Disrupting Crystal Growth Through Molecular Recognition: Designer Therapies for Kidney Stone Prevention

CONSPECTUS:

Aberrant crystallization within the human body can lead to several disease states or adverse outcomes, yet much remains to be understood about the critical stages leading to these events, which can include crystal nucleation and growth, crystal aggregation, and adhesion of crystals to cells. Kidney stones – aggregates of single crystals with physiological origins – are particularly illustrative of pathological crystallization, with 10% of the U.S. population experiencing at least one stone occurrence in their lifetimes. The human record of kidney stones is more than two thousand years old, noted by Hippocrates in his renowned oath and much later by Robert Hooke in his treatise Micrographia. William Hyde Wollaston – physician, chemist, physicist, crystallographer – was fascinated with stones, leading him to discover an unusual stone that he described in 1810 as cystine oxide, later corrected to cystine. Despite this long history, however, fundamental understanding of the stages of stone formation, as well as the rational design of therapies for stone prevention, has remained elusive.

This Account reviews discoveries and advances from our laboratories that have unraveled the complex crystal growth mechanisms of L-cystine, which forms L-cystine kidney stones in more than 20,000 individuals in the U.S. alone. Although L-cystine stones affect fewer individuals than common calcium oxalate stones, they are usually larger, recur more frequently, and are more likely to cause chronic kidney disease. Real-time in situ atomic force microscopy (AFM) reveals that crystal growth of hexagonal L-cystine is characterized by a complex mechanism in which six interlaced anisotropic spirals grow synchronously, emanating from a single screw dislocation to generate a micromorphology with the appearance of stacked hexagonal islands. In contrast, proximal heterochiral dislocations produce features that appear to be spirals but actually are closed loops, akin to a Frank-Read source. These unusual and aesthetic growth patterns can be explained by coincidence of the dislocation Burgers vector and the crystallographic 6₁ screw axis. Inhibiting L-cystine crystal growth is key to preventing stone formation. Decades of studies of “tailor-made additives” – imposter molecules that closely resemble the solute and bind to crystal faces through molecular recognition – have demonstrated their effects on crystal properties such as morphology and polymorphism. The ability to visualize crystal growth in real time by AFM enables quantitative measurements of step velocities and, by extension, the effect of prospective inhibitors on growth rates, which can then be used to deduce inhibition mechanisms. Investigations with a wide range of prospective inhibitors revealed the importance of precise molecular recognition for binding of L-cystine imposters to crystal sites, which results in step pinning and inhibition of step advancement, as well as growth of bulk crystals. Moreover, select inhibitors of crystal growth, measured in vitro, reduce or eliminate stone formation in knockout mouse models of cystinuria, promising a new pathway to L-cystine stone prevention. These observations have wide-ranging implications for the design of therapies based on tailor-made additives for diseases associated with aberrant pathological crystallization, from disease-related stones to “xenostones” that form in vivo because of crystal growth of low-solubility therapeutic agents such as antiretroviral agents.

Graphical Abstract

INTRODUCTION

The human record of kidney stones dates to Hippocrates (460–370 BC), who in his renowned oath stated, “I will not use the knife, not even on sufferers from stone, but will withdraw in favor of such men as are engaged in this work.” Using a compound microscope fashioned after Galileo’s to examine minute bodies, Robert Hooke remarked in Micrographia,⁵ “How great an advantage it would be to such as troubled with the stone, to find some menstruum that might dissolve them without hurting the bladder.” But this attribution of stones to the bladder was repeated, mistakenly, in 1810 by William Hyde Wollaston in his essay On Cystic Oxide, a New Species of Urinary Calculus.^6,7 Wollaston collected kidney stones and received some unusual specimens from local physicians and surgeons, one an egg-shaped stone weighing 270 “grains” (18 grams), extracted from a 36-year-old man (Figure 1). The stone was 3 cm 4 cm (!), compact, yellowish, and glistening, “appearing as one mass confusedly crystallized throughout its substance.” From its solubility, appearance after combustion, peculiar odor under a blowpipe, and component hexagonal crystals, Wollaston recognized this stone as different from others. Believing the stone originated from the bladder, he named it cystic oxide, after the Greek Kýsti, for bladder. Later, Alexander Marcet correctly assigned the origin of Wollaston’s cystine oxide stone to the kidney.⁸ Rather than renaming the stone nephritic oxide, Marcet argued against changing established nomenclature. In 1833, Berzelius corrected the chemical record, renaming the material cystine. Consequently, cystine, and its amino acid reduction product, cysteine, owe their names to a misnomer that persists to this day. Marcet also noticed a familial connection among cystine stone formers – an early glimpse of a genetic disease.

