Reverse Engineering the Intracellular Self-Assembly of a Functional Mechanopharmaceutical Device

Weakly basic, poorly soluble chemical agents could be exploited as building blocks for constructing sophisticated molecular devices inside the cells of living organisms. Here, using experimental and computational approaches, we probed the relationship between the biological mechanisms mediating lysosomal ion homeostasis and the self-assembly of a weakly basic small molecule building block (clofazimine) into a functional, mechanopharmaceutical device (intracellular Crystal-Like Drug Inclusions – “CLDIs”) in macrophage lysosomes. Physicochemical considerations indicate that the intralysosomal stabilization of the self-assembled mechanopharmaceutical device depends on the pHmax of the weakly basic building block and its affinity for chloride, both of which are consistent with the pH and chloride content of a physiological lysosomal microenvironment. Most importantly, in vitro and in silico studies revealed that high expression levels of the vacuolar ATPase (V-ATPase), irrespective of the expression levels of chloride channels, are necessary and sufficient to explain the cell-type dependent formation, stabilization, and biocompatibility of the self-assembled mechanopharmaceutical device within macrophages.


Thermodynamic equilibrium model of CFZ's solubility.
For weak electrolytes, such as weakly basic drugs, the protonation state of the drug (B) is dictated by the relationship between the drug's association constant (K a ) and the acidity of its environment (H 3 O + ). This is represented by the following equilibrium expression: where BH + is the protonated form of the drug as a result of the interaction of the free base form of the drug with hydronium ion, and H 2 O is the byproduct of the reaction and remains constant. Thus, the equilibrium constant (K) is multiplied by H 2 O to obtain K a . Therefore, equation (1) is re-written using K a . According to thermodynamic laws of mass action, the total amount of drug is comprised of both the neutral and ionized forms at any given pH of its environment 19 . Thus, the total solubility (S T ) of the drug is written as: T However, depending on the pH of the solution, the total drug solubility equation must be slightly modified to account for the distinction of the primary species in the solid versus solution phase 19 . For pH < pH max , the ionized form of the drug is the saturated species that is in equilibrium with the salt from of the drug in the solid phase; therefore, its solubility remains constant while the solubility of the neutral form of the drug varies with SCIenTIfIC REPORTS | (2018) 8:2934 | DOI: 10.1038/s41598-018-21271-7 respect to pH. Therefore, by substituting equation (5) into equation (7), the total drug solubility for pH < pH max is re-written as: where the solid phase is denoted by the subscript "s".
To the contrary, for pH > pH max , the neutral form of the drug is in the solid phase; thus, its solubility remains constant whereas the solubility of the ionized form of the drug varies with respect to pH. Therefore, by substituting equation (6) into equation (7), the total solubility for pH < pH max can be re-written as: S [B] ( 1 10 ) (9) Moreover, at pH = pH max , both the ionized and neutral forms of the drug are the saturating species, where both the salt and free base precipitate forms are in the solid phase. Thus, equation (7) can be re-written as: T S S For our purpose, we substituted our model drug CFZ in the above equations (1)(2)(3)(4)(5)(6)(7)(8)(9)(10). Rewriting equation (7) results in the following total CFZ-H + Cl − solubility equation: T S where the solid phase, which in this case is the free form of CFZ, is denoted by the subscript "s", and the ionized form of the drug, [CFZH + ], is in the solution phase, and similar to equation (6), is dictated by pK a and pH as follows: Thus, by substituting equation (12) into equation (11), we obtain the following: Furthermore, by using different combinations of any two solubility-pH data points from the experimental total solubility-pH measurements, the pK a and the intrinsic free base solubility values were simultaneously solved using equation (13). Then, by substituting these values into equation (13), the total CFZ-H + Cl − solubility was calculated for the given range of pH 4.5 to 8.9, and the experimentally obtained total drug solubility-pH curve was fitted. Furthermore, in order to generate the total solubility-pH dataset for pH below pH max , we used the following equation: T S Similar to equation (8), equation (14) can be further expressed as: However, we first had to determine [CFZH + ] S . Because we only used one form of the drug (CFZ-H + Cl − salt) as the starting material in the experimental measurements of total drug solubility as a function of pH, we used a mathematical proof approach to determine [CFZH + ] S and thereby pH max (see Supplementary Methods) without necessarily having to experimentally generate a total drug solubility-pH curve for pH below pH max . Moreover, the salt solubility product (K sp ) of CFZ-H + Cl − is given by the following equation: However, because of the 1:1 stoichiometric relationship of CFZH + and Cl − in CFZ-H + Cl − 7 , the K sp of CFZ-H + Cl − in an aqueous media can be expressed as following: was centrifuged (10 min at 400 × g, 4 °C), the supernatant removed, and the cell pellet was resuspended in RPMI 1640 media (Life Technologies) with 5% FBS (Life Technologies) and Penicillin/Streptomycin (Thermofisher). Cells were plated onto 4 or 8 chamber coverglass (#1.5, Lab-Tek II, Nunc, Rochester, NY) in RPMI for imaging studies. The cells were allowed to attach overnight, washed, and imaged in fresh RPMI.
