Novel injury scoring tool for assessing brain injury following neonatal hypoxia-ischemia in mice

Abstract

The variability of severity in hypoxia ischemia (HI) induced brain injury among research subjects is a major challenge in developmental brain injury research. Our laboratory developed a novel injury scoring tool based on our gross pathological observations during hippocampal extraction. The hippocampi received scores of 0–6 with 0 being no injury and 6 being severe injury post-HI. The hippocampi exposed to sham surgery were grouped as having no injury. We have validated the injury scoring tool with T2-weighted MRI analysis of percent hippocampal/hemispheric tissue loss and cell survival/death markers after exposing the neonatal mice to Vannucci’s rodent model of neonatal HI. In addition, we have isolated hippocampal nuclei and quantified the percent good quality nuclei to provide an example of utilization of our novel injury scoring tool. Our novel injury scores correlated significantly with percent hippocampal and hemispheric tissue loss, cell survival/death markers, and percent good quality nuclei. Caspase-3 and Poly (ADP-ribose) polymerase-1 (PARP1) have been implicated in different cell death pathways in response to neonatal HI. Another gene, sirtuin1 (SIRT1), has been demonstrated to have neuroprotective and anti-apoptotic properties. To assess the correlation between the severity of injury and genes involved in cell survival/death, we analyzed caspase-3, PARP1, and SIRT1 mRNA expressions in hippocampi 3 days post-HI and sham surgery, using RT-qPCR. The ipsilateral (IL) hippocampal caspase-3 and SIRT1 mRNA expressions post-HI were significantly higher than sham IL hippocampi, and positively correlated with the novel injury scores in both males and females. We detected a statistically significant sex difference in IL hippocampal caspase-3 mRNA expression with comparable injury scores between males and females with higher expression in females.

Introduction

The developing brain has been shown to be especially vulnerable to hypoxia ischemia (HI) related brain injury in humans and experimental animals [1]. The selective vulnerability of the brain seen following early HI has been linked to the age and the sex of the subject, duration and type of the insult, as well as the presence of associated comorbidities at the time of the injury such as infection [2–7]. Although it is very hard to control for these variables when conducting research in humans, experimental research has provided tools to control for the majority of these variables in the laboratory setting [8]. However, one of the challenges of developmental brain injury research in the laboratory is that the severity of the HI-induced injury can vary among subjects, despite controlling for the genetic background, age, sex, and strain of the experimental animals, and duration of the hypoxia [9]. Attempts to address this variability have been made by multiple investigators and efforts have been made to identify tools to stratify the experimental animals according to the severity of injury [10–14]. The main goal of using these tools is to control for severity of injury to decrease the variability in the research targets being studied. For example, expression of certain genes and proteins may be influenced by the severity of the injury and if the data is not stratified accordingly, valid conclusions may fail to be drawn. In addition, translational neuroprotective therapies have shown to be more successful following mild to moderate injury and not following severe injury [15, 16]. Thus, development of new therapies that target mechanisms mediating neuroprotective actions in mildly or moderately injured brains may hold the most promise.

Hypoxia ischemia induced brain injury activates multiple pathological mechanisms such as oxidative stress, excitotoxicity, inflammation, and cell death, proceeding from within minutes of injury and extending beyond a year [17]. Investigators who study translational developmental brain injury research target their therapies to minimize or eliminate the early and late neuronal cell death [18–24]. Cell death mechanisms following brain injury have been extensively studied in experimental animals [4, 25–31]. Unlike normal physiological conditions, following brain injury, both in neonates and adults there is activation of caspase-dependent and caspase-independent cell death pathways. A growing number of studies have shown that the activation of caspase-dependent and caspase-independent apoptotic pathways following neuronal injury are sexually differentiated [28]. Previous rodent studies demonstrate increased caspase-3 and PARP1 activation within hours following global and focal ischemia [5, 28, 29, 32]. Sirtuin1 (SIRT1) is a widely-studied member of the sirtuin family that has been shown to have a regulatory role in apoptosis, inhibition of neuroinflammatory signaling pathways, and neuroprotection [33, 34]. SIRT1 has been shown to interact with tumor suppressor genes such as p53, and transcription factors such as NF-kB and the FOXO family [35]. SIRT1 decreases the p53 activity, which results in inhibition of apoptosis [36, 37]. Several studies have demonstrated SIRT1 upregulation following different neuronal injury models, enhancing neuronal survival [38, 39].

In this paper, we are reporting a novel injury scoring tool that we have validated the injury scores with different methods such as brain imaging and cell survival/death gene expressions. In addition, we are reporting an example of utilization of the novel injury scoring tool by quantifying the percent good quality nuclei that are isolated from sham and mildly, moderately, and severely injured hippocampi. Our novel injury scores correlate with percent tissue loss obtained by using T2 weighted MRI, caspase-3, and SIRT1 mRNA expressions in both male and female neonatal mice hippocampi 3 days post-HI. In addition, the percentage of good quality nuclei decreases as the injury severity increases. We have also identified a sex difference in caspase-3 mRNA expressions between males and females which is comparable with the previous literature.

Materials and Methods

Animal use

All procedures were carried out in adherence with NIH Guide for the Care and Use of Laboratory Animals using protocols reviewed and approved by the Institutional Animal Care and Use Committee at the School of Medicine and Public Health of the University of Wisconsin-Madison. Total of 37 male and 36 female postnatal day 9 (P9) mice pups (total 73) were used in the study.

