All water was 18 MΩ-cm MilliQ purified and was used to produce all aqueous solutions. Hydrochloric acid solutions were prepared from concentrated HCl (TraceSelect, Sigma).

A 55 g/cm² tantalum foil target was irradiated by a 1.4 GeV proton beam, creating a multitude of radioactive and stable spallation products. The product nuclei, lanthanides in particular, diffused from the ≈2,000°C target to a ≈2,000°C tungsten surface ionizer. The ions were extracted at 30 kV and mass-separated with a 70° sector magnet (1.5 m mean bending radius). The A = 140 beam was implanted into two Zn-coated gold foils. The zinc, totaling 2–3 mg had been electrodeposited onto the gold foils over approximately 0.5 cm². The entire procedure is nearly identical to the methodology recently described for projects collecting Tb isotopes at ISOLDE (3, 17).

The following methodology was carried out two times with small variations between each run.

The zinc layer containing ¹⁴⁰Nd was etched briefly with aq. HCl (200 µL, 2 M), and the resulting solution was diluted to 2.2 mL with aq. HCl (2 M). This was heated to 98°C, cooled, and passed over AG1x8 anion exchange resin (300 mg, Biorad, 200–400 mesh, initially formate-form) packed in a 4 mm internal diameter (ID), fritted polypropylene column (Supelco) that had been prepped by three times sequential washing with water (3× bed volume), 2 M HCl (3× bed volume), and 6 M HCl (3× bed volume), finishing with equilibration in 2 M HCl. The resin has a high affinity for [ZnCl₄]²⁻ ions in 2 M HCl (25) and was intended to remove any zinc impurity from the ¹⁴⁰Nd. An additional 0.5 mL of 2 M HCl was used to rinse residual ¹⁴⁰Nd from the column, and the entire effluent was collected and adjusted to pH 5–6 with aq. ammonium acetate (1 M, pH 7) and aq. ammonium hydroxide (28%, TraceSelect, Sigma) to a final acetate concentration of roughly 200 mM. This solution was passed over a hydroxamate functionalized Waters CM resin bed (26) (100 mg, 4 mm ID) to trap the ¹⁴⁰Nd and was washed with water (7 mL), then eluted with aq. HCl (600 µL, 0.1 M). The eluent was kept as the stock ¹⁴⁰Nd solution for further radiolabeling and formulation.

For the DOTA-LM3 preparation, 220 µL of the ¹⁴⁰Nd stock solution was added to aq. ammonium acetate (780 µL, 300 mM). Two productions were made, the first with 9 µg DOTA-LM3 (added from a stock solution of 1 mg/mL in water) and the second with 18 µg DOTA-LM3. The solution was heated in a sealed container to 95°C and incubated for 25 min. After cooling to room temperature, the labeling reaction was quenched with aq. DTPA (12 µL, 0.5 mM, pH 7) and let stand for 5 min. The solution was passed over a Waters C18 light sep-pak (prepped with 10 mL ethanol and 10 mL water) to trap the labeled product, rinsed with 1 mL water, and then eluted with 2 mL ethanol. The ethanol solution was taken almost to dryness (residual volume was approximately 50 µL) with the residual being diluted with 900 µL HEPES buffered isotonic saline (10 mM HEPES, 150 mM NaCl, pH 7.4) and was used directly for injections.

The neodymium “chloride” injections were prepared by diluting 100 µL of ¹⁴⁰Nd stock (in 0.1 M HCl) with 390 µL HEPES buffered isotonic saline (10 mM HEPES), and neutralizing with 10 µL 1 M Na-HEPES.

Samples of ¹⁴⁰Nd-DOTA-LM3 were injected onto a reverse-phase C-18 (Luna 3uC18(2)(n) 100 A 100 × 2 mm 3 μm, Phenomenex) column at a flow rate of 0.5 mL/min starting from 0% acetonitrile in water, and reaching 100% over a 15 min gradient. Elution was monitored with a radio detector. The entire effluent was collected in 1 min intervals (500 µL each) and quantified 4 days after collection by liquid scintillation counting on a HIDEX 300 SL spectrometer.

RadioTLC was performed by spotting 1 µL of the DOTA-LM3 solutions (before and after C-18 purification) onto aluminum-backed silica TLC sheets. The sheets were eluted in 10% (w/v) aq. CH₃COONa:CH₃OH (1:1). Unreacted ¹⁴⁰Nd remained at the origin, and ¹⁴⁰Nd-DOTA-LM3 moved to R_f ~0.5.

