Metabolic Profiling of S -praziquantel: Structure Elucidation Using the Crystalline Sponge Method in Combination with Mass Spectrometry and Nuclear Magnetic Resonance S

Praziquantel (PZQ) is the drug of choice for treatment of the neglected tropical disease schistosomiasis. Although the drug has been extensively used over several decades and its metabolism well studied (several oxidative metabolites are known from literature), the knowledge of the complete structure of some of its metabolites remains elusive. Conventional techniques, such as nuclear magnetic resonance or liquid chromatography mass spectrometry were used in the past to investigate phase I and phase II metabolites of PZQ. These techniques are either limited to provide the complete molecular structure (liquid chromatography mass spectrometry) or require large amount of sample material (NMR), which are not always available when in vitro systems are used for investigation of the metabolites. In this study, we describe new structures of S -PZQ metabolites generated in vitro from human liver microsomes using the crystalline sponge method. After chromatographic separation and purification of the oxidative metabolites, ultra-performance liquid chromatogra-phy-quadrupole


Introduction
Schistosomiasis is a neglected tropical disease caused by a parasitic flatworm of the genus Schistosoma and affects almost 240 million people worldwide, with a high prevalence in Africa. The World Health Organization implemented a program for schistosomiasis control over the past 40 years and recommends the anthelmintic drug praziquantel (PZQ) as the treatment of choice for all forms of the disease (WHO, 2011(WHO, , 2020. PZQ is currently used as a racemic formulation in therapy; however, the activity is mainly associated with the R-enantiomer (Meister et al., 2014;Kova c et al., 2017). Although the drug has been known and extensively used since the early 1980s, the information content of the metabolic pathway is limited. Enantioselective transformation of R-PZQ and S-PZQ has been explored in various studies, showing different metabolic profiles for both enantiomers (Wang et al., 2014;, Park et al., 2021. Several mono-oxidized (116 Da) and secondary oxidative metabolites (132 and 114 Da) are known from the literature, but most of their complete structures are still unknown. The main metabolite in human is 4'hydroxy-PZQ and the position of hydroxylation was identified both in cis and trans configuration (Nleya et al., 2019). Two more metabolites have been described as 8-hydroxy-PZQ and 11b-hydroxy-PZQ by isolation from human urine and in vitro recombinant human P450 reactions, using NMR for data analysis (Schepmann and Blaschke, 2001;. Additional metabolites have been analyzed using different liquid chromatography mass spectrometry (LC-MS) systems, giving information of the type of metabolism (i.e., hydroxylation, dehydrogenation and glucuronidation) and allowing to narrow down the site of metabolism to a certain part of the molecule (Lerch and Blaschke, 1998;Huang et al., 2010;Wang et al., 2014). Despite the advantage of LC-MS techniques to reach high sensitivity for sample analysis from in vitro origin, these techniques failed to identify the complete structure of metabolites.
A new approach for structural elucidation of compounds was introduced in 2013 by Makoto Fujita and is commonly known as the "crystalline sponge method" (CS-XRD) (Inokuma et al., 2013). The method enables crystal structure determination without the limitation of crystallization with only nanogram to a few microgram of analyte material. A metal coordination complex [(ZnX 2 ) 3 ·(tpt) 2 ] n ·x(solvent) m (X5I; tpt52,4,6-tris(4-pyridyl)-1,3,5-triazine) is used as a pre-existing crystal and functions as a host. During a "soaking process", the analyte (guest) is absorbed into the pores of the three-dimensional framework via diffusion and regularly ordered by intermolecular, noncovalent interactions and thus accessible for X-ray diffraction analysis Brunet et al., 2017;Sakurai et al., 2017). We recently demonstrated that CS-XRD is a valuable tool, which can be applied to identify the exact position of metabolism for phase I and phase II metabolites generated in low amounts from in vitro incubation (Rosenberger et al., 2020).
In this study, we apply the CS-XRD for the determination of the absolute structures of S-PZQ and six of its hydroxylated metabolites after incubation with human liver microsome (HLM).
Metabolism of S-PZQ by HLM. The hydroxylation of S-PZQ was conducted with an incubation mixture containing 0.5 mg/ml of HLM, 50 mM of potassium phosphate buffer (pH 7.4), 1 mM of magnesium chloride, and 200 mM of substrate. After 5 minutes of preincubation (37 C, 150 rpm), the reaction was initiated by the addition of nicotinamide adenine dinucleotide phosphate (NADPH, 1.5 mM), and the mixture was incubated for another 24 hours (final volume: 5.40 ml). After 6 hours, the reaction was boosted by addition of NADPH (1.5 mM).
The reaction was quenched by adding one volume of ice-cold acetonitrile and the precipitated proteins were removed by centrifugation at 4000 g for 1 hour at 4 C. The supernatant was evaporated to dryness (nitrogen flow, 40 C) and resolubilized in acetonitrile/water (20:80). 100 ml aliquots were used for purification and fractionation by high-performance liquid chromatography coupled with mass spectrometry (HPLC-MS).
