Mecp2 deletion alters the brain profile of choline-containing phospholipids and cardiolipin

Our previous in vivo ¹H MRS study revealed altered relative levels of total choline (tCho) in Mecp2-deficient mice vs. controls [20], where tCho represents several metabolites predominantly involved in choline phospholipid metabolism (choline, Cho; phosphocholine, PC; glycerophosphocholine, GPC), plus substantial contributions from phosphoethanolamine and taurine [25]. The phosphodiester, glycerophosphocholine, is a phospholipid (PL) degradation product while choline and the phosphomonoester, phosphocholine, are both phospholipid anabolites and catabolites. Since individual choline compounds and phospholipid classes cannot be quantitated by in vivo MRS, we determined their (absolute) tissue levels by ¹H MRS of whole-brain extracts (Fig. 1). Each of the three choline species mentioned above showed a trend towards decreased levels in Mecp2-deficient mice vs. controls (Table 1); the difference was close to statistical significance for choline (ca. −40%; p = 0.077), but not for the phosphorylated choline species. For the sum of these three choline compounds, a slight decrease was observed (not significant), in agreement with our previous in vivo results obtained at a short echo time [20]. Interestingly, the PC/GPC ratio was clearly increased by 30% (p = 0.034), mostly due to a GPC decrease (Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Metabolite levels in Mecp2-/y and control mouse brain (µmol/g wet weight, and molar ratios for selected metabolites)

https://doi.org/10.1371/journal.pone.0000157.t001

³¹P MRS was able to distinguish 16 phospholipid classes and subclasses (Fig. 2), fourteen of which were successfully assigned. The analysis of PL profiles by ³¹P MRS showed significant differences in PL levels between Mecp2-deficient mice and controls solely for choline-containing PL (diacyl-phosphatidylcholine, PtdC; alkyl-acyl-phosphatidylcholine, AAPtdC; choline plasmalogen, PtdC_plasm; lyso- phosphatidylcholine, lyso-PtdC) (Table 2; see also Table S1 for cardiolipin, CL).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1.

Typical high-resolution ¹H NMR spectrum of the brain extract from a Mecp2-deficient mouse (water-soluble metabolites). The entire spectral region of interest (upper left spectrum) as well as magnified subregions are shown. Creatinine (Crn) was not quantified since the weak singlet was not sufficiently separated from other peaks forming a broad hump between 3.01 and 3.06 ppm. Abbreviations: ac, acetate; ala, alanine; asp, aspartate; Cho, choline; Cr, creatine; GABA, γ-aminobutyrate; gln, glutamine; glu, glutamate; gly, glycine; GPC, glycerophosphocholine; lac, lactate; myo-Ins, myo-inositol; NAA, N-acetylaspartate; NANA, N-acetylneuraminate; PC, phosphocholine; scyllo-Ins, scyllo-inositol; suc, succinate; tau, taurine.

https://doi.org/10.1371/journal.pone.0000157.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2.

Typical high-resolution ³¹P NMR spectrum of the brain extract from a Mecp2-deficient mouse (phospholipids). The entire PL region (upper left spectrum) as well as magnified subregions are shown. Processing parameters were varied to highlight important spectral details. Abbreviations: AAPtdC, alkyl-acyl-phosphatidylcholine; AAPtdE, alkyl-acyl-phosphatidylethanolamine; CL, cardiolipin; PLx,y, unassigned phospholipids; PtdI, phosphatidylinositol; PtdA, phosphatidic acid; PtdC, (diacyl)phosphatidylcholine; PtdC_plasm, choline plasmalogen; PtdE, (diacyl)phosphatidylethanolamine; PtdE_plasm, ethanolamine plasmalogen; PtdG, phosphatidylglycerol; PtdS, phosphatidylserine; SM, sphingomyelin.

