Association between cold-inducible RNA binding motif 3 and hypothermia effect in murine hypoxia-ischemia model measured by metabolic MRI

Abstract

Introduction:

The efficacy of therapeutic hypothermia (TH) for neonatal hypoxic-ischemic encephalopathy (HIE) is inconsistent, and the cause remains unclear. This study aimed to explore the role of cold stress protein in the TH-induced neuroprotection following hypoxia-ischemia (HI) using metabolic MRI.

Methods:

Postnatal day 10 (P10) mice underwent unilateral HI followed by treatments with therapeutic hypothermia (TH) or normothermia (NT). HI and sham mice were scanned at 4 h and 22 h following TH after injection of hyperpolarized ¹³C-1 labeled pyruvate. The dynamic HP-¹³C MRSIs were acquired to examine the cerebral metabolic profile, i.e., the conversion rate from pyruvate to lactate (k_PL) and the ratio of lactate to pyruvate (Lac/Pyr) in the injured hemisphere. T₂-weighted images (T₂WI) and diffusion MR images (dMRIs) were acquired to identify the anatomical structures and assess the injury. Mice brains were collected during and at 0 h, 4 h, 12 h, 18 h and 22 h after treatments for western blot to investigate the time course of the levels of the cold stress protein (RNA binding motif 3, RBM3) and cell death markers (spectrin 145/150 and spectrin 120) changes. The cerebral metabolic profile, RBM3 and spectrin levels, and injury size were compared across groups and between specific timepoints. The relationship between the cerebral metabolic profile and RBM3 levels in HI+TH group was also evaluated.

Results:

We observed the upregulation of RBM3 during TH at 4 h and 22 h after TH. The spectrin 145/150 and spectrin 120 were unchanged over time in HI+TH group, whereas they significantly increased at 18 h and 22 h in HI+NT group. Additionally, the injury size was noticeably larger at 22 h in HI+NT group. Lower k_PL and Lac/Pyr were observed at 4 h and 22 h after TH, with a negative correlation to RBM3 levels in HI+TH group.

Conclusion:

This study demonstrates that RBM3 may be one of the key factors associated with TH-induced neuroprotection by reducing the anaerobic glycolysis process in HI mice, suggesting RBM3 upregulation may enhance the efficacy of TH for neonatal HIE.

Graphical Abstract

Introduction

Neonatal hypoxic-ischemic encephalopathy (HIE) results from oxygen and blood (glucose) supply deprivation during the perinatal period, leading to a significant rate of mortality and neurological morbidity in children. With a high incidence of 1.5 to 2.5 per 1000 live births, neonatal HIE has become a worldwide health issue[1]. Therapeutic hypothermia (TH) applied within 6 hours (hrs) of life and continued for 72 hrs is currently the standard of care for moderate to severe HIE. However, the therapeutic efficacy of TH is inconsistent with an estimated 40–50% adverse outcomes after TH[2]. Therefore, it is important to understand the molecular mechanism of hypothermia-induced neuroprotection and to identify key factors for the success of hypothermia in order to improve the current therapy as well as to develop new intervention strategies.

Hyperpolarized carbon-13 MR spectroscopic imaging (HP-¹³C MRSI) is a non-invasive imaging technique that measures the real-time metabolic process in tissue, by introducing a ¹³C labeled substrate[3]. Our previous studies have demonstrated the ability of HP [1-¹³C]pyruvate MRSI to assess the cerebral metabolic profile, the conversion rate from pyruvate to lactate (k_PL) and the ratio of lactate to pyruvate (Lac/Pyr) during neonatal brain development in mice with hypoxia-ischemia (HI) and following TH. We found various temporal changes in the cerebral metabolic status between TH good-outcome group and TH poor-outcome group, which could help identify the therapeutic effect of TH [4].

Cold shock proteins (CSPs) is a small group of proteins that are expressed widely in the neonatal rodent brain and are synthesized in response to either hypothermia or other conditions of mild stress, i.e. hypoxia[5,6]. Among the CSPs, the RNA-binding protein (RBM3) is a glycine-rich protein that promotes cell survival by stabilizing specific mRNA and has recently gained increasing attentions[7–9]. In perinatal asphyxia models, RBM3 has been shown to be upregulated, providing protection against apoptotic neuronal death during therapeutic cooling and contributing to favorable neurodevelopmental outcomes [10,11]. We hypothesized that upregulation of RBM3 is associated with the TH-induced neuroprotective effect on HI by the suppression of cerebral metabolism, and the subjects identified as the TH-responders with HP [1-¹³C]pyruvate MRSI are expected to have a significantly robust expression of RBM3. In this study, we first investigated the time course of changs in RBM3 during and after TH. To further understand the early neuroprotective effect of TH, we measured the spectrin breakdown products (SBDPs), i.e., 145/150kD (indicative of necrotic and excitotoxic neuronal death) and 120kD (represents apoptotic death) correspondingly and evaluated the injury size after TH. Then we assessed the correlation between the cerebral metabolic profile (k_PL and Lac/Pyr) and levels of RBM3 at specific timepoints after TH in the HI mice model.