Ultra-Performance Liquid Chromatography-Quadrupole Time-of-Flight Mass Spectrometry. The supernatants were analyzed on a Waters Acquity ultra-performance liquid chromatography (UPLC) system combined with a Xevo G2-S qTOF (Waters Corporation). Analysis was performed with an electrospray ion source in positive ion mode. For MS scan, the quadrupole timeof-flight mass spectrometry (qTOF) was operating with a source temperature at 150 C, a desolvation temperature at 600 C and a capillary voltage at 0.5 kV. For full-scan MS mode, the collision energy was set to 4 V, and the mass range was set to m/z 100 to 1000. For MS/MS studies, qTOF MS e acquisition was conducted using a collision energy ranging from 25 to 40 V.
Metabolite separation was achieved on a chiral Lux Cellulose-2 column (150-2 mm, 3 mm; Phenomenex) equipped with a Lux Cellulose-2 precolumn (3 mm inner diameter; Phenomenex) at a flow rate of 0.45 ml/min over a period of 20 minutes using solvent A (water 1 0.1% formic acid) and solvent B (acetonitrile 1 0.1% formic acid) as mobile phases. The column oven temperature was set to 40 C. Elution was performed using a gradient from 20 to 90% B in 15 minutes, 90 to 95% B in 0.2 minutes, 95% B for 2.1 minutes, returning to 20% B in 0.1 minutes and re-equilibration at 20% B for 2.6 minutes. The software UNIFI version 1.9.4 (Waters Corporation) was used to support the identification of metabolites. The metabolites were evaluated for mass error (# 4 ppm), fragmentation pattern and retention time.
HPLC-MS. The supernatants were analyzed and fractionated on an Acquity Arc HPLC system (Waters Corporation). The equipment components were described in detail in our previous work (Rosenberger et al., 2020).
Samples were analyzed with electrospray ionization mass spectrometry in positive ionization mode within an acquisition range from m/z 100 to m/z 650 in continuum mode and a sampling frequency of 2 Hz. The capillary voltage and the cone voltage were set to 0.8 V and 10.0 V, respectively. The probe temperature was set to 600 C. The mass value of m/z 329 was registered by the MS detector and triggered the fractionation of the target analytes. The software MassLynx 4.2 and FractionLynx were used for data acquisition and sample fractionation.
HPLC separation for M4 and M6 PZQ was performed with two Chromolith Performance RP18e columns (100-4.6 mm; Merck KGaA) for primary chromatographic purification and one Purospher Star RP18e Hibar HR column (100-2.1 mm, 2 mm; Merck KGaA) as second chromatographic system. The column oven temperature was set to 25 C. For the first purification, elution was performed at a flow rate of 1.0 ml/min using solvent A (water) and solvent B (acetonitrile) as mobile phases. The analytes were eluted over a period of 30 minutes, using a multi-segmented gradient from 0% to 25% B in 7 minutes, 25% to 30% B in 3 minutes, 30% to 80% B in 10 minutes, 80% B for 5 minutes, returning to 0% B in 0.1 minutes, and re-equilibration at 0% B for 4.9 minutes. The eluent was directed to the waste during the first 5 minutes to reduce contamination. For the second purification, elution was performed at a flow rate of 0.45 ml/min over a period of 15 minutes, applying the following gradient: mobile phase 0% to 60% for 10 minute, 60% B to 100% B in 2 minute; 100% to 0% B in 0.1 minutes and re-equilibration at 0% B for 2.9 minutes.
HPLC separation for M3, M5, and additionally M6 PZQ was achieved using an additional third purification step. The further purification was performed using the Lux Cellulose-2 column and gradient described as before. Differences were the column oven temperature of 25 C and the solvent A (water) and solvent B (acetonitrile) as mobile phases.
Following each chromatographic separation step, the collected fractions were pooled, evaporated to dryness at 40 C under nitrogen flow and resolubilized in acetonitrile/water (20:80) for further purification. After final purification, the collected fractions from each metabolite were combined and evaporated to dryness to conduct the crystalline sponge (CS) soaking and NMR measurement.
For guest soaking of S-PZQ, a single crystal of 1a was transferred with 50 ml of cyclohexane to a glass vial. After addition of 2 ml of dissolved analyte solution (1 mg/ml in 1,2-dimethoxyethane), the sample vial was closed with a screw cap and the septum seal was pierced with a syringe needle, before placing it in an incubator at 50 C for 1 day. Following evaporation of solvent, the vial was placed without a needle for 2 days at 4 C to increase guest occupancy and hence the chirality transfer from the guest molecules to the CS framework. For S-trans-4'-hydroxy-PZQ, one crystal of 1b was used and analyzed after evaporation of nhexane after 1 day. The soaking condition for S-cis-4'-hydroxy-PZQ was adjusted by reducing the volume of n-hexane to 25 ml, the temperature for solvent evaporation to 25 C and the soaking time to 3 days by using a syringe needle with smaller diameter. Soaking of incubation samples (about 1-2 mg) was conducted similarly to the reference material. Two microliters of 1,2-dimethoxyethane, 50 ml of cyclohexane and one CS (1a) were pipetted into the vial, which contained the pooled metabolite material. The incubator temperature was set to 50 C and the evaporation time was one day.
