Comparison between proton magnetic resonance spectroscopy and high- performance liquid chromatography to quantify muscle carnosine in humans

Proton magnetic resonance spectroscopy (1H-MRS) has been used as a non-invasive alternative to quantify carnosine in human muscle. It is unclear whether 1H-MRS is a valid and reliable method. 1H-MRS validity and reliability was examined in a series of in vitro and in vivo studies. In the in vitro study, phantoms containing different concentrations of carnosine, imidazole, histidine and bovine serum albumin (BSA) were submitted to 1H-MRS to verify: 1) signal linearity; 2) whether other sources of imidazole could contribute to carnosine signal. In the in vivo study, carnosine was determined in the m. gastrocnemius by 1H-MRS and by high-performance liquid chromatography (HPLC, a reference method) in muscle biopsy samples from 16 young men. Test-retest reliability was determined with (n=10) and without (n=5) voxel repositioning and re-shimming. Convergent validity (n=16) was determined by comparing carnosine values obtained with 1H-MRS vs. HPLC. Discriminant validity (n=14) was determined by measuring carnosine before and after 4 weeks of β-alanine supplementation. In vitro carnosine signal showed excellent linearity (Pearson correlation: r=0.999). Histidine and imidazole, but not BSA, emitted quantifiable signals in the same chemical shift of carnosine. A clear loss in signal quality was shown in the signal obtained in vivo. 1H-MRS coefficient of variation without repositioning voxel was 6.6% and increased to 16.9% with voxel repositioning. 1H-MRS was able to detect a significant increase in muscle carnosine after β-alanine supplementation, both a substantial disagreement with HPLC was shown. 1HMRS showed adequate discriminant validity, but limited reliability and poor agreement with the reference method.


Introduction
Carnosine is a multifunctional dipeptide abundantly found in human skeletal muscle cardiac muscle and in some neuronal cells . Carnosine has numerous properties that confers performance enhancing effects (Saunders et al., 2017), as well as a wide-range of potential therapeutic applications ). Such properties include hydrogen cation (H + ) buffering (Dolan et al. 2018), scavenging of reactive species (Carvalho et al. 2018), and protection against glycation end products (Ghodsi & Kheirouri,2018). Several studies have demonstrated the beneficial effects of A non-invasive alternative method based on proton magnetic resonance spectroscopy (1H-MRS) has been developed to quantify carnosine in human skeletal muscle (Ozdemir et al. 2017). 1H-MRS has been considered advantageous to assess muscle carnosine because it is non-invasive, virtually free of risk and suitable to be used in any population. In 1H-MRS, carnosine is quantified from two unique detectable signals emitted by the carbon four (C4-H) and the carbon two (C2-H) of the imidazole ring, which resonate at seven and eight ppm of the magnetic resonance spectrum (Ozdemir et al. 2007).
Although 1H-MRS has been used to quantify carnosine in numerous investigations , there has not been any comprehensive investigation of the validity of 1H-MRS for muscle carnosine assessment against a reference method, such as the chromatographic determination in muscle biopsy samples.
Carnosine quantification by 1H-MRS has several limitations that warrant a thorough experimental examination. Firstly, in vivo carnosine signals are broad, of small amplitude, often close to the noise level, and tend to suffer dipolar coupling, in particular the signal emitted by C4-H. This makes C4-H quantification unfeasible in most cases (Boesch & Kreis, 2001). Also, the in vivo spectrum is crowded with metabolite peaks, thereby making carnosine identification particularly challenging, even when prior knowledge-based approaches are used (Kreis, 1997;Tkac et al, 2002) and, as such, carnosine quantification appears to be more difficult than other abundant muscle metabolites, such as creatine, taurine and lactate (Just Kukurová et al, 2016). Secondly, the signals emitted by the imidazole ring could, in theory, also be detected in other imidazole-containing molecules, such as free imidazole, free histidine, carnosine analogues and histidine residues in proteins. In fact, previous investigations have reported problems in differentiating signals from carnosine and its analogue homocarnosine in human brain (Solis et al. 2015 To address the potential limitations of 1H-MRS to quantify carnosine in human skeletal muscle, the present investigation examined the reliability, accuracy and sensitivity of 1H-MRS for the determination of muscle carnosine in humans using in vitro and in vivo experiments. Carnosine determination by high-performance liquid chromatography (HPLC) in extracts from human muscle biopsy samples was used as the reference method.