NCI-H727 lung carcinoid cancer cells (ATCC CRL-5815, LGC Standards) were cultured in RPMI-1640 media supplemented with 10% fetal bovine serum and 1% penicillin-streptomycin (Invitrogen) at 37° C and 5% CO₂. Cells in their exponential growth phase and at 80–90% confluence were harvested by trypsinization and resuspended in 1:1 media and matrigel (BD Biosciences) at 5 × 10⁷ cells/mL. Subcutaneous tumors were established in female NMRI nude mice (Taconic, Denmark) by inoculation of 5 × 10⁶ cells in 100 µL on each flank above the hind limbs in the subcutaneous space. All animal experiments were performed under a protocol approved by the National Animal Experiments Inspectorate of Denmark.

Longitudinal small animal PET/CT imaging (Inveon Multimodality PET/CT scanner, Siemens) was performed with NCI-H727 tumor bearing mice injected intravenously with 3.3–4.3 MBq ¹⁴⁰Nd-DOTA-LM3 (n = 8) or 2.7–3.1 MBq ¹⁴⁰Nd-chloride (n = 3) in 150 µL. Mice were anesthetized with sevoflurane (Abbott Laboratories) during injection and PET/CT imaging. PET data was acquired for 600 s in list mode at 1, 3, and 16 h after injection. The mice were sacrificed after the 16 h time-point and a PET acquisition was performed 2 h post-mortem. Images were reconstructed using a 3D maximum a posteriori algorithm with CT based attenuation correction. CT images were acquired with the following settings: 300 projections, 65 kV, 500 µA, and 400 ms exposure, and reconstructed with an isotropic voxel size of 105 µm. Image analysis was performed using the Inveon Software (Siemens). Region of interests (ROIs) were drawn manually over the tumor regions and other organs based on the CT images and the uptake of ¹⁴⁰Nd-DOTA-LM3 or ¹⁴⁰Nd-chloride quantified as % injected dose per gram tissue (%ID/g).

Conventional ex vivo biodistribution was performed after the post-mortem scan. Tumors and organs were resected, weighted and the radioactivity was counted in a gamma counter (Wizard², PerkinElmer).

¹⁴⁰Nd was produced by 1.4 GeV proton induced spallation of tantalum at ISOLDE. The process for vaporization and ionization of lanthanides at the ISOLDE facility is well described (15), and proceeded without complication. The electromagnetic separation of proton rich A = 140 spallation products led to a total of about 530 MBq (in two productions) of >99% radionuclidic purity ¹⁴⁰Nd in two Zn-coated gold foils. The ¹⁴⁰Nd implanted foils were briefly etched (without fully dissolving the entire Zn layer) with aq. hydrochloric acid (HCl, 2 M) and the carrier Zn (approximately 1 mg Zn²⁺ in 2 mL 2 M HCl) was removed by passage over AG1x8 anion exchange resin. ICP-OES measurement showed that Zn was completely adsorbed onto the resin, with <30 ng Zn remaining in the purified ¹⁴⁰Nd stock solution (~60 MBq). ¹⁴⁰Nd was concentrated by trap-and-release on a small mixed-bed hydroxamate/carboxylate-functionalized resin. Trapping was only efficient after heating the solution for several minutes at 95°C, indicating that the ¹⁴⁰Nd may not have been completely dissolved during the initial Zn etching. When the 2 M HCl etch solutions containing the Zn were heated prior to purification, the trapping on the hydroxamate/carboxylate resin exceeded 99% efficiency. The release of ¹⁴⁰Nd from the resin was accomplished by elution with aq. HCl (600 µL, 0.1 M) at 98% efficiency.

The eluted ¹⁴⁰Nd was reacted with DOTA-LM3 in ammonium acetate buffer (300 mM, pH 4.8), and after 30–60 min at 95°C radioTLC indicated that ¹⁴⁰Nd-DOTA-LM3 had formed in a 75% radiochemical yield. Quenching with DTPA and C-18 sep-pak purification led to an ultimate combined radiochemical yield and recovery of 60% in HEPES-buffered saline (pH 7.4). RadioTLC after C-18 purification indicated >95% radiochemical purity of ¹⁴⁰Nd-DOTA-LM3. The production, purification, and labeling procedure was performed twice, where the amount of peptide relative to radioactivity was selected based upon titration. Although it would have been ideal to have all samples at identical specific activity, it was not possible, and in this case, the final radiolabeled specific activities for ¹⁴⁰Nd-DOTA-LM3 were 5.0 and 2.5 MBq/nmol, each with sufficient activity for injection of four mice (see below).

