1H-NMR, 1H-NMR T2-edited, and 2D-NMR in bipolar disorder metabolic profiling
Sethi et al. Int J Bipolar Disord
1 1 H-NMR, H-NMR T -edited, and 2D-NMR 2 in bipolar disorder metabolic profiling
Sumit Sethi 0 4
Mariana Pedrini 0 4
Lucas B. Rizzo 0 4
Maiara Zeni‑Graiff 0 4
Caroline Dal Mas 3
Ana Cláudia Cassinelli 2
Mariane N. Noto 0 2 4
Elson Asevedo 0 4
Quirino Cordeiro 2
João G. M. Pontes 1 6
Antonio J. M. Brasil 1 6
Acioly Lacerda 0 4
Mirian A. F. Hayashi 3
Ronei Poppi 5
Ljubica Tasic 1 6
Elisa Brietzke 0 4
0 Department of Psychiatry, Universidade Federal de São Paulo‐ UNIFESP , Rua Borges Lagoa, 570. Vila Clementino, São Paulo CEP 04038‐020 , Brazil
1 Laboratório de Química Biológica, Department of Organic Chemistry, Institute of Chemistry, Universidade Estadual de Campinas‐ UNICAMP , Caixa Postal 6154, Campinas, São Paulo CEP 13083‐970 , Brazil
2 Department of Psychiatry, Irmandade da Santa Casa de Misericórdia de São Paulo (ISCMSP) , Rua Major Maragliano, 287. Vila Mariana, São Paulo CEP 04017‐030 , Brazil
3 Depart‐ ment of Pharmacology, Universidade Federal de São Paulo‐UNIFESP, Rua Três de Maio, 100. Vila Clementino, São Paulo CEP 04044‐020 , Brazil
4 Department of Psychiatry, Universidade Federal de São Paulo‐UNIFESP , Rua Borges Lagoa, 570. Vila Clementino, São Paulo CEP 04038‐020 , Brazil
5 Department of Analytical Chemistry, Institute of Chemistry, Universidade Estadual de Campinas‐UNI‐ CAMP, Caixa Postal 6154, Campinas, São Paulo CEP 13083‐970 , Brazil
6 Laboratório de Química Biológica, Department of Organic Chemistry, Insti‐ tute of Chemistry, Universidade Estadual de Campinas‐UNICAMP , Caixa Postal 6154, Campinas, São Paulo CEP 13083‐970 , Brazil
Background: The objective of this study was to identify molecular alterations in the human blood serum related to bipolar disorder, using nuclear magnetic resonance (NMR) spectroscopy and chemometrics. Methods: Metabolomic profiling, employing 1H‑ NMR, 1H‑ NMR T2‑ edited, and 2D‑ NMR spectroscopy and chemometrics of human blood serum samples from patients with bipolar disorder (n = 26) compared with healthy volunteers (n = 50) was performed. Results: The investigated groups presented distinct metabolic profiles, in which the main differential metabolites found in the serum sample of bipolar disorder patients compared with those from controls were lipids, lipid metabolism‑ related molecules (choline, myo‑ inositol), and some amino acids (N‑ acetyl‑ l ‑ phenyl alanine, N‑ acetyl‑ l ‑ aspartyll ‑ glutamic acid, l ‑ glutamine). In addition, amygdalin, α‑ ketoglutaric acid, and lipoamide, among other compounds, were also present or were significantly altered in the serum of bipolar disorder patients. The data presented herein suggest that some of these metabolites differentially distributed between the groups studied may be directly related to the bipolar disorder pathophysiology. Conclusions: The strategy employed here showed significant potential for exploring pathophysiological features and molecular pathways involved in bipolar disorder. Thus, our findings may contribute to pave the way for future studies aiming at identifying important potential biomarkers for bipolar disorder diagnosis or progression follow‑ up.
