Accelerator mass spectrometry of small biological samples

Authors

  • Mehran Salehpour,

    Corresponding author
    1. Ion Physics, Ångström Laboratory, Department of Engineering Sciences, Box 534, Uppsala University, SE-751 21 Uppsala, Sweden
    • Ion Physics, Ångström Laboratory, Department of Engineering Sciences, Box 534, Uppsala University, SE-751 21 Uppsala, Sweden.
    Search for more papers by this author
  • Niklas Forsgard,

    1. Ion Physics, Ångström Laboratory, Department of Engineering Sciences, Box 534, Uppsala University, SE-751 21 Uppsala, Sweden
    Search for more papers by this author
  • Göran Possnert

    1. Ion Physics, Ångström Laboratory, Department of Engineering Sciences, Box 534, Uppsala University, SE-751 21 Uppsala, Sweden
    Search for more papers by this author

Abstract

Accelerator mass spectrometry (AMS) is an ultra-sensitive technique for isotopic ratio measurements. In the biomedical field, AMS can be used to measure femtomolar concentrations of labeled drugs in body fluids, with direct applications in early drug development such as Microdosing. Likewise, the regenerative properties of cells which are of fundamental significance in stem-cell research can be determined with an accuracy of a few years by AMS analysis of human DNA. However, AMS nominally requires about 1 mg of carbon per sample which is not always available when dealing with specific body substances such as localized, organ-specific DNA samples. Consequently, it is of analytical interest to develop methods for the routine analysis of small samples in the range of a few tens of µg. We have used a 5 MV Pelletron tandem accelerator to study small biological samples using AMS. Different methods are presented and compared. A 12C-carrier sample preparation method is described which is potentially more sensitive and less susceptible to contamination than the standard procedures. Copyright © 2008 John Wiley & Sons, Ltd.

Accelerator mass spectrometry (AMS) is an analytical method for determining isotopic ratios with exceptional dynamic range and sensitivity. Medium energy particle accelerators (∼1 MeV/nucleon) are used to eliminate background molecular isobaric species, through non-adiabatic collisions which is, by and large, the limiting sensitivity factor in conventional mass spectrometry techniques. Another unique aspect of AMS is the capability of detecting the isotopes, utilizing different methods. For example, 14C is detected with a single particle detector, whereas 12C is measured using a current detector achieving routine isotopic ratio determination in the 1:1015 range. In an AMS setup, the whole accelerator is used as a mass spectrometer.

Some biological applications of AMS involve the measurement of 14C-labeled organic compounds. The use of AMS for biological studies has been pioneered by Vogel et al. at the Lawrence Livermore National Laboratory, LLNL.1, 2 We have recently applied AMS to pharmaceutical substances in human blood and demonstrated sensitivity down to about 100 zeptomole (10−21 mole) of the drug3 and concentrations approaching attomolar levels in human blood.4 An interesting biological application of AMS is the so-called Microdosing method, which involves the administration of small amounts of drugs (far below therapeutic level) in order to minimize toxic effects. When applied directly to humans it is possible to obtain pharmacokinetic information such as ADME (Absorption, Distribution, Metabolism and Excretion) data. As the data can be obtained at a very early stage, before the phase 1 clinical trials, this is also referred to as the Human Phase 0 Clinical Trials. Microdosing is of great interest to the pharmaceutical industry, as it offers the potential of shortening the expensive path to new drugs. Microdosing has been approved by the FDA5 and the EMEA.6

Another application of AMS is in stem-cell and regenerative medicine research. There are over 200 different types of cells in the human body each fulfilling a particular function and each having specific regenerative properties. Some cells such as hair, nail, skin and intestinal cells are continuously replaced. Other cells have a much more limited or seemingly non-existent turn-over rate such as neurons in the human brain.7 It is still unknown, for example, whether the heart can regenerate new cells. In regenerative medicine and stem-cell research, it is imperative to have data on the regenerative properties of the different cells in the body but this data is not yet available. Frisén's group at the Karolinska Institute (KI) in Stockholm have pioneered the application of AMS to the determination of the age of the human cells,8–10 and the cells can now be unambiguously dated with an accuracy of a few years. They have shown, among other things, that some parts of the brain produce new cells whereas other parts are as old as the individual and are never replaced.