Induction of neonatal HI

Neonatal HI was induced as previously described with some modifications [40]. P9 C57BL/6J mice were anesthetized with isoflurane (Butler Schein Animal Health Supply, Reno, NV) (5% for induction, 3% for maintenance) in nitrous oxide/oxygen [40]. The body temperature of the pups was maintained at 36°C using a heated surgical table (Molecular Imaging Products, Bend, OR). Under a surgical microscope (Nikon SMZ-800 Zoom Stereo, Nikon, Melville, NY), a midline neck skin incision was made followed by elevation of the submandibular salivary glands bilaterally. The left carotid sheath was then visualized between the trachea and the left sternocleidomastoid muscle. The left common carotid artery was freed from the carotid sheath by blunt dissection, electrocauterized with a bipolar electrocoagulator (Vetroson V-10 Bi-polar Electrosurgical Unit, Summit Hill Laboratories, Navesink, NJ) and cut. The surgical site was flushed with 0.5% bupivacaine for pain control and closed with a single 6–0 silk suture. Mice were returned to their home cages, and the home cages with dam were placed into a normoxic chamber with a temperature set point of 36.5°C and monitored continuously for a 2 h recovery period so that the pups continued to nurse and not deprived of nutrients. To induce unilateral ischemic injury following the 2 h recovery period, the mice pups were placed in a hypoxic chamber (BioSpherix Ltd, Redfield, NY) equilibrated with 10% O₂ and 90% N₂ at 36.5°C for 50 minutes. This is a well-characterized model of neonatal HI and results in reproducible brain injury ipsilateral (IL) to the electrocauterized left common carotid artery [41–43]. In this model, unilateral sectioning of the common carotid artery alone does not induce ischemic injury due to collateral circulation from the contralateral (CL) side through the circle of Willis. Only subsequent exposure to hypoxia results in IL hypoxic ischemic reperfusion injury due to the preferential decrease of blood flow to the IL hemisphere secondary to cerebral vasoconstriction induced by hypocapnia [44]. However, the injury severity may differ even if the pups are produced by the same dam. The injury severity may range from mild to severe injury. Sham-operated mice receive anesthesia and exposure of the left common carotid artery without electrocauterization or hypoxia induction as previously described [41].

Extraction of hippocampi

Hippocampi were extracted as described previously by our laboratory and others [45, 46]. Briefly, following decapitation of the mouse without anesthesia the skull was removed, brain was extracted and the whole brain was placed into a petri dish in a dorsal orientation. Using curved forceps, the cerebellum and brainstem were removed. Then a sagittal incision was made to separate the hemispheres [47]. The basal ganglia and thalamus were removed from the medial surface of the hemisphere [41, 48, 49]. Both hippocampi were then identified and harvested [50, 51]. The whole brain and bilateral hippocampi photographs were taken using a Nikon SMZ-800 Zoom Stereo (Nikon, Melville, NY) microscope and iPhone 12 Pro (Apple, CA, USA).

Novel injury scoring tool

The injury scores were assigned using our novel injury scoring tool during the hippocampal extraction (Fig. 1). As the brain is extracted from the skull the cortical injury is assessed, followed by the assessment of hippocampal opacification and liquefaction. As the severity of injury increased in Vannucci’s HI model of neonatal mouse, the cortical injury started becoming more evident. However, because the cortical injury can be inconsistent, we based our injury scores of 0–6 on the hippocampal morphology. The injury scores were then grouped together based on severity of injury. Injury scores 0 and 1 were assigned as mild, 2 and 3 were assigned as moderate, and 4, 5, and 6 were assigned as severe. Sham samples were assigned as having no injury. The personnel assigning the injury scores during extraction was blinded to the percent tissue loss calculations performed using MRI.

Fig. 1. Novel injury scoring tool.

The freshly harvested whole brain, ipsilateral (IL) and contralateral (CL) hippocampi samples were examined under a microscope and injury scores were assigned from 0 to 6. Hippocampi are scored as 0, if the IL hippocampus has identical tissue consistency and morphology with the CL hippocampus. Hippocampi are scored as 1, if the head of the IL hippocampus (approximately 1/3 of the length of the hippocampus) is slightly opaque compared to the CL hippocampus. Hippocampi are scored as 2, if the opacity in the IL hippocampus extends to the body (approximately 1/2 of the length of the hippocampus) and slight liquefaction in the head of IL hippocampus is present. Hippocampi are scored as 3, if the opacity in the IL hippocampus constitutes 2/3 of the length of the hippocampus and slight liquefaction is observed in the entire IL hippocampus. Hippocampi are scored as 4, if there is opacity and moderate liquefaction in the entire IL hippocampus, and if there is a slight loss of shape due to liquefaction. Hippocampi are scored as 5, if opacity and moderate liquefaction is present in the entire IL hippocampus along with moderate shape loss. Hippocampi are scored as 6, if there is total liquefaction in the IL hippocampus that causes total shape loss. The injury scores were then grouped together based on severity of injury. Injury scores 0 and 1 were assigned as mild, 2 and 3 were assigned as moderate, 4, 5, and 6 were assigned as severe. Sham samples were assigned as having no injury. Whole brain photographs are shown with white arrows denoting cortical infarcts.

MRI parameters and post-processing

In order to validate our novel injury scoring tool, MRI was performed using an Agilent 4.7-tesla Small Animal MRI scanner with a Varian 200-MHz quadrature mouse radiofrequency coil (38 mm internal diameter) prior to hippocampal extraction and injury scoring in mice 3 days post-HI. At 3 days post-HI, mice were anesthetized with 1.5% isoflurane in an oxygen/air mixture administered through a nose cone and then secured in a cradle position within the center of the magnet bore. The respiratory rate and body temperature were monitored with an MR-compatible physiology monitoring unit, and temperature was maintained within physiologic limits (37.0 ± 0.2°C) using a heated airflow unit. T2-weighted fast spin echo images (TR = 3,500 ms; effective TE = 60 ms; echo train length = 8; matrix size: 192 × 192; averages = 8) were acquired from 14–16 contiguous axial slices with a field of view of 20 × 20 mm and a slice thickness of 0.8 mm.

T2-weighted MRIs were then manually assessed for artifacts. Rician noise was removed from T2-weighted images using an adaptive non-local means algorithm as part of the Advanced Normalization Tools software suite [52, 53]. Images were subsequently corrected for inhomogeneous spatial intensities using the N4BiasFieldCorrection tool [54].

