Synthesis and characterization of ⁸⁹Zr-labeled ultrasmall nanoparticles

Abstract

The ultrasmall nanoparticle AGuIX is a versatile platform that tolerates a range of chemical diversity for theranostic applications. Our previous work showed that AGuIX clears rapidly from normal tissues, while durably accumulating within the tumor microenvironment. On this basis, AGuIX was used to detect tumor tissue with Gd³⁺ enhanced MRI, and can sensitize tumors to radiation therapy. As we begin the translation of AGuIX, we appreciated that coupling AGuIX to a long-lived radioisotope would help to more completely measure the magnitude and duration of its retention within the tumor microenvironment. Therefore, we developed ⁸⁹Zr-DFO-AGuIX. AGuIX was coupled to DFO and then to ⁸⁹Zr in ~99% radiochemical yield. Stability studies showed that ⁸⁹Zr-DFO-AGuIX did not dissociate after 72 hours. In animals bearing U87MG xenografts, it was detectable at levels above background for 72 hours. Lastly, ⁸⁹Zr-DFO-AGuIX did not accumulate in inflammatory abscesses in vivo, highlighting its specificity for well vascularized tumors.

Introduction:

Owing to their chemical versatility and encouraging properties in vivo, nanotechnologies have been aggressively developed and refined over the past two decades for biomedical applications^{1, 2}. We have previously developed AGuIX, an ultrasmall nanoparticle (NP) for theranostic applications in cancer^2–5. AGuIX is ~10 kDa NP with a hydrodynamic diameter of ~3 nm. The core of the NP is a polysiloxane matrix, onto which chemically diverse chelators can be grafted. For instance, about 10 DOTAGA chelators have been engineered per NP to bind Gd³⁺ for contrast enhanced MRI in vivo^{6, 7}. Moreover, alternative chelators can be attached onto AGuIX for nuclear imaging, and we recently engineered one additional NODAGA chelator to bind ⁶⁸Ga³⁺ to AGuIX for PET in vivo⁸. Lastly, near-infrared fluorescent molecules can be conjugated to the NP to track its tissue localization with microscopy or to follow them by optical imaging^{9, 10}.

Applying these constructs in vivo to study the biodistribution of AGuIX longitudinally with MRI and PET has shown that the construct does not accumulate substantially in normal tissues, and rapidly clears from rodents via the kidneys. After 24 hours, ~90% of the NPs are cleared in the urine, with ~10% of the injected NPs bound within the kidneys and ~0.2% distributed in all other normal tissues. Microscopy with fluorescently tagged AGuIX showed retention within the proximal tubules of the kidneys, with subsequent excretion of the native or partially degraded NPs.¹¹. Lastly, sufficient accumulation of the NP has been observed in the microenvironment of preclinical tumor models to amplify the therapeutic effects of external beam radiation therapy^{11, 12,}. Collectively, these observations strongly justify first in man studies with AGuIX.

Long lived radioisotopes (e.g. zirconium-89, iodine-124) are essential to fully study the pharmacokinetics and biodistribution of molecules with higher order structure in vivo¹³. Because AGuIX can be readily engineered with chemically discrete chelators, we reasoned that ⁸⁹Zr could be incorporated onto chelates grafted at the surface of the polysiloxane core for nuclear imaging. Indeed, we and others have shown that ⁸⁹Zr is an excellent radionuclide to characterize monoclonal antibodies in preclinical models of cancer, and that ⁸⁹Zr can be stably coupled to large biomolecules with a desferrioxamine (DFO) chelator^{13, 14}. These considerations motivated us to design a synthetic route to DFO-conjugated AGuIX, and to evaluate the properties of ⁸⁹Zr-DFO-AGuIX in vitro and in vivo.

