Journal of Neuroscience Methods 175 (2008) 25–35

Contents lists available at ScienceDirect

Journal of Neuroscience Methods

journa l homepage: www.e lsev ier .com/ locate / jneumeth

Novel analysis for improved validity in semi-quantitative 2-deoxyglucoseautoradiographic imaging

Neil Dawsona,∗, Linda Ferringtonb, Henry J. Olverman, Paul A.T. Kellya Centre for Cognitive and Neural Systems, University of Edinburgh, 1 George Square, Edinburgh EH8 9JZ, United Kingdomb Centre for Neuroscience Research, University of Edinburgh, 1 George Square, Edinburgh EH8 9JZ, United Kingdom

a r t i c l e i n f o

Article history:Received 16 May 2008Received in revised form 24 July 2008Accepted 29 July 2008

Keywords:2-DeoxyglucoseBrain metabolism

a b s t r a c t

The original [14C]-2-deoxyglucose autoradiographic imaging technique allows for the quantitative deter-mination of local cerebral glucose utilisation (LCMRglu) [Sokoloff L, Reivich, M, Kennedy C, DesrosiersM, Patlak C, Pettigrew K, et al. The 2-deoxyglucose-C-14 method for measurement of local cerebralglucose utilisation—theory, procedure and normal values in conscious and anestherized albino rats.J Neurochem 1977;28:897–916]. The range of applications to which the quantitative method can bereadily applied is limited, however, by the requirement for the intermittent measurement of arterialradiotracer and glucose concentrations throughout the experiment, via intravascular cannulation. Somestudies have applied a modified, semi-quantitative approach to estimate LCMRglu while circumventingthe requirement for intravascular cannulation [Kelly S, Bieneman A, Uney J, McCulloch J. Cerebral glucoseutilization in transgenic mice over-expressing heat shock protein 70 is altered by dizocilpine. Eur J Neu-rosci 2002;15(6):945–52; Jordan GR, McCulloch J, Shahid M, Hill DR, Henry B, Horsburgh K. Regionallyselective and dose-dependent effects of the ampakines Org 26576 and Org 24448 on local cerebral glu-cose utilisation in the mouse as assessed by C-14-2-deoxyglucose autoradiography. Neuropharmacology2005;49(2):254–64]. In this method only a terminal blood sample is collected for the determinationof plasma [14C] and [glucose] and the rate of LCMRglu in each brain region of interest (RoI) is esti-mated by comparing the [14C] concentration in each region relative to a selected control region, whichis proposed to demonstrate metabolic stability between the experimental groups. Here we show thatthe semi-quantitative method has reduced validity in the measurement of LCMRglu as compared to thequantitative method and that the validity of this technique is further compromised by the inability ofthe methods applied within the analysis to appropriately determine metabolic stability in the selectedstandard region. To address these issues we have developed a novel form of analysis that provides an

index of LCMRglu (iLCMRglu) for application when using the semi-quantitative approach. Provided thatthe methodological constraints inherent in 2-deoxyglucose autoradiography (e.g. normoglycaemia) aremet this analytical technique both increases the validity of LCMRglu estimation by the semi-quantitative

r its b

1

pmebtbT

us(

co

0d

method and also allows fo

. Introduction

Under normal conditions the adult brain uses glucose as therincipal substrate for the production of high energy phosphateolecules with which to fuel the many biochemical reactions nec-

ssary for cerebral function, and a direct link has been established

etween functional activation and glucose catabolism in neuralissues (Sokoloff, 1981) to the extent that the glucose consumedy a single action potential can be determined (Sokoloff, 1999).he intimate relationship between cerebral function and glucose

∗ Corresponding author. Tel.: +44 131 651 1903; fax: +44 131 651 1835.E-mail address: N.Dawson@ed.ac.uk (N. Dawson).

tacwnctb

165-0270/$ – see front matter © 2008 Elsevier B.V. All rights reserved.oi:10.1016/j.jneumeth.2008.07.020

roader experimental application.© 2008 Elsevier B.V. All rights reserved.

tilization forms the basis from which the 2-deoxyglucose, and sub-equent fluoro-deoxyglucose, imaging techniques were developedSokoloff et al., 1977; Reivich et al., 1979).

2-Deoxy-d-glucose (2-DG) is a structural analogue of native glu-ose differing only in the substitution of a hydrogen atom in placef the hydroxyl group normally attached to the second carbon ofhe hexose ring. In the presence of tracer concentrations of thisnalogue, the systems of brain uptake and phosphorylation areommon to both glucose and 2-DG, but the isomerase enzyme

hich converts glucose-6-phosphate to fructose-6-phosphate doesot recognise the anomalous structure of 2-DG-6-phosphate whichonsequently accumulates in brain tissues. With a knowledge ofhe relevant rate constants for the processes which take blood-orne glucose and 2-DG to phosphorylation in the tissue and of

2 roscie

to2rgcqtatdeetgap

pstccnarnsvnetettsqtflftLtebtsrsapmtttgt

2mtsaq

Luattpotptsa

2

qLteWoocpaa

2

3f11grssi0obsttslrvi2iEt

ptp

6 N. Dawson et al. / Journal of Neu

he blood concentrations of both substances, then by applicationf the operational equation (Sokoloff et al., 1977) the rate at which-DG-6-P accumulates (typically radiolabelled with 14C) may beelated directly to the rate at which glucose itself passes through thelycolytic pathway. The capacity to calculate rates of cerebral glu-ose utilization (LCMRglu) in absolute terms, in conjunction withuantitative autoradiographic imaging of the radiolabelled tracero localise these to quite specific regions of the brain, has providedpotent experimental tool with which to investigate brain func-

ion in a variety of animal models. The operation equation for theetermination of LCMRglu was developed in and had been mostxtensively applied to rats. Amongst other things, the operationalquation takes account of the fact that at the time of killing, theotal tracer concentrations in brain tissue measured by autoradio-raphy comprises both residual, unphosphorylated 2-DG as wells the accumulated 2-DG-6-phosphate, the concentration of whichrovides an index of LCMRglu.

While the fully quantitative approach is undoubtedly veryowerful the rigors required by the operational equation imposeerious limitations upon the experimental design. In particular,he requirement to generate profiles of plasma glucose and 2-DGoncentrations using intravascular cannulation either precludesertain types of experiment, or at least makes them more tech-ically demanding to perform. This is particularly so in very smallnimals such as mice where the dangers of hypovolaemic stressesulting from repeated blood sampling are very real. In mice aumber of attempts have been made to circumvent these limits. Inome studies it is assumed that there is no significant inter-animalariation in plasma tracer and glucose levels (although these wereot actually measured) and LCMRglu in each brain region of inter-st (RoI) is estimated by the ratio of [14C] in each region relativeo a pre-selected reference region (Miyamoto et al., 2000; Quelvent al., 2004). In other, more rigorous, studies an approximation tohe complete plasma history profile of tracer and glucose concen-rations is gained from data derived from a single, terminal bloodample (Kelly et al., 2002; Jordan et al., 2005). In this latter, semi-uantitative approach the concentrations of tracer and glucose inhe terminal blood sample are expressed as a ratio ([14C]/[glucose])or each animal, and where this ratio can be identified as an “out-ier” from the overall group mean, the animal can be excluded fromurther analysis. This process of exclusion provides some validity tohe assumption of minimal inter-animal variability. Subsequently,CMRglu is estimated using autoradiographic imaging by derivinghe ratio of [14C] in each RoI relative to that of a selected refer-nce region, where the rate of LCMRglu is assumed to be constantoth within and between the experimental groups. The choice ofhis reference region is validated with confirmation that local tis-ue [14C] concentrations remain relatively constant in this brainegion between experimental groups, but this can only be con-idered where the plasma variables (particularly plasma glucose)re not significantly altered by the experimental manipulation. Inractice, metabolic stability in the standard region across experi-ental groups is confirmed empirically when it can be shown that

here is a linear relationship between tissue [14C] in this region andhe blood [14C]/[glucose] ratio for each animal, and that this rela-ionship remains constant between the control and experimentalroups. However, no statistical data is ever offered to substantiatehis.

