Identification of proteins from 4200-year-old skin and muscle tissue biopsies from ancient Egyptian mummies of the first intermediate period shows evidence of acute inflammation and severe immune response

Abstract

We performed proteomics analysis on four skin and one muscle tissue samples taken from three ancient Egyptian mummies of the first intermediate period, approximately 4200 years old. The mummies were first dated by radiocarbon dating of the accompany-\break ing textiles, and morphologically examined by scanning electron microscopy of additional skin samples. Proteins were extracted, separated on SDS–PAGE (sodium dodecyl sulfate polyacrylamide gel electrophoresis) gels, and in-gel digested with trypsin. The resulting peptides were analysed using nanoflow high-performance liquid chromatography–mass spectrometry. We identified a total of 230 unique proteins from the five samples, which consisted of 132 unique protein identifications. We found a large number of collagens, which was confirmed by our microscopy data, and is in agreement with previous studies showing that collagens are very long-lived. As expected, we also found a large number of keratins. We identified numerous proteins that provide evidence of activation of the innate immunity system in two of the mummies, one of which also contained proteins indicating severe tissue inflammation, possibly indicative of an infection that we can speculate may have been related to the cause of death.

This article is part of the themed issue ‘Quantitative mass spectrometry’.

1. Background

The use of the polymerase chain reaction (PCR) in forensic archaeological research has produced a large amount of highly informative data and results. However, it is not without limitations [1]. One of the main issues is that DNA is degraded at least 10 times faster than protein [2]. Hence, remains of ancient mammoths and mastodons, which are typically 5000–100 000 years old, can be analysed for the presence of either DNA or protein [3–5]. However, no detectable amount of DNA remains in dinosaur fossils, which are tens of millions of years old, so they are not amenable to analysis using PCR—despite the prevailing belief in popular culture. Proteins, in contrast, have been detected in numerous dinosaur fossils dated at 60–70 million years old [6–10], although these results are not without controversy [11].

The amplification process used in PCR analysis is a double-edged sword. When it is done correctly and all works well, it is possible to amplify measurable amounts of DNA from vanishingly small amounts of starting material [12]. However, the question of contamination looms large, especially in ancient materials. A small amount of contaminating DNA may be amplified just as readily, which can lead to erroneous results [13]. The analysis of ancient proteins using proteomics techniques does not rely on amplification. This means that significantly more protein material needs to be present in a tissue sample in order to be detected. It is only in recent years that the sensitivity of mass spectrometric (MS) detection of proteins has improved to such an extent that it is now possible to analyse such small amounts. The issue of contamination is also present in shotgun proteomics analysis of ancient samples [14], but in a different way. Materials extracted from ancient burial sites may well have been contaminated by handling by people involved in the sample collection process, and we have no way of controlling for that.

In recent years, forensic molecular analysis techniques involving both DNA and protein have been used in a variety of investigations. These have uncovered some very important evidence that has helped to reveal important archaeological insights. Examples include the confirmation of the presence of Yersinia pestis in suspected mediaeval plague victims [15–17], and the discovery of evidence of Mycobacterium tuberculosis in Late Neolithic–Early Copper Age skeletal human remains from Hungary [18], and tissue samples from 500-year-old mummified Inca remains [19].

One very famous set of mummified remains discovered in more recent times is that of the 5300-year-old Tyrolean Iceman Oetzi found trapped in a glacier in 1991. This well-preserved specimen has been the subject of numerous studies, including: analysis of furs from his accoutrements, which were found to come from sheep, goat, cattle, deer, chamois, wolf or fox and possibly bear [20,21]; detailed study of the brain proteome [22]; detection of Treponema denticola in a tissue biopsy [23]; and a complete genome sequence for an ancient isolate of Helicobacter pylori [24]. These studies involved application of a variety of techniques, including gel-based and liquid chromatography-based proteomics approaches, extraction of ancient DNA, PCR amplification of selected target molecules and next-generation genome and metagenome sequencing technologies.

Great progress has been made in recent years on using MS analysis of ancient proteins in an archaeological context. Much of this work involves analysis of collagens in animal bone material, which provides a naturally protective matrix to ensure against degradation [25]. Recent studies, for example, used this technique to conduct a phylogenetic tree of ancient South American ungulates and helped resolve one of the unanswered questions from Darwin's work in the area [26,27]. Phylogenetic analysis based on collagen protein sequences was also used in a study involving fossilized bone material from ancient extinct giant ground sloths [28]. In-depth analysis of collagen sequences has also been used to identify hominin bones in the context of highly fragmentary archaeological assemblages [29]. A somewhat different paradigm was employed in a study involving analysis of tissue samples from well-preserved Andean mummies [19]. The authors used unbiased shotgun proteomic analysis to uncover evidence that appeared to indicate a severe pulmonary bacterial infection, and confirmed this by detection of Mycobacterium using PCR. Archaeologists are not the only people interested in ancient Egypt; the general public also have an enduring fascination with ancient Egyptian culture, especially the iconic pyramids and mummies. Scientific investigations into subjects such as the harem conspiracy surrounding the Pharaoh Ramesses III [30] or the life and death of Tutankhamen [31] invariably also find their way into the popular press. Some of the groundbreaking work in this area has involved the use of gas chromatography–mass spectrometry (GC-MS) for detailed chemical analysis, which reveals much about the nature of the resins used in the embalming materials used in mummification [32,33]. This approach has led to a substantial revision of the previously estimated earliest dates of mummification in ancient Egyptian culture [34].

