Allopurinol Reprograms Uric Acid Self-Assembly by Disrupting Cytotoxic Fibrils and Redirecting Crystal Growth

Abstract

The supramolecular self-assembly of uric acid (UA), the end product of purine metabolism, underlies crystal deposition in gout and kidney diseases. However, the intermediate states linking soluble UA to crystalline phases remain poorly defined. Here, we report that UA self-assembles into amyloid-like fibrils that coexist with crystalline forms and exhibit cytotoxicity. Native ion mobility spectrometry–mass spectrometry (IMS–MS) reveals discrete UA oligomers up to 60-mers, suggesting a stepwise assembly process. Optical and electron microscopy distinguish between fibrous and crystalline morphologies, with the fibrillar network acting as a potential scaffold for nucleation. We demonstrate that allopurinol, beyond its known function as a xanthine oxidase inhibitor, directly perturbs UA aggregation. Allopurinol alters the thermodynamics of self-assembly, suppressing fibril formation and promoting crystallization into a more stable anhydrous polymorph. In contrast, epigallocatechin gallate (EGCG) suppresses both fibrillation and crystallization. X-ray diffraction confirms the formation of a distinct anhydrous crystal phase in the presence of allopurinol, analogous to that found in patient-derived deposits. These findings expand the chemical understanding of UA phase behavior and polymorphism and establish cytotoxic UA fibrils as drug-modifiable intermediates. Modulating small-molecule-driven metabolite self-assembly provides a mechanistic basis for rational intervention in gout and other disorders characterized by metabolite aggregation.

Introduction

Uric acid (UA) accumulation and crystallization are central to the pathogenesis of both gout and UA nephrolithiasis, yet they manifest through distinct chemical pathways. Gout, a severe form of arthritis within the family of inborn errors of metabolism (IEMs), is a unique medical condition with a long historical record. It is caused by the accumulation of UA crystals in the joints, affecting approximately 1–2% of the Western population. Dietary causes account for approximately 12% of gout cases, and three genes contribute to high UA levels. UA is a breakdown product of purines and is found in many of the foods we consume. An abnormality in metabolizing UA and crystallization of these compounds in joints can cause attacks of painful arthritis, kidney stones, and blockage of the kidney filtering tubules, leading to kidney failure. Today, treatment is primarily achieved through anti-inflammatory drugs aimed at relieving gout symptoms. The American College of Rheumatology recommends allopurinol, a xanthine oxidase (XOD) inhibitor, as the first-line therapy for gout. By inhibiting XOD, allopurinol prevents the conversion of hypoxanthine to xanthine and xanthine to UAs, two key steps in purine catabolism, thereby reducing serum urate levels and promoting the dissolution of monosodium urate crystals in gout. Despite its widespread use as a frontline XOD inhibitor, allopurinol often falls short; many patients experience adverse effects, resistance, or insufficient response. These limitations underscore the broader challenge of relying solely on UA-lowering strategies to manage the multifaceted pathophysiology of gout. Furthermore, despite UA’s central role in gout, the mechanism by which it aggregates into crystals remains poorly understood. UA crystals can accumulate asymptomatically, making it challenging to connect initial deposition with acute flare-ups.

UA accumulation, while widely recognized as a key contributor to gout, is also implicated in UA kidney stones, Lesch-Nyhan syndrome and glycogen storage disease type I. Several other diseases extend beyond their classic role in crystal-induced arthritis. Hyperuricemia, characterized by elevated levels of UA in the blood, has been increasingly linked to cardiovascular diseases. Studies have shown that UA can promote oxidative stress, inflammation, and endothelial dysfunction, all of which contribute to hypertension, atherosclerosis, and coronary artery disease. − Elevated UA levels are considered both a biomarker and a potential mediator of these conditions, although the exact mechanisms underlying this relationship remain under investigation.

In the realm of metabolic disorders, UA accumulation is associated with type 2 diabetes (T2D) and metabolic syndrome. − Hyperuricemia can exacerbate insulin resistance, a hallmark of T2D, through its pro-inflammatory effects and by impairing endothelial nitric oxide production. Furthermore, UA has been implicated in the development of fatty liver disease by contributing to lipid accumulation and hepatocyte injury. These connections suggest that UA plays a broader role in metabolic dysregulation beyond gout.

Kidney diseases also exhibit a strong association with UA. − Chronic kidney disease (CKD) and UA nephropathy, for instance, highlight the damaging effects of UA on renal function. In CKD, decreased renal clearance of UA leads to its accumulation, which in turn exacerbates renal injury through crystal deposition and inflammatory pathways.

Emerging research also suggests a role for UA in neurological diseases. For example, while some studies indicate that UA’s antioxidant properties might have a protective effect against neurodegenerative conditions like Parkinson’s and Alzheimer’s diseases, , others highlight its potential to promote neuroinflammation under certain pathological conditions. , These seemingly paradoxical roles underscore the complexity of UA’s impact on the nervous system.

Together, these findings highlight that UA accumulation leads not only to classical crystallopathies, such as gout and kidney stones, but also engages broader supramolecular assembly pathways whose outcomes, including fibril formation and polymorph selection, remain underexplored.

Gout, hyperoxaluria, and cystinuria are characterized by the pathological formation of crystals or stones, specifically UA in gout, oxalate in hyperoxaluria, and cystine in cystinuria. These crystalline deposits are hallmark features of these diseases and often lead to severe complications. Our research has demonstrated that precursors to these crystals, such as oxalate and cysteine , (the precursor to cystine), can form amyloid-like fibrils. This finding suggests a potential link between fibrillization and subsequent crystal formation, providing a promising avenue for understanding the molecular mechanisms underlying crystal-related diseases. Although the mechanism by which fibrils may “seed” crystal growth remains unknown, this evidence offers a novel perspective on the fundamental processes of crystal growth and polymorphic selectivity. Should UA be found to utilize fibril assemblies for rapid crystal growth, the concept of “fibrils seeding crystals” would represent more than a mere coincidence.

