Capturing an elusive but critical element: Natural protein enables actinium chemistry.

Actinium-based therapies could revolutionize cancer medicine but remain tantalizing due to the difficulties in studying and limited knowledge of Ac chemistry. Current efforts focus on small synthetic chelators, limiting radioisotope complexation and purification efficiencies. Here, we demonstrate a straightforward strategy to purify medically relevant radiometals, actinium(III) and yttrium(III), and probe their chemistry, using the recently discovered protein, lanmodulin. The stoichiometry, solution behavior, and formation constant of the 228Ac3+-lanmodulin complex and its 90Y3+/natY3+/natLa3+ analogs were experimentally determined, representing the first actinium-protein and strongest actinide(III)-protein complex (sub-picomolar Kd) to be characterized. Lanmodulin’s unparalleled properties enable the facile purification recovery of radiometals, even in the presence of >10+10 equivalents of competing ions and at ultratrace levels: down to 2 femtograms 90Y3+ and 40 attograms 228Ac3+. The lanmodulin-based approach charts a new course to study elusive isotopes and develop versatile chelating platforms for medical radiometals, both for high-value separations and potential in vivo applications.

INTRODUCTION

The use of radionuclides in medicine has been continuously growing over the past decades, especially in the case of short-lived radiometals, as they are among the most efficient means for medical imaging [e.g., positron emission tomography (PET) and single-photon emission computed tomography (SPECT) scans] and treatment of a wide array of diseases (e.g., bone metastasis, prostate cancers, and non-Hodgkin’s lymphoma) (–). In particular, actinium (Ac3+) is largely seen as the most promising candidate for next-generation cancer treatments, i.e., targeted alpha therapy (, , ). The nuclear properties of 225Ac make it an ideal projectile against tumor cells (fig. S1): 225Ac is a rare case of an alpha emitter with only short-lived daughters, decaying without leaving radioactive materials in the patient’s body; its 9.95-day half-life is suitable for medical applications; and its decay chain delivers, on average, four high-energy alpha plus two beta particles—leading to treatment efficacy hundreds of times higher (–) than current medical radiometals. Therefore, if successful, the combined large-scale production, purification, and assembly of 225Ac pharmaceuticals has the potential to revolutionize cancer medicine.

However, no 225Ac-based drug has reached the market because of substantial scientific challenges, from isotope production to the design of specific chelators for radiolabeled drugs. Ac3+ chemistry remains largely unknown compared to cations currently used in medicine (Ga3+, Zr4+, Tc5+/7+, Y3+, Gd3+, Lu3+, Sc3+, In3+), other heavy radioelements for oncology (223Ra2+, 227Th4+) (, ), or even synthetic actinides discovered decades after actinium (Np, Pu, Am), which impairs the development of any application involving actinium. For example, only very recently was the coordination number of Ac3+ in HCl solution determined (), and only a handful of Ac3+ complexes have been thermodynamically characterized, all with simple ligands (e.g., AcCl3, AcF3 [Ac(C2O4)2]−, and [AcEDTA]−) (). In addition, since only one Ac3+ complex with a small organic chelator has ever been isolated [i.e., Ac2(C2O4)3·nH2O(s) (, )] and partially characterized via powder x-ray diffraction, the properties of Ac complexes are typically extrapolated from those of the easier-to-study analogs with lanthanides and heavy actinides (Am3+ and Cm3+). Another barrier is the low availability of all Ac materials even at the research scale. All Ac isotopes are radioactive (Fig. 1A). The longest-lived one, 227Ac (21.8 years), would be suitable for chemical studies, but it is cost prohibitive (>1000 USD/μg), is difficult to produce and purify, and requires specialized handling (i.e., a glovebox), even relative to many other radioisotopes (, ). 225Ac is the main focus of medical research, and multiple production routes are being investigated worldwide, as the current isotope availability is much lower than the projected demand. Current production methods require large instruments like particle accelerators and cyclotrons or necessitate decades-old legacy nuclear materials, limiting 225Ac availability. In addition, access is largely limited to clinical development rather than the fundamental studies of Ac3+ chemistry that would be helpful to better design those clinical applications (, –). The 235U/231Pa and 233U/229Th decay chains include 227Ac and 225Ac, respectively, but these radionuclides have very long secular equilibrium periods (fig. S1 and table S1) and present nuclear proliferation issues, so they are tightly controlled and unlikely to make actinium more accessible. Consequently, there is no widely available generator from which Ac isotopes could be periodically extracted, forcing research laboratories willing to embark on actinium studies to purchase limited, expensive, and short-lived 225Ac sources.

Fig. 1.

Actinium isotopes and protein-based radiometal recovery.