1H‑ NMR; Biomarkers; Bipolar disorder; Metabolic profiling; Chemometrics; 2D NMR
Bipolar disorder (BD) is a potentially debilitating
mental disorder affecting 1–3% of the population worldwide
(Goodwin and Jamison 2007; Marohn 2011; Nierenberg
et al. 2013; Phillips and Kupfer 2013). Previous studies
have suggested a number of theories to explain the
etiology of BD, including the influence of genetic factors,
deficits in neurodevelopmental processes, implication of
neurodegenerative pathologies, changes in monoamine
neurotransmission, and abnormalities in neuroplasticity
(van Enkhuizen et al. 2015; Goes 2016; Passos et al. 2016).
A number of medications have been used for several
decades to treat this disorder and/or to ameliorate the
symptoms, though the mechanisms of action of most of mood
stabilizers still remain to be fully elucidated (Malhi et al.
2013; Yatham et al. 2013). Increased accuracy in early
diagnosis of BD is the key to improve the mental health
outcomes, to ameliorate the clinical course, and to make
better the treatment response of patients with this
disorder. However, the diagnosis of BD, especially in early
stages and in predominantly depressive presentations,
remains challenging (Brietzke et al. 2016). These
challenges could possibly be overcome by the identification of
differentially expressed biomarkers, which could reflect
© The Author(s) 2017. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License
(http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium,
provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license,
and indicate if changes were made.
or convey the pathophysiological processes (Oswald et al.
2007; Marmol 2008).
Hydrogen-1 nuclear magnetic resonance (1H-NMR)
spectroscopy-based metabolomics can be used to
monitor a wide range of metabolites in biological samples,
allowing for a sensitive, high-throughput molecular
screening (Beckonert et al. 2007). 1H-NMR has an
exceptional reproducibility and is quantitative to the extent
that a given peak area is directly proportional to the
concentration of the corresponding metabolite, turning
this technique a well-established approach for
metabolomics (Maher et al. 2008). Although many platforms for
metabolic profiling (for instance, the LC–MS, GC–MS,
NMR, capillary electrophoresis-MS) are available, none
technology alone provides the complete coverage of the
metabolome. Previous employment of these techniques
to study the brain of patients with psychiatric disorders
was reported (Lan et al. 2009; Strzelecki et al. 2015), but
it was scarcely applied for serum metabolome studies of
individuals with mental illnesses.
In this study, a NMR metabolomics profiling approach
was used to identify molecular alterations in the human
blood serum of BD patients. Our intention was to
conduct analysis to distinguish the individuals of type 1 BD
patients and healthy control groups, based on their
metabolic profiles. Also, as previously reported by us
(Sussulini et al. 2009), with time distance and new sampling,
NMR-based metabolomics has been applied as to verify
if it is possible to be used for precise classification of
Twenty-six euthymic outpatients with bipolar disorder
(BD) type 1 (BD group) and 50 healthy control (HC)
individuals (control group) were included in the research. All
individuals in the BD group, with age between 18 and
65 years old, fulfilled the DSM-IV (Diagnostic and
Statistical Manual of Mental Disorders) criteria according
to SCID-I (Structured Clinical Interview for DSM
Disorder). Euthymia was defined as not fulfill criteria for any
mood episode and present, at the same time, with scores
below 8 in Young Mania Rating Scale (YMRS) and
Hamilton Depression Rating Scale (HDRS)-17 items. Exclusion
criteria were as follows: be acutely suicidal, comorbidity
with substance use disorders, except nicotine, presence
of acute or unstable or chronic general medical
comorbidity, inability to read and understand study procedures,
pregnancy, and postpartum period. All subjects were
inquired on medical history, including lifetime use of any
medication. Body mass index (BMI) was also measured
using the formula BMI = Weight (kg)/Height (m)2.
For comparison, a control group of 50 healthy
volunteers was recruited from the community. To be eligible
for the HC group, individuals were required to have age
between 18 and 65 years old, negative history of any
psychiatric condition current or along lifetime, and never
made use of psychiatric medication. In addition, only
volunteers with no family history of major mental disorders
(schizophrenia-spectrum disorders, mood disorders, and
suicide) were included. The same exclusion criteria of BD
group were applied to HC group.
