To chelate thallium( I ) – synthesis and evaluation of Krypto ﬁ x-based chelators for 201 Tl

While best known for its toxic properties, thallium has also been explored for applications in nuclear diag-nostics and medicine. Indeed, [ 201 Tl]TlCl has been used extensively for nuclear imaging in the past before it was superceded by other radionuclides such as 99m Tc. One reason for this loss of interest is the severe lack of suitable organic chelators able to e ﬀ ectively coordinate ionic forms of Tl and deliver it to speci ﬁ c diseased tissue by means of attached biological vectors. Herein, we describe the synthesis and characterisation of a series of Krypto ﬁ x 222-based chelators that can be radiolabelled with 201 Tl( I ) in high radiochemical yields at ambient temperature. We demonstrate that from these simple chelators, targeted derivatives are readily accessible and describe the synthesis and preliminary biological evaluation of a PSMA-targeted 201 Tl-labelled Krypto ﬁ x 222-peptide conjugate. While the Krypto ﬁ x system is demonstra-bly capable of binding the thallium cation, no PSMA-mediated cell-uptake could be detected with the PSMA conjugate, suggesting that this targeting moiety may not be ideal for use in conjunction with 201 Tl.


Introduction
Thallium is most commonly known for its detrimental properties such as its former use in rat poison or as a popular means to murder in both fiction and non-fiction. 1 However in academic and medical settings, thallium, and in particular its radioactive isotopes have been recognised for their potential utility in healthcare. In 1975 Lebowitz et al. discussed the possible application of thallium-201 for myocardial imaging due to the similarities in size between Tl(I) and K(I). 2 Thallium-201 ( 201 Tl) has a half-life of 73 hours and emits γ-photons at 72 keV (X-ray, 94% abundance), 135 keV (3% abundance) and 167 keV (10% abundance) making it suitable for application in single photo emission computed tomography (SPECT). As predicted, 201 Tl initially found widespread use as a myocardial imaging agent until it was replaced by emerging technetium-99 m tracers. 3 In a few instances [ 201 Tl]TlCl has also been employed for tumour detection and monitoring. 4 Additionally, 201 Tl decays through electron capture (EC) to 201 Hg, emitting an average of 36.9 Meitner-Auger electrons per decay event. Meitner-Auger electron-emitting agents have shown significant potential for application in systemic, targeted radionuclide therapy of cancer. 5 With its high decay yield of Meitner-Auger electrons, 201 Tl has promise as a radiotherapeutic isotope and possibly for theranostic applications. 6 The clinical application of 201 Tl is currently limited to its use as the "free" or "unchelated" metal ion. Many other radioactive metallic isotopes have demonstrated clinical utility but are largely used in a complexed or chelated form. For example, [ 68 Ga]Ga 3+ and [ 177 Lu]Lu 3+ can be complexed to chelators which are appended to peptides that target receptors expressed in cancer tissue. The ensuant 68 Ga and 177 Lu radiopharmaceuticals are used for PET imaging and systemic radiotherapy respectively, and have had clinical impact in management and treatment of neuroendrocrine cancer and prostate cancer. 7 Advancements in the development of new 201 Tl-based radiopharmaceuticals are currently prevented by an almost complete lack of studies into the formation of stable compounds between thallium and organic chelators that could feature biological targeting functions. Thus far, only two studies have investigated the coordination of thallium to popular radioimaging chelators. In 2011, Hijnen et al. Unfortunately, the DTPA compound was found to rapidly decay in human blood serum. Although the DOTA complex was found to be stable in serum, the biodistribution of the compound mirrored that of non-chelated or "free" [ 201 Tl]Tl(I), suggesting limited stability of [ 201 Tl]Tl(III)-DOTA in vivo. 8 Interestingly, Fodor et al. described the same complex as an "extraordinarily robust macrocyclic complex" in a 2015 report. 9 However no in vivo experiments were reported, suggesting that while [ nat Tl]Tl(III)-DOTA may possess favorable stability in vitro, the Tl(III) core is readily reduced in vivo, leading to rapid decomplexation.
