Increased cerebral expressions of MMPs, CLDN5, OCLN, ZO1 and AQPs are associated with brain edema following fatal heat stroke

Human brain samples were collected from 46 autopsy cases, including 23 fatal heat stroke cases and 23 age-matched controls. Nine candidate reference genes (PES1, POLR2A, IPO8, HMBS, SDHA, GAPDH, UBC, B2M, ACTB) were evaluated in the cerebral cortex of 10 forensic autopsy cases (5 heat stroke and 5 controls), using the geNorm module in qBaseplus software. SDHA, POLR2A, IPO8 and HMBS were identified as the most stable reference genes. Using these validated reference genes, mRNA expressions of Matrix metalloproteinases (MMPs, MMP2 and MMP9), Claudin5 (CLDN5), Occludin (OCLN), Zona occludens protein-1 (ZO1) and Aquaporins (AQPs, AQP1 and AQP4) in the cerebral cortex were examined. Relative mRNA quantification using Taqman real-time PCR assay demonstrated increased calibrated normalized relative quantity (CNRQ) values of MMP9, CLDN5, OCLN, ZO1 and AQP4 in heat stroke cases. Heat stroke cases showed an increase in brain water content, which was found to be positively correlated with MMP9, OCLN, ZO1 and CLDN5 mRNA. When using one conventional reference gene (GAPDH or ACTB) for normalization, no difference was detected between heat stroke and controls. In immunostaining, only AQP4 showed more intense staining in most heat stroke cases. The present study, for the first time, reports increased cerebral MMP9, CLDN5, OCLN, ZO1 and AQP4 in heat stroke and suggest a crucial role of reference gene selection when using postmortem human tissues.

MMP9-mediated BBB disruption may be a crucial step in the pathogenesis of some neuroinflammatory diseases [6][7][8] . Claudin5 (CLDN5), Occludin (OCLN) and Zona occludens protein-1 (ZO1) are key tight junction (TJ) proteins that play an important role in modulation of BBB permeability 9 . Decreased expression of CLDN5, OCLN and ZO1 has been reported to be closely associated with BBB damage 10 . Aquaporins (AQPs) are water channels that facilitate water transport. AQP1 and AQP4 are presumed as major contributors to participate in brain water homeostasis 11 . AQPs, in particular AQP4, appear to play a crucial role in cerebral volume regulation following trauma, inflammation, ischemia, tumors as well as metabolic disturbances 12,13 .
Experimental studies reported that hyperthermia caused by heat exposure is closely associated with the breakdown of the BBB followed by brain edema 14 . The disruption of the BBB is characterized by the degradation of the junctional complex proteins and increase in multiple matrix metalloproteinases. However, these studies have generally been restricted to rodent models that have inevitably inherent deficiencies, including a lissencephalic brain and small head size relative to body size. Furthermore, heat stroke in human is rarely as pure as in experimental models, it is necessary to investigate the human materials following fatal heat stroke.
The present study analyzed the gene expressions of MMP2, MMP9, CLDN5, OCLN, ZO1, AQP1 and AQP4, in the brains of forensic autopsy cases, using reverse transcription quantitative PCR (RT-qPCR), combined with immunohistochemical detections, to investigate the molecular pathology of brain edema in fatal heat stroke cases with special regard to the importance of reference gene selection.

Results
Brain water content. There was a significant increase in brain water content in the heat stroke group as compared to the control group (Table 1, heat stroke: 82.3% ± 2.9%; control: 79.3% ± 2.6%, p < 0.05, Student's t test). RNA concentration, purity and integrity. RNA concentrations ranged from 25.9 to 348.7 ng/μL (mean 194.1 ng/μL). There were no age or postmortem interval dependences on Pearson correlation analysis (p > 0.05). RNA concentrations showed no significant differences between the heat stroke group and control group.
RNA purity, determined using 260/280 absorbance (A 260 /A 280 ) ratios, ranged from 1.75 to 2.19 (mean 1.98). There were no age or postmortem interval dependences on Pearson correlation analysis (p > 0.05). RNA purity showed no significant differences between the heat stroke group and the control group.
RIN values showed substantial variations in each group. There were no age or postmortem interval dependences on Pearson correlation (p > 0.05). RIN values were evidently lower in the heat stroke group as compared to the control group (Table 1, heat stroke: 3.6 ± 1.0; control: 4.9 ± 1.1, p < 0.05, Student's t test). Amplification efficiency. The amplification efficiencies of targets and reference genes ranged from 86.2% (ACTB) to 104.6% (UBC), showing small inter-individual variations (standard deviation, SD < 5%). Details are shown in Table 2.
Reference genes validation. The geNorm module in qBase plus software ranked the 9 reference genes.
The most stable one was SDHA followed by POLR2A, IPO8 and HMBS. The least stable one was ACTB (Fig. 1). Pairwise variation (V) was calculated based on normalization factor values (NFn and NFn + 1) after the inclusion of the least stable reference gene and indicated if the extra reference gene added to the stability of the normalization factor. The V-value was the lowest when the fifth most stable gene (PES1) was added (Fig. 2). However, when the fourth most stable gene (HMBS) was added, V3/4 showed a V-value of 0.153, near the threshold of 0.15. Therefore, to save on cost and time, four reference genes SDHA, POLR2A, IPO8 and HMBS, were selected for normalization.