The foundation of this study was the availability of a small number of very small skin and muscle tissue samples from mummies from the first intermediate period (FIP; ca 2181–2055 BC), approximately 4200 years old. The mummies originated from Assiut and Gebelein, two provincial sites in Middle Egypt. The necropolis of Assiut is situated approximately 375 km south of Cairo, whereas Gebelein is 28 km south of modern Luxor, on the west bank of the Nile.

We set out to determine if it was possible to detect measurable amounts of proteins from these materials using state-of-the-art high-resolution liquid chromatography–mass spectrometry (LC-MS) techniques. We undertook this as a pilot study aimed at determining which, if any, proteins could be reliably identified in mummified skin and muscle tissue of this age. We first characterized the mummies by morphological and radiocarbon examination, and then employed a quantitative MS approach, which is based on many previous studies in our laboratory examining changes in the proteome of cells in response to the imposition of external stresses. In those studies, we have employed label-free quantitation using spectral counting based on normalized spectral abundance factors (NSAF) [35], and performed high-stringency protein identification using an approach we developed based on reproducibility of peptide identification across biological replicate samples. In this case, we did not have the luxury of biological replicate samples owing to the very limited amount of sample material available. Instead, we relied on triplicate technical replicates, and used a spectral counting approach to determine which proteins could be reproducibly identified. We started with a set of five tissue samples for proteomic analysis, consisting of four skin biopsies and one muscle biopsy, which allowed us to perform relative quantitative comparisons between them. We were able to find large amounts of collagens and some keratins present, which is consistent with previous literature. However, we were also able to detect numerous other proteins at significant levels that were suggestive of the presence of acute inflammation and systemic immune response. Our results show that it is indeed feasible to identify proteins from extremely limited amounts of mummified skin and muscle samples that are over 4000 years old.

2. Material and methods

(a). Sampling

Tissue samples were collected from three mummies stored in the permanent collection of the Museo Egizio in Turin, Italy, in 2014, as follows.

Adult unknown mummy

Adult mummy, sex unknown, S.16742 Gebelein, (FIP) ca 2160–2055 BC.

Proteomics samples

Sample 1. Skin biopsy from chest.

Sample 2. Skin biopsy from tibia.

Microscopy sample

Skin biopsy from the right humerus

Radiocarbon dating sample

Textile collected from outer textile, at feet area.

Female mummy known as Khepeshet

Female adult mummy (known as Khepeshet), S.14381/1 RCGE 52218, Assiut, (late FIP) ca 2160–2055 BC.

Proteomics sample

Sample 3. Skin biopsy from right hand.

Microscopy sample

Skin biopsy from right zygomatic.

Radiocarbon dating sample

Textile collected from outer shroud, at hip area.

Male (uncertain) mummy known as Idi

Male (uncertain) adult mummy (known as Idi), S.14391/2 RCGE 52090, Assiut, (FIP) ca 2160–2055 BC.

Proteomics samples

Sample 4. Skin biopsy from right shoulder/rib.

Sample 5. Muscle biopsy from right shoulder/rib.

Microscopy sample

Skin biopsy from right rib cage.

Radiocarbon dating sample

Textile collected from outer shroud, at chest area.

(b). Scanning electron microscopy

The scanning electron microscopy (SEM) analysis was performed according to established methods [36,37]. Briefly, the samples were pre-fixed in 2% paraformaldehyde/2.5% glutaraldehyde in neutral 0.1 M phosphate buffer. They were then post-fixed for 2 h in 1% (w/v) osmic acid dissolved in phosphate-buffered saline (PBS). The samples were further treated in a graded ethanol series and isoamyl acetate, dried in a critical point dryer (HCP-2, Hitachi, Japan), and Pt–Pd coated using an anion coater (E-1030, Hitachi, Japan). They finally were observed under an S-4700 scanning electron microscope (Hitachi, Japan).

(c). Radiocarbon dating

Radiocarbon dating of textile samples was performed at the Oxford Radiocarbon Accelerator Unit. Calibrated dates at 95.4% probability were calculated using the IntCal v. 13 curve [38] in OxCal v. 4.2.4 [39].

(d). Protein extraction and separation, trypsin in-gel digestion, nanoflow LC–MS/MS and protein identification

Details of methods used for these steps are included in the electronic supplementary material. Briefly, proteins were extracted into SDS–PAGE (sodium dodecyl sulfate polyacrylamide gel electrophoresis) sample buffer, fractionated on gels and in-gel digested with trypsin using established protocols. Peptides were analysed and fragmented on a Q-Exactive Orbitrap mass spectrometer, and identified by matching spectra to the NCBI human protein sequence database. A representative SDS–PAGE gel image showing three protein extracts of the proteomics sample number 3 is also included in the electronic supplementary material.

