Drug screening in human physiologic medium identifies uric acid as an inhibitor of rigosertib efficacy

The nonphysiological nutrient levels found in traditional culture media have been shown to affect numerous aspects of cancer cell physiology, including how cells respond to certain therapeutic agents. Here, we comprehensively evaluated how physiological nutrient levels affect therapeutic response by performing drug screening in human plasma-like medium. We observed dramatic nutrient-dependent changes in sensitivity to a variety of FDA-approved and clinically trialed compounds, including rigosertib, an experimental cancer therapeutic that recently failed in phase III clinical trials. Mechanistically, we found that the ability of rigosertib to destabilize microtubules is strongly inhibited by the purine metabolism end product uric acid, which is uniquely abundant in humans relative to traditional in vitro and in vivo cancer models. These results demonstrate the broad and dramatic effects nutrient levels can have on drug response and how incorporation of human-specific physiological nutrient medium might help identify compounds whose efficacy could be influenced in humans.


INTRODUCTION
Tumor growth is influenced by both cell intrinsic and cell extrinsic factors, and nutrient availability is emerging as a critical environmental factor that can shape the metabolic fitness and proliferative capacity of tumors (1)(2)(3).Concurrent with these discoveries has been a renewed realization that standard cell culture media were not designed to mimic the nutrient environment found in vivo, but were rather designed to provide excess amounts of the minimal nutrients required to sustain cancer cell growth in vitro (4)(5)(6)(7)(8)(9).As a result, the nutrients present in these traditional culture media do not accurately recapitulate the complexity or abundance of nutrients found in vivo.Importantly, it has recently been shown that the highly non-physiological nutrient levels found in culture media can contribute to inconsistencies between in vitro and in vivo experiments, especially for those directly related to cellular metabolism (10)(11)(12)(13).It has also been observed that nutrient availability can affect the response to a variety of cancer therapies (14), including traditional chemotherapies (15,16), metabolic inhibitors (17,18), targeted therapies (19,20), and immunomodulatory checkpoint inhibitors (21).
Because of the importance of nutrient availability in influencing tumor metabolic phenotypes and therapeutic vulnerabilities, there is considerable interest in targeting systemic metabolism either alone or in combination with existing therapies to treat cancer.This includes several dietary interventions that are being investigated as components of cancer therapies (22,23), including amino acid starvation (24)(25)(26), ketogenic diet (27,28), caloric restriction (29)(30)(31), and fasting-mimicking diets (32).The success of dietary intervention studies is critically dependent on mouse models and has led to tremendous interest in translating these findings to patients.And while mouse models provide an essential platform to study interactions between systemic and tumor metabolism, there are a number of metabolic differences between mice and humans that influence tumor biology (33)(34)(35)(36), including how cancer cells respond to cancer therapeutics (15).These issues have motivated the development of novel culture media that specifically mimic the nutrient composition found in human plasma as platforms for studying therapeutic response under more physiological human nutrient conditions (12,15).
Here, we sought to determine the extent to which nutrient availability impacts the sensitivity of cancer cells to diverse therapeutic agents by utilizing a high-throughput differential sensitivity drug-screening platform to profile therapeutic sensitivity in cancer cells growing in traditional versus physiological human plasma-like medium (HPLM).This screen revealed dramatic nutrient-dependent changes in sensitivity to a wide variety of drugs in cells cultured in HPLM.Among these differences were changes in sensitivity to the experimental therapeutic rigosertib (ON-01910), the efficacy of which was strongly antagonized by the purine degradation product uric acid.

