|Home | About | Journals | Submit | Contact Us | Français|
Early diagnosis and follow-up of neurodegenerative diseases are often hampered by the lack of reliable biomarkers. Neuroimaging techniques like magnetic resonance spectroscopy (MRS) offer promising tools to detect biochemical alterations at early stages of degeneration. Intracellular pH, which can be measured noninvasively by 31P-MRS, has shown variations in several brain diseases. Our purpose has been to evaluate the potential of MRS-measured pH as a relevant biomarker of early degeneration in Huntington's disease (HD). We used a translational approach starting with a preclinical validation of our hypothesis before adapting the method to HD patients. 31P-MRS-derived cerebral pH was first measured in rodents during chronic intoxication with 3-nitropropionic acid (3NP). A significant pH increase was observed early into the intoxication protocol (pH=7.17±0.02 after 3 days) as compared with preintoxication (pH=7.08±0.03). Furthermore, pH changes correlated with the 3NP-induced inhibition of succinate dehydrogenase and preceded striatum lesions. Using a similar MRS approach implemented on a clinical MRI, we then showed that cerebral pH was significantly higher in HD patients (n=7) than in healthy controls (n=6) (7.05±0.03 versus 7.02±0.01, respectively, P=0.026). Altogether, both preclinical and human data strongly argue in favor of MRS-measured pH being a promising biomarker for diagnosis and follow-up of HD.
Huntington's disease (HD) is a neurodegenerative disorder characterized by abnormal movements and dementia associated with degeneration of the striatum and to a lesser extent cerebral cortex (Brouillet et al, 2005; Harper, 1991). Even if mechanisms of neurodegeneration in HD remain partially unknown, defects in energy metabolism—in particular anomalies in succinate dehydrogenase (SDH) activity—might play a key role in the pathogenesis of this disorder (Beal, 2005; Brouillet et al, 1999; Damiano et al, 2010). This observation has driven numerous studies aiming at validating functional imaging measurements as biomarkers for HD. Positron emission tomography and magnetic resonance spectroscopy (MRS) studies have evidenced anomalies in cerebral glucose consumption (Kuhl et al, 1982; Powers et al, 2007), tricarboxylic acid (TCA) cycle rate (Boumezbeur et al, 2005; Henry et al, 2002), and concentrations of brain metabolites including lactate (Jenkins et al, 1998), N-acetylaspartate, and myo-inositol (Dautry et al, 2000; Hoang et al, 1998), either in animal models or in patients. However, none of these measurements has proven sensitive and robust enough to serve as an early biomarker of HD, so that diagnosis and monitoring of the disease remain impossible before irreversible brain damage.
Among magnetic resonance approaches, 31P-MRS has the advantage of detecting phosphorus-containing compounds, which are directly involved in cell energy metabolism, making it possible to assess the cerebral energetic status noninvasively. In particular, 31P-MRS makes it possible to accurately assess the intracellular pH of living tissues (Petroff and Prichard, 1983). Magnetic resonance spectroscopy-measured pH anomalies have been reported in brain diseases such as epilepsy (van der Grond et al, 1998), brain tumor (Albers et al, 2005; Gerweck and Seetharaman, 1996), ischemia (Welch et al, 1992), chronic hypoxia (Hamilton et al, 2003), psychiatric disorders (Hamakawa et al, 2004; Jensen et al, 2008; Kato et al, 1994), and Parkinson's disease (Rango et al, 2006).
In this context, our aim was to test the hypothesis that cerebral pH is modified in HD using 31P-MRS and therefore can be used as a biomarker of HD. The hypothesis was first validated in a well-characterized rat model of HD (Bizat et al, 2003a) using a 4-T preclinical MRI, demonstrating correlation of pH changes with brain SDH enzymatic activity. This approach was then translated to a clinical 3T MRI to measure brain pH in a group of HD patients.