Peritoneal macrophage isolation. Following euthanasia, a small incision was made in the lower abdomen. The peritoneal cavity was then flushed with 10 mL of ice cold DPBS containing 5% FBS (Sigma) and collected. The peritoneal lavage was centrifuged for 10 min at 400 × g, 4 °C, and then re-suspended in DMEM media (Life Technologies) with 5% FBS and Penicillin/Streptomycin and counted. The cells were plated into Mat-tek dishes overnight in a serum-free growth media, and washed five times with phosphate-buffered saline (PBS).
Macrophage depletion. Mice were fed CFZ or control diet continuously for a four-week period. Following two weeks of feeding, mice were treated with either liposomes containing either 7 mg/mL of clodronate or PBS (FormuMax Scientific Inc., Sunnyvale, CA) for two weeks. Mice were initially treated with 200 µL of liposomes followed by 100 µL injections twice per week (or a matching volume of PBS) to ensure continual macrophage depletion. Liposomes were injected intraperitoneally, as previously described 20 . After completing four weeks of feeding and two weeks of liposome treatment, mice were sacrificed and tissues were collected.
Biochemical analysis of CFZ in tissues. The concentration of CFZ in the organs of mice was determined spectrophotometrically. After four weeks of CFZ-or vehicle-diet treatment, mice were euthanized via CO 2 asphyxiation, and organs were collected. Tissue (20-30 mg) was homogenized in 500 µL of radioimmunoprecipitation assay buffer (Sigma) with added protease inhibitors (Halt protease and phosphatase inhibitor cocktail and 0.5 M EDTA; Thermo Pierce, Rockford, IL). Drug was extracted from homogenate (350 µL) with three washes with xylenes (1 mL). The drug was then extracted from the xylene with three 1 mL passes of 9 M sulfuric acid. The concentration of CFZ present in the tissue was determined using a 96-well plate reader (Biotek Synergy 2, Winooski, VT) (wavelength 540 nm). To account for extraction yield, untreated liver and spleen samples were spiked with known amounts of CFZ prior to extraction, and were analyzed in the same plate as the CFZ treated samples. The mass of CFZ per organ was determined using the yield-corrected concentration of CFZ in the mass of tissue analyzed, as determined through use of a standard curve with known concentrations of CFZ, and is reported as mg CFZ/g tissue.

Biochemical analysis of CFZ in plasma.
Blood was collected and centrifuged (7,000 × g for 5 minutes). Deep-etch, freeze-fracture electron microscopy. At the time of euthanasia, the liver was collected and kept cold (4 °C). The organ was frozen against a copper block, cooled with liquid helium, and stored in liquid nitrogen. The sample was fractured with Balzers 400 nitrogen cooled vacuum evaporator and freeze-etched for two minutes (−100 °C). A rotary replica was generated with 2 nm platinum and backed with 10 nm carbon film support. It was cleaned with chromo-sulfuric cleaning solution (Fisher Scientific, cat# SC88) for twelve hours and rinsed with DI water. The sample was picked up on formvar coated grids and viewed on a JEOL 1400 electron microscope with an AMT camera (JEOL USA, Inc., Peabody, MA).

Spectral confocal microscopy.
For the preparation of slides, CFZ-H + Cl − crystals suspended in PBS (20 µl) were placed on a glass slide and a cover-slip was applied onto the sample prior to imaging. Spectral confocal microscopy was performed on a Leica Inverted SP5X confocal microscope system with two-photon FLIM (Leica Microsystems, Buffalo Grove, IL) using excitation wavelengths (470-670 nm). Image analysis and quantification were performed on Leica LAS AF. Several regions of interest of individual crystals were used to obtain fluorescence data, which were imported into MS-Excel for further analysis. All fluorescence yields were normalized to the maximum fluorescence yield measured across the tested spectral range, and background subtracted using data obtained from a blank slide. confirmation, RAW264.7 cells were seeded at 30,000 cells/well in an 8-chamber multiwell plate (500 µl/well growth media). Twenty-four hours later, the growth media was replaced with media containing varying concen- Quantitative cytometric analysis. Following macrophage staining, the population of macrophages was determined by taking the ratio of the total F4/80 signal to the total nuclear signal across an image. Because of the unique Cy5-specific fluorescence property of CLDIs 11 , the prevalence of CLDIs was determined by taking the ratio of the total Cy5 signal to the total nuclear signal across an image. 5 images per mouse per organ were analyzed.

Epifluorescence microscopy of cells incubated with CFZ and
Drug stability in a lysosomal microenvironment. CLDIs were isolated from spleen and liver of 8-week CFZ treated mice based on previously published protocols 10,21 . Free base CFZ crystals were pure CFZ crystals (Sigma-Aldrich, C8895). A speck (≪1 mg) of isolated CLDIs, CFZ-H + Cl − crystals, and free base CFZ crystals were placed in 2 mL of each of the stability testing medias: 1× PBS (pH 7.4), lysosomal buffer without sodium chloride (LB-, pH 4.5), and lysosomal buffer with sodium chloride (LB+, pH 4.5). Lysosomal buffers were prepared according to a previously published protocol 22 . Samples were stirred in media for 5-7 days; after which the stability of crystals from each sample was monitored using brightfield and fluorescence microscopy using a Nikon Eclipse Ti inverted microscope (Nikon Instruments, Melville, NY). Look-up-tables (LUTs) and exposure times were maintained the same throughout the samples.
