Magnetic Resonance Spectroscopy: Basic Principles and Application in White Matter Disorders

108.1 Basic Principles

Pauli was the first, in 1924, to suggest that electrons spin at high speed. Because the spinning electrons have mass, they have a certain angular moment and, thus, as spinning electric charges, a magnetic mo- ment. Later, it was observed that certain nuclei also have magnetic moments. What these nuclei have in common is the fact that they are isotopes with an odd atomic mass and/or odd atomic number, whereas nuclei with both an even atomic number and an even atomic mass do not have a magnetic moment. Subse- quently it was noted that when these nuclei are placed in a powerful magnetic field, they show a precession- al motion about the axis of the field. In 1946 Bloch and Purcell simultaneously discovered the possibility of resonant energy absorption and emission of pre- cessing nuclei, the basis of magnetic resonance imag- ing (MRI) and spectroscopy (MRS). For this work they were awarded the Nobel Prize in 1952. MRS has developed into a very important tool in molecular chemistry and physics to reveal molecular structure, chemical reaction rates, and diffusion processes. The first spectroscopy experiments on living systems were performed on small-bore systems with tiny objects, such as red blood cells and excised tissue. Wider bores have since been developed, allowing the study of muscle disorders and experimental work with small animals. Subsequently, in vivo MRS of human brain and other organs has also become possible.

Atomic nuclei are composed of protons and neu- trons. These nuclear particles exhibit spin.A proton is positively charged, and this spinning charge gener- ates a small magnetic field. Although a neutron is electrically neutral, its component electrical charges are not uniformly distributed within its volume, and thus the neutron also generates a magnetic field when spinning, but smaller than that produced by a spin- ning proton. The magnetic moments of these parti- cles are directed randomly, and they cancel each oth- er out in nuclei with an even number of protons and neurons. When the nucleus has an odd number of neutrons and/or protons, the nucleus has a net spin, and this spinning charge has a magnetic moment.

These nuclei exhibit the magnetic resonance phe- nomenon. Nuclei of interest for in vivo spectroscopy are

H,

P,

Na, and also

C,

N,

O, and

F as mark- ers of biochemically interesting compounds. If there is no external magnetic field, the nuclear magnetic

moments have a random direction and there is no net magnetic vector. The intrinsic magnetic phenomena can be detected by applying an external magnetic field. The nuclear magnetic moments tend to align themselves parallel or antiparallel to the external magnetic field. On the other hand, however, nuclei tend to be in constant random motion due to thermal effects. The percentage of nuclei that align with the external magnetic field depends on the strength of the magnetic field relative to the random thermal effects. Slightly more nuclei align with the external magnetic field than against it, because the first posi- tion is more stable than the second. Stability is relat- ed to the amount of energy that a nucleus possesses.

Stability is greater at low energy levels. In fact, there is a continuous transition of nuclei between high- and low-energy positions, but there is a slight net surplus in the low energy position (1:10

at 1.5 T).

A spinning nucleus precesses about the axis of the magnetic field. This motion is called Larmor preces- sion, and its frequency the Larmor frequency. The Larmor frequency (w) depends on the magnetic field strength H

and characteristics of the particular nu- cleus, as expressed in the gyromagnetic ratio g. This dependency is represented in the Larmor equation w = g H

. Resonance is a property of physical systems to oscillate at a preferred frequency which is charac- teristic of the system. The characteristic frequency is referred to as the resonance frequency. The most effi- cient energy transfer to atomic particles precessing in magnetic fields occurs in their resonance frequency, the Larmor frequency. For magnetic field strengths used in MRS these resonance frequencies are within the radiofrequency (RF) band of the electromagnetic spectrum. A short burst of RF energy is known as an RF pulse. When an RF pulse is administered, this en- ergy is absorbed, and more nuclei move into the high- energy position. When the RF pulse is terminated, the nuclei return to equilibrium and energy is released at the same frequency. The RF signal emitted can be detected.

The basis of MRS is the chemical shift phenome- non. The atomic nucleus is surrounded by an electron cloud and other atomic nuclei. If an external magnet- ic field is applied, precession is also induced in this electron cloud, resulting in a small magnetic field.

This small magnetic field has an influence on the atomic nucleus and modifies the effect of the external magnetic field at the site of the nucleus. Because of

Magnetic Resonance Spectroscopy: Basic Principles and Application in White Matter Disorders

M.S. van der Knaap, P.J.W. Pouwels

Chapter 108

this slight change in the local magnetic field, the nu- cleus resonates at a slightly different frequency. This shift in frequency is called chemical shift. The chemi- cal shift is determined by the electrochemical envi- ronment of the resonating nucleus, and therefore the precise shift is characteristic of a particular atomic nucleus in a particular compound. The bandwidth of the RF pulse, of course, must contain all the frequen- cies that are necessary to excite the particular atomic nucleus in compounds of interest. The RF signal ob- tained in an MRS experiment contains many slightly different frequencies. Fourier transformation of the signal transforms a complex time-domain signal into a complex frequency-domain signal. The resulting spectrum depicts all the nuclear resonances as a func- tion of their frequency. Each peak in the spectrum represents the atomic nucleus in a particular com- pound. The area under each peak is proportional to the number of nuclei producing that peak. MR spec- tra can be calibrated to yield absolute concentrations of the metabolites represented. The chemical shift de- pends on the magnetic field strength. However, in MRS the resonance frequencies are not expressed in absolute units (hertz) but in relative units (parts per million, ppm) related to the resonance frequency of a given reference compound. The relative units, ppm, are independent of magnetic field strength. In this way the results obtained in experiments performed at different magnetic field strengths become directly comparable with respect to chemical shift. Another phenomenon that determines the appearance of a spectrum is the influence of other nuclei in the same molecule on the nucleus of interest, which is called J-coupling. If there is no coupling, the resonance re- mains a single peak (a so-called singlet). If a certain nucleus in particular compound – in the case of

1

H-MRS a proton – is influenced by other protons only a few chemical bonds apart, coupling will induce a splitting of the resonance. Depending on the num- ber of neighboring protons, the resonance may be split into a doublet, a triplet, or more complex pat- terns. Due to J-coupling spectra will look different at different field strengths.

The nuclei that one would like to observe can be se- lected by using the appropriate RF pulse. MRI most commonly makes use of the magnetic properties of protons and is in fact

H MRI. The protons that con- tribute to the signal intensity on the images are main- ly present in water and fat. MRS, however, focuses not only on protons (

H-MRS), but also on phosphorus (

P-MRS) and, less frequently, on other nuclei such as

13

C,

N,

F, and

Na. In human tissues, these nuclei are present in much lower concentrations than are protons, and the overall sensitivity of the nuclei is much lower than that of protons. Table 108.1 lists the resonant frequencies and relative MR sensitivities of various nuclei at a magnetic field strength of 1.5 T.