This study was conducted on five male Lewis rats (336±9g, 12-week-old; Iffa Credo, L'Arbresle, France). All experimental procedures were performed in strict accordance with the recommendations of the European Community (86/609) and the French National Committee (87/848) for care and use of laboratory animals. A well-characterized model of chronic intoxication using 3-nitropropionic acid (3NP) was chosen (Bizat et al, 2003a): 3NP was delivered by continuous infusion at a dose of 54mg/kg per day using osmotic minipumps (Alzet, Palo Alto, CA, USA) implanted subcutaneously in the back of the animals.
For all nuclear magnetic resonance experiments, animals were secured in the prone position in a custom-built cradle. A water-heating pad was used to keep body temperature stable. Anesthesia was induced with 5% isoflurane in a 1.5ml/min oxygen flow and maintained with 2% isoflurane in a 1.5ml/min oxygen flow applied with a face mask allowing free breathing. Rodent experiments were performed on a 4-T magnet (Magnex, Abingdon, UK), equipped with high-performance gradients (20cm ID, 200mT/m, 260microseconds rise time) and interfaced to an Avance (Bruker, Ettlingen, Germany) console. A custom-built 31P radiofrequency surface coil (Ø2.5cm, 68.97MHz) was used to measure brain pH. Shimming procedure and anatomical imaging were performed using a 1H surface coil (Ø5cm, 170.28MHz; Bruker).
Each animal underwent MR sessions at four different stages of the disease: before pump implantation (Ctrl), 1 day after (D1), 3 days after (D3), and 5 days after (D5). For each session, the MR experiment was conducted as follows: scout fast low angle shot (FLASH) images were acquired to position the 31P detection volume (10 × 7 × 8mm3) in the rat brain. First- and second-order shimming was performed on this volume, leading to a typical ~15Hz line width on water. Then two 31P spectra were acquired using an optimized point resolved spectroscopy (PRESS) sequence (echo time/repetition time=8.1/4,000milliseconds, 1,024points, 5,900Hz spectral width, NT=1,024, acquisition time=68minutes). Considering the long acquisition time, a scout image was acquired between the two 31P acquisitions to check for animal position. In case of motion, the voxel of interest was repositioned and shimming procedures were repeated. The two 31P spectra were then phased, corrected from frequency shift and summed. Finally, T2-weighted images optimized to detect cerebral lesions (rapid acquisition with refocused echoes (RARE), echo time/repetition time=60/3,000milliseconds, 192 × 192 matrix, 300μm in-plane resolution) were acquired. The total acquisition time including shimming, 31P spectroscopy and anatomical imaging was ~3hours.
All 31P spectra were zero filled to 2,048 points and analyzed using an Advanced Method for Spectral Fitting (AMARES) within jMRUI software (MRUI Project, European Union Program) (Naressi et al, 2001). Thirteen 31P multiplets were included in the basis set (Jensen et al, 2002), assuming lorentzian line shapes (S=singlet, D=doublet, T=triplet): PE (phosphoethanolamine) (S), Pser (phosphoserine) (S), PC (phosphocholine) (S), DPG (2,3-diphosphoglycerate) (D), Pi (inorganic phosphate) (S), GPE (glycerophosphoethanolamine) (S), GPC (glycerophosphocholine) (S), MP (membrane phospholipids) (S), PCr (phosphocreatine) (S), ATPγ (adenosine triphosphate γ) (D), α ATPα (D) and β ATPβ (T), and NAD (nicotinamide adenine dinucleotides) (S). Resonance frequencies and line widths used as prior knowledge were estimated using the values of T2 and J-coupling constants taken from the literature (Govindaraju et al, 2000). Strong constraints were applied to the amplitude (Amp) and relative resonance frequencies of double and triple multiplicity metabolites to maintain the structure of the corresponding multiplets: Amp(DPG)1=Amp(DPG)2, Amp(ATPα)1=Amp(ATPα)2, Amp(ATPγ)1=Amp(ATPγ)2, Amp(ATPβ)1=0.5 × Amp(ATPβ)2=Amp(ATPβ)3. Relative amplitudes of all phospholipids resonances were estimated from a high signal-to-noise ratio (SNR) spectrum (sum of spectra from the four imaging time points), and empirically set as: Amp(PSer)=0.35 × Amp(PE), Amp(MP)=1.2 × Amp(GPC), Amp(GPE)=0.56 × Amp(GPC). The relative phase of each resonance was calculated from the estimated phase of the PCr peak, assuming a first-order phase of 1.5° per p.p.m. (corresponding to the delay between excitation and first acquisition point). Soft constraints were applied on the resonance frequencies of all metabolites to allow for pH-induced chemical shift variations. Importantly, no strong constraint was applied to the quantification of the Pi peak. Given the relatively low sensitivity of 31P detection in rodents, 31P spectra were summed over the five animals for each time point (Ctrl, D1, D3, and D5). Indeed, fitting the average spectra minimizes numerical instabilities associated with the fit of noisy data sets (Boumezbeur et al, 2004; Henry et al, 2002). Then, pH was calculated for the four time points with jMRUI from the chemical shift of Pi relative to PCr. The parameters of Pi–PCr system were set to pK=6.77, δHA=3.23, and δA=5.70p.p.m. (Petroff and Prichard, 1983).
