Preparative capillary gas chromatography (PCGC) is the central technique used for the purification of volatile or semi-volatile organic compounds for radiocarbon analysis using accelerator mass spectrometry (AMS). While thicker film columns offer efficient separations, column bleed of cyclic poly(dimethyl siloxane) (PDMS) stationary phase has been highlighted as a potential source of contaminant carbon in ‘trapped’ compounds. The dimethylpolysiloxane CH3 groups are of ‘infinite’ radiocarbon age due to the fossil carbon origin of the feedstock used in production. Hence, column bleed, if present at sufficiently high concentrations, would shift the radiocarbon ages of trapped compounds to older ages. Quantification of the column bleed in trapped samples, however, is extremely challenging and up to now has only been achieved through indirect 14C determinations of chromatographic blanks, which are used for post 14C determination ‘corrections’. As part of wider investigations aimed at better understanding the chemical nature of contamination in compound-specific 14C-determinations, herein, we report a rigorous approach to column bleed identification and quantification. Using reference fatty acid methyl esters (FAMEs) 1H nuclear magnetic resonance spectroscopy (NMR), employing a 700 MHz instrument equipped with a 1.7 mm microcryoprobe optimised for 1H observation, was able to detect low sub-microgram amounts of low molecular weight compounds (<500 Da). Direct quantification of PCGC ‘trapped’ FAMEs was achieved based on the recorded 1H NMR spectra. Gravimetrically prepared calibration mixtures of cyclic PMDSs and FAMEs, showed column bleed abundance to be below 0.03% w/w of the ‘trapped’ FAMEs, which would lead to a maximum shift in radiocarbon age of <3 years toward older values. We therefore conclude that column bleed contamination has a negligible effect on the 14C determination of FAMEs prepared using the chromatographic method described. The 1H NMR analysis also revealed the absence of other protonated carbon-containing components that would affect radiocarbon determinations at the precisions achievable by AMS. INTRODUCTION A critical concern when preparing samples for radiocarbon dating is contamination of the sample through the introduction of exogenous carbon. Such contamination from sample treatment can lead to significant offsets (older or younger) of the actual sample age leading to erroneous dates1,2,3. Contamination becomes particularly problematic with the small sample sizes, i.e. less than 1 mg of C, that are increasingly commonly analysed due to advances in sample preparation methods and AMS technologies. Identification and quantification of exogenous carbon in samples for radiocarbon analysis has been attempted using various approaches, e.g. FTIR has been used to identify contamination in bone collagen4 and Raman to determine soil carbon contamination of charcoal5,6. However, these techniques are ineffective in the case of low-level contamination of small samples, due to a fundamental lack of sensitivity, precluding quantification of contamination at the part per thousand level, which would affect radiocarbon determinations. The question of exogenous contamination is especially critical in compound-specific 14C determinations in which preparative capillary gas chromatography (PCGC) is used to isolate compounds from extracts of various environmental matrices7,8,9,10, and archaeological pottery vessels11,12,13. The compound-specific approach routinely involves trapping sub-milligram amounts of analyte for 14C determinations, hence, the use of PCGC requires assessment of all potential sources of exogenous carbon likely to arise during the sample pre-treatment. As the analytes are purified exogenous carbon could potentially be introduced either from the PCGC used for compound purification, the handling of compounds between isolation, oxidation and graphitisation. A recognised source of exogenous carbon to compounds isolated by PCGC is column “bleed”14,15,16 (figure 1), derived from thermal degradation of the commonly used PDMS stationary phase coating the column through heating in the GC oven15,16. The cyclic degradation products of the polymer released typically are n = 3 and n = 4 cyclic oligomers of the monomer unit (-[Si(CH3)2-O]n-), with the possibility of higher homologues up to an n = 717,18. Several approaches have been considered to identify, limit, and correct for the effects of ‘column bleed’ from the GC column. Eglinton and co-workers7, who reported the first use of PCGC to isolate compounds for radiocarbon dating, suggested using columns coated with a thin film of ultra-low bleed stationary phase (≤0.5 µm). Stott et al.12 attempted to determine the column bleed concentration by the preparation of chromatographic blanks when isolating C16:0 and C18:0 FAMEs from archaeological pottery. This was achieved by trapping the column eluent for almost an entire run after injection of solvent only (~40 times longer than typical trapping windows). The blanks contained insufficient carbon for radiocarbon analysis (~0.9 µg), thus, the column bleed was assumed to exert an insignificant effect on compound-specific 14C age determinations. In a later study Ziolkowski et al.19 used a column coated with a thicker stationary phase film (1.5 