Figure 1.

(A) The Wollaston “cystic oxide stone,” now residing in the Gordon Museum (London). The blue color is associated with indigo derivatives produced upon exposure to light over time. Reproduced with permission from the Gordon Museum, Guy’s Hospital Campus, King’s College London. (B) Scanning electron microscopy image of a cystine stone, illustrating the hexagonal plate-like habit of individual cystine crystals. Insert: Molecular structure of L-cystine.

Fast forward to present, kidney stones comprising L-cystine affect at least 20,000 individuals in the United States.⁹ Although this number is substantially smaller than those afflicted by calcium oxalate monohydrate (COM) stones (approximately 10% of the U.S. population), L-cystine stones are larger, recur more frequently, and are more likely to cause chronic kidney disease. The formation of L-cystine stones is a consequence of excessive levels of L-cystine in the urine because of defective reabsorption of filtered cystine.¹⁰ This condition is the result of a genetic disorder caused by mutations in one of two genes that code for components of the major proximal renal tubule cystine and dibasic amino acid transporter.¹¹ The condition is exacerbated by the low solubility of L-cystine, which favors crystallization and subsequent aggregation of crystallites into centimeter-size stones.

Current therapies for the prevention of L-cystine stones rely on the oral administration of thiol drugs, which exchange with the disulfide group of L-cystine to form more soluble asymmetric disulfides that can be excreted more readily. These compounds, however, often require large doses, cause unpleasant side effects, or are simply ineffective. This prompted us to explore L-cystine mimics – molecular imposters – which could arrest crystal growth by binding to active growth sites on the surface of L-cystine crystals through molecular recognition. Along the way, this effort informed the crystal growth mechanisms at the nanoscale attributable to the hexagonal crystal symmetry first recognized by Wollaston. Moreover, the efficacy of crystal growth inhibition by these molecular imposters promises a new strategy for the prevention of L-cystine stones, supported by our studies in mouse models of cystinuria.

L-CYSTINE CRYSTAL STRUCTURE: A PRIMER

A brief review of the L-cystine crystal structure, micromorphology, and growth is a prerequisite to any discussion of growth inhibition (Figure 2). L-cystine stones are aggregates of plate-like L-cystine crystals with a hexagonal habit (Figure 1B). Crystals grown in the laboratory¹² are typically thin (. 30 microns) hexagonal plates with {0001} basal faces up to 400 μm wide, bounded by six equivalent $\left\{10\overline{1}0\right\}$ faces, identical to the morphology observed in actual stones. This form crystallizes in the hexagonal P6₁22 space group, with lattice constants a = b = 0.5422 nm and c = 5.6275 nm.¹³ The L-cystine molecules wind around the 6₁-screw axis as a hydrogen-bonded helix, each successively rotated by 60^o, such that six cystine molecules span the unit cell along the c axis. The helices associate with each other through hydrogen bonding and S…S interactions, which occur at intervals of c/2 along each of the six equivalent $\left\{10\overline{1}0\right\}$ directions in the basal plane. The structure also can be viewed as six-sided stacked layers of L-cystine molecules, each “elementary” layer having a thickness of c/6, the length of a L-cystine molecule. Each layer contains a 2-fold axis in-plane but lacks symmetry elements (e.g., rotation, mirror) parallel to the [0001] direction, resulting in six crystallographically inequivalent edges, each with a unique presentation of the cystine molecule. These edges can be denoted as A(), B() and C() ( denotes chemically identical presentations of the cystine molecule related by the two-fold axis orthogonal to the screw axis). Each successive layer is rotated by because of the 6₁-screw symmetry, which results in six crystallographically unique cystine presentations along the unit cell height.

Figure 2.

(A) Three adjacent helices of L-cystine molecules, winding about the [0001] direction (c axis), viewed perpendicular to the $(01\overline{1}0)$ plane. Six layers, denoted as C1 to C6, produce six unique L-cystine presentations on this face that span the 5.6 nm c axis. (B) Schematic of a hexagonal L-cystine crystal with Miller indices. (C) Schematic of the C1 elementary layer viewed down the c axis. The six unique sectors and their growth directions are denoted A(+), B(+), C(+), A(−), B(−), and C(−), each having a unique presentation of L-cystine. (D) Schematic of the six molecular layers in the L-cystine unit cell, with colors corresponding to unique L-cystine orientations. For example, green denotes a unique orientation of L-cystine on the $(10\overline{1}0)$ face in layer C1, winding around the c axis according to the 6₁-screw axis symmetry. € Space-filling model of a single cystine c/6 layer. (F) The crystal structure of a layer viewed down the c axis. A, B, and C denote to slices corresponding to $\left\{10\overline{1}0\right\}$ faces expected from Hartman−Perdok theory. The hydrogen bonds can be visualized more clearly by inspection of the crystallographic information file (CIF), which can be found at the Cambridge Structural Database (REFCODE = LCYSTI10). Adapted with permission from reference 20. Copyright 2015 American Chemical Society.

