Synthesis and tumour cell uptake studies of gadolinium(III)–phosphonium complexes

The synthesis of a new series of Gd(III)-arylphosphonium complexes is described and the solution stability of selected compounds is reported. Their lipophilicity and uptake in human glial (SVG p12) and human glioblastoma multiforme (T98G) cell lines are presented. The in vitro cytotoxicity of all complexes was determined to be low at therapeutically-relevant concentrations. Selected Gd(III) complexes are potential candidates for further investigation as theranostic agents.

MRI relaxivity of representative complex 1. Relaxivity studies were performed on the representative parent complex 1 and calculations showed that r 1 = 7.0 ± 1.1 mM −1 s −1 (9.4 T, 298 K) which is somewhat higher than the value reported for the Gd-DO3A analogue 11 (5.1 mM −1 s −1 ) 8 . Furthermore, the relaxivity value of 1 lies at the higher end of clinical Gd MRI contrast agents such as Magnevist, probably related to the higher MW of 1 compared to that of clinical contrast agents. The complete relaxivity data are presented in Table S6 and Figure S1 (Supplementary Information).

Determination of Gd(III) complex lipophilicity.
Lipophilicity is a vital parameter to consider regarding DLCs as it underpins their membrane permeability, and also strongly influences absorption, distribution, metabolism and excretion of the Gd(III)-phophonium complex within the body 33,34 . Hence, it is important to  Table 1. Competitive stability of complexes 1 and 11. a Slope of the line of best fit resulting from a plot of the average data using 10 different concentrations (2.0-250 μM). b Regression analysis of the line of best fit. c Percentage of Gd 3+ retained by the phosphonium complex over a period of 24 h, as determined by the shift in absorbance from 433 to 573 nm, averaged across all data points. www.nature.com/scientificreports/ consider and analyse the effect that the lipophilicity of these complexes has on their efficacy to assist in future Gd theranostic design. LogP (where P is the partition coefficient) is the most common measurement of lipophilicity and it was determined for 1-10 by means of a reverse-phase HPLC method involving a suite of standard reference compounds. A summary of the results is shown in Table 2.
All of the Gd(III) complexes prepared in this work were shown to be reasonably lipophilic, with logP values ranging from 1.12 to 1.63. These values are consistent with those previously reported for a range of related DO3A-phosphonium complexes and lie within the pharmacologically optimal value of 1-3, as outlined by Waring et al. 7,8,33 . Furthermore, these logP values lie within the predictive model on the optimised uptake of lipophilic cations into the mitochondria reported by Trapp and Horobin 35 . This model indicates the optimal lipophilicity for DLCs lie above logP = − 2, precluding any kinetically-limited uptake due to the high energy of desolvation, and also fall below logP = 2 in order to avoid non-specific lipid binding 35 . In vitro cytotoxicity studies. In vitro cytotoxicity was determined for each of the new complexes by means of standard 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) colourimetric assays over a 62.5 μM-2 mM range. A summary of the results in Table 3 reports the half-maximal inhibitory concentration (IC 50 ) of each of the assessed complexes for the human glial (SVG p12) and human glioblastoma multiforme (T98G) cell lines. Ideally, Gd theranostic agents should demonstrate relatively low cytotoxicity to ensure any uptake into healthy tissue causes only minimal damage, as the intended cytotoxic event for Gd-based binary agents results from either the thermal neutron (NCT/NCEPT) or X-ray photon (PAT) irradiation.
The IC 50 values for complexes 1-10 were found to lie in the low mM range, in agreement with those previously-reported for related DO3A complexes [7][8][9] . A one-tailed Mann-Whitney U test demonstrated that these complexes were slightly more cytotoxic toward the T98G cells over their healthy SVG p12 cell line, albeit the overall toxicity of the complexes was only marginal (mM).
In order to evaluate the relative cytotoxicities of complexes 1-10 with a known (albeit no longer clinicallyused) Gd theranostic, Motexafin-Gd (MGd), was chosen. In vitro cell experiments involving MGd were reported to result in 50% cell arrest of the HF1 tumour cell line at concentrations of less than 100 μM, and the Gd agent www.nature.com/scientificreports/ exhibited significant cytotoxicity at concentrations of ca. 50 μM (in the absence of any X-ray irradiation) 36,37 . Evidently, complexes 1-10 appear to exhibit far lower in vitro cytotoxicities than the clinical agent MGd.