Multi-parameter microscopy. Multi-parameter polarization, brightfield, and fluorescence microscopy was conducted using a Nikon Eclipse Ti inverted microscope (Nikon Instruments, Melville, NY). Polarization microscopy was performed using the LC-PolScope 23 , with the illuminating light narrowed to 623 nm by an interference filter (623 ± 23 nm, Semrock Optics, Rochester, NY). Polarization images were captured using an Abrio imaging system (Cambridge Research & Instrumentation, Inc, Woburn, MA). Brightfield images were captured using the Nikon DS-3 camera (Nikon Instruments) and fluorescence images were taken with the Photometrics CoolSnap MYO camera system (Photometrics, Tuscon, AZ) under the control of Nikon NIS-Elements AR software (Nikon Instruments). Illumination for fluorescence imaging was provided by the X-Cite 120Q Widefield Fluorescence Microscope Excitation Light Source (Excelitas Technology, Waltham, MA). Images were acquired and analyzed as previously described 24 .
Raman sample preparation and measurements. Isolated mouse alveolar macrophages were transferred onto pre-sterilized silicon chips (16008; Ted Pella, Inc., Redding, CA) and incubated for 1 h (@ 37 °C and 5% CO 2 ) to allow adherence to chip. The cell-containing chips were washed by brief submersion into isotonic NaCl (0.9%) solution followed by DI water, then allowed to air-dry. Raman measurements were acquired with the WiTec alpha300R confocal Raman microscope (WITec, Ulm, Germany) equipped with two excitation lasers: a 532 nm solid-state sapphire and a 785 nm wavelength-stabilized diode (0-55 mW and 0-88 mW tunable intensity ranges, respectively). A 100X air objective (Zeiss Epiplan-NEOFLUAR, NA (numerical aperture) = 0.9) coupled to a CCD detector via a multi-mode fiber of 100 μm diameter serving as the confocal pinhole, produced 0.72 μm and 1.06 μm illumination spots (for 532 nm and 785 nm lasers respectively). To minimize fluorescence background from pure CFZ and CFZ-H + Cl − reference crystals, samples were excited with 785 nm. The 532 nm laser was utilized for excitation of biological samples due to its elicitation of a stronger Raman signal from the microscopic inclusions/CLDIs of interest. Point spectra were acquired (n ≥ 40 cells/group) by focusing a laser spot on cytoplasmic inclusions, CLDIs, or pure reference crystals: at each point, the laser was tuned to optimum intensity, and acquired point spectra over an integration time of 25 seconds. Individual raw spectra were baseline-subtracted and normalized using a MATLAB ® processing algorithm developed in-house. For single-cell Raman imaging, the 532 nm excitation laser was raster-scanned across a 50 × 50 micron area with a step-size of 0.5 microns, yielding spectral datasets consisting of 10,000 spectra per cell. Exploiting the dramatic spectral differences arising from fluorescence of different CFZ forms, the un-processed Raman spectra were linearly deconvoluted via WiTec ProjectFOUR software's basis component analysis using representative un-processed reference spectra obtained from untreated cells, pure free base CFZ crystals, pure CFZ-H + Cl − crystals, and the silicon substrate. Dataset acquisition, processing, and image display parameters were performed equivalently for each cell specimen.

CLDI injection and stabilization assay.
To determine how macrophages stabilize CLDIs, mice were treated with either liposomal PBS or liposomal clodronate, as previously described 25 . 48 hours after liposome administration, mice were injected I.P. with 200 µg of CLDIs suspended in 1 mL of PBS (n = 3 mice per group per time point). At time points ranging from 0 to 48 hours, the mice were euthanized, the peritoneal lavage was collected and pelleted, and the drug content within the pellet was analyzed using the previously described spectrophotometric analysis method. Using a simple exponential regression with recovered drug content, the half-life was estimated. CLDI loading within individual macrophages. Using the total recovered mass of CLDIs within the liver and spleen, the CLDI loading within individual xenobiotic-sequestering macrophages was estimated using literature reported values for macrophage numbers 26 within the liver and spleen, corrected for the percentage of cells which contained a CLDI. For the clodronate depleted organs, the reduction was accounted for with the quantified reduction in total F4/80 signal. Drug uptake measurements. Measurement of CFZ uptake by RAW264.7 cells was performed using a modified absorbance spectroscopy method using 9 M H 2 SO 4 (pH ≪ 0.1) to digest the entire cell population and extract CFZ from the cells. At two time points (2 and 4 h) after CFZ exposure, growth media was aspirated from the culture dish and the cells were washed twice with PBS before adding 9 M H 2 SO 4 (100 µl/well). The plate was incubated at room temperature for 30 min before spectrophotometric quantification of CFZ using a Synergy 2 plate reader (BioTek, Winooski, VT) at 540 nm (Abs 540 ) and 750 nm (Abs 750 ) (for background). Standards were prepared on the same plate by adding 100 µl/well pre-determined standards of CFZ. Total CFZ uptake is reported as total intracellular drug (in picomoles) as measured and calculated using the standard curve. Assuming that the doubling time of 50,000 cells/well is ~11 h 27 , and that no compromise of cell viability occurs because of the optimum dose used for both CFZ and the pharmacological inhibitors, the total number of cells/well at the time of cellular drug uptake measurement is ~68,000.

