Rapid Communication

Imaging of brain hypoxia in permanent andtemporary middle cerebral artery occlusion in therat using 18F-fluoromisonidazole and positronemission tomography: a pilot study

Masashi Takasawa1,2,*, John S Beech3,*, Tim D Fryer4, Young T Hong4, Jessica L Hughes1,2,Keiji Igase1,2, P Simon Jones1, Rob Smith4, Franklin I Aigbirhio1,4, David K Menon3,John C Clark4 and Jean-Claude Baron1

1Stroke Research Group, Department of Clinical Neurosciences, University of Cambridge, Cambridge, UK;2Department of Clinical Neurosciences, Centre for Brain Repair, University of Cambridge, Cambridge, UK;3Division of Anaesthesia, Department of Medicine, University of Cambridge, Addenbrooke’s Hospital,Cambridge, UK; 4Department of Clinical Neurosciences, Wolfson Brain Imaging Centre, University ofCambridge, Cambridge, UK

In acute stroke, the target of therapy is the severely hypoxic but salvageable tissue. Previous humanstudies using 18F-fluoromisonidazole and positron emission tomography (18F-FMISO PET) haveshown high tracer retention indicative of tissue hypoxia, which had normalized at repeat scan > 48 hlater. In the only validation study of 18F-FMISO, using ex vivo autoradiography in thread middlecerebral artery occluded (MCAo) rats, there was unexpected high uptake as late as 22 h afterreperfusion, raising questions about the use of 18F-FMISO as a hypoxia tracer. Here we report a pilotstudy of 18F-FMISO PET in experimental stroke. Spontaneous hypertensive rats were subjected todistal clip MCAo. Three-hour dynamic PET was performed in 7 rats: 3 normals, 1 with permanentMCAo (two sessions: 30 mins and 48 h after clip), and 3 with temporary MCAo (45 mins, n = 1;120 mins, n = 2; scanning started 30 mins after clip removal). Experiments were terminated byperfusion–fixation for standard histopathology. Late tracer retention was assessed by bothcompartmental modelling and simple side-to-side ratios. In the initial PET session of the permanentMCAo rat, striking trapping of 18F-FMISO was observed in the affected cortex, which had normalized48 h later; histopathology revealed pannecrosis. In contrast, there was no demonstrable tracerretention in either temporary MCAo models, and histopathology showed ischemic changes only.These results document elevated 18F-FMISO uptake in the stroke area only in the early phase ofMCAo, but not after early reperfusion nor when tissue necrosis has developed. These findingsstrongly support the validity of 18F-FMISO as a marker of viable hypoxic tissue/penumbra afterstroke.Journal of Cerebral Blood Flow & Metabolism (2007) 27, 679–689. doi:10.1038/sj.jcbfm.9600405; published online11 October 2006

Keywords: hypoxia; ischemia; penumbra; positron emission tomography; SHR; stroke

Introduction

In acute stroke, the penumbra, that is, the severelyhypoxic but potentially salvageable region sur-rounding the ischemic core, is the main target fortherapy (Baron, 2001a, b). It is characterized byreduced perfusion in the face of relatively main-tained oxygen consumption, which translates asreduced tissue partial O2 tension, increased O2

gradient from capillary to tissue, and elevated O2

extraction fraction.Received 26 June 2006; revised 23 August 2006; accepted 27August 2006; published online 11 October 2006

Correspondence: Professor JC Baron, Department of Neurology,Cambridge University, Addenbrooke’s Hospital, Box 83, HillsRoad, Cambridge CB2 2QQ, UK.E-mail: jcb54@cam.ac.uk

MT was supported by Japan Society for the Promotion of Science

(JSPS) Post Doctoral Fellowships for Research Abroad (17-307).

This study was supported by Medical Research Council (MRC)

Grant G0001219 to JCB. No conflicts of interest to declare.

*Both authors contributed equally to this work.

Journal of Cerebral Blood Flow & Metabolism (2007) 27, 679–689& 2007 ISCBFM All rights reserved 0271-678X/07 $30.00

www.jcbfm.com

Using invasive devices such as oxygen electrodesor electron paramagnetic resonance, tissue pO2

has been directly measured in animal strokemodels (Crockard et al, 1976; Liu et al, 2004). It hasoccasionally been possible to measure tissue pO2

directly in man by insertion of oxygen electrodesinto the parenchyma during craniotomy (Dings et al,1998), but this is not practical in the clinical setting.

Recently, the hypoxia-sensitive nitroimidazoleshave been used for direct mapping of the hypoxictissue after stroke (Nunn et al, 1995; Baron andWarach, 2005). Nitroimidazole compounds such as18F-fluoromisonidazole (18F-FMISO) diffuse freelyacross cell membranes and in living cells are thenreduced by intracellular reductases into a radicalanion. In normoxic conditions, this compound israpidly reoxidized and diffuses back out of cells.Under hypoxic conditions, that is, inadequateoxygen supply relative to demand, however, furtherreduction steps occur, and the reduced compoundbecomes irreversibly bound to intracellular macro-molecules. Thus, reduced nitroimidazoles are sup-posed to be trapped within hypoxic cells, but not bynecrotic or nonhypoxic cells (Nunn et al, 1995).

