Development of a non-radiometric method for measuring the arterial input function of a 11C-labeled PET radiotracer

Positron emission tomography (PET) uses radiotracers to quantify important biochemical parameters in human subjects. A radiotracer arterial input function (AIF) is often essential for converting brain PET data into robust output measures. For radiotracers labeled with carbon-11 (t1/2 = 20.4 min), AIF is routinely determined with radio-HPLC of blood sampled frequently during the PET experiment. There has been no alternative to this logistically demanding method, neither for regular use nor validation. A 11C-labeled tracer is always accompanied by a large excess of non-radioactive tracer known as carrier. In principle, AIF might be obtained by measuring the molar activity (Am; ratio of radioactivity to total mass; Bq/mol) of a radiotracer dose and the time-course of carrier concentration in plasma after radiotracer injection. Here, we implement this principle in a new method for determining AIF, as shown by using [11C]PBR28 as a representative tracer. The method uses liquid chromatography-tandem mass spectrometry for measuring radiotracer Am and then the carrier in plasma sampled regularly over the course of a PET experiment. Am and AIF were determined radiometrically for comparison. The new non-radiometric method is not constrained by the short half-life of carbon-11 and is an attractive alternative to conventional AIF measurement.

Positron emission tomography (PET) is a uniquely valuable molecular imaging modality for noninvasively exploring physiology and biochemistry in health and disease 1,2 , and has an expanding role in drug development 3 and medical diagnosis 4 . PET has notable importance for neuropsychiatric research both for the study of pathophysiology and for drug development. Appropriately designed radiotracers 5 permit sensitive imaging and quantification of many of the proteins in brain 6 that are implicated in neuropsychiatric 7 , neurological 8 , and neurodegenerative disorders 9 , as well as substance dependence 10 . These proteins include various neurotransmitter receptors, transporters, enzymes, and amyloid plaques 3 . PET radiotracers may also be used to verify protein target engagement by experimental drugs and how target engagement varies with dosing regimen 11 . Such information can be critical for establishing meaningful clinical trials. Short-lived carbon-11 (t 1/2 = 20.4 min) or fluorine-18 (t 1/2 = 110 min) are the two most commonly used radionuclides for labeling radiotracers for PET imaging of brain 5 .
Typically, in each PET scanning session, measurement of the radiometabolite-corrected arterial input function (AIF) of the radiotracer is required for use in conjunction with a biomathematical model to robustly quantify a radiotracer target within brain 12 . Nearly all PET radiotracers generate radiometabolites in plasma 5 . The AIF is the time-course of non-metabolized radiotracer in plasma. Virtually all AIF measurement has been based on a single methodology, namely fast radio-high performance liquid chromatography (radio-HPLC) separation of the parent radiotracer from radiometabolites in plasma from multiple blood samples taken serially throughout a PET scanning session for radiometric quantification. When imaging with a 11 C-labeled PET tracer, the time available for measuring AIF by this means is severely constrained to a few half-lives (typically ~ 90 min). This is very logistically demanding. Whereas the dose of a radiotracer administered to a human subject (~ 750 MBq) is measured with an ionization chamber, the low levels of radioactivity found in plasma samples (of the order of kBq) are typically measured with a sensitive γ-counter. The accuracy of radioactivity measured with an ionization chamber or γ-counter depends, among other factors, on the surrogate radioisotope(s) used to calibrate 11 C-labeled tracers: 13 C to 12 C ratio at the radiolabeling site and dependence of this ratio on A m . As part of method development, we set out to measure the ratio of 13 C to 12 C in the carrier of some 11 C-labeled tracers to determine whether there was any variability that might impact on the proposed use of LC-MS/MS for measuring A m and AIF. We showed that we could measure simultaneously all three isotopologues, [ 20 ). The obtained values were within the expected range of about 1.1% per carbon atom in a measured product ion (reference standards, Table 1).