https://doi.org/10.1371/journal.pone.0000157.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Metabolite levels in Mecp2-/y and control mouse brain (µmol/g wet weight, and molar ratios for selected metabolites)

https://doi.org/10.1371/journal.pone.0000157.t002

All other phospholipid classes and subclasses determined were present at very similar levels in the brains of both animal groups (diacyl-phosphatidylethanolamine, PtdE; alkyl-acyl-phosphatidylethanolamine, AAPtdE; ethanolamine plasmalogen, PtdE_plasm; sphingomyelin, SM; lyso-sphingomyelin, lyso-SM; phosphatidylserine, PtdS; phosphatidylinositol, PI; phosphatidic acid, PtdA; phosphatidylglycerol, PtdG). Choline plasmalogen (PtdC_plasm) was decreased in Mecp2-deficient mouse brain (Table 2); however, PtdC_plasm values may be less accurate than the other PL values since the PtdC_plasm peak areas measured may be influenced by the presence of the large PtdC signal directly upfield from PtdC_plasm. PtdS and PtdE showed a trend towards increased values (not significant). PL_tot (total phospholipids) levels did not vary between groups, nor did the molar ratio between choline and ethanolamine-containing phospholipids, PC-L/PE-L (Table 2). As a matter of fact, molar fractions of phospholipids are a direct measure of the balance of phospholipid composition that affects the fluidity and function of membranes (Farooqui and Horrocks 2001).

Cerebral profiles of neurotransmitters in Mecp2-null mice

Choline, introduced above as a phospholipid metabolite, is also involved in neurotransmitter metabolism as it is an anabolite and catabolite of acetylcholine. Other neurotransmitters and their metabolites are closely linked to glucose (glc) metabolism. Glutamate (glu), generally acknowledged to be the most important transmitter for normal brain function [26], can be produced from the citric acid cycle following pyruvate formation from glucose. We did not detect a significant difference in glutamate between Mecp2-deficient and control mice based on ¹H spectra of brain extracts. However, the presynaptic precursor of glutamate, glutamine (gln), was significantly increased in Mecp2-deficient mice by almost 40%. The glutamine/glutamate ratio also was significantly increased in Mecp2-deficient mice (Table 1). Aspartate (asp) is a neurotransmitter and a glutamate metabolite, and, like glutamate, does not differ between Mecp2-deficient and control mice. Similarly, the inhibitory neurotransmitter, γ-aminobutyrate (GABA), is derived from glutamate and only shows an insignificant decrease in Mecp2-deficient vs. control mice. Glycine (gly), another neurotransmitter [27], shows almost identical concentrations in both groups.

Mecp2 deletion alters cerebral osmolyte profiles

NAA is considered a neuronal marker and has also been proposed to act as an osmolyte [28]. NAA is involved in asp metabolism and lipid synthesis, and was found in large quantities in mouse brain (ca. 4 µmol/g). However, no differences were detected between the two groups (Table 1). Besides GPC and glutamine mentioned above, we quantitated two additional compounds known as osmolytes, taurine (tau) and myo-Ins. This polyol that is also important in phospholipid (phosphatidylinositol) metabolism, was decreased by one third in Mecp2-deficient mice (p = 0.034 for both significance tests, Table 1). In contrast, there was no difference in taurine levels between the two groups studied. Neither the NAA and ketone body-derived metabolite, acetate (ac), nor the glycolysis products, lactate (lac) and alanine (ala), nor citric acid cycle-derived succinate (suc) were different for the two groups analyzed. In a previous paper, we raised the possibility that the NAA signal measured by in vivo MRS may overlap with an unresolved signal from N-acetylaspartylglutamate (NAAG) [20]. The latter should be readily detectable as a separate resonance in brain extract spectra. Nevertheless, we did not find any peak next to NAA that would be intense enough to explain the additional signal measured in vivo (max. 5–6% of NAA). Therefore, the extra peak observed in vivo is, in all likelihood, not due to NAAG. Another N-acetyl compound, N-acetylneuraminate (NANA) or sialic acid, is moderately decreased in Mecp2-/y mice (near significance, p = 0.077, Table 1). NANA is synthesized from glucose and is involved in the synthesis of gangliosides that are implicated in signal transduction in eukaryotic cells [29]. Knockout mice lacking the major gangliosides of brain are highly susceptible to induced seizures [30]; however, little is known about a specific role of NANA in Rett syndrome.

How are membrane phospholipids and brain (cell) growth linked in Mecp2-/y mice?