(e). Data processing and quality control

The 15 lists of proteins obtained from triplicate analyses of the five different tissue samples were filtered using the Scrappy program [40], according to two criteria. A protein was retained as a valid hit in the final dataset for each sample if (i) the protein was identified in all three replicates and (ii) the total number of spectral counts in triplicates of at least one condition was a minimum of six. This transforms the low-stringency protein identification data from individual replicate analyses into a single high-stringency dataset of reproducibly identified proteins present at each time point [41,42]. Additional filtering was then applied based on total peptide counts summed across the three replicates, to minimize reverse database identifications. The peptide cut-off used varied from 6 to 21 depending on the total number of identifications and reverse database hits in the sample. Each sample was filtered so that the peptide false discovery rate was less than 0.3%. Protein abundance data were calculated using NSAF [35], with an addition of a spectral fraction of 0.5 to all spectral counts to compensate for null values and enable log transformation for subsequent statistical analyses [43]. Summed NSAF values were used as a measure of relative protein abundance to allow for quantitative comparison between replicates and samples.

(f). Bioinformatic analysis of identified proteins

Gene ontology (GO) classifications for UniProt protein identifiers were generated using the GO Term Mapper (http://go.princeton.edu/cgi-bin/GOTermMapper) with searching performed against the goa_human (generic GO slim) terms for cellular compartment and biological process. The numbers of proteins annotated within each category were then plotted as a percentage of the total number of UniProt identifiers in the dataset. Ensemble database protein headers were mapped to UniProt identifiers, using the open-access Protein Identifier Cross Reference tool (http://www.ebi.ac.uk/tools/picr), with searches mapped to the Swiss-Prot and TrEMBL databases and restricted to human proteins.

3. Results

(a). Scanning electron microscopy of skin fragments

In SEM analysis, all three samples examined in this study share common ultramicroscopic morphologies. The three skin tissues from the mummified remains were mainly composed of collagen fibres, while structures such as cells or vessels could not be observed, as shown in figure 1.

Figure 1.

SEM images of skin tissue from three Egyptian mummies: (a) S14381/1, female known as Khepeshet; (b) magnified image of (a); (c) S14391/2, male (uncertain) known as Idi; (d) S16742, adult unknown. Scale bars, (a) 100 µm; (b)–(d) 30 µm. The tissue was clearly composed mainly of collagen fibres.

(b). Radiocarbon dating

The Italian Archaeological Mission (MAI), led by Ernesto Schiaparelli with Giovanni Marro, excavated cemeteries at Assiut in 1911–1913 and Gebelein in 1920, largely to augment the collections of the Museo Egizio, Turin, Italy [44–46]. The Assiut mummies known as Khepeshet and Idi were elite burials interred in inscribed coffins, with grave goods such as head-rests, a bow, staves and a cartonnage mask on the face of the female ‘Khepeshet’ [44]. In contrast, Mummy S.16742 (unknown) was buried at Gebelein in a hollowed-out tree trunk with no record of artefacts [47].

Radiocarbon dating was performed on small textile fragments collected from each mummy at the same time as the skin and muscle materials were sampled. As shown in table 1, the dates as indicated by radiocarbon dating for all three mummies referred to in this paper are in the range of 2000–2450 BC. Thus, although attributed to the Middle Kingdom in the museum inventory, radiocarbon dating has shown that ‘Khepeshet’ actually dates earlier, to the late FIP. Calculating an average of the midpoint of the 95% confidence date ranges gives a figure of 2191 BC, which is why we have used the value of approximately 4200 years old for the age of the samples throughout the paper.

Table 1.

Radiocarbon dating of textile samples from each mummy examined in this study.

sample	OxA	calibrated age BC (68.2% range)	calibrated age BC (95.4% range)
adult mummy, sex unknown, S.16742	32 851	2338–2206	2456–2154
female adult mummy, Khepeshet, S.14381/1	32 759	2197–2045	2204–2033
male (uncertain) adult mummy, Idi, S.14391/1	32 760	2203–2053	2279–2036

(c). Quantitative summary of protein identifications from the whole proteomics sample set

The number of proteins reproducibly identified in the different samples was highly variable, which appears to correlate with the expected variations in sample quality. The number of proteins identified ranged from 18 to 132, as shown in table 2. In total, across the five tissue samples, there were 230 protein identifications in total after removing reversed database false positives, at a peptide false discovery rate of 0.15% and a protein confidence level of greater than 95%. The final non-redundant dataset contained 132 unique protein identifications, not including trypsin. Full details of proteins identified, with the number of peptides and NSAF values, are included in the electronic supplementary material, table S1.

Table 2.

