Magnetic resonance imaging of noradrenergic neurons

Noradrenaline is a neurotransmitter involved in general arousal, selective attention, memory, inflammation, and neurodegeneration. The purpose of this work was to delineate noradrenergic neurons in vivo by T1-weighted MRI with magnetization transfer (MT). In the brainstem of human and mice, MRI identified the locus coeruleus, dorsal motor vagus nucleus, and nucleus tractus solitarius. Given (1) the long T1 and low magnetization transfer ratio for the noradrenergic cell groups compared to other gray matter, (2) significant correlation between MT MRI signal intensity and proton density, and (3) no correlation between magnetization transfer ratio (or R1) and iron, copper, or manganese in human brain, the high MRI signal of the noradrenergic neurons must be attributed to abundant water protons interacting with any T1-shortening paramagnetic ions in active cells rather than to specific T1-shortening molecules. The absence of a high MRI signal from the locus coeruleus of Ear2(−/−) mice lacking noradrenergic neurons confirms that cell bodies of noradrenergic neurons are the source of the bright MRI appearance. The observation of this high signal in DBH(−/−) mice, in 3-week-old mice, and in mice under hyperoxia/hypercapnia/hypoxia together with the general absence of neuromelanin (NM) in noradrenergic neurons of young rodents further excludes that it is due to NM, dopamine β-hydroxylase, their binding to paramagnetic ions, blood inflow, or hemoglobin. Instead, these findings indicate a high density of water protons whose T1 is shortened by paramagnetic ions as the relevant source of the high MRI signal. In the brain of APP/PS1/Ear2(−/−) mice, a transgenic model of Alzheimer’s disease, MRI detected noradrenergic neuron loss in the locus coeruleus. Proton magnetic resonance spectroscopy revealed that a 60–75% reduction of noradrenaline is responsible for a reduction of N-acetylaspartate and glutamate in the hippocampus as well as for a shortening of the water proton T2 in the frontal cortex. These results suggest that a concurrent shortage of noradrenaline in Alzheimer’s disease accelerates pathologic processes such as inflammation and neuron loss. Electronic supplementary material The online version of this article (10.1007/s00429-019-01858-0) contains supplementary material, which is available to authorized users.


3
in the medulla oblongata, and A6, i.e., the locus coeruleus (LC) in the pons (Moore and Bloom 1978). Cell bodies of NA neurons are rich in copper (Cu 2+ ) ions, because Cu 2+ is required for the electron transfer in an enzymatic reaction catalyzed by dopamine β-hydroxylase that converts dopamine to NA (Kaufman 1974). Accordingly, the copper concentration in LC is 6-to 10-fold higher than in other gray matter or 14-to 20-fold higher than in white matter (WM) (data from human, Prohaska 1987). Thus, NA neurons can be assumed to contain abundant water protons whose T 1 is shortened by paramagnetic ions.
The purpose of this work was to (a) delineate NA neurons in human and mice in vivo by combining T 1 -weighted MRI with magnetization transfer (MT), which does not affect the MRI signal increase induced by paramagnetic ions but suppresses signals from water molecules in contact with diamagnetic macromolecules (Henkelman et al. 2001;Watanabe et al. 2012), (b) compare MRI of wild-type and transgenic mice, i.e., Ear2(−/−) and DBH(−/−), to verify the source of the image contrast, (c) apply MRI to characterize APP/PS1/Ear2(−/−) mice, a transgenic model of Alzheimer's disease, and (d) perform proton magnetic resonance spectroscopy (MRS) to further characterize the transgenic model.

Human brain MRI
All procedures involving human participants (adult male, n = 12) in this study were approved by the ethics committee of the Georg-August-Universität Göttingen and performed in accordance with the ethical standards of the institutional and national research committee and with the 1964 Helsinki declaration and its later amendments. All participants gave written informed consent before each examination.
For magnetization transfer gradient-echo MRI of LC and A2, the data were accumulated twice (TR/TE = 863/4.4 ms, total acquisition = 9 min 10 s). The slices are positioned perpendicular to the anterior wall of the fourth ventricle. The LC is observable in the 5th-7th slices from the top, while A2 is observable in the 16th-18th slices. For comparison, interleaved multi-slice T 1 -weighted fat-suppressed 2D fast spinecho MRI at the same spatial resolution was performed with TR/TE = 597/7.3 ms, echo train length = 3, flip angle = 150°, pixel bandwidth = 400 Hz, number of slices = 3 or 21, total acquisition = 6 min 22 s.

