2-Nitroimidazole-Furanoside Derivatives for Hypoxia Imaging—Investigation of Nucleoside Transporter Interaction, 18F-Labeling and Preclinical PET Imaging

The benefits of PET imaging of tumor hypoxia in patient management has been demonstrated in many examples and with various tracers over the last years. Although, the optimal hypoxia imaging agent has yet to be found, 2-nitroimidazole (azomycin) sugar derivatives—mimicking nucleosides—have proven their potential with [18F]FAZA ([18F]fluoro-azomycin-α-arabinoside) as a prominent representative in clinical use. Still, for all of these tracers, cellular uptake by passive diffusion is postulated with the disadvantage of slow kinetics and low tumor-to-background ratios. We recently evaluated [18F]fluoro-azomycin-β-deoxyriboside (β-[18F]FAZDR), with a structure more similar to nucleosides than [18F]FAZA and possible interaction with nucleoside transporters. For a deeper insight, we comparatively studied the interaction of FAZA, β-FAZA, α-FAZDR and β-FAZDR with nucleoside transporters (SLC29A1/2 and SLC28A1/2/3) in vitro, showing variable interactions of the compounds. The highest interactions being for β-FAZDR (IC50 124 ± 33 µM for SLC28A3), but also for FAZA with the non-nucleosidic α-configuration, the interactions were remarkable (290 ± 44 µM {SLC28A1}; 640 ± 10 µM {SLC28A2}). An improved synthesis was developed for β-FAZA. For a PET study in tumor-bearing mice, α-[18F]FAZDR was synthesized (radiochemical yield: 15.9 ± 9.0% (n = 3), max. 10.3 GBq, molar activity > 50 GBq/µmol) and compared to β-[18F]FAZDR and [18F]FMISO, the hypoxia imaging gold standard. We observed highest tumor-to-muscle ratios (TMR) for β-[18F]FAZDR already at 1 h p.i. (2.52 ± 0.94, n = 4) in comparison to [18F]FMISO (1.37 ± 0.11, n = 5) and α-[18F]FAZDR (1.93 ± 0.39, n = 4), with possible mediation by the involvement of nucleoside transporters. After 3 h p.i., TMR were not significantly different for all 3 tracers (2.5–3.0). Highest clearance from tumor tissue was observed for β-[18F]FAZDR (56.6 ± 6.8%, 2 h p.i.), followed by α-[18F]FAZDR (34.2 ± 7.5%) and [18F]FMISO (11.8 ± 6.5%). In conclusion, both isomers of [18F]FAZDR showed their potential as PET hypoxia tracers. Differences in uptake behavior may be attributed to a potential variable involvement of transport mechanisms.


Introduction
Nowadays, the value of tumor hypoxia imaging for patient stratification, targeted and individualized therapy, and the monitoring thereof is undeniable [1][2][3][4][5][6]. The presence of tumor hypoxia is associated with increased tumor-aggressiveness, invasiveness, and therapy resistance [1,7,8]. While considerable effort has been made in the past decades, the full potential of hypoxia imaging, especially using specific tracers for positron emission tomography (PET) still has to be unraveled, as detection sensitivity, PET tracer target specificity, and the underlying molecular mechanisms by which these can be achieved are not entirely clear. Thus, there is still the need for optimization of hypoxia tracers; the mechanistic questions of the underlying principles of PET hypoxia tracer uptake and image contrast need to be answered as well. Of the numerous hypoxia tracer classes that have evolved, we investigated the class of 2-nitroimidazoles in the past, with a specific focus on 18 F-labeled 1-(5 -deoxy-5 -fluoro-α-D-arabinofuranosyl)-2-nitroimidazole ([ 18 F]FAZA) [9][10][11][12][13]. To date, the uptake mechanism of e.g. [ 18 F]FAZA has not been identified in detail, yet it is assumed to be via passive diffusion, as an alpha-configurated nucleoside derivative should not be transported actively. It has been conceptualized in the past [13][14][15][16] that uptake and retention of 2-nitroimidazole-sugars in hypoxic tumor tissue can be altered by permutation of both, sugar moiety and stereochemistry at the anomeric carbon atom (2-nitroimidazole linked)-with the rationale to take advantage of transport mechanisms involving nucleoside transporters. Involvement of nucleoside transporters in the tissue uptake-process of 2-nitroimidazole-sugars was only investigated recently with β-allofuranose as C 6 -sugar [15]. Transport of nucleosides and nucleoside analog drugs is mediated by two unrelated protein families in humans, the SLC28 family of concentrative nucleoside transporters (hCNTs) and the SLC29 family of equilibrative nucleoside transporters (hENTs) [17]. The SLC28 family has three concentrative (hCNT1/2/3) members and the SLC29 family four equilibrative members (hENT1/2/3/4), respectively. The roles of human nucleoside transporters (hNTs) in transport of nucleosides and nucleoside drugs are summarized in recent reviews [18,19].
Here, we aimed at a systematic investigation of the interaction of selected 2-nitroimidazolefuranoses, resembling nucleoside analogs, with each of the five recombinant hNTs produced in a yeast model system (hENT1/2 and hCNT1/2/3). Transporters that may play a role in uptake of these nitroimidazoles into hypoxic tumor cells or normoxic control tissue like e.g. the muscle should be identified. The four compounds chosen for evaluation of their interaction with nucleoside transporters are found in Table 1, together with references to published data and the work list for the actual study. Besides the transporter interaction investigation, for β-FAZA, a novel synthesis route should be developed and for 1 -α- [2 ,5 - [20,21] and β-[ 18 F]FAZDR, of which positive PET imaging results have been recently obtained [13]. The mouse colon carcinoma model could be chosen, as mouse transporters exhibit close similarity to human nucleoside transporters [22]. With this comparative study, we sought to gain further insight into the importance and influence of the sugar moiety and configuration of 2-nitroimidazole at the anomeric carbon atom on tracer uptake and image contrast.

