Effect of methylene blue on the genomic response to reperfusion injury induced by cardiac arrest and cardiopulmonary resuscitation in porcine brain

Animals

Experiments were performed on 30 piglets of Swedish triple breed of both sexes aged 12-14 weeks and with a mean weight of 25.8 kg ± 1.3 kg.

Cardiac arrest model

Previous use of the same model of extended cardiac arrest (12 min of untreated cardiac arrest (CA) and 8 min of cardiopulmonary resuscitation (CPR) has already been described [19]. Here we used the same anaesthesia, fluid administration surgical preparation and experimental protocol to induce CA. Return of spontaneous circulation and post-resuscitative treatments either with normal saline or with MB were administered following the protocol previously described by our group [18]. After completion of study, all animals received an injection of 10 mL potassium chloride 20 mmol/mL and were sacrificed. In all the animals the skull was opened in prone position immediately before death and the brain was taken out and frozen in liquid nitrogen within 1 min after death. Thirty pigs were divided into three groups. The first group (CA, n = 12) underwent untreated CA. The brain of the animals was removed either immediately after CA (Ca0 n = 3), or after 5 min (Ca5 n = 3), 20 min (Ca20 n = 3) and 30 min (Ca30 n = 3) respectively (Figure 1A). The other two groups underwent CA followed by CPR and received either an infusion saline (ROSC n = 9) or an infusion of MB (10 mg/mL Metyltioninklorid, Apoteket, Umeå, Sweden) with saline (MB n = 9) as described in Figure 1B. Brain samples from resuscitated animals were collected at 30 min (Rosc30 and MB30 groups), 60 min (Rosc60 and MB60 groups) or 180 min (Rosc180 and MB180 groups) after ROSC.

RNA extraction

Total RNA was extracted using RNeasy Lipid Tissue kit (Qiagen)according to the manufacturer's instructions. Contaminating genomic DNA was removed during RNA purification by addition of an on-column DNase digestion step with RNase-Free Dnase I (Qiagen). RNA quality was verified on an Agilent 2100 Bioanalyzer (Agilent).

Affymetrix gene profiling and microarray analysis

Total RNA (10 μg) was first converted to double-stranded cDNA (One-cycle cDNA synthesis kit, Affymetrix UK Ltd) and then to biotinylated cRNA (IVT labelling kit, Affymetrix UK Ltd) as described in the One-cycle Target Labelling protocol (Affymetrix UK Ltd). After fragmentation and quality confirmation (Agilent 2100 Bioanalyzer), 10 μg of the biotinylated cRNA were hybridized for 16 hours to an Affymetrix GeneChip Porcine Genome Array. The array was then washed and stained with streptavidin-phycoerythrin solution using the Fluidics Station 450 (Affymetrix UK Ltd.) before it was scanned using the Affymetrix GeneChip Scanner 3000. Data were normalized and comparative analysis results were obtained from Affymetrix GeneChip Operating Software (GCOS v1.4; Affymetrix UK Ltd). The results were deposited to GEO database under accession number GSE22165. Statistical analysis was performed using the Significance Analysis of Microarrays (SAM) statistics program to determine genes that were significantly regulated at each time point [22]. SAM uses a permutation method to determine the significance of changes in expression by assimilating a set of gene-specific t-tests and hence reports a false-discovery rate (FDR) rather than a probability. Data query, mining, filtering and sorting as well as hierarchical cluster analysis and principle component analysis (PCA) were performed using Spotfire DecisionSite^® for Functional Genomics software (Tibco-Spotfire Inc., Cambridge, MA, USA).

Gene filtering and selection

In a first step, after normalization, results from the GeneChip Operating Software (GCOS) were considered. Genes with raw intensity values that did not meet a minimum threshold (more than 30 in at least one of the samples) or described as absent in all 30 samples were excluded from the results. The selected probe sets were then submitted to statistical analysis to identify significant changes in gene expression. Pair-wise comparison analysis based on GCOS statistical algorithm was carried out to compare animals of the experimental group to each animal of the corresponding control group. The group means were also compared using the SAM software. By combining these 2 approaches, we obtained a list of modulated genes for each condition. The stringent criteria for the lists of genes were that the FDR values obtained with the SAM software were inferior to 5% and there were 9 significant change calls generated by GCOS. Furthermore, the lists of genes were restricted to include only genes for which expression changes were ≥ ± 1.5-fold. Changes which occurred 5, 20 or 30 min after cardiac arrest or 30, 60 or 180 min after ROSC were analyzed using the Ca0 as control group. Changes after 30, 60 or 180 min of MB treatment following ROSC were analyzed using the Rosc30, Rosc60 and Rosc180 as control groups, respectively.