T2-weighted MRI to quantify percent tissue injury

The T2-weighted MRI percent injury quantification was manually done using ImageJ and ITK-SNAP, to ensure that we have both the area and volumetric measures, respectively (Fig. 2). On average, 9 slices were used for hemispheric measurements and 3–4 slices were used for hippocampal measurement. In order to analyze the T2-weighted MRIs for percent area loss, we manually segmented images using ImageJ (NIH). The T2-weighted MRIs were converted to the 8-bit format using the Image/Type tool. The threshold was adjusted using the Image/Adjust/Threshold tool on each image to isolate the hemispheres. The hemispheres were then traced with the wand tool. The ventricles, cysts, and injured areas were excluded from the area measurement with thresholding. The hemisphere was divided into contralateral and ipsilateral halves using the pencil tool. The same steps were followed for the hippocampal segmentation. The areas of the hemispheres and hippocampi were then calculated using the Analyze/Analyze particles tool. The area loss was converted into a percent change using the following formula: [((CL Area) − (IL Area)) ÷ (CL Area)] × 100. These values were used for subsequent analyses. The T2-weighted MRI were also manually segmented using ITK-SNAP [55]. Specifically, preprocessed T2-weighted MRI were loaded into the ITK-SNAP viewer and slices containing hippocampi were visually identified and labeled. Only viable tissue was labeled for volumetric analysis, with lesions, cysts, or tissues with liquefaction being excluded. Manual segmentations of the entire right and left hemisphere of the brain were additionally labeled to account for cortical damage. After segmentation, volumes of the manually labelled hippocampus and brain hemispheres were automatically calculated in ITK-SNAP by multiplying the voxel count in each slide by the voxel volume (0.004883 mm³). The volume loss was converted into a percent change using the following formula: [((CL Volume) − (IL Volume)) ÷ (CL Volume)] × 100. These values were used for subsequent analyses. The personnel quantifying the percent tissue loss via T2-weighted MRI were blinded to the injury scores assigned to each brain.

Fig. 2. MRI segmentation methodology with ImageJ and ITK-SNAP.

The T2-weighted MRIs were segmented using the ImageJ and ITK-SNAP programs. (a) Representative MRI coronal slices obtained from mouse brain, (b) hippocampal and (b’) hemispheric segmentation using ImageJ and (c) hippocampal and (c’) hemispheric segmentation using ITK-SNAP are shown. Red and blue represents the contralateral and ipsilateral sides, respectively (scale bar: 1 mm).

Definition of anterior and posterior hippocampus

In order to determine whether our novel injury scores were influenced by spatial vulnerability to the HI injury, we have also correlated our novel injury scores with the MRI measures of anterior and posterior percent hippocampal and hemispheric loss. To define the anterior and posterior hippocampi in our MRI percent tissue loss analysis, the Allen Mouse Brain Atlas for P56 coronal sections were used as a reference. The ventral hippocampal commissure (vhc) in image 64 from the Allen Mouse Brain Atlas was determined to denote the start of the anterior hippocampus and the dorsal hippocampal commissure (dhc) in image 77 from the Allen Mouse Brain Atlas was determined to denote the start of the posterior hippocampus [56]. Similarly, the dorsal hippocampal commissure was used to denote the end of the anterior hemisphere and the start of the posterior hemisphere (Fig. 2).

Quantitative reverse transcription polymerase chain reaction (RT-qPCR)

The CL and IL hippocampi were taken from −80° C and placed on ice immediately. Total RNA was extracted from a single hippocampus using the RNeasy Mini Kit from Qiagen. The purity and concentration of the total RNA were determined from the optical density measurements at 260 nm and 280 nm (Nanodrop, Thermo Scientific, Wilmington, DE). For the reverse transcription reaction (RT); 1000 ng of total RNA, 1 ug of random hexamers (Applied Biosystems #N8080127), 1 ug of dNTP mix (Promega #U1511) and RNase free water were heated for 5 min at 65° for denaturation. 4 ug of First Strand Buffer, 2 ug of DTT, 1 ug of M-MLV (Invitrogen #28025013), and 1 ug of RNase inhibitor (Promega, #N2511) were then added and the samples heated up to 42° C for an hour followed by 5 min at 95° C to inactivate the reverse transcriptase. The resulting cDNA samples were then diluted with 40 ul of RNase free water and stored at −80° C until use. qPCR was performed using TaqMan Gene Expression Assay probes (Applied Biosystems, Waltham, MA). A master mix solution was made consisting of 0.5ul TaqMan gene expression assay probe (20X), 7.5 ul TaqMan Gene expression master mix, and 3.0 ul nuclease free water. Samples are loaded in duplicate on a 384-well plate with a final reaction volume of 15 ml, consisting of 4.0 ul of cDNA and 11.0 ul of master mix solution. Each reaction for a single gene was done in duplicate and consisted of predesigned gene-specific primers and probes for Sirt1 (Mm01168521_m1), Casp3 (Mm01195085_m), Parp1 (Mm01321084_m1), Gapdh (Mm99999915_g1), Hprt (Mm03024075_m1), and TaqMan® Master Mix (Applied Biosystems #43–695-14) on a 384-well plate. PCR amplification was accomplished using a ViiA 7 Real-Time PCR System (Thermo Fisher Scientific, Waltham, MA) and QuantStudio Real-Time PCR Software v1.3 (Applied Biosystems) running a standard amplification protocol (50° C 2 min, 95° C 10 min, 95° C 15 sec, 60° C 1min, 40 cycles). Following the completion of the qPCR run, relative gene expression levels were calculated using the 2^−ΔΔCt method upon normalization to the expression levels of reference genes (Gapdh and Hprt). Our preliminary data showed comparable results when we normalized gene expressions with housekeeping genes Gapdh and Hprt separately. Thus, we report our gene expressions using only Hprt as the housekeeping gene and hypothalamus as the cross-plate control [45].