Calculation of signal intensity in gradient-echo MRI of the brain
In MRI of the brain in vivo, the signal originates exclusively from water protons. With negligible signal contributions from T 2 coherence, the observable signal in the steady state of a spoiled gradient-echo sequence yields: with flip angle α, repetition time TR, echo time TE, spin-lattice relaxation rate R 1 , effective spin-spin relaxation rate R * 2 , and initial magnetization M 0 . The observable signal in magnetization transfer MRI yields: S = S 0 (1 − MTR) with S 0 the observable signal in gradient-echo MRI without MT and MTR the magnetization transfer ratio.

Anesthesia
After induction of anesthesia with 5% isoflurane, animals were intubated with a purpose-built polyethylene endotracheal tube (0.58 mm inner diameter, 0.96 mm outer diameter) and artificially ventilated using an animal respirator (TSE, Bad Homberg, Germany) with a respiratory rate of 25 breaths per minute and an estimated tidal volume of 0.35 ml as previously described (Schulz et al. 2002;Watanabe et al. 2004Watanabe et al. , 2016a. The animals were then placed in a prone position on a purpose-built palate holder equipped with an adjustable nose cone. The Göttingen animal bed (Tammer et al. 2007) secured a reproducible and reliable fixation of the mouse head and receiver coil in the magnet isocenter. Respiratory movement of the abdomen as well as rectal temperature was monitored by a unit supplied by the manufacturer (Bruker Biospin MRI GmbH, Ettlingen, Germany).
In addition, to examine whether a change in blood supply influences the MRI signal intensity, the animals were anesthetized by intraperitoneal injection of ketamine (200 mg/kg body weight) and xylazine (16 mg/kg body weight). Hyperoxia was induced by changing the inspiratory gas from ~ 90% air and ~ 10% O 2 to ~ 20% air and ~ 80% O 2 , while hypoxic hypercapnia was induced by ~ 60% air, ~ 10% O 2 , and ~ 30% CO 2 as described previously (Watanabe et al. 2016a).
For evaluation of signal intensities, anatomically defined cross-sections were obtained from the original 3D MRI data sets by multiplanar reconstructions using software supplied by the manufacturer (Paravision 5.0, Bruker Biospin MRI GmbH, Ettlingen, Germany). The plane of the anterior commissureposterior commissure served as a reference for the selection of standardized sections to facilitate comparisons with minimized intra-and inter-individual variability. For LC or A2, a rectangular region-of-interest of six pixels was taken in the center of the delineated structures. For the brainstem, a circular region-of-interest of 1004 pixels was taken in the brainstem between LC and A2. The SNR was defined as the mean MRI signal intensity divided by the standard deviation of the noise. The analysis followed a strategy previously developed for intra-individual comparisons of MR images obtained after manganese administration (Watanabe et al. 2004).

Mouse brain MRI at 9.4 T
High-field MRI measurements were carried out at 9.4 T (Bruker Biospin MRI GmbH, Ettlingen, Germany). RF excitation was accomplished with the use of a birdcage resonator (inner diameter 70 mm), while signals were received by a four-channel phased-array surface coil. An off-resonance Gaussian RF pulse with a frequency offset of − 1.4 kHz, a duration of 12 ms, and a flip angle of 240° was incorporated into a fat-suppressed gradient-echo MRI sequence (3D FLASH) with: TR/TE = 33/7.5 ms, flip angle 30°, spectral bandwidth 21 kHz, field-of-view 20.5 × 15.4 × 20.5 mm 3 , matrix 256 × 192 × 256, at an isotropic resolution of 80 µm, 2 averages, total acquisition 54 min. In addition, fat-suppressed low-resolution MRI (spectral bandwidth 10 kHz, field-of-view (15 mm) 3 , matrix 128 3 , isotropic resolution of 117 µm, 2 averages, total acquisition 9 min) was performed with MT (flip angle of 120°) to examine whether a change in blood supply influences the signal intensity.

