Content uploaded by Elaine L Bearer
All content in this area was uploaded by Elaine L Bearer on Mar 11, 2022
Content may be subject to copyright.
February 2022, NMR in Biomedicine, DOI: 10.1002/nbm.4675
Longitudinal manganese-enhanced magnetic resonance imaging of neural projections
Taylor W. Uselman1, Christopher S. Medina1, Harry B. Gray2, Russell E. Jacobs3 and Elaine L. Bearer1,2
1 University of New Mexico Health Sciences Center, Albuquerque, NM 87131
2 Beckman Institute, California Institute of Technology Pasadena, CA 91125
3 Zilkha Neurogenetic Institute, Keck School of Medicine, University of Southern California, Los Angeles, CA 90089-
Elaine L. Bearer, MD-PhD, FAAAS, FCAP
Department of Pathology
MSC08-4640, 1 University of New Mexico
University of New Mexico Health Sciences Center
Albuquerque, NM 87131
firstname.lastname@example.org; email@example.com; firstname.lastname@example.org
Russell E Jacobs, Ph.D.
Zilkha Neurogenetic Institute
Keck School of Medicine
University of Southern California, Los Angeles, CA 90089-2821
Key words: Manganese, Longitudinal imaging, Magnetic resonance imaging, Rodent brain imaging, Tract-tracing,
brain-wide activity mapping, T1-weighted imaging, Computational image analysis
Running Title: Longitudinal MEMRI
Uselman et al. Longitudinal MEMRI
Manganese-enhanced magnetic resonance imaging (MEMRI) holds exceptional promise for preclinical studies of
brain-wide physiology in awake-behaving animals. The objectives of this review are to update the current information
regarding MEMRI and to inform new investigators as to its potential. Mn(II) is a powerful contrast agent for two main
reasons: 1) High signal intensity at low doses; 2) Biological interactions, such as projection tracing and neural activity
mapping via entry into electrically active neurons in the living brain. High-spin Mn(II) reduces the relaxation time of
water protons: at Mn(II) concentrations typically encountered in MEMRI, robust hyperintensity is obtained without
adverse effects. By selectively entering neurons through voltage-gated calcium channels, Mn(II) highlights active
neurons. Safe doses may be repeated over weeks to allow for longitudinal imaging of brain-wide dynamics in the same
individual across time. When delivered by stereotactic intracerebral injection, Mn(II) enters active neurons at the
injection site and then travels inside axons for long distances, tracing neuronal projection anatomy. Rates of axonal
transport within the brain were measured for the first time in "time-lapse" MEMRI. When delivered systemically
Mn(II) enters active neurons throughout the brain via voltage-sensitive calcium channels and clears slowly. Thus
behavior can be monitored during Mn(II) uptake and hyper-intense signals due to Mn(II) uptake captured
retrospectively, allowing pairing of behavior with neural activity maps for the first time. Here we review critical
information gained from MEMRI projection mapping about human neuropsychological disorders. We then discuss
results from neural activity mapping from systemic Mn(II) imaged longitudinally that have illuminated development of
the tonotopic map in the inferior colliculus as well as brain-wide responses to acute threat and how it evolves over
time. MEMRI posed specific challenges for image data analysis that have recently been transcended. We predict a
bright future for longitudinal MEMRI in pursuit of solutions to the brain-behavior mystery.
Uselman et al. Longitudinal MEMRI
This comprehensive review of MEMRI details: the chemical and physical properties of Mn (II); considerations for safe
dosing; perspectives on scanning parameters; discussions of biological mechanisms of projection and neural activity
mapping; contributions yielded by MEMRI; procedures for image processing; strategies for computational analysis;
and concludes with remarks for future directions. This paper will provide a resource to guide investigators when
applying MEMRI to critical neurobiological problems.
Uselman et al. Longitudinal MEMRI
1-Mn, Mn(III)-5,10,15-tris(pentafluorophenyl)-corrole ; 3D-PCNN, Three-dimensional pulse couple neural networks ;
ACA, anterior cingulate cortex ; AD, Alzheimer’s disease ; AFNI, Analysis of Functional NeuroImages ; AIDAmri,
Atlas-based Imaging Data Analysis Pipeline ; Amy, amygdala ; ANTS, Advanced Normalization Tools ; AON,
anterior olfactory nucleus ; APP, amyloid precursor protein ; aq, aqueduct ; Aβ, amyloid-β ; BBB, blood brain barrier ;
BNST, bed nuclei of the stria terminalis ; BOLD, blood oxygenation level dependent ; CA1-3, cornu ammonis areas 1-
3 of the hippocampus ; Cav1.2, L-type calcium channel 1.2 ; CNR, contrast-to-noise ratio ; CNS, central nervous
system ; COINS, Collaborative Informatic and Neuroimaging Suite ; DAIM, dynamic activity induced MEMRI ;
DMT, dipositive metal transporter ; DOI, 2,5-dimethoxy-4-iodoamphetamine ; DPDP, dipyridoxyl diphosphate ; DR,
dorsal raphe ; DREADD, Designer Receptor(s) Exclusively Activated by Designer Drug(s) ; EM, electron microscopy
; FA, flip angle ; fMRI, functional MRI ; FMRIB, Functional Magnetic Resonance Imaging of the Brain ; FSE,
fast/turbo spin echo ; FSL, FMRIB Software Library ; GP, globus pallidus ; HerMn, HerPBK10 protein encapsulating
1-Mn ; IC, inferior colliculus ; ICA, independent component analysis ; IEG, immediate early gene ; IP, intraperitoneal ;
IV, intravenous ; Kif1a, kinesin-3 ; KLC1, kinesin light chain 1 ; KO, knockout ; LGN, Lateral geniculate nucleus ;
MDT, minimal deformation target ; MEMRI, manganese-enhanced magnetic resonance imaging ; MP-RAGE,
magnetization prepared gradient echo ; MRI, magnetic resonance imaging ; NET, DAT, SERT/5-HTT, norepinephrine,
dopamine and serotonin reuptake transporters ; NIfTI, Neuroimaging Informatics Technology Initiative; NSF,
nephrogenic systemic fibrosis ; ON, optic nerve ; PCA, principal component analysis ; PFC/mPFC, prefrontal
cortex/medial prefrontal cortex ; R1, spin-lattice relaxation rate ; R2, spin-spin relaxation rate ; RATS, Rapid
Automatic Tissue Segmentation ; RBET, Rodent Brain Extraction Tool ; RDA, rhodamine dextran-amine ; RIXS,
Resonant X-ray Inelastic Scattering ; ROI, region of interest ; RSC, retrosplenial cortex ; S/B, Signal over background
; SC, Superior colliculus ; SHERM, SHape descriptor selected Extremal Regions after Morphologically filtering ;
SNR, signal-to-noise ratio ; SPGR, spoiled gradient echo ; SPM, Statistical Parametric Mapping ; SubQ/SC,
subcutaneous ; T1-FLAIR, T1-weighted fluid attenuated inversion recovery ; TE, echo time ; TMT, 2,3,5 Trimethyl-3-
thiazolineTR, repetition time ; VEP, visual evoked potential ; WT, wild type
Uselman et al. Longitudinal MEMRI
2. History of Mn(II) MR imaging
3. Manganese properties
3.1. Physical and chemical properties
3.2. Safety considerations
4. Methodology for manganese-enhanced magnetic resonance imaging
4.1. Contrast agents for MR
4.2. Pulse sequence considerations
4.3. Evaluation of images for optimization of scanning parameters and power analysis
5. Applications of MEMRI: Tracing neuronal projections and axonal transport dynamics by longitudinal imaging
5.1. Rationale and background: Role of axonal transport in tracing and stereotactic injection for localized Mn(II)
5.2. Trans-synaptic tracing
5.3. Applications: Hippocampal projection tracing
5.3.A. Example 1: Rates of Mn(II) accumulation in distal hippocampal projections: Down syndrome
5.3.B. Example 2: Rates of Mn(II) accumulation in distal hippocampal projections: Alzheimer's disease
5.4. Applications: Forebrain projections and reward circuitry
5.4.A. Example 3: Prefrontal cortical projections
6. Applications of MEMRI: Longitudinal imaging for brain-wide activity-induced Mn(II) intensity changes
6.1. Mn(II) enters via Cav1.2, an L-type voltage-gated calcium channel and Mn(II)-induced increased T1-signal
intensity correlates with, and is proportional to neuronal activity
6.2. Dose and timing for longitudinal MEMRI after systemic administration
6.3. Systemic Mn(II) and intermediate gene expression
6.4. MEMRI versus fMRI/BOLD: Limitations and benefits
6.5. Applications: Mn(II) to enhance contrast for analysis of volumetric changes across lifespans
6.6. Applications: Sensory systems
6.6.A. Examples: Longitudinal MEMRI for the auditory system
6.7. Applications: Longitudinal MEMRI for brain-wide activity mapping over time before and after experiences.
6.7.A. Example: Effect of acute fear and serotonin on brain-wide activity
7. Statistical and computational image analysis
7.1. Pre-processing of images
7.1.A. Correction of intensity non-uniformity
7.1.B. Automated brain extraction (Skull-stripping)
7.1.C. Gray scale normalization
7.2. Statistical analysis: Region-of-interest (ROI) measurements and Statistical Parametric Mapping (SPM)
7.2.A. Region-of-interest (ROI)
7.2.B. Statistical Parametric Mapping (SPM)
7.3. Emerging strategies for functional brain network identification with MEMRI
8. Concluding remarks and future directions
Uselman et al. Longitudinal MEMRI
The brain has been a mysterious black box since the beginning of human history. Because it is large (1400 g on
average in the human male), encased in a bony skull and opaque, traditional optical imaging techniques of living brains
were not feasible. Most human brain anatomy was originally discovered post-mortem. Magnetic resonance imaging
(MRI) technology exploded in the latter part of the 20th century, with the Nobel Prize in Physiology and Medicine
awarded to Paul Lauterbur and Sir Peter Mansfield in 2003 for "their discoveries concerning magnetic resonance
imaging"1. MRI yielded detail at less than 1 cm resolution and was further enhanced with the discovery that blood
oxygenation level changes modulate the MR signal2-6. These exciting advances allowed investigators to peer into the
living brain in new ways.
Originally missing from MR imaging was physiological and biochemical labeling. While intensity values produced
anatomical detail, identification of structures relied on visual correlations with histologic atlases of fixed brains. Early
on, Lauterbur (and others developing MR systems) proposed the use of paramagnetic metal ions as contrast agents7-9.
Contrast agents were coming into vogue for other imaging modalities also being developed in the mid to late 20th
century, including electron- and fluorescence microscopy, which also produce intensity patterns without
biochemical/molecular definition. Solutions for molecular imaging employing those approaches provided inspiration
for MR contrast agent development. For example, heavy metal particles, opaque to the electron beam in electron
microscopy, such as those of lead, uranium, lanthanum, and gold10, became useful. These particles could be
specifically deposited locally due to enzymatic activity or derivatized to label particular molecular content11.
Immunogold labeling identified individual molecules within subcellular structures, such as membrane compartments
and intracellular viruses12. Platinum evaporated onto frozen, fractured cells revealed contours whose shape identified
structures, where instead of contrast agents, deformation of contours based on their composition revealed substructures
such as lipid rafts13-15. These chemical modifications revolutionized imaging by adding dimensions of physiological
and biochemical/molecular information to structure.
In MR, proton spins are polarized in a strong magnetic field (the B0 field) and excited with radiofrequency energy.
Alignment is "tipped" by pulsing another magnetic field perpendicular to the B0 field. Precession about B0 emits
radiofrequency energy. For MR contrast, paramagnetic ions such as Mn(II) and Gd(III) interact with water protons,
thereby enhancing proton relaxation, giving a hyper-intense signal in T1 weighted imaging protocols. While eight
Gd(III) chelates are used in health care today to reveal vascular anatomy, Mn(II) is used in pre-clinical settings to
reveal physiological processes at tissue and organismal levels. Mn(II) selectively enters active neurons via voltage-
gated Ca(II) channels. Although Mn(II) may also enter cells via the dipositive metal transporter (DMT), this does not
appear to be responsible for Mn(II)-dependent signal in brain16. Because MR is non-invasive and non-destructive,
repeated imaging sessions of the same individual over time are possible. Mn(II), which is relatively safe at low doses
that give robust MR signals, can be delivered continuously or repeatedly to the same animal with good tolerance.
These properties allow longitudinal imaging of the same individual over time, either at short or long intervals.
Importantly, knowledge of both individual and of group averages also can be obtained, provided residual Mn(II) is
accounted for17-20. Together with computational processing and statistical analysis, longitudinal imaging of neural
activity will illuminate the influences of drugs, experience, genetics or disease progression on the brain. The beauty of
MRI is that it is non-invasive and non-destructive which allow physiological observations in intact living organisms
Application of Mn(II) as a contrast agent for MRI has been steadily increasing (Fig. 1), owing to its unique
physiological attributes of selectively entering active neurons through voltage-gated calcium channels and of
delineating neuronal projections through its transport by endogenous axonal transport machinery. A PubMed search
with query terms ‘manganese’ and ‘magnetic resonance imaging’, first performed in 200421, now reveals a rapid
Uselman et al. Longitudinal MEMRI
increase to 2010 to reach a current plateau of ~100 per year (Fig. 1A) with a cumulative rise of over 2000 publications
in 2020 (Fig. 1B). Some MEMRI-related publications were likely missed by this search, and some may have been
included that were about other manganese MRI uses. A deeper understanding of the molecular mechanisms of Mn(II)
transport may have contributed to this rise in publication numbers22-25. A search in Web of Science, using the same key
words as for PubMed, revealed a similar trend. Many publications fall in radiology and neuroscience fields. Chemistry
and materials science make up the majority of those remaining, with several articles published in journals of American
Chemical Society26-36. With literature in the field growing rapidly, a need to assimilate this information became
apparent, which is a primary goal of this review. Among reviews published over the past 10 years37-50, many focus on
nanoparticles and few incorporate detailed discussions of both chemical and biological roles of Mn(II) in neuroscience.
Newer MEMRI studies have largely shifted away from technology development to report on applications in basic
biomedical and clinical research, where the technology has tremendous value. This newer focus is presented here.
We here explore applications of manganese to examine structures and physiology in the living brain, including
intracellular axonal transport, neural projections, and brain-wide neural activity, reviewing pertinent historical
developments and focusing on recent breakthroughs. The power of MEMRI to investigate how experience and disease
affect brain physiology is emerging as a beautiful way to penetrate the complexities of whole brain dynamics.