Although data have been published using the semi-quantitative-DG technique outlined above, there have been no attempts

ade to either validate the inherent assumptions, or to confirm

he accuracy of the results. In this study we have examined theemi-quantitative approach in the same control and drug-treatednimals in which LCMRglu was measured using the original, fullyuantitative method. This allowed us not only to compare the

t(tat

nce Methods 175 (2008) 25–35

CMRglu data generated from each approach, but also alloweds to test the assumptions upon which the semi-quantitativepproach is based; in particular the extent to which a singleerminal blood sample provides an indication of the plasma his-ories of tracer and glucose levels (from which the integratedrecursor specific activity value is calculated in the quantitativeperational equation) and of the criterion used for selection ofhe standard area against which autoradiographic analysis is com-ared. Finally, we have devised a novel analytical method by whichhe quality of the semi-quantitative 2-DG data can be improvedo as to better match that generated from the fully quantitativepproach.

. Materials and methods

In order to assess the quality of data generated using the semi-uantitative 2-DG technique as currently published we measuredCMRglu using this method in parallel with the fully quantita-ive method in the same animals following acute injection ofither saline or 3,4-methylenedioxymethamphetamine (MDMA).e chose this experimental treatment because MDMA has previ-

usly been shown to produce a robust response, with different areasf the brain showing increased LCMRglu, decreased LCMRglu, or nohange at all (Wilkerson and London, 1989; Quate et al., 2004). Thisattern of response thus allowed us to identify both false-positivend false-negative data being generated by the semi-quantitativepproach.

.1. Experimental procedures

A total of 10 male Dark-Agouti rats, weighing between 230 and00 g, were prepared for the measurement of LCMRglu using theully quantitative protocol as originally published (Sokoloff et al.,977) and as previously detailed from this laboratory (Kelly et al.,995). Briefly, femoral arteries and veins were cannulated undereneral anaesthesia, the animals lightly restrained, and allowed toecover for two hours before any subsequent manipulation. Mea-urement of LCMRglu was initiated 15 min after the injection ofaline (0.5 ml i.p.; n = 5) or MDMA (15 mg/kg i.p.; n = 5) with anntravenous (i.v.) injection of [14C]-2-deoxyglucose (1.3 MBq/rat in.75 ml saline; Sigma–Aldrich, UK) administered at a constant ratever the first 30 s of the experiment. A series of 14 timed arteriallood samples were collected at pre-determined intervals over theubsequent 45 min. Samples were centrifuged in order to separatehe plasma, aliquots of which were removed for the determina-ion of plasma glucose (10 �l) and 14C concentrations (20 �l) byemi-automated glucose oxidase assay (Beckman Glucose Ana-yzer) and liquid scintillation analysis (Packard, Tricarb 2900TL),espectively. At 45 min, the animals were killed by rapid intra-enous injection of sodium pentobarbitone, and the brains weremmediately dissected out intact and rapidly frozen in pre-cooled-methylbutane (−45 ◦C). Frozen brains were mounted onto spec-

men holders with embedding medium (Shandon M-1; Thermolectron Corp., Pittsburgh, PA, USA) and stored at −70 ◦C until sec-ioned.

Frozen brains were serially sectioned (20 �m) in the coronallane. A series of three sections were retained from every 200 �m,haw mounted onto glass cover slips and rapidly dried on a hotlate (75 ◦C). Autoradiograms were prepared by applying these sec-

ions, together with a series of eight precalibrated [14C]-standards40–1069 nCi/g tissue equivalents; Amersham International, UK)o X-ray film (Kodak, SB-5) for 7 days. A computer based imagenalysis system (MCID/M5+) was used to determine the localissue [14C] concentration in each RoI from the optical density

N. Dawson et al. / Journal of Neuroscience Methods 175 (2008) 25–35 27

Fig. 1. The operational equation for quantitative [14C]-2-deoxyglucose autoradiography (adapted from Sokoloff et al., 1977). Ri represents local cerebral glucose utilization(LCMRglu). C∗

i(T) represents the total [14C] present in a single brain region of interest. C∗

p and Cp represent the arterial plasma concentrations of [14C]-2-DG and glucose,r ∗ ∗ ∗ of [14

r d consi lycolyo

oLf[1

2

meiaiesvaatlqprIt

itadbtcaio

df

2

taaofqf

miierapvoa(bfaofiob

espectively. k1, k2 and k3 represent the constants for carrier mediated transportepresents the time at which the experiment is terminated (∼45 min). In the lumpen the tissue, while ˚ represents the fraction of glucose which continues along the gf hexokinase for [14C]-2-DG and glucose, respectively.

f autoradiographic brain images relative to the [14C]-standards.CMRglu (�mol/100 g/min) in each brain region was calculatedrom the local [14C] concentration and the plasma history of14C] and [glucose] using the operational equation (Sokoloff et al.,977).

.2. Semi-quantitative analysis

The terminal [14C]/[glucose] ratio was determined for each ani-al from the 45-min blood samples collected at the end of each

xperiment. To investigate the validity of the assumption, inherentn the semi-quantitative approach, that this terminal ratio providesn accurate estimation of the integrated precursor specific activ-ty (included in the denominator of the quantitative operationalquation) we performed a correlation analysis of the relation-hip between the terminal [14C]/[glucose] ratio and the integralalue generated from the operational equation using the tracernd glucose concentrations measured from all 14 timed samples. Inddition correlation analysis was also used to investigate the rela-ionship between the terminal plasma [14C]/[glucose] ratio and theevel of tissue residual [14C]-2-DG (included in the numerator of theuantitative operational equation), also calculated from the com-lete plasma history profile, to investigate if the terminal plasmaatio provides an accurate reflection of residual [14C]-2-DG levels.n this correlation analysis we used data from all of the animals inhe study, irrespective of treatment.

Tissue [14C] concentrations were generated by quantitativemage analysis for each of the 59 RoI included in this study andhe criteria used to identify an appropriate “standard” region werepplied. Thus, two regions were selected (corpus callosum andorsal tegmental nucleus, DTN) in which the linear relationshipetween terminal [14C]/[glucose] ratios and tissue [14C] concen-

rations was unchanged between treatment groups. The selectionriterion was provided with added rigour by applying regressionnalysis to the data to ensure that there was no statistically signif-cant effect of treatment upon the regression lines. Thus, two setsf semi-quantitative LCMRglu data were generated from the two

tapb(

C]-2-DG from plasma to tissue, tissue to plasma and hexokinase, respectively. Ttant � represents the ratio of the distribution space between [14C]-2-DG to glucosetic pathway. V ∗

m, K∗m and Vm, Km represent the Michaelis–Menten kinetic constants

ifferent standard regions. Finally, the standard areas were testedor metabolic stability by fully quantitative analysis.

.3. Semi-quantitative index of LCMRglu (iLCMRglu) analysis

In the final part of the study, we generated an approxima-ion to the fully quantitative operational equation which may bepplied with the limited blood data ([14C]/[glucose] ratio) gener-ted from the semi-quantitative approach, and tested the accuracyf the LCMRglu data thus generated against that produced usingull plasma [14C] and [glucose] histories and the application of theuantitative operational equation. The operational equation for theully quantitative method is shown in Fig. 1.