Cell culture and pharmacological treatment. For pharmacological inhibition experiments
Experimental lysosomal pH measurements. Lysosomal pH was measured fluorometrically in peritoneal macrophages and macrophage-derived RAW264.7 cells, using methods described previously 28 . In brief, macrophages were incubated for 40 h with CFZ crystals. 20 h prior to measuring pH, cells were incubated for 15-18 h with 150 µg/ml Oregon Green-labeled dextran-10 kD (OGDx; Thermo-Fisher Scientific), and transferred to unlabeled culture medium for 3-5 h to allow OGDx to traffic fully to lysosomes. Following transfer to Ringer's buffer, cells on Mat-tek dishes were observed in a Nikon TE300 microscope equipped for multichannel fluorescence microscopy with a 60X objective lens (N.A. 1.4). Phase contrast and three fluorescence images (excitation/emission: 440 nm/535 nm; 485 nm/535 nm; and 580 nm/630 nm) were collected from each field. The 580 nm/630 nm image was used to quantify clofazamine fluorescence. The ratio of the 485 nm/525 nm to 440 nm/525 nm images were used to measure lysosomal pH. Calibration of fluorescence ratios was obtained from cells incubated in ionophores and buffers as described in Davis and Swanson 28 . Average lysosomal pH was determined for individual cells.

Modeling lysosomal CFZ-H + Cl − accumulation. A Lysosomal ion regulation model which incorporates
lysosomal membrane proteins, such as V-ATPase, CLC7, and membrane proton permeability was utilized 16 , and further elaborated and modified (see Supplementary Methods online) to explore the relationship between lysosomal ion homeostasis and lysosomal CFZ-H + Cl − accumulation and stabilization. In order to incorporate the lysosomal accumulation of CFZ-H + Cl − into the model as a function of lysosomal proton and chloride ions, the rate of overall drug accumulation obtained from experimental findings 4,29 was used to define the rates of both lysosomal proton (H sequestered ) and chloride (Cl sequestered ) sequestrations by CFZ, in units of molecules per second. We assumed these rates to be equal due to the proposed equal contribution of both proton and chloride ions in the intracellular formation and accumulation of CFZ-H + Cl − salt precipitates: Model parameterization. The model incorporated 23 parameters: four were adjustable and the remaining 19 were fixed (see Supplementary Table S1). Fixed parameters were given values obtained from the literature 16,[30][31][32][33][34] , and are associated with physiological lysosomal ion homeostasis. Therefore, these parametric values are interchangeably referred hereon as "baseline input values" or "physiological baseline input values". The adjustable parameters are those values that were varied from their respective baseline input values in order to investigate their individual as well as combined effects on the physiological lysosomal pH, Cl − , and membrane potential readout values, as discussed in the following subsections. These parameters include the number of active V-ATPase and CLC7 molecules per lysosome, as well as the cytoplasmic chloride concentration. In addition, the rates of proton and chloride sequestrations by CFZ are considered adjustable parameters as they are foreign to the lysosome.
Simulating lysosomal CFZ-H + Cl − accumulation. The previously mentioned rates of proton and chloride sequestrations by CFZ were adjusted to either 0.01 or 0.1 picomoles/cell/day 4,29 . However, in our model simulations, we converted the units of these rates to picomoles/lysosome/day assuming there are ~100 lysosomes in a cell. From here onwards, we interchangeably refer to these rates, 0.01 and 0.1 picomoles/cell/day, as 1-fold (1X) and 10-fold (10X) rates of lysosomal CFZ-H + Cl − accumulation, respectively. Simulating changes in lysosomal membrane proteins and cytoplasmic chloride amounts. In order to study the roles of V-ATPase, CLC7, and cytoplasmic chloride concentration on lysosomal accumulation of CFZ-H + Cl − , we varied the total numbers of active V-ATPase and CLC7 molecules per lysosome, and the cytoplasmic chloride concentration, while fixing the values of other lysosomal parameters at their baseline physiological input values in the presence as well as absence of lysosomal CFZ-H + Cl − accumulation. More specifically, to simulate the simultaneous inhibition of V-ATPase and CLC7, the total number of CLC7 molecules per lysosome was varied from 0 to 5,000 (resulting in 7-16 data-generating points) while, one parametric simulation at a time, the total number of V-ATPase molecules per lysosome was manually varied from 0 to 300 (resulting in 4-7 data-generating points). Moreover, to simulate the simultaneous inhibition of V-ATPase and cytoplasmic chloride, the cytoplasmic chloride concentration was varied from 0 to 10 mM (resulting in 16 data-generating points) while, one parametric simulation at a time, the total number of V-ATPase molecules per lysosome was manually varied from 0 to 300 (resulting in 4-7 data-generating points).
Using the aforementioned ranges of the adjustable lysosomal parameters, the corresponding lysosomal parameter inhibition range of 0 to 100% was calculated; where 0% represents no change from respective physiological baseline input value, and 100% represents the input value set to ~zero. The inhibition range was calculated as follows by comparing the corresponding input value (Adjusted Input Value) from the aforementioned given range with its respective physiological input value (Baseline Input Value): Baseline Input Value Adjusted Input Value Baseline Input Value 100% (19) Calculating Confirmation of steady-state and mass balance. For all of the aforementioned simulations, the final readout values were confirmed that they were steady-state values by performing the simulations for >24 h. We confirmed that mass balance was maintained in all of the simulations in the presence and absence of lysosomal CFZ-H + Cl − accumulation as long as a physiological pH gradient of up to 4.6 pH units was maintained across the lysosomal membrane.

