The systemic treatment was designed as blinded, placebo-controlled study (Figure 1). Mice were randomly assigned to a treatment group, and starting from PD10–11 they received either vehicle (phosphate buffered saline, PBS), low-dose Trolox (10 mg/kg body weight) or high-dose Trolox (40 mg/kg body weight). Compounds were administered by i.p. injection every 48 h for a duration of ~6 weeks, i.e., until the mice had reached an age at which characteristic Rett symptoms are clearly evident (~PD47–50). In case of complications (massive weight loss, infections, edema, obvious signs of pain or suffering, tremors, lethargy, ruffled fur, difficulty breathing) treatment was terminated and the respective mouse euthanized. Injections, phenotypic analyses, in vitro experiments and data analyses were team efforts, with some overlap among those performing injections, running experiments and analyzing data. Everyone actively involved was blinded with respect to both mouse genotype and type of treatment. Yet, in view of the severe phenotype of the Mecp2^−/y mice and the differences in body weight, their genotype became more and more obvious once they reached PD30 and beyond.

Details on the pharmacokinetics of Trolox are not available, and its high blood brain barrier transferability has been demonstrated only very recently (Wada et al., 2016). Therefore, our compound dosing was chosen in accordance to earlier reports in which Trolox was i.p. injected or orally fed to rats/gerbils and successfully ameliorated neurotoxic effects as well as ischemia-related neurodegeneration (Gupta and Sharma, 2006; Al Mutairy et al., 2010). At the same time, the higher Trolox dose of 40 mg/kg is the consequence of the maximum soluble amount in saline (~4 mg/ml without additional solvents) and the maximum allowable volume for i.p. injections in mice. The 48 h injection interval is based on a report on mice, which shows that upon administration, the levels of the vitamin E compound tocopherol in blood, brain, lung, kidney and liver remain stable for at least 24 h upon a single dose administration (Baxter et al., 2012). Furthermore, we were concerned that i.p. injections performed on a daily basis for several weeks—especially in view of the early onset of treatment (PD10/11)—would be too stressful for the fragile Rett mice.

All chemicals and assays were obtained from Sigma-Aldrich, unless indicated otherwise. Trolox ((±)-6-hydroxy-2,5,7,8-tetramethylchromane-2-carboxylic acid) was dissolved in PBS (1 mg/ml and 4 mg/ml injection solution), stored at 4°C, and used within 1 week. Artificial cerebrospinal fluid (ACSF) required for in vitro experiments contained (in mM): 130 NaCl, 3.5 KCl, 1.25 NaH₂PO₄, 24 NaHCO₃, 1.2 CaCl₂, 1.2 MgSO₄ and 10 glucose. Constant aeration with carbogen (95% O₂, 5% CO₂) adjusted pH to 7.4. Cyanide (CN⁻) was dissolved as aqueous 1 M stock and stored at −20°C. FCCP (carbonyl cyanide-4-(trifluoromethoxy) phenylhydrazone, Tocris Bioscience) and rhodamine 123 (Rh123) were dissolved as 10 mM stocks in dimethyl sulfoxide (DMSO) and stored at 4°C. Final DMSO concentration was ≤0.02%.

For detailed phenotyping, all mice underwent various behavioral tests prior to final experiments, to rate motor function and activity, environmental exploration, and working memory. However, to prevent an “overtesting” not every animal underwent each test available, but typically only three of the behavioral tests. Motor coordination was tested on three consecutive days (~PD43), using a mouse Rotarod (Ugo Basile). The first day allowed the mice to adapt to the treadmill, the second day defined motor performance, and improvement from the second to the third day represents motor learning. Mice were placed on the rotating drum (non-skid surface, 30 mm diameter), the speed continuously accelerated from 5 rpm to 50 rpm within 5 min, and the time each animal stayed on the rod was measured as latency to fall. Occasionally, mice clasped to the rotating drum without running. In these cases the timing was stopped after five full rotations of clasping.

Motor activity and environmental exploration were rated in open field and hole board tests (~PD47–50). In the open field (45 cm × 45 cm), the motility of mice was monitored for a duration of 5 min. Any lateral movements were detected by a grid of 16 infrared beams and tracked using ActiTrack v2.7.13 software (Panlab, Harvard Apparatus). For detailed analyses, the open field was divided in three zones (a 95 mm wide periphery, the respective 95 mm × 95 mm corners and the remaining 260 mm × 260 mm center) which were defined as a virtual mask in the tracking software (see Figure 3B). In the hole board configuration, a base-plate with 16 equally spaced holes (25 mm diameter) was inserted into the arena, and an additional grid of infrared beams detected the head-dipping behavior of the mice.

Working memory was analyzed in a Y-maze (32.5 cm long and 8.5 cm wide arms, equally spaced at 120° angles), in which mice were tracked by video camera and ANY-maze software (Stoelting). Around PD47–50, mice were placed in the start arm and allowed to explore the maze for 10 min while one arm was blocked. After 1 h the test was repeated for a duration of 5 min with all arms being accessible, and the number of spontaneous alterations was counted. An alternation is defined as consecutive entry of all three arms in any order. Thus, the number of maximum possible alternations is the total sum of arm entries minus two. To compare mice, the number of spontaneous alternations was normalized to the maximum possible alternations.

