The Synthesis of Rotaxane Probes for Magnetic Resonance Imaging (MRI)

Rotaxanes are simple molecules made by an interaction between a host and a guest. Research has shown that rotaxanes have the capability of exchanging a hyperpolarized inert gas atom for use in Magnetic Resonance Imaging (MRI). Currently, MRI contrast is enhanced by the injecting a magnetic gadolinium (III) [Gd(III)] ions into the human body prior to the imaging experiment. Unfortunately, these contrast agents are expensive and toxic; as a result, there is need for a cheaper and less toxic imaging agent. Additionally, it would be beneficial to develop targeted contrast agents, i.e. macromolecules that bind specific analytes, proteins, or cellular receptors within the body. By combining rotaxanes with Hyperpolarized Xe Chemical Exchange Saturation Transfer (HyperCEST), Xe MRI technology should be capable of imaging specific areas in the human anatomy, thus facilitating the study and diagnosis of diseases or injuries. Currently Xe MRI is being used to genertate images of the lungs and the brain, but with a synthetic molecule we hope to broaden this capability to include high-resolution molecular imaging. The manuscript, “Cyclodextrin-based Pseudo-rotaxanes: An Easily Conjugatable Scaffold for Hyperpolarized Xenon Magnetic Resonance Imaging Biosensors” is the result of our preliminary work to develop a viable molecular probe. The manuscript focuses on the development of a new class of xenon-129 MRI contrast agents based on rotaxanes of γ-cyclodextrin, and the application of this new technology to the synthesis of a potential biosensor for imaging the -amyloid plaques that are associated with Alzheimer’s disease..


Introduction
Hyperpolarized (HP) xenon-129 based magnetic resonance imaging (MRI) biosensors have the potential to become a molecular imaging modality with similar sensitivity to positron emission tomography (PET), but with theoretically better spatial resolution, no ionizing radiation and lower cost. HP gas MRI takes advantage of the signal enhancement provided by the hyperpolarization of gas, that is, the aligning of the spins of a majority of nuclei with an external magnetic field, providing a signal enhancement of up to 100,000 times above thermally polarized nuclei. 8 HP gas MRI is an ideal imaging modality for imaging of the lungs. 9 Xe diffuses throughout the whole body following inhalation, and because Xe is hydrophobic, it deposits particularly well in lipid-rich tissues. High-resolution, three-dimensional images are obtained by detecting the HP Xe that is deposited in various fatty tissues, such as the brain. 10-12 HP Xe atoms cannot, by themselves, be tuned to target particular regions in the body; however, targeted HP Xe molecular probes that are capable of binding both biochemical receptors and xenon atoms in vivo have been postulated as a way to perform molecular imaging, and numerous prototypes have been developed, though targeted HP Xe biosensors have yet to be used for imaging in a living animal.
Like all molecular probes, HP Xe MRI biosensors consist of two functional parts, a binding component and a detection component that are joined by a covalent tether ( Figure 1A). The binding component is an affinity tag or antibody that binds to a specific analyte or biochemical receptor, allowing for the detection and characterization of specific biochemical phenomena. The detection component for a HP Xe biosensor is usually a supramolecular cage-like structure that can encapsulate a xenon atom. For in vivo imaging, the magnetically active HP Xe can be inhaled by the subject and the imaging biosensor can be administered via an injection. After the dissolved Xe gas has circulated through the subject's body and the targeted biosensor has attached to the desired molecular target, the Xe will be reversibly encapsulated by the biosensor in a traditional host-guest interaction. If the reversible encapsulation is slow on the NMR time scale, it will produce a unique chemical signal in the 129 Xe NMR spectrum which can be transformed into a three-dimensional image.
The Hyperpolarized gas Chemical Exchange Saturation Transfer (HyperCEST) pulse sequence takes advantage of the continual diffusion of Xe atoms in and out of a Xe-encapsulating cage molecule, and allows for the detection of picomolar concentrations of a HP Xe biosensor. [13][14][15] Because the binding of the Xe is reversible, but slow on the NMR timescale, the 129 Xe spectrum of a biosensor contains two peaks, one for unbound Xe and one for the xenon that is encapsulated by the biosensor. By exciting the HP Xe atoms inside the supramolecular cage at their unique chemical shift offset frequencies, the Xe atoms inside the cage molecules become depolarized. When these depolarized Xe atoms exchange with the HP Xe atoms from the pool of dissolved Xe atoms, there is a reduction in signal from the pool of dissolved phase Xe atoms ( Figure 1B). The HyperCEST technique, combined with hyperpolarization of the nuclei, provides a theoretical signal enhancement of up to a billion times above thermally polarized nuclei. 16 Despite the seminal work of Pines and colleagues over two decades ago, 14 HP-Xe MRI biosensor technology has yet to be translated into a clinical imaging technique.
We recently disclosed the first in vivo images of the MR contrast portion of a xenon MR imaging biosensor in a live animal using the non-targeted Xe cage, CB6. 17 In our opinion, the development of targeted HP Xe MRI biosensors has not been slowed by a lack of interest in the techniques or by barriers in MRI technology. The problem is more fundamental: the supramolecular hosts that encapsulate xenon are difficult to synthesize and derivatize, so sufficient quantities of biosensors for in vivo imaging are simply not available.
Herein, we describe the development on a new class of water-soluble HP Xe biosensors that use a pseudo-rotaxane to encapsulate HP Xe and can be imaged using a HyperCEST pulse sequence. We have developed a new biosensor based on this scaffold that has the potential to image the amyloid fibrils that are associated with Alzheimer's disease (AD), which we call Xenon Cage Amyloid Ligand-1 (XCAL-1). The modular nature of the synthetic scheme should allow for the rapid synthesis of not only XCAL-1, but a wide variety of biosensors, thus paving the way for numerous clinical applications.