Confirmatory Real-Time PCR analysis

Data generated from microarrays were confirmed using real-time PCR (qPCR).

All experiments were performed in triplicates for every sample and contained negative controls. Results were evaluated by MyiQ single-colour Real-time PCR detection system software (Bio-Rad Laboratories, USA). The expression levels for Hsp22, Hsp27, Hsp40, Hsp70, Hsp72, Hsp90, rgs2, cart, ndrg2 as well genes similar to dusp1, ddit3, gadd45b, nfil3, set and dhcr24 mRNA transcripts were described as the ratios of the targets normalized to the endogenous reference (sdha) and then the ratios for fold change relative to the calibrator were obtained. The calibrator was constituted from brain samples of piglets belonging to the Ca0 group.

Gene annotation

The Affymetrix porcine genome microarray is minimally annotated. Approximately 11% of the probe sets on this array are described with gene names, posing a challenge to biological interpretation of data. Lack of annotation is likely due to the limited availability of full-length porcine cDNA sequence. To gain further information about the identity and/or function of the selected sequences which were not associated to a protein name, BLAST sequence-similarity searches were performed. After careful verification of the resulting alignments, probe sets sharing sequence similarity with known proteins from other species were annotated as "similar to". To avoid redundancy, only gene symbols are mentioned in the text while full names are used in the figures and tables. Furthermore genes annotated "similar to" will be considered as sharing the same function as the gene they were associated with.

Gene expression profiles after untreated CA

Gene expression profiles after ROSC

Transcriptional changes for Rosc 30, 60 and 180 were analyzed using Ca0 as baseline. For changes specifically induced by MB treatment, MB30, 60 and 180 groups were compared to their corresponding ROSC (with saline) controls i.e. Rosc 30, 60 and 180, respectively. After gene filtering, we obtained a final list of 518 probe sets showing significant changes of expression after either ROSC or MB treatment. Manual annotations gave a total of 317 probe sets (61%) similar to a known protein, 80 (16%) identified genes and 121 (23%) lacking any sequence similarity in any database and therefore called "unknown" (for a detailed list see Additional file 1).

We reorganized this large data set by performing hierarchical clustering (see Figure 2). The data set was split into 9 parts named cluster 1 to 9 varying in size from 2 to 207 genes. The relevance and the relationship between the clusters was studied using PCA (Figure 3A) and expression patterns (Figure 3B). The results showed 5 distinct subsets of transcripts named as cluster 1, 5, 6, 7 and 8 which contained 34, 130, 41, 205 and 99 genes, respectively. Four more diffuse and heterogeneous very small subsets (4 members or less) were not considered as valid clusters. Therefore their members were associated to the main cluster they appeared to be most similar to, based on the PCA 3D representation. "Clusters" 4 (2 genes) and 9 (3 genes) were grouped with cluster 1 (Figure 4), "cluster" 3 (4 genes) with cluster 8 (Figure 5) and "cluster" 2 (4 genes) with cluster 7 (Figure 6) while cluster 6 (Figure 7) and cluster 5 (Figure 8) remained unaffected. Relationships between the main clusters were clearly displayed by the PCA analysis: cluster 8, corresponding to genes up-regulated after ROSC, opposite to cluster 1 (down-regulated after ROSC) and cluster 7, corresponding to genes induced by MB, opposite to cluster 5 (decreased by MB) (see Figure 3A). Cluster 6 represented transcripts with an early increase of expression after ROSC and was standing by itself. Annotated genes were classified by functional category. The results for differentially expressed proteins after ROSC and by MB treatment, respectively, were summarized in Figure 9. A search for overrepresented functional classes within each cluster showed significant results only for the group formed by cluster 8. In this group of transcripts induced after ROSC, 6 functional classes corresponding to response to stress (p < 0.001), regulation of transferase activity (p = 0.001), regulation of protein kinase activity (p = 0.001), apoptosis/type I programmed cell death (p = 0.034), activation of mitogen-activated protein (MAP) kinase kinase kinase (p = 0.044) and programmed cell death (p = 0.044) were overrepresented. A list of the processes affected after ROSC and/or by MB (addressed in the discussion) can be found in Additional file 2.

