Postmortem proteomics to discover biomarkers for forensic PMI estimation

To evaluate the relevance of new marker proteins for time since death estimation in humans, we additionally analyzed thigh muscle samples from humans using SDS-PAGE and Western blotting. Muscle samples were gathered from routine autopsy cases at the Department of Forensic Medicine of the University of Salzburg, Austria. Only cases without prior history of muscle diseases and cooling (4 °C) times under 24 h were included. Sample H1 was taken from a 76-year-old woman who died on respiratory failure in a hospital (Table 1). Intermediate PMI is represented by a sample (H2) from a 75-year-old female who was strangled in bed and was found 2 days later (Table 1). H3 was taken from a 67-year-old man who was found dead in his apartment. He showed signs of advanced putrefaction. PMI could only be roughly estimated by non-biomedical evidence through police investigation (Table 1). Accumulated degree days (ADD), as a measure of energy affecting a system, were calculated as follows: ADD [°d] = time [d] × temperature [°C]. The presented cases remained in 4 °C cooling environment for 13.9 ± 3.0 h.

In all cases, skin, fat, and fasciae were opened by a small incision at the lateral thigh and a biopsy sample of the vastus lateralis muscles was taken from 4 to 8 cm depth (approximately half the distance to the femur, depending on individual physique). Samples were snap frozen and stored in liquid nitrogen until further processing. Sampling of human muscle tissue for scientific purposes was approved by the ethics commission of the University of Salzburg (EK-GZ: 11/2017).

The protein extracts were prepared with RIPA buffer (details of products and specific solutions are given in Online Resource 1) followed by sonication. Samples were desalted with Zeba Spin Desalting Columns and denatured in 5 M urea and 5 mM DTT (dithiothreitol) at 65 °C for 30 min and then reduced in 40 mM iodoacetamide at room temperature for 30 min. The lysates were again desalted by Zeba Spin Desalting Columns and digested with 20 μg/mL of trypsin for 2 h at 37 °C. The tryptic peptides were quantitated using Quantitative Colorimetric Peptide Assay kit. Twenty micrograms of tryptic peptides were purified by ziptip and resuspended in 20 μL of injection buffer containing 2% acetonitrile and 0.1% trifluoroacetic acid (TFA). Peptides were analyzed using a LC-MS/MS system consisting of an Easy-nLC 1200 and an Orbitrap Fusion Lumos mass spectrometer equipped with a nano-electrospray source. Protein quantitation was performed using MaxQuant Software (ver. 1.2.2.5) [24] as described previously [25]. The intensities of each protein were normalized (log2 scale) based on total intensity employing normalizer tool [26], and the normalized protein list was further processed by R and MATLAB software. For additional details including data analysis procedure refer to Online Resource 2.

Antibody binding was visualized by adding chemiluminescence substrate and documented with a digital gel analysis system. Protein band intensities were measured using ImageJ software. Bands in the 0-hpm samples were considered the native form, and all alterations (disappearance of the native bands, or appearance of additional bands) were considered degradation processes. All signals < 5% the intensity of the native bands at 0 hpm were considered background and thus no band.

For Western blot analysis, Spearman correlation values (Spearman’s ρ, p value) were calculated for each band to assess statistically significant band changes within the investigated PMI. Kolmogorov-Smirnov tests and visual checks for normal distribution and homogeneity of variance were applied to calculated data of relative GAPDH and eEF1A2 intensities of proteomics and Western blot approaches. In case of normally distributed data, one-way analysis of variance (ANOVA) was used to test for significant differences between 0-hpm group and postmortem groups. Significance was determined by Tukey’s post hoc multiple comparisons tests. In case the data failed to satisfy the assumptions of parametric homoscedastic additive models, differences in measured variables between groups were tested by nonparametric methods. Intergroup differences were tested by the one-way rank-based analog to analysis of variance (Kruskal-Wallis test), followed by all pairwise multiple comparisons by the Bonferroni correction method. For all analyses, SPSS 23.0 software (IBM Corp) was used. p values ≤ 0.05 were considered statistically significant, p values ≤0.001 were considered highly significant.