Magnetic resonance spectroscopy
At 9.4 T, localized proton MRS (STEAM, TR/TE/ TM = 6000/10/10 ms) was performed with the use of the birdcage resonator and a saddle-shaped quadrature surface coil (Bruker Biospin MRI GmbH, Ettlingen, Germany) on anesthetized mice. A 1.8 × 1.8 × 1.0 mm 3 volume-of-interest was localized in the frontal cortex or 1.8 × 1.8 × 1.2 mm 3 volume-of-interest was centered on the hippocampal formation. Water saturation was accomplished by means of three Gaussian-shaped CHESS RF pulses (90°-90°-180°).
Metabolite quantification involved spectral evaluation by LCModel (Provencher 1993) and calibration with brain water concentration (Duarte et al. 2014), for which the unsuppressed water proton signal served as an internal reference. The attenuation of the unsuppressed water signal was calibrated with T 2 relaxation times of water protons in the volume-of-interest that was determined by a multi-echo spin-echo MRI (TR/TE = 2500/10-123 ms). T 1 relaxation times were determined with the use of a spin-echo saturation recovery sequence and 7 TR values from 0.15 to 6 s. Metabolites with Cramer-Rao lower bounds above 20% were excluded from further analysis.

Statistical analysis
Statistical evaluation was performed using SPSS ® (version 21.0, IBM ® ) and Microsoft Excel ® . Alpha level (criterion of significance) was set to 0.05. Significant differences between two groups of data were determined by the Mann-Whitney's U test.

Magnetization transfer MRI delineates noradrenergic neuron groups based on proton-density contrast while saturating long-T 1 extracellular protons
In human brainstem, MT-MRI delineated assemblies of NA neurons (Fig. 1). Figure 2 compares various quantitative maps of the human brain. The M 0 map (M 0 ), proton-density-weighted MRI with MT (α15 MT), and T 1 -weighted MRI with MT (α70 MT) show a similar contrast except for cerebrospinal fluid (CSF). The MT ratio maps (α15 MTR and α70 MTR) and R 1 map (R 1 ) also show a similar contrast. Quantitative evaluations (Supplementary Fig. 1, Table 1a) confirm these observations. Mean regional R 1 of gray matter (GM) structures correlate linearly with water content (Supplementary Fig. 1a). The upward deviation of R 1 of the WM from the linear relationship is attributable to its large interaction with water protons (Koenig et al. 1990). This effective cross-relaxation in WM is apparently cancelled out by the more pronounced spin-lock effect on GM structures, which results in an apparent linear relationship between the MT ratio (α70) and the water content including both GM and WM (Supplementary Fig. 1b). The values of the GM structures with significantly higher iron content, e.g., the globus pallidus and the red nucleus, and those with significantly higher copper content, e.g., the substantia nigra (SN) and LC, are well in line with those of other GM structures. Accordingly, regional signal intensities in T 1 -weighted MRI with MT (α70 MT) correlate significantly (p < .005) with those in M 0 in every subject ( Supplementary Fig. 1c), which also indicates that the regional spin density of intracellular water protons correlates significantly with the regional water content. Here, the NA cell groups were shown to have low MT ratios and high T 1 and T 2 values compared to other GM structures (Table 1). These findings indicate that the NA cell groups have high water content and that their highly concentrated copper is not the source of the high MRI signal intensity. For LC and A2 which are relatively unclear to delineate, the data were accumulated twice (TR/TE = 863/4.4 ms, total acquisition = 9 min 10 s) from another set of subjects that turned out to confirm the above findings (Table 2). Figure 3a-d shows the variables R 1 , R * 2 , and MTR as a function of tissue water content, which can be used for the calculation of signal intensities in gradient-echo MRI (Fig. 3e) for a selected combination of acquisition parameters α, TR, and TE. A strong on-resonance irradiation saturates the long-T 1 water protons in the extracellular fluid with a water content of 0.99 (Fig. 3e). Additional MT suppresses the MRI signal in proportion to the macromolecular content (i.e., in inverse proportion to the water content), while the saturation of the extracellular water protons is preserved. As a result, the MT effect dominates for brain tissue (water content < 0.9), while the T 1 effect dominates for extracellular fluid (water content > 0.95). When saturating extracellular long-T 1 protons, the significant correlation between the signal intensities in MT-MRI and the proton density ( Supplementary Fig. 1c) indicates that the signal intensities reflect the spin density Fig. 1 MRI delineates noradrenergic neuron groups in human brain in vivo. a (Left) Mid-sagittal human brain MRI illustrating the fieldof-view selected for imaging the locus coeruleus and A2. (Right) Transversal MRI of A2 cell groups (arrows) in six different subjects. b Transversal T 1 -weighted MRI without (α70) and with MT (α70MT), MT ratio map (α70MTR), T 1 map (T 1 ), M 0 map (M 0 ), and T 2 map (T 2 ) showing the locus coeruleus (arrows) and c the A2 cell group (arrows) ◂ of intracellular water protons whose T 1 is shortened. In contrast to tissue water content, paramagnetic ion concentrations correlate neither with R 1 nor with MT ratios (Fig. 3f-k). These findings indicate that the regional water content determines both MT ratio and R 1 . Figure 4 shows that LC can be delineated by increasing the number of slices in multi-slice 2D fast spin-echo MRI in a similar way as by additional MT pulses (cf., α70 and α70 MT in Fig. 1b). Although each slice is acquired in an interleaved manner, substantial MT effect is induced by the repetitive 150° refocusing pulses with frequency offset corresponding to location offset for each of the 21 slices.