Single-crystal X-ray diffraction measurements were performed by using a Rigaku Oxford Diffraction XtaLAB Synergy-R diffractometer (Cu-K a , k 5 1.54184 Å) at a temperature of 100 K and crystal structure modeling using OLEX2 , SHELXT, and SHELXL  were conducted as described in detail in our previous work (Rosenberger et al., 2020). The software ShelXle was used to generate the electron density maps of S-PZQ and its hydroxy metabolites (H€ ubschle et al., 2011).
NMR of PZQ Metabolite M3. Structural elucidation of metabolite M3 was performed at 298 K using a 700 MHz Bruker Avance III equipped with a 5-mm cryocooled triple resonance probe. The isolated metabolite (about 15 mg) was dissolved in 200 ml of DMSO-d 6 and transferred into a 3 mm NMR tube. The 1 H NMR spectrum was acquired with 64 k time domain points, a spectral width of 20 ppm, a relaxation delay of 10 seconds, and 1024 scans. Edited heteronuclear single quantum coherence NMR spectrum was recorded with 1024 × 256 time domain data points over a spectral width of 12 ppm in the t2 and 165 ppm in the t1 dimension. The homonuclear correlation spectroscopy NMR spectrum was acquired with 1024 × 256 time domain data points over a spectral width of 12 ppm in the t2 and t1 dimension. The rotating frame Overhauser enhancement effect spectroscopy NMR spectrum was recorded with 1024 × 256 time domain data points over a spectral width of 12 ppm in the t2 and t1 dimension. The software MestReNova version 14.1.0 (Mestrelab Research) was used for data analysis.

Results
UPLC-qTOF Analysis of S-PZQ Metabolites Prepared by Incubation with HLM. Both enantiomers of PZQ ( Fig. 1) were incubated with HLM for 24 hour respectively, for metabolite profiling and identification (data not shown). S-PZQ showed an increased formation of metabolites and was therefore chosen for further analysis. During the study of metabolic stability, the UPLC-qTOF chromatogram of S-PZQ, incubated with HLM in the presence of NADPH, showed the formation of several oxidative metabolites (116, 132 and 114 Da), which is in line with already published data (Huang et al., 2010;Wang et al., 2014). The incubation of S-PZQ without cofactor was conducted in parallel and showed no formation of metabolites, indicating their generation by enzymatic reaction. The analysis presented in this study, however, focuses on six monohydroxylated metabolites of S-PZQ (M1-M6).
The metabolites M1-M6 were detected at a protonated molecular mass [M1H] 1 of m/z 329. The increased m/z value of 16 Da, compared with the [M1H] 1 ion of the parent (m/z 313), suggested the addition of one oxygen to the molecular scaffold.
The mono-oxidized metabolites were further analyzed based on their MS/MS spectra and accurate mass measurements (Fig. 2, Supplemental Table 1). The mass spectrum ([M1H] 1 ) of M1 and M2 showed similar fragment ion peaks as the parent compound. The main fragment ions at m/z 203 and m/z 174 were assigned to the "core moiety" (i.e., hexahydro-pyrazino [2,1-a]isoquinolin-4-one), indicating that the oxidation occurred at the cyclohexane ring. Metabolites M1 and M2 were identified as S-trans-4'-hydroxy-PZQ and S-cis-4'-hydroxy-PZQ by comparison of the respective retention time with the reference material.
The main fragment ions at m/z 219, m/z 190, m/z 162 and m/z 148, observed in the spectra of both M3 and M5, were assigned to an oxidized core moiety, indicating that the hydroxylation occurred somewhere at the tetrahydro-isoquinoline ring system. The full-scan mass spectrum of metabolites M4 and M6 showed the abundant ion m/z 311, representing a predominant loss of water (329 -> 311). In addition, both metabolites fragmented to form characteristic ions of m/z 201, m/z 144 and m/z 130, indicating a spontaneous water loss of the fragment ions observed for M3 and M5. The loss of water is predominantly observed for aliphatic hydroxylation, whereas it is less common for hydroxylation reactions, which occur on phenyl rings (Ramanathan et al., 2000;Hol capek et al., 2010). Therefore, it is likely that the addition of one oxygen occurred in the aliphatic part of the core moiety and not on the aromatic benzene.
The tandem mass spectrometry fragmentation pattern allowed to limit the oxidation to a certain part of the molecule but could not reveal the actual positions of hydroxylation. The metabolite structures of S-PZQ were therefore further analyzed using CS-XRD and NMR.
