith increased [11C](R)-PK11195 binding. The red square represents Patient 8 and the red triangle Patient 6, who were both in relapse at thetime of the scans. The yellow diamond represents Patient 9, who had secondary progressive multiple sclerosis. T1*, black holes.

thus they probably represent a more active disease volume,i.e. PBBS-expressing tissue.

Despite the finding that the [11C](R)-PK11195 bindingwithin the T1-weighted MRI lesions (hypointensitiesexcluding black holes) appeared to be unrelated to acuterelapse, it correlated with increasing disability (total EDSS).This fraction of [11C](R)-PK11195 binding within T1-weighted MRI lesions may, therefore, reflect the secondaryneurodegenerative (Wallerian) pathology that underliesprogressive disability.

In concordance with the post-mortem data, our in vivoobservation of increased [11C](R)-PK11195 signals wellbeyond focal lesions supports the notion that additionalmechanisms apart from the focal macrophage accumulationsfound in the areas of blood–brain barrier leakage (Nesbitet al., 1991; Gay et al., 1997) must contribute to theprogression of the disease. There is mounting evidencethat hitherto undetected disease processes in brain tissue thatappears normal in MRI may explain the imperfect correlationbetween clinical neurological disability and the extent ofdisease defined by different MRI spin-echo protocols (Hustedet al., 1994; Riahi et al., 1998; Gonen et al., 2000; Tortorellaet al., 2000).

Although patients with high disability scores generally hadeither regionally and/or globally increased [11C](R)-PK11195binding in our study, we found only a limited correlationwith clinical disability measured by standardized clinicalassessment scales (EDSS) (Kurtzke, 1983). This may in partreflect the frequently discussed limitations of clinical scaleswhen applied to heterogeneous disease entities such asmultiple sclerosis (Riahi et al., 1998). It is also likely that,as a marker of current disease activity, [11C](R)-PK11195binding would relate better to parameters of clinical change(e.g. speed of disease progression) than to cumulativemeasures of long-standing and recent disability. This isdemonstrated in Patient 6, who was scanned during relapseand suffered from transient speech apraxia (Mohr et al.,1978). Here, the increased [11C](R)-PK11195 binding in thevicinity of the left Broca’s area established the clinico-pathological correlation which was not obvious in either theEDSS scores or the global lesion loads, as defined by MRIor PET imaging.

2334 R. B. Banati et al.

Fig. 7 Focal [11C](R)-PK11195 binding. (A and B) A focus of elevated [11C](R)-PK11195 binding (arrow)in the brainstem co-localizing with the medial longitudinal fasciculus (MLF) is seen in this patient (Patient9), who was suffering from internuclear ophthalmoplegia at the time of the PET scan. MR � medial rectusmuscle. (C and D) Patient 2, who had a recent history of optic neuritis. There is increased [11C](R)-PK11195 binding in the lateral geniculate bodies (LGB), which are more pronounced on the right side.The schematic drawing indicates that optic neuritis caused an anterograde neuronal reaction of the opticnerve (NII) in the lateral geniculate body. VC � visual cortex. Extracerebral signals from regions such asthe retro-orbital glands have been removed from the [11C](R)-PK11195 images for the sake of clarity.(E–H) Patient 6, who had transient speech dyspraxia (reduced fluency of spontaneous speech with onlymild language disturbance) at the time of the scans. There was increased [11C](R)-PK11195 binding in theleft frontal operculum close to Broca’s area in the anterior portion of the large white matter tract formed bythe superior longitudinal fascicle (SLF). A small but clinically silent lesion, which is gadolinium-enhancingin the T1-weighted MRI (E), is seen to have a similarly confined signal in the [11C](R)-PK11195 PET (redarrow).