Drug screening identifies nutrient-dependent effects on drug response
Commercial culture media such as DMEM and RPMI-1640 (RPMI) contain nutrients at non-physiological levels and lack many critical components present in human plasma (37).Due to these deficiencies, several labs have recently developed media that more accurately mimic physiological nutrient levels found in human circulation (10)(11)(12)15).We have made use human plasma-like medium (HPLM) (15) to culture breast cancer cell lines, where after a two-week adaptation period we observed similar or slightly decreased growth rates (Fig. S1a), and consistent remodeling of intracellular metabolite abundance (Fig. S1b, c).Because of the recent observation of the impact of nutrient availability and cellular metabolism on the response to a variety of cancer therapies (15,16,20), we hypothesized that culturing cancer cells in HPLM would change how they respond to therapeutic agents on a larger scale.To address this question, we utilized a highthroughput differential sensitivity drug screening platform containing a library of 626 metabolic inhibitors and anticancer compounds arrayed in 10-point dose curves (38).This platform contains compounds targeting a wide variety of cancer-relevant pathways, many of which are FDA-approved or have been evaluated in clinical trials.We screened the triple-negative breast cancer (TNBC) cell line SUM149 growing in RPMI or HPLM media, where we observed dramatic changes in sensitivity to a variety of compounds (Fig. 1a, b and Table s1).Interestingly, while very few drugs are more effective in HPLM, a large proportion of drugs are less effective in physiological medium.
Among the drugs most strongly affected by culture in HPLM are four inhibitors of the de novo purine biosynthesis pathway -lometrexol, azathioprine, 6-thioguanine (6-TG), and 6-mercaptopurine (6-MP) -all of which are less effective at reducing cell numbers in HPLM (Fig. 1a-f).We found that both SUM149 cells and another TNBC cell line, HCC1806, are able to proliferate when treated with lometrexol in HPLM, but not RPMI (Fig. 1g, h).Based on these observations, we investigated the level of purine nucleotides by LC-MS analysis in HCC1806 cells treated with and without lometrexol in both RPMI and HPLM.As expected, we found that lometrexol causes a large drop in the abundance of most purine nucleotides in RPMI; however, this drop is significantly blunted in HPLM (Fig. 1i).In addition to de novo biosynthesis, cells can acquire purines through the purine salvage pathway (Fig. 1j), and the presence of substrates for the purine salvage pathway, such as hypoxanthine, are known to reduce the efficacy of purine synthesis inhibitors (39,40).While traditional media formulations do not contain salvage substrates, HPLM contains hypoxanthine at 10 µM as is found in human plasma.Indeed, we found that addition of hypoxanthine to RPMI is sufficient to provide resistance against these compounds (Fig. 1k-n), and the removal of hypoxanthine from HPLM strongly increases the sensitivity of SUM149 cells to purine biosynthesis inhibitors and (Fig. 1o-r).While the ability of hypoxanthine to rescue purine synthesis inhibitors is known, these results demonstrate the power and utility of our screening platform to identify physiological nutrients that modify cancer cell sensitivity to therapeutic agents.

Uric acid in HPLM reduces cancer cell sensitivity to rigosertib in vitro
Another top hit from our screen was the experimental cancer therapeutic rigosertib, which is significantly less effective against cells growing in HPLM than in RPMI (Fig. 1a, 2a).
We validated these results by performing rigosertib dose-response analyses in HCC1806 and SUM149 cells, where we observed 2711 and 283-fold increases (respectively) in the IC50 for rigosertib in HPLM (Fig. 2b, c).Similar results were obtained in two lung cancer cell lines, A549 and Calu6, suggesting that this effect is likely general and not restricted to breast cancer cells (Fig. 2d, e).Rigosertib's anti-cancer effects have been shown to be mediated by induction of both G2/M cell cycle arrest and cell death (41,42).Accordingly, we treated the HCC1806 cells with 150 nM rigosertib, a dose we identified to reduce cell number only in RPMI (Figure 2b), and observed that rigosertib induces phosphorylation of histone H3, G2/M cell cycle arrest in RPMI but not in HPLM.Similarly, induction of cell death by treating HCC1806 cells with 200 nM rigosertib was blocked in HPLM (Fig. 2f-i).
Next, we sought to determine the component(s) of HPLM that antagonizes rigosertib activity.Like RPMI and other traditional media, HPLM consists of glucose, amino acids, and salts, albeit at different concentrations (15).HPLM also contains 27 additional components that are not found in RPMI but are found in human plasma.Most of these unique ingredients are organized into 11 stock solutions numbered 8 through 18.To determine whether a unique component of HPLM is responsible for the reduced sensitivity to rigosertib, we combined HPLM stocks 8-18 and added them to RPMI and performed dose-curve analyses, where we found that stocks 8-18 were able to recapitulate the effect of HPLM on rigosertib sensitivity (Fig. 3a, b).We then analyzed stocks 8-18 individually and found that addition of stock 18 alone was sufficient to protect against rigosertib in RPMI (Fig. 3c, d).Importantly, stock 18 contains only one component: the purine metabolism waste product uric acid, which is present in human plasma and HPLM at 350 µM.Indeed, we found that removal of uric acid from HPLM was sufficient to dramatically increase cancer cell sensitivity to rigosertib (Fig. 3e).We confirmed the broad protective effects of uric acid by performing rigosertib dose curves on multiple cancer cell lines of different origin, including lung, renal, and CML, where we observed the protective effects of uric acid in all cases (Fig. s2).To determine whether uric acid protect cells from rigosertib in a dose-dependent manner, we performed a dose curve of uric acid in RPMI in the presence of 80 nM rigosertib.Interestingly, we found that uric acid concentrations as low as 27 µM are able to partially protect against rigosertib (Figures 3f, g).Similar to HPLM, physiological concentrations of uric acid alone were sufficient to block the ability of 150 nM rigosertib from inducing histone H3 phosphorylation, G2/M cell cycle arrest.
Similarly, induction of cell death by treating HCC1806 cells with 200 nM rigosertib was blocked in the presence of uric acid (Fig. 3h-k).