The study was conducted on 13 participants: seven HD patients (five males, two females, aged 47±11 years) presenting global brain atrophy, as qualitatively assessed on T1-weighted clinical MR images, and six controls matched for age and sex (four males, two females, aged 43±8 years). The duration of HD since the onset of symptoms was comparable for all seven patients (7±1 years). Patients were recruited in the framework of the MIG-HD (multicentric intracerebral grafting in HD). The protocol was approved by the ethics committee of Henri Mondor Hospital. All subjects gave their written informed consent. They had no psychiatric or neurologic disorders except HD. Table 1 summarizes the characteristics of all patients included in the study (age, gender, body mass index, and number of cytosine-adenine-guanine (CAG) trinucleotide repeats in Huntingtin gene) as well as their UHDRS (Unified Huntington's Disease Rating Scale) clinical scores (Huntington Study Group, 1996). The age and gender of the control subjects are also reported in Table 1.
Human experiments were performed on a whole-body 3T system (Bruker) equipped with a 31P quadrature birdcage coil (Ø27cm, 50.7MHz). Contrary to rodents, the volume of skeletal muscle surrounding the human skull is negligible as compared with the brain volume. For this reason, 31P signal can be localized in the brain by the radiofrequency coil only (Hamakawa et al, 2004; Kato et al, 1994). The edge of the cylindrical coil was positioned at the level of the upper lip so that the detection volume contained the brain while limiting contamination from masseter, sternocleidomastoid, and trapezius muscles. First-order shimming was performed manually on the 31P signal, leading to a typical ~15Hz line width on PCr. Spectra were then collected using a pulse-acquire 31P sequence without localization (100microseconds 90° hard pulse, SW 10k, 1,024 points, repetition time=7.5seconds, NT=28). The total acquisition time including shimming and collection of the pulse-acquire spectra was <10minutes.
Pulse-acquire 31P spectra of the upper head exhibit the superimposition of metabolite peaks to a broad baseline associated with bone signal (line width≈50p.p.m.) (McNamara et al, 1994), which must be eliminated for valuable quantification of metabolites. To do so, the truncating function of the AMARES package was used, as previously described (Vanhamme et al, 1997). Briefly, this function allowed to left shift the free induction decays (FIDs) 12 points (1.2milliseconds) to eliminate the background signal, quantify the truncated signal, and finally automatically correct the amplitudes and line widths of the estimated components for the truncated points through adjustment of the estimated lorentzian line shapes of all metabolites to the original signal. Spectra quantification and pH calculation were performed as in rodents. However, the higher sensitivity of human 31P acquisitions allowed for individual pH calculation for each human subject. Given the small number of subjects, a Wilcoxon statistic analysis was performed to assess the significance of pH changes between HD and control groups.
Animal behavior was monitored over the protocol duration (5 days) to assess the effects of the 3NP chronic intoxication. At D1 (24hours after implantation of 3NP-delivering pumps), no obvious toxic effect was visible. However, all 3NP-treated rats showed symptoms of drowsiness, slowness of movement, and general uncoordination at D3. After 5 days, all animals were lying in a recumbent position and showed typical paddling movements of hind limbs, in agreement with previous description of this intoxication model (Bizat et al, 2003a; Garcia et al, 2002; Mittoux et al, 2002; Ouary et al, 2000).