2. HISTORY OF Mn(II) MR IMAGING
Early in the development of NMR spectroscopy, many investigators explored the effects of metal ions on nuclear
resonances7,51-53. In the 1950s, Mildred Cohn investigated the NMR spectra of biological molecules in solutions of
metal(II) complexes54; and, in her 1962 JBC paper, she reported that Mn(II) broadened the NMR spectra of ADP and
ATP, owing to dramatic effects on proton relaxation times55. Importantly, she later showed that Mn(II)-perturbed
proton NMR spectra could be employed to determine metal(II) binding sites on enzymes52. What is more, in the 1970s,
Lauterbur, Mendoca-Diaz and co-workers demonstrated that Mn(II) uptake in cardiac muscle could be detected by
MR8,9. Working with Mansfield, they developed methods that allowed assignment of MR signals to specific locations
within these tissues56. !
3. MANGANESE PROPERTIES
3.1. Physical and chemical properties. Two properties of the Mn(II) contribute to its power as an MR contrast agent:
1) it is strongly paramagnetic (there are five unpaired electrons in its electronic ground state (Fig. 2A); and 2) there is
rapid dissociative interchange of inner and outer shell water molecules (Fig. 2B and 2C). The first manganese
compound, potassium permanganate (KMnO4), was synthesized over 300 years ago57. Although oxo-manganese
complexes are generated during photosynthesis58, Mn(II) and Mn(III) are more commonly encountered in biology59.
Many enzymes have manganese as the essential component in their active centers, with manganese superoxide
dismutase a prominent example59. These Mn-binding proteins likely act as a "sink" for persistence of radioactive
54Mn(II) after systemic infusion60,61. Such sequestered manganese does not seem to produce MR signals, as
radioactivity can be detected for 45 days or more61, whereas very little hyper-intense signal remains detectable in most
regions at 14 days in rat brain62.
In aqueous biological media, one or more water molecules reside in the Mn(II) coordination inner sphere. These
coordinated water molecules are quite labile, as there is Mn-O bond weakening attributable to interactions with the two
3d -sigma-antibonding electrons in the high-spin Mn(II) ligand field63. It follows that the barrier to water dissociation
is low, and that exchange of inner sphere waters with those in the media is very rapid, with residence times in the
submicrosecond range64. For most imaging applications, properties of the protons in the water molecule, the most
abundant ions in biological systems, are measured. Thus, the low barrier ligand exchange property of high-spin Mn(II)
has been exploited in magnetic resonance imaging. Water exchange rates are altered by tissue properties that produce
Uselman et al. Longitudinal MEMRI
differences in MR signal intensities of water protons and give anatomical information, as we document in this review.
Note: We write Mn(II) in the text, tables, and most figures to emphasize that neither the structure nor the composition
of the metal complex in biological media is known. We write Mn2+ only for the six-coordinate aquo ion, as in this case
both the composition and the overall charge on the metal complex are known
Very little work has been done on Mn(III) for MR imaging. Not surprising, because the aquo ion is a very powerful
oxidant, and even in acidic aqueous media is not stable. Only in a chelated environment can water exchange be
measured. Interestingly, axial water exchange in Mn(III) porphyrins is as fast as the reaction in aquo Mn(II), which
requires a special explanation65. Normally, water dissociation from metal(III) ions is much slower than from
corresponding dipositive centers, as would be expected from simple electrostatic considerations. Axial Mn(III)-
OH2 bonds in high-spin Mn(III), however, are exceptionally labile, with exchange rates in the nanosecond range,
owing to severe sigma-antibonding interactions in the tetragonal ligand field63. In low-potential chelates that can cross
the blood brain barrier and have functional groups that target desired brain tissues, then Mn(III) could be a champion
MR imaging agent.
The aforementioned ligand exchange property of Mn(III) makes it a very promising candidate for MR imaging. In one
study of note, Sims et al. employed a 3T MRI system to test the suitability of sulfonated Mn(III)-5,10,15-
tris(pentafluorophenyl)-corrole (called 1-Mn) as an MRI contrast agent66. Notably, 1-Mn exhibited high relaxation
rates as the Mn(III) concentration increased. 1-Mn at an accumulation of 1 mM in tumors could be clearly identiﬁed in
the T1-weighted MR image of the living mouse. Moreover, 1-Mn when encapsulated into the tumor-targeting/cell-
penetrating protein, HerPBK10 (called HerMn), accumulated in breast tumors preferentially. Originally identified as a
receptor tyrosine kinase that confers poor prognosis on 30% of breast cancers, Her2 has now been found in many other
cancer types. HerMn exhibited an increased T1 relaxation rate and decreased relaxation time in tumor regions, thereby
enhancing the tumor contrast in the T1-weighted MR image. It was found that contrast enhancement of tumor regions
in the MR mouse image was 20% higher for the HerMn conjugate compared to 1-Mn. These results, which indicated
that HerMn can facilitate tumor-selective imaging, demonstrated that HerMn possesses great potential as an MRI
contrast agent for detection of Her2 positive tumors67. Current clinical imaging typically detects 0.5-1 cm or larger.
Thus, an ability to detect tumors at 1 mm diameter would be a major advance for finding and treating breast cancer
recurrence, since newer therapies can halt progression at earlier stages when metastases are smaller.
3.2. Safety considerations. Previous concerns surrounding safety delayed exploration of Mn(II) for MR imaging of
neural physiology. In recent work re-visiting this potential problem, fewer Mn(II) side effects were found than some
earlier reports had suggested; and, notably, in other early work, Alan Koretsky found strong T1 signals at doses far
below those previously reported in toxicology studies of LD5068,69. Subsequently, virtually all MEMRI investigators
focused on determining safe doses and monitoring side effects. Multiple studies by many different laboratories have
confirmed safety of aqueous MnCl2 doses sufficient to produce robust MR signal intensity when delivered
intraperitoneally (IP), intravenously (IV), or subcutaneously (SubQ) for systemic distribution, or directly into the
vitreous of the eye or into precise locations within the brain in preclinical studies (Tables 1-3). Also exciting are
human studies that have begun with chelated Mn(II). One chelate in particular, mangafodipir (a.k.a. Teslascan), which
is Mn(II) in a complex with dipyridoxyl diphosphate (DPDP), although FDA approved in 2003 for MR imaging in
humans, was originally not commercially successful, despite showing usefulness in retinal imaging in rats70. Very
recently mangafodipir demonstrated success for imaging normal human brain71 and for detection of lesions in humans
with multiple sclerosis72. We emphasize that a safe dose of unchelated MnCl2 for humans has not yet been determined. !
Various chelates of Mn(II) also may be useful for imaging vasculature and digestive systems in human
MR26,27,35,36,73,74, but these applications have been overshadowed by gadolinium-based compounds74. Chelated Gd(III),
Uselman et al. Longitudinal MEMRI
not Mn(II), is most commonly used for human medical imaging36. While un-complexed Gd(III) is known to be very
toxic75, Gd(III) chelates remain in the vasculature and thus have become standard contrast agents for imaging
vasculature. In the brain, Gd(III) chelates are routinely used for diagnosis of invasive brain tumors that break the blood
brain barrier76. When not chelated, Gd(III) would be expected to be substantially more dangerous than Mn(II)
dichloride75. And indeed, LD50 values for various chelates depend on their relative affinities for Gd(III)77. Recent
reports of nephrogenic systemic fibrosis (NSF) and retention in the endothelium in the brain78 after Gd(III) imaging
have led to proposals to revisit manganese chelates for safer vascular imaging36.
Because manganese has a wide variety of industrial uses and its ore is abundant worldwide, effects of exposure in the
workplace have received much attention, with over 450 publications listed in PubChem and dedicated pages on EPA,
OSHA and CDC websites. These guidelines are intended to protect workers by regulating predominantly respiratory
workplace exposures to manganese dust. Manganism, a neurological disability resembling Parkinsonism, may arise in
humans exposed for years to mixed manganese dust above 0.5 mg/mL379. While previous Material Safety Data Sheets
(MSDS) from Sigma Aldrich, a supplier of Mn(II) dichloride and MnCl2-tetrahydrate, were reproduced in reviews of
MEMRI40,50,62, the supplier, now Merck, has replaced this safety list with the statement "To the best of our knowledge,
the chemical, physical, and toxicological properties have not been thoroughly investigated" (Safety Data Sheet V. 6.5,
revised 1/15/2020 (Product number 244589 Aldrich, Millipore SiFTDgma). The earlier Sigma Datasheet had relied on
toxicology reports from the 1960s. This new statement does not indicate that Mn(II) is entirely safe. While PubChem
lists toxicity studies for Mn(II), many citations date from early in the last century and the methods used to acquire data
are not easily retrievable. LD50 concentrations reported in PubChem are many-fold higher than used for MEMRI. In
contrast, careful work on MR imaging using uncomplexed Mn(II) as a contrast agent has defined safe levels showing
little evidence for acute or chronic side effects attributable to uncomplexed Mn(II) at the low doses that produce good
MR signals, although various side effects are found at higher doses (see Table 1-3). Caution must be taken in older
mice as there is some suggestion that age may increase vulnerability to side effects of Mn(II), as it also increases
Mn(II) uptake in the hippocampus80. At least one transgenic, SERT-KO, is more vulnerable to systemic Mn(II)18. Care
also must be taken to monitor dosage, by observing effects on cardiac function, behavior, weight and even
electrophysiology in the living animal, as well as post-mortem histology after stereotactic injections looking for
inflammation and necrosis. When doses are kept low, safety becomes less of a concern.
While Gd(III) is not normally found in the body, Mn(II) is a required participant in biological systems. Normal
manganese ion concentrations in the human body differ among biological tissues evaluated. Normal manganese
concentrations are about 4–12 µg/L in whole blood, 1–8 µg/L in urine, and 0.4–0.85 µg/L in serum81. Because
manganese ore and other manganese products have been in industrial use for over 200 years, an extensive literature on
chronic high dose exposures exists. Prolonged respiratory exposure to high amounts of dust from manganese ore in
miners was originally found to result in a variety of disorders including Parkinsonism, first described by Couper in
1837, while the actual oxidation states of exposure and presence of other contaminants were not considered. Past
exposures to manganese ore were an order of magnitude higher than modern exposures in developed countries;
therefore, the clinical syndrome seen in the time of Couper is no longer typical of modern Mn-exposed workers82,83.
Importantly, safe levels for manganese exposure in the workplace are being re-written today. Consequences of Mn(II)
exposure are dose-dependent, and thus safe doses can be defined.
Mn(II) is delivered in two ways for MEMRI: Systemically by delivery into the circulatory system for brain-wide
imaging of neural activity, and locally by stereotactic injection. Systemic delivery involves the highest amount of
Mn(II) per body weight (Table 1), while localized injections use very little but at higher concentrations (Table 2 and
3). Thus, systemic Mn(II) could carry the highest risk for Parkinsonism or other brain and organ damage, while local
injections may alter electrophysiology, induce inflammation or cell death. Hence, an evolving question has been
Uselman et al. Longitudinal MEMRI
whether enough Mn(II) can be safely delivered for imaging. When calculating Mn(II) dose in weight, the compound
being weighed is critical. While some publications report dose in mg/kg body weight, often whether the weight (mg)
refers to the molecular weight of MnCl2 or its hydrate, MnCl2-4H2O, is not specified. We include calculations for both
molecular species in Tables 1-3.
For systemic delivery, Koretsky and coworkers, who considered neural and cardiac toxicity, found that low doses of
Mn(II) produce robust MR intensity changes without demonstrable side effects in mice68. To obtain long-term elevated
systemic Mn(II) safely, two strategies have been employed: fractionated doses given either IP or IV at intervals to
obtain a total dose in rat of 180 mg/kg MnCl2-4H2O (a.k.a. Mn(II) chloride tetrahydrate, also termed manganese
dichloride tetrahydrate). To reduce stress on the heart, single doses of 90 mg/kg were delivered by slow IV infusion
into the tail vein, whereas lower doses were IP-delivered as a single bolus84. Side effects, including weight loss, were
noted acutely at higher doses, while delivery of the same total amount in six doses at 48h intervals resulted in
equivalent intensity patterns in the brain without side effects. In rats, manganese treatment via chronically implanted
osmotic pumps (11.4 mg/kg/24h for 7 days (MnCl2-4H2O) also did not produce detectable adverse effects on voluntary
running, food or water intake or body weight; all these parameters being quantitatively analyzed. In contrast, a single
SubQ or IP dose of this amount of MnCl2-4H2O resulted in abrupt and long lasting side effects, such as a decrease of
voluntary running activity already at 3h post-injection; and weight loss 24h post-injection85. In mouse, continuous
SubQ infusion by osmotic pump was shown to improve contrast and reduce or eliminate side effects, such as weight
loss86, decreased muscle strength and elevated cortisol87. Continuous osmotic pump delivery over 15 days of up to 50
mg MnCl2-4H2O/kg/day (total dose 750 mg MnCl2-4H2O/kg) had no effect on spatial learning but did cause skin
ulcers at the pump site88. Two IP injections separated by 9 days of 0.3 mmol/kg of aqueous MnCl2-4H2O (59 mg/kg)
had no effect on behavior in the light-dark box at 24h post-injection18. Similarly in rats, two IP injection of 44 mg/mL
MnCl2 after a one-week interval had no effect on retinal electrophysiology or other evidence of injury in the eye89.
In early work on the visual system, MnCl2 was injected into the vitreous of the eye22,90,91 (Table 2). Bearer et al. tested
the dose/volume effect on electrophysiology in the visual system by measuring visual evoked potentials22. A low
volume at low concentrations of Mn(II) (0.125µL of 50 mM MnCl2 ) depressed visual evoked potentials (VEP)
initially with full recovery at 24 h. In contrast, a larger volume with a higher Mn(II) concentration (0.5 µL of 200 mM)
depressed VEPs completely both acutely and at 24 h, which resulted in loss of axons in the optic nerve. A volumetric
effect of 0.5 µL of saline on acute VEP was also noted, which recovered fully at 24 h. Intravitreal Mn(II) injections
may produce some acute injury at all dosage levels, while topical administration may be safer and yet generate
sufficient signal along the optic nerve92-94.
