Contributions of Magnetic Resonance Spectroscopy to Understanding Development: Potential Applications in the study of Adolescent Alcohol Use and Abuse

Abstract

A growing body of research has documented structural and functional brain development during adolescence, yet little is known about neurochemical changes that occur during this important developmental period. Magnetic resonance spectroscopy (MRS) is a well-developed technology that permits the in vivo quantification of multiple brain neurochemicals relevant to neuronal health and functioning. However, MRS technology has been underused in exploring normative developmental changes during adolescence, and the onset of alcohol and drug use and abuse during this developmental period. This review begins with a brief overview of normative cognitive and neurobiological development during adolescence, followed by an introduction to MRS principles. The subsequent sections provide a comprehensive review of the existing MRS studies of development and cognitive functioning in healthy children and adolescents. The final sections of this article address the potential application of MRS in identifying neurochemical predictors and consequences of alcohol use and abuse in adolescence. MRS studies of adolescent populations hold promise for advancing our understanding of neurobiological risk factors for psychopathology by identifying the biochemical signatures associated with healthy brain development, as well as neurobiological and cognitive correlates of alcohol and substance use and abuse.

Significant structural and functional brain development occurs during adolescence, supporting the concurrent development of emotional and cognitive regulation (Casey, Jones, & Hare, 2008; Dahl, 2001; Spear, 2000; Steinberg, 2005). However, brain development during adolescence may also potentiate the occurrence of a variety of forms of psychopathology, including alcohol abuse and dependence (Paus, Keshavan, & Giedd, 2008; Spear, 2000). Adolescent neurological changes have been observed via the use of non-invasive structural and functional magnetic resonance imaging techniques (MRI and fMRI, respectively). Such adolescent neuroimaging research has suggested a developmental trajectory towards increasingly refined neural circuitry and signaling efficiency. Less is known, however, about developmental neurochemical changes occurring during this period, or the impact of these changes on cognition, behavior, or the risk for alcohol addiction. Brain neurochemistry can be explored via magnetic resonance spectroscopy (MRS), a technology that permits non-invasive in vivo detection of multiple brain metabolites relevant to neuronal health and functioning, as well as key neurotransmitter systems such as gamma amino-butyric acid (GABA) and glutamate (Glu). To date, there have been a number of MRS studies exploring normative developmental changes in neurochemistry in the neonatal period, infancy and early childhood, but a paucity of studies focusing on adolescence. There have been significantly more studies conducted using MRS to investigate metabolite abnormalities associated with pediatric psychopathology, including, but not limited to pediatric depression (Gabbay, et al., 2012; Olvera, et al., 2010), bipolar disorder (Davanzo, et al., 2003; Sikoglu, et al., 2013), obsessive-compulsive disorder (Whiteside, Abramowitz, & Port, 2012), generalized anxiety disorder (Strawn, et al., 2013) and attention deficit-hyperactivity disorder (Soliva, et al., 2010; Yeo, Hill, Campbell, Vigil, & Brooks, 2000). In contrast, no MRS data are available that provide insight into the effects of alcohol use and abuse on neurochemistry during adolescence, despite the high prevalence of consumption in this age group (SAMHSA, 2004). Accordingly, the purpose of this review is to 1) provide a brief overview of key cognitive, behavioral, and neurological changes during adolescence; 2) outline the basic principles and methods of MRS; 3) review the existing MRS literature on normative neurochemical development and the relationships between neurochemistry and cognitive functioning in normative development and in adults with alcohol abuse disorders; and 4) present potential applications of MRS in understanding the effects of alcohol consumption and alcohol use disorders on neurochemistry in adolescent populations. Developmental characterization of neurochemistry via MRS during adolescence would help fill in important knowledge gaps and could yield mechanistic hypotheses regarding normative trajectories of adolescent neurochemical development, and importantly, non-normative neurochemical trajectories associated with adolescent alcohol abuse disorders.

Adolescent Development

While structural and functional brain development, along with concurrent cognitive changes, are widely evident and documented during adolescence, less is known about changes in brain chemistry during this age period that may contribute to healthy developmental changes in cognition and behavior, as well as adolescents’ elevated risk for multiple forms of psychopathology (Paus, Keshavan, & Giedd, 2008). Structural, functional and neurochemical changes in the adolescent brain do not occur independently, but rather interact to produce the cognitive and behavioral changes observed during adolescence. Thus, the section below provides a very brief overview of adolescent cognitive and neurobiological development, providing necessary context for the subsequent discussion of MRS findings. More extensive discussions of the many cognitive, behavioral and neurological changes that occur during adolescence are available in multiple existing reviews (e.g. Blakemore & Choudhury, 2006; Casey, et al., 2008; Dahl, 2001; Spear, 2000; Steinberg, 2005, 2010).

Cognitive development in adolescence is predominantly characterized by the maturation of executive functions and the consequent consolidation of collaborative cognitive systems, allowing for flexible thinking, complex problem solving and cognitive control (Hooper, Luciana, Conklin, & Yarger, 2004; Huizinga, Dolan, & van der Molen, 2006; Johnstone, Pleffer, Barry, Clarke, & Smith, 2005; Katya Rubia, et al., 2006; Toga, Thompson, & Sowell, 2006). Adolescents also demonstrate higher levels of sensation-seeking than children or adults, often leading to experimentation with alcohol and drugs, as well as other risky behaviors (Spear, 2000). Adolescence also features interacting and changing relationships between emotion and cognitive control (Cohen-Gilbert & Thomas, 2013; Somerville, Jones, & Casey; Van Leijenhorst, et al., 2011), and elevated risk for multiple forms of psychopathology, including alcohol and substance abuse, mood disorders, eating disorders and conduct disorder (Paus, Keshavan, & Giedd, 2008).

These important cognitive and behavioral changes during adolescence are linked to concurrent structural and functional changes occurring in the adolescent brain (Casey, et al., 2008; Toga, et al., 2006). Structural MRI and histological studies have demonstrated a continuous increase in white matter volume during adolescence, thought to reflect ongoing myelination of axons within the brain (Giedd, et al., 1999; Gogtay, et al., 2004; Sowell, Thompson, Holmes, Jernigan, & Toga, 1999; Sowell, Trauner, Gamst, & Jernigan, 2002; Yakovlev & Lecours, 1967). Diffusion tensor imaging (DTI) data also support the existence of ongoing changes in white matter structural integrity during adolescence (Barnea-Goraly, et al., 2005; Bava, et al., 2010; Liston, et al., 2006; Schmithorst, Wilke, Dardzinski, & Holland, 2002). Higher-order association areas and prefrontal cortex show a peak in grey matter volume in early adolescence, followed by a gradual decline that may reflect continued synaptic pruning (Giedd, et al., 1999; Gogtay, et al., 2004). As a number of brain metabolites, detectable using MRS, have physiological relevance to the expression of structural changes, e.g. choline (Cho) reflective of myelination, N-acetyl-aspartate (NAA) reflective of grey matter integrity, and phosphomonoesters (PME) and phosphodiesters (PDE) reflective of membrane synthesis and breakdown (described in detail below, see also Table 1), integration of measures across structural and neurochemical studies in healthy adolescents and adolescents with psychopathology are warranted.

Table 1.