Typical 31P spectra collected before 3NP administration and on the third day of 3NP administration are presented in Figure 1. Cerebral pH values are plotted versus time in Figure 2A. An increase in brain pH was detected on the first day of 3NP intoxication as compared with the control value (pHD1=7.11±0.02 versus pHCtrl=7.08±0.03, P=0.1). On the third day of 3NP intoxication, brain pH became significantly higher than the control value (pHD3=7.17±0.02 versus pHCtrl=7.08±0.03, P<0.001). After the onset of striatal lesions on D5, brain pH decreased but remained significantly higher than before 3NP administration (pHD5=7.12±0.02 versus pHCtrl=7.08±0.03, P=0.04).
Using a well-characterized 3NP model allowed us to compare changes in brain pH with SDH inhibition previously reported on the exact same model of chronic intoxication: Figure 2A shows the time course of SDH inhibition reported by Bizat et al (2003a). Comparison with our study showed significant correlation of 31P-measured pH with SDH inhibition (P<0.05).
Figure 2B shows T2-weighted MRI acquired on the same animal. No major lesion was visible in any animal before D5. In contrast, massive striatal lesions were detected at D5 on the five animals, as shown by strong T2 hypersignal in both striata.
31P measurement of brain pH was successfully conducted for all HD patients included in the study. Our acquisition procedure made it possible to accurately measure brain pH in <10minutes, with a low sensitivity to head motion appropriate for patients suffering from abnormal movements. Typical nonlocalized 31P spectra acquired in one HD patient and in one healthy subject are shown in Figure 3. The spectra demonstrate upfield Pi shift toward alkaline values in HD. Brain pH was significantly higher in HD than in controls (pHHDpatients=7.05±0.03 versus pHcontrol=7.02±0.01, P=0.026). Note that no correlation was found between pH and the number of CAG repeats, likely due to the narrow range of CAG repeats (41 to 54) in the HD group.
For controls and HD patients, the relative concentration of each 31P-detected metabolite was calculated as the ratio of metabolite peak area to the sum of all 31P peak areas. No significant difference was observed in metabolite concentrations between HD patients and controls (data not shown).
All pH values reported in this study are global cerebral values. In the 3NP rodent model, pH measurements were actively localized to the brain using a single voxel PRESS sequence to avoid signal contamination from the surrounding skeletal muscles. Given the negligible amount of muscle tissue surrounding the human skull (Hamakawa et al, 2004; Kato et al, 1994), the human study was based on nonactively localized pH measurements, the cerebral localization being passively defined by the geometry of the volume coil.
Huntington's disease being characterized by an early striatal neurodegeneration preceding the onset of cognitive and motor symptoms (Harper, 1991), 31P-MRS measurements localized to the striatum would be of high interest. However, one is faced with intrinsic difficulties when trying to perform localized 31P-MRS in HD patients striata. On the technical side, localized 31P-MRS is an intrinsically low sensitivity method. Furthermore, because of the major atrophy of HD patients striata, this structure is of low cellular density. The combination of these methodological and biological limitations make it impossible to obtain localized striatal measurements in experimental conditions compatible with clinical examination of HD neurologic patients, as 31P-MRS low sensitivity combined with striatal low cellular content leads to poor signal-to-noise and strong partial volume effect.
Nonetheless, global brain pH measurements may be a useful indicator of brain energy status in HD. Even though HD is characterized by early striatal neurodegeneration, energy deficits in this pathology have been shown to affect the entire brain, particularly, the cortex and the basal ganglia, and even peripheral organs (Mochel and Haller, 2011; Thibaud et al, 2010). As a consequence, the global cerebral pH measured in this study are weighted means of pH values from all cerebral regions, which all are, at least partially, affected by neurodegeneration or energy deficit. In other words, the measurement of whole-brain cerebral pH reflects global energetic deficiency in HD.