Design of pseudo-rotaxane HP Xe probe
Our group, as well as others, have tested a number of different xenon hosts including cryptophanes, 18-21 cucurbiturils, 17,22-24 liposomes, 25,26 gas vesicles, 27 and bacterial spores. 28 Of these xenon-capturing scaffolds, cryptophane-A has been studied the most. For example, cryptophane-A derivatives have been conjugated to affinity tags to bind a wide variety of targets, 29-31 such as CD14 cancer cells, 32 the cancer-associated HER2 receptor, 27 zinc, 33 toxic metal ions, 34 and the inflammation-marking peripheral benzodiazapene receptor (PBR) 35 . These reports have not come without a great deal of effort. In one recent report, the Dmochowski group conjugated a cryptophane-A to folic acid to yield a potential biosensor for cancer. 36 While notable not only for its scientific novelty, the tenacity of the research associate who conducted 20 non-linear steps to synthesize the final product is remarkable in and of itself. Surely there must be a simpler and higher yielding synthesis of a HP-Xe biosensor! Until recently, the hollow, ball-shaped molecular cages known as cryptophanes appeared to be privileged structures in the field of HP-Xe imaging. Cryptophane-A contains a hydrophobic core with a volume of 95 Å 3 , 37 and is capable of reversibly binding xenon with a ka of 3 kM -1 , with a residence time of 0.5-8 ms at room temperature 38 . Despite it's more tube-like structure, cucurbit[6]uril (CB[6]) is also capable of binding xenon with a comparable affinity (ka = 200 M -1 ), 39 but its larger derivatives, cucurbit [7]uril (CB [7])and cucurbit [8]uril (CB [8]), do not show any affinity for xenon, as observed by 129 Xe NMR. Additionally, ,  and -cyclodextrins (CD), which are truncated cone-shaped macrocycles composed of six, seven or eight D-glycopyranoside units, have hydrophobic cavities with minimum diameters of 5.3 Å, 6.5 Å and 8.3 Å, respectively ( Figure 2A). CDs are some of the most commonly used hosts in supramolecular chemistry, 13a and they would be ideal components of 129 Xe biosensors because they are non-toxic and water-soluble. Unfortunately, ,  and -CD either fail to bind xenon in aqueous media, or the reversible binding has too fast of an exchange rate to be observed by 129 Xe NMR at room temperature. 40b We hypothesized that macrocycles that were too large to bind xenon on their own, such as the cyclodextrins, could be threaded with long alkyl chains to create rotaxane-type complexes that were capable of forming a ternary complex with xenon ( Figure 2B). 13c Rotaxanes are well-known supramolecular species composed of a molecular axle that is threaded through a tube-shaped host, creating a non-covalently bound structure. In order to serve as a molecular probe, the inner diameter of the rotaxane's macrocycle must be large enough to fit both the molecular axle and a xenon atom in its hydrophobic core. However, a macrocycle that is too large would not be detectable using HyperCEST because the xenon would exchange in and out of the host at a rate that is too high to support HyperCEST detection. Prior to this work the only precedent for the formation of this kind of ternary complex with xenon was reported by Cohen, who showed that a CB6 derivative could simultaneously bind hexane and xenon. 40 Consequently, we designed three classes of pseudo-rotaxanes to determine their potential as the imaging component of HP-Xe biosensors: CB-based pseudorotaxanes, pillararene-based pseudo-rotaxanes, and CD-based pseudo-rotaxanes. In addition to the novelty of the ternary xenon complexes, the design shown in Figure 2B could also expedite the development of a wide variety of biosensors because they would not be synthesized by covalently tethering the affinity tag to the xenon host; rather, the affinity tag would be conjugated to the axle of the rotaxane, which is presumably a straightforward process, and the tethering of the affinity tag to the macrocycle would then be accomplished via classic supramolecular chemistry relying primarily on hydrophobic interactions, which, coincidentally, are the same forces that are required for efficient xenon binding. For these preliminary studies, we defined two criteria for success: firstly, the molecule had to be readily synthesized and conjugatable.
Secondly, the molecule had to be MR detectable by displaying a HyperCEST effect.
Towards the first goal, we found that pillararenes were relatively easy to synthesize and CB and CD macrocycles are both commercially available. Each of the host molecules were threaded with five, eight and ten-carbon molecular threads that contained terminal ethylimidazolium groups, which served to enhance the water solubility of the greasy alkanes and enabled facile detection by mass spectrometry (Scheme 1). Furthermore, one can easily imagine methods to create similar moieties that were attached to affinity tags. In all cases studied, the rapid formation of threaded complexes was observed by NMR, though most cases showed rapid host-guest exchange on the NMR time scale.
Subsequent analysis by 129 Xe NMR quickly identified the promising scaffolds that combined the desired attributes of facile synthesis with MR detectability via HyperCEST.
For the HP-Xe studies, we used a custom-built fritted phantom inside of a custom dual tuned 1 H/ 129 Xe radiofrequency (RF) coil to acquire all free induction decay (FID) spectra ( Figure 3). Cage molecules, dissolved in water and/or DMSO, were placed inside the fritted phantom, and HP 129 Xe was introduced below the fritted phantom which created microbubbles that bubbled vertically through the sample. A series of saturation pre-pulses at a variety of chemical shift offsets were loaded into the user interface software of the GE Achieva 3T MR scanner. Spectra with different saturation pre-pulses were acquired approximately every six seconds.