Replication of gene expression profiles by RT-qPCR

Validation of differentially expressed genes according to the microarray results was conducted by RT-qPCR. We selected genes with different expression patterns and the results showed a good correlation with the findings from microarrays (Figure 10)

Effects of cardiac arrest/global ischemia

Interruption of blood flow after CA limits the delivery of oxygen and glucose to neurons causing ATP reduction which has numerous consequences including neurotoxicity, DNA damage, and production of free radicals [23]. In our model, we found only 28 significantly differentially expressed transcripts in total after global ischemia. Among the moderately up-regulated genes, it is noteworthy that 7 transcripts affected by ischemia belonged to ion transport and calcium signalling functions as it is well known that one of the first consequences of ischemia-induced apoptosis is a disruption of ionic gradients across membranes with increase in extracellular potassium and influx of sodium, chloride and calcium into the cells. Interestingly our list included 2 zinc transporters: SLC30A9 (ZNT9) and SLC30A10 (ZNT10). These findings are in line with an earlier study showing that another member of the SLC30 family, ZNT1, was induced after transient forebrain ischemia in the Gerbil [24]. The SLC30 proteins are CDF (cation diffusion facilitatator) proteins and plasma membrane efflux transporters [25]. The mechanism of transport for many CDF proteins appears to be via zinc/H + or K + antiports. Therefore, despite their name, CDF proteins do not serve as diffusion facilitators but rather as secondary active transporters, using the gradient of other ions to drive the transport of zinc [26]. Identification of ZNT9 and ZNT10 as ischemia-induced genes adds therefore to the growing evidence suggesting that the role of zinc efflux transporters in cellular zinc homeostasis may be particularly important in postischemic neuronal death where high levels of zinc are accumulated by neurons [27].

Effects of ROSC/reperfusion

Cardiac arrest and CPR elicit global brain ischemia and reperfusion causing immediate and delayed cell death. I/R injury results in oxidative stress leading to further generation of free radicals. This in turn activates numerous major signalling pathways, in particular through changes in gene expression which influence the survival or death of the cells.

The use of microarrays has been applied to study the effects of permanent or transient ischemia on total brain or distinct regions in rodent models but, to our knowledge, this is the first time that a genome-wide transcript profiling study is undertaken in porcine brain to study the effect of I/R. Our results showed that 141 genes were induced and only 39 were repressed after ROSC. Such an imbalance between up- and down-regulated genes could be due to a technical bias. Yet it is in agreement with a recent report where the authors monitored the genomic response to global ischemia and reperfusion in the rat brain and concluded that the ischemia/reperfusion response of the rat brain is therefore an active process [28]. A functional analysis of the clusters revealed that apoptosis/programmed cell death and stress response (34% of all induced genes) were overrepresented among the genes up-regulated after ROSC. Among them we found proteins belonging to the caspase-dependent apoptotic pathway, p53 and mitogen-activated protein kinase pathways, playing a pivotal role in stress and apoptosis. Consistent with findings in rodent, heat shock proteins are among the most prominently upregulated genes after the ischemic insult [29, 30]. Overall our gene lists are in line with previous microarray studies monitoring cerebral ischemic insult [28–34] including proteins such as ICAM-1, a main actor for the regulation of ischemia-induced injury since its inhibition induces neuroprotection after I/R injury [35, 36]; VEGFA as a main indicator for brain ischemia [37]; IFRD1, an early gene stimulating p53 and up-regulating iNOS after oxidative stress [38]; inducible HMOX1 and constitutive HMOX2, two rate-limiting enzymes for the catabolism of the prooxidant heme [39, 40] and CHOP. a main marker of ER stress [41, 42].

Effects of MB treatment

One of the responses to cerebral ischemia is an increase in the production of NO, catalyzed by nitric oxide synthase (NOS) expression. There has been increasing evidence that neuronal nitric oxide synthase (nNOS)-induced NO is an important mediator in ischemic brain injury [43]. Methylene blue (MB) exerts an antioxidant effect by binding and inactivating the action of NO [44], blocking the soluble guanylate cyclase [45] - the primary receptor for NO - as well as inhibiting Nitric Oxide Synthase (NOS) [46]. Although transcriptional levels of NOS did not show any changes after treatment by MB in our study, we have previously reported that NOS enzymes were regulated by MB at the protein level, in porcine brain, already after 30 min (Miclescu et al., submitted). These results concur with the conclusion that NOS enzymes are preferentially regulated post-transcriptionally, thus allowing a more rapid adaptation of the enzyme activity to cellular stress variations [47]. As could be expected, we identified the sGC transcript in the set of genes regulated by MB together with HSP90, a chaperone forming a complex with both sGC and endothelial nitric oxide synthase (eNOS) for the modulation of NO production [48]. Although the modulation of cGMP/NO pathway is considered the most significant effect of MB, recent studies indicate that it has multiple cellular and molecular targets.