The translational study presented here demonstrates significant pH increases in both animal model and HD patients. First of all, the use of a well-characterized animal model made it possible to follow HD progression and to look for parallels between pH changes and development of brain lesions. In all animals, no striatal lesion was detected by T2-weighted MRI at D3, the time point when cerebral pH reached its highest value. In contrast, all rats exhibited bilateral striatal lesions at D5. Therefore, significant changes in cerebral pH can be measured using 31P-MRS before any detection of brain lesion by conventional anatomical imaging. Furthermore, pH changes correlated with the inhibition of the mitochondrial enzyme SDH, showing a regular increase from D1 to D3 followed by a slight decrease. It must be kept in mind that defect in SDH activity—which has been consistently found in several post mortem biochemical analyses of the HD striatum (Damiano et al, 2010)—is a key feature of this disease. As a consequence, 31P-derived pH appears as an early biomarker of neurodegeneration in the animal model, reflecting the metabolic impairment status consecutive to SDH inhibition.
Under normal physiological conditions, a large majority of brain Pi is intracellular, so that MRS measurement of pH is considered as a measure of intracellular pH. One may wonder whether this holds true in HD, knowing that extracellular pH is significantly higher than intracellular pH (7.30 versus 7.05) (Kaila and Bruce, 1998), and that brain atrophy may significantly increase extracellular contribution. Quantitative estimate of intracellular and extracellular contributions to the Pi peak (Veech et al, 1979) shows that intracellular contribution under physiological conditions is ~40 times as high as extracellular contribution: intracellular Pi concentration (~2.5mmol/L) is five times as high as extracellular concentration (~0.5mmol/L) (Luyten et al, 1989; Veech et al, 1979) and intracellular volume is eight times as high as extracellular volume in the brain (Mascalchi et al, 2004). Therefore, extracellular Pi could contribute to the measured pH increase only in case of strongly increased extracellular volume. This possibility is not supported by currently available data: the relaxation time T2, which is highly sensitive to extracellular volume, does not exhibit any change in our rats during pH increase (first 3 days of 3NP intoxication). In addition, no histological alteration was observed in the rat brain at the same stage of intoxication (Bizat et al, 2003a). Human data also show that HD patients exhibit T2 values, which are similar or shorter than healthy subjects (Vymazal et al, 2007). Consequently, pH values measured in our study are likely to represent intracellular pH, under physiological as well as under pathological conditions. Recent in vivo MRS data suggest that a large majority of brain Pi is located in neurons as opposed to glial cells, based on the fact that Pi is in a compartment where oxidative metabolism strongly dominates glycolytic metabolism (Chaumeil et al, 2009; Lei et al, 2003). Altogether, these observations allow us to interpret our results as an increase in neuronal pH in HD.
Cerebral alkalosis has been reported in several other brain diseases such as transient ischemia (Kogure et al, 1980; Mabe et al, 1983), acute ischemic stroke (Chopp et al, 1990), chronic cerebral infarction (Hugg et al, 1992; Sappey-Marinier et al, 1992), hypoxic encephalopathy (Mitsufuji et al, 1995), and neonatal encephalopathy (Robertson et al, 2002). The most common explanation for alkaline pHi includes the activation of the electroneutral Na+/H+ exchanger: Na+ influx into the cell induces the efflux of H+ to maintain electroneutrality. Triggers of Na+/H+ antiporter activation are still investigated, as this transporter is regulated by several extracellular factors such as hormones, growth factors, and cytokines. Interestingly, brain pHi was also positively correlated with brain tissue lactate levels in most studies. One hypothesis is that systemic accumulation of sodium lactate would be the trigger to the intracellular alkaline shift through a compensatory mechanism involving Na+/H+ exchangers. Another hypothesis in that astrocytes alkalinization could increase the glycolitic rate through the pH-dependent activation of phosphofructokinease, inducing in turn an excess of lactate pumped out of the cellular compartment. Decrease in the PCr/Pi ratios reported in hypoxia supports the hypothesis of an increased glycolysis. In the case of HD, increased lactate levels have been reported by Jenkins et al (1993, 1998) but no changes in 31P-containing metabolites were found in our study. Further studies will therefore be required to establish whether lactate activated Na+/H+ antiporter significantly contributes to pH changes in HD.
