Introduction
Cell tracking by scintigraphy with radionuclides has been routine in nuclear medicine for 30 years [
1] for tracking autologous leukocytes to detect sites of infection/inflammation [
2,
3]. The standard radiolabelling methodology has been non-specific assimilation of lipophilic, metastable complexes of
111In (with oxine [
4], tropolone [
5] and occasionally other bidentate chelators [
6]), and later
99mTc [
7].
New insight in immunology is creating interest in imaging the migration of individual immune cell types (e.g. eosinophils [
8,
9], neutrophils [
8,
9], T-lymphocytes [
10‐
12], and dendritic cells [
13]) in cancer, atherosclerosis, stroke, transplant and asthma. Regenerative medicine and cell-based therapies are creating new roles for tracking stem cells and chimeric antigen receptor-expressing T-lymphocytes [
14,
15]. Conventional labelling methods have been applied in some of these areas, but for clinical use some of these new applications will require detection of small lesions and small numbers of cells beyond the sensitivity of gamma camera imaging with
111In (e.g. coronary artery disease, diabetes, neurovascular inflammation and thrombus), creating a need for positron-emitting radiolabels to exploit the better sensitivity, quantification and resolution of clinical positron emission tomography (PET).
So far the search for positron-emitting radiolabels for cells has met with limited success. The near-ubiquitous presence of glucose transporters allows labelling with [
18F]fluorodeoxyglucose (FDG), but labelling efficiencies are highly variable, the radiolabel is prone to rapid efflux and the short half-life (110 min) of
18F allows only brief tracking [
16‐
18].
68Ga can be used to label cells [
19] but it too has a short half-life (68 min).
64Cu offers a longer (12 h) half-life and efficient cell labelling using lipophilic tracers (complexes of PTSM [
11,
20‐
22], GTSM [
23], diethyldithiocarbamate [
24] and tropolonate [
25]), but rapid efflux of label from cells is a persistent problem and a still longer half-life would be preferred. A “PET analogue” of [
111In]oxinate
3, capable of cell tracking over 7 days or more, would be highly desirable but is not yet available.
89Zr is a long half-life positron emitter that could meet this need [
26,
27]. The favoured oxidation state of zirconium is 4+ (compared to 3+ for indium), but the parallels between the two metals in reactivity and preferred ligand types suggest that the mechanism exploited to label cells with
111In (i.e. lipophilic metastable chelates entering cells and subsequently dissociating) might be exploited in the case of
89Zr. Tetravalent zirconium forms ZrL
4 complexes with monobasic bidentate ligands such as oxinate [
28], tropolonate [
29] and hydroxamates [
30], analogous to InL
3 [
31,
32]. Here we describe the first synthesis of [
89Zr]oxinate
4 and comparison with [
111In]oxinate
3 for labelling several cell lines and tracking eGFP-5T33 cells in mice. eGFP-5T33 is a syngeneic murine multiple myeloma model originating from the C57Bl/KaLwRij strain [
33], engineered to express enhanced green fluorescent protein (eGFP). It was chosen for this work because the fate of the cells after i.v. inoculation is known from the literature [
34‐
36] and prior work in our laboratory. Like human multiple myeloma, 5T33 develops elevated serum immunoglobulin levels and osteolytic disease [
33,
37]. Intravenously injected cells migrate exclusively to the liver, spleen and bone marrow [
37].
Materials and methods
Radiochemistry
89Zr was supplied as Zr4+ in a 0.1 M oxalic acid (PerkinElmer, Seer Green, UK), brought to pH 7 with 1 M Na2CO3 and diluted to 500 μl with water. This “neutralised [89Zr]oxalate” was used as a control in cell labelling experiments in vitro. To prepare [89Zr]oxinate4 the above neutralised [89Zr]oxalate solution (typically 20–90 MBq) was added to a glass reaction vessel containing 500 μl of a 1 mg/ml 8-hydroxyquinoline solution in chloroform. The vessel was shaken for 15 min and the product [89Zr]oxinate4 was recovered from the chloroform phase by evaporation, redissolved in dimethyl sulfoxide (DMSO, 10–20 μl) and diluted with phosphate-buffered saline (PBS, 1–3 ml) or cell culture medium. Full details are provided in the Electronic Supplementary Material (ESM).