Methods

All procedures and MR scans were approved by the Institutional Animal Care and Use Committee at the University of California San Francisco, in accordance with the National Institutes of Health guidelines for the Care and Use of Laboratory Animals.

HI mice model establishment and intervention.

A total of 153 CD-1 mice at postnatal day 10 (P10) were included. Mice were anesthetized by 2–2.5% isoflurane during surgery. The HI group pups underwent electrical coagulation of the left common carotid artery (CCA) and then hypoxia (in a hypoxia atmosphere of 10% O₂ / 92% N₂ at 36.5°C) for 1 hr as previously described according to the Vannucci model[12]. The sham group pups had the left CCA exposed without coagulation and hypoxia. Pups were returned to the dam after HI insult. Then, pups were randomizely assigned to two treatment groups: HI+TH group and HI+NT group. Pups were treated with TH or normothermia (NT) by placing them in chambers submerged in a water bath at 30°C (HI+TH group) or 36.5°C (HI+NT) for 3.5 hrs. The body temperature range of pups was kept at 29.0–30.5°C during TH. After TH, the pups were allowed to rewarm to 35.5–36.5°C for 30 minutes (mins) before being returned to the dam. All pups were randomly allocated to two batches of experiments: (1) Western blot for examining the time course of RBM3 and spectrin levels during and after TH (experiment 1); (2) MRI scans+Western blot for investigating the anaerobic metabolic status at specific timepoint after TH and its relationship with RBM3 levels (experiment 2) (Figure 1)

Figure 1.

Experimental design. In experiment 1, Mice underwent hypoxia-ischemia (HI) condition or sham surgery and HI mice were then treated with therapeutic hypothermia (TH) or normothermia (NT) for 3.5 hrs, then mice brains were collected during treatment (−1 hr) and at 0 h, 4 h, 12 h, 18 h and 22 h after treatment for western blots to measure the expression levels of cold stress protein (RNA binding motif 3, RBM3) and cell death markers (spectrin 145/150kD and spectrin 120kD). In experiment 2, Mice underwent HI or sham surgery and HI mice were then treated with TH or NT for 3.5 hrs. Then, The sham and HI groups were performed MRI scans at 4 h and 22 h after treatment to acquire the hyperpolarized metabolic images, and anatomic and diffusion MR images. Mice brains were subsequently collected after scans for Western blots to measure the expression levels of RBM3.

Western blotting (Experiment 1)

The cortices from the sham, HI+NT and HI+TH group animals were dissected at multiple time points: during TH or NT (−1h), at 0 h, 4 h, 12 h, 18 h and 22 h after TH or NT. The tissue were snap frozen and stored at −80 °C until use. Nuclear and cytoplasmic proteins were extracted with the NE-PER^™ Nuclear and Cytoplasmic Extraction Reagents (Pierce Biotechnology, Rockford, IL) according to the manufacturer’s protocol. Briefly, the tissue was homogenized in 250 μl ice-cold CER I buffer with the 100x Halt^™ protease and phosphatase inhibitors (Pierce Biotechnology). After incubation with 13.75μl CER II buffer and vortex, the sample was centrifuged for 10 min at 16,000×g at 4 °C and the supernatant was saved as the cytoplasmic extract. The pellet was resuspended in 100 μl ice-cold NER buffer, shaked at 1,500 rpm in the cold room for 20 min, and centrifuged at 16,000×g at 4 °C for 30 min. The resultant supernatant was considered as the nuclear extract. The protein concentrations were measured by BCA assay kit (Pierce Biotechnology).

