Pyruvate to lactate metabolic changes during neurodevelopment measured dynamically using hyperpolarized ¹³C imaging in juvenile murine brain

Abstract

Hyperpolarized ¹³C MRI has recently been used to dynamically image metabolism in vivo. This technique provides the capability to investigate metabolic changes in mouse brain development over multiple time points. In this study, we used ¹³C MR spectroscopic imaging and hyperpolarized ¹³C-1 labeled pyruvate to analyze its conversion into lactate. We also applied T2-weighted anatomical imaging to examine brain volume changes starting from postnatal day 18. We combined these results with body weight measurements for a comprehensive interpretation of mouse brain maturation. Both the produced lactate level and pyruvate to lactate conversion rate decreased with increasing age in a linear manner. Total brain volume remained the same after P18, even though body weight continued to grow exponentially. Our results have shown that the rate of metabolism of ¹³C-1 pyruvate to lactate in brain is high in the young mouse and decreases with age. The brain at P18 is still relatively immature and continues to develop even as the total brain volume remains the same.

Introduction

Understanding normal brain maturation is essential to establish a baseline for the assessment of developmental diseases that involve the brain. The biomarkers of normal brain development include global and local volume changes, increase in cortical complexity, microstructural changes resulting from white matter growth and myelination, and variations in metabolic rate and metabolite production. Magnetic Resonance Imaging (MRI) provides a non-invasive method to study in vivo brain development and subsequently enables quantification of the aforementioned biomarkers.

Most of MRI studies of normal brain maturation use T1- and T2-weighted and diffusion-weighted imaging (DTI) to assess morphological and microstructural changes, and subsequently, their inference on anatomical growth and functional segregation [1–3]. ¹H Magnetic Resonance Spectroscopy (MRS) offers the ability to examine metabolic properties during brain maturation [4–6] but, until most recently, it has only provided steady state information of metabolites due to low metabolic concentrations and the long acquisition time needed to obtain diagnostic information. ¹³C spectroscopy provides another perspective, allowing investigation of metabolic exchanges and labeled neurotransmitters such as glutamate, GABA and aspartate in the developing murine brain [7–9]. Sensitivity of ¹³C had been low because its natural abundance is only about 1% in living organisms, and is, therefore, hard to detect within a limited scan time. Thus, most of the ¹³C MRI studies require a long scan time, which can only be achieved in cultured cells and extracted tissues [10–13], hampering the feasibility of longitudinal scanning to study maturation. Some of the in vivo studies require constant infusion of metabolites for minutes or hours [14, 15], which is impossible for young mice with small body volumes. Moreover, long scan time hinders the capability of measuring dynamic changes of metabolites.

In 2003, Golman et al [16] introduced a new method to dynamically quantify metabolism using hyperpolarized ¹³C, which enabled an entirely different avenue to study metabolism in vivo, in real time. Dynamic nuclear polarization (DNP) has been proven to increase ¹³C NMR signal more than 10,000 fold, allowing investigation of ¹³C metabolic exchanges in vivo [17]. DNP depends on nuclear spin polarization in the solid state. It requires a free radical, which is an unpaired electron, to transfer its electron spin polarization to the nuclear spins by microwave irradiation at low temperature. Once a hyperpolarized substrate is injected into a living mammal, we are able to collect real-time ¹³C metabolic imaging to study metabolism by following the substrate as it participates in a biochemical process. However, the hyperpolarization of the nuclear spins return to thermal equilibrium after a very short time at room temperature. Therefore, a metabolite must be carefully chosen in order to visualize the metabolic pathway within this time period. Hyperpolarized ¹³C-1 labeled pyruvate has been shown to demonstrate in vivo metabolism in murine [18], non-human primate [19], and human [20] due to its rapid delivery and uptake by the cells in various organs, and its quick conversion into lactate, alanine and/or carbon dioxide due to various metabolic pathways.

In this study, we applied DNP using ¹³C-1 labeled pyruvate to investigate the changes in pyruvate to lactate conversion over the course of mouse brain maturation, and combined this with T2-weighted MRI to obtain voxel-based volume changes in relation to maturation in mouse brain.

Methods

Experimental procedure

8 normal mice were scanned starting on postnatal day 18 and repeated every 10 days. A few time points were delayed due to technical issues. Mice were anesthetized with 1.5% isoflurane and 1 L/min oxygen during scans. A catheter was inserted into the tail vein for proceeding injection of hyperpolarized pyruvate. During the experiment, 0.3mL of mixed saline and heparin was injected into the mouse every 15 minutes to prevent blood clot. All experiments were conducted on a vertical 14.1T (Agilent) 600WB NMR spectrometer with 55mm 1000mT/m gradients. A 38mm diameter ¹H and ¹³C dual-tuned coil was used, where ¹H frequency was for main field shimming and T2-weighted anatomical imaging, and ¹³C coil was used for hyperpolarized ¹³C spectroscopic imaging.