Number of proteins identified in each sample.

proteomics sample number	no. proteins identified	peptide false discovery rate (%)
1 (30)	17	0
2 (31)	25	0.03
3 (41)	45	0.07
4 (36)	24	0.29
5 (37)	119	0.29
	total = 230	average = 0.14

(d). Ancient skin samples contain large numbers of collagens and keratins

The proteins identified in each tissue sample are discussed in the following sections. All five contained large numbers of collagens and keratins, confirming our hypothesis that these very long-lived proteins would be the most abundant detectable in samples of this age.

(e). Proteomics sample number 1—skin biopsy from chest of adult mummy, sex unknown

The first skin sample analysed contained only 17 proteins, which included nine collagens, six keratins, albumin and an osteoclast transporter protein. The latter is at the lower end in terms of abundance, so it may well be random noise, although it does also appear in the second skin sample analysed. The more abundant collagens and keratins found in the tissue samples from this mummy are identified with hundreds and thousands of peptides, so they are very high-confidence identifications, and provided the first evidence that we could indeed identify proteins present in such ancient tissue samples.

(f). Proteomics sample number 2—skin biopsy from tibia of adult mummy, sex unknown

The second skin sample analysed contained 25 proteins, which included 18 different collagens, four keratins, plus a ubiquitin ligase and a deubiquitinase, both of which are identified by a relatively small number of peptides. The identification of many more collagens than in the first sample, whereas the number of keratins does not increase, suggests that there may have been more protein extracted from this particular tissue sample.

One important point to reiterate is that this is the only one of the three mummies, in this study, that was considered a non-elite burial. The remains were interred in a hollowed-out log rather than a sealed coffin. The other two mummies were elite burials in sealed and decorated wooden coffins. This may account for the difference in our observed results. The first mummy would have been more exposed to the elements over time, which may have caused greater protein degradation than occurred for the other two, resulting in fewer proteins being identified.

(g). Proteomics sample number 3—skin biopsy from right hand of female adult mummy known as Khepeshet

The third skin sample analysed contained 45 proteins, which included 27 collagens, six keratins and albumin, but also contained a number of other proteins that are present at high abundance, because they are identified by a large number of peptides. The proteins identified in this sample are summarized in table 3, sorted into functional categories. The most abundant of these are myeloperoxidase, eosinophil peroxidase, cathepsin G and proteinase 3. Proteins present at a slightly lower abundance than these include defensin alpha one, S100 calcium binding protein A8 (S100A8), ferritin, cathepsin G and proteinase 3.

Table 3.

Summary of results for proteomics skin sample number 3.

functional category	proteins identified
collagens	collagen, type I, alpha 1; collagen, type X, alpha 1; collagen, type XXIII, alpha 1; collagen, type XXVII, alpha 1; collagen, type I, alpha 2; collagen, type II, alpha 1; collagen, type III, alpha 1; collagen, type III, alpha 1; collagen, type III, alpha 1; collagen, type IV, alpha 1; collagen, type IV, alpha 2; collagen, type IV, alpha 4; collagen, type IV, alpha 5; collagen, type IX, alpha 1; collagen, type V, alpha 1; collagen, type V, alpha 2; collagen, type VI, alpha 2; collagen, type VI, alpha 3; collagen, type VII, alpha 1; collagen, type VIII, alpha 1; collagen, type X, alpha 1; collagen, type XI, alpha 1; collagen, type XI, alpha 2; collagen, type XXII, alpha 1; collagen, type XXIV, alpha 1; collagen, type XXVIII, alpha 1; collagen, type XXVIII, alpha 1
keratins	keratin 1, type II; keratin 10, type I; keratin 14, type I; keratin 2, type II; keratin 5, type II; keratin 9, type I
immune response proteins	cathepsin G, defensin alpha 1, eosinophil peroxidase, ferritin, fibrosin, myeloperoxidase precursor, proteinase 3, S100 calcium binding protein A8
other	albumin, BCL6 corepressor-like protein 1, dermatopontin

Myeloperoxidase is a part of the host defence system in leucocytes, which is responsible for microbicidal activity against a wide range of organisms. It catalyses the production of hypochlorous acid, along with other cytotoxic intermediates, which are used by neutrophils to kill bacteria and other pathogens [48]. Eosinophil peroxidase is a related enzyme that mediates tyrosine nitration of secondary granule proteins in mature resting eosinophils. It has been shown to be responsible for killing of Mycobacterium tuberculosis by inducing bacterial fragmentation and lysis [49]. The defensins, as the name implies, are a family of proteins known to have antibacterial and fungicidal activity [50]. Neutrophil defensins are members of a family of cysteine-rich cationic proteins that are found predominantly in leucocytes, and are specifically associated with macrophages involved in inflammation response in lung tissue [51]. Ferritin is an iron-complexing protein that stores iron in a readily available soluble form. It is important for iron homeostasis, as iron is taken up in the ferrous form and then oxidized and deposited as ferric hydroxides [52].