Magnetization transfer delineates noradrenergic neuron groups in the brainstem of mice in vivo
In the brainstem of mice, MRI delineated assemblies of NA neurons (Fig. 5). In young Ear2 (−/−) mice, MRI detected substantially less LC signal than in age-and gender-matched controls (Fig. 5, Supplementary Fig. 2) in agreement with the reduction in LC neuron numbers (Warnecke et al. 2005). Histologically, there was no neuromelanin (NM) observable in LC of control animals (Kummer et al. 2014;Warnecke et al. 2005) in agreement with previous reports (Barden and Levine 1983). This indicates that the NM is not the source of the high MRI signal intensity in mice. Figure 6 shows that dopamine β-hydroxylase has no influence on the MRI signals of NA cell groups. The SNR of LC is not significantly different in DBH(−/−) (47.9 ± 4.4, n = 5) compared with DBH(+/−) mice (46.3 ± 2.6, n = 5). This indicates that neither the presence of dopamine β-hydroxylase nor its binding to Cu 2+ ions (Blumberg et al. 1965) is the source of the high MRI signal intensity. Figure 7a confirms that MRI of NA neurons in vivo at higher spatial resolution and higher magnetic field is in agreement with light microscopy of cell bodies. Figure 7b, c shows that alterations in tissue oxygenation or perfusion do not substantially influence the signal intensity of LC. Increased deoxyhemoglobin content induced by hypoxic hypercapnia clearly accelerates T * 2 -dephasing (Fig. 7c) in brain tissue associated with dilated vasculature, whereas the high signal intensity of LC is barely affected. These findings indicate that the main source of the high signal intensity of NA neurons is neither presence nor inflow of hemoglobin, water, or other component of blood.

