Imaging of a Clinically Relevant Stroke Model
Glucose Hypermetabolism Revisited
Background and Purpose—Ischemic stroke has been shown to cause hypermetabolism of glucose in the ischemic penumbra. Experimental and clinical data indicate that infarct-related systemic hyperglycemia is a potential therapeutic target in acute stroke. However, clinical studies aiming for glucose control in acute stroke have neither improved functional outcome nor reduced mortality. Thus, further studies on glucose metabolism in the ischemic brain are warranted.
Methods—We used a rat model of stroke that preserves collateral flow. The animals were analyzed by [2-18F]-2-fluoro-2-deoxy-d-glucose positron emission tomography or magnetic resonance imaging during 90-minute occlusion of the middle cerebral artery and during 60 minutes after reperfusion. Results were correlated to magnetic resonance imaging of cerebral blood flow, diffusion of water, lactate formation, and histological data on cell death and blood–brain barrier breakdown.
Results—We detected an increased [2-18F]-2-fluoro-2-deoxy-d-glucose uptake within ischemic regions succumbing to infarction and in the peri-infarct region. Magnetic resonance imaging revealed impairment of blood flow to ischemic levels in the infarct and a reduction of cerebral blood flow in the peri-infarct region. Magnetic resonance spectroscopy revealed lactate in the ischemic region and absence of lactate in the peri-infarct region. Immunohistochemical analyses revealed apoptosis and blood–brain barrier breakdown within the infarct.
Conclusions—The increased uptake of [2-18F]-2-fluoro-2-deoxy-d-glucose in cerebral ischemia most likely reflects hypermetabolism of glucose meeting increased energy needs of ischemic and hypoperfused brain tissue, and it occurs under both anaerobic and aerobic conditions measured by local lactate production. Infarct-related systemic hyperglycemia could serve to facilitate glucose supply to the ischemic brain. Glycemic control by insulin treatment could negatively influence this mechanism.
Increased local brain glucose metabolism in a rat model of middle cerebral artery (MCA) occlusion was described by Ginsberg et al1 in a seminal article from 1977. Studies using different experimental approaches have investigated the role of regional hypermetabolism of glucose in the pathophysiology after hypoperfusion of the brain.2–11 This phenomenon has been shown in positron emission tomographic (PET) studies of acute human ischemic stroke (IS) and of postasphyctic infants.12,13 Various explanations for the frequent observations of hyperglycemia in IS and the detrimental effects of hyperglycemia on brain tissue have been proposed.14 Higher levels of blood glucose have been shown to be predictive for a more severe stroke and an increased functional impairment in humans, and therefore hyperglycemia is increasingly being considered as a therapeutic target in IS.15 However, in a recent clinical trial in which insulin treatment was used in patients with acute IS, a poorer outcome in the normoglycemic control group was observed.16 A recent review by Cochrane concluded that there is no reliable evidence for glucose control in acute IS.17 Considering that the pathophysiological mechanisms for hyperglycemia in stroke remain largely unknown, further preclinical studies are warranted.
The study of glucose hypermetabolism in experimental stroke has been limited by the lack of clinically relevant stroke models. The frequently used filament MCA occlusion technique for inducing IS in rodents causes obstruction of collateral flow from MCA-adjacent vascular territories. In other words, the filament occludes the anterior cerebral artery and the posterior communicating artery, which in the rodent is the main contributor to the posterior cerebral artery. This results in infarctions of entire cortical and subcortical regions.18,19 Furthermore, models commonly used for IS provide reperfusion that is unpredictable, which makes imaging studies during reperfusion difficult. In the present study, we used a recently described model for inducing a small focal cortical infarction that preserves collateral flow,20 and analyzed the metabolism using a small animal PET scanner. This experimental setup made it possible to determine the distribution and temporal changes in the use of [2-18F]-2-fluoro-2-deoxy-d-glucose ([18F]FDG) in the same animal during occlusion and reperfusion. Using the same experimental protocol as for the small animal PET experiments, we performed high-field magnetic resonance imaging (MRI) to study changes in cerebral blood flow (CBF), diffusion of water (diffusion-weighted imaging), and lactate formation by MR spectroscopy. Further validation was performed with histological analyses of cell death and blood–brain barrier (BBB) breakdown.