Uric acid inhibits the microtubule destabilizing activity of pharmaceutical grade rigosertib
While the mechanism of action of rigosertib remains controversial (43)(44)(45)(46), several recent reports have demonstrated that rigosertib is a microtubule destabilizing agent that binds tubulin dimers at the colchicine binding site (41,47,48).To verify the ability of rigosertib to destabilize microtubules, we performed short-term (4 hr) treatments of HCC1806 and SUM149 cells cultured in RPMI with increasing doses of commercial-grade rigosertib, where we observed increased levels of α-tubulin in the soluble fraction of cell lysates, suggesting that rigosertib does indeed destabilize microtubules (Figs.4a, b and S3a, b).Importantly, however, the ability of rigosertib to destabilize microtubules in cells grown in HPLM is strongly inhibited (Fig. 4a, b and S3a, b).To determine whether uric acid prevents rigosertib-mediated microtubule destabilization, we treated cells cultured in HPLM with and without 350 µM uric acid with commercial-grade rigosertib, which results in a dose-dependent increase in the level of soluble α-tubulin only in the absence of uric acid (Fig. 4c, d S3c, d).
Previous work has shown that the presence of a contaminant in commercial-grade rigosertib may contribute to its anti-cancer effects (49).To determine whether HPLM blocks the effect of rigosertib or a potential contaminant, we made use of pharmaceuticalgrade rigosertib that lacks the potentially active contaminant.Similar to commercial-grade rigosertib, culture of cells in HPLM strongly reduces the cellular sensitivity to pharmaceutical-grade rigosertib (Fig. 4e, f).Importantly, addition of 350 µM uric acid to RPMI prevented sensitivity to pharmaceutical-grade rigosertib in a panel of renal cancer cell lines (Fig. 4g, h), indicating that uric acid is protective against rigosertib and not a contaminant found in the commercial-grade compound.Similarly, treatment of 786-O cells with pharmaceutical-grade rigosertib results in a dose dependent decrease in αtubulin found in the pellet (microtubule fraction) when compared to total tubulin in RPMI, but the addition of uric acid to RPMI prevented rigosertib-mediated destabilization of microtubules (Fig. 4i, j).

Uric acid may weaken the rigosertib:b-tubulin interaction
The acute ability of uric acid to prevent rigosertib-mediated destabilization of microtubules motivated us to explore the potential molecular effects of rigosertib and uric acid on tubulin structure.As a benchmark, we compared rigosertib to colchicine in our analyses.We started by performing molecular dynamics (MD) simulations of colchicine-bound, rigosertib-bound and apo-tubulin (non-drug bound control).After equilibrating each structure for 0.5 μs, we then performed four independent simulations of each complex resulting in more than 6 μs of total simulation time.Using principal component analysis (PCA) to evaluate the large-scale structural differences induced by colchicine and rigosertib, we found that both colchicine and rigosertib produce a similar "kink" in the dimer that likely explains their ability to prevent microtubule polymerization (Fig. S4 and Movies S1 & S2).In addition, rigosertib induces a unique conformational change that alters the relative orientation of α and β-tubulin (Fig. S4, and Movies S1 & S2).
Specifically, the colchicine-bound simulation features a persistent salt bridge formed between αR221 and bE328 that is directly adjacent to the colchicine binding site (Fig. 5a).
Since rigosertib alters the intradimer interface and creates a greater distance between αR221 and bE328 (Fig. 5b), this salt bridge cannot be formed in rigosertib-tubulin (Fig. 5a).Helix H10 in b-tubulin contains E328, and loss of this salt bridge makes H10 more dynamic and creates a pocket between H10 and strand S9 (Fig. 5c).Docking studies revealed that uric acid could potentially bind within this pocket via hydrogen bonding with residues in both H10 and S9 (Fig. 5c).Importantly, S9 also interacts with the carboxyl group of rigosertib, and using free energy calculations we found that the likely effect of uric acid binding would be to weaken the binding affinity of rigosertib to β-tubulin.We evaluated this possibility using the cellular thermal shift assay (CETSA) (50), where we observed denaturation and precipitation of both α and β-tubulin at 60°C that was strongly reduced in the presence of rigosertib, suggesting that rigosertib is capable of binding to tubulin (Fig. 5d, e), as has been shown by multiple other labs (41,47,48).However, addition of uric acid to the culture media significantly reduces the stabilization of α and βtubulin by rigosertib (Fig. 5d, e).This data is suggestive of a potential mechanism by which uric acid antagonizes rigosertib activity by weakening the interaction between rigosertib and b-tubulin, thereby acting as an uncompetitive inhibitor.However, additional studies will be required to determine whether uric acid directly interacts with rigosertibbound β-tubulin or functions through other mechanisms.