We recently described a more robust protocol for the oxidation of commercially available [ 201 Tl]Tl(I)Cl to [ 201 Tl]Tl(III) Cl 3 . In this work the ability of the readily available chelators, DTPA, DOTA and EDTA, for binding [ 201 Tl]Tl(III) was investigated. Radiolabelling was possible with all three chelators as evidenced by iTLC, however only [ 201 Tl]Tl(III)-DOTA showed sufficient stability in human serum, ruling out the other two chelates for in vivo applications. 10 While the ability to chelate [ 201 Tl](III) with sufficient stability required for in vivo applications is promising (though yet to be fully realized), this approach will always be susceptible to in vivo reduction of [ 201 Tl]Tl(III) to [ 201 Tl]Tl(I), and likely subsequent metal dissociation, as has been indicated by the work of Hijnen. We therefore started to explore the possible chelation of "native" [ 201 Tl]Tl(I). This presents some challenges as the monovalent thallium is a very large ion with an ionic radius of 150 pm and, like the alkali metals, its solution chemistry is dominated by aquation. 11 We initially looked at two classes of compounds for which there was some evidence in the literature that suggested potential for radiolabelling [ 201 Tl]Tl(I), crown ethers and cryptophanes. Crown-ethers are well-documented to bind a wide range of cations. 12 However in our hands, we could not detect any binding of [ 201 Tl]Tl(I) by 18-crown-6, dibenzo-18-crown-6 and dicyclohexyl-18-crown-6 under radiolabelling conditions. The group of Brotin has reported an impressive series of different cryptophane derivatives over the last years, including extensive characterization of their ability to bind different cations, including non-radioactive thallium. 13 As the synthesis of these compounds represents a significant challenge, their group kindly provided us with two samples of cryptophanes with which to conduct preliminary studies. While we did observe the appearance of a new peak in the radio-HPLC upon reaction of Crypt1 with [ 201 Tl]Tl(I)Cl, this could only be achieved under strongly basic conditions (Fig. 1). Adjusting the pH to ∼7 or exposure to human serum resulted in the release of the thallium cation, hence we did not pursue these chelators any further.
Due to the similarities in ionic radii between K(I) and Tl(I) we turned our attention to 4,7,13,16,21,24-hexaoxa-1,10-diazabicyclo [8.8.8]hexacosane, more commonly known as Kryptofix 222. This compound was first reported by Sauvage et al. in 1969. In this first report the authors described the remarkable propensity of these compounds to form complexes with "metallic cations". 14 Since then, this compound class has found applications in many different fields ranging from crystallography to chemical sensing. 15 Interestingly, its ability to encapsulate potassium ions led to its use as a phase-transfer catalyst for the radiosynthesis of [ 18 F]-fluoro2-deoxy-2-Dglucose ([ 18 F]FDG). 16 Based on these findings we have developed a series of Kryptofix 222 derivatives that can be conjugated to a biological targeting function of choice. Herein we report the synthesis and characterization of these derivatives as well as the preparation of a prostate-cancer targeted analogue. Furthermore, the ability of the Kryptofix compound class to coordinate both non-radioactive and radioactive [ 201 Tl] Tl(I) has been investigated.

Results and discussion
Before embarking on the synthesis of functionalized derivatives, the ability of commercially available Kryptofix 222 (K222) and Kryptofix 22 (K22, Fig. 2) to complex [ nat Tl]Tl(I) was investigated by NMR spectroscopy. Addition of TlPF 6 in D 2 O to a solution of the cryptand in D 2 O resulted in a clear shift in resonances for K222 but only a very small shift for K22 (ESI Fig. S1 and S2 †). A proof-of-concept radiolabelling study was conducted with K222. Incubation of the cryptand with [ 201 Tl]TlCl in saline led to almost quantitative formation of a new peak with a significantly higher retention time compared to free [ 201 Tl]Tl(I) on reverse-phase radioHPLC (ESI Fig. S3 †). Encouraged by these preliminary data we pursued the synthesis of bifunctional K222 chelators. We focused on K222-NH 2 as this compound could be conjugated to carboxylic acids via amide-coupling, whilst also offering the possibility to convert it to other synthons for "click"-type coupling reactions i.e. K222-N 3 and K222-DBCO (Fig. 2).