This study allowed us to investigate the regional dynamics of [18F]FDG uptake as an indicator of glucose metabolism in relation to CBF, diffusion of water, and immunohistochemical outcome, in a clinically relevant rat model of focal stroke with preserved collateral circulation.
Materials and Methods
All experiments were conducted according to the regulations of the Karolinska Institutet and were approved by the local laboratory animal ethics committee. Male Sprague-Dawley rats (355–450 g; Scanbur, Sollentuna, Sweden) were subjected to microwire occlusion of the M2 branch of the MCA (M2CAO), as previously described (groups 1–6; n=25).20 Imaging studies and postmortem analyses were performed as shown in Table. Animals were anesthetized using 2% isoflurane (Virbac, Carros Cedex, France) blended with air–O2 (7:3) during surgery and imaging. In group 2, blood glucose levels were assessed immediately before and 60 minutes after retraction of the microwire by analyzing tail artery blood samples with a glucometer (HemoCue 210+/201RT; HemoCue, Ängelholm, Sweden).
Positron Emission Tomography
In vivo PET investigations were performed on a Focus 120 (CTI Concorde Microsystems, Knoxville, TN) small animal PET scanner. Animals in groups 1 and 2 were placed in the PET scanner with the brain in the field of view within 5 minutes after placement of the microwire in the MCA. [18F]FDG was administered via the tail vein (20–40 MBq, 500 μL). Data were collected continuously during 90 minutes from the time of injection. Next, the microwire was retracted, and a second injection of [18F]FDG was administered via the tail vein (20–40 MBq, 500 μL) followed by data collection every second during 60 minutes from the time of reperfusion. Animals in group 2 also underwent a follow-up MRI 24 hours after M2CAO. Animals in group 3 were placed in the PET scanner with the brain in the field of view 24 hours after 90 minutes of M2CAO.
A macroparameter analysis method was used in addition to time–activity curves of the standardized uptake values (SUV).21 A volume of interest (VOI) with a predetermined size was drawn over the innominate artery using a PET image summed during the first 5 minutes of data collection. The Patlak method gives the Ki value, the so-called influx constant as a quantification of the net uptake of the radiotracer. The VOIs were chosen using a dual-modality approach, where both a PET and an MRI image were obtained for some animals (n=6).
The MRI experiments were conducted using a horizontal 9.4-T magnet (Varian, Yarnton, United Kingdom). Within 5 minutes after placement of the microwire in the M2 branch of the MCA, animals in groups 4 and 5 were placed in the MRI scanner and were imaged using diffusion-weighted imaging, arterial spin labeling, and point-resolved spectroscopy. Diffusion tensor images were acquired using multislice 3-shot spin-echo echo-planar imaging sequence with repetition time (TR) 3 s, and echo time (TE) 25 ms; diffusion-sensitizing gradients were applied along 12 directions with 2 diffusion-sensitizing factors b=0 and 1000 s/mm2. Perfusion measurements were performed using 3-shot gradient echo-planar imaging with a TE of 10.5 ms and 14 slices of 1-mm thickness with no gap in between. 1H MRI spectra were acquired using point-resolved spectroscopy sequence from a VOI (2.5×1.7×2.5 mm3). Imaging data were collected during 90 minutes. After 90 minutes, the microwire was retracted followed by data collection for an additional 60 minutes.
The animals in groups 1 to 3 were euthanized at 24 hours after M2CAO, and animals in group 4 were euthanized 48 to 96 hours after M2CAO. The brains from animals in groups 2, 4, and 6 were removed and snap frozen. The animals in groups 2 and 6 were injected with fluorescein isothiocyanate (FITC)–dextran (4 kDa, Sigma-Aldrich; 500 mL of 50 mg/mL; group 6, n=3) and after 24 hours (group 2, n=4). The animals were euthanized 15 minutes after the injection. Animals in group 6 were euthanized 10 minutes after reperfusion, after the 90 minutes of M2CAO. Six coronal 2 mm cryosections were taken throughout the brain and immersed in a 2% solution of 2,3,5-triphenyltetrazolium chloride for photography.
To visualize apoptotic staining, the ApopTag Fluorescein or ApopTag Red Detection kits were used. After an overnight incubation, sections were incubated in blocking buffer containing antirabbit and antirat secondary antibodies (1/200, Jackson). All sections were counterstained with the nuclear marker Hoechst (1/5000) and mounted with polyvinyl alcohol/glycerol containing 2.5% DABCO (Sigma).