DISCUSSION
Despite promising pre-clinical data and extensive evaluation in early stage studies, rigosertib has thus far failed to improve outcome in the two phase 3 clinical trials in which it has been investigated (51,52).And while there are likely numerous factors that have contributed this poor clinical performance, our discovery that uric acid strongly antagonizes the microtubule-destabilizing activity of rigosertib in vitro suggests that the elevated levels of uric acid characteristic of humans may also contribute.Most species, including mice and others commonly used in cancer research (e.g., bovine serum), have a functional uricase gene that converts uric acid to the more soluble allantoin, resulting in relatively low circulating uric acid concentrations (Fig. 5d).However, due to the evolutionarily recent pseudogenization of the uricase gene in humans and other closely related apes, humans have circulating uric acid levels that are an order of magnitude higher than other mammals (Fig. 5d) (53)(54)(55)(56)(57)(58).Further, cancer patients, including those with myelodysplastic syndromes where rigosertib has been most thoroughly investigated, often present with hyperuricemia (59, 60), and therefore they may have uric acid levels that are even higher than those found in HPLM.Importantly, given that uric acid is the underlying cause of gout, there are numerous approved therapies to reduce uric acid levels in patients.Our work suggests that such therapies, including a low-purine diet, xanthine oxidase inhibitors (e.g., allopurinol, febuxostat), and uric acid degrading enzymes (e.g., rasburicase, pegloticase) could be candidates to improve the therapeutic response to rigosertib (60)(61)(62)(63).
As previously mentioned, identifying the precise mechanistic target of rigosertib has been challenging.Rigosertib was initially identified as a PLK1 inhibitor (41,(64)(65)(66)(67)(68), and has also been proposed to act as an inhibitor of RAS (69) and PI3K (70).Use of an unbiased CRISPRi/a chemical-genetic approach combined with structural biology studies identified rigosertib as a microtubule destabilizing agent that binds to the colchicine binding site on β-tubulin (47,48), a finding that has now been corroborated by other groups (45).Our molecular dynamic simulation studies also suggest that rigosertib binds to the colchicine binding site of β-tubulin and induces conformational changes that are similar to, but distinct from, those induced by colchicine.It is important to note, however, that the crystal structures of tubulin with colchicine or rigosertib are strongly affected by the presence of stathmin, and this limits the interpretation of these structures (47).In addition, docking results suggest that uric acid may act as an uncompetitive inhibitor of rigosertib through interaction with residues in loop S9 -a conclusion that is further confirmed by our CETSA results.Together, these results suggest that the effect of uric acid on rigosertib efficacy is mediated through microtubules and not the other proposed targets of rigosertib.
In vitro tissue culture models offer several advantages over in vivo tumor models, including the ability to perform large-scale screening studies.However, there has always been a large bottleneck of promising in vitro cancer findings that turn out to be irrelevant in human tumors.While there are many factors that contribute to this bottleneck, our work and that of others has shown that the non-physiological nutrient levels found in traditional culture media likely contribute to some in vitro and in vivo discrepancies.Importantly, unnatural nutrient levels are not an inherent problem of tissue culture, and it is becoming clear that replacement of traditional media with more physiological media can rectify some of the problems with tissue culture systems.And while mouse models will continue to be the gold-standard of pre-clinical cancer studies, it is important to consider that there are also differences between mice and humans that could contribute to discrepancies in how cancer cells respond to therapies.Our work demonstrates that use of human physiological media can lead to identification of drug:metabolite interactions that otherwise might be missed using traditional tissue culture or mouse models, and we support the notion that use of physiological media is a highly valuable addition to the cancer research pipeline.