For Western blot analysis, an equal amount of protein samples (25–30 μg of cytoplasmic, 8–10μg of nuclear protein) was applied to 4–12% Bis-Tris SDS polyacrylamide gel electrophoresis and transferred to polyvinyl difluoride membrane. The blots were probed with the following primary antibodies overnight at 4°C: rabbit anti-RBM3 (1:500, Cat# HPA003624, MilliporeSigma, Burlington, MA); mouse anti-α spectrin (1:4000, Cat# MAB1622, Millipore, Billerica, MA) and mouse β-actin (1:6000, Cat# sc-47778, Santa Cruz Biotechnology Inc.). Horseradish peroxidase-conjugated goat anti-rabbit or goat anti-mouse secondary IgG antibodies (Santa Cruz Biotechnology Inc.) were used, and signal was visualized with enhanced chemiluminescence (Amersham, GE Healthcare). Image J software was used to measure the mean optical densities (OD) and the areas of protein signal on radiographic film after scanning. The protein levels of HI+NT and HI+TH group samples were normalized to the values of the sham group.

MRI acquisition (Experiment 2)

The sham and HI group pups were scanned at 4 h and 22 h after TH or NT on a Bruker multinuclear 3T MRI system with a 38mm diameter ¹H-¹³C dual-tuned head volume coil. Before MRI scans, 48μL of C-1 labeled ¹³C pyruvic acid was polarized in an Oxford Hypersense^™ DNP instrument at 3.5T under 1.5K for an hour. The hyperpolarized ¹³C-1 pyruvate sample was then rapidly dissolved by a heater and mixed with 4.5 mL of NaOH buffer, which resulted in a 160 mM pyruvate solution with pH~7.5. The solution was injected into the tail vein through a catheter over a span of 12 seconds (s). The total volume injected was 300 μL(150 μL into the pups and 150 μL left in the catheter). Pups were anesthetized with 1%–1.5% isoflurane and 1 L/min oxygen during MRI scans. Vital signs were carefully monitored, maintaining the respiratory rate around 30 breaths per minute (bpm) and the body temperature within a range of 36.5–37.0 °C. The dynamic HP-¹³C MRSI data was acquired 24mm × 24mm × 5mm slice centered on the brain with 2D chemical-shift imaging (CSI). Center-out k-space trajectory was used with 8 × 8 phase encoding. 128 spectral points were acquired with 2500 Hz bandwidth. The acquisition was repeated every 4 s for a total of 60 s with a constant flip angle of 10°. Anatomical images and diffusion MR images (dMRIs) were acquired accordingly for anatomical reference and injury identification. Anatomical images were acquired using a 2D T₂-fast spin echo (FSE) sequence with TR/TE=2000 ms/60 ms and 10 echo train lengths, Average=16, slice thickness=1 mm with 1 mm slice space, FA=90°, FOV was 40 mm × 40mm with 192 frequency and phase encodes, resulting in a 0.2 mm × 0.2 mm in-plane resolution. dMRIs were acquired using an Echo-planar imaging (EPI) sequence with TR/TE=2500 ms/20 ms and the same FOV, slice thickness and slice space and geometry as T2WI, Average=4, FA=90°, acquisition matrix=128 mm × 128 mm, resulting in a 0.3 mm × 0.3 mm in-plane resolution, b value=1000 s/mm² with 30 directions, which was used for identifying the brain injury. After MRI scans, brain sample was collected and prepared for Western blotting to measure the RBM3 levels.

MRI data processing and analysis

The 2D CSI data was apodized with a 10-Hz Lorentzian filter, then zero-filled to double the resolution in both spatial dimensions before Fourier reconstruction using SIVIC software (https://sourceforge.net/projects/sivic/) and then were read into MATLAB software (version 2019a; https://www.mathworks.com/products/matlab.html) for the tensor denoising and the phase and baseline corrections. The multi-voxel MR spectral dimension was zero-filled to twice the original size prior to Fourier transformation to better match the demention of the brain. The voxel shift was performed on the 2D grid and overlaid onto the anatomical T₂-weighted image (T₂WI) to locate the voxels on the left and right hemispheres. The pyruvate and lactate images were recorded for each of the 20 timepoints for the construction of dynamic curves and then the dynamic area under the curve (AUC) maps for pyruvate and lactate were obtained by taking the sum of magnitude values in the peak ranges. k_PL map was generated by fitting the dynamic curve and peak height on a voxel-wise basis for each dynamic voxel. Only voxels inside the brain regions were used for downstream analysis to avoid the inclusion of k_PL values from the noisy spectra and the areas of poor signal-to-noise ratio (SNR). Subsequently, the mean k_PL values and the total sum of AUC of pyruvate and lactate through time were calculated in the left and right hemispheres. The mean Lac/Pyr values of both hemispheres were calculated by the summed AUC of lactate to the summed AUC of pyruvate. Apparent Diffusion Coefficient (ADC) map was generated using the reconstruction streamline of diffusion tensor imaging (DTI) implemented in DSI studio software (https://dsi-studio.labsolver.org/download.html). ADC quantifies the rate of water diffusion, offering detailed insights into brain injury during the acute phase following HI insult. The ADC color map clearly delineates the border of the injured area, providing better accuracy in measuring the injury size compared to T2WI. The region of interest (ROI) was defined based on the identified lesion on the ADC map, and the injury size was calculated as the area of ROI in the slice with the largest lesion.