Physiological significance of MRS metabolites

1H MRS	Physiological significance
N-acetyl-aspartate (NAA)	Marker of neuronal integrity. Reductions often indicate tissue pathology.
Choline (Cho)	Involved in pathways of cellular membrane synthesis and degradation.
Creatine + phosphocreatine (tCr)	Markers of cellular energetic state.
Myo-Inositol (mI)	Involved in phospholipid metabolism and maintenance of osmotic equilibrium.
Glutamate (Glu)	Excitatory neurotransmitter and key molecule in cellular metabolism. Also a precursor for GABA synthesis.
Glutamine (Gln)	Precursor for glutamate and plays a role in protein synthesis.
GABA	Inhibitory neurotransmitter and key molecule in cellular metabolism.
“GLX”	Combination of Glu, Gln, GABA resonances.
Lactate (LAC)	By-product of anaerobic metabolism.
31P MRS
Phosphocreatine (PCr)	High-energy phosphate, contributes to the maintenance of β-NTP levels.
Nucleotriphosphates: α-, γ-, β-NTP	Level of energy available as ATP in brain (β-NTP).
Phosphomonoesters (PME)	Building blocks of membrane phospholipids.
Phosphodiesters (PDE)	Major catabolic products of membrane phospholipid degradation.
Inorganic Phosphate (Pi)	High-energy phosphate that combines with creatine to form PCr, but also is released from phosphocreatine to synthesize nucleoside triphosphates (ADP to ATP). Chemical shift of inorganic phosphate can be used to calculate intracellular pH. PCr/Pi ratio provides a measure of the energy status in brain, as it is a ratio of the most labile form of high-energy phosphate (PCr) to the ultimate breakdown product of all high-energy phosphate compounds (Pi).

Functional neuroimaging studies suggest that the brain regions recruited for executive tasks also change significantly during the second decade of life. Studies using fMRI have documented a shift from diffuse to more focal activity and/or greater reliance on the prefrontal cortex (PFC) between late childhood and early adulthood during performance of tasks requiring executive functions (Casey, et al., 1997; Crone, Wendelken, Donahue, van Leijenhorst, & Bunge, 2006; Durston, et al., 2006; Scherf, Sweeney, & Luna, 2006). This shift is thought to reflect improved efficiency in the recruitment of necessary regulatory circuitry, along with reductions in the overall effort necessary to perform executive tasks (Durston, et al., 2006; Luna, Padmanabhan, & O'Hearn, 2010). In addition to changes observed in prefrontal function, developmental changes occur during adolescence in wide-spread circuits including parietal, temporal, cerebellar and thalamic regions. Development of these cortical-subcortical circuits also supports age related improvements in inhibitory control (Rubia, Smith, Taylor, & Brammer, 2007). Functional neuroimaging studies also show evidence that brain circuits involved in reward processing and emotion regulation are differentially activated in adolescence versus in childhood or adulthood (Ernst, et al., 2005; Galvan, et al., 2006; Hare, et al., 2008). It has been suggested that there is a developmental shift towards reliance on more frontal regions during reward processing and decision-making during adolescence (Casey, et al., 2008), supporting improvements in self-regulation and future-oriented decisions. As with structural studies, brain metabolites that can be acquired using MRS can be linked to processes associated with changes in functional brain activation. For example, phosphocreatine (PCr) reflects available energy resources, and beta nucleoside triphosphate (β-NTP) reflects the brain’s available ATP. Neurochemicals measurable via MRS also reflect excitatory and inhibitory neurotransmission: Glu is the brain’s primary excitatory neurotransmitter while GABA reflects inhibitory neurotransmission, both of which are also involved in glucose metabolism. Also potentially impacting functional brain activation, lactate is indicative of anaerobic metabolism when energy substrates are taxed or unavailable due to neurological abnormalities (described in detail below, see also Table 1).

Thus, to extend structural and functional developmental findings, MRS can be employed to assess neuronal health and integrity, bioenergetics, neurotransmission, and cell membrane synthesis and degradation. Use of this technique to identify neurobiological processes operating on a cellular level will extend our understanding of adolescent developmental changes in brain structure and function, as well as cognition, behavior, and risk for psychopathology.

Magnetic Resonance Spectroscopy

Magnetic resonance spectroscopy permits in vivo detection of multiple brain metabolites and some neurotransmitters. MRS data can be acquired using most standard scanners and often spectra can be acquired within the same imaging session as other structural and functional scans. In all forms of MRI, when a sample is placed into a large static magnetic field (B₀), each nucleus precesses (or spins) about its axis, with some nuclei aligned with, and some nuclei aligned against the direction of B₀. The frequency of this nuclear precessional rotation is unique to the resonant (Larmor) frequency of a specific nucleus at a specific magnetic field strength. When a pulse from a radiofrequency (RF) coil is transmitted at the Larmor frequency for a particular nucleus (e.g. hydrogen), a transient magnetic field (B₁) is generated, causing the targeted nuclei to transition to a higher energy state by absorbing the energy from the RF pulse. When the RF pulse is turned off, the nuclei within the sample release their energy in the form of a radiofrequency signal that contains spectral information encoded within it. In MRS, this information-rich signal is detected by the RF coil (acting as a receiver) and subsequently plotted as energy released (signal intensity) as a function of time, and then converted to a series of spectral peaks, whereby signal intensity can be visualized in the frequency domain (Hz), which is independent of magnetic field strength.

Protons (¹H) (Figure 1) and phosphorus (³¹P) (Figure 2) are the two nuclei most commonly studied via MRS, although detection of resonance intensities is also possible for carbon (¹³C), sodium (²³Na), sulfur (³³S), fluorine (¹⁹F), and lithium (⁷Li). The use of MRS is limited to these and several other nuclei, based on their physical properties, i.e., possessing a resonance frequency that responds, or aligns, within a magnetic field, and the release of energy after perturbation following application of an RF pulse, all of which make these nuclei MR visible. For a given nucleus (e.g. ¹H), chemically distinct groups within a molecule that contain the given nucleus possess minor differences in their local resonant frequencies due to the inhomogeneous and unique distribution of electrons within the molecule (Bovey, 1988). These small differences in resonant frequency, or chemical shift, make it possible to differentiate molecules based on their distinct spectral signatures (e.g., Figure 1 - ¹H spectrum, Figure 2 - ³¹P spectrum). The area of the resonance intensity, or the area under the peak, is proportional to the concentration of molecules that contribute to the resonance, thus allowing for the estimation of the concentration of a particular molecule within the sample.

Figure 1. Sample ¹H MR spectrum.

Sagittal (left) and axial (right) anatomical images illustrating the placement of a single voxel (2 × 2 × 3cm) in the anterior cingulate cortex of a healthy subject, and the associated ¹H spectrum (below). Abbreviations: tCr, total creatine; Gln/Glu, glutamine/glutamate; mI, myo-inositol; tCho, total choline; GABA, gamma amino butyric acid; Glu, glutamate; NAA, N-acetylaspartate.

Figure 2. Sample ³¹P MR spectrum.

Sagittal (left) and axial (right) anatomical images illustrating extracted spectral data from a 3D CSI voxel grid (2.1 × 2.1 × 2.1cm) from a region placed in the anterior cingulate cortex of a healthy subject, and the associated ³¹P spectrum (below). Abbreviations: PME, phosphomonoesters; PE, phosphoethanolamine; PC, phosphocholine; Pi, inorganic phosphate; PDE, phosphodiesters; GPE, glycerophosphoethanolamine; GPC, glycerophosphocholine; PCr, phosphocreatine; NTP, nucleoside triphosphate.