Structural Elucidation of S-PZQ Metabolites (M1-M4 and M6) by CS-XRD. For successful structure determination using CS-XRD, the optimization of soaking conditions (e.g., CS type, solvent, temperature, time) is the crucial step. Thereby, various parameters need to be examined for every analyte . The optimal soaking parameters for S-PZQ were selected by conducting a CS affinity screening and examining the affinity of target analyte to CS framework using two CS types [(ZnI 2 ) 3 ·(tpt) 2 ] n ·x(solvent) m and [(ZnCl 2 ) 3 · (tpt) 2 ] n ·x(solvent) m ; solvent5cyclohexane, n-hexane) at three different temperatures (50 C, 25 C, 4 C) (data not shown). The screening method has been described in detail in our previous work (Rosenberger et al., 2021). The optimal soaking conditions were then applied for S-PZQ and its mono-oxidized metabolites. To increase the occupancy of the analyte in the CS pores and to maximize the intermolecular interactions between CS framework and guest molecules, the soaking time was further adjusted for some analytes.
The parent active pharmaceutical ingredient and the reference metabolites S-trans-4'-hydroxy-PZQ (M1) and S-cis-4'-hydroxy-PZQ (M2) could successfully be structurally elucidated with CS-XRD using a conventional X-ray diffractometer. The crystal structure of S-PZQ revealed one guest molecule with an occupancy of 64% and one additional cyclohexane molecule with final residual factor (R 1 ) and internal R-factor (R int ) values of 8.23% and 3.08%, respectively. A Flack parameter of 0.234(8) was obtained, suggesting that the absolute configuration PZQ is S. One molecule of metabolite M1 was assigned by its electron density in the asymmetric unit with an occupancy of 52% (R 1 6.58% and R int 3.86%, v 5 0.229 (8)). First analysis of the reference material M2 resulted in the determination of the hydroxy group at the 4'-cyclohexyl moiety in trans configuration. Pharmacokinetic data from previous studies showed the preferred formation of M1 and a cis to trans interconversion (Nleya et al., 2019;. To influence the less favored axial position of the substituent as little as possible, the soaking process was further optimized by decreasing the temperature to 25 C. As a result, one guest molecule in cis configuration could be assigned to M2 (occupancy 54%, R 1 5.85%, R int 2.86%, v 5 0.328 (8)).
Next, CS-XRD was applied to the mono-hydroxylated metabolites of S-PZQ generated by in vitro incubation. The formation of several metabolites in different quantities was observed, necessitating an extensive chromatographic separation and purification process to fractionate the very closely eluting metabolites (Fig. 3A). To generate the sample amount in high purity, M4 and M6 were purified using two different chromatographic columns. M3 and M5 were thereby detected as one peak, requiring further separation using a third column (Fig. 3B). To estimate the amounts of M3-M6, the MS (for M3 and M5) and UV (for M4 and M6) signals of the reference material M2 were used. Thereby, an increase in the UV-VIS absorption maxima was observed for M3 (278 nm) and M5 (281 nm) compared with S-PZQ or M6 (263 nm), indicating an aromatic hydroxylation at the benzene moiety.
To reveal the exact position of hydroxylation, the soaking experiments were conducted under similar conditions as used for the parent compound to elucidate the structure of metabolites M3-M6. XRD measurements of the soaked CS allowed the crystallographic analysis of the metabolite structures. M6 shows high electron density at position 11b of  the isoquinoline ring system, confirming the structure elucidated via NMR in previous works . High electron density in close proximity to the isoquinoline part was also observed for M4 at position 7 and M3 at position 9. Metabolite M5 shows diffuse electron density in the CS pores, but no structure could be successfully assigned. The electron density map of the "core moiety" from S-PZQ and its successfully elucidated metabolites are shown in Fig. 4.
The crystal structure of M4 revealed one molecule with an occupancy of 76%, three cyclohexane molecules, and final R 1 and R int values of 4.99% and 1.45%, respectively. The addition of one oxygen at position 7 of the molecule caused the creation of a second chiral center, which could be determined as S configuration indicated on the Flack parameter value of 0.191(7).
As already observed for the other metabolites and the parent compound, M6 was expected to be enantiopure. Surprisingly, an inconclusive Flack parameter of 0.5, indicating a racemic mixture, was observed. Furthermore, when analyzing the CS pores, both S-and R-11b-hydroxy-PZQ (occupancy 40% and 30%) were observed (R 1 5.92%; R int 2.48%). The observation was further confirmed by chiral chromatography with the separation of both enantiomers. After chromatographic separation and purification, both enantiomers were dmd.aspetjournals.org examined separately by CS-XRD. The soaking experiment of the first enantiomer of M6 showed the presence of one S-11b-hydroxy-PZQ molecule, but due to low occupancy of the analyte and therefore an insufficient chirality transfer from guest molecules to the host framework, the resulting inconclusive Flack value did not allow the confirmation of the chiral center. However, when analyzing the second enantiomer, one molecule of R-11b-hydroxy-PZQ (occupancy 100%) could be observed in the CS pores (R 1 7.32%; R int 3.69%, v 5 0.221(8)), hence indicating the identification of the first enantiomer to be in S configuration.