To analyze breathing (PD47–50), mice were placed in a whole-body plethysmograph (Data Sciences International) and allowed to breathe naturally under conscious and unrestrained conditions. Breathing patterns were recorded by Ponemah v5 control software. After a ~10 min adaptation, breathing was analyzed for the following 3 min. The individual respiratory cycles were detected by the threshold crossing method (Clampfit 10.3, Molecular Devices), and breathing frequencies calculated as the reciprocal of the averaged inspiratory interval. The irregularity score was determined as the normalized difference between a pair of subsequent breaths (Barthe and Clarac, 1997; Telgkamp et al., 2002; Wegener et al., 2014).

To obtain brain tissue, mice (~PD50) were deeply anesthetized with ether and decapitated. The brain was isolated and placed in chilled ACSF for ~2 min. Coronal brain slices of 400 μm thickness were cut using a vibroslicer (752M Vibroslice, Campden Instruments) and split along the sagittal midline. Depending on the experiment they were transferred to an interface recording chamber or a submersion-style storage chamber. Prior to the experiments, slices were left undisturbed for at least 90 min, to ensure their recovery from dissection and slicing. Any brain tissue not required for acute analyses was immediately frozen in liquid nitrogen for later biochemical assays. Blood samples collected upon decapitation served to determine hematocrit and blood glucose levels (Contour^® blood glucose meter; Bayer).

For transcardial perfusion, deep ether anesthesia was confirmed by the absence of foot-pinch reflexes. The chest was opened, the heart exposed, and an injection needle inserted in the left ventricle. Remaining blood was removed from the vascular system by PBS perfusion, then perfusion fixation was performed by 4% paraformaldehyde (PFA) in PBS. Once isolated, brains were postfixed for 3 days in 4% PFA and long-term stored in PBS with 0.02% NaN₃.

Perfusion-fixed brains were cut into coronal sections (30 μm) using a microtome (VT1200S Leica). The sections were placed on microscope slides, passed through a descending ethanol series (95% for 15 min; 70% and 50% for 1 min each) and washed twice in demineralized water. They were then stained in an aqueous cresyl violet/acetate solution (5 mg/ml cresyl violet, 30 μl/ml acetic acid) for 2 min. Remaining dye was washed out by water, followed by an ascending ethanol series (50% for 1 min, 70% plus 1% acetic acid for 2 min, 95% for 2 min, 99.9% for 2 min). Finally, sections were cleared in xylol (5 min), dried (15 min), and coverslipped with Eukitt quick-hardening mounting medium. For the morphometric studies a rostrocaudal location of about −1.7 mm with respect to Bregma was chosen, and three sections per brain were analyzed using a digital microscope (Coolscope, Nikon) with NIS-Elements 3.2 cell analysis software. Quantification of neuronal size is based on soma area, and it was determined by randomly selecting five cells in each section analyzed. Those cells were chosen, which showed clearly identifiable boundaries and were not obstructed by neighboring/overlapping objects. The cross-sectional area was determined in the best focal plane by drawing along the outer cell boundaries (NIS-Elements analysis software).

Electrophysiological recordings were run in an Oslo style interface chamber, which was constantly aerated with carbogen (95% O₂, 5% CO₂; 400 ml/min) and supplied with oxygenated ACSF (3–4 ml/min). Synaptic plasticity and hypoxic responses were analyzed at temperatures of 31–32°C and 35–36°C, respectively. Recording electrodes were pulled from borosilicate capillaries (GC150TF-10, Harvard Apparatus) on a horizontal electrode puller (Model P-97, Sutter Instruments) and filled with ACSF. Their tip resistance was trimmed to ~5 MΩ. Schaffer collaterals were stimulated with stainless steel electrodes (50 μm diameter microwire, AM-Systems), delivering 0.1 ms unipolar stimuli (S88 stimulator with PSIU6 stimulus isolation units, Grass Instruments). Field excitatory postsynaptic potentials (fEPSPs) and extracellular DC potentials were recorded in CA1 st. radiatum, using a locally constructed field potential amplifier (Hepp et al., 2005). Data were sampled with an Axon Instruments Digitizer 1322A and analyzed by PClamp 9.2 software (both Molecular Devices). Synaptic function was rated based on input/output (i/o) curves and long-term potentiation (LTP). For LTP experiments, stimulus intensity was set to obtain half-maximum responses. Stable LTP was induced by three subsequent trains of high-frequency stimulation (100 Hz, 1 s, 5 min train interval; Janc and Müller, 2014). Hypoxia-induced spreading depression (HSD) was elicited by switching the interface chamber’s gas supply to nitrogen (95% N₂ and 5% CO₂); O₂ was resubmitted 30 s after HSD onset (Hepp et al., 2005).