Mice hippocampal nuclei isolation

2–10 mg of IL hippocampal tissue stored in −80° C was placed on dry ice immediately for the nuclei isolation. All buffers were prepared fresh and maintained on ice. Tissue was pulverized using a pellet pestle (Fisherbrand #12141364) in 250 μL of chilled lysis buffer containing 250 mM sucrose (Sigma #S0389), 25 mM KCl (Sigma #60142), 5 mM MgCl2 (Sigma #M1028), 20 mM Tris-HCl (pH 7.4) (Sigma #T219), 0.08 U/μL RNase Inhibitor (Sigma #56968100), 1x Protease Inhibitor (Sigma #56079100), 1 mM DTT (Sigma #646563), and nuclease-free water (Americanbio #18722100). The tissue suspension was transferred to a dounce tissue grinder (1 mL volume, Wheaton #357538; autoclaved, chilled). Another 250 μL of chilled lysis buffer was added to the original tube for rinsing purposes, and then transferred to the 1 mL dounce tissue grinder. The tissue suspension was homogenized thirty times with loose pestle and thirty times with tight pestle, ensuring that constant pressure and no air was introduced to the sample. The homogenized solution was filtered through a 40 μm cell strainer (Falcon #352340) pre-wetted with approximately 100 μL of chilled lysis buffer into a 50 mL tube. Then, 500 μL of 50% Iodixanol consisting of optiprep (Alere Technologies #00121), 25 mM KCl, 5 mM MgCl2, 20 mM Tris-HCl (pH 7.4), 0.08 U/μL RNase inhibitor, 1x protease inhibitor, 1 mM DTT, and nuclease-free water were added. The solution was pipette mixed ten times and pipetted into a 15 mL tube. Tubes were centrifuged at 1000 rcf for 30 minutes at 4°C on a swing out rotor. After spinning, the supernatant was removed and pellets were resuspended in chilled resuspension buffer (250 mM sucrose, 25 mM KCl, 5 mM MgCl2, 20 mM Tris-HCl [pH 7.4], 0.08 U/μL RNase inhibitor, 1x protease inhibitor, 1% BSA [GeminiBio #C021004], 1 mM DTT, and Nuclease-Free water. 1.2 mL of wash buffer (0.1 % Tween-20 [Biorad #1662404], 25 mM KCl, 5 mM MgCl2, 20 mM Tris—HCl [pH 7.4], 0.08 U/μL RNase inhibitor, 1% BSA, 1 mM DTT, 1x DPBS [Gibco #2421018]) was added to the sample, and then transferred into a 1.5 mL LoBind Eppendorf tube (Eppendorf #K195217H). 10 μL of sample were set aside for first counts on hemocytometer (Fisher Scientific #0267151B). The tubes were centrifuged at 500 rcf for 10 minutes at 4°C on a fixed rotor. Based on the first quantification, nuclei were diluted to 1 million nuclei/mL in sample-run buffer (25 mM KCl, 5 mM MgCl2, 20 mM Tris-HCl [pH 7.4], 0.08 U/μL RNase inhibitor, 1% BSA, 1 mM DTT, 1x DPBS). 10 μL of sample was put into hemocytometer for second quantification, three random fields per sample consisting of 8 hemocytometer squares per field (dimensions: 0.25 mm × 0.25 mm for each square) were imaged with ZEISS Axio Vert.A1 light microscope (ZEISS, Jena, Germany) under 20X magnification.

ImageJ quantification of isolated hippocampal nuclei

We have assessed the morphology of the nuclei obtained from mildly, moderately, severely injured and sham hippocampi, and quantified the percent good quality nuclei using the ImageJ software to provide an example for utilization of our novel injury scoring tool. Particles with an area greater than 150 pixels qualified as nuclei. The nuclei that had an intact membrane and more than 75% circularity were considered as good quality nuclei per 10x Genomics® Single Cell Preparation Guide (Rev C, CG00053, Pleasanton, CA). The nuclei that lacked membrane integrity and had less than 75% circularity were considered as suboptimal due to nuclei damage. Particles with an area less than 150 pixels were considered as debris and were not included in the nuclei quantification. Three separate 8-bit images consisting of 8 hemocytometer squares (dimensions: 0.25 mm × 0.25 mm for each square) were used for the quantification of each hippocampus. The images were processed using the Adjust/Threshold tool. The same brightness/contrast settings were used for all images. The total nuclei were quantified by using the “Analyze/Analyze Particles/Size: 150-Infinity, Circularity: 0.00–1.00, Show: Overlay” setting. The good quality nuclei were quantified by using the “Analyze/Analyze Particles/Size: 150-Infinity, Circularity: 0.75–1.00, Show: Overlay” setting. The percent good quality nuclei was calculated with this formula: [(good quality nuclei count ÷ total nuclei count) × 100]. Negligible number of cells that show clumping were not included in the cell counts.

Statistical analysis

Linear regression analyses were conducted to validate our novel injury scoring tool by regressing hemispheric and hippocampal damage (measured using MRI) on the injury score. These analyses were conducted both across spatial regions and stratified by spatial regions (anterior and posterior). The strengths of the associations were quantified by calculating the corresponding R2 values. Analogously, correlations between ImageJ and ITK-SNAP percent tissue loss measurements were evaluated using linear regression analyses. Multi-factorial analysis of variance (ANOVA) was conducted to evaluate the caspase-3, PARP1 and SIRT1 mRNA expressions. Sex (male vs. female), treatment (Sham vs. HI) and side (CL vs. IL) and the corresponding two-way interaction effects were included as factors. Model adjusted means were calculated and compared between the treatment, sex, and side interaction combinations. Analogously, the associations between caspase-3, PARP1, and SIRT1 mRNA expression and injury score were evaluated by regressing the mRNA expression levels on the severity of injury scores, quantified as 0= “No Injury”, 1= “Mild Injury”, 2= “Moderate Injury” and 3= “Severe Injury”. The R2 values and corresponding p-values were reported and the results were summarized in graphical format using scatter plots. All reported p-values are two-sided and p<0.05 was used to define statistical significance. Statistical analyses were conducted using SAS software (SAS Institute, Cary NC), version 9.4.