¹⁴⁰Nd in an unchelated form was prepared for injection by pH adjustment of the eluted ¹⁴⁰Nd stock solution with sodium HEPES and diluted to a final formulation in pH 7.4, 150 mM NaCl, and 10 mM HEPES. Characterization of the ¹⁴⁰Nd chemical species was not performed, and it is herein referred to as “¹⁴⁰Nd-chloride” for convenience.

Purified ¹⁴⁰Nd-DOTA-LM3 was analyzed on reverse-phase HPLC to highlight the parent-daughter dechelation effect. The relative quantifications of the HPLC effluent are shown overlaid in Figure 2. The product ¹⁴⁰Nd-DOTA-LM3 eluted at 7.9 min, which agreed with the equilibrated liquid scintillation counter (LSC) trace, confirming >95% radiochemical purity of the ¹⁴⁰Nd-DOTA-LM3. It should be noted that parent-daughter transient equilibrium was reached before the LSC samples were counted, meaning that the LSC signal was representative of the parent ¹⁴⁰Nd elution profile. In contrast, the online radio-detector trace demonstrated the daughter ¹⁴⁰Pr behavior. This is because the detector was more sensitive to gamma radiation arising from positron annihilation after ¹⁴⁰Pr decay, and because of the relatively low abundance of penetrating radiation arising from ¹⁴⁰Nd decay. Here, ¹⁴⁰Pr was shown to elute with the solvent front with an elevated baseline between the solvent peak and elution of the parent ¹⁴⁰Nd-DOTA-LM3. Since the ¹⁴⁰Pr³⁺ eluted faster than the radiolabeled peptide, the solvent front peak was evidence of the formation of ¹⁴⁰Pr³⁺ in the injected solution, and the elevated baseline showed the in situ formation and washout of ¹⁴⁰Pr³⁺ from ¹⁴⁰Nd decaying on the column. From the trace, it was evident that the release of ¹⁴⁰Pr from DOTA-LM3 after ¹⁴⁰Nd decay is >95% efficient, matching the observations of Zhernosekov and co-workers (13). Furthermore, the column behavior illustrated the expected in vivo behavior: a rapid redistribution of the positron-emitting daughter after the decay of the parent.

Eight mice bearing dual-flank NCI-H727 lung carcinoid tumors were injected with 3–4 MBq ¹⁴⁰Nd-DOTA-LM3: four at a specific activity of 5 MBq/nmol and four at 2.5 MBq/nmol. A further three tumor-bearing mice were injected with 3 MBq ¹⁴⁰Nd-chloride. PET quantifications were obtained at 1, 3, and 16 h postinjection. After the last scan the animals were euthanized, and following equilibration of the daughter, the mice were rescanned. Finally, the animals were dissected, and tissue samples were weighed and counted.

For the unbound ¹⁴⁰Nd-chloride injections, the ex vivo biodistribution (16 h) showed high levels of accumulation in the lungs, spleen and liver, and to a lesser extent bone and tumor (Figure 3). This is a typical biodistribution for free +3 oxidation state radiolanthanides [for femur and liver vs. tumor, see, Ref. (27)] and for other hard radiometals [e.g., see, Ref. (24)]. The tissues with high accumulation of ¹⁴⁰Nd were also expected to accumulate released ¹⁴⁰Pr, owing to the chemical similarities between praseodymium and neodymium. Thus, this biodistribution serves as a descriptor of which tissues we expect to exhibit “sink” behavior in the pre-mortem PET scans.

The ex vivo biodistribution of ¹⁴⁰Nd-DOTA-LM3 was very distinct from that of ¹⁴⁰Nd-chloride with the pancreas, a known somatostatin receptor positive organ, being particularly interesting (21). The pancreas was found to take up ¹⁴⁰Nd-DOTA-LM3 without accumulating the free radiometal (13% ID/g with high specific activity ¹⁴⁰Nd-DOTA-LM3, compared to 0.1% ID/g with the free ¹⁴⁰Nd³⁺). Therefore, in the PET study, the pancreas was expected to have a lower signal in the pre-mortem ¹⁴⁰Nd-DOTA-LM3 scans compared to the post-mortem scans. The liver, spleen, and lungs had the opposite behavior, accumulating the free radiometal but not the peptide, and were expected to have a higher signal in the pre-mortem scans than in the post-mortem.