The Melbourne group have investigated thepotential of 18F-FMISO for detecting in vivo thehypoxic area and convincingly showed areas ofincreased 18F-FMISO uptake in acute stroke patientsup to 48 h after clinical onset, fulfilling the opera-tional criteria for penumbra (Read et al, 1998, 2000;Markus et al, 2004). Further, abnormal 18F-FMISOtrapping was not observed at follow-up 18F-FMISOscan 6 to 11 days after onset (Read et al, 1998). Theseresults were entirely consistent with the biochem-ical behavior outlined above. In the only experi-mental study on 18F-FMISO, Saita et al (2004), usingex vivo autoradiography, showed the expected highuptake during occlusion (using a 120 mins occlusiontime), but also reported residual areas of increased18F-FMISO retention in the affected areas as late as20 h after removal of the middle cerebral artery(MCA) thread occlusion. The finding of residualtracer trapping long after removal of the occludingdevice raised questions about the mechanism(s) of18F-FMISO trapping and its use as a hypoxia marker,but the authors proposed periinfarct edema, smallvessel occlusion, or the ‘no-re-flow’ phenomenon as

possible explanations for possibly persistent hypoxia.No in vivo study on 18F-FMISO in ischemic strokemodels has been reported to date.

In the present study, we report pilot results of18F-FMISO positron emission tomography (PET) inexperimental stroke. Rats were subjected to clipMCA occlusion (MCAo), followed by full histo-pathology. Our primary aim was not to documenttrapping of 18F-FMISO in the situation of acuteischemia, which is well established (Read et al,1998, 2000; Markus et al, 2004; Saita et al, 2004), butto investigate whether significant trapping wouldoccur when the tissue is necrotic, for example, 48 hafter permanent MCAo, or when it has been througha phase of severe ischemia but is neither necroticnor hypoxic, for example, after brief MCAo.

Materials and methods

Animals and Experimental Design

All animal experiments were in accordance with UKHome Office guidelines, and were approved the Univer-sity of Cambridge Animal Ethical Review Panel.

Seven male spontaneous hypertensive rats (CharlesRiver Laboratory, Margate, UK) weighing 295 to 305 gwere used. The experimental design, including the timingof 18F-FMISO injection relative to the start of MCAo, isshown in Table 1. In all rats, we implemented 3-hour18F-FMISO dynamic PET scanning. Three normal rats wereused as controls. One rat was subjected to permanentMCAo (2 PET sessions: 1 early after clip placement and theother 48 h later), and 3 rats to temporary MCAo (1 PETsession each after 45- or 120-minutes temporary MCAo;n = 1 and n = 2 rats, respectively). Experiments wereterminated by perfusion–fixation immediately after com-pletion of the final PET session.

Anesthesia

Anesthesia was induced with 4% isoflurane administeredin a 0.3 L/mins O2 and 0.7 L/mins N2O mix, andmaintained with 2% isoflurane during all surgicalprocedures. The left femoral vein was cannulated for theinjection of 18F-FMISO, and blood samples in both normaland stroke rats were obtained via right femoral artery

Table 1 Experimental design

Rat Group No. of PET scans 18F-FMISO injection after the start of MCAo Blood sampling Histology

1 Normal 1 — Whole TAC N/A2 Normal 1 — Whole TAC N/A3 Normal 1 — — N/A4 Permanent 2 39 mins 4 samples N/A

48 h 4 samples +5 Temporary (45 mins) 1 73 mins 4 samples +6 Temporary (120 mins) 1 159 mins 4 samples +7 Temporary (120 mins) 1 158 mins 4 samples +

MCAo, middle cerebral artery occlusion; N/A, not applicable; TAC, time–activity curve.

18F-FMISO PET in rat MCA occlusionM Takasawa et al

680

Journal of Cerebral Blood Flow & Metabolism (2007) 27, 679–689

cannulation. A heparinized saline flush was used tomaintain the arterial line patency. Body temperature wasmaintained at 371C throughout the surgery and PETscanning using a rectal temperature-regulated heating pad.

Focal Ischemia

Middle cerebral artery occlusion was performed using themethod described by Buchan et al (1992). Briefly, the rightcommon carotid artery was isolated through a ventralmidline neck incision, and a 4-0 surgical silk ligatureloosely placed around the artery. With the rat positionedonto its left flank, a 1-cm incision perpendicular to andbisecting a line between the lateral canthus of the right eyeand the external auditory canal was made. The underlyingtemporalis muscle was excised, and under direct visuali-zation, the right MCA was exposed through a 2-mmburr hole drilled 2 to 3 mm rostral to the fusion of thezygomatic arch with the squamosal bone. Drilling wasperformed under saline irrigation.

The MCA was visualized where it crosses the inferiorcerebral vein, which lies within the rhinal fissure. The siteof the occlusion is proximal to the MCA bifurcation, butdistal to the origin of the lenticulostriate arteries. The duraoverlying the MCA was cut and retracted. A #1 micro-aneurysm clip (Codman, Sundt AVM, Raynham, MA,USA) was placed on the MCA, the wound was closed, andthen the right common carotid artery was ligated, asprepared beforehand.

In the permanent MCAo model (Rat #4), the micro-aneurysm clip remained in situ for the duration of thestudy until perfusion fixation. Positron emission tomo-graphy scanning was carried out twice, shortly afterocclusion and 48 h later. In the temporary MCAo models,the clip was removed 45 mins (Rat #5) and 120 mins(Rat #6 and #7) after the start of MCAo. In each rat,effective recanalization of the MCA was visually verified.