(DIS)SYMMETRY AND DISLOCATIONS

Crystal growth is commonly described by the terrace-ledge-kink model,¹⁴ in which steps advance across crystal terraces by the addition of solute molecules to kink sites along ledges defined by the intersection of a step and terrace. Screw dislocations, which Burton, Cabrera and Frank (BCF) posited as essential to crystal growth near equilibrium where most crystals grow,¹⁵ often are revealed as spirals emanating from a central core that continually extrudes steps to which molecules can attach. In this manner, the steps at the boundary of each spiral behave as “living ends” for solute attachment, resolving the paradox of fast growth from solutions at low (near equilibrium) supersaturation. Dislocations are common in molecular crystals,¹⁶ and often are associated with the inclusions of adventitious particles.^17,18

In the case of L-cystine, real-time in situ atomic force microscopy (AFM)¹⁹ of the (0001) face during growth in supersaturated (0.7 mM) aqueous L-cystine solutions reveals BCF theory in action. Growth steps spin clockwise out of each dislocation core, creating hexagonal island with features resembling a pinwheel (Figure 3A) created by six minor steps that radiate outward from the core.^1,2,20 The height of each minor step is c/6, equivalent to the thickness of a single elementary layer, and they are successively rotated by 60^o. Consequently, the minor steps are crystallographically equivalent by the 6₁-screw axis.

Figure 3.

(A) AFM image of a growth hillock generated by a single dislocation on the L-cystine (0001) surface. Six unique growth layers are labeled from 1 to 6; each contains a pair of three edges (A,B,C) growing in directions, i.e. A(+), B(+), C(+), A(−), B(−) and C(−). Step edges of layer 1 are traced with a white line. Starting from the center of the spiral, edges 1B(+) and 1A(+) are within the boundary of the hexagonal island, the edges 1C(−) and 1B(−) have grown to the boundary of the island, and the edges labeled 1A(−) and 1C(+) have grown to the boundary and are concealed by layers above them. The outer 1B(+) edge, which is equivalent to the 1B(+) near the core, also is concealed. Video of the spiral growth is provided in the Supporting Information. (B) The slowest-velocity step prevents advancement of the high-velocity step in a layer above, resulting in step bunching. Panel A reproduced with permission from reference 1. Copyright 2013 United States National Academy of Sciences.

The minor steps in the pinwheel advance more slowly than the five other inequivalent steps, which have already advanced to the hillock edges. According to Hartman−Perdok theory,²¹ planes that sever the fewest strong bonds will have the lowest surface energy. Analysis of the intermolecular hydrogen bonds suggests that the lowest energy steps are parallel to $\left\{10\overline{1}0\right\}$ planes, corresponding to slices A, B, and C in Figure 2F. Minor step A contains two in-plane N−H···O hydrogen bond pairs (1.80 Å) and short S···S contacts (3.47 Å) along the [112̅0] direction and truncates the fewest hydrogen bonds (two N−H···O hydrogen bonds pairs (1.91 Å). Minor steps B and C, parallel to other <112̅0> directions, each truncate four N−H···O hydrogen bond pairs (two stronger bond pairs, 1.80 Å, and two weaker bond pairs, 1.91 Å), suggesting a higher step surface energy and velocity of these steps compared with step A. Step A truncates the fewest hydrogen bonds, leading to the smallest step edge energy and, as a result, the slowest step velocity. Step A can advance in either the forward or reverse direction, denoted A(+) and A(−). Larger velocities would be expected along directions B(+)/B(−) and C(+)/C(−).

The white trace in Figure 3A denotes the six unique steps of a single elementary layer emanating from the core; four are concealed under the top surface. The step velocities increase with increasing L-cystine concentration, and the velocities of the six inequivalent steps differ. For example, at [L-cystine] = 1.2 mM, these velocities were A(+) = 1.8, B(+) = 13, C(+) = 9.5, A(−) = 5.1, B(−) =12.7, and C(−) = 10.9 nm/s. The rank ordering of the velocities, particularly the low values for A(+)/A(−), agrees with the Hartman-Perdok model as well as the step energy ranking calculated for L-cystine binding to the six unique molecular steps.