Data analysis
Normalization against validated reference genes. Raw Ct values and amplification efficiencies of targets and 4 validated reference genes, SDHA, POLR2A, IPO8 and HMBS, were imported into the qBase plus software. CNRQ values were exported and statistically investigated.
There were no gender-related differences, or age or postmortem interval dependence in CNRQ values of target genes on Pearson correlation (p > 0.05).
CNRQ values of AQP4, CLDN5, OCLN, ZO1 and MMP9 were significantly higher in the heat stroke group as compared to the control group (Fig. 3).
CNRQ values of CLDN5 and MMP9 were found to be positively correlated with brain water contents (r 2 = 0.1225 and 0.1486, p < 0.05).  Table 2. Introduction of the PCR primers and probes. "_m" indicates the assay's probe spans an exon junction and will not detect genomic DNA. "_g" indicates the assay may detect genomic DNA. Detailed information for each TaqMan Assay is available from Applied Biosystems.  Immunostaining. Immunostaining showed substantial interindividual variations in each group. AQP1 ( Fig. 6a and b) and AQP4 ( Fig. 6c and d, Fig. 7) were mainly detected in glial cells which were morphologically identified as astrocytes, and only AQP4 showed more intense staining in most heat stroke cases. CLDN5 was strongly positive in capillary endothelia, and no significant differences in distribution or intensity were detected between heat stroke and control group ( Fig. 6e and f, Fig. 7). MMP2 was detected clearly in the neurons, showing no significant differences in distribution or intensity between heat stroke and control group ( Fig. 6g and h, Fig. 7). MMP9 was located in glial cells, neurons and capillary endothelia, and no significant differences in distribution or intensity were detected between heat stroke and control group ( Fig. 6i and j, Fig. 7). OCLN was positive in capillary endothelia, sporadically in neurons and glial cells, and no significant differences in distribution or intensity were detected between heat stroke and control group ( Fig. 6k and l, Fig. 7). ZO1 was located in capillary endothelia, and no significant differences in distribution or intensity were detected between heat stroke and control group ( Fig. 6m and n, Fig. 7).