Discovery of pseudo-rotaxane HP Xe probes
Our initial studies commenced with the analysis of cucurbiturils, as we have previously had success using the most common member of this family of macrocycles, CB6, in HyperCEST studies. As expected, irradiation at +128 ppm (relative to the peak corresponding to dissolved xenon) produced a 67% depletion, thus confirming that our experimental method was reliable. Various threaded complexes of CB6, CB7 and CB8 were then synthesized (see Supporting Information) and subjected to the same HyperCEST protocol. Unfortunately, none of the pseudo-rotaxanes could be detected by HyperCEST, indicating that there is not sufficient space in the cavity of the supramolecular complex (likely true for complexes like 1CB6) or that that xenon exchanges too rapidly in and out of the complex to be detected (possibly true for larger complexes like 1CB8).
Two different pillarene structures were also tested (see Supporting Information for structures), but both suffered from poor water solubility. Consequently, organic cosolvents or non-ionic diazide bars (Scheme 4) had to be employed. All pillararenederived threaded complexes failed to produce a HyperCEST signal. This was surprising because pillararene-based pseudo-rotaxane reported to be was capable of binding xenon. 41 We attempted to further these studies by applying HyperCEST saturation pulses and acquiring a HyperCEST depletion spectrum, but we were unable to observe a HyperCEST effect in pseudo-rotaxanes based on the pillararene macrocycle. We were neither able to detect the presence of a peak corresponding to a xenon-pillararene complex, nor a HyperCEST effect.
The maximum HyperCEST depletion for all three cyclodextrin-based rotaxanes occurred at approximately +128 ppm from the Xe gas phase signal. HyperCEST depletion spectra for 1γ-CD are shown in Figure 4. Importantly the HyperCEST depletion for (1γ-CD) was comparable to that of CB6, a xenon cage that we have recently shown to be amenable to in vivo HP Xe MRI.
3T clinical whole-body MR scanner, which proves significant advantages over conventional NMR spectrometers, namely the ability to perform whole-body imaging experiments. Unfortunately, these advantages come with some trade-offs, specifically the ability to acquire HyperCEST depletion spectra with a Lorenzian fit line, such as in the data processed for cryptophane and cucurbituril agents as demonstrated by the Schroeder and Dmochowski groups using high-field, high-resolution NMR spectrometers. 42  + C10 diethylimidazolium bar (1) 50% +128 a Samples (2 mL, 10 mM) were dissolved in water, 1D 129Xe spectrum were initially recorded and then a series of HyperCEST spectra were sequentially recorded using a series of off resonance pulses varying by 5 ppm. b Performed in H2O/DMSO. c Performed in CHCl3. NMR titration studies were performed to assess the nature of the host:guest interaction for 1γ-CD. 46