MRS methods are relatively insensitive. Weak MR signals are measured from relatively low-concentra- tion compounds. For this reason spectra are obtained from relatively large volumes of interest (so-called voxels). The voxel size necessary to produce a spec- trum with a reasonable signal-to-noise ratio is larger for nuclei with low abundance and low inherent sen- sitivity. For

H-MRS voxels typically have a size of 4–12 ml, whereas for

P typical voxel sizes are in the range of 24–63 ml. A further improvement of the spectral signal-to-noise ratio is obtained by signal av- eraging. For

H-MRS 64–128 acquisitions are com- mon, whereas 128–256 acquisitions are generally used for

P-MRS. It is also essential to use MR instru- ments with optimal technical equipment to ascertain optimal primary sensitivity. The magnetic field should be extremely homogeneous over the volume analyzed. Any significant inhomogeneity in the mag- netic field spreads out and blurs chemical shift spec- tral lines due to the spread of Larmor frequencies across the volume. The resulting spectral line broad- ening is undesirable because it reduces the signal-to- noise ratio and hampers the ability to distinguish two closely neighboring resonance lines. The frequencies and frequency separations increase linearly with field strength. High magnetic field strengths are therefore necessary to optimize spectral resolution. High spec- tral resolution is particularly important for proton spectroscopy, where most interest is focused on metabolites within a narrow chemical shift range, and where the peaks of the metabolites often overlap. For MRS in human beings, a magnetic field strength of 1.5–4 T is used. In experimental work with animals, magnetic field strengths of 4–8 T or even higher are preferred.

For the purpose of cleaning the final spectrum or obviating unwanted resonances, the spectroscopist has access to many suppression and editing tech- niques. The largest peak in the

H spectrum re- presents water. The concentration of water is about 10

–10

times higher than the concentration of the other metabolites. Consequently, the water peak dom-

Table 108.1. Resonance frequencies and relative MR sensitiv- ities at 1.5 T

Nucleus Larmor frequency Relative

in MHz MR sensitivity

1H 63.86 1

31P 25.85 0.066

19F 60.08 0.83

23Na 16.89 0.093

13C 16.06 0.016

15N 6.47 0.001

inates the spectrum, while the other resonances are barely visible. For this reason, water suppression techniques are invariably applied to remove the water peak from the spectrum and make the other peaks better visible and quantifiable. Lipids present in the skull and subcutaneous tissue are another poten- tial problem, as their concentration is also high and their resonances overlap with those of interest- ing brain metabolites. Several techniques are avail- able to overcome the contamination of a spectrum by fat: location of the volume of interest far from the skull, fat suppression techniques, and outer vol- ume suppression when an entire brain slice is investi- gated.

Even with all these precautions, MRS remains a rel- atively insensitive technique requiring concentra- tions at least in the millimolar range for metabolites to be visualized. Many important but too diluted compounds have so far remained undetectable by in vivo MRS. In addition, only molecules sufficiently small and mobile to tumble freely render MR signals useful for in vivo work. No useful signals can be obtained from large molecules such as proteins, even soluble ones, nor from most membrane compounds or small molecules bound to large ones. The broaden- ing of spectral lines makes differentiation of these molecules impossible. However, under certain condi- tions they will contribute to the spectrum as a

“macromolecular baseline.”

For in vivo MRS it is important that a volume of in- terest can be selected from which the spectrum is de- rived. A simple technique is the placement of a sur- face coil over the volume of interest. The spatial defi- nition of such coils is rather poor since the intensity of signal reception decreases with distance. Several techniques have been devised to improve the spatial response of surface coils and to make positioning of the volume of interest more accurate. One major dis- advantage remains, which is that surface coils are on- ly suitable for volumes of interest near the surface.

Image-localized single voxel spectroscopy by gradi- ent control allows direct and precise definition of a volume of interest on a proton image. Major advan- tages are the ability to select volumes located in the deeper parts of the brain and the ability to define a volume of interest guided by the abnormalities seen on the image, avoiding contamination of the spec- trum by signals from undesired regions. In chemical shift imaging (CSI), also called spectroscopic imag- ing, spectra are obtained from many contiguous vox- els at the same time, covering more or less an entire brain slice.

H CSI is usually applied to investigate one of multiple brain slices of 1–2 cm thickness with an in-plane resolution of 1 ¥ 1 cm

. The information thus obtained can be used to construct maps for separate metabolites, which display the distribution of a par- ticular metabolite over the brain slice.

108.2 Metabolites

In a spectrum from the human brain,

P-MRS reveals seven major peaks (Fig. 108.1): phosphomonoesters (PME), inorganic phosphate (P

, including H

PO

, H

PO

, HPO

, and PO

), phosphodiesters (PDE), phosphocreatine (PCr), and the g-, a-, and b-phos- phate groups of adenosine triphosphate (ATP), res- onating at 6.5, 4.9, 2.6, 0, –2.6, –8.0, and –16.5 ppm respectively. ADP contributes to the g- and a-ATP peaks. NAD and NADH contribute to the a-ATP peak.

The main component of the PME peak is phospho- ethanolamine. Other components are phospho- choline, glycerol 3-phosphate and sugar phosphates.

Minor components are phosphoserine and phospho- inositol. Phosphoethanolamine is a precursor in the anabolism of phosphatidylethanolamine and a prod- uct in the catabolism of sphingomyelin. Phospho- choline is a precursor of phosphatidylcholine and sphingomyelin and a product in the catabolism of sphingomyelin. Glycerol 3-phosphate is a precursor of phosphatidylethanolamine, phosphatidylcholine, and plasmalogens and a product in the catabolism of the same compounds. The PME peak is elevated in all rapidly growing tissues with rapid membrane synthe- sis, such as tumors and the growing brain. It is proba- ble that the elevation is caused by the enhanced pres- ence of compounds meant for the production of membrane phospholipids. Generally, the PME peak is considered to reflect phospholipid anabolic activity.

This notion has been confirmed by the finding that the level of PME correlates linearly with the rate of phospholipid synthesis.

The

P spectrum of the brain contains a large, broad peak centered in the PDE region, underlying the remainder of the spectrum (Fig. 108.2). This large, broad peak originates from large, relatively immobile membrane phospholipids. This broad component is commonly removed from the spectrum by filtering

108.2 Metabolites 861

Fig. 108.1. Localized ³¹P spectrum of the human brain. The spectrum is obtained at 1.5 T with the ISIS technique, TR 3750 ms, 256 measurements, voxel size 63 ml

techniques as a standard processing procedure. The remaining PDE peak contains mainly contributions from more mobile phospholipids and relatively minor contributions from phospholipid breakdown products, including glycerophosphocholine and glyc- erophosphoethanolamine, freely soluble cytosolic molecules. These latter compounds are catabolic products of phosphatidylcholine, phosphatidyletha- nolamine, and plasmalogens.

The

P spectrum contains multiple peaks repre- senting high-energy phosphorus compounds (ATP and PCr) and P

. Intracellular PCr acts via the creatine kinase reaction as a buffer to maintain a constant ATP level in the face of variable energy demands. In condi- tions of decreasing ATP, PCr reacts with ADP to pro- duce ATP and creatine with creatine kinase as catalyz- ing enzyme. The reaction is reversible and PCr can be re-synthesized from creatine and ATP. This so-called creatine phosphate shuttle facilitates energy distribu- tion and responds to energy demands. The PCr/P

ratio is often used as a measure of the energy status of the brain. The ATP level only drops when PCr is ex- hausted. In view of the metabolic stability of ATP, it has often been used as an internal reference for quan- titation. In particular the b-ATP peak, which does not contain any contributions of other compounds

known to be present in brain tissue in significant amounts, is used as a reference.

The chemical shift of P

relative to PCr is a function of pH. The pH determines the relative amounts of H

PO

,and HPO

in the equilibrium H

PO

AHPO

+ H

, and the precise resonance frequency of P

as a whole. Using the equation of Petroff et al. (1985), the pH can be deduced from the chemical shift of P

rela- tive to PCr. This pH holds for the compartment in which these phosphates are present, which in the brain is the intracellular compartment.