Cell lines
Cell lines were acquired from the Brugge Lab at Harvard Medical School, Boston, MA (HCC1806, SUM149), the Kim Rathmell Lab at Vanderbilt University Medical Center, Nashville, TN (A498 and Caki2), the Vadim Gaponenko (K562) Lab at the University of Illinois at Chicago, Chicago, IL, and the ATCC (A549 and Calu6).Cell lines were tested for mycoplasma using the MycoAlert Mycoplasma Detection Kit (Lonza) and were authenticated by STR analysis.Cells were grown in human plasma-like medium according to the published formulation (15) with 5% dialyzed fetal bovine serum (FBS) (Sigma) and pen/strep (Invitrogen) at 37°C with 5% CO2.Media was changed at least every two days.As needed, cells were incubated in RPMI media (Thermo Fischer, 11875-093) with or without uric acid (Sigma) with 5% dialyzed FBS and pen/strep.

Growth curve analysis
For growth curves 10,000 cells were plated in 12-well plates and were treated with the indicated drugs the following day.Fresh media and drug were added every 2 days.After 5-7 days cells were counted using a Z1 Coulter Particle Counter (Beckman Coulter).

Intracellular tubulin polymerization assay
50,000 cells per well were plated in a 12 well plate, 24 hours before treatment with increasing concentrations of rigosertib with and without uric acid (Sigma U2625) in corresponding media.After drug treatment, the cells were lysed in a hypotonic lysis buffer per well of 384-well plates.After 24 hr, a Seiko Compound Transfer Robot pin transferred 100 nL of each drug library into wells with plated cells.Following pin-transfer, 20 µL of cell culture medium was added to all wells, resulting in each drug being applied at a final 10-point concentration series ranging from 20 µM to 1 nM.After 72 hr of drug treatment, the cells were washed with PBS, fixed with 4% formaldehyde, and stained with 5 mg/mL bisbenzimide.An Acumen Cellista plate cytometer was used to image plates and determine the cell numbers in individual wells.XY plots were generated comparing relative numbers of surviving RPMI and HPLM cells with concentrations of each drug tested.Area under the curve (AUC) values were calculated for each plot and drugs were ranked based on the difference between the AUCs for RPMI and HPLM cells.

LC-MS metabolite analysis
LC-MS metabolite analysis was performed as previously described (71).Metabolites were extracted using 80% ice cold methanol.A Vanquish UPLC system was coupled to a Q Exactive HF (QE-HF) mass spectrometer equipped with HESI (Thermo Fisher).
The mobile phase gradient was as follows: 0-13min: 80% to 20% of mobile phase B, 13-15min: 20% of mobile phase B. ESI ionization was performed in both positive and negative modes.The MS scan range was 60-900m/z.The mass resolution was 120,000 and the AGC target was 3x10 6 .The capillary voltage was 3.5 KV and the capillary temperature was 320°C.5 µL of sample was loaded.LC-MS peaks were manually identified and integrated with EL-Maven (Elucidata) by matching with an in-house library.
MetaboAnalyst was used to normalize the peak areas of target metabolites to the median fold change across all identified metabolites, calculate fold changes, and calculate pvalues.