The principle used is based on the release of large amounts of carbon-14 in the atmosphere during the 1950s and 1960s as a result of the nuclear bomb tests which nearly doubled the 14C to 12C isotopic ratio. This increased isotopic ratio has been transcribed in the DNA molecules of living organisms as a result of the ecological cycle. As the DNA molecule is stable and the chronological profile of the 14C to 12C ratio in living matter is well documented, one can use the AMS isotopic ratio measurement to date the DNA molecules. The method has also been applied in forensic science to determine the ages of the 2004 South East Asia Tsunami victims by performing AMS analysis of the teeth.10

A number of diseases are caused by the build-up of diverse tissues in the body's fluid transportation channels, which in due course can give rise to diverse dysfunctions. A classical case is the clogging of the coronary arteries. The human nervous system, and in particular the brain, is exceptionally susceptible to the clogging of channels which can lead to different diseases. One example is Alzheimer's disease where the neuritic plaque (also called senile dendritic or armyloid plaque) and the neurofibrillary tangles (consisting of e.g. tau proteins) are believed to interfere with cellular functioning. Some preliminary work has been done on this topic11 but it is still not known whether such structures are developed in late adulthood or if it is a continual process from childhood. These formations contain information about the time of deposition and build-up time profile which can be accessed using AMS. As with DNA, it is not possible to extract large amounts of this material from the human brain. To be able to study obstructive tissues with AMS and find the formation time would be of great fundamental value and of importance for the understanding of the disease. Similarly, in multiple sclerosis the formation of inclusions and lesions in the central nervous system is well known but the time-frame over which they are developed is unknown.

In spite of the interesting development in these applications, one analytical limitation of the biological AMS measurement method is the requirement to have about 0.5–1 mg of carbon per sample. Most of the DNA samples or the obstructive tissues are, however, only available in the sub-100 µg range. This is the topic of this manuscript in which data for different methods of sample preparation are presented.

EXPERIMENTAL

The schematic diagram of the experimental setup is shown in Fig. 1. The samples are prepared by graphitization and are introduced into the ion source of the accelerator mass spectrometer. Secondary ions are produced and accelerated to high energies and the isotopic ratio is measured. The different components of the experimental procedure will be separately described below.

Figure 1.

Schematic diagram of a biological accelerator mass spectrometry experiment, showing the different steps involved in the measurement of small samples.

Sample preparation

Sample preparation for the accelerator mass spectrometry (AMS) analysis of organic samples is a well-established discipline which has gone through methodological characterization and semi-empirical improvements since the mid 1980s. The sample preparation methods face the stringent requirement of producing µAs of negatively charged carbon ions, 12,13,14C, and sustaining the sputtering yield in a reproducible and predictable manner. The different methods have the commonality of turning the organic sample into CO2 and subsequently reducing the gas catalytically into graphite form which is ideal to use as targets in the AMS ion sources. We have used three sample preparation methods, all based on oxidation of the samples to CO2 with subsequent reduction into graphite, as discussed below.

Method 1

For routine, high-throughput sample production, we use the method of Ognibene et al.,12 which has been modified for the preparation of small samples3 and will only be briefly described here. The samples are placed in quartz tubes with CuO powder (∼100 mg), dried, evacuated to about 5 · 10−4 mbar and vacuum-sealed. The quartz tubes are then heated to 950°C for 3 h and allowed to cool slowly. The produced CO2 gas is transferred into septa-sealed borosilicate vials containing zinc (100 mg) and iron powder (1 mg) by cryogenic trapping through a hypodermic needle. The pressure prior to gas transfer is about 2 · 10−5 mbar. The septa-seal vials are removed and heated to 530°C for 6 h where the graphitization takes place as catalytic deposition of carbon (graphite) aggregates on the iron powder. The samples are finally collected and pressed into aluminium cathode targets (1 mm in diameter), placed in a holder and sent for AMS analysis. Nominally, about 1 mg of carbon is used for each sample corresponding to 10 µL of blood (10% carbon content), 25 µL plasma (4.5% carbon content), 100 µL of urine (about 1% carbon) or 3 mg of dried DNA (30% carbon).