(A) Half-lives of known actinium isotopes. 225Ac is the principal actinium isotope studied for targeted alpha therapies. 227Ac is a research isotope and only produced at the microgram scale. Traces of 227Ac are also present in aged 235U or 231Pa materials. 228Ac occurs in the decay chain of natural thorium (232Th) at very low concentrations, requiring a highly efficient extraction system. The other isotopes are short-lived and rarely studied. (B) Principle of the lanmodulin (LanM)–based one-step radioisotope separation method via centrifugation using the “spin filters” or “protein concentrators.” In this configuration, the protein is retained with the target isotope (e.g., Y3+ and Ac3+), while the low–molecular weight fraction is eluted. The filtrates contain the purified unbound metals (e.g., radium, strontium, and thorium). (C) LanM-based metal separation was performed using a size exclusion column (also called “desalting column”). In this configuration, the radiolabeled LanM elutes before the unbound metals. The separation can be performed by gravity elution, pumped elution, or centrifugation. In both (B) and (C) scenarios, no matrix change or acidity adjustment is needed to recover the elements in separate fractions. The radiometal can be desorbed from LanM by decreasing pH (<2) or adding a chelator in excess.

By contrast, 228Ac3+ (6.1 hours) occurs in the decay chain of 232Th4+, an inexpensive radioisotope that is abundant and already available at numerous industrial sites and research institutions. 228Ac3+ has been largely overlooked due to the lack of practical extraction methods, since the mass of actinium isotopes in aged radioactive sources is extremely low (table S1). After 10 years, 1 g of 232Th only contains ~3.5 × 10−14 g of 228Ac—enough for radiochemical studies but beyond the capabilities of current Ac chelators. 228Ac could become the most accessible Ac isotope and help to develop Ac chemistry as a whole, but only if an efficient and practical purification method can be implemented.

Here, we show that such a method can be realized using the protein lanmodulin (LanM), the first natural macrochelator selective for rare earth elements to be discovered (). Taking advantage of LanM’s size, selectivity, and affinity, we purify 228Ac3+ and also 90Y3+, a rare earth radioisotope currently in clinical use in medical PET and SPECT imaging, from their generators (232Th4+/228Ra2+ and 90Sr2+, respectively) with unprecedented ease of implementation (Fig. 1, B and C). Furthermore, our method allows us to study the chemistry of Ac3+ with LanM—by far the most complex ligand for Ac3+ characterized to date—and other relevant bioligands. These studies reveal a higher affinity even than for the protein’s natural substrates (lanthanides) and the highest stability constant reported for any Ac3+ complex, small molecules included (). Because many radiometals are trivalent rare earths or actinides that LanM selectively binds and that would otherwise require cumbersome purification processes, our work suggests a general strategy for rapid radioisotope purification, detection, study, and perhaps even theranostic applications.

RESULTS

Macromolecular purification of radioisotopes

The use of biomacromolecules, such as proteins, has largely been absent from purification strategies targeting radiometals like yttrium and actinium, as most known metal-binding proteins are not selective for f-elements (–). Even previous studies investigating interactions of proteins with uranium or heavy actinides (e.g., plutonium, americium, and curium) (–) have largely focused on bacterial and human iron or calcium transport machineries, which not only form relatively weak complexes with trivalent f-elements but also only operate under a very narrow set of conditions (typically pH >6 and requiring a synergistic ligand like a carbonate or a siderophore molecule). By contrast, LanM displays unprecedented selectivity for trivalent lanthanides against most other cations (, , ). Instead of using minor differences in extraction reaction constants [as with current extractants, resulting in arduous metal ion separations (, )], a selective macromolecular extractant such as LanM [~12 kDa, hydrodynamic radius ~4 nm ()] would enable a new and more robust mechanism for recovery and purification of radiometals based on size differences. We hypothesized that protein-bound versus unbound metal ions could easily be separated using inexpensive methods like spin filtration or size exclusion chromatography (Fig. 1, B and C).

Short-lived radioisotopes require special handling, but their actual concentration is very low (table S1), necessitating extremely efficient and selective metal-binding molecules. To evaluate the protein-based radioisotope purification strategy, we first attempted to recover 90Y3+ from 90Sr2+ sources using LanM. 90Y3+ is used for radioimmunotherapy and medical imaging (e.g., SPECT/CT, gamma, and PET) (, ). 90Sr2+ has a long half-life (28.9 years) and β-decays to 90Y3+ (2.66 days). At secular equilibrium, the ratio 90Sr2+/90Y3+ is ~4000 mol/mol. The 90Sr2+/90Y3+ pair can be used as a generator for the short-lived 90Y3+ but, despite a sufficient stockpile of 90Sr2+, its global deployment has been hindered by difficulties in separating these isotopes efficiently, rapidly, and inexpensively ().