Fluorescence microscope images were acquired on a Vslide slide scanning microscope (MetaSystems, Alltlussheim, Germany). Whole microscope slides were scanned at ×2.5, and tissue was detected based on the Hoechst 33342 signal. Images were stitched to generate a large 4-channel fluorescence image of the entire specimen with microscopic resolution.
Image and Statistical Analysis
PET data were processed with small animal PET manager and evaluated using the Inveon Research Workplace (Siemens Healthcare, Erlangen, Germany) software. MRI data were processed with VnmrJ software (Agilent Technologies, Palo Alto, CA) and evaluated using ImageJ (National Institutes of Health, Bethesda, MD). T2-weighted MR images and PET images from animals in group 1 and 2 were coregistered and analyzed using the Inveon Research Workplace. Three sets of VOIs were defined for PET images. (1) In groups 1 and 2 (n=9), a VOI encompassing the brain regions with elevated [18F]FDG uptake including neighboring regions was manually traced by visual assessment. Next, thresholding was performed to select a final VOI including voxels with elevated [18F]FDG uptake. (2) In group 2 (n=6), 2 sets of VOIs were generated using coregistered images from MRI and PET. One VOI was generated from manually tracing regions of interest matching the final infarct. The other was generated by subtracting the infarct VOI from the VOI containing regions showing elevated [18F]FDG uptake.
For spectroscopy analyses in animals in group 4 (n=4), 1 voxel was centered in the area displaying restricted diffusion on the diffusion tensor images. In animals in group 5 (n=3), 2 separate voxels were centered in the area displaying restricted diffusion and in brain regions adjacent to the restricted diffusion.
For CBF analyses in animals in group 4 (n=4), we coregistered sequentially acquired arterial spin labeling images with the diffusion-weighted imaging image acquired at 60 minutes after reperfusion. One VOI was generated by manual tracing of the infarct region showing restricted diffusion. Another VOI was generated by manual tracing of the region showing visually assessed reduction of CBF outside of the infarct VOI.
The Wilcoxon matched-pairs signed-rank test was performed to assess significance levels in time–activity curves and Patlak compartmental analyses from PET, comparing both SUV and macroparameters in the infarct region to the corresponding region in the contralateral hemisphere after 90-minute M2CAO and after 60 minutes of reperfusion (GraphPad Prism, San Diego, CA). Any value of P<0.05 was considered significant.
One animal in group 1 did not show any infarct lesion detectable by 2,3,5-triphenyltetrazolium chloride staining. Two animals developed large infarcts covering >2/3 of the MCA territory during in-bore MRI. All 3 animals were excluded from the study. Thus, the failure rate of this method for inducing focal IS was 12% in the present study.
PET Scans of Acute Ischemia
Distribution and Dynamics of [18F]FDG Uptake
The mean blood glucose (mmol/L±SD) in the tail artery before insertion of the microcatheter was 13.0±3.69, and the mean blood glucose in the tail artery at cessation of imaging was 14.5±3.1 (group 2, n=4). The time interval between placement of the microwire tip in the distal MCA and injection of [18F]FDG in the tail vein was 8.2±2.4 minutes (groups 1 and 2, n=9).
First, we analyzed [18F]FDG-PET images summed >90 minutes of occlusion (n=9). By visual assessment, we detected a markedly elevated [18F]FDG uptake in all animals (n=9) in regions within the targeted MCA territory (Figure 1). Next, we coregistered [18F]FDG-PET images with T2-weighted MRI images from follow-up MRI (n=6). The regions showing elevated [18F]FDG uptake encompassed parietal and frontal brain regions including the ischemic region succumbing to infarction in the parietal cortex defined by an increased signal at T2-weighted MRI and a decrease in the apparent diffusion coefficient (ADC) at diffusion-weighted MRI 24 hours after occlusion (Figure 1). Furthermore, we found elevated [18F]FDG uptake in frontoparietal peri-infarct regions (Figure 1), showing normal T2 signal and diffusion in MRI acquired at 24 hours. The region showing an elevated [18F]FDG uptake was traced to generate a VOI (ELEVVOI) for kinetic and compartmental analysis. The infarct detected by MRI and the region with elevated [18F]FDG uptake outside the final infarct were traced to generate VOIs for kinetic analysis (ISCHEMICVOI and PENUMBRAVOI).