Experimental Section:

Chemicals:

Sodium hydroxide (NaOH, 99.99%), hydrochloric acid (HCl, 36.5−38%) and dimethylsulfoxide (DMSO, >99.5%) were purchased from Aldrich Chemical (France), acetonitrile (CH3CN, >99.9%) was purchased from Carlo Erba (France), trifluoroacetic acid (TFA, >99%) was purchased from Alfa Aesar (United Kingdom), copper sulfate pentahydrate (CuSO₄.5H₂O, 98%) was purchased from Merck (Germany). AGuIX particles were purchased from Nano-H (Saint-Quentin Fallavier, France). The desferrioxamine chelate, p-NCS-Bz-DFO (N1-hydroxy-N1-(5-(4-(hydroxy(5-(3-(4-isothiocyanatophenyl)thioureido)pentyl)amino)-4-oxobutanamido)pentyl)-N4-(5-(N-hydroxyacetamido)pentyl)succinamide) was purchased from ChemaTech (France). All products were used without further purification. Only Mili-Q water (ρ>18MΩ.cm) was used for the aqueous solution preparation.

Synthesis of DFO-AGuIX:

AGuIX nanoparticles (500 μmol based on [Gd³⁺]_, 1 eq.) were dispersed in 6.67 mL of water for 1 hour at a pH of 7.4. 37.6 mg of p-NCS-Bz-DFO (50 μmol, 1 eq.) were dissolved in 4.80 mL of DMSO. The DMSO solution was then gradually added over 8 hours to the AGuIX solution under stirring at 50°C. The addition was performed gradually to prevent precipitation of the ligands and of the functionalized product. After addition was complete, the pH was decreased to 5.0 with 1M HCl solution and stirred at room temperature overnight to precipitate free unreacted DFO. Unreacted DFO was firstly removed by centrifugation; all other other unreacted products were removed by tangential filtration trough Vivaspin® membranes (molecular weight threshold = 5 kDa, Sartorius Stedim Biotech). The DFO-AGuIX solution was concentrated to approximately [Gd³⁺] = 100 mM after many cycles of tangential filtration. The solution was sterile filtered through 0.2 μm syringe filter in order to remove the largest impurities. It was then freeze-dried for storage, using a Christ Alpha 1–2 lyophilizer. The Gd³⁺ yield for the synthesis of AGuIX prior to conjugating to DFO was 48%, which was determined with relaxometry and inductively coupled plasma mass spectrometry as reported previously⁸. After conjugation of DFO to AGuIX, the yield of the final molecule was determined to be 68% using ICP-MS.

Radiochemistry:

To a solution of [⁸⁹Zr]Zr-oxalic acid (1.2mCi; 12 μl) and 2 M Na₂CO₃ (12 μl) was added 0.30 ml of 0.5 M HEPES (pH 7.1–7.3), 0.50 ml of DFO-AGuIX (pH = 7, 22.5 μmoles Gd) or 0.50 of DFO (Deferoxamine mesylate salt, pH = 7, 15 mg), and 0.70 ml 0.5 M HEPES (pH 7.1–7.3) into the reaction vial were added. After incubation (120 minutes) at 37° C, the radiolabeling yield was determined by iTLC. The final product was filtered through a 0.2 micron filter into a sterile vial. Radiolabeling efficiency (typically > 98.5%) was determined by ITLC using chromatography strips and 20 mM citric acid (pH 4.9–5.1) as the mobile phase. In vitro stability of ⁸⁹Zr-AGuIX was assessed in serum for up to 3 days. For the serum stability studies, 50 μL ⁸⁹Zr-AGuIX at 50 mM in Gd³⁺ were added to 450 μL serum and the mixture was incubated at 37 °C. Samples were taken at 1h, 4h, 24h, and 75h and assessed by ITLC. ⁸⁹Zr-transferrin was prepared as previously described¹⁵.

Serum stability determination:

Blood samples were immediately subjected to centrifugation at 2300g for 15 min at 4 °C, and the serum supernatants were collected. The supernatant was immediately aliquoted and studied. In vitro, 50 μCi ⁸⁹Zr-DFO-AGuIX were incubated with 100% fetal bovine serum and aliquots were reserved for analysis at 1 hr, 4 hr, 8 hr, 24 hr and 72 hr. Healthy nu/nu mice (8–10 weeks old, n = 3) were injected with 50 ± 3 uCi of ⁸⁹Zr-DFO-AGuIX preparation in 100 μL HEPES solution (pH=7.4). Blood was sampled from the submandibular vein at 1 hr, 4 hr, 8 hr, 24 hr and 72 hr post injection. Activity within the blood samples was resolved by ITLC and each experiment was carried out in triplicate.