Figure 2 (A to D) shows the LanM-driven separation obtained after a single elution of an aqueous solution at pH 7 containing a mixture of 90Sr2+/90Y3+, initially at secular equilibrium. Efficient and quantitative separation of the two isotopes was observed after a single step and without requiring any eluant variation, loading/washing/elution cycles, or highly acidic extraction conditions. Such separation takes <5 min and could be even faster if automated. The presence of 90Y3+ in the high–molecular weight fraction and 90Sr2+ in the low–molecular weight fraction was confirmed by monitoring the activity of the samples by liquid scintillation counting (LSC) (Fig. 2, B to D). The decay profiles of the protein fraction (Fig. 2C) and small-molecule fraction (fig. S2) match exactly those of pure 90Y3+ and 90Sr2+, respectively. The energy spectra of the protein and low–molecular weight fractions also match the fingerprints of the high-energy β emitter 90Y and low-energy β emitter 90Sr, respectively (Fig. 2D). After this one-step separation, the radiopurity of 90Y increases from 50 to 99.9% (fig. S3). The robustness of the LanM-based separation method and stability of the protein against radiolysis were also evaluated. No decrease in separation performance was observed even when exposing LanM to 90Sr/90Y at 5 μCi/ml for a period of 150 hours (fig. S4).

Fig. 2.

One-step extraction and purification of yttrium and actinium using LanM.

(A) Typical elution profile of a 90Sr/90Y solution, initially at secular equilibrium, in the presence of LanM. [LanM] = 50 μM. Activity = 37 kBq/ml (1 μCi/ml). Buffer: 90 mM NaCl and 10 mM Hepes (pH 7.0). (B and C) Activity profile monitored at regular intervals, confirming the presence of high-purity 90Y in the LanM fraction and quantitative separation 90Sr/90Y. (D) Energy spectra, measured by liquid scintillation, for the starting 90Sr/90Y solution (top), the low–molecular weight fraction (middle), and LanM fraction (bottom). Spectra are normalized to the count maxima. (E) Typical elution profile of a 228Ra/228Ac solution, initially at secular equilibrium, in the presence of LanM. [LanM] = 25 μM. Buffer: 90 mM NaCl and 10 mM Hepes (pH 7.0). Note that pure 228Ac inherently gives some counts in the low-energy window (green curve) due to the continuous nature of the beta emission. (F and G) Activity profile monitored at regular intervals, confirming the presence of high-purity 228Ac in the LanM fraction (error bars or marker size = 2σ). (H) Typical energy spectra of a solution containing the 232Th decay chain (top), the solution after hydroxide precipitation (middle), and the purified 228Ac fraction using the LanM-based method (bottom).

Similar attempts to separate 90Y3+ and 90Sr2+ using human apo-transferrin—a protein considered as the main f-element transporter in mammals (, , )—failed owing to the lower affinity and lack of selectivity of transferrin relative to LanM. Even under optimal conditions and at transferrin concentrations well above the level found in the bloodstream, less than 10% binding to 90Y3+ was observed (fig. S5), highlighting that LanM is much more effective and selective than previously known lanthanide-binding proteins.

The efficiency and selectivity of LanM for 90Y3+ chelation suggests that it could be leveraged to complex actinium, even for its very short-lived isotope, 228Ac3+. Commercially available 232Th4+ salts exhibit measurable activity of 228Ac3+ after a few months (fig. S6), albeit at extremely low concentration (table S1). Being able to selectively recover 228Ac3+ could make Ac chemistry readily available and independent from the cost-prohibitive and scarce 227/225Ac sources. The 232Th4+ decay chain comprises 228Ra2+ (5.7 years), which decays to 228Ac3+, then 228Th4+ (1.9 years) and 224Ra2+ (3.6 days) (fig. S1). Starting from a 232Th4+ nitrate solution, neutralization to pH 6 to 8 precipitates 232Th4+ and 228Th4+ as hydroxide and coprecipitates the initial 228Ac3+, leaving the two radium isotopes in solution. Without regeneration by 228Th4+, 224Ra2+ and its daughters naturally decay within ~20 days (fig. S7). 228Ra2+ regenerates 228Ac3+ within ~30 hours, due to secular equilibrium, so it represents a potential 228Ra2+/228Ac3+ generator. Our process is shown in detail in fig. S8.