The S100A8 protein is responsible for a plethora of different functions, many of which are related to other proteins discussed earlier [53]. Like ferritin, it is a metal-binding protein, which binds both calcium and zinc and plays a major role in the regulation of immune response and inflammatory processes. It can function as an amplifier of inflammation in autoimmunity, as well as in tumour development in numerous cancers [54]. It also possesses antimicrobial activity, towards both bacteria and fungi, which is believed to function by chelation of zinc ions, which are essential for microbial growth [55]. In addition, it is known to be involved in antioxidant scavenging and nitrosylation of multiple proteins [56]. Lastly, the S100A8 protein is commonly expressed at high levels in chronic and acute lung inflammations, and has been used as a monitoring biomarker for pulmonary-related diseases such as acute respiratory distress syndrome [57].

Cathepsin G is a serine protease with trypsin- and chymotrypsin-like specificity, which is known to have significant antibacterial activity [58]. Protease 3, also known as myeloblastin, is known to disrupt immune silencing in autoimmune vasculitis [59], and interestingly, the two main forms of vasculitis are associated with either myeloblastin or myeloperoxidase [60]. Cathepsin G is found in the azurophil granules of polymorphonuclear neutrophilic leucocytes and has been strongly linked to pathogenesis of a variety of diseases associated with inflammation of the lungs and airways, including the bacterial form of chronic obstructive pulmonary disease [61–63]. Polymorphonuclear neutrophils form a primary line of defence against bacterial infections using both oxidative and non-oxidative pathways to destroy phagocytized pathogens. Proteins that have been reported to mediate tissue injury at sites of inflammation, and hence have been suggested as a potential biomarker for lung inflammation, include the neutrophil serine proteinases cathepsin G, proteinase 3 and elastase and the defensins [61]. These three proteases are major components of the neutrophil primary granules, and hence play an important role in intracellular pathogen destruction [62]. Cathepsin G has also been found in neutrophil extracellular traps, which degrade virulence factors and kill bacteria [64].

Considered together, the presence of this group of neutrophil-associated proteins in this sample suggests very strongly that the individual was undergoing a severe immune response event, coupled with a major inflammation response, most likely in response to a microbial infection. The identification of a specific subset of proteins in this sample suggests there is a strong possibility that Khepeshet was suffering from a bacterial pulmonary infection at the time of death.

(h). Proteomics sample number 4—skin biopsy from right shoulder/rib of male (uncertain) adult mummy known as Idi

The fourth skin sample analysed contained 24 proteins, which included 13 collagens, three keratins and albumin, and four other relatively abundant proteins: CD300e immune receptor, surfactant protein D, fibrosin and complement C1q tumour necrosis factor-related protein 2. The proteins identified in this sample are summarized in table 4, sorted into functional categories. Fibrosin is a lymphokine that functions as a soluble mediator of chronic inflammation and serves as a molecular link between chronic inflammatory cells and tissue scarring [65]. The CD300e immune receptor is an immunoglobulin-like protein capable of regulating innate immunity responses [66], whereas surfactant protein D is a component of the lung innate immunity system that is known to enhance clearance of pathogens and also to be involved in modulation of inflammatory responses [67]. This skin sample was taken from the right shoulder area. Given the passage of time and the extensive post-mortem processing involved in mummification, it is quite plausible that this skin sample could contain some lung tissue. The presence of these three proteins at such high levels offers a glimpse into the health status of the individual involved. These would seem to indicate that the immune response had been activated, possibly in response to a pathogenic infection.

Table 4.

Summary of results for proteomics skin sample number 4.

functional category	proteins identified
collagens	collagen, type I, alpha 1; collagen, type 1, alpha 2; collagen, type II, alpha 1; collagen, type III, alpha 1; collagen, type III, alpha 1; collagen, type IV, alpha 4; collagen, type V, alpha 1; collagen, type V, alpha 2; collagen, type VII, alpha 1; collagen, type X, alpha 1; collagen, type XXII, alpha 1; collagen, type XXIV, alpha 1; collagen, type XXVIII, alpha 1
keratins	keratin 1, type II; keratin 10, type 1; keratin 10, type 1
immune response proteins	CD300e immune receptor, fibrosin, surfactant protein D
other	albumin, complement C1q tumour necrosis factor-related protein 2, OTU deubiquitinase 7A, SHC (Src homology 2 domain containing) transforming protein 1, WIZ isoform 1

(i). Proteomics sample number 5—muscle biopsy from right shoulder/rib of male (uncertain) adult mummy known as Idi

Sample number 5 was different from the first four samples analysed, because it came from a piece of muscle tissue, rather than from skin. It was also different because it produced more than twice as many protein identifications as any of the other samples, with 119 proteins identified in total. Similar to the skin samples, it contained 34 different collagens and 14 different keratins. In addition to these, there were 72 other proteins identified, which are summarized in table 5, sorted into functional categories. It is clearly not practicable to discuss every protein in detail, so in the following section we discuss the more abundant proteins, especially those which can reveal something about the physiological state of the tissue sample.

Table 5.