Method 2

Method 2 is an adaptation of method 1 above with the distinct difference of not using any plastic parts such as the tubes, connections, seals and valves, which have been replaced with stainless steel. The reason for this choice is that the small samples are susceptible, specifically to 12C contaminants from miscellaneous plastic parts which are introduced during handling and gas transfer.

The biological samples are dried in vacuum and are then pumped with a turbo-molecular pump until a pressure of less than 5 · 10−4 mbar is reached. The larger liquid samples such as DNA solutions in water are freeze-dried in vacuum. The samples are weighed and placed in prebaked quartz tubes with about 100 mg CuO powder and are then vacuum-sealed using a high-temperature hydrogen and oxygen torch. The sealed quartz tubes, which have one end as a break-seal, are placed in an oven and heated to 950°C for 3 h and allowed to cool. The quartz tubes are then connected to a vacuum system with the break-seal end into a metallic bellow which can be bent to puncture the break-seal and release the gas. Once the vacuum has reached the 2 · 10−5 mbar region, the vacuum pump valve is closed and the sample gas (mostly CO2) is released and is cryogenically transferred into a vial through a vacuum plug valve (Swagelok, Solon, OH, USA). After the gas transfer, the valve is closed, trapping the gas in the graphitization reactor consisting of a 1 mL borosilicate vial with 80 mg of zinc and a smaller vial containing iron powder (1 mg) separated by a few borosilicate balls. The vial containing the sample gas is removed and placed in a heater block with the upper part of the vial, including the valve, remaining at room temperature. The vial is heated to 530°C for 6 h where the graphitization takes place through catalytic deposition of carbon (graphite) aggregates on the iron powder. The samples are finally collected and pressed into an aluminum holder and sent for AMS analysis.

Carbon-carrier method

The carbon-carrier method is used whenever the amount of carbon in the sample is insufficient to prepare graphite targets. This method has been traditionally applied exclusively to high-performance liquid chromatography (HPLC) fractions,13 e.g. body fluids, containing 14C-labeled drugs. The HPLC fractions typically have very small amounts of the labeled drug: in the pg to fg range. The specific activities of the drugs are, however, very high, usually in the tens of Ci/mol; this translates into the 109 Modern range, or equivalent to almost every drug molecule having a 14C atom (1 Modern corresponds to present day 14C content which is approximately 98 attomole 14C/mg 12C). The carbon-carrier method therefore facilitates accurate isotopic dilution which allows mass determination of the HPLC fraction. The carriers are chosen based on their chemical properties and 14C-deficiency. Samples are prepared with mg-size composites of the carbon carrier and the HPLC fraction. The AMS measurements register the increase in the isotopic ratio caused by the labeled drug. The increase is dependent on both the activity and the amount of the drug. Here, we apply the carrier method to samples with much lower activities in the 1–2 Modern range, but at higher masses in the µg range

As the carbon carrier, we have chosen a petrochemical compound tributyrin (TRB, C15H26O6). The carbon-14 content varies significantly depending on the vendor, by as much as a factor of 50. We have used TRB from MP Biomedicals Inc. (Solon, OH, USA) which had the lowest 14C content (<500 zmol of 14C per mg 12C). We have used typically 1.5 mg of TRB corresponding to about of 1 mg of carbon per sample. The composite sample of TRB and the DNA samples are then treated as an AMS sample and undergo the usual oxidation and reduction process to convert the sample into graphite according to method 1 or 2. The addition of DNA will increase the isotopic ratio which is measured and the amount is determined.

In all cases special care must be taken to avoid contamination by any carbon-containing substances or tools in the laboratory which could affect the isotopic ratio. All vials are prebaked; 3 h at 950°C for quartz vials and 6 h at 450°C for borosilicate vials and parts.

Accelerator mass spectrometry

The experimental setup has been described in detail elsewhere3 and will only be briefly described here. For the AMS studies we have used the Uppsala University 5 MV Pelletron tandem accelerator (NEC Inc., Middleton, WI, USA) which was commissioned for use in 2001. The ion sources used for these studies are negative ion sputter ion sources which use primary ions in the keV energy range to cause the emission of sputtered secondary negative ions. Two independent cesium (Cs+) sputter ion sources are used, both with automated multiple sample holders. One ion source is home-made (25 samples) and is used for samples with high 14C contents. The other is a high-throughput commercial ion source (40 samples, MC-SNICS; NEC Inc.) providing about 3 times higher current than the other ion source. The latter is usually used for samples from archaeological studies with low 14C contents, as well as carrier samples.