LC-MS/MS investigation of the carrier in
However, when this method was applied to the carrier present in prepared radiotracer doses, we observed [ 13 C] i to [ 12 C] i ratios that were well above the normal range for natural abundance (Table 1). These findings showed that 13 C enrichment had occurred during radiotracer production, which we now attribute to a known nuclear reaction, 14 N(p,2n) 13 C, that would co-exist with the 14 N(p,α) 11 C reaction during the cyclotron production of carbon-11 through irradiation of nitrogen with 16 MeV protons (see "Discussion" section) 21 . In our production setting, the A m value of a PET radiotracer strongly reflects the amount of radioactivity produced during  www.nature.com/scientificreports/ cyclotron irradiation, which in turn depends on the integrated proton beam current (μA × min) 22 . Consequently, we predicted that the [ 13 C] i to [ 12 C] i ratio would increase with the measured A m value of the 11 C-labeled tracer.
In each of four radiotracers, when measuring the fragment ion that contained the radiolabeling site in two separately prepared doses, the [ 13 C] i to [ 12 C] i ratio was higher in the dose that had the higher A m value (determined radiometrically) ( Table 1). One radiotracer (R)-PK11195 gave abundant product ions (m/z 238; 239) that lacked the 11 C-labeling site (i.e., lacked the amido N-methyl group). The [ 13 C] i to [ 12 C] i ratio for these ions was found to be in the range of natural abundance and invariant with the A m value of the radiotracer preparation (Table 1), thereby affirming that changes in 13 C-enrichment were confined to the fragment containing the 11 C-labeling site.  Fig. S1).  Fig. S2). This direct measurement of all three types of isotopologue yielded A m values that closely matched those obtained by measuring only the two isotopologues [ 11 C] i and [ 13 C] i . Therefore, measurement of [ 11 C] i and [ 13 C] i alone plus a separate measurement of the 13 C to 12 C ratio in a radiotracer dose sufficed to provide an accurate A m value.

Plot of [ 13 C] i to [ 12 C] i ratio versus A m (by LC-MS/MS) for [ 11 C]PBR28.
A m values, including the ratio of [ 13 C] i to [ 12 C] i , were determined in 16 preparations of [ 11 C]PBR28. Variations in radiosynthesis time were negligible (36.3 ± 0.82 min; mean ± SD; n = 16) and dose radioactivity was therefore decay-corrected to the end of each synthesis. As a control, the ratio of [ 13 C] i to [ 12 C] i was also measured in reference (natural abundance) PBR28 on each occasion of radiotracer analysis. The ratios of [ 13 C] i to [ 12 C] i for [ 11 C]PBR28 preparations correlated strongly with A m values determined with LC-MS/MS (r = 0.975; p < 0.0001; n = 16) (Fig. 2). The Y-axis intercept of 9.09% (A m = 0) for this curve was almost identical to the mean ratio of [ 13 C] i to [ 12 C] i measured for reference PBR28 (9.16 ± 0.05%; mean ± SD; n = 16; represented by the red line in Fig. 2). The small standard deviations in the latter value and those in Table 1 demonstrate the high precision with which such ratios could be determined with LC-MS/MS.    Figure 3 shows ion chromatograms from the analysis of plasma at baseline, and at 10 and 90 min after intravenous injection of [ 11 C]PBR28 (710 MBq; A m : 383.5 GBq/µmol) in one subject. The plasma matrix did not interfere with the ionization or detection of carrier PBR28 or of the internal standard, and no interfering peak was observed in the ion chroma-  Fig. S4).   Fig. 4c,d, respectively. To examine the impact of different types of AIF measurement on the input function for kinetic modeling of PET data, we compared areas under the curve (AUCs) for the plasma time-activity curves. On average, AUC calculated from LC-MS/MS was 31% lower than that calculated with the radiometric method (t = − 11.9, p < 0.001), and 8% lower than that calculated with the corrected radiometric method (t = − 6.0, p < 0.001) ( Table 4). VAR, Pearson correlation coefficient, and ICC, were respectively 37%, 0.97, and 0.07 for LC-MS/MS versus direct radiometric method, and 8%, 0.99, and 0.91 for LC-MS/MS versus the corrected radiometric method.