Experimental Design
The study was approved by the institution's Ethics Committee and conformed to the 2013 version of the Declaration of Helsinki. This study comprised two investigations.
In the first investigation, we performed a series of in vitro 1H-MRS acquisitions in phantoms aiming to 1) determine the linearity of the carnosine signal, 2) examine the influence of the presence of the imidazole ring in other compounds (i.e., in free histidine and in histidine residues in protein) to the carnosine signal, thereby gathering knowledge on the contribution of other sources of imidazole ring to the signal obtained in vivo, and 3) compare the signal quality obtained in vitro vs. in vivo.
The second investigation aimed to evaluate the test-retest reliability, as well as the discriminant and convergent validity of the 1H-MRS technique to measure muscle carnosine. HPLC quantification in muscle extracts was chosen as the reference method.
To account for the major sources of error in both methods, test-retest reliability was assessed in two different conditions for 1H-MRS (i.e., with and without removing the participant from the scanner, repositioning and re-shimming the voxel) and in three different conditions for HPLC (i.e., same extract from the same sample analysed on two separated runs, different extracts from the same sample analysed on two separated runs and different extracts from two samples analysed on two separated runs). To assess discriminant validity, muscle carnosine was determined in a group of participants before (PRE) and after (POST) β-alanine supplementation. This intervention was intentionally chosen due to its highly consistent effects on muscle carnosine (Harris et al. 2006;Hill et al. 2007;Carvalho et al. 2018), thereby allowing the assessment of whether 1H-MRS is able to discriminate two knowingly different carnosine concentrations. To assess convergent validity, muscle samples were obtained from the same group of participants immediately after 1H-MRS, both before and after β-alanine supplementation, so that the results obtained with 1H-MRS could be compared with those obtained with HPLC.
Muscle carnosine concentrations obtained with 1H-MRS were converted to the same unit of muscle carnosine content (i.e., from mmol·L -1 to mmol·kg -1 of dry tissue) for a clearer comparison between methods. Both 1H-MRS and HPLC techniques were performed by well-trained researchers, with large experience in carnosine determination. The 1H-MRS in vivo and muscle biopsy assessments were individually standardized so that each participant performed their PRE and POST-sessions at the same time of day. Participants were requested to abstain from alcohol and unaccustomed exercise in the 48 hours prior to the experimental sessions. Participants were instructed to arrive at the laboratory at least 2 hours following their last meal. Ad libitum water consumption was allowed before and after the sessions. containing BSA (the equivalent to 12.5 mmol·L -1 of imidazole) was prepared to assess whether imidazole-residues in large size molecules (e.g., protein) could emit a signal at the same chemical shift (7 and 8 ppm). All phantoms were solidified by melting agarose 2% w:v in autoclaved ultra-pure water prior to adding carnosine, imidazole, histidine or BSA. Phantom concentrations were calculate based on the imidazole content so that all concentrations were equimolar to 12.5 mmol·L -1 of imidazole. The 12.5 mmol·L -1 concentration for BSA was chosen because this is nearly the maximum achievable within the solubility of BSA and it represents a mid-range physiological concentration of carnosine in human skeletal muscle.

In vitro investigation
For the present investigation, a 3 Tesla whole-body magnetic resonance scanner (Achieva, Philips, Best, The Netherlands) equipped with an 8-channel knee coil was used.