Clinical assessment included SCID-I for confirmation
of diagnosis and assessment of psychiatric
comorbidities. YMRS and HDRS-17 items were used to determine
severity of manic and depressive symptoms, respectively.
Functioning was estimated by General Assessment of
Serum collection and storage
All blood samples were taken in the morning (between 8
and 10 a.m.) after 12 h of fasting. Blood was drawn into
Vacutainer tubes, and immediately allowed to clot for
at least 30 min at room temperature, before it was
centrifuged at 1500×g, for 5 min. The serum was then
aliquoted, transferred into clean polypropylene tubes, and
stored at −80 °C until use. The maximum period of
storage was of two weeks.
NMR spectroscopy analyses: 1H‑NMR, 1H‑NMR T2‑edited,
and 2D NMR
For NMR spectroscopy analyses, serum samples were
thawed and centrifuged at 12,300×g, for 10 min, at 4 °C
to separate any precipitate. Aliquots of 250 μL of the
supernatants were diluted with 250 μL of deuterated
water (D2O) or PBS buffer (250 μL containing 10% D2O)
and transferred into 5.0 mm diameter NMR tubes. NMR
tubes were placed into a Bruker 600 NMR
spectrometer (Bruker Advance III, Bruker GmBH,
Rheinestennen, Germany) using TBI (Triple resonance Broadband
Inverse) probe at 25 °C. 1H-NMR spectra were recorded
as three independent measurements for each sample
using the CH3-lactate signal (δ1.33, 3H, d, 3J = 7.0 Hz)
as a reference, and applying the pulse sequence
WATERGATE (p3919gp) with 128 ns (Piotto et al. 1992).
T2-edited NMR spectra were recorded using the CPMG
(Carr-Purcell-Meiboom-Gill) sequence (Meiboom and
Gill 1958), where a fixed total spin–spin relaxation delay
2nτ of 100 ms was used to attenuate the broad NMR
signals from slowly tumbling molecules such as proteins
and lipids retaining those from low molecular weight
compounds and some lipid components. For each
spectrum, 128 transients were acquired into 32,000 data
points, with a spectral width of 12 kHz. To confirm the
assignments made from 1H-NMR and T2-edited spectra,
some blood serum samples were also examined using 2D
[1H–13C] HSQC. For each 2D spectrum, 256 increments
with 64 transients per increment were collected and
extended to 4 K data points. The signal assignments were
based on the literature and/or databases such as
Biological Magnetic Data Bank (BMRB) (Ulrich et al. 2008) and
Human Metabolome Database (HMDB) (Wishart et al.
2007) and are indicated on the T2-edited spectrum and
confirmed by the 2D fully assigned [1H–13C] HSQC data.
Chemometrics of 1H NMR spectral data
1H-NMR data were transported to a data matrix, and
chemometrics analysis, based on principal component
analysis (PCA) and partial least-squares discriminant
analysis (PLS-DA), were performed using MATLAB (The
MathWorks, Natick, MA) or Pirouette (Infometrix, USA)
software. T2-edited spectra were not used in
chemometric analysis. The spectral region of chemical shifts, from
1.00 to 4.40 ppm, was used in chemometrics (Nørgaard
2009). PLS-DA was used as a supervision method for
The summary of the collected sample characteristics is
presented in Table 1. A predominance of females over
males was observed in our BD sample, and the
percentages were statistically different between the BD and HC
groups. In addition, we observed, as expected, a high
prevalence of tobacco smoking individuals in our BD
group compared with HC group. Functioning was also
impaired in the BD group compared with HC group,
which reflects a well-known incomplete association
between mood symptoms and functionality in
individuals with BD. In addition, there was a clear predominance
of chronic patients, with a mean duration of illness of
about 15 years.