48μL of C1 labeled ¹³C pyruvate was polarized using an Oxford Hypersense^TM DNP instrument at 3.5T under 1.5K for an hour. Before injecting into the mouse, the hyperpolarized ¹³C-1 pyruvate was mixed with 4.5mL of NaOH buffer, which resulted a 160mM pyruvate solution with pH~7.5. The dissolution mixture was then injected into the tail vein through a catheter over a span of 12 seconds. The total volume injected was 300 μL with 150 μL into the mouse and 150 μL left in the catheter.

Data acquisition

¹³C spectroscopic imaging

Data were acquired on a 24mm×24mm×5mm slice centered on the brain, with 2D chemical-shift imaging. Center-out k-space trajectory was used with 7x7 phase encoding (zero-filled to 8x8). 128 spectral points were acquired with 2500 Hz bandwidth. The acquisition was started simultaneously with the pyruvate injection and repeated every 4s (3s acquisition time with 1s delay between each repetition) for a total of 60s (or 15 repetitions) with constant flip angle of 10°.

T2-weighted anatomical imaging

A 2D fast spin echo (FSE) sequence was applied for anatomical imaging with TR/TE=1.3s/12ms, and 8 echo train length with 12 ms echo spacing. The field of view was 30mmx30mm with 256 frequency and phase encodes, resulting in a 0.12mmx0.12mm in-plane resolution. The slice thickness was 1mm, and 10 slices were acquired without gap, covering most of the brain. The total scan time was 11 minutes.

Data processing

¹³C spectroscopic imaging

A 5Hz Lorentzian apodization was applied to each free-induction decay before Fourier transforming the data. After Fourier transformation, each voxel could be referred to a 15-time-point spectrum (Figure 1a). The peak height of pyruvate and lactate at each time point for each voxel were measured. Then we chose the six voxels matching the location from the T2-weighted anatomical image (Figure 1b), and averaged signal intensities among the six voxels for pyruvate and lactate separately. The pyruvate signal intensity depended on the polarization level, which varied between experiments, and consequently, the lactate signal intensity was affected, due to a linear relationship between pyruvate and lactate [21]. In order to compare lactate signal at different ages, we normalized the data by dividing lactate signal by the total carbon signal in the slab at each time point. We recorded the 5 highest intensities of normalized lactate in each scan for every mouse.

Figure 1.

The 15-time-point dynamic spectrum (a) is depicted for a voxel. The six voxels matched with the brain (b) were chosen for ¹³C analysis.

T2-weighted anatomical imaging

To investigate brain volume changes through maturation, we used linear co-registration between the brains of two consecutive scans (i.e., 1^st to 2^nd, 2^nd to 3^rd, and so on) for each individual mouse (Figure 2). To achieve this, we first manually masked the brain mask of the 1^st scan from non-brain tissues and parts including other organs of the head, the feet and water tubes running under the mouse to keep its body temperature. Then we linearly registered the 2^nd scan to the 1^st brain mask. This process subsequently created a brain mask of the 2^nd scan by transforming the mask of the 1^st scan. In the same manner, we registered the 3^rd scan to the 2^nd scan to create the brain mask of the 3^rd scan, and so on. Multiplication of the three scaling parameters (in x, y, and z-axes) in the transformation matrix measured brain volume changes.

Figure 2.

Linear registration scheme between subsequent scans within each subject. The multiplication of the 3 scaling parameters T₁, T₂ and T₃ from the resultant transformation matrix was used for volume change calculation.

Kinetic modeling

The hyperpolarized ¹³C-pyruvate was converted into lactate rapidly in the brain, allowing us to investigate real-time change of this metabolic process. To quantify the rate of this change, we applied the kinetic model from Zierhut et al [22] to fit the pyruvate and lactate signal over time. We averaged the peak height of pyruvate and lactate separately from the six voxels in the brain for each time point, and then we could obtain a curve of signal build up and decay for pyruvate and lactate individually (Figure 3).

Figure 3.

Dynamic data of the averaged peak height of pyruvate (x) and lactate (*) from six voxels in the brain. Kinetic models were fitted to pyruvate (—) and lactate (—).