Nuclei that are capable of producing a resonance signal, and are therefore considered MR visible, differ in detection sensitivity based on their electromagnetic properties and concentrations, which must be high enough (millimolar range) to be detected. Sensitivity, or intensity of the MR signal, can be amplified by increasing the size of the voxel (or volume) sampled and/or by increasing the strength of the magnetic field. While larger voxels have more limited anatomical specificity, the use of stronger magnetic fields is limited by the availability of high field scanners. While a large number of MRS publications have included spectral data acquired at low field (1.5T), more recently, MRS data have been reported at increasingly higher field strengths (3T, 4T, 7T and above) (Mangia, et al., 2006; Mekle, et al., 2009; Tkac, et al., 2001; Tkac, Oz, Adriany, Ugurbil, & Gruetter, 2009). To date, however, published MRS studies in healthy children and adolescents have only been based on data acquired at 4T (one study) or lower field strengths, the majority of which have been acquired at 1.5T. Increases in field strength over the past decade have been particularly valuable for technological advancement of MRS methods, as sensitivity increases linearly with field strength, due in part to greater spectral dispersion, or separation of metabolite peaks, which allows for improved quantification. Although the current trend is to acquire spectral data at relatively short echo times (e.g., 2–20msec) in order to minimize signal loss, in some instances a metabolite-specific echo time is necessary to optimize metabolite detection (e.g., TE=68msec for GABA).

In addition to identifying the nucleus of interest, hypotheses regarding anatomical structures of interest should also be considered when choosing the sequence for spectral acquisition. Spectra are acquired from either single anatomical volumes (single voxel) or from multiple volumes in two- or three-dimensions within the brain, each of which has their own strengths. Multi-dimensional techniques such as chemical shift imaging (CSI) permit the collection of a matrix of spatially-resolved spectral data from a large region of tissue selected using high resolution anatomical MRI. CSI permits the extraction of data from multiple voxels during acquisition, and during post-processing allows for regional examination of metabolite levels (Brown, Kincaid, & Ugurbil, 1982; Wiedermann, et al., 2001). Thus CSI is well suited for exploratory studies, such as those characterizing major metabolites (NAA, Cho, creatine (Cr)) across multiple regions during brain development. Single voxel MRS, on the other hand, permits focus on a small number of discreet brain regions with higher spatial and spectral resolution, which is necessary for regions with increased susceptibility to field inhomogeneties, such as the frontal lobe or the hippocampus. Single voxel acquisition is also necessary when employing specialized acquisition schemes that are optimized for detecting complex metabolites, such as GABA or Glu.

The choice of MRS quantification strategy depends on the nucleus of interest to be examined. The most commonly used commercially available software program employed for the quantification of in vivo proton MR spectra is Linear Combination of Model Spectra (LCModel) (Provencher, 1993, 2001). LCModel fits spectra by comparing in vivo raw spectral data either with in vitro metabolite data collected under identical conditions or with simulated basis sets. Phosphorous spectra are generally fitted in a similar fashion, typically using in-house software that fits spectra in the frequency or time domain using linear and interactive algorithms, similar to those used in LCModel.

Some MRS studies report absolute values while others report ratios of one metabolite to another. In practice, calculating the absolute values of metabolite concentrations requires knowledge of the T₁ and T₂ relaxation times of the molecule of interest, the repetition time (TR) and echo time (TE) acquisition parameters, the tissue volume of interest, and the efficiency of signal detection. Calculation of absolute concentrations in LCModel uses a theoretical fitting routine (basis set) that incorporates the spectroscopic imaging parameters (TR, TE) and nuclei-specific relaxation times. However, other factors specific to the volume of interest (partial volume effects or tissue content) and to the MR scanner are not incorporated in these calculations, which make metabolite concentrations very liberal, and perhaps unreliable, estimates of true metabolite levels. Given these difficulties, it is common practice to report relative values of MRS data as metabolite ratios. There remains, however, a significant debate regarding the optimal reference standard for determining metabolite ratios. Metabolite ratios can be calculated using a number of strategies. For example, metabolite peaks can be quantified relative to an external standard, typically a small capsule containing known amounts of metabolites placed near the subject’s head. This method increases scan time, as spectra collected from the in vitro, external standard must be acquired within the same scanning session with the patient in the scanner. Alternatively, reference data can also be acquired from a phantom following removal of the subject from the scanner. However, this method is subject to variations due to differences in coil loading between the subject’s head position and that of the phantom. More recently, the use of an Electric Reference To access In vivo Concentrations (ERETIC) method has been applied for normalizing metabolite levels (Chen, Pavan, Heinzer-Schweizer, Boesiger, & Henning; Heinzer-Schweizer, et al.). This method introduces an additional synthetic reference signal during the spectral acquisition, and is not only sensitive to patient coil loading, but can be quantified independent from in vivo metabolites.

In terms of normalizing to internal standards, the unsuppressed peak arising from water is routinely used to calculate metabolite ratios, but is complicated by the need to discriminate between water in tissue versus water in CSF, which can differ by up to 30–40%. Another common strategy is to use a relatively stable metabolite peak, such as Cr. In proton MRS, the Cr peak represents the concentration of Cr plus PCr, which is typically maintained at constant levels in healthy tissue. However, like water, Cr levels vary between brain regions and tissue types and may also differ between subject populations, complicating interpretations of metabolite ratios when Cr serves as the denominator. Furthermore, because the chemical composition of grey and white matter differ, it is important to assess the tissue content of the sampled brain regions. High resolution anatomical MR images used to define voxels for spectral acquisition allow for tissue segmentation and subsequent calculation of relative metabolite levels in grey versus white matter (Pouwels & Frahm, 1998). Nonetheless, choosing a quantification strategy remains a significant challenge. It is likely that this technological issue will continue to be a major focus of research for the next several years.

While studies utilizing MRS yield an abundance of information about both the structure and chemical composition of tissues, MRS technology is limited by a number of factors. Optimal signal to noise ratio and a homogenous magnetic field are necessary to obtain narrow resonance peaks for quantification, which may require 10 to 45 minutes of scanning, free of motion artifact. As with other MR techniques, subject or patient comfort is critical for minimizing subject motion, which degrades spectral quality. Perhaps the most significant limitation of MRS is low sensitivity. The signal strength of a particular nucleus depends upon its inherent signal intensity and the externally applied magnetic field strength. Sensitivity limitations can be minimized by acquiring spectra at higher field strengths and by applying optimized strategies for spatial localization.

MRS during Healthy Brain Development & Relationships with Cognition

Few studies have been published regarding age-related changes in neurochemical metabolites focused on the first two decades of life. As reviewed in subsequent sections, these studies have used cross-sectional research designs and focus on childhood or the full life span, with little data available regarding neurochemical changes specifically occurring during adolescence. While some data are available regarding the relationships between MRS-visible metabolite concentrations and cognitive functioning in childhood and adolescence, the majority of such studies have been conducted in adults. Regardless, these studies have great value and offer promising new directions for future research in adolescent populations, as brain metabolite levels are predictive of cognitive performance, but also of impairments in cognitive functioning that are associated with psychopathological conditions, such as alcohol use disorders. Accordingly, results from selected studies examining metabolite-cognition relationships in adults are included here in order to elucidate the potential cognitive relevance of developmental changes in MR-visible metabolites.

Development and Cognition: ¹H MRS Findings

In vivo ¹H MRS permits the detection and quantification of important amino acids that contain proton nuclei, including NAA, cytosolic choline compounds (Cho), Cr, myo-inositol (mI), and lactate (Figure 1). Albeit near the lower limit of detection, Glu, glutamine (Gln), and GABA can also be quantified in the proton spectrum (Behar & Rothman, 2001; Jensen, Frederick, & Renshaw, 2005; Licata, et al., 2009; Petroff, Mattson, & Rothman, 2000; Shulman, Rothman, Behar, & Hyder, 2004). Physiological relevance of proton and phosphorous brain metabolites are provided in Table 1 and developmental ¹H MRS findings are summarized in Table 2 and Figure 3.

Table 2.

