One night with Venus, a lifetime with mercury

In the following section, provenance and proof of authenticity of the samples is described briefly. More information can be found in the supplementary section.

A few single hairs were removed with sterile and pre-annealed tweezers, transferred into small paper envelopes and labelled accordingly: Schubert A 1.W, Schubert A 2.W, Schubert A 3.W, Schubert B and Schubert B 1.W. A lock of Franz Schubert’s hair, which originates from the collection of Ignaz Weinmann [14], is on loan from the Vienna City Library and currently exhibited at the Schubert Memorial Site in the castle Atzenbrugg, Lower Austria. A picture of the medallion is depicted (see Fig. 1) which is labelled as follows:

“Medallion with Franz Schubert’s hair. Light-coloured bronze, diameter 64 mm, circular engraving ‘Hair of Franz Schubert’. In the centre a depression (diameter 44 mm) containing hair remnants mixed with wood particles. Domed cover glass above.” (WBR, MS, ZPM 338, Bibliotheca Schubertiana Ignaz Weinmann)

Two single hairs of the loose strands of approximately 3 and 4 cm in length were removed from the medallion individually using sterile and annealed splinter tweezers and subsequently transferred in paper envelopes labelled as follows: Schubert C 1—tox and Schubert C 2—DNA.

Three single hairs were provided for molecular genetic analyses applying mtDNA analysis. Hair C was submitted in two segments with approximately 1.5 cm (without root) and 1 cm (with root) in length. The root portions of hair A (approx. 2.3 cm) and hair B (approx. 7 cm) were excised. The shaft of hair B was cut into 1 cm long sections prior to DNA extraction and root and shaft segments were investigated independently. In the first step, the surface of the hair shafts (not roots) was cleaned with aqua bidest. and then subjected to a mild lysis. The protocol consists of incubation (2 h, 56 °C) in 900 µL of a lysis buffer (10 mM Tris pH 8.0, 100 mM NaCl, 1 mM CaCl₂, 2% SDS, 36 µl proteinase K) under agitation to remove external contamination. Full lysis of the purified hair shafts and uncleaned roots was performed overnight in 900 µL of the lysis buffer (see above), 10 µL proteinase K and 10 µL 1 M DTT. On the following day, 10 µL proteinase K and 10 µL 1 M DTT were additionally added and the samples were incubated until no hair structures were visible anymore. The DNA extraction was performed following a modified protocol developed by Dabney et. al which is outlined in [15]. In our study, 2 mL binding buffer and 80 µL 3 M sodium acetate were used. The mtDNA analysis was performed according to the PEC protocol detailed in [16]. Haplogrouping was performed using SAM2 via EMPOP following the procedure outlined in [17]. Quantification of nuclear DNA and mtDNA was performed using the SDquant system [18].

Laser ablation inductively coupled plasma sector field mass spectrometry (LA-ICP-SFMS) has provided a valuable tool for obtaining spatial information of samples with a high lateral resolution, high sensitivity and multielement capability. Single hairs were analyzed from the root of the hair to the tip using a NWR213™ Nd:YAG 213 nm laser ablation system (ESI Elemental Scientific Inc., Omaha, NE, USA) coupled to an ICP-SFMS (Element 2, Thermo Fisher Scientific, Bremen, Germany). The ablated material was transported out of the ablation cell in a continuous flow of helium (550 mL min⁻¹). The carrier gas was mixed with argon (1.2–1.25 L min⁻¹) via a Y-connection of the transport line and the analyte-gas mixture was directed into the ICP. Laser scanning was performed with a spot size of 50 µm in a first scan and 20 µm in a second scan in the same line with a scan speed of 50 and 10 µm s⁻¹. The applied laser energy (fluence) was 27.8–29.1 J cm⁻². The ICP-SFMS was operated in low mass resolution mode (m/∆m = 300). Masses monitored were: ³⁴S, ⁷⁵As, ¹¹¹Cd, ¹²¹Sb, ¹⁹⁵Pt, ²⁰⁰Hg, ²⁰²Hg and ²⁰⁸Pb. Data acquisition was performed in E‑scan mode (50% mass window, 60 samples per peak, 10 ms sampling time), resulting in a cycle time per pass of 3.524 s (50 µm s⁻¹ scan speed) and 1.924 s (10 µm s⁻¹ scan speed). For the best possible focusing of the laser beam onto the sample surface, individual sections of approx. 10 mm were analyzed with an overlap of 200 µm between the individual sections. Each measurement cycle started with the acquisition of a blank. The average of the obtained blank intensities was subtracted from the subsequent analyte signal. In this study, no absolute quantification of the investigated elements was performed. Instead, the signal intensities obtained for the investigated samples were assessed via comparison to those obtained for hair samples from healthy individuals. Sulfur was used as internal standard for normalization of the metal signals as its concentration has been reported to be relatively constant in hair [19]. Prior to analysis, all hair samples were precleaned by soaking in 1:10 diluted NovaClean™ AFX (pure11 GmbH, Grünwald, Germany) general purpose cleaner for 1 h, rinsing with ultrapure water followed by acetone in three cycles and drying in a laboratory oven at 75 ± 5 °C overnight. The hair samples were adhered onto glass slides using double-sided tape “Pritt non-permanent”. Starting points and directions of the analysis were indicated on the glass slide. Thus, growth directions and the arrangement of the time coordinates could be subsequently confirmed using SEM analyses.