The discovery that metabolites, including amino acids (Phe, Tyr, Met, Cys, , Trp, Leu/Ile), secondary metabolites (orotic acid, homocysteine, oxalate), lipids (glucosylceramide and sphingolipids), and nucleobases, can form amyloid-like nanostructures has reshaped our understanding of IEMs. This includes adenine, a purine metabolite known to form amyloid-like structures. These assemblies share key physicochemical properties with protein amyloids, such as morphology, dye binding, electron diffraction patterns, immunological responses, and cytotoxicity. Notably, fibril formation has been observed in systems previously thought to form only crystals, such as cysteine/cystine and oxalate. , In these cases, fibrils form before crystals appear. Another example is biogenic guanine crystals in zebrafish eyes, where plate-like crystals grow from crystal leaflets on preassembled protein fibers. These systems highlight the relatively unexplored interplay between fibrils and crystal polymorphs, suggesting that fibril formation, which is kinetically favored, may direct the formation of novel crystal structures. In addition, this provides a new perspective on the metabolostasis network in pathology.

A key effort in recent research is to determine the clinical values of amyloid fibrils, especially those of metabolites. Efforts to develop novel treatments for amyloidogenic diseases have included harnessing the humoral immune response to produce antibodies against amyloid deposits. In the latter half of the twentieth century, humoral immunity was recognized as a double-edged sword: while intravenous immunoglobulins derived from human plasma could protect immunodeficient patients from infections, antibodies targeting autologous molecules, autoantibodies, were implicated in autoimmune diseases affecting the skin, joints, endocrine organs, and nervous system. Interestingly, autoantibodies were also found in the serum of patients with amyloidogenic diseases, specifically targeting amyloid assemblies such as amyloid β-protein fibrils, tau tangles, and prion plaques. − The generation of antibodies against amyloid assemblies has enabled further investigation into amyloid pathology, utilizing techniques such as immunostaining, Western blotting, and immunohistochemistry. This antibody reactivity could be a viable strategy for assessing in vivo formation of UA fibrils and their relevance to disease progression.

Despite advances in electron microscopy, capturing the early stages of crystal or fibril formation remains a challenging task. Our research underscores the crucial role of biophysical approaches in elucidating the interactions between inhibitors and metabolite clusters, which are vital for understanding drug mechanisms and refining therapeutic strategies. We have recently pioneered the use of ion mobility spectrometry-mass spectrometry (IMS-MS) to investigate the self-assembly and coassembly of metabolites into fibrils and crystals. ,,− We preserve weakly bound, noncovalent metabolite clusters for structural analysis by employing gentle electrospray ionization and soft ion transfer. This method has successfully characterized the fibril formation of homocysteine and cysteine/cystine and the crystal formation of adenine-sulfate. Expanding IMS-MS to study drug-inhibitor mechanisms, particularly the interactions of UA with allopurinol and the amyloid inhibitor epigallocatechin gallate (EGCG), is of great importance (see Figure S1 for their chemical structures). Inhibitors may target metabolite monomers or aggregated clusters, with their mechanisms reflected in the dynamics of cluster distributions, which can be retained in the gas phase for detailed analysis.

In this work, we investigate the assemblies and interactions of UA with inhibitors, with a focus on their early stages of aggregation. We test the hypothesis that UA fibril formation is a critical precursor to crystallization and plays a significant catalytic and biological role in the progression of the disease. The high dynamics of cluster assembly, disassembly, and redistribution during metabolite aggregation add complexity to the interpretation of the data. Thus, this study focuses on both the structural and biological characterization of UA aggregation and its interactions with inhibitors. Advancing our understanding of the principles driving fibril-to-crystal transitions and developing technological innovations are interrelated goals of this work.

Materials and Methods

Ion Mobility Spectrometry-Mass Spectrometry

The stock solutions of UA and allopurinol (≥98% purity by HPLC, VWR) were prepared in LC-MS-grade water (1 mg/mL), followed by the addition of 100 μL of a 1:5 NH₄OH:H₂O solution and vortexing to aid dissolution. EGCG (≥95% by HPLC, Sigma-Aldrich) stock solution was reconstituted in ammonium acetate, pH 7. For each sample, further dilution to the desired working concentration (1 mM) was made in ammonium acetate, pH 7. All multifield IMS-MS experiments were performed using an Agilent 6560 IMS-Q-TOF instrument (Agilent Technologies, Santa Clara, CA). The mass calibration and CCS were validated using the Agilent ESI-L tuning mix (diluted in 95:5 v/v ACN:H₂O). The ions were generated by electrospray ionization (ESI) using a dual ESI/Agilent Jet Stream source and a syringe pump at a rate of 30 μL/min. Instrument parameters were tuned based on Gabelica et al. The ions were stored in a source funnel and then pulsed into a 78.1 cm drift cell filled under a drift-gas pressure of 3.94 Torr. The ions then traversed through the drift tube under the influence of a weak electric field and simultaneously collided with the stationary buffer gas (nitrogen). The size, shape, and net charge of the ions determined the velocity at which they drifted through the cell. Data were obtained in positive polarity over 5.2 min with drift cell voltages ΔV = 1490, 1390, 1290, 1190, 1090 V in nitrogen. For each species, the drift velocity is related to the reduced ion mobility K ₀ and used to calculate the momentum-transfer collision integral through the Mason-Schamp equation, which is reported as the experimental CCS. The mass spectra were analyzed on IM-MS Browser B.08.00 software (Agilent, Santa Clara). The ATD and the mass spectra graphs were made using OriginPro. The instrument method parameters are listed in the Supporting Information (Tables S1 and S2). Model structures of various UA “block”-like clusters were built from the crystal data (see below) and geometry optimized using the AVOGRADO software and the MMFF94s force field. The theoretical CCS values were calculated using the trajectory method (TM). ,

An Isotropic Model for the Assembly of Nonglobular Monomers

To reconcile a “nonglobular” monomer with isotropic (sphere-like) n-mer packing, we rescale the monomer CCS by comparing its orientationally averaged projected area to that of a volume-equivalent sphere. For a circular cylinder (“disk”) of radius R and thickness t (ε = t/R), the surface area is S _disk = 2πR ² + 2πRt. By Cauchy’s formula, the mean projected area is ⟨A⟩ = S/4, giving

$${⟨A⟩}_{\mathrm{d}\mathrm{i}\mathrm{s}\mathrm{k}}=\frac{1}{4}{S}_{\mathrm{d}\mathrm{i}\mathrm{s}\mathrm{k}}=\frac{\pi }{2}(R^2+Rt)$$ 1

Let the volume-equivalent sphere satisfy $\frac{4}{3}\pi r_s^3=\pi R^{2}t\Rightarrow r_s^3=\frac{3}{4}{R}^{2}t$ . Its surface area is $S_{\mathrm{sphere}}=4\pi r_s^2$ and $⟨A{⟩}_{\mathrm{sphere}}=\pi r_s^2$ .