Positron Emission Tomography Scanning

Animals were imaged using a micro-PET P4 scanner(Concorde Microsystems, Knoxville, TN, USA) (Tai et al,2001). After the surgical procedures (described above), theanimal was placed prone on the scanning bed, and locatedin a purpose-built plastic frame incorporating ear bars anda bite bar. In all cases, the brain was centered in the fieldof view of the scanner (78 mm axial� 200 mm diameter)to maximize sensitivity and resolution (B2 mm). Oncepositioned on the bed and for the duration of PETscanning, anesthesia was reduced to 1% to 1.5% isofluranein 0.3 L/mins O2 and 0.7 L/mins N2O mix.

In the permanent MCAo rat, the first PET sessioncommenced as quickly as possible after clip placement(39 mins later). In the temporary MCAo rats, scanning wasalso started as quickly as possible after clip removal (28 to39 mins later) (Table 1).

In each rat, around 80 MBq of 18F-FMISO were injectedintravenously as a bolus. The PET timeframes were asfollows: 10� 30 secs, 15� 1 mins, 5� 2 mins, and 30� 5mins (total: 60 frames, up to 180 mins after injection).

The energy and timing windows used were 350 to 650 keVand 6 nsecs, respectively. Emission data acquisition wasfollowed by transmission scans for attenuation correction(acquisition time: 20 mins) with a rotating germanium-68/gallium-68 point source. The transmission scans wereperformed after the emission scans to start the latter assoon as possible after clip placement/removal. Windowedcoincidence mode transmission scanning was used toreduce emission contamination of the transmission datato < 1%.

Throughout the imaging protocol, SaO2 and heart rateswere continuously monitored using noninvasive pulseoximetry. The latter were maintained at 96% to 98% and350 to 360 beats/mins, respectively.

Blood Sampling

To determine arterial blood 18F-FMISO kinetics, twonormal rats underwent serial blood sampling (Table 1).Blood samples were withdrawn every 3 secs for the first3 mins, then every minute for the next 7 mins, every5 mins for 20 mins, and every 30 mins for the final150 mins. To limit blood loss, blood samples were takenat 30, 60, 120, and 180 mins after tracer injection from theMCAo rats. These samples were used to scale an inputfunction determined from a control rat.

All samples were stored in 0.2 mL microcentrifuge tubesin ice-cold water. The radioactivity concentration of eachsample (in kBq/mL) was measured using a sodium iodidewell counter (Canberra Harwell, Didcot, UK). To study thedistribution of 18F-FMISO in blood, immediately afterradioactivity counting in whole blood, the samples werecentrifuged and the radioactivity concentration in plasmawas measured.

Image Analysis

The images were reconstructed using the PROMIS 3Dfiltered backprojection algorithm (Kinahan and Rogers,1989), adapted in-house to work with data from the micro-PET P4 scanner.

Corrections for randoms, dead time, background, nor-malization, attenuation, sensitivity, and decay wereapplied to the data during reconstruction. Images werereconstructed into 0.5� 0.5� 0.5 mm3 voxels in an array of180� 180� 151 and a Hanning window cutoff at theNyquist frequency was incorporated into the reconstruc-tion filters to give an image resolution of B2.3 mm full-width at half-maximal.

The acquired data were transformed and analyzed usingthe software package Analyze 6.0 (AnalyzeDirect Inc.,Lenexa, KS, USA). First, the 0 to 180 mins added PETimage sets from all rats were coregistered to a magneticresonance imaging (MRI) T1 template of a healthyspontaneous hypertensive rat rat of similar age andweight, using a six-parameter rigid manual coregistrationas previously reported by Hughes et al (2005) in ourlaboratory using the same rat MCAo model. An ‘affectedcortex region of interest (ROI)’ was defined across 36contiguous coronal slices as the intersection between the

18F-FMISO PET in rat MCA occlusionM Takasawa et al

681

Journal of Cerebral Blood Flow & Metabolism (2007) 27, 679–689

upper 80% threshold of the 18F-FMISO added scan of thefirst PET session of the permanent MCAo rat (wherehypoxia and thus tracer trapping were expected), and theborders of the cortex as defined on the coronal MRI (seeFigure 2E for illustration). A mirror ROI was thengenerated on the unaffected cortex.

This set of ROIs (with their mirrors) was then applied tothe resliced dynamic PET data of all the rats. Theweighted mean radioactivity concentration (in kBq/mL)in the set of ROIs on each side was then calculated toproduce a mean time–activity curve (TAC) for each side.

Kinetic Modelling

Based on the biochemical behavior of 18F-FMISO, anirreversible two-tissue compartmental model was used inthis study. The arterial input function for rats other thanRat #1 was approximated by scaling the arterial inputfunction from Rat #1 through least-squares fitting with thefour discrete blood samples (Hughes et al, 2005).

K1 (transport rate from plasma to the ‘free’ tissuecompartment) and Ki (influx rate from plasma to the‘bound’ compartment) were computed from the meanTACs (one for each hemisphere) and arterial input functionusing in-house compartmental modelling software. Ki wasdetermined from [K1 � k3/(k2 + k3)], where k2 indicates thetransport rate from the ‘free’ compartment to plasma andk3 describes the influx rate constant from the ‘free’ to the‘bound’ compartment. The rate constant k4 was assumed tobe zero because of the chemically covalent binding inhypoxic tissue. Fitting for K1, k2, and k3 used iterativelyreweighted least squares, which adjusts the weightingfactors according to noise estimation (Muzic and Christian,2006). Right/left (controls) or affected/unaffected side(MCAo) ratios for K1 and Ki were then calculated.