Little information has been published on the recovery of 228Ac3+ from macroscopic amounts of 232Th4+, besides a recent liquid-liquid extraction method performed in concentrated HNO3 media (), and an ion-exchange protocol including >30 steps (fig. S9) (), which we deemed impractical for such a short-lived radioisotope. The liquid-liquid extraction method developed by Kazakov and co-workers () consists of a multistage process where a 232Th(IV) salt is first dissolved in 4 M HNO3, and then three consecutive liquid-liquid extraction batches are performed using the extractant di-(2-ethylhexyl)phosphoric acid and toluene as diluent (each batch necessitating three steps: mixing, centrifugation, and phase separation using a funnel) followed by two column chromatography separations and an evaporation to dryness and redissolution step (fig. S9A). The ion-exchange protocol developed by Aldrich et al. () is even more complicated, as it requires four ion-exchange resin separations and nine evaporation-redissolution operations (fig. S9B). Both methods also produce an acidic solution of 228Ac whose pH conditions must be further adjusted before use for biological applications. The variety and complexity of the operations involved in these conventional hydrometallurgical methods (liquid-liquid extraction in acidic media, evaporation to dryness…) preclude their implementation for a fast, convenient, and periodic harvesting of short-lived radioisotopes such as 228Ac. In addition, we experimentally observed that the activity of 228Ra2+/228Ac3+ solutions increases after ~28 days due to reformation of 228Th4+ from 228Ra2+—a phenomenon predicted by the Bateman equation but overlooked in the literature—rendering previously proposed 228Ra2+/228Ac3+ generators rapidly inoperable (fig. S7). By contrast, our LanM-based size exclusion method offers direct, rapid, and practical extraction and purification of 228Ac3+ (Fig. 2, E to H). In none of the conditions tested did we observe interactions between Ra2+ and LanM, yielding high-purity 228Ac(III)-LanM fractions, as confirmed by radioactive decay monitoring and spectral identification (liquid scintillation and gamma spectroscopy; Fig. 2H and figs. S3 and S10), even when starting from ~8200 equivalents of 228Ra2+ or 224Ra2+/228Ra2+/228Ac3+ mixtures (fig. S11).

We were able to prepare 228Ac3+ solutions with ≥99.5% radiopurity (99.9% for 90Y3+) on a daily basis and at low cost, paving the way for investigating Ac3+ chemistry. Similar to 90Y3+/90Sr2+, attempts to recover actinium using transferrin were unsuccessful (fig. S5). These results establish that LanM uniquely enables practical purification of valuable radioisotopes from generators, making 90Y3+ and 228Ac3+ more accessible for medical and chemical studies and potentially extendable to other isotopes.

LanM enables facile detection of radiometals even at ultratrace levels

To demonstrate the efficacy of the LanM-based extraction system and evaluate its limits, tests were performed with metal concentrations in the subnanomolar range, mimicking nuclear medicine conditions (). Single-step separations of 228Ra2+/228Ac3+ and 90Sr2+/90Y3+ mixtures were performed using just 1 μM LanM at pH 7 and without sample preconcentration. Even under these unfavorable conditions for metal binding, LanM enables the efficient and selective scavenging of 228Ac3+ and 90Y3+. The 90Y3+ recovery yield in the protein fraction was maintained at >75% for concentrations as low as ~45 fM, or just ~2 femtograms of yttrium (Fig. 3A). Below this level, uptake of 90Y3+ by LanM was still observed, albeit with a lower yield (50 to 60%). LanM has higher affinity for actinium, and it allows its selective recovery to a concentration as low as 0.3 fM—just 40 attograms under the studied conditions (Fig. 3B). This extremely low mass was only twice the instrumental quantification limit for 228Ac so that more dilute samples were not investigated, but it is likely that the proposed LanM-based system works at even lower Ac3+ levels (fig. S12).

Fig. 3.

Recovery of Y and Ac radioisotopes at ultralow levels.

(A) Activity and mass of 90Y recovered in the protein fraction, compared to the total initial amount, following a single-step elution of 90Sr/90Y samples at various concentrations. [LanM] = 1 μM. pH 7.0. (B) Similar experiments with 228Ra/228Ac samples. (C and D) Activity recovered in the protein fraction and radiolabeling yield following a single-step elution of 228Ra/228Ac samples as a function of the incubation period. [LanM] = 10 μM. Room temperature. Buffer: 10 mM Hepes and 90 mM NaCl (pH 7). (E) Radioactive decay curve of the protein fractions obtained following different incubation periods (5 to 120 min) and elution through a size exclusion column. The model radioactive decay for pure 228Ac (crosses) is plotted for comparison with the experimental data (circles). The error bars correspond to ±3σ.

Furthermore, the metal uptake reactions with LanM are fast. After only a 5-min incubation at room temperature and neutral pH, the Ac3+-LanM binding reaction is almost complete (Fig. 3, C to E). The results reinforce that the LanM-based approach could enable faster, more efficient, and low-cost purification of radioisotopes for medicine and research. It could also be adapted for the field-deployable analysis of radioisotopes, especially f-elements resulting from nuclear activities.