Statistical analysis

We assessed normality using the Shapiro-Wilk test, which indicated a deviation from a normal distribution. Mann-Whitney U test was used to compare the protein levels (RBM3, spectrin 145/150kD, spectrin 120kD) and the injury size between HI+NT and HI+TH groups, as well as the metabolic profile (k_PL and Lac/Pyr) and the injury size between the timepoints 4 h and 22 h. One-way Kruskal-Wallis test was used to compared the protein levels (RBM3, spectrin 145/150kD, spectrin 120kD) across various timepoints as well as the metabolic profile (k_PL and Lac/Pyr) across sham, HI+NT and HI+TH groups. The Pearson correlation was used to evaluate the correlation between the metabolic profile (k_PL and Lac/Pyr) and RBM3 levels. SPSS software v.29.0 was used for statistical analysis, and R studio software was used for creating graphs. Statistical significance was defined as a P-value of ≤ 0.05.

Results

A total of 145 pups were included in the study with 88 pups (male=53, female=35) in Experiment 1 and 57 pups (male=30, female=27) in Experiment 2. Eight pups were excluded due to death during the experiment or poor Western blot results and image quality.

Time course of the RBM3, spectrin 145/150kD and spectrin 120kD expression during and after treatment in HI+NT group and HI+TH group

The Western blot pictures of RBM3, spectrin 145/150kD and spectrin 120kD expression during treatment and 0 h, 4 h, 12 h, 18 h and 22 h after treatment were presented in Supplemental figure. The bar plots (Figure 2A–C) shows the time course of the changes in the RBM3, spectrin 145/150kD, and spectrin 120kD levels in HI+NT group and HI+TH group. In HI+NT group, RBM3 levels peaked at 12 h after NT, with significant higher level compared to other timepoints (P<0.05). In HI+TH group, RBM3 levels peaked at during (−1 hr, P<0.05) and at 4 h post-TH (P<0.05), then declined to the lowest levels at 18 h post-TH (P<0.05) and started to rise again at 22 h post-TH. Compared to HI+NT group, RBM3 levels were significantly higher during TH (P<0.05) as well as at 4 h (P<0.05) and 22 h (P<0.01) after TH compared to HI+NT group (Table 1). As the marker of necrotic and apoptotic cell death, spectrin 145/150kD and spectrin 120kD levels increased at early timepoints (0h and 4 h) post-TH (P<0.05) and remained relatively lower levels at later timepoints (12–22 h) post-TH in the HI+TH group. The spectrin 120kD levels at 0 h post-TH in HI+TH group was significantly higher than that in HI+NT group (P<0.05). In HI+NT group, in which spectrin 145/150kD and spectrin 120kD levels notably increased at 18 h and 22 h and their levels were significantly higher compared to HI+TH group (Table 1).

Figure 2.

The time course of the expressions of the RNA binding motif 3 (RBM3) (A) and the cell death markers, i.e., spectrin 145/150 (B) and spectrin 120 (C) during and after treatment with therapeutic hypothermia (TH) or normothermia (NT). In HI+NT group, RBM3 levels peaked at 12 h, with significant higher level compared to other timepoints (P<0.05). In HI+TH group, RBM3 levels peaked at during (−1 h, P<0.05) and 4 h post-TH (P<0.05), then declined to the lowest levels at 18 h post-TH (P<0.05) and started to rise again at 22 h post-TH. Compared to HI+NT group, RBM3 levels significantly higher during TH (P<0.05) as well as at 4 h (P<0.05) and 22 h (P<0.01) after TH compared to HI+NT group. Additionally, spectrin 145/150 and spectrin 120 levels increased at early timepoints (0 h and 4 h) post-TH (P<0.05) and remained relatively lower levels at later timepoints (12–22 h) post-TH in the HI+TH group. Spectrin 120 levels at 0 h post-TH in HI+TH group was significantly higher than that in HI+NT group (P<0.05). While in HI+NT group, spectrin 145/150 and spectrin 120 levels notably increased at 18 h and 22 h post-NT and their levels were significantly higher compared to HI+TH group (P<0.05).