The ratio of orientationally averaged areas (and thus the geometric part of CCS in TM approximations) is

$$f(\epsilon )\frac{\mathrm{C}\mathrm{C}{\mathrm{S}}_{\mathrm{s}\mathrm{p}\mathrm{h}\mathrm{e}\mathrm{r}\mathrm{e}(\mathrm{v}\mathrm{o}\mathrm{l}-\mathrm{e}\mathrm{q})}}{\mathrm{C}\mathrm{C}{\mathrm{S}}_{\mathrm{d}\mathrm{i}\mathrm{s}\mathrm{k}}}=\frac{{⟨A⟩}_{\mathrm{s}\mathrm{p}\mathrm{h}\mathrm{e}\mathrm{r}\mathrm{e}}}{{⟨A⟩}_{\mathrm{d}\mathrm{i}\mathrm{s}\mathrm{k}}}=\frac{\pi r_s^2}{\left(\frac{\pi }{2}\right)(R^2+Rt)}=\frac{2r_s^2}{R^2+Rt}=\frac{2{(\frac{3}{4})}^{2/3}{\epsilon }^{2/3}}{1+\epsilon }$$ 2

We therefore use $\mathrm{C}\mathrm{C}{\mathrm{S}}_1^{(\mathrm{s}\mathrm{p}\mathrm{h}\mathrm{e}\mathrm{r}\mathrm{e}-\mathrm{e}\mathrm{q})}=f(\epsilon )\mathrm{C}\mathrm{C}{\mathrm{S}}_1$ in the compact-aggregate scaling.

$$\mathrm{C}\mathrm{C}{\mathrm{S}}_n=f(\epsilon )\mathrm{C}\mathrm{C}{\mathrm{S}}_1\times n^{2/3}$$ 3

Let $\epsilon =\frac{t}{R}=0.25$ for UA, f(0.25) = 0.524, i.e., a ∼48% reduction in monomer CCS.

Powder X-ray Diffraction

The X-ray diffraction pattern was collected using a Bruker D8 Discover diffractometer. The setup employed was a Debye–Scherrer configuration, with a sealed X-ray tube featuring a copper anode (40 kV, 40 mA). A Goebel mirror parallelized the divergent beam, and the detector was a LynxEye XE linear detector. The diffraction patterns were collected between 15° and 40° 2θ with a step size of 0.02° 2θ for 2 s per step. The sample was loaded into a quartz capillary with a 0.7 mm diameter. The capillary was positioned at the center of the goniometer, and it was spinning to minimize the sample’s preferred orientation.

X-ray Crystallography

The single-crystal X-ray diffraction data for UA crystals and allopurinol crystals were measured on a Bruker D8 Venture diffractometer equipped with a Photon 100 detector (Cu Kα radiation, λ=1.54178 Å). Each crystal was coated in Paratone oil and mounted on a MiTeGen loop. Data were collected at 100 K after the crystal was placed in a cold stream of nitrogen generated by an Oxford Cryostream low-temperature apparatus. Data were reduced with Bruker SAINT and corrected for absorption using SADABS. The structure was solved and refined using SHELXT and SHELXL, respectively. All non-hydrogen atoms were refined with anisotropic displacement parameters. Crystal and refinement data are listed in Tables S4–S6.

Transmission Electron and Optical Microscopy

UA (99% purity by HPLC, Sigma-Aldrich) was dissolved at concentrations ranging from 1.0 to 2.0 mg/mL at 90 °C in PBS buffer to obtain a homogeneous monomeric solution. Once fully dissolved, the solution was mixed with 1 mM EGCG or 150 μM allopurinol (both >98% pure by HPLC, Sigma-Aldrich). UA diluted in PBS served as a control for the EGCG-treated sample, while 0.005% DMSO served as a control for the allopurinol-treated sample. The solutions were gradually cooled down and incubated overnight at room temperature to facilitate self-assembly. For the TEM images, samples at the indicated UA concentrations were vortexed, and 10 μL of each was drop-cast onto 400-mesh copper grids (Electron Microscopy Sciences, Hatfield, PA, USA). The grids were incubated at room temperature for 5 min, and excess fluid was absorbed. Samples were visualized using two microscopes, a JEOL 1200EX electron microscope operating at 80 kV and a Talos F200i (S)­TEM equipped with a field emission gun and a Ceta-M Camera. The accelerating voltage was 80 kV, and the beam current was ∼0.5 nA. For the optical images, 2.0 mg/mL UA-treated samples were placed in a 24-well glass-bottom plate (Cellvis, P24-1.5H-N) and visualized using a Leica TCS SP8 confocal microscope with a × 40 1.1 NA water-immersion objective (Leica Microsystems, Wetzlar, Germany). The results displayed are representative of three biological experiments.

Cytotoxicity Assay

HEK293, an immortalized human cell line originally generated by adenoviral transformation of human embryonic kidney cells, was selected as a model due to the kidneys’ central role in the pathogenesis of gout. Notably, gout is a well-documented contributor to kidney failure. Cells were seeded at a density of 2 × 10⁵ cells/mL in Dulbecco’s Modified Eagle Medium (DMEM, high glucose) supplemented with 10% fetal bovine serum (FBS; Biological Industries) and cultured in 96-well tissue culture plates (100 μL per well). Cells were allowed to adhere overnight at 37 °C in a humidified incubator.

Treatment solutions were prepared by dissolving UA in serum-free DMEM at several concentrations: 0.5 mg/mL (2.97 mM), 1.0 mg/mL (5.95 mM), 1.5 mg/mL (8.9 mM), and 2.0 mg/mL (11.9 mM). The dissolving was done by heating to 90 °C, followed by gradual cooling to room temperature. For the 1 mg/mL condition, the solution was either applied directly or fractionated into supernatant and pellet fractions by centrifugation (5,000 × g, 1 min). Medium subjected to the same heating and cooling process without UA was used as a negative control. In parallel, media prepared for tissue culture treatments were dispensed into 96-well plates for visualization of UA crystals at various tested concentrations. The following day, culture medium was replaced with treatment solutions (100 μL per well), and cells were incubated overnight at 37 °C. This short incubation minimized the formation of large UA crystals, ensuring that the treatment medium contained primarily UA monomers, oligomers, fibrils, and small crystals. A time-course TEM analysis confirmed this distribution. Cell viability was assessed using the MTT assay [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide; Sigma] according to the manufacturer’s instructions. Absorbance was measured using a microplate reader at 570 nm with background subtraction at 680 nm. Results represent three independent biological replicates and are expressed as mean ± SD (Figure S2 and Figure ).