To our knowledge, the present study contains the first detailed quantitative analysis of membrane phospholipid classes and subclasses (phospholipidomics) in brain tissue with a mutated Mecp2 or MECP2 gene. PL metabolism plays an essential role in cell growth since PL form the matrix of cell membranes, and membrane synthesis is indispensible for cell growth. In fact, brain growth defects resulting in decreased head size and reduced neuronal cell size are one of the hallmarks of RS [7]. Global reduction in brain size and several of the morphological anomalies detected in RS patients have also been observed in the Mecp2-deficient mice used in this study [20]. Moreover, knockdown of Mecp2 levels in dissociated embryonic non-neuronal cells from mice has been shown to reduce their proliferation [19]. These effects point to growth perturbation in Mecp2-deficient brain tissue, and should therefore be reflected by parameters characterizing the biosynthesis and degradation of cellular membranes. Nonetheless, studies of cerebral membrane lipids in RS have been scarce, and have not included quantitation of individual phospholipid classes [31]–[33].

The individual steps of PtdC synthesis and degradation are well known (Kennedy pathway, Figure S1), and the role of PtdC turnover in membrane production and cell growth is well established. Therefore, the PtdC metabolite pattern observed in Mecp2-deficient mouse brain can be interpreted in terms of membrane PtdC turnover and cell growth, as outlined in the legend to Figure S1. The metabolic fingerprint of a reduced PtdC turnover rate indicates that the brain cells of Mecp2-deficient mice possess a reduced potential to boost PtdC production when cells are in a growing state [34], [35]. This may limit the ability of cells to grow, a mechanism potentially underlying brain growth arrest in Mecp2-deficient mice (see scheme below).

Autopsy of both RS patients and Mecp2 mutant mice has revealed that the brain size is reduced without reduction of the number of neurons [36]. However, the size and dendritic arborization of neurons are reduced. On the other hand, there are up to ten times less neurons than glial cells (mainly astrocytes) in the brain. Recently published evidence suggests that, in fact, suppression of astrocyte growth may contribute to microcephaly in RS [19]. Thus, the choline PL pattern suggesting reduced growth potential in Mecp2-deficient mouse brain may represent in part non-neuronal cells. Indeed, reduced levels of the astrocyte marker, myo-inositol, in the brains of Mecp2-/y mice would be consistent with the assumption of a reduced number of astrocytes (see discussion of osmolytes below). While the interpretation of our PtdC data in terms of reduced cell growth is straight forward and rather compelling, the extent to which astrocytes and neurons are responsible for growth reduction awaits further clarification (Table S2). Experiments on co-cultured Mecp2-/y astroglial cells and neurons could be performed to study the relationship between metabolism and growth in each cell type [37]. Immunohistochemistry of brain tissue sections from Mecp2-/y mice may be employed to test the hypothesis of a more pronounced growth impairment in astrocytes vs. neurons, although correct quantitation of astrocytes may be rather difficult for reasons outlined elsewhere [38].

Potential role of PAF-remodeling in neurotransmission of Mecp2-deficient mice

PL metabolism is not only essential to cell growth. PL remodeling plays an important role in the generation of lipid messengers and in intracellular signaling. The characteristic differences in choline PL containing diacyl and acyl-alk(en)yl moieties between Mecp2-deficient mice and controls point to modifications of the platelet activating factor (PAF) remodeling pathway or “PAF cycle” (Figure S2) [39]. PAF is an acyl-alkyl PL that modulates a range of cellular processes. PAF effects relevant to the CNS may include enhancement of glutamate exocytosis, neuronal migration, memory formation, and collapse of growth cones essential to the formation of synapses (Table S2) [39]. While our results support the hypothesis of an altered PAF cycle in the Mecp2-null mouse brain, we have no direct evidence of changed PAF concentrations since PAF levels are below the detection threshold of our MRS experiments. Highly sensitive techniques such as mass spectrometry should be employed to reveal whether the levels of PAF (and a number of PAF metabolites) are indeed modified as a result of the perturbations detected in this report. This would be of particular interest considering that PAF is known to activate excitatory synapses by increasing presynaptic glutamate release. Future studies have to determine if PAF plays a role in RS.

Mecp2 deletion causes characteristic changes in glutamate metabolism

Altered glutamine/glutamate ratios found in the Mecp2-null mouse brain may be explained based on previously published evidence of heightened excitatory activity in girls with RS [40]. The latter finding indicates that glutamate production and release from neurons are enhanced, while normal postsynaptic signal transduction may be disrupted due to overexcitation by increased glutamate release. Perturbed signaling and/or dysregulation mediated by astrocyte-associated glutamate receptors [41] may trigger increased glutamate uptake and glutamate-glutamine conversion in astrocytes, increasing glutamate-glutamine cycling (Figure S3). This would be consistent with (i) data showing an increase in metabolic activity in glia upon increased glutamate release from neurons [42], and (ii) previously reported reduction of brain ATP levels in Mecp2-null mice [20] since ATP is also consumed in the synthesis of glutamine from glutamate [43]. Glutamate levels may be kept constant by compensatory glutamate synthesis from glucose and by anaplerosis (Table S2) [44], [45]. In fact, the glutamate-glutamine cycle and neuronal activity are tightly linked to glucose metabolism (Figure S3), and elevated glutamate concentrations found in the CSF of RS patients have also been interpreted in terms of enhanced glucose utilization [22].