Results
The weakly basic building block has pH-dependent solubility properties. CFZ is a weak base with two amines, which can be protonated depending on the pH of the immediate environment (Fig. 1a). To study its pH-dependent solubility behavior, we used an established approach 19 (see Supplementary Methods online for detailed explanation), which allowed an accurate measurement of the total solubility of CFZ-H + Cl − as a function of pH. From the experimental measurements, we calculated the solubility properties of CFZ, in an aqueous media at 25 °C, which include its apparent pK a,2 (pK a' = 6.08 ± 2.43 × 10 −3 ; 95% Confidence Interval = 6.07, 6.09) and intrinsic free base solubility (S 0 = 0.48 ± 4.05 × 10 −6 µM; 95% CI = 0.48, 0.48) (Fig. 1b). Moreover, to determine the effect of lysosomal Cl − (Fig. 1c) on the stability of the free base versus salt form of the drug, we performed drug stability experiments (Fig. 1d) in the presence and absence of lysosomal buffer (pH 4.5, Cl − concentration = 0-100 mM). We monitored drug stability by relying on the unique fluorescence profile of the different forms of CFZ: CFZ free base exhibits green and red fluorescence; peak excitation: 540-560 nm, peak emission: 560-600 nm, whereas CFZ-H + Cl − exhibits red and far red fluorescence; peak excitation: 560-600 nm, peak emission: 650-690 nm 11 . Indeed, the fluorescence of CLDIs resembles the fluorescence of the synthesized CFZ-H + Cl − crystals (peak excitation: 560-600 nm, peak emission: 650-690 nm) (see Supplementary Fig. S1). A low pH environment in the absence of Cl − (LB−) induced the destabilization of both CFZ-H + Cl − and free base CFZ, without affecting the stability of CLDIs. To the contrary, a low pH environment in the presence of a physiological lysosomal Cl − (LB+) induced the stabilization of CFZ-H + Cl − and CLDIs, whereas the free base CFZ precipitated out as CFZ-H + Cl − (Fig. 1d). Thus, these findings reveal that the precipitation of CFZ as a hydrochloride salt was highly sensitive to small, but physiologically relevant variations in pH and chloride concentrations (pH max = 4.5 ± 7.11 × 10 −15 , 95% CI = 4.5, 4.5; and K sp = 332.3 ± 3.71 µM 2 , 95% CI = 323.1, 341.5) (Fig. 1b,d); where the pH max reveals the pH below which CFZ-H + Cl − salt formation occurs, and the K sp elucidates the Cl − concentration, which is >(K sp ) 1/2 µM, that would further stabilize the salt form of the drug through a common-ion effect 35 . Thus, at the pharmacologically-relevant CFZ concentrations that have previously been measured in serum (1 to 20 µM), local variations in pH (~7.4 extracellular; ~7.2 cytosolic, and ~4.5 lysosomal) and chloride concentrations (~80 mM extracellular; ~10-20 mM in cytosol; and ~50-200 mM in lysosomes) can lead to a greater propensity of CFZ to precipitate as a hydrochloride salt in the lysosomal microenvironment (Fig. 1c,d).
The intracellular organization of the self-assembled mechanopharmaceutical device. Prolonged administration of CFZ results in its accumulation in both human and animals as microscopic, insoluble, and membrane-bound aggregates, known as Crystal Like Drug Inclusions (CLDIs), which are primarily found within tissue macrophages [3][4][5] . CLDIs are primarily comprised of a protonated, hydrochloride salt of CFZ (Fig. 2a) 7 , which forms faceted structures possessing three orthogonal cleavage and multiple fracture planes with lamellar spacing of 6 nm to 14 nm 21 (Fig. 2b). These orthogonal cleavage and fracture planes correspond to the orthogonal crystallographic planes of the CFZ-H + Cl − unit cell (Fig. 2c). When examined under polarized light (623 nm wavelength), the orthorhombic crystal structure of the deep red CLDIs 11,29,36 exhibits dichroism (Fig. 2d) 24 . Due to the biological origin of the CLDIs 7 , the interaction between the cell and crystal often leads to significant variations in its organization, impacting the manner in which it interacts with polarized light (Fig. 2d). Furthermore, reference spectra of pure CFZ and CFZ-H + Cl − identified characteristic Raman peaks for clear distinction between the two forms (Fig. 2e). Spectra obtained from cytoplasmic regions of untreated mouse alveolar macrophages were typical of biological specimens with no apparent contributions from drug. However, spectra obtained from cytoplasmic regions of 4-and 8-week CFZ-fed mouse alveolar macrophages revealed the presence of CFZ-H + Cl − , which dominated any Raman signal associated with biological specimens, and exhibited  unique peak intensities and peak widths, which attribute to the geometric organization of the supramolecular structures: peaks generally become narrower and more intense with higher degree of order in samples, which suggests that by 8 weeks, the macrophages have organized the accumulated drug into supramolecular crystalline packages of CFZ-H + Cl − salt. Moreover, Raman imaging of single-cells revealed cytoplasmic accumulation of both forms of the drug after just 1 week and continuing through 4 weeks; at 8 weeks CLDI formation was apparent and clearly distinguishable from free base CFZ, as evidenced by non-overlapping Raman signals in the biocrystal regions of Raman image (Fig. 2f).