Animal studies:

All animal protocols were approved by the UCSF Administrative on Laboratory Animal Care. For tumor imaging studies, 8-wk-old male nude (nu/nu) mice (Charles River Laboratories) were injected subcutaneously with 1 × 10⁶ cells containing 50% Matrigel (v/v) in the flank while anesthetized with 2% isoflurane. Animals were used for imaging studies with MRI and PET once the tumors became palpable. For MRI studies, AGuIX (20 μmol) and DOTAREM (20 μmol) were administered via tail vein injection. Animals were imaged with MRI 20 minutes post injection. For biodistribution studies with inflammatory abscesses, ~50 μL of turpentine oil (Sigma Aldrich) was injected subcutaneously in the right hind limb per a previously established methodology¹⁶. Fifteen minutes was allowed to pass to generate the acute phase response, whereupon the mice were injected with ⁸⁹Zr-transferrin or ⁸⁹Zr-DFO-AGuIX. Animals were euthanized 30 min post injection with CO₂ (g) asphyxiation.

Magnetic resonance imaging:

All MRI acquisitions were performed on a 7T Small Animal MRI facility at UCSF, using a 7T 300 MHz Horizontal Bore Varian MR System. Pre and Post-treatment MR imaging was performed. A multislice spin-echo sequence was used to generate spin-density weighted images, with a TR of 3000 ms and a TE of 18 ms. Sample scans had an acquisition matrix of 128×128 points, with the field of view measuring 19.2×19.2 mm. Prior to injection in mice, a concentrated colloid (AGuIX in water, [Gd³⁺] = 100 mM) was diluted by saline solution in order to obtain an intravenous use solution ([Gd³⁺] = 50 mM). The pH was adjusted to 7.4. Before use, this solution was filtered onto syringe filter with nylon membrane (pore diameter 0.22 μm). The chelate used was DOTAREM® (laboratories Guerbet, Aulnay sous Bois France, 0.5 mM/mL) as available in MRI units. The aqueous AGuIX was manually injected in the tail vein at 200 μL volume. The gadolinium chelates were injected via the tail vein at the same concentration (50mM) at 200 μL volume.

Results:

AGuIX NPs were synthesized according to our previously established, top-down protocol. Briefly, gadolinium oxide cores were first synthesized in diethylene glycol. The oxide core was coated with a polysiloxane shell using hydrolysis–condensation of aminopropyl triethoxysilane (APTES) and tetraethyl orthosilicate (TEOS). DOTAGA groups (1,4,7,10-tetra-azacyclododecane-1-glutaric anhydride-4,7,10-triacetic acid) were then covalently ligated onto the nanoparticles via a primary amine from APTES. To induce core dissolution, the NPs were transferred from DEG to water. The resulting polysiloxane hollow cores were collapsed and fragmented into small and rigid platforms of polysiloxane. The resulting NPs bear on their surface DOTAGA molecules that are chelated to Gd³⁺ cations.⁷

To prepare DFO-AGuIX, 1-(4-isothiocyanatophenyl)-3- [6,17-dihydroxy-7,10,18,21-tetraoxo-27-(N-acetylhydroxylamino)-6,11,17,22-tetraazaheptaeicosine] thiourea (p-NCS-Bz-DFO) was incubated with AGuIX for one hour at room temperature. DFO was covalently ligated to the solvent exposed primary amines contained within APTES, and the resulting functional group was a thiourea. DFO-AGuIX was subjected to tangential filtration over a 5 kDa membrane to remove any unconjugated p-NCS-Bz-DFO at pH 5. DFO-AGuIX NPs were then freeze−dried for storage.