Results

Novel injury scoring tool allows for injury stratification

Hippocampus is an important structure for memory and learning. Hippocampi are injured consistently in rodents and humans following HI. Thus, different mechanisms that are involved in memory and learning can be triggered according to the severity of injury. Stratification of the injury helps with more accurate interpretation of specific gene and protein expressions that may be altered as the severity of injury changes.

In order to stratify the hippocampal injury, we have developed a novel injury scoring tool that relies on gross pathological observations (Fig. 1). The injury scores of 0–6 were given based on hippocampal morphology.

Hippocampi are scored as 0, if the IL hippocampus has identical tissue consistency and morphology with the CL hippocampus. Hippocampi are scored as 1, if the head of the IL hippocampus (approximately 1/3 of the length of the hippocampus) is slightly opaque compared to the CL hippocampus. Hippocampi are scored as 2, if the opacity in the IL hippocampus extends to the body (approximately 1/2 of the length of the hippocampus) and slight liquefaction in the head of IL hippocampus is present. Hippocampi are scored as 3, if the opacity in the IL hippocampus constitutes 2/3 of the length of the hippocampus and slight liquefaction is observed in the entire IL hippocampus. Hippocampi are scored as 4, if there is opacity and moderate liquefaction in the entire IL hippocampus, and if there is a slight loss of shape due to liquefaction. Hippocampi are scored as 5, if opacity and moderate liquefaction is present in the entire IL hippocampus along with moderate shape loss. Hippocampi are scored as 6, if there is total liquefaction in the IL hippocampus that causes total shape loss. In summary, as the severity of injury increases, the opacification and liquefaction in the hippocampi extends from the head to the tail, eventually leading to shape loss. The cortical injury becomes more evident as the hippocampal injury severity increases to 4 or beyond. However, according to our observations, for a given hippocampal injury score equals or greater than 4, the cortical injury may vary. For analysis, the injury scores are grouped together based on severity of injury. Injury scores 0 and 1 are assigned as mild, 2 and 3 are assigned as moderate, and 4, 5, and 6 are assigned as severe. Sham samples are assigned as having no injury (Fig. 1).

Percent hemispheric and hippocampal tissue loss positively correlates with novel injury scores

In order to validate our novel injury scoring tool, MRI was performed to quantify the hemispheric and hippocampal tissue loss prior to brain extraction (Fig. 2.a). The T2 weighted images obtained from 19 neonatal mice 3 days post-HI were analyzed and the percent tissue loss was quantified using ImageJ and ITK-SNAP to determine the percent area (Fig. 2.b, b’) and volume loss (Fig. 2.c, c’), respectively. The results were then correlated with the assigned injury scores obtained from extracted hippocampi. Two of the 19 MRI scans were excluded due to tilting of the brain to avoid inaccurate analysis (n=17).

For ImageJ, the percent area loss measurements obtained for hemispheres (R²=0.84, p<0.0001) and hippocampi (R²=0.65, p=0.0001) significantly correlated with the assigned novel injury scores (Fig. 3.a–b). Similarly, the percent volume loss measurements obtained with ITK-SNAP for hemispheres (R²=0.88, p<0.0001) and hippocampi (R²=0.82, p<0.0001) correlated with our assigned novel injury scores (Fig. 3.c–d).

Fig. 3. Correlation of the hemispheric and hippocampal percent tissue loss with the injury score.

Percent tissue loss analysis of the brain MRIs were performed to validate the novel injury scoring tool. Using ImageJ, (a) hemispheric (R²=0.84, p<0.0001) and (b) hippocampal (R²=0.65, p=0.0001) area loss was quantified and correlated with injury score. The data was stratified into anterior and posterior (a’) hemispheric (R²_anterior=0.80, p_anterior<0.0001, R²_posterior=0.59, p_posterior=0.0003) and (b’) hippocampal (R²_anterior=0.62, p_anterior=0.0002, R²_posterior=0.42, p_posterior=0.0051) regions and correlated with injury score. Using ITK-SNAP the (c) hemispheric (R²=0.88, p<0.0001) and (d) hippocampal (R²=0.82, p<0.0001) percent volume loss was quantified and correlated with injury score. The data was stratified into anterior and posterior (c’) hemispheric (R²_anterior=0.87, p_anterior<0.0001, R²_posterior=0.58, p_posterior=0.0004) and (d’) hippocampal (R²_anterior=0.68, p_anterior<0.0001, R²_posterior=0.61, p_posterior=0.0002) regions and correlated with injury score (n=17).

We have further stratified our data into anterior and posterior hemisphere/hippocampi and correlated with our novel injury scores. In ImageJ, the anterior (R²=0.80, p<0.0001) and posterior (R²=0.59, p=0.0003) hemispheric percent area loss measurements had a statistically significant positive correlation with our novel injury scores (Fig. 3.a’). The anterior (R²=0.62, p=0.0002) and posterior (R²=0.42, p=0.0051) hippocampal percent area loss measurements also had a statistically significant positive correlation with our novel injury scores (Fig. 3.b’). In ITK-SNAP, the anterior (R²=0.87, p<0.0001) and posterior (R²=0.58, p=0.0004) hemispheric percent volume loss measurements had a statistically significant positive correlation with our novel injury scores (Fig. 3.c’). Similarly, the anterior (R²=0.68, p<0.0001) and posterior (R²=0.61, p=0.0002) hippocampal percent volume loss measurements had a statistically significant positive correlation with our novel injury scores (Fig. 3.d’). In summary, our novel injury scores correlated with percent hemispheric/hippocampal tissue loss measurements and there was no statistically significant spatial difference detected.

Finally, we correlated the area and volume loss measurements quantified with ImageJ and ITK-SNAP, respectively. Our results showed that ImageJ (area loss) and ITK-SNAP (volume loss) measurements of the hemispheric (R²=0.82, p<0.0001) and hippocampal (R²=0.82, p<0.0001) percent tissue losses correlated significantly with each other (Fig. 4.a–b). Thus, we confirm that both techniques can be utilized to validate novel injury scores using percent tissue loss.