The complexity of the fully quantitative operational equation isuch reduced by the fact that as the time following acute admin-

stration of [14C]-2-DG increases, the concentration of [14C]-2-DGn the plasma approaches 0. Thus, those factors in the operationalquation that contain the rate constants (k∗

1, k∗2 and k∗

3), the cor-ection for lag in the equilibration between tissue and plasmand the concentration of unphosphorylated [14C]-2-deoxyglucoseresent in the tissue, also approach 0. Indeed, at infinite time thesealues will be equal to 0, but even at 45 min following a pulsef [14C]-2-DG their contribution is small, and any errors associ-ted with their empirical determination are deemed insignificantSokoloff et al., 1977). Therefore, we propose that these factors maye removed from the equation when applying simplified analysisor use with the semi-quantitative methodology. As long as a suit-bly long experimental period is adopted, the errors associated withur simplified operational equation, in which these factors do notgure, should be small. The lumped constant in the quantitativeperational equation is proposed to remain the same between allrain regions and between all animals and so we have also chosen

o remove this factor from the equation in order to further simplifynalysis. Therefore, we suggest that LCMRglu may be most appro-riately estimated when using the semi-quantitative methodologyy comparing the total tissue concentration in each brain regionC∗

i (T)) to the terminal plasma [14C]-2-DG (C∗p(t)) and glucose (Cp(t))

2 roscie

coifatirretg

bliaoWotca1igttyd

to

R

wC

Cbtrpfm

bitgdattaitgtgs

t

taaar

watm

2

qpmllLaceadcttiimaMt

3

3

tRobsRpRaiefr

8 N. Dawson et al. / Journal of Neu

oncentrations. This suggestion underlies the first ratio applied inur novel operational equation. In addition, as under normal phys-ological conditions the values of the lumped and rate constantsor [14C]-2-DG uptake and metabolism are consistent betweennimals any variability in the plasma [14C]-2-DG and glucose his-ories between animals is the primary factor dictating variabilityn the level of [14C] measured in tissues where the metabolicate is constant. This means that adjustment may be made to theatio in accordance with the plasma [14C]-2-DG/[glucose] ratio ofach individual animal relative to both the other animals withinhe same experimental group and also between experimentalroups.

Although the accumulation of tracer within the tissues of therain is directly proportional to LCMRglu, it is also influenced to a

arge extent by the glycaemic status of the animal over the exper-mental period and thus in a hyperglycaemic animal, for example,larger proportion of the total tracer found in tissues at the pointf killing will be in the form of residual, unphosphorylated 2-DG.hilst this factor is accommodated within the numerator of the

perational equation, semi-quantitative approaches assume thathe contribution made by residual 2-DG to total tissue tracer con-entrations is negligible. Whilst the theoretical implications of thisssumption have been described previously (Kelly and McCulloch,983), in practice it may not be an important issue so long as theres little difference between the glycaemic status of animals within aroup and no systematic effect of treatment. However, any methodhat does take account of differences in residual 2-DG will reducehe variability of data, and thus increase the power of the anal-sis to detect real changes, and widen the range of experimentalesign.

In parallel to the fully quantitative analysis of LCMRglu, usinghe operational equation, we calculated a semi-quantitative indexf LCMRglu (iLCMRglu) using the following equation:

i =(

C∗i (T)

(C∗p(t)/Cp(t))individual

)×

((C∗

p(t)/Cp(t))individual

mean(C∗p(t)/Cp(t))group

)

×(

mean(C∗p(t)/Cp(t))

group

mean(C∗p(t)/Cp(t))control group

)

here Ri = index of local cerebral glucose utilization (iLCMRglu),∗i (T) = the total [14C] present in a single brain region of interest,∗p(t) = the concentrations of [14C]-2-DG present in the terminallood sample, Cp(t) = the concentrations of glucose present in theerminal blood sample, (C∗

p(t)/Cp(t))individual = the [14C]/[glucose]atio for an individual animal, mean(C∗

p(t)/Cp(t))group = the meanlasma ratio of the appropriate experimental (treatment) groupor the given individual animal, mean(C∗

p(t)/Cp(t))control group = theean plasma ratio of the chosen control group.C∗

i (T) represents the total [14C] concentration present in eachrain region as determined from the analysis of autoradiographic

mages. From the terminal blood sample, taken at 45 min afterracer injection, the plasma [14C]-2-DG concentration (C∗

p(t)) andlucose (Cp(t)) concentration for each individual animal can also beetermined. The terminal plasma data for each individual animalre then used in the form of a ratio, [14C]-2-DG concentration rela-ive to glucose concentration (C∗

p(t)/Cp(t)individual), as a reflection ofhe complete plasma history profile for these factors. In iLCMRglunalysis the initial ratio for each brain region is calculated by divid-ng the total concentration of [14C] present in each brain region by

his individual plasma ratio. The remaining two ratios in the iLCMR-lu equation are generated from the plasma ratios of the individual,he respective mean plasma ratio for that individuals experimentalroup (mean(C∗

p(t)/Cp(t))group), and the mean plasma ratio for theelected control group (mean(C∗

p(t)/Cp(t))control group; in the case of

rgts(

nce Methods 175 (2008) 25–35

his analysis the control group were the saline-treated animals).By introducing the ratios

((C∗

p(t)/Cp(t))individual

mean(C∗p(t)/Cp(t))group

)×

(mean(C∗

p(t)/Cp(t))group

mean((C∗p(t)/Cp(t))control group

)

he equation normalises for inter-animal variation in blood tracernd glucose concentrations within any single experimental group,nd normalises for inter-group variation with respect to control (fornalysis in which the animal belongs to the control group the finalatio in this equation equals 1).

Data generated using this equation were compared directlyith those from the fully quantitative approach to assess their

ccuracy, and with data from the semi-quantitative approacho assess whether this represented a qualitative improve-

ent.

.4. Statistical analysis

In assessing the validity of control region selection in semi-uantitative analysis the possibility of a significant difference inlasma data and tissue [14C] concentration between the experi-ental groups was analysed using Student’s t-test with acceptable

evels of significance set at p < 0.05. In linear regression and corre-ation analyses significance was also set at p < 0.05. The resultingCMRglu data from each form of analysis (mean ± S.E.M.) werenalysed using Student’s t-test with acceptable levels of signifi-ance set at p < 0.05. In keeping with previous studies of similarxperimental design anatomically discrete brain regions weressumed to represent variables that can be analysed indepen-ently within each measure (McCulloch et al., 1982) and noorrection was applied for multiple comparisons. If the assump-ion of inter-regional independence in the analysis is not valid,hen one would expect significant differences to be generatedn 4 of the 71 regions analysed by chance alone. The valid-ty of each analytical method was assessed by comparing the

agnitude, directionality, localisation and statistical outcome oflterations in indices of LCMRglu detected between saline andDMA-treated animals compared to those detected by the quan-

itative method.

. Results

.1. Quantitative determination of LCMRglu

Data generated using the fully quantitative approach showedhat acute MDMA treatment produced significant changes in LCM-glu when compared to saline-injected controls in 19 (Tables 1–3)f the 71 RoI included in the study (cf. four regions that mighte expected to arise by chance alone). Of the brain areas whereignificant drug effects were found, significant increases in LCM-glu (range 19–88%) were found in all bar one region, theosterior cingulate cortex, where a significant decrease in LCM-glu was detected (−19%). These effects were both qualitativelynd quantitatively consistent with previously reported MDMA-nduced effects on LCMRglu (Wilkerson and London, 1989; Quatet al., 2004). Of particular interest were the quantitative dataor two regions that met the criteria for use as a standardegion in semi-quantitative analysis (Fig. 2). Whilst LCMRglu was

elatively constant between the saline and MDMA treatmentroups in the dorsal tegmental nucleus (mean difference = −5%),here was a significant increase in LCMRglu in the corpus callo-um (+19%) of MDMA-treated animals relative to saline controlsTable 3).