¹H MR Spectroscopic findings in healthy children and adolescents

Investigators	Sample	MRS	Region(s) of interest	Metabolites/findings
van der Knaap et al. (1990)	1 mo – 16 yrs. mean = 5.9 yrs. (n = 41) no sedation	SV, 1.5 T 7 × 3 × 3 cm	paraventricular region, predominantly WM	-NAA/Cho and NAA/Cr increases with age -Cho/Cr decreases with age -Most rapid changes from 1 to 3 years
Kreis et al. (1993)	35 wks – 17.8 yrs. (n = 109 scans) healthy (18%) recovered infants (21%) cerebral pathology (61%) chloral hydrate	SV, 1.5 T 3–8 cm3 young 8–16 cm3 older	occipital cortex (GM) parieto-occipital (WM)	-Cr, myo-I increases until 2 yrs -NAA and Cho increases until 7 yrs.
Toft et al. (1994)	infants 259 – 295 days mean = 277 days (n = 8) adolescents 10 – 15 yrs. mean = 12.3 yrs. (n = 8) no sedation	SV, 1.5 T 2 cm3	infants: 1 voxel including caudate, putamen, globus pallidus adolescents: 4 voxels occipetal, basal ganglia, temporal, frontal. All voxels included GM and WM.	-NAA, PCr+Cr, Glu+/H2O higher in adolescents -Cho higher in infants -Inositols similar across groups
Pouwels et al. (1999)	0 – 18 yrs. (n = 97) healthy (8%) neuropediatric (92%) 18 – 39 yrs. (n = 72) <6 yrs. chloral hydrate	SV, 2.0 T 8–18 mL 8–18 mL 4–5 mL 4–6 mL 4–6 mL	parietal GM parieto-occipital WM cerebellum, vermis thalamus basal ganglia	-NAA increases in GM, cerebellum, thalamus with age up to 2 yrs, then stabilizes -NAA constant in WM, basal ganglia -Cr no change after 1 yr., highest level in cerebellum, followed by basal ganglia, thalamus, parietal GM, parieto-occipital WM -Cho declines in WM after 5 yrs
Choi et al. (2000)	3 – 14 yrs. mean = 9 yrs. (n = 30) young chloral hydrate	SV, 1.5 T 1.8 × 2 × 2 cm	allocortex (hippocampus, parahippocampal) isocortex (medial, frontal and parietal)	-NAA/Cr lower in allocortex than isocortex -Cho/Cr and mI/Cr higher in allocortex than isocortex -No age effects
Kadota et al. (2001)	4 – 88 yrs. mean=45.6 yrs. (n = 90) no sedation	CSI, 1.5 T 1.125 cm3	superior to corpus callosum, 12 voxels: 6 WM, 6 mesial GM, placed bilaterally in anterior, middle and, posterior regions	-NAA/Cho in WM rapidly increases until 10–20yrs. -NAA/Cho in WM declines starting in late 30s Laterality in males for NAA/Cho in WM -Steeper increase in WM NAA/Cho in males than in females -NAA/Cho in GM gradually declines linearly with age
Horska et al. (2002)	3 – 19 yrs. mean = 12.3 yrs. (n = 15) n = 2 nembutal	CSI, 1.5 T .8cm3	frontal, parietal (WM) basal ganglia (GM) thalamus (GM)	-NAA/Cho in GM peaks at 10 yrs. -NAA/Cho in WM increases with age
Costa et al. (2002)	3 – 18 yrs. (n = 37) no sedation	SV, 1.5T 8cm3	parieto-occipital WM cerebellar hemisphere	-NAA/Cr, Cho/Cr lower in cerebellum than parieto-occipital WM -NAA/H2O, Cr/H2O and Cho/H2O higher in the cerebellum than parieto-occipital WM -NAA/H2O increased with age in the cerebellum and parieto-occipital WM -Cho/H2O increased with age in the cerebellum
Goldstein et al. (2009)	6 – 18 yrs. (n = 105) no sedation	CSI, 1.5 T 1.5 × 1.5 × 2 cm	prefrontal cortex basal ganglia superior temporal cortex inferior parietal cortex centrum semiovale occipetal regions	-No significant differences in NAA between youngest (6–9.5 yrs.) vs. oldest (12–18 yrs.) group.
Raininko & Mattsson (2010)	13 – 72 yrs. (n=57) no sedation	SV, 1.5T 4 × 1.2 × 1.8 cm	supraventricular WM	-NAA increased with age -Glx showed U-shaped age dependance with highest concentrations in youngest and oldest subjects -No age dependence in Cho or Cr -No gender differences
Silveri et al. (2013)	12–14 yrs, (n = 30) 18–24 yrs. (n = 20) no sedation	SV, 4T 2 × 2 × 3 cm	anterior cingulate cortex parieto-occipetal cortex	-No differences between age groups for Cr -GABA/Cr lower in ACC but not POC in younger vs. older group -Gln/Glu lower in ACC but not POC in younger vs. older group

Abbreviations: SV, single voxel; CSI, chemical shift imaging; WM, white matter; GM, gray matter.

Note: Subjects in all studies were healthy, normally developing individuals except where noted otherwise. Findings reported in this table reflect healthy subjects only.

Figure 3. Summary of normative developmental ¹H MRS studies.

Age spans investigated across all available ¹H MRS studies of normative development. Ages in parentheses indicate plateau of age effects.

NAA contributes the largest signal in the proton spectrum at a chemical shift of 2.009 ppm. NAA is primarily found in mature neurons and has been viewed as an indicator of neuronal integrity (Birken & Oldendorf, 1989; Demougeot, Marie, Giroud, & Beley, 2004; Inglese, et al., 2008; Moffett, Ross, Arun, Madhavarao, & Namboodiri, 2007; Pouwels & Frahm, 1997; Sullivan, et al., 2001). Across the published developmental ¹H MRS studies, in vivo levels of NAA have most often been reported to increase with age, whether reported as a concentration, or as a ratio relative to Cho, Cr or the unsuppressed water signal (Costa, Lacerda, Garcia Otaduy, Cerri, & Da Costa Leite, 2002; Grachev & Apkarian, 2000; Horska, et al., 2002; Hüppi, et al., 1991; Kadota, Horinouchi, & Kuroda, 2001; Kreis, Ernst, & Ross, 1993; Pouwels, et al., 1999; van der Knaap, et al., 1990). NAA levels have been reported to increase most rapidly within the first few years of life, and plateau anywhere from 2 to 30 years of age depending on the region and tissue type studied. Two published studies, however, have failed to find any significant associations between NAA and age in samples spanning from 3 to 14 years (Choi, Ko, Lee, Lee, & Suh, 2000) and from 6 to 18 years (Goldstein, et al., 2009), using moderate to large sample sizes. NAA has been most widely accepted as a biomarker of neuronal integrity, but may also index neuronal density. Thus, early NAA increases with age may reflect the rapid changes in brain tissue volume occurring early in life. However, production of NAA also is related to energy metabolism, suggesting that developmental increases in NAA may also result from the global increases in glucose metabolism associated with brain maturation (Chugani, 1998).