Solution thermodynamics of actinium-LanM

The combined preparation of 228Ac3+ solutions, efficient binding to LanM, and separation of the LanM complexes from small molecules afford a unique opportunity to probe the chemistry of actinium. While solution chemistry and crystallographic studies have been performed on f-element/protein species since the 1960s (, ), including with heavy actinides (, ), no actinium-protein complex has ever been characterized, even from a solution thermodynamic standpoint. This situation is due to the low availability of Ac isotopes, lack of Ac3+-specific spectroscopic features, and incompatibility of proteins with classic techniques used for radiotracers, like liquid-liquid extraction (, ). Even for small Ac3+ complexes, no stability constant has been experimentally determined since the 1970s, with the strongest species reported being [AcEDTA]− (log β11 = 14.2) (, ).

Here, we developed a combined radiochemical and size exclusion chromatography method relying on the natural dissociation of metal-chelator complexes at low concentration (Materials and Methods). To establish the method, the formation constant of 90Y3LanM (log β31) was determined to be 33.9 (±0.5) at pH 7, corresponding to an average dissociation constant (Kd) of 4.3 (±1.5) pM per binding site (Fig. 4, A and B, and table S2). This value is in excellent agreement with results obtained using nonradioactive yttrium [Kd,app(natY3LanM) = 4.1 ± 0.3 pM; fig. S13] in ligand-protein titrations followed by circular dichroism (CD), confirming the applicability of the radiometry/dilution-based method. CD cannot be applied to actinium, as each experiment would require milligram quantities of the element (>1000× the current world’s supply) and the stability constants between Ac3+ and the competing ligand, which are currently unknown (). By contrast, the radiometry/dilution-based experiments only require femtograms of the radioisotope and do not require a competing ligand. This approach revealed that LanM has very strong affinity for Ac3+, with a log β31 of 36.2 ± 0.5 for 228Ac3LanM at pH 7, equivalent to an average Kd of 865 fM per site (Fig. 4, C to F). We previously showed (, , ) that lanthanide ions form polymetallic complexes with LanM, Ln3LanM, at macroscopic concentrations (micromolar to millimolar range). However, LanM has four potential metal binding sites (), and it was unclear whether the fourth site could play a role in the binding of actinium. Combining the results obtained by the dilution-induced dissociation technique and speciation calculations confirms that the actinium-LanM complex is Ac3LanM (fig. S14). Ac3LanM not only is the first polynuclear actinium species to be studied but also represents the most stable actinide(III)-protein complex.

Fig. 4.

Stability of the actinium-LanM complex.

(A) Examples of elution profiles obtained with 90Y/90Sr at various concentrations of LanM. Buffer: 10 mM Hepes and 90 mM NaCl (pH 7.0), 22°C. Size exclusion column: Sephadex PD-10 (5 kDa MWCO). (B) Radioactivity detected in the low–molecular weight fraction (90Sr) and protein fraction (90Y) as a function of LanM concentration. (C) Examples of elution profiles obtained with 228Ac/228Ra at various concentrations of LanM. Conditions similar to the 90Y/90Sr experiments. (D) Radioactivity detected in the low–molecular weight fraction (228Ra) and protein fraction (228Ac) as a function of LanM concentration. (E) Comparison of the experimental results and speciation models comprising 228Ac3LanM with various stability constants, covering four orders of magnitude. The best fit of the experimental results is for log β31(228Ac3LanM) = 36.2. (F) Correlation between the Lewis acidity of the cations and the stability of their complexes with LanM. The average Kd per site is plotted against the first hydrolysis constant of the cation. The points for Ac3+ and Y3+ in magenta correspond to the radiochemical method; the points for La3+ and Y3+ in gold correspond to the CD method. Previous data (CD, pH 7.2) () are in blue. A similar correlation is given for transferrin in fig. S19.

CD experiments using La3+, a nonradioactive Ac3+ analog, yielded a Kd,app of 1.8 (±0.1) pM for natLa3LanM (fig. S13). The subpicomolar stability of Ac3LanM extends a trend previously noted for the early lanthanides (, ) where LanM displays higher affinity for cations with lower Lewis acidity (Fig. 4F). This reversed preference is remarkable when compared to small chelators () used in medicine, since they typically exhibit decreasing affinity () as the metal gets larger (fig. S15), leading to weak La3+ or Ac3+ complexes. LanM’s combination of high Ac3+ affinity, fast kinetics, and selectivity are unparalleled when compared to ligands currently used in clinical studies. For instance, 2,2′,2′′,2′′′-(1,4,7,10-Tetraazacyclododecane-1,4,7,10-tetrayl)tetraacetic acid (DOTA), which is largely seen as the gold standard for medical radiometals, exhibits slow kinetics and requires elevated temperatures (>70°C) (, ) to trigger metal chelation, hindering procedures involving biomolecules. From a thermodynamic perspective, DOTA and other small chelators are also orders of magnitude less selective than LanM for Ac3+ against cations naturally present in the bloodstream (fig. S16). Hence, contrary to what has been observed for small f-element complexes used in medicine (–), Ac3LanM is inherently less prone to in vivo dissociation via transmetallation reactions. Moreover, LanM’s multiple binding sites enable multi-isotope radiolabeling (verified here with 90Y3+/228Ac3+; fig. S17 to S18), which could allow theranostic strategies using a single chelator. The thermodynamic stability of Ac3LanM also makes it resistant against endogenous metal-binding proteins like transferrin, whose two binding sites can accommodate trivalent f-elements (, , ). While the stability of the Ac2Transferrin complex has never been measured, a literature review allowed its estimation based on an empirical correlation: log β(Ac2Transferrin) = 10.9 ± 2 or Kd = 3.5 μM (fig. S19). Despite the large uncertainty inherent to the latter estimation, LanM’s affinity for actinium is clearly several orders of magnitude higher than that of transferrin, and transferrin is unable to bind to Ac3+ at low concentrations (fig. S12), which is in line with experimental observations. The results suggest that LanM could be the basis for new actinium pharmaceuticals, if sufficiently kinetically inert.