N.D

awson

etal./JournalofN

euroscienceM

ethods175

(2008)25–35

29

Table 1LCMRglu measurements gained by quantitative, semi-quantitative and semi-quantitative index of LCMRglu (iLCMRglu) analysis in cortical and motor regions of MDMA-treated and control (saline-treated) Dark-Agouti rats

Quantitative analysis (�mol/100 g/min) Semi-quantitative analysis iLCMRglu

Saline MDMA % [14C] RoI/[14C] DTN [14C] RoI/[14C] corpus callosum Saline MDMA %

Saline MDMA % Saline MDMA %

Cortical areasAnterior cingulate 119 ± 3 116 ± 9 −2 1.12 ± 0.07 1.15 ± 0.06 3 3.68 ± 0.19 3.55 ± 0.09 −4 15.07 ± 0.41 14.7 ± 1.13 −2Frontal layer IV 95 ± 3 103 ± 8 9 0.97 ± 0.05 1.05 ± 0.04 9 3.23 ± 0.13 3.08 ± 0.17 −5 12.02 ± 2.22 13.1 ± 0.96 9Frontal layer VI 87 ± 1 86 ± 5 −1 0.91 ± 0.03 0.90 ± 0.02 −2 3.02 ± 0.07 2.62 ± 0.08 −13* 10.99 ± 0.51 10.87 ± 1.01 −1Somatosensory layer IV 114 ± 4 148 ± 7 30** 1.15 ± 0.03 1.44 ± 0.06 25** 3.79 ± 0.10 4.44 ± 0.12 17** 14.46 ± 0.51 18.77 ± 0.86 30**Somatosensory layer VI 83 ± 3 81 ± 7 −3 0.88 ± 0.01 0.84 ± 0.06 −5 2.91 ± 0.07 2.57 ± 0.06 −12 10.57 ± 0.20 10.29 ± 0.89 −3Posterior cingulate 117 ± 5 95 ± 8 −19* 1.18 ± 0.05 0.96 ± 0.07 −18* 3.88 ± 0.16 2.95 ± 0.06 −24** 14.87 ± 0.61 12.05 ± 0.92 −19*Parietal layer IV 94 ± 2 94 ± 8 1 0.95 ± 0.03 0.95 ± 0.06 0 3.18 ± 0.04 2.93 ± 0.07 −8* 11.86 ± 0.28 11.92 ± 0.91 1Parietal layer VI 83 ± 1 83 ± 8 0 0.86 ± 0.03 0.84 ± 0.04 −2 2.87 ± 0.09 2.67 ± 0.08 −7 10.50 ± 0.11 10.50 ± 0.96 0Piriform 54 ± 4 71 ± 1 32 0.61 ± 0.04 0.74 ± 0.04 22 2.01 ± 0.12 2.30 ± 0.13 14 6.80 ± 0.53 9.01 ± 0.97 32Entorhinal 72 ± 3 81 ± 8 12 0.77 ± 0.03 0.84 ± 0.06 8 2.55 ± 0.08 2.58 ± 0.14 1 9.16 ± 0.39 10.30 ± 1.01 12Auditory 133 ± 5 150 ± 11 13 1.32 ± 0.66 1.43 ± 0.10 11 4.35 ± 0.19 4.50 ± 0.22 4 16.82 ± 0.67 19.03 ± 1.34 13Occipital IV 109 ± 4 107 ± 10 −2 1.10 ± 0.04 1.07 ± 0.08 −3 3.65 ± 0.14 3.29 ± 0.10 −10 13.88 ± 0.54 13.62 ± 1.25 −2

Motor areasStriatum

Medial 93 ± 3 152 ± 10 64*** 0.96 ± 0.03 1.47 ± 0.09 54*** 3.16 ± 0.10 4.56 ± 0.21 44*** 11.77 ± 0.38 19.28 ± 1.24 64***Lateral 103 ± 4 129 ± 9 25* 1.05 ± 0.03 1.27 ± 0.10 21 3.48 ± 0.12 3.90 ± 0.11 12* 13.2 ± 0.49 16.39 ± 1.12 25*Ventrolateral thalamus 92 ± 4 111 ± 10 21 0.95 ± 0.03 1.11 ± 0.06 17 3.13 ± 0.09 3.45 ± 0.23 10 11.64 ± 0.50 14.13 ± 1.21 21Subthalamic nucleus 80 ± 3 95 ± 10 20 0.84 ± 0.03 0.96 ± 0.06 14 2.77 ± 0.06 2.97 ± 0.19 7 10.10 ± 0.43 12.08 ± 1.33 20Lateral habenula 109 ± 10 87 ± 6 −20 1.09 ± 0.06 0.89 ± 0.08 −18 3.61 ± 0.25 2.75 ± 0.15 −24* 13.80 ± 1.22 10.98 ± 0.79 −20

Substantia nigraPars reticulata 53 ± 3 96 ± 16 81* 0.60 ± 0.02 0.95 ± 0.10 59* 1.98 ± 0.05 2.96 ± 0.34 50* 6.70 ± 0.35 12.13 ± 2.06 81*Pars compacta 64 ± 2 95 ± 8 48** 0.70 ± 0.03 0.97 ± 0.09 37* 2.33 ± 0.06 2.98 ± 0.21 28* 8.17 ± 0.21 12.08 ± 0.98 48**

CerebellumParamedian lobule 52 ± 1 67 ± 5 29* 0.59 ± 0.02 0.71 ± 0.04 20* 1.96 ± 0.03 2.20 ± 0.06 12** 6.57 ± 0.07 8.50 ± 0.59 29*Copula pyramis 60 ± 2 96 ± 8 61** 0.66 ± 0.02 0.97 ± 0.06 47** 2.18 ± 0.05 2.99 ± 0.12 37*** 7.56 ± 0.20 12.20 ± 1.05 61**Vermis 85 ± 1 138 ± 9 62*** 0.89 ± 0.02 1.34 ± 0.05 50*** 2.94 ± 0.06 4.15 ± 0.14 41*** 10.81 ± 0.17 17.45 ± 1.09 62***Red nucleus 70 ± 2 85 ± 7 21 0.75 ± 0.02 0.87 ± 0.05 15 2.49 ± 0.04 2.68 ± 0.12 8 8.90 ± 0.30 10.73 ± 0.93 21Inferior olive 79 ± 1 103 ± 6 30** 0.83 ± 0.01 1.03 ± 0.06 24* 2.76 ± 0.03 3.19 ± 0.08 16** 10.00 ± 0.18 13.01 ± 0.78 30**

Semi-quantitative analysis was completed with the corpus callosum (CC) and dorsal tegmental nucleus (DTN) as the selected control region. Data shown as mean ± S.E.M. and % difference between MDMA-treated and controlanimals. * denotes p < 0.05, ** denotes p < 0.01 and *** denotes p < 0.001 significant difference from saline control (t-test).