We expected that adding DFO to AGuIX would change its retention on reverse phase HPLC. On this basis, we subjected DFO-AGuIX to RP-HPLC to determine its purity. Naked AGuIX, DFO-AGuIX, and free DFO were loaded into the injection loop separately. The corresponding chromatogram from the DFO-AGuIX showed peaks at 2.5 min corresponding to degradation fragments of DFO-AGuIX due to the hydrolysis of silyl ether bonds in diluted medium¹⁷. The major peak at 13–14 min corresponded to DFO-AGuIX (Figure 1A). Moreover, DFO-AGuIX eluted later than AGuIX and no free DFO was detected in the DFO-AGuIX formulation. The final colloidal solution was characterized by dynamic light scattering and DFO-AGuIX was determined to have a hydrodynamic diameter of 4.4 ± 1 nm (slightly larger to the hydrodynamic diameter of naked AGuIX that is 3.6 ± 0.8 nm, Figure 1B), which is suitable for renal excretion.

Figure 1.

Synthesis and characterization of DFO-AGuIX. A. An overlay of reverse phase HPLC traces showing the resolution of DFO-AGuIX at ~15 min compared to naked AGuIX (~13 min) and free DFO (~16 min). These data also underscore the purity of DFO-AGuIX after filtration. The early eluting peaks on the AGuIX HPLC trace are minor degradation products. They do not appear on the DFO-AGuIX trace because they are removed with reaction purification. B. Dynamic light scattering data showing that the hydrodynamic diameter of DFO-AGuIX is larger than naked AGuIX, as expected. These data were acquired using the purified DFO-AGuIX material. C. Determination of the number of DFO chelates per AGuIX nanoparticle using Cu²⁺ titration and UV/Vis spectroscopy. The area under the peak corresponding to free Cu²⁺ was calculated after addition of Cu²⁺ on DFO-AGuIX and injection in HPLC (see also Supplemental Figure 1). Under sub-saturating conditions (1), no peak was detected, while increasing the concentration of Cu²⁺, free Cu²⁺ ions were detected as the concentration exceeded ~7 mM (2). This assay was conducted with 38.4 mM AGuIX (in [Gd³⁺]).

To quantify the number of DFO molecules per AGuIX, we titrated Cu²⁺ into a preparation of DFO-AGuIX ([Gd³⁺] = 38.4 mM) and resolved the free and AGuIX bound Cu²⁺ with HPLC. We chose to utilize Cu²⁺ for several reasons. First, the constant of complexation between Cu²⁺ and DFO has been calculated¹⁸, and is only two fold less than Fe³⁺ (log β = 23.98 for [CuLH]⁰, log β = 41.01 for [FeLH]⁺). Moreover, Cu(II) sulfate is easily detected at 700 nm. Lastly, we have previously developed, validated and published this HPLC based approach using Cu(II) sulfate to determine the number of NODA chelators on the AGuIX scaffold⁸.

On titration, a free fraction of Cu(II) sulfate was initially detected at ~7 mM of Cu²⁺, and this point of inflection allowed us to estimate between 1–2 DFO chelators per DFO-AGuIX (Figure 1C and Supplemental Figure 1). A detectable but negligible amount of unknown Cu(II) conjugated degradation products were also observed (Supplemental Figure 2). These early eluting peaks did not interfere with the integration of the Cu(II) sulfate peak on the HPLC trace. ICP-MS performed on DFO-AGuIX also resulted in a relative composition consistent with 1–2 DFO molecules per AGuIX: Gd₁₀APTES*_27.8TEOS*_26.6DOTAGA*_10.2DFO*_1.4. Lastly, a separate set of experiments with Cu²⁺ and naked AGuIX lacking DFO showed almost no evidence of binding (Supplemental Figure 3). This indicates that the DOTAGA chelators within AGuIX are saturated with Gd³⁺ (free chelate < 2%), and the Cu²⁺ binding observed with DFO-AGuIX is due to an interaction between Cu²⁺ and DFO.

DFO-AGuIX was then metallated with ⁸⁹Zr-oxalate in one hour to a radiochemical yield of 99%. Unreacted ⁸⁹Zr-oxalate was removed by centrifugal membrane filtration per a previous protocol¹⁹. The radiochemical purity was determined to be 100% by iTLC using 20 mM citric acid as the mobile phase (Figure 2A and Supplemental Figure 4).

Figure 2.