GC-MS metabolite analysis
Polar metabolites were prepared for analysis by first drying the samples in individual microcentrifuge tubes, then adding 15 µL of methoxy amine in pyridine (MOX) (Thermo Fisher) and incubating at 40ºC for 90 min.The samples were then further incubated with 20 µL of N-(tert-butyldimethylsilyl)-N-methyl-trifluoroacetamide with 1% tert-Butyldimethylchlorosilane (TBDMS) (Sigma) at 60 ºC for 60 min.The resulting derivatized solution was vortexed briefly, centrifuged, and transferred into polypropylene GC/MS vials (Agilent).Subsequent metabolite abundance analysis was conducted using an Agilent 6890N GC coupled with 5975B Inert XL MS.An Agilent J&W DB-35 ms column was used.
Chromatography grade helium (Airgas) was used as the carrier gas, flowing at a rate of 1 mL/min.Depending on sample abundances, either 1 or 2 µL of samples were injected using either split or splitless modes.The 6890N GC inlet temperature was set to 270 ºC, and the oven temperature was initially set to 100 ºC, then raised to 300 ºC at a rate of 2.5 ºC/min.Electron ionization mode with 70 eV was used for 5975B MS measurement.
Acquisition was performed using scan mode with a detection range of 150-625 m/z.Mass isotopomer distributions (MIDs) were corrected for natural isotope abundance.Detailed methods are as published (72).each) for docking of uric acid.The following workflow was used for each structure: 1) the ligand (uric acid) and protein (tubulin-drug complexes) were prepared to be compatible with Maestro applications using LigPrep and ProteinPrep, respectively, 2) generated possible binding sites on the tubulin complex using SiteMap, 3) created receptor grids to be used for docking via Glide, and 4) docked the prepped ligand (uric acid) to the tubulin complex using ligand docking by Glide.To evaluate the binding free energy of rigosertib with and without uric acid, we utilized the Molecular Mechanics Generalized Born Surface Area methods (MMGBSA) in Prime.

Cellular thermal shift assay (CETSA)
K562 cells were treated with 40 µM Rigosertib for 4 hr in the corresponding media after which cells were washed with 1X PBS.Next, cells were resuspended in PBS containing 1X Halt protease inhibitor cocktail (Thermo Fisher, PI87786) and counted using a Z1 Coulter Particle Counter (Beckman Coulter).700,000 cells were dispensed in PCR tubes and heated at the indicated temperatures for 3 minutes in a thermocycler.After heating, cells were cooled to 20°C and lysed by thee cycles of freeze/thaw in liquid nitrogen.
Following lysis, denatured proteins were separated by centrifugation at 15,000 rpm for 10 min at 4°C.The lysate was dissolved in 6X loading buffer and run on SDS-PAGE as described in the western blot section.

Fig 1 .
Fig 1. Culture in HPLM changes sensitivity to a variety of therapeutic agents (a) Percent difference in the area under curve (% Difference in AUC) data for SUM149 cells cultured in either RPMI or HPLM after treatment with anti-cancer and metabolic inhibitor libraries.Only compounds with an Emax >50% in either medium are shown.(b) The same data as in (a) categorized based on target pathway.(c -f) Dose-response curves of the purine biosynthesis inhibitors lometrexol (c), azathioprine (d), 6-mercaptopurine (e), and 6-thioguanine (f) on SUM149 cells growing in RPMI vs HPLM.(g & h) Growth curves of HCC1806 (g) and SUM149 (h) cells treated with lometrexol in RPMI vs HPLM.(i) LC-MS analysis to quantify purine nucleotide abundance in HCC1806 cells treated with lometrexol in RPMI vs HPLM.* indicates p < 0.05 for HPLM + lometrexol relative to RPMI + lometrexol (unpaired two-tailed t-test).(j) Schematic representation of purine synthesis and salvage pathways.(k -n) Dose-response curves of the purine biosynthesis inhibitors lometrexol (k), azathioprine (l), 6-mercaptopurine (m), and 6-thioguanine (n) on SUM149 cells grown in RPMI with and without hypoxanthine (HXN).(o -r) Dose-response curves of the purine biosynthesis inhibitors lometrexol (o), azathioprine (p), 6-mercaptopurine (q), and 6-thioguanine (r) on SUM149 cells grown in HPLM with and without hypoxanthine (HXN).For all panels data represents the means ± SD of triplicate samples.

Fig 2 .
Fig 2. Culture in HPLM reduces sensitivity to rigosertib (a) Dose response curve of SUM149 cells treated with rigosertib from the high-throughput screen described in Fig 1.Data are the mean ± SD of triplicate samples.(b -e) Dose response curves for rigosertib treatment of HCC1806 (b), SUM149 (c), A549 (d) and Calu6 (e) cells growing in RPMI vs HPLM.Data are the mean ± SD of triplicate samples.(f) Representative western blot of phospho-Histone H3 in HCC1806 cells treated with 150 nM rigosertib in RPMI vs HPLM.(g & h) Cell cycle analysis of HCC1806 cells treated with 150 nM commercial-grade rigosertib in RPMI (g) and HPLM (h).(i) Cell death analysis of HCC1806 cells treated with 200 nM commercial-grade rigosertib in RPMI vs HPLM.Cell death and cell cycle data are the means ± SD of triplicate samples.* indicates p < 0.05 from unpaired two-tailed t-test.NS (not significant) indicates p > 0.05.