Magnetization transfer generates proton-density contrast in "T 1 -weighted" MRI of the brain
In principle, signal intensity in spoiled gradient-echo MRI reflects the spin density, T 1 , and T * 2 of water protons. In T 1 -weighted MRI, repetitive strong on-resonance irradiation saturates the extracellular long-T 1 protons while providing intracellular water protons with high signals. With additional MT, image contrast now primarily reflects the proton density (cf., equations in "Calculation of signal intensity in gradient-echo MRI of the brain") and only secondarily the T 1 relaxation (and T * 2 even less). This is because the water protons whose T 1 is shortened by diamagnetic molecules are proportionally saturated by the off-resonance irradiation ( Supplementary Fig. 1a, b). The proton density effect dominates the observable signal for intracellular water (the water content of brain cells < 0.9), while the T 1 effect applies to extracellular water (the water content of cerebrospinal fluid > 0.95) (Fig. 3e).
As a result, apart from long-T 1 protons of extracellular water, additional MT actually turns "T 1 -weighted" MRI into "proton-density-weighted" MRI as far as the brain tissue itself is concerned. Here, "T 1 -weighted" applies only for the contrast between CSF and brain tissue itself. In other words, signal intensity in gradient-echo MRI with strong on-resonance and off-resonance irradiation reflects the density of water protons whose T 1 is shortened by paramagnetic ions because the T 1 -shortening effect induced by diamagnetic molecules is cancelled by MT, whereas the T 1 -shortening effect induced by paramagnetic ions is not. For example, SN appears brighter than the red nucleus or cerebral peduncles in α15 and α70 images in Fig. 2 because of higher proton density in SN. The tissue surrounding SN contains water molecules bound to diamagnetic macromolecules where MT suppresses signal (more of them compared to SN).
So far, a number of studies reported the delineation of SN and LC by multi-slice 2D fast spin-echo MRI (Sasaki et al. 2006;Clewett et al. 2016). As shown in the present showing magnetization transfer effect induced by repetitive refocusing pulses. Also note that the signals of the subcortical white matter (black arrows) are predominantly suppressed by magnetization transfer effect in a 21-slice acquisition and, thus, its contrast to the cortex has disappeared or is even reversed study (Figs. 1b, 4), the mechanism for contrast generation is basically the same for MT and for increasing the number of slices in 2D MRI with strong refocusing pulses: T * 2 or T 2 effect, respectively, is dominated by other factors, e.g., proton density. In this regard, spin-echo MRI is theoretically less efficient for delineating SN or LC than gradientecho MRI because the time spent for refocusing hardly contributes to the delineation. Further, the MT effect generated by refocusing pulses is less controllable than the use of specific MT pulses because the MT effect depends on many other parameters, e.g., the number, direction, order, and gap of slice acquisition (Thomas et al. 2004). In more detail, the frequency offset of refocusing pulses is commonly smaller than that of a specific MT pulse, so that a more pronounced spin-lock effect is involved in "MT effect" by refocusing pulses. The generally higher T 1 /T 2 ratios for cell assemblies than for WM or CSF in vivo (Table 1) indicate that the spin-lock effect is unfavorable for the delineation of cell assemblies from surrounding WM or CSF. Higher T 1 /T 2 ratio results in a more pronounced spin-lock effect, i.e., lower signal intensities for the cell assemblies, whereas the lower T 1 /T 2 ratio results in a less pronounced spin-lock effect, i.e., higher signal intensities for WM and CSF (cf., Supplementary Fig. 1a, b). Potential off-resonance artifacts in gradient-echo MRI may be minimized by increasing the spatial resolution.
The present study also showed that LC and A2 can be well delineated regardless of the field strength: the measurements of mice were carried out either at 2.35 or 9.4 T. In fact, this supports the view that the delineation of LC and A2 does not rely on T 1 or T 2 contrast but on the proton-density contrast because the relaxation times are field dependent.

Paramagnetic ions in active cells
Every active cell needs paramagnetic ions as enzyme cofactors for electron transfer which serves essential processes (Que et al. 2008). In mammalian brain, these include cellular respiration, anti-oxidant defense, and catecholamine synthesis/metabolism, e.g., by electron transport chain, superoxide dismutase, tyrosine hydroxylase, and dopamine β-hydroxylase. As a consequence, metals such as iron, copper, and manganese, which possess a high capacity to gain and donate electrons, are the most abundant endogenous paramagnetic ions.
Most copper ions in brain are bound to proteins and thus their distribution among different sub-cellular compartments is tightly regulated (Lutsenko et al. 2010;Que et al. 2008). Therefore, the T 1 relaxivity of Cu 2+ in brain in vivo may be high compared to other paramagnetic ions, because Cu 2+ has a long electron spin relaxation time (Burton et al. 1979). At 50 MHz, T 1 relaxivity of free Cu 2+ is about 0.5 L mmol −1 s −1 (Vymazal et al. 1998) that may be enhanced upon binding (Eisinger et al. 1961) to be about 5 L mmol −1 s −1 in vivo, whereas that of free Fe 3+ is 6.9 L mmol −1 s −1 (Vymazal et al. 1998) that may be reduced (Eisinger et al. 1961) to about 3 L mmol −1 s −1 . Given this favorable property of efficient T 1 -shortening, it is possible that the exceptionally high copper concentration plays a role in MT-MRI signal intensity of NA neurons, although the present study shows that it is the water proton density that plays a dominant role.

