Analysis of cerebral glucose metabolism following experimental subarachnoid hemorrhage: an [18F]FDG-PET study over 7 days using an MRI-template based analysis tool


 Little is known about changes in brain metabolism following SAH, possibly leading towards secondary brain damage. Despite sustained progress in the last decade, analysis of in vivo acquired data still remains challenging. The present interdisciplinary study uses a semi-automated data analysis tool analyzing imaging data independently from the administrated radiotracer. The uptake of 2-[18F]Fluoro-2-deoxy-glucose ([18F]FDG) was evaluated in different brain regions in 14 male Sprague-Dawley rats, randomized into two groups: (1) SAH induced by the endovascular filament model and (2) sham operated controls. Serial [18F]FDG-PET measurements were performed. Quantitative image analysis was performed by uptake ratio using a self-developed MRI-template based data analysis tool. SAH animals showed significantly higher [18F]FDG accumulation in gray matter, neocortex and olfactory system as compared to animals of the sham group, while white matter and basal forebrain region showed significant reduced tracer accumulation in SAH animals. All significant metabolic changes were visualized from 3 hours, over 24 hours, day 4 and day 7 following SAH/sham operation. This [18F]FDG-PET study provides important insights into glucose metabolism alterations following SAH - for the first time in different brain regions and up to day 7 during course of disease. The present tool improves PET image quantification and provides more flexible data analysis advocating its clinical application.


Introduction
Despite new discoveries in research and sustained advances in the treatment of aneurysmal subarachnoid hemorrhage (aSAH), morbidity and mortality remain high 1,2 . The world-wide incidence of aSAH is approximately 7.9 per 100,000 person-years with aSAH accounting for about 5% of all strokes 3,4 .
Overall mortality rate is estimated to be approximately around 40% 4,5 . Brain injury following aSAH is multimodal and occurs directly as early brain injury, and secondarily as delayed brain injury 6 . During the rst 2 weeks following aSAH, angiographic cerebral vasospasm (CVS) occurs in about 70% of SAH patients, but only 30% of patients develop delayed cerebral ischemia (DCI). Therefor DCI remain the major cause of morbidity and mortality among patients who survived the initial bleeding and treatment of the ruptured aneurysm 7,8 . Several mechanisms during the acute phase of SAH contribute to DCI and poor outcome. These include neuroin ammation, microthrombosis, cortical spreading depolarizations, disrupted blood-brain barrier (BBB) integrity, microvascular dysfunction and metabolic derangement 9-12 . Changes in brain metabolism and accumulation of metabolites such as lactate, pyruvate and glutamate are described, possibly responsible for a derangement of oxidative brain metabolism, leading to secondary brain injury and poor outcome 2,10,12 . Glucose is the main energy substrate of the brain and crucial for normal brain function. However, the effects of SAH on cerebral glucose metabolism are not fully understood yet. In this study we aimed to analyze how glucose metabolism in a well-established Page 3/15 SAH rat model is affected by SAH in vivo using [ 18 F]FDG-PET. Thereby, data evaluation and analysis was performed using a self-developed, semi-automated MRI-template based data analysis tool, as previously described in 13 . This interdisciplinary study provides data on metabolic processing of glucose in the early phase following SAH, where so far only limited data is available, and for the rst time measurements of [ 18 F]FDG brain metabolism over course of disease up to day 7 following SAH and as well for the rst time in separation of different brain regions. These gained insights on brain glucose metabolism should help to develop new targets for translational neuroprotective therapies following aSAH.