Peripheral benzodiazepine binding in CNS pathology 2335

Limitation of the study and outlookBinding of [11C](R)-PK11195 was measured with a simpli-fied reference region approach and cluster analysis thatpermitted the robust extraction of a normal referencekinetic without prior anatomical definition of a normalreference region. However, one patient with SP multiplesclerosis had to be excluded from this study since noappropriate concentration time–activity curve could beextracted to serve as an input function. This may indicatethat, in this patient with extremely severe brain disease,the amount of healthy tissue from which a normal inputfunction could have been derived was too low.

The presence of small amounts of specifically bound[11C](R)-PK11195 (Petit-Taboue, 1991) in normal white andgrey matter would lead to slight underestimation of the truebinding. Since similar reference tissue curves were used bothfor patients and for controls, the small reduction in BP wouldbe consistent across all subjects.

This study has focused on the expression of the PBBS asa marker of active disease, yet some evidence points to thepossibility that the PBBS may play a more active role ininflammation and may thus itself be a target for therapeuticintervention. The PBBS is known to regulate steroidogenesis(Krueger and Papadopoulos, 1993) and modulate macrophagefunctions (Pawlikowski, 1993), while some PBBS ligandsappear to possess anti-inflammatory properties (Torres et al.,1999). Finally, microglia are activated not only in inflamma-tory diseases, such as multiple sclerosis, but also in chronicneurodegenerative diseases, such as Alzheimer’s dementia,for which the potentially beneficial effects of anti-inflammatory drug treatment are already under investigation(McGeer and McGeer, 1995). Thus, in future, a broaderconcept of activated microglia as mediators of ‘neuro-inflammation’ together with new methods for the in vivodetection of microglial activation may also be applied toprimarily non-inflammatory CNS diseases.

AcknowledgementsWe wish to thank our patients, who volunteered to participatein this study, Dr V. Cunningham for advice on PET dataquantification, J. Chalcroft, D. Buringer, K. Bruckner,M. Falkenberg and I. Milojevic for technical assistance, theradiographers A. Blythe, D. Griffith, J. Holmes, H. McDevittand L. Schnorr, and Professor S. McNeish. R.B.B. is supportedby a grant from the Deutsche Forschungsgemeinschaft andthe Multiple Sclerosis Society of Great Britain and NorthernIreland. Material and financial support was given by theMax-Planck-Institute of Psychiatry, Martinsried and by DrHammer and Dr Schneider of Boehringer Ingelheim,Ingelheim, Germany. A.C. is supported by a Marie Curiefellowship from the European Community in the Training andMobility of Researchers Programme for Biomedicine (BMH4/CT98/5100).

ReferencesAnholt RR, Pedersen PL, DeSouza EB, Snyder SH. The peripheral-type benzodiazepine receptor. Localization to the mitochondrialouter membrane. J Biol Chem 1986; 261: 776–83.

Ashburner J, Haslam J, Taylor C, Cunningham VJ, Jones T. Acluster analysis approach for the characterization of dynamic PETdata. In: Myers R, Cunningham V, Bailey D, Jones T, editors.Quantification of brain function using PET. San Diego: AcademicPress; 1996. p. 301–6.

Banati RB, Gehrmann J, Schubert P, Kreutzberg GW. Cytotoxicityof microglia. [Review]. Glia 1993; 7: 111–8.

Banati RB, Myers R, Kreutzberg GW. PK (‘peripheral benzo-diazepine’) binding sites in the CNS indicate early and discretebrain lesions: microautoradiographic detection of [3H]PK11195binding to activated microglia. J Neurocytol 1997; 26: 77–82.

Banati RB, Goerres G W, Myers R, Gunn RN, Turkheimer FE,Kreutzberg GW, et al. [11C](R)-PK11195 positron emission topo-graphy imaging of activated microglia in vivo in Rasmussen’sencephalitis. Neurology 1999; 53: 2199–203.

Benavides J, Cornu P, Dennis T, Dubois A, Hauw J-J, MacKenzieET, et al. Imaging of human brain lesions with an omega3 siteradioligand. Ann Neurol 1988; 24: 708–12.