Scanning electron microscopy (SEM) is a powerful method to investigate the microstructure and topography of both inorganic and organic matter. To obtain detailed information about the state of condition of the strands of hair, the samples investigated by LA-ICP-SFMS were subsequently analysed with a JEOL JSM-IT200 scanning electron microscope (JEOL (Germany) GmbH, Freising, Germany), under high vacuum. In order to preserve the samples in their original appearance, the hairs were not coated by carbon prior to the investigation. To avoid charging and damage caused by the electron beam on the surface of the nonconductive hairs and simultaneously achieve images with a reasonable resolution and quality, a very low acceleration voltage (i.e., 1 kV) and a slow speed of scanning process (50 s) were used.

Molecular genetic analyses were carried out on three individual hair samples. Their authenticity is based on independently documented sampling processes. The mtDNA and DNA concentrations were very low as expected for such old hair fragments, except for the root of hair A (Table 1). While the shaft segments were precleaned, the roots were not cleaned as the cleaning procedure would have removed genuine (mt)DNA from the root. Therefore, root portions are more likely to show extraneous contamination. Indeed, the root portion of hair A resulted in higher sequence read depths than the shaft portions and also the mitotype differed. This mitotype was also present in our local contamination exclusion database which contains mtDNA sequences of staff, technical personnel, students and visitors. These results suggest that the higher values obtained for the root portion of hair A could represent contamination. The remaining hair segments showed higher values for the 69 bp target than for the 143 bp target which is in line with expectations for severely degraded DNA in such old specimens [16]. Hair segments which yielded > 69 bp were subjected to mtDNA sequencing. Hair segments A‑S and B‑S yielded mtDNA concentrations of approx. 100 mitogenomes µL⁻¹ and resulted in mitotypes which cannot be excluded as deriving from the same individual and lineage (Table 1). Hair C‑S did not yield sequencing results above the required reads. The identical mitotype 263G 315.1 C 750G 15784C 16519C (range 1–841, 8211–8476, 15784–16569) which was shared by hair A‑S and hair B‑S belonged to haplogroup H65. This mitotype was observed 9 times in a database of 75,838 mitogenomes (mitoTree; Nicole Huber, Institute of Legal Medicine, Medical University of Innsbruck, Innsbruck, Austria, personal communication), which equals 0.0145% (augmented counting method; x + 2 / n + 2, where x is the number of observations and n is the database size). Applying a 99.5% confidence interval (BetaInv) results in an upper limit of 0.0232% or 1:4300 for the probability that a randomly selected individual matches this mitotype. Despite the lack of maternally related reference samples from a (living) relative of Franz Schubert, we can assume that the two curls in Raimund Hofbauer’s possession come from one and the same person (Fig. 2).

The LA-ICP-SFMS is the method of choice for sensitive and selective measurement of the distribution of trace elements in hair samples [20]. The hair sample is placed in a closed ablation chamber which is continuously passed through by a carrier gas. The laser beam with defined energy and size ablates, i.e., vaporizes, a certain amount of material, which is directly transported to the ICP-MS where the ablated material is atomized, ionized and detected after mass spectrometric separation in real time. To monitor longitudinal variations of relative elemental concentrations in the individual hairs, the obtained transient analyte signals were converted to hair length and lifetime. Assuming an average hair growth rate of 10 mm per month, a time resolution of approximately 9.8 h (50 µm s⁻¹ scan speed) and 1.4 h (10 µm s⁻¹ scan speed) has been obtained. Analyte/sulfur ratios of controls (unexposed healthy individuals with natural hair color) were calculated and compared to those obtained from the hair samples of Franz Schubert. In view of the very limited amount of hair samples available, it was not possible to determine the absolute concentrations of mercury or lead by acid digestion. Figure S1 (supplementary section) shows the relative intensities of mercury (Hg) and lead (Pb) in a single hair of a male nonsmoking individual, obtained by ablating the sample from the root to the end of the hair at the surface (cuticle). Average ratios of 0.0025 ± 0.0004 (²⁰²Hg/³⁴S) and 0.0008 ± 0.0004 (²⁰⁸Pb/³⁴S) were calculated. Variations of these metals longitudinally in a hair sample of Franz Schubert (Schubert A 3.W) are depicted in Fig. 3. A significant decrease in concentrations, in particular 8–5 months before his death, clearly reflect exposure of toxic metals. Elevated concentrations of mercury and lead up to a factor of 50 and 2000, respectively (signal spikes were considered as outliers and thus, not included), were detected during growth of all hair samples analysed. Figure 4 shows relative intensities of mercury and lead in the cuticle (surface scan) and the core of a hair sample (Schubert B) from close to the root to the tip of the hair. Decreased concentrations in the core relative to the cuticle could be determined for these elements, whereby the greatest difference in relative concentrations was observed for mercury.

The SEM analysis showed that even at low magnifications it could be clearly observed that the laser had milled a 20 µm wide longitudinal groove into the cortex of all hairs, exposing the medulla. In addition, individual measurements taken at 80 μm intervals revealed regular punch-outs of 10 μm in the exposed hair pith, providing data at time intervals of approximately 5–6 h (see Fig. 5).

Detailed examinations of the hairs from the two Schubert/Hofbauer locks showed a thickness of 46–95 μm and a well-preserved scale structure of the hair surfaces. There were no impurities. Hairs from the medallion (Weinmann/Vienna City Library), on the other hand, showed a severe loss of scale structure of the surface and a fluctuating hair thickness between 45 and 75 μm. Extensive microbial contamination and spongy, fine-pored matrix loosening was found both on the surface and in the medullary substance.

Finally, the observations by SEM enabled determination of the arrangement of the cuticle, consequently, the direction of hair growth could clearly be determined in all samples. This information was of outmost importance, because it helped to verify the results of the LA-ICP-SFMS analysis (i.e., the direction of the spot measurements related to the direction of the growth of the hairs).