30N

.Daw

sonet

al./JournalofNeuroscience

Methods

175(2008)

25–35

Table 2LCMRglu measurements gained by quantitative, semi-quantitative and semi-quantitative index of LCMRglu (iLCMRglu) analysis in limbic and hippocampal regions of MDMA-treated and control (saline-treated) Dark-Agoutirats

Quantitative analysis (�mol/100 g/min) Semi-quantitative analysis iLCMRglu

Saline MDMA % [14C] RoI/[14C] DTN [14C] RoI/[14C] corpus callosum Saline MDMA %

Saline MDMA % Saline MDMA %

Limbic areasNucleus accumbens 87 ± 3 118 ± 7 35** 0.91 ± 0.02 1.16 ± 0.05 28** 3.00 ± 0.09 3.62 ± 0.19 21* 11.06 ± 0.37 14.92 ± 0.86 35**Septal nucleus 61 ± 4 83 ± 5 35* 0.68 ± 0.04 0.85 ± 0.06 25* 2.25 ± 0.14 2.63 ± 0.06 17* 7.79 ± 0.49 10.48 ± 0.69 35*Anterior thalamus 105 ± 5 93 ± 7 −11 1.06 ± 0.04 0.95 ± 0.08 −11 3.53 ± 0.19 2.92 ± 0.17 −17* 13.29 ± 0.61 11.82 ± 0.88 −11Mediodorsal thalamus 102 ± 5 118 ± 9 16 1.04 ± 0.05 1.17 ± 0.06 12 3.44 ± 0.16 3.62 ± 0.14 5 12.96 ± 0.63 15.02 ± 1.16 16Medial amygdala 50 ± 5 68 ± 12 37 0.58 ± 0.06 0.71 ± 0.09 23 1.90 ± 0.15 2.16 ± 0.24 14 6.32 ± 0.58 8.63 ± 1.58 37Lateral amygdala 75 ± 2 91 ± 4 21** 0.80 ± 0.03 0.93 ± 0.06 17 2.64 ± 0.10 2.86 ± 0.08 9 9.46 ± 0.31 11.49 ± 0.49 21**Ventral tegmental area 45 ± 2 57 ± 4 28* 0.53 ± 0.01 0.61 ± 0.02 16* 1.75 ± 0.04 1.96 ± 0.08 12* 5.71 ± 0.26 7.28 ± 0.46 28*Mamillary body 100 ± 4 104 ± 8 3 1.02 ± 0.02 1.04 ± 0.07 2 3.39 ± 0.09 3.21 ± 0.13 −5 12.71 ± 0.49 13.14 ± 1.03 3

HippocampusMolecular layer 87 ± 8 94 ± 8 9 0.91 ± 0.08 0.96 ± 0.07 6 2.98 ± 0.23 2.96 ± 0.18 −1 11.04 ± 1.01 11.97 ± 0.97 9Dorsal subiculum 76 ± 5 84 ± 5 9 0.81 ± 0.05 0.87 ± 0.05 6 2.69 ± 0.17 2.68 ± 0.14 0 9.69 ± 0.59 10.60 ± 0.64 9Dentate PO 56 ± 2 60 ± 6 6 0.63 ± 0.02 0.64 ± 0.05 2 2.08 ± 0.08 1.98 ± 0.09 −5 7.10 ± 0.26 7.56 ± 0.77 6Dorsal CA1 58 ± 2 75 ± 6 30* 0.64 ± 0.02 0.78 ± 0.03 21** 2.13 ± 0.07 2.43 ± 0.13 14 7.31 ± 0.28 9.51 ± 0.82 30*CA2 56 ± 1 62 ± 4 11 0.63 ± 0.02 0.67 ± 0.04 6 2.09 ± 0.04 2.07 ± 0.07 −1 7.15 ± 0.16 7.91 ± 0.53 11Ventral CA1 65 ± 3 74 ± 4 15 0.7 ± 0.0.3 0.78 ± 0.05 10 2.32 ± 0.07 2.39 ± 0.07 3 8.19 ± 0.35 9.39 ± 0.52 15Ventral subiculum 64 ± 5 67 ± 6 5 0.70 ± 0.05 0.72 ± 0.05 2 2.32 ± 0.14 2.21 ± 0.09 −5 8.14 ± 0.62 8.55 ± 0.81 5CA3 67 ± 2 66 ± 6 −1 0.73 ± 0.01 0.71 ± 0.06 3 2.40 ± 0.04 2.17 ± 0.11 −10 8.50 ± 0.26 8.40 ± 0.75 −1

Semi-quantitative analysis was completed with the corpus callosum (CC) and dorsal tegmental nucleus (DTN) as the selected control region. Data shown as mean ± S.E.M. and % difference between MDMA-treated and controlanimals. * denotes p < 0.05 and ** denotes p < 0.01 significant difference from saline control (t-test).

N.D

awson

etal./JournalofN

euroscienceM

ethods175

(2008)25–35

31

Table 3LCMRglu measurements gained by quantitative, semi-quantitative and semi-quantitative index of LCMRglu (iLCMRglu) analysis in sensory and multi-modal brain regions of MDMA-treated and control (saline-treated)Dark-Agouti rats

Quantitative analysis (�mol/100 g/min) Semi-quantitative analysis iLCMRglu

Saline MDMA % [14C] RoI/[14C] DTN [14C] RoI/[14C] corpus callosum Saline MDMA %

Saline MDMA % Saline MDMA %

Sensory areasSuperior olive 113 ± 4 109 ± 8 −3 1.13 ± 0.02 1.09 ± 0.06 −4 3.74 ± 0.10 3.37 ± 0.13 −10 14.28 ± 0.54 13.83 ± 0.96 −3Trigeminal nucleus 61 ± 1 84 ± 8 39* 0.67 ± 0.02 0.85 ± 0.05 27** 2.22 ± 0.04 2.63 ± 0.10 19** 7.71 ± 0.18 10.71 ± 0.97 39*Lateral geniculate 92 ± 3 89 ± 10 −3 0.95 ± 0.02 0.90 ± 0.06 −5 3.14 ± 0.10 2.79 ± 0.15 −11 11.62 ± 0.44 11.31 ± 1.29 −3Superior colliculus

Superficial layers 109 ± 6 97 ± 10 −11 1.10 ± 0.05 0.98 ± 0.06 −11 3.64 ± 0.13 3.03 ± 0.16 −17* 13.88 ± 0.59 12.33 ± 1.27 −11Deep layers 86 ± 3 102 ± 9 19 0.90 ± 0.04 1.03 ± 0.07 15 2.96 ± 0.11 3.17 ± 0.15 7 10.88 ± 0.35 12.99 ± 1.17 19Inferior colliculus 142 ± 5 131 ± 10 −8 1.40 ± 0.05 1.23 ± 0.13 −13 4.65 ± 0.20 3.74 ± 0.27 −19* 18.06 ± 0.66 16.62 ± 1.33 −8

