Head-to-head comparison of DFO* and DFO chelators: selection of the best candidate for clinical 89Zr-immuno-PET

Purpose Almost all radiolabellings of antibodies with 89Zr currently employ the hexadentate chelator desferrioxamine (DFO). However, DFO can lead to unwanted uptake of 89Zr in bones due to instability of the resulting metal complex. DFO*-NCS and the squaramide ester of DFO, DFOSq, are novel analogues that gave more stable 89Zr complexes than DFO in pilot experiments. Here, we directly compare these linker-chelator systems to identify optimal immuno-PET reagents. Methods Cetuximab, trastuzumab and B12 (non-binding control antibody) were labelled with 89Zr via DFO*-NCS, DFOSq, DFO-NCS or DFO*Sq. Stability in vitro was compared at 37 °C in serum (7 days), in formulation solution (24 h ± chelator challenges) and in vivo with N87 and A431 tumour-bearing mice. Finally, to demonstrate the practical benefit of more stable complexation for the accurate detection of bone metastases, [89Zr]Zr-DFO*-NCS and [89Zr]Zr-DFO-NCS-labelled trastuzumab and B12 were evaluated in a bone metastasis mouse model where BT-474 breast cancer cells were injected intratibially. Results [89Zr]Zr-DFO*-NCS-trastuzumab and [89Zr]Zr-DFO*Sq-trastuzumab showed excellent stability in vitro, superior to their [89Zr]Zr-DFO counterparts under all conditions. While tumour uptake was similar for all conjugates, bone uptake was lower for DFO* conjugates. Lower bone uptake for DFO* conjugates was confirmed using a second xenograft model: A431 combined with cetuximab. Finally, in the intratibial BT-474 bone metastasis model, the DFO* conjugates provided superior detection of tumour-specific signal over the DFO conjugates. Conclusion DFO*-mAb conjugates provide lower bone uptake than their DFO analogues; thus, DFO* is a superior candidate for preclinical and clinical 89Zr-immuno-PET. Electronic supplementary material The online version of this article (10.1007/s00259-020-05002-7) contains supplementary material, which is available to authorized users.