Cell labelling initial evaluation
[
89Zr]Oxinate
4 was evaluated in vitro in three cell lines: cultured J774 mouse macrophages [
38], MDA-MB-231 breast cancer cells [
39] and eGFP-5T33 murine myeloma cells [
40]; and in leukocytes from healthy volunteers. Studies were approved by an independent UK National Research Ethics Committee and complied with the Declaration of Helsinki. Culture methods and leukocyte preparation and labelling are described in the ESM. Labelling of cell lines was initially evaluated as follows: To triplicate suspensions of 10
6 cells in 450 μl of serum-depleted [i.e. not supplemented with foetal bovine serum (FBS)] culture medium in glass tubes were added 0.05 MBq of [
89Zr]oxinate
4 (or neutralised [
89Zr]oxalate) in 50 μl of serum-depleted medium. After incubation for up to 60 min at room temperature the tubes were centrifuged (5 min, 490
g), 450 μl of supernatant was removed and both supernatant and pellet were counted in a Wallac 1282 Compugamma Universal Gamma Counter. Pellet activity was corrected for the residual 50 μl of supernatant. Similar methods were used with higher activities (up to 40 MBq) and cell numbers (up to 5 x 10
7) and with [
111In]oxinate
3 for comparison. After these initial evaluations a standard labelling incubation period of 30 min was adopted for all subsequent biological evaluation.
Efflux of radioactivity from cells
Cells were labelled with [89Zr]oxinate4 or [111In]oxinate3 in glass tubes (106 cells and 0.05 MBq 89Zr or 0.1 MBq 111In per tube), washed three times with PBS, resuspended in culture medium (500 μl) and incubated at 37 °C. Samples taken at intervals up to 24 h were centrifuged and pellets and supernatants counted.
Cell viability
To assess the effect of [89Zr]oxinate4 on viability of eGFP-5T33 cells, samples of 1.2–1.4 × 106 cells (initially 93 % viable based on trypan blue exclusion) were radiolabelled as described above, washed and incubated in 20 ml of RPMI-1640 (supplemented with 10 % FBS, 200 U/l penicillin, 0.1 g/l streptomycin and 2 mM L-glutamate) in T75 tissue culture flasks at 37 °C (5 % CO2). After 24 h, viability of these cells and of controls (treated similarly except for omission of [89Zr]oxinate4) was determined by trypan blue exclusion.
Labelling eGFP-5T33 cells for in vivo cell tracking
Cells were labelled with [111In]oxinate3 or [89Zr]oxinate4 as described above [4 × 106 and 5 × 107 eGFP-5T33 cells/tube; ca. 0.5 MBq per tube for ex vivo organ counting and up to 40 MBq per tube for PET/single photon emission computed tomography (SPECT) imaging], washed three times with PBS and resuspended in 0.2–1 ml of sterile PBS ready for inoculation. Cell viabilities were determined by trypan blue exclusion. To obtain radiolabelled cell lysates, labelled eGFP-5T33 cells were subjected to three cycles of flash-freezing in liquid nitrogen and rapid heating to 90 °C before resuspending in 200–500 μl PBS and repeatedly passing through 27- and 29-gauge needles until visibly homogeneous.
Ex vivo cell tracking of eGFP-5T33 murine multiple myeloma
Animal experiments complied with the Animals (Scientific Procedures) Act (UK 1986) and Home Office (UK) guidelines. Male C57Bl/KaLwRij mice, 6–7 weeks old (Harlan, UK) were acclimatised for >7 days with ad libitum access to water and diet. To assess early tissue distribution ten mice were inoculated via the tail vein with 2 × 106 cells labelled with 0.6–0.8 MBq [89Zr]oxinate4 in 100 μl sterile PBS and culled by cervical dislocation at 9 (n = 3), 24 (n = 4) and 48 h (n = 3) post-injection for ex vivo tissue counting. The longer-term biodistribution of [89Zr]oxinate4- and [111In]oxinate3-labelled cells was compared in three mice injected with 0.9 MBq [89Zr]oxinate4 in 6 × 106 cells and three with 1.7 MBq [111In]oxinate3 in 5 × 106 cells), culled by cervical dislocation after 7 days. Major thoraco-abdominal organs, the left femur and thigh muscle were excised, weighed and gamma-counted.