In adults, NAA concentrations in a left occipito-parietal voxel have been found to correlate positively with a composite measure of neuropsychological performance (Jung, Yeo, Chiulli, Sibbitt Jr, & Brooks, 2000) and with full-scale IQ, as measured by the Wechsler Adult Intelligence Scale, 3^rd edition (WAIS-III) (Jung, et al., 2005). A later study using CSI reported both a positive relationship between NAA values in right posterior grey matter and performance IQ and a negative correlation between NAA in right anterior grey matter and verbal IQ (Jung, et al., 2009). Pfleiderer and colleagues (Pfleiderer, et al., 2004), however, found a positive relationship between vocabulary scores and NAA in the left dorsolateral prefrontal cortex, though this effect only reached significance in women. Studies exploring metabolite-cognition relationships in healthy elderly populations have revealed positive relationships between NAA and NAA/Cr ratios and performance on tests of executive function, attention, processing speed and memory (Charlton, McIntyre, Howe, Morris, & Markus, 2007; Ferguson, et al., 2002; Pfefferbaum, Adalsteinsson, Spielman, Sullivan, & Lim, 1999; Ross, Sachdev, Wen, Valenzuela, & Brodaty, 2005; Valenzuela, et al., 2000). Several studies have also reported associations between NAA and cognitive functioning within normative pediatric samples. In a sample of healthy 10 to18 year-olds, (Gimenez, et al., 2004) positive correlations were reported between measures of NAA/Cho in left hippocampus and measures of both verbal and non-verbal learning and memory. Similarly, (Yeo, et al., 2000) reported positive relationships between NAA and performance on a 2-back memory task in a sample of 7 to 12 year-olds. This study, however, found no relationship between NAA concentrations and a second memory measure - the Visual Temporal Order Memory Test. In contrast with earlier research (Jung, et al., 1999), neither of these studies found a significant relationship between metabolite measures and general intelligence. In a third developmental study, a large sample of healthy individuals between 6 and 18 years of age were examined (Ozturk, et al., 2009). Along with increased NAA/Cr with age in both parietal grey matter and white matter, this study reported a relationship between the NAA/Cr ratio in left frontal white matter and right-handed scores on the Perdue pegboard task of dexterity, while the same metabolite ratio in the right frontal white matter correlated with scores on a bead memory task. In a similarly aged sample (6 – 18 years), Goldstein and colleagues (2009) found a positive relationship between NAA levels and performance on a battery of visual-spatial construction tasks (Goldstein, et al., 2009). Taken together, these studies suggest both an increase in neural integrity with age, as indexed by NAA, and a positive relationship between NAA concentrations and cognitive functioning in both children and adults.

The largest Cho signal resonates at 3.22 ppm and reflects a number of choline-containing compounds, including free choline, phosphorylcholine, and glycerophosphocholine (PC and GPC, respectively) (Barker, et al., 1994). Choline-containing compounds (Cho) are involved in pathways of cellular membrane synthesis and degradation (Miller, 1991; Pouwels & Frahm, 1998), with most of the choline in the brain being membrane-bound as phosphatidylcholine, which is MR invisible (Miller, 1991). Declines in measurable Cho are consistent with accelerated myelination, as NMR visible choline residues become incorporated into MR invisible macromolecules associated with myelin production. Developmental findings in childhood and adolescence with regard to Cho are inconsistent, with some studies reporting an age-related decrease in Cho (Pouwels, et al., 1999; Toft, Christiansen, Pryds, Lou, & Henriksen, 1994; van der Knaap, et al., 1990), others reporting an age-related increase in Cho (Costa, et al., 2002; Kreis, et al., 1993), and some claiming no significant relationship between Cho levels and age (Raininko & Mattsson, 2010). These disparate results may be partly accounted for by differences between studies in voxel size and location. Looking at the ratio of NAA/Cho, a number of studies have found age-related increases, which may reflect both increases in NAA and decreases in Cho (Horska, et al., 2002; Kadota, et al., 2001; van der Knaap, et al., 1990). Not all studies, however, report developmental changes in this metric (Choi, et al., 2000; Costa, et al., 2002; Gimenez, et al., 2004). MRS studies of developmental delay in children have reported reduced NAA and elevated Cho levels, thought to reflect delayed myelination (Fayed & Modrego, 2005). Thus, simultaneous changes in NAA and Cho may reflect maturational processes critical to normative cognitive development. Consistent with these findings, in adults, Cho concentrations in a temporal voxel, were found to be negatively correlated with performance on visual memory and digit span tasks (Buckley, et al., 1994). In contrast, however, a positive relationship between vocabulary scores and Cho in the left dorsolateral prefrontal cortex (DLPFC) has been reported (Pfleiderer, et al., 2004). The conflicting results and lack of replication of MRS studies examining both developmental changes in measurable Cho and relationships between Cho concentrations and cognition make it difficult to draw firm conclusions regarding the roll of this metabolite in typical neurological and cognitive development at this time.

Singlet Cr resonances at 3.03 ppm and 3.93 ppm arise from protons in Cr and phosphorylated Cr (PCr). Cr plays a major role in brain energy metabolism, acting both as an energy buffer, by maintaining constant brain adenosine triphosphate (ATP) levels through the creatine kinase reaction, and by distributing energy (via mitochondria) within the brain (Kemp, 2000; Miller, 1991). There is some evidence that developmental changes in Cr levels plateau by the second year of life (Kreis, et al., 1993; Pouwels, et al., 1999), with the exception of a study documenting higher Cr levels when comparing infants to adolescents (Toft, et al., 1994). Infants tested in this study, however, were less than one year old and thus the results do not preclude the possibility that the Cr changes observed between infancy and adolescence occurred by two years of age. In healthy adults, a positive relationship has been observed between temporal Cr levels and visual memory span performance (Buckley, et al., 1994). In a sample of 7 to 12 year-olds, a similarly positive relationship was observed between Cr and 2-back memory task performance but not between Cr concentrations and a temporal order memory task (Yeo, et al., 2000). In elderly populations, however, Cr concentrations were found to correlate positively with increasing age and negatively with memory and executive function (Charlton, et al., 2007; Ferguson, et al., 2002; Valenzuela, et al., 2000). Thus, Cr increases may index metabolic changes both in the first two years of life and in old age. Furthermore, Cr concentrations show differing relationships with cognitive skills at different stages of development, suggesting a changing significance of this metabolic marker with age.

The most pronounced mI resonance occurs at 3.54 ppm. mI is involved in the synthesis and turnover of phospholipid membranes, maintenance of osmotic equilibrium, and is a phosphorylated derivative in the synthesis of secondary messengers (Berridge & Irvine, 1989; Brand, Richter-Landsberg, & Leibfritz, 1993; Downes & Macphee, 1990; Kim, McGrath, & Silverstone, 2005; Moore, et al., 1999; Wolfson, et al., 2000). mI concentrations have been reported to increase with age until reaching a plateau by age 2 (Kreis, et al., 1993). However, increases in mI have also been detected in early adulthood (Grachev & Apkarian, 2000) and across the lifespan (Raininko & Mattsson, 2010). Concentrations of mI, as measured via MRS, are yet to be linked to cognitive performance in pediatric or young adult populations. However, increases in the mI/Cr ratio have been observed in cases of mild cognitive impairment and Alzheimer’s disease in aging populations (Catani, et al., 2001; Kantarci, et al., 2002).

Lactate, or lactic acid, increases significantly when the brain is deprived of oxygen, or anaerobic respiration increases. Thus, lactate is an important metabolic marker. The lactate resonance is a doublet at 1.33 ppm, which is obscured by a resonance arising from lipids (Auer, Gossl, Schirmer, & Czisch, 2001; Behar, Rothman, Spencer, & Petroff, 1994), but can be reliably resolved when a long time to echo (TE) is employed (Behar, et al., 1994). To date, no age or cognition-related changes have been associated with lactate concentrations in healthy children or adolescents. An age related increase in lactate has, however, been reported in a young adult sample (19–31 years), when data were averaged across multiple voxels (Grachev & Apkarian, 2000).