In addition, the right/left or affected/unaffected ratiosof radioactivity concentrations (to be referred to as AR)were also computed, for the first 10 PET frames (0 to5 mins) to obtain an index of perfusion, and for the last10 frames (130 to 180 mins) to assess late uptake/retention.The right/left or affected/unaffected ratios of radioactivityconcentrations were obtained as they can be less prone tocalculation errors than K1 and Ki, and also because onlylate scans ARs have been reported in previous publica-tions on 18F-FMISO in stroke (Read et al, 1998, 2000;Markus et al, 2004; Saita et al, 2004).

Histology

In the MCAo rats, immediately after the completion of thePET scanning protocol (Table 1), the rat was transcardiallyperfused with 4% paraformaldehyde, the brain harvestedand postfixed in paraformaldehyde overnight. The follow-ing day, the brain was immersed in 30% sucrose solutionuntil it sank to the bottom of the vial, then 40-mm-thickcoronal sections were cut on a cryostat, mounted on poly-L-lysine-coated glass slides, and stored at �201C untilanalysis.

The mounted sections were carefully dehydrated in70%, 95%, and 100% ethanol for 15 mins, and then

immersed back through 95%, 70% ethanol and distilledwater. The slides were immersed in Cresyl Violet (CV)solution for 30 mins. They were briefly washed in distilledwater, 70%, 95%, and 100% ethanol, cleared with xylene,and then coverslipped (Tureyen et al, 2004). Cresyl Violet-stained tissue was evaluated by light microscopy.

Results

Normal Rats

The TAC of the whole blood in a normal rat isillustrated in Figure 1A. The whole blood/plasmaradioactivity ratio was essentially stable throughoutthe experiment (1.0670.05, mean value71 s.d.) (Fig-ure 1B). The results in the other rat were very similar,with a whole blood/plasma ratio of 0.9870.04.

18F-fluoromisonidazole distribution (weightedsum across the entire 60 frames) is illustrated inFigure 1C. The brain was well delineated, butsummed uptake was higher in some extracerebraltissues. Time–activity curves for both cerebral hemi-spheres are illustrated in Figure 1D. Activity peakedaround 30 mins after tracer administration, followedby a gradual decline. The results for the other ratswere very similar.

Results for K1, Ki, their ratios, and AR are shownin Table 2. The values for both hemispheres weresimilar for all variables.

Permanent Middle Cerebral Artery Occlusion

The quantitative values are shown in Table 2 and theTACs and PET images illustrated in Figure 2.

In the first PET session, early uptake in theaffected hemisphere was lower than in the unaf-fected hemisphere (Figure 2A, left panel; Figure 2B;early AR = 0.70). Subsequently, there was gradualretention of the tracer on the affected side, reachinggreater than twice that in the unaffected side nearthe end of scanning (Figure 2A, right panel; Figure2B; late AR = 2.63). The PET images fused to the MRItemplate illustrate that the high tracer uptakecentered on the cortical ribbon, as expected withthis MCAo model (Figure 2E, top row). There wasalso some 18F-FMISO accumulation in the overlyingsurgically damaged temporalis muscle (Figure 2A).

In the second PET session (48 h later), early traceruptake in the affected hemisphere was again lowerthan in the unaffected hemisphere (Figure 2C, leftpanel; Figure 2D; early AR = 0.78). However, therewas no subsequent tracer retention in the affectedside, nor in temporalis muscle (Figure 2C, rightpanel; Figure 2E, lower row). The TAC for theaffected side displayed a slow upslope during theearly phase, which then reached a plateau tosuperimpose onto the unaffected side at the end ofscanning (Figure 2D; late AR = 1.09). Both the K1 andKi values for the unaffected side were similar tothose of the two normal rats (Table 2). In the affected

18F-FMISO PET in rat MCA occlusionM Takasawa et al

682

Journal of Cerebral Blood Flow & Metabolism (2007) 27, 679–689

cortex, there was a markedly increased Ki value inthe first PET session (Ki ratio = 8.08), which was backwithin the normal ranges for the second session. The

K1 ratio was decreased at both first and second PETsessions (ratio = 0.74 and 0.62; Table 2). Figure 3illustrates the results of the fitting procedure for the

0

100

200

300

400

500

0 30 60 90 120 150 180

Whole Blood

Rad

io a

ctiv

ity C

once

ntra

tion

(kB

q/m

l)

Who

le B

lood

/Pla

sma

A

C D

B

Left hemisphere

Right hemisphere

0

1000

2000

3000

1801206003 hours

2SD

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

0 20 40 60 80 100 120 140 160 1800

1000

2000

3000

0 5 10 15 20 25

Rad

ioac

tivity

Con

cent

ratio

n (k

Bq/

ml)

Time Post-injection (min)

Time Post-injection (min)Time Post-injection (min)

CoronalTransverseRtLt

Nose

Figure 1 Blood tracer kinetics and PET data in a normal rat (#1). (A) Decay-corrected whole-blood TAC; inset shows the entire 3-hour TAC; (B) whole blood/plasma radioactivity ratio over time. The dotted line shows the 72 s.d. confidence interval; (C)distribution of 18F-FMISO in the head (0 to 3 h weighted-sum image); (D) TACs for both hemispheres, showing peak activity around30 mins after tracer injection, followed by a gradual decline.