Macrophages are necessary for the self-assembly of the mechanopharmaceutical device.
To investigate whether macrophages play an active role in stabilizing CLDIs in vivo, macrophages were depleted by injecting mice (n = 3-4 per treatment group) intraperitoneally with clodronate liposomes. This treatment depletes phagocytic cells without affecting non-phagocytic cells 20 . Moreover, the aforementioned CFZ's unique fluorescence characteristics were used to monitor the effect of macrophage depletion on the accumulation of CFZ-H + Cl − crystal. Following liposome therapy, the relative number of F4/80 positive macrophages within the liver and spleen of each treatment group was quantified using the total ratio of F4/80 to nuclear signal, revealing a significant reduction in the total macrophage population in the clodronate-CFZ fed group when compared to the PBS-CFZ treated group (p < 0.05, ANOVA, Tukey's HSD) (Fig. 3a). Consequently, using the ratio between the intrinsic far-red fluorescence of CFZ-H + Cl − 37 and total nuclear signal, there was a significant reduction in the accumulation and stabilization of CFZ-H + Cl − within the liver and spleen (p < 0.05, ANOVA, Tukey's HSD) compared to the PBS-CFZ treated group (Fig. 3b). More specifically, within the PBS-CFZ treated mice, the livers accumulated an average of 0.68 ± 0.27 mg CFZ/g tissue, while the clodronate-CFZ livers accumulated an average of 0.23 ± 0.09 mg CFZ/g tissue (p < 0.05, two-tailed Student's t-test); a reduction of nearly 66%. Similarly, the spleens of PBS-CFZ treated mice accumulated an average of 2.31 ± 0.40 mg CFZ/g tissue, compared to 0.70 ± 0.12 mg CFZ/g tissue in the clodronate-CFZ groups (p < 0.01, two-tailed Student's t-test); a reduction of 70%.
The individual drug loading within the macrophages of the liver and spleen was estimated based off of literature-reported values for macrophage populations in those two organs 26 and the total recovered drug within each organ, and there was no difference in total drug loading within either population of macrophage as a result of clodronate therapy (Fig. 3c). Furthermore, there was no significant difference in mean plasma concentrations of CFZ at the end of therapy (Fig. 3d). Taken together, this provides evidence that, while clodronate therapy was reducing the macrophage population and subsequent bioaccumulation within the liver and spleen, it was not impacting the general bioavailability and absorption of CFZ in the body. To determine how macrophages stabilized the self-assembled mechanopharmaceutical devices, isolated CLDIs were injected I.P. into clodronate and PBS liposome treated mice and collected at time points from 0 to 48 hours (n = 3 mice per group per time point). Without macrophages present in the peritoneal cavity to internalize the CLDI, there was a significant reduction in the estimated half-life of the biocrystal (p < 0.05, ANOVA) (Fig. 3e). This was confirmed via microscopic analysis of the peritoneal exudate 48 hours post-injection, with CLDIs becoming internalized by the macrophage and retaining their far-red fluorescence in the PBS-treatment group, while the clodronate-treated group showed reduced CLDI content within the exudate (Fig. 3f). Additionally, in order to confirm that clodronate did not impact drug accumulation within other macrophage populations, alveolar macrophages were collected and analyzed. Within the alveolar macrophage population, there was no significant difference in total cell number collected between the PBS and clodronate-treated groups (p = 0.48, two-tailed Student's t-test), as well as in the ability of the cells to accumulate CFZ in the form of a CLDI; where each alveolar macrophage population contained approximately 20% CLDI-positive cells (p = 0.36, two-tailed Student's t-test) (Fig. 3g). Thus, the specific effect of clodronate liposomes on liver and spleen CLDIs and CFZ content, as well as the increased rate of CLDI destabilization and solubilization in the absence of peritoneal macrophages provide evidence that these cells are directly responsible for the in vivo accumulation, self-assembly, and stabilization of CFZ-H + Cl − .