The refined crystallographic structures of S-PZQ and its hydroxy metabolites are shown in Fig. 3. Crystallographic data, ORTEP diagrams of the asymmetric unit including CS framework and analytes, reciprocal h0l layers and Flack parameters before and after using solvent masking are shown in the Supplemental Data (Supplemental Fig.  1-22).

Downloaded from
Characterization of M3 by NMR. Additional to CS-XRD experiments, M3 was further isolated to characterize the structure by NMR analysis (Supplemental Data). The results confirmed the position of hydroxylation at position 9 of the benzene moiety. The following signals were assigned to the structure in the 1 H NMR measurement (700 MHz, DMSO-d 6 ). The NMR spectra indicated the presence of two conformers, which are indicated in the following assignment by superscript A and B. The numbering is depicted in Fig. 5 . The 1 H NMR spectrum of M3 showed three aromatic proton signals (one singlet at d 6.58 and two duplets at d 7.25 and d 6.72-6.64). The coupling pattern indicated a substitution of one benzene hydrogen with a hydroxyl group at either position 9 or 10. rotating frame Overhauser enhancement effect spectroscopy cross-correlations of H-7 to H-8 and H-11b to H11 were observable. Since H-8 showed singlet multiplicity and H-11 duplet multiplicity, the substitution can be unambiguously assigned to position 9 (Fig. 5, Supplemental Fig. 23-27).

Discussion
In summary, we were able to elucidate the absolute structure of three known and two unknown hydroxy metabolites of S-PZQ using CS-XRD. UPLC-qTOF analysis and target-structure prediction of the fragmentation pattern allowed us to determine the type of metabolism and to limit the position of hydroxylation to a certain part of the molecule. However, only the crystallographic data could give information about the stereochemistry and pinpoint the exact position of the inserted hydroxy groups at the 4'-C position of the cyclohexane moiety (M1 and M2) as well as the 7-C (M4), 9-C (M3) and 11b-C (M6) positions at the core moiety. The structure of metabolite M3 was furthermore confirmed by NMR analysis.
The large number of different metabolites and impurities from the in vitro incubation matrix (e.g., buffer salts, cofactor, microsomal stock solution) required a lengthy semi-preparative sample separation and purification process as well as the selection of different columns. Metabolites M3, M4, and M5 were coeluting in the first chromatographic purification step. The application of a second column allowed the separation of M4, which could be successfully identified as S-7hydroxy-S-PZQ. First experiments of the second peak using CS-XRD were unsuccessful. Only diffuse electron density could be observed in the CS pores, indicating successful guest inclusion, but no regular order of analyte molecules, which is a necessary requirement for the assignment of guest compounds. From our experience, additional impurities in analyte material investigated with CS-XRD could lead to such behavior. Therefore, a third chromatographic step was conducted for further purification, which allowed the detection of an additional metabolite and thus the separation of M3 and M5. CS-XRD analysis of M3 allowed to identify the position of hydroxylation at 9-C of the aromatic ring. The crystallographic analysis of M5, however, did not reveal the Downloaded from structure of the metabolite as a guest molecule in the CS pores, but MS and UV-VIS data strongly indicated that the hydroxylation occurred at the aromatic benzene ring. The structure of metabolite M6, which is known from literature, was confirmed, and the presence of both enantiomers in vitro was shown. The analysis of both enantiomers as a racemic mixture and individually after chiral separation allowed the determination of the respective configuration as S-11b-hydroxy-PZQ and R-11b-hydroxy-PZQ.
The application of CS-XRD for S-PZQ and its metabolites demonstrates that this technology can be applied for metabolite identification studies in drug discovery in which metabolites are only available in trace amounts from in vitro incubation. In comparison with the CS analysis, a high amount of substance was needed to conduct the NMR spectrum of M3, prolonging the lengthy sample purification process from days to weeks to collect sufficient analyte material. Furthermore, no information of the absolute configuration was obtained. The low abundancy of metabolite M4 in the incubation solution only allowed the chromatographic preparation of 1-2 mg of sample material, which is insufficient for complete structure analysis by NMR. However, the CS-XRD could be applied and allowed the successful determination of M4.
Our results show that CS-XRD in combination with MS data are a valuable tool for absolute structure identification of metabolites in early phases of drug discovery and development, it constitutes a powerful tool for human metabolite identification from in vitro incubations and allows earlier risk mitigations associated with metabolites in safety testing guidelines (FDA, 2016). The technology was able to provide new knowledge of the metabolism of PZQ, which has been used for the treatment of schistosomiasis for decades. The structural information at the atomic level were gained using only trace amounts of analyte material, outlining a significant advantage in comparison with other techniques like MS or NMR. Thus, we successfully confirmed the structure of three known hydroxy metabolites and elucidated the complete molecular scaffold of two unknown metabolites of S-PZQ.