Non-specific areasNucleus gelatinosus 110 ± 3 117 ± 9 6 1.12 ± 0.05 1.16 ± 0.08 4 3.69 ± 0.14 3.59 ± 0.15 −3 13.98 ± 0.44 14.87 ± 1.17 6Cerebellar white 25 ± 1 30 ± 2 21* 0.33 ± 0.01 0.37 ± 0.03 13 1.08 ± 0.03 1.13 ± 0.04 5 3.17 ± 0.16 3.83 ± 0.22 21*Corpus callosum 22 ± 1 26 ± 2 19* 0.30 ± 0.01 0.33 ± 0.02 7 n/a n/a n/a 2.74 ± 0.10 3.27 ± 0.20 19*Nucleus reunines 83 ± 6 95 ± 12 15 0.87 ± 0.04 0.95 ± 0.06 10 2.87 ± 0.17 2.99 ± 0.33 4 10.53 ± 0.82 12.08 ± 1.49 15Medial habenula 78 ± 3 75 ± 4 −4 0.83 ± 0.04 0.78 ± 0.04 −6 2.74 ± 0.11 2.41 ± 0.06 −12* 9.89 ± 0.37 9.46 ± 0.50 −4Lateral habenula 109 ± 10 87 ± 6 −20 1.09 ± 0.06 0.89 ± 0.08 −18 3.61 ± 0.25 2.75 ± 0.15 −24* 13.80 ± 1.22 10.98 ± 0.79 −20Hypothalamus 54 ± 2 60 ± 4 13 0.61 ± 0.02 0.65 ± 0.03 7 2.01 ± 0.07 2.01 ± 0.04 0 6.80 ± 0.30 7.66 ± 0.53 13Internal capsule 26 ± 1 34 ± 3 31 0.34 ± 0.01 0.40 ± 0.03 18 1.11 ± 0.02 1.23 ± 0.05 10 3.30 ± 0.12 4.31 ± 0.43 31Pontine reticular formation 53 ± 1 62 ± 4 17 0.6 ± 0.02 0.67 ± 0.04 11 1.99 ± 0.05 2.07 ± 0.09 4 6.72 ± 0.16 7.84 ± 0.52 17Dorsal raphe 81 ± 1 85 ± 6 4 0.86 ± 0.02 0.87 ± 0.05 2 2.83 ± 0.06 2.69 ± 0.09 −5 10.32 ± 0.17 10.76 ± 0.80 4Median raphe 90 ± 2 89 ± 7 −1 0.93 ± 0.02 0.09 ± 0.05 −3 3.08 ± 0.06 2.79 ± 0.10 −9* 11.39 ± 0.22 11.26 ± 0.92 −1Dorsal tegmental nucleus 98 ± 3 93 ± 7 −5 n/a n/a n/a 3.31 ± 0.06 3.13 ± 0.23 −5 12.41 ± 0.43 11.80 ± 0.93 −5Nucleus ambiguus 62 ± 1 86 ± 5 40** 0.68 ± 0.02 0.89 ± 0.04 30** 2.26 ± 0.04 2.75 ± 0.11 22** 7.84 ± 0.13 10.96 ± 0.58 40**

Semi-quantitative analysis was completed with the corpus callosum (CC) and dorsal tegmental nucleus (DTN) as the selected control region. Data shown as mean ± S.E.M. and % difference between MDMA-treated and controlanimals. * denotes p < 0.05 and ** denotes p < 0.01 significant difference from saline control (t-test). n/a denotes measurements not available due to form of analysis being applied.

32 N. Dawson et al. / Journal of Neuroscience Methods 175 (2008) 25–35

F rence[ idencet

3

3

bticritsprttlpfidrl

3

rpsip[ar[ptFsg

3

Fa

ig. 2. Relationship between tissue [14C] (nCi/mg) concentration in the selected refe14C]/[glucose] ratio in saline and MDMA-treated Dark-Agouti rats. There was no evhe different experimental groups (linear regression, see Table 4).

.2. Semi-quantitative analysis

.2.1. Evaluation of terminal plasma ratiosThe plasma [14C]/[glucose] ratios generated from the terminal

lood sample of each animal correlated significantly with bothhe level of residual tissue [14C]-2-DG (r = 0.71, p = 0.02) and thentegrated precursor specific activity values (r = 0.78, p = 0.008)alculated for grey matter. In white matter, however, while the cor-elation between the terminal plasma [14C]/[glucose] ratio and thentegrated precursor activity was also significant (r = 0.76, p = 0.01),he correlation with residual tissue [14C]-2-DG failed to reachignificance (r = 0.56, p = 0.09) (Fig. 3). Overall, these correlationsrovides support for the assumption that the terminal plasmaatio provides an appropriate index of the complete plasma historyhroughout the experimental time period within each animal andhat this ratio also provides a reflection of residual tissue [14C]-2-DGevels at the experimental endpoint. However, the scatter of dataoints around the regression line does indicate inter-animal dif-

erences in the exact proportionality of the relationship. This maympact particularly in white matter where variability in the ratioata may have contributed to the failure to detect a significant cor-elation between the terminal plasma ratio and residual [14C]-2-DGevels.

rsqi

ig. 3. Correlation between the terminal plasma [14C]/[glucose] ratio (nCi/mg) and thectivity (IPSA; nCi·min/�mol), in both grey and white matter, as calculated from the full a

region: (a) dorsal tegmental nucleus, (b) corpus callosum, and the terminal plasmafor a significant difference in the relationship between these parameters between

.2.2. Evaluation of criteria for reference region selectionUsing the published criteria for the selection of a reference

egion for semi-quantitative image analysis both the DTN and cor-us callosum were identified as being apparently metabolicallytable between experimental groups. Thus, there was no signif-cant difference between groups in any of the terminal plasmaarameters and also no significant difference in the mean tissue

14C] concentration of these regions. With an objectivity not usu-lly applied in previously published work, application of linearegression analysis also confirmed that the relationship between14C] concentrations in these two regions, and the [14C]/[glucose]lasma ratios were not significantly different between experimen-al groups in terms of slope, intercept or elevation (Fig. 2, Table 4).urthermore, the [14C] concentration of these regions showed aignificant correlation with the plasma ratio in both experimentalroups.

.2.3. Evaluation of semi-quantitative estimation of LCMRglu

Although the choice of DTN as a metabolically stable reference

egion was validated by the fully quantitative approach, the data setubsequently generated did not match that produced by the fullyuantitative method. Whilst in 17 RoI this approach did accuratelydentify significant drug-induced changes that were equivalent in

residual tissue [14C]-2-DG concentration (nCi/g) and integrated precursor specificrterial plasma history of each animal using the quantitative operational equation.

N. Dawson et al. / Journal of Neuroscie

Table 4Control region and plasma parameters used in semi-quantitative analysis from salineand MDMA-treated Dark-Agouti rats

Saline MDMA

Dorsal tegmental nucleus [14C] (nCi/mg) 493.8 ± 63.1 316.8 ± 56.7Corpus callosum [14C] (nCi/mg) 146.5 ± 21.6 126.2 ± 23.5

Plasma variables[14C] concentration (nCi/ml) 93.7 ± 20.8 55.2 ± 5.5[Glucose] concentration (mg/ml) 12.2 ± 1.5 11.9 ± 1.4[14C]:[glucose] ratio (nCi/mg) 7.4 ± 1.1 4.8 ± 0.5

Linear regression analysisDorsal tegmental nucleus

Intercept 58.31 ± 92.32 22.66 ± 96.20Slope 55.24 ± 11.30 70.30 ± 21.77

Corpus callosumIntercept 18.22 ± 26.2 27.9 ± 22.4Slope 16.6 ± 3.2 17.2 ± 5.1

Correlation analysis

D

dtTmiqo

raiar−feafTgm

3

naaiaftdwe

4

Sib

biwicatm(rmantgeardaia

sbmdwaRieuoauiitmrwerolateiytfmawIbst

Dorsal tegmental nucleus R2 0.89* 0.78*Corpus callosum R2 0.89* 0.79*

ata presented as mean ± S.E.M. * indicates significant correlation (p < 0.05).

irectionality and magnitude to those detected using fully quanti-ative analysis, the data set did include three false-negative results.hus, in the lateral amygdala, lateral striatum and cerebellar whiteatter, where the fully quantitative approach found significant

ncreases in LCMRglu of 21%, 25%, and 21%, respectively, the semi-uantitative approach indicated that there was no significant effectf the drug (Tables 1–3).