4.

Uric acid reduces HEK293 cell viability through fibrils and crystals. (A) MTT viability assay in HEK293 cells shows a significant reduction in viability in unfractionated media containing UA fibrils and crystals. Separation of fibrils (supernatant) from crystals (pellet) by centrifugation also reduced viability, though to a lesser extent. Data represent one representative experiment out of three biological repeats; error bars denote SD. Student’s t test, **p < 0.01, *p < 0.05. (B) Representative light microscopy images of media alone, unfractionated UA (1 mg/mL), supernatant, and pellet fractions. Crystals are evident in the unfractionated and pellet fractions. Scale bar: 1 mm. (C) Light microscopy images of media alone, UA (1 mg/mL), and UA (2 mg/mL) after 24, 48, and 72 h (hrs) of incubation at room temperature. Crystals are evident in UA samples, and their abundance markedly increases over time. Scale bar: 1 mm.

Quantitative Analysis of Fibril Morphologies

Fibril widths were obtained from ImageJ measurements. Line selections were drawn orthogonal to the local fiber axis. All analyses were performed in Python.

For each condition (UA alone and UA with allopurinol), we computed descriptive statistics including sample size, mean, standard deviation (SD), standard error (SE), median, normalized median absolute deviation (MAD), quartiles (Q1, Q3), interquartile range (IQR), minimum, maximum, and coefficient of variation (CV). Distributional assumptions were assessed using the Shapiro–Wilk test for normality and Levene’s test (median-centered) for homoscedasticity. Because the data were not guaranteed to be normal or homoscedastic, between-group comparisons included both Welch’s t test (robust to unequal variances) for mean differences and the Mann–Whitney U test (two-sided) for distributional/median differences. Effect sizes were summarized as Hedges’ g (small-sample corrected Cohen’s d). A nonparametric bootstrap was used to obtain percentile-based 95% confidence intervals (CI) for mean differences, while the Hodges–Lehmann estimator provided a robust estimate of the median difference. Unless otherwise stated, all tests were two-sided with α = 0.05, and no multiple-testing correction was required for the single primary comparison.

Results and Discussion

IMS-MS Probing of Uric Acid Cluster Formation Correlated with Nanostructure Assemblies Observed by TEM

Figure A shows representative ESI mass spectra of UA in both positive (top) and negative (bottom panel) polarity modes. The concentration of UA in IMS-MS experiments was reduced by ∼60% compared to the lowest concentration typically employed in other experiments (>0.5 mg/mL) to prevent clogging of the nebulizer, saturation of the detector, and to slow down the assembly process. For context, 0.5 mg/mL UA corresponds to ∼3.0 mM. Amyloid proteins are typically analyzed at ∼1–10 μM monomer concentration. When normalized per residue, 3.0 mM UA is equivalent to ∼50 μM of a 60-residue polypeptide (since 50 μM × 60 residues ≈ 3 mM in “residue equivalents”). In our MS experiments, UA was analyzed at concentrations ranging from 500 μM to 1 mM, corresponding to ∼8–17 μM of the 60-residue polypeptide. These concentrations, therefore, fall within a reasonable building-block range while remaining sufficiently low to minimize saturation effects and reduce ESI artifacts. Mass spectral peaks with distinct isotopic spacing are observed and tentatively assigned to nominal cluster (oligomer) size-to-charge ratios (n/z). More and higher-intensity peaks were observed in the high m/z region when mass spectra were collected in negative mode than in positive mode.

1.

Early stage of UA self-assembly probed by IMS-MS. (A) Mass spectra of UA in positive (top) and negative (bottom) modes of early UA oligomers (clusters). Each mass spectral peak is annotated with a nominal cluster size (n) to charge (z) ratio. MS peaks with n/z colored in red are those with a mixture of clusters at different charge states. Multiple clusters may share the same n/z ratio. (B) A plot of experimental cross-sections as a function of cluster size. Post-IM dissociation of UA clusters. (C) Illustrations showing neutral losses in clusters with the same charge state and charged losses in clusters with different charge states. These clusters travel through the drift cell at velocities proportional to their size and charge but are detected at the same mass-to-charge (m/z) ratio. (D) Representative IM profiles of UA clusters. 2D plots of drift time vs m/z for selected clusters observed at monoisotopic m/z 169 and m/z 1345.

As discussed in our previous work on metabolite assemblies, drift-tube (DT)-IMS-MS offers a gentle platform for preserving noncovalent assemblies for IM measurements. ,, Our DT-IMS-MS configuration enables ions to quickly enter the drift cell for IM measurements, thereby minimizing the transition through ion optics and guides that could potentially disrupt their structures. Nonetheless, post-IM dissociation inevitably occurs after the ions exit the drift tube and before they reach the detector, leading to very complex mobilograms. In general, ion mobilograms (or arrival time distributions; ATDs) of small clusters are broad and often contain multiple features corresponding to the dissociation of larger clusters. On the other hand, large clusters, which are more structurally stable, will exhibit ATDs with well-resolved peaks. The analysis of post-IM mobility data relies on the fact that precursor and fragment ions will have the exact drift times since dissociation occurs after the precursor ions exit the drift tube. Furthermore, neutral losses result in IM features with longer-than-expected arrival times, whereas charged losses produce IM features with shorter-than-expected arrival times (Figure C,D). Thus, by aligning the drift times of the ATDs of consecutive clusters, we can systematically identify and assign cluster sizes and charge states to most of the IM features. ,

Figure B shows a summary of experimental CCS values as a function of cluster size (see also Table S3). The experimental CCS values were compared to two models (Table S4). The first model, represented by the dashed line, is based on UA clusters constructed from X-ray crystallographic data (see the following section and representative structures and CCS values in Figure S3). The second model is a “contracted” isotropic model. In the classical isotropic model, the cross-section of an oligomer of size n is predicted from the monomer cross-section; CCS_n = CCS₁ × n ^2/3. , This model assumes the monomer is spherical, which works well for peptides and proteins. However, small molecules tend to pack more tightly than globular proteins. As a result, the classical isotropic model often overestimates the CCS of small-molecule clusters. , To address this, the “contracted” isotropic model reduces the CCS of the monomer by ∼50% (from 128 Ų to 68.5 Ų) and uses this adjusted value to approximate the lower limit of the CCS values for small-molecule clusters (as detailed in eqs -). The overall CCS data reveal that UA clusters are disordered when n <10. Above this range, the cluster growth follows a linear pattern in CCS, indicating assembly into ordered structures. , Notably, there is a sharp increase in CCS as the charge state rises from z = −1 to −2, but the change becomes more gradual with further increases to z = −3 and −4. This pattern suggests that the nucleation size of UA clusters likely falls between n = 5 and 10, with clusters becoming more stable beyond n = 10.