In vitro cellular uptake of Gd. As the therapeutic application of Gd(III) complexes relies upon maximising Gd accumulation within the mitochondria of the target tumour cells, comprehensive in vitro cellular uptake studies were performed on complexes 1-10 using both SVG p12 and T98G cell lines at four different concentrations (complexes 1-5: 125 μM, 250 μM, 500 μM, 1 mM; complexes 6-10: 62.5 μM, 125 μM, 250 μM, 500 μM). The harvested cells were analysed for accumulated Gd by means of ICP-MS and these values were normalised to protein content, as determined by means of a modified Lowry protein assay. Representative cell uptake results are presented in Figs. 3, 4, 5 (and Supplementary Tables S1-S5) and are expressed as ng Gd/mg protein. Figures 3, 4, 5 demonstrate that uptake of complexes 1-10 occurs in both tumour and healthy cell lines. Table 4 highlights the tumour-to-normal cell uptake ratios. Statistical analysis of the data was performed by means of a two-way ANOVA, demonstrating significantly greater uptake of Gd in the T98G glioma cell line compared with the SVG p12 glial cell line for complexes 2 and 6-9 but no statistically significant difference was found between the two cell lines for complexes 1, 3-5 and 10. Indeed, for complexes 1-3, the tumour cell uptake and selectivity results are not as significant as our previous research reporting related DO3A complexes whereby a reduced delocalisation at the phosphonium centre in moving from, for example, R 1 = Ph (1) to Me (2) was strongly correlated with a reduced tumour-cell selectivity and significantly increased Gd uptake 9 . Furthermore, statistical analysis demonstrated that at the lowest treatment dose for each Gd(III) complex (with the exception of complex 3), a significantly higher Gd uptake in the T98G tumour cell line over the SVG p12 cell line was found. This was also true at the second-lowest treatment dose for each complex, with the exception of   www.nature.com/scientificreports/ the triethylene glycol (TEG) complex 10 which was found to exhibit high tumour cell selectivity (2.8) but only at the lowest treatment dose (62.5 μM). However, the propyl-bridged complex 6 exhibited the highest tumour cell selectivity (3.7) of all complexes studied in this work (at 125 μM). None of the complexes examined in this work demonstrated any significant selectivity for the T89G cell line at the two highest treatment doses (500 μM or 1 mM). The tendency for lower concentration treatment doses resulting in higher Gd uptake into the tumour cells over the healthy cells is consistent with the tumour cell's elevated ΔΨ m which is responsible for tumourselective accumulation up to a point whereby high levels of Gd complex accumulation within mitochondria leads to depolarisation, after which passive Gd influx may become the dominant mechanism of mitochondrial  www.nature.com/scientificreports/ accumulation 2,38,39 . This hypothesis is further supported by the tendency that the highest concentration treatment resulted in no significant difference in Gd uptake between the two different cell lines. The electron-rich P-aryl rings in complexes 4 and 5 led to a somewhat increased cell uptake compared to the parent 1 at lower treatment doses, but no significant improvement in tumour-cell selectivity was observed. Furthermore, replacement of the archetypal xylyl linker in 1 with various alkyl and TEG linkers (6-10) resulted in a significant increase in Gd uptake and tumour cell selectivity at lower treatment doses. In general, no statistically significant trend was established for complexes 6-10, and Gd uptake did not appear to be related to the relative distance between the phosphonium and Gd centres, although a very high Gd uptake was observed for 7 (5349.0 ± 478.1 ng Gd/mg protein), containing the n-butyl linker, at one of the assessed doses (250 μM). The incorporation of a hydrophilic TEG linker in complex 10, however, did lead to an increased Gd uptake at higher concentrations (250 and 500 μM) when compared to the more lipophilic alkyl linkers found in 6-9. However, the higher Gd uptake in 10 is countered by the reduced tumour cell selectivity when compared to most other Gd(III) complexes assessed in this study, except at the lowest treatment dose (62.5 μM).