When the corpus callosum was used as the selected referenceegion, the pattern of MDMA-induced changes in LCMRglu wasltered dramatically in magnitude, localisation, and directional-ty when compared to the fully quantitative method. From thisnalysis, 10 false-positive results were generated, all of which rep-esented significant drug-induced decreases in LCMRglu of up to24% (Tables 1–3). Moreover, this approach also produced three

alse-negative results, failing to detect significant MDMA-inducedffects in the hippocampal dorsal CA1 region, lateral amygdaland cerebellar white matter, where the fully quantitative approachound increases in LCMRglu of 30%, 21%, and 21%, respectively.hese results are not unexpected given the fact that the data areenerated using a reference area in the brain which is not trulyetabolically stable between treatment groups.

.3. Modified quantitative iLCMRglu analysis

Changes in iLCMRglu induced by MDMA and detected by ourovel approach were remarkably similar in magnitude, direction-lity and localisation to those generated from fully quantitativenalysis. Of particular note was the fact that both MDMA-inducedncreases (in 21 of the 71 RoI) and decreases (in 1 RoI) were detectedccurately by this form of analysis with neither false-negative noralse-positive results being generated. Moreover, the magnitude ofhe changes, increases ranging between range 20% and 81% and theecrease of −19%, were almost exactly the same as those foundith the fully quantitative approach (see Fig. 4 for representative

xamples).

. Discussion

Within the constraints of the model originally developed byokoloff et al. (1977) the 2-DG quantitative autoradiographic imag-ng method provides an accurate index of functional activity in therain under a variety of experimental conditions. However, a num-

dereT

nce Methods 175 (2008) 25–35 33

er of practical modifications have been introduced over the yearsn an attempt to widen the range of experimental paradigms in

hich the method might be applied (Meibach et al., 1980) but thesenvariably have the potential to introduce systematic errors asso-iated with the limitations of the assumptions upon which theyre based (Kelly and McCulloch, 1983). More recently, the poten-ial benefits of 2-DG functional brain imaging in transgenic mouse

odels have been recognised, and a protocol has been publishedJordan et al., 2005; Kelly et al., 2002) which tries to avoid theequirement to measure plasma tracer and glucose levels by inter-ittent sampling of arterial blood. This approach does represent

n advance upon previous, similar modifications in that it recog-ises the influence of blood tracer and glucose concentrations uponotal brain tissue tracer accumulations. However, the informationathered from a single terminal blood sample is used merely toxclude those animals that deviate from the experimental norm onpost hoc basis and to provide validation of metabolic stability in a

eference region of the brain that is used as an intra-animal stan-ard for subsequent image analysis. Given that these proceduresre designed to identify and eliminate outliers from the analysis, its surprising that no objective criteria are proposed (e.g. statisticalnalysis) upon which to base the decision-making process.

In the studies reported here, we have evaluated the validity ofemi-quantitative 2-DG analysis by comparing the data generatedy that approach against that produced from exactly the same ani-als using the fully quantitative methodology. The experimental

esign included an acute drug treatment (MDMA), the response tohich has been characterised previously by ourselves and others

nd has been found to induce both increases and decreases in LCM-glu in specific brain areas, as well as a large number of brain areas

n which the drug has no effect (Wilkerson and London, 1989; Quatet al., 2004). Whilst some imprecision might have been expectedsing the less rigorous semi-quantitative approach, a comparisonf the data revealed inaccuracies in the magnitude, localisation,nd directionality of drug-induced changes in the index of glucosese generated in this way, and both Type I and Type II errors were

dentified. Moreover, these inaccuracies could be exacerbated by annappropriate choice of reference region for internal standardisa-ion of image analysis. Although both the DTN and corpus callosum

et the a priori criteria used to select an appropriate referenceegion, the fully quantitative analysis revealed that only the DTNas truly metabolically stable between treatment groups. How-

ver, despite the DTN being confirmed as an appropriate referenceegion, image analysis relative to this area generated a numberf false-negative results (Type II errors), suggesting at the veryeast a reduced statistical power in comparison to the quantitativepproach. Whilst in this instance it might be possible to increasehe accuracy and statistical power of the data set by the simplexpedient of increasing sample size, this is not a solution that willnvariably resolve the underlying issues. In a more extensive anal-sis (data not shown) we used a number of other brain areas ashe reference region for image analysis, on the basis that they bothulfilled the required selection criteria by the semi-quantitative

ethod and were also confirmed by fully quantitative analysiss being metabolically stable between treatment groups. Almostithout exception these analyses also generated a number of Type

errors, which are likely to be exacerbated rather than eliminatedy increasing sample size. In fact, out of all of these brain regions,emi-quantitative data generated from analysis performed usinghe DTN showed the greatest concordance with those alterations

etected by the quantitative approach. However, the problem isven more fundamental in that the selection criteria for referenceegions in semi-quantitative analysis are not sufficiently robust tonsure that the region in question is indeed metabolically stable.hus, we have demonstrated that it is possible to select a reference

34 N. Dawson et al. / Journal of Neuroscience Methods 175 (2008) 25–35

Fig. 4. LCMRglu responses to acute MDMA as detected by quantitative, semi-quantitative and semi-quantitative index of LCMRglu (iLCMRglu) analysis. Data shown as %d cancei neratt he sels resema

rbbonliLuiaruigl

mptatoslfssaemdwd

BeeatrtmeL

ittlttrtorqtsHlfkd

ifference from control. Representative data shown from regions in which the signifin which false-positive (Type II error) and false-negative (Type I error) results are geegmental nucleus and SQ:CC represents data analysed with the corpus callosum as taline control (t-test). Note how LCMRglu alterations detected by iLCMRglu analysisnd significance.

egion in which a significant alteration in LCMRglu is not detectedy the methods used in semi-quantitative analysis, as exemplifiedy the corpus callosum in this study. In this case the weaknessesf this analytical approach are further compounded, and a greaterumber of Type I and Type II errors are generated. The corpus cal-

osum was chosen in this analysis as it represented the brain regionn which the smallest magnitude of significant alteration (19%) inCMRglu was detected between the experimental groups whensing quantitative analysis. This clearly results in profound changes

n the LCMRglu alterations detected when using semi-quantitatives compared to quantitative analysis. However, when other brainegions, which met the same criteria as the corpus callosum, weresed in semi-quantitative analysis (data not shown) the alterations

n LCMRglu detected showed even greater divergence from thoseenerated by semi-quantitative analysis when using the corpus cal-osum.

Despite the obvious problems with the semi-quantitative 2-DGethodology identified above, the simplicity of the experimental

rotocol does remain attractive. Having identified where the poten-ial sources of inaccuracy lie, we have developed an alternativenalytical approach to the semi-quantitative 2-DG methodologyhat presents several advantages over the approaches used previ-usly, not the least of which is a level of accuracy that we havehown matches the fully quantitative approach. The close corre-ation between terminal plasma [14C]/[glucose] ratios generatedrom a single, terminal blood sample and the integrated precur-or specific activity (Fig. 3) derived from multiple, timed bloodamples in the same animals, indicates that these ratios representn accurate index of the denominator of the original operational

quation. This ratio is then normalised to both the intra-groupean and the inter-group mean as a way of accounting for any

ifferences in the amounts of residual (unphosphorylated) 2-DGhich, together with the tracer concentrations measured autora-iographically, make up the numerator of the operational equation.

aciIa

of LCMRglu responses to MDMA is unaltered by the different analyses and regionsed by semi-quantitative analysis. SQ:DTN represents data analysed with the dorsalected control region. *p < 0.05, **p < 0.01 and ***p < 0.001 significant difference fromble those determined by quantitative analysis in terms of localisation, directionality

y approximating as closely as possible to the original operationalquation and deriving an index of LCMRglu that is independent forach brain area, there is no need to use a reference brain regiongainst which to standardise regions of interest, thus avoidinghe potentially erroneous assumption of metabolic stability in thisegion across treatment groups. One further consequence of this ishat our novel approach can be used in a wider variety of experi-

ental conditions, including those in which treatments result inither alterations in plasma parameters or global alterations inCMRglu.