The synthetic strategy towards K222-NH 2 was based on works by Gansow, 17 Pettit,18 and Allen 19 adjusted with our own optimisations (Scheme 1). Compound 1 was obtained by reacting 4-nitrocatechol with ethyl bromoacetate in acetone in the presence of sodium carbonate. The reaction could be shortened to 60 min if conducted in a microwave reactor at 180°C. Hydrolysis to 2 with excess LiOH was straightforward and the chlorination to 3 with thionyl chloride could be achieved after optimization i.e. it was important to keep the reaction refluxing overnight and to directly use the crude product after evap- oration of thionyl chloride. For the preparation of 4 both reagents (3 and Kryptofix 22) were dissolved in fresh anhydrous toluene in separate round-bottom flasks and sonicated until dissolved. Both were added via syringe to a third roundbottom flask filled with more anhydrous toluene cooled by means of an ice bath. The speed of addition was not important for this reaction as both slow addition over 1 h by means of a syringe pump and fast addition over 1 minute gave roughly equivalent yields. The reaction was complete after 1 h at room temperature and did not require overnight stirring as described in some literature reports. 19 Compound 5 was obtained by refluxing 4 in BH 3 ·THF overnight and was used without further purification. Reduction of the nitro group of 5 was achieved by using fine zinc dust instead of powder, giving K222-NH 2 after purification on a C18 cartridge. FSO 2 N 3 was prepared as described in the literature 20 and reacted with K222-NH 2 to form K222-N 3 after purification by column chromatography on silica. NHS-ester activated dibenzocyclooctyne (DBCO-NHS) was reacted with K222-NH 2 , forming K222-DBCO as evidenced by LC-MS. This compound was then utilised further without purification.
As a model targeting function, we chose peptidomimetic 6 which has a high affinity for the prostate specific membrane antigen (PSMA), a receptor that is highly overexpressed in prostate cancer cells and for which many cell assays and murine models have been developed over the last decades. 21 The PSMA-targeting precursor 6 was prepared as described elsewhere. 22 Utilizing FSO 2 N 3 , the terminal amine on 6 was transformed into the azide 7 (Scheme 2). Deprotection with TFA cleaved off all tert-butyl protecting groups yielding 8 which could be reacted with K222-DBCO under copper-free click-conditions forming the final conjugate K222-PSMA which was purified by preparative HPLC. The quantity of K222-PSMA obtained was not sufficient for NMR assignment. The compound was characterized by 1 H NMR, LC-MS and HR-MS (ESI †).
While this project was ongoing, another group reported on the synthesis and application of K222 derivatives as chelators for 203 Pb, an emerging radiometal for SPECT applications. 23 While the synthetic derivatives reported by McDonagh et al. are slightly different to the ones reported here, their general findings that successful radiolabelling is possible with this class of compounds and that the resulting Pb(II) complexes possess suitable stability in human blood serum, encouraged us to proceed with the investigation of the thallium-binding ability of our compounds.
With all target compounds in hand we proceeded to investigate the ability of these derivatives to bind Tl(I). First, we repeated the NMR experiments conducted with K222 and K22 with the new derivative 5. Compound 5 was dissolved in CD 3 OD and an initial 1 H NMR was recorded. TlPF 6 (1.5 equivalents) was pre-dissolved in CD 3 OD, added to the solution of 5 and a second proton spectrum was recorded immediately after (Fig. 3). A clear change in the chemical shifts pertaining to the alkyl-protons of the K222 scaffold can be observed in the 1 H NMR spectra. On the other hand, the three aromatic protons show little to no change. These experimental findings were supplemented with density functional theory (DFT) calculations performed on both the Tl(I) as well as the K(I) complexes with ligand 5. For reference, DFT calculations were also performed for the already reported complex of K222 with potassium (ESI Fig. S19 †). Comparison of the oxygen-and nitrogen-metal bonds showed good agreement between the calculated and experimentally determined structures. The structures for Tl(5) + (Fig. 4) and K(5) + are similar, with all 6 oxygens and 2 nitrogen atoms of the macrocycle involved in interactions with the metal centre (ESI Fig. S20 †).

View Article Online
Slightly longer distances between putative coordinating atoms and metal centre were detected for the calculated Tl(I) complex structure compared to the K(I) analogue (ESI Table S2 †). Comparison of the Tl(I) with the K(I) coordination environments revealed that the K(I) ion sits more centrally in the cryptand scaffold whereas in the case of Tl(I), the ion is located near the plane formed by the four non-aromatic oxygen and nitrogen atoms. These calculations suggest that the binding mode of K(I) and Tl(I) are similar with regards to 5 but that in the case of Tl(I), the metal ion is less encapsulated by the cryptand which could indicate lower stability. Altogether, these experimental and computational studies strongly suggest that an interaction between the chelator and Tl(I) is occurring and that it is, as expected, between the oxygen and nitrogen atoms of the K222 cryptand and the thallium cation. As additional confirmation, the NMR solution was analysed by ESI-MS and the mass peak for a thallium ion bound by 5 was detected (ESI Fig. S18 †).