1H NMR spectroscopy data analysis
To explore all the potential differences in the metabolic
profiles between the BD and HC groups, the 1H-NMR
spectra (Fig. 1a) were subjected to PCA, i.e., the
spectral data obtained for 76 blood serum samples that were
acquired as 228 spectra were used in PCA. T2-edited
spectra were not used in chemometrics analysis because
Table 1 Clinical and demographic characteristics of the sample
Healthy controls (n = 50)
BD (n = 26)
Fig. 1 a 1H NMR Spectra with water suppression (Pulse sequence: p3919gp, ns = 128) blood serum sample from HC (black), and BD group (gray).
Spectral region highlighted δ 4.40 to 1.00 with higher loadings. b 1H NMR Spectra recorded with T2‑filter (Pulse sequence: cpmgpr1d, ns = 128)
blood serum sample from HC (black), and BD group (gray). Assignments: Lactate (δ 1.33 and 4.11), glucose (δ 3.41, 3.46, 3.77 and 3.82), leucine (Leu;
δ 1.71 and 3.73), proline (Pro; δ 3.34), glycerylphosphocholine (GPCho; δ 3.24), myo‑inositol (δ 3.25), creatine (δ 3.03), glutamine (Gln; δ 2.44), gluta‑
mate (Glu; δ 2.14), lipids (δ 1.29 and 2.04), lysine (Lys; δ 1.68), alanine (Ala; δ 1.47), valine (Val; δ 1.04)
these spectra show altered peak areas in relation to the
peaks of 1H-NMR spectra. Based on the PCA results, it
was possible to observe a distinction of the samples into
two groups: BD and HC using the spectral region δ 1.00
to 4.40 (Fig. 2).
Considering the PCA results and the previous
observations about the spectral regions not relevant for analyses,
PLS-DA was performed using the same spectral range
employed in the PCA. PLS-DA of 1H NMR spectroscopy
data revealed an excellent separation between BD and
Through the assignments from 1H-NMR edited with
T2 filter (Fig. 1b) and the correlations from the HSQC
contour maps (Fig. 3), it was possible to compare the
obtained data with databases. Thereby, based mainly on
spectral data (chemical shift, coupling constants, and
multiplicity) and in accordance with biochemical
knowledge, seven key-metabolites (Fig. 4) were found to be
crucial for the two-group separation. Furthermore, it
was observed that there are significant differences in
some metabolite concentrations when the two
investigated groups were compared, in which important
significant variations in some metabolites were observed,
and compounds were then identified through T2-edited
spectra. The blood serum samples from the BD group
had shown to be richer in lipids, whose amounts were up
Fig. 2 Metabolomics discriminate bipolar disorder patients from healthy individuals. a Plots of cross‑ validated PCA scores, 50 samples (25 BD + 25
HC) is equivalent to 150 spectra (75 BD + 75 HC). b 2D samples scores plot of PLS‑DA, 76 samples (26 BD + 50 HC) is equivalent to 228 (78
BD + 150 HC). c Loadings Graph of spectral region δ 1.00 to 4.40 used in chemometrics. d 3D samples scores plot of PLS‑DA
to 1/3 higher compared with those from the HC group.
Also, metabolites such as d-glucose, l-alanine, and lac
tate were more pronounced in BD group compared with
HC group. All these metabolites were found to be more
abundant in BD group than in HC group, as their peaks
areas were at most presenting a ratio of about 3:2
(Additional file 1: Table S3).
1H-NMR-based metabolomics were used to
chemometric analysis. PCA and PLS-DA analyses were employed
to examine the metabolic profiles of blood serum of
individuals with BD compared with HC volunteers. The
loading plot (Fig. 2) highlighted the most significant variables
with the highest loading values, which corresponded to
the spectral region from δ 1.00 to 4.40. The 1H-NMR
spectra of BD blood serum samples are characterized by
overlaps that resulted in broad and intense peaks. So, we
acquired T2-edited 1H-NMR spectra (Fig. 1b) to achieve a
better spectral resolution and assign peaks referred to the
metabolites with the low molecular weight (<1500 Da),
such as: l-valine, l-alanine, and creatine. The peaks
assignments were performed in comparison with
chemical shifts values that were published previously (Misra
and Bajpai 2009). It is important to mention that the
peaks in T2-edited spectra are not appropriate for
integration. Through comparative analysis of the 2D NMR
spectra, we observed important differences in metabolic
profiles among serum samples from the two, BD and HC,
groups. These differences are indicated in Fig. 3, in which
it is possible to observe that some metabolites such as
lipoamide, l-glutamine, among others, were detected
only in BD group, while N-acetyl-l-alanine was observed
only in HC group (Fig. 4).