Peak assignment in the

H spectrum is problemat- ic because of overlapping resonances. Straightfor- ward assignment is possible for the singlet represent- ing the N-acetyl methyl resonance of N-acetylaspar- tate (NAA) at 2.02 ppm, the methyl and the methylene resonance of total creatine (Cr), including free crea- tine and phosphocreatine, at 3.02 ppm and 3.93 ppm, respectively, and the methyl resonances of choline- containing compounds (Cho) at 3.22 ppm (Fig. 108.3).

The methyl resonance of N-acetylaspartyl glutamate (NAAG) is a shoulder on the methyl resonance of NAA with only a slight difference in chemical shift (2.05 as opposed to 2.02 ppm). It is responsible for about 10–25% of the NAA peak. Since the singlet res- onances of NAA, Cr, and Cho exhibit relatively long T

relaxation times, they may be specified in spectra at long echo times (135 or 270 ms) (Fig. 108.3a). When shorter echo times (15–30 ms) are used, a consider- ably increased number of resonances with short T

relaxation times can be visualized (Fig. 108.3b). Most obvious are the appearance of a strong signal from multiple collapsed resonances of myo-inositol (mIns) at 3.56 ppm.A complex pattern of coupled resonances between 2.1 and 2.5 ppm together with a further group of resonances around 3.8 ppm are assigned to glutamine and glutamate. Resonances of g-aminobu- tyric acid (GABA) are overlapped by the larger reso- nances of Cr, glutamate, and NAA (GABA resonances at 1.90, 2.30, and 3.03 ppm). Additional resonances are seen originating from the aspartyl group of NAA (2.48, 2.60, and 2.66 ppm), glucose (3.43 and 3.80 ppm) and scyllo-inositol (3.35 ppm). The singlet of glycine (3.55 ppm) overlaps with the mIns peak.

The quartet of taurine (centered at 3.35 ppm) partly overlaps with scyllo-inositol. Lactate is not usually visible under normal conditions, but can be visual- ized as a doublet centered at 1.33 ppm if elevated. Ala- nine can only be seen when increased in concentra- tion, then giving rise to a doublet at 1.47 ppm. Pyru- vate is below the level of detection under normal cir- cumstances, but when elevated it gives rise to a single peak at 2.36 ppm. In case of elevated tissue levels of free lipids, for instance as a result of myelin break- down or spectral contamination by fat from the skull, broad resonances are seen at 0.9 and 1.3 ppm origi- nating from the methyl and methylene groups of the

Fig. 108.2. A An unfiltered ³¹P spectrum of the brain. B The same spectrum after filtering.The spectra are obtained at 1.5 T with the ISIS technique, TR 3750 ms, 256 measurements, voxel size 63 ml

lipids. Because of severe overlap of resonances in spectra recorded with short echo times, it is difficult to quantify metabolite concentrations. Special tech- niques such as spectral editing have been developed to quantify glutamine, glutamate, and GABA sepa- rately. Alternatively, a spectrum can be quantified by describing it as a linear combination of model spectra of all individual metabolites included in a database using prior knowledge.

NAA is a compound found at high concentration in the CNS, whereas only traces are found in other tis- sues. In the mature brain NAA is almost entirely con- fined to neurons and their axons: 40% of the NAA is in the nerve-ending mitochondrial fraction and 60%

is in the cytosol. In cell cultures NAA has also been demonstrated in oligodendroglial precursor cells.

NAA and its derivative NAAG are synthesized in neu- ronal mitochondria. However, the catabolic enzyme for NAA hydrolysis, aspartoacylase, is expressed only in oligodendrocytes, and the catabolic enzyme for NAAG is expressed only in astrocytes. Little is known about the role of NAA in normal function or disease states. Possible functions include a role as precursor of the putative neurotransmitter NAAG, as storage form of the neurotransmitter aspartate, as acetyl donor in lipid synthesis (important in myelin synthe- sis), as stabilizer of the concentration of acetyl-CoA, as molecular water pump, and as having a role in os- motic regulation, protein synthesis, and cell signal- ing. NAA is considered to be a neuron- and axon-spe- cific marker. In disease conditions NAA is often low, related to neuronal or axonal dysfunction or loss. Par- tial recovery of NAA level has been described in im- proving cerebral disorders. It is, therefore, not correct to interpret a decrease in NAA as a straightforward indication of irreversible neuronal loss. However, it is difficult to separate the effects of recovery of NAA concentrations in recovering neurons from the effects of tissue contraction after removal of damaged tissue elements. Temporary reductions in NAA have been ascribed to temporary metabolic depression and de- creased mitochondrial energy production.

The Cr peak represents the total amount of crea- tine and phosphocreatine present in the creatine ki- nase shuttle. The total creatine pool remains fairly constant under a variety of conditions. For this reason Cr has often been used as internal reference for quan- titation. However, since Cr is not constant under all conditions, other means of referencing are preferable.

The Cr concentration of glia is relatively high and consequently increased Cr is seen in conditions of gliosis. Cr is only present intracellularly, and Cr has also been considered a marker for cellular density.

The Cho peak contains contributions from various compounds. Water-soluble choline-containing com- pounds in the brain, including choline, glycerophos- phocholine, and phosphocholine, are responsible for the largest part of the peak. A smaller contribution to the peak comes from choline-containing phospho- lipids, including sphingomyelin and phosphatidyl- choline. Acetylcholine is also a water-soluble com- pound, but has a low concentration. Elevated Cho is seen in conditions of high cell density and enhanced membrane turnover, such as brain growth, myelina- tion, demyelination, inflammation, and tumor growth.

Cho is also an osmolyte and its level may reflect com- pensation for osmotic changes.

Lactate occupies a special position in energy me- tabolism. Being the end product of glycolysis, it rises in concentration whenever the glycolytic rate in a vol- ume of tissue exceeds the tissue’s capacity to catabo- lize lactate or export it to the blood stream. Lactate levels are increased under conditions of anaerobic

108.2 Metabolites 863

Fig. 108.3. Localized ¹H spectra of parietal cortex (12 ml) of a 4-year-old healthy girl, obtained at 1.5T. A At a long echo time of 135 ms (PRESS, TR/TE = 3000/135 ms) only singlet reso- nances of NAA, Cr, and Cho are visible. B At a short echo time of 20 ms (STEAM, TR/TE = 6000/20 ms) additional resonances of mIns, Glx (glutamate and glutamine), and other metabolites are present in the spectrum

glycolysis, for example in failure of energy supply or in respiratory chain defects. Macrophages are highly dependent on anaerobic glycolysis for their energy production. Elevated lactate is also seen in conditions characterized by the presence of increased numbers of macrophages, such as active demyelination and tissue necrosis. Increased lactate secondarily leads to increased alanine, which may become visible in the spectrum.

The mIns peak is also a composite peak, containing contributions from myo-inositol as main component, and, additionally, inositol monophosphate, phos- phatidyl inositol (a membrane phospholipid), and inositol diphosphate as minor components. Inositol diphosphate is, however, a very important compo- nent, functioning as a second or third messenger for various hormone actions. Inositol diphosphate is ac- tive in releasing calcium from the endoplasmic retic- ulum and mitochondria. Many enzymes are depen- dent upon the inositol-induced calcium release. The function of myo-inositol itself is still largely un- known, but it may be the storage form of the inositol- dependent messenger system. Another important function of myo-inositol is that of osmoregulator. In diabetes mellitus, inositol phosphate has been con- sidered as a mediator of some of the secondary com- plications, in particular polyneuropathy. In galac- tosemia some of the late complications have been at- tributed to depletion of myo-inositol. Also in hepatic encephalopathy a depletion of myo-inositol is seen.

myo-Inositol is only present in glia and mIns can therefore be used as a glial marker. In conditions of gliosis, the mIns level is increased. scyllo-Inositol is a stereoisomer of myo-inositol and usually the varia- tion in myo-inositol level and scyllo-inositol level change simultaneously. However, in some individuals high scyllo-inositol levels are found in the presence of normal myo-inositol levels.