Introduction
Positron emission tomography with 89 Zr-labelled antibodies ( 89 Zr-immuno-PET) is a valuable tool to characterise the in vivo behaviour of monoclonal antibodies (mAbs) as well as other drugs with slow clearance from the blood, such as other types of proteins, mAb conjugates, nanoparticles and cells. In this way, immuno-PET can be used to (i) assess target expression; (ii) evaluate the in vivo behaviour of the drug in relation to efficacy and toxicity; (iii) optimise dose, route and schedule of administration; (iv) optimise drug design and (v) select patients with the highest chance of benefiting from drug treatment [1][2][3]. Over the last decade, the number of clinical studies using 89 Zr-labelled mAbs has increased enormously, confirming the importance of immuno-PET imaging, especially in oncology [4]. With a half-life of 78.4 h, 89 Zr is well suited for studying the kinetics of molecules with relatively long plasma half-lives [2,5]. Currently, the majority of 89 Zrimmuno-PET studies employ desferrioxamine B (DFO). DFO has been administered to thousands of patients, in free form for the treatment of iron overload or attached to slow kinetic drugs as a chelator for 89 Zr labelling. However, preclinical studies revealed that the hexadentate 89 Zr-DFO complex is prone to dissociation in vivo with free 89 Zr 4+ predominantly accumulating in bones and joints because of its high affinity for strong electronegative donor atoms such as oxygen and phosphorus in hydroxyapatite present in components of the bones [6][7][8][9][10][11][12]. Although the deposition of 89 Zr in bone has not been routinely observed in the published clinical studies to date, a systematic review is required to further evaluate this question. A resolution to this problem is particularly important since non-specific bone uptake will provide an increased radiation burden to the patient and may contribute to the misidentification of bone metastases. These challenges have prompted the development of a variety of octadentate chelators for 89 Zr that should result in increased stability of the 89 Zr complexes [13][14][15][16][17][18][19]. Two proprietary chelators, DFO*-NCS [20] and DFOSq [21], both derivatives of DFO, showed preliminary improvements in performance in vitro and in vivo compared with DFO and are currently under consideration for clinical use.
The synthesis of DFO* was first reported by Patra et al. [22] in 2014, and aimed to extend DFO with a fourth hydroxamic acid group. This derivative of DFO presents the advantage of using the oxygen-rich hydroxamate functional group in order to fully coordinate Zr 4+ [7]. Thus, in 2017, Vugts et al. [23] introduced the bifunctional chelator DFO*-NCS and compared [ 89 Zr]Zr-DFO*-NCS-trastuzumab and [ 89 Zr]Zr-DFO-NCS-trastuzumab in vitro and performed a first in vivo pilot study in N87-tumour-bearing nude mice.
[ 89 Zr]Zr-DFO*-NCS-trastuzumab presented promising properties compared with its DFO counterpart with strikingly lower accumulation in bones. In 2016, in parallel to the development of DFO*-NCS, Rudd et al. [24] introduced the bifunctional chelator DFOSq and explained the increased stability of the 89 Zr-chelator complex by the dione oxygen of the squaramide moiety contributing to a putative octadentate coordination of 89 Zr. In vitro, [ 89 Zr]Zr-DFOSq-Taur was more stable than [ 89 Zr]Zr-DFO-p-Ph-SO 3 H when challenged with EDTA. In vivo PET imaging and ex vivo biodistribution studies with [ 89 Zr]Zr-DFOSq-trastuzumab also revealed reduced liver and bone uptake compared with [ 89 Zr]Zr-DFO-NCStrastuzumab as well as satisfactory HER2 tumour targeting.
As DFO* exhibits promising preliminary performance, we compare herein the in vitro and in vivo properties of [ 89 Zr]Zr-DFO*-mAb conjugates to their [ 89 Zr]Zr-DFO analogues using the isothiocyanate and squaramide linker forms. For this purpose, DFO*Sq, a derivative of both DFO* as well as DFOSq, was synthesised and also included in this head-to-head comparison, as this bifunctional chelator might provide additional insight into the mutual contribution of an extra hydroxamate group or squaramide group to 89 Zr complexation (Fig. 1). To allow comparative in vitro and in vivo studies, [ 89 Zr]Zr-DFO*-NCS-trastuzumab, [ 8 9 Zr]Zr-DFOSq-trastuzumab, [ 89 Zr]Zr-DFO-NCS-trastuzumab and [ 89 Zr]Zr-DFO*Sqtrastuzumab were synthesised based on previously described procedures [23]. Their stability was assessed at 37°C in serum and formulation solution (± competing chelators such as EDTA, DFO and DFO*). Next, the biodistribution of the four [ 89 Zr]Zr-trastuzumab conjugates was assessed in HER2-expressing NCI-N87 tumour-bearing nu/nu mice. Superior performance of DFO* over DFO was further confirmed in a second model using a fast growing, highly internalizing tumour model (EGFR-expressing A431 xenograft). Furthermore, chelator stability ([ 89 Zr]Zr-DFO* and [ 89 Zr]Zr-DFO) for competing metals was evaluated in vitro using a panel of nine metals either known for their chelating capacity with DFO or their natural abundance in the human body. Finally, to evaluate the practical advantages of stable 89 Zr coupling for the accurate detection of bone metastases, a mouse model of intratibial breast bone metastases was developed and evaluated using trastuzumab and non-binding control mAb B12 comparing the chelators DFO* and DFO.

Quality controls
Radiochemical purity, protein integrity, binding assays as well as determination of chelator-to-mAb ratio can be found in the electronic supplementary information.  Zr]Zr-DFO*Sq-trastuzumab (0.2 mg mAb/mL, 10 MBq/mL) were incubated in the presence of human serum (Sigma-Aldrich) following conditions described previously [23]. Five hundred microlitres of each of the radioimmunoconjugates in formulation buffer adjusted to pH 7 with 2 M Na 2 CO 3 were incubated in triplicate with 500 μL of serum. All samples were incubated over a week at 37°C in a CO 2 incubator in the presence of 0.02% NaN 3 (8 μL, 25 mg/mL in MilliQ water) to maintain sterility. The initial pH was 7, and this stayed constant during the course of the incubation period (measured values between 7.0 and 7.3). Samples were taken before incubation (day 0), and at 1, 3 and 7 days in a laminar flow hood to avoid contamination. Radiochemical purity was determined by SE-HPLC and spin filter analysis (as described in the electronic supplementary information). Immunoreactivity was checked at day 0 and 7 days by a binding assay as described in the electronic supplementary information.

EDTA, DFO and DFO* challenge of the radioimmunoconjugates
The four radioimmunoconjugates were challenged with EDTA, DFO and DFO* as described in the electronic supplementary information.