Fig. 4. Correlations of ImageJ and ITK-SNAP percent tissue loss measurements.

Following MRI, percent area loss was calculated using ImageJ and percent volume loss was calculated using ITK-SNAP. The percent (a) hemispheric and (b) hippocampal tissue loss obtained using each program were correlated with each other. The hemispheric and hippocampal measurements had a statistically significant positive correlation (p<0.0001) with an R² value of 0.82 and 0.82, respectively (n=17).

Ipsilateral hippocampal SIRT1 and caspase-3 mRNA expressions increase in IL hippocampi in both male and female mice post-HI

In order to validate our novel injury scoring tool, we have measured the cell survival (SIRT1) and cell death marker (caspase-3, PARP1) gene expressions in the neonatal mice hippocampi 3 days post-HI using RT-qPCR (Sham n=6–8 per sex, HI n=10–12 per sex). Following HI, the male IL hippocampal SIRT1 mRNA expressions were statistically significantly higher compared to corresponding CL HI (p=0.0002) and IL sham hippocampi (p=0.0001) (Fig. 5.a). Similarly, following HI, the female IL hippocampal SIRT1 mRNA expressions were statistically significantly higher compared to corresponding CL HI (p=0.0006) and IL sham hippocampi (p=0.0002) (Fig. 5.a’). We did not detect a statistically significant sex difference in hippocampal SIRT1 mRNA expressions. The SIRT1 mRNA expressions in the CL hippocampi exposed to only hypoxia were not statistically significantly different among the experimental groups.

Fig. 5. SIRT1, Caspase-3, PARP1 mRNA expressions in CL and IL hippocampi at 3 days post-HI.

At 3 days post-HI, hippocampi were extracted and the CL and IL hippocampi were probed for SIRT1, caspase-3, PARP1, and Hprt. Hippocampal mRNA expressions were quantified using RT-qPCR. The mean CL and IL SIRT1 (a, a’), caspase-3 (b, b’), and PARP1 (c, c’) mRNA expression ± SEM values are reported (Sham n=6–8 per sex, HI n=10–12 per sex) with 2^−ΔΔCt values on the Y-axis and the experimental groups on the X-axis. The average injury scores were 2.18±1.30 for male HI hippocampi and 2.25±1.26 for female HI hippocampi (p=0.856). The IL hippocampal SIRT1 and caspase-3 mRNA expressions were statistically significantly higher than CL hippocampi in both males (p_SIRT1=0.0002, p_caspase-3=0.0016) and females (p_SIRT1=0.0006, p_caspase-3=0.0001) post-HI. There was a statistically significant difference in IL hippocampal SIRT1 and caspase-3 mRNA expressions between sham and HI groups in male (p_SIRT1=0.0001, p_caspase-3=0.007) and female (p_SIRT1=0.0002, p_caspase-3=0.0002) mice. Asterisk (*) represents higher IL hippocampal caspase-3 mRNA expressions in females compared to males post-HI (b’) (p=0.0204). No statistically significant induction of PARP1 mRNA expression was detected post-HI (c, c’). There was no statistically significant difference detected among the CL groups.

As shown in Fig. 5.b, following HI, the male IL hippocampal caspase-3 mRNA expressions were statistically significantly higher compared to the corresponding CL hippocampi (p=0.0016) and IL sham hippocampi (p=0.007). Similarly, following HI, the female IL hippocampal caspase-3 mRNA expressions were statistically significantly higher compared to corresponding CL hippocampi (p=0.0001) as well as IL sham hippocampi (p=0.0002) (Fig. 5.b’). There was a statistically significant sex difference between male and female IL hippocampal caspase-3 mRNA expressions post-HI (p=0.02) (Fig. 5.b’). There was no statistically significant difference in the CL hippocampal caspase-3 mRNA expressions among the experimental groups. The PARP1 mRNA expressions did not differ between HI and sham groups (Fig. 5.c, c’).

Ipsilateral hippocampal SIRT1 and caspase-3 expressions positively correlate with injury severity in both male and female mice post-HI

After determining the induction of the caspase-3 and SIRT1 in male and female IL hippocampi, we moved on to determine the correlation between injury severity and SIRT1 and caspase-3 mRNA expressions by using our novel injury scoring tool (Fig. 6) (Sham n=6–8 per sex, HI n=10–12 per sex). Our HI model results in mild and moderate hippocampal injury more than severe hippocampal injury. Average injury scores were 2.18±1.30 for male HI hippocampi and 2.25±1.26 for female HI hippocampi (p=ns). Because we did not detect any statistically significant difference in the CL hippocampal caspase-3 and SIRT1 mRNA expressions among the experimental groups, we are reporting only the correlations of IL hippocampal gene expressions with injury severity. The IL hippocampal SIRT1 (Fig. 6.a, a’), and caspase-3 (Fig. 6.b, b’) mRNA expressions positively correlated with injury severity in both male (p_SIRT1=0.0333, p_caspase-3=0.0125) and female (p_SIRT1=0.0495, p_caspase-3<0.0001) mice at 3 days post-HI.

Fig. 6. Correlations of IL hippocampal SIRT1 and caspase-3 mRNA expressions with injury severity post-HI.

The male and female IL hippocampi were grouped as no injury, mild, moderate, and severe according to their injury severities as explained in the methods section. Regression analyses were performed to assess the correlation between IL hippocampal SIRT1 (a, a’), caspase-3 (b, b’) mRNA 2^−ΔΔCt values and the injury severity (Sham n=6–8 per sex, HI n=10–12 per sex). The 2^−ΔΔCt values are shown on the Y-axis and the injury severity groups are shown on the X-axis. The IL hippocampal SIRT1 mRNA 2^−ΔΔCt values positively correlated with injury severity in (a) males (p<0.05, R²=0.25) and (a’) females (p<0.05, R²=0.21). IL hippocampal caspase-3 mRNA 2^−ΔΔCt values positively correlated with injury severity in (b) males (p<0.05, R²=0.31) and (b’) females (p<0.05, R²=0.68).