Synthesis and characterization of ⁸⁹Zr-DFO-AGuIX. A. A representative ITLC showing the near complete metallation of DFO-AGuIX with ⁸⁹Zr-oxalate. The large peak corresponds to activity at the baseline, interpreted to be ⁸⁹Zr-AGuIX, and an arrow indicates the expected R_f for ⁸⁹Zr-oxalate (see also Supplemental Figure 5). B. Representative ITLC traces showing the stability of ⁸⁹Zr-DFO-AGuIX over time in neat fetal bovine serum. No peaks were resolved from baseline, suggesting that ⁸⁹Zr-DFO-AGuIX is not metabolized to smaller radioactive byproducts. C. Representative iTLC traces showing the stability of ⁸⁹Zr-DFO-AGuIX over time in in mouse serum. Mice were injected with ~50 μCi of ⁸⁹Zr-DFO-AGuIX, and blood was harvested at the indicated time point. The isolated serum was spotted and resolved by iTLC.

We next evaluated the stability of ⁸⁹Zr-DFO-AGuIX in vitro and in vivo. No detectable degradation of ⁸⁹Zr-DFO-AGuIX was observed after 72 hours of incubation at 37° C in bovine serum (Figure 2B). Moreover, no evidence of degradation was observed in the serum harvested from mice injected intravenously with ⁸⁹Zr-DFO-AGuIX (Figure 2C and Supplemental Figure 5). Because ⁸⁹Zr-DFO-AGuIX was very stable, we determined the biological half-life of the construct in normal mice. Serial measurements of total activity in tumor naïve nu/nu mice injected with ⁸⁹Zr-AGuIX intravenously showed that the biological half-life is ~ 67 hours (Supplemental Figure 6).

We next asked if ⁸⁹Zr-DFO-AGuIX accumulates within a subcutaneous tumor model in mice. Male nu/nu mice inoculated with subcutaneous U87MG tumors were injected with ⁸⁹Zr-DFO-AGuIX intravenously, and the biodistribution was monitored over time. After observing evidence of accumulation in the tumor with MRI 20 min post injection due to the positive contrast agent properties of DFO-AGuIX (r₁ = 16.7 mmol⁻¹.s⁻¹ and r₂/r₁ = 1.5 at 37°C and 1.4 T) (Supplemental Figure 7), we investigated the tumor associated activity from 24 – 72 hours with biodistribution studies. Durable retention of the NPs (~2% ID/g) was observed in human glioma U87MG tumors, and was statistically greater than the accumulation of ⁸⁹Zr-DFO (~0.5% ID/g at 24 hours post injection, see Figure 3A and 3B, and Supplemental Figures 8 and 9 for biodistribution from a larger panel of normal tissues.). The tumor to blood and tumor to muscle ratio for ⁸⁹Zr-AGuIX incrementally increased over time to values greater than 10 at 72 hours post injection (Figure 3C). The uptake in other normal tissues was low, as expected, with the notable exception of the kidneys (Supplemental Figure 9). Autoradiography of the tumor slices ex vivo also showed that the highest degree of activity within the tumor slices co-localized with areas harboring visually obvious pericellular compartments (Supplemental Figure 10). Lastly, we tested whether ⁸⁹Zr-DFO-AGuIX accumulated in inflammatory abscesses in vivo. After inducing an acute phase response with an intramuscular injection of turpentine, normal nu/nu mice were treated with ⁸⁹Zr-DFO-AGuIX or ⁸⁹Zr-transferrin, a molecule we previously showed to localize to sites of inflammation¹⁵. After 30 min, a biodistribution study was conducted to determine the amount of activity in the inflamed muscle and normal tissues. The uptake of ⁸⁹Zr-Tf was higher in the inflamed muscle compared to the untreated contralateral muscle, as expected. However, no difference in activity was observed between the inflamed and untreated muscles of mice treated with ⁸⁹Zr-AGuIX (Figure 3D and Supplemental Figure 11). Moreover, the level of tissue associated activity in the inflamed muscle was ~0.5% ID/g, or roughly equal to what was observed for ⁸⁹Zr-DFO in U87MG tumors.