Table 2 PET results and histology

K1 Ki AR

Rat Rt. (affected) Lt. (unaffected) Ratio Rt. (affected) Lt. (unaffected) Ratio Earlya Lateb Histological findings

Normal rats1 0.0468 0.0395 1.19 0.0016 0.0015 1.09 1.07 1.00 N/A2 0.0374 0.0352 1.06 0.0013 0.0010 1.31 1.04 1.00 N/A3 N/A N/A N/A N/A N/A N/A 0.98 1.05 N/A

Stroke rats4

1st PET 0.0328 0.0442 0.74 0.0114 0.0014 8.08 0.70 2.63 N/A2nd PET 0.0294 0.0477 0.62 0.0014 0.0013 1.07 0.78 1.09 Pannecrosis

5 0.0262 0.0309 0.85 0.0009 0.0010 0.86 0.88 1.08 Early change6 0.0300 0.0338 0.89 0.0017 0.0016 1.10 0.79 1.03 Severe change7 0.0259 0.0343 0.75 0.0010 0.0013 0.82 0.69 1.03 Severe change

AR, side-to-side asymmetry ratio, radioactivity in right (affected cortex)/radioactivity in left (unaffected cortex); N/A, not applicable; ratio, right (affected side)/left (unaffected side) in K1 and Ki values; K1, Ki unit, mL plasma/min (mL tissue).a0–5 mins after injection (first 10 frames).b130–180 mins after injection (the last 10 frames).

18F-FMISO PET in rat MCA occlusionM Takasawa et al

683

Journal of Cerebral Blood Flow & Metabolism (2007) 27, 679–689

affected-side and unaffected-side TACs of first PETsession.

Histopathology revealed tissue infarction withclear pannecrosis (Figures 2F and 2G).

Temporary Middle Cerebral Artery Occlusion

The quantitative values are shown in Table 2 and theTACs and PET images are given in Figures 4 and 5.

0

100

200

300

400

500

0 30 60 90 120 150 180

Late PhaseEarly Phase

Late PhaseEarly Phase

A

C

E

first PET

first PET

second PET

second PET

Ischemic Core Unaffected Side

CV × 2

CV × 40

Affected Cortex

Unaffected Cortex

Affected Cortex

Unaffected Cortex

G

F

D

B

Rad

ioac

tivity

Con

cent

ratio

n (k

Bq/

ml)

0

100

200

300

400

500

Rad

ioac

tivity

Con

cent

ratio

n (k

Bq/

ml)

Time Post-injection (min)

0 30 60 90 120 150 180

Time Post-injection (min)

18F-FMISO PET in rat MCA occlusionM Takasawa et al

684

Journal of Cerebral Blood Flow & Metabolism (2007) 27, 679–689

In the 45 mins MCAo rat (#5), no significant tracerretention was detected in the affected MCA territory(Figures 4A and 4C). However, there was some traceraccumulation in the surgically damaged temporalismuscle (Figure 4A). The TACs (Figure 4B) showedslightly lower initial uptake in the affected ascompared with unaffected side (early AR = 0.88),but subsequently the tracer kinetics became vir-tually identical in both hemispheres (late AR = 1.08).Kinetic analysis showed slightly reduced K1 in theaffected cortex, but essentially symmetrical Ki

values (Table 2). Histopathology revealed no cleararea of infarction (Figure 4D), but early ischemicchanges such as irregular shape and triangulation ofneurons (Figure 4E).

The findings in the 120-minute MCAo rats areillustrated for Rat #6 in Figure 5. Again nosignificant 18F-FMISO brain trapping was shown(Figures 5A and 5C), but there was some tracerretention in the temporalis muscle (Figure 5A).The TACs were similar to those obtained for the

45-minutes MCAo rat (Figure 5B and Table 2; earlyAR = 0.79, late AR = 1.03). Histopathology (Figure5D) revealed severe ischemic changes with diffuseneural cell loss (Figure 5E). Rat #7 showed verysimilar results (Table 2).

The K1 and Ki values (Table 2) were essentiallysymmetrical in Rat #6 and slightly decreased in theaffected cortex in Rat #7.

Discussion

This is the first study to investigate the use of 18F-FMISO PET in vivo to assess the hypoxic brainduring and after MCAo in the small animal. Theresults show the expected trapping in the affectedcortex early after MCAo, that is, when hypoxia isprominent, and the lack of demonstrable trappingboth when ischemic necrosis has fully developed48 h after permanent MCAo, and when tissue isnot necrotic and has been reperfused after briefMCAo, that is, when hypoxia is not expected onpathophysiologic grounds.

Although a small sample of animals could beinvestigated in this study, these were complexexperiments and their results were clear. Further-more, this is the first study to report 18F-FMISOtissue kinetics after stroke either in human or theanimal, providing much-awaited information rela-tive to previous work where late ARs only have beenreported (Read et al, 1998, 2000; Markus et al, 2004;Saita et al, 2004). Finally, because of the slowkinetics of 18F-FMISO (see below), it was notpossible to assess hypoxic trapping during tempor-ary MCA occlusion.

Findings in Normal Rats

This is the first study to show the 18F-FMISOkinetics in blood and brain in the normal rat (Figure1). In the brain, the TACs peaked around 30 minsafter injection, followed by a gradual decline. Thesebrain kinetics are similar to, although faster than,normal human data, which show a gradual increaseuntil 60 to 90 mins after intravenous injection (Valket al, 1992; Bruehlmeier et al, 2004). In addition, weshow here that the radioactivity in plasma andwhole blood were very similar and their ratio didnot significantly change over time (within 2 s.d.)