Compound
Transporter interaction 18

Organic Chemistry
FAZA [23,24], α-FAZDR and β-FAZDR [13] were prepared by known procedures; for β-FAZA, a novel synthesis route was developed, although β-FAZA is a known compound [14]. Figure 1 gives the structures of the non-radioactive fluorinated 2-nitroimidazole sugars that were synthesized and evaluated for their interaction with human nucleoside transporters (see Section 2.3). An improved synthesis of the β-arabinose-derived nucleoside analog β-FAZA from the known starting material β-1 (prepared as described by Kumar et al. [25]) was developed (Scheme 1). Selective silylation of the primary hydroxyl group at C-5´ with TBDMSCl/imidazole in pyridine, followed by acetylation, furnished crystalline and fully protected nucleoside β-2 (80%). Bis(2-methoxyethyl)aminosulfur trifluoride [26] (Deoxo-Fluor ® ) converted it to 5´-fluoro nucleoside β-3 in 80% yield. Deprotection with MeONa in MeOH gave the desired arabinose-derived nucleoside β-FAZA in 75%. β-FAZA was recently prepared by a different approach [14]. Here, we were able to improve the overall process compared to [14] by a reduction of synthetic steps (for initial reaction steps from the commercially available 1-β-D-(ribofuranosyl)-2-nitroimidazole, see [14] and [25]), overall reaction times, and an increase in yield to 48% for the three steps shown in Scheme 1, compared to 21% for the last three steps in [14]. The improvement was obviously due to the utilization of the commercial fluorination agent, Deoxo-Fluor®, in the penultimate step. An improved synthesis of the β-arabinose-derived nucleoside analog β-FAZA from the known starting material β-1 (prepared as described by Kumar et al. [25]) was developed (Scheme 1). Selective silylation of the primary hydroxyl group at C-5 with TBDMSCl/imidazole in pyridine, followed by acetylation, furnished crystalline and fully protected nucleoside β-2 (80%). Bis(2-methoxyethyl)aminosulfur trifluoride [26] (Deoxo-Fluor ® ) converted it to 5 -fluoro nucleoside β-3 in 80% yield. Deprotection with MeONa in MeOH gave the desired arabinose-derived nucleoside β-FAZA in 75%. β-FAZA was recently prepared by a different approach [14]. Here, we were able to improve the overall process compared to [14] by a reduction of synthetic steps (for initial reaction steps from the commercially available 1-β-D-(ribofuranosyl)-2-nitroimidazole, see [14,25]), overall reaction times, and an increase in yield to 48% for the three steps shown in Scheme 1, compared to 21% for the last three steps in [14]. The improvement was obviously due to the utilization of the commercial fluorination agent, Deoxo-Fluor ® , in the penultimate step.
Radiolabeling, hydrolysis, and purification were carried out similar to the synthesis of   [13]). Maximum yield obtained was 10.3 GBq at the end of the synthesis. As these amounts were sufficient for the planned mouse experiments and quality of the product was also satisfactory, the process and reaction parameters were not optimized further.
Radiolabeling, hydrolysis, and purification were carried out similar to the synthesis of β-[ 18 F]FAZDR (Scheme 2). Radiochemistry was performed on the same automated synthesis module (TRACERlab FXF-N, GE, non-cassette based) as for the other two PET tracers. Radiochemical yields were 15.9 ± 9.0% (n = 3) with higher variations compared to β-[ 18 F]FAZDR (10.9 ± 2.4%, n = 4, uncorrected for decay [13]). Maximum yield obtained was 10.3 GBq at the end of the synthesis. As these amounts were sufficient for the planned mouse experiments and quality of the product was also satisfactory, the process and reaction parameters were not optimized further.
Radiochemical (RCP) and chemical purity (CP) were determined using HPLC and TLC. The latter method showed RCP of α-[ 18 F]FAZDR of > 97%; in HPLC radiochromatograms, only the product peak was visible, corresponding to > 98% of RCP. With UV detection at 220 nm, only matrix peaks were visible (< 5 min retention time), with no other chemical impurities above the detection limit. The product solutions were always clear, colorless, and free of visible particles with a pH of 6.7 and a molar activity > 50 GBq/µmol.