Discussion
RT-qPCR is increasingly applied to determine changes in gene expressions due to the high sensitivity and accuracy of the technique. The most common procedures in RT-qPCR are relative measurements of genes of interest after normalization with the endogenous reference gene(s). Accurate and reliable relative RT-qPCR requires ideal reference gene(s). However, expressions of several conventional reference genes were shown to vary due to  nutritional or hormonal factors, biological processes, and/or tissue or cell types; a single endogenous reference gene cannot meet the criteria of an ideal reference gene 15 . In the field of molecular neurobiology, RT-qPCR, using postmortem autopsy tissues, has become a hotspot [16][17][18] . It can provide novel biomarkers and disease-modifying therapeutic targets for some CNS diseases. Previous studies, using different normalization strategies, demonstrated contradicting results, suggesting the importance of normalization 19,20 .
In the present study, with the help of geNorm module in qBaseplus software, nine reference genes were evaluated. The V-value was the lowest when the fifth most stable gene (PES1) was added. Further addition of genes increased V-values, indicating a negative influence on the normalization process. Ideally, a threshold V-value of 0.15 is recommended as a cut-off value by geNorm to determine the optimal number of reference genes 21 . However, when the fourth most stable gene (HMBS) was added, V3/4 showed a V-value of 0.153, near the threshold of 0.15. Therefore, to save on cost and time, four reference genes SDHA, POLR2A, IPO8 and HMBS, were selected for normalization. In the present study, CLW/H ratios were higher in the heat stroke group as compared to the control group, indicating brain edema in heat stroke cases. However, using mRNA measurements of intracerebral MMPs, CLDN5, OCLN, ZO1, and AQPs as markers of brain edema, inconsistent results were detected by different normalization methods. When those four validated reference genes, SDHA, POLR2A, IPO8 and HMBS, were used for normalization, increased cerebral expressions of AQP4, CLDN5, OCLN, ZO1 and MMP9 were detected in heat stroke group. However, these findings cannot be detected when GAPDH or B2M alone was used for normalization. Expression stability values of these five reference genes calculated by geNorm showed ACTB as the least stable one, followed by GAPDH. Therefore, gene expression levels that normalized against four validated reference genes were believed to be accurate and reliable; ACTB and GAPDH, two conventional reference genes, were not suitable for normalization of human postmortem brain tissues.
Another considerable factor influencing the accuracy of gene expression analysis using RT-qPCR is the integrity of RNA 22 . However, unlike animal experimentation, RNA degradation is inevitable and unpredictable for human tissues collected at autopsy. In the present study of the human brain tissues, RIN values showed no postmortem interval-dependent changes but were significantly lower in the heat stroke group as compared to the control group, indicating that RNA quality was more seriously affected in cases of hyperthermia.
The up-regulations of MMP2 and MMP9 in brain are associated with an increase of BBB permeability by degrading the endothelial basal lamina of the BBB which results in vasogenic edema 23 . Despite the well documented effects of systemic inflammatory response, the impact of hyperthermia on the BBB has been overlooked and the probable mechanism has not been fully addressed. In the present study, brain tissues in heat stroke group showed evidently higher CNRQ values of MMP9, but not MMP2. These findings suggest independent contributions of MMP2 and MMP9 in the brain tissues of heat stroke group, which require further investigation. MMP9 was regarded as a key player in the alteration of BBB permeability. Several studies in animal models have shown that increased MMP9 is closely related to the breakdown of BBB, by digesting ECM 24,25 . Both inhibiting MMP9 and deleting MMP9 gene can attenuate the BBB disruption 26,27 .
In addition, intercellular junctions between endothelial cells are essential for vascular integrity and function. Breaching of endothelial barriers is a key event in the development of brain edema. The loss of CLDN5, OCLN and ZO1 in TJs can open the BBB, lead to brain edema and neuronal cell death. Decreased expression of CLDN5, OCLN and ZO1 is closely associated with BBB damage 28 . However, in the present study, increased CLDN5, OCLN and ZO1 mRNA expression was detected in the heat stroke group. These findings might be considered a compensatory mechanism to mend junctional complexes and restore barrier function.
AQP4, the principal AQP in mammalian brain, highly correlates with a variety of pathophysiological processes of brain edema. Increased expression of AQP4 in the brain indicated that AQPs participated in the formation of brain edema 29 . In the vasogenic edema resolution phase, an increase of AQP4 was observed in several studies 30,31 . Therefore, in heat stroke cases, AQP4 might play a beneficial role in eliminating accumulating water from the extracellular space of the CNS, suggesting an activation of the self-protective system.
In the present study, the immunostaining did not detect any evident differences in distribution or intensity among all the causes of death except for AQP4. These findings may be caused by a lower sensitivity of immunostaining in detecting changes in gene products than that with quantitative analyses of gene expressions using RT-qPCR. The major limitation of the present study is that the protein levels of all targets have not been examined because of the limitation of the postmortem materials, which need further investigation.
In conclusion, the present study shows that brain edema is evident in heat stroke, and human brain retains a self-protective response capacity in victims who died due to heat stroke. Systematic analysis of gene expressions using RT-qPCR is a useful procedure and validation of reference genes is crucial.