Gln, Glu, and GABA all play multiple critical roles in neurological functioning. Gln serves as a precursor for Glu, which is a major excitatory neurotransmitter found in all brain cell types, with the highest concentrations typically observed in neurons. GABA is the major inhibitory neurotransmitter found in the mammalian brain (McCormick, 1989). Gln, Glu and GABA also play important roles in glucose metabolism, neuronal energetics and ammonia detoxification (Behar & Rothman, 2001; Patel, et al., 2005). While brain Glu is present at much higher concentrations than GABA, only a small fraction of Glu participates in neurotransmission. These metabolites each consist of multiple peaks that have strong spectral overlap, and thus are obscured by peaks of higher concentrations, such as Cr. For instance, GABA resonates at three chemical shift positions, triplet peaks at 2.31 and 3.01 ppm and a quintet at 1.91pm. Glu also consists of a number of multiplet peaks, although the majority of the Glu signal is observed at 2.35 ppm. Since there is considerable overlap of Gln and Glu, predominantly around 3.75 ppm, a combined "GLX" resonance intensity is often reported particularly when spectra are acquired at low field strength (1.5 Tesla). For example, higher GLX concentrations have been reported to predict inferior global cognitive functioning in patients with Type 1 diabetes (Lyoo, et al., 2009). Recently, increasing availability of high field MR scanners and specialized editing techniques have enhanced MR visibility of these metabolites by facilitating separation of resonances, allowing more accurate quantification of individual Glu, Gln and GABA resonances (Jensen, et al., 2009; Keltner, Wald, Frederick, & Renshaw, 1997; Mescher, Merkle, Kirsch, Garwood, & Gruetter, 1998; Weber, Trabesinger, Duc, Meier, & Boesiger, 1997). One such study in found increased GABA in the frontal eye fields (FEF) predicted diminished impact of distractors on directed visual saccades in healthy adults (Sumner, Edden, Bompas, Evans, & Singh). Furthermore, in an all-male adult sample, higher GABA concentrations in DLPFC were found to correlate with lower self-report based scores of rash impulsivity (Frederic Boy, et al., 2011). Higher Glu in medial prefrontal cortex was found to be associated with poorer performance on a continuous performance task (CPT) in an adult sample including both the healthy siblings of patients with schizophrenia and healthy control subjects (Purdon, Valiakalayil, Hanstock, Seres, & Tibbo, 2008). In a study of young adults, GABA, Glu and Gln were all found to be higher in a 21–31 year-old group when compared to a 19–20 year-old group (Grachev & Apkarian, 2000). Our preliminary study (Silveri, et al., 2013) provides the first in vivo human brain evidence of lower GABA/Cr in adolescence relative to adulthood, measured at 4T, in 30 healthy adolescents relative to 20 emerging adults in the anterior cingulate cortex (ACC), but not in a control region (parieto-occipital cortex). Higher ACC GABA significantly predicted better accuracy on the NoGo trials of a Go NoGo task and lower overall BIS trait impulsivity. Given this developmental increase in ACC GABA observed during adolescence, and the predictive nature of brain GABA levels on response inhibition, motor control and impulsivity (F. Boy, et al., 2010; Frederic Boy, et al., 2011; Silveri, et al., 2013), further characterization of brain metabolites and potential links to alcohol abuse disorders in adolescence is warranted.

In a young adult sample (19–31 years), Grachev & Apkerian (2000) detected elevated NAA in females relative to males in sensorimotor cortex and higher lactate in males relative to females in the dorsolateral prefrontal cortex. To date, no published studies explore sex differences in ¹H metabolites in healthy children or adolescents. Using ¹H MRS to explore neurochemical changes paralleling the sexual differentiation that occurs in the neuro-endocrine system during puberty may prove to be a particularly fruitful area of study. Increased sexual dimorphism during this period may also result in divergent cognition-metabolite relationships in men and women. Exploring such relationships may provide key insights into sex differences in emotion and reward processing that emerge during this period.

Development and Cognition: ³¹P MRS Findings

In vivo ³¹P MRS permits the detection of high-energy phosphate metabolites and constituents of membrane synthesis, which reflect cellular bioenergetic state and cell membrane integrity and function, respectively (Table 1). Phospholipid metabolites associated with high-energy intracellular metabolism that are detectable include PCr, inorganic phosphate (Pi), and β-NTP, primarily reflecting adenosine triphosphate (ATP) in the brain. Components of cell membranes detectable with ³¹P MRS include PME and PDE (see Figure 2 for a sample ³¹P spectrum). For ³¹P MRS, the chemical shift of each metabolite is referenced relative to PCr, which is used as an internal standard. Physiological relevance of each phosphorous metabolite is listed in Table 1 and developmental ³¹P MRS studies are summarized in Table 3 and Figure 4.

Table 3.

³¹P MR Spectroscopic findings in healthy children and adolescent

Investigators	Sample	MRS	Region(s) of interest	Metabolites/findings
Boesch et al. (1989)	33 wks – 6 yrs. (n = 40, 48 exams) healthy (25%) cerebral pathology (75%) chloral hydrate	SV, 2.35 T	bilateral fronto-temporal regions	-PME/PDE declines until 1.3 yrs. -PCr/β-NTP increases until 1.3 yrs.
van der Knaap et al. (1990)	1 mo – 16 yrs. mean=5.9 yrs. (n = 41) no sedation	SV, 1.5T	paraventricular region, predominantly WM	-PME/β-NTP decreases with age -PDE/β-NTP, PCr/β-NTP, PCr/Pi increases with age
Moss & Talagala (1997)	mean = 13.4 yrs. (n = 29) sedation not specified	CSI, 1.5 T 3.5cm3	mesial frontal lobe mesial occipital lobe	-females, β-NTP higher occipital than frontal -males, β-NTP higher in frontal than occipital -males, higher PDE in occipital than frontal
Hanaoka et al. (1998)	4 mo – 13 yrs. (n = 37) sedation not specified	SV, 2.0 T 60–90cm3 40–60cm3	bilateral fronto-parietal cerebrum bilateral cerebellar hemispheres	-PME/PDE decreases until 2 yrs in cerebellum -PME/PDE decreases with age in cerebrum -PCr/γ-NTP increases 1–2 yrs, peaks at 8 yrs
Goldstein et al. (2009)	6 – 18 yrs. (n = 105) no sedation	CSI, 1.5 T 3 × 4.5 × 3 cm	prefrontal cortex basal ganglia superior temporal cortex inferior parietal cortex centrum semiovale occipetal regions	-PCr increases from 6–9.5 yrs., plateaus between 9.5–12 yrs. and increases again between 12–18 yrs. -PME/PDE lower in youngest (6–9.5 yrs.) vs. oldest (12–18 yrs.) age group.

Abbreviations: SV, single voxel; CSI, chemical shift imaging; WM, white matter; GM, gray matter.

Note: Subjects in all studies were healthy, normally developing individuals except where noted otherwise. Findings reported in this table reflect healthy subjects only.

Figure 4. Summary of normative developmental ³¹P MRS studies.

Age spans investigated across all available ³¹P MRS studies of normative development. Ages in parentheses indicate plateau of age effects.