Mitochondrial metabolism and membrane potentials were monitored in CA1 st. radiatum using computer-controlled epifluorescence imaging systems. They were composed of a polychromatic xenon-light source (Polychrome II, Till Photonics), a CCD-camera (SensiCam QE, PCO) and an upright microscope (either Axiotech or Axioskop I, Zeiss). Hippocampal slices were imaged in a submersion-style chamber (30–33°C), using a 40× 0.8NA water immersion objective (Achroplan Zeiss). FAD and NADH tissue autofluorescence were monitored in a ratiometric approach (Duchen and Biscoe, 1992), by alternating excitation at 360 nm (NADH, 70 ms exposure) and 445 nm (FAD, 40 ms exposure). Image pairs were taken every 5 s (0.2 Hz frame rate), and autofluorescence was separated by a 450 nm dichroic mirror and a 510/80 nm emission bandpass filter (Gerich et al., 2006; Janc and Müller, 2014).

For Rh123 imaging of mitochondrial membrane potential changes, slices were dye-loaded (5 μM, 15 min) in a miniaturized staining chamber (Funke et al., 2007). As Rh123 was used in quenching mode, depolarization of mitochondria is indicated by an increase in Rh123 fluorescence (Emaus et al., 1986; Foster et al., 2006). Rh123 was excited at 480 nm using frame rates of 0.2 Hz and 5 ms exposure times; its emission was separated using a 505 nm dichroic mirror and a 535/35 nm bandpass emission filter.

To assess oxidative tissue damage, lipid peroxidation and protein carbonylation were quantified in cryopreserved tissue samples of frontal cortex. Lipid peroxidation was determined in ~10 mg neocortical tissue using a colorimetric assay (MAK085), which is based on the reaction of MDA with thiobarbituric acid (TBA). Tissue samples were homogenized on ice in 300 μl MDA lysis buffer containing butylhydroxytoluene (100×), and debris was removed by centrifugation (13,000 g, 10 min). Then 0.2 ml of homogenate was added to 600 μl TBA solution. This mixture was incubated for 60 min at 95°C, cooled down on ice to room temperature (RT), and centrifuged again (13,000 g, 30 s) before absorbance at 532 nm was determined against a MDA standard curve.

Protein oxidation was determined using a protein carbonyl assay (MAK094), which reports the interaction of protein carbonyl groups with 2,4-dinitrophenylhydrazine (DNPH). The resulting stable dinitrophenyl (DNP) hydrazone adducts are detected spectrophotometrically. Tissue samples were homogenized in 500 μl CelLytic M buffer and incubated for 15 min before debris was removed by centrifugation (14,000 g, 15 min). The supernatant was collected and its protein content determined by Bradford assay (B6916). To quantify protein carbonyl levels, 0.15 ml of tissue supernatant (~10 mg/ml protein) was mixed with 0.15 ml of DNPH solution and incubated at RT for 10 min. Then, protein was precipitated by adding 0.45 ml of 100% trichloroacetic acid (5 min incubation on ice) and subsequent centrifugation at 13,000 g for 2 min. The pellet was dissolved in ice-cold acetate, incubated for 5 min at −20°C and centrifuged again at 13,000 g for 2 min, before the previous acetate washing step was repeated. Then the precipitated protein was re-dissolved in 0.33 ml of 6 M guanidine solution and incubated for 20 min at 60°C. Samples were allowed to cool down to RT; their carbonyl content was calculated from the absorbance at 375 nm, and their final total protein content determined by bicinchoninic acid assay (BCA1 AND B9643).

For electrophysiological and optical recordings slices of at least five different brains of each genotype and treatment group were used. In behavioral experiments at least six mice per group were tested. All numerical values are given as mean ± standard deviation; the number of experiments (n) refers to the number of slices or mice analyzed. The significance of the changes observed was tested by two-way ANOVA followed by post hoc all-pairwise multiple comparisons (Holm-Sidak method); statistical calculations were performed with Sigmaplot 12.5 (Systat Software Inc.). In the diagrams, statistically significant differences among treatment groups of the same genotype are indicated by asterisks (^*P < 0.05, ^**P < 0.01; ^***P < 0.001), and those among WT and Mecp2^−/y mice are identified by crosshatches (^#P < 0.05, ^##P < 0.01, ^###P < 0.001).

In vivo treatment started at the presymptomatic stage (PD10–11), to detect potential differences in disease onset or progression. Injections of either PBS, low-dose (10 mg/kg) or high-dose Trolox (40 mg/kg) were performed every 48 h for ~6 weeks (Figure 1). In total, 177 mice (88 WT and 89 Mecp2^−/y mice) received blinded Trolox or PBS injections for the entire treatment duration. Another 17 mice (3 WT, 14 Mecp2^−/y) died or had to be euthanized before the end of treatment. The deceased WT mice had received 10 mg/kg (n = 2) and 40 mg/kg Trolox (n = 1); the deceased Mecp2^−/y mice belonged to PBS (n = 1), low-dose Trolox (n = 7) and high-dose Trolox-treated groups (n = 6). For summary and detailed survival rates see Table 1.