Discussion
This study assessed the feasibility of using LC-MS/MS for measuring AIF in human subjects undergoing PET scanning with a 11 C-labeled radiotracer. LC-MS/MS analysis was found to provide a convenient and sensitive method for measuring the AIF of a 11 C-labeled tracer without measuring its radioactivity. This method can be performed on multiple blood samples without the time and logistical constraints of the conventional radiometric method. The A m measured during the production of a radiotracer may be used to transform plasma carrier concentrations into AIF radioactivity data.    13 C-enrichment had occurred in the carrier at the same position that had been labeled with carbon-11 during radiotracer synthesis. As clearly shown for [ 11 C]PBR28, the degree of isotopic enrichment correlated with A m value. Therefore, the augmented [ 13 C] i to [ 12 C] i ratio was not due to an isotope effect in the synthesis or purification of the radiotracer. The carbon-11 for labeling each radiotracer was produced by the 14 N(p,α) 11 C reaction on nitrogen with a 16 MeV beam of protons which degrade in energy on progressing through the gas target. An explanation for the A m -related 13 C-enrichment is the co-occurrence of the 14 N(p,2p) 13 C reaction. For irradiations of nitrogen gas with 13.2 MeV protons, the 14 N(p,2p) 13 C reaction has a total cross section of 74.2 mb (millibarn) which is very similar to that for the 14 N(p,α) 11 C reaction (68.9 mb) 23 . Therefore, the mass of generated carbon-13 is expected to be similar to the mass of carbon-11 produced. For a typical 40-min irradiation producing about 75 GBq of carbon-11, this amount would be about 10 nmol, or roughly enough to explain the 13 C-enrichment seen for carrier in doses of [ 11 C]PBR28.
The A m value of a radiotracer needs to be determined accurately in order to derive mass of carrier from radioactivity or vice versa. A previous study from our laboratory reported an MS/MS technique for isolating [ 11 C] i and [ 13 C] i to measure the A m values for 11 C-labeled tracers 15 . Each A m value was calculated using a constant [ 13 C] i to [ 12 C] i ratio that had been measured for the reference natural abundance ligand. In the present study, the   Table 2). The A m value measured with a γ-counter was used to compare these two sets of data.
These findings further demonstrated that the LC-MS/MS technique can isolate and measure very low kBq levels of [ 11 C] i and, consequently, that this capability can be used to evaluate the accuracy of radioactivity measurements from radiation detectors such as ionization chambers and γ-counters. Building on this work, we compared the A m values of [ 11 C]PBR28 measured with LC-MS/MS with those determined using an ionization chamber and a γ-counter. The A m (mean ± SD) value measured by ionization chamber was close to that obtained via LC-MS/MS, whereas that obtained via γ-counter was appreciably higher. Thus, comparing the three sets of A m data revealed a significant difference in radioactivity estimates between the two commonly used radiometric methods.
Typically, an ionization chamber is calibrated for measuring carbon-11 with a pair of surrogate radioisotope standards, 137 Cs (t 1/2 = 30.17 years; β − , γ 662 keV) and 57 Co (t 1/2 = 271.79 days; ε, γ 122,136 keV), and a γ-counter with a different standard, 68 Ge (t 1/2 = 270.8 days; decays to 68 Ga; t 1/2 = 67.6 min, β + , ε, γ). Clearly, these surrogate isotopes have decay modes that are very different from those of 11 C (β + , ~ 99.8%). Moreover, the accuracy of radioactivity measured with an ionization chamber or γ-counter is well known to be influenced by sample volume and geometry effects 14,24,25 . Studies seeking to measure positron-emitters in ionization chambers have been conducted with fluorine-18 (t 1/2 = 109.8 min) 24,25 , but none have used shorter-lived carbon-11. Indeed, we have previously observed in our facility that identical ionization detectors calibrated in the same way with the same surrogate standards can give estimates of carbon-11 radioactivity differing by up to 12% 15 . During the course of the present study, we thoroughly investigated whether such differences could be ascribed to detector dead-time and linearity, sample geometry, and volume effects, to the material of the sample container (glass or polypropylene), or to measurement time. None of these factors accounted for the observed differences. These results underscore the role that a sensitive MS/MS technique may play in checking radioactivity measured with radiometric techniques. Notably, the MS/MS technique obviates sample volume and geometry concerns.
In the conventional radiometric measurement of AIF, plasma radioactivity is measured with a γ-counter and then corrected for the contribution of radiometabolites, as determined with radio-HPLC analysis 17,18 (for radiochromatogram, Supplementary Fig. S5). The accuracy of AIF determined in this manner depends on the calibration of the γ-counter. The AIF data so generated can be compared with those from the LC-MS/MS of the carrier if equivalence between measured radioactivity and mass of [ 11 C] i has been demonstrated. Here, radioactivity based on the mass of [ 11 C] i was determined from the A m and concentration of carrier (from LC-MS/ MS) in a sample whose radioactivity had been measured with a γ-counter. The γ-counter measurement gave a higher value. Accordingly, it was necessary to adjust the radioactivity measured with a γ-counter to achieve a meaningful comparison of AIF determined from radiometric and LC-MS/MS methods.
LC-MS/MS, with the use of an internal standard labeled with stable isotopes ( 13 C and deuterium), was able to quantify carrier PBR28 in small volumes of plasma (200 µL) taken from human subjects undergoing PET experiments with [ 11 C]PBR28. Calibration of the method used reference PBR28 whose [ 13 C] i to [ 12 C] i ratio is lower than that of carrier PBR28. The unexpectedly skewed [ 13 C] i to [ 12 C] i ratio in the carrier was found to contribute negligible error (< 1%) to measurements of PBR28 concentrations. The LC-MS/MS method was sensitive and specific enough to measure carrier PBR28 up to 90 min after injection of radiotracer with A m values in the range Table 4. AUCs (kBq × min/mL) for the plasma time-activity curves from LC-MS/MS and radiometric measurements. a AUCs from radiometric measurements before (A) and after (B) the correction of radioactivity (for the systematic difference between LC-MS/MS and γ-counter measures). b Between the AUC from LC-MS/ MS and that from the radiometric method for AUC dataset A and AUC dataset B. www.nature.com/scientificreports/ of 139 to 631 GBq/µmol. 11 C-Labeled radiotracers are typically administered with molar activities at the lower end of this range. The entire range of quantification was achieved by injecting as little as 1/20th of each plasma sample onto the LC-MS/MS. If using a radiotracer with an exceptionally higher A m value, quantification could still be achieved by increasing the injection volume, from for example 10 to 25 μL, or concentrating the plasma sample two-fold, although the LC procedure might consequently need some modification. In addition, it is expected that quantification limits would vary with the type of radiotracer carrier being measured. With regards to measuring AIF for [ 11 C]PBR28, when the radioactivity was corrected for the difference between radioactivity measured by γ-counter and by LC-MS/MS, the plasma concentration curves from the two methods matched (Fig. 4). The difference likely occurred because LC-MS/MS performs absolute quantification of the carrier whereas the γ-counter measures radioactivity relative to the surrogate radioisotope used for calibration. The radioactivity (Bq) is given by the product of the decay constant of the radionuclide and the number of un-decayed radioactive atoms. Thus, the LC-MS/MS measurement described here is expected to give absolute radioactivity, given that it is derived from the mass of [ 11 C] i , the decay constant of carbon-11, and Avogadro's number.
The AUCs for plasma time-activity curves from the LC-MS/MS method were 8.1 ± 3.6% (n = 8) lower than those from the radiometric method with corrected radioactivity. Nonetheless, the %RSD of AUCs calculated for 8 subjects was the same for the two methods and showed good correlation (Pearson r = 0.987; p = 0.01). Thus, the LC-MS/MS method is as reproducible as the radiometric method for measuring AIFs.