The metabolites that had their peaks assigned based
on the 2D NMR and that were classified as “healthy”
means that these metabolites have a higher concentration
in blood serum samples from the HC (healthy control)
Fig. 3 Two‑ dimensional NMR Spectra: Contour maps of HSQC (pulse sequence: hsqcedetgpsp.3) blood serum sample from (a) HC (ns = 64,
SD = 16), and b BD group (ns = 64, SD = 16)
group compared with the BD group individuals
(Additional file 1: Table S1). The same reasoning can be applied
to the BD group. However, it is important to remark that
we have not quantified the identified biomarkers up to
the present moment.
Despite the large overlapping between the peaks in
the 1H NMR spectra, it was possible to estimate the
concentration ratios for some metabolites, such as
d -glucose, l -alanine, lactate, and lipids, through
comparison of the mean value rates of peaks integration of
Fig. 4 Chemical structures of seven key‑metabolites (1 from HC group + 6 from BD group) identified by 2D NMR spectroscopy
1H-NMR spectra (Additional file 1: Table S3). For com
parison of HC and BD profiles, the greatest variation
in concentration of metabolites was observed for the
Partially, our results are in accordance with numerous
previous studies that have demonstrated changes in brain
metabolites in BD by in vivo 1H NMR. Davanzo et al.
(2001) have shown an increase in myo-inositol in anterior
cingulate cortex in children with BD. Acute lithium(I)
treatment was associated to a significant reduction in the
myo-inositol/creatine ratio, especially among
responders to this medication (Davanzo et al. 2001). This result
was also reproduced in adults (Machado-Vieira et al.
2015). Several other studies produced mixed results
with choline, GABA, and glutamate, probably influenced
by methodological issues, chosen brain areas or
differences in subpopulations of individuals with the disorder
The increase of risk of hyperlipidemia in BD patients
was reported by Hsu et al. (2015), which corroborates
with our results, in which increases in the lipids
levels in BD were also noticed (Fig. 1b). Furthermore,
currently Yoshimi et al. (2016) studied the metabolism of BD
patients through analysis of blood serum samples using
capillary electrophoresis-time-of-flight mass
spectrometry (CE-TOF/MS). They reported changes in amino acids
metabolism (valine, alanine, glutamine) and metabolites
belonging to the citric acid and urea cycles, such as the
α-ketoglutarate and N-acetylglutamic acid whose
levels were increased in BD patients. In our analyses, the
presence of α-ketoglutaric acid, α-ketoisovaleric acid
and N-acetyl-l-aspartyl-l-glutamic acid (NAAG) was
observed, as pointed in Fig. 4.
Glutamate and glutamine are well-known markers
(Manji et al. 2003). Also, Pålsson et al. (2015) reported
the increase in the levels of these metabolites in blood
serum and cerebrospinal fluid (CSF) of BD patients using
HPLC with fluorescence detection. This increase can be
due to mitochondrial dysfunction or alterations in the
metabolism of the cell (Hsu et al. 2015).
However, to our knowledge, the 1H-NMR
spectroscopy-based metabolomics analysis of serum was less
studied. Sussulini et al. (2009) investigated differential
metabolites in human serum sample of patients with
BD (n = 25) under different drug treatments: lithium(I)
(n = 15) versus other medications (n = 10). This
strategy showed significant potential for exploring
pathophysiological and toxicological features involved in BD.