In terms of body weights, WT and Mecp2^−/y mice were indistinguishable up to the third postnatal week. Yet, from PD22 on, all Mecp2^−/y mice—independent of the treatment—showed a reduced growth and gained less body weight than the WTs (Figure 2A). Their final body weight at PD50 was about half the weight of WT mice, and their body mass index (BMI, determined as body weight divided by body length²) was also significantly lower (Figure 2B). Genotype-matched comparison did not reveal any effects of the type of treatment on the growth parameters among the different WT or Mecp2^−/y groups.

Blood analyses detected ~29% lower blood glucose levels in PBS-treated Mecp2^−/y mice than in the corresponding WTs. Those Mecp2^−/y mice receiving low or high Trolox doses, showed a trend towards higher glucose levels and they did no longer differ significantly from PBS-treated or Trolox-receiving WTs (Figure 2C). In contrast, Trolox tended to reduce glucose levels in WTs, which reached the level of significance in the high-dose treated group (Figure 2C). The increased hematocrit, which we reported already earlier for Mecp2^−/y mice (Fischer et al., 2009), was still evident in all Mecp2^−/y groups; any modulation by Trolox was not observed (Figure 2D).

Rett patients and Rett mice show complex motor impairment. To assess whether Trolox may exert any positive effects on motor coordination, rotarod tests were performed on three consecutive days (Figure 3A). After habituation (first day), all Mecp2^−/y groups showed a poorer performance on the rotarod on the second day and only stayed on the treadmill for markedly ( >40%) shorter intervals than the respective WTs. Motor learning is expected between the second and third day of testing. Indeed, WT mice regardless of the treatment showed a trend towards increased running times of ~20%. In the Mecp2^−/y groups, any improvement of motor learning between the second and third day was not seen (Figure 3A).

Exploratory behavior and locomotor activity of mice was determined in an open field arena (Figure 3B). This test mimics the natural conflict of mice between exploring a novel environment and avoiding illuminated, open areas. All mice avoided the center region, rather preferring the periphery and even more the corners (Figures 3B,C). Marked differences in the relative times spent within the three zones were not obvious between Mecp2^−/y and WT mice. Mecp2^−/y mice treated with PBS and low-dose Trolox were significantly less active than WTs, as indicated by their reduced average velocity (Figure 3D), longer relative resting time (Figure 3E), and shorter total distance traveled (Figure 3F). Interestingly, these activity parameters did not significantly differ among high-dose Trolox-treated WT and Mecp2^−/y mice, as WTs tended to become less active under this treatment (Figures 3D,E,F).

Environmental exploration of mice was tested by the hole board. Mecp2^−/y mice receiving PBS or low-dose Trolox showed significantly fewer head-dips than the respective WTs and rested longer (Figures 3G,H). High-dose Trolox-treated Mecp2^−/y mice became clearly more active than the other Mecp2^−/y groups, showing more head-dips and resting less than PBS-treated Mecp2^−/y mice. Also their head dipping did no longer differ from high-dose treated WTs (Figure 3G).

For the assessment of spatial working memory, the Y-maze is considered a useful initial test (Dudchenko, 2004). It can be performed relatively easily, does not require multiple repeats, and does not impose high levels of stress or pronounced motor challenge on the mice. In view of the clear learning/memory deficits of MeCP2-deficient mice (Pelka et al., 2006) and their motor impairment, we therefore established this test. Against expectation, the number of spontaneous alterations did not differ among the respective WT and Mecp2^−/y groups, nor did genotype-matched comparison indicate any effects of Trolox (Figure 3I). This suggests that the spatial complexity of only three arms is not sufficiently challenging to detect clear genotypic effects.

To evaluate the breathing pattern, unrestrained whole-body plethysmography was performed. Breathing was clearly disturbed in PBS-treated Mecp2^−/y mice as compared to WTs (Figure 4A), and Trolox treatment failed to improve the regularity of breathing in Mecp2^−/y mice. All Mecp2^−/y groups still showed a higher incidence of breathing arrests and lower breathing frequencies than WTs (Figures 4B,C). Accordingly, the irregularity score, i.e., the variability of respiratory cycle duration, remained increased in all Mecp2^−/y mice (Figure 4D).

To screen for differences in hypoxia susceptibility, acute hippocampal tissue slices isolated from treated WT and Mecp2^−/y mice were exposed to transient O₂ withdrawal to trigger an HSD episode, which represents the concerted network response to severe hypoxia. HSD was recorded as negative extracellular DC-potential shift, and its characteristic parameters amplitude (ΔV), duration at the half amplitude level (t_1/2), and time to onset (Δt) were analyzed (Figure 5A). In hippocampal slices of low-dose Trolox-treated Mecp2^−/y mice, the time to HSD onset was significantly increased as compared to the respective WTs and PBS-treated Mecp2^−/y mice (Figure 5B). The amplitude and duration of the DC potential shift did not differ among genotypes and treatment groups (Table 2).