The secondary negative ions from the sample are accelerated to a potential of about −50 kV and are mass-to-charge ratio (m/z) analyzed through a 90° magnet. The ions are subsequently steered, focused and injected into the accelerator towards the positive potential of the accelerator terminal (nominally at + 3 MV). The high-energy ions collide with the gas molecules in a low-pressure cell where all molecular entities, including the interfering isomers (e.g. 12,13CH), undergo non-adiabatic ion-electron collisions. Consequently, they are stripped of their electrons resulting in a distribution of atomic positive charge states with no molecular species surviving the collisions. The positively charged ions are now accelerated once again, this time away from the terminal and are mass-to-charge analyzed through another 90° magnet. The beam goes through a number of focusing elements and a 30° switch-magnet. The ions are unambiguously identified in a surface barrier detector where the energy of the ions and the energy loss in a foil are measured. The vacuum throughout the system is in the order of 3 · 10−8 mbar. Isotopic ratios are then measured by single particle counting of 14C in a solid-state detector and current measurements of 13,12C in a Faraday cup. The 14C/12C ratio, R, is presented in units of Modern (M) or as a percentage of Modern (pMC). 1 Modern corresponds to present day 14C content which is approximately 98 attomole 14C/mg 12C, 13.56 dpm/g 12C or 6.1 fCi/mg 12C, depending on the unit of preference.

One 14C-free sample (old carbon) and 2–3 standard reference material samples (oxalic acid II, obtained from NIST, Boulder, CO, USA) are used for every AMS experiment. The samples and the reference samples are normally run for 3 × 5 min after which the reference samples are measured. The precision is in the range of 0.4–0.8% depending on the absolute current from the ion source which affects the statistical fluctuations of the Poisson distribution and hence the measured error.

The combusted samples (CO2 gas) are also analyzed by electron ionization in an isotope ratio mass spectrometer (Fisons/VG-Isotech 652-Optima, Manchester, UK) where the 13C to 12C ratios are measured. Any fractionation effects are compensated for according to the formulation of Stuiver and Polach14 and Mook and van der Plicht.15 For biological samples, however, the isotopic fractionation, δ13C, is normally not significant compared with the experimental uncertainties.

RESULTS AND DISCUSSION

There are two major difficulties when addressing small AMS samples below 100 µg. First, the contamination from the sample preparation and handling becomes significant compared with the size of the sample and hence introduce inaccuracies. The amount of the contamination, depending on the system, is typically between 1 and 10 µg C per sample depending on the specific laboratory contamination level, routines, etc. Secondly, small samples produce lower secondary ion currents than standard samples,16 and this affects the measurement precision detrimentally. One of the parameters in an AMS experiment that is often optimized is the ion current produced by the sample. Higher currents facilitate faster data acquisition times, better statistics and consequently higher precision. We have measured 13C ion current, after a 5 min warm up, from DNA samples from herring sperm (Invitrogen, Carlsbad, CA, USA) in the post-acceleration region, after the analyzing magnet, as a function of the amount of carbon in the sample. Figure 2 shows two different measurements using standard sample preparation methods 1 and 2. It can be seen that for large samples above a few hundred µg of carbon the current is constant at about 1000 nA which is typical for a standard sample in our laboratory. Data for large samples in the 1–10 mg range is not presented as it is well established that large samples have stable current output, independent of sample mass (see, e.g.,16). As the sample size is reduced, the current decreases. Furthermore, the current reproducibility for small samples is poor. Data points for same-size samples show considerable variations. Low current outputs require longer data acquisition times and, therefore, as a rule of thumb, any sample producing less than 10 nA is rejected as the noise, and the statistical errors, become too large.

Figure 2.

The measured 13C current after the acceleration stage as a function of the amount of carbon in herring sperm DNA samples using two different sample preparation methods, method 1 (closed circles) and method 2 (open rhombi), as described in the text.