Analysis of Whole Region With Elevated [18F]FDG Uptake (ELEVVOI)
The time–activity curves from the VOI traced around the final infarct showed that the SUV for the ELEVVOI had a slower rate of increase, reaching maximum tissue radioactivity concentrations at later time points compared with contralateral hemispheres (mean±SD) 52.3±13.9 minutes and 13.3±12.5 minutes, respectively (Figure 2A). At the end of 90 minutes of PET registration during M2CAO, the SUV (mean±SD) for the ELEVVOI was significantly higher (P=0.0039) compared with the corresponding contralateral cortex 1.660±0.3615 and 1.264±0.2060, respectively (Figure 2A). At the end of 60 minutes of PET registration during reperfusion, the SUV (mean±SD) for the ELEVVOI was significantly higher (P=0.0117) compared with the corresponding contralateral cortex 1.288±0.3040 and 1.154±0.2799, respectively (Figure 2B). Furthermore, Patlak compartmental analysis of ELEVVOI revealed a significantly increased (P=0.0039) net flux of [18F]FDG to the intracellular compartment during occlusion; Ki (mean±SEM) for the metabolic lesion was 0.012±0.00067 versus 0.0063±0.00085 for corresponding contralateral cortex (Figure 2C). Our findings indicate that the slower inflow of [18F]FDG during occlusion causes a slighter slope of the time–activity curves, although the collateral flow is sufficient to provide [18F]FDG. This results in an elevated [18F]FDG uptake at later time points suggestive of accelerated glycolysis. The smaller differences, although statistically significant, detected during reperfusion may reflect cessation of glycolysis because of infarction.
Analysis of Increased [18F]FDG Uptake in Regions Undergoing Infarction (ISCHEMICVOI) and Outside Infarction (PENUMBRAVOI)
The SUV (mean±SD) for the ISCHEMICVOI was significantly higher (P=0.0313) at the end of occlusion compared with the corresponding contralateral cortex, 1.506±0.3057 and 1.119±0.1290, respectively (Figure 3A). At the end of reperfusion, the SUV (mean±SD) for the ISCHEMICVOI was higher although not statistically significant (P=0.0625) compared with the corresponding contralateral cortex 1.698±0.5431 and 1.455±0.3800, respectively (Figure 3B).
The SUV (mean±SD) for the PENUMBRAVOI was significantly higher (P=0.0313) at the end of occlusion compared with the corresponding contralateral cortex 1.319±0.2899 and 1.130±0.1668, respectively (Figure 3C). At the end of reperfusion, the SUV (mean±SD) for the PENUMBRAVOI was higher although not statistically significant (P=0.0625) compared with the corresponding contralateral cortex 1.364±0.5181 and 1.259±0.5018, respectively (Figure 3D). [18F]FDG metabolism was thus markedly increased both in the ischemic tissue gradually undergoing infarction (Figure 3A) and in the surrounding penumbra (Figure 3D).
One group of animals was imaged with [18F]FDG at 24 hours after M2CAO (n=3). In all animals, a glucose hypometabolic region was detected in the targeted hemisphere. This region corresponded to the infarct lesions verified by 2,3,5-triphenyltetrazolium chloride staining (Figure I in the online-only Data Supplement).
MRI Scans of Acute Ischemia
To understand the hyperglycolytic response in stroke, we assessed factors contributing to infarct development by MRI during occlusion and reperfusion. Using the same time frame and brain ischemia protocol as for PET, we investigated changes in CBF by arterial spin labeling, restricted diffusion of water by diffusion weighted imaging, and formation of lactate by MR spectroscopy.
All animals in group 4 (n=6) showed cortical focal infarcts defined by a decrease in the ADC reflecting flux of water to the intracellular compartment (Figure 4B). A decrease in ADC was detected at the first imaging time point 10 minutes after M2CAO. The mean volume (mm3±SD) of the ADC lesion at 10 and 150 minutes after M2CAO was 14.8±17.0 and 25.8±22.8, respectively. Thus, in this model, preserved collateral flow results in gradual increase of infarction size in the acute phase.