Conclusions
Complexes 1-10 were synthesised in good to high yields and in high purity. Stability assays comparing the DOTA-based complex 1 to the previously-reported, archetypal DO3A-phosphonium complex 11 at mildly acidic pH conditions demonstrated a significant increase in the stability of 1, and these findings are consistent with those previously reported for Gd(III)-DOTA complexes by Caravan et al. 26 . Complexes 1-10 were all found to have low (mM) cytotoxicity at therapeutically-relevant concentrations, with significant in vitro tumour cell uptake and reasonable tumour cell selectivity at lower concentrations. The reduced tumour cell selectivity of the complexes at higher concentrations is consistent with previous findings which proposes a different mechanism for uptake beyond a certain threshold, although it may also be indicative of ΔΨ m depolarisation 9 .
No significant trends were observed in cell uptake when various phosphonium targeting vectors were used (1-5), in direct contrast to previous studies involving Gd(III)-DO3A-phosphonium complexes where, for example, the DO3A analogue of 2 was found to have very high T98G cell uptake compared to the parent triphenylphosphonium complex but with a much lower tumour cell selectivity [7][8][9] . Similarly, a related trend was also observed for the alkyl and TEG linker complexes 6-10 which were found to have reduced tumour cell selectivities but showed significantly higher Gd uptake than the para-xylyl linker complexes 1-5 in both cell lines but even more so in the T98G tumour line. Indeed, the para-xylyl DOTA complexes 1-5 were not found to differ significantly from their previously-reported DO3A counterparts in terms of cytotoxicity and tumour cell uptake 7-9 , although, in general, tumour cell selectivity was found to be reduced for the DOTA family of complexes, a factor that might be related to a decreased lipophilicity and diminished ability to traverse both the cell and mitochondrial membranes owing to the presence of the amide linkage. However, the logP values in Table 2 indicate that the lipophilicities of the DO3A and DOTA complexes are very similar and, indeed, the logP difference between the archetypal complexes 1 and 11 is only marginal at best. In vivo uptake and biodistribution studies with selected Gd(III) complexes are currently underway and the results of this work will be reported in due course.

Methods
All precursor chemicals were commercially available. For experiments requiring H 2 O, ultrapure H 2 O was collected from a Milli-Q water purification system. Anhydrous MeCN, DMF, and PhMe were obtained using a Puresolv system. Et 2 O was dried over sodium wire, distilled and stored under N 2 over microsieves. MeOH and EtOH were stored under N 2 over 4 Å sieves. All other solvents were used without further purification. Reactions requiring an inert atmosphere were performed under dry N 2 and employed conventional Schlenk techniques 40 . Characterisation. All 1 H, 13 C{ 1 H}, and 31 P{ 1 H} NMR spectra were recorded at 300 K on a Bruker Avance 300 (5 mm QNP probe, 1 H at 300 MHz, 13 C at 75 MHz, and 31 P at 121 MHz), a Bruker Avance III 400 (5 mm BBFO probe, 1 H at 400 MHz, 13 C at 100 MHz, and 31 P at 162 MHz), or a Bruker Avance III 500 (5 mm BBFO probe, 1 H at 500 MHz, 13 C at 125 MHz, and 31 P at 202 MHz). All NMR signals (δ) are reported in ppm. 1 H and 13 C NMR spectra were referenced according to their solvent residual peaks. 13 C{ 1 H} NMR spectra in D 2 O were referenced according to an internal standard of DMSO-d 6 (39.39 ppm). 31 P{ 1 H} NMR spectra were referenced to external 1% TMS in CDCl 3 using the unified reference scale 41 . Coupling constants are reported in Hz. Peak multiplicities have been abbreviated as s (singlet), d (doublet), t (triplet), q (quartet), qi (quintet), br (broad), and m (multiplet -unassignable multiplicity). Proton and carbon positions within aryl rings have been abbreviated i-(1, 1), o-(1, 2), m-(1, 3), and p- (1,4). The xylyl linker has been abbreviated to 'xy' for clarity's sake when assigning proton and carbon nuclei belonging to this linker group.
FT-IR spectra were run on a benchtop Bruker ALPHA FTIR spectrometer. All 96 and 384 well plates were read using a BMG LABTECH POLARstar Omega UV spectrometer. Low resolution ESI-MS were recorded on a Bruker amaZon SL mass spectrometer. MALDI-MS were recorded on a Bruker autoflex speed LRF MALDI-TOF mass spectrometer. High resolution ESI-FT-ICR-MS data were recorded on a Bruker Apex Qe 7 T FTICR mass spectrometer. Quantitative MS was recorded on a Perkin Elmer NexION 350X ICP-MS with a quadrupole analyser.