Glutamate is the most important excitatory neuro- transmitter, whereas GABA, formed by decarboxyla- tion from glutamate, is the most important inhibitory neurotransmitter. In the presynaptic neuron gluta- mine is converted to glutamate and ammonia by glu- taminase. Glutamate can either be converted to GABA by glutamic acid decarboxylase or be released in the synaptic cleft. After release by the presynaptic neu- ron, glutamate is taken up by the astrocyte in which it is processed by glutamine synthetase into glutamine, which is transported back to the presynaptic neuron.

This cycle operates between the presynaptic neuron, the neuronal cleft, and the astrocyte. Hyperammone- mia has a great impact on this cycle by stimulating glutamine synthesis via glutamine synthetase, by pos- sible inhibition of glutaminase, and by inhibition of glutamate reuptake by the astrocyte. In hyperam- monemia glutamate decreases and glutamine in- creases. Glutamate may be elevated in conditions of

active tissue damage, such as hypoxia–ischemia, as one of the so-called excitatory amino acids. Apart from its role as neurotransmitter and precursor of GABA, glutamate is important as a Krebs cycle com- ponent via a-ketoglutarate. Considering its predomi- nantly glial localization, glutamine may be considered a glial marker.

Taurine is a key component of the cytosol buffer and serves to optimize cytosol buffering capacity at physiological pH. It is involved in neurotransmission, osmoregulation, and brain growth. It controls magne- sium homeostasis and calcium homeostasis.

108.3 Normal Values: Age Dependency and Regional Variability

The major biochemical changes related to processes of brain maturation are reflected in spectroscopic changes (Figs. 108.4 and 108.5).

The most striking finding in

P-MRS (Fig. 108.4) of neonatal brain is a very high PME peak, in particu- lar in fetuses and preterm babies, in whom the PME peak is higher than all other peaks in the spectrum.

Relative to b-ATP, PME decreases with ongoing brain maturation, to reach final values at about 2 years of age. In terms of absolute metabolite concentrations also, PME decreases with increasing age. The eleva- tion of the PME peak in early development is mainly related to elevated phosphoethanolamine. The high PME peak can be attributed to active membrane phospholipid synthesis, in particular active myelina- tion. Myelination is almost complete at the age of 2 years.

PDE is very low in fetal brain. PDE increases with ongoing brain maturation, but is still relatively low in term neonates. The PDE/b-ATP ratio and absolute concentrations of PDE increase with age. Final values are reached at the age of about 2 years. PDE originates to a large extent from phospholipids and to a smaller extent from low-molecular-weight soluble metabo- lites in phospholipid breakdown. The increase in PDE with increasing brain maturation is primarily related to the increasing membrane density of the brain, which is largely caused by progressing myelination.

With increasing membrane content of the brain, membrane turnover also increases, and, accordingly, so does the concentration of intermediary products of phospholipid breakdown, contributing to the height of the PDE peak.

The PME/PDE ratio can be used as a maturation index for brain development.

There are major changes in energy metabolism re-

lated to brain maturation. On the basis of findings in

animal experimental research it was assumed that the

cerebral concentration of ATP, reflected in b-ATP,

does not change with age and is constant irrespective

of maturational stage. However, evidence has been provided by Buchli et al. (1994) that the ATP concen- tration of the brain increases significantly with age during early development. Relative to b-ATP, PCr increases rapidly after birth and reaches final values after a few months. P

/b-ATP does not undergo signif- icant changes. Buchli et al. (1994), using absolute quantitative data, found some postnatal increase in P

and a more marked increase in PCr and ATP. So, there is a maturational increase in the creatine kinase reac- tion rate with age. PCr/P

is considered a measure of the phosphorylation potential and energy status of the examined tissue; the increase may, therefore, im- ply an increase in energy reserve in infant brain tis- sue. This increase would be compatible with the in- crease in rate of energy metabolism observed in post- natal cerebral maturation.

Human cerebral pH shows some decrease with on- going cerebral maturation.

1

H spectra also undergo major developmental changes. In neonates, in particular in preterms, the spectrum is dominated by Cho and mIns peaks, whereas in the adult the NAA peak is largest (Fig.

108.5).

NAA is low at birth. NAA/Cr ratio and absolute NAA values have been shown to increase after birth.

The steepest increase occurs during the first year of life, with very gradual increases during childhood, and final values are reached around the age of 10 years. The increase in NAA can be attributed to processes of neuronal maturation, including increase in number of axons, dendrites, and synaptic connec- tions.

There is a rapid increase in the cerebral Cr concen- tration before and around term, and final values are reached after a few months. The difference from adult values is small but significant. Therefore, Cr is an in- ternal reference of limited value, especially in early development.

Cho is high in neonates and decreases thereafter.

Cho/Cr ratio and absolute Cho values decrease with increasing cerebral maturation, and final values are reached after 3–5 years. The high Cho in early devel- opment can be attributed to a high rate of membrane synthesis and turnover, in particular related to the process of myelination.

The mIns concentration is high at birth and de- creases rapidly. Final values are reached within the

108.3 Normal Values: Age Dependency and Regional Variability 865

Fig. 108.4. Four ³¹P spectra of the brain obtained at different ages after birth at term: 1 month (A), 4 months (B), 2 years (C), and 15 years (D). Note the decrease in PME and increase in PDE

and PCr relative to b-ATP with increasing age. The spectra are obtained at 1.5 T with the ISIS technique, TR 3750 ms and 256 measurements, voxel size 54–63 ml

first year of life. The significance of the high mIns concentration in early development is not known.

The resonances originating from taurine are high at birth, and gradually decrease.

The peaks originating from glutamine and gluta- mate do not change significantly.

Metabolite concentrations show regional variabili- ty, determined by the structures included in the study (Fig. 108.6). Depending on the size of the individual voxels, partial volume effects are important. Due to the cortical foldings it is not possible to obtain spec- troscopic data from a voxel containing cortex only, and it is difficult to choose a voxel containing white matter only. The impact of the problem is more severe on

P-MRS than on

H-MRS, because larger voxels are used. In

P-MRS, PME, PDE, and P

are higher in white matter than in gray matter, whereas PCr is high- er in gray matter. In

H-MRS it is possible to obtain spectra from voxels containing mainly cortex or white matter or central nuclei. It has been shown that NAA, Cr, mIns, glutamate, and glutamine are higher in gray matter spectra than in white matter spectra, whereas Cho is higher in white matter spectra (Fig. 108.6). The normal spectroscopic findings in the intact brain stem, cerebellum, and basal ganglia are again different. The maturational changes as de-

scribed above are valid for brain structures in gener- al, but they may differ in detail for different brain structures. These regional and age-dependent spec- tral changes make it necessary to obtain region-spe- cific normal values for different ages.

108.4 MRS in Cerebral Disorders:

Process-Specific Abnormalities Two types of spectroscopic abnormalities are seen in white matter disorders:

1. Process-specific spectroscopic abnormalities, re- lated to delayed maturation and tissue damage.

2. Disease-specific spectroscopic changes, directly related to the particular disorder under investiga- tion.

In this section, the process-specific abnormalities are discussed.