Figure 2 Positron emission tomography data and histology in the permanent MCAo rat (Rat #4). (A) Early phase (sum of first 10frames, that is, 0 to 5 mins) and late phase (sum of last 10 frames, that is, 130 to 180 mins) for the first PET session (started39 mins after clip on), showing prominent late trapping of the tracer in the affected cortex (arrow), and some trapping as well in thesurgically injured temporalis muscle (arrowheads); (B) TACs for the affected and unaffected sides for the first PET session. (C) Earlyphase and late phase accumulated images for the second PET session (48 h later; same frames as first session), showingsymmetrical late retention; (D) TACs for the affected and unaffected sides for the second PET session. (E) Fusion images with MRItemplate (coronal images), also illustrating the affected cortex ROI on two slices (see Materials and Methods), and the surgicallyinjured temporalis muscle (arrowheads); (F) illustrative CV coronal sections at approximately the same levels as in (E); (G) CV at highmagnification (magnification shown in each panel).

0

100

400

200

300

500

15060300 90 120 180

Affected CortexUnaffected Cortex

Rad

ioac

tivity

Con

cent

ratio

n (k

Bq/

ml)

Time Post-injection (min)

Figure 3 This illustrates the results of the fitting procedure(continuous line) for the affected- and unaffected-side TACs forRat # 4, first PET session. As can be seen, the fit wassatisfactory for both TACs.

18F-FMISO PET in rat MCA occlusionM Takasawa et al

685

Journal of Cerebral Blood Flow & Metabolism (2007) 27, 679–689

(Figure 1B). These novel findings suggest thatthe distribution of 18F-FMISO is nearly equal inblood cells and plasma and is stable after traceradministration.

Findings in Permanent Middle Cerebral ArteryOcclusion

In the rat with permanent MCAo, striking changeswere obtained at the first PET session. There wasreduced initial uptake in the affected area (earlyAR = 0.70; K1 ratio = 0.74), consistent with theexpected reduced perfusion during occlusion. Thiswas followed by continuous tracer accumulation inthe affected area, quite distinct from the washoutobserved in the unaffected side and in normal rats,with an AR of 2.63 in the final 50 mins (Figure 2A).

Accordingly, kinetic modelling showed a Ki > 8times the mirror ROI value. These findings areclearly indicative of severe tissue hypoxia, which isentirely expected early after MCAo (Crockard et al,1976; Liu et al, 2004). They are also consistent with18F-FMISO human studies performed early afterstroke (Read et al, 1998; Markus et al, 2004), andwith the findings of Saita et al (2004), who reportedhigh ex vivo 18F-FMISO uptake 2 h after tracerinjection administered 0.5 and 1 h after onset of a2-h thread-up temporary MCAo. However, thepresent study is the first to document in the livingrat that this increased uptake during occlusionreflects abnormal dynamic trapping of 18F-FMISO.

In previous rat studies, the assessment of tracertrapping has been delayed at least 2 h (Saita et al,2004) because of the slow brain kinetics and longequilibration time of 18F-FMISO (Bruehlmeier et al,

0

100

200

300

400

500

0 30 60 90 120 150 180

Ischemic Core Unaffected Side

CV × 2

CV × 40

A

C

Late PhaseEarly Phase

Affected CortexUnaffected Cortex

E

D

B

Rad

ioac

tivity

Con

cent

ratio

n (k

Bq/

ml)

Time Post-injection (min)

Figure 4 Positron emission tomography data and histology in the 45-minute temporary MCAo rat (Rat #5). The same layout as inFigure 2. (A) Early phase (0 to 5 mins) and late phase (130 to 180 mins) PET images (scanning started 28 mins after clip removal),showing no prominent late retention of the tracer in the affected area (arrow). Some 18F-FMISO accumulation was seen only in thesurgically injured temporalis muscle (arrowheads); (B) TACs for the affected and unaffected sides. (C) Fusion images with MRItemplate (coronal images), also showing the surgically injured temporalis muscle (arrowheads); (D, E) Illustrative CV coronalsections at approximately the same levels as in (C) (D) and at high magnification (E) (magnification shown in each panel).

18F-FMISO PET in rat MCA occlusionM Takasawa et al

686

Journal of Cerebral Blood Flow & Metabolism (2007) 27, 679–689

2004). This in principle precludes the use of 18F-FMISO to assess tissue hypoxia during MCAo of lessthan 3 h duration (taking into account the inevitabletime needed to position the animal inside thescanner after clip placement, typically around30 mins). However, the TACs (Figure 2B) suggestthat it may be possible to assess 18F-FMISO trap-ping even before 2 h, which may have practicalimplications.