While stereotactic local injections deliver a much lower total amount of Mn(II) compared to systemic applications,
higher amounts are present at the injection site (Table 3). For stereotactic injections into brain parenchyma, effects of
MnCl2 concentration, pH and volume are critical in diminishing side effects, evidenced by reactive gliosis and
neuronal cell injury in histologic sections of the injection site95. Since 20 nmol of Mn(II) induced local tissue damage,
Canals et al. tested 1, 4, 8 and 16 nmol of MnCl2 with injection volumes ranging from 20 to 160 nL from stock MnCl2
concentrations of 50 and 100 mM in 10 mM Tris–HCl pH 7.3. These concentrations produced minimal to no tissue
damage95. In a series of tract-tracing publications, we found a single injection of 3 nL aqueous 200 mM MnCl2
produced no gliosis or inflammation at 3 weeks after injection22-24,96-99, including absence of reactive microglia by Iba1
In summary, Mn(II) concentrations that produce statistically significant hyperintense signals do not result in tissue
injury by electrophysiology, histopathology, or immunology, and, importantly, have no deleterious effect on behavior,
and no acute or chronic cardiac or neurological effects in preclinical studies. It should be noted that the studies cited
Uselman et al. Longitudinal MEMRI
here used MnCl2, and there may be a contribution from the chloride ion to tissue effects. Other Mn(II) salts exist with
partners that have less biological activity than chloride, such as Mn(II) acetate (Mn(CH3CO2)2, but these have yet to be
explored. Important considerations in addition to the total amount of Mn(II) delivered are timing of delivery (slow,
fractionated), the concentration of Mn(II), the volume injected, and the pH and osmolality of the Mn(II) solution.
Numerous clinical trials conducted in humans in the USA and internationally have demonstrated minimal side effects
for chelated Mn(II), mangafodipir (a.k.a. Teslascan)100, which is why it received FDA approval.
4. METHODOLOGY FOR MANGANESE-ENHANCED MAGNETIC RESONANCE IMAGING
4.1. Contrast agents for MR. Contrast in MR images arises from at least three factors: local concentrations of species
being imaged (e.g., 1H, 31P, 19F); their local spin-lattice relaxation rates (R1); and their local spin-spin relaxation rates
(R2). Typically, water protons are the species being imaged. Here this discussion will be limited to 1H MRI where
contrast agents are used to modulate the R1 and/or R2 of nearby water molecules. Thus, contrast agents fall into two
classes: those that primarily affect R1 and those that primarily affect R2. The R1 contrast agents are most often
paramagnetic compounds of Gd(III) and Mn(II) that enhance 1H relaxation rates, resulting in brighter signals when
detected by particular pulse sequences. The relaxivity per ion of Gd(III) and Mn(II) can be quite large, ranging up to
80 mM-1s-1 73,101. The R2 agents are represented by superparamagnetic compounds, such as iron oxide particles, that
enhance R2 relaxation times resulting in strong signal losses102-106. We note that in his seminal 1973 publication7,
Lauterbur indicated that alterations in T1 relaxation rates produced by MnSO4 can be used to alter contrast in an image.
The reason that researchers might choose to measure T1 versus a T1-weighted signal is that if measured properly, T1
should only depend on the field strength and chemical environment of the water protons. So, the main advantage to
making T1 maps for MEMRI is that the results would be comparable across studies at the same field strength, the
values would not be influenced by the choice of pulse sequence, and it would not require signal intensity normalization
or correction for B1 non-uniformity (see Section 7). The disadvantage is the increased imaging time. Knowledge of
the relaxivity of a contrast agent allows its local concentration to be determined via measurement of changes in water
proton T1 (1/R1) values107. In practice, this measurement is seldom done for in vivo brain applications of MEMRI for
two principal reasons: First, T1 measurements (T1 maps) require increased imaging time than simple imaging; and
second, Mn(II) relaxivity needs to be determined for the brain locations of interest. Simply assuming Mn(II) relaxivity
in tissue is the same as in buffer is not sufficient, because pH, viscosity, macromolecular concentration etc., in tissue
will affect that relaxivity. Thus, MEMRI studies generally use straightforward measurements of the local increase in
image intensity with a T1 weighted imaging protocol108-110. Nevertheless, T1 mapping is still being used successfully
for both systemic uptake and tract-tracing102,108,111,112.
4.2. Pulse sequence considerations. There are many and varied T1 weighted protocols, each with particular sets of
parameters to be determined for the specific hardware employed. Imaging parameters are optimized under the
constraints of time, required spatial resolution and desired signal-to-noise and contrast-to-noise ratios (SNR &
CNR)113,114. MEMRI has been performed over a range of magnetic field strengths up to 11.7 T with varying types of
gradients and RF coils. Best whole brain signal-to-noise ratios (SNR) often are obtained at higher field strengths with a
small volume RF coil encompassing only the head, although surface coils for mouse heads are becoming available that
may improve detection in the cortex and other superficial structures. Another good setup for MEMRI in rodents is to
use a larger RF volume coil for transmission to improve B1 homogeneity, then a small, phased array of surface coils
for sensitive detection.
Fast spin-echo or gradient-echo sequences with short repetition times (TR) and/or large flip angles (FA) are often
utilized, but many more sophisticated methods are available, e.g., fast/turbo spin echo (FSE), T1-weighted fluid
attenuated inversion recovery (T1-FLAIR), spoiled gradient echo (SPGR), and magnetization prepared gradient echo
Uselman et al. Longitudinal MEMRI
(MP-RAGE). Although many sequences in use in MEMRI also contribute T2 and/or T2* weighting, they are employed
to minimize acquisition time. Imaging parameters are optimized under the constraints of time, required spatial
resolution and desired SNR and CNR113,114.
4.3. Evaluation of images for optimization of scanning parameters and power analysis.!Longitudinal!imaging of
the living brain requires collection times as short as possible. Rodent studies typically aim for 100-200 µm voxel
dimensions where resolution is fine-grained enough to delineate most brain nuclei if SNR & CNR allow115. If an
experimental goal does not require minimal imaging time with high isotropic resolution, direct T1 measurements can
be a more sensitive alternative approach. For instance, Fontaine and colleagues used T1 maps with MEMRI for
identification of early stage neurodegeneration in aging and tauopathy112. While this protocol was able to limit imaging
time to 45 min, 3D resolution was sacrificed, illustrating the tradeoff between scan time and voxel dimensions. So, if
the goal is to obtain high 3D resolution images of brain-wide activity in the shortest time, T1-weighted imaging is often
preferable to T1 mapping.
For MEMRI there is no standard formula for setting pulse sequences for T1-weighted images to gain the greatest
contrast enhancement from the Mn(II)116. Many parameters depend on the hardware: field strength, radiofrequency
coil, gradient characteristics and other features of the particular scanner. Optimization of the collection parameters to
acquire the greatest contrast from the Mn(II) at the highest resolution within the shortest imaging time requires the
usual implementation of an appropriate T1-weighted protocol and adjusting TE, TR, flip angle, the number of averages
and setting the field of view just large enough to encompass the whole brain115. That parameters have been optimized
for the Mn(II) contrast is best monitored by an image-based approach, where Mn(II)-enhancement resulting from
specific parameter sets for that unique scanner and its specific hardware can be measured. Optimization of parameters
should aim for the highest CNR to distinguish Mn(II)-enhanced regions while maintaining good SNR. Typically a non-
Mn(II) enhanced image is compared to a Mn(II)-enhanced image of the same animal. The parameters that give the
greatest SNR and CNR difference between these two conditions may be considered optimal.
Once the imaging parameters are optimized for the best Mn(II)-enhancement, an imaging study can be designed,
including a power analysis to determine the numbers of samples needed to obtain adequate statistical power. Since
most MEMRI publications omit mention of this step, we give some guidance here (note that grant reviewers often ask
for details of such analyses). Power is the probability of detecting a true association in contrast to the decision error
when one mistakenly rejects the null hypothesis (α, commonly set at 5% or 1%). Each research hypothesis is tested
using a statistical model that requires a separate calculation of predicted sample size for a pre-specified α and desired
power for the study (typically 80% or higher)117. Additionally, the mode of analysis (region of interest (ROI) or voxel-
wise) and any planned correction for multiple comparisons should be considered when selecting the α level. When the
effect of interest is the degree of Mn(II)-enhancement over baseline, images of a set of animals captured before and
after Mn(II) will provide pilot data for estimating sample size from a power analysis. Intensity measurements should
be acquired from regions of interest in three areas: those less visually enhanced by Mn(II) (muscle, thalamus), those
with greater enhancement (hippocampus, olfactory), and a non-tissue region (for noise). If a particular region is under
investigation, that region should also be measured. Selection of the ROI can also be based on an expected projection
(for tract-tracing), measuring at the same location in a series of longitudinal images22. All ROIs must be adjusted for
noise by dividing each measurement by the standard deviation of ROI measured in an empty, non-tissue region of that
image. A power analysis from a pilot experiment should use input images that have undergone the same pre-processing
pipeline as planned for the fully powered study. Normalization of both spatial and intensity scales prior to the power
analysis not only helps with accuracy, but it is also critical for interpretation of MEMRI results. We discuss
normalization in Section 7.
Uselman et al. Longitudinal MEMRI
In a paired t-test the null hypothesis proposes that there are no differences in intensities between the paired pre- and
post-Mn(II) images. To find the minimal sample size required to disprove the null at desired power (1-β) of 0.8 or
more and confidence level (1-α) of 0.95 or higher (set a priori), the effect size of these differences must first be
estimated117. Differences between paired pre- and post-Mn(II) ROIs/images for each animal are determined, then the
average and the standard deviation of those differences are calculated. Values from these calculations can be entered
into various web-based calculators such as http://powerandsamplesize.com or
http://statulator.com/SampleSize/ss2PM.html118,119. Statistical software, such as Python, R, SAS/STAT, SPSS, etc.,
also provides functions for power analyses and sample size estimation.
Typically this type of power analysis predicts that 10-15 animals imaged longitudinally are needed for a paired t-test to
reach a power of 0.8-0.95 at a 95% confidence level for difference in intensity before and after Mn(II), delivered by
either localized stereotactic injections or systemically in the expected locations18,23,24,96-99,120-122. Similarly, these
calculations predict sample sizes for a voxel-wise SPM paired t-test, with clustering and smoothing. When images are
captured from different cohorts with and without Mn(II), an unpaired t-test or ANOVA is necessary and sample size
needed to reach statistical significance is increased. So, in addition to the benefit of each animal serving as its own
control, another advantage of longitudinal imaging is the smaller number of animals needed for statistical power.
5. APPLICATIONS OF MEMRI: TRACING NEURONAL PROJECTIONS AND AXONAL
TRANSPORT DYNAMICS BY LONGITUDINAL IMAGING
5.1. Rationale and background: Role of axonal transport in tracing and stereotactic injection for localized
Mn(II) delivery. Investigators in Koretsky’s lab pioneered biological applications of MEMRI in the late 1990s. First,
Yi Jen Lin, in her thesis69 and publication68, demonstrated activity-dependent uptake of Mn(II) in the brain that could
be imaged by T1-weighted MRI. The following year Robia Pautler, also a member of Koretsky's team, completed her
thesis and published the first report of Mn(II) neuronal tracing from the nares into the olfactory system and from the
eye into the visual system90. In this section, we review the latest development in MEMRI-based tracing of neural
projections. In the next section (Section 6), we discuss systemic MEMRI for volumetric comparisons during
development, aging and neurodegeneration, and the emerging concepts and results from activity-induced manganese-
dependent imaging of brain-wide neural activation. We note that manganese-enhanced MRI has also found
applications in cardiac and tumor imaging123-127.
Tract-tracing or projection mapping with contrast agents, whether by optical microscopy or MRI, depends on the
endogenous neuronal axonal transport system (Fig. 3). In this context, tract-tracing refers to highlighting projections
from neurons in one region to their distal projections somewhere else, which differs from MR-based "tractography",
which uses diffusion anisotropy in diffusion-tensor images, reporting on fiber tract anatomy but not axon
directionality. Axonal transport is critically important for neuronal function, and even survival128. Transport carries
crucial components to and from the synapse, including molecules involved in sustaining and growing neuronal
connections, signaling molecules that keep the neuronal cell body informed, and the synapse supplied with
neurotransmitters. The axonal transport system depends on three components: microtubules that form "tracts",
molecular motors that translate chemical energy into force, and cargo that are packets of subcellular materials that
associate with motors for transport. Microtubules are polymers of tubulin, a protein with two isoforms, which form
tubules with a diameter of 24 nm and lengths that can extend for millimeters129. Kinesin-1, the first microtubule-based
motor to be identified, was discovered in the giant axon of the squid130, a powerful experimental system for axonal
biology where vesicular transport was originally observed131-133. Subsequently, kinesin-1 was found to be a member of
a much larger family of highly conserved microtubule-based motor proteins that perform multiple key cellular
functions: chromosomal separation in mitosis and meiosis, cell division, and partitioning of components throughout
Uselman et al. Longitudinal MEMRI
eukaryotic cells134. Use of MEMRI tract-tracing in the optic nerve and the hippocampal-forebrain projections revealed
the role of kinesin-1 in the central nervous system of the living animal for the first time22,24, as detailed below.
Clues to the cargo-motor receptor, which remained a mystery for two decades after the discovery of kinesin-1, came
from biochemical studies135-137 and genetic work in Drosophila138-140. The specific mechanism of this cargo transport
was finally determined through experimental work on the squid giant axon at the Marine Biological Laboratory in
Woods Hole, MA. When injected into the intact axon, fluorescently labeled, negatively charged polystyrene beads,
similar in size to transport vesicles, travel anterograde towards the synapse (Fig. 3A1)141. Using chemically modified
polystyrene beads covalently bound to synthetic peptides injected into the axon and imaged by time-lapse confocal
microscopy, Satpute and coworkers identified a peptide zipcode within amyloid precursor protein (APP, whose
proteolytic product, Aβ, is a major component of senile plaques) that was sufficient to bind cargo to kinesin for
anterograde axonal transport142 and was more efficient than mere negative charge141. Further work with transgenic
animals demonstrated that at least some Mn(II) transport in axons within the brain depends on this transport system.
After entering active neurons, locally injected Mn(II) secondarily enters subcellular compartments including those that
are transported, such as mitichondria, the endoplasmic reticulum, and other membrane-bound organelles (Fig. 3A2).
This secondary uptake into intracellular transport vesicles may depend on DMT, which unlike voltage-gated Ca(II)
channels is found in intracellular membranes143. If a large amount of Mn(II) is locally delivered, enough is taken up to
be imaged at long distances from the injection site, for example from the eye to the midbrain (Fig. 3A3).
Early work demonstrating this transport, a feature of Mn(II) not predicted by previous work, employed the optic nerve
and the olfactory system for development of the technology and for an understanding the underlying biology90,91,144-146.