Utilization of the novel injury scoring tool for single nucleus transcriptomics studies

As an example for researchers who would like to utilize our novel injury scoring tool, we have stratified isolated hippocampal nuclei quality according to severity of injury. Recent technology that examines single nucleus transcriptomics defines the nuclei that are not undergoing apoptosis, necrosis, or membrane lysis as healthy and good quality nuclei. We isolated nuclei from IL sham and IL HI hippocampi with mild, moderate, and severe injury 3 days post-HI (Fig. 7.e–g). The nuclei suspensions with consistent nuclei counts of ~1×10⁶/mL were quantified using the ImageJ software to determine the good quality nuclei that have intact membranes and more than 75% circularity (Fig. 7.a–d). Three bright field images per hippocampus were analyzed and the good quality nuclei values were quantified. The average percent good quality nuclei was % 56 ± 3 for sham, % 42 ± 3 for mildly injured, % 33 ± 2 for moderately injured, and % 17 ± 2 for severely injured hippocampi (n=4 hippocampi per experimental group). The percent good quality nuclei negatively correlated with severity of injury (p<0.0001, R²=0.71) (Fig. 7.h).

Fig. 7. Quantification of percent good quality hippocampal nuclei post-HI.

At 3 days post-HI, the male and female IL hippocampi were extracted and utilized for nuclei isolation. Three fields per hippocampus were imaged using 20X magnification and the images were quantified using the ImageJ software. (a) The bright field nuclei images were converted to 8-bit format and then (b) threshold images were created. (c) The particles with an area greater than 150 pixels were considered as nuclei. (d) The nuclei with an intact membrane and >75% circularity were considered as good quality nuclei. Nuclei obtained from the IL sham, (e) mildly, (f) moderately, and (g) severely injured hippocampi were used to assess the changes in nuclei morphology according to the injury severity (scale bar: 0.05 mm). The red arrowheads show good quality nuclei, and the white arrowheads show damaged nuclei. (h) The percentage of good quality nuclei were calculated and correlated with the injury severity. Counts obtained from each field per injury severity were plotted in the figure. The injury severity was shown on X-axis and the percent good quality nuclei was shown on Y-axis. The percent good quality nuclei negatively correlated with the injury severity (p<0.0001, R²= 0.71) (n=4 hippocampi per group, total n=16). The average percent good quality nuclei was % 56 ± 3 for sham, % 42 ± 3 for mildly injured, % 33 ± 2 for moderately injured, and % 17 ± 2 for severely injured hippocampi.

Discussion

In this report, we describe a novel injury scoring tool that can be utilized in future research to stratify hippocampal gene or protein expressions according to the injury severity. This will enable us to identify the pathways that get activated differentially as the severity of injury increases.

The utility of novel injury scoring tool

One of the challenges of developmental brain injury research in the laboratory is that despite controlling for the genetic background, age, sex, strain of experimental animals, duration of hypoxia, and consistency among observers, severity of the injury varies [57]. Thus, in behavioral and therapeutic research there is a need for a robust and versatile injury scoring tool that allows researchers to stratify their data by injury severity.

Previous injury scoring tools have been primarily based on immunohistochemistry or MRI analysis to assess the severity of injury [58, 59]. These injury scoring tools can provide detailed injury scores of certain regions of the brain such as cortex, thalamus, hippocampus, as well as the specific locations within the hippocampus such as CA1 or CA2 regions. However, they require significant tissue processing, time, or resources [11, 14]. Injury scoring tools that utilize immunohistochemistry makes the tissue unavailable for additional assays such as RT-qPCR and western blot. Our injury scoring tool is based on gross pathological observations during brain extraction, which is efficient, cost effective, and allows an injury score to be assigned before further tissue processing and analysis. However, the efficiency of our injury scoring tool comes with its own limitations. Our novel injury scoring tool cannot be used alone for studies that utilize neuroprotective interventions, since the injury score cannot be assigned prior to the intervention. However, our novel injury scoring tool can be paired with MRI since the percent tissue loss measurements strongly correlate with our novel injury scores. Thus, therapeutic efficacy of any intervention can be gauged by comparing pre-treatment MRIs and post-extraction injury scores.

Specific regions of the brain can be more vulnerable following neonatal HI. After pediatric traumatic brain injury (TBI), MRI studies have shown that the severity of hippocampal head injury negatively correlated with the Glasgow Coma Scale, suggesting that the head of the hippocampus is more vulnerable to injury compared to the tail [60]. This spatial vulnerability suggests that a similar injury pattern could be possible in other models of injury such as neonatal HI. Our data suggests that the hippocampal injury starts from the head of the hippocampus, which corresponds to the anterior hippocampus in the MRI. As the severity increases the injury extends towards the tail of the hippocampus that corresponds to the posterior hippocampus in the MRI. However, our injury scores correlated with both anterior and posterior hippocampal and hemispheric percent tissue loss, suggesting that our scores are not influenced by anterior or posterior spatial orientation at 3 days post-HI. Given the nature of our experimental design and one time point assessment, we cannot make any conclusions about spatial evolution of injury or recovery. However, it is well established that HI injury evolves over time in both rodent and human neonates [61][62].

The correlation of ITK-SNAP and ImageJ shows that the T2 weighted in vivo MRIs can be used to quantify percent tissue loss in the hippocampus and hemisphere. We quantified the tissue loss with both of these readily available and common programs to provide different options to researchers who may choose use our novel injury scoring tool. ImageJ and ITK-SNAP percent tissue loss calculations are well correlated and that the scoring tool correlates with MRI performed in living mice, regardless of the MRI quantification approach. This injury scoring validation method can be applied to gross pathological injury scoring tools for other brain regions such as the striatum, thalamus, and amygdala. Resolution in these images acquired using a 4.7T MRI scanner was not sufficient for us to reliably visualize and segment some brain regions such as striatum, thalamus, and amygdala. Imaging with higher magnetic field strength MRI scanners may allow this method to be extended to these brain regions.