In retarded brain maturation, MRI only shows delayed myelination, because, apart from gyration, processes of neuronal maturation cannot be visual- ized. In MRS one expects metabolite levels appropri- ate for a younger age than the age of the patient. The use of MRS to provide quantitative measures for

Fig. 108.5. Four ¹H spectra of parietal white matter obtained at 1.5 T (STEAM, TR/TE = 2500/20 ms) and at different ages:

32 weeks postconceptional age (A), 3 months after term birth

(B), at 4 years (C), and at 9 years (D). Note the increase in NAA and decrease in Cho and mIns relative to Cr with increasing age

processes of cerebral maturation is limited by the rather wide normal variation.

In demyelinating disorders, it is primarily the myelin sheath that is lost; secondarily, axonal damage and loss occurs. Histologically, demyelinating disor- ders are characterized by rarefaction of the white matter with respect to its normal constituents. Gliot- ic scar tissue fills in the free spaces that arise as a con- sequence of the loss of myelin and axons. Conse- quently, demyelinating disorders do not lead to sig- nificant cerebral atrophy in the early stages. Only when the demyelination is severe and long-standing does atrophy ensue. The rarefaction of white matter implies that the total amount of membrane phospho- lipids per volume of brain tissue decreases, resulting in a decrease in PDE in the

P spectrum (Fig. 108.7).

The decrease in the ratio of PDE to b-ATP has been shown to be proportional to the extent of the demyeli- nation. The PME peak, which is proportional to the rate of membrane phospholipid synthesis, remains normal until demyelination is very severe. In severe demyelination a variable decrease in PME/b-ATP ra- tio is observed. The neuronal damage and loss accom-

panying demyelination are reflected in a decrease in NAA in the

H spectrum. The decrease in NAA occurs early in the process. The extent of NAA loss can vary considerably, depending on the stage of the disease

108.4 MRS in Cerebral Disorders: Process-Specific Abnormalities 867

Fig. 108.6. ¹H spectra of a 4-year-old healthy girl, obtained with STEAM (TR/TE = 6000/20 ms) at 1.5 T. Spectra are from (A) parietal cortex (10 ml), (B) parietal white matter (4 ml),

(C) basal ganglia (5 ml), and (D) cerebellar vermis (8 ml). The spectra are plotted on the same vertical scale to allow a quali- tative comparison

Fig. 108.7. ³¹P spectrum of a patient with severe demyelina- tion. Note the very low PDE. Compare this spectrum to that of Fig. 108.1, which was obtained from a normal child of the same age

and destructiveness of the process, as illustrated in Fig. 108.8 for patients with infantile and juvenile Krabbe disease. In active demyelination Cho is elevat- ed, related to enhanced membrane lipid turnover (Fig. 108.9). In addition, elevated protein and lipid peaks at 0.9 and 1.3 ppm are occasionally seen, relat- ed to the enhanced presence of myelin breakdown products (Fig. 108.9b). A variable increase in lactate occurs in active demyelination, presumably related to the infiltration of the tissue by macrophages (Fig.

108.9b). Increased levels of glutamate/glutamine have been observed, which may be related to processes of active tissue degeneration. The concentration of mIns increases as a consequence of gliosis (Fig.

108.10a). In the end-stage phase of demyelination, Cho decreases again, lactate and lipids disappear, and the spectrum becomes “empty,” dominated by a high mIns peak. The cortex spectrum remains relatively intact (Fig. 108.10b).

In primary neuronal degenerative disorders, neu- rons undergo degenerative changes, die, and disap- pear, together with their axons and myelin sheaths.

White matter rarefaction does not occur and gliosis is usually inconspicuous. Tissue loss is primarily evi- dent as atrophy.As expected, the PDE/b-ATP ratio has been found to remain within the normal range, even in far advanced disease. The neuronal damage is reflected in a decrease in NAA, but the decrease in

Fig. 108.8. ¹H spectra obtained with STEAM (TR/TE = 6000/

20 ms) at 1.5 T. A The white matter spectrum (5 ml) of a 9-month-old girl with advanced-stage infantile Krabbe dis- ease shows almost complete absence of NAA, increased Cho and mIns, as well as a clear elevation of lactate, represent- ed by the doublet at 1.33 ppm. B The white matter spectrum (5 ml) of a 4-year-old girl with the much slower disease course of late-infantile/early-juvenile Krabbe disease is characterized by a strong elevation of mIns and a less severe decrease of NAA

Fig. 108.9. ¹H spectra of white matter affected by demyelina- tion (4–5 ml, STEAM, TR/TE = 6000/20 ms, 1.5 T) show variable extents of metabolic abnormalities. A In a 4-year-old X-linked adrenoleukodystrophy patient with nonprogressive lesions, metabolic abnormalities in the affected white matter are lim- ited to a small decrease of NAA and increases of Cho and mIns.

B In a 10-year-old X-linked adrenoleukodystrophy patient with highly active demyelination, a complete absence of NAA, a strong increase of Cho, and lipid resonances at 0.9 and 1.3 ppm, overlapping with a doublet of lactate, are found

NAA underestimates the neuronal/axonal loss as compared to primary demyelinating disorders be- cause of tissue contraction (Fig. 108.11a). However, in neuronal degenerative disorders the decrease in NAA in the cortex spectrum is more marked (Fig. 108.11b) than in demyelinating disorders (Fig.

108.10b). In neuronal degenerative disorders the NAA decrease occurs before evident atrophy on MRI.

So, in this respect, MRS is more sensitive than MRI. In neuronal degenerative disorders, Cho may be elevat- ed related to enhanced membrane turnover. Some elevation of mIns is sometimes seen, related to some gliosis.

In hypomyelination, the level of Cho tends to be low, indicative of decreased (myelin) membrane syn- thesis and turnover (Fig. 108.12). In the case of con- comitant white matter gliosis, mIns and Cr are elevat- ed, as is seen in H-ABC (hypomyelination with atro- phy of the basal ganglia and cerebellum) (Fig. 108.13).

In the case of concomitant axonal damage and loss,

108.4 MRS in Cerebral Disorders: Process-Specific Abnormalities 869

Fig. 108.10. The¹H spectrum of affected white matter (A;

5 ml, STEAM, TR/TE = 6000/20 ms, 1.5T) of a 10-year-old girl with an advanced stage of metachromatic leukodystrophy only contains signals of mIns, Cho, and Cr, indicating the pure gliotic contents of the tissue. B In contrast, the spectrum of the cortex (8 ml) still contains a relatively high signal of NAA

Fig. 108.11. The¹H spectrum of white matter (A; 5 ml, STEAM, TR/TE = 6000/20 ms, 1.5 T) of an 8-year-old boy with the Turk- ish variant of late-infantile neuronal ceroid lipofuscinosis shows a marked decrease of NAA and an increase of lactate.

B The loss of NAA is also pronounced in the cortex spectrum (8 ml). In the cortex spectrum all signals are relatively low due to atrophy and admixture of more CSF than usual

Fig. 108.12. The¹H spectrum of white matter (5 ml, STEAM, TR/TE = 6000/20 ms, 1.5 T) of a 25-year-old male with hypo- myelination shows a decrease of Cho whereas all other metabolite concentrations are within normal ranges

white matter NAA is decreased. Some have found an increased NAA in the white matter of patients with Pelizaeus–Merzbacher disease, which can be ex- plained by the higher axonal density per tissue vol- ume in the absence of myelin sheaths.