High-energy PCr serves as a buffer for maintenance of β-NTP levels and shuttles energy from sites of production to sites of utilization (Bessman & Geiger, 1981; Wallimann, Wyss, Brdiczka, Nicolay, & Eppenberger, 1992). Thus, availability of PCr stimulates the production of β-NTP via conversion to Cr and high-energy phosphate (Pi) (Wallimann, et al., 1992). This process results in a reduction of PCr levels, while the levels of ADP and Pi increase to support maintenance of β-NTP levels (Gyulai, Roth, Leigh, & Chance, 1985). The ratio of PCr relative to Pi, therefore, is thought to reflect phosphorylation potential, or available energy within a cell (Nioka, et al., 1990). The chemical shift of Pi can also be used to calculate internal pH level, which serves to modulate synaptic transmission and plasticity, and neuronal excitability, and can aid in the discrimination between diseased and healthy tissue (de Graaf, 2002). In general, the ratio of PCr to both β-NTP and γ-NTP increase most rapidly within the first few years of life and then increase more gradually, reaching a plateau around age eight (Boesch, Gruetter, Martin, Duc, & Wuthrich, 1989; Hanaoka, Takashima, & Morooka, 1998; van der Knaap, et al., 1990). More recent work, however, suggests that while PCr values may level off in late childhood and early adolescence, with the positive relationship between PCr and age once again becoming significant between 12 and 18 years of age (Goldstein, et al., 2009). Therefore, it has been posited that increases in cellular energy stores enable synaptic remodeling – an energetically demanding process. Goldstein and colleagues (2009) also reported positive relationships between PCr levels and age-related improvements in composite measures of language abilities, visuospatial construction and memory, but not executive functioning. In contrast to these findings in adolescents, Voltz and colleagues (Volz, et al., 1998) found a negative relationship between energy cycling, indicated as PCr/ATP, and executive function performance measured via the Wisconsin Card Sorting Test, in healthy adults. Taken together, these findings suggest a changing relationship between energy cycling and executive functioning skills, with developmental changes likely reflecting a change over to increased neuronal efficiency (Casey, et al., 2008; Durston, et al., 2006), which requires fewer energy resources, such as PCr and β-NTP, for the individual to perform at a higher level. This represents a maturational change from the more diffuse activation typically observed during performance of higher order cognitive functions during childhood and early adolescence, where PCr levels are not directly coupled to executive functioning abilities.

Components of cell membranes detectable with ³¹P MRS include PME and PDE. The PME resonance arises primarily from phospholipid membrane precursors (Pettegrew, et al., 1991), while the PDE resonance arises from the presence of phospholipid catabolites. Therefore, the PME peak reflects membrane synthesis and the PDE peak reflects products of membrane breakdown while the ratio of PME/PDE indexes membrane phospholipid turnover. At high field (4.0 Tesla and higher), subpeaks within the PME peak can be further quantified into phosphoethanolamine (PE) and phosphocholine (PC) anabolites, and in the PDE peak, glycerophosphoethanolamine (GPE) and glycerolphosphocholine (GPC) catabolites, allowing more precise quantification of metabolites and reflecting concurrent changes in the biophysical state of membrane phospholipids. A decline in the ratio of PME/PDE has been observed with age, until approximately two years of age, in both the cerebrum and the cerebellum (Boesch, et al., 1989; Hanaoka, et al., 1998; van der Knaap, et al., 1990). Significant reductions in PME/PDE with age have also observed between childhood and adolescence (Goldstein, et al., 2009). While these age-related changes in the ratio of PME/PDE may reflect changing rates of membrane synthesis and degradation, likely associated with developmental profiles of pruning and myelination, decreases in PME may also be due in part to the incorporation of membrane precursors into larger membrane macromolecules (e.g., synaptic receptors comprised of large protein structures), which are largely MR invisible.

From the limited data on sex differences in ³¹P metabolites during childhood and adolescence, males do not appear to differ from females in terms of overall age-related changes in metabolite levels. However, sex differences in the regional distributions of metabolites have been reported during this age period in a single study (Moss & Talagala, 1997). Higher β-NTP has been observed in frontal versus occipital regions in males, whereas females exhibit higher β-NTP in the occipital region relative to frontal regions. Furthermore, males demonstrate higher PDE levels in the occipital region than the frontal region (Moss & Talagala, 1997). Sex differences in ³¹P metabolites may therefore reflect the well-documented sex differences in structural and functional brain changes during childhood and adolescence, as well as the cognitive differences between males and females. For example, the more rapid increase in white matter in males versus females and an earlier peak in grey matter volume in females versus males may be reflected in differential neurometabolite profiles. Thus, sex differences in the timing of structural brain changes may result in variable energy demands and rates of membrane turnover in various brain regions, affecting the concentrations of β-NTP and PDE, respectively. Further research is needed to explore this possibility.

MRS in Alcohol and Substance Abuse and Dependence

Risk-taking is a key feature of adolescence, which includes experimentation with alcohol and drugs. Prevalence of alcohol use increases from 2.6% at age 12 to 67.5% at age 21 (SAMHSA, 2004). Quantity of alcohol consumed likewise increases during adolescence, sometimes reaching binge-like levels. A study of of risk-taking behaviors among U.S. teenagers found that 25.5% of all students had drunk more than 5 drinks on one occasion (i.e., a binge episode) during the previous month (Eaton, et al., 2006). Both heavy episodic use and early initiation of alcohol use are associated with higher rates of abuse and dependence (Grant & Dawson, 1997; Nigg, et al., 2006). It is clear that the onset of drinking during adolescence raises a serious public health concern, as it increases dangerous risk-taking behaviors already potentiated by adolescent brain development and can have a negative impact on neurological development during an important period of brain reorganization.

Cognitive alterations are frequently reported in association with alcohol abuse and dependence in adult populations (Oscar-Berman, 1990; Oscar-Berman, 2000; Oscar-Berman & Marinkovic, 2003; Parsons & Nixon, 1998). Accordingly, MR studies have documented a wide variety of associated neurobiological changes, which highlight the frontal lobe and hippocampus as being the most compromised by chronic alcohol use in adult cohorts (Harris, et al., 2008; Marlene Oscar-Berman & Marinkovic, 2007; Paulus, Tapert, Pulido, & Schuckit, 2006; Pfefferbaum, Sullivan, Mathalon, & Lim, 1997; Sullivan & Pfefferbaum, 2005). More recently, a growing body of literature has likewise demonstrated alterations in brain structure and function in frontal networks and the hippocampus, in adolescents and young adults with alcohol use disorders (DeBellis, et al., 2000; Tapert, et al., 2001; Tapert, Pulido, Paulus, Schuckit, & Burke, 2004; Tapert, Schweinsburg, et al., 2004) and those reporting binge alcohol consumption (McQueeny, et al., 2009; Schweinsburg, McQueeny, Nagel, Eyler, & Tapert, 2010; Squeglia, Schweinsburg, Pulido, & Tapert, 2011), as well as those who have not yet begun to drink alcohol, but who are family history positive for alcoholism (Hill, et al., 2009; Silveri, Rogowska, McCaffrey, & Yurgelun-Todd, 2011; Spadoni, Norman, Schweinsburg, & Tapert, 2008).

While no studies to date have used MRS to investigate the consequences of alcohol use on neurochemistry during adolescence, numerous MRS studies conducted in adult cohorts have demonstrated abnormalities in brain metabolites associated with alcohol use, abuse and dependence. In the majority of these published MRS studies, which have largely been confined to recently detoxified alcoholic adult males (>35 years old), substantial evidence for reductions in NAA and Cho, mostly in frontal lobe regions of interest, has been reported (for a comprehensive review and table of MRS studies in alcohol abusing and dependent populations, see Meyerhoff, Durazzo, & Ende, 2011). There is less evidence of abnormalities in Cr and mI associated with chronic alcohol use, although when reported, decreased Cr likely reflects altered cell bioenergetics and elevated mI reflects proliferation of glial cells (Meyerhoff, et al., 2011). Alterations in metabolite levels have been observed during acute abstinence, with recovery of metabolite levels increasing with length of alcohol abstinence in dependent populations that exhibit physiologic withdrawal (Meyerhoff, et al., 2011). Importantly, recovery of brain metabolites is predictive of restoration of cognitive functioning, underscoring the dual utility of MRS, to capture the impact of alcohol use on brain chemistry, but also to identify neurochemical correlates associated with recovery.