Against expectation, the earlier observed increased hypoxia susceptibility of Mecp2^−/y hippocampal slices (Fischer et al., 2009; Janc and Müller, 2014) was not obvious among PBS-treated WT and Mecp2^−/y mice. To clarify, whether the lack of this genotypic difference may be due to frequent animal handling and i.p. injections, seasonal or breeding issues, HSD was triggered also in slices from untreated (naïve) Mecp2^−/y and WT mice. These controls reproduced the hastened onset of HSD by ~28% in slices from Mecp2^−/y mice with almost the identical times to HSD onset (Table 2) reported previously for naïve Mecp2^−/y and WT slices (Janc and Müller, 2014; Table 3).

Synaptic function and plasticity were assessed in acute hippocampal tissue slices by recording fEPSPs. Basal synaptic function, as judged on normalized i/o curves, did not noticeably differ between PBS-treated WT and Mecp2^−/y mice (Figure 6A). The pronounced hyperexcitability observed earlier in slices of naïve Mecp2^−/y mice (Janc and Müller, 2014), was not present among the PBS-injected groups. Neither did Trolox treatment induce any significant effects on basal synaptic function (Figure 6A). Trolox treated WT mice showed, however, a trend towards reduced fEPSP amplitudes.

Synaptic plasticity was rated upon high-frequency stimulation, which mediated a robust short-term potentiation (STP) and LTP of fEPSPs in all Mecp2^−/y and WT groups, regardless of treatment (Figures 6B,C). The amount of STP, determined right after the third stimulus train, was significantly more pronounced in low-dose Trolox-treated Mecp2^−/y mice than in the respective WTs and in Mecp2^−/y mice receiving PBS or high Trolox (Figure 6D). Other differences between genotypes and/or treatment groups were not detected. The degree of LTP (determined at 50–60 min upon train stimulation) did not differ significantly among the various groups, but the final level of LTP reached, tended to be higher in low-dose Trolox-treated than in PBS injected Mecp2^−/y mice (Figure 6D).

RTT is accompanied by mitochondrial alterations. To assess whether Trolox-treatment may mediate any effects on mitochondrial metabolism, we monitored FAD/NADH autofluorescence as well as Rh123 fluorescence in acute hippocampal tissue slices of the treated mice. In correspondence to the electrophysiological recordings, these imaging experiments were performed in CA1 st. radiatum. The basal ratio of FAD/NADH autofluorescence did not differ among PBS and high-dose Trolox treated Mecp2^−/y and WT mice; but it was slightly decreased in low-dose Trolox-treated WTs (Figure 7A). Pharmacological inhibition of mitochondrial respiration by low (100 μM) and high (1 mM) doses of CN⁻ provoked the expected decreases in FAD/NADH ratio, whose magnitudes were similar among treatment groups and genotypes (Figure 7A).

Mitochondrial polarization and anoxia vulnerability were assessed in Rh123-loaded slices, by evoking maximum depolarization with the uncoupler FCCP (5 μM, 5 min) and blocking mitochondrial respiration with CN⁻ (100 μM, 2 min), respectively. The maximum mitochondrial depolarization, indicated by a massive increase in Rh123 fluorescence, did not differ among genotypes and treatment groups (Figure 7B). Neither did CN⁻-mediated inhibition of respiration evoke any different mitochondrial depolarizations (Figure 7C). It therefore can be concluded that Trolox treatment—even when applied systemically over a longer period—does not generally interfere with mitochondrial metabolism and function.

The redox alterations associated with RTT were reported to culminate in oxidative damage in the brains of Rett mice (De Felice et al., 2014). Therefore, we determined the extent of lipid peroxidation and protein carbonylation in brain tissue of the treated mice (~PD50), to rate the antioxidative capacity of Trolox. These tests were performed on neocortex, as the hippocampi had already been used for the previously described FAD/NADH and Rh123 imaging experiments. First, we verified the feasibility of the chosen assays, and incubated neocortical tissue samples with the oxidant tert-butyl hydroperoxide (TBHP; 500 μM, 10 min) to induce oxidative damage. As compared to control tissue, TBHP increased lipid peroxidation and protein carbonyl content by 59.4 ± 8.2% and 79.7 ± 16.1% (each n = 3), respectively, confirming that both assays are functional.

Comparing neocortical samples from PBS-treated WT and Mecp2^−/y mice, did not reveal any genotypic differences in lipid peroxidation (Figure 8A). In Mecp2^−/y mice treated with high-dose Trolox, however, the degree of lipid peroxidation was significantly dampened as compared to PBS- and low-dose Trolox injected Mecp2^−/y mice and high-dose Trolox treated WT mice (Figure 8A). In WTs, high-dose Trolox just showed a tendency to decrease lipid peroxidation. Protein carbonyl content was increased in PBS-treated Mecp2^−/y mice as compared to the corresponding WTs, whereas in low- and high-dose Trolox treated Mecp2^−/y mice this genotypic difference was no longer detectable (Figure 8B). Accordingly, the degree of oxidative damage in brain tissue can be ameliorated at least partly by Trolox treatment.