Application to the synthesis of XCAL-1
To demonstrate the utility of this new class of xenon-binding agents for the synthesis of targeted biosensors, we synthesized a potential molecular probe using Thioflavin T (ThT). ThT is a fluorescent dye that binds to the -amyloid plaques that are associated with Alzheimer's disease (AD). 47 Using HP Xe biosensors like XCAL-1, we envision potential clinical applications for studying the progression of AD or the efficacy of treatments for it (Scheme 5).
The synthesis of the biosensor involves a simple acylation, followed by the formation of the pseudo-rotaxane. Both steps are nearly quantitative. As before, formation of the 1:1 complex was confirmed by 1 H NMR studies, and the association constant for the pseudo-rotaxane was determined to be 2.0 x 10 4 M -1 . Thus far, we have synthesized XCAL-1 on a scale of hundreds of milligrams, but the reactions are simple

Summary
In conclusion, we have discovered a novel method for synthesizing potential HP    consists of a cage molecule, linker, and targeting moiety ( Figure 6). 16,52 The 129 Xe will exchange quickly in and out of the cage to produce a defined NMR signal different to that of free HP 129 Xe. 14,52,56,57 The caged molecules can be functionalized to bind to specific biological receptor. 14 Previously, cryptophane-A has been functionalized and shown exchanging HP 129 Xe (Figure 8). 16,52 A 129 Xe biosensor allows for the study and diagnosis of various biochemical phenomena. delivery, and cell transport agents. 59 Rotaxanes can form a ternary complex with HP 129 Xe to create a molecular biosensor. We are capable to detect, in small amounts, the 129 Xe biosensors in the body using Hyperpolarized 129 Xe Chemical Exchange Saturation Transfer (HyperCEST). 19 Right now we are trying to synthesize a Folic Acid probe to bind to cancer tumors and a Thioflavin-T probe to image cells found in Alzheimer's disease.