There have been fewer MRS studies that have quantified GABA and Glu metabolite levels in heavy alcohol-using adult populations, which is extremely relevant given that these compounds are central targets of alcohol action, and that heavy alcohol use, dependence and withdrawal have been shown to be associated with neurobiological alterations in GABAergic and glutamatergic systems (Krupitsky, et al., 2007; Krystal, Petrakis, Limoncelli, et al., 2003; Krystal, Petrakis, Mason, Trevisan, & D'Souza, 2003; Krystal, et al., 2006; Tsai & Coyle, 1998). Of the limited studies available, relative to healthy comparison subjects, detoxified alcoholics exhibited significantly lower occipital GABA levels, which was negatively correlated with impaired verbal recall (Behar, et al., 1999). More recent studies have reported no significant group differences in ACC or occipital GABA in alcohol dependent patients studied early in abstinence (Abe, et al., 2012; Mason, et al., 2006; Mon, Durazzo, & Meyerhoff), although smoking history emerged as an important mediator of metabolite changes (Mason, et al., 2006). In a study of young adult social alcohol drinkers, a 13 ± 8% reduction in occipital GABA was observed after an acute alcohol infusion, which is consistent with facilitation of the GABA_A receptor by alcohol (Gomez, et al., 2011).

In stark contrast, only one study to date has used MRS to explore the neurochemical predictors of risk for substance abuse in adolescents. Moss and colleagues used ³¹P spectroscopic imaging to study a sample of peripubertal children (mean age of 12.5 years) at varying degrees of risk for developing a substance use disorder (Moss, Talagala, & Kirisci, 1997). Subjects in this study were split into three groups: 1) adolescents with a paternal history of substance use disorder and a personal diagnosis of disruptive behavior disorder, 2) adolescents with a paternal history of substance use disorder but no personal history of disruptive behavior disorder and, 3) adolescents with no paternal history and no disruptive disorders - a healthy comparison group. The first group, predicted to be at highest risk of developing substance use problems, was found to have significantly lower PDE in right parietal cortex when compared to the low-risk control subjects. Furthermore, PDE in this brain region correlated positively with the Information scale score of the Wechsler Intelligence Scale for Children, 3^rd edition (WISC-III) in only the highest risk group. It was speculated that the difference in PDE concentrations could reflect heightened synaptic pruning or decreased activity-dependent synaptogenesis in right parietal cortex, consequently influencing information retrieval abilities. The group possessing only a single risk factor of family history of substance abuse did not differ significantly from either of the other two groups for any metabolite measure and no effects of age were observed. A significant main effect of sex for PCr levels did emerge, with females exhibiting lower levels than males across brain regions and risk groups, a result that may reflect differences in white and grey matter proportions within the regions studied (Moss, et al., 1997). Females also had lower frontal β-NTP levels than males, but higher β-NTP concentrations overall, mirroring findings from a separate study by the same group in older subjects (Moss, et al., 1997).

While there have been thirty-two MRS studies published investigating alcohol-related neurochemical consequences in alcohol abusing and dependent adult populations (see also, Meyerhoff, et al. 2011), no studies have been published to date that have examined neurochemistry using MRS in adolescents either with significant alcohol use histories or alcohol use disorders. Notably, the metabolites detected as abnormal in adult patients with alcohol abuse disorders overlap with metabolites that demonstrate maturational changes during healthy adolescence (e.g., NAA, Cho, Cr, and GABA). Furthermore, metabolite alterations observed in alcohol abuse cohorts occur widely in the frontal lobe, which is in line with other neuroimaging results demonstrating that this late-maturing brain region is particularly susceptible to alcohol effects. Given the impact of alcohol use and abuse observed across multiple domains of cognitive functioning in adults, the paucity of MRS studies investigating neurochemical correlates of adolescent alcohol use further highlights the critical need for research in this area.

Conclusions and Future Directions

Structural, functional, and neurochemical maturation in the healthy adolescent brain occurs in regions and in cognitive domains that overlap with those observed to be most vulnerable to alcohol abuse and dependence. While a substantial body of neuroimaging research exploring the impact of alcohol on the structure and functional activation of the adolescent brain exists, less is known about neurochemical changes occurring in the human adolescent brain that are associated with healthy development, or that are associated with adolescent alcohol use. Recent advances have made MRS technology more amenable to the study of pediatric populations, offering promise for understanding typical brain neurochemical development and maladaptive developmental trajectories associated with the manifestation of alcohol disorders or other adolescent psychopathologies.

Developmental characterization of neurochemistry via MRS during adolescence could yield mechanistic hypotheses regarding the neurobiological consequences of alcohol use, as well as the propensity for future, continued and heavy alcohol use. For example, animal models suggest reduced GABA in prefrontal control regions in adolescents versus adults may reduce the sedative effects of alcohol, allowing for longer and heavier bouts of drinking (Silveri & Spear, 2004), which is consistent with developmental changes in GABA concentrations measured in frontal regions of human adolescents (Silveri, et al., 2013). In addition, lower prefrontal concentrations of NAA have been reported in children and adolescents compared to adults (Costa, et al., 2002; Horska, et al., 2002; Kadota, et al., 2001), but also in recently detoxified alcoholics relative to healthy controls (Meyerhoff, et al., 2011), suggesting that alcohol use during adolescence may present a double-threat to neuronal integrity and volume in prefrontal regions critical to self-regulation. Given that PCr concentrations have been reported to increase with age (Goldstein, et al., 2009), the developing adolescent brain may be more susceptible to alcohol effects on this important high-energy resource pool, leading to deleterious alterations in brain function. Moreover, MRS can be integrated with other imaging techniques and behavioral assessments to provide an additional level of analysis and construct a more comprehensive profile of brain development. This profile would provide a foundation for quantifying the effects of alcohol use on the adolescent brain and for the development of biomarkers that could help identify adolescents at risk for initiating and escalating alcohol consumption, even before they start drinking.

For the majority of the currently published developmental studies using MRS, the sparse distribution of sampling from different age groups makes results difficult to interpret. Larger samples and longitudinal studies are needed to map the time course of age-related metabolite changes associated with structural and functional brain development (Figures 3, 4), as well as with cognitive and behavioral changes during adolescence. Furthermore, the majority of existing pediatric MRS studies were conducted at a relatively low field strength (1.5 Tesla), which limits the number of proton- and phosphorous-containing metabolites that can be detected and reliably quantified. Significant future advances in understanding pediatric neurochemical development will depend on a number of factors: 1) increased uniformity in protocols across studies and sites (improving feasibility and reliability in multi-site studies); 2) the inclusion of large sample sizes to achieve statistical power to detect significant group differences; and 3) collection of longitudinal data sets. It also will be important to consider the influence of sex differences and brain laterality on metabolite levels. The measurement of additional nuclei (e.g., ¹³C, ¹⁹F) and the integration of metabolite data with data from other MR imaging modalities, such as brain tissue volumes (MRI), indices of white matter microstructure integrity (DTI) and neuronal activation (fMRI), will provide a more comprehensive understanding of neurochemical brain development.

Taken together, MRS studies of adolescent populations hold promise for advancing developmental neuroscience by identifying the biochemical signatures associated with healthy brain development, as well as neurobiological predictors and consequences of alcohol use and abuse. Given the paucity of extant developmental MRS data, a better characterization of healthy neurochemical development is needed to identify potential risk factors for the manifestation of alcohol and substance abuse disorders, which can also be applied to better understand neurobiological alterations associated with a number co-occurring psychopathological conditions that often emerge during adolescence. Thus, the continued evolution of MRS is expected to contribute to our understanding of pathophysiology, mechanisms of treatment response and ultimately to contribute to advances in treatment development in both low and high-risk adolescent populations.