Ex vivo fluorescence-activated cell sorting (FACS)
Mice were culled by CO2 asphyxiation 2 (n = 3) and 7 days (n = 3) after inoculation with 107 eGFP-5T33 cells labelled with 1.5 MBq [89Zr]oxinate4. Livers, spleens and femora were harvested. Livers and spleens were homogenised. Homogenates were filtered through a 40-μm cell strainer (BD, USA), diluted to 107 cells/ml in ice-cold FACS buffer [1 v/v% FBS, 2 μM ethylenediaminetetraacetate (EDTA) in PBS] and kept on ice. Marrow was flushed from the femora with 5–6 ml of ice-cold FACS buffer, filtered through a 40-μm cell strainer and kept on ice. Samples were analysed on a BD FACSAria III cell sorter (BD, USA) collecting 10,000 each of eGFP-positive and eGFP-negative cells (day 2) and 20,000 each of eGFP-positive and eGFP-negative cells (day 7) for gamma counting.
PET imaging
Preclinical PET/CT images were acquired in a nanoScan® PET/CT (Mediso, Budapest, Hungary) scanner with mice under isoflurane (2 % in oxygen) anaesthesia, starting 60 s before inoculation with 107 eGFP-5T33 cells labelled with 5 MBq [89Zr]oxinate4 in a 200-μl bolus via the right lateral tail vein. Scanning was continued for 2 h and repeated 1, 2, 5, 7 and 14 days after inoculation, acquiring a CT scan after each PET scan. Control mice were inoculated with [89Zr]oxinate4 (4 MBq in 100 μl saline) but no cells, and lysate from 1.1 × 106 cells labelled with 0.9 MBq [89Zr]oxinate4, and scanned between 5 and 50 min post-injection and at 24 h post-injection.
SPECT imaging
SPECT scans were acquired for 4 h immediately after inoculation of 107 eGFP-5T33 cells labelled with 10 MBq [111In]oxinate3 in a 200-μl bolus, using a nanoSPECT/CT (silver upgrade, Mediso, Budapest, Hungary) with four high-resolution multi-pinhole collimators. Further SPECT scans were acquired 1, 2, 3, 5 and 7 days post-injection, each followed by a CT scan. As a control, cell lysate from ca. 1.1 × 106 cells labelled with 1.1 MBq [111In]oxinate3 was injected intravenously followed by SPECT/CT scanning at 60 and 105 min post-injection.
Statistics
Statistical significance was tested using a two-tailed Student’s t test with significance defined as a confidence level of 95 % or above.
Discussion
A general-purpose radiolabel that allows the advantages of PET to be implemented for tracking cells in vivo over several days should satisfy several criteria: the radioisotope must have a suitably long half-life and appropriate positron energy and abundance, convenient and economic availability, a simple and efficient labelling procedure, no major selectivity for different cell types, minimal effects on cell survival and function in vivo and minimal efflux from cells. These criteria should be met at least as well as [
111In]oxinate
3 meets them and better than positron-emitting alternatives reported to date. We have synthesised a lipophilic, metastable complex, [
89Zr]oxinate
4, evaluated it biologically in a range of cell lines in vitro and selected one cell line (eGFP-5T33 myeloma cells) for evaluation in vivo. The results indicate that [
89Zr]oxinate
4 meets these criteria. The half-life of
89Zr is 3.3 days (longer than that of
111In and potential positron-emitting competitors
64Cu,
18F and
68Ga); only
124I (4.2 days) is comparable in this respect but cell labelling with
124I has not been reported. The positron abundance and energies and gamma emissions of
89Zr, although not ideal compared to
18F,
64Cu and
68Ga, proved adequate to give better-resolved images than SPECT with
99mTc in a direct comparison (human sentinel node imaging) [
41].