Materials and Methods
Sample collection. A total of 46 human forensic autopsy cases were selected from autopsy documents. The demographics of study subjects are described in Table 1. In each case, the cause of death was carefully diagnosed on the basis of autopsy examination, including macromorphological, histological, toxicological and biochemical analyses. Cases were divided into two groups as follows: 23 fatal heat stroke cases and 23 age, gender and postmortem interval (PMI)-matched control subjects, including three hanging cases, three strangulation cases, five blunt injury cases, six fire fatality cases, and six acute cardiac death cases. A thorough neuropathological analysis was performed as part of our routine investigation, and cases with any preexisting neurological pathologies were excluded in the present study.
In the present study, clearly accountable cases without any other complications that may have contributed to the death, supported by well-established circumstantial evidence, were collected. Postmortem interval was defined as the estimated time from death to autopsy which was estimated on the basis of autopsy findings and circumstantial evidence recorded in autopsy documents. The definition and criteria of deaths due to heat stroke are described in our previous report 32 ; drug abusers and chronic alcoholics were excluded from the heat stroke group. This work was approved by our institutional Ethics Committee of Southern Medical University. All sampling methods were followed anatomical practices and carried out in accordance with regulations of Methods of extraction, fixation, packing and inspection of forensic pathology of The PRC Public Safety Industry Standard (GA/T 148-1996) and Forensic pathology materials extraction, fixed operating instructions of Southern Medical University (NYSJ-JS-BL04). As one of our routine work, the informed consent paper was obtained from the immediate family members of deceased before starting autopsy.
Brain water content. Brain water content was measured by a halogen moisture analyzer (model HB43, METTLER TOLEDO, Switzerland) automatically according to the manufacturer's instructions 20 . Briefly, brain tissue samples were taken from consistent sites in the parietal lobe of left cerebral hemispheres at autopsy. About one gram of brain tissue was weighted first to obtain a wet weight (WW), then placed in the analyzer at 150 °C for about 30 min and weighted again to obtain a dry weight (DW). The water content was calculated using the equation: (WW − DW)/WW × 100%.
Toxicological analyses. The procedures of drug testing and analysis, including chemicals and reagents, sample preparation and conditions of the instrument were performed by gas chromatography/mass spectrometry 33 . Extraction of total RNA and cDNA synthesis. Tissue specimens were taken from consistent sites in the central anterior of left cerebral hemispheres (precentral gyrus) at autopsy, then immediately submerged in 1 ml of RNA stabilization solution (RNAlaterTM, Ambion, Austin). Total RNA was isolated from 100 mg of sample using RNAiso Plus (Takara Bio, Inc., Shiga, Japan) according to the manufacturer's instructions. After extraction, the RNA concentration was estimated by spectrophotometric analysis using a NanoDrop 1000 (Thermo Scientific, Wilmington, USA). cDNA copies of total RNA were obtained using a High Capacity RNA-to-cDNA kit (Applied Biosystems Japan, Ltd.), then were adjusted to a concentration equivalent to 5 ng/μl of total RNA using nuclease-free water.
Evaluation of the quality and integrity of RNA samples. RNA purity was determined using 260/280 absorbance (A 260 /A 280 ) ratios. The RNA integrity number (RIN) was determined using a RNA 6000 Nano Labchip kit in an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, USA) following the manufacturer's protocol.
Amplification efficiency calculation. Amplification efficiencies were calculated from raw fluorescent data (Rn values), using a completely objective and noise-resistant algorithm, the Real-time PCR Miner program 36 (http://ewindup.info/miner/).

Data analysis
Normalization against validated reference genes. Raw Ct values and calculated amplification efficiencies of these 9 reference genes were imported into the qBase plus software 37 . In qBase plus software, geNorm module was used to identify the most stable reference genes and determine the minimum number of reference genes 21 .
After determining the minimum number of reference genes, raw Ct values and amplification efficiencies of targets and validated reference genes were imported into the qBaseplus software again. Using a calibrator case (peracute death due to blunt injury, 52-year-old male; 27 h postmortem), calibrated normalized relative quantity (CNRQ) values were exported from the qBaseplus software, and statistically investigated.
Normalization against conventional reference gene (ACTB or GAPDH). The expression levels of mRNA transcripts are described as the ratios of the targets normalized to the endogenous reference (GAPDH or B2M), using the 2 −ΔΔCt method, as the ratios for fold change relative to the above mentioned calibrator. The 2 −ΔΔCt method assumes that the amplification efficiency of the reaction is ideal (100%) and constant for each sample.
Immunostaining. The brains were fixed in buffered 4% formaldehyde for 2 weeks and the cerebral hemispheres were cut coronally into 1-cm-thick sections, and the sections of cerebellum and brain stem were also prepared. Paraffin-embedded brain tissue specimens were taken from the standardized anatomical regions. Serial sections (5 μm thick) were cut and stained with hematoxylin-eosin (HE) as part of routine laboratory investigation. In the present study, parietal lobes of left cerebral hemispheres were used for immunostaining.
Image J. 1.46 R (NIH, USA) was used to calculate the intensity and extent of staining for the detected molecules. The per-area density of staining was calculated in this manner to reflect the percentage of the positive staining, resulting in a semi-quantitative analysis. A total of 5 microscopic fields were randomly selected, and their images were cropped. The integral optical density (IOD) levels of the stained cells in the tissue samples were then calculated by image analysis. Results were expressed as the mean ± standard deviation (SD) per tissue examined. Using a calibrator case (peracute death due to blunt injury, 52-year-old male; 27 h postmortem), ratios for fold change relative to the calibrator were used for statistical analysis.
Statistic. All the RT-qPCR experiments were performed in triplicate, and results are reported as the mean ± SD. Correlation analyses between pairs of parameters were performed using linear regression (Pearson correlation analysis). The Student's t test (two-tailed) was used to compare groups. Statistical analyses were performed using GraphPad Prism version 5.0 (GraphPad Software, San Diego, USA). Values of p < 0.05 were considered as statistically significant.