Metals and other cation challenge of [ 89 Zr]Zr-DFO* or [ 89 Zr] Zr-DFO
This experiment was adapted from described methods [26,27] and can be found in the electronic supplementary information.  Zr]Zr-DFO*Sq-trastuzumab was evaluated in tumourbearing mice. Forty female nu/nu mice were injected subcutaneously (s.c.) in both flanks with 2 × 10 6 NCI-N87 cells. Tumour growth was monitored on a daily basis, and tumour volume was assessed with a calliper ((length × width × depth) / 2) at least twice a week as soon as tumours became detectable. When tumours reached an average volume of 100-200 mm 3 , mice were randomised and divided in 8 groups of 5 mice for injection of 100 μg of radioimmunoconjugate in 100-200 μL. Of radioimmunoconjugate, 1.1 MBq was administered intravenously (i.v.) via the retro orbital plexus to animals under anaesthesia with inhalation of 2-4% isoflurane/O 2 . At 2, 24, 48, 72 and 144 h post-injection (p.i.), blood samples were drawn, and at 72 and 144 h p.i., 5 mice per group were anaesthetised, bled, euthanised and dissected. Additionally, among the mice sacrificed at 144 h p.i., two mice per group were imaged at 24, 72 and 144 h p.i. For all mice, blood, tumours and organs of interest were collected, weighed and the amount of radioactivity in each sample was measured in the gamma counter. Radioactivity uptake was calculated as the percentage of the injected dose per gram of tissue (%ID/g).

Biodistribution studies in a BT-474 bone metastasis model
To evaluate the performance of DFO* and DFO mAb conjugates in the accurate detection of bone metastases, an intratibial bone metastasis model was developed following a procedure adapted from Campbell et al. [28]. For this purpose, BT-474, a highly proliferating HER2 positive cell line that can present osteoblastic and osteoclastic bone remodelling properties, was employed [29]. One day before surgery until 2 days after surgery, the drinking water of nu/nu mice (11 weeks old) was replaced with water containing carprofen (Rimadyl®) at a concentration of 0.067 mg/mL. On the day of surgery, all mice received 0.1 mg/kg of buprenorphine (Temgesic®) s.c., 30 min before the procedure. Mice were closely monitored during anaesthesia, and an incision was made in the skin to expose the left tibia of each mouse followed by injection of 1.5 × 10 6 (in 10 μL PBS) luciferase transfected BT-474 cells directly in the tibia. After suture, the same procedure was performed with PBS in the right tibia of every mouse as a negative control. Disease progression was followed by bioluminescence (In Vivo Xtreme, Bruker, The Netherlands) and CT imaging using a preclinical NanoPET/CT (Mediso, Hungary) until tracer injection approximately 6 weeks later.

PET imaging
PET imaging was performed with a dedicated small animal NanoPET/CT scanner (Mediso Ltd., Hungary). Mice were anaesthetised by inhalation of 2-4% isoflurane/O 2 during the whole scanning period (1-h duration per time point). A 5-min CT scan was acquired prior to each PET scan and used for attenuation and scatter correction purposes. Reconstruction was performed using a 3dimensional reconstruction algorithm (Tera-Tomo; Mediso Ltd.) with four iterations and six subsets, resulting in an isotropic 0.4-mm voxel dimension.

Statistics
The Grubbs outlier test was used to check and remove outliers, and statistical analysis was performed on the tissue uptake values of the different groups of mice with the Welch's t test. For biodistribution data, the Grubbs test is useful to determine if one value within a group of mice deviates too much (lower or higher) from the mean. Welch t test is a t test for small groups which does not assume that the variances are equal between the groups. Both assume normal Gaussian distribution of the values. Two-sided significance levels were calculated, and p < 0.05 was considered to be statistically significant. All graphs were generated using GraphPad Prism 5.02 software.

Synthesis of 89 Zr-labelled compounds
All radioimmunoconjugates were obtained with a non-decay corrected radiochemical yield between 76 and 85%, a radiochemical purity above 98% and a preserved immunoreactive fraction (> 95%). The chelator to mAb ratio was on average 1 (0.7-1.2) for all radioimmunoconjugates.