The growth rate anisotropy within each elementary layer produces curious, and illusory, crystal growth modes. Figure 3B illustrates two adjacent elementary layers, which by symmetry are rotated by 60^o. A faster moving step in the upper layer cannot advance past the slow-moving step below, leading to step “bunching.” In the same way, the slow-moving step in the second layer impedes the advancement of a faster-moving step in the third layer, and so on, eventually bunching six elementary layers to form a hexagonal island with a height equivalent to the 5.6 nm translational the c-axis repeat.

Growth from a single dislocation produces a micromorphology that resembles stacks of islands organized about the dislocation core (Figure 4A). Attachment of L-cystine molecules to the steps results in outward spread of hillocks from the dislocation core. Islands near the top are smaller because they emerge after the ones below. Occasionally, spirals generated by a closely spaced pair of heterochiral dislocations collide and coalesce to form Frank-Read loops, creating the impression of a macro-spiral (Figure 4C). According to BCF theory, a single screw dislocation generates a spiral, but a Frank-Read source generates closed loops (a.k.a. islands). The L-cystine AFM images seem to contradict BCF, but this is an illusion that arises because each island formed from a single dislocation consists of six interlaced spirals that bunch to produce well-defined islands. The Frank-Read source gives the illusion of a spiral because it is produced by stacks of closed-loop six-sided islands wherein each side has a different length because of the anisotropic step velocities. The successive 60^o rotations of each island results in a spiral-like micromorphology. Related features have been postulated for “symmetry-induced interlacing” of elementary growth layers sharing a screw axis,²² including 6₁, and observed for lower-order screw axes.^23,24 The dynamics and evolution of these growth modes is revealed in video simulations that use measured step growth rates (see Supporting Information). The end points of these simulations faithfully reproduce the observations (Figures 4B,D). While at first glance the AFM growth images appear to contradict BCF theory, they do not.

Figure 4.

L-cystine growth hillocks captured by real-time in situ AFM. (A) Hillocks generated by a single dislocation and (C) by a pair of dislocations, a so-called Frank–Read source. (B,D) Respective simulations using measured step velocities. Six different colors are used to depict the individual minor step edges of the c/6 elementary layers. Companion videos of growth and their simulations are provided in the Supporting Information. Reproduced with permission from reference 1. Copyright 2013 United States National Academy of Sciences.

MOLECULAR IMPOSTERS AND GROWTH INHIBITION

As mentioned above, the step velocity of the major 5.6 nm steps is equivalent along all six directions because the slowest step with each layer limits the advancement of each bunch. Measurement of the step velocities proved critical for determining the efficacy of crystal growth inhibition by “molecular imposters”– molecules that closely resemble the solute and bind to crystal faces through molecular recognition – a.k.a. “tailor-made additives/auxiliaries.”^25,26 The molecular imposter, L-cystine dimethyl ester (L-CDME; Figure 5), retains the core structure of L-cystine but is decorated with steric methyl “blockers” on both ends. The introduction of L-CDME during cystine crystal growth results in dramatic roughening of the $\left\{10\overline{1}0\right\}$ steps, signaling binding to the steps.

Figure 5.

Subset of molecular imposters that inhibit L-cystine crystal growth: L-CDME = cystine dimethyl ester); L-CDMOR cystine dimorpholide; L-CDNMP = cystine di(N-methylpiperazide); L-CDPE = cystine diphenyl ester. L-HCME translates as “half-cystine methyl ester”.

Figure 6A illustrates the inhibitory effect of a subset of molecular imposters in Figure 5 on the $\left\{10\overline{1}0\right\}$ major step velocities measured on the (0001) basal plane. In the absence of inhibitors, the step velocity exhibits a “dead zone” in which steps advance negligibly as [L-cystine] is increased, until c_d,min, which corresponds to the minimum supersaturation required for a nonzero step velocity. In the absence of inhibitors, c_d,min = 0.8 mM, slightly above the equilibrium solubility of L-cystine (c_eq = 0.7 mM), signifying the presence of an intrinsic impurity that inhibits growth. The impurity may be the stereoisomer of cystine, the reduction product cysteine, or a foreign agent introduced during manufacturing of the commercial form. The dead zone expands upon addition of L-CDME to supersaturated L-cystine solutions, however, moving c_d,min to 1.1 and 1.6 mM for [L-CDME] = 0.056 and 0.149 mM, respectively. This is tantamount to an effective increase in L-cystine solubility within this dead zone, although the origins of this behavior are kinetic. Above c_d,min, step velocities increase slowly until reaching, a threshold value above which step velocity increases quickly and linearly with increasing L-cystine concentrations. The existence of two threshold values bracketing a slow growth region has been reported for other compounds crystallizing in the presence of impurities.²⁷

Figure 6.