When injected into the eye, Mn(II) enters retinal ganglion cells. Axonal projections from these cells form the optic
nerve and project to midbrain structures, superior colliculus and lateral geniculate (Fig. 3A3). Because of its large size
(>1 mm in diameter), the optic nerve is readily visualized in MR images of living rodents 22,90,91. All of the axons in the
optic nerve arise in the retinal ganglion cells and project towards the midbrain. Hence all transport from the ete
towards the midbran is "anterograde" with respect to the neuronal cell body and the microtubule polarity, and thus
mediated by anterograde moledular motors, i.e., the kinesins.
Localized injection, where the position of the injectate is carefully controlled, i.e., "stereotactic injection" (Fig. 3B1),
can deliver sufficent Mn(II) for long-range tracing. Early work showed Mn(II) lit up known projections from the nasal
passages into the olfactory system as far as the anterior olfactory nucleus (Fig. 3B2)90, and from the vitreous of the eye
to the midbrain90,91. These reports led to studies of Mn(II) transport from the amygdala of the living mouse brain to its
projections147, and between the song learning centers in the songbird (Fig. 3B3)148-153. In the songbird, MEMRI
mapped the seasonal changes of the central control of song and effect of testosterone on the vocal control system and
song learning centers (Fig. 3B3)151-153.
To interpret and extend these exciting studies, it was necessary to learn more about the biological and chemical basis
of Mn(II) transport in living axons--was transport of Mn(II) mediated by the kinesin-based axonal transport system?
Two groups reported that colchicine, a microtubule-destabilizing agent, decreases Mn(II) distal accumulation,
suggesting a microtubule transport mechanism that would invoke kinesin154,155. In 2007, tackling this question by
genetic means, the Bearer and Jacobs team tested kinesin-based axonal transport by measuring the rate that Mn(II)-
induced intensity increased along the optic nerve by scanning at 6 min intervals longitudinally--a MEMRI "time-lapse"
study as the time between image captures is short, with the animal remaining in the scanner for a single session (Fig.
3C1)22. A mouse with a deletion in a kinesin-1 subunit (kinesin light chain 1, KLC1) displayed delayed transport of
Mn(II) compared to wild-type littermates in the optic nerve (Fig. 3C1). This result, the first direct evidence of kinesin-
1's contribution to central nervous system transport in the living animal and evidence that Mn(II) transport was
Uselman et al. Longitudinal MEMRI
mediated by kinesin, was further confirmed by decreased transport of Mn(II) in hippocampal to forebrain projections
in these KLC1 knock-out mice24. Thus at least some component of the Mn(II) transport along neuronal projections
depends on microtubules and the microtubule-based anterograde transport motor, kinesin-1. However, knockout of
KLC1 did not fully abolish Mn(II) transport, so other mechanisms must operate as well, including engagement of other
kinesins such as Kif1a (Kinesin-3), also abundant in axons, or even unconventional myosins156. Unfortunately,
knockout of these motors is embyronic lethal, so as yet no genetically altered experimental system is available to test
APP was also found to be involved in Mn(II) transport (Fig 3C2)97. Thus, MEMRI detects alterations in axonal
transport in animal models of Alzheimer's disease (AD), particularly models carrying the dominantly inherited
mutations of human APP, such as the APPSwInd transgenic98,99, Tg2576 with APPK670N,M671L mutation155, or mice
carrying human mutations in tau protein linked to several neuropathologies157. Transport in CNS neuronal projections
in other neuropathologies are aggressively being investigated with MEMRI techniques, such as frontal-temporal
dementia158,159, Parkinson's disease160 and other forms of neurodegeneration40,161. Rates of Mn(II) transport from the
nares into the brain were delayed in an AD mouse model155 and the effect was reversed by anti-Aβ treatment with R-
florbiprofen162 but not by gamma-secretase inhibitors25. Insight into the physiological mechanisms of these diseases
enabled by MEMRI tract tracing has provided new insights for prognostics and therapeutics. Future work will likely
pursue these diseases, as MEMRI reports on APP-dependent transport, and as transgenic mice replicating features of
AD promise to provide valuable experimental systems for mechanistic understanding and drug discovery. To assist in
future work by investigators new to MEMRI, protocols describing techniques for tract-tracing are now available163,164.
5.2. Trans-synaptic tracing. Whether a track is mono- or polysynaptic is an important consideration when performing
Mn(II) tract-tracing or interpreting projections highlighted by Mn(II) (Fig. 3D). Intracerebral Mn(II) injections into the
amygdala revealed accumulatons in hippocampus, septal nucleus, entorhinal cortex, subiculum, medial habenulla and
fornix147, and injections into locus coeruleus demonstrated distal accumulation different from known projections165.
These findings suggect polysynaptic tracing. By tracing transport with MEMRI in blind compared to sighted mice, it
was discovered that Mn(II) preferentially accumulates at distant synapses when the pre-synaptic neuron is active22
(Fig. 3D). This finding was interpreted as indicating that Mn(II) preferentially traces functional (i.e., electrically
active) poly-synaptic circuits. The amount of accumulation of Mn(II) in post-synaptic targets can be measured as a test
for trans-neuronal dysfunction166. Thus MEMRI provides information about active projections but may miss
anatomical connections in silent circuits, and accumulation may be proportional to the degree of activity within the
circuit and not its anatomy.
5.3. Applications: Hippocampal projection tracing.
5.3.A. Example 1: Rates of Mn(II) accumulation in distal hippocampal projections: Down syndrome: Down
syndrome, the most common genetic cause of human developmental disability167,168, is a consequence of inheriting
three copies of chromosome 21 (trisomy). A transgenic mouse trisomic for 13.4 Mb of DNA syntenic with human
ch21, Ts65Dn, has deficits reminiscent of human Down Syndrome169-173. These mice have decreased cholinergic
neurons in the medial septal nuclei174-177, the target of hippocampal CA3 projections178-182. Abnormal transport within
these projections is thought to contribute to cholinergic neuron survival and cognitive dysfunction183, yet whether
axonal transport from the hippocampus to the medial septum was affected by the trisomy had not been tested.
Subsequently, longitudinal images of Mn(II) transport dynamics demonstrated that transport is more robust in trisomic
Ts65Dn mice than in their normal diploid littermates, as evidenced by increased accumulation of Mn(II) in the basal
forebrain after CA3 injection in Ts65Dn compared to littermates120. Images were captured at successive time points
(0.5 h, 6 h, and 24 h) after stereotactic injection into CA3 of the hippocampus. To detect Mn(II)-dependent intensity
increases along the hippocampal projections, resultant images (n = 12 for each genotype) were analyzed by statistical
Uselman et al. Longitudinal MEMRI
parametric mapping, comparing intensity values between time points (Fig. 4). This revealed differences in the rate of
Mn(II) accumulations between the two genotypes, which was interpreted as being due to an altered physiology of
transport, although altered anatomy could also play a role. In subsequent studies, anatomical variations between
several other genotypes were more thoroughly investigated23,24,99,184.
5.3.B. Example 2: Rates of Mn(II) accumulation in distal hippocampal projections: Alzheimer's disease.
Effect of amyloid precursor protein (APP) mutations: Hippocampal projections to septum and basal forebrain are
large fiber tracts that course along the fimbria to the fornix where they dive downwards. The diameter of these fiber
bundles renders them detectable with 100 µm voxel resolutions, especially when highlighted by Mn(II). Taking
advantage of this effect, rates of transport within these projections were measured in a series of images captured before
and at successive time points after Mn(II) stereotactic injections into CA3 of the hippocampus (Fig. 5). Initially,
addressing the biological basis of Mn(II) transport, the Bearer-Jacobs team followed up on the Down syndrome story
by exploring Mn(II) transport rates in other mutant mice. Knock-out of APP, one of the triplicated genes in Down
Syndrome, resulted in decreased transport in both hippocampal and retinal projections23, further suggesting that APP
mediates a proportion of Mn(II) fast axonal transport (See Fig. 3 above). These studies suggested that Mn(II) tract-
tracing and transport dynamics may yield insights into Alzheimer's disease, since mutant APP is a known risk factor
for dominant inherited Alzheimer's disease; in addition, previous studies speculated that transport defects underlie
cognitive impairments as well as neuronal survival139,142,185. In a double transgenic mouse expressing a mutated form of
APP found in dominantly inherited AD, APPSwInd, under control of the Tet-off system186, transport of Mn(II) was
altered both in rates and location of accumulations (Fig. 5)98,99. Aging affected the Mn(II) distribution at 24 h after
stereotactic injection99 (Fig. 5C and D). Toggling expression of APPSwInd to decouple plaques from expression
provided the beginnings of a mechanistic understanding of its effect on transport (Fig. 5E). Cohorts of mice with
plaques when expression was repressed compared to those with expression without plaques, and to those with both
expression and plaques showed that mutant APP expression alone in the absence of plaques, disturbs transport98.
Plaques alone, a condition replicating sporadic AD, also disturbed transport, but differently. Caveats that were
examined included location of injection (Fig. 5A), variation in amount of Mn(II) injection at these tiny volumes (3-
5nL), although the amount injected in all cases saturated the uptake system, and normalization artifacts. These studies
support previous work in the olfactory system which also showed decreased transport with aging187. A limitation of
Mn(II) is its unknown uptake and clearance rate in these neurons, which could affect distal accumulations. New
mechanistic understandings from these results are: 1) Axonal transport decreases during normal aging; 2) Plaques
alone replicate the condition of sporadic AD with transport defects independent of mutant APP expression; and 3)
Mutant APP affects transport rate and location of post-synaptic targets independent of plaque.
5.4. Applications: Forebrain projections and reward circuitry.
5.4.A. Example 3: Prefrontal cortical projections: Pyramidal neurons of the prefrontal cortex (PFC) project deep into
subcortical, limbic, and brain stem structures. Their activity is thought to be regulated reciprocally by monoaminergic
systems (noradrenergic, serotonergic and dopaminergic) that arise in these distant regions, since the PFC is heavily
innervated by these systems. MEMRI was applied to determine whether deregulation of monoaminergic systems altered
functional anatomy of PFC projections (Fig. 6)96,97,188. As an example, in the NET-KO mouse, Mn(II) (3-5 nL of 0.6 M
MnCl2) was stereotactically injected into the PFC (x = 0.59 ± 0.13 mm (R of midline), y = +0.72 ± 0.49 mm (bregma), z = -
1.04 ± 0.36 mm (deep) (Fig. 6A)97, a medial region recently named the anterior cingulate area (ACA) 189. MR images were
captured before injection and then at successive time points over the following 24h. Images captured 0.5 h after injection
served to locate the actual injection sites. Monoamine systems were altered by genetic disruption of transporters for each
(NET, DAT and SERT/5-HTT) (Fig. 6B). Littermates for each knockout were imaged in parallel, for a total of 72 animals
each imaged at 4 or more time points, yielding more than 288 images in total. The image of each brain was extracted from
the whole head image of the living animals. Gray scale normalization and alignments were performed and the progress of
Mn(II) accumulation from injection site into the brain determined by voxel-wise paired t-tests yielding statistical
Uselman et al. Longitudinal MEMRI
parametric maps between successive time points. Results were published in a series of three papers96,97,188. While these
genetic alterations are lifelong and thus may have unknown secondary effects, each genetic disruption produced a
different signature of altered mPFC projections compared to its littermate wild type non-transgenic or congenic
animals. Because these projections are diffuse and not a fiber bundle, as in the hippocampal projections, the Mn(II)
accumulation is also less focal. It is also not clear whether these differences reflect anatomical changes or altered
neural activity in these projections. The Bearer-Jacobs team addressed these questions in two ways: 1) with diffusion
tensor imaging contrasting wild type littermates with their respective knock-outs; and post-mortem histopathology.
While no anatomical differences could be identified in the SERT-KO, the NET-KO displayed some significant
anatomical changes. A next step will be an inducible conditional disruption of the NET gene, such as through a floxed
NET mouse mated to a transgenic with Cre driven by a noradrenergic neuron-specific promoter (M. Hahn, personal
communication). The central noradrenergic system arising from the locus coeruleus in the brain stem is lost in
Alzheimer's and other dementias190-192. This system is also critically important for mental health and addiction. Hence
more knowledge of its actions through MEMRI applications will contribute essential information to many critical
questions of brain science.
Alterations in these monoaminergic systems due to genetic disruption of their respective reuptake transporters result in
altered projections of one of their downstream targets, PFC-ACA pyramidal neurons. Whether these alterations are
directly due to the effect of disruption on monoamine reuptake in the adult, to formation of projection anatomy during
early development or to secondary effects on other processes remains to be explored. Furthermore, whether such
alterations can be induced in the adult by experience or drugs is an open question. Knowing how other types of more
controllable genetic manipulations, such as conditional inducible knockouts or chemogenetic activation/inhibition of
monoaminergic systems triggered at a specific time in specific neurons alter brain-wide activity patterns as detected by
MEMRI, will help us attribute brain-wide neural activity to molecular/genetic functions.
6. APPLICATIONS OF MEMRI: LONGITUDINAL IMAGING FOR BRAIN-WIDE ACTIVITY-INDUCED
6.1. Mn(II) enters via Cav1.2, an L-type voltage-gated calcium channel, and Mn(II)-induced increased T1-signal
intensity is correlated with, and proportional to, neuronal activity. Chemical similarity to Ca(II) provides Mn(II)
the unique ability to enter excitable cells via voltage-gated calcium channels in an activity-dependent manner. Early
work by London et al. reported Mn(II) uptake in rat brains after a single IP injection by MRI. However, they did not
experimentally associate MR signal changes with neural activity, but rather began to characterize Mn(II) distribution in
the brain193. In 1990, Narita et al. found that tetanic stimulation of frog motor neurons immersed in Ca(II)-free Mn(II)-
containing buffer, where Mn(II) was the sole dipositive ion present, produced miniature end-plate potentials194. In the
presence of a calcium channel blocker, verapamil, these potentials were reduced194. Nothing more was done until 1997,
when Lin and Koretsky reported these combined properties of Mn(II) in hopes of directly measuring neural activity by
MR68,69. From these experiments, they demonstrated for the first time that Mn(II) accumulation occurs in the brain
after neurostimulation with the excitatory neurotransmitter, glutamate (Fig. 7A)68,69. In these pioneering studies,
MnCl2 was directly injected into the brain within the scanner and immediate effects detected within minutes. These
investigations provided the first instance of activity-dependent accumulation of Mn(II) and ultimately led the group to
coin the term, "Dynamic Activity Induced MEMRI,” (DAIM)69. Lu et al. followed up on stimulation within the
scanner using cocaine19,20, while Jimmy Bell's lab investigated other appetite-controlling compounds on Mn(II) uptake
after peripheral injection within the scanner195,196. However, the DAIM approach of stimulations within the scanner has
not been widely used. Instead, systemic Mn(II) was shown to highlight brain regions in the normally behaving rodent,
with the most intense signal found 24 h or more after Mn(II) delivery, depending on route of administration68,107,109,197.