In vivo investigation
Sixteen young, healthy, physically active men volunteered to participate, two of whom could not complete the entire study due to personal reasons. Therefore, 14 participants (age:27±5 years; body mass: 82.9±11.8 kg; stature: 1.77±0.06 m; body mass index: 26.3±2.4 kg·m 2 ) completed all tests. Participants were fully informed of possible risks and discomforts associated with participation before providing informed consent.
They were requested to maintain similar levels of physical activity and dietary patterns for the duration of the study. Exclusion criteria were: i) use of supplements containing creatine or β-alanine in the 3 months and 6 months prior to the study; ii) use of anabolic steroids; iii) chronic use of glucocorticoids; iv) chronic-degenerative disease and/or condition that affected the locomotor apparatus, and v) any condition that would prevent them from undertaking the proposed tests (e.g., metallic prostheses that could interfere with 1H-MRS quality).
Participants were assessed for muscle carnosine before and after a 4-week period of β-alanine supplementation. β-alanine was provided in 800-mg tablets (CarnoSyn™, NAI, USA) and the participants were asked to take 2 tablets along with meals, four times per day, totalling 6.4g·d -1 of β-alanine. All 16 participants completed a baseline assessment for carnosine quantification in the medial portion of m. gastrocnemius of the right leg using both 1H-MRS and HPLC. Carnosine quantification via 1H-MRS was not possible in one participant due to a peak of very small amplitude with baseline below zero. Therefore, the analysis of convergent validity was conducted on 15 participants. To minimise differences between methods owing to variations in sampling sites, biopsy sites were intentionally taken from the closest possible sites to those where the spectra were obtained. This was ensured with the physician examining the image of the voxel position before defining the location and the depth the biopsy needle would be inserted. Following supplementation, the 14 participants who completed the entire study were again assessed for muscle carnosine using both 1H-MRS and HPLC. The responses to supplementation were used to compare the discriminant validity between methods.
A sub-sample of 10 participants volunteered for the test-retest reliability assessment of 1H-MRS with voxel repositioning and re-shimming. They undertook the first 1H-MRS, left the room, waited for 5 minutes and then were repositioned back on the machine for the second 1H-MRS. The voxel was repositioned as closely as possible to the site where it was positioned in first test; this was achieved using an image of the voxel position obtained in the first 1H-MRS as a guide. Another sub-sample of 5 participants volunteered for the test-retest reliability assessment of 1H-MRS without voxel repositioning and re-shimming. They undertook the first 1H-MRS and stood still for the second 1H-MRS, which was performed immediately after the first.
To assess inter-assay reliability of HPLC determination of muscle carnosine, muscle extracts obtained from 15 biopsy samples randomly chosen from a collection of containing 144 muscle samples, using an online random number generator. These were analysed in duplicate in two independent runs performed on different days. To assess "inter-extract" reliability of HPLC, two different muscle extracts obtained from 11 biopsy samples randomly chosen from this same collection were analysed in duplicates on different days. To assess "inter-biopsy" reliability of HPLC, two consecutive muscle samples were obtained from the same incision in a sub-sample of 7 participants who volunteered for this study. The second biopsy location was changed by rotating the needle's window guillotine in 90 degrees, thereby sampling the collateral site of the first biopsy.

1H-MRS in vivo assessment
Spectra were acquired using single voxel point-resolved spectroscopy (PRESS) localisation with the following parameters: TR/TE=3000/30 ms, voxel size=10×10×30 mm 3 , number of averages (NEX)=256, 2048 data points with a spectral width of 2000 Hz. The total acquisition time of the 1H-MRS was 13.9 min. The 50 mmol·L -1 carnosine phantom was used as an external reference. To that end, an acquisition was made using the same parameters as those used in vivo, except for the TR, which was 12000 ms. In each in vivo measurement, the right leg of each participant was positioned and was firmly immobilised in the knee coil, such that the gastrocnemius muscle was in the centre of the coil. The left leg was supported outside the coil to improve comfort and thus minimise leg movement. Voxel location was standardised on the larger calf region in the centre of the medial portion of the gastrocnemius muscle of the right leg. The same well-trained and experienced biomedical technician was responsible for placing the voxel in all conditions. After placing the voxel, a set of images depicting individual voxel location was saved and used to guide positioning in all further exams of that individual.