When oligomer shapes (CCSs) are smaller than isotropic predictions suggest, it indicates that the monomer units are more tightly packed than expected. This may result from initial stabilization through unfavorable interactions in the gas phase, such as hydrogen bonding and solvation, leading to partial collapse of the assemblies upon desolvation. This consideration becomes particularly important in the next section, where we examine the behavior of UA clusters in the presence of allopurinol.

Uric Acid Assembles into Fibrils and Crystals

UA crystals, including those from in vitro and from patient samples, exist in anhydrous, monohydrate, or dihydrate forms. , Here, we use freeze-dried samples of self-assembled UA, which typically give rise to a microcrystalline powder of the anhydrous phase. This powder was analyzed using powder X-ray diffraction (PXRD) to confirm the atomic structure. The refined experimental data showed good agreement with the calculated pattern using the reported structure in Table S4 (Figure A), with only minor differences in cell parameters: a = 6.192 ± 2 Å, b = 7.407 ± 3 Å, c = 13.081 ± 5 Å, and β = 90.436 ± 8°. This arrangement is consistent with the previously reported crystal structure of UA without hydration. As reported, the structure consists of two layers of UA chains, oriented parallel to the (210) and (21̅0) crystallographic planes, respectively. The two chain families enclose an angle of approximately 60° along the c-axis (Figure B–C). Within each chain, UA molecules are stabilized by N–H···O hydrogen bonds, whereas adjacent chains are associated through π–π stacking interactions. The interconnection between the two chain families is mediated by additional hydrogen bonds. Notably, the molecular packing of UA is consistent with the development of acicular crystal morphologies and the subsequent formation of fibrillar assemblies.

2.

PXRD data of UA showing parallel β-sheet structures. (A) X-ray diffraction of UA fibrils (red line) and the fitted pattern (blue line) according to the reported structure. (B) Parallel β-sheet-like layer secondary structure of UA fibrils. (C) Crystal structure of UA by a parallel set of head-to-tail H-bonded chains along the b axes.

TEM and optical imaging revealed that UA forms both fibrillar and crystalline structures, depending on concentration (Figures and ). At concentrations of 0.5–1.5 mg/mL, fibrillar networks predominated (Figure ), whereas at 2.0 mg/mL, large crystals were observed (Figure ). Fibril widths were quantified from n = 190 measurements. The distribution was approximately symmetric with a slight right skew (skewness = 0.36) and near-mesokurtic shape. Central tendency and spread were as follows: median = 12.08, mean = 12.36 ± 2.21 (SD), IQR = 2.83 (Q1 = 10.77; Q3 = 13.60), and 95% CI for the mean = 12.04–12.67. The range was 7.0–19.1, with two Tukey outliers. These values indicate a relatively homogeneous distribution of widths with limited outliers. Notably, the observed widths are on the same order of magnitude as the canonical 10–20 nm typically reported for amyloid fibrils. Overall, the measured widths are consistent with mature amyloid morphologies.

3.

Uric acid assembles into fibrils at low concentrations. Representative TEM images of UA fibrils at (A) 0.5 mg/mL, (B) 1.0 mg/mL, and (C) 1.5 mg/mL. Statistical characterization of fibril widths observed at 0.5 mg/mL. (D) Histogram with kernel density estimation (KDE), which provides a smoothed curve showing the overall distribution shape; multiple peaks suggest possible subpopulations. (E) Empirical cumulative distribution function (ECDF), showing the proportion of fibrils below each measured width and providing a complete view of distribution spread for cross-condition comparisons. (F) Normal Q–Q plot comparing observed quantiles with those expected for a normal distribution; data falling near the diagonal indicate approximate normality, while systematic curvature or tail deviations reveal skewness or heavy tails.

7.

Allopurinol and EGCG differentially remodel UA fibrillization and crystallization. Representative transmission electron microscopy (TEM) images (A–D) and optical microscopy images (E–H) show the effects of EGCG and allopurinol on UA aggregation. Statistical analyses of UA fibrils observed in the pure samples and in the presence of allopurinol (I–J). Panels A and E show UA fibrils and crystals formed under control conditions (1 mg/mL UA in PBS); panels B and F show UA with EGCG (1 mg/mL UA + EGCG); panels C and G show UA with DMSO vehicle control (0.005% DMSO); and panels D and H show UA with allopurinol (1 mg/mL UA + allopurinol). TEM images reveal that EGCG-treated samples produce short, clumped fibrils, whereas allopurinol-treated samples yield longer, coarser fibrils. Optical images show that EGCG suppresses crystal formation, leading to fewer but larger, more transparent crystals, whereas allopurinol promotes the formation of flat, elongated crystals. Scale bars: 500 nm (TEM), 50 μm (optical microscopy). (I) Violin and box plots of fibril widths. Distribution of fibril widths measured from three independent data sets of pure UA fibrils and three independent data sets of UA fibrils in the presence of allopurinol. Violin plots depict the full distribution of widths, while overlaid box plots indicate the interquartile range and median. (J) ECDF curves for all six data sets, showing the cumulative fraction of fibrils below a given width. A consistent rightward shift is observed for UA fibrils in the presence of allopurinol relative to pure UA fibrils, indicating thicker fibrils across the entire distribution.

These results suggest that fibrils are the kinetically favored product under conditions where crystal nucleation is rare. In contrast, crystals represent the thermodynamically stable product that emerges at higher supersaturation or longer times. Importantly, we do not suggest that the activation barrier itself decreases with reaction progress; instead, the probability of crossing the nucleation barrier increases with supersaturation, allowing crystals to form.

Thus, fibrillation and crystallization should not be viewed as mutually exclusive or strictly sequential, but rather as parallel yet coupled pathways. Fibrils may persist and coexist with crystals, as observed in our samples, and their morphology can bias subsequent crystal growth. Crystallization can also occur independently under favorable conditions.