The investigated groups (HC and patients with BD under
different treatments) could be distinguished
according to their metabolic profiles, and the main differential
metabolites found were glycoprotein lipids, mono- and
polyunsaturated lipids, acetate, choline, glutamate, and
The results of this research should be interpreted at
light of some limitations. First, the small sample size
precludes subgroup analysis including potential
medication effects in metabolic profile. In addition, the
inclusion of only euthymic individuals makes impossible to
know which metabolic changes are related to traits and
which are related to mood states in BD. The different
proportion of men and women between the BD and HC
groups is a result of a female predominance in the
mental health service where the study was conducted. On the
other hand, no significant differences in metabolic profile
between female and male were noticed here. Strengths
of the current study included the very careful selection
of cases and controls, including the inclusion of patients
without any significant comorbidities and HCs without
any evidence of personal or family mental illness history.
Considering the main results of this study, we can
conclude that 1H NMR spectroscopy-based metabolomics
analysis of serum could be a useful strategy to investigate
the pathophysiology of BD, as well as to identify and
validate potential biomarkers for diagnosis and/or for
In this study, it was possible to the identify seven
key-metabolites, which showed to be in accordance
with some other works described in the literature that
used different analytical platforms for monitoring BD.
Although the greatest differences between the two
studied groups were observed in lipid ratios (more abundant
in the BD group), due to the overlapping, some
integration of the peak areas can not be taken as sufficiently
precise for quantification of some metabolites. Nevertheless,
it is important to mention that the spectral region used
was similar to the work of Sussulini et al. (2009)
indicating a greater accuracy in this spectral range when looking
for the biomarkers and also showing the high
reproducibility of the NMR spectroscopy technique in
metabolomic profiling of BD individuals.
Additional file 1. HSQC and Description of 1H‑NMR spectra for individu‑
als with bipolar disorder and for healthy controls, HSQC contour map
correlations and identification of possible key metabolites.
SS participated in data collection, data analysis, and writing of manuscript. MP,
ACC, EA, QC, and MNN participated in recruitment and assessment of patients
and controls, data analysis, and writing. LR, MZG, and CDM participated in
data collection, analysis, and writing. GMP and AJMB participated in NMR data
collection, analysis, and writing. AL participated in conception, fund raising,
and writing. MAFH, RP, LT, and EB participated in all the steps of conception,
execution, and diffusion of results. All authors read and approved the final
The authors declare that they have no competing interests.
Ethics approval and consent to participate
The Local Ethics Committee approved this study (Numbers
35376414.8.0000.5505), and all subjects provided written informed consents
before their inclusion in the study. The subjects consented in publication of
data preserving anonymity.
We thank the Conselho Nacional de Desenvolvimento Científico e Tecnológico
(CNPq, Brasília, Brazil) for financial support and fellowships. SS received a
Young Talent Scholarship from the CNPq. We also acknowledge the scientific
grant from FAPESP (Process Number 2014/18938‑8).
Springer Nature remains neutral with regard to jurisdictional claims in pub‑
lished maps and institutional affiliations.
Beckonert O , Keun HC , Ebbels TM , Bundy J , Holmes E , Lindon JC , Nicholson JK . Metabolic profiling, metabolomic and metabonomic procedures for NMR spectroscopy of urine, plasma, serum and tissue extracts . Nat Protoc . 2007 ; 2 : 2692 - 703 .
Brietzke E , Rosa AR , Pedrini M , Noto MN , Kapczinski F , Scott J. Challenges and developments in research of the early stages of bipolar disorder . Rev Bras Psiquiatr . 2016 ; 38 : 329 - 37 .
Bustillo JR . Use of proton magnetic resonance spectroscopy in the treatment of psychiatric disorders: a critical update . Dialogues Clin Neurosci . 2013 ; 15 : 329 - 37 .
Davanzo P , Thomas MA , Yue K , Oshiro T , Belin T , Strober M , McCracken J. Decreased anterior cingulate myo‑inositol/creatine spectroscopy resonance with lithium treatment in children with bipolar disorder . Neuropsychopharmacology . 2001 ; 24 : 359 - 69 .