Probing actinium chemistry using LanM

The solution behavior of 228Ac3LanM and 90Y3LanM was also evaluated. Size exclusion chromatography assays showed that both complexes are resistant to acidic conditions, with LanM releasing its radioisotopes only below pH ~2.8 (Fig. 5, A and B). Fitting of the binding curves as a function of pH using the Hill equation yields a high Hill coefficient (18.2 for Ac and 17.6 for Y). The sharp release of the metal upon acidification further supports the binding of Ac3+ (and Y3+) to the three binding sites of LanM. Following selective binding to LanM, both radiometals and protein can be recovered for further use by either lowering pH or adding a chelator in excess; we prepared HCl, HNO3, and EDTA and diethylenetriamine pentaacetic acid (DTPA) solutions of 228Ac3+, with quantitative yield via one-step spin filtration from 228Ac3LanM (fig. S20).

Fig. 5.

Solution chemistry of actinium in the presence of LanM.

(A) Activity detected in the low–molecular weight fraction as a function of pH, after filtration (20 min, 12,000g, VivaSpin, 3 kDa) of 228Ra/228Ac solutions, initially at secular equilibrium. 22°C. I = 0.1 M (NaCl). The dotted line corresponds to the fit using the Hill equation (n = Hill coefficient). (B) Same experiments with 90Y. (C) 228Ac or 90Y recovered in the LanM fraction in the presence of Ca2+, Mg2+, Mn2+, Zn2+, and Cu2+. The x-axis value is the concentration of each metal. PD-10 column, 20 μM LanM, 10 mM Hepes, and 90 mM NaCl (pH 7.4). (D) Percentage of 228Ac and 90Y bound to LanM in the presence of carbonate ions. pH 7.4, 10 mM Hepes, 90 mM NaCl, and 0 to 100 mM NaHCO3. (E and F) Similar experiments in selenite and phosphate media. (G) Activity partition in the presence of biorelevant anions. Twenty micromolar LanM (pH 7.4), 90 mM NaCl, 10 mM Hepes, and 25 mM biorelevant anions. 228Ac and 90Y experiments were done independently. Initial samples were at secular equilibrium. A perfect separation would yield 50% in each fraction. (H) Stability of La3+-bound complex of LaMP1 Förster resonance energy transfer (FRET) sensor in 50% human serum. Two micromolar LaMP1 and 4 μM La, 37°C. The raw FRET ratios are also given in fig. S21.

Extending our observation that neither 228Ra2+ or 90Sr2+ interact substantially with LanM, LanM’s chelation of Ac3+ and Y3+ is also not affected by the presence of cations that are ubiquitous in vivo and that interact with endogenous proteins and small chelators: Ca2+, Mg2+, Mn2+, Zn2+, and Cu2+. The use of short-lived 228Ac3+ and 90Y3+ allowed testing LanM’s selectivity with mixtures containing up to 3 × 1011 mol equivalents of each competing cation (Fig. 5C). Even under these extreme conditions, selective uptake of Ac3+ by LanM was still observed. For 90Y3LanM, a slight decrease in radiolabeling yield occurs above a billion-fold excess of divalent cations, corroborating the higher stability determined for 228Ac3LanM versus 90Y3LanM. Chelators currently used in nuclear medicine are unable to sustain such an excess of competing ions (fig. S16).

Differences were observed between Ac3+ and Y3+ relative to their interactions with small anions present in the bloodstream (Fig. 5, D to H). A high concentration of sulfate, acetate, or glycine does not destabilize 228Ac3LanM or 90Y3LanM. However, at high concentration, carbonate and selenite ions compete with LanM and decrease its radiolabeling efficiency for 90Y, whereas Ac3LanM is unaffected. In the case of phosphate ions, 228Ac3LanM was found to be more sensitive than 90Y3LanM. The observed trends (PO43− > CO32− ≈ SeO32− for actinium versus CO32− > SeO32− > PO43− for yttrium) shed light on subtle chemical differences between actinium and rare earth elements that could be leveraged for separation purposes.