An alternative explanation for increased glutamine levels in Mecp2-null mice may be a reduction in glutaminase activity, caused by a decreased need for glutamate neurotransmitter production due to a reduced number of synapses [17]. The latter might be a consequence of reduced dendritic arborization [40]. Also in this model, compensatory glutamate synthesis may maintain stable glutamate levels as suggested above. While a perturbation of the glutamate-glutamine cycle in the Mecp2-null mouse brain is well supported by our data, mechanistic details remain to be elucidated. A quantitative investigation of the relationship between glutamate-glutamine cycling, glucose and energy metabolism in RS would have to include flux measurements based on ¹³C-labeled glucose as a precursor, and would also have to consider a possible role of GABA [46].

Osmoregulation is perturbed in Mecp2-deficient mice

Myo-inositol levels were markedly decreased in Mecp2-deficient mice vs. controls, in both in vivo [20] and in vitro MRS measurements. Among all the water-soluble compounds quantitated in this study, the myo-inositol decrease and glutamine increase occurred with the highest statistical significance. In addition, the levels of both compounds were negatively correlated for all samples combined (p = 0.05, Spearman-Rank test), and myo-inositol was decreased by about the same amount (1.21 µmol/g tissue wet weight) as glutamine was increased (1.16 µmol/g). Since both compounds exert an osmoregulatory function, one may speculate that the increased glutamine accumulation discussed above may have triggered a compensatory decrease in myo-inositol. Such an effect has been described for cultured astrocytes [45], and may also occur when intracellular glutamine concentration is increased due to other mechanisms caused by Mecp2-deletion. Since myo-inositol is considered to be an astrocyte marker, a decreased myo-inositol level in Mecp2-deficient mice may alternatively indicate a lower number of astrocytes rather than a compensation for glutamine increase (see above and Table S2). However, a drastic reduction of the number of astrocytes vs. neurons is unlikely because in this case, a significant decrease in creatine would have been observed for Mecp2-deficient mice as creatine is more expressed in astrocytes than in neurons [47]. Further studies are necessary to clarify this point, as suggested above in the discussion of PL metabolism and cell growth. Future studies with alternative Mecp2 mutations (nonsense/missense, overexpression) should reveal whether the observed myo-inositol effect is specific to the particular animal model employed [20]. Reduced myo-inositol levels have not been reported in the brains of RS patients to date; however, only a few case reports are available on subjects with variable age, while brain myo-inositol concentrations are known to change with age even in healthy children [48].

Conclusion

The quantitative analysis of a large number of brain metabolites allowed us to obtain, for the first time, a comprehensive characterization of the metabolic phenotype associated with Mecp2 mutation. Most importantly, metabolic fingerprints of reduced choline phospholipid turnover were detected, implying diminished cell growth in agreement with reduced brain size (see also Table S2). Moreover, indicators of altered osmoregulation in Mecp2-null mouse brain suggest another link to changes in brain size, potentially caused by neurons and astrocytes. In addition, changed neurotransmitter metabolism hints at specific mechanisms of brain dysfunction due to Mecp2 deletion. These hypotheses open new avenues for studying the genotype-phenotype correlation in RS, and for the identification of novel molecular targets for more efficient pharmacological RS treatment. They also serve as a basis for future studies investigating brain metabolism across the age spectrum of the animals, including measurements at a presymptomatic stage with the potential to distinguish between primary and secondary effects of Mecp2 deletion.