Fani et al. showed that in comparing the biodistribution of ⁶⁸Ga-NODAGA-LM3, ⁶⁸Ga-DOTA-LM3, ⁶⁴Cu-NODAGA-LM3, and ⁶⁴Cu-CBTE2A-LM3, tumor uptake relative to the pancreas, stomach, and kidney is highly variable depending on the metal and chelator used (23, 28). In the current work, we injected 300 pmol (5 MBq/nmol) or 600 pmol (2.5 MBq/nmol) in order to get enough signal for the PET scans. For the biodistributions from Fani et al., 10 pmol was injected, and blocking was performed with 200 nmol of excess DOTA-LM3. The overall effect of the blocking was to reduce the uptake in sst2 expressing tissues, which is exactly what was observed between the higher and lower specific activity injections of the current study: a modest reduction in the uptake in sst2 expressing tissues relative to the kidney, and overall faster excretion.

Based upon the ex vivo biodistribution, the tumors had an intermediate behavior, weakly concentrating both the free ¹⁴⁰Nd and the labeled peptide, meaning that the PET signal was expected to change little between the pre- and post-mortem scans (i.e., washout of ¹⁴⁰Pr from the tumor volume could be compensated by uptake of ¹⁴⁰Pr from the bloodstream). This behavior is in many ways an undesirable outcome, because signals arising from the tumors can be attributable to both the free metal ions and the targeted peptide. Future work with a different vector/target system could give a result with a simpler interpretation. Nevertheless, the pancreas remains an interesting tissue within the present set of experiments.

Positron emission tomography data were analyzed to quantify the ¹⁴⁰Pr signal from the tumors, pancreas, kidney, lung, and liver (Table 1). The PET signals for the tumors are given as the percent of the injected signal per gram (%IS/g) (Figure 4). This unit is non-standard and is described further in the Supplementary Material. %IS/g was chosen for two reasons: first because the PET scans quantify the redistributed daughter ¹⁴⁰Pr, not the ¹⁴⁰Nd-tracer, and second because the data were not corrected for point spreading due to extensive positron range. This means that there is a substantial partial volume distortion when converting from annihilation signal (%IS/g) to tracer concentration (%ID/g), and that the well-counter based ex vivo biodistribution is not directly comparable to the post mortem PET results. A discussion of the partial volume effect due to the high energy positrons from ¹⁴⁰Pr is included in the Supplementary Material. The tumor time-activity curve shows that in this model (as expected from the ex vivo biodistribution) very little redistribution is observed. For the present case, the tumor was known to accumulate the trivalent lanthanide to some small degree (from the Nd-chloride injection data above), meaning that ¹⁴⁰Pr³⁺ produced by decay in the tumor region had some tendency to remain. In other organs, however, due to the dechelation effect, there was a dramatic change in the PET signal in the pre- and post-mortem imaging quantifications. The redistribution effects are displayed in Figure 5 as the ratio of the mean organ signal between the post-mortem, and pre-mortem (16 h) images. This effect is most evident in the pancreas and liver for the high specific activity injections, where the decrease in liver signal is matched by an increase in the pancreas signal, confirming the behaviors of the peptide and free metal observed in the ex vivo biodistribution.

It should be noted that the injected mass of the tracer had a large effect upon the PET quantifications and ex vivo biodistribution (Figures 4 and 5). This indicates that the receptor-specific accumulation was becoming saturated as the injected mass increased from 0.5 nmol (5.0 MBq/nmol) to 1 nmol (2.5 MBq/nmol). This saturation behavior is non-desirable for probing the internalization behavior of the peptide as non-specific interactions begin to dominate the tracer distribution.

Figure 6 shows an example PET/CT reconstruction, qualitatively illustrating the redistribution effect. The most striking differences in the images between the daughter distribution (16 h, left panels) and the parent distribution (post-mortem, right panels) are seen in the liver, lungs, and pancreas. As expected, the 16 h pre-mortem PET images more closely reflected the biodistribution observed for the free ¹⁴⁰Nd-chloride injections, while the post-mortem images showed the biodistribution for the intact ¹⁴⁰Nd-DOTA-LM3 tracer.