Serum stability of radioimmunoconjugates
When the four radioimmunoconjugates were incubated with serum for a week in a CO 2   f o r m u l a t i o n b u f f e r a d j u s t e d t o p H 7 , a l l f o u r radioimmunoconjugates presented a radiochemical purity above 95% except for [ 89 Zr]Zr-DFO-NCS-trastuzumab, which only had a radiochemical purity of 70 ± 9% (Fig. 3a). When challenged with 375 equivalents of EDTA disodium salt, DFOmesylate or DFO* (Fig. 3b- (Fig. 3b) while for the DFO radioimmunoconjugates, stability was lower with 82 ± 4% for the DFOSq radioimmunoconjugate followed by 68 ± 1% for the DFO-NCS radioimmunoconjugate. With 375 equivalents DFO-mesylate instead of EDTA disodium salt as challenging chelator, more than 70% intact tracer was left in the case of the DFO* radioimmunoconjugates after incubation for 24 h, while the DFO radioimmunoconjugates were less than 10% intact already after 4 h of incubation (Fig. 3c). With 375 equivalents of DFO*, the effect on stability was very similar to that observed for DFO (Fig. 3d). When the radioimmunoconjugates were challenged with 3750 equivalents of EDTA disodium salt (Fig. 3e) or DFO-mesylate (Fig. 3f), the same trends were observed, although with a more drastic decrease in radiochemical purities, while radiochemical purities of the DFO* radioimmunoconjugates were always higher than those of the comparable DFO radioimmunoconjugates.
Comparable experiments were also performed in formulation buffer at pH 5.5 instead of pH 7.0 (see Fig. S1 Table 1). Niobium caused a drastic decrease for both DFO* and DFO with radiochemical purities below 10% within the first hour. Iron induced a gradual decrease in radiochemical purity with 93 ± 1% for [ 89 Zr]Zr-DFO* and 89 ± 2% for [ 89 Zr]Zr-DFO (significant difference, p < 0.05) after 1 week.
In the bones (sternum, thigh bone and knee), the two DFO* radioimmunoconjugates presented significantly lower uptake than the two DFO radioimmunoconjugates (Fig. 5). In the knee, the difference was the largest and increased over time in favour of the DFO* radioimmunoconjugates compared with the two DFO radioimmunoconjugates. At 144 h p.i., [ 89 Zr]Zr-DFO*-NCS-trastuzumab presented the lowest uptake in the knee (1.2 ± 0.3%ID/g), followed by [ 89 Zr]Zr-DFO*Sq-trastuzumab (1.6 ± 0.6%ID/g), [ 89 Zr]Zr-DFO-NCS-trastuzumab (7.9 ± 0.7%ID/g) and finally [ 89 Zr]Zr-DFOSq-trastuzumab (12.1 ± 4.6%ID/g) (Fig. 5b). PET   Fig. S2, and the images obtained at 144 h p.i. are presented in Fig. 6. Notably, the spinal cord, which was not collected for ex vivo tissue distribution, showed higher uptake of the DFO conjugates, especially the DFOSq radioimmunoconjugate. Concerning metabolic organs, at 72 and 144 h p.i., a tendency was observed for a lower liver uptake of the DFO*-NCS and DFO*Sq radioimmunoconjugates compared with the DFO-Sq and DFO-NCS radioimmunoconjugates, which was only significant at 144 h when the two DFO* radioimmunoconjugates were compared with  Table S3. At 72 and 144 h p.i., a significantly lower uptake in bone was observed for [ 89 Zr]Zr-DFO*-NCS-cetuximab compared with [ 89 Zr]Zr-DFO-NCS-cetuximab (p < 0.05 or p < 0.01), which was already present at 24 h p.i., except for the sternum and the head, confirming the superiority of the DFO* form as chelator as observed with trastuzumab.