Table 1.

Time course of the changes in the levels of RBM3 and spectrin proteins during and after treatment in HI + NT group and HI+TH group

Nuclear protein Levels (~folds of sham value)		−1 h (NT: n=4, TH: n=4)	0 h (NT: n = 5, TH: n = 7)	4 h (NT: n=4, TH: n=6)	12 h (NT: n=6, TH: n=6)	18 h (NT: n = 5, TH: n=6)	22 h (NT: n=3, TH: n = 12)	P-value
RBM3 (median±SD)	NT	1.8±0.3	1.3±0.7	1.9±0.5	3.1±0.4	0.9±0.8	0.7±0.4	−1 h vs. 12 h, 0.05* 0 h vs. 12 h, 0.02* 12 h vs. 18 h, <0.001*** 12 h vs. 22 h, <0.001***
	TH	8.6±1.6	3.5±2.2	8.6±2.2	2.7±1.1	0.9±0.3	2.1±0.4	−1 h vs. 0 h, 0.04* −1 h vs. 18 h, <0.001*** −1 h vs. 22 h, <0.002** 4 h vs. 0 h, 0.03* 4 h vs. 18 h, <0.001*** 4 h vs. 22 h, <0.001*** 18 h vs. 0 h, 0.01* 18 h vs. 12 h, 0.009**
	P-value	0.03*	0.11	0.01*	0.93	0.79	0.004**
Spectrin 145/150 (median±SD)	NT	2.9±1.4	2.8±1.0	2.3±0.7	2.8±0.5	6.2±1.4	6.7±3.7	−1 h vs. 18 h, 0.03* 0 h vs. 18 h, 0.004** 4 h vs. 18 h, 0.002** 12 h vs. 18 h, 0.02* 0 h vs. 22 h, 0.03* 4 h vs. 22 h, 0.02*
	TH	1.6±0.4	3.7±0.6	3.0±0.3	1.4±0.1	2.2±1.5	2.1±0.8	−1 h vs. 0 h, 0.008** −1 h vs. 4 h, 0.01* 0 h vs. 22 h, 0.04*
	P-value	0.03*	0.43	0.17	0.43	0.004**	0.03*
Spectrin 120 (median±SD)	NT	0.9±0.2	1.1±0.5	1.5±0.3	1.3±0.5	4.0±0.9	2.9±2.0	−1 h vs. 18 h, <0.001*** −1 h vs. 22 h, 0.02* 0 h vs. 18 h, 0.004** 4 h vs. 18 h, 0.04* 12 h vs. 18 h, 0.01*
	TH	1.9±0.3	2.0±0.5	1.8±0.2	1.5±0.4	1.5±0.4	1.7±0.5	0 h vs. 12 h, 0.05*
	P-value	0.03*	0.01*	0.11	0.79	0.004**	0.37

RBM3 = RNA-Binding Motif Protein 3, SD=Standard Deviation, NT=Normothermia, TH=therapeutic hypothermia, HI = Hypoxia-ischemia.

^*P<0.05

^**P<0.01

^***P<0.001

Comparison of injury size between HI+NT group and HI+TH group

Figure 3A–D shows the injured areas on T2WI and ADC maps at 22 h after treatment. Figure 3E shows the comparison of injury size between HI+NT group and HI+TH group at 4 h and 22 h after treatment. We detected a significantly larger injured size at 22 h compared to 4 h in HI+NT group (P<0.05), but no significant difference of the injury size between 4 h and 22 h in HI+TH group (Table 2). Additionally, the injury size in HI+NT was bigger than that in HI+TH at both 4 h and 22 h, but the differences didn’t reach statistical significance (Table 2).

Figure 3.

Comparison of the injury size at 4 h and 22 h after treatment between HI+NT group and HI+TH group. The injured areas at 22 h after treatment show hyperintensities on T₂-weighted images (T2WIs) and dark blue on color-coded ADC maps in one HI+NT mouse (A, B) and one HI+TH mouse (C, D). The box plot (E) shows the injury size was relatively bigger in the HI+NT group compared to the HI+TH group at both 4 h and 22 h after treatment and a significant increase in the injury size at 22 h post-NT in the HI+NT group (P<0.05).

Table 2.