At the second PET session, performed 48 h afterictus, there was no evidence of 18F-FMISO trapping(Figure 2C). Despite mildly reduced perfusion, assuggested by decreased early uptake and K1, thetracer did adequately reach the affected tissue.Subsequently, the TACs in the affected and unaf-fected areas became superimposed until the end ofscanning, that is, there was no demonstrable tracertrapping in the affected area, as also supported by

the kinetic modelling. These novel findings fromkinetic 18F-FMISO data are entirely consistent withhuman reports (Read et al, 1998) showing normal-ization of 18F-FMISO uptake at follow-up PETperformed 6 to 11 days after stroke onset, while thelatest observed high 18F-FMISO uptake was 47.5 hafter onset (Markus et al, 2004). Histopathology ofthis rat obtained immediately after the second PETsession revealed pannecrosis in the affected MCAarea. Thus, complete infarction may induce a loss ofthe biochemical processes that underlie nitroimida-zole binding in hypoxic conditions, principallythe nitroreductase enzymes that reduce 18F-FMISOintracellularly (Nunn et al, 1995). An alternative, butnot mutually exclusive, hypothesis is that there isnormoxia in the necrotic tissue 2 days after stroke,because of wreckage of the respiratory chain pro-cesses with very little oxygen consumption.

0

100

200

300

400

500

0 30 60 90 120 150 180

Ischemic Core Unaffected Side

CV × 2

CV × 40

A

C

Late PhaseEarly Phase

Affected CortexUnaffected Cortex

E

D

B

Rad

ioac

tivity

Con

cent

ratio

n (k

Bq/

ml)

Time Post-injection (min)

Figure 5 Positron emission tomography data and histology in one 120-minute temporary MCAo rat (rat #6). The same layout as inFigure 2. (A) Early phase (0 to 5 mins) and late phase (130 to 180 mins) PET images (scanning started 39 mins after clip removal),showing no prominent late retention of the tracer in the affected area (arrow). Some 18F-FMISO accumulation was seen only in thesurgically injured temporalis muscle (arrowheads); (B) TACs for the affected and unaffected sides. (C) Fusion images with MRItemplate (coronal images), also showing the surgically injured temporalis muscle (arrowheads); (D, E) Illustrative CV coronalsections at approximately the same levels as in (C) (D) and at high magnification (E) (magnification shown in each panel).

18F-FMISO PET in rat MCA occlusionM Takasawa et al

687

Journal of Cerebral Blood Flow & Metabolism (2007) 27, 679–689

Findings in Temporary Middle CerebralArtery Occlusion

In this investigation, we administered 18F-FMISOwithin 40 mins after clip removal. There was nodemonstrable tracer trapping in the affected cortexin any subject, despite differences in the degree ofhistologic damage. In the 45-minutes MCAo rat,there was reduced initial uptake, suggestive ofmoderate hypoperfusion, consistent with [14C]io-doantipyrine autoradiography results in indepen-dent spontaneous hypertensive rats subjected to thesame model (JL Hughes et al, unpublished). Therewas also reduced initial uptake in the 120-minuteMCAo rats. In all rats, however, tracer concentrationlater became identical between the affected andunaffected sides. These results are consistent withearlier studies showing normal or increased tissuepO2 early after reperfusion (Crockard et al, 1976; Liuet al, 2004). In the Saita et al (2004) study, 18F-FMISO was administered intravenously at varioustimes after withdrawal of the occluding thread.These authors report persistent cortical 18F-FMISOretention up to 4 h after thread removal, and up to20 h in the striatum. To explain these unexpectedresults, the authors proposed persistent tissuehypoxia despite MCA recanalization, because of,for example, periinfarct edema, small vessel occlu-sion, or the ‘no-reflow’ phenomenon (del Zoppoand Mabuchi, 2003). Although CBF using, forexample, laser Doppler could not be implementedin the present study because of the PET set up,early 18F-FMISO kinetics consistently documentedadequate tracer delivery to the affected tissue, rulingout complete ‘no-reflow’ even after 120-minuteMCAo.

General Comments

The present study highlights two distinct mechan-isms leading to lack of 18F-FMISO trapping afterstroke, namely early reperfusion restoring tissueoxygenation, and subsequent tissue necrosis remov-ing the cellular processes responsible for reductionand binding of the tracer, regardless of the oxygena-tion level. Our findings have clinical relevance asthey probably explain why 50% of the patientsstudied with 18F-FMISO within 48 h of stroke onsetby Markus et al (2004) did not exhibit high traceruptake. It would obviously be of considerableinterest to measure tissue pO2 directly in conjunc-tion with 18F-FMISO PET to refine these notions.

Preliminary modelling of 18F-FMISO kinetics innormal and tumoric brain tissue has been reported(Bruehlmeier et al, 2004). However, the biochemicalprocesses underlying 18F-FMISO trapping are com-plex (Casciari et al, 1995; Nunn et al, 1995;Thorwarth et al, 2005), suggesting accurate quanti-fication of the 18F-FMISO trapping process in brainhypoxia might prove challenging. In this study, weused the irreversible two-tissue compartment model

as a first approach. Our preliminary results areencouraging. Firstly, consistently good fitting wasobtained (see Figure 3 for illustration), and sec-ondly, the data were consistent with the simpleasymmetry ratio method. There was, however,higher sensitivity of the modelling method to tracertrapping in hypoxic tissue, with an eight-foldincrease in Ki as compared with an AR of 2.63 inthe first session of the permanent MCAo rat; theresults both for K1 and Ki for the other stroke studieswere also consistent with the asymmetry ratios.However, these results should be considered pre-liminary pending development of a validated com-partment model.

Some 18F-FMISO accumulation was also seen inthe temporalis muscle early after MCAo (see Figures2A, 4A and 5A, arrowheads). This muscle istransected during the surgical exposure of theMCA and is therefore potentially hypoxic acutely.Accumulation of 18F-FMISO in abnormal muscle hasbeen reported in sarcoma of the extremities (Bentzenet al, 2003).