While these results suggest a functional connection between UA fibrils and crystals, they do not provide direct visualization of fibrils converting into crystals. The most consistent interpretation is that fibrils act as scaffolds or catalyst-like assemblies that bias nucleation and polymorph selection without necessarily being consumed. This model explains their coexistence with crystals at equilibrium across different concentrations. If disrupting fibril formation alters crystal formation, it would further support this link. We can gain insight into this process by examining how other small molecules perturb UA fibril formation and crystallization.

Uric Acid Fibrils are Cytotoxic to HEK293 Cells

Previous studies have shown that amyloid-like structures formed by metabolites exhibit cytotoxic effects. To evaluate the cytotoxicity of UA fibrils, cultured HEK293 cells were exposed to these assemblies. The fibrils were generated by dissolving various concentrations of UA in culture medium and then gradually cooling the solution. Cells were incubated with fibril-containing media for 24 h, while media lacking fibrils served as controls. Cytotoxicity was assessed using the MTT assay. The results showed a dose-dependent decrease in cell viability with increasing concentrations of UA assemblies (Figure S4). Because crystal formation becomes unavoidable at high UA concentrations (>1.0 mg/mL), the treatment samples were separated into fractions containing either fibrils or crystals. Subsequent MTT assays revealed that both UA fibrils and crystals are cytotoxic to HEK293 cells (Figure A,B). In addition, a time-course analysis of UA in the media supports the dynamic transition to crystals (Figure C).

Allopurinol “Remodels” Uric Acid Assembly

The robust formation of UA clusters observed by IMS-MS, along with the detection of fibrillar structures by TEM, underscores the need to elucidate how drugs such as allopurinol modulate UA assembly pathways. Allopurinol is the first-line therapy for gout, primarily inhibiting UA formation in the purine metabolic pathway. However, UA deposits likely already exist at the initial diagnosis, making it crucial to understand the interactions between UA aggregates and allopurinol to develop more effective treatments for UA-related disorders.

We prepared an equimolar mixture of UA and allopurinol to investigate allopurinol’s inhibitory mechanism. We analyzed it alongside pure UA using IMS-MS immediately after mixing and at t = 7 days. The mass spectra showed no obvious binding or interaction between allopurinol and small UA clusters (Figure A). Only mass spectral peaks corresponding to the monomer and dimer of allopurinol were observed alongside UA oligomer peaks.

5.

Allopurinol exhibits a unique mode of action. It may remodel the assembly of UA clusters, leading to the observation of new features in the ATDs. Mass spectrum of (A) UA and (B) UA with allopurinol in positive mode showing the presence of allopurinol without significant changes in the intensities of small UA cluster peaks. The inset shows a comparison of IMS-MS data for pure UA and UA with allopurinol. The appearance of new IM features at longer arrival times suggests the formation of new structure types that are otherwise suppressed in pure UA. (C) For comparison, EGCG disrupts the formation of UA clusters.

The ATDs of UA peaks in the absence of allopurinol exhibit multiple IM features, corresponding to a range of clusters of varying sizes that dissociate after exiting the drift cell (as discussed in Figure ). The ATDs for small cluster size-to-charge ratios (e.g., n/z 1/1) are broad, whereas they become well resolved as the cluster size increases. These data suggest that smaller clusters are more prone to disassemble in the gas phase, while larger clusters exhibit greater stability. Although this behavior is anticipated, its clear representation in the ATDs is noteworthy.

In the presence of allopurinol, the ATDs displayed complex features at longer drift times (see the inset of Figures and S5), revealing a distinct mode of action for this inhibitor. This observation is unprecedented, as it has not been reported in any earlier studies on amyloid inhibition. The absence of mass spectral peaks corresponding to UA-allopurinol heterocomplexes suggests that these oligomers are either unstable and therefore not detected at distinct m/z values or that their interactions occur at cluster sizes beyond our detectable mass range. However, if these interactions were weak, we would not have observed such major changes in UA’s ATDs. We propose that allopurinol remodels UA self-assembly by interacting with very high-order UA clusters.

Further evidence is provided by a detailed analysis of Figure , which presents the ATDs of the major peaks with nominal n/z values ranging from 1/1 to 10/1. Because these are nominal values, clusters with different charge states may appear at the same m/z, and post-IM dissociation could result in clusters of varying sizes being detected at the same m/z. Additionally, UA can form protomers, as indicated by the two IM features observed in the ATD of n/z = 1/1, making the ATDs of higher-order clusters more complex.

6.

(A) Allopurinol exhibits a distinct mode of action, potentially weakening (“redissolving”) UA clusters and leading to the emergence of unique species in the ATDs of UA clusters. These features arise from the post-IM dissociation of large UA clusters, in which the preferred dissociation products depend on initial cluster size but generally favor large clusters because of their initial abundance. This is evident in the drift times of the red-marked features, which remain unchanged with increasing cluster size, suggesting they originate from post-IM dissociation of the same clusters. In contrast, the drift times of early UA cluster features (gray) increase as cluster size grows, following expected trends. Additionally, in the ATD of n/z = 1/1, the protomer of protonated UA (green) appears alongside features corresponding to post-IM dissociated clusters (blue). (B–C) The increase in intensity of features at long arrival times (40–50 ms) in the presence of allopurinol is statistically significant. Examples of the features at 47.76 and 50.45 ms are shown for the ATDs with a nominal n/z ratio of 10/1. *** p < 0.001, n = 5–12.

Beginning at a nominal n/z ratio of 5/1, the arrival time distributions exhibit two distinct features corresponding to charge states z = 2 and z = 1. The z = 2 species are consistently observed regardless of allopurinol, whereas the z = 1 species appear only in its presence. For example, at nominal n/z = 9/1, the z = 2 peak corresponds to n = 18, while a new feature at longer arrival times (highlighted in red in Figure A) is assigned to z = 1. As expected, the drift times of the z = 2 series (e.g., n/z = 6/2, 7/2, 8/2, ...) increase progressively with cluster size, consistent with larger oligomers requiring more time to traverse the drift cell. In contrast, the drift times of the z = 1 features remain constant as nominal n/z increases from 3/1 to 11/1. This behavior rules out the possibility that these novel signals originate from small z = 1 oligomers isobaric with the z = 2 series (e.g., n/z = 9/1 vs n/z = 18/2).