Whereas the direct interactions between Ac3+ and carbonate, selenite, or phosphate ions have never been probed under biorelevant conditions, our results indicate that Ac3+ exhibits stronger interactions with softer oxygen donors. In line with our experimental results, recent density functional theory calculations () showed that replacing DOTA’s carboxylate groups by phosphonate groups would impart enhanced stability for the Ac3+ complex. However, the design of new actinium complexants should also take into account the selectivity against natural cations, a challenge for most small molecules. Considering LanM’s advantages relative to these chelators (e.g., fast kinetics, high selectivity and affinity, and pH stability), we posit that the introduction of softer residues in LanM could lead to even more suitable Ac3+ macrochelators. Last, challenging 2 μM La3+-LaMP1, a LanM-based fluorescent sensor (Kd,app = 9.4 pM) (), with 50% human serum at 37°C reveals ~5% loss of fluorescence response over 24 hours, indicating only minimal metal dissociation (Fig. 5H and fig. S21). Given that Ac3LanM is ~10-fold tighter than La3LaMP1, these results suggest that Ac3+-LanM may be sufficiently stable for in vivo applications.

DISCUSSION

This study represents the first characterization of an actinium-protein complex, 228Ac3LanM, as well as an experimental study of coordination preferences of Ac3+ that previously have largely been probed computationally. Our LanM-based separation strategy allows the efficient, practical, and concomitant chelation/purification of actinium (and 90Y3+) down to the femtomolar level. The simplicity of our approach, starting from abundant and relatively available starting materials and requiring minimal steps, promises to make 228Ac3+ more available for fundamental research and chelator development. Furthermore, the short half-life of 228Ac relative to other actinium isotopes means that 228Ac is the most stringent test for the affinity and selectivity of the LanM-based system, such that this approach should be extendable to purification of other actinium isotopes, such as the medically valuable 225Ac3+ from 229Th4+ or 232Th4+. In addition, this biomacromolecular strategy is likely translatable for production of other medically relevant radiometals such as 44/47Sc3+, purification and detection of strategic actinides such as 241Am3+ and 244Cm3+, and study of other elusive elements, notably, Es3+, Fm3+, and Md3+, of which several picograms can be produced but which have remained inaccessible with current methodologies. Last, the stability of LanM’s complex with actinium is similar (in fact slightly tighter) to those of its native substrates (lanthanides) and in the same range as siderophores considered to play a role in the mobilization of actinides and radiolanthanides in the environment (). Therefore, our results raise questions about the influence of natural proteins such as LanM on the binding and transport of natural or anthropogenic radionuclides in the environment. Just as LanM has helped to incorporate the lanthanides into the bioinorganic periodic table, it promises to facilitate understanding of the interactions between actinides and biological systems, as well as the development of the chemistry of actinium and other rare isotopes for both fundamental research and applications in medicine.

MATERIALS AND METHODS

Caution! 90Sr and 232Th, and concentrates of their decay products (228Ra, 228Ac, 228Th, etc.), constitute serious health hazards because of their radioactive and chemical properties. Radiochemical experiments were conducted at Lawrence Livermore National Laboratory in laboratories designed for the safe handling of short-lived and long-lived radioactive materials and associated waste.

Materials

Starting 90Sr2+/90Y3+ solutions were prepared by dilution of a primary standardized stock (Eckert & Ziegler, USA). Starting 232Th4+ solutions were prepared by direct dissolution of Th(NO3)4·nH2O salts purchased from SPEX Industries (USA) and Strem Chemicals (USA). The age of the different 232Th salts varied from ~0.5 to 20 years. Liquid scintillation and gamma analysis confirmed the presence of decay products. All aqueous solutions were prepared using deionized water purified by reverse osmosis. Buffers (VWR), nonradioactive chemicals (MilliporeSigma), PD-10 size exclusion columns (GE Healthcare), and Vivaspin protein concentrators (Cytiva) were purchased and used as received. Wild-type LanM and LaMP1 were expressed and purified as previously described (, , ).

Radioanalysis

LSC was performed with a PerkinElmer Tri-Carb 5110TR instrument equipped with an alpha/beta discriminator. Calibration was performed on a daily basis. Samples were diluted in Ultima Gold, and results are background-corrected. Error bars displayed on the decay profile correspond to ±2σ unless otherwise indicated. Gamma analyses were performed using high-purity germanium coaxial p-type semiconductor detectors, each rated at 21 and 37% standard efficiency relative to 3 × 3 NaI. Energy and efficiency calibrations were performed with National Institute of Standards and Technology traceable standards. Samples were counted as ~10-ml solutions in plastic vials (10 cm2 cross-sectional area) at distances of 4 to 5 cm from the face of the detector endcaps for up to 6 hours. Spectra were processed using GAMANAL ().