Bocchini V, Mazzolla R, Barluzzi R, Blasi E, Sick P, KettenmannH. An immortalized cell line expresses properties of activatedmicroglial cells. J Neurosci Res 1992; 31: 616–21.

Conway EL, Gundlach AL, Craven JA. Temporal changes in glialfibrillary acidic protein messenger RNA and [3H]PK11195 bindingin relation to imidazoline-I2-receptor and alpha 2-adrenoreceptorbinding in the hippocampus following transient global forebrainischaemia in the rat. Neuroscience 1998, 82: 805–17.

Cremer JE, Hume SP, Cullen BM, Myers R, Manjil LG, TurtonDR, et al. The distribution of radioactivity in brains of rats given[N-methyl-11C]PK 11195 in vivo after induction of a corticalischaemic lesion. Int J Rad Appl Instrum B 1992, 19: 159–66.

Cuzner ML. Molecular biology of microglia. In: Russell WC, editor.Molecular biology of multiple sclerosis. Chichester (UK): Wiley;1997. p. 97–120.

Dubois A, Benavides J, Peny B, Duverger D, Fage D, Gotti B,et al. Imaging primary and remote ischaemic and excitotoxic brainlesions. An autoradiographic study of peripheral type benzodiazepinebinding sites in the rat and cat. Brain Res 1988; 445: 77–90.

Filippi M, Horsefield MA, Ader HJ, Barkhof F, Bruzzi P, Evans A,et al. Guidelines for using quantitative measures of brain magneticresonance imaging abnormalities in monitoring the treatment ofmultiple sclerosis. Ann Neurol 1998; 43: 499–506.

Gay FW, Drye TJ, Dick GW, Esiri MM. The application ofmultifactorial cluster analysis in the staging of plaques in earlymultiple sclerosis. Identification and characterization of the primarydemyelinating lesion. Brain 1997; 120: 1461–83.

Gonen O, Catalaa I, Babb JS, Ge Y, Mannon LJ, Kolson DL, et al.Total brain N-acetylaspartate: a new measure of disease load inMS. Neurology 2000; 54: 15–9.

2336 R. B. Banati et al.

Grootoonk S, Spinks TJ, Sashin D, Spyrou NM, Jones T. Correctionfor scatter in 3D brain PET using a dual energy window method.Phys Med Biol 1996; 41: 2757–74.

Gunn RN, Lammertsma AA, Hume SP, Cunningham VJ. Parametricimaging of ligand-receptor binding in PET using a simplifiedreference region model. Neuroimage 1997; 6: 279–87.

Gunn RN, Lammerstma AA, Cunningham VJ. Parametric imagingof ligand–receptor interactions using a reference tissue modeland cluster analysis. In: Carson RE, Daube-Witherspoon ME,Herscovitch P, editors. Quantitative functional brain imaging withpositron emission tomography. San Diego: Academic Press; 1998.p. 401–6.

Hardwick M, Fertikh D, Culty M, Li H, Vidic B, Papadopoulos V.Peripheral-type benzodiazepine receptor (PBR) in human breasttissue: correlation of breast cancer cell aggressive phenotype withPBR expression, nuclear localization and PBR-mediated cell prolif-eration and nuclear transport of cholesterol. Cancer Res 1999; 59:831–42.

Hertz L. Binding characteristics of the receptor and coupling totransport proteins. In: Giesen-Crouse E, editor. Peripheral benzodiaz-epine receptors. London: Academic Press; 1993. p. 27–57.

Hewson AK, Smith T, Leonard JP, Cuzner ML. Suppression ofexperimental allergic encephalomyelitis in the Lewis rat by thematrix metalloproteinase inhibitor Ro31-9790. Inflamm Res 1995;44: 345–9.

Husted CA, Goodin DS, Hugg JW, Maudsley AA, Tsuruda JS, deBie SH, et al. Biochemical alterations in multiple sclerosis lesionsand normal-appearing white matter detected by in vivo 31P and 1Hspectroscopic imaging. Ann Neurol 1994; 36: 157–65.