To fully benefit from this technology, further investigation of the CS method is still needed to expand the chemical space of possible guest analytes by applying new types of CS with different binding sites and pore sizes to successfully encapsulate a wide range of target analytes in the CS pores and thus to successfully elucidate their complete structure.

Single-crystal X-ray diffraction experiments
Single crystal X-ray diffraction measurements were conducted on a Rigaku Oxford Diffraction XtaLAB Synergy-R diffractometer using Cu-Kα X-ray radiation (λ = 1.54184 Å), equipped with a HyPix-Arc 150° Hybrid Photon Counting (HPC) detector (Rigaku, Tokyo, Japan) at a temperature of 100 K using a Cryostream 800 nitrogen stream (Oxford Cryostreams, UK). The software CrysAlisPro ver. 171.41.68a (S-/R-M6 praziquantel), ver. 171.41.99 (M1 praziquantel, R-M6 praziquantel) and ver. 171.41.113a (S-praziquantel, M2 praziquantel, M3 praziquantel and M4 praziquantel) was used for calculation of measurement strategy, data reduction (data integration, empirical and numerical absorption corrections and scaling) and the generation of the h0l layers. The software ShelXle was used to generate the electron density maps Fo to verify the position of metabolism (Hübschle et al., 2011). Isocontour levels are described as σ (square root of the average variance of the density).

Crystal structure analysis
All crystal structures were modeled using OLEX2 , solved with SHELXT ver. 2014/5 and refined using SHELXL ver. 2018/1 . Figures of framework and analyte were created using OLEX2. Non-hydrogen atoms were refined anisotropically. Hydrogen atoms were fixed using the riding model. Populations of the guests in the crystal were modelled by least-square refinement of a guest/solvent disorder model under the constraint that the sum of them should equal to 100%. The number of used DMD-AR-2021-000663 3 constraints and restraints was tried to minimize and applied without changing the standard deviation. Solvent cyclohexane molecules in the pores were found in the difference electron density map and refined using the restraints (DFIX, DANG, SIMU, RIGU and SADI). These molecules are expected to be severely disordered due their high thermal motion. Due to their averaged structure of various geometry and orientation, some cyclohexane molecules are distorted to energetically-unfavorable (boat-shaped or twisted) structures. One "Alert A" notification was found by the validation program CheckCIF. This alert is derived from undefined solvent molecules is unavoidable due to severe disorder. The comments for the alerts are described in the CIFs using the validation response form (vrf). Solvent masking algorithm was applied during structure refinement. The Flack parameters before and after solvent masking including their estimated standard deviations are described (Classic Flack method).
The present crystal structures are not used for exact structure analysis. They are used to confirm structure proposals derived from information obtained from mass spectrometry and knowledge of the parent structure. Therefore, details of the crystals structure (bond lengths, angles, etc.) will not be discussed.

Crystallographic data for 1a•S-praziquantel (after solvent masking)
Crystal size: 342 × 180 × 97 µm 3 , refined formula: C 87.57H69.78I12N25.28O1.28Zn6, formula weight (Mr) = 3411.76 g/mol, light yellow needle, crystal system: monoclinic, space group C2, Z = 4, 27315 unique reflections merged from recorded 145718 ones (2.579° < θ < 67.078°) were used for structural analysis (Rint = 0.0308). Lattice parameters, R-factor on F 2 > 2σ (F 2 ), weighted R-factor, and goodness-of-fit are as follows: a = 35.2195 (4)  The framework is refined using the constraint EADP for some disordered ZnI2 moieties, as well as the commands SADI and DANG for some tpt rings. Three ZnI2 moieties are disordered and refined using disorder model. One guest molecule (A) with an occupancy of 64% and one solvent molecule (B) were found in the asymmetric unit. The benzene ring was fixed using AFIX 66. The C=O double bonds, C-C and C-N single bonds were restrained using SADI command. DANG was used for some angles of the guest. In addition, FLAT was applied for the two cyclic amide functional groups. SIMU was applied for all analyte molecules. Applied restraints can be taken from Supplemental Fig. 1. Additional electron density was found that could not be assigned due to disorder. Contribution of this electron density was removed using solvent masking algorithm implemented in OLEX2. A solvent mask was calculated and 428 electrons were found in a volume of 1622 A 3 in one void per asymmetric unit. This is consistent with the presence of 9 cyclohexane molecules per asymmetric unit. Supplemental Fig. 2 shows the electron density map of the S-praziquantel molecule taken from the crystal structure. The parent compound interacts with the CS framework by C-H···O interactions (distance 2.49 Å) between the aromatic ring of the tpt ligand and the carbonyl group of the analyte.