The number of protein markers currently used for forensic PMI estimation is still limited, thus, identification of additional protein markers is necessary to facilitate future application of this approach. To discover additional protein biomarkers, we analyzed postmortem alterations in skeletal muscle proteome harnessing mass spectrometry (MS)–based proteomics.

By that, we identified and quantified a total of 896 and 579 proteins from mouse and rat, respectively. An unsupervised hierarchical clustering was conducted to classify the protein intensities, resulting in several clusters in both mouse and rat cases (Fig. 1a). We focused on identifying proteins whose quantity consistently decreased (from each time point to the subsequent) as hours postmortem (hpm) increased. By that, we identified 36 and 10 consistently decreasing proteins, from mouse and rat skeletal muscle proteome, respectively. We found that GAPDH and eEF1A2 were commonly decreasing proteins in both mouse and rat (Fig. 1b). When normalizing the protein intensities to that of the 0-hpm samples in both mice, GAPDH abundance slightly decreased to 95.4% from the initial level at 24 hpm. At 48 and 72 hpm, the intensities declined by more than 80%, reaching 19.5 and 18.5% of the initial protein abundance. At 96 hpm, only 6.3% of the initial concentration was measured. In both rats, the rate of GAPDH reduction was largely consistent over time. GAPDH abundance decreased by a mean of 29.9% at 24 hpm. At 48 hpm, the concentration dropped to 59.1% from the initial value, at 72 hpm to 35.2% and reached 22.7% from the initial level at 96 hpm (Fig. 1c). The reduction pattern of eEF1A2 was similar between mouse and rat at 24 hpm decreasing by 39.2% and 33.6%, respectively. In mice, the eEF1A2 abundance declined sharply to 26.8% from the initial protein level at 48 hpm and continued to decrease to 16.9% (72 hpm) and 14.9% (96 hpm). In rats, the eEF1A2 protein decrease was continuous from 66.45% at 24 hpm to 56.5% at 48 hpm and the further decreased levels remained roughly constant at 27.9% and 24.3% at 72 hpm and 96 hpm, respectively (Fig. 1d). This decreasing trend of both proteins did not apply for statistical analysis due to the small sample size. However, to prove the applicability of those two proteins as a biomarker, we examined the postmortem alterations in another experimental setting using a larger sample size.

To validate postmortem degradation of GAPDH and eEF1A2 and to compare the degradation patterns to that of other well-established proteins on behalf of postmortem degradation (tropomyosin, desmin, vinculin), we ran standard Western blot analysis on a larger sample set, as described earlier [15, 27]. Tropomyosin western blots resulted in characteristic double bands in all animals and time points investigated (Fig. 2a). Desmin depicted a distinct single native band at approximately 50 kDa in all 0-hpm samples (Fig. 2b). In samples with PMIs between 24 and 72 hpm, the presence frequency decreased and at 96 hpm no native band was detectable in any of the samples. This represented a highly significant correlation with the time since death (ρ = 0.510, p = 0.009) (Fig. 3a). In addition to the gradual disappearance of the native band, all samples with a PMI of 24 hpm or more depicted characteristic degradation products. One band at approximately 41 kDa was present in all samples at 24 and 48 hpm, but gradually decreased in frequency in later time points. This transient appearance cannot be sensibly described using Spearman correlation. A second desmin degradation product with a molecular weight of approximately 38 kDa appeared initially in some of the 48-hpm samples and increased its occurrence frequency until 96 hpm, representing a significant correlation with the time since death (ρ = 0.495, p = 0.012) (Fig. 3a). Similarly, analysis of vinculin resulted in distinct native bands at approximately 117 and 135 kDa in all of the 0-hpm samples (Fig. 2c). These bands were recognized to represent the native form of vinculin (approximately 117 kDa) and the slightly larger splice variant meta-vinculin (approximately 135 kDa). Only few of the samples depicted a loss of the native vinculin band at 96 hpm. However, abundance of meta-vinculin bands significantly decreased, starting at 24 hpm and none of the samples collected from animals after 72 hpm depicted this band anymore. Spearman’s ρ for this protein band was very high and the PMI correlation highly significant (ρ = 0.766, p < 0.001) (Fig. 3a). Similar values were obtained for the appearance of two vinculin degradation products. While one degradation product at approximately 84 kDa was present in all samples with a PMI of 24 hpm and more (ρ = 0.707, p < 0.001), the second band at approximately 75 kDa gradually appeared in samples with PMIs of 24 to 72 hpm and was present in all animals at 96 hpm (ρ = 0.648, p < 0.001) (Fig. 3a).