The initial experiments aimed to characterize the dependence of the isotopic ratio on the sample amount and to determine the level at which the contamination becomes significant. We have chosen method 2 for this purpose: The choice is motivated by the fact that the method avoids use of various plastic parts which are potential contamination sources of carbon, particularly detrimental for small samples. Furthermore, the current produced by this method is higher for small samples (see Fig. 2). Figure 3 shows the measured isotopic ratio (in pMC) as a function of the amount of carbon in herring sperm samples, using sample preparation method 2. Data is presented for three separate sets of samples at different experimental occasions over a period of about 3 months. There are a few observations to be made. (1) The measurements for sample amounts over 100 µg C are quite consistent and different experimental data sets give the same value; (2) Below this limit, different fluctuations and variations are observed. There is a dependence of the isotopic ratio, R, on the sample amount. Although the precision of this method is better than 1% (1σ, standard deviation), the accuracy for small samples is not better than 30% in the few µg C range. In the 50–100 µg C range the accuracy is better than 5%.

Figure 3.

The measured isotopic ratio in percent Modern (pMC) measured as a function of the amount of carbon in herring sperm DNA samples, using method 2 for three different experiments shown in open triangles, half-open squares and closed circles, denoted arbitrarily by B30, B37 and B31.

We are planning a number of improvements, in an effort to reduce the contaminations for small samples, such as (1) better vacuum (<10−6 mbar): during the cryogenic transfer of the CO2 gas; the residual gas in the vacuum system is also frozen and collected which could have a significant contamination contribution; (2) replacement of all the borosilicate parts with quartz. Carbon contaminants, including atmospheric CO2, are adsorbed onto the glass surfaces; therefore, using quartz vials facilitates prebaking at a much higher temperature and rids us of carbon contamination; (3) prebaking in an oxygen environment to improve the oxidation efficiency of the contaminant carbon into CO2; (4) replacement and/or purifications of the chemicals including the CuO, Zn and Fe to reduce the carbon amount; and (5) introducing valves and parts in the system without o-rings which are known to contain 14C-abundant contamination.

Another way of measuring the isotopic ratios for small samples is utilizing the so-called carbon-carrier method13 where tributyrin is added to the sample, as explained in an earlier section. We have adapted this method to DNA samples. Unlike the case with HPLC fractions with very high activities, the DNA samples have much lower activities (1–2 Modern) but are available in considerably higher sample amounts (µg) which are subsequently introduced into the carbon carrier.

The amount of 14C in the composite sample (Σ) consists of 14C originating from the tributyrin (TRB), the DNA and the contamination (γ) introduced during the sample preparation and handling. As the 14C amount is the product of the isotopic ratio (R) and the 12C mass (M), this can be written as:

equation image(1)

As MΣ = (MDNA + Mγ + MTRB), the isotopic ratio of the sample, RΣ, can be written as:

equation image(2)

Here the contamination is defined as the amount of carbon introduced during sample preparation, including sample handling, drying, combustion, graphitization and loading. We are making the following approximations. As the amount of TRB is about 1 mg and Mγ and MDNA are in the µg range, i.e., MDNA ≪ MTRB and Mγ ≪ MTRB, we can approximate MDNA + Mγ + MTRB ≈ MTRB. Since the same sample preparation procedure is implemented for all samples, it can be assumed that the carbon contamination introduced (Mγ) is the same for all samples. Differentiating Eqn. (2) with respect to the carbon mass of the DNA, i.e., δ/ δMDNA, will give rise to:

equation image(3)

This implies that by measuring the isotopic ratio (R) of the composite sample as a function of the carbon mass of the DNA sample, the DNA isotopic ratio can be deduced from the gradient of the curve, independent of the contamination. This is valid as long as Mγ ≪ MTRB, which for a 1 mg tributyrin carrier translates to a contamination about 10 µgC. In our laboratory we are well below this value.16 We have consequently prepared a number of samples with different amounts of DNA which have been added to the carbon carrier, TRB, ensuring that the same amount is used (1.56 ± 0.05 mg) for all samples. After the carrier addition, the samples have been prepared according to method 1 or 2 and measured by AMS. The same procedure was used for all samples. The results for herring sperm are shown in Fig. 4 for two sets of data. Open circles show the measured isotopic ratio for the DNA and carrier samples prepared according to method 1. The closed rhombi represent similar data using method 2. The first point that should be noted is that there is a linear relationship between RΣ and MDNA, as predicted by Eqn. (2). The linear regression fit has a correlation coefficient of 0.99 in both cases. Secondly, it should be noted that the gradients of the two sets of data are similar, indicating that the two sample preparation methods give similar results. Method 1 indicates a higher offset, indicating a higher level of 12C contamination. The latter also has a characteristically higher stochastic scatter.

