|
How is Magnetic Resonance Spectroscopy
Applied?
To a certain degree, MRS is a complex technique, but this is due
in part to its versatility in application. One has the choice of
selecting 1) a particular nucleus of interest, 2) the magnetic field
strength of the MR system to conduct the experiments, and 3) the
transmit and receive coil configuration and pulse sequence for localization
(1).
Nuclei of interest
The chosen nuclei will determine what biochemical information can
be assessed and the spatial resolution capability as part of the
localization. The MR signal sensitivity of the more popular 1H
spectroscopy is about 15 times greater than that of 31P spectroscopy.
As a result, the special resolution of 1H spectroscopy at 1.5 tesla,
for example, tends to be within the 1-to-8 cm3 range, while typical
31P spectroscopy voxels are between 27 and 60 cm3. By
choosing 1H spectroscopy, one can assess the viability of neurons
(5,6), glutamate-glutamine neurotransmission cycling (7,8), the
g-aminobutyric acid (GABA) neuronal system, and the second messenger
metabolism by measuring, respectively, the metabolite levels of
N-acetylaspartate (NAA), glutamate, glutamine, GABA, and myo-inositol
(Figure
1a) (9,10). The 1H spectral peaks, phosphocreatine (PCr) and
creatine (Cr), are indistinguishable, as indicated in Figure
1, and PCr and Cr are reactants in the creatine kinase high-energy
phosphate reaction. Therefore, 1H spectroscopy is not the preferred
choice for assessing the high-energy phosphate metabolism unless
the equilibrium is altered.
On the other hand, 31P spectroscopy can measure the metabolite
levels of adenosine triphosphate (ATP), PCr, and inorganic orthophosphate
(Pi), which are associated with high- energy phosphate metabolism
(Figure
1b) (11,12). One would expect decreased PCr with increased high-energy
utilization or a deficit in PCr production. Further, in 31P spectroscopy,
membrane phospholipid (MPL) synthesis and membrane degradation can
be assessed by measuring the freely mobile, water-soluble phosphomonoesters
(termed “free-PME,” and including primarily phosphocholine
[PC] and phosphorylethanolamine [PE]) and phosphodiesters (termed
“free-PDE,” and including glycerolphosphocholine [GPC]
and glycerolethanolamine [GPE]), respectively (13,14) (Figure
1). In a rat model study of neuronal degeneration and regeneration
(using neonatal lesions in the entorhinal cortex), higher levels
of free-PMEs were observed at the time and at the site of neuritic
sprouting, suggesting that the free-PME levels directly reflect
membrane synthesis (15). 1H spectroscopy can also assess limited
information on MPL metabolism.The trimethylamine 1H peak or the
choline-containing peak is primarily composed of GPC (the breakdown
product of MPL) and PC (the MPL precursor). The contribution of
choline (Cho) is below the detection limit (16); therefore, in this
review, the choline-containing peak is termed “GPC+PC.”
Stanley, Pettegrew, and Keshavan provide a more detailed review
of the 1H and 31P metabolites (1).
Magnetic Field Strength
The signal-to-noise ratio (per acquisition time) is critical for
accurate and reliable quantification. In general, increasing the
magnetic field strength leads to an approximate linear increase
in the MR signal amplitude; consequently, conducting the spectroscopy
at a relatively higher magnetic field strength, such as 3, 4, or
even 7 tesla, leads to smaller localized voxel sizes (that is, greater
spatial resolution), which minimizes the degree of partial volume
of different tissue types within the localized voxel. Moreover,
at higher field strengths, the degree to which different peaks overlap
one another is much less (or the chemical shift dispersion is greater),
thus improving the accuracy and precision of quantifying these overlapping
peaks, including glutamate and glutamine (17–21).
Localization Method
Methods available for localization differ, ranging from using a
single-loop transmit or receive coil with a single RF pulse as a
sequence to using a dual-tuned (1H and 31P) volume head coil with
slice-selective RF pulses and gradient pulses for spatially encoding
the field of view. The stimulated acquisition mode (STEAM) and the
point-resolved spectroscopy (PRESS) pulse sequences are the 2 most
commonly used for localization with in vivo 1H spectroscopy (10,21,22).
Both these sequences acquire the MR signal from the intersection
of 3 orthogonal slices or slabs and can be applied to localize either
a single voxel or, combined with phase-encoding gradients, to localize
multiple voxels simultaneously in 2 or 3 dimensions. Also, the time
of echo (TE) parameter of these sequences or the time given for
the MR signal to exponentially attenuate prior to acquisition (4)
may be long (for example, 135 or 272 ms), to acquire only the 1H
metabolites with singlets (NAA, PCr+Cr, and GPC+PC), or it may be
short (35 ms or less), to acquire the singlets plus the multiplets,
such as glutamate, glutamine, myo-inositol, GABA, aspartate, n-acetylaspartylglutamate
(NAAG), taurine, glucose, and scyllo-inositol (9,10). Choices of
localization sequences for in vivo 31P spectroscopy include image-selected
in vivo spectroscopy (ISIS) applied as a single- or multiple-voxel
technique (23), spin echo (24,25), and chemical shift imaging (CSI)
sequences (26). The surface coil provides the most sensitivity,
followed by the Helmholtz coil and the volume coil.