The most important result from the PET imaging was the pancreatic signal. In this case, in the pre-mortem image, it was difficult to delineate the pancreas due to the elevated-background in the liver. This is despite the fact that the ex vivo biodistribution revealed that the pancreas contained 13%ID/g of the tracer: which should be easily distinguishable from the 3%ID/g of the liver. The discrepancy is resolved in the post-mortem imaging, where the accumulation of the tracer in the pancreas, and not in the liver, is apparent. The effects observed are consistent with a hypothesis that ¹⁴⁰Nd-DOTA-LM3 localized on the surface of pancreatic cells (via non-internalizing interactions with the somatostatin receptor) releases ¹⁴⁰Pr³⁺ into the blood stream, where it is quickly redistributed to the liver, spleen and lungs. In fact, results from the higher specific activity ¹⁴⁰Nd-DOTA-LM3 studies show that 56 ± 7% (n = 4, mean ± SD) of the in situ produced ¹⁴⁰Pr washed out of the pancreas before decay.

For statistical analysis the pre- and post-mortem PET quantifications were compared with a paired-difference two-tailed t-test. The tabulated data are presented in Table 1, along with the p values. In this case, the t-test was used to determine the significance of the absolute difference between the pre- and post-mortem signals in %IS/g. Admittedly, there are many ways to analyze these data, and in this case, the paired-difference test was selected because it adds to the statistical power by comparing the tissues in a single subject to themselves after intervention. Clearly from the p values in Table 1, the use of DOTA-LM3 as a tracer led to significant changes from the pre- to post-mortem images, while the non-targeted ¹⁴⁰Nd-chloride remained largely unaltered. Of note in the free ion ¹⁴⁰Nd-chloride injections, is that the only tissue with a statistically significant difference between the pre- and post-mortem quantification is the kidney, whereas in the targeted DOTA-LM3 images the only tissue lacking a significant difference was the muscle. The kidney is interesting because in all cases the post-mortem signal was higher, by 25–50% than the pre-mortem value. While it could be suggested that this is due to rapid excretion of ¹⁴⁰Pr³⁺ from the kidney to the bladder in vivo, the fact that this is also observed with the ¹⁴⁰Nd-chloride injections indicates that within the kidney the chemical form of the neodymium is not necessarily as a free cationic lanthanide. However, in all other tissues, the distribution of ¹⁴⁰Nd from the ¹⁴⁰Nd-chloride injections strongly resembles that of redistributed ¹⁴⁰Pr.

The pancreatic washout reveals a potential benefit derived from the ¹⁴⁰Pr’s delayed positron. Specifically, the degree of redistribution of ¹⁴⁰Pr³⁺ may be affected by its location and access to blood flow. This means that the PET signal observed with ¹⁴⁰Nd labeled vectors might be highly dependent on their cellular internalization status, as ¹⁴⁰Pr³⁺ cations originating from decays occurring on the surface of the cell or in circulation may be transported away by the blood flow, whereas¹⁴⁰ Pr³⁺ cations released from ¹⁴⁰Nd decay inside of a cell have an additional diffusion barrier. A conceptual graphic demonstrating the idea is depicted in Figure 7. As many promising new classes of pharmaceuticals, in particular nanoparticle drug formulations, gene therapy and targeted Auger emitting radionuclides, are expected to be most effective when internalized, a PET radiolabel for determining internalization would be a valuable tool for drug development.

When searching for long-lived PET radionuclides, the utility of ¹⁴⁰Nd is immediately evident in its half-life and lack of concurrent gamma emissions. However, due to the nature of the delayed positron from the short-lived daughter, it is important to understand how imaging may be affected by dechelation. The redistribution of ¹⁴⁰Pr in the present case was clearly visible in the pre- and post-mortem PET images. While the tumor signal was significantly changed, the magnitude of change was small which may preclude application. However, this behavior might be model dependent, and not general for all tumor types or locations. Nevertheless, the signals from the other tissues show the potential for using the daughter-delay to determine the in vivo internalization status of new probes, as highlighted by the pancreatic signal. These data support a hypothesis that in certain cases, PET imaging with ¹⁴⁰Nd provides a localized signal only if a vector is internalized. This capability may prove useful in future drug development where in vivo internalization is critical for drug action.

Overall, the statistical analysis proves that the images generated using a DOTA-based ¹⁴⁰Nd/¹⁴⁰Pr in vivo generator are significantly altered from the true distribution of the tracer. With more development, it may be possible to use this technique to determine specific details of the interaction between the tracer and its molecular target. While the p-values are often less than the nominal 0.05 designation for statistical significance, practical significance is weakened by a large inter-subject variability. This means that further development in the ¹⁴⁰Nd in vivo generator system will, for the time-being, remain an invasive procedure relegated to pre-clinical drug development.