The ClickZip principle.
Fast complexation and extremely high inertness are desirable but contradictory attributes to be achieved with conventional chelators. To circumvent this problem, we devised ClickZip as an unconventional molecular mechanism to irreversibly trap LnIII ions inside a coordination cage (Fig. 1A). A macrocyclic chelator was equipped with azide and alkyne moieties strategically placed on two opposing pyridine pendant arms. Although uncatalysed azide-alkyne cycloaddition is thermodynamically favoured, it is typically slow and requires rather forcing conditions32. The chelator thus remains open to readily accept LnIII ions for complexation. Upon complexation, the coordination of the pyridines sterically forces the azide-alkyne reaction, forming an intramolecular 1,5-triazole bridge locking the coordination cage. This reaction is notably different from copper(I)-catalysed click (the metal is not a catalyst—see further) and from the classical Huisgen cycloaddition (1,5-triazole is formed, compared to 1,4-/1,5-isomer mixture)32, and can be best described as a metal-templated Huisgen cycloaddition. Practically speaking, the ClickZip process runs as one-pot reaction in fully aqueous solution under moderate heating (80°C).
Note on abbreviated notation.
For brevity, the following notation will be used throughout the text for specific chemical species (see Supplementary Fig. 1): L1 (open chelator); RL1 (L1 derivatized with R); [M(RL1)] (open chelate of metal M with RL1); 1,4-cz-RL1 and 1,5-cz-RL1 (empty cages with 1,4- and 1,5-triazole bridge, respectively); 1,4-cz-[M(RL1)] and 1,5-cz-[M(RL1)] (ClickZip chelates with 1,4- and 1,5-triazole bridge, respectively). Unless important for isomer distinction, the most discussed 1,5-cz-[M(RL1)] is further abbreviated to R{M}.
Metal ion role and preferences.
Several observations point towards a templating rather than catalytic role of the LnIII ion in the ClickZip reaction. Firstly, the ClickZip rates and yields strongly depend on ionic radius, increasing from LaIII to LuIII (Fig. 1B-C), with YIII falling between DyIII and HoIII, confirming this trend. Secondly, using excess metal expediates the complexation, but the unchelated metal does not promote intermolecular triazole cross-linking. Thirdly, the purity, yield, and reaction rate of ClickZip are remarkably independent of concentration (5 µM–50 mM), with deviations notable only at the extreme limits (Supplementary Fig. 2). Overall, these results indicate that the chelated metal ion exerts indirect steric effects through coordination of the pyridines rather than participating directly in the azide-alkyne cycloaddition.
1,4-/1,5-Triazole selectivity.
The surprising regioselectivity of ClickZip towards 1,5-triazole products regardless of the LnIII choice and ligand derivatizations (Supplementary Fig. 3) prompted investigation with computational chemistry methods. The largest LaIII, smallest LuIII, and selected non-lanthanides (CaII, LiI, NaI, KI) were compared in terms of the calculated Gibbs free energies of their open intermediate [M(L1)] and bridged 1,5-cz-[M(L1)] and 1,4-cz-[M(L1)] products. For all these metals there was a clear thermodynamic drive to both products, as expected for Huisgen cycloaddition. The 1,5-isomer was favoured in all cases, except for the large LaIII and KI ions (Extended Data Fig. 2). However, experimental data proved that even LaIII provided exclusively the 1,5-isomer (Fig. 1). This discrepancy was explained by considering the reaction mechanism and kinetics. The transition state leading to the 1,5-triazole product was significantly lower in energy and therefore kinetically preferred in both LaIII and LuIII, in agreement with the experimental results. The reason seems to be partial de-coordination of the pyridines required for both transition states, which is more pronounced and energetically demanding for the 1,4-isomer (Extended Data Fig. 2). The peculiar case of alkaline metals will be discussed next.
Empty cage synthesis and conventional complexation.
To compare the ClickZip synthesis with conventional complexation, it was necessary to prepare the empty cage 1,5-cz-PhL1. Interestingly, in the absence of LnIII ions, the intramolecular azide-alkyne reaction of PhL1 yielded a mixture of species dependent on pH and buffer (Extended Data Fig. 3). The 1,5-cz-PhL1 was the major product under acidic pH or alkaline pH in presence of NaI ions. In the absence of NaI (using KI, RbI or CsI in the buffer), alkaline pH resulted in dominant formation of the 1,4-cz-PhL1 isomer not observed with LnIII ions. In contrast, LiI ions suppressed formation of both products, largely preserving the open ligand PhL1. Thus, both empty-cage isomers could be obtained in high yields under specific optimized conditions, confirmed by X-ray analysis (Extended Data Fig. 4, Supplementary Fig. 5). These results indicate that complexation of a size-matched metal ion (here NaI) or a specific degree of protonization (intramolecular hydrogen bonds) have similar templating effects towards 1,5-triazole formation, while the absence of metal templating may provide the other isomer.