Figure 3.

⁸⁹Zr-DFO-AGuIX accumulates in the tumor microenvironment, but not in inflammatory abscesses. A. Biodistribution data at 24 hours post injection of ⁸⁹Zr-DFO-AGuIX or ⁸⁹Zr-DFO shows significantly higher uptake of the NP in the microenvironment of subcutaneous U87MG tumors compared to ⁸⁹Zr-DFO. No substantial differences were observed in muscle or bone, two normal reference tissues, from the cohorts receiving either radionuclide. B. Biodistribution data showing that ⁸⁹Zr-DFO-AGuIX persists in the tumor microenvironment for several days post injection. At 72 hours, the tumor associated activity was ~1.0% ID/g, which is above background. C. A graphical representation of the mean tumor to muscle and tumor to blood ratios over time for mice treated with ⁸⁹Zr-DFO-AGuIX. D. Biodistribution data showing no uptake of ⁸⁹Zr-DFO-AGuIX in the inflamed muscles within the hindlimbs of a mouse cohort. By comparison, ⁸⁹Zr-transferrin showed robust uptake in the inflamed muscle, presumably owing to the abundant expression of the transferrin receptor on peripheral mononuclear blood cells. I.M. = inflamed muscle, C.M. = contralateral unmanipulated muscle.

Discussion:

In this report, we describe the synthesis and characterization of AGuIX functionalized with DFO, and conjugated to ⁸⁹Zr. The construct was synthesized from AGuIX in two steps to >98% radiochemical yield. Consistent with our expectations, the radiolabeled construct is stable out to 72 hours in vitro and in vivo, and durably accumulates into subcutaneous tumors during this time period. Finally, ⁸⁹Zr-DFO-AGuIX did not accumulate in inflammatory abscesses in vivo, as expected. Collectively, these data support the use of ⁸⁹Zr-DFO-AGuIX as a tool to measure the biodistribution and pharmacokinetics of AGuIX in a first-in-man study.

To our knowledge, this is the first example of a ⁸⁹Zr-labeled ultrasmall nanoparticle, though others have conjugated ⁸⁹Zr to larger nanoparticles with longer circulation times^20–24. As we demonstrate, the use of ⁸⁹Zr is justified, as AGuIX can persist for days in the tumor microenvironment due to the enhanced permeability and retention effect. Consistent with previous MR and microscopy data, accumulation within the kidney also persists for days, extending the biological half-life of AGuIX²⁵. Moreover, we felt a radiometal-based approach (as opposed to a halogen) was appropriate, as we have previously shown that chemically discrete chelators can be added to the polysiloxane core after functionalization with DOTAGA and saturation with Gd³⁺. Lastly, ⁸⁹Zr was an attractive radionuclide to work with, owing to its long half-life (~78 hours), and the recently clinical data showing it is well tolerated in patients when attached to monoclonal antibodies^{26, 27}.

Incorporating a long lived radioisotope for PET onto AGuIX presents a more realistic approach to quantify the amount of NPs in cancer at later time points post injection. While our preclinical experience with AGuIX shows that its concentration in tumors can be calculated with MRI within minutes to hours post injection, contrast becomes more difficult to detect in tumors >24 hours post injection⁴. By comparison, a ~1 hour PET acquisition should be sufficient to acquire the ~10 million coincident events needed for tumor contrast based on the magnitude of radioactivity detected in the biodistribution study. PET also broadens the dynamic range of concentrations that can be measured by lowering the limit of NP detection compared to contrast enhanced MRI. All of these considerations are significant, as the pharmacokinetics of the NP will dictate the design of a pending clinical trial to determine its value as a radiosensitizer.

Using our current synthetic scheme, approximately 1 molecule of DFO was incorporated per AGuIX NP. In the future, we anticipate more DFO chelators can be added to the unreacted primary amines on the polysiloxane core, which in turn could favorably raise the specific activity of the construct. Along these lines, our data also argues strongly for the ligation of ⁸⁹Zr-AGuIX onto large biomolecules (e.g. IgG, Fab) for immunoPET applications¹⁴. We are currently exploring these frontiers.