Agarose gel electrophoresis of PCR products during the initial optimization experiments revealed single bands for all primer sets (not shown). Moreover, melting curve analysis was performed after each run. This always demonstrated a single homogenous melt peak, confirming specific amplification. The melting temperatures of all PCR products are given in Tab. 1. Since PCR products of identical size and melting temperature may also arise due to the existence of processed pseudogenes on the genomic DNA 30, which potentially contaminates the samples, additional control experiments were performed. Indeed, several primer pairs (Gapdh, Actb, CypA, Pgk1 and Rpl13a) resulted in amplification products of equal size and melting temperature with genomic DNA as template, whereas the other primer pairs did not yield any signal under these conditions. Therefore, though all primer pairs (except for Actb) span introns, minus RT controls were performed for the respective primer pairs. No signals were detected in the minus RT controls. Calibration curves were generated using relative concentrations vs. the threshold cycle (Ct). The RSq value (R², linear correlation coefficient), an indicator of fit for the standard curve plotted to the standard data points of all genes ranged from 0.995 to 1.000. Based on the slopes of the standard curves, the amplification efficiencies ranged from 92 % to 105 % (Tab. 1), (derived from the formula PCR efficiency = 10 ^1/slope -1, calculated by the Mx3005P software). Efficiencies higher than 100 % may result from this calculation method, which is an overestimate of the "real efficiency" 31.

GeNorm was used to identify the most stable reference gene for each of the three analyzed time points separately. Fig. 1 shows the average expression stability values M of the remaining control genes. GeNorm identified CypA and Pgk1 as the most stable pairwise combination of reference genes for the experimental groups four days after ACA treatment (M value for combination of best two genes 0.145), CypA and Gapdh for the groups 7 days after the insult (M value for combination of best two genes 0.097) and CypA and Pgk1 for the 21 day time point (M value for combination of best two genes 0.141). Interestingly, for all the analyzed time points B2m shows the highest M value (generated using all ten genes 0.454, 4 days; 0.531, 7 days; 0.651, 21 days). Moreover, we analyzed the pairwise variation values (V) between two sequential normalization factors containing an increasing number of genes (Fig. 2). The pairwise variation value of V2/3 is only 0.046 for the experimental group 4 days after ACA, 0.045 for the 7 days experimental groups and 0.049 for the 21 days groups. According to Vandesompele et al. 19 the ideal pairwise variation value is less than 0.15. Thus, although including further reference genes the V value further decreases, there is no need to include more than two genes into the normalization factor, because this would not improve normalization dramatically.

When NormFinder was used to analyze the same data set a slightly different order of gene stability was calculated (Tab. 3). More stable gene expression is indicated by lower average expression stability values. For the experimental groups 4 days after ACA treatment Actb was identified as the most stable gene with a stability value of 0.035. For the groups 7 days after ACA Gapdh was calculated to be the most stable gene with a stability value of 0.032 and for 21 days Tbp with 0.027. Interestingly, B2m was identified to be the most unstable gene with stability values of 0.132, 0.117 and 0.126 at 4 days, 7 days and 21 days after ACA, respectively.

After having identified the most stable combination of two reference genes by geNorm, we calculated the expression of the other genes for each of the three different time points. Data were normalized to the normalization factor calculated by geNorm, taking into account the combination of the two best performing housekeeping genes. Most of the analyzed candidate reference genes are not significantly changed by either of the treatments at any of the three time points (Fig. 3).

In contrast, B2m transcript, which was identified as the most unstable gene by geNorm as well as by NormFinder, is 1.7-fold upregulated in the ACA group and twofold upregulated in the ACA with minocycline treatment group 4 days after treatment compared to the sham treated group. At 21 days after treatment there is a 2.4 fold upregulation in B2m expression in the ACA with minocycline treatment group vs. control. Though not significant, there is a trend for B2m mRNA upregulation at 7 days after treatment too (Fig. 4).

Using immunohistochemistry, NeuN staining is significantly reduced due to ACA compared to sham-operated animals (Fig. 5A, B). This reflects the neuronal damage within the CA1 and dentate gyrus regions. The neuronal loss is paralleled by massive microgliosis, as demonstrated by OX42 immunostaining (Fig. 5C, D).

Moreover, in order to evaluate our experimental paradigm also via qRT-PCR, we analyzed the expression of the microglial marker Mac-1 (synonymous names Cd11b or Itgam). This gene is constitutively expressed by resting microglia as well as by macrophages and is known to be upregulated upon microglia activation 32. We found an about 10- to 20-fold upregulation of Mac-1 four days after the ACA insult which persisted during the later time points analyzed (Fig. 6).