All initial radiolabelling studies were conducted with compound 5 as it was available in larger amounts than the other derivatives. Initial attempts at radiolabelling 5 with [ 201 Tl](I) in water were unsuccessful. Reactions were monitored by analytical radioHPLC utilizing a C18 column and a 0-100% water/ acetonitrile gradient with 0.1% TFA. During the synthesis of the K222 derivatives, the last step involves purification on reverse phase columns and through trial and error it was found that neutralisation of the collected fractions via addition of 1 M NaOH was necessary, otherwise no chelation of TlPF 6 could be observed by NMR. We hypothesised that this is due to protonation of the nitrogen atom in the macrocycle ring, as supported by DFT calculations. Visualisation of the HOMO for 5(Tl) revealed overlap of nitrogen and Tl-based orbitals, originating from the p and d orbitals, respectively (ESI Fig. S19, † left image). This interaction was no longer observed in the HOMO upon protonation of the nitrogen (ESI Fig. S19, † right image), and is reflected in the N-Tl distances where an elongation from 3.04 Å to 3.83 Å was observed upon protonation. In the absence of this stabilisation from N-Tl interactions, Tl(I) was found to reside further away from the aromatic oxygen donors, resulting in a noticeably less enveloped structure. Based on these findings we assume that concentration of collected fractions in the presence of small amounts of HPLC mobile phase acid additives (TFA or formic acid) resulted in the formation of a protonated cryptand which is unable to coordinate [ nat Tl]Tl(I).
With this knowledge in mind, further [ 201 Tl]Tl(I) radiolabelling reactions were conducted in water with the addition of small amounts of 1 M Na 2 CO 3 (final pH = 11). As no new peaks could be detected by radioHPLC, we hypothesized that the acidic mobile phase resulted in decomplexation of 5 from Unfortunately, under these conditions, we could not obtain reproducible radio-HPLC traces. We attribute this to adsorption of hydrophilic [ 201 Tl(5)] + to silica as in some repeated experiments, radioactive species were retained on the C18 column and did not elute. Such behaviour is sometimes observed in HPLC experiments using mobile phases without inclusion of a buffering or ion pairing reagent or acid. 24 Through further extensive experimentation it was found that addition of 0.01% formic acid to the HPLC mobile phase ( pH ∼ 4) enabled reproducible radio-HPLC chromatography. With the HPLC conditions optimised, radiosynthetic protocols for preparation of putative [ 201 Tl(5)] + were found to be robust: exclusive formation of [ 201 Tl(5)] + in >95% radiochemical yield was observed upon reaction of 5 with [ 201 Tl]Tl(I)Cl at pH = 11, in aqueous solution at room temperature after 15 minutes reaction time (Fig. 5 left).
Comparison of the UV-trace of 5 with the γ-trace of [ 201 Tl (5)] + shows a large difference between the retention times of the free ligand (8.0 min) and radioactive complex (4.2 min). The Tl(I) complex of 5 is significantly more hydrophilic than that of the free ligand, which we attribute in part to the single positive charge of [ 201 Tl(5)] + , as well as conformational differences between chelated 5 and non-chelated 5.
With the optimized [ 201 Tl]Tl(I) radiolabelling conditions determined for compound 5 we proceeded to the reaction of bioconjugate K222-PSMA with [ 201 Tl]Tl(I)Cl, which also led to the formation of a single new peak in the radioHPLC (Fig. 5  right). So far, all evidence pointed to K222-based chelators complexing Tl(I) under both non-radioactive (stoichiometric) and radioactive (non-stoichiometric tracer) conditions. A crucial benchmark for any radiochelator is not only the ability to bind a given radiometal but stably retain it under physiologically relevant conditions. A common assay to evaluate the stability of radiolabelled bioconjugates is to incubate them with human blood serum over time. After incubation, cells were washed and lysed, and cell lysates were collected and counted for radioactivity using a γ-counter ( Fig. 6 and ESI, Table S1 †).