The correlation of cell survival/death genes with novel injury scores following HI

As a second validation tool, we have identified cell survival/death genes and correlated them with novel injury scores. Recent studies demonstrated that SIRT1 is a critical player in the mitigation of cell death following neuronal injury and may act to inhibit apoptosis, conferring neuroprotection [33, 34, 63, 64]. In a TBI study, SIRT1 levels started to increase from 30 minutes to 24 hours after trauma and then gradually decreased to the normal levels both in vivo and in vitro. In addition, inhibition of SIRT1 significantly promoted apoptotic neuronal death after TBI [63]. Similarly, serum SIRT1 levels were found to be increased after traumatic spinal cord injury [65]. Rodent studies that are focusing on ischemic injury have also shown increased SIRT1 protein levels at 18, 24, 48 hours and 7 days after middle cerebral artery occlusion (MCAO) [38, 39]. In this study, we have shown that SIRT1 mRNA expression increases in both male and female IL hippocampi at 3 days post-HI. We have also demonstrated that the magnitude of the HI-induced SIRT1 mRNA expression positively correlated with the severity of injury using our novel injury scoring tool. We speculate that mechanisms exist by which injured cells respond to the HI insult and proportionately induce SIRT1 upregulation, which in turn functions to limit cellular damage and/or promote repair. Future studies will provide clarity on the pathways that mediate the induction of SIRT1 transcription in injured cells, and thereby confer some degree of neuroprotection and/or cellular repair in specific cell populations.

Many investigators have shown the activation of the caspase system, particularly caspase-3, in response to neonatal HI [5, 26, 66–68]. Caspase-3 has been identified as the most abundant effector caspase in the immature brain and it functions as the executing molecule in various apoptotic pathways [4][66]. While there is a dramatic increase in caspase-3 in the immature brain, the increase is slight in the mature brain following HI [4]. In this study, we have shown that caspase-3 mRNA expression increases in both male and female IL hippocampi compared to sham at 3 days post-HI. In addition, the increase in hippocampal caspase-3 mRNA expression positively correlated with the severity of injury in both males and females.

Sex differences in cell death pathways following HI

Over the past several years, an increasing number of studies have focused on the sex differences in cell death mechanisms [27, 31]. Previous data demonstrates that the progression of apoptosis is driven more heavily by caspase-dependent pathways in females and caspase-independent pathways in males. Caspase inhibitor Q-VD-OPh, has shown to decrease the caspase-3 activation in both male and female mice, however the effect was more robust in females and resulted in neuroprotection following MCAO [69]. In the present study, we report that HI resulted in caspase-3 mRNA upregulation in both male and female hippocampi 3 days post-HI, and expression is higher in female compared to male hippocampi with comparable injury scores. This is in contrast to our previously reported findings that caspase-3 protein expressions were higher in the male hippocampi compared to the female hippocampi at 1 day post-HI [45]. However, in that study, hippocampi were not stratified according to the severity of injury, so it remained unknown whether this represented a primary sex difference in cell death pathways at early time points, or whether this difference in expression was driven by a more severe injury in male hippocampi. Creation of the novel injury scoring tool described here allows us to address questions such as this. Indeed, when using this scoring tool to stratify injury, we found that when male and female hippocampi have comparable injury scores, the caspase-3 expression was higher in female hippocampi compared to male. We speculate that higher levels of HI-induced caspase-3 expression in females may contribute to enhanced neuroprotection in female versus male neonates. The mechanisms that mediate enhanced expression in females remain to be determined.

Male neurons undergo apoptosis predominantly via PARP1/apoptosis inducing factor (AIF) dependent pathways while apoptosis proceeds via caspase/cytochrome-c dependent pathway in the female neurons both in vivo and in vitro [27, 69]. The increasing levels of nitric oxide following neuronal injury activates the DNA repair enzyme PARP1. Overactivation of PARP1 leads to nicotinamide adenine dinucleotide (NAD⁺) consumption, Poly ADP-ribose (PAR) accumulation, and subsequent AIF mobilization [69, 70]. Previous data has demonstrated an early increase of PAR accumulation and NAD⁺ reduction 4 hours after neonatal HI [70]. Another study in mice reported that the number of AIF positive hippocampal cells peaked at 24h post-HI then decreased 72h post-HI [4]. We did not detect a statistically significant increase in PARP1 expression 3 days post-HI which may be secondary to the fact that PARP1 induction happens early, within hours following HI Thus, we were unable to show any correlation between PARP1 and severity of injury.

Utilization of our novel injury scoring tool in single nucleus transcriptomics studies

We have tested our novel injury scoring tool to assess the effect of severity of injury on nuclei morphology prior to initiation of our single nucleus transcriptomics studies and reported here as an example for researchers who might utilize this novel injury scoring tool in their experiments. We utilized the extracted nuclei from our single nucleus transcriptomics experiments and stratified the percent good quality nuclei according to severity of injury. Our data showed negative correlation between the percent good quality nuclei and severity of injury. Highly apoptotic or necrotic samples showed high percentage of damaged nuclei that would impact the data quality. Thus, stratifying the nuclei according to the injury severity will not only decrease the variability of the data but also will help identify pathways that are activated differently as the severity of the injury changes.

Limitations of the study and suggestions for researchers:

In addition to previously discussed limitations, in our hands, Vannucci’s HI model generates more mild and moderate injury, as opposed to severe injury. This resulted in having lower number of severely injured hippocampi samples. Thus, the number of tested hippocampi was not equally distributed among the mild-moderate and severely injured hippocampi. Our previous experience showed that when we used the cDNAs obtained at different time points in the same plate, we encountered variability within experimental/sham groups. Thus, we recommend converting the cDNA from mRNA at the same time for all samples that are going to be run in the same plate prior to running qPCR. In this study, this methodology helped reduce the variability within and across the experimental conditions, i.e. sham CL and IL mRNA expressions, and obtain more consistent and reproducible data.

Data Availability Statement

All data generated or analyzed during this study are included in this article. Further enquiries can be directed to the corresponding author.