Recent Advancements in the Interpretation of Spectroscopy Data
The Broad Underlying 31P Peak
In addition to the free-PME and free-PDE metabolites, a typical
in vivo 31P spectrum of the brain also can contain a
relatively broad underlying peak over the PDE and PME spectral region
(Figure
1b), owing to larger and less mobile molecules with PDE and
PME moieties, which can be quantified using a post-processing method
(27). These molecules, termed “broad-PDE” (or PME [i-tc]
+ PDE [i- tc]) (27), may reflect signals arising from small
MPL structures, including micelles, synaptic vesicles, transport
and secretory vesicles associated with the Golgi and endoplasmic
reticulum (28–32), and small phosphorylated proteins (31,32).
Synaptic vesicles are enriched in grey matter nerve terminals, and
transport and secretory vesicles are enriched in white matter axons.
As discussed later, the ability to quantify the broad-PDE component
is important to the study of schizophrenia.
The Role of N-acetylaspartate
Over the years, the clinical usefulness of 1H spectroscopy has
grown by demonstrating decreased NAA in many different neuropathologies,
including psychiatric disorders (5,6). NAA is synthesized in neuronal
mitochondria from acetyl-CoA and aspartate by the membrane-bound
enzyme L-aspartate N-acetyltransferase (33,34). Several studies
have confirmed the neuronal localization of NAA (35-38), including
the study by Urenjak and colleagues (39), which demonstrated the
presence of NAA in both neurons and oligodendrocytes in developing
brains but only in neurons in mature brains. Consequently, NAA is
commonly considered to be a putative neuronal marker. There are,
however, inconsistencies of NAA as a marker of viable neurons. Increased
NAA has been reported in Canavan’s disease (40). In this disease,
there is a deficiency in the NAA catabolic enzyme (aspartoacylase),
which leads to neurodegeneration of white matter, including demyelination
(41,42). In addition, decreases in NAA have been shown to be reversible
in neurological diseases involving white matter (43,44). Recent
reinvestigations into the localization of NAA have revealed that
NAA can be expressed in mature oligodendrocytes (that is, myelin)
(45), and there is evidence of intercompartmental cycling of NAA
between neurons and oligodendrocytes (45–48).
|
|
The NAA is synthesized in one compartment and catabolized in the
other and may function as a molecular water pump (49). Considering
these recent findings, the role or interpretation of NAA may no
longer indicate neuronal viability but may reflect the formation
and maintenance of myelin (46). To a certain degree, this does complicate
the interpretation of NAA alterations, especially in studies of
neuropsychiatric disorders in children and adolescents.
Recent Methodological Advancements
The in vivo concentration of GABA in human brain is less than one-half
that of glutamine, and the 1H spectrum of GABA contains complex
multiple peaks that overlap with PCr+Cr and other multiple peaks,
including glutamate and glutamine (50). Consequently, the in vivo
GABA quantification has poor reliability unless a customized pulse
sequence is used to isolate a particular peak of GABA from other
overlapping peaks (that is, to apply a spectral-difference, editing-type
sequence) (51–55). This technique has shown encouraging results
in the study of psychiatric disorders. For the first time, decreased
in vivo GABA levels in the occipital cortex have been observed in
medication-free subjects with major depressive disorder (MDD), compared
with control subjects (56). This provides evidence that associates
altered GABAergic neurotransmission with depression (57). Although
not detailed in this review, Goddard and others applied this technique
to study individuals with a panic disorder, and observed decreased
GABA levels in the occipital cortex of unmedicated subjects with
a panic disorder but without major depression, compared with matched
control subjects (58). This finding is consistent with studies showing
lower GABA levels in animals with anxiety-like behaviours.
Further, the spectral-editing technique can be customized to isolate
and quantify other less prominent metabolites including glutathione
(59), an enzyme-catalyzed antioxidant whose role is to protect the
brain against oxidative stress (60). In support of an altered antioxidant
defense and increased oxidative injury in schizophrenia (61), in
vivo glutathione levels were significantly lower in the medial prefrontal
cortex of schizophrenia subjects compared with control subjects
(62). Do and others hypothesize that the glutathione deficit in
schizophrenia may lead to degeneration of neuronal processes and
loss of connectivity in the prefrontal cortex (62).