It is important to be aware that there are several assumptionsmplicit in our novel approach but while these are also true forhe fully quantitative approach, it is worth considering whetherhese assumptions might be rendered invalid by our novel ana-ytical method. A principal assumption of our novel approach ishat the relationship between the rate of [14C]-2-DG incorpora-ion into the cerebral tissue and the plasma [14C]/[glucose] historyemains consistent between experimental groups. This assump-ion provides the foundation for adjustment of the initial ratio inur analysis to accommodate individual animal and group plasmaatio variability, and directly parallels the assumption in the fullyuantitative approach that the rate constants for carrier-mediatedransport and phosphorylation of [14C]-2-DG, and the lumped con-tant, are consistent between the different experimental groups.owever, it is accepted that the certainty of this assumption is

ikely to diminish if the glycaemic state of animals within the dif-erent experimental groups is vastly different. Indeed, both theinetic and lumped constants have been shown to alter under con-itions of severe hyperglycaemia (Sokoloff et al., 1977; Schuier et

l., 1981). The requirement that all animals are in the normogly-aemic range for the fully quantitative approach to 2-deoxyglucosemaging is likely to be even more important for our novel approach.nfact, it is also possible that the semi-quantitative approach isctually more sensitive to minor violations of these assumptions

roscie

aavlp[ttaaMsc[ebsnpmgb

ia1safnmbtaa

A

DF

R

J

K

K

K

K

K

M

M

M

Q

Q

R

S

S

S

S

N. Dawson et al. / Journal of Neu

rising from altered 2-DG kinetics than is the fully quantitativepproach. In quantitative analysis the contribution of these kineticalues to the operational equation is minimised by the relativelyong experimental time, by the end of which only low levels oflasma [14C]-2-DG are present relative to the large peak of plasma

14C]-2-DG in the seconds after i.v. bolus injection. However, inhe semi-quantitative method [14C]-2-DG is often injected i.p. Inhis case the clearance of [14C]-2-DG from the plasma is prolongednd as a consequence higher residual levels of tracer may remaint the experimental end-point (Meibach et al., 1980; Kelly andcCulloch, 1983). This will impact upon the operational equation

uch that any error present in those factors containing the kineticonstants will be increased, and any error that results from altered14C]-2-DG kinetics upon the determination of LCMRglu will benhanced. Although some of the uncertainty can be eliminatedy ensuring that there are no inter-group differences in glycaemictate that could introduce systematic errors into the determi-ation of iLCMRglu, the absence of definitive, accurately timedlasma histories of glucose and tracer in the semi-quantitativeethodology runs the risk of missing any transient alterations in

lycaemic state that could alter the uptake of [14C]-2-DG into therain.

Although the data presented here are derived from one exper-mental manipulation we have applied this analytical approach tonumber of previously published data sets (Kelly and McCulloch,

982; McCulloch et al., 1982; Kelly et al., 1988) and have foundimilar degrees of concurrence between alterations in LCMRglus detected by our novel approach and those detected by theully quantitative method (data not shown). In conclusion, theovel analytical approach to the semi-quantitative [14C]-2-DGethod outlined here represents an increase in accuracy that

rings data very close to that generated by the fully quantita-ive technique. Any loss in accuracy from the fully quantitativepproach is offset by the wider experimental design that can beccommodated.

cknowledgements

This work was funded by European Commission Researchirectorates, Research & Technological Development Project, 6thramework Programme, LSHM-CT-2003-503474.

W

nce Methods 175 (2008) 25–35 35

eferences

ordan GR, McCulloch J, Shahid M, Hill DR, Henry B, Horsburgh K. Regionally selectiveand dose-dependent effects of the ampakines Org 26576 and Org 24448 on localcerebral glucose utilisation in the mouse as assessed by C-14-2-deoxyglucoseautoradiography. Neuropharmacology 2005;49(2):254–64.

elly S, Bieneman A, Uney J, McCulloch J. Cerebral glucose utilization in transgenicmice over-expressing heat shock protein 70 is altered by dizocilpine. Eur J Neu-rosci 2002;15(6):945–52.

elly PAT, Davis CJ, Goodwin GM. Differential patterns of local cerebral glu-cose utilisation in response to 5-hydroxytryptamine-1 agonists. Neuroscience1988;25:907–15.

elly PAT, McCulloch J. Effects of the putative GABAergic agonists, muscimoland THIP, upon local cerebral glucose utilisation. J Neurochem 1982;39:613–24.

elly PAT, McCulloch J. A potential error in modifications of the [C-14]-2-deoxyglucose technique. Brain Res 1983;260(1):172–7.

elly PAT, Ritchie IM, Arbuthnott GW. Inhibition of neuronal nitric oxide synthase by7-nitroindazole: effects upon local cerebral blood flow and glucose utilisationin the rat. J Cereb Blood Flow Metab 1995;15(5):766–73.

cCulloch J, Kelly PAT, Ford I. Effect of apomorphine on the relationship betweenlocal cerebral glucose utilisation and local cerebral blood flow (with an appendixon its statistical analysis). J Cereb Blood Flow Metab 1982;2(4):487–99.

eibach RC, Glick SD, Ross DA, Cox RD, Maavani S. Intraperitoneal administra-tion and other modifications of the 2-deoxy-d-glucose technique. Brain Res1980;195(1):167–76.

iyamoto S, Leipzig N, Lieberman JA, Duncan GE. Effects of ketamine, MK801 andamphetamine on regional brain 2-deoxyglucose uptake in freely moving mice.Neuropsychopharmacology 2000;22:400–12.

uate L, McBean D, Ritchie I, Olverman H, Kelly P. Acute methylenedioxymtham-phetamine administration: effects on local cerebral blood flow and glucoseutilisation in the Dark-Agouti rat. Psychopharmacology 2004;173:287–95.

uelven I, Roussin A, Zajac JM. Functional consequences of neuropeptide FFreceptor stimulation in mouse: a cerebral glucose uptake study. Neuroscience2004;126:441–9.

eivich M, Kuhl D, Wolf A, Greenberg J, Phelps M, Ido T, et al. The[18F]fluorodeoxyglucose method for the measurement of local cerebral glucoseutilization in man. Circ Res 1979;44:127–37.

chuier F, Orzi F, Suda S, Lucignani G, Kennedy C, Sokoloff L. Influence of plasmaglucose on the lumped constant of the deoxyglucose method—effects of hyper-glycaemia in the rat. J Cereb Blood Flow Metab 1981;10:765–73.

okoloff L. Localisation of functional activity in the central nervous system by mea-surement of glucose utilisation with radioactive deoxyglucose. J Cereb BloodFlow Metab 1981;1:7–36.

okoloff L. Energetics of functional activation in neural tissues. Neurochem Res1999;24:321–9.

okoloff L, Reivich M, Kennedy C, Desrosiers M, Patlak C, Pettigrew K, et al.

The 2-deoxyglucose-C-14 method for measurement of local cerebral glucoseutilisation—theory, procedure and normal values in conscious and anestherizedalbino rats. J Neurochem 1977;28:897–916.

ilkerson G, London ED. Effects of methylenedioxymethamphetamine on localcerebral glucose utilisation in the rat. Neuropharmacology 1989;28:1129–38.