Total RNA isolation was performed by adding 500 µL of guanidinium thiocyanate-phenol-chloroform extraction method (TRI Reagent, Sigma-Aldrich, St. Louis, MO) to samples. All specimens were homogenized by mechanical disruption using the Ultra-Turrax Mixer (IKA®) instrument and total RNA was extracted in RNAse-free environment and kept at −80°C until use. A DNase digestion step (RNase-Free DNase Set, Qiagen) was included and all RNA was ressuspended with ultrapure water (Qiagen). This ultrapure water was similarly used for all subsequent dilutions of RNA samples.

Nucleic acid concentrations were measured in a NanoDrop® ND-1000 Spectrophotometer (NanoDrop Technologies, USA). Purity of the total RNA extracted was quantified by measuring the absorbance at 230, 260 and 280 nm. Through the concentration of each sample assessed in NanoDrop®, all samples were diluted to use the same amount of total RNA (200 ng) to assess their quality. RNA quality of each sample was assessed by a microfluidic-based electrophoresis system using the Experion™ Automated Electrophoresis System (Bio-Rad Laboratories, USA), which automatically assesses the integrity of RNA samples expressed as a RNA quality indicator (RQI) number.

All RNA extracted from murine samples was retrotranscribed to cDNA using the iScript cDNA Synthesis Kit® (Bio-Rad) according to the manufacturer's protocol. For the cDNA synthesis, approximately 1 µg of each sample total RNA were used. Each 20 µL reaction mix contained: 4 µL of 5× iScript reverse transcription supermix, 1 µg of total RNA and RNase-free water to make up 20 µL. The cDNA synthesis was performed in a Mastercycler ep gradient S (Eppendorf®) with following steps: priming at 25°C for 5 min, reverse transcription at 42°C for 30 min and enzyme inactivation at 85°C for 5 min. All cDNA samples were stored at −20°C until quantitative real-time PCR (qPCR) analysis.

qPCR was performed in IQ™ 5 Real-Time PCR detection System (Bio-Rad) in 96-well plates (Bio-Rad) with a reaction volume of 20 µL and runs up to 40 cycles using iQ™ SYBER® Green Supermix. The final PCR reaction mixture of 20 µL contained 0.5 µL of cDNA sample, 10 µL of iQ™ SYBR® Green Supermix, 0.5 µL of each primer and 8.5 µL of RNase-free water. The cycling conditions were set as follows: Taq DNA polymerase activation at 95°C for 3 min, amplification steps: denaturation at 95°C for 15 s, annealing at 60°C for 15 s, and extension at 72°C for 15 s with fluorescence acquisition. The cycling conditions were equal for all genes except for the Bhmt gene, which has an annealing temperature of 64°C. Genomic DNA contamination was verified by reaction without reverse transcription in randomly tested samples, which resulted in Ct values >38. Based on literature, we have chosen for our analysis 11 genes (Table 1). Five commonly used reference genes were chosen (Actb, Gapdh, Hprt, Cyp2E1 and Ppia) as widely used stable expressed genes [16], [17]. Alb, Bhmt, Mylk and Tpm1 were chosen as chosen as being the most expressed tissue-specific genes [18]. We also included in our panel the 40S ribosomal protein S29 (Rps29) and the 72 kDa subunit of the signal recognition particle (Srp72) genes, which proved to be more stable than other commonly used reference genes [19], [20]. Details on the measurements of all this transcripts are provided in Figures S1A and S1B. All cDNA samples were measured in duplicate, and mean values of the quantification cycle Ct were used for calculations. Data were normalized according to the ΔCt model with the following formula: ΔCt = Ct _{(target gene)}−Ct _{(reference gene)}. Gene expression changes were analyzed using the built-in iQ5 Optical system software (version 2). All real time experiments were performed following the recently defined Quantitative Real-Time PCR Experiments (MIQE) guidelines [21]. Details are provided in Table S1. All primers (HPLC purified, Stabvida) were designed using Beacon Designer software (version 7.2, PREMIER Biosoft International, Palo Alto, CA) and thoroughly tested.

Quantification and expression data were statistically processed using GraphPad Prism 5. The determined p values of the statistical significance were examined using linear regression and the Pearson correlation (r). An acceptable Pearson correlation value of r>0.900 was considered for posterior analysis. Mathematical model was developed from the 11 hour kinetics, by performing regression analysis using Tool Analysis supplement of Microsoft Excel 2010 for Windows, in order to estimate the slope (m) and y-intercept (b) errors. Level of significance was always set to p<0.05.