Magnetic Resonance Spectroscopy and Neuropsychiatric
Disorders
Schizophrenia
Despite decades of research, the biological basis of schizophrenia
remains unclear. Neuroimaging studies using MRI primarily have shown
cortical grey matter reductions in frontal and temporal lobes, increased
ventricular and sulcal cerebrospinal fluid (CSF), and alterations
in basal ganglia, thalamus, and cerebellum volumes, suggesting that
schizophrenia is a “network disorder” that involves the
heteromodal association cortex and the corticothalamocerebellar
circuits (63–65). Functional neuroimaging studies have shown
reduced function of critical brain structures such as the frontal
cortex, suggesting “hypofrontality” (66). Recent models
suggest that the excitatory neurotransmitter, glutamate, and the
inhibitory neurotransmitter, GABA, play an important role in schizophrenia
(67–70). Peripheral measurements of MPL also have implicated
neuronal cell membranes in schizophrenia (71). Likewise, there is
growing support for a neurodevelopmental abnormality underlying
the neuropathophysiology leading to schizophrenia (72–78).
Because of the neurodevelopmental focus of many studies, the discussion
of spectroscopy studies in schizophrenia will be limited to those
studies looking at the first-episode stage prior to medication and
at the premorbid stage.
First-Episode, Medication-Naive Schizophrenia Studies
The seminal in vivo spectroscopy study by Pettegrew and colleagues
(78) observed altered MPL metabolism and high- energy phosphate
metabolism in the combined right and left prefrontal region of first-episode,
medication-naive (FEMN) schizophrenia subjects, compared with control
subjects. This was the first study to support Feinberg’s hypothesis
of abnormal neurodevelopment in schizophrenia caused by an exaggeration
of normal preadolescent synaptic pruning in the prefrontal region
(75–78). The evidence was based on the developmental profile
of PME and PDE levels (that is, the normal decreasing MPL precursor
levels and increasing MPL breakdown product levels with age are
exaggerated in FEMN schizophrenia patients). Recent neuropathological
studies showing reduced neuropil in the dorsolateral prefrontal
cortex (DLPFC) in schizophrenia is consistent with the overexaggeration
of synaptic pruning (79). Subsequent studies of FEMN schizophrenia
patients demonstrated similar MPL alterations in the left prefrontal
(80,81) and in the right and left temporal lobe (82). The latter
suggests that the exaggerated synaptic pruning also may involve
the temporal lobe.
In view of recent evidence on the PDE peak, if the broad-PDE component
is not segregated or quantified, then the quantified PDE (and PME
to a lesser degree) may be a reflection of both breakdown products
of MPL and the larger, less-mobile molecules with PDE (and PME)
moieties. Consequently, it is unclear whether the increased PDE
in persons with FEMN schizophrenia is due solely to increased free-PDE
levels. In a preliminary study using the method described in (27),
decreased free-PME levels and increased broad-PDE levels were observed
in the prefrontal region of persons with FEMN schizophrenia, compared
with control subjects (83). Interestingly, with a 1H decoupling
method, an increased broad-PDE component, but not the free-PDE,
also has been reported in the prefrontal region of chronic, medicated
schizophrenia sufferers (84). This implies that the originally reported
increased PDE is not due to increased breakdown products of MPL
but to increased phosphorylated proteins, increased content of synaptic
and transport vesicles, increased fluidity in the MPL, leading to
a buildup of vesicles, or increased motion in these larger molecules
with PME and PDE moieties. Similarly, since the observed amount
of the broad-PDE component depends on the acquisition parameters,
field strength, and post-processing (27), these factors may account
for the inconsistencies observed in the PDE results of several chronic
schizophrenia studies (85).
With evidence pointing to an alteration in development, the timing
of the insult leading to the developmental alteration remains controversial;
it may occur either early in life (72–74) or late postnatally
(75–77,86). Determining whether similar MPL alterations are
present earlier in the disorder (for example, during the premorbid
phase) is one approach to addressing this issue. A 31P spectroscopy
study reported increased PDE levels in the combined right- and left-frontal
region of 14 adolescents at increased genetic risk for schizophrenia
(offspring of schizophrenia sufferers), compared with control subjects
(87). Any subjects with schizophrenia symptoms were excluded, and
2 offspring subjects had an adjustment disorder. In a similar study,
decreased free-PME and increased broad-PDE were observed in the
prefrontal region of children and adolescents whose parents had
schizophrenia, compared with control subjects (88). Interestingly,
the decreased free-PME was absent in 5 of the 16 at-risk offspring
subjects who were free of psychiatric illness, which is consistent
with the Klemm and others study (87), suggesting that the precursor
levels of MPL in the prefrontal region may be indicators of early
(nonspecific) psychiatric symptoms. In addition, those offspring
subjects with psychiatric illness (excluding schizophrenia) show
PME deficits and increased broad-PDE levels similar to those found
in FEMN schizophrenia patients, suggesting that the deficits may
be present even earlier than those found in the adolescent years,
at least in subjects with an increased genetic risk for schizophrenia.
This would be consistent with the early developmental model of schizophrenia
(72–74). It is possible, however, that subjects who go on to
develop schizophrenia and who do not have the genetic predisposition
for the disorder may differ in neurodevelopment neuropathology.
|
