Computed tomography (CT) scanning and radiological examinations of the Maiden revealed that all her organs, including the eyes and the brain, were intact 
. Both radiological and visual examination revealed pathologies consistent with a range of infectious diseases: 1) a radiolucent area in the upper lobe of the right lung, 2) a mucosal enlargement of the left maxillary sinus consistent with sinusitis, 3) a zoster-like lesion on the right calf, and 4) streaks of mucus under both nostrils 
. Similar exams on the boy revealed no lesions, and no mucosal enlargement or other signs of upper respiratory infection. To identify the proteins on the lips of both mummies, and assess the presence of pathogens we collected mouth swabs. Proteomics analyses of the mouth swabs based on high-resolution mass spectrometry revealed the presence of proteins expected in nasal secretions: serum proteins (i.e. albumin, hemoglobin and serotransferrin) in both mummies. The nasal mucus protein (PLUNC) level was three times higher in the Maiden’s sample than in the boy’s.
In addition to serum proteins, we found several proteins that are not normally present in blood or saliva, but are consistent with host immune response to infectious disease in the Maiden’s sample (see ). Cathepsin G is a specialized neutrophilic polymorphonuclear leukocyte serine protease found in the azurophil granules and its function has been linked to pathogenesis of diseases associated with inflammation and neutrophil infiltration of the airways, such as bacterial COPD (Chronic Obstructive Pulmonary Disease) 
. Cathepsin G and neutrophil elastase have also been found in neutrophil extracellular traps (NETs) that degrade virulence factors and kill bacteria 
. A marker of chronic lung inflammatory diseases, α-1 antitrypsin, is a strong indicator of mycobacterial infection 
. It protects tissues against inflammatory, cytotoxic proteases, such as those from neutrophils. Neutrophil defensin 1 and 3 are part of the defensin family of cysteine-rich cationic proteins found in leukocytes and are specifically associated with macrophages involved in lung tissue inflammation response 
Immune system proteins and their respective accession numbers in the swab sample of the Maiden identified by mass spectrometry.
The proteomic analysis of the Maiden sample also uncovered two groups of proteins consistent with severe inflammation of the lungs. The first group of proteins included S100 A8/A9, apolipoprotein A1 and A2, and transthyretin. The second group of proteins included vitamin-D-binding protein (VDB), serine protease inhibitor (SERPIN) and transthyretin (TTR). The first proteins are commonly expressed in chronic and acute lung inflammations, and have been used as monitoring biomarkers for pulmonary related diseases 
. The second group of proteins is also involved in acute lung inflammation, specifically in mycobacterial infections 
. The presence of the full complement of these proteins in the mouth swab of the Maiden provided strong evidence of response to a severe respiratory bacterial infection. The external visible symptoms and the gamut of immune response proteins obtained from the mouth swab supported the hypothesis of pulmonary infection caused by Mycobacterium
The boy did not show signs of upper respiratory or pulmonary infections based on CT-scans and radiology analyses, despite the presence of blood in the mouth swab and cloth samples. For these reasons, we inferred that the boy did not have a respiratory infection, and the presence of blood was the result of trauma. Proteomic analysis of the boy’s mouth swab revealed that his α-1 antitrypsin levels were high, and neutrophil defensin levels were low. These results supported the inference that the boy was not suffering from a lung infection (see table S2
). The comparison of the spectral counts in the inflammatory/immune category was highly significant (Z
0.0003), while the comparison for all other proteins was not significant (Z
0.7206). Inflammatory and immune response proteins were elevated in the Maiden (median spectral count
27.03) relative to the boy (median spectral count
2.14). Levels of all other proteins detected in the samples were similar (median spectral count of Maiden
273.33; for the boy
To determine the etiology of the disease, we amplified the heat-shock protein (hsp65
) gene using Mycobacteria
-specific primers 
and DNA extracted from the mouth swab taken from lips of the Maiden. The PCR assay followed by direct sequencing of PCR products confirmed the presence of Mycobacterium sp.
in one of the mouth swab samples, as well as the presence of non-pathogenic Bifidobacterium sp.
(). The presence of Bifidobacterium sp.
on the lips of the Maiden cannot be a result of fluid deposition during decomposition, since the bodies of the children of Llullaillaco did not decompose. We interpreted the detection of Bifidobacterium sp.
as an indication of vomit shortly prior to her death, rather than as a result of postmortem contamination.
The position of the recovered sequence at the base of the Mycobacterium
genus could be caused by the large amounts of missing data in the sequence (84%) relative to the genomic sequences (alignment was 1,759-bp long). Based on the best phylogeny (), we compared alternative trees seeking to further refine the placement of our sequence (). These comparisons ruled out cross-contamination of the Mycobacterium sp
. sequence with DNA from non-pathogenic Bifidobacterium sp.
≤0.002), as well as the sequence corresponding to the non-pathogenic Mycobacterium smegmatis
complex (Bayesian posterior probability [BPP]
0.008, more conservative approximately unbiased [AU] and weighted Shimodaira-Hasegawa [WSH] tests P
≥0.117), but could not rule out the recovered sequence belonging to the pathogenic Mycobacterium avium-bovis
0.230, AU and WSH P
≥0.475). The phylogenetic analyses indicate a higher probability for the hypothesis that the sequence corresponded to the pathogenic Mycobacterium avium-bovis
than to the non-pathogenic Mycobacterium