Goes FS . Genetics of bipolar disorder: recent update and future directions . Psychiatr Clin North Am . 2016 ; 39 : 139 - 55 .
Goodwin FK , Jamison KR . Manic‑ depressive illness: bipolar disorders and recurrent depression . 2nd ed . New York : Oxford University Press ; 2007 .
Hsu JH , Chien IC , Lin CH . Increased risk of hyperlipidemia in patients with bipolar disorder: a population‑based study . Gen Hosp Psychiatry . 2015 ; 37 : 294 - 8 .
Lan MJ , McLoughlin GA , Griffin JL , Tsang TM , Huang JT , Yuan P , Manji H , Holmes E , Bahn S. Metabolomic analysis identifies molecular changes associated with the pathophysiology and drug treatment of bipolar disorder . Mol Psychiatry . 2009 ; 14 : 269 - 79 .
Machado‑ Vieira R , Gattaz WF , Zanetti MV , De Sousa RT , Carvalho AF , Soeiro‑ de‑Souza MG , Leite CC , Otaduy MC . A longitudinal (6‑ week) 3T 1H‑MRS study on the effects of lithium treatment on anterior cingulate cortex metabolites in bipolar depression . Eur Neuropsychopharmacol . 2015 ; 25 : 2311 - 7 .
Maher AD , Crockford D , Toft H , Malmodin D , Faber JH , McCarthy MI , Barrett A , Allen M , Walker M , Holmes E , Lindon JC , Nicholson JK . Optimization of human plasma 1H NMR spectroscopic data processing for high‑throughput metabolic phenotyping studies and detection of insulin resistance related to type 2 diabetes . Anal Chem . 2008 ; 80 : 7354 - 62 .
Malhi GS , Tanious M , Das P , Coulston CM , Berk M. Potential mechanisms of action of lithium in bipolar disorder . Current understanding. CNS Drugs . 2013 ; 27 : 135 - 53 .
Manji HK , Quiroz JA , Payne JL , Singh J , Lopes BP , Viegas JS , Zarate CA . The underlying neurobiology of bipolar disorder . World Psychiatry . 2003 ; 2 : 136 - 46 .
Marmol F. Lithium : bipolar disorder and neurodegenerative diseases Possible cellular mechanisms of the therapeutic effects of lithium . Prog Neuropsychopharmacol Biol Psychiatry . 2008 ; 32 : 1761 - 71 .
Marohn S. Natural medicine guide to bipolar disorder . Charlottesville: Hampton Roads Publishing Company ; 2011 .
Meiboom S , Gill D. Modified spin‑ echo method for measuring nuclear relaxation times . Rev Sci Instrum . 1958 ; 29 : 688 - 91 .
Misra D , Bajpai U. Metabolite characterization in serum samples from normal healthy human subjects by 1H and 13C NMR spectroscopy . Bull Chem Soc Ethiop . 2009 ; 23 : 211 - 21 .
Nierenberg AA , Kansky C , Brennan BP , Shelton RC , Perlis R , Iosifescu DV . Mitochondrial modulators for bipolar disorder: a pathophysiologically informed paradigm for new drug development . Aust N Z J Psychiatry . 2013 ; 47 : 26 - 42 .
Nørgaard L. The iToolbox for MATLAB; KVL: Frederiksberg . 2009 . http://www. models.life.ku.dk.
Oswald P , Souery D , Kasper S , Lecrubier Y , Montgomery S , Wyckaert S , Zohar J , Mendlewicz J. Current issues in bipolar disorder: a critical review . Eur Neuropsychopharmacol . 2007 ; 17 : 687 - 95 .
Pålsson E , Jakobsson J , Södersten K , Fujita Y , Sellgren C , Ekman CJ , Ågren H , Hashimoto K , Landén M. Markers of glutamate signaling in cerebrospinal fluid and serum from patients with bipolar disorder and healthy controls . Eur Neuropsychopharmacol . 2015 ; 25 : 133 - 40 .