The intracellular accumulation, protonation, and self-assembly of the building block in cytoplasmic vesicles does not interfere with lysosomal pH. Taking advantage of the aforementioned CFZ's unique fluorescence properties, we studied the accumulation of the un-protonated free base and protonated hydrochloride salt form of the drug in macrophages. Incubation of macrophages with CFZ in vitro revealed vesicular staining pattern consistent with CFZ free base (peak excitation: 540-560 nm, peak emission: 560-600 nm) as well as CFZ-H + Cl − (peak excitation: 560-600 nm, peak emission: 650-690 nm) 11 . Both fluorescent forms were visible in CFZ-treated cells and absent in untreated, control cells ( Fig. 4a and see Supplementary Fig. S2). Based on the observed little overlap of their orthogonal fluorescence signals, CFZ-H + Cl − did not co-localize with free base CFZ (Pearson's Colocalization Coefficient -PCC between green and far-red fluorescence = 0.30 ± 0.29) (see Supplementary Fig. S3). In addition, when cells were co-incubated with both CFZ and Lysotracker Blue (LB), LB fluorescence signal (excitation: 373 nm, emission: 422 nm) was inhibited in the cytoplasm and displaced 38 to the nucleus (based on the observed nuclear staining), whereas the punctate green and far-red fluorescence of CFZ were clearly visible in the cytoplasm of these cells (Fig. 4c, and see Supplementary Figs S2 and S3). As a control experiment, we measured the lysosomal pH of macrophages following phagocytosis of CLDIs in macrophage-derived RAW264.7 cells as well as in peritoneal macrophages of 8-week CFZ-fed mice (Fig. 4d). In both cases, the lysosomal pH was not physiologically affected by the presence of the CFZ, as reflected by <0.1 pH unit difference between the drug-free and drug-containing peritoneal macrophages and RAW264.7 cells. Thus, CFZ uptake in macrophage lysosomes does not perturb physiological lysosomal pH homeostasis, which is required for various lysosomal functions, such as lysosomal trafficking and degradation of intrinsic and extrinsic materials within the lysosome by various pH-dependent hydrolytic lysosomal enzymes, which are essential for overall cellular homeostasis and viability 39 . Moreover, because the fluorescence signals of Lysotracker ® probes are pH-independent 40 , we interpret this to mean that intralysosomal CFZ accumulation interferes with Lysotracker probe accumulation through a mechanism that is independent of changes in lysosomal pH. The building block's self-assembly mechanism is biocompatible with the lysosome. To identify the candidate mechanism driving CFZ-H + Cl − accumulation and stabilization in lysosomes, we used a well-established systems-based mathematical model of lysosomal ion regulation. We first performed a simultaneous variation of the number of active V-ATPase and chloride channels per lysosome (resulting in 0 to 100% inhibition of both parameters from their respective baseline physiological input values of 300 and 5000) in the presence and absence of dose-dependent lysosomal CFZ-H + Cl − accumulation (0 to 0.01 picomoles/cell/day), as measured in vitro 4,29 . The simulation results indicated that CFZ-H + Cl − accumulates at the measured rate of 0.01 picomoles/cell/day exerting a negligible effect on lysosomal pH, Cl − , and membrane potential from respective physiological baseline readout values (4.53 pH units, 224.6 mM, and 0.79 mV). Only a significant (≥97%) reduction in the total V-ATPase amount per lysosome led to changes in these values (Fig. 5a). Consistent with this, <97% V-ATPase inhibition per lysosome was required for the physiological accommodation of lysosomal CFZ-H + Cl − accumulating at 0.01 picomoles/cell/day (Fig. 5b). However, unlike the number of V-ATPase, the number of chloride channel molecules per lysosome did not show any effect on the physiological CFZ-H + Cl − accumulation unless this chloride current was completely absent (Fig. 5b).
In dose-dependent simulations, CFZ-H + Cl − accumulating at a ten-fold greater (non-physiological) rate of 0.1 picomoles/cell/day led to more pronounced changes in lysosomal pH, Cl − , and membrane potential (Fig. 6a). At least <68.3% V-ATPase inhibition per lysosome was required to sustain physiological ion concentrations in the presence of this 10-fold greater rate of CFZ-H + Cl − accumulation (Fig. 6b). In contrast, variations in the numbers of chloride channels did not affect the physiological lysosomal accumulation of CFZ-H + Cl − at a rate of 0.1 picomoles/cell/day, unless there was a complete inhibition of the chloride current (Fig. 6b).
V-ATPase is essential in the building block accumulation and self-assembly. To verify our model and simulation findings, we performed experiments to detect the effects of V-ATPase and chloride channels on the intracellular accumulation of CFZ. All experiments regarding the incubation and accumulation of CFZ were performed under conditions in which cellular viability was maintained. In the absence of any inhibitor, 8 µM extracellular CFZ accumulated at a rate of 0.01 picomoles/cell/day, as previously reported 4,29 . However, inhibiting the V-ATPase with BafA1 reduced cellular CFZ accumulation by 80 ± 13% (p < 0.01) in a 4 h incubation, relative to control, uninhibited cells (Fig. 7a). This observation is consistent with our simulation results, where up to 50% V-ATPase inhibition resulted in a lysosomal pH increment (Fig. 5a), which associates with a reduction in cellular drug accumulation. In contrast, when cells were treated with the chloride channel inhibitor NPPB and exposed to CFZ, no significant change was measured in the accumulation of CFZ following a 4 h incubation (Fig. 7b). This contrasts with the effect observed with BafA1 (Fig. 7a), and is consistent with our simulation results (Fig. 5a), thereby indicating that in the presence of CFZ uptake, lysosomal ion homeostasis is insensitive to chloride channel inhibition.