Discussion
In this study, we evaluated in depth the in vitro as well as the in vivo performance of DFO* as a future clinical candidate chelator for 89 Zr-immuno-PET to replace the current gold standard: DFO. The combination of 89 Zr with DFO leads to unwanted bone uptake in preclinical studies. In clinical studies, this has not been discussed to date, most probably because either uptake in bones is less obvious or because uptake in the case of bone lesions is largely unknown. Furthermore, a new improved chelator has not as yet been directly compared in humans with the current gold standard DFO. The observed preclinical instability has been repeatedly reported in the literature [8,10] and has led to the development of many new trivalent and tetravalent chelators. In this study, DFO* and DFO radioimmunoconjugates were compared consisting of two different linkers (squaramide and isothiocyanate) between the chelator and mAb.  (Figs. 2 and 3). This was further confirmed in vivo by assessing the uptake in bone; [ 89 Zr]Zr-DFO*-NCS-trastuzumab showed a significantly lower uptake than [ 89 Zr]Zr-DFOSq-trastuzumab (Figs. 4, 5 and 6). These data are in line with a recent study from Berg et al. [25], in which rhesus monkeys were injected with 89 Zr-labelled mAbs and followed by total-body PET imaging over a period of 30 days. The results showed that [ 89 Zr]Zr-DFOSq-trastuzumab had substantial and significant bone uptake compared with [ 8 9 Zr]Zr-DFO*-NCStrastuzumab at all imaging time points after day 0 (p value < 0.05). i. of 100 μg per construct. Uptake expressed as %ID/g (mean ± SD, n = 5-6 animals per group). Significant differences between the four constructs are marked with asterisks (*p < 0.05; **p < 0.01) In the comparison between DFO*-NCS and DFOSq, DFO-NCS was included as the current clinical gold standard and benchmark for the studies presented herein. While the in vitro and in vivo performance of DFO*NCS appeared clearly superior to DFO-NCS, no obvious advantages were found for DFOSq over DFO-NCS. The superiority of DFO*-NCS in comparison with DFO-NCS became apparent in vitro in serum as well as in chelator challenge experiments. Furthermore, in vivo, a clearly decreased bone uptake using DFO* was observed in two xenograft models with [ 89 Zr]Zr-DFO*-trastuzumab and cetuximab. These results are in line with our preliminary results reported previously [23,30]. Moreover, the impact of more stable 89 (Fig. S4).
In our study, DFOSq did not perform better than DFO-NCS in vivo, while a trend towards higher stability was observed for DFOSq in vitro (Figs. 2 and 3a, b) Fig. 3b-f). This is in line with the reported in vitro results of Rudd et al. [24]. They observed a higher stability for DFOSq than DFO-NCS in a challenge experiment with 500 equivalents EDTA (pH 7 for 24 h) comparing [ 89 Zr]Zr-DFOSqTaur (88% stability) with [ 89 Zr]Zr-DFO-p-PhSO 3 H (70%) in water-soluble conditions. In vivo, Rudd et al. [24] showed that the tumour-to-bone ratio was better for DFOSq than for DFO-NCS and also the tumour uptake was higher for DFOSq, but no actual uptake (i.e. in %ID/g) levels in bone were reported.
To evaluate the coordination chemistry of 89 Zr with DFOSq, we introduced a hybrid chelator, called DFO*Sq, consisting of octadentate chelator DFO* and the same squaramide linker as in DFOSq. While DFOSq and DFO-NCS showed comparable stability in vitro and in vivo, DFO*-NCS and DFO*-Sq exhibited the same superior stability compared with the DFO radioimmunoconjugates, indicating that an extra hydroxamate group is contributing more strongly than a squaramide group to 89 Zr coordination (Fig.  1, option B). Recently, Holland [31] investigated with density functional theory (DFT) the different coordination isomers for 89 Zr. While with DFO* eight-coordinate isomers were the most likely, only one of the oxygen atoms of the squaramide moiety of DFOSq seemed to be involved, resulting in a sevencoordinate complex.
In literature, several chelators have been described that aimed at improving coordination of 89 Zr 4+ , but only few showed promising properties and conjugation to mAbs for in vivo application [7,[32][33][34][35]. For example, two bifunctional chelators, p-SCN-Bn-HOPO (based on 3,4,3-(LI-1,2-HOPO) [12] and more recently DFO-cyclo*-pPhe-NCS [36], have been suggested as second-generation clinical candidate chelators. In the case of p-SCN-Bn-HOPO, the synthesis has remained problematic [37], which limits clinical utilisation. DFO-cyclo*-pPhe-NCS, reported during the course of our studies, is a racemic compound combining DFO with an additional cyclic hydroxamate moiety and the same linker as used in DFO*-NCS. This chelator demonstrated promising in vitro properties in EDTA and DFO challenge experiments; however, in vivo [ 89 Zr]Zr-DFO-cyclo*-NCS-trastuzumab did not show superiority over [ 89 Zr]Zr-DFO*-NCS-trastuzumab in HER2+ SKOV-3 tumour-bearing mice. Finally, DOTA, a well-known chelator for other radiometals such as 111 In, 177 Lu and 90 Y, has recently been used to complex 89 Zr, but has not yet been evaluated as a bifunctional chelator variant [26,38]. DOTA may be limited by the high temperature (reported temperature 95°C) required for efficient radiolabelling which may necessitate a prelabeling strategy to generate [ 89 Zr]Zr-DOTA-mAb complexes. Furthermore, 89 Zr in oxalic acid needs to be converted to 89 Zr in HCl to allow efficient complexation with DOTA.
The question remains as to whether DFO* is the ideal chelator and which linker is most suited for coupling 89 Zr-chelator to biomolecules. In other words, what are the advantages and disadvantages with respect to (i) solubility of linker-chelator, (ii) reactivity of linker-chelator, (iii) radiolabelling of chelator-mAb complexes, (iv) stability of radioimmunoconjugate and (v) availability of the linker-chelator. Firstly, DFO*-NCS and DFO-NCS are known to be soluble in DMSO but poorly in water, which is also the case for DFO*Sq, while DFOSq is relatively soluble in water (90/10% water/DMSO). Concentrations as used in the conjugation (i.e. 5 mM) are easily achievable in all cases. Thus, solubility is not a limiting factor for conjugation to mAbs in our opinion, since a small percentage of DMSO is allowed in these reactions and coupling is efficient. Also, from a GMP perspective, since DMSO can be easily and efficiently removed from the radioimmunoconjugate before formulation, e.g. during PD-10 purification. Secondly, for modifications of mAbs, DFOSq and DFO*Sq seem to require longer incubation times: overnight at RT, using 3 and 5 equivalents of DFOSq and DFO*Sq, respectively, resulted in 1 DFOSq per mAb molecule on average. This could be improved by increasing the chelator-to-mAb ratio in the conjugation or by increasing the reaction temperature, but to date, no data is available showing to what extent this could improve the conjugation protocol. This constitutes a difference with DFO*-NCS and DFO-NCS for which coupling to mAbs is fast (typically 30 min to 2 h, at 37°C resulting in case of 3 equivalents of DFO(*)-NCS in one chelator per mAb). Thirdly, radiolabelling of mAbs with 89 Zr via DFO* or DFO is equally efficient and the same sensitivity is observed for metal impurities: Fe 3+ and Nb 3+ can interfere with 89 Zr binding not only to DFO, but also to DFO*. Although these metals do not affect DFO* radioimmunoconjugate stability in vivo, their presence as impurities might affect the labelling efficiency and the stability of radioimmunoconjugates in vitro, as previously observed by Pandya et al. [26] and Deri et al. [27] with DFO and iron. These findings were not surprising, since DFO is clinically used for the treatment of iron overload, while it has also been used for stable radiolabelling of monoclonal antibodies with 59 Fe [39] and 95 Nb, as in studies with [ 95 Nb]Nb-DFO-bevacizumab [40,41]. Fourthly, the in vitro stability of radioimmunoconjugates consisting of DFO* is better than the ones consisting of DFO, and therefore, the need for anti-oxidant additives is less. This is especially important in the case of central tracer manufacturing and distribution to clinical sites. Finally, as DFO*-NCS is now commercially available with a structure very close to the current standard DFO-NCS, its translation to the clinic is facilitated and expected soon.