In 2005, the Turnbull lab used this understanding to map auditory stimuli-evoked activity in awake behaving mice198.
Many studies, too many to enumerate here, have since demonstrated the utility of systemic MEMRI, to capture neural
Uselman et al. Longitudinal MEMRI
activity induced by a variety of stimuli, often adapting the Turnbull approach. Some examples will be described and
Recently, investigators have begun to take advantage of clever experimental systems to further address remaining
questions on the activity-dependent nature of Mn(II) accumulation. For instance, Bedenk and colleagues were able to
pinpoint specific mechanisms of Mn(II) accumulation using a genetic mouse line lacking different L-type calcium
channels in the brain199. These mice helped them reveal the L-type calcium channel 1.2 (Cav1.2) as a primary mode of
Mn(II) entry into cells, as MR signal intensities were reduced only in Cav1.2 lacking mice (Fig. 7B and C). Still,
Cav1.2 may not be the only mechanism of Mn(II) entry into active neurons, since some enhancement occurs in its
absence, and the relative health of the knock-outs argues for some voltage-gated calcium entry into neurons.
Another recent contribution came from combining electrophysiology of single neurons with MEMRI200. In this study,
neurons from isolated Aplysia buccal ganglia displayed Mn(II)-MR signal enhancement after electrical or chemical
stimulation (Fig. 7D-F)200. Quantitative measures showed correlations between the numbers of action potentials,
stimulated either electrically or by a neuroactive drug (dopamine) with T1-weighted intensity increases. Thus, the
degree of T1-signal increase is directly proportional to the number of action potentials.
These examples demonstrate approaches that can be deployed for future studies seeking to uncover remaining
biological mechanisms of activity-dependent entry of Mn(II) into neurons, accumulation and clearance. Some
questions remain: For example, is Mn(II) transported along projections to distant locations after being taken up by
active neurons after systemic delivery? For Mn(II) to be taken up and then transported in sufficient amounts to be
detected is hard to imagine, but some evidence suggests that Mn(II) does accumulate in nearby post-synaptic neurons
from regions with high uptake199. Further, is Mn(II) accumulation proportional to neural activity in all brain regions
equally? The cortex is commonly less highlighted by Mn(II) than brain stem regions. What is the basis for this
difference? And finally, do differences in vascular perfusion influence Mn(II) brain-wide distribution? Since uptake
occurs over 24h, it is generally assumed that by that time point after systemic delivery, interstitial Mn(II)
concentrations have reached equilibrium and would be uniformly distributed throughout the brain parenchyma.
However, testing this point may require use of vasoactive drugs, such as acetazolamide for vasodilation and
vasopressin for vasoconstriction, coupled with MEMRI.
6.2. Dose and timing considerations for longitudinal MEMRI after systemic administration. Continuous imaging
experiments conducted by Koretsky and co-workers demonstrated early on that Mn(II), delivered IV, begins to
distribute in the brain within minutes68,201 and dissipates around 14 days from the rat brain (Fig. 8A)109. Since then,
many studies have characterized Mn(II) distribution in the brain with a variety of dosing regimens18,84-88,109,199,202,203.
Some of these studies have been listed in Section 3.2 (Table 1), where investigators have carefully monitored safety.
In this section, we will discuss Mn(II)-dose responses for developing longitudinal MEMRI experiments.
For systemic delivery, the Mn(II)-enhanced signal first becomes apparent in the cerebral spinal fluid of the ventricular
system, then extensively distributes throughout the brain (Fig. 8A)109. While early work suggested that disruption of
the blood brain barrier (BBB) may be necessary, it was later shown that Mn(II) enters the murine brain over 24 h
without the need to compromise the BBB204. Analysis of the distribution of Mn(II) between plasma, interstitial and
intracellular compartments, a.k.a. "compartment modeling", during DAIM focused on the pituitary, which has an
incomplete BBB205. By measuring plasma Mn(II) concentrations and comparing with MR signal in the pituitary using
the Patlak and Logan graphical analysis, Leuze et al. reported quantitative measurements of changes in Mn(II) passage
into the interstitium and subsequent entry into active neurons following pharmacological modulation of calcium
currents and of neural activity. It would be useful going forward to perform compartmental modeling in regions of the
brain where the BBB is complete to determine what proportion of the Mn(II) signal arises from which compartment
Uselman et al. Longitudinal MEMRI
and whether all such brain regions are equivalent. Deeper understanding of the dynamics of Mn(II) interstitial
concentrations and any regional variations would contribute valuable information for interpretation of neural activity
Single doses of Mn(II) by three different routes have been studied: Intraperitoneal (IP), intravenous (IV) and
subcutaneous (SubQ). After a single dose, the maximum signal occurs 24 to 72 h after administration, as demonstrated
by significantly shorter T1 relaxation times in grey matter over this period18,107,109,197,206 (Fig. 8B and C). Mn(II)
delivered by IP injection may enter the ventricles and the brain more slowly than by IV delivery, as evidenced by
significantly shorter T1 relaxation times in the ventricular system compared to pre-injection at 24h, as compared with
IV delivery where ventricular signal returns to pre-injection levels more quickly197 (Fig. 8B). This slight delay after IP
compared to direct IV delivery could be a consequence of the time needed for Mn(II) transit from peritoneal cavity
into vasculature. While a "peak" of Mn(II)-enhanced intensity occurs between 22-48 h in mouse, even 19 days after
delivery Mn(II) signal remains slightly above pre-Mn(II) levels in many locations in the mouse brain (Fig. 8C).
Although this residual signal may represent some of the Mn(II) detected by more sensitive autoradiography using
54Mn(II)60,61, it is thought that the metal ions detected by autoradiography might be stably sequestered by complexing
with enzymes. Such sequestered Mn(II) does not affect the relaxivity of protons, which is established by injections of
radioactive Mn(II). Such stably sequestered Mn(II) also would not be available to enter into activated neurons. Several
studies have compared MEMRI of Mn(II) with 52Mn(II) imaged by PET207-209. Little 52Mn(II) is detected in brain after
IP injection when co-injected with Mn(II)208,209. These reports underscore the importance of capturing an image before
each successive Mn(II) dose to detects residual signal when performing longitudinal imaging experiments. Although
Mn(II) may enter cells through DMT, which is not voltage dependent, this is not likely to be an important avenue for
Mn(II) uptake in brain16.
Maternal IP injection of MnCl2 has been shown to result in sufficient uptake and enhancement for both fetal (in
utero)210 and neonatal211 brain imaging. These exciting reports led to the potential for future MEMRI studies of mouse
Longitudinal tumor imaging by MEMRI by repeated Mn(II) IP and IV doses has been shown to be more useful than
Gd-DOTA to detect growth of metastases213 and to identify primary brain tumors induced genetically in rodents127,214.
Increased Mn(II) signal was expected due to increased Ca(II) pathways of cancer cells that would not be detected by
chelated Gd(III). This approach has recently been applied as the first MEMRI-guided preclinical drug trial for
medulloblastoma, which showed that inhibition of tumor-associated macrophages attenuates tumor progression in the
Sonic-Hedgehog-medulloblastoma mouse model215.
Other methods, such as slow continuous infusions or multiple fractionated doses, allow larger doses of Mn(II) without
risk of acute side effects18,84,88,202,203,216. Application of these methods for longitudinal studies is of particular interest
because the continuous availability of interstitial Mn(II) for uptake into newly activated neurons allows experiments
testing the effect of experience or drugs on brain activity18,121. Mn(II) distribution kinetics, even for single doses, allow
researchers to exploit maximum signal enhancement measuring activity dependent changes before and after
neurostimulation or experience.
Investigators employing longitudinal MEMRI need to carefully consider the timing, degree, and mode of Mn(II)
dosing in relation to delivery of neurostimulation and behavioral experiences (Fig. 8D). Some studies have imaged at
4-6h, before Mn(II) intensity has reached its peak. Our group has used the extensive Mn(II) distribution around and
after 24 h to observe both maximal signal enhancement of basal uptake and to map acute changes occurring due to
neurostimulation18,121. This timeline indicates a pre-injection image for baseline, non-Mn(II)-dependent intensity, and a
post-injection image for Mn(II)-dependent intensity present prior to the stimulus. To mitigate the possibilities of non-
Uselman et al. Longitudinal MEMRI
stimulus-specific signals from brain activity not related to the experimental stimulus, both stimulus and post-stimulus
imaging should begin soon after pre-stimulus imaging and stay within the broad peak of Mn(II) brain
Limitations of systemic Mn(II) for neural activity include low temporal resolution, such that MEMRI represents an
integrated composite of activity occurring between Mn(II) delivery and imaging. So, regions with low levels of activity
or with intermittent activity may not be identified. Safety concerns may preclude saturating the uptake system.
Residual Mn(II) from previous injections needs to be accounted for when doing repeated administrations, as Mn(II)
clearance is slow. Despite these limitations, systemic MEMRI offers the best opportunity currently available to witness
brain-wide neural activity coupled with behavioral features.
6.3. Systemic Mn(II) and Immediate Early Gene Expression. Immediate early gene (IEG) expression detected by
immunohistochemistry or in situ hybridization for RNA transcripts provides an alternative read-out of neural
activity217-219. Systemic Mn(II) does not activate IEG expression indiscriminately (Fig. 9A)121. However, locations with
neural activity detected by MEMRI correspond to those where IEG are expressed, either egr-1 or c-fos18,121,220. This
correspondence is nearly perfect for SPM of MEMRI images and c-fos expression in the olfactory bulb (Fig. 9Bi-iv),
the anterior olfactory nucleus (Fig. 9Bv) and the dorsal raphe nucleus (Fig. 9Bvi). In this case, the same mice (n = 12)
were first imaged by MR, then fixed, embedded, sectioned and stained for c-fos by immunohistochemistry. Both c-fos
expression and Mn(II) uptake depend on calcium currents, and thus c-fos is a particularly useful marker for the same
neural activity detected by MEMRI. Staining for c-fos thus serves as an alternative to electrophysiology for validation
that MEMRI signal represents neural activity, and can be performed in the same brains when fixed after MR scanning.
However, MEMRI is less sensitive than c-fos, and thus c-fos activity may be detected where no MEMRI signal is
found. It follows that MEMRI allows for in vivo experiments similar to c-fos experiments, i.e., integrating calcium
currents over time.
6.4. MEMRI versus fMRI/BOLD: Limitations and benefits. As a functional marker, systemic MEMRI has some
advantages and disadvantages for imaging brain-wide neural activity compared to fMRI/BOLD221,222. When comparing
cortical activity imaged by BOLD with that of MEMRI under continuous imaging during Mn(II) infusion, activity
patterns essentially match221. However, while BOLD imaging must be performed in the scanner, Mn(II) can be used to
label active neurons in awake behaving animals during performance of a task. Activity patterns correlating to the
behavior are retrospectively imaged. Some additional advantages include Mn(II) being more of a direct report of
neural activity, while BOLD is a hemodynamic signal that secondarily reports on O2 metabolic demand presumed to
arise from active neurons. Such secondary signals are likely to be affected by vascular dynamics independent of
neuronal activity. The degree of BOLD signal may be dependent on the volume of brain regions occupied by
capillaries in addition to neuronal O2 demand, confounding interpretation. Mn(II)-induced intensity increases appear to
be proportional to neuronal action potentials200, provided the interstitial Mn(II) concentration is consistent and all
neurons express the same type and amount of Ca(II) channels. For an as yet undefined reason, the hippocampus takes
up Mn(II) more quickly than other brain regions. Furthermore, Mn(II) gives a stronger MR intensity signal than
BOLD222. Its relative safety and robust signal allow delivery multiple times during a longitudinal experiment87,216,223 or
continuously with an osmotic pump for as long as three weeks without effects on behavior in mouse88. MEMRI can be
performed at higher resolution due to its higher CNR (even 50 µm isotropic voxels of the whole brain in living mice)
than fMRI, which typically has 250 µm voxel resolution and is often collected as slices or anisotropic slabs rather than
3D isotropic whole brain images. Since an accumulated manganese signal is very slowly changing, longer acquisition
times can be used to good advantage, unlike the fast events studied with BOLD.
That Mn(II) uptake occurs in the awake behaving animal allows anesthesia to be used only during retrospective
“readout". This eliminates worries of any major effect of anesthesia on neural activity but carries the downside that
Uselman et al. Longitudinal MEMRI
distribution is averaged over time with low temporal resolution. Recent studies have demonstrated that Mn(II) uptake
induced by stimuli can be measured during a 2-3 h window 22-26 h post IP, at the height of Mn(II) concentration in
brain18,121. In comparison, BOLD uses no extraneous contrast agent and detects neural activity occurring in the scanner
with relatively high temporal resolution (seconds to minutes). This poses a challenge for studying behavior, since
rodents cannot self-report their internal state, which is typically inferred from behavior. Furthermore, while in the
scanner, the animal is often under anesthesia, such as isoflurane and medetomidine, which interferes with the
noradrenergic system's vascular regulation224,225 as well as with CSF outflow226,227, and thus anesthesia likely affects
BOLD signals in multiple ways. For some applications, such as neural activity in response to cocaine or
methamphetamine (drugs that affect vascular dynamics)228,229, MEMRI is the method of choice.
MEMRI detects cocaine-, oxycodone-, sound-, acute threat-, psychotropic drug and odor-induced
activities19,62,121,198,230-233. A complete picture of neural activity dynamics in many different situations will likely require
several different imaging approaches, with both MEMRI and BOLD as well as optical techniques.
6.5. Applications: Mn(II)-enhancement for analysis of volumetric changes across lifespans
Previous work on development using MRI without contrast reported brain development involves dramatic volumetric
changes234,235. Similarly in aging, multiple parameter MRI without contrast also found volumetric and structural
changes236. Systemic MEMRI greatly improves anatomical contrast for analysis of brain volume changes. MEMRI can
be performed longitudinally over short and long time periods without deleterious effects210,211,237,238. MEMRI for in
utero brain development revealed volumetric changes in embryonic structures210, and illuminated changes in neuronal
density in small brain regions, such as the developing granular cell layer of the cerebellar cortex during post-natal
development238. For finer brain structures, it is important to consider the relative contribution of Mn(II) uptake versus
neuronal density. A 4D atlas of post-natal mouse brain development from PD1-11 was prepared using Mn(II) delivered
to the pups via their dam211. Each of these MEMRI studies demonstrated the absence of adverse effects on
development, using safe doses commonly used for adults. In Szulc et al., controls not treated with Mn(II) and imaged
at 21 days showed no differences from treated pups, demonstrating the absence of adverse effects from Mn(II),
anesthesia or imaging protocols on development. Results revealed interesting dynamics of regional brain volumes
during development as various brain structures changed dramatically.