The enormous potential of RNA technologies, together with reports of unexpectedly high stability in certain conditions, have stimulated forensic researchers to explore the RNA world [14]. Among its potentialities, time dependent RNA decay analysis was suggested to assess the PMI [7], [22]. Nevertheless, the integrity of RNA among different post mortem tissues or organs is known to behave differently [23], [24] adding a new level of complexity to such analysis. Our strategy began by analyzing the tissue-specificity of RNA integrity loss during PMI. One has to be conscious of the existence of several parameters that may influence the rate of RNA decay in post mortem organs. Environmental temperature and microbiological contamination are among the most relevant variables. Since we aimed to define a panel of RNA transcripts that significantly correlate with PMI, it was mandatory to experimentally control these two parameters. Therefore, immediately after euthanasia, samples of the 8 murine tissues; heart, lung, spleen, femoral quadriceps, liver, stomach, pancreas and skin were removed and maintained in aseptic conditions at controlled temperature (21°C). Four and twenty hours post mortem, the extracted RNA was evaluated for purity and integrity. These organs were chosen due to its high potential in forensic area and their applicability for gene expression studies using post mortem human tissue [15], [23], [25], [26], [27]. As a method for RNA quality assessment, we determined the RNA Quality Indicator (RQI) that quantifies the integrity of the RNA samples. According to the RQI values obtained, three different groups were formed (Figure S2). The first group (I) was constituted by heart, spleen and lung, which displayed the highest stability, even for longer PMI, maintaining RQI values above 6. The group II was composed by the femoral quadriceps, liver and stomach. The RNA recovered from these organs has a faster degradation rate, as observed with the time dependent decrease of the RQI values. Pancreas and skin composed the third group (III), where low RNA integrity (RQI below 4) was observed since represent ribonuclease-rich organs [28]. This confirms the tissue specificity for correlative analysis between time and RNA degradation. Based on these results, we pursued our study by performing kinetic analysis during an eleven hour period (checkpoint at every single hour) of representative organs of group I and II (heart, femoral quadriceps and liver). Accordingly some authors, to minimize the impact of RNA integrity, only RQI values above 5 are acceptable for successful and reliable qPCR quantification [26], [29], [30], which led us to exclude the third group from the study. RNA recovered from the heart demonstrated to be extremely stable post mortem, as shown by RQI values above 7 even after 11 h (Fig. 1A). This reflects the quality of this tissue for PMI assessment. Remarkably, no significant differences were observed in the RQI levels during the first 4 hours post mortem, although a time dependent decrease in RNA integrity was observed until 11 hours post mortem. Indeed, RNA from heart tissue samples showed a significant correlation (p = 0.0006) with the PMI until 11 h. In opposition to heart tissue, post mortem RNA recovered from both femoral quadriceps (Fig. 1B) and liver (Fig. 1C) immediately began to decay. Nevertheless, the loss of RNA integrity in both tissues significantly correlated with PMI (p<0.0001) with a Pearson correlation higher than 0.900.

To evaluate RNA decay over the PMI, we have defined a list of eleven genes for quantitative PCR analysis (Table 1). Based on the knowledge that gene transcription is tissue-selective and context-dependent a panel of both evolutionarily-conserved (Rps29 and Srp72) [19], [20], most expressed tissue-specific mRNA (Bhmt and Alb for the liver and Tpm1 and Mylk for the femoral quadriceps and cardiac muscle) and more ubiquitous transcripts (Actb, Gapdh, Hprt, Ppia and Cyp2E1) were chosen. The threshold cycle (Ct) value is defined as the point at which fluorescence rises above the background fluorescence. The Ct values and thus the copy number of template molecules varied among the transcripts and PMI (Table S2). To eliminate this potential bias and to control the variations in extraction or reverse transcription yield and efficiency of amplification, we have normalized all target genes against Rps29, which was chosen taking into account its high degree of stability. Indeed, we observed that the transcription of this gene showed the lowest variations in the gene expression between all PMIs and organs. As an example, the coefficient of variation (CV) for the Ct values obtained for Rps29 transcripts on all femoral quadriceps samples was lower than 1% (ranging from 25.4 to 26.0), while it varies between 2.7–5.5% for all the other analyzed transcripts. After normalization, the Pearson correlation (r) and p value were assessed for each gene within each organ (Table 2). Interestingly, no significant correlation was found between any analyzed transcript in heart samples and PMI. The ΔCt values were found constant for all genes in all PMIs (Table S3), although the heart RNA integrity was found to linearly decrease with time (Fig. 1A). On the other hand, the decay of several femoral quadriceps and liver transcripts were found to positively correlate with the PMI (Table 2) and ΔCt values showed greater variation (Tables S4, S5, S6, S7). Aiming to assure the maximum significance between gene decay and PMI only those genes with Pearson correlations higher than 0.900 and found to be statistically significant were considered for posterior analysis (Fig. 2 and represented in bold in Table 2). This approach allowed the identification of 4 genes (Actb, Gapdh, Ppia and Srp72) in the femoral quadriceps and two genes (Alb and Cyp2E1) in the liver as the most reliable dependent variables that correlate with PMI. The need for method uniformization and systematization, accompanied with narrow confidence intervals led us to focus the construction of a mathematical model on the transcripts identified in the femoral quadriceps, which displayed the larger number of correlational variables.