MeCP2-deficiency is associated with decreased brain size, thinning of neocortical layers and smaller neurons (Belichenko et al., 2008; Fischer et al., 2009). To screen for changes in these parameters, we performed morphological analyses on perfusion-fixed brains of the treated mice. Nissl staining of thin coronal sections (30 μm) revealed the expected reduction in total brain size in Mecp2^−/y mice (Figure 9A). As referred to PBS-treated WTs, the total hemisphere size in the respective Mecp2^−/y group was smaller by an average of 9.4%, the hippocampal formation was reduced by 7.0% and neocortical layers—as analyzed here for the primary somatosensory cortex (trunk region)—were thinned by 9.1% (Figures 9B–D). The soma of CA1 pyramidal neurons was also smaller in PBS-treated MeCP2-deficient mice by an average of 8.7% than in the respective WTs (Figure 9E). Systemic Trolox treatment did not mediate any effects on these morphological parameters in WT and Mecp2^−/y mice, nor did it reverse the genotypic differences.

The detailed assessment of body weights and BMI revealed that Trolox treatment did not ameliorate the reduced growth of Mecp2^−/y mice. Once separated from their foster mothers and required to feed on their own, all Mecp2^−/y groups gained less weight than their WT littermates, as was reported earlier also for non-treated Mecp2^−/y mice (Guy et al., 2001).

Weaning marks a switch in metabolism from milk (high fat diet) to mouse chow (carbohydrate rich diet; Ferré et al., 1986), and this transition seems problematic in Mecp2^−/y mice. Loss of MeCP2 function increases insulin levels, decreases insulin signaling and results in a dissociation of insulin levels and their appropriate metabolic effects on glucose regulation (Pitcher et al., 2013; Krishnan et al., 2015). Also in Rett patients, increased glucose utilization indicates altered glucose regulation and/or metabolism (Villemagne et al., 2002). We found ~29% lower blood glucose levels in PBS-treated Mecp2^−/y mice than in WTs. This difference was abolished in Mecp2^−/y mice receiving high or low Trolox doses, indicating that chronic Trolox treatment rescued the altered glucose regulation/metabolism.

Interestingly, it was shown in rats that glucose-mediated production of mitochondrial ROS is mandatory for insulin secretion; higher mitochondrial ROS levels increased insulin secretion, whereas antioxidants decreased it (Leloup et al., 2009). Thus the lower glucose levels observed in placebo treated Mecp2^−/y mice may at least in part be a consequence of the systemic oxidative stress in these mice, which is then antagonized by Trolox. In contrast in WT mice, high-dose Trolox slightly lowered blood glucose levels by an as yet unidentified mechanism. For male C57BL6 mice (PD 56–70), the vendor specifies a blood glucose content of 13.7 ± 3.4 mM (n = 114; see http://www.criver.com/files/pdfs/rms/c57bl6/rm_rm_d_c57bl6n_mouse.aspx). High-dose Trolox treated WTs showed glucose levels of 12.1 ± 2.1 mM (n = 23), which are well within this range. As neither the growth curves of high-dose Trolox treated WT mice, nor any other of their behavioral parameters showed a significant decline, adverse systemic consequences of these moderate blood glucose changes seem unlikely.

The hematocrit was not affected by Trolox, and remained increased in all Mecp2^−/y groups. As the high hematocrit represents an adaptive response of Mecp2^−/y mice to irregular breathing and the associated intermittent systemic hypoxia (Stettner et al., 2008; Fischer et al., 2009), this suggests that Mecp2^−/y mice, despite chronic Trolox treatment, still suffered from intermittent systemic hypoxia. The unchanged breathing irregularities detected in all Mecp2^−/y groups during plethysmography support this notion.

Motor problems and insufficient motor skills are evident in various MeCP2-mutant mice (Guy et al., 2001; Shahbazian et al., 2002; Moretti et al., 2005; Wegener et al., 2014). Compared to WTs, the treated Mecp2^−/y mice failed markedly earlier on the rotarod, and their impaired motor function was not ameliorated by Trolox. The poor rotarod performance likely arises from an insufficient motor coordination, as is suggested by the pronounced hind limb clasping in symptomatic Mecp2^−/y mice (Guy et al., 2001). General muscle weakness is a less likely cause. We found most Mecp2^−/y mice being able to climb around in their cages, and others confirmed normal muscle strength of Mecp2^−/y mice in wire suspension tests (Santos et al., 2007).

Mecp2^−/y mice are also less active (Guy et al., 2001), which in part may be due to their motor deficits. They traveled shorter distances than WTs, which in the open field and hole board tests was accompanied by a decline in mean velocity and prolonged resting times. Marked genotypic differences in the relative times spent in the defined areas were not found, but low-dose Trolox-treated Mecp2^−/y mice stayed slightly more in the periphery and less in the corners than the respective WTs. Environmental exploration in the hole board was also less pronounced in Mecp2^−/y mice. Yet, the number of head-dips was increased in high-dose Trolox-treated Mecp2^−/y mice as compared to the other Mecp2^−/y groups, and they did no longer differ from high-dose Trolox-treated WTs. This was paralleled by a shortened mean resting time of the respective Mecp2^−/y mice. Accordingly, environmental exploration in Mecp2^−/y mice slightly improved upon Trolox treatment.