Comparison of the ADC value of injury and the injury size between HI+NT group and HI+TH group

Measurements	Timepoint	NT (4 h: n = 3, 22 h: n = 5)	TH (4 h: n=4, 22 h: n=8)	P-value
Injury size (mm2) (median±SD)	4 h	9.00±3.36	6.45±8.63	4h: NT vs. TH: P=0.63 22h: NT vs. TH: P=0.07
	22 h	17.97±8.49	12.90±5.58	NT: 4 h vs.22 h: P=0.04* TH: 4 h vs.22 h: P=0.46

ADC = Apparent diffusion coefficient, SD=Standard Deviation, NT=Normothermia, TH=therapeutic hypothermia, HI = Hypoxia-ischemia.

^*P<0.05.

Comparison of cerebral metabolic profile between HI+NT group and HI+TH group

Figure 4A–F shows the HP ¹³C MRSIs and the corresponding spectrum of the selected voxel in the injured hemisphere at 4 h after treatment in one HI+NT mouse (A-C) and one HI+TH mouse (D-F). Figure 5A–B shows the comparisons of k_PL and Lac/Pyr in the ipsilateral hemisphere (injured side) across HI+NT, HI+TH and sham groups at different timepoints after treatment. We found both k_PL and Lac/Pyr were significantly higher at 4 h and 22 h in HI+NT group compared to HI+TH (P<0.05) and sham groups (P<0.05) (Table 3). There was no significant difference in k_PL and Lac/Pyr between HI+TH group and sham group. Additionally, both k_PL and Lac/Pyr were not significantly different between 4 h and 22 h in any of the groups (Table 3).

Figure 4.

Hyperpolarized ¹³C MR spectroscopic images (MRSIs) at 4 h after treatment in one HI+NT mouse (A-C) and one HI+TH mouse (D-F). A and D show a 2D grid with zero-filled 16×16 voxels overlaid onto T₂-weighted images (T2WIs). B and E show the MRSIs covering the entire brain area (red rectangular, a total of 12 voxels). C and F show the spectrum of the selected voxel in the injured region (black square).

Figure 5.

Comparison of the cerebral metabolic profile, i.e., k_PL and Lac/Pyr at 4 h and 22 h after treatment between HI+NT group and HI+TH group (A, B) and the correlation between the metabolic profile and RBM3 levels in HI+TH group (C, D). k_PL and Lac/Pyr were significantly higher in HI+NT group compared to HI+TH group and sham group at both 4 h and 22 h (P<0.05). There’re significant negative correlations between k_PL and RBM3 levels as well as Lac/Pyr and RBM3 levels in HI+TH group (P<0.01).

Table 3.

Comparison of metabolic profile in the injured ipsilateral hemisphere between HI+NT group and HI+TH group.

Anaerobic metabolic profile	Timepoint	NT (4 h: n=4, 22 h: n=4)	TH (4 h: n = 12, 22 h: n = 14)	Sham (4 h: n=3, 22 h: n=3)	P-value
kPL (median±SD)	4 h	0.10±0.04	0.03±0.04	0.01±0.03	4h: NT vs. TH: P=0.03*  NT vs. Sham: P=0.01*
	22 h	0.11±0.02	0.02±0.02	0.01±0.001	22h: NT vs. TH: P=0.005**  NT vs. Sham: P=0.004** NT: 4 h vs.22 h: P=1.00 TH: 4 h vs.22 h: P=0.23 Sham: 4 h vs.22 h: P=0.70
Lac/Pyr (median±SD)	4 h	1.09±0.23	0.28±0.38	0.16±0.35	4h: NT vs. TH: P=0.03*  NT vs. Sham: P=0.02*
	22 h	1.01±0.10	0.20±0.20	0.14±0.03	22h: NT vs. TH: P=0.004**  NT vs. Sham: P=0.008** NT: 4 h vs.22 h: P=0.49 TH: 4 h vs.22 h: P=0.16 Sham: 4 h vs.22 h: P=1.00

k_PL= Conversion rate from pyruvate to lactate, Lac/Pyr = Ratio of lactate to pyruvate, SD=Standard Deviation, NT=Normothermia, TH=therapeutic hypothermia, HI=Hypoxia-ischemia.

^*P<0.05

^**P<0.01

Correlation between cerebral metabolic profile and RBM3 levels.

Figure 5C–D shows the correlation between k_PL (5C) and Lac/Pyr (5D) in the ipsilateral hemisphere (injured side), and RBM3 levels in HI+TH group, respectively. We found significantly negative correlations between k_PL and RBM3 as well as Lac/Pyr and RBM3; the correlation coefficient was −0.761 (P<0.01) and −0.792 (P<0.01), respectively (Table 4). There were no significant correlations between k_PL and RBM3 as well as Lac/Pyr and RBM3 in HI+NT group.

Table 4.