General phosphonium salt procedure (GP1). Phosphonium salts of interest were prepared by means of dropwise addition of the precursor arylphosphines in toluene to a stirred solution of the relevant dibromo-p-xylyl, alkyl or TEG linkers in toluene, followed by heating at reflux (unless specified). The crystallized product was filtered off and washed with toluene and diethyl ether to yield the phosphonium salt in high purity.
General DOTA linking procedure (GP2). The bromine functional groups of the phosphonium salts prepared in GP1 were replaced with azide groups by stirring with NaN 3 in DMF at RT for 72 h. The DMF was removed under a flow of N 2 , and the residue was then dissolved in CHCl 3 and any excess NaN 3 filtered off. The filtrate was reduced in vacuo to yield the phosphonium azide in high purity (> 95%). The azide group was reduced to an amine using 10 mol% Pd/C stirred in MeOH under a H 2 atmosphere at RT for 6 h. The catalyst was removed by means of filtration, and the solvent volume reduced to afford the phosphonium-amine in high purity. The phosphonium targeting vectors were coupled to the DOTA chelating component by means of a peptide coupling reaction performed in peptide-grade DMF with the HATU coupling reagent in the presence of excess NMM and stirred at RT for 24 h. The DMF and NMM were removed under N 2 , and the residue was deprotected by the addition of neat TFA upon stirring the mixture for a further 12 h at RT. The volatiles were removed in vacuo, and the crude residue dissolved in H 2 O and washed with CHCl 3 . The aqueous fraction was reduced in vacuo to give a hygroscopic solid which was purified by means of reverse-phase HPLC, and the product lyophilised to afford the pure ligand. www.nature.com/scientificreports/ 2,2′,2″-(10-(2-((4-((Methyldiphenylphosphonio)methyl)benzyl)amino)-2-oxoethyl)-1,4,7,10-tetraazacyclododecane-1,4,7-triyl)triacetatogadolinium(III) trifluoroacetate (2): GP1 using methyldiphenylphosphine (1.00 mL, 1.08 g, 5.37 mmol) and α,α'-dibromo-p-xylene (1.60 g, 6.06 mmol) each in 20 mL of toluene at reflux for 2 h to yield 2.0 g (81.4%) of the phosphonium salt. Following GP2, the phosphonium salt (384 mg, 0.83 mmol) was added to excess NaN 3 (177 mg, 2.72 mmol) in DMF (15 mL 13  www.nature.com/scientificreports/ 25 mM stock solutions of each GdCl 3 , 1 and 11 were prepared in the various buffer solutions. Two serial dilutions of each solution were freshly prepared, the first ranging from 25 mM to 19.5 μM, and the second ranging from 20 mM to 31.2 μM. To the wells of a 384 well plate were added 80 μL of the desired pH buffer and 10 μL of the selected serial dilution (in triplicate). The plates were agitated in a plate reader and allowed to equilibrate for 24 h. To each of the wells was added 10 μL of the 10 × concentrated xylenol orange stock solution in the absence of light. The plates were again agitated in a plate reader and allowed to equilibrate for 15 min at RT. The absorbance values were then determined at both 573 nm and 433 nm by means of a BMG LABTECH POLARstar Omega, UV Spectrometer, and the data plotted and processed using Graphpad Prism 7.
Cell culture studies. The human glioblastoma multiforme (T98G) and human glial (SVG p12) cell lines were purchased from the ATCC and were maintained as monolayers in a minimum essential medium supplemented with foetal bovine serum (10% v/v), penicillin (100 units mL −1 ), streptomycin (100 μg mL −1 ) and L-glutamine (2.5 mM), within incubators at 37 °C in a humidified 5% CO 2 atmosphere. Centrifugation was performed at 3000 rpm for 5 min. Earls MEM, FCS, PBS, Trypsin, L-glutamine, and the antibiotic solutions were purchased from ThermoFisher scientific.