µm) and generated blanks based on 400 dry injections and trapping 7 min retention time ‘windows’ then normalised the amount of exogenous carbon to 50 injections. This approach assumes that the column releases its stationary phase at constant rate regardless of time or temperature, thus the amount of column bleed associated with a trapped compound is proportional the length of the trapping window. It is however known that that the release of PDMS from GC columns increases with temperature. Another approach adopted by Ziolkowski et al.19, and probably the most effective of the approaches used up to now20,21 to assess the effect of column bleed, involved the isolating of reference compounds of known modern age then investigating shifts in 14C content by radiocarbon analysis. The influence of column bleed was estimated from deviations in the 14C content of reference materials compared to those of dry injections, i.e. blanks. The summary is that all the methods used to date to assess the potential effects of column bleed on 14C determinations are indirect and unable to identify and quantify the specific chemical contaminants of the PCGC trapped analytes. The specific analytical hurdle thwarting direct characterisation and quantification until now is the small quantity of compound isolated by PCGC, typically sub-milligram, combined with the low concentration of contaminant, i.e. ppt, that could potentially effect 14C determinations. The latter highlights the need for a new approach for assessing analyte purity for compound-specific 14C determinations. Nuclear magnetic resonance (NMR) offers the possibility of detecting and quantifying the presence of contaminant species but has not typically been employed due to perceived problems with sensitivity. The latest generation of high field NMR spectrometers equipped with microcryoprobes, however, extends the lower limits of detection of protonated species to the picomol scale22,23. The extreme sensitivity makes these probes ideal tools to study mass limited samples such as isolated low abundance proteins, peptides and small molecules and difficult to express proteins. For example, this technology has been used to determine the structures of molecules that can only be isolated in minute amounts (micrograms or less) from natural sources such as deep sea sponges22,23, red algae24 and plants25. It is also used by structural genomic consortia for high throughput microscale screening of protein targets26,27. In addition to analysing the structure and purity of molecules, NMR can also be used to determine the concentration of one more chemical species in solution with a high level of precision in a non-destructive manner23,28. qNMR has proven to be a reliable, specific and linear over a wide concentration range with limits of accuracy and precision in the order of 0.5-1%29,30,31. The method is particularly suited for the simultaneous determination of the active constituents and impurities in samples from the food, pharmaceutical and chemical industries30,32,33,34. Furthermore a number of studies have used qNMR to specifically identify and quantify impurities in agrochemicals35, pharmaceuticals36 and amino acids and peptides37, however this is the first time to our knowledge that NMR has been used in the field of radiocarbon analysis. The advent of this analytical technology opens the way for a new approach to the qualitative and quantitative assessment of contamination in analytes for radiocarbon analysis. The sensitivity and dynamic range of high field NMR instruments, combined with their capability for compound identification and quantification, offers hitherto unattainable potential for use in assessing contamination at the ‰ level in sub-milligram amounts of purified compounds, trapped in ca. 1 min ‘windows’, instead of extended sequences. This allows for the first time, the assessment of exogenous carbon in trapped analytes, specifically the definitive assessment of the degree to which column bleed and other potential sources of carbon could affect radiocarbon date determinations Herein, we demonstrate: (i) the use of microcryoprobe-equipped 700 MHz 1H NMR to determine the purity of compounds isolated with PCGC, (ii) identification of exogenous carbon in the trapped analytes by comparisons with authentic standards, (iii) quantification of the contamination present using calibration mixtures, and (iv) the impact of such contamination on high precision 14C dates using AMS and conclude whether corrections are required. EXPERIMENTAL Reference materials and sample preparation All reference materials were purchased from Sigma Aldrich (Poole, UK). HPLC grade solvents were purchased from Rathburn Chemical Ltd (Walkerburn, UK) and deuterated chloroform (“100 %”, 99.96 atom % D) from Sigma Aldrich (Poole, UK). Organic residues from the pottery vessel were extracted using the method described by Correa-Ascensio and Evershed38. Stock solutions containing known concentrations of C18:0 FAME and hexamethylcyclotrisiloxane were prepared to produce a NMR calibration curve for its quantification by NMR. These solutions contained the FAME at 1 mg.mL-1 and siloxane in varying concentrations, from 1 mg.mL-1 to 1.10-6 mg.mL-1 (see table 1), diluted in chloroform-d solvent.
- Bristol BioDesign Institute
- SYNTHETIC BIOLOGY