Figure 4.

The measured isotopic ratio of carrier samples in percent Modern (pMC) measured as a function of the amount of carbon in herring sperm DNA samples using method 1 (open circles) and method 2 (closed rhombi). The lines show least squares fits to the data points.

The value of RDNA can be deducted based on the gradient of the line according to Eqn. (1) as RDNA = MTRB(δRΣ/δMDNA). MTRB is known, as the amount of TRB has been measured to be 1.56 ± 0.05 mg and is deliberately kept the same for all samples. The amount of carbon in TRB is also known (59.6%) and therefore MTRB = 930 µg. We have not measured the carbon content in hydrated DNA but assume it to be 30%.17 The uncertainties with respect to this parameter will be addressed separately. The gradients of the linear regression lines are 0.102 ± 0.01 pMC/µgC for the data points for method 1 and 0.110 ± 0.006 pMC/µg C for the data points for method 2. Correspondingly, the RDNA values are 95 ± 9 pMC and 102 ± 5 pMC, respectively. For the latter we have excluded the data point with the highest carbon mass which could challenge the approximation MDNA ≪ MTRB. The RDNA value including this point would be 96 ± 3 pMC. It should be mentioned that due to experimental time constraints the acquisition times were shorter than normal for the data using method 1, which is the reason for the larger error bars. The values obtained using this method are well within the corresponding values obtained using standard AMS sample preparation using mg amounts, namely R = 101.2 ± 0.4 pMC. The total amount of sample used with the carbon-carrier method has been about 50 µg for five data points for subsequent linear regression. Potentially, the carrier method requires significantly less sample amount as three data points at lower masses should suffice for the determination of RDNA.

In an effort to demonstrate the absolute reproducibility of the data, after about 6 months we repeated the measurements in Fig. 4 in a different sample preparation setup. The data is shown as open circles in Fig. 5, using sample preparation method 2. The dotted line represents the least squares fit that was implemented to the data in Fig. 4 for comparison. It can be seen that the data is reproducible. It is of interest to compare the data for additives with different isotopic ratios in order to study the resolving characteristics of this method. Similar experiments have been performed with human DNA and 14C-labeled human blood. The human DNA used was from the cerebellum of an individual born in 1983 and was purified as described in Spalding et al.8 (obtained from K. Spalding and J. Frisén, Karolinska Institute, Stockholm, Sweden) The DNA sample has an expected R of about 120 pMC, as the cerebellum DNA is expected to be as old as the individual.8 The corresponding carrier method data is shown in Fig. 5 as closed rhombi. The line can clearly be resolved from the herring sperm data (101 pMC). The least squares fit has a correlation coefficient of 0.98 and the gradient is 0.1475 ± 0.006 pMC/µg carbon corresponding to 137 ± 5 pMC. The discrepancy will be addressed shortly. Similar measurements were performed for a 14C-labeled anti-psychotic drug, remoxiprid, C16H23BrN2O3 (MW = 371, AstraZeneca, Södertälje, Sweden), in human plasma. The drug has been 14C-marked with a specific activity of 2.035 GBq/mmol, dissolved in water and then diluted in human plasma. The plasma sample of interest was available in large quantities and the isotopic ratio was first measured using standard AMS as 235.6 ± 1.1 pMC. Carbon-carrier measurements were performed with this solution and the data is shown in Fig. 5 as half-closed squares. The gradient of the linear regression is 0.2316 ± 0.019 which gives an R value of 215 ± 17 pMC, which is just outside the 1σ error bar of the standard AMS sample. It is noted that the line has an offset which is significantly higher than the measured TRB value. We have not identified the origin of this contamination which is a carbon source independent of the sample size. However, it is interesting to note that the method is insensitive to such contaminations, as shown in Eqn. (3), which otherwise can notably offset single measurements using standard AMS.