In one case of aging, a longitudinal MEMRI study of volumetric changes examined early brain atrophy in a transgenic
Alzheimer's model mouse, where following both wild-type and transgenics over-time determined statistical age x
genotype interactions, yielding details on the timing of abnormalities over the lifespan237, . Following volumetric
changes from post-natal period into old age with MEMRI could identify alterations in both gross brain volume and fine
gray matter structure, as fluctuations occur throughout the lifespan. Repeated brief exposures to systemic Mn(II) across
a lifespan were reported to have no side-effects on neuroanatomy or survival in wild-type or even in the more fragile
Alzheimer's model mice237. Anatomical regions with complex foliations patterns such as the cerebellum, other regions
that undergo structural changes over the lifespan are important sources of anatomical variation. We will consider these
challenges in Section 7 below.
6.6. Applications: Sensory systems.
6.6.A. Examples: Longitudinal MEMRI for the auditory system: Systemic MEMRI has been applied to understand
various sensory systems. Perhaps the first study to apply systemic Mn(II) to witness brain activity changes in response
to stimuli over time comes from work in Daniel Turnbull's lab studying the effect of sound on midbrain neural activity
(Fig. 10). These longitudinal MEMRI experiments revealed a tonotopic map for two pitches in the inferior colliculus
(IC) (Fig. 10A and B)198,231. Mice exposed to a broad band width of noise (20-50 kHz) for 24 h after Mn(II) IP
injection and then imaged were re-injected a second time 48 h after the first injection, exposed to a specific pitch (40
kHz) and re-imaged 24 h after the second injection (i.e., 72 h after the first injection). This revealed different locations
Uselman et al. Longitudinal MEMRI
of Mn(II) accumulation at the two time points collected longitudinally. Exposure to two different pure pitches (16 kHz
and 40 kHz) in separate cohorts of mice demonstrated Mn(II) accumulation in different anatomical locations within the
IC (Fig. 10C and D)--thereby demonstrating for the first time that a tonotopic map exists in midbrain prior to
transmission to auditory cortex. This organization of IC occurs during the first three weeks of life, as documented in
these studies, again for the first time using MEMRI. This major contribution may underlie maternal-pup
communications necessary for survival239.
6.7. Applications: Longitudinal MEMRI for brain-wide activity mapping over time before and after
experiences: How the brain responds over time to ethological stimuli, neurodegeneration or addiction towards
recovery or progression is a major quest across all areas of neuroscience. MRI, because it is non-toxic and non-
destructive, holds the promise of fulfilling this pursuit. Until now, rodent behavior has been employed as a proxy for
emotional states and neural activity in the brain and is a poor read-out. Much excitement has attended new
experimental procedures involving electrode arrays and optogenetic stimulation. Sadly, these techniques are invasive
and give only a narrow window on the brain, thus lacking the global picture of integrated brain activity possible with
Much of the focus of MEMRI investigators has been technological. However, a number of applications were designed
with the goal of witnessing events in the live animal that had previously been described by other means in fixed and
processed tissue. These applications should now be followed with longitudinal imaging, which cannot be done with
methods requiring sacrifice. While solving the narrow window of microscopic and electrophysiological approaches to
obtain whole brain images, fixing specimens for MEMRI does not provide dynamic information about processes, but
only about the moment in time when brains were fixed. Systemic Mn(II) followed by fixation and imaging has been
used to show brain-wide responses to pain240, hippocampal changes after early life adversity241 and brain circuitry
involved in response to acute novel stress in rats242. Longitudinal studies of brain-wide progression in living mice
across time are rare.
Only very recently has MEMRI begun to be deployed to address questions of how brain-wide activity evolves over
time under experimental conditions, such as confrontation with predator odor, evolution of a disease state or of
substance abuse. Evolution of disease processes in neurodegenerative diseases, such as Huntington's243 and
Alzheimer's40,112 witnessed by MEMRI, have been reported although much more needs to be done. Opioid addiction
studies in rats have primarily used fMRI-BOLD instead of MEMRI although the problems of movement within the
scanner during drug withdrawal despite anesthesia have led in one instance to application of MEMRI, as this can be
performed retrospectively when deeper anesthesia is possible without affected the activity pattern233.
6.7.A. Example: Effect of acute fear and serotonin on brain-wide activity over time: A critical issue for
understanding the brain is how an acute experience plays out over time. Vulnerability to substance use disorder, mental
illness, and possibly even dementia accrue to victims experiencing a life-threatening fear event. Yet how such an event
impacts brain activity and whether that impact persists, evolves or resolves to baseline remains unknown. Longitudinal
MEMRI in experimental animals, where the pre-threat condition can be captured, and the sequence of brain activity
followed non-invasively over time offers a chance to peer into the brain and witness this critical process. Uselman et
al. 202018 developed an experimental procedure to capture response to acute threat and its outcomes (Fig. 11). While
this protocol provokes brain activity with a predator odor, other stimuli are possible, including drugs121. The key is that
Mn(II) concentrations clear from the brain slowly, such that enough remains at 24-26 h to enter newly activated
neurons. Thus both a pre-exposure and post-exposure image of Mn(II) enhancement can be obtained from the same
animal and compared.
Uselman et al. Longitudinal MEMRI
This experimental design allows us to quantify brain-wide activity as it evolves from before a threat to 9 days later, the
time when wild type mouse behavior is known to return to baseline. By witnessing the whole brain, mapping activity
with statistical parametric mapping, and parsing its individual segments with a new InVivo Atlas, Uselman et al.
202018 revealed the balance of activity between 86 sub-regions for the first time (Fig. 12). The degree of intensity
change in statistically significant regions was confirmed with region of interest measurements. A mouse with genetic
disruption of the serotonin transporter (SERT) is known to be especially vulnerable to acute threat and does not return
to baseline activity at 8-9 days244. This SERT-deficient mouse allowed the investigators to compare the normal acute
threat response with a response that fails to resolve, an "anxiety-like" state. Not surprisingly the SERT mouse had
more activity at baseline, with further activity induced by predator odor, which progressed over time to a different
anatomical pattern, but still at a heightened level. As detailed above, secondary effects of a life-long knockout
confound attribution of this effect specifically to the SERT gene deletion. Although this SERT-KO provides a
complicated and extreme example of persistent fearfulness after an acute life-threatening event, this mutant mouse may
yet provide insights otherwise not obtainable about the brain responses that lead to persistent fearful behavior.
Interestingly, amygdala activation occurred after activity in the bed nucleus of the stria terminalis. The areas with
greatest differences between genotypes, such as striatum and ventral pallidum, paraventricular nucleus, ventral
tegmental area, midbrain, dorsal raphe and pontine reticular nucleus, may encompass neural network(s) that correlate
with fear persistence as can be predicted from assembly of multiple optogenetics and electrophysiological studies245,246.
The patterns of activity may also correspond to networks identified with independent component analysis of mouse
fMRI/BOLD images224,247-249. Especially noteworthy in this MEMRI study is detection of activity in deeper brain
regions of the limbic system and brain stem, regions less accessible than cortical regions to more invasive
electrophysiological approaches, or as easily imaged by fMRI-BOLD.
Further computational analyses of resulting images will yield "network" information, yet whether that will be
confirmative of fMRI-BOLD or additive is unclear. Unlike fMRI-BOLD, MEMRI is not affected by anesthesia as the
Mn(II) uptake is imaged retrospectively. And because MEMRI signal is stronger than the BOLD signal and MEMRI
imaging session can be longer, spatial resolution is improved. Hence we predict that networks uncovered by MEMRI
will be more granular than those described by BOLD and reveal new information about the contribution of smaller
regions than can be defined by fMRI. However, both techniques will probably be required to fully define the integrated
brain-wide activity that produces vulnerability to acute threat, or the evolution over time of other neuropsychiatric
7. STATISTICAL AND COMPUTATIONAL ANALYSES
The enormous benefit of longitudinal imaging whether across different scanning sessions or within a single scanning
session, is that it reduces the sample size needed to reach statistical significance and uses each mouse's own images are
its controls. These advantages are best realized during computational image analysis. To exploit this benefit, images of
each individual within a cohort captured at successive time points must be brought into a standardized image format,
with normalization of intensity and anatomy. Although many of these steps are standard nowadays, MEMRI analysis
poses unique challenges for normalization, which we discuss here. Although many approaches developed for human
MR imaging may be repurposed for rodents, this can sometimes also be tricky even without MEMRI challenges.
When performing multiple processing steps investigators should keep the actual underlying biology in mind and
examine each step for possible introduction of non-biological signals or signals not relevant to the hypothesis. One
way to do that is to obtain behavioral measures interleaved with the MEMRI scanning. ROIs validate SPM results by
determining the degree of signal enhancement. Confirming measurements on un-processed raw images is also useful.
Below we summarize the benefits and unique challenges posed by normalization and analysis of longitudinal MEMRI
images of mouse brain and describe a few ways to surmount them.
Uselman et al. Longitudinal MEMRI
7.1. Pre-processing of images. Preprocessing involves a series of steps that transform a set of MR images from raw,
whole head images fresh off the scanner to more refined, standardized images that can be spatially aligned, intensity
normalized for statistical analysis. In 2003, a series of steps for normalization of MR images was compiled into an
automated pipeline by the Toga group250-253. Other methods for mouse brain image analysis were emerging at the same
time, including application of human image analysis to mouse96,188,254-258. Standardization of image processing would
facilitate analysis of large cohorts and multi-center studies259, especially now that mouse brain MR image repositories
are coming online, such as "open neuro" (http://openneuro.org) at the Stanford Center for Reproducible Neuroscience.
A number of processing environments exist which are continuously being updated, such as AFNI (Analysis of
Functional NeuroImages)260, ANTS (Advanced Normalization Tools)261, AIDAmri
(https://github.com/maswendt/AIDAmri)262 and COINS-TReNDS (https://coins.trendscenter.org). However, while
even the more recent AIDAmri that promises a complete processing environment, including alignments to an
annotated atlas, none of these pipelines consider the unique challenges for MEMRI. A stepwise protocol specifically
for MEMRI preprocessing is available from Medina et al. 2017, in Current Methods in Molecular Biology, which
includes a 3D T1-weighted Mn(II)-enhanced template image. The Medina protocol can be performed locally after end
user installation of multiple freeware scripts and is based on command lines rather than a graphical user interface263,
and thus is transparent and readily adapted.!
7.1.A. Correction of intensity non-uniformity: Before performing any processing, image headers must be identical
with images oriented similarly. Typically images are converted to NIfTI format (Neuroimaging Informatics
Technology Initiative)260,264. A first step in normalization is elimination of intensity inhomogeneities arising from
image acquisition265,266. This correction is commonly referred to as N3 correction, (non-parametric, non-uniformity
normalization) and updated to N4ITK267. For MEMRI, this step should be performed on the raw whole head image, as
performing bias field corrections after skull stripping effaces some Mn(II) signal18.
7.1.B. Automated brain extraction (Skull-stripping): For accurate alignments of the brains in a dataset to enable
statistical precision, the brain must be isolated from non-brain tissue in which anatomical structures dramatically differ
between individuals, such as nares and teeth. With longitudinal imaging datasets containing dozens, or even hundreds
of images, as required for statistical power, manual extraction, such as BrainSuite268,269 and fsleyes from FMRIB
Software Library, FSL 270,271, is prohibitively time-consuming. The benefits of automated compared to manual
extraction are (1) decreasing time required, and (2) increasing consistency across datasets, which in turn improves
MEMRI in rodents poses two challenges for automated brain extraction. First, the hyperintense Mn(II) signal interferes
with intensity-based brain extraction, which is especially notable in brains after stereotactic injection. Second, the
shape of the mouse brain is different from human, being elongated rather than spherical and having little distance
between skull and brain, limiting repurposing of human extraction procedures.
Unfortunately, most methods for skull-stripping developed for human fail both challenges, as these methods use
intensity patterns to identify boundaries between brain and skull or other segmentation-type strategies and are based on
the more spherical human brain shape268-270,272. Other programs applicable to rodents similarly use intensity-based
segmentations, including AFNI 3dSkullStrip260, three dimensional pulse coupled neural networks (3D-PCNN)273,
Rodent Brain Extraction Tool (RBET)274, Rapid Automatic Tissue Segmentation (RATS)275, SHape descriptor selected
Extremal Regions after Morphologically filtering (SHERM)276, and deep learning programs that use a U-Net
architecture277,278. One such U-Net skull-stripping program was developed specifically to address distortions unique to
blood-oxygenation-level-dependent (BOLD) imaging that occur in the setting of lower spatial resolution and
susceptibility-induced distortions that may occur during BOLD imaging278. Brain extraction procedures are further
reviewed by Feo and Giove, 2019279,280.
Uselman et al. Longitudinal MEMRI
SkullStrip, developed specifically for rodent MEMRI by the Bearer lab in 2015281, is a template-based brain extraction
which does not rely on intensity patterns within the brain of the Mn(II)-enhanced image. A whole-head reference
image!of a non-enhanced image is registered to a new whole-head image from the experimental dataset captured prior
to Mn(II) injection. The control point grid from that anatomical registration is then applied to the template mask
created for that reference image. For each Mn(II)-enhanced image of that same animal captured in separate imaging
sessions, its unique pre-Mn(II) mask is rotated and applied, thereby avoiding the influence of hyperintense Mn(II)
signal on the mask. Automation of brain extraction via SkullStrip accelerates brain extraction by processing multiple
images in a matter of minutes once a template mask has been created. The software and instructions for SkullStrip can
be found in Delora et al. (2016)281.
7.1.C. Gray scale normalization: A critical step for MEMRI analysis is normalization of gray scale intensity values
between T1-weighted images, since this is the parameter affected by Mn(II) that will be used to report transport or
neural activity. Many different approaches exist to adjust all images in the dataset to the same scale263,282,283. Note that
the gray scale must be normalized consistently across all images in the dataset, which is critical for subsequent
statistical analysis of Mn(II) signal intensity changes. Early work used a three-step global normalization process to
remove intensity drift between images captured at different time points of the same animal284. These three steps
involve 1) setting all images to an intensity scale of 1000 and averaging all images from the same animal; 2)
Normalizing all scans from each animal to their average image; 3) Matching intensity of averaged image sets for each
subject to intensity of a common MR template removing inter-subject differences in global intensity. A scaling factor
was then applied to these individual normalized image sets that had been corrected for intra-subject drift in intensity.