Thioflavin-T Molecular Probe
Thioflavin-T (ThT) is a fluorescent molecule with a high binding affinity to betaamyloid (Aβ) plaques used to diagnose Alzheimer's Disease (AD), a type of dementia, in the brain. [60][61][62] Amyloid plaques are developed from amyloid precursor proteins (APPs), also known as senile plaques, found exclusively in AD patients. Aβ plaques are a type of amyloidosis, a disease, caused by the misfolding of peptide or protein that is unable to remain in its natural state. 61,63 Along with AD, amyloidosis has been associated with Parkinson's Disease, type-II diabetes, and cataracts. 61,63,64 As an individual ages, amyloid plaques build up inside neurons and blood vessels in the brain to reduce the functionality of the brain. 64 In order to diagnose AD, physicians need to detect the presence of Aβ along with any cognitive deterioration. 60 Pharmaceutical companies have developed ThT derivatives to be used in vitro allowing for the identification and quantification of Aβ. 65 Once the ThT binds to Aβ, ThT fluoresces with an excitation at 440nm and with emission at 490nm. Unbound ThT has a very low fluorescence with an blue-shifted excitation at 350nm and emission at 440nm. 63 Using the capability to synthesize various ThT derivations, we wanted to synthesize a molecular probe, as stated earlier, to image Aβ using MRI. We chose 4aminobenzothiazole because it is the easiest derivative to synthesize (Scheme 1). Based on the determined inclusion complexes, 8-carbon and 10-carbon alkyl guests were chosen to make up the core of our molecular probe.
The first approach was to bind ThT (compound 7) to the alkyl chain by an azidealkyne "click" reaction (Scheme 6) 66 or through a nucleophilic acyl substitution (Scheme 7). 67,68 The "click" reaction route required the synthesis of a propargylated ThT (10) and a diazido alkyl chain (4 and 5). Unfortunately the "click" reaction was unable to be purified or confirmed by NMR; as a result, the acyl substitution (Scheme 4) was attempted using ThT with sebacoyl chloride (route 1), and disuccinimidyl sebacate (9, route 2). Synthetic route 1 was able to produce compound 8 with a yield of 98%. Route 2 produced no product, only starting material. With the bar for the rotaxane (8) in hand, the next step was to perform binding studies with γ-cyclodextrin and 129 Xe to see if we have successfully created a new pseudo-rotaxane molecular probe.
Scheme 7: Synthetic routes for the thioflavin-T probe.

Folic Acid Molecular Probe
Folic acid (FA) was chosen to make a biosensor due to its high binding affinity to folate receptors (FR) which are upregulated in many tumors, particularly ovary, lung, breast, kidney, brain, endometrium, and colon cancers. 5 Currently, FA conjugated drugs are being used as cancer treatments by binding to the FR located on cancer tumors. FA is water soluble and naturally forming, which makes it cheap, readily available, and applicable to a wide range of cancer tumors. 5 Having a probe that specifically binds to one of these tumors would allow researchers and physicians to study and treat these diseases more efficiently.
Using the same approach as the ThT molecular probe, we wanted to bind FA to an alkyl chain by functionalizing FA. We have made several unsuccessful attempts to bind dibromopropanyl chloride (DBPC) to the primary amine in order to perform a "click" reaction to either compound 4 or 5 (Scheme 8). 69,6 The next approach would be to perform a coupling reaction to activate γ-carboxylic acid of FA with N-hydroxysuccinimide and to react this activated ester with 1,10-diaminodecane to form a molecular thread containing FA on both ends (Scheme 9). 67 The preliminary development of a ThT probe was successful. After performing the binding studies, we determined a pseudo-rotaxane, not a rotaxane, was formed. The pseudo-rotaxane allows for the guest molecule to slip in and out of the host molecule.
Our probe can be altered to make a rotaxane by conjugating a large end cap to one side

Experimental
Known and novel compounds were synthesized according to the following procedures.

Association studies for Thioflavin-T probe
The NMR titration experiments were conducted per the following procedures.
The host concentrations were kept constant while the guest concentration was increased periodically. Stock solutions of 10 mM host (gamma CD) and 100 mM guest (ThTprobe) were prepared. A series of NMR samples were prepared ranging from 1:1 (Host: guest ratios are mM concentrations) to 1:5, increasing the guest ratio by 1 mM for each sample. The aromatic proton of the ThT (circled in red) was monitored. Data was processed using Wolfram Mathematica software. The chemical shift values were used to determine the correct fit from the equations shown below (Table 2). 75 The plots are shown below. All samples were prepared at 45 o C. In these equations G and H represent host and guest. PGo and PGH are constants.
Pobs (NMR shift value) is the y-axis, and the guest concentration is used as the x-axis to plot the data. Curve fitting Equation 3 produced a negative value indicating the data set does fit the given equation; hence, the combination of 1:1 and 1:2 stoichiometry won't be present in that system. Using the calculated values, the association constants of the complexes were obtained.