89Zr is now commercially and economically available as a GMP product. The cell labelling procedure described here is simple, indeed identical to that used with [
111In]oxinate
3; good labelling efficiencies are achieved within a few minutes. The labelling method is effective for several different cell lines and human leukocytes. The survival of labelled eGFP-5T33 cells in vitro is comparable to or better than that of [
111In]oxinate
3-labelled cells and not significantly different to that of control cells treated identically except for omission of radioactivity. While leukocytes are efficiently labelled and show good retention of radioactivity at 24 h, the effects of labelling on survival and function have not been determined, and further investigation of these aspects in the different leukocyte types are required. The in vivo biodistribution and FACS sorting data suggest that labelled eGFP-5T33 cells remain viable for at least 7 days, since the
89Zr-labelled cells continue to express eGFP after this period in vivo. Retention of
89Zr by cells studied here in vitro over 24 h compares well with published cell radiolabels; it is significantly better than that of
111In and markedly superior to that of cells labelled with
64Cu complexes of PTSM [
11,
20‐
22] and GTSM [
23] or with
18F-FDG. The superior retention of
89Zr compared to
111In by eGFP-5T33 cells in vivo is inferred from the much slower transfer of
89Zr than
111In from liver, spleen and bone marrow to kidneys and suggests that replacing
111In SPECT with
89Zr PET can extend the period over which cells can be reliably tracked in vivo; indeed, we have been able to acquire good PET images up to 2 weeks post-inoculation. The 14-day image showed a similar biodistribution of radioactivity to that seen at 7 days in the same mouse suggesting that the radiolabel continued to be retained by 5T33 cells in vivo (Fig.
S4).
The overall biodistribution observed with both
111In- and
89Zr-labelled cells (initial uptake in lung followed by migration to liver, spleen and bone marrow) is consistent with earlier studies of the migration of related murine myeloma cells, by histology and by radiolabelling with
51Cr [
35]. Organ to organ uptake ratios among the main target organs (liver, spleen, bone marrow) did not vary significantly throughout the study, either for
89Zr or
111In, suggesting that labelled cells, or radioactivity, did not migrate from one haematopoietic tissue to another and that the radiolabel used did not change the tissue preference of eGFP-5T33 cells. To test the implicit assumption that migration of radioactivity from liver, spleen and bone marrow to kidney and bladder is an indicator of gradual efflux of radioactivity from living or dead labelled cells, we examined the distribution of activity after injection of [
89Zr]oxinate
4 and that of cells that had been killed/lysed after radiolabelling with
89Zr and
111In. The resulting biodistribution qualitatively resembled that seen with labelled healthy cells, but both
89Zr- and
111In-labelled cell lysate showed greatly increased uptake in kidney and also liver and reduced uptake in spleen, even at early time points (30 min, 24 h, Fig.
3), compared to labelled healthy cells. This contrast in behaviour of
89Zr injected in the form of living labelled cells and dead/lysed labelled cells is consistent with the hypothesis that migration of radioactivity from initially targeted organs to kidney and bladder is a non-invasive marker for efflux of radioactivity from living or dying labelled cells in vivo.
89Zr injected in the form of neutralised [
89Zr]-oxalate also shows marked differences from the behaviour of labelled cells, consistent with the assumption that the images obtained after injection of
89Zr-labelled cells reflect the location of the cells. With [
89Zr]oxinate
4 significant uptake in heart was observed that was not seen with labelled cells, while [
89Zr]oxalate shows gradual skeletal accumulation: radioactivity can be imaged in the joints as early as 15 min post-injection [
42,
43].
The novel experimental approach of ex vivo FACS sorting of eGFP-positive and eGFP-negative populations from organ homogenates, followed by radiation counting of these fractions, showed that the radioactivity in the target tissues remains associated with the originally labelled eGFP-expressing cells (Fig.
4), and hence that these labelled cells remain alive during the 7 days in vivo; 96 % of liver radioactivity, 99 % of spleen radioactivity and >99 % of bone marrow radioactivity remains associated with eGFP-positive cells at 2 days, falling to 95, 92 and 98 %, respectively, by 7 days. Although we observed excellent in vivo survival of labelled eGFP-5T33 cells in this study, a more thorough understanding of the possible cytotoxic effects of the wider use of both
89Zr and
111In labelling of cells is required for a proper interpretation of cell tracking data generally. This ex vivo FACS-based method provides a potential experimental approach to this challenging problem.
These highly promising results warrant further development including refinement of the radiosynthesis and labelling process to improve radiochemical and cell labelling yields, select the most suitable medium for cell labelling and eliminate undesirable use of chloroform, and determine the functional and survival effects of labelling in a variety of cell types, prior to evaluation of the method for cell tracking in humans. While estimates of likely dosimetry in humans have not yet been performed, it will also be important to assess the potential advantages of cell tracking by PET with 89Zr against the likely increased radiation burden per megabecquerel associated with 89Zr, balancing the increased dose per decay, the low positron yield of 89Zr and the increased detection efficiency of PET compared to SPECT.