Material And Methods
Animals and ethical statements PET data of 14 male Sprague-Dawley rats (Harlan Winkelmann GmbH, Borchen, Germany) weighing 250-300 g (mean age 8 weeks) were analyzed. The animals were maintained for at least one week for acclimatization before induction of SAH, including a 12/12-hour light/dark cycle. Access to food and water was given all time, except that at least six hours before PET measurement, food was removed.
Induction of SAH was performed via the endovascular lament model as described previously 12,14 . In summary, a right paramedian longitudinal incision on the ventral neck was performed. The external carotid artery was ligated and exiting branches were coagulated. Afterwards, a temporary aneurysm clip was placed on the common and internal carotid arteries. The external carotid artery was incised about 6 mm distal from the carotid bifurcation and a 3.0 Prolene lament (Ethicon, Inc. Somerville, New Jersey) was inserted and secured with a silk ligature. Next, temporary clips were removed and the external carotid artery was cut. Thus, the lament could be moved intracranial via the internal carotid artery. The lament was then advanced 3 mm further perforating the vessel in the area of the anterior cerebral artery (ACA). Immediately, the lament was quickly pulled back into the external carotid artery ascertaining reperfusion of the internal carotid artery.
Induction of anesthesia was performed with 4% iso urane, followed by oral intubation and mechanical ventilation with an air-oxygen mixture to provide normal blood gases. After induction of anesthesia, iso urane was lowered to 2.5% for surgical procedures and to 1.5% from 30 minutes before induction of SAH until 30 minutes after induction of SAH. Thereafter, animals were woken up, returned to separate boxes for recovery and only got iso urane anesthesia during [ 18 F]FDG PET scans ( 4% iso urane for induction and 1-2% iso urane for maintenance, in oxygen at 2L/min). Temperature was constantly measured throughout the whole experiment, keeping the temperature level at 37°C. Continuous arterial blood pressure measurement was performed by cannulating the tail artery. The division of the animals into group (1) subarachnoid hemorrhage (SAH, n = 7) and (2) sham operated controls (sham, n = 7) was

Software programs
Reconstructed data was analyzed on a self-developed semi-automated nuclear medicine data processing analysis tool (NU_DPA) implemented in the software program Matlab (version 2018a, MathWorks, United States) 13 . In addition to pre-implemented functions in Matlab, the following functions, datasets as well as toolboxes were used for the evaluation of our data: Sprague Dawley T2* and atlas MRI data sets provided by Papp et al. [17][18][19] for anatomical sub-classi cation of the PET data, a tool for NIfTI data import by Shen et al. 20 , the Medical Image Registration Toolbox (MIRT) by Myronenko 21 for a ne registration processes, the VolumeViewer3D of the Medical Image Reader and Viewer Toolbox by Schaefferkoetter 22 for data visualization and Export_ g by Altman 23 for exporting high-resolution graphics.

Statistical data analysis
Statistical evaluation was executed using the semi-quantitative parameter uptake ratio (UR), according to the data analysis methods by Miederer et al. 24 . The UR was calculated from the ratio of the measured radioactivity in a subregion of the VOI to the measured radioactivity of the total VOI.

UR = measuredradioactivityinVOI Target measuredradioactivityinVOI Full
Due to the small number of animals (n = 7), no normal distribution was assumed and data was statistically evaluated using the non-parametric Mann-Whitney U-test. All signi cant effects found between sham and SAH group were observed over the full 20 min measurement time period. Bonferroni corrections were made for serial measurements and con rmed signi cance levels.