Uptake of 201 Tl activity in DU145-PSMA+ cells measured 4.7 ± 0.2% AR ( percentage added radioactivity to 500 000 cells), however uptake in the PSMA-negative DU145 cell line similarly measured 4.9 ± 0.2% AR, and uptake in DU145-PSMA+ cells blocked with 2-PMPA measured 5.0 ± 0.1% AR. This suggested that uptake of 201 Tl radioactivity was not mediated by PSMA receptors. To further probe this, both DU145-PSMA+ and DU145 cells were treated with solutions of "free"/"unchelated" There are many endogenous biomolecules that are likely to compete with ligands such as K222 derivatives for binding to K(I)-mimics such as Tl(I). For example, potassium ion channels transport K(I) across cell membranes. These channels possess high affinity for K(I) to compensate for the energetic cost of dehydration/desolvation of K(I), which is required to enable  While these data with [ 201 Tl]Tl(I) are encouraging, we believe that other radiometals might be better suited to complexation with K222-based bioconjugates and therefore, we are exploring the application of K222-based compounds to 223 Ra, 213 Bi and 111 In.

Materials and methods
Commercially available reagent grade solvents and chemicals were used without further purification. Anhydrous solvents were acquired from solvent towers within the department and stored over 3 Å molecular sieves. 1 H, 13 C, COSY, TOCSY, HSQC, HMBC NMR spectra were recorded on a Bruker AVIII 400 or 700 spectrometer. Chemical shifts are reported in ppm and referenced to residual protonated impurities in the solvent for NMR spectra. High Resolution Electrospray Mass Spectrometry was carried out by Dr Lisa Haigh of the mass spectrometry service at Imperial College, or independently on an Agilent 6200 TOF LC-MS instrument. LC-MS for K222-PSMA was measured on a Waters 3100 Mass Detector equipped with a Waters 2998 Photodiode Array Detector, Waters 515 HPLC Pump, Waters 2545 Quarternary Gradient Module, and a Waters 2767 Sample Manager. LC-MS solvents were Millipore water and acetonitrile (with 0.1% formic acid). A 18 min gradient was run from 5% to 100% Acetonitrile on a XBridge BEH C18 Column, 130 Å, 5 µm, 4.6 mm × 100 mm. Flash chromatography used silica gel (60 Å pore size). Where specified, automated flash chromatography was performed using a Biotage Isolera Four unit and 10 g or 25 g SNAP KP-Sil/Sfar duo cartridge. HPLC was performed on an Agilent 1200 Series Liquid Chromatograph with UV and LabLogic Flow-Count detector with a sodium-iodide probe (B-FC-3200). Unless otherwise mentioned the mobile phase A contained H 2 O with 0.1% FA, and mobile phase B contained MeCN with 0.1% FA. Preparative reverse-phase HPLC was conducted using a Pursuit XRs C18 column (250 × 10 mm, 5 μm) and UV spectroscopic detection at 250 nm. 201 TlCl in saline was purchased from Curium Pharma, UK and used without further processing. Fresh human serum was obtained from a healthy volunteer.

Synthesis of K222 derivatives
Compound 1. Nitrocatechol (1.00 g, 6.45 mmol) was dissolved in acetone (30 mL). Ethyl bromoacetate (1.5 mL, 13.56 mmol) and K 2 CO 3 (4 g, 29 mmol) were added to the yellow solution. The reaction mixture was refluxed overnight (72°C, 16 h). The solution was then filtered and the filtrate was dried in vacuo. Automated column chromatography on silica (dry loading after dissolution in DCM; gradient 0-100% EtOAc : hexane) yielded 1 as a brown oil (0.96 g, 2.93 mmol, 45%). Instead of thermal heating the reaction can also be conducted via microwave reactor (400 mg scale in a 2-5 mL vial) at 120°C for 1 h. The NMR data matched that reported previously. 26   Compound 5. Compound 4 (110 mg, 0.22 mmol) was dissolved in 1 M borane THF solution (2 mL) and refluxed under a nitrogen atmosphere overnight. The mixture was then allowed to cool to room temperature before 3 M HCl (10 mL) was added and the solution was refluxed for another 3 h. After cooling to room temperature, the solution was neutralised by addition of NH 4 OH (15 mL). All solvent was removed in vacuo. The solid residue was re-dissolved in ACN/H 2 0 (1 : 1, 2 mL) and purified by automated column chromatography on a C18 cartridge (gradient 0-100% ACN : H 2 O with 0.1% TFA). The fractions containing product were neutralised by addition of 1 M NaOH and dried, yielding 5 as a reddish oil (58.0 mg, 0.12 mmol, 55%). The NMR data matched that reported previously. 