Quantification
Absolute quantification of the carnosine resonance was determined using the Pt is the temperature correction factor applied as the signal decreases by 6% between the room temperature (i.e., phantom temperature) and body temperature (Davies, 2003). For the in vitro signal, it is not necessary to correct the T1 relaxation, since the acquisition was performed with a sufficiently long TR (TR=12000 ms) to neglect this factor. Signals were also corrected by water content; since the water content in phantoms is ~100%, a correction factor=1 was used. For the in vivo analyses, a correction factor=0.66 was used, assuming that ~2/3 of the muscle is water (Schoeller, 1989).
The relaxation correction factors were calculated using the following equations:

Muscle biopsies
Muscle samples (~70-100 mg) were obtained under local anaesthesia (3 mL, 1% lidocaine) from the mid-portion of the m. gastrocnemius using the percutaneous needle biopsy technique with suction (Bergstrom, 1962). Samples were obtained from the same leg for all experiments. PRE and POST supplementation biopsies were taken from incisions made as close as possible to one another. Samples taken for the inter-biopsy reliability analyses were obtained from the same incision, but from slightly different sites, as described above. All samples were snap frozen in liquid nitrogen and were subsequently stored at -80°C until analyses. Samples were freeze-dried and dissected free of any visible blood, fat and connective tissue before being powdered and further submitted to HPLC determination of carnosine.

Chromatographic determination of histidine-containing dipeptides in whole muscle
Deproteinised muscle extracts were obtained from 3-5 mg freeze-dried samples according to the protocol described by Harris

Reliability of HPLC for carnosine quantification in human skeletal muscle
Inter-assay reliability showed very similar values for both test and retest (mean ± 1SD difference=0.6±4.0%). No statistically significant differences between measures were shown (t=0.144; p=0.887), suggesting that HPLC is free of systematic errors when the same muscle extracts are analysed. A high ICC and low CV were seen between test and retest values (ICC=0.996, 95%CI=0.987-0.999; CV=2.72%) (Figure 4, panel A).
Thirteen of the 15 samples had less than 5% variation between measurements, with the remaining two samples having less than 10% variation.
Eight of the 11 samples were below 5% of variation between measurements, with the remaining 3 samples being below 10% variation.
Inter-biopsy reliability analysis also showed very similar measurements (mean ± 1SD difference =1.1±6.0%). No statistically significant differences were shown between measures (t=-0.588; p=0.578). A high ICC and low CV were seen between test and retest values (ICC=0.957, 95%CI=0.750-0.993; CV=3.95%) (Figure 4, panel C). Five of the seven samples were below 5% of variation between measurements with the remaining 2 samples being either below or at 10% of variation.

Reliability of 1H-MRS determination of muscle carnosine
The mean carnosine values obtained via 1H-MRS for test and retest without voxel repositioning showed similar measurements (mean±1SD difference=5.0±6.7%) ( Figure   5, panel A). No statistically significant differences were shown (t=-1.0; p=0.37), indicating that 1H-MRS is free of systematic errors. Intraclass coefficient correlation was 0.924 (95%CI=0.451-0.992) and the CV was 6.6%. The variation between tests was below 5% in 4 out of the 5 participants.
The variation between tests was below 5% in only one participant, between 5-10% in only two participants, and above 10% in the remaining 7 participants ( Figure 5, panel B).

Convergent validity of 1H-MRS vs. HPLC
Both methods can detect the increase in muscle carnosine in response to β-alanine supplementation (both p<0.05; figure 6, panel A). Although no statistically significant differences for the POST-PRE deltas were shown between methods (neither for the absolute, nor for the relative delta - Figure 6, panel B), a large disagreement was shown between methods regarding the ability to detect changes in muscle carnosine in response to supplementation (Figure 6, panels C and D).
When both PRE-and POST-supplementation measures were pooled, the Bland-Altman plot showed a visible disagreement between HPLC and 1H-MRS, which increased when carnosine values were <10 mmol·kg -1 dm ( Figure 6, panel E). Only two of the 27 measures were below 5% difference between techniques; 17 out of 27 measures were above 20% difference, and 7 measures were above 50% ( Figure 6, panel F). Absolute (left chart) and relative (right chart) post-pre delta values for muscle carnosine measured using both methods in response to β-alanine supplementation. Panel C: Absolute post-pre delta values obtained using both methods plotted against the identity line (i.e., representing 100% agreement between methods). Panel D: Relative post-pre delta values obtained using both methods plotted against the identity line (i.e., representing 100% agreement between methods). Panel E: Bland-Altman plot for percent differences between methods. Panel F: pooled pre and post data for muscle carnosine values obtained using both methods plotted against the identity line (i.e., representing 100% agreement between methods).