Summary of results for proteomics muscle sample number 5.

functional category	proteins identified
collagens	collagen alpha-1(XVI) chain; collagen alpha-3(IV) chain; collagen and calcium binding EGF domains CCBE1; collagen, type I, alpha 1; collagen, type X, alpha 1; collagen, type XXVII, alpha 1; collagen, type I, alpha 1; collagen, type I, alpha 2; collagen, type II, alpha 1; collagen, type III, alpha 1; collagen, type III, alpha 1; collagen, type III, alpha 1; collagen, type IV, alpha 1; collagen, type IV, alpha 2; collagen, type IV, alpha 4; collagen, type IV, alpha 5; collagen, type IX, alpha 1; collagen, type V, alpha 1; collagen, type V, alpha 2; collagen, type VI, alpha 2; collagen, type VI, alpha 3; collagen, type VII, alpha 1; collagen, type VIII, alpha 1; collagen, type X, alpha 1; collagen, type XI, alpha 1; collagen, type XI, alpha 2; collagen, type XVII, alpha 1; collagen, type XVIII, alpha 1; collagen, type XXII, alpha 1; collagen, type XXII, alpha 1; collagen, type XXIII, alpha 1; collagen, type XXIV, alpha 1; collagen, type XXVIII, alpha 1; collagen, type XXVIII, alpha 1
keratins	keratin 1, type II; keratin 10, type I; keratin 13, type I; keratin 14, type I; keratin 15, type I; keratin 16, type I; keratin 2, type II; keratin 5, type II; keratin 6A, type II; keratin 6B, type II; keratin 76, type II; keratin 77, type II; keratin, type II; cytoskeletal 79, keratin 9, type I
immune response proteins	annexin A1, defensin alpha 1, Dmbt-1, deleted in malignant brain tumours 1, fibrosin, immunoglobulin heavy constant alpha 1, immunoglobulin heavy constant gamma 1 (G1 m marker), immunoglobulin kappa constant, immunoglobulin kappa variable 3–20, immunoglobulin kappa variable 3D-20, myeloperoxidase precursor, polymeric immunoglobulin receptor, prolactin-induced protein, S100 calcium binding protein A8, surfactant protein D, Thy-1 cell surface antigen
stress response proteins	heat shock protein 90 kDa alpha family class B member 1, heat shock protein 90 kDa beta family member 1, heat shock protein family D (Hsp60) member 1
muscle structural proteins	actin gamma 1, actin, beta, actin, beta-like 2, decorin, filamin C, muscle-specific, myosin, heavy chain 1, skeletal muscle, adult, myosin, heavy chain 2, skeletal muscle, adult, myosin, heavy chain 3, myosin, heavy chain 7, cardiac muscle, beta, profilin 1, tubulin alpha 1b, tubulin beta class I, titin
metabolic enzymes	aldolase, enolase 1 alpha, fructose-bisphosphate A, aldolase, fructose-bisphosphate C, amylase, alpha 1A, ATP synthase, H+ transporting, mitochondrial F1 complex, alpha subunit 1, cardiac muscle, ATP synthase, H+ transporting, mitochondrial F1 complex, beta polypeptide, carbonic anhydrase III, glyceraldehyde-3-phosphate dehydrogenase, lactate dehydrogenase A, lactate dehydrogenase B, lysozyme, phosphoglycerate kinase 1, pyruvate kinase, muscle, transglutaminase 3
other	albumin, BCL6 corepressor-like protein 1, Bex5, nerve growth factor receptor-associated protein 2, chaperonin containing TCP1, subunit 4 (delta), cystatin 2A, dermatopontin, eukaryotic translation elongation factor 2, filaggrin, haemoglobin subunit beta, heterogeneous nuclear ribonucleoprotein A2/B1, heterogeneous nuclear ribonucleoprotein H1, histone cluster 1, H2bj, histone cluster 1, H2bm, insulin receptor substrate 1, kinesin family member 26B, lymphocyte cytosolic protein 1 (L-plastin), OTU domain containing 7A deubiquitinase, peptidoglycan recognition protein 2, phosphofurin acidic cluster sorting protein 1, polypyrimidine tract binding protein 1, protein X-ray repair cross-complementing protein 6, transient receptor potential cation channel, subfamily V, member 1, transmembrane protein 11, zymogen granule protein 16

There were two proteins clearly identified, myosin-7 and ATP synthase, H⁺ transporting, mitochondrial F1 complex, alpha subunit 1, which are annotated as being cardiac-specific isoforms. This suggests that the sample may have been a piece of cardiac muscle tissue, as cardiac-specific isoforms of muscle proteins have been previously used to determine the presence of cattle heart tissue in ancient cosmetics [68]. Given that the tissue was sampled from the right shoulder/rib area, it is plausible that it might contain some cardiac tissue. This opens up the possibility of further intensive bioinformatics analysis of the spectra we have already collected, to search for specific mutations that may be indicative of disease states. For example, point mutations in the myosin-7 sequence have been found to be indicative of cardiomyopathy [69].