Passos IC , Mwangi B , Vieta E , Berk M , Kapczinski F. Areas of controversy in neuroprogression in bipolar disorder . Acta Psychiatr Scand . 2016 ; 134 : 91 - 103 .
Phillips ML , Kupfer DJ . Bipolar disorder diagnosis: challenges and future directions . Lancet . 2013 ; 381 : 1663 - 71 .
Piotto M , Saudek V , Sklenár V. Gradient ‑tailored excitation for single ‑ quantum NMR spectroscopy of aqueous solutions . J Biomol NMR . 1992 ; 2 : 661 - 5 .
Sena MM , Chaudhry ZF , Collins CH , Poppi RJ . Direct determination of diclofenac in pharmaceutical formulations containing B vitamins by using UV spectrophotometry and partial least squares regression . J Pharm Biomed Anal . 2004 ; 36 : 743 - 9 .
Strzelecki D , Grzelak P , Podgórski M , Kałużyńska O , Stefańczyk L , Kotlicka‑Antczak M , Gmitrowicz A. Comparison of metabolite concentrations in the left dorsolateral prefrontal cortex, the left frontal white matter, and the left hippocampus in patients in stable schizophrenia treated with antipsychotics with or without antidepressants. 1H‑NMR spectroscopy study . Int J Mol Sci . 2015 ; 16 : 24387 - 402 .
Sussulini A , Prando A , Maretto DA , Poppi RJ , Tasic L , Banzato CE , Arruda MA. Metabolic profiling of human blood serum from treated patients with bipolar disorder employing 1H NMR spectroscopy and chemometrics . Anal Chem . 2009 ; 81 : 9755 - 63 .
Ulrich EL , Akutsu H , Doreleijers JF , Harano Y , Ioannidis YE , Lin J , Livny M , Mading S , Maziuk D , Miller Z , Nakatani E , Schulte CF , Tolmie DE , Wenger RK , Yao H , Markley JL. BioMagResBank . Nucleic Acids Res . 2008 ; 36 : D402 - 8 .
van Enkhuizen J , Janowsky DS , Olivier B , Minassian A , Perry W , Young JW , Geyer MA . The catecholaminergic‑ cholinergic balance hypothesis of bipolar disorder revisited . Eur J Pharmacol . 2015 ; 753 : 114 - 26 .
Wishart DS , Tzur D , Knox C , Eisner R , Guo AC , Young N , Cheng D , Jewell K , Arndt D , Sawhney S , Fung C , Nikolai L , Lewis M , Coutouly MA , Forsythe I , Tang P , Shrivastava S , Jeroncic K , Stothard P , Amegbey G , Block D , Hau DD , Wagner J , Miniaci J , Clements M , Gebremedhin M , Guo N , Zhang Y , Duggan GE , MacInnis GD , Weljie AM , Dowlatabadi R , Bamforth F , Clive D , Greiner R , Li L , Marrie T , Sykes BD , Vogel HJ , Querengesser L. HMDB: the human metabolome database . Nucleic Acids Res . 2007 ; 35 : D521 - 6 .
Yatham LN , Kennedy SH , Parikh SV , Schaffer A , Beaulieu S , Alda M , O'Donovan C , Macqueen G , McIntyre RS , Sharma V , Ravindran A , Young LT , Milev R , Bond DJ , Frey BN , Goldstein BI . The evolution of CANMAT Bipolar Disorder Guidelines: past, present, and future . Bipolar Disord . 2013 ; 15 : 58 - 60 .
Yoshimi N , Futamura T , Kakumoto K , Salehi AM , Sellgren CM , Holmén‑Larsson J , Jakobssong J , Pålssong E , Landén M , Hashimoto K. Blood metabolomics analysis identifies abnormalities in the citric acid cycle, urea cycle, and amino acid metabolism in bipolar disorder . BBA Clin . 2016 ; 5 : 151 - 8 .