Transporter Studies in Saccharomyces Cerevisiae
The interaction of 2-nitroimidazole sugars with nucleoside transporters was determined by ability of the 2-nitroimidazole sugars to inhibit [ 3 H]uridine uptake in yeast cells producing recombinant human nucleoside transporters and results expressed as IC50 values. Among the four Radiochemical (RCP) and chemical purity (CP) were determined using HPLC and TLC. The latter method showed RCP of α-[ 18 F]FAZDR of > 97%; in HPLC radiochromatograms, only the product peak was visible, corresponding to > 98% of RCP. With UV detection at 220 nm, only matrix peaks were visible (< 5 min retention time), with no other chemical impurities above the detection limit. The product solutions were always clear, colorless, and free of visible particles with a pH of 6.7 and a molar activity > 50 GBq/µmol.

Transporter Studies in Saccharomyces cerevisiae
The interaction of 2-nitroimidazole sugars with nucleoside transporters was determined by ability of the 2-nitroimidazole sugars to inhibit [ 3 H]uridine uptake in yeast cells producing recombinant human nucleoside transporters and results expressed as IC 50 values. Among the four compounds tested, only the β-sugar compounds inhibited hENT1 and hENT2 (SLC29A1/2), but at high concentrations. For FAZA and α-FAZDR, no interaction was detected. β-FAZDR inhibited hCNT1 (SLC28A1) and hCNT3 (SLC28A3) potently with IC 50 values of 269 ± 4 µM and 124 ± 33 µM, respectively. FAZA inhibited hCNT1 with an IC 50 value of 290 ± 44 µM and FAZA was the only compound to show considerable inhibitory activity against hCNT2 (SLC28A2) with an IC 50 value of 640 ± 10 µM. Individual IC 50 values for all compounds and investigated transporters are summarized in Table 2.
In summary, FAZA inhibited [ 3 H]uridine transport by hCNT1 and hCNT2, in contrast to the commonly accepted theory of FAZA entering cells by passive diffusion. Although some of the compounds inhibited [ 3 H]uridine uptake by nucleoside transporters, transporter interaction does not necessarily imply transport through the cell membrane. To confirm this uptake, experiments with the corresponding radioactive compounds needed to be undertaken, which was not possible here, due to their lack of availability. Keeping this in mind, it is nevertheless possible that our results challenge the current understanding of FAZA image contrast, as tumor-type dependent transporter expression might especially alter the early-phase uptake in tumor tissue and contribute to PET signal heterogeneity. This should be clarified in additional studies in the future. In contrast to FAZA, β-FAZA with the nucleosidic β-configuration showed weak interaction with the nucleoside transporters. Among the tested compounds, β-FAZDR showed best inhibition of uridine uptake by hCNT1 and hCNT3, while α-FAZDR showed no interaction with any transporter. Thus, both the deoxyribose sugar moiety and the β-configuration of the 2-nitroimidazole at the anomeric carbon atom are important molecular properties, if nucleoside transporter mediated uptake is envisaged in order to obtain higher TMR at early imaging time-points.  Figure 2A). Here, we were especially interested in the direct comparison of α-[ 18 Figure 2D). While the washout from both, tumor and muscle tissue of [ 18 F]FMISO was limited, we observed an increased washout for α-[ 18 F]FAZDR from both, tumor and muscle tissue ( Figure 2D).
Conclusively, we observed the highest TMR for β-[ 18