One of the most prominent symptoms in RTT is erratic breathing (Julu et al., 2001; Weese-Mayer et al., 2006; Stettner et al., 2008; Katz et al., 2009). Each episode of irregular breathing evokes a drop in arterial O₂ saturation, which is restored only after normal ventilation resumes. Such intermittent systemic hypoxia may provoke damage and/or induce redox stress in highly vulnerable neuronal tissue. Also the regulation of breathing itself might be affected. H₂O₂-evoked oxidative stress modulates the in vitro breathing pattern generated in the isolated medullary pre-Bötzinger Complex of neonatal WT mice: Biphasic responses were reported, consisting of an initial depression followed by augmented respiratory activity (Garcia et al., 2011). Therefore, we also assessed whether Trolox may improve breathing. As expected, Mecp2^−/y mice showed highly irregular breathing patterns. High- or low-dose Trolox did not improve the regularity of breathing, and the number and duration of apneas did not decrease. Yet, a further deterioration of breathing was not observed either. It should be mentioned though that the individual Mecp2^−/y mice varied greatly; some displayed only mild breathing disturbances with just 1–2 apneas during the evaluated 3 min interval, whereas others showed up to 49 apneas exceeding 750 ms.

In contrast to our earlier studies on naïve mice (Fischer et al., 2009; Janc and Müller, 2014), PBS-treated WT and Mecp2^−/y mice did not differ in their hypoxia susceptibility assessed as time to HSD onset. Nevertheless, treatment with low-dose Trolox increased the hypoxia tolerance of Mecp2^−/y mice as compared to the respective WT and the PBS-treated Mecp2^−/y group. This effect was not seen with high-dose Trolox, emphasizing the importance of well-adjusted cellular redox balance and careful antioxidant dosing. In view of the irregular breathing and intermittent hypoxia, the Trolox-mediated increase in hypoxia tolerance is clearly of merit, as it may prevent additional complications especially in highly anoxia-vulnerable neuronal networks such as the hippocampus and cortex.

Synaptic plasticity is impaired in Mecp2-mutant mice (Asaka et al., 2006; Moretti et al., 2006; Fischer et al., 2009; Janc and Müller, 2014). Yet, the differences among PBS-treated Mecp2^−/y and WT mice were less evident than in naïve mice (Janc and Müller, 2014). As suggested by the normalized fEPSP amplitudes recorded in CA1 st. radiatum (Table 3), this seems to result from PBS-treated WTs showing a higher degree of neuronal excitability than naïve WT mice. This variability of genotypic differences in synaptic plasticity may be explained by earlier reports identifying stress and stress-hormones as a cause for impaired hippocampal LTP (Kim and Yoon, 1998; Kim and Diamond, 2002). Repeated i.p. injections for several weeks could represent a chronic stressor (Ryabinin et al., 1999) which induces stress hormone release and dampens LTP in slices of frequently handled/injected WT mice. In Mecp2^−/y mice, LTP is already impaired to a degree that it may not have responded to any additional stress.

Nevertheless, in vivo Trolox treatment mediated some positive effects on synaptic plasticity. Low-dose Trolox-treated Mecp2^−/y mice showed a significantly improved STP as compared to the respective WTs and the other Mecp2^−/y groups. The final amount of LTP also tended to increase, but did not reach the level of significance. High-dose Trolox treatment failed to improve synaptic plasticity in Mecp2^−/y or WT mice. This may result from a probably too high concentration mediating adverse effects. As a certain ROS level is necessary for signal transduction cascades during normal physiological processes (Serrano and Klann, 2004), optimum synaptic plasticity can be achieved in a well-balanced redox environment only. Accordingly, superoxide scavenging and overexpression of superoxide dismutase 3 impair hippocampal LTP (Klann et al., 1998; Thiels et al., 2000). Along this line, high-dose Trolox may have shifted hippocampal redox balance beyond the optimum. Similar effects we found earlier in naïve mice, in which Trolox rescued STP and LTP in Mecp2^−/y hippocampal slices, but dampened it in WTs (Janc and Müller, 2014).

Trolox did not mediate any consistent changes in mitochondrial metabolism or polarization. A slight decrease of the FAD/NADH baseline ratio was found in low-dose Trolox treated WTs, but this was not accompanied by any changes in mitochondrial polarization, estimated here as Rh123 fluorescence. Hence, general negative side effects of this compound on mitochondrial function can be excluded. In contrast to our earlier studies on naïve mice (Großer et al., 2012; Janc and Müller, 2014), PBS injected/handled WT and Mecp2^−/y did no longer differ in their hippocampal FAD/NADH ratios and Rh123 responses (Table 3). Again, the stress associated with the frequent i.p. injections, might have affected mitochondrial metabolism, thereby diminishing the genotypic differences. Most primary mediators of stress responses exert numerous effects on mitochondrial metabolism and ROS generation and may even induce apoptosis (Manoli et al., 2007).