(A) Step velocities, V, of bunched $\left\{10\overline{1}0\right\}$ steps on a {0001} L-cystine crystal surface without additive and with additives (0.056 mM). The dead zone corresponds to the region c_eq < c < c_d,min. Beyond c_d, the slope is proportional to the kinetic coefficient, β. (B) Dependence of the L-cystine residual concentration after crystallization on inhibitor concentration. Using EC_2x as a metric, L-CDMOR and L-CDNMP are more effective at sustaining metastable L-cystine concentrations than L-CDME (EC_2x(L-CDMOR) = 0.86 μM, EC_2x(L-CDNMP) = 0.26 μM, EC_2x(L-CDME) = 6.37 μM). (C,D) The dependence of R/R_o, the growth rate normalized to growth in the absence of inhibitor, for $\left\{10\overline{1}0\right\}$ surfaces on the (1000} face, on [L-CDME], as measured from step velocities. (D) The dependence of R/R₀ on [L-CDME] for growth on the $\left\{10\overline{1}0\right\}$ face, as measured for bulk crystals. Panels A and B reproduced with permission from references 3 and 4. Copyright 2017 and 2016 American Chemical Society.

The slope beyond c_d is nearly identical for [L-CDME] = 0, 0.056, and 0.149 mM, and for various molecular imposters. This behavior is consistent with the Cabrera−Vermilyea mechanism of impurity action,^28,29 in which inhibitor molecules at the step edge act as stoppers that block step propagation through pinning (Figure 7).^30,31 Pinning induces curvature of the step edge as it advances between the adsorbates. This curvature increases the step edge energy – tantamount to a decrease in supersaturation – as described by the Gibbs−Thomson law. Increasing the inhibitor concentration decreases the average distance between pinning sites, further increasing the step edge curvature. When the separation between posts is smaller than the diameter of the critical nucleus, step curvature is sufficiently high to stop growth. This is observed as a shift of the dead zone to higher L-cystine concentration with increasing L-CDME concentration.

Figure 7.

Step pinning by adsorbed inhibitor molecules and their incorporation into the crystal lattice during overgrowth. Adapted with permission from reference x (Chem. Rev. 2017, 117, 14042−14090). Copyright 2017 American Chemical Society.

Measurements of residual L-cystine concentrations in solution after crystallization using a fluorescent tag protocol during bulk crystal growth revealed substantial differences (Figure 6B) among the inhibitors.^4,32 Here, the concentration effect of an inhibitor that elevates supersaturation of L-cystine following crystallization is benchmarked using EC_2x, the average value of the L-cystine concentration at low and high inhibitor concentration. L-CDMOR and L-CDNMP were superior to L-CDME with significantly lower EC_2x values, corresponding to higher supersaturations and suggesting more effective kinetic inhibition of L-cystine growth compared with L-CDME. Notably, the supersaturations with respect to hexagonal L-cystine achieved with all three of these imposters were sufficient for crystallization of the less stable tetragonal polymorph. The concentration at which only the tetragonal polymorph crystallizes, albeit in small amounts, is yet another metric for the inhibition of the hexagonal form.

The disparity in the EC_2x values measured for the different imposters, despite comparable reductions in the $\left\{10\overline{1}0\right\}$ step velocities on the {0001} basal plane, suggested an important role for growth sites on the $\left\{10\overline{1}0\right\}$ faces. The small size of these faces, combined with their high step density, precluded the measurement of step velocities by in situ AFM. Therefore, the effect of molecular imposters on the growth of $\left\{10\overline{1}0\right\}$ surfaces was evaluated by comparing AFM data acquired for $\left\{10\overline{1}0\right\}$ step motion on the {0001} L-cystine surface with the height-to-width aspect ratios of L-cystine crystals obtained from bulk crystallization. In both cases, inhibition through binding to growth sites on the $\left\{10\overline{1}0\right\}$ surface can be characterized by R/R₀, where R and R₀ is the growth rate normal to the $\left\{10\overline{1}0\right\}$ surface in the presence and absence of inhibitor, respectively (Figure 6C,D). These data, exemplified here for L-CDME, revealed that the inhibitors were significantly more effective in arresting growth perpendicular to the $\left\{10\overline{1}0\right\}$ face than the $\left\{10\overline{1}0\right\}$ steps on the {0001} face. This was corroborated by geometry-optimized molecular mechanics computations, which revealed significantly stronger binding of the imposters at kink sites on the $\left\{10\overline{1}0\right\}$ face than at kink sites on {0001} (Table 1, Figure 8). The strong binding of the inhibitors to the $\left\{10\overline{1}0\right\}$ face also was evident from the increasing aspect ratio (c/a) of the hexagonal crystals with increasing concentration of inhibitor, eventually resulting in the exclusive formation of the tetragonal form (Figure 9).