Results
According to 13 , the PET image data were aligned to a resized MRI T2* as well as MRI atlas image data that was adjusted to match the spatial resolution of the PET image data. Therefore, the MRI datasets were resized by a scale factor of 1/20 to match the resolution of the PET image data (  and 6 hours after induction of experimental SAH compared to a sham group, indicating an increase in anaerobic glycolysis and a global hyper-glycolysis to balance the mismatch of energy need and supply, already in the rst hours following SAH. In a primary analysis, we also evaluated imaging data by calculating the standardized uptake value (SUV), whereby similarly to Song et al. only signi cant effects were observed between the (1) experimental SAH and (2) sham operated groups 3 hours following experimental SAH, and only a trend in elevated [ 18 F]FDG uptake in the SAH group on day 1, day 4 and day 7 could be observed (data not shown). However, when comparing the ratio of measured radioactivity using the uptake ratio, we could observe further signi cant differences between both experimental groups and in different brain regions.
We assume that the strong differences in the statistical data analysis using SUV compared to UR are to be attributed to two aspects: First, SAH is a disease affecting the entire metabolism in the brain.
Therefore, the tracer accumulation of [ 18 F]FDG might be (severely) impaired. Second, the SAH disease resulted in higher variances due to the weight variation required to calculate SUV. The calculation of SUV refers to radiotracer concentration [MBq / cm 3 ≙ g] divided by the ratio of injected radioactivity [MBq] to body weight [g]). Due to these aspects, an evaluation by SUV might be unfavorable. Since the UR excludes the factors weight and injected radioactivity but only focuses on the enriched radiotracer in general, a data analysis by UR could be more suitable and speci c than one by SUV.
A crucial aspect of the evaluation according to UR was the adequate detection of the VOI (whole brain), as well as the exact subdivision into the individual subregions. By adding and aligning the data on a suitable MR data atlas, it was possible to achieve such a revaluation according to UR. The alignment of the acquired PET data on a PET data template created of the own PET sham datasets and then aligning them onto the MR atlas using the transformation matrices resulting from the two times a ne coregistration process, also increased the validity of the data. Due to the described characteristics of the self-made analysis tool we were able to distinguish metabolic changes in several brain regions.
In addition, this is the rst study that shows the quantitative [ 18  In general, they could show that the presence of whole brain ischemia and/or regional ischemia within the region of the MD probe was associated with increased levels of energyrelated metabolites (lactate, pyruvate, glucose, adenosine, inosine and hypoxanthine) and excitatory amino acids (EAAs) retrieved by MD 29 . When PET did not show any signs of ischemia or when signs of regional ischemia were found remote from the MD probe region, only occasionally increased levels of energy-related metabolites and EAAs were seen. They suggested that PET may be of use in de ning critical ischemic regions (tissue at risk) where the MD probe could be inserted for chemical monitoring.  11,31,32 . In previous studies, we found, that despite incomplete recovery of CBF for more than 6 hours after experimental SAH, tissue oxygenation recovered to baseline level 2 hours after SAH and signi cantly exceeded this level up to 140% and more of baseline level after 6 hours 33 . The incomplete recovery of CBF which exceeded the decline in tissue oxygenation (ptiO 2 ) seems to mirror the disability of O 2 -utilization of the brain. Taken together these ndings with our discovery of reduced pyruvate dehydrogenase enzyme (PDH) -key enzyme to the TCA cycle and oxidative phosphorylation -3 hours following experimental SAH in rats 12 , we assume that there is a switch from aerobic to anaerobic metabolism. This shift from aerobic to anaerobic metabolism might be due to maintain the energy supply in consideration of the metabolic disturbance starting already in the early phase after SAH (3 hours) and lasting at least up to day 7, re ected by the signi cantly increased [ 18 F]FDG uptake in gray matter, neocortex and olfactory system. The fact that the [ 18 F]FDG uptake is signi cantly reduced in white matter and basal forebrain region might be due to the fact, that the ruptured aneurysm, respectively the endovascular punctured vessel to induce SAH, is anatomically located in the frontobasal region and the free ruptured amount of blood ows through the subarachnoid space along the frontobasal olfactory region and further along the outer CSF space reaching gray matter and neocortex. Through diffusion of blood and blood degradation products the white matter is reached, possibly re ected by a reduced [ 18 F]FDG uptake. In addition, our ndings that aSAH increases the glucose usage in gray matter but decreases it in white matter would be consistent with an impaired function but increased apoptosis -of course with limitations in useful interpretation due to possible underlying anesthetic effects.
Although there might be a bias via the usage of anesthetics such as iso urane 34 , this experiment would not be possible without anesthesia. The mortality rate of SAH animals would drastically increase without sedation and intubation because of the sharp initial increase of intracranial pressure (ICP) immediately after SAH and the possible consecutive breathing arrest 35 . In addition, PET-scans in SAH rats would not be possible without anesthesia or only with results of enormous blurring. Translational research approaches in SAH patients could try to analyze glucose metabolism via PET scans without anesthetic effects -although not possible in poor grade SAH patients, too.
With regard to novel neuroprotective therapy options in aSAH, the derangement of cerebral metabolism illustrated in this study could be a potential therapy target. Calcium-channel blockers, like nimodipine, for example, were found to have a bene cial effect in terms of metabolic disruption, histological damage and clinical outcome after cerebral ischemia 36 . While most of the underlying mechanisms are still unknown, also a closer look at the cellular level could be worth it. With regard to our discovery of reduced PDH function, which could play a critical role in the development of an early brain injury following aSAH 12 , the prevention from this inactivity may have a neuroprotective effect. Bypassing the PDH e.g. with dichloroacetate (DCA) or entering "new fuel" to the TCA cycle e.g. via acetyl-L-carnitine (ALCAR) could have neuroprotective effect and attenuate or prevent from secondary brain injury following aSAH 12 . Finally, mitochondrial dysfunction following aSAH is shown to activate the autophagy of neuronal cells, possibly leading to early brain injury and DCI 37 . Therefore, targeting the autophagy-lysosomal system could also have a neuroprotective effect 38 .