Analysis of postmortem GAPDH band patterns resulted in distinct single bands at approximately 40 kDa (Fig. 4a). Bands with an intensity of > 5% of the native band were detected in all tested animals within the investigated time period of 96 hpm, and thus no correlation with the PMI could be calculated using these strict standard criteria of a significant qualitative change. However, there was a decrease of band intensity already at 24 hpm (mean 59.8% ± 2.9% of the native band) (Fig. 3b). This trend continued throughout the investigated period of 96 hpm and reached significance levels at 48 and 96 hpm respectively (p = 0.012 and p = 0.016). At 96 hpm, the mean intensity of the native GAPDH band was 42.9 ± 13.8% of the native band at 0 hpm.

eEF1A2 was detected as a distinct single band at approximately 50 kDa in all of the 0-hpm samples, and in significantly lower intensity in all 24-hpm samples (p < 0.001) (Fig. 3b, Fig. 4b). While in the 48-hpm samples, two of the five animals (40%) depicted bands above the threshold value of 5% of the native band (significantly decreased intensity, p < 0.001), no signal was detected in the other 48-hpm samples and in any of the 72- and 96-hpm samples, representing a highly significant correlation with the PMI (p < 0.001) with a Spearman ρ of 0.849 (Fig. 3a).

Similar to the animal model, we used the standard proteins tropomyosin, desmin, and vinculin to compare and validate postmortem degradation of GAPDH and eEF1A2 in human cases. Analyses of tropomyosin resulted in double bands in H1 and H2. In H3, tropomyosin bands were lacking (Fig. 2d). Western blot analysis of desmin showed a characteristic native band with a molecular weight of approximately 50 kDa in H1 and H2 (Fig. 2e). H2, in addition, depicted characteristic desmin degradation products of approximately 41 kDa and 38 kDa. In H3, only the 38 kDa band was present. Analysis of vinculin resulted in the native form of the protein in H1 and H2 (approximately 117 kDa) (Fig. 2f). In H1, additionally the higher molecular weight band was present (meta-vinculin, approximately 135 kDa). Both H1 and H2 samples also depicted protein bands at approximately 90 kDa and 84 kDa. Analysis of H3 revealed bands at approximately 84 kDa, 75 kDa, and 63 kDa.

To test the relevance of application in forensic cases, we analyzed the expression of GAPDH and eEF1A2 in three human cases representing different PMIs (Table 1). GAPDH protein patterns depicted distinct native bands in all three cases without major differences in intensity (Fig. 4c), confirming general applicability of this specific antiserum for human muscle tissue. In H3, several distinct degradation products of molecular weights between approximately 37 and 35 kDa were detected in addition to the native band. When analyzing postmortem eEF1A2, a distinct single band in H1 and H2 was detected (Fig. 4d), proving general applicability of this specific antiserum for forensic analysis. Notably, there was no obvious difference in band intensity detectable between both of these cases. However, in H3, no eEF1A2 band signal was detectable, suggesting similar processes in humans compared with the results from proteomic analysis and Western blot experiments in the animal model.