Itzhak Y, Baker L, Norenberg MD. Characterization of the peripheral-type benzodiazepine receptors in cultured astrocytes: evidence formultiplicity. Glia 1993; 9: 211–8.

Kreutzberg GW. Microglia: a sensor for pathological events in theCNS. [Review]. Trends Neurosci 1996; 19: 312–8.

Krueger KE, Papadopoulos A. The role of mitochondrialbenzodiazepine receptors in steroidogenesis. In: Giesen-Crouse E,editor. Peripheral benzodiazepine receptors. London: AcademicPress; 1993. p. 87–109.

Kuhlmann AC, Guilarte TR. Cellular and subcellular localization ofperipheral benzodiazepine receptors after trimethyltin neurotoxicity.J Neurochem 2000; 4: 1694–704.

Kurtzke JF. Rating neurologic impairment in multiple sclerosis: anexpanded disability status scale (EDSS). Neurology 1983; 33:1444–52.

Lammertsma AA, Hume SP. Simplified reference tissue model forPET receptor studies. Neuroimage 1996; 4: 153–8.

Li H, Cuzner ML, Newcombe J. Microglia-derived macrophages inearly multiple sclerosis plaques. Neuropathol Appl Neurobiol 1996;22: 207–15.

Lublin FD, Reingold SC. Defining the clinical course of multiplesclerosis: results of an international survey. National MultipleSclerosis Society (USA) Advisory Committee on Clinical Trials ofNew Agents in Multiple Sclerosis. Neurology 1996; 46: 907–11.

Ludwin SK, Johnson ES. Evidence for a ‘dying-back’ gliopathy indemyelinating disease. Ann Neurol 1981; 9: 301–5.

McGeer PL, McGeer EG. The inflammatory response system of brain:implications for therapy of Alzheimer and other neurodegenerativediseases. [Review]. Brain Res Brain Res Rev 1995; 21: 195–218.

Myers R, Manjil LG, Cullen BM, Price GW, Frackowiak RS, CremerJE. Macrophage and astrocyte populations in relation to [3H]PK11195 binding in rat cerebral cortex following a local ischaemic lesion.J Cereb Blood Flow Metab 1991a; 11: 314–22.

Myers R, Manjil LG, Frackowiak RS, Cremer JE. [3H]PK 11195and the localisation of secondary thalamic lesions following focalischaemia in rat motor cortex. Neurosci Lett 1991b; 133: 20–4.

Mohr PJ, Pessin MS, Finkelstein S, Funkenstein HH, Duncan GW,Davis KR. Broca aphasia: pathologic and clinical. Neurology 1978;28: 311–24.

Nesbit GM, Forbes GS, Scheithauer BW, Okazaki H, Rodriguez M.Multiple sclerosis: histopathologic and MR and/or CT correlation in37 cases at biopsy and three cases at autopsy. Radiology 1991; 180:467–74.

Newcombe J, Cuzner ML. Monoclonal antibody 14E identifies theoligodendrocyte cell body in normal adult human and rat white matter.J Neuroimmunol 1988; 19: 11–20.

Pappata S, Levasseur M, Gunn RN, Myers R, Crouzel C, SyrotaA, et al. Thalamic microglial activation in ischemic stroke patientsdetected in vivo by PET and [11C]PK11195. Neurology. In press 2000.

Pawlikowski M. Immunomodulating effects of peripherally actingbenzodiazepines. In: Giesen-Crouse E, editor. Peripheralbenzodiazepine receptors. London: Academic Press; 1993. p. 125–35.

Petit-Taboue MC, Baron JC, Barre L, Travere JM, Speckel D,Camsonne R, et al. Brain kinetics and specific binding of [11C]PK11195 to omega 3 sites in baboons: positron emission tomographystudy. Eur J Pharmacol 1991; 200: 347–51.

Poser CM, Paty DW, Scheinberg L, McDonald WI, Davis FA, EbersGC, et al. New diagnostic criteria for multiple sclerosis: guidelinesfor research protocols. Ann Neurol 1983; 13: 227–31.