Other studies used internal references, such as the pons, to normalize between animals at each time point, such as
Bertrand et al. 2013159. For region of interest analyses, normalized intensities were plotted and fitted over time to a
tract-tracing bolus model159,187. Another method employs an external phantom placed next to the head during all scans
and the intensity of this phantom is used to normalize intensity across the images, since the phantom's intensity in
relation to that of the tissue is assumed to be constant across time points and samples285. Another approach is to choose
an ROI in the image believed to be unaffected across time and samples (e.g., muscle outside the brain)109,286 and to
determine a scaling factor to normalize intensities across all images. In the case of stereotactic injection of Mn(II)
where MR intensity changes are highly localized, the mode of the histogram of a pre-Mn(II) MR image will reflect
intensity of unenhanced regions that will be constant across time and samples120. In this study Bearer and colleagues
chose one wild-type pre-injection image as a template and all 52 images (all animals at all time points) were spatially
registered to it using an affine transformation and interpolated to 50 µm voxel size. The average of aligned pre-injected
images from the wild-type group and from the genetically altered group were each calculated and used as template
images for intensity normalization for their respective groups, removing bias due to selection of any individual image
as template. Thus scaling the histogram of all voxel intensities in each post-Mn injection image to the mode of the
average pre-Mn(II) template image's histogram normalizes intensities across all post-Mn(II) images in a
7.1.D. Alignments: For MEMRI images, two potential issues arise in the anatomical alignments needed for
subsequent statistical analysis. First, spatial registration could carry a risk of degrading intensity information due to
interpolation287,288. However, this risk is minimized when aligning congenic mouse brains due to their anatomic
similarity, but greater when aligning brains with a high degree of anatomical differences, such as occurs in some
transgenics263, during development210,211,238 or aging237, and in some brain regions such as foliations in the cerebellum,
as discussed further below. In the case of mutants with consistent anatomical differences from wild type, such as the
KLC-KO mouse263, a better strategy for statistical comparison is to align within each group. Intensity measurements
obtained from atlas-based ROIs separately aligned to each group can be used to compare between groups when
anatomical differences exist263. Use of more computationally intensive interpolation methods, such as B-spline, further
Uselman et al. Longitudinal MEMRI
serve to retain the original distribution of intensities between the voxels288-290. Creation of a minimal deformation
“target” image (MDT) from the dataset also helps to reduce the degree of transformation needed for each image291.
A second concern is potential misalignment of pre- and post-Mn(II) images acquired in different scanning sessions
when post-Mn(II) images have hyperintense signal that could bias alignment algorithms. However, with longitudinal
imaging when the same animal is imaged at each time point, this becomes a non-issue as long as there is no anatomical
change between time points. This possibility can be mitigated by capturing a pre-Mn(II) image at each time point18
and/or a T2-weighted anatomical image, which is mostly unaffected by Mn(II)292. These images can be used to
generate a control point grid to align post-Mn(II) images to other time points. For further circumvention of potential
distortions or misalignments due to localized hyper-intense signal from stereotactic injections, a two-step, indirect
alignment approach can be deployed99. In this process, pre-Mn(II) images are first aligned linearly to their MDT
image, and then post-Mn(II) images of the same animal’s brain are aligned to their respective MDT-aligned pre-Mn(II)
image. Linear rigid body alignment (with 6 parameters), used in these first alignment steps, has minimal effect on
voxel arrangement or relative intensity between adjacent voxels288,289. Rigid-body registration also eliminates potential
misalignment due to hyperintense signal since gross brain features, such as orientation and location, are the core of the
algorithm and not local signal intensities270,293,294. These images can then undergo non-linear alignment. This second
step is best carried out by an indirect process since local signal intensities in the post-Mn(II) images may influence
non-linear alignments. In this process, subject-specific control point grids are generated from a non-linear alignment of
pre-Mn(II) images to the MDT image263. These grids are then applied to each animal’s respective post-Mn(II) image,
which were previously linearly registered to their respective pre-Mn(II) image. This approach results in satisfactory
alignment of post-Mn(II) to the same animal’s pre-Mn(II) image, with no distortions due to Mn(II)-enhanced
intensities, and with little to no loss of intensity information due to interpolation. See Medina et al. (2017)263 for more
details on this process. Alternatively averaged pre-Mn(II) images can be aligned to a higher resolution atlas template
and then the inverse of that deformation applied to the template atlas295,296. Once all post-Mn(II) images are aligned to
the averaged pre-Mn(II) image, the atlas template will also be aligned with all.
Anatomical differences pose problems that interfere with spatial alignment, a problem not unique to MEMRI which
needs brief mentioned here as the increased contrast in MEMRI may increase detection of these differences.
Anatomical regions with complex foliation patterns, such as cerebellum and developing brain, are significant sources
of such anatomical variation as are some genetic alterations that cause anatomical abnormalities, such as deletions of
KLC-1 in aging mice24, or Nkx2.1 in prenatal stages210. During post-natal stages this problem has been tackled
effectively211,238 as well as in comparisons of male and female mouse brains212, in tumor growth127 and in
neurodegeneration where volumetric analyses are performed237 or when intensity patterns are statistically
compared97,99,263. In many cases, investigators have aligned genotypes separately, either to different reference
atlases210 or to different template atlases created from within each genotypic cohort 24,96,97,120,188. Results are either
projected side-by-side or superimposed on a gray scale template for visual assessment of anatomy. No statistical
conclusions can be made from such visual assessments, since statistical between-group comparisons are not possible
when differing anatomy introduces alignment artifacts. However, strong conclusions are justified when the images are
intensity scaled to the same template and the statistical threshold for each within genotype pairwise t-test is the same.
7.2. Statistical analysis: Region-of-interest (ROI) measurements and Statistical Parametric Mapping (SPM).
MEMRI, much like human fMRI, identifies locations of neural activity. Quantifying the statistical significance of
intensity differences between pre- and post-Mn(II) is important for attributing intensity to Mn(II) uptake, while
comparing signal between time points reveals effects of stimuli or disease process.
Once images have been preprocessed, they are ready to undergo statistical analysis. Two primary ways most often
used to compare MEMRI intensities within or between cohorts are ROI measurements and unbiased voxel-wise
Uselman et al. Longitudinal MEMRI
statistics. While statistical methods for unbiased voxel-wise comparisons such as SPM are often limited to pairwise t-
test analyses for reasons discussed below, ROI measurements can be analyzed by linear mixed-effects models which
take into account correlations. SPM is an unbiased voxel-wise detection of the statistical significance of intensity
differences. In contrast, ROIs are typically candidate regions where the degree of intensity change is measured. In
some MEMRI studies, ROI is needed in addition to SPM to determine biological relevance of statistically significant
differences or to compare between cohorts when pairwise t-tests cannot be applied. Since ROIs are pre-selected
regions, an SPM locating regions that differ significantly may best target ROIs for intensity measurements.
7.2.A. Region of interest analysis: MEMRI enhancements have frequently been measured using "region of interest"
(ROI) analysis, which are pre-selected areas expected to take up Mn(II) and display T1 intensity increases compared to
those that do not. To compare the signal within an anatomical structure or neuronal track, a small area can be selected
and its coordinates propagated across all images in the dataset, best done after images have been rigorously aligned,
although in some reports investigators used the image anatomy compared to an atlas or directly aligned to digital atlas
to different images to identify areas297. The average intensity value from the voxels in these ROIs can be measured and
statistically compared. Programs such as Fiji/ImageJ and those in the FMRIB Software Library may be used to
automate ROI measurements298,299. The intensity values can then be compared using mixed model ANOVAs or other
statistical approaches to determine the significance of intensity differences. An advantage of ROI intensity values is
that unpaired t-tests and other statistical models can be readily applied for comparisons between two genetically
different cohorts. The downside of ROI measurements is that they rely on pre-selected candidate regions and are not
unbiased (or comprehensive) whole brain comparisons.
7.2.B. Statistical parametric mapping (SPM): The Statistical Parametric Mapping (SPM) suite is a free program that
runs on the proprietary platform of MATLAB289,300,301. It performs unbiased voxel-wise statistical comparisons of
whole brain image intensity values. SPM may perform paired or unpaired t-tests or ANOVA. Typically voxel
intensities of higher resolution images are smoothed to the full-width at half-maximum of a selected Gaussian kernel,
which decreases the spatial resolution of the image. When running an analysis, a p-value threshold is selected that will
generate a brain-wide map of voxel-wise T-values for t-test analyses, or F-values for an ANOVA. Detection of
significant voxels can be limited to a cluster of voxels, such that isolated individual voxels are not mapped18. These
maps can be displayed in a variety of visualization programs with min/max thresholds set according to T- or F-values.
Overlay onto an aligned gray scale image reveals anatomy. T-tests are more commonly used, as multivariate ANOVAs
or General Linear Models have raised issues in appropriately testing model effects and following up with post-hoc
tests302. However, univariate approaches are becoming more readily applied for analysis of longitudinal imaging data.
Flexible factorial analysis also poses challenges. The complexity of these factorial models renders them unable to
partition variance accurately303,304, since any image with significantly greater signal than the others, as occurs with
MEMRI, would be the factor accounting for the most variance. There would be advantages for development and
utilization of additional models exploit the full power of longitudinal MEMRI. One such model would be a flexible
statistical model that could incorporate all parameters of interest, allowing for assessment of many possible
combinations of contrasts across a longitudinal timeline while also accounting for any potential covariation.
7.3. Emerging strategies for functional brain network identification with MEMRI. Multi-scale network analysis
of brain activity is a widely used approach to illuminate modular neuro-architecture based on BOLD imaging305-310 but
has yet to be applied to MEMRI. Such network analysis from rodent fMRI has revealed a number of resting state
networks224,247,249,311-313. Since MEMRI is independent of anesthesia, longitudinal MEMRI will reveal networks
responsive to experiences, drugs including those of abuse, and acute disease processes occurring in the awake
behaving animal outside of the scanner. In addition to these advantages of MEMRI over BOLD for network analysis,
MEMRI can produce higher resolution than BOLD (50-100 µm isotropic voxels compared to 100 x 250 anisotropic or
Uselman et al. Longitudinal MEMRI
greater), approximating that of neuropixels, which are invasive and have a 70 µm probe circumference314. Thus
MEMRI adds finer detail to network subunit identification with its higher sensitivity for smaller nuclei.
An annotated atlas facilitates identification of networks. Many rodent brain atlases created with various methods are
available243,248,282,315-322. An atlas based on a high-resolution MEMRI image (56 µm isotropic voxels) hand-drawn and
annotated with reference to the histological atlas from the Allen Institute for Brain Science, Mouse Brain Reference
Atlas323, is optimal for alignment with MEMRI data18,263. This atlas has more segments than others created from MRI
of fixed brain320, and higher resolution than any available commercially (Ekam Solutions, Boston, MA), but is not as
fine-grained as those based on microscopy323. While human brain atlases incorporate many individuals due to high
inter-individual variance, this is not necessary for the highly convergent anatomy of mouse brains282. When aligned
with Mn(II) enhanced images, digital atlases allow automated quantification of segment-specific intensity values or of
statistically significant changes between time points or conditions18,292,324-326. A future direction will be creation of a
multimodal atlas integrating histologic, functional imaging, electrophysiological and, possibly, behavioral information.
Segment-wise intensity values can be further processed with methods like principal component analysis (PCA) or
cross-correlations240,297. Other potential strategies yet to be applied to longitudinal MEMRI include repurposing fMRI-
type voxel-wise measurements, such as independent component analyses, which will reveal network dynamics327-330.
when! analytical! approaches! like! ICA! network! analysis,! originally! adapted! for!fMRI,! is! applied! to!MEMRI.! Other!
promising! frameworks! for! studying! networks! revealed! by! MEMRI! come! from! whole-brain! imaging! of! immediate!
early!gene!expression331,332!or!structural!covariance!networks333-335.!Further adaptation of these analytical strategies
for longitudinal MEMRI data could lead to the illumination of large-scale brain networks dynamics before, during and
after stimuli, under conditions not ethical in human studies.
Translation of such rodent brain networks to human brain activity may be more straightforward in regions with highly
conserved anatomy, such as the deeper brain structures of the limbic system and brain stem, which are better
elucidated by MEMRI than by BOLD, and best understood in human by lesions336-338 rather than by BOLD imaging
which is best for neocortical activity.
8. CONCLUDING REMARKS
So, what have we learned? First and foremost, it is apparent that Mn(II)-enhanced MRI can shed new light on cellular
(brain-wide) physiology. Unique Mn(II) imaging applications include intracerebral localized injections to study
intracellular transport dynamics in the living brain and delineation of neuronal projection anatomy, as well as systemic
delivery to reveal dynamics of retinal response to light, organization of the tonotopic map of the midbrain during
development, and brain-wide responses to acute threat and its evolution.
That said, we emphasize that manganese has not yet reached its full potential as an MRI contrast agent. Chelated
Mn(II) or Mn(III) could replace Gd(III) as an intravascular contrast agent, thereby avoiding complications of brain
retention and nephrogenic systemic fibrosis that accompany chelated Gd(III) exposure78. And Mn(II)-based imaging
goes beyond applications of chelated Gd(III), as uncomplexed Mn(II) enters cells via calcium channels and is thus a
tool for cellular physiology in the intact living organism. Numerous studies in animals have shown that Mn(II) can be
safely used at concentrations sufficient for imaging when side effects are closely monitored.
We also would like to know precisely where Mn(II) goes once it enters cells. Autoradiography of 54Mn(II) by electron
microscopy (EM) by layering silver halide films on frozen sections of brains collected after systemic delivery
conceivably could identify intracellular compartments, but only if preparation of sections can be done before
radioactive decay of the isotope60,61. Alternatively, Mn(II) cellular distribution could be determined by subcellular
Uselman et al. Longitudinal MEMRI
fractionation of brains after 54Mn(II) systemic delivery, followed by measurements of radioactive emissions from
54Mn(II), MR imaging, biochemical analysis for markers, and negative stain EM to determine subcellular
distribution339. This approach would have an additional advantage, as it will reveal how different subcellular fractions
affect the Mn(II) MR signal. A real need that many MEMRI studies point to is for a bench top assay that can measure
manganese accumulation, thereby identifying conditions that can be tested in vivo. Importantly, a bench top assay also
would allow labs without imaging ability to acquire functional information.