Correlation between the metabolic profile and RBM3 protein levels in HI+TH group

	r	P-value
RBM3 — kPL	−0.76	0.003**
RBM3 — Lac/Pyr	−0.79	0.001**

RBM3 = RNA-Binding Motif Protein 3, k_PL= Conversion rate from pyruvate to lactate, Lac/Pyr = Ratio of lactate to pyruvate, TH=therapeutic hypothermia, HI = Hypoxia-ischemia.

^**P<0.01

Discussion

This is the first study investigating the time course of RBM3 levels during TH and within 24 hrs after TH, as well as evaluating the relationship between the in vivo cerebral metabolism and RBM3 levels after TH in neonatal HI mice. We found an upregulation of RBM3 levels during TH as well as at 4 h and 22 h after TH in HI+TH group. TH significantly lowered cerebral metabolic profile (k _PL and Lac/Pyr) at 4 h and 22 h, and there were significant negative correlations between the cerebral metabolic profile and RBM3 levels in the HI+TH group. These findings support our hypothesis that the upregulation of RBM3 may contribute to the TH neuroprotective effect by reducing anaerobic glycolysis.

RBM3 is a member of the glycine-rich RNA-binding protein family and is involved in mRNA biogenesis that is essential for cell proliferation and protests against adverse condition-induced cell death in both non-neural cells and neural cells, such as hypothermia and hypoxia[7]. In this study, we observed the upregulation of RBM3 in HI mice with and without TH, which is consistent with the findings in the previous study[8,10]. In HI+NT group, we found that RBM3 levels peaked at 12 h and declined to the lowest levels at 22 h after NT, indicating that the responses of RBM3 to HI conditions lasted for approximately 10 hrs. In HI+TH group, RBM3 levels peaked at during TH and at 4 h post-TH, suggesting the early response of RBM3 to TH. The lowest levels of RBM3 were seen at 18 h post-TH, implying the effect of hypothermia lasted for approximately 14 hrs. Intrerestingly, RBM3 levels began to rise again at 22 h post-TH. However, since our observation window was limited to 24 hrs, the underlying reason remains unclear. Studying the temporal changes of RBM3 expression over a longer time frame (beyond 24 hrs) would help us gain a more comprehensive understanding the RBM3 responses to hypothermia.

Previous studies have illustrated that RBM3 played a cellular protective role by regulating apoptosis and promoting neurogenesis in cell and rodent models with HI[9,10,14–16]. Hypothermia increased cold-inducible protein levels and improved cerebellar-dependent learning were also reported in neonatal rat with HI [17]. To further investigate the neuroprotective effect of TH, we measured the levels of two spectrin cleavage products as a marker of the two major types of cell death. Necrotic cell death mediated by calpain is evaluated by the spectrin cleavage products at 145/150kD, and apoptotic cell death mediated by caspase-3 is assessed by the spectrin cleavage products at 120kD[18]. We observed changes in spectrin 145/150kD and spectrin 120kD increased at early timepoints (0–4 h after TH, approxiamtely 4–8 h post-HI) and remained relatively lower levels at later timepoints (12–22 h after TH, approximately 16–26 h post-HI) in the HI+TH group. In HI+NT group, spectrin 145/150kD and spectrin 120kD levels significantly increased at 18–22 h (22–26 h post-HI). This finding suggests TH exerts neuroprotective effects of anti-necrosis and anti-apoptosis during the secondary energy failure following HI. A previous ex-vivo study reported that neuroprotection via RBM3 is regulated by hypothermia but not by hypoxia in human SK-N-SH neurons[13]. Our finding provides the evidence in HI mice model to support this and may enhance the understanding of the potential molecular mechanism of hypothermia-induced neuroprotection. Additionally, we assessed the injury size using ADC map derived by dMRIs. We found the injury sizes at both 4 h and 22 h after treatment in HI+NT group were relatively larger compared to HI+TH group. In HI+NT group, the injury size at 22 h was significantly bigger than that at 4 h, which was not seen in HI+TH group. This finding suggests that the brain injury worsened in 24 h post-insult without TH, further supporting the neuroprotective effect of TH from a different perspective.