Discussion
As a weakly basic, poorly soluble small-molecule chemical that is FDA-approved, orally bioavailable, slowly cleared, and administered at high doses for prolonged periods of time, CFZ affords us the opportunity to reverse engineer the construction of a mechanopharmaceutical device that was found to form naturally inside cells. We report that, because of its physicochemical properties, CFZ is highly prone to precipitation in acidic  . Intracellular CFZ accumulation. Total intracellular CFZ accumulation in picomoles (pmoles) in RAW264.7 cells (~68,000 cells/well) when pre-incubated with (a) BafA1 and (b) NPPB and initial concentration of CFZ = 4, 6, and 8 μM at t = 4 hours. Significant inhibition of CFZ accumulation was measured at t = 4 hours when cells were exposed to BafA1. To the contrary, no such inhibition was measured when cells were exposed to NPPB at t = 4 hours. Data are the mean ± S.D. of six experiments. *p < 0.05, **p < 0.01, ***p < 0.001. Each data point was compared in a pair-wise independent Student t-test. endolysosomal compartments in macrophages, where it self-assembles into a highly ordered CFZ-H + Cl − biocrystal, possessing many interesting optical and biomechanical features. Indeed, the preferential accumulation of CFZ in macrophage lysosomes is evident based on the displacement of Lysotracker Blue fluorescence from the cytoplasmic vesicles (Fig. 4). Furthermore, fluorescence imaging revealed that CFZ was present both as a free base and hydrochloride salt form within intracellular vesicles (Fig. 4). Thus, we performed computational simulations and experiments to probe the relationship between CFZ's pH-dependent solubility properties, the formation of salt and free base crystals, and the molecular, ion transport mechanisms responsible for this mechanopharmaceutical phenomenon, which could be exploited for developing other weakly basic small molecule building blocks as cell-targeted, self-assembling mechanopharmaceutical devices 41,42 .
Despite the complexities of lysosomal ion homeostasis, computational simulations and experimental results provide a straightforward explanation for the observed behavior of CFZ: the number of V-ATPase molecules that drive the lowering of macrophage lysosomal pH 43 plays a critical role in determining the intracellular bioaccumulation of CFZ in lysosomes without affecting lysosomal ion homeostasis. Our results show that very high levels of V-ATPase expression, that are typical of macrophages, are both necessary and sufficient to explain the accumulation and stabilization of CFZ-H + Cl − salt in these cells because they allow the cells to maintain physiological lysosomal pH and membrane potential in spite of the accumulation of the drug. In addition, the intrinsic solubility properties of CFZ are such that the protonated form of the drug strongly interacts with chloride in a concentration range that is typically observed in lysosomes, which leads to the physiological accumulation of the drug in its hydrochloride salt form. Accordingly, V-ATPase expression establishes the ability of the cell to maintain physiological lysosomal pH 44 , which, in relation to the drug's pH max and K sp (Fig. 1), can sustain the continuous accumulation of the hydrochloride salt form of the drug 19 in the lysosome.
More importantly, we found that experimental results 4,29 were highly consistent with model simulation results. First, the inhibition of V-ATPase activity resulted in the reduction of the total intracellular CFZ accumulation (Fig. 7a), which was consistent with the model simulation results (Figs 5 and 6). In contrast, NPPB-mediated reduction of chloride channel activity did not affect the accumulation and stabilization of the salt form of the drug (Fig. 7b), which was also consistent with the model simulation results (Figs 5 and 6, and see Supplementary Figs S5 and S6). Furthermore, the measured K sp of CFZ-H + Cl − and our CLDIs/CFZ-H + Cl − stability results ( Fig. 1) indicate that the physiological lysosomal chloride levels are sufficient to stabilize CFZ-H + Cl − salt formation through a common-ion "salting-out" mechanism 35 . Therefore, our findings collectively provide the first mechanistic framework that explains the accumulation of highly ordered, self-assembled CFZ-H + Cl − salt biocrystals in macrophages (Figs 2 and 3); which depends on the high levels of V-ATPase expression that is unique to this cell type 16 .
It is also noteworthy that macrophages have lysosomes which are an order of magnitude greater in size than that of other cell types 45 . This provides them with the capability to accommodate massive accumulation of weakly basic drugs 46 . Even in cases where the initial lysosomal size is small (<1 μm, in radius), the macrophage lysosomes can undergo biogenesis 47,48 . A master regulator of lysosomal biogenesis and V-ATPase expression is transcription factor EB (TFEB) 39,46,49 . Moreover, given the recent findings regarding the role of TFEB in upregulating lysosomal genes during inflammation 50 , the role of TFEB in the bioaccumulation and stabilization of CFZ-H + Cl − salt in macrophage lysosomes, in relation to CFZ's anti-inflammatory activity 9,10 , could be further investigated.
To conclude, we have discovered how lysosomal ion transport pathways play a major role in the self-assembly of a biocompatible, cell-targeted supramolecular structure formed by a weakly basic, small molecule building block. Together with the physicochemical properties of the freely soluble building blocks, these transport pathways can be exploited to develop novel types of drug depots and cell-directed drug delivery systems [51][52][53] . Indeed, interest has been raised in macrophages as drug delivery vehicles. As our studies have shown, macrophages may allow drugs to be locally stabilized and undergo a sustained release in a site-specific manner, offering a strategy for reducing off-target side effects 54,55 . Macrophages are also one of the few cell types, which are primarily involved in the phagocytosis of foreign particles of different sizes and shapes, including drugs 56,57 , which offer yet another opportunity for loading these cells with massive amounts of chemotherapeutic agents. To this end, additional investigations into the manner in which these cells mechanically synergize with the self-assembled weakly basic small molecule building blocks in order to stabilize functional mechanopharmaceutical devices are warranted.