Table 1.

Binding energies (kcal/mol) of select inhibitors at kink sites on the {0001} and $\left\{10\overline{1}0\right\}$ faces.

Crystal face	{0001}	$\left\{10\overline{1}0\right\}$
	cc	ac	aa
L-HCME	30	−74	−148
L-CDME	35	−365	−387
L-CDPE	16	−401	−396
L-CDMOR	43	−397	−423
L-CDNMP	14	−386	−416

Figure 8.

(left) Schematic representation of the kink sites on the {0001} and $\left\{10\overline{1}0\right\}$ faces of L-cystine. Although the three kink sites may appear identical, they differ with respect to the steric environment experienced by the adsorbed inhibitor. (right) Lowest energy binding modes of (A) L-CDPE, (B) L-CDMOR and (C) L-CDNMP for kink sites on $\left\{10\overline{1}0\right\}$ steps on the (0001) face (the cc kink site). Adapted with permission from reference 3. Copyright 2017 American Chemical Society.

Figure 9.

Dependence of L-cystine crystal morphology on the concentration of L-CDMOR inhibitor. Note that at 2 μM of L-CDMOR L-cystine crystallizes as tetragonal polymorph.

The step-pinning molecular imposters can be overgrown and permanently sealed in L-cystine crystals, to concentrations as high as 0.45 mol%. Much stronger growth inhibition (Figure 6) and adsorption of inhibitors (Table 1) on {10–10} face compared to {0001} face suggest much higher incorporation of inhibitors into $\left\{10\overline{1}0\right\}$ face. This incorporation can be described by a distribution coefficient, K_d = [Inhibitor]_crystal/[Inhibitor]_solute, which was estimated to be one hundred times greater for the $\left\{10\overline{1}0\right\}$ face than for the {0001} growth sectors. L-cystine crystals grown in the presence of molecular imposters exhibited anomalous birefringence under crossed polarizers, with six well-defined biaxial $\left\{10\overline{1}0\right\}$ sectors corresponding to 60^o rotations of the extinction directions in adjacent sectors (Figure 10). These observations illustrate a remarkable fidelity of the stereospecific binding of the imposters at the same unique crystal site in each sector, such that the imposter orientation in each sector is rotated by 60^o with respect to its adjacent sectors.

Figure 10.

Optical micrographs between crossed polarizers with a first order red waveplate of (A) L-cystine crystals grown in the presence of 0.0025 mM L-CDPE and (B) tetragonal L-cystine crystals grown in the presence of 0.005 mM L-CDME. The sectoral zoning is consistent with attachment to a unique crystal site in each example. Adapted with permission from reference 3. Copyright 2017 American Chemical Society.

The stereochemical specificity of molecular imposter binding to crystal sites was tested by measuring step velocities in the presence of various molecules from a library of more than fifty compounds (some examples are shown in Figure 5).^3,32 This library included molecules in which the terminal hydroxyl group of L-cystine were replaced or its backbone core modified, as well as compounds based on L-cysteine. The step velocity measurements revealed that the most effective inhibitors were L-cystine diesters and diamides that retained essential structural elements of cystine. Small changes or deletions to the key molecular recognition elements of cystine – swapping sulfur with carbon, lengthening the core by using esters of homocystine, eliminating the carboxyl termini, and substituting enantiomers (D-CDME) – led to substantially reduced, and often negligible, inhibition. Cysteine methyl ester (L-HCME) was a reasonably effective inhibitor, suggesting stereospecific binding to a portion of the important crystal site. Cystamine, which is commensurate with the binding site but lacks the terminal carboxylic group, was moderately effective, consistent with an important role for intermolecular S···S interactions between inhibitor disulfide moieties and the crystal site. Notably, common cystinuria therapeutic compounds did not inhibit crystal growth. Collectively, these observations demonstrated that inhibition relies on a rather strict structural and stereochemical recognition between an inhibitor and specific L-cystine crystal sites.