Neuromelanin
Similar MRI contrast described for LC in human so far has been attributed to a water proton T 1 -shortening induced by NM (Sasaki et al. 2006;Clewett et al. 2016). NM bound to iron in vitro was shown to shorten T 1 and thus to reduce MT ratio (Trujillo et al. 2017). However, it is unlikely that the MT contrast of specific cellular assemblies shown in the present study is due to unique T 1 -shortening molecules like NM. Firstly, if the high signal intensity of those structures (including LC) was due to the presence of T 1 -shortening molecules, then the T 1 of those structures would have been shorter than other gray matter or surrounding structures. The present study showed that exactly the opposite is the case (Table 1). In fact, T 1 relaxation times of some GM structures that contain no NM (e.g., thalamus, globus pallidus, red nucleus, and subthalamic nucleus) are even shorter than those of SN, LC, or A2 (Table 1) in line with a recent report (Priovoulos et al. 2018). If the high signal of SN and LC resulted from the T 1 -shortening by NM, the GM structures These apparent contradictions can be explained as follows based on the present data (Figs. 1, 2, 3; Table 1): it is not the T 1 relaxation but the water proton density that predominantly determines the in vivo MT-MRI contrast within the brain (cf., "Magnetization transfer MRI delineates noradrenergic neuron groups based on proton-density contrast while saturating long-T 1 extracellular protons" and "Magnetization transfer generates proton-density contrast in "T 1 -weighted" MRI of the brain"). The in vitro evidence of T 1 -shortening by NM (Trujillo et al. 2017) does not necessarily mean that the delineation of NM-containing structures by in vivo MT-MRI results from the T 1 -shortening. Theoretically (cf., 2 under (from left to right) ketamine anesthesia, hyperoxia, 2% isoflurane, and hypoxic hypercapnia. Neither of these conditions that induce alterations in deoxyhemoglobin or blood flow substantially influences the signal intensity of the locus coeruleus (white arrows). c (Left) Horizontal MRI of animal no. 2 without and (right) with hypoxic hypercapnia. Note the preserved high signal intensity of the locus coeruleus (white arrows) despite the pronounced signal loss caused by increased deoxyhemoglobin content in dilated vasculature throughout the brain equations in "Calculation of signal intensity in gradient-echo MRI of the brain"); T 1 -shortening plays a role, but signal intensity is dominated by M 0 . In this regard, low MT ratios and high T 1 and T 2 values of SN, LC, and A2 as well as of caudate nucleus and putamen (Table 1) indicate that these structures have high water content and thus yield high signal intensities in c,2a). Conversely, high MT ratios and low T 1 and T 2 values of thalamus, globus pallidus, red nucleus, and subthalamic nucleus (Table 1) indicate that these structures have low water content and thus yield low signal intensities in MT-MRI (Fig. 2b, c).
Secondly, the present study demonstrated high MRI signal intensities in LC and A2 of 3-and 4-week-old mice 7b,c). In SN of 15-to 24-month-old rats, electron microscopy revealed a few small NM granules in only 5% of SN neurons (DeMattei et al. 1986) while NM content increases with age. Another study showed that chronic exposure of rats to continuous florescent light increased NM deposition in SN (Romeo et al. 2013). However, given the lack of NM reported for NA neurons in rodents or for neurons in young rodents (Barden and Levine 1983), the high signals of LC and A2 in 4-week-old mice are not attributable to NM. In fact, although LC and A2 in the same kind of animals as those used in the present study have been intensively examined by a variety of histological techniques (Hammerschmidt et al. 2013;Kummer et al. 2014;Warnecke et al. 2005), the presence of NM, which could be readily detected under a simple light microscopy, has not been observed.
Thirdly, there is a substantial pool of paramagnetic ions in SN and LC that are not associated with NM (cf. "Paramagnetic ions in active cells"). Even when assuming that paramagnetic ions contributed to a brighter signal in T 1 -weighted images, the NM is unlikely to play a central role. The concentration of NM Fe in LC (SN) is only about 3.6 (20) ng/mg wet tissue, whereas the concentration of Fe in LC (SN) is as much as about 25 (150) ng/mg wet tissue (Zecca et al. 2004). The concentration of NM Cu in LC (SN) is only about 1.2 (0.4) ng/mg wet tissue, whereas the concentration of Cu in LC (SN) is as much as 31 (16) ng/mg wet tissue. In more detail, total Fe in LC of human is 5-45 ng/mg wet tissue. Fe bound to NM in LC is 1777 ± 92 ng/mg NM dry weight. NM concentration in LC is 1000-3200 ng/mg wet tissue. 1777 ng/mg NM dry weight multiplied by 1000-3200 ng/mg wet tissue yields 1.8-5.4 ng/mg wet tissue. Total Cu in LC is 8-64 ng/mg wet tissue. Cu bound to NM in LC is 605 ± 79 ng/mg of NM dry weight. 605 ng/mg NM dry weight multiplied by 1000-3200 ng/mg wet tissue yields 0.6-1.8 ng/mg wet tissue.
In general, NM as an effective metal chelator primarily serves to trap iron and provide neuronal protection from oxidative stress . Excess dopamine or of the hippocampal formation) of (left) APP/PS1 (n = 8, 18.8 ± 5.8 months) and (right) APP/PS1/Ear2(−/−) (n = 10, 18.8 ± 3.6 months) mice in vivo (see Table 2). Glu glutamate, NAA N-acetylaspartate, ↓ significant decrease noradrenaline can be removed by converting it into a stable compound like NM, and this process rescues the cell. Thus, it must be noted that the present study using a mouse model of Alzheimer's disease has to be cautiously translated into human medicine, because the amount and effect of NM may be different between mice and human.