(i). Dmbt-1 and transglutaminase

The protein known as deleted in malignant brain tumour-1 (Dmbt-1) is especially interesting. It is not something normally seen at appreciable levels in muscle tissue, and it is known to be a protein that performs a wide variety of functions in different contexts [70]. It functions as a tumour suppressor, mediates interactions between tumours and surrounding cells, plays a role in host–pathogen defence, and, furthermore, increased abundance of it has been reported to be correlated with cardiac amyloidosis [71]. The concomitant increased abundance of this protein with increased abundance of transglutaminase has been found to be correlated with pancreatic cancer progression, and both of those proteins were found in significant levels in this muscle sample [72]. The fact that we find both of these proteins still present after all this time suggests that they may have been present at higher than usual level of abundance in the original tissue sample. This again opens up avenues for further exploration, because it offers a tantalizing possibility that Idi may have suffered from pancreatic cancer, or some other form of cancer.

(ii). Functions of the more abundant proteins

Given the age of the sample, it is not really possible within the constraints of the current study to compare these results against those from freshly sampled muscle tissue to try and estimate which proteins are expressed at higher levels, or lower levels, than would be expected. Because it is not practicable to discuss all of the proteins involved individually, in the following section, we present functional evidence for the more abundant proteins present in the sample. Applying an arbitrary cut-off of a total of 150 peptides identified across the three technical replicates, which is extremely high and has a false discovery rate of zero, allows us to focus on 22 proteins identified that were not collagens or keratins.

Actins and tubulins are structural proteins that are highly conserved across species and are ubiquitously expressed in all eukaryotic cells. Myosin proteins are similarly found in all muscle tissue and are responsible for muscle contractions. Filamin C is a muscle-specific protein that plays a central role in muscle cells, by functioning as a large actin-cross-linking protein [73]. Decorin is the best characterized member of the extracellular small leucine-rich proteoglycan family present in connective tissues, and binds strongly to collagens [74].

Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) is a metabolic enzyme that plays a role in glycolysis and nuclear functions, and is known to be a component of the gamma interferon-activated inhibitor of translation complex, which is activated by inflammation and mediates interferon-gamma-induced transcript-selective translation inhibition [75]. Additionally, it has been shown to play a role in tumour progression, by interaction with TNF-alpha [76]. Enolase is a multifunctional enzyme that plays a role in glycolysis, and also takes part in various cellular processes including allergic response and hypoxia tolerance. Both GAPDH and enolase are part of the same important metabolic pathway, along with pyruvate kinase, that synthesizes pyruvate from d-glyceraldehyde 3-phosphate.

Immunoglobulin alpha is the major immunoglobulin in bodily fluids and functions in defence against local infection by preventing access of foreign antigens to the immune system [77]. Ig kappa chain C region is part of the same protein complex. Polymeric immunoglobulin receptor binds polymeric immunoglobulin A and M at the basolateral surface of epithelial cells. Prolactin-inducible protein not only performs a variety of functions in the reproductive system, but also has the ability to bind immunoglobulins and T cell receptors, implicating it in a wide range of immunological functions [78].

Amylase is a relatively abundant enzyme, especially in saliva, which is responsible for endohydrolysis of alpha 1–4 glucosidic linkages in polysaccharides. In serum, elevated levels of amylase are strongly associated with inflammation, and levels of the enzyme are used as an indicator of pancreatitis [79]. Annexin A1 plays an important part in the innate immune response as it affects glucocorticoid-mediated responses and thus regulates inflammatory process. It is also known to contribute to the adaptive immune response by enhancing various signalling cascades that are triggered by T-cell activation [80].

(iii). Gene ontology classification

The 119 proteins reproducibly identified in muscle sample number 5 were subject to GO classification based on biological process. After removal of unannotated entries, 109 protein sequence entries were classified. The top 10 biological process categories assigned within the sample, along with the genome frequency for each category, are shown in figure 2. This functional categorization allows proteins to be sorted into more than one category where evidence exists that they perform multifunctional roles, so this should be considered as a more quantitatively accurate representation than the categorizations in table 5.

Figure 2.

Gene ontology (GO) based on biological process for proteins reproducibly identified in proteomics muscle sample number 5, from S.14391/2 (male, uncertain, known as Idi). The top 10 most abundant categories of identified proteins are presented, with the dark bars indicating GO term usage in the list of identified proteins as a percentage, and the light bars indicating genome frequency as a percentage.

The top five biological process categories reported in terms of abundance were anatomical structure development (57.8%), response to stress (34.9%), cell differentiation (33.0%), immune system processes (33.0%) and catabolic processes (32.1%). All five were significantly over-represented compared with the number of proteins in that category in the human genome, with the enrichment varying from 3.5- to 5.3-fold. The proteins in the categories of anatomical structure development and cell differentiation were predominantly the collagens and keratins, as expected. The presence of such a large number of proteins in the categories of stress response and immune system processes is a striking feature of these data, and indicates that the cellular immune response was highly activated.

Taken together, the data from both of these tissue samples (4 and 5) indicate the presence of serious inflammation and a highly active immune response, which suggests that Idi was most likely suffering from a serious infection, which we can speculate may be linked to the cause of death.