STOPPING CYSTINE KIDNEY STONES IN VIVO

Kidney stone formation typically is associated with four critical phenomena – crystal nucleation, crystal growth, adhesion to epithelial cells, and crystal aggregation.³³ Blocking one of these steps may be sufficient to prevent stone formation. The effective inhibition of L-cystine growth by molecular imposters suggested a pathway to a new class of small molecule therapeutics for the prevention of cystine stones, particularly if the inhibitors can arrest growth immediately following the first appearance of crystal nuclei. The efficacy of L-CDME, L-CDNMP and L-CDMOR was studied in a Slc3a1 knockout (KO) mouse model of cystinuria. Initial studies involved treatment of KO male mice with L-CDME (by gavage).^34,35 These studies revealed a significant decrease in stone size compared with a water-only control group (Figure 11). Although the number of stones was greater for mice treated with L-CDME, the overall mass was reduced by approximately 50%. Overall, cystine excretion in urine was the same between the two groups, indicating that L-CDME did not interfere with cystine metabolism. Scanning electron microscopy analysis of cystine stones from the L-CDME group revealed hexagonal L-cystine crystals that were numerous but reduced in size. Curiously, these effects were observed despite the negligible amount of L-CDME detected in the urine of the treated mice. Surprisingly, however, L-cysteine methyl ester (L-HCME, above), was detected in stones from the L-CDME group, indicating cleavage of the disulfide bond. Given the inhibition observed for L-HCME in vitro, its presence in mouse urine suggests a possible inhibiting role in vivo.

Figure 11.

Analysis of cystine stones. (A). Microcomputed tomography images of bladders from two Slc3a1 knockout male mice treated with water (left panel) or L-CDME (right panel). For each mouse, the top panel is a cross-sectional image and the bottom panel is the intact organ. (B). Stones retrieved from the bladder of a water- or L-CDME-treated mouse. Adapted with permission from reference 34. Copyright 2017 Elsevier.

As described above, L-CDMOR and L-CDNMP were more effective as sustaining high supersaturations of L-cystine, which anticipates more effective inhibition of L-cystine crystal growth by these compounds in vivo. Moreover, these compounds were presumed to have greater stability in vivo than L-CDME, which may be hydrolyzed readily by esterases and explain its negligible presence in the urine of treated mice. Two groups of Slc3a1 KO male mice were treated with either compound through daily gavage for four weeks, and a third control group received water only. Five out of the seven mice in the control group formed stones. All six mice treated with L-CDMOR formed stones, but only one out of the group of seven mice treated with L-CDNMP formed stones. These results demonstrate that L-CDNMP is dramatically more effective than L-CDMOR with respect to preventing stone formation in cystinuria mice. While this observation is somewhat surprising given the similar crystal growth inhibition observed by AFM, it is consistent with the greater inhibition by L-CDNMP suggested by the crystallization inhibition assays. Moreover, significantly higher concentrations (ca. 3.6 times) of L-CDNMP were found in mice with cystinuria compared with L-CDMOR at equivalent oral doses (Figure 12). These unexpected results suggest that the activities of other transporters are elevated after knocking out the Slc3a1 gene, which worked in the favor of L-CDNMP but not L-CDMOR. Nonetheless, combined with the lack of adverse pathological effects, the dramatic inhibition effect observed in the mouse model for L-CDME and L-CDNMP support the strategy of using molecular imposters as small molecule therapeutics, with L-CDNMP appearing to be a leading candidate. The poor in vivo activity of L-CDMOR is a cautionary tale, as it demonstrates that promising leads can be frustrated by the complexity of biology.

Figure 12.

Average drug concentration in mouse urine after 7 daily oral dosing of L-CDMOR and L-CDNMP. SLC(+/+) and SLC(−/−) denote wild-type and mutant mice, respectively.

OUTLOOK

Aberrant crystallization within the human body can lead to several disease states, although the pathogenesis of such diseases is not studied often from a materials science perspective.³⁶ The observations in this Account illustrate a unique, and powerful, combination of fundamental studies of crystal growth and their translation to the pre-clinical setting, promising a new strategy for treating cystinuria, a disease for which an effective therapy has been elusive. The competitive binding approach embodied in the imposter principle has been used to arrest the nucleation and growth of monosodium urate crystals,^37,38 the sole crystalline phase found in gout deposits.^39,40 This would have impressed Galileo, who suffered from Saturnine gout,^41,42 a form of gout associated with lead exposure (in Galileo’s case likely from wine consumption). Saturnine gout has been attributed to lead inhibition of tubular urate transport but lead diurate crystals also have been implicated in the rapid nucleation of MSU.⁴³ It remains to be seen if the molecular imposter principle can be translated universally to other diseases caused by pathological crystallization. Recent work in our lab pertaining to a PX23 receptor antagonist for treatment of chronic pain has illustrated the promise of molecular imposters for arresting the crystallization of low-solubility active pharmaceutical ingredients (APIs) in the urinary or gastrointestinal tract.⁴⁴ The effect of small molecules on calcium oxalate monohydrate crystallization – the major ingredient of the most common kidney stone – has been studied exhaustively,^45,46 although these are not molecular imposters per se. Although not a direct cause of human disease, hematin crystallization in the food vacuole of malarial parasites removes heme that otherwise would be toxic to the parasite. This has prompted several elegant mechanistic studies of the inhibiting role of small molecules, including quinoline antimalarials, on hematin crystallization.^47,48,49,50 These, among other studies, continue to reveal the value of studying crystal growth at the near-molecular level for improving human health challenges associated with crystallization.