Cerebral Blood Flow
We measured CBF during M2 occlusion and reperfusion to determine the impact of the stroke model on perfusion (group 4, n=4). The mean blood flow ratio±SD between the VOI containing ischemic tissue undergoing infarction and a corresponding VOI in the contralateral cortex during occlusion was 0.054±0.053 (30 minutes), 0.052±0.12 (60 minutes), and 0.056±0.034 (90 minutes; Figure 4D). In the VOI corresponding to peri-infarct regions showing hypoperfusion, the ratio of blood flow±SD during reperfusion was 0.14±0.056 (30 minutes), 0.18±0.12 (60 minutes), and 0.22±0.053 (90 minutes; Figure 4D). The differences in blood flow between the ischemic VOI and the peri-infarct VOI were statistically significant at 90 minutes of M2 occlusion (P=0.024; Figure 4D). After reperfusion, blood flow in the ischemic VOI and peri-infarct VOI increased to 0.60±0.33 (P=0.046) and 0.65±0.35 (P=0.0066), respectively (Figure 4D). The differences in blood flow at 30 and 60 minutes after M2 occlusion were not significant (Figure 4D).
The temporal evolution of lactate was measured in a VOI placed in the center of the ADC lesion (group 4, n=4; Figure 4A and 4B). Lactate levels were first elevated at 30 minutes after M2 occlusion and remained unchanged throughout occlusion and reperfusion (Figure 4C). In a separate group of animals (n=3), the peri-infarct region was investigated by lactate MR spectroscopy. No lactate was detected within the peri-infarct region during 90 minutes of occlusion, whereas lactate was detected consistently within the ischemic region succumbing to infarction in the same animals (Figure 4A and 4B). This data, in combination with [18F]FDG PET data, show that increased glycolysis occurs under aerobic and anaerobic conditions.
Cortical regions from animals in groups 1 and 4 (n=6 and 6, respectively) were analyzed by immunohistochemistry to assess cellular DNA fragmentation and BBB injury. Stainings were compared with the lesion on diffusion-weighted imaging in the corresponding MRI images. Apoptosis defined by terminal deoxynucleotidyl transferase dUTP nick end labeling stainings and morphological criteria corresponded directly to the area of restricted diffusion. DNA-fragmentated cells in general, detected by terminal deoxynucleotidyl transferase dUTP nick end labeling stainings, were detected within the lesion defined by T2 MRI (Figure 5A). No terminal deoxynucleotidyl transferase dUTP nick end labeling–positive cells were detected outside of the T2 lesion (Figure 5B). Similarly, immunoreactivity for Rat IgG was confined to the T2 lesion (n=6; 24 hours; Figure 5C). Two subsets of animals were intravenously injected with FITC–dextran (4 kDa) either 10 minutes (n=3) or 24 hours (n=4) postreperfusion after M2CAO. In the 10-minute group, FITC–dextran extravasation was strictly confined to the infarct lesion. No FITC–dextran extravasation was detected in the 24-hour group (data not shown).
Hypermetabolism of glucose in the ischemic brain has previously been detected by autoradiography and by [18F]FDG-PET.1,3,6–9,11,22 The increase in glucose metabolism has been reported to occur in peri-infarct brain regions in different models of IS in the rat.7,9,11,22 In contrast to those studies, we used a stroke model in which collateral circulation is preserved to a larger extent. We performed PET/MRI at early time points, and we observed increased [18F]FDG uptake in ischemic brain regions that undergo infarction as well as in brain regions adjacent to the final infarction volume. These findings are important because understanding the glucose metabolism pathophysiology in the context of IS is fundamental for designing optimal treatment regimens.
In this study, the experimental protocol includes PET imaging at earlier time points after vessel occlusion, ie, the tissue that later succumbs to infarction has not yet died at the time of imaging. Furthermore, we apply an occlusion model20 for focal cortical ischemia that does not impede collateral flow to the same extent as other models. Using MRI, we found significant blood flow differences between the VOI containing the final infarct and the VOI containing the region showing hypoperfusion outside of the ischemic region undergoing infarction, suggesting the presence of an ischemic region and a zone with reduced CBF supported by collateral flow. This would better simulate the situation in human stroke and provide a larger and more differentiated penumbral zone with a greater potential to increase glucose metabolism. Because this occlusion model preserves collateral flow to a larger extent compared with other models of stroke, the increased [18F]FDG uptakes in tissue that later undergo infarction could demarcate a penumbral zone, as infarct progression is slower under collateral flow.