Size exclusion separations

PD-10 columns prepacked with Sephadex G-25 media [8.3 ml, 5 kDa molecular weight cutoff (MWCO)] and Vivaspin 500 concentrators (membrane with 3 kDa MWCO) were preequilibrated with five bed volumes of buffer, as per the manufacturers’ manuals. Sample volumes ranged from 0.4 to 3.0 ml. For thermodynamic measurements, samples were equilibrated at least 1 hour before elution. For PD-10 columns, the sample was injected, followed by addition of buffer (~12 ml), and eluted by gravity. The elution lasts 5 to 10 min. The macromolecules elute at ~3 ml, and the small molecules elute at ~6 to 9 ml. Individual fractions of 0.3 to 0.8 ml were collected and weighed for mass balance (0.1 mg accuracy). For spin filters, samples were centrifuged at 12,000g for 20 min. Each fraction was radioanalyzed immediately after separation and monitored over time as needed. See below for radiopurity calculations.

Protein Kd determination by a radiometric- and dilution-based method

Solution thermodynamics stipulate that metal-ligand complexes dissociate at low concentrationsIn the presence of a large excess of ligand, [Ligand]free ≈ [Ligand]total

Size exclusion separations (pH 7.0; 10 mM Hepes + 90 mM NaCl, 22 ± 0.5°C, PD-10) were performed with varying LanM and metal concentrations. The initial solutions of 228Ac3+/228Ra2+ were aged 28 days to avoid any contamination with 224Ra2+ and its daughters. At least 15 independent column separations were performed for both 90Y3+/90Sr2+ and 228Ac3+/228Ra2+ (12 to 15 fractions per column). Each fraction was measured over time by LSC to confirm the radionuclide identity. 90Sr2+ and 228Ra2+ were used as internal tracers. Each column separation allows determination of the fraction of bound and unbound Y3+ (table S2) or Ac3+ (Fig. 4E). As the metal and/or protein concentrations decrease, lower radiolabeling yields are observed for Y3+ and Ac3+, albeit at different levels. Knowing the metal partition and the total concentrations, each column experiment allows evaluation of the formation of the metal-protein complex. Speciation calculations were performed with the computer program Hyperquad Simulation and Speciation (HySS) () and taking into account the metal hydrolysis constants (). Under the studied conditions, the metal hydroxide species account for <1% for Ac3+ and <15% for Y3+. The reported log β(M3LanM) values are the average of at least 15 column experiments (±1σ).

Protein Kd determination by CD

The conformational response of LanM in the presence of EGTA-buffered La3+ and ethylenediamine N,N′-disuccinic acid (EDDS)–buffered Y3+ solutions was measured using CD as previously described ().

Stability of La(III)-LaMP1 in serum

A solution of 10 mM LaCl3 was added to 140 μM LaMP1 to a final metal:protein ratio of 2:1 and incubated at room temperature for 10 min. Buffer A [20 mM Mops-KOH and 140 mM NaCl (pH 7.4), sterile filtered] and sterile-filtered human serum were individually incubated at 37°C. La(III)-LaMP1 or apo-LaMP1 was added to prewarmed buffer A to 4 μM, mixed 1:1 with serum, and incubated at 37°C. At each time point, 100 μl of each reaction mixture was placed in a Greiner 96-well μCLEAR half-area plate and assayed on a BioTek Synergy H1 plate reader with 433-nm excitation and 460- to 550-nm emission (1-nm steps). FRET ratios were calculated by taking the ratio of an average of three wavelengths at the peak of each fluorophore emission [474 to 476 nm for enhanced cyan fluorescent protein (ECFP) and 528 to 530 nm for enhanced yellow fluorescent protein (EYFP)].

Radiopurity calculations

The purity of the 228Ac and 90Y fractions was evaluated on the basis of the decay profile determined by LSC, using the total counts (0 to 2000 keV), to take into account all the isotopes eventually present. Decay curves contained 25 to 65 time points, each corresponding to an LSC measurement and individual background. Decay curves were fitted with the following function

Since 90Sr and 228Ra have far longer half-lives than 90Y and 228Ac, their activity remains constant and can be determined from the residual activity after the decay of initial 90Y and 228Ac, respectively. Because of the secular equilibria 90Sr/90Y and 228Ra/228Ac, the residual activity corresponds to twice that of 90Sr or 228Ra, if present. For example

The fitted values of Ainitial and Aresidual are directly linked to the purity of initial sample.