RTT is closely linked to oxidative stress and redox imbalance (Sierra et al., 2001; De Felice et al., 2011, 2012b; Großer et al., 2012; Müller and Can, 2014). Unclear is, however, which particular brain regions are affected and whether the oxidative damage mainly targets proteins, lipids or both. Earlier studies on whole brains of various Rett mouse models indicated that in particular lipids are subject to oxidative damage, whereas protein damage is not detectable in each model (De Felice et al., 2014). Our comparison of neocortical tissue of PBS-treated WT and Mecp2^−/y mice did not reveal obviously different levels of lipid peroxidation, but high-dose Trolox decreased the extent of lipid peroxidation in Mecp2^−/y neocortex below WT levels. Furthermore, the increased protein carbonyl content detected in PBS treated Mecp2^−/y mice, was no longer evident upon Trolox treatment. This confirms that systemic Trolox administration is capable of ameliorating oxidative brain tissue damage. However, it also should be noted here that the sensitivity range of these assays may not be ideal to detect reliably also more subtle changes in protein and lipid oxidation. We verified both assays by provoking oxidative tissue damage via direct oxidant (TBHP) treatment. This resulted in a clear peroxidation of lipids and carbonylation of proteins in neocortical samples, but it also revealed that despite direct oxidant stress only marked but not massive changes in spectrophotometric readout were obtained.

Various earlier studies verified that antioxidants and free-radical scavengers ameliorate certain aspects of the complex clinical appearance of RTT (Chapleau et al., 2013). In female Rett mice, alterations in vascular endothelium were reversed by administration of curcumin, an anti-inflammatory compound which also mediates antioxidative effects (Panighini et al., 2013). In Rett patients, ω-3 PUFA-supplementation led to an improvement of motor function, non-verbal communication and breathing regularity (De Felice et al., 2012a; Maffei et al., 2014). Ultimate proof for a pathogenic role of oxidative alterations in RTT is the brain-specific reactivation of the Mecp2 gene, which resulted in a clear phenotypic improvement in concert with a normalization of various oxidative stress markers (De Felice et al., 2014).

As vitamin E levels are decreased in blood serum of Rett patients (Formichi et al., 1998) and as this vitamin as well as its derivatives act as chain-breaking antioxidants, which prevent the propagation of free radicals in membranes and plasma lipoproteins (Traber and Stevens, 2011; Alberto et al., 2013), these compounds appear especially promising. Applied in vivo by oral uptake or i.p. injection, Trolox ameliorates ischemia-induced hippocampal cell loss and 3-NPA-mediated striatal lesions (Gupta and Sharma, 2006; Al Mutairy et al., 2010) as well as oxidative-stress related retinal damage (Dorfman et al., 2006). Our previous in vitro tests confirmed the efficacy of Trolox also in a Rett mouse model, in which it improved synaptic function as well as hypoxia tolerance and opposed neuronal hyperexcitability in Mecp2^−/y hippocampal slices (Janc and Müller, 2014).

In the present study, the systemic administration of Trolox improved some of the tested parameters as well. Yet, this in vivo Trolox treatment did not noticeably change the most characteristic Rett symptoms. Obvious signs that the disease progression was slowed down in Mecp2^−/y mice were not obtained either. It has to be considered though that these male Rett mice are very severely affected. Therefore, to understand completely the merits of systemic Trolox administration, it needs to be assessed in future trials, how efficient this treatment would be in heterozygous female Rett mice. They represent the conditions of a cellular MeCP2 mosaicism, truthfully reflecting the genetic conditions of the female patients. As disease progression and symptom severity are milder in female Rett (Mecp2^+/−) mice, this will require, however, prolonged treatment durations of several months or even over a year. This certainly cannot be achieved by i.p. injections and the associated stress, as we learned in this present study.

In conclusion, this detailed in vivo study confirms some pharmacotherapeutic merit of systemically applied antioxidants in MeCP2-deficient brains. Unfortunately, frequent animal handling and repeated injections apparently modulated the phenotypic appearance of mice and diminished the genotype-related differences of numerous parameters. Therefore, the route of drug administration and the frequent mouse handling are critical issues in study design, which need to be further optimized in future trials. Also, as Trolox ameliorated only a few of the multiple complex Rett symptoms, one may consider its use as adjuvant, i.e., its combination with other compounds improving different symptoms, in order to complement the spectrum of effects. For example, respiration, which was not improved by Trolox was shown earlier to be normalized by norepinephrine uptake blockers, GABA uptake inhibitors, and serotonin 1a receptor agonists (Zanella et al., 2008; Abdala et al., 2010) or the antidepressant mirtazapine which furthermore prevented cortical atrophy (Bittolo et al., 2016). In view of the complex phenotype of RTT, such combination therapies, designed to the particular needs of each individual patient are probably the most promising but also the most challenging strategy.