Cytotoxicity assays. The in vitro cytotoxicities of all Gd(III) complexes were assessed in both T98G and SVG p12 cell lines using the colourimetric 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay 47 . Briefly, cells were harvested with trypsin (0.1% v/v), the trypsin diluted with complete medium, and cell pellets isolated via centrifugation. Any remaining traces of trypsin were removed from the pellet via a suspension/centrifugation step. The pellets were then completely re-suspended, the number of cells mL −1 was determined using a hemocytometer (Weber), and the solution diluted such that 90 μL aliquots of cell suspension per well of a 96-well plate yielded 1 × 10 4 cells per well. The cells were incubated for 24 h to allow them to adhere.
The cells were subsequently dosed with 10 μL per well of a sterile 10 × stock of the serial dilution (2 mM-62.5 μM) of Gd(III) complex (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11) or the relevant vehicle (control). Each dilution/vehicle pair was repeated in triplicate within a plate and each plate was repeated in triplicate, for both cell lines. After 72 h incubation, MTT solution in phosphate-buffered saline (PBS; 30 μL, 0.17% w/v) was added to all wells in the absence of light and the incubation was continued for a further 4 h. The culture medium and excess MTT solution were carefully removed so as not to disrupt any MTT-formazan crystals formed during the incubation and were subsequently dissolved in 150 μL of DMSO.
Cell viability was determined by measuring the absorbance at 600 nm using a BMG LABTECH POLARstar Omega UV spectrometer. All readings were corrected for absorbance from the background control wells, and the level of MTT was expressed relative to the corresponding vehicle-treated controls (as % viability). Corresponding IC 50 values for each of the compounds assessed were then determined as the dose required to give a 50% decrease in cell viability. IC 50 values were reported with standard errors.
Cell uptake studies. 25 cm 3 cell culture flasks were seeded with 3 mL of a cell suspension to yield 2 × 10 5 cells per flask and were incubated for 72 h until ~ 80% confluent. Sterile 20 mM stock solutions of the Gd(III) complexes 1-10 were diluted with warm (37 °C) culture medium to yield a series dilution of final concentrations including 1000 μM, 500 μM, 250 μM, 125 μM and 62.5 μM. The non-dosed culture medium in each flask was removed and replaced with Gd-dosed media or a relevant vehicle (control), each concentration was repeated in triplicate for both cell lines along with the relevant control. The cells were incubated for a further 48 h after which they were harvested.
The culture medium was removed, and the monolayers washed with warm (37 °C) PBS (1 mL) prior to treatment with trypsin (0.1% v/v). The trypsin was diluted with medium, and cell pellets isolated by means of centrifugation before the supernatant was removed and the pellets re-suspended in warm PBS. Further centrifugation produced washed pellets, eliminating the possibility of residual Gd(III) complex remaining outside of the cells, which were subsequently re-suspended in warm PBS (1 mL). From this suspension, 100 μL was set aside for protein analysis and the remaining 900 μL was centrifuged to isolate the cell pellets. The supernatant was removed, and the cells were digested in concentrated HNO 3 (110 μL, 69%) at 40 °C in a heating block for 24 h. 100 μL of the digest solution was diluted to 10 mL to result in a 1% HNO 3 solution which was measured for Gd content by means of ICP-MS. ICP-MS was run on a Perkin Elmer ELAN 6100 Inductively Coupled Plasma Emission Mass Spectrometer (ICP-MS). Gd uptake is reported as μg Gd/mg protein ± standard error.

Protein analysis.
A Bio-Rad DC protein assay kit was used to determine protein concentrations 48 . A bovine serum albumin (BSA) protein standard curve was prepared each time the assay was performed, ranging from 1.0 to 0.1 mg mL −1 in PBS. Lysis of the 100 μL cell suspension was achieved using three snap freeze-heat cycles and thorough pipette mixing. The protein content of each solution was then analysed by pipetting 5 μL samples of the blanks, vehicle (control), and treated cell solutions into a 96-well plate, adding 25 μL of the alkaline copper tartrate solution and 200 μL of the Folin reagent solution included in the Bio-Rad kit in the absence of light. Each sample was prepared in triplicate. The plates were incubated at 37 °C for 15 min and the absorbance measured at 750 nm using a BMG LABTECH POLARstar Omega UV spectrometer. The protein concentration was determined by correcting all measurements with the background controls and comparing the absorbance with that of the BSA standard curve.