Price GW, Ahier RG, Hume SP, Myers R, Manjil LG, Cremer JE,et al. In vivo binding to peripheral benzodiazepine binding sites inlesioned rat brain: comparison between [3H]PK11195 and[18F]PK14105 as markers for neuronal damage. J Neurochem 1990;55: 175–85.

Ramsay SC, Weiller C, Myers R, Cremer JE, Luthra SK, LammertsmaAA, et al. Monitoring by PET of macrophage accumulation in brainafter ischaemic stroke [letter]. Lancet 1992; 339: 1054–5.

Riahi F, Zijdenbos A, Narayanan S, Arnold D, Francis G, Antel J,et al. Improved correlation between scores on the expanded disabilitystatus scale and cerebral lesion load in relapsing–remitting multiplesclerosis. Results of the application of new imaging methods. Brain1998; 121: 1305–12.

Robb RA, Hanson DP. A software system for interactive andquantitative visualization of multidimensional biomedical images.Australas Phys Sci Eng Med 1991; 14: 9–30.

Rodriguez M, Scheithauer B. Ultrastructure of multiple sclerosis.Ultrastruct Pathol 1994; 18: 3–13.

Peripheral benzodiazepine binding in CNS pathology 2337

Shah F, Hume SP, Pike VW, Ashworth S, McDermott J. Synthesis ofthe enantiomers of [N-methyl-11C]PK11195 and comparison of theirbehaviours as radioligands in rats. Nucl Med Biol 1994; 21: 573–81.

Simon JH, Kinkel RP, Jacobs L, Bub L, Simonian N. A Walleriandegeneration pattern in patients at risk for MS. Neurology 2000; 54:1155–60.

Smith T, Schmied M, Hewson AK, Lassmann H, Cuzner ML.Apoptosis of T cells and macrophages in the central nervous systemof intact and adrenalectomized Lewis rats during experimental allergicencephalomyelitis. J Autoimmun 1996; 9: 167–74.

Snedecor GW, Cochran WG. Statistical methods. 7th ed. Ames (IA),Iowa State University Press; 1980.

Sorensen JC, Dalmau I, Zimmer J, Finsen B. Microglial reactions toretrograde degeneration of tracer-identified thalamic neurons afterfrontal sensorimotor cortex lesions in adult rats. Exp Brain Res 1996;112: 203–12.

Sriram S, Rodriguez M. Indictment of the microglia as the villain inmultiple sclerosis. [Review]. Neurology 1997; 48: 464–70.

Stephenson DT, Schober DA, Smalstig EB, Mincy RE, Gehlert DR,

Clemens JA. Peripheral benzodiazepine receptors are colocalizedwith activated microglia following transient global forebrain ischemiain the rat. J Neurosci 1995; 15: 5263–74.

Studholme C, Hill DL, Hawkes DJ. Automated three-dimensionalregistration of magnetic resonance and positron emission tomographybrain images by multiresolution optimization of voxel similaritymeasures. Med Phys 1997; 24: 25–35.

Torres SR, Nardi GM, Ferrara P, Ribeiro-do-Valle RM, Farges RC.Potential role of peripheral benzodiazepine receptors in inflammatoryresponses. Eur J Pharmacol 1999; 385: R1–2.

Tortorella C, Viti B, Bozzali M, Sormani MP, Rizzo G, Gilardi MF,et al. A magnetization transfer histogram study of normal-appearingbrain tissue in MS. Neurology 2000; 54: 186–93.

Truyen L, van Waesberghe JH, van Walderveen MA, van Oosten BW,Polman CH, Hommes OR, et al. Accumulation of hypointense lesions(‘black holes’) on T1 spin-echo MRI correlates with diseaseprogression in multiple sclerosis. Neurology 1996; 47: 1469–76.

Received March 27, 2000. Revised June 7, 2000.Accepted June 29, 2000