The ultimate goal of this study was to develop an abstract model using mathematical language to describe the behavior of gene transcripts decay over PMI. In that sense, we have performed regression analysis with the objective of building descriptive mathematical laws to be used by forensic experts for valid and reliable determination of the PMI. The linearity of the method was determined by evaluation of the regression curve (ΔCt versus time) and expressed by the correlation coefficient (r). We used three independent calibration curves (y = mx+b) for each of the 4 selected genes to obtain the mean slopes (m) and y-intercept (b). Linearity was accepted if r≥0.900. The goal of linear regression was to determine the best estimates for the slope and y-intercept. This was accomplished by minimizing the residual error between the experimental y values, and those values predicted by regression line equation. The most commonly used form of linear regression is based on three assumptions: (1) that any difference between the experimental data and the calculated regression line is due to indeterminate errors affecting the y values, (2) that these indeterminate errors are normally distributed, and (3) that the indeterminate errors in y do not depend on the value of x. The derivation of equations for calculating the estimated slope and y-intercept can be found elsewhere [31]. This allows us to define a regression equation to calculate the PMI (Fig. 3A). Once known, it is possible to determine the PMI and to estimate the error associated to time. For that, we measured an average signal for our sample () and use it to calculate the value of x. The standard deviation for the calculated value of x (S_x) is given by the equation in Fig. 3B, where K is the number of replicate samples (K = 15) used to establish , n is the number of measured endpoint times (n = 12), is the average signal for each endpoint time, is the summation of individual (x−)². Once S_x is known, the confidence interval for the PMI can be calculated attending the value of t determined by the desired level of confidence (p<0.05) for n−2 degrees of freedom. Figure S3 evidence the interval of confidence for each sampled endpoint and it is observed that similar results were obtained during all sample period, giving more relevance to our results. The highest error was registered for time zero, which is explained by inherent difficulty to collect samples instantaneously.

As explained above, our model was tested using controlled environmental conditions. We decided to challenge our system by sampling femoral quadriceps tissues from intact mouse corpses at 1, 4 and 10 h after death. RQI values ≥7.9 were obtained. The PMI was calculated using the equation shown in Fig. 3A, substituting the by the average signal of the replicates for each time point (K = 5) and considering ΔCt values for all 4 transcripts (Actb, Gapdh, Ppia and Srp72). Remarkably, the medium PMI values obtained were 1.90±0.01, 4.10±0.87 and 9.80±1.87 hours, demonstrating the high predictive power of our mathematical model.

Overall, we observed a time and organ dependent decrease of the RQI values. A significant correlation was found between total RNA isolated from the heart, femoral quadriceps or liver samples and the PMI. A quantitative analysis of several transcripts on these organs led to the identification of 4 genes that when analyzed on a quadriceps specimens correlate with PMI. Furthermore, we developed a mathematical model to estimate the PMI with an error mean of 51.4 minutes. We have successfully challenged our method using femoral quadriceps samples recovered from intact mouse corpses. This study may represent a new paradigm to estimate PMI and become a complementary tool for traditional methods, with the ultimate goal to increase the accuracy of the PMI estimation. The suitability of these strategies for post mortem human autopsy material as well as the influence of the various parameters (e.g. cause of death, age at death, gender and body mass index, duration of agony, storage conditions of the body) will be addressed in future studies.