In several disorders, white matter pathology is dominated by white matter rarefaction and cystic degeneration. Examples are vanishing white matter disease and severe variants of sulfite oxidase deficien- cy/molybdenum cofactor deficiency. In particular in vanishing white matter, the process of progres- sive white matter rarefaction can be followed with

1

H-MRS. The peaks representing metabolites normal- ly present in the spectrum gradually become lower and disappear (Fig. 108.14). The peaks representing lactate and glucose become visible (Fig. 108.14). Lac- tate and glucose are normally below the level of detec- tion in brain tissue, but they have a higher concentra- tion in the CSF, which makes them detectable in high- ly rarefied and cystic white matter, where most of the white matter has been replaced by tissue water or CSF.

Acute inflammatory white matter lesions, related to encephalitis, acute disseminated encephalomye- litis, and multiple sclerosis, are characterized by a de- crease in NAA and variable increases in Cho, lactate, lipids, and mIns (Fig. 108.15). The findings are indis- tinguishable from those seen in active demyelinating metabolic disorders and related to the same basic pathological processes, including axonal damage and loss, enhanced membrane turnover, macrophage in- filtration, enhanced presence of membrane break- down products, and gliosis.

As MRS allows in vivo assessment of cerebral ener- gy metabolism, it can be applied to evaluate cerebral consequences of cellular energy failure in mitochon- drial defects. Inborn errors involving mitochondrial oxidative phosphorylation and electron transport may lead to failure to synthesize sufficient ATP, accu- mulation of ADP, and failure of PCr synthesis from Cr.

In addition, mitochondrial dysfunction may lead to

Fig. 108.13. The¹H spectrum of white matter (5 ml, STEAM, TR/TE = 6000/20 ms, 1.5 T) of an 8-year-old boy with H-ABC at 1.5 T, shows a slight decrease of Cho, as well as an increase of Cr and especially of mIns

Fig. 108.14. ¹H spectra of white matter (6 ml, STEAM, TR/TE

= 6000/20 ms, 1.5 T) of patients with vanishing white matter.

A In the spectrum of a 10-year-old girl with mild disease all metabolites are only slightly decreased, indicating a rarefac- tion of the white matter. B The spectrum of a 3-year-old girl with totally cystic white matter on MRI only contains lactate (doublet at 1.33 ppm) and glucose (signals at 3.4 and 3.8 ppm), indicating the replacement of white matter by CSF

Fig. 108.15. The ¹H spectrum (6 ml, STEAM, TR/TE = 6000/

20 ms, 1.5 T) of a white matter lesion of a 44-year-old female with relapsing-remitting multiple sclerosis shows a decrease of NAA and an increase of mIns

accumulation of unoxidized reducing equivalents in the form of NADH, NADPH, and lactate. Alanine may be elevated secondary to the lactate elevation. The pattern of metabolic disturbances in mitochondrial

disorders is, in most respects, indistinguishable from that of hypoxia. In

H-MRS in a variety of mitochon- drial encephalopathies, including Leigh syndrome, NARP syndrome, Kearns–Sayre syndrome (Fig.

108.16), MELAS (Fig. 108.17), MERRF, and Leber hereditary optic atrophy, variable elevations of cere- bral lactate are found, but lactate is not unequivocally elevated in all patients and/or in all brain structures (Fig. 108.18). Cross et al. (1993) found a good correla- tion between CSF levels of lactate: when the CSF lac- tate was below 2.5 mmol/l, no lactate was seen in

H spectra of the brain, whereas in all cases with a CSF lactate above 4.0 mmol/l, elevated lactate was found in MRS. However, patients have been described repeat-

108.4 MRS in Cerebral Disorders: Process-Specific Abnormalities 871

Fig. 108.16. The¹H spectrum of white matter (5 ml, STEAM, TR/TE = 6000/20 ms, 1.5 T) of a 15-year-old boy with Kearns–

Sayre syndrome shows a decrease of NAA and mIns, as well as an increase of lactate (doublet at 1.33 ppm) and glucose (res- onances at 3.43 and 3.8 ppm)

Fig. 108.17. A The¹H spectrum of an acute lesion in the occipital cortex (8 ml, STEAM, TR/TE = 6000/20 ms, 1.5 T) of a 12-year-old boy with MELAS shows a prominent doublet of lactate and complete loss of NAA. B The spectrum of an adjoining older lesion in the parietal cortex (6 ml) contains only some small signals from remaining lactate, Cho, and Cr

Fig. 108.18. A The¹H spectrum of white matter (5 ml, STEAM, TR/TE = 6000/20 ms, 1.5 T) of a 2-year-old boy with Leigh syn- drome (Leigh/NARP mutation) is completely inconspicuous.

B A spectrum of the lesion in the basal ganglia (4 ml) in the same patient contains a small lactate doublet. It should, however, be noted that lactate is not necessarily detected in patients with mitochondrial disorders, not even in areas which are abnormal on MRI

edly with absence of lactate in

H spectra of the brain, while CSF lactate was high, as well as patients with el- evated lactate in

H spectra of the brain while CSF lac- tate was normal. There is a regional variability in ele- vations of cerebral lactate, lactate being most pro- nounced in regions where MRI shows fresh structur- al abnormalities, as illustrated in Fig. 108.17. The lactate decreases again when the lesion turns into in- active scar tissue. In some patients the elevated ala- nine can be visualized. Elevated cerebral glucose lev- els may be seen (Fig. 108.16), ascribed to a high de- gree of nonoxidative glycolysis. In pyruvate dehydro- genase complex deficiency, elevated cerebral pyruvate may be demonstrable in addition to the elevated lac- tate (Fig. 108.19).

P-MRS in mitochondrial disorders may reveal a decrease in PCr and increase in P

in the presence of a normal pH, while calculations yield an increase in ADP, a decrease in phosphorylation po- tential, and an increased percentage of the maximal rate at which ATP is being synthesized, the latter re- flecting the increased demand on the mitochondria able to participate in oxidative phosphorylation.

These abnormalities are, however, again not invari- ably present. The above

H and

P spectroscopic fea- tures are a measure of the derangement of cerebral energy metabolism and can be used to monitor the course of disease in mitochondrial disorders, in par- ticular to evaluate the effects of treatment. In addi- tion, nonspecific spectroscopic changes can be pre- sent in mitochondrial disorders, in particular a de- crease in NAA, glutamate, and Cr in lesions, related to neuronal dysfunction and loss.

Similar spectroscopic changes related to cellular energy failure are present in hypoxic–ischemic en- cephalopathy of neonates. In the acute phase of cere-

bral hypoxia–ischemia, acute reciprocal changes oc- cur in the concentrations of PCr and P

, so that the PCr/P

ratio falls. Only when PCr/P

reaches a low val- ue does the concentration of ATP also decline. Intra- cellular pH is reduced. These changes are rapidly re- versed when cerebral perfusion and oxygenation are restored. Over the subsequent few days, spectral ab- normalities develop again in spite of the absence of any ongoing hypoxia–ischemia. Again there are roughly reciprocal changes in PCr and P

and, in se- verely affected infants, a fall in ATP. However, intracel- lular pH tends to rise rather than to fall. Later P

in- creases, well out of proportion to the fall in PCr, and in severely damaged brains PCr and ATP are absent, leaving only a large residual P

peak in the spectrum.

In less severely affected infants who recover, the

P metabolite ratios return to normal over the course of about 2 weeks, but sometimes the total phosphorus signal is reduced, indicating permanent loss of cells.

The explanation of the presence of two stages is that the initial acute hypoxic–ischemic episode initiates a series of reactions, which later cause progressive dis- ruption of oxidative phosphorylation in brain tissue.

This disruption is termed “secondary energy failure.”