All animals were maintained in accordance with the guidelines of the German Animal Welfare Act. The study was approved by the Animal Care and Use Committees of the State of Saxony-Anhalt and the University of Magdeburg under the permit number G/1/06. The animals were housed under temperature -controlled conditions at 21 ± 1°C, a 12 h light/dark cycle, and free access to standard rat chow (Altromin 1324™, Altromin GmbH, Lage, Germany) and water.

Sixty age-matched (15 weeks, 300–350 g) and strain-matched (Wistar, inbred, Harlan-Winkelmann; Borchen, Germany) male rats were divided into four treatment groups: (i) asphyxial cardiac arrest (ACA); (ii) ACA with minocycline treatment; (iii) sham operated; (iv) sham operated, with minocycline treatment. Three different survival times were analyzed per group: 4 days, 7 days and 21 days. Thus, 5 animals were used per treatment and time point.

The surgery protocol has been described in detail previously by Ebmeyer et al. 3 with the slight modification that resuscitation was started exactly after 5 min of asphyxiation. During preparation, the insult and the first hour post return of spontaneous circulation (ROSC) body temperature was controlled and maintained at 37°C. Rats were then further kept at normal body temperature (37°C) by placing them in an incubator cage for 24 h post-resuscitation. Blood pressure values were measured according to the protocol. The mean artery pressure (MAP) values were similar in all animals included in the study and the course over time closely resembled the published data for intra-insult temperature-controlled Wistar rats 3. Animals that showed critical low blood pressure levels (MAP < 50 mmHg) were excluded from further analysis.

Minocycline hydrochloride (Sigma, St. Louis, USA) was administered twice a day, i.p., at a dosage of 6 mg every 12 hours for a period of 5 days. First minocycline injection was performed 60 min after ROSC. The drug was dissolved in PBS at a concentration of 6 mg/ml. Between the injections the solution was stored at 6°C for five days. The volume to be injected was adapted to room temperature before the injection. Sham operated animals received a PBS injection (1 ml).

Rats were sacrificed by decapitation during deep anesthesia. Brains were quickly removed and, for qRT-PCR, hippocampus was dissected on ice, immediately frozen on dry ice and stored at -80°C until RNA extraction.

For immunohistochemistry, brains were quickly removed from the cranium, postfixed in 4 % buffered paraformaldehyde (pH 7.4) at 4°C overnight, cryoprotected in a solution of 30 % sucrose (Merck) in 0.4 % buffered paraformaldehyde (pH 7.4) for 2 days and rapidly frozen at -20°C using 2-methylbutan (Roth, Karlsruhe, Germany). Serial sagittal sections (20 μm thick) were cut on a cryostat (Jung Frigocut 2800 E, Leica, Bensheim, Germany). Free-floating sections were washed and incubated with monoclonal mouse anti-NeuN (Chemicon, Temecula, CA, 1:100), polyclonal rabbit anti-MAP 2 (microtubule-associated protein 2, Chemicon 1:2.500), or monoclonal mouse anti-CD 11b/c (OX42, PharMingen, Hamburg, Germany, 1:800) in PBS with 0.3 % Triton X-100 and 1 % normal goat serum overnight at 4°C. Following incubation with primary antibodies, slices were washed in PBS (3 × 5 min), and incubated overnight with the respective secondary antibodies (1:500, goat anti-mouse-IgG Alexa Fluor 488, goat anti-rabbit-IgG Alexa Fluor 488, goat anti-mouse-IgG Alexa Fluor 546, Molecular Probes, Göttingen, Germany), mounted and examined on a fluorescence microscope (Axiophot, Zeiss). Control reactions (substitution of the primary antisera by phosphate buffered saline) yielded negative results (i.e. no specific immunostaining was seen in these sections).

Candidate reference genes were selected from those most commonly used in the literature. To reduce the chance that these genes might be coregulated, ten genes belonging to different functional classes were selected. Primers for Actb, CypA and Hprt were as published by Peinnequin et al. 46. Mac-1 primers were as published by Raghavendra et al. 47. All other primers were designed by the primer 3 software 48 based on rat sequences in the database 49. Here primers were chosen, that bind to different exons in order to avoid non-specific product formation from potentially contaminating genomic DNA. The specificity of the primers was checked using a BLAST search against nucleotide collection (nr) of the NCBI database.

All primers were synthesized by MWG Biotech (Ebersberg, Germany). The sequences of primers are listed in Tab. 1.