Conclusion
The LC-MS/MS of fast-decaying PET radiotracers provides interchangeable mass and radioactivity data and offers the convenience of measuring AIF through the carrier of the radiotracer instead of radioactivity. The method, here exemplified with [ 11 C]PBR28, circumvents possible radiometabolite interference and error due to volume and geometry effects associated with radiometric measurements. Potentially, this non-radiometric method might allow measurement of AIF on stored plasma samples by analytical service laboratories that perform LC-MS/MS quantifications. In such instances, AIFs in radioactivity unit can be derived from the A m measured during production of the radiotracer. Taken together, the LC-MS/MS technique poses a convenient, non-radiometric, reproducible, and sensitive method for measuring AIF, deserving of widespread application in the expanding PET imaging field. . Analysis was performed using LC method and MS/MS settings already described 15 . A second transition, m/z 349 → 121, was included in the method requiring acquisition of [ 11 C] i , [ 12 C] i , and [ 13 C] i . In A m measurements based on acquisition of [ 11 C] i and [ 13 C] i , the 13 C peak area of the carrier was converted into the 12 C peak area using the ratio of [ 13 C] i to [ 12 C] i measured in the carrier, where [ 13 C] i includes [ 12 C] i having a single natural abundance 2 H or 17 O atom. A m was determined from the peak areas of radioactive and carrier species as (A*)/(A + A*) × A m *, where A* is the sum of peak areas for [ 11 C] i and for the calculated area for the same species containing carbon-13, A is the sum of the peak areas for [ 12 C] i and [ 13 C] i , and A m * is the theoretical carrier-free A m of carbon-11 (3.413 × 10 20 Bq/mol, the product of ln2/t 1/2 and Avogadro's number) 15 .  Table 1). Radiotracer samples were analyzed after full radioactive decay. Specifically, the sample was diluted (100-500 fold) and injected (5 µL; n = 4) onto the LC-MS/MS. Radiotracer's carrier was chromatographed on a C18 column (2 × 20 mm, 3 µm; Phenomenex, Torrance, CA) using a similar water-acetonitrile (10 mM ammonium acetate or 0.2% acetic acid) gradient as previously reported 15 . The ratio of peak areas for product ion from [ 13 C] i to that from [ 12 C] i , multiplied by 100, gave the ratio [ 13 C] i to [ 12 C] i as a % value.

Ratio of [
Reference PBR28, (R)-rolipram, DPA713, and (R)-PK11195 were analyzed similarly, and the ratio of [ 13 C] i and [ 12 C] i was determined for each.