Blood flow and hemoglobin
MT is commonly used for magnetic resonance angiography to better delineate signal intensity in T 1 -weighted multislice 2D gradient-echo MRI caused by initial excitation of unsaturated blood water protons that have just flown into the imaging plane, i.e., time-of-flight effect. Here, in MRI of mouse brain, LC and A2 are well within (and off the margin of) the volume that is excited only as a 3D slab by a moderate flip angle of 22°. Thus, time-of-flight effect is not so pronounced as in 2D MRI. Further, high signal intensity is observed (not in vessels but) in nerve cell assemblies like LC and A2 which gain blood supply through capillary beds. In such brain tissue (e.g., the visual cortex), the signal intensity change was shown to be independent from the excitation flip angle, i.e., from the time-of-flight effect (Frahm et al. 1994).
By contrast, the signal intensity in the visual cortex was shown (Frahm et al. 1994) to be significantly dependent on brain activation (i.e., visual stimulation by light) that is associated with washout of T * 2 -shortening paramagnetic deoxyhemoglobin from capillary beds by inflowing diamagnetic oxyhemoglobin. Although the preserved high signal intensity of LC despite the pronounced signal loss caused by increased deoxyhemoglobin content in dilated vasculature (Fig. 7c) indicates that hemoglobin is not the main source of the high signal of NA neurons in the present study, the blood-oxygenation-level-dependent effect is theoretically expected and cannot be excluded for LC. Given the high exposure of LC neurons to blood circulation through dense capillaries (Mather and Harley 2016), the T * 2 effect could have been well observed if the gradientecho MRI was performed with a longer echo time. However, the present study sought for proton-density rather than blood-oxygenation-level-dependent contrast. Table 3 Concentration (mM) and concentration ratio of major cerebral metabolites as well as T 1 and T 2 relaxation times of water protons in APP/ PS1/Ear2(−/−) and wild-type control mice

Dopaminergic neurons
The present study shows that SN as well as LC yields high signal intensities in MT-MRI of human brain (α70MT in Fig. 2d) in agreement with a number of previous studies (Sasaki et al. 2006). This clear delineation is probably associated with the large size of the pars compacta in SN as well as with the tight packing of the large dopaminergic cells in the pars compacta. In primates, the pars compacta is as large as the rest of SN, the pars reticulata (Voogd 1998). By contrast, SN was not delineated in MT-MRI of mouse brain in the present study. This can be explained by partial volume effect because, in contrast to primates, the pars compacta is substantially smaller than the pars reticulata in rodents (Voogd 1998).

Conclusions
To summarize, this is the first report about the in vivo MRI visibility of A2 cell assemblies in human as well as of NA neurons in animals. The absence of the high MRI signal from LC of Ear2(−/−) mice confirms the cell bodies of NA neurons as a source of the high signal. The preservation of the high signal of NA neurons in DBH(−/−) mice indicates that the source of the high signal is neither dopamine β-hydroxylase nor their binding to paramagnetic ions. The detection of these NA cell groups in young wild-type mice by T 1 -weighted MRI with MT, the reduced contrast in Ear2 (−/−) mice, and the preserved contrast in DBH (−/−) mice together indicate that the source of the high MRI signal is neither NM nor dopamine β-hydroxylase, but a high density of water protons whose T 1 is shortened by paramagnetic ions. Application of this MRI method to a transgenic model of Alzheimer's disease illustrated its potential use in neuroradiology. Given that the decline in LC neuron numbers is associated with aging, dementia, Aβ plaque load, and the progression of Alzheimer's disease (Kummer et al. 2014), it is foreseeable that MRI of NA neurons will play an increasing role in translational biomedical research of neurodegenerative diseases.