The model was first fit into the pyruvate curve:

$$M_{\mathit{pyr}}(t)=\{\begin{array}{ll}\frac{{\mathit{rate}}_{\mathit{inj}}}{k_{\mathit{pyr}}}(1-e^{-k_{\mathit{pyr}}(t-t_{\mathit{arrival}})}),\hfill & \hfill t_{\mathit{arrival}}\le t<t_{\mathit{end}}\\ M_{\mathit{pyr}}(t_{\mathit{end}})e^{-k_{\mathit{pyr}}(t-t_{\mathit{end}})},\hfill & \hfill t\ge t_{\mathit{end}}\end{array}$$

where M_pyr (t) is the pyruvate signal at time t, and t_end is the time when pyruvate signal reaches the highest value. The parameters estimated from the fitting are: rate_inj, the of pyruvate injection; k_pyr, the rate of pyruvate signal decay, which is $\frac{1}{T_{1\mathit{pyr}}}$ , where T_1pyr is the T1 constant of hyperpolarized pyruvate at 14T field strength; and t_arrival, the time when pyruvate arrives in the brain. Then we used these estimated parameters to fit into the lactate curve:

$$M_{\mathit{lac}}(t)=\{\begin{array}{ll}\frac{k_{pl}{\mathit{rate}}_{\mathit{inj}}}{k_{\mathit{pyr}}-k_{\mathit{lac}}}\left(\frac{1-e^{-k_{\mathit{lac}}(t-t_{\mathit{arrival}})}}{k_{\mathit{lac}}}\right)-\frac{1-e^{-k_{\mathit{pyr}}(t-t_{\mathit{arrival}})}}{k_{\mathit{pyr}}},\hfill & \hfill t_{\mathit{arrival}}\le t<t_{\mathit{end}}\\ \frac{M_{\mathit{pyr}}(t_{\mathit{end}})k_{pl}}{k_{\mathit{pyr}}-k_{\mathit{lac}}}(e^{-k_{\mathit{lac}}(t-t_{\mathit{end}})}-e^{-k_{\mathit{pyr}}(t-t_{\mathit{end}})})+M_{\mathit{lac}}(t_{\mathit{end}})e^{-k_{\mathit{lac}}(t-t_{\mathit{end}})},\hfill & \hfill t\ge t_{\mathit{end}}\end{array}$$

where M_lac(t) is the lactate signal at time t. We could estimate k_lac, the rate of lactate decay, and k_pl, the conversion rate from pyruvate to lactate from this fitted model. The conversion rate, k_pl, is another parameter besides normalized lactate level to imply metabolic rate in maturation.

Statistical analysis

To examine the association between age and the lactate level, pyruvate to lactate conversion rate or brain volume, we used a mixed effect linear model on the 5 highest intensities of normalized lactate, the pyruvate to lactate conversion rate, and the volume scaling separately using SurfStat (http://www.math.mcgill.ca/keith/surfstat/) [23]. This model permitted multiple measurements per subject while controlling for between-subject variation, thus increasing statistical power. We tested linear and nonlinear polynomial fitting.

In addition, we recorded body weight for each experiment, because body weight can also be an empirical indicator of maturation besides brain volume. The body weight was measured before experiment since the injection of hyperpolarized pyruvate would change the total body weight.

Results

By plotting the normalized lactate level versus age for each individual mouse, we observed significant decrease in normalized lactate signal with increasing age using a mixed effect model on the normalized lactate intensities (t=−3.84; p=0.001) (Figure 4).

Figure 4.

Linear fitting (—) of the mixed effect model to the peak normalized lactate level against age.

We plotted the estimated pyruvate to lactate conversion rate, k_pl, against age for each individual mouse (Figure 5). Linear fitting resulted in the highest significance when we evaluated the mixed effect model on k_pl. Similar to lactate level, k_pl also decreased with increasing age (t=−2.11; p=0.044).

Figure 5.

Linear fitting (—) of the mixed effect model to the pyruvate to lactate conversion rate, K_pl^, against age.

Analysis of linear regression on brain volume changes (i.e., multiplication of 3D scaling factors) showed no relationship between volume and age (t=0.82; p>0.4). However, total body weight increased with age (Figure 6) even though the global brain volume did not change after P18. Body weight almost doubled from P18 to P30, and the rate of weight growth slowed down after P30 (Figure 6). It follows an exponential growth (t=5.07; p=0.000025).

Figure 6.

The plot of body weight against age for each individual subject, with general-exponential fitting ( ).

Discussion

Our ¹³C metabolic imaging results demonstrated that pyruvate to lactate conversion was higher at younger age, and decreased linearly with increasing age. At a young age, brain and body consume more energy for brain development and body growth. Lactate can be used as an energy fuel for brain activation [24–32], and thus, its production is higher at younger age to support neuronal and other functional needs. It has been shown in cultured cells that both neurons and astrocytes utilize lactate as a source of energy and a precursor of lipids [33] and these results further were supported by ¹³C MR studies [11–13]. Moreover, the utilization of lactate by oligodendrocytes is three times higher than that by neurons and astrocytes [34]. The combination of these published findings suggests that lactate is greatly used for lipogenesis during myelin synthesis and may support a further explanation that long-lasting microstructural changes can occur in relation to brain maturation even at the stage where macroscopic volume growth stops, which is demonstrated in our results.