In conclusion, our results show trapping of 18F-FMISO in the stroke area only in the early phase ofMCAo, but not when tissue necrosis has developedor if early reperfusion has occurred. These findingsstrongly support the validity of 18F-FMISO as aspecific marker of the viable hypoxic brain/penumbraafter stroke.

References

Baron JC (2001a) Mapping the ischemic penumbra withpet—a new approach. Brain 124:2–4

Baron JC (2001b) Perfusion thresholds in human cerebralischemia—historical perspective and therapeuticimplications. Cerebrovasc Dis 11(Suppl 1):2–8

Baron JC, Warach S (2005) Imaging. Stroke 36:196–9Bentzen L, Keiding S, Nordsmark M, Falborg L, Hansen

SB, Keller J et al (2003) Tumor oxygenation assessed by18F-fluoromisonidazole PET and polarographic needleelectrodes in human soft tissue tumours. RadiotherOncol 67:339–44

Bruehlmeier M, Roelcke U, Schubiger PA, Ametamey SM(2004) Assessment of hypoxia and perfusion in humanbrain tumors using pet with 18F-fluoromisonodazoleand 15O-H2O. J Nucl Med 45:1851–9

Buchan AM, Xue D, Slivka A (1992) A new model oftemporary focal neocortical ischemia in the rat. Stroke23:273–9

Casciari JJ, Graham MM, Rasey JS (1995) A modellingapproach for quantifying tumor hypoxia with [F-18]fluoromisonidazole PET time–activity data. Med Phys22:1127–39

Crockard HA, Symon L, Branston NM, Juhasz J (1976)Changes in regional cortical tissue oxygen tensionand cerebral blood flow during temporary middlecerebral artery occlusion in baboons. J Nurol Sci 27:29–44

Dings J, Meixensberger J, Jager A, Roosen K (1998) Clinicalexperience with 118 brain tissue oxygen partialpressure catheter probes. Neurosurgery 43:1082–95

18F-FMISO PET in rat MCA occlusionM Takasawa et al

688

Journal of Cerebral Blood Flow & Metabolism (2007) 27, 679–689

del Zoppo GJ, Mabuchi T (2003) Cerebral microvesselresponses to focal ischemia. J Cereb Blood Flow Metab23:879–94

Hughes JL, Beech JS, Fryer TD, Smith R, Golovko O, CleijM et al (2005) Imaging post-ischaemic cellular changesusing 11C-flumazenil & microPET followingtemporarydistal MCA occlusion in the spontaneous hypertensiverat (SHR). J Cereb Blood Metab 25:S603

Kinahan PE, Rogers JG (1989) Analytic 3D image recon-struction using all detected events. IEEE Trans Nucl Sci36:964–8

Liu S, Shi H, Liu W, Furuichi T, Timmins GS, Liu KJ(2004) Interstitial pO2 in ischemic penumbra and coreare differentially affected followingtransient focal cere-bral ischemia in rats. J Cereb Blood Metab 24:343–9

Markus R, Reutens DC, Kazui S, Read S, Wright P, PearceDC et al (2004) Hypoxic tissue in ischaemic stroke—persistence and clinical consequences of spontaneoussurvival. Brain 127:1427–36

Muzic RFJ, Christian BT (2006) Evaluation of objectivefunctions for estimation of kinetic parameters. MedPhys 33:342–53

Nunn A, Linder K, Strauss HW (1995) Nitroimidazolesand imaging hypoxia. Eur J Nucl Med 22:265–80

Read SJ, Hirano T, Abbott D, Markus R, Sachinidis JI,Tochon-Danguy HJ et al (2000) The fate of hypoxic

tissue on 18F-fluoromisonidazole positotron emissiontomography after brain stroke. Ann Neurol 48:228–35

Read SJ, Hirano T, Abbott DF, Sachinidis JI, Tochon-Danguy HJ, Chan JG et al (1998) Identifying hypoxictissue after acute ischemic stroke using PET and18F-fluoromisonidazole. Neurology 51:1617–21

Saita K, Chen M, Spratt NJ, Porritt MJ, Liberatore GT, ReadSJ et al (2004) Imaging the ischemic penumbra with18F-fluoromisonidazole in a rat model of ischemicstroke. Stroke 35:975–80

Tai YC, Chatziioannou A, Siegel S, Young J, Newport D,Goble RN et al (2001) Performance evaluation of themicroPET P4—a PET system dedicated to animalimaging. Phys Med Biol 46:1845–62

Thorwarth D, Eschmann SM, Paulsen F, Markus A (2005)A kinetic model for dynamic [18F]-FMISO pet data toanalysis tumour hypoxia. Phys Med Biol 50:2209–24

Tureyen K, Vemuganti R, Sailor KA, Dempsey RJ (2004)Infarct volume quantification in mouse focal cerebralischemia—a comparison of triphenyltetrazolium chlo-ride and cresyl violet staining techniques. J NeurosciMethods 139:203–7

Valk PE, Mathis CA, Prados MD, Gilbert JC, Budinger TF(1992) Hypoxia in human gliomas—demonstration bypet with fluorine-18-fluoromisonidazole. J Nucl Med33:2133–7

18F-FMISO PET in rat MCA occlusionM Takasawa et al

689

Journal of Cerebral Blood Flow & Metabolism (2007) 27, 679–689