Direct complexation of LnIII ions with 1,5-cz-PhL1 was unsuccessful, with no Ph{Ln} product detectable after heating to 80°C for 1 week (Extended Data Fig. 5) or 6 months (Supplementary Fig. 6). Instead, formation of Ph{Ca} was observed, detected previously in trace amounts in reactions with LnIII ions (Fig. 1) and synthesis of empty cages (Extended Data Fig. 3), likely due to CaII ions leaching from the glassware. This is in stark contrast to the 1,4-cz-PhL1 isomer, which provided 1,4-cz-[Ln(PhL1)] chelates by direct complexation of LnIII ions under the same conditions, though with mediocre yields. Lanthanide chelates of both types thus could be accessed via different strategies (Supplementary Fig. 7).
Kinetic inertness.
The kinetic inertness of the chelates was tested by acid-assisted dechelation under pseudo-first-order conditions with excess HCl, quantitatively monitored by LC-MS and expressed as half-lives. Chelates of a DOTA derivative [Ln(NO2BnDOTA)], amenable to LC-MS detection, served as a reference33. Four increasingly demanding conditions from 0.1 M HCl at 25°C to 6.0 M HCl at 80°C were used to cover a broad range of half-lives. This revealed a remarkable increase in inertness across the series of Ph{Ln} chelates, spanning 10 orders of magnitude from La to Lu (Fig. 2A). Starting from Sm, Ph{Ln} surpassed inertness of the DOTA system, steadily improving up to Lu (Supplementary Figs. 8–12). Ph{Lu} showed remarkable resistance to dechelation even under the harshest conditions (Fig. 2B). With an estimated half-life of 3 years in 6.0 M HCl at 80°C, it is the most inert lanthanide chelate reported to date27–29. In contrast, the isomeric 1,4-cz-[Ln(PhL1)] chelates were much less kinetically inert, approximately 2 orders of magnitude worse than the DOTA system.
Solid-state structures, isostructurality, isomerism.
The striking differences in properties between the 1,5-triazole- and 1,4-triazole-bridged chelates are best understood from their solid-state structures (Fig. 3). In the case of Ph{Lu}, the 1,5-triazole is part of an 18-membered ring, where all five donor N-atoms (three from cyclen, two from pyridines) can coordinate tightly to the LuIII ion. On the other hand, the 1,4-triazole in 1,4-cz-[Lu(PhL1)] chelate increases the size of this ring to 19 atoms, bringing steric strain and chain conformations that disfavour simultaneous coordination of both pyridines. This mismatch explains why the 1,4-cz-[Ln(PhL1)] chelates are not produced via the ClickZip reaction and are much less inert. The orientation of the triazole hydrogen relative to the coordination cage may also play a role (Fig. 3).
Solid-state structures for all Ph{Ln} chelates from Sm to Lu (including Y) revealed exceptional similarity of the molecules (Supplementary Fig. 13), despite the dramatic differences in kinetic inertness. Structural parameters showed very small relative changes in response to the lanthanide contraction; the coordination environments and overall shapes of the molecules remained the same. This isostructurality was corroborated by the behaviour of the chelates in reversed-phase chromatography, where a mixed sample of Ph{Ln} chelates (Ln = Sm to Lu, Y) showed a single peak with no sign of separation (Supplementary Fig. 14). Only the early lanthanides from La to Nd demonstrated notable deviations.
The ClickZip chelates are inherently chiral, with defined rotation of the pendant arms (Λ or Δ) and of the four ethylene units in the cyclen ring (λλλλ or δδδδ), analogous to the DOTA system15,27,33. However, in contrast to DOTA, they only adopt enantiomeric Λλλλλ and Δδδδδ forms with twisted-square antiprismatic (TSA) arrangements (Extended Data Fig. 6). To probe whether these enantiomers can interconvert, we modified {Lu} with L-cysteine residues to produce a pair of distinguishable diastereomers. These were chromatographically separated and their epimerization was followed by NMR, revealing that they interconvert within hours at 37°C (Supplementary Fig. 15), much slower than the DOTA chelates30. The enantiomers of {M} can thus be regarded either as one or two compounds, depending on the time scale.
Post ClickZip synthesis.
Their exceptional stability and synthetic accessibility make ClickZip chelates interesting substrates for further chemical transformations. A range of reactions could be performed on the organic ligand while keeping the LnIII ion safely locked inside (Extended Data Fig. 7). Heating Ph{Lu} with excess strong base (DBU) in D2O led to full deuteration of all five CH2 groups in the pendant arms by proton exchange. Extensive treatment with NaBH4 in MeOH quantitatively reduced one of the coordinated pyridines to a piperidine ring, which remained coordinated to the chelated metal. Functional groups exposed to the exterior could undergo efficient transformations including Suzuki coupling, nucleophilic aromatic substitutions, and copper(I)-catalysed click, showcasing the versatility of ClickZip chelates as synthetic building blocks.