Discussion
While a lot of effort has been put in the development of hypoxia specific PET tracers (based on 2-nitroimidazole as hypoxia-selective moiety), and a wealth of studies indicate both hypoxia specificity and added value for patient stratification and treatment monitoring [1][2][3][4][5][6]; the exact image contrast generating mechanisms are still poorly understood. While the commonly accepted underlying theory assumes trapping of 2-nitroimidazole-based PET tracers in hypoxic tissue and washout from normoxic control tissue [12,23,27], a plethora of studies (both clinical and preclinical) have demonstrated that this is most probably not the only mechanism explaining the obtained image contrast for various 2-nitroimidazole-based PET tracers in diverse tumor-entities [9,13,[28][29][30][31]. Especially, the fact that PET tracer kinetics of 2-nitroimidazole-based compounds can indicate both reversible or irreversible uptake in patients with the selfsame tumor type, e.g. head and neck cancer, challenges the concept of hypoxia-selective trapping [31]. This ambiguous behavior was also observed by others in both mouse and man [28,32,33]. Substantial tumor and muscle washout could recently be observed by us for both [ 18 F]FAZA [9] and β-[ 18 F]FAZDR [13]. While rapid washout from normoxic control tissue (in this case muscle tissue) was anticipated, we were again struck by the high washout rates from hypoxic tumor tissue, especially for β-[ 18 [13]. Thus, one could conclude from this body of data that hypoxia tracer image contrast consists of three different components, which are theoretically influenced by the contribution of nucleoside transporters (detailed analysis of hypoxia tracer kinetic modeling complexity can be found in [30] However, theoretically, if the sum of free and non-specifically bound tracer in target tissue is higher than the actual specific binding in target tissue, the observed clearance could be dominated by the washout of the non-specific or free tracer, while the specific fraction of the PET signal does not necessarily need to be reversible to generate the observed contrast; this could also be caused by a smaller fraction of irreversible binding. In addition, nucleoside transporters actively contribute to early and, thus also to late PET image contrast. While we cannot draw conclusions on the dominance of free, non-specific or specific fractions, we can conclude that nucleoside transporters definitely influence image contrast, especially early image contrast, here at 1 h p.i. Furthermore, the hypoxia specificity of α-[ 18 F]FAZDR could be indirectly proven in this manuscript, as α-[ 18 F]FAZDR-uptake perfectly resembled β-[ 18 F]FAZDR-uptake in the same tumor, imaged on consecutive days (Figure 2A).
We additionally validated this finding by an exemplary scan of one CT26 colon carcinoma bearing mouse with [ 18 F]FAZA, α-[ 18 F]FAZDR and β-[ 18 F]FAZDR on three consecutive days (Figure 4), displaying very similar uptake patterns. This can be concluded as the hypoxia selectivity of β-[ 18 F]FAZDR, as recently shown by us [13], while [ 18 F]FAZA hypoxia selectivity was shown by us and others in a plethora of publications [1,5,7,9,10,12,13,28,29,31]. In perfect concordance with the transporter data from this study, TMR at 1 h p.i. were highest for β-[ 18 F]FAZDR, probably due to the involvement of nucleoside transporters.
Conclusively, the exact mechanism generating 2-nitroimidazole PET hypoxia tracer image contrast is still elusive, while we add knowledge on stereochemistry-dependent involvement of nucleoside transporters regarding both early and late PET signals of 2-nitroimidazole-sugars. For the first time, we could show an interaction of FAZA with nucleoside transporters, thus cellular FAZA-uptake may not solely be attributable to passive diffusion-although our results still need to be taken with caution, as Pharmaceuticals 2019, 12 9 4), displaying very similar uptake patterns. This can be concluded as the hypoxia selectivity of β-[ 18 F]FAZDR, as recently shown by us [13], while [ 18 F]FAZA hypoxia selectivity was shown by us and others in a plethora of publications [1,5,7,9,10,12,13,28,29,31]. In perfect concordance with the transporter data from this study, TMR at 1 h p.i. were highest for β-[ 18 F]FAZDR, probably due to the involvement of nucleoside transporters. Conclusively, the exact mechanism generating 2-nitroimidazole PET hypoxia tracer image contrast is still elusive, while we add knowledge on stereochemistry-dependent involvement of nucleoside transporters regarding both early and late PET signals of 2-nitroimidazole-sugars. For the first time, we could show an interaction of FAZA with nucleoside transporters, thus cellular FAZA-uptake may not solely be attributable to passive diffusion-although our results still need to be taken with caution, as interaction might not necessarily be related to active transport. These findings open up novel avenues for hypoxia imaging and might help clarifying the underlying principles of hypoxia imaging.

General
Acetonitrile for azeotropic drying before 18 F-radiolabeling was from Merck (DNA synthesis grade, Darmstadt, Germany). Dimethylsulfoxide (DMSO, dried over molecular sieves) as solvent for labeling was used from Fluka (Germany). Kryptofix 2.2.2. was purchased from Merck.