The pathophysiological mechanism of HIE involves a complex cascade of cellular and molecular processes triggered by energy failure and mitochondrial dysfunction which centers on a decline in oxidative metabolism and the reliance on anaerobic glycolysis for maintaining cerebral energy demands. These processes consequently result in accumulation of lactic acid and excitatory amino acids, cytotoxic edema, oxidative stress and cell death[19]. TH is currently employed for neuroprotection in neonates with HIE; one plausible theory regarding hypothermia neuroprotection is that hypothermia might reduce cerebral metabolism and thus decrease oxygen demand in neurons[2]. Our previous study examined the temporal changes of cerebral metabolic profile (k_PL and Lac/Pyr) in HI mice using HP [1-¹³C]pyruvate MRSI and discovered TH suppressed anaerobic metabolic process (no significant change of k_PL and Lac/Pyr) within 24 h post-HI in the TH good-outcome group, indicating the therapeutic effect of TH[4]. In this study, we provide additional evidence for the effect of TH on HI in reducing anaerobic glycolysis. Specifically, k_PL and Lac/Pyr levels were significantly lower at both 4 h and 22 h in HI+TH group compared to HI+NT group, with values approaching those observed in the sham group. Moreover, we found significantly negative correlations between k_PL and Lac/Pyr, and RBM3 level after TH in HI+TH animals. Despite previous rodent study reporting RBM3 enhanced glycolysis and reduced apoptosis in skeletal muscle under cold expresure[20], our finding initially demonstrated the relationship between RBM3 levels and cerebral metabolism in HI mice brain following TH, which implies that overexpression of RBM3 might contribute to the TH-induced neuroprotective effects through suppressing anaerobic metabolic processes. Currently, RBM3 has attracted extensive attention regarding its upstream (e.g., Tropomyosin receptor kinase B, TrkB and fibroblast growth factor 21, FGF21) and downstream (reticulon-3, RTN3 and erythroid-derived 2-like 2, Nrf2) pathways in neuroprotection[21,22]. This study identifies the potential role of RBM3 in TH-induced neuroprotection for HI and introduces an in vivo MRI imaging tool that can be used to assess the efficacy of the RBM3-related therapeutic strategies that enhance the success rate of TH.

Interestingly, we noticed the individual variabilities of k_PL and Lac/Pyr were larger in HI+HT group compared to HI+NT group and the sham group, especially at 4 h after TH, suggesting the various responses to TH at the early phase of treatment. This aligns with our previous study, which demonstrated the various temporal changes in k_PL and Lac/Pyr within 24 h after HI between TH good-outcome group and TH poor-outcome group[4]. Further studies are needed to better understand the gene modulation of RBM3 and its causal role in enhancing TH neuroprotective effect for the treatment of neonatal HIE.

A major limitation of this study is the lack of histology and behavioral assessments at later timepoints to evaluate long-term outcomes following TH. While it is well established that TH can delay the progress of injury at the early phase after HI insult, it does not always lead to favorable long-term outcomes. Moreover, previous animal studies have shown that males were more adversely affected by HI and showed a better response to the TH (as measured by MRI and behaioral test), suggesting the potential sex difference in response to both HI and TH[23,24]. However, we didn’t analyze the sex difference in this study due to the mismatched sample sizes by sex for protein levels assessments and the small number of animal in each gender group for cerebral metabolism and injury size measurements. This is another limitation of this study and warrants furture investigation.

Conclusion

This study demonstrates that RBM3 overexpression may be one of the key factors associated with TH-induced neuroprotection by reducing the anaerobic metabolism in HI mice, suggesting RBM3 upregulation may enhance TH efficacy for neonatal HIE. In addition, this study demonstrated the ability of HP [1-¹³C]pyruvate MRSI approach to assess the cerebral anaerobic glycolysis status after the effective TH and the relationship between the cerebral metabolic profile and RBM3 levels. These findings support the use of metabolic imaging as a noninvasive marker to monitoring the therapeutic effect of RBM3-related treatment strategies to improve the success rate of TH.

Supplementary Material

Suppl-01

Supplemental Figure The Western blot pictures of RBM3, spectrin 145/150 kD and spectrin 120kD expressions during and at 0 h, 4 h, 12 h, 18 h and 22 h aDer treatments (NT or TH) in HI+NT group and HI+TH group.

Plain Language Summary.

This study aimed to explore the role of the cold shock protein RNA-binding motif 3 (RBM3) in therapeutic hypothermia (TH)-induced neuroprotection following hypoxia-ischemia (HI) using metabolic MRI. The findings demonstrated that RBM3 may be one of the key factors associated with TH-induced neuroprotection by reducing the anaerobic glycolysis process in HI mice, suggesting that RBM3 upregulation may enhance the efficacy of TH for hypoxic-ischemic encephalopathy (HIE).

Data Availability Statement

Data will be shared in compliance with relevant data protection and confidentiality guidelines. Requests for data access can be directed to the corresponding author