Compatibility with peptide hydrolysis.
While the acid-assisted dechelation of lanthanide DOTA chelates proceeds with a rate similar to hydrolysis of a peptide bond24, the ultra-inert ClickZip chelates of LuIII and near lanthanides should outlast complete hydrolysis of peptides and proteins. To test this, we prepared model hexapeptide conjugates with two ClickZip derivatives, HO2CPh{Ln} or HO2CBn{Ln}, which differed in their amide linkage to the peptide N-end (Extended Data Fig. 8). Hydrolysis in 2 M HCl at 80°C digested the peptide to individual amino acids; the chelates were cleaved from the peptide with no loss of the metal (LuIII or TmIII), or only small loss (5% for YIII chelate). Faster cleavage of HO2CBn{Ln} shows that the tags can be optimized for this step.
Multiplexing in vitro.
For their resistance to acidic hydrolysis and isostructural character, ClickZip chelates are ideal tags for multiplexed bioanalysis. We selected cell internalization as a biological effect to test whether tags carrying different metals are distinguishable to living cells (Fig. 4A). Two cell-penetrating peptides (CPPs), based on hexaarginine and its partially protected derivative, were labelled with {Lu} and {Tm} tags at the N-termini, resulting in a total of four conjugates (Fig. 4B). Paired comparisons for compounds carrying different tags were performed simultaneously on the same sample of cells (Fig. 4C). Tags released from acidic digestion of the cells were quantified by LC-MS (Fig. 4D). As expected, the peptide with partially protected guanidine groups was internalized less efficiently than the naked hexaarginine. However, compounds with identical peptides but different tags were internalized into cells to a similar extent, regardless of their cell-penetrating efficiency. When compounds with different peptides were compared, permutating the tags provided a mirror image of the results. This demonstrates that the observed differences in the biological effect (internalization) were due to differences in the peptide part and the choice of metal had a negligible effect. The LC-MS quantification was verified by ICP-OES (Fig. 4E).
Ex-vivo quantification in animal tissue.
Compared to in-vitro cell cultures, quantification of mass tags in animal tissue must deal with much more complex matrices of interferences. To demonstrate that ClickZip chelates are compatible with such conditions, we used two 31-amino acid prolactin-releasing peptide (PrRP31) analogues previously studied for their anti-obesity effect34, which differ in the presence of a biodistribution-altering fatty acid residue: native PrRP31 or PrRP31 palmitoylated at position 11 (palm11-PrRP31). The two peptides were labelled with derivatized {Lu} and {Tm} tags (Fig. 5A) and intravenously (i.v.) administered to C57BL/6J mice independently or in a mixture (Fig. 5B). Liver tissue was selected for ex-vivo analysis, as it has the highest natural lanthanide background of all organs, presenting a particularly difficult matrix12. Although pre-purification of the tags was possible, we opted for analysis of the full lysate, removing only insolubles (Supplementary Figs. 19–20). Both tags were confidently quantified on an LC-MS/MS system primarily used for proteomics (Fig. 5C). However, independent verification with elemental analysis on ICP-MS provided a different view. The levels of Lu and Tm were higher than expected, even in samples to which they were not applied (Fig. 5D). Here, the ability to distinguish lanthanides in different chemical forms proved essential. Using liquid chromatography prior to ICP-MS detection, the {Ln} tags could be discerned from the unchelated LnIII ions, confirming the same {Ln} content as determined by LC-MS/MS (Fig. 5E, Supplementary Fig. 24). Because other lanthanides were found in the liver samples only at much lower levels, it was suspected that the unchelated Lu and Tm were contaminants, rather than biological background. The exact source could not be traced, but likely originated from environmental contamination in a laboratory that regularly works with lanthanides. In conventional trace-metal analysis, contamination may completely obscure the desired measurement. However, the {Ln} tags demonstrate a remarkable analytical robustness, allowing quantitative separation and subtraction of the unchelated lanthanides, regardless of their origin.
In vivo multiplexing capabilities.
To probe the degree of similarity between ClickZip tags under in vivo conditions, we labelled the lipidized peptide palm11-PrRP31 with {Tm}, {Yb}, or {Lu} tags, and followed them with ICP-MS for 60 minutes after i.v. co-administration to Wistar rats (Fig. 6A-B). The pharmacokinetics of all three conjugates in blood plasma were nearly identical, with the highest concentration observed immediately after injection, followed by rapid clearance35 (Fig. 6C). The three lanthanides were further determined in two liver lobes (caudate and left lateral) after the animals were sacrificed. The same amounts were found in both lobes and the relative ratios of the metals were not significantly different (Fig. 6D). Thus, varying the LnIII ion in ClickZip tags has a negligible effect on both pharmacokinetics and biodistribution of tagged biomolecules, allowing reliable tracking and quantification of multiple labelled molecules simultaneously in vivo.