In this study, the analysis of total body weight showed that the mouse continued to grow after P18 and the rate of growth slowed down after P30. Our volumetric analysis based on T2-weighted images, on the other hand, did not show a significant change of overall brain volume after P18. This is consistent with findings of a previous high-resolution ex vivo T2-weighted MRI study that global and local volumes of mouse brain remained constant in size after P18 [3]. This may imply that the mouse brain reaches its maximum volume earlier than the whole body does. Another study has, however, shown that the lipid composition of myelin and microsomal fractions of mouse brain appeared to greatly increase from P14 to P24 and continued growing until P47 [35]. Protein composition of mouse brain myelin also continues to increase until P300, with the highest rate of increase between P20–P25 and decreasing rate of increase afterwards [36]. Although there is cross-subject variation in determining maturation stage [37], these published findings together with our result suggest that the microstructural growth of the mouse brain may continue even after adolescence although its total volume did not change.

In addition to total lactate level, pyruvate to lactate conversion rate is another parameter to measure the degree of metabolic change during brain development. The higher the conversion rate, the faster the metabolism takes place, and consequently, more lactate is produced within a limited amount of time. The maximum reaction rate of pyruvate, V_max, is a more accurate and commonly recognized parameter to measure the rate of metabolism than the pyruvate to lactate conversion rate. If we can vary the concentration of pyruvate, we will get different k_pl from different pyruvate concentration. Based on the different conversion rates and pyruvate concentrations, we can apply Michaelis-Menten kinetics model to estimate the maximum reaction rate of pyruvate, $V_{\mathit{max}}:K_{pl}=\frac{V_{\mathit{max}}[\mathit{pyr}]}{K_m+[\mathit{pyr}]}$, where [pyr] is the concentration of pyruvate and K_m is the Michaelis constant which refers to the concentration of pyruvate at half V_max. However, the main obstacle of achieving this is that the mouse is too small to receive multiple shots of pyruvate within one experiment at young age. This can be improved by reducing the dead volume of saline and heparin in the catheter. We are working on a technical solution to divert the dead volume before injection through a custom-built cog system for this purpose. With minimal dead volume, we will be able to perform multiple experiments at a younger age when brain volume is still growing and access the correlation between metabolic rate and brain growth.

A notable strength of the current study lies in the longitudinal in vivo MRI imaging, yielding multiple measurements per individual. Our analysis was based on the global measurement of brain metabolism and volume. Although regional metabolism may be important to understand pattern of brain development, low resolution in the hyperpolarized ¹³C acquisitions is a major technical challenge, limiting the ability for regional analyses. The hyperpolarized substrate returns to polarization at thermal equilibrium very quickly after taken out of the polarizer. Moreover, T1 of hyperpolarized substrate is shorter at the higher field strength of our 14T animal scanner as compared to lower fields such as 1.5T or 3T. Nevertheless, we achieved a very important advance in obtaining a dynamic spectroscopic image, where the acquisition was repeated every 4 seconds for a total of 60 seconds. The image spatial resolution of 3mm x 3mm was the highest we could achieve within 4 seconds by current technique. There is a tradeoff between temporal resolution and spatial resolution. With future improvement in the MRI pulse sequence, we will be able to reach comparable ¹³C spectroscopic image resolution with anatomical image resolution, enabling the analysis of brain local changes in metabolism and volume from a young age to P18 and later ages.

Our study demonstrates the effectiveness of hyperpolarized ¹³C MRI to examine the dynamic metabolic changes of pyruvate to lactate conversion during brain maturation, suggesting other substrates can be used in the same manner to explore dynamic changes of other metabolic pathways. In our kinetic modeling, we neglected the oxidation from lactate to pyruvate by assuming its rate was negligible. Sometimes it is important to consider the lactate to pyruvate conversion rate, especially at very young age or with brain abnormalities such as hypoxia or hypoglycemia, in which lactate is utilized as substitute for glucose as energy source [38]. In a future study, we plan to evaluate this by injecting hyperpolarized lactate into the mouse to estimate its conversion rate to pyruvate [39]. This can also be measured by using Metabolic Activity Decomposition Stimulated-echo Acquisition Mode (MAD-STEAM), which results a more robust and accurate modeling fitting with both conversion rates of pyruvate to lactate and lactate to pyruvate simultaneously [40, 41].

Overall, our longitudinal study using hyperpolarized ¹³C MR spectroscopic imaging is the first to show in vivo dynamic changes of pyruvate to lactate conversion in the brain during normal brain maturation. The results from normal brain studies can serve as a baseline for future metabolic studies of injured developing brain such as hypoxic ischemic encephalopathy.