Conclusion
Aneurysmal SAH is a complex cerebrovascular disease with continuing high mortality and morbidity rates. To our best knowledge, this is the rst time that metabolic changes following experimental SAH have been examined up to day 7 in time course of disease via [ 18 F]FDG-PET and for the rst in different brain regions. We have shown signi cantly increased [ 18 F]FDG uptake in gray matter, neocortex and olfactory system as compared to animals of the sham operated group, while white matter and basal forebrain region showed signi cantly reduced tracer accumulation in SAH animals. Our data analysis tool allows a variety of data analysis methods offering a more exible data evaluation, regardless of whether the radiotracer uptake is homogeneous in brain or target-selective. Thus, also in view of novel tracer developments, this tool represents a very promising alternative for the evaluation of preclinical data. A better understanding of the underlying metabolic derangement hopefully helps to identify new targets for translational neuroprotective therapies.  Figure 1 Co-registered PET image template based on the sham-operated controls as well as following anatomical image information: MRI T2* image (A), gray matter (B), white matter (C), the neocortex (D), the basal forebrain region (E) as well as the olfactory system (F).

Figure 3
Signi cant differences in UR between sham operated and SAH animals for white matter. p ≤ 0.05 *; p ≤ 0.01 **; p ≤ 0.001 ***. [ 18 F]FDG uptake is highly signi cant (p≤ 0.001) reduced in SAH animals compared to the sham operated group 3 hours following the SAH/sham operation. The UR of SAH animals stays signi cantly (p ≤ 0.05 ) reduced in white matter up to day 4 following SAH.

Figure 4
Signi cant differences in UR between sham operated and SAH animals for neocortex. p ≤ 0.05 *; p ≤ 0.01 **; p ≤ 0.001 ***. The [ 18 F]FDG uptake in SAH animals is highly signi cant (p ≤ 0.001) elevated compared to the sham operated control and stays signi cantly higher (p ≤ 0.001) up to day 7 following SAH.

Figure 5
Signi cant differences in UR between sham operated and SAH animals in the basal forebrain region.

Figure 6
Signi cant differences in UR between sham operated and SAH animals in the olfactory system. p ≤ 0.05 *; p ≤ 0.01 **; p ≤ 0.001 ***. [ 18 F]FDG uptake is signi cantly elevated 24 hours respectively on day 1 following SAH compared to sham operated controls. This more-uptake gets even more signi cant on day 4 (p ≤ 0.01) and on day 7 (p ≤ 0.001) following SAH/sham operation.