Discussion
In light of the growing attention that muscle carnosine has been receiving due to its potential ergogenic and therapeutic properties, quantifying this dipeptide in muscle tissue is becoming an increasingly necessary procedure. As such, the use of an accurate, reliable and sensitive method for carnosine quantification is of the utmost importance.
Although 1H-MRS has emerged as a non-invasive alternative for analytical methods that require muscle biopsies, no study to date has examined its validity against a wellestablished reference method. In the present study, we performed a series of in vitro and in vivo experiments to examine several aspects of 1H-MRS validity (i.e., signal linearity, matrix effect, reliability, discriminant and convergent validity). In order to be certain that HPLC is a reliable method for muscle carnosine determination and could be used as the reference in this study, we also conducted a thorough reliability examination of the its reliability, which showed excellent repeatability in all instances (i.e., inter-assay, interextract, and inter-biopsy).
The present investigation revealed important methodological issues that must be resulting in lower signal-to-noise ratio (Hoult & Richards, 1969). In individuals with low muscle carnosine content, increased error is to be expected, since peak amplitude is naturally lower and, therefore, very close to the basal noise. This is supported by the increased disagreement between 1H-MRS and the reference method shown in the Bland-Altman plot when carnosine concentrations are near to the lowest range. One could suggest to measure, as an alternative, the carnosine subpeaks. However, identifying all subpeaks in the spectrum may not be possible, since they might be totally covered by noise, especially in volunteers who present low muscle carnosine levels.
To investigate the potential impact of the imidazole ring present in other molecules (e.g., free imidazole, free histidine, carnosine analogues and histidine residues in proteins) on the carnosine signal detected by 1H-MRS, the in vitro signal was compared between carnosine, imidazole, histidine, and BSA. Although the best signal quality was obtained with carnosine, quantifiable signals were also obtained with imidazole and free histidine. This indicates that small imidazole-containing molecules might constitute a potential source of error, although they are likely of low relevance for the skeletal muscle since they are expressed in very low concentrations in comparison with carnosine (Parkhouse et al. 1985). Conversely, no signal was obtained with BSA, probably due to its large size (~66 kDa). Increasing molecular size leads to slower tumbling and correspondingly shorter spin-spin relaxation times (T2), resulting in to a more complex spectrum with very broad peaks of low amplitude that do not surpass noise level.
Accordingly, 1H-MRS experiments become unreliable at room temperature for proteins larger than 30 kDa and largely fail for proteins above 35 kDa in the absence of elevated temperature (Wand et al. 1998 representation of studies assessing muscle carnosine before and after an intervention. Such an increase in variation indicates that voxel positioning and shimming are major sources of random error in 1H-MRS. In the present study, all possible measures were taken to ensure that voxel would be positioned in the same location (i.e., the same experienced technician was responsible for voxel positioning in all exams; voxel positioning at "retest" was done with the image depicting voxel position at "test" as a guide  Figure 6). The disagreement seemed to increase when participants showed smaller carnosine content, possibly due to 1H-MRS signal characteristics. As A potential limitation of this study is that the results were obtained from the human calf muscle (i.e., gastrocnemius medialis), known to have a mixed proportion of type I and II fibers; hence, it remains to be elucidated if 1H-MRS shows better measurement performance for carnosine quantification in a more homogeneous muscle tissue. In addition, considering that physically active participants were recruited for the present study and that local adipose tissue appeared to be an important factor interfering with 1H-MRS signal, we cannot rule out the possibility that 1H-MRS is more reliable in individuals with low body fat, such as athletes.

Conclusion
1H-MRS is capable of measuring carnosine in muscle tissue and is sensitive to detect overall changes in muscle carnosine brought about by β-alanine supplementation.
However, 1H-MRS has a high-degree of variation due to random error associated with voxel positioning, and poor convergent validity. This makes quantification problematic, particularly in regions surrounded by fat tissue, in individuals with high levels of body fat, and in individuals with low muscle carnosine levels. Caution should be exercised when interpreting muscle carnosine quantification data obtained with 1H-MRS.