This course of events explains why

P-MRS findings are often normal on the first day of life, to become ab- normal over the subsequent few days. The

P-MRS findings appear to have prognostic value. Of the chil- dren whose PCr/P

falls below the 95% confidence limits for normal infants, two-thirds die and almost all survivors have a serious neurological handicap.

Nearly all infants with a decrease in ATP die. With re- spect to

H spectroscopic findings, it has been shown that lactate is elevated during the hypoxia–ischemia, to recover partially soon after restoration of normal oxygenation of the brain and rise again during the first 2 days, mirroring the phenomenon of secondary energy failure seen in

P-MRS. NAA begins to de- crease several hours after the hypoxic–ischemic episode. mIns may also increase within the first few days. Glutamate may be relatively elevated for a while after the acute hypoxic–ischemic incident, suggesting excitotoxic effects. In contrast to the normalization of the

P spectra,

H spectra remain abnormal. NAA may recover partially, but does not return to normal.

The increased lactate may persist for weeks.

H spec- troscopic findings also have prognostic value. The lowest NAA levels are found in the children with the poorest outcome. The presence of elevated lactate and also of mIns carries a poor prognosis with respect to survival and neurological handicap.

In conditions of chronic hypoxia–ischemia, such as subacute arteriosclerotic encephalopathy and CADASIL, damage-specific spectroscopic abnormali- ties are found, including decreased NAA and in- creased Cho and mIns, whereas lactate may be nor- mal or mildly elevated.

Fig. 108.19. The¹H spectrum of frontal cortex (5 ml, STEAM, TR/TE = 6000/20 ms, 1.5 T) of a 7-month-old girl with pyruvate dehydrogenase complex deficiency shows not only a promi- nent lactate doublet (1.33 ppm), but also a clear singlet at 2.36 ppm, which can be assigned to pyruvate. The spectrum is further characterized by decreased signals of NAA and mIns

The

H spectrum contains resonances of several metabolites that have a function as idiogenic os- molytes. mIns is probably the most important among them, but Cho, Cr, NAA, glutamine and glutamate al- so have a role in maintaining iso-osmosis. It has been shown that in hypernatremia mIns, scyllo-inositol, Cho, Cr, and glutamine/glutamate are elevated, re- turning to normal when the serum sodium level re- turns to normal. In hyponatremia, mIns, Cho, Cr, and NAA are decreased and return to normal with nor- malization of the serum sodium level. In our patient with extremely elevated levels of

-arabitol and ribitol in the brain (see below), NAA, mIns, Cr, and Cho are decreased, probably a compensatory reaction to the osmotic effects of the accumulating polyols.

In cases of hyperammonemia of any origin, in- cluding urea cycle defects, propionic acidemia, Reye syndrome, Wilson disease, and other causes of hepat- ic failure,

H spectra show changes directly related to the high levels of ammonium. Hyperammonemia has a great impact on the equilibrium between glutamate, glutamine, and GABA by stimulating glutamine syn- thesis from glutamate, resulting in an accumulation of glutamine and a depletion of glutamate. Elevated levels of glutamine can be visualized easily by

H- MRS (Fig. 108.20). Other consistent and striking spectroscopic changes in hyperammonemia concern a decrease in mIns and Cho (Fig. 108.20). As mIns is an osmolyte, depletion of mIns is probably an osmot- ic response to glutamine accumulation. In addition, mIns is a precursor of glucuronic acid, which helps to detoxify xenobiotics by conjugation. Depletion of mIns could partially be the result of excessive detoxi- fication. The decrease in Cho is probably also an os- motic response. In cases of neuronal damage, NAA is decreased, but in many patients with hyperammone- mia, NAA is normal. The spectroscopic changes di- rectly related to hyperammonemia are reversible af- ter normalization of ammonia levels, implying that

1

H-MRS is a very useful tool for monitoring cerebral abnormalities in hyperammonemias.

108.5 MRS in Cerebral Disorders:

Disease-Specific Abnormalities Specific changes, related to the specific disorder, can be found in particular in some of the inborn and ac- quired errors of metabolism.

The relative increase in NAA in Canavan disease is well known (Fig. 108.21). It is specifically the metabo- lism of NAA that is disturbed in this disorder. When absolute quantitation is used and metabolites are ex- pressed in millimoles per liter of brain tissue, Cho and Cr appear to be low and NAA can vary between the normal range and up to a two-fold elevation. In direct biochemical measurements the increase in the level of cerebral NAA expressed in micromoles per milligram protein is much more pronounced. The apparent discrepancy between MRS and biochemical findings can be explained by the different way of ex- pressing the concentration: per volume (MRS) or per dry weight (biochemical analyses). In Canavan dis- ease, the white matter has a reduced myelin content and cellularity, and the white matter is converted into a loose, vacuolated meshwork, with glia and bare axons left as tissue elements. This serious white mat- ter rarefaction implies that a large part of the selected volume of interest for MRS consists of water. The low Cho with low Cho/Cr ratio in Canavan disease is probably related to the very low membrane content of the white matter. In itself the elevation of NAA is

108.5 MRS in Cerebral Disorders: Disease-Specific Abnormalities 873

Fig. 108.20. A The¹H spectrum of basal ganglia (8 ml, STEAM, TR/TE = 6000/20 ms, 1.5 T) of a 6-year-old boy with a porto- caval shunt and serum ammonia levels of 150–200 mmol/l shows the effects of hyperammonemia in the brain: apart from an increase in glutamine, the spectrum is characterized by low signals from mIns and Cho, compensating the osmotic pres- sure. B The resonances from glutamine dominate the spec- trum obtained from basal ganglia and white matter (27 ml, STEAM, TR/TE = 2500/20 ms, 1.5 T) in a 1-week-old neonate suffering from argininosuccinate lyase deficiency (urea cycle defect) and serum ammonia levels of more than 1000 mmol/l.

The signals of the osmolytes mIns and Cho are small in com- parison to the usually much higher levels in neonates

specific for Canavan disease, as in all other conditions NAA is normal or decreased, but not increased. In ad- dition, mIns is increased and in some of the Canavan patients lactate is elevated.

One patient has been described who had no de- tectable NAA in the brain (Martin et al. 2001), proba- bly related to an as yet unidentified defect in NAA synthesis (Fig. 108.22). Surprisingly, the clinical pic- ture was characterized by psychomotor retardation only, whereas MRI revealed delayed myelination at the age of 3 years.

Salla disease and severe infantile sialic acid storage disease are characterized by accumulation of N-acetyl- neuraminic acid (sialic acid). N-acetylneuraminic acid resonates close to the methyl resonance of NAA, lead- ing to a pseudo-elevated NAA peak in the

H white matter spectrum. In addition, the Cho peak is abnor- mally low, related to hypomyelination, and the Cr peak is elevated, probably related to white matter gliosis.

In maple syrup urine disease,

H-MRS during acute metabolic decompensation shows elevated lac- tate and resonances around 0.9 ppm corresponding to cerebral accumulation of branched-chain amino acids and keto acids (Fig. 108.23). NAA is decreased.

With treatment and clinical recovery the

H spectra normalize.

In nonketotic hyperglycinemia,

H spectra show an elevated glycine signal at 3.55 ppm at both short and long echo times. Using a short echo time, the glycine peak co-resonates with the mIns peak, but using echo times of 135 or 270 ms, the peak at 3.55 ppm repre- sents glycine only, glycine having a much longer T

than mIns (Fig. 108.24). There is evidence that the time course of glycine content in brain tissue, as shown by MRS, correlates more closely with the clinical course than do plasma and CSF glycine val- ues.