Hippocampal tissue was homogenized in peqGOLD TriFast (Total RNA Isolation Reagent, PeqLab, Erlangen, Germany) using an Ultra-Turrax Homogenizer (IKA Labortechnik, Staufen i. Br., Germany). Total RNA was prepared according to the manufacturers instructions. The concentration of total RNA was determined by measuring the optical density at 260 nm and the purity was checked as the 260 nm/280 nm ratio with expected values between 1.8 and 2.0. The integrity of total RNA was assessed by electrophoresis on 1.2 % (w/v) agarose gels. To remove potential DNA contamination, the RNA samples were DNase treated (TURBO DNA-free Kit, Ambion, Austin, USA). First strand cDNA was prepared from 1 μg DNase treated total RNA in a total volume of 20 μl using the RevertAid First Strand cDNA Synthesis Kit (Fermentas, St. Leon-Roth, Germany). Oligo(dT)₁₈ primers were used and all samples were stored at -80°C until further analysis. Additionally, cDNA was prepared from a mixed RNA sample (later referred to as pool cDNA) for each of the tree time points. For this purpose equal amounts of DNase treated total RNA of all twenty samples of each time point were mixed. Moreover, minus RT controls were prepared for each sample using the identical procedure except for the omission of the enzyme.

The real-time PCR was performed using the MX3005P device (Stratagene, La Jolla, CA, USA). The reactions contained 1× Brilliant SYBR Green QPCR Master Mix (Stratagene), 30 nM ROX reference dye, each primer at 200 nM and prediluted cDNA (according to 10 ng total RNA) in a 25 μl reaction. After an initial denaturation step at 95°C for 10 min amplification was performed with 40 cycles of denaturation at 95°C for 30 s, annealing at 60°C for 40 s and extension at 72°C for 40 s. Amplification was followed by a melting curve analysis to confirm PCR product specificity.

No signals were detected in no-template controls. Minus RT controls were run for those primer pairs, that have been shown to give a PCR product of same size and melting temperature in initial test experiments with genomic DNA (0.1–50 ng/well). No signals were detected in the minus RT controls.

The experimental threshold (Ct) was calculated using the algorithm enhancements provided by the MxPro Mx3005P v3.00 software: amplification based threshold, adaptive baseline, moving average. All samples were run in duplicate and the mean value of each duplicate was used for all further calculations.

During optimization of the protocol PCR products were loaded on 2 % agarose gels to confirm specificity of amplification and the absence of primer dimer formation. PCR efficiencies were estimated by running standard curves with the above described pool cDNA (5 points, cDNA amount between 25 ng and 0.04 ng according to initial total RNA concentration).

Two publicly available software tools, geNorm 19 and NormFinder 11 were used to analyze gene expression stability. Both tools require the transformation of Ct values to linear scale expression quantities. The average Ct values of the duplicates were therefore exported into Microsoft Excel from the Mx3005P software. Ct values were converted into relative quantities (Q) via the delta-Ct method by the formula Q = (E)^dCt, with dCt = Ct of the highest abundant sample-Ct of the sample and E (efficiency) as determined by linear regression (calculated by the Mx3005P software). The quantities were then imported into geNorm software 50 (version 3.4), which was used as described in its manual. Twenty data points were used per time point (five biological replicates per treatment). Each of the three time points was analyzed separately. GeNorm calculates a gene-stability measure M, which is the average pairwise variation of a particular gene with all other control genes. Genes are ranked according to the determined M value from the least stable (highest M value) to the most stable (lowest M value).

Moreover, the calculated quantities were entered into a second software tool, NormFinder 51. NormFinder estimates the overall expression variation of the candidate normalization genes and the variation between sample subgroups of the sample set using a model-based approach 11. According to the resulting stability value the candidate reference genes can be ranked based on their expression stability.

A workflow diagram schematically depicting the process of stable gene selection is provided (Additional file 1).

The relative expression software tool (REST^©) 5253 was adopted to calculate expression ratios. This software allows for a group-wise comparison of expression differences. Moreover, the expression ratio results are tested for significance by a Pair Wise Fixed Reallocation Randomisation Test ^©. Out of the several software versions, REST-MCS was chosen, because this allows the comparison of up to six experimental conditions against one reference condition for up to ten samples per group. Each of the three time points was analyzed separately. The experimental group (iii), i.e. sham operated, served as the reference condition. The two most stable genes, as calculated in advance by geNorm, were set as reference genes. The absolute gene regulation values along with the corresponding standard error, calculated by the REST-MCS software, were then forwarded to Microsoft Excel to create the graph presented in Fig. 3. For comparison, the reference condition, i.e. sham operated, was assigned a value of 1.

Ct values were converted into relative quantities (Q) via the delta-Ct method by the formula Q = (E)^dCt, with dCt = Ct of the highest abundant sample-Ct of the sample and E (efficiency). Relative quantities were normalized by dividing by the normalization factor calculated by geNorm from the two most stable genes. Normalized relative quantities were then rescaled by dividing by the arithmetic mean of the normalized relative quantities of the control group (sham w/o). The calculated values were imported into Prism 4 program (GraphPad Software, Inc., San Diego, CA) to generate the scatterplots presented in Fig. 4 and Fig. 6.