We identified three major features of this class of z = 1 species, which become more distinct and resolved as cluster size increases, as post-IM dissociation products. At n = 6, the leftmost feature is dominant, but at n = 9 and 10, the preference shifts progressively to the middle and rightmost features. What could explain this trend? Allopurinol may interact with large UA clusters, weakening their assembly. These heterocomplexes undergo post-IM dissociation in the gas phase, where the distribution of product species depends on the size and structure of the initial hetero-oligomers. Larger clusters are more likely to dissociate than smaller ones, explaining why the z = 1 species become more intense and well-resolved as cluster size increases.

To estimate the oligomer size (n) associated with the slow-mobility features, we used an n = 20 oligomer at z = 2, which exhibits K ₀ = 0.825 cm² V^–1s^–1 and Ω = 510 Ų, as a reference. Because mobility scales inversely with CCS according to the Mason-Schamp equation, the slowest feature at K ₀ = 0.570 cm² V^–1s^–1 corresponds to an effective CCS of 713 Ų. If oligomer growth is compact and roughly spherical, where CCS scales with n ^2/3, the observed CCS corresponds to an oligomer of ∼36 subunits. In contrast, if the assemblies extend in a rod-like fashion, where CCS scales approximately linearly with n, the same CCS corresponds to ∼29 subunits. Taken together, the observed mobility is consistent with oligomers of 29–36 subunits, with n = 29 representing the strict minimum size.

This effect is akin to allopurinol “(re)­dissolving” large UA clusters that would otherwise form fibrils or the dihydrate crystal polymorph. Such a “destabilization” effect of allopurinol on UA clusters may induce the formation of a different assembly. This hypothesis will be further examined in light of the X-ray crystallography data discussed below.

In summary, IMS-MS enables direct observation of UA oligomers and their remodeling in response to small molecules. In our measurements, we detected UA clusters spanning from dimers to 60-mers, including both low-order oligomers and higher-order aggregates that exhibit defined, noncovalent architectures. The ATDs and mobility data reveal that these assemblies are not simple desolvated remnants of solution-phase species, but structurally informative intermediates that reflect different stabilization mechanisms. Putative hydrated UA clusters, such as those involved in the formation of UA dihydrate (UAD) crystals, could undergo substantial structural collapse upon desolvation due to the loss of water-mediated hydrogen bonds. These typically reorganize into compact structures with smaller CCS values. In contrast, UA anhydrous (UAA) assemblies, stabilized predominantly through direct intermolecular interactions such as hydrogen bonding and π-stacking, exhibit greater kinetic stability in the gas phase and maintain extended conformations. This results in larger CCS values that preserve their supramolecular architecture and render them detectable by IMS-MS.

In parallel, to further validate that IMS-MS reveals a unique inhibitory mechanism for allopurinol, we examined the effect of the canonical amyloid inhibitor EGCG. EGCG is known to interact with small clusters, preventing their assembly into larger structures. Typically, EGCG and amyloid interactions form stable heterocomplexes observable in the gas phase. In our experiments, mixtures of UA and EGCG exhibited a decrease in the mass spectral peaks of UA clusters (Figure C). At t = 0, EGCG likely quenches the signals from UA clusters, but by t = 6 days, no small UA clusters are detected, even though EGCG signals have vanished. Thus, it is reasonable to conclude that EGCG interacts with early oligomers of UA and prevents them from nucleating into stable clusters.

The distinct differences observed in the data for allopurinol and EGCG demonstrate that IMS-MS is both powerful and sensitive to the differing interaction mechanisms of these inhibitors with UA assemblies. We posit that allopurinol may disrupt UA fibril formation and potentially facilitate the reorganization or accelerated formation of new UA clusters. In contrast, EGCG appears to act in a more classical manner, targeting early UA oligomerization to prevent fibril formation. To support these predictions, we used TEM and optical microscopy to image UA in the presence of both inhibitors under various conditions (Figure ). Additional optical images are shown in Figure S6. The addition of EGCG reduced the presence of typical fibrils (Figure A-B), similar to previous work on other metabolites. The fibrils in the presence of EGCG are clumpy and much shorter. EGCG also has a strong effect on UA crystallization (Figure E–H). Fewer, but broader and more transparent crystals were observed.

Since allopurinol dissolves in DMSO, to avoid any solvent effects, as previously demonstrated, its effect was compared to a solution containing the same concentration of DMSO (Figure C). TEM and optical imaging corroborate the IMS–MS findings, showing that allopurinol remodels both fibril morphology and crystal polymorph (Figure C−D, G−H). Specifically, we performed statistical analysis (Figure I–J) on three different TEM images from each fibril group. Across all three pure UA fibril data sets, pairwise mean differences in fibril width fell entirely within the ±3 nm equivalence margin (Welch 90% CIs contained in [−3, +3] nm), indicating that fibril widths are statistically equivalent across replicates. Similarly, all three UA fibrils in the presence of allopurinol data sets were mutually equivalent within ±3 nm by Welch 90% CIs.

When pooled, pure UA fibrils were significantly thinner than UA fibrils in the presence of allopurinol. Normal Q–Q plots indicated mild deviations from normality; therefore, both parametric (Welch) and nonparametric (Mann–Whitney) approaches were reported, and both consistently demonstrate that pure UA fibrils are thinner than UA fibrils in the presence of allopurinol. Welch’s t test gave t = −41.57, p < 0.001, with a mean difference of −11.61 nm [95% CI −12.16, −11.06]. A nonparametric Mann–Whitney U test confirmed this result (U = 736, p < 0.001). Bootstrap resampling yielded a mean difference CI of [−12.17, −11.07] nm, further supporting the robustness of this conclusion.

Complementing these results, ECDFs highlight a systematic right-shift for UA fibrils in the presence of allopurinol relative to pure UA fibrils across the entire width distribution. A two-sample Kolmogorov–Smirnov test confirmed significant distributional differences (KS = 0.933, p < 0.001). These ECDF findings reinforce that the presence of allopurinol not only increases the mean fibril width but also shifts the entire width distribution toward thicker fibrils.

These microscopy data further support the idea that fibril and crystal formation are coupled processes, with changes in self-assembly pathways, detected at the cluster level by IMS–MS, manifesting in both fibrillar and crystalline outcomes.