Another unknown is the ligation of Mn(II) when being imaged. Contrast in MEMRI is thought to be dependent on the
composition of Mn(II) environment, and in part represents ligation states. Thus the intensity of the Mn(II) signal is
relative to biochemical properties of the tissue. It is likely that the inner coordination sphere includes ligands other than
water, as the brain contains many small molecules and ions, as well as lipids and proteins. We suggest that Resonant
X-ray Inelastic Scattering (RIXS) could provide information about Mn(II) ligation associated with MRI images. Note
that 2p3d RIXS revealed the electronic structure of low-spin ferrous and ferric complex ions340 including their ligand
field spectra341. RIXS also successfully identified spin-allowed and forbidden d--d excitations in vanadium
complexes342. Employing 2p3d RIXS (excitation at the manganese L-edge), both spin-forbidden and spin-allowed d-d
excited states of Mn(II) could in principle be detected, allowing determination of the octahedral ligand field splitting,
which in turn would shed light on the extent of water coordination in the Mn(II) inner sphere and on the molecular
composition of its local environment.
One overarching question remains: how can MEMRI translation to humans be more ubiquitously deployed? When
taking this major step, it is our view that Mn(II) chelators that can release the paramagnetic ion selectively at specific
targets must be further developed. Teslascan/mangafodipir, a chelated Mn(II) reagent for MRI, was approved by the
FDA in 2003 but is not currently manufactured due to lack of commercial interest, despite applications in common
diseases, such as retinal imaging, where its potential usefulness was demonstrated in rats70. Like chelated Gd, chelated
Mn(II) probably would not cross the blood brain barrier and thus would remain in the vasculature. However,
mangafodipir may release uncomplexed Mn(II) ion in regions of neural activity for uptake into active brain regions in
human71. Mangafodipir safely identified plaques in human multiple sclerosis72. Application of MEMRI to non-brain
tissue, such as non-CNS organs in mouse343, cardiac muscle in human126,344, and diagnoses for cancer123,124,345, as well
as others mentioned in Section 5.1, are emerging. Analogies of brain-wide activity using functional atlases to compare
rodent networks with human networks obtained by fMRI are more likely than invasive techniques such as high-density
electrode arrays or optogenetic manipulations to lead to translation of mechanistic knowledge from preclinical to
human disease states. More MEMRI-BOLD comparisons that go beyond matching acute Mn(II) uptake with BOLD
signals in the cortex are needed to examine correspondence between "resting state" patterns obtained by BOLD with
24 h Mn(II) uptake with MEMRI with both technologies performed on the same individuals. These comparisons will
assist translation of MEMRI results to human fMRI and may further our understanding of the basis of the BOLD
Where to next? MEMRI has two unique features that can be deployed to address biological problems that have escaped
solutions by other means: Tracing functional projections and behavior-neural activity pairs. Both of these are brain-
wide, and neural activity is non-invasive. First, projection tracing after stereotactic injection of Mn(II) detects active
circuits, the impact of disease progression on their anatomy and the physiological properties of axonal transport.
Further elucidation of the impact of experience, genetics or disease processes on brain-wide circuitry in living animals
will propel the field of circuitry investigations. Second, systemic Mn(II) labels active neurons in the awake behaving
animal, which are subsequently imaged by MR without requiring sacrifice, enabling capture of behavior-image pairs
repeatedly over time. This allows behavior at different time points to be matched with neural activity patterns using
such approaches as causal inference theory, yet to be developed for MEMRI. Application of these unique features of
Uselman et al. Longitudinal MEMRI
MEMRI toward understanding trajectories of addiction is in its infancy, with only 5 publications on cocaine19,20,346-348,
and one on oxycodone233 so far. Other applications with long trajectories and behavioral correlates will also be gold
mines for MEMRI, such as the impact of early life adversity on vulnerability to post-traumatic stress349, depression,
anxiety241 or addiction; Alzheimer's progression40,112,161,350; and appetite control351. The emergence of chemogenetic
tools, such as Designer Receptors Exclusively Activated by Designer Drugs (DREADDs)352-355, will also provide
fodder for MEMRI experiments since DREADD effects on neural activity are slower than electrophysiology and best
validated by behavior. To date only a single publication reports effects of DREADD inhibition on MEMRI signal356
and none on DREADD activation.
In closing, we predict with confidence that MEMRI will soon join other key technologies for understanding brain-wide
neural dynamics, an area where chemists, physicists, neuroscientists and bioengineers will continue to innovate.
We thank the many scientists who generously sent us their publications about MEMRI during the preparation of this
manuscript. We encourage anyone whose work might have been inadvertently omitted to send us their publications for
inclusion in future articles. We are indebted to Orrin Myers, UNM Department of Family and Community Medicine,
Associate Professor of Biostatistics, for assisting us with the statistical sections. We gratefully acknowledge the
Caltech Beckman Institute Biological Imaging Center for access to the MR scanner, support from Zilkha
Neurogenetics Institute at Keck School of Medicine of USC (REJ), and funding from the Harvey Family Endowment
(ELB). Funding was also obtained from National Institutes of Health: National Institute on Drug Abuse (RO1
DA018184, REJ), National Institute for Neurological Disease and Stroke (RO1NS062184, ELB), National Institute
for Mental Health (RO1MH096093, ELB) and National Institute on Aging (P20 AG068077 ELB). Partial support
came from the UNM Brain and Behavioral Health Institute (BBHI 201920005, ELB and TWU), National Institute
of General Medical Science (P20 GM121176, ELB) and National Center for Advancing Translational Sciences CTSC
(TR001449, ELB). Work at Caltech (HBG) was supported by the Arnold and Mabel Beckman Foundation.
Uselman et al. Longitudinal MEMRI
1. Gore JC. Out of the Shadows--MRI and the Nobel Prize. NEJM. 2003;349(24):2290-2292.
2. Ogawa S, Lee TM, Nayak AS, Glynn P. Oxygenation-Sensitive Contrast in Magnetic Resonance Image of
Rodent Brain at High Magnetic Fields. Magnetic Resonance in Medicine. 1990;14:68-78.
3. Ogawa S, Lee TM, Kay AR, Tank DW. Brain magnetic resonance imaging with contrast dependent on blood
oxygenation. Proc Natl Acad Sci U S A. 1990;87(24):9868-9872.
4. Ogawa S, Lee TM. Magnetic resonance imaging of blood vessels at high fields: in vivo and in vitro
measurements and image simulation. Magn Reson Med. 1990;16(1):9-18.
5. Ogawa S, Menon RS, Tank DW, et al. Functional brain mapping by blood oxygenation level-dependent
contrast magnetic resonance imaging. A comparison of signal characteristics with a biophysical model.
Biophys J. 1993;64(3):803-812.
6. Ogawa S, Tank DW, Menon R, et al. Intrinsic Signal Changes Accompanying Sensory Stimulation: Functional
Brain Mapping With Magnetic Resonance Imaging. Proceeding of the National Academy of Science, USA.
7. Lauterbur PC. Image Formation by Induced Local Interactions: Examples Employing Nuclear Magnetic
Resonance. Nature. 1973;242(5394):190-191.
8. Lauterbur PC, Mendonca MH, Rudin AM. Augmentation of tissue water proton spin-lattice relaxation rates by
in vivo addition of paramagnetic ions. In: Frontiers of Biological Energetics. Vol 1. Academic Press;
9. Mendoca-Dias MH, Gaggelli E, Lauterbur PC. Paramagnetic Contrast Agents in Nuclear Magnetic Resonance
Medical Imaging. Seminars in Nuclear Medicine. 1983;13(4):364-376.
10. Faulk WP, Taylor GM. An immunocolloid method for the electron microscope. Immunochemistry.
11. Stirling JW. Immuno- and affinity probes for electron microscopy: a review of labeling and preparation
techniques. The journal of histochemistry and cytochemistry : official journal of the Histochemistry Society.
12. Cheng SB, Ferland P, Webster P, Bearer EL. Herpes simplex virus dances with amyloid precursor protein
while exiting the cell. PLoS One. 2011;6(3):e17966.
13. Bearer EL, Friend DS. Anionic lipid domains: correlation with functional topography in a mammalian cell
membrane. Proc Natl Acad Sci U S A. 1980;77(11):6601-6605.
14. Bearer EL, Friend DS. Modifications of anionic-lipid domains preceding membrane fusion in guinea pig
sperm. The Journal of cell biology. 1982;92(3):604-615.
Uselman et al. Longitudinal MEMRI
15. Friend DS, Bearer EL. beta-Hydroxysterol distribution as determined by freeze-fracture cytochemistry. The
Histochemical journal. 1981;13(4):535-546.
16. Crossgrove J. Manganese Distribution Across the Blood–Brain Barrier III The Divalent Metal Transporter-1 is
not the Major Mechanism Mediating Brain Manganese Uptake. NeuroToxicology. 2004;25(3):451-460.
17. Silva AC, Bock NA. Manganese-Enhanced MRI: An Exceptional Tool in Translational Neuroimaging.
Schizophrenia Bulletin. 2007;34(4):595-604.
18. Uselman TW, Barto DR, Jacobs RE, Bearer EL. Evolution of brain-wide activity in the awake behaving mouse
after acute fear by longitudinal manganese-enhanced MRI. Neuroimage. 2020;222(116975):1-22.
19. Lu H, Xi ZX, Gitajn L, Rea W, Yang Y, Stein EA. Cocaine-induced brain activation detected by dynamic
manganese-enhanced magnetic resonance imaging (MEMRI). Proc Natl Acad Sci U S A. 2007;104(7):2489-
20. Lu H, Yang S, Zuo Y, Demny S, Stein EA, Yang Y. Real-time animal functional magnetic resonance imaging
and its application to neuropharmacological studies. Magn Reson Imaging. 2008;26(9):1266-1272.
21. Koretsky AP, Silva AC. Manganese-enhanced magnetic resonance imaging (MEMRI). NMR Biomed.
22. Bearer EL, Falzone TL, Zhang X, Biris O, Rasin A, Jacobs RE. Role of neuronal activity and kinesin on tract
tracing by manganese-enhanced MRI (MEMRI). Neuroimage. 2007;37 Suppl 1:S37-46.
23. Gallagher JJ, Zhang X, Ziomek GJ, Jacobs RE, Bearer EL. Deficits in axonal transport in hippocampal-based
circuitry and the visual pathway in APP knock-out animals witnessed by manganese enhanced MRI.
24. Medina CS, Biris O, Falzone TL, Zhang X, Zimmerman AJ, Bearer EL. Hippocampal to basal forebrain
transport of Mn(2+) is impaired by deletion of KLC1, a subunit of the conventional kinesin microtubule-based
motor. Neuroimage. 2017;145(Pt A):44-57.
25. Wang FH, Appelkvist P Fau - Klason T, Klason T Fau - Gissberg O, et al. Decreased axonal transport rates in
the Tg2576 APP transgenic mouse: improvement with the gamma-secretase inhibitor MRK-560 as detected by
manganese-enhanced MRI. (1460-9568 (Electronic)).
26. Gale EM, Atanasova IP, Blasi F, Ay I, Caravan P. A Manganese Alternative to Gadolinium for MRI Contrast.
Journal of the American Chemical Society. 2015;137(49):15548-15557.
27. Gale EM, Mukherjee S, Liu C, Loving GS, Caravan P. Structure–Redox–Relaxivity Relationships for Redox
Responsive Manganese-Based Magnetic Resonance Imaging Probes. Inorganic Chemistry.
28. Gale EM, Wey H-Y, Ramsay I, Yen Y-F, Sosnovik DE, Caravan P. A Manganese-based Alternative to
Gadolinium: Contrast-enhanced MR Angiography, Excretion, Pharmacokinetics, and Metabolism. Radiology.
Uselman et al. Longitudinal MEMRI
29. Kim T, Momin E, Choi J, et al. Mesoporous silica-coated hollow manganese oxide nanoparticles as positive
T1 contrast agents for labeling and MRI tracking of adipose-derived mesenchymal stem cells. J Am Chem Soc.
30. Yu M, Ambrose SL, Whaley ZL, et al. A Mononuclear Manganese(II) Complex Demonstrates a Strategy To
Simultaneously Image and Treat Oxidative Stress. Journal of the American Chemical Society.
31. Lelyveld VS, Brustad E Fau - Arnold FH, Arnold Fh Fau - Jasanoff A, Jasanoff A. Metal-substituted protein
MRI contrast agents engineered for enhanced relaxivity and ligand sensitivity. (1520-5126 (Electronic)).
32. Pan D, Caruthers Sd Fau - Hu G, Hu G Fau - Senpan A, et al. Ligand-directed nanobialys as theranostic agent
for drug delivery and manganese-based magnetic resonance imaging of vascular targets. J Am Chem Soc.
33. Ren Y, Sedgwick AC, Chen J, et al. Manganese(II) Texaphyrin: A Paramagnetic Photoacoustic Contrast Agent
Activated by Near-IR Light. J Am Chem Soc. 2020;142(38):16156-16160.
34. Taylor KM, Rieter WJ, Lin W. Manganese-based nanoscale metal-organic frameworks for magnetic resonance
imaging. J Am Chem Soc. 2008;130(44):14358-14359.
35. Troughton JS, Greenfield MT, Greenwood JM, et al. Synthesis and Evaluation of a High Relaxivity
Manganese(II)-Based MRI Contrast Agent. Inorganic Chemistry. 2004;43(20):6313-6323.
36. Wahsner J, Gale EM, Rodríguez-Rodríguez A, Caravan P. Chemistry of MRI Contrast Agents: Current
Challenges and New Frontiers. Chemical Reviews. 2018;119(2):957-1057.
37. Cacace AT, Brozoski T, Berkowitz B, et al. Manganese enhanced magnetic resonance imaging (MEMRI): a
powerful new imaging method to study tinnitus. Hearing research. 2014;311:49-62.
38. Cai X, Zhu Q, Zeng Y, Zeng Q, Chen X, Zhan Y. Manganese Oxide Nanoparticles As MRI Contrast Agents In
Tumor Multimodal Imaging And Therapy. International journal of nanomedicine. 2019;14:8321-8344.
39. Caspani S, Magalhães R, Araújo JP, Sousa CT. Magnetic Nanomaterials as Contrast Agents for MRI.
Materials (Basel, Switzerland). 2020;13(11).
40. Cloyd RA, Koren SA, Abisambra JF. Manganese-Enhanced Magnetic Resonance Imaging: Overview and
Central Nervous System Applications With a Focus on Neurodegeneration. Frontiers in aging