Fig 3 .
Fig 3. Uric acid prevents the activity of rigosertib (a & b) Dose response curves of HCC1806 (a) and SUM149 (b) cells treated with rigosertib in RPMI vs RPMI + HPLM stocks 8-18.(c & d) Cell growth assays of HCC1806 (c) and SUM149 (d) cells treated with 80 nM rigosertib in the presence of individual HPLM stocks 8-18.R = RPMI and H = HPLM.(e) Dose response curve of MCF7 cells treated with rigosertib in HPLM vs HPLM -UA.(f & g) Dose response curves of uric acid on HCC1806 (f) and SUM149 (g) cells treated with 80 nM rigosertib.(h) Representative western blot of phospho-Histone H3 in HCC1806 cells treated with 150 nM rigosertib in HPLM vs HPLM -UA.(i & j) Cell cycle analysis of HCC1806 cells treated with 150 nM commercial-grade rigosertib in HPLM (i) and HPLM -UA (j).(k) Cell death analysis of HCC1806 cells treated with 200 nM commercial-grade rigosertib in HPLM and HPLM -UA.For all panels, data is represented as mean ± SD of triplicate samples.* indicates p < 0.05 from unpaired two-tailed t-test.NS (not significant) indicates p > 0.05.

Fig 4 .
Fig 4. Uric acid inhibits the microtubule destabilizing activity rigosertib (a) Western blot of soluble α-tubulin from SUM149 treated with increasing doses of rigosertib (0.1 µM, 0.5 µM and 1 µM ) for 4h in RPMI and HPLM.(b) Quantification of western blots from (a).Data is represented as mean ± SD from three independent experiments.**** indicated p < 0.0001, * indicates p < 0.05 from one way ANOVA followed by Tukey's multiple comparison test.NS (not significant) indicates p > 0.05.(c) Western blot of soluble α-tubulin from SUM149 treated with increasing doses of rigosertib (0.1 µM, 0.5 µM and 1 µM) for 4h in HPLM and HPLM -UA.(d) Quantification of western blots from (c).Data is represented as mean ± SD from three independent experiments.* indicates p < 0.05 from one way ANOVA followed by Tukey's multiple comparison test.NS (not significant) indicates p > 0.05.(e & f) Dose response curves of HCC1806 (e) and SUM149 (f) cells treated with pharmaceutical-grade rigosertib in RPMI vs HPLM.(g & h) Dose response curves of a panel of renal cancer cell lines treated with pharmaceutical-grade rigosertib in RPMI (g) vs RPMI + UA (h).(i) Western blot of soluble and pellet α-tubulin from 786-O cells treated with increasing doses (5 nM, 50 nM, 100 nM, 500 nM, 1000 nM) of pharmaceutical-grade rigosertib for 4 hr in RPMI and RPMI + UA. (j) Quantification of western blots from (i).Data is represented as means ± SD of three independent experiments.* indicates p < 0.05 from two-way ANOVA.

Fig 5 .
Fig 5. Uric acid inhibits rigosertib activity by reducing the affinity of rigosertib for b-tubulin (a) Structural comparisons of colchicine-bound and rigosertib-bound tubulin.Colchicine and rigosertib are colored orange and cyan, respectively.The salt bridge between bE328 and aR221 found in colchicine structure is absent in the rigosertib structure, allowing H10 (green) to move away from the dimer body and create a pocket for uric acid (yellow) to bind.(b) Distance between bE328 and aR221 in the colchicine and rigosertib simulations.When this ionic bond is not formed, H10 becomes untethered which creates the binding pocket for uric acid.(c) Molecular details of uric acid binding in the pocket between H10 (green) and S9 (magenta).Residues that form hydrogen-bonds with uric acid are labeled.(d) CETSA analysis of K562 cells treated for 4 hr with 40 µM pharmaceuticalgrade rigosertib in RPMI at the indicated temperature.(e) Quantification of b-tubulin melting at increasing temperature in the absence of uric acid and rigosertib.N = 5 independent experiments.(f) Quantification of b-tubulin at 60C in the presence and absence of rigosertib and uric acid.Data is represented as means ± SD of five independent experiments.** indicates p < 0.01 from unpaired twotailed t-test.(g) Unlike mice and other model organisms/systems, humans do not express uricase resulting in uniquely high uric acid levels.