Whether a presumed glucose hypermetabolism occurs under aerobic or anaerobic conditions is a subject of controversy. Previous studies of the ischemic rim in rats under conventional MCA occlusion suggest anaerobic as well as aerobic glycolysis.6,11 In the present study, we used MR spectroscopy to assess lactate formation in the ischemic region succumbing to infarction and in regions adjacent to the final infarct. We detected lactate in the ischemic region in all animals but not in brain regions directly adjacent to the final infarct in any animal. Considering that [18F]FDG uptake was increased in both the ischemic and peri-infarct regions at the same time points, we conclude that these increases occur under both aerobic and anaerobic conditions. Although accelerated glycolysis is the most plausible mechanism for ischemia-related glucose hypermetabolism, other explanations have been proposed, as, for example, a change in affinity for [18F]FDG over glucose, ie, a changed lumped constant in ischemic and hypoperfused tissues.23
In this study, we show evidence for BBB disruption by the detection of rat IgG and terminal deoxynucleotidyl transferase dUTP nick end labeling–positive cells directly in regions directly matching the T2 MRI lesion. This corroborates the definition of ischemic and penumbral VOIs made using in vivo MR and PET. Furthermore, we detected intravenously administered FITC–dextran (4 kDa) within the infarct at 150 minutes after M2CAO but not at 24 hours, indicating BBB disruption within the infarcted tissue itself at 150 minutes after M2CAO with subsequent repair some time before 24 hours after reperfusion. A disruption of the BBB could theoretically change [18F]FDG kinetics by increased diffusion over the BBB. However, this should equally affect the forward and reverse capillary transport of [18F]FDG. In several studies, 2-DG localization has been shown to be independent of BBB disruption in tumors, indicating that the net influx of 2-DG is not increased because the reverse capillary transport is equally increased.24,25 Our data support these observations.
We show that focal cortical stroke produces a significantly increased uptake of [18F]FDG during occlusion in ischemic cortical regions succumbing to infarction and in cortical regions adjacent to the final infarct displaying reduction of CBF. We show that lactate formation is detectable within brain regions featuring increased [18F]FDG uptake, misery perfusion, and restricted diffusion of water, although not in regions displaying reduced CBF, demonstrating that increased glycolysis occurs under both aerobic and anaerobic conditions. These findings, in this model of focal ischemia with preserved collateral circulation, strongly suggest that hypermetabolism of glucose occurs in the ischemic penumbra. Hypermetabolism of glucose in this situation could well be a reactive response to increased energy demands, thereby aiming for limitation of the lesion development. This knowledge can be important for the interpretation and design of clinical studies aiming for glucose control in the setting of acute IS.
Sources of Funding
This study was supported by the Swedish Society of Medicine, Söderbergska Stiftelsen, Uppdrag Besegra Stroke (supported by the Swedish Heart-Lung Foundation, Karolinska Institutet, Friends of Karolinska Institutet US and the Swedish order of St John), Åhlén-stiftelsen, Thurings stiftelse, Tore Nilsons stiftelse, and KERIC.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.114.008407/-/DC1.
- Received December 8, 2014.
- Revision received December 8, 2014.
- Accepted December 31, 2014.
- © 2015 American Heart Association, Inc.
- Sako K,
- Kobatake K,
- Yamamoto YL,
- Diksic M
- Welsh FA,
- Greenberg JH,
- Jones SC,
- Ginsberg MD,
- Reivich M
- Choki J,
- Greenberg J,
- Sclarsky D,
- Reivich M
- Belayev L,
- Zhao W,
- Busto R,
- Ginsberg MD
- Carmichael ST,
- Tatsukawa K,
- Katsman D,
- Tsuyuguchi N,
- Kornblum HI
- Yuan H,
- Frank JE,
- Hong Y,
- An H,
- Eldeniz C,
- Nie J,
- et al
- Rosso C,
- Corvol JC,
- Pires C,
- Crozier S,
- Attal Y,
- Jacqueminet S,
- et al
- Carmichael ST
- Ginsberg MD,
- Busto R
- Arnberg F,
- Lundberg J,
- Söderman M,
- Damberg P,
- Holmin S
- Nakai H,
- Yamamoto YL,
- Diksic M,
- Worsley KJ,
- Takara E
- Hawkins RA,
- Phelps ME,
- Huang SC