19  K222-NH 2 . Compound 5 (98.7 mg, 0.21 mmol) was dissolved in a mixture of DCM (5 mL) and acetic acid (2 mL). Fine zinc dust (150 mg) was added to the solution and the mixture was stirred at room temperature for 3 h. The solution was diluted by addition of MeOH and water (1 : 1, 20 mL) and the DCM was evaporated. The zinc dust was filtered off and the filtrate was dried in vacuo. The crude product was purified by automated column chromatography on a C18 cartridge (gradient 0-100% ACN : H 2 O with 0.1% TFA). The fractions containing product were neutralized by addition of 1 M NaOH and dried, yielding K222-NH 2 as a brownish oil (41.5 mg, 0.094 mmol, 43%). The NMR data matched that reported previously. 19  K222-N 3 . The fluorosulfuryl azide reagent was prepared as described in a literature report. 20 K222-NH 2 (20 mg, 46 µmol) was dissolved in DMF and MTBE (1 : 1, 1 mL). Fluorosulfuryl azide (100 µL, 0.7 mM) and KHCO 3 (70 µL, 3 M) were added, and the solution was stirred overnight. LC-MS analysis revealed formation of the target compound. The solvent was evaporated and the crude product was purified by automated silica column chromatography (dry loading after dissolution in DCM/MeOH; gradient 0-10% MeOH : DCM) yielding K222-N 3 as a reddish oil (8.1 mg, 17 µmol, 37%). 1 H NMR (400 MHz, CDCl 3 ) δ 6.89 (d, 3 J H-H = 8.7 Hz, 1H), In a typical experiment, [ 201 Tl]TlCl (∼1 MBq, 50 μL) was added to an Eppendorf vial together with 50 µL of the organic chelator (0.5-2 mM). Na 2 CO 3 (25 µL of 1 M) and water (75 µL) were added to the vial and the reaction was shaken at room temperature for 15 min.
Serum stability. A solution containing [ 201 Tl][Tl(K222-PSMA)] + (100 μL, ∼1 MBq) was added to filtered human serum (Sigma-Aldrich, 300 μL) and incubated at 37°C for either 0 min, 30 min or 120 min. After the indicated time the sample was treated with ice-cold acetonitrile (300 μL) to precipitate and remove serum proteins. Acetonitrile in the supernatant was then removed by evaporation under a stream of N2 gas (30 min). The final solution was then analysed by reversephase analytical HPLC.
To assess PSMA targeting, PSMA-expressing cells DU145-PSMA and non-PSMA-expressing cells DU145 cells were seeded in 6-well plates at a density of 0.5 × 10 cells per well 1 day prior to the experiment. Three technical replicates were performed for each condition. Cell medium was replaced with 1 mL complete medium 1 hour before the cells were treated. [ 201 Tl][Tl (K222-PSMA)] + (50 kBq, in 10 μL of phosphate buffered saline) was added to each well, and the cells incubated at 37°C for 15 min and 1 h. Competition studies were also performed following co-incubation with the PSMA-inhibitor 2-( phosphonomethyl)pentane-1,5-dioic acid (PMPA; 30 μL of 750 mM PMPA solution per well). After incubation, plates were placed on ice, the supernatant was removed and the cells were washed with ice cold phosphate buffered saline solution (2 × 1 mL). The cells were lysed with ice cold radioimmunoprecipitation assay (RIPA) buffer (Thermo Fisher Scientific Inc., 500 μL), and radioactivity content was determined by gamma counter (1282 Compugamma; LKB, window set to channels 35-110 to measure 201 Tl gamma emissions). Cell uptake was also performed following [ 201 Tl]TlCl incubation (50 kBq, 10 μL) for 15 min and 1 h. Cells were harvested and radioactivity measured as described above.

DFT calculations
All calculations were performed using the Gaussian 16 package (Revision C.01). 27 Full geometry optimisations were performed using the ωB97X-D functional 28 in gaseous state with the LANL2DZ basis set and effective core potentials for Tl, and 6-311+G** for all other atoms. 29 Structures were optimised without using symmetry constraints. Frequency analysis of the optimised structures confirmed a true minimum energy structure by the absence of imaginary frequencies.

Conflicts of interest
There are no conflicts of interest to declare.