Crystallographic data for 1b•S-trans-4'-hydroxy praziquantel (M1) (after solvent masking)
Crystal size: 253 × 213 × 97 µm 3 , refined formula: C81.95H60.57I12N25.05O1.57Zn6, formula weight (Mr) = 3336.32 g /mol, yellow needle, crystal system: Monoclinic, space group C2, Z = 4, 27092 unique reflections merged from recorded 134281 ones (2.631° < θ < 67.074°) were used for structural analysis (Rint = 0.0386). Lattice parameters, R-factor on F 2 > 2σ (F 2 ), weighted Rfactor and goodness-of-fit are as follows: a = 34.1212 (5)  The framework is refined using the restraint SIMU and constraint EADP for some carbon and nitrogen atoms of the tpt rings. Three ZnI2 moieties are disordered and refined using disorder model. One guest molecule was found in the asymmetric unit with an occupancy of 52%. The benzene ring was fixed using AFIX 66 and DFIX, DANG, SADI and FLAT were used for some bonds and angles of the guest (see Supplemental Fig. 5). In addition, the guest molecule was refined by applying SIMU (for the complete molecule). A solvent mask was calculated and 255 electrons were found in a volume of 1024 A 3 in one void per asymmetric unit. This is consistent with the presence of 5 n-hexane molecules per asymmetric unit. The oxidation of the cyclohexane ring in position 4' in trans configuration is clearly visible from electron density maps (Supplemental Fig. 6).
S-trans-4'-hydroxy praziquantel was used as enantiomerically pure substance. The reciprocal space of 1b•M1 was inspected for the h0l layer, revealing Bragg spots h=2n and l=n (Supplemental Fig.7).
SIMU was applied for the analyte molecule. A solvent mask was calculated and 440 electrons were found in a volume of 1544 A 3 in one void per asymmetric unit. This is consistent with the presence of 8.5 n-hexane molecules per asymmetric unit. The oxidation of the cyclohexane ring in position 4' in cis configuration is clearly visible from electron density maps (Supplemental Fig. 9).
S-cis-4'-hydroxy praziquantel was used as enantiomerically pure substance. The reciprocal space of 1b•M2 was inspected for the h0l layer, revealing Bragg spots h=2n and l=n (Supplemental Fig.10).
Supplemental Fig. 8. ORTEP diagram with 50% probability of the asymmetric unit of 1b•M2 praziquantel; restraints and constraints applied in the refinement of M2 praziquantel Supplemental Fig. 9. Electron density map Fo of M2 praziquantel (contoured at the 3.70σ level)
The framework is refined using the constraint EADP for some disordered ZnI2 moieties and the restraint SIMU for some carbon atoms. Five ZnI2 moieties are disordered and refined using the disorder model. One guest molecule (A) and two cyclohexane molecules (B and C) were found in the asymmetric unit. The analyte occupancy is 60%. Solvent molecule C is disordered with one ZnI2 moiety and refined using PART 1 command. Some restraints were applied for refinement of the analyte and solvent molecules (DFIX, SADI, FLAT and DANG). The guest molecule was refined by applying SIMU (for complete molecule). The restraints used for the refinement are summarized in Supplemental Fig. 11. The benzene part of the molecule could be refined with minor constraints (AFIX 66) and the presence of oxidation at position 9 of the aromatic ring is clearly visible from the electron density map (Supplemental Fig. 12). The position of hydroxylation had to be restrained to its theoretical values by applying DFIX and DANG commands due to high mobility and not resolved disorder with solvent molecules at this part. Intermolecular hydrogen bonds could be observed by the C-H···O distance (2.17 Å) between the framework and the carbonyl of the analyte. A solvent mask was calculated and DMD-AR-2021-000663 15 418 electrons were found in a volume of 1342 A 3 in one void per asymmetric unit. This is consistent with the presence of 8.5 cyclohexane molecules per asymmetric unit.
In the present study, the task of the crystalline sponge method was the determination of the hydroxy positions after in vitro incubation of the parent compound. The position of metabolism could be determined unambiguously. The combination of known parent structure, mass spectrometry and crystalline sponge analysis leads therefore to a successful structure elucidation. Furthermore, NMR analysis of the metabolite confirms the structure.
S-9-hydroxy praziquantel was isolated from incubation of enantiomerically pure Spraziquantel. The reciprocal space of 1b•M3 was inspected for the h0l layer, clearly revealing Bragg spots h=2n and l=n (Supplemental Fig.13).