Crystallization of Uric Acid in the Presence of Allopurinol Reveals Structural Remodeling

We further elucidate the UA crystal polymorphs in detail. Single crystals of UA were obtained in both the absence and the presence of allopurinol, revealing distinct crystallization behaviors. In UA-allopurinol mixtures, two separate crystal types were consistently observed: UA crystals and allopurinol crystals, confirming that the two compounds crystallized independently. Notably, UA crystals formed from the pure UA samples or the UA with EGCG samples were exclusively UAD, while those from allopurinol-containing samples were exclusively UAA. This conclusion is based on X-ray diffraction analysis of multiple crystals from each sample. These findings suggest that allopurinol has a major effect on UA crystallization, accelerating the transformation of UAD to a more thermodynamically stable UAA.

Notably, Grases et al. reported that UAD crystals transform into UAA when in contact with liquid. The transition is accompanied by the formation of hexagonal, bulky UAA crystals and the appearance of cracks in UAD crystals. They also suggested that UAA was the more thermodynamically stable crystal form. This was supported by Liu et al., who showed that the UAA crystals were substantially harder and more brittle than UAD. Presores and Swift suggested that UAD is metastable under physiological conditions and converts to UAA. Through seeding and det experiments, they identified UAD dissolution as the rate-limiting step in this transformation. Nonetheless, while the transformation from UAD to UAA is established, the mechanism remains unclear. Our data suggest that allopurinol assisted the dissolution of UA clusters, accelerating the formation of UAA and bypassing UAD.

Notably, unlike previous studies, which prepared UAD crystals and then studied the transformation of UAD to UAA, our experiments began with a mixture of soluble UA and allopurinol, with UA alone serving as a control. We find that the crystallization of UA in a UA-allopurinol mixture (weeks to months) is significantly slower than that of UA alone (days). We also suspect that allopurinol competes with water for hydrogen bonding with UA, as well as possible π-π stacking. Such interactions could prevent UAD from crystallizing, leading to the formation of UAA. This agrees with our IMS-MS data, which show the presence of smaller but distinct UA clusters in the presence of allopurinol. The interaction of UA and allopurinol is not strong enough for heterocomplexes to be observed by IMS-MS, as well as for cocrystallization to be seen with X-ray crystallography.

Together, these findings also suggest a possible inverse relationship between fibril stability and the stability of the crystal polymorphs that ultimately form. In pure UA, we observed relatively uniform fibrils with widths of ∼10–12 nm, consistent with stable amyloid-like assemblies that predominantly formed the metastable dihydrate polymorph. In contrast, the presence of allopurinol produced fibrils that were thicker and more heterogeneous in width, indicating reduced stability; yet, these conditions favored the eventual formation of the thermodynamically stable anhydrous crystals. These data raise the intriguing possibility that more stable fibrils bias the system toward metastable crystal states. In contrast, less stable fibrils leave the system kinetically free to reorganize into more stable polymorphs. Although further time-resolved studies are needed, this model provides a structural framework for understanding how fibril stability influences polymorphic outcomes in UA crystallization. The structural transformation directly induced by allopurinol in UAA may have substantial implications for the physicochemical properties and pathological behavior of UA deposits in patients with gout.

Conclusion

Our study reveals a striking new dimension to UA biology: UA can self-assemble into cytotoxic, amyloid-like fibrils, mirroring the behavior of cystine and oxalate, two other metabolites that are linked to crystal-related diseases. Using IMS-MS, we directly captured UA oligomers up to 60-mers, uncovering a rich spectrum of intermediate assemblies. We discovered that allopurinol, a frontline treatment for gout, not only influences crystallization but also reshapes the molecular assembly landscape. By establishing fibril formation as a drug-responsive intermediate in UA aggregation, our work reframes the pathogenesis of gout and expands the therapeutic landscape. IMS-MS proves to be a valuable tool not only for dissecting assembly pathways but also for revealing molecular mechanisms of small-molecule modulators. This framework not only advances our understanding of UA pathophysiology but also opens new therapeutic avenues to intercept pathogenic crystallization at its molecular origin. Our data further raise the possibility that fibril stability dictates crystal outcomes: stable, uniform fibrils in pure UA bias toward the less stable UAD polymorph, whereas less stable, heterogeneous fibrils formed with allopurinol ultimately yield the more stable UAA crystals.

Our results also underscore the power of IMS-MS as an analytical platform for probing molecular self-assembly and its modulation by small molecules. Beyond resolving the size and shape distributions of noncovalent UA oligomers, IMS-MS captured the conformational remodeling of these clusters in response to allopurinol. The differential behavior of hydrated versus anhydrous assemblies, where the former collapse upon desolvation and the latter persist due to strong intermolecular stabilization, provides structural insight into the supramolecular logic that underpins UA aggregation. IMS-MS thus emerges not only as a discovery tool for identifying fibrillar and oligomeric intermediates but also as a means to infer their stability, hydration compatibility, and crystal-forming potential.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacsau.5c00812.

• IMS-MS instrument parameters (Table S1–S2); experimental and theoretical CCS of UA clusters (Tables S3–S4); crystallographic data collection and refinement statistics (Tables S5–S7); chemical structures (Figure S1); additional TEM images of UA fibrils and crystals (Figure S2); representative model structures of UA clusters (Figure S3); toxicity of UA aggregates on HEK293 cells (Figure S4); additional ATDs of UA with and without allopurinol (Figure S5); additional optical images of UA with and without inhibitors (Figure S6) (PDF)

T.D.D., E.G., and D.L.B.-Y. conceived, designed, and supervised the project. D.S.O. and H.A.S. collected and analyzed MS data. H.A.S. obtained the single-crystal X-ray diffraction data. H.A. and O.S.T performed TEM and optical imaging. I.S.-B., D.L., D.Z., and D.L.B.-Y. conducted cell viability and PXRD experiments. D.S.O., H.A.S., T.D.D., and D.L.B.-Y. wrote the manuscript with input from all coauthors. CRediT: Dana Laor Bar-Yosef conceptualization, investigation, writing - original draft, writing - review & editing; Damilola S. Oluwatoba data curation, formal analysis, investigation, writing - original draft; Hanaa Adsi formal analysis, investigation; Happy Abena Safoah data curation, formal analysis, investigation; Om Shanker Tiwari investigation; Ilana Sogolovsky-Bard investigation; Dor Zaguri investigation; Davide Levy investigation; Ehud Gazit conceptualization, project administration, resources, supervision, writing - review & editing; Thanh D Do conceptualization, formal analysis, project administration, resources, supervision, writing - original draft, writing - review & editing.

The authors declare no competing financial interest.