Figure 5.

The measured isotopic ratio of carrier samples in percent Modern (pMC) measured as a function of the amount of added carbon from three different samples: (1) Herring sperm DNA shown in open circles. The dotted line displays the least squares fit from the earlier data from Fig. 4. (2) Human cerebellum DNA (closed rhombi) and (3) spiked human plasma: 14C-labeled drug in human plasma (half-closed squares). All data points were obtained using method 2.

Blank samples were also studied, i.e. preparing carrier samples with the same water as was used for the DNA samples. We have prepared 10, 60 and 120 µL water samples according to the carrier method described above and no measurable differences were observed compared with the isotopic ratio of the tributyrin carrier (0.25 ± 0.2 pMC). The three data points are shown in Fig. 5 (zero added carbon). It should also be noted that the data in Fig. 5 demonstrate that changes in the isotopic ratio can be measured for DNA samples down to a few µg. The point with the lowest mass of 1.4 µg C has an isotopic ratio which can clearly be separated from the blank sample. This corresponds to 140 zmol of 14C.

The isotopic ratio determined for the human cerebellum DNA in Fig. 5 is overestimated by almost 15% compared with the value measured by standard AMS. Examining the experimental protocol it was found that prior to sealing the quartz tube for combustion, the TRB and the samples were pumped for considerably longer than normal (2 h). We measured the evaporation rate of TRB (boiling point of 305°C) in the quartz tubes in vacuum and estimated it to be about 0.1–0.2 mg/h. This is significant enough to change the mass of tributyrin. It is known that by reducing the mass of the carrier the isotopic ratio is increased (e.g. see Salehpour et al.16). Another point which should be noted is that we have not measured the carbon content of the hydrated DNA sample. Any deviations from the assumed value would affect the deduced isotopic ratios. Hence, in order to improve the accuracy of this method, the carbon content of the sample should be determined. Finally, the precision of the carbon-carrier method can be improved further by implementing more stringent laboratory routines to prevent contamination which affect the sample-to-sample fluctuations. The sensitivity of the method can be potentially increased by more than a factor of ten using smaller TRB carrier amounts, which will directly affect the response of the carrier sample. These mentioned improvements would facilitate the use of the carbon-carrier method for smaller samples, in the µg C range.

SUMMARY AND CONCLUSIONS

In the first set of experiments, we have measured the isotopic ratio of a number of samples using standard AMS sample preparation. Samples with masses below 100 µg of carbon should be treated with care as both the precision and the accuracy are compromised; the former due to lower currents and thus higher statistical errors, and the latter due to external contaminations during sample preparation. Improvement pathways are presented and discussed. In the second set of experiments we have used a carbon-carrier method. This method has traditionally been used for samples, such as HPLC fractions from labeled compounds, with very small masses (fg) but very high specific activities (109 Modern). We have used a similar setup compromising samples with lower activities (1–2 Modern) but higher masses (µg). This method is very sensitive to small amounts of DNA or 14C-labeled drugs in human blood, e.g. DNA samples containing 1.4 µg C, corresponding to 140 zmol of 14C, give rise to clearly measurable changes in the isotopic ratio.

The carbon-carrier method shows a precision of a few percent (1σ) for DNA samples but the accuracy is, at the moment, not better than 15% as complementary measurements are required. Explicitly, the amount of carbon and the carrier need to be determined more accurately. The method has the potential to improve the accuracy to levels similar to the standard AMS method. In particular, the carrier method is an alternative for small samples in the few µg C range. Specifically, the relative insensitivity of the method to systematic contamination including handling and sample preparation is of interest. Alternatively, this method can be applied to biomedical applications where very small amounts of 14C-labeled samples, with moderate specific molar activities, are accessible in very small amounts down to 100 ag.

Acknowledgements

Uppsala BIO, Uppsala, Sweden, is gratefully acknowledged for the funding of the project and Dr Kirsty Spalding and Prof. Jonas Frisén for providing the DNA samples. We also would like to thank Dr Ira Palminge-Hallén (from the Swedish Medical Products Agency) and Lars Ståhle (from AstraZeneca in Södertälje) for providing us with the labeled substance.

Ancillary