Fig. 108.21. A The¹H spectrum of white matter (4 ml, STEAM, TR/TE = 6000/20 ms, 1.5 T) of a 4-month-old boy with Canavan disease shows elevated signals of NAA and mIns, whereas Cr and Cho are decreased. B Only NAA and mIns are visible in the white matter spectrum (3.5 ml, STEAM, TR/TE = 6000/20 ms, 1.5 T) of the same patient at the age of 1 year.

Fig. 108.22. The¹H spectrum of white matter (6 ml, PRESS, TR/TE = 6000/30 ms, 2 T) of a 3-year-old with an NAA deficien- cy does not contain signals from NAA or NAAG. From Martin et al. (2001), with permission

Fig. 108.23. The¹H spectrum of predominantly white matter (27 ml, STEAM, TR/TE = 1500/270 ms, 1.5 T) of a 3-year-old boy with maple syrup urine disease in the acute state shows next to the doublet of lactate a doublet around 0.9 ppm, assigned to the methyl group of branched-chain amino acids (probably leucine) or keto acids. From Felber et al. (1993), with permission

Succinate dehydrogenase deficiency is a mitochon- drial disorder with a defect in complex II. Elevated succinate, represented by a resonance at 2.40 ppm, can be found in some patients (Fig. 108.25), providing evidence for the diagnosis (Brockmann et al. 2002).

A severe reduction of the Cr peak in

H MR spec- tra of the brain is a central feature of the so-called cre- atine deficiency syndromes (Fig. 108.26). The lack of creatine in the brain can be caused by a defect in the synthesis of creatine due to either guanidinoacetate

108.5 MRS in Cerebral Disorders: Disease-Specific Abnormalities 875

Fig. 108.24. A The¹H spectrum of cortex and white matter (12 ml, STEAM, TR/TE = 2500/20 ms, 1.5 T) of a 1-week-old neonate with nonketotic hyperglycemia shows a very high signal at 3.55 ppm. B A spectrum obtained with a long echo time (PRESS, TR/TE = 2500/270 ms) shows that this signal is largely due to glycine, since mIns does not contribute to the spectrum at such long echo times

Fig. 108.25. The¹H spectrum of white matter (4 ml, STEAM, TR/TE = 6000/20 ms, 2 T) of a 17-month-old girl with succinate dehydrogenase deficiency is dominated by the singlet from succinate at 2.40 ppm.In addition, NAA is severely reduced,and lactate is elevated.From Brockmann et al.(2002),with permission

Fig. 108.26. A The ¹H spectrum of white matter (4 ml, STEAM, TR/TE = 6000/20 ms, 1.5 T) of a 2-year-old boy with guanidi- noacetate methyltransferase deficiency does not contain sig- nals from Cr. B The presence of Cr 3 months later shows the positive effect of treatment with oral creatine supplementa- tion in patients with a creatine synthesis defect

䊳

methyltransferase (GAMT) deficiency or glycine amidinotransferase (AGAT) deficiency, or by a defect in the transport of creatine across the blood–brain barrier (X-linked creatine transporter defect). In GAMT deficiency, the spectrum may also reveal a res- onance at 3.78 ppm, representing guanidinoacetate, which is not seen in the other creatine deficiency syn- dromes. In the defects in creatine synthesis, oral crea- tine substitution leads to striking improvement of the symptoms of the patients and a return of the Cr peak into the spectrum, although the Cr concentration re- mains below normal. Oral creatine substitution does not influence the cerebral creatine level in patients with a transporter defect.

In Sjögren–Larsson syndrome a prominent singlet is found at 1.3 ppm, the lipid region of the spectrum (Fig. 108.27). The peak is seen at both short and long echo times. In addition, in most patients increased resonances at 0.8–0.9 ppm are seen. These findings are compatible with the presence of an abnormal amount of lipids, the methylene protons resonating at 1.3 ppm and the methyl protons resonating at 0.8–

0.9 ppm. The peaks are only found in the cerebral white matter with the highest levels in the periven- tricular region, where T

-weighted images show sig- nal abnormalities, and not in the cerebral gray matter or cerebellum (Willemsen et al. 2004). The identity of the lipids responsible for the peaks has not yet been elucidated, but most likely the resonances represent long-chain fatty alcohols.

1

H-MRS allows monitoring of the cerebral pheny- lalanine levels in phenylketonuria (PKU). The a and b protons of phenylalanine give rise to complex multi- plets of low intensity at the region between 3 and 4 ppm, which cannot be distinguished from overlying strong signals of abundant metabolites under in vivo conditions. In contrast, all signals of the phenyl pro- tons collapse into a single peak at 7.36 ppm of suffi- cient intensity for quantitation. However, this region of the spectrum, too, normally contains several over- lapping peaks between 6.5 and 8.5 ppm. The most efficient way of removing these background signals is to use the spectra of normal controls for subtraction.

As the cerebral phenylalanine level in classical phenylketonuria patients under free nutrition is typ- ically below 1 mmol/l, selection of a large volume of interest for MRS, of the order of 25 ml or larger, is re- quired to obtain a sufficient signal-to-noise ratio for reliable quantitation. PRESS is preferable to STEAM for better signal to noise. In addition, short echo times of 20 ms or less are necessary to minimize sig- nal loss due to T

-weighting of the spectrum.

Blood–brain barrier kinetics for phenylalanine entry into the brain can be studied using

H-MRS. Brain phenylalanine levels do not correlate closely with blood phenylalanine levels. The individual variation in brain-to-blood phenylalanine ratio is large, even among patients with similar blood levels. This indi- vidual variability may contribute to the different neu- rological outcomes in PKU patients.

Fig. 108.27. The¹H spectrum of white matter (8 ml, STEAM, TR/TE = 3000/20 ms, 1.5 T) of a 16-year-old boy with Sjögren–

Larsson syndrome shows a prominent signal at 1.3 ppm, prob- ably representing long-chain fatty alcohols. Courtesy of Dr.

M.A.A.P. Willemsen, Department of Pediatric Neurology, Uni- versity Medical Center St Radboud,Nijmegen,The Netherlands

Fig. 108.28. The¹H spectrum of basal ganglia (3.5 ml, STEAM, TR/TE = 1600/10 ms, 1.5 T) of a 9-day-old patient with galac- tosemia due to a deficiency of galactose-1-phosphate uridyl- transferase. Resonances at 3.67 and 3.74 ppm correspond to galactitol. From Wang et al. (2001), with permission.

Untreated galactosemia is characterized by elevat- ed galactitol concentrations in the brain, which disap- pear with treatment. Galactitol gives rise to reso- nances at 3.67 and 3.74 ppm (Fig. 108.28).

One patient has been described with highly elevat- ed levels of the polyols ribitol and

-arabitol in the brain (van der Knaap et al. 1999). The patient suffered from a slowly progressive encephalopathy and pe- ripheral neuropathy. MRI of the brain revealed exten-

108.5 MRS in Cerebral Disorders: Disease-Specific Abnormalities 877

Fig. 108.29. ¹H spectra of white matter (5 ml) of a 14-year-old boy with deficiency of ribose-5-phosphate isomerase leading to polyol accumulation, obtained with (A) STEAM (TR/TE = 6000/20 ms) and (B) PRESS (TR/TE = 3000/135 ms) at 1.5 T. The large signals between 3.6 and 3.8 ppm are partly inverted due

to J-coupling in the long echo time spectrum. Comparison with model spectra of D-arabitol (C, D) and ribitol (E, F), acquired with corresponding short echo times (C, E) and long echo times (D, F) illustrates the contribution of the two polyols to the spectra in A and B