Animals

Experiments were performed using the mouse model strain B6.129P2(C)-Mecp2 ^tm1-1Bird as described previously [20]. Briefly, the mice obtained from Jackson Laboratory were maintained on a C57BL/6 background. Hemizygous mutant males were generated by crossing heterozygous knockout females to C57BL/6 males. All experiments were performed in hemizygous Mecp2-deficient males of 5–8 weeks of age. Although RS in humans predominantly affects female patients, most researchers use Mecp2-/y male mice for their studies. This choice is dictated by the fact that the Mecp2 gene is X-linked in mouse and in humans, and females will thus have a different amount of normally Mecp2-expressing cells depending on their X-chromosome inactivation profile. Since the proportion of Mecp2-deleted X chromosomes that will be inactivated in a given female animal is unpredictable, we decided to use Mecp2-/y male mice that correspond to a complete absence of the Mecp2 gene product in all cells (i.e., a real null phenotype) [20]. One consequence of complete Mecp2 suppression is that these animals are extremely frail, and therefore are difficult to reproduce in large numbers. Hence, while Mecp2-/y males are a widely accepted RS model, it is not an ideal model. Interestingly, a recent publication comparing a male and a female RS model demonstrated that both models exhibited very similar somatic growth and behavioral changes [49]. In this pilot study, four Mecp2-/y mice and four C57BL/6 controls were used. The experimental procedures were carried out in keeping with the European Guidelines for the Care and Use of Laboratory Animals (Council Directive 86/6009/EEC).

Sample preparation

Animals were sacrificed by cervical dislocation. The brain was swiftly and entirely removed through an incision in the skull, freeze-clamped, weighed, and stored at −80°C until extraction with methanol/chloroform/water (4∶4∶4 ml). The frozen brain was broken up into several pieces in a mortar under liquid nitrogen. The brain pieces obtained were consecutively added to 4 ml ice-cold methanol and immediately homogenized using a Polytron tissue homogenizer (Kinematica, Kriens-Lucerne, Switzerland). The sample was incubated on ice for 15 min, then 4 ml of ice-cold CHCl₃ and 4 ml 4°C water were added sequentially, under vigorous shaking after each addition. Samples were thoroughly vortexed and placed at −20°C overnight for phase separation. Subsequent centrifugation at 11,000 rpm (ca. 13,000 g at max. radius) in a Sorvall Evolution RC centrifuge (Thermo Electron Corporation, Waltham, MA, USA) for 40 min at 4°C completed the phase separation.

Solvents were evaporated from the chloroform/methanol phase under a nitrogen stream, and methanol was removed from the water/methanol phase by bubbling nitrogen through the solution while samples were kept on ice. The aqueous solution was lyophilized overnight, and both lyophilizate and lipid samples were stored at −80°C. For NMR spectroscopy, lipids were dissolved in a ²H-chloroform/methanol/water solution as described elsewhere [50]. The lyophilizate was redissolved in D₂O, and pH was adjusted to ca. 7.0. The sample was placed in a 5-mm Wilmad NMR tube (528-PP; Carlo Erba-SDS, Val de Reuil, France) containing a stem coaxial insert (2 mm O.D.) filled with a D₂O solution (pH 7.7) of 20 mM sodium 3-(trimethylsilyl)-2,2′,3,3′-tetradeuteropropionate (TSP-d₄) for ¹H MRS, and 20 mM methylenediphosphonate (MDP) for ³¹P MRS, for chemical-shift referencing and quantitation. Chemicals were purchased from Sigma-Aldrich (Saint Quentin, Fallavier, France), except for a number of phospholipids used for MRS signal assignment (Doosan Serdary Research, Toronto, ON, CA).

NMR spectroscopy

¹H and ³¹P NMR spectra at 400.1 and 162.0 MHz, respectively, were obtained on a 9.4 T AVANCE 400 WB FT-NMR spectrometer from Bruker (Wissembourg, France), and analyzed with Bruker's Topspin software (for more technical details see Text S1). Metabolite concentrations are given as µmol/g brain tissue (wet weight). In addition, ratios of metabolite levels are presented where appropriate.

Statistics

The statistical methods used to detect significant differences and correlations between groups included nonparametric tests (Statview 5.0.1, Cary, NC, USA). Parametric tests such as analysis of variance are appropriate only where the conditions of equal variances and normal distribution are met. However, Statview's F-test procedure yielded unequal variances for several metabolite concentrations and metabolite ratios, and common tests for normal distribution do not give meaningful results when applied to small sample groups [51], [52]. Consequently, we consistently used the nonparametric Mann-Whitney U test to determine the significance of differences between two groups, and the nonparametric Spearman-Rank test for correlations between metabolite levels (for further statistical details see Table S1).