Conclusion
In the evaluation of next-generation chelator candidate for clinical 89 Zr-immuno-PET, DFO* showed a superior in vitro and in vivo performance over the current clinical gold standard DFO, regardless of the linker used (NCS and Sq).
[ 89 Zr]Zr-DFO*-mAbs appeared the most stable in vivo with the lowest uptake of 89 Zr in bones, which might be highly relevant to avoid misdiagnosis in the case of bone metastases as was shown herein in an in vivo model of bone metastases. In addition, as DFO*-NCS is now commercially available, the clinical translation is under development.
Funding Open access funding provided by Amsterdam UMC (Vrije Universiteit Amsterdam). This research has received funding from the European Union's Horizon 2020 research and innovation programme under the Marie Sklodowska-Curie grant agreement no. 675417. This work was funded by the Swiss National Science Foundation (grant Nr. SNSF 205321_157216 to G.G. and T.L.M.).
Data availability Not applicable.

Compliance with ethical standards
Conflict of interest The authors do not have any conflict of interest. Herman Gill, Jan Marik and Simon Williams are paid employees of Genentech, Inc. Patents are filed for DFO*-NCS and DFOSq, which are referenced in this manuscript [20,21].
Ethics approval All animal experiments were performed according to the NIH Principles of Laboratory Animal Care and Dutch national law ("Wet op de dierproeven," Stb 1985, 336). All cell lines used are obtained from the American Type Culture Collection (ATCC).

Consent to participate Not applicable.
Consent for publication All authors agreed with the content and gave explicit consent to submit this manuscript; they obtained consent from their responsible authorities at their institute/organisation.

Code availability Not applicable.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not