Supplemental Fig. 11. ORTEP diagram with 50% probability of the asymmetric unit of 1a•M3 praziquantel; restraints and constraints applied in the refinement of M3 praziquantel

Crystallographic data for 1a•S-7-hydroxy S-praziquantel (M4) (after solvent masking)
Crystal size: 198 × 142 × 89 µm 3 , refined formula: C100.54H94.49I12N25.51O2.27Zn6, formula weight (Mr) = 3611.29 g /mol, colorless needle, crystal system: Monoclinic, space group C2, Z = 4, 24684 unique reflections merged from recorded 54687 ones (2.573° < θ < 67.08°) were used for structural analysis (Rint = 0.0145). Lattice parameters, R-factor on F 2 > 2σ(F 2 ), weighted Rfactor, and goodness-of-fit are as follows: a = 35.2848 (5)  The framework exhibits disorder of five ZnI2 parts refined using the disorder model. One guest molecule (A) with an occupancy of 76% and three additional cyclohexane molecules (C, D and E) can be found in the asymmetric unit. The benzene ring of molecule A was fixed using AFIX 66 and the non-aromatic rings had to be restrained using SADI, DFIX, DANG and FLAT (Supplemental Fig. 14). SIMU was applied for all atoms of the analyte molecule. A solvent mask was calculated and 272 electrons and 50 electrons were found in a volume of 1040 A 3 and 147 A 3 , respectively, in two voids per asymmetric unit. This is consistent with the presence of 5.5 and 1 cyclohexane molecules per asymmetric unit. The electron density map clearly showed the introduction of an additional atom at position 7 of the non-aromatic part of the structure (Supplemental Fig. 15). The distance of the C-O bond had to be restrained using DFIX.
S-7-hydroxy S-praziquantel was isolated from incubation of enantiomerically pure Spraziquantel. The reciprocal space of 1b•M4 was inspected for the h0l layer, clearly revealing Bragg spots h=2n and l=n (Supplemental Fig.16).

Crystallographic data for 1a•S-11b-hydroxy praziquantel and R-11b-hydroxy praziquantel (S-/R-M6)
Crystal size: 339 × 172 × 90 µm 3 , refined formula: C45.46H35.94I6N12.99O1.49Zn3, formula weight (Mr) = 1745.61 g /mol, yellow needle, crystal system: Monoclinic, space group C2/c, Z = 8, 14979 unique reflections merged from recorded 53191 ones (2.610° < θ < 74.592°) were used for structural analysis (Rint = 0.0248). Lattice parameters, R-factor on F 2 > 2σ(F 2 ), weighted Rfactor, and goodness-of-fit are as follows: a = 34.4208 (4)  The framework is refined without restraints. The framework exhibits disorder of three ZnI2 part refined using the disorder model. Two guest molecules (A) and (B) can be found in the asymmetric unit. Guest A is disordered at special position (inversion) and refined using PART -1 command and s.o.f. = 0.5x(FVAR#5). This leads to an analyte occupancy of 40%. An additional guest molecule (B), disordered with one ZnI2 moiety, was observed, and is refined as PART 1 with a fixed occupancy of 30%. AFIX 66 was applied to the benzene rings, and DFIX, SADI, DANG and FLAT were used for some bonds and angles of the guest (Supplemental Fig. 17). The guest molecules were refined by applying SIMU. Additional electron density was observed for both guest molecules in position 11b in either S or R configuration of the non-aromatic ring system and was assigned as hydroxylation (+16 Da) (Supplemental Fig. 18). The introduced oxygens were restrained using the command DFIX (A, B) and DANG (B).
The reciprocal space of 1a•S-/R-M6 was inspected for the h0l layer, revealing the absence of Bragg spots h=2n and l=n (Supplemental Fig.19) and confirming the centrosymmetric space group C2/c.

Crystallographic data for 1a•R-11b-hydroxy praziquantel (R-M6) (after solvent masking)
Crystal size: 236 × 168 × 103 µm 3 , refined formula: C81.5H59.78I12N25O1.5Zn6, formula weight (Mr) = 3328.34 g /mol, light yellow needle, crystal system: Monoclinic, space group C2, Z = 4, 28422 unique reflections merged from recorded 185893 ones (2.575° < θ < 67.078°) were used for structural analysis (Rint = 0.0369). Lattice parameters, R-factor on F 2 > 2σ(F 2 ), weighted Rfactor, and goodness-of-fit are as follows: a = 35.2910 (6)  The framework is refined using the constraint EADP for some disordered ZnI2 moieties and carbon atoms. DFIX and SADI was used for some tpt parts. Five ZnI2 moieties are disordered and refined using the disorder model. One guest molecule (A) can be found in the asymmetric unit near the center of symmetry (C2 axis) and is refined using PART -1 command with s.o.f. = 0.5. The analyte occupancy was modeled to 100%. AFIX 66 was applied to the benzene ring, and SADI and DANG were used for some bonds and angles of the guest (Supplemental Fig. 20). The guest molecule was refined by applying SIMU. A solvent mask was calculated and 422 electrons were found in a volume of 1538 A 3 in one void per asymmetric unit. This is consistent with the presence of 8.5 cyclohexane molecules per asymmetric unit. Additional electron density was observed in position 11b of the non-aromatic ring system and was assigned as hydroxylation (+16 Da) (Supplemental Fig. 21).
The reciprocal space of 1a•R-M6 was inspected for the h0l layer, revealing Bragg spots h=2n and l=n (Supplemental Fig.22).