4. Discussion

A recent study analysing the proteome of human skin identified 159 proteins using an SDS–PAGE gel slice proteomics approach, and an additional 616 proteins using an iTRAQ labelling approach [81]. In that study, skin biopsy samples were collected from volunteers, and washed to remove excess blood. Clearly, this is a very different situation from the mummified samples in this paper. The presence of residual amounts of blood with the tissue may well complicate the interpretation of their results, because blood and serum are known to contain very high levels of a small number of proteins. Nonetheless, their results provide a very useful baseline in terms of what one would expect to find in the proteome of human skin from a living volunteer. The most commonly found proteins in their results were different isoforms of keratin, which comes as no surprise to anyone who has ever prepared samples in the laboratory for protein identification by MS; human keratins are a ubiquitous and highly abundant contaminant resulting from sample handling.

In contrast, in our results, the most abundant protein observed was various isoforms of collagen. While it is not possible to distinguish the source of the keratins present, the fact that they are present at much lower amounts than the collagens identified suggests that they originate from the mummy. Contamination from sample handling would be expected to introduce keratins in much greater abundance. Our morphological study clearly showed that these skin tissues of Egyptian mummies are mainly composed of collagen fibres. Apart from the collagen fibres, very few intact microscopic structures were observed in the mummified tissues. The typical cross-striations of collagen fibres were not apparent on these SEM images; this indicates that the collagen fibres were at least partly degraded. However, the collagen fibres observed profusely in these images strongly corroborate our MS findings that large numbers of collagens still remained in these ancient skin and muscle samples from Egyptian mummies. This is in accordance with previous analysis of similar samples using microscopic and histological techniques, which clearly showed that collagen fibres were well preserved. Collagens are abundant in many ancient tissue samples, which is in agreement with the prevailing hypothesis that there are intrinsic structural features, including extensive cross-linking via hydroxyproline residues, which make them more resistant to degradation and hence longer-lasting.

The FIP, Egypt's first ‘dark age’, was characterized by increasing aridity, resulting in megadrought and famine [82,83]. There are no reliable historical records of outbreaks of infectious diseases in this period. Texts allude to chaos and death (‘… no shortage of dead, the shroud calls out before one comes near it’) [84]. To the best of our knowledge, the health of this population has not been investigated by modern scientific methods. However, it is recognized that food and water shortages weaken the immune system and increase risk factors for spread of deadly infectious diseases such as malaria, tuberculosis, visceral leishmaniasis and other parasitic intestinal infections [85]. Groups with these underlying chronic conditions are at increased risk of contracting, for example, cholera, typhoid fever and acute respiratory infections. Infectious diseases identified previously in ancient Egyptian mummies include malaria and tuberculosis [86,87].

Disease detection in ancient tissue samples is mostly performed using PCR-based or metagenomics approaches, which can definitely establish pathogen presence, but do not provide direct evidence of an active disease. In a recent study, Corthals et al. used quantitative analysis of proteins identified between two different 500-year-old Inca mummies, based on spectral counting and non-parametric statistics [19]. This enabled them to show that one mummy displayed evidence of a severe pulmonary infection, because numerous biomarker proteins were present at much higher levels in one than in the other. We are not in a position to perform that type of analysis in this study, because we do not have a comparable control. Nonetheless, we can make a reasonable estimate that any proteins observed at higher abundance in tissue samples of this age must have been expressed at relatively high levels in the original tissue. Using that approach, we have been able to show that many of the proteins still present in these samples are linked to inflammation and immune response.

5. Conclusion

In this study, we have conclusively identified more than 230 proteins present in a set of small skin and muscle tissue samples taken from three Egyptian mummies of the FIP, approximately 4200 years of age. The samples were found to contain large amounts of collagen, which is in agreement with morphological analysis by scanning electron microscopy. The protein expression profile we have observed and characterized for two of the mummies, those known as Khepeshet and Idi, is consistent with tissue inflammation and a severe immune system response to infection at the time of death. A subset of those proteins detected suggests there is a strong possibility that Khepeshet may have been suffering from a bacterial pulmonary infection.

Data accessibility

The electronic supplementary material file contains details of additional methods used, along with details of all peptides and proteins identified from each of the five proteomics samples analysed. This includes protein identifiers, descriptions, number of peptides identified in each replicates and in total, and calculated NSAF values.

Authors' contributions

J.J. designed the study, was present for the tissue sampling, and provided critical discussion with historical context and insights, R.B. was present for the tissue sampling and provided critical discussion, D.S.L. and D.H.S. performed scanning electron microscopy, M.M. and D.X. performed mass spectrometry and database searching, P.R. performed protein extraction, electrophoresis and database searching, and assisted with mass spectrometry, P.A.H. performed protein extraction, mass spectrometry, database searching and bioinformatics analysis, and assembled the manuscript.

Competing interests

The authors declare that they have no competing interests.

Funding

J.J. received funding in the form of a Macquarie University Research Fellowship, and none of the other authors received any specific funding for this work.