Download video file ^{(2.3MB, mov)}

Download video file ^{(5.1MB, mov)}

Download video file ^{(555.8KB, mov)}

Download video file ^{(6.5MB, mov)}

Download video file ^{(1.6MB, mov)}

Download video file ^{(9.7MB, mov)}

Key References.

• Shtukenberg A. G.; Zhu, Z.; An Z.; Bhandari, M.; Song, P.; Kahr, B.; Ward, M. D. Illusory spirals and loops in crystal growth. Proc. Natl. Acad. Sci. 2013, 110, 17195–17198. (1) Spirals and loops on L-cystine crystal surfaces may appear to violate classical theories of dislocation-actuated growth, but this is an illusion that can be clarified by simulations.

• Rimer, J. D.; An, Z.; Zhu Z.; Lee, M. H.; Goldfarb, D. S.; Wesson, J. A.; Ward, M. D. Crystal Growth Inhibitors for the Prevention of L-Cystine Kidney Stones through Molecular Design. Science 2010, 330, 337–341. (2) Esters of L-cystine dramatically reduce the growth velocity of steps on the (0001) surface because of specific binding, signaling rational design of crystal growth inhibitors.

• Poloni, L. N.; Zhu, Z. Garcia-Vázquez, N.; Yu, A.; Connors, D.; Hu, L.; Sahota, A.; Ward, M. D.; Shtukenberg, A. G. The Role of Molecular Recognition in L-Cystine Crystal Growth Inhibition. Cryst. Growth Des. 2017, 17, 2767−2781. (3) L-cystine mimics, particularly L-cystine diesters and diamides, inhibit L-cystine crystal growth through a common inhibition mechanism consistent with Cabrera−Vermilyea step pinning.

• Hu, L.; Yang, Y.; Aloysius, H.; Albanyan, H.; Yang; M.; Liang, J.-J.; Yu, A.; Shtukenberg, A.; Poloni, L.; Kholodovych, V.; Tischfield J.; Goldfarb, D. S.; Ward, M. D.; Sahota, A. L-cystine diamides as L-cystine crystallization inhibitors for cystinuria. J. Med. Chem. 2016, 59, 7293 – 7298. (4) L-Cystine diamides were effective inhibitors of L-cystine crystallization and more stable than esters, with L-cystine bis(N′-methylpiperazide) particularly effective in vivo in a knockout mouse model of cystinuria.

Biographies

Alexander G. Shtukenberg received a specialist degree in 1993 and Candidate of Science degree in 1997 at the Geological Faculty of Saint Petersburg State University, Russia, and continued to work at the Geological Faculty as a researcher and faculty member, earning the Doctor of Science degree in 2009 and the title of professor in 2010. Since 2010 he has been associated with the Molecular Design Institute in the Department of Chemistry at New York University, where he holds the title of Research Professor.

Longqin Hu received a bachelor’s degree in pharmacy in 1984 and a Master’s degree in Medicinal Chemistry in 1987 from the Second Military Medical University in Shanghai. He studied medicinal chemistry at the University of Kansas, (Ph.D., 1993), after which he was an NIH postdoctoral fellow at the University of Delaware. He began his academic career in 1996 at the University of Oklahoma, and in 1999 moved to Rutgers University, where he is currently a Professor and Chair in the Department of Medicinal Chemistry, Ernest Mario School of Pharmacy.

Amrik Sahota received his undergraduate and graduate education in the UK, including a PhD in medical genetics from Guy’s Hospital Medical School, University of London, followed by postdoctoral training in medical genetics at the Indiana University Medical Center. He is a Professor in the Department of Genetics at Rutgers University.

Bart Kahr studied chemistry at Middlebury College and then Princeton University (Ph.D., 1988), followed by a postdoctoral appointment at Yale University. He was a faculty member at Purdue University from 1990 to 1996 and at the University of Washington, Seattle from 1997 to 2009. He then moved to New York University, where he is a Professor of Chemistry in the Molecular Design Institute.

Michael D. Ward studied chemistry at the William Paterson College of New Jersey and Princeton University (Ph.D., 1981). After a postdoctoral appointment at the University of Texas, Austin, he held research positions at Standard Oil of Ohio and Dupont Central Research, joining the faculty of the Department of Chemical Engineering and Materials Science at the University of Minnesota in 1990. Ward moved to New York University in 2006, where he is a Silver Professor of Chemistry in the Molecular Design Institute.