The 3NP model used here has been previously characterized in our laboratory and by other groups worldwide (Bizat et al, 2003a, 2003b; Brouillet et al, 2005). One important characteristic of this 3NP model is that all animals respond homogenously to the toxic effects of the neurotoxin. Therefore, pH changes measured in our study can be interpreted with regard to biochemical events previously reported for this model: from the beginning of 3NP treatment to D3, no histological brain lesion was observed. However, on day 3, biochemical evaluation of the striatum in 3NP-treated animals showed that cytochrome C was released from mitochondria to the cytoplasm, reflecting mitochondrial impairment in particular loss of mitochondria membrane potential in the striatum. Later on day 4, caspase-9 was transiently found in its active form as seen using enzymatic, biochemical, and immunohistochemical analysis (Bizat et al, 2003a, 2003b). Activation of caspase-9 is an ATP-dependent process, so that at day 3 ATP levels are likely to be maintained in the striatum of 3NP-treated rats. On day 4, deregulation of Ca2+ homeostasis, resulting from mitochondrial dysfunction, produced massive activation of the protease calpain leading to cell demise. At day 4 to day 4.5, histological lesions appeared in the striatum (Bizat et al, 2003a). These data confirm, in addition to our MRI data showing no T2 change in the striata before D5, that pH increase precedes striatal lesions and might be concurrent to early mitochondrial dysfunction associated with maintained ATP level.
Increased pH has recently been observed in muscle disease using the same 31P-MRS approach (Thibaud et al, 2010). The sensitivity of 31P detection being much higher in skeletal muscle than in the brain, the authors observed that the Pi peak was not simply shifted toward alkaline values, but that a second Pi peak appeared nearby the ‘physiological' Pi peak (~0.3p.p.m. upfield) under pathological conditions. This second Pi peak was ascribed to a pool of suffering cells that could not maintain proper ionic homeostasis. Given the detection sensitivity and spectral resolution of our brain studies, it is very likely that such a peak would not be discriminated from the ‘physiological' Pi peak on our data, resulting in an enlarged combined Pi peak which maximum intensity is shifted toward alkaline chemical shift.
In conclusion, the translational approach used here shows that neuronal pH (1) is increased in an animal model of HD as well as in HD patients, (2) correlates with enzymatic impairment, and (3) increases before histological and MRI-visible brain lesion in rodents. Given the increasing availability of 31P-MRS on clinical MR scanners, these results make intracellular pH a promising biomarker for HD. Longitudinal measurements on the same patients will be needed to determine whether 31P-derived pH is a reliable biomarker of disease progression in the individual patient.
Fundamental studies are also required to better characterize the structure of the Pi peak observed on the 31P spectrum under pathological conditions. Given the relatively low sensitivity and spectral resolution at 3 and 4T, our study does not rule out the possibility of an additional Pi peak appearing upfield under degeneration, as reported in muscle disease (Thibaud et al, 2010). In vivo studies at higher magnetic field will be necessary to address this issue.
Finally, alternative approaches are available for in vivo pH measurement, based on PE detection (Corbett et al, 1987; Petroff et al, 1985) or on nucleoside triphosphates detection (Williams and Smith, 1995). Although the sensitivity of our low-field study did not make it possible to implement these approaches, further high-field studies will be of major interest to confirm and help interpret our results.
The MIG-HD trial is granted through two PHRC AOM00139 and AOM 04021 from the DRCD (Assistance Publique-Hôpitaux de Paris) and the support of the AFM. The management center involves ACB-L (principal investigator), S Palfi, P Remy, M Peschanski, P Hantraye, JP Lefaucheur, D Challine, and P Maison. CROs were AR and D Schmitz. The sites PI are ACB-L, S Palfi (Créteil), P Krystkowiak, S Blond (Lille), JF Démonet, Y Lazorthes (Toulouse), CV, P Menei (Angers), P Damier, Y Lajat (Nantes), F Supiot, M Levivier (Bruxelles).
The authors declare no conflict of interest.