Prodrugs for colon-restricted delivery: Design, synthesis, and in vivo evaluation of colony stimulating factor 1 receptor (CSF1R) inhibitors

The ability to restrict low molecular weight compounds to the gastrointestinal (GI) tract may enable an enhanced therapeutic index for molecular targets known to be associated with systemic toxicity. Using a triazolopyrazine CSF1R inhibitor scaffold, a broad range of prodrugs were synthesized and evaluated for enhanced delivery to the colon in mice. Subsequently, the preferred cyclodextrin prodrug moiety was appended to a number of CSF1R inhibitory active parent molecules, enabling GI-restricted delivery. Evaluation of a cyclodextrin prodrug in a dextran sodium sulfate (DSS)-induced mouse colitis model resulted in enhanced GI tissue levels of active parent. At a dose where no significant depletion of systemic monocytes were detected, the degree of pharmacodynamic effect–measured as reduction in macrophages in the colon–was inferior to that observed with a systemically available positive control. This suggests that a suitable therapeutic index cannot be achieved with CSF1R inhibition by using GI-restricted delivery in mice. However, these efforts provide a comprehensive frame-work in which to pursue additional gut-restricted delivery strategies for future GI targets.


Introduction
Colony-stimulating factor 1 receptor (CSF1R), a tyrosine kinase that is highly expressed in myeloid cells, including monocytes, macrophages and osteoclasts, regulates their differentiation, proliferation, and survival. CSF1R serves as the receptor for two ligands: colony stimulating factor 1 (CSF1) and interleukin . Inhibition of CSF1R to deplete macrophages, either via small molecule intervention or by using antibodies to the receptor or ligands, has been targeted as a potential therapeutic in the field of oncology. [ (CK) have been observed pre-clinically and clinically with anti-CSF1R and anti-CSF antibodies [20][21][22] and small molecule CSF1R inhibitors. [23,24] Typically, monitoring of these enzymes is used in clinical practice to indicate liver damage, and this on-target effect of increased serum enzymes would make clinical development challenging for chronically ill IBD patients who may taking other concurrent medication. [25] Other on-target effects are predicted from CSF1R deficient mice, and include reduced fertility, abnormal bone metabolism, and impact on brain development. [26][27][28] In humans, various mutations in the CSF1R protein kinase domain give rise to a predisposition to hereditary diffuse leukoencephalopathy with neuroaxonal spheroids. [29] In the context of inflammatory GI disease, systemic CSF1R inhibition would be expected to deplete circulating monocytes and colonic macrophages, both involved in colitis disease process To address the aforementioned on-mechanism safety concerns, we hypothesized that minimizing systemic exposure to a CSF1R inhibitor would improve safety margins by avoiding exposure to the liver, bone, reproductive tract, and brain. Colon-localized delivery of a CSF1R inhibitor, defined as maintaining exposures over the CSF1R cellular EC 50 in the colon and remaining considerably below in the blood, should not deplete circulating monocytes but may prevent monocytes that traffic into the colon from differentiating into macrophages, and decrease survival and activation of existing tissue macrophages. Our goal was to explore whether sufficient efficacy would be maintained with colon-localized CSF1R inhibition by comparing the effect of local delivery of small molecule CSF1R inhibitors to that of a known systemic CSF1R inhibitor in a preclinical mouse model of colitis.
Two broad strategies for GI-restriction were evaluated: (1) limit oral absorption in the upper GI and (2) maximize hepatic and/or non-hepatic metabolism to inactive metabolites. In terms of limiting oral absorption, two tactics were considered: design of compounds with non-traditional drug-like properties-high molecular weight, hydrophilic compounds, [30] and utilization of prodrugs that are metabolized to release active drug in the colon. For the high metabolism approach, strategies exist to maximize metabolism to inactive metabolites including engagement of intestinal or blood esterases and amidases. Consideration was given to identifying compounds that are substrates of cytochrome P450 3A4 (CYP3A4). This drug metabolizing enzyme is highly expressed along the GI tract in human and rodents. [31] Thus, compounds known to undergo metabolism by CYP3A4 could be metabolized to inactive metabolites upon absorption into intestinal epithelial cells, resulting in minimized systemic exposure of active parent compound. However, a key concern around this approach was the potential for saturation of CYP3A4, as well as potential drug-drug interactions associated with co-administered CYP3A4 substrates or inhibitors, and therefore this approach was deprioritized.
The high expression of P-glycoprotein (P-gp or Multi-drug Resistance 1 (MDR1) protein) efflux transporter along the GI tract [32,33] suggested an additional potential tactic to restrict systemic exposure. Following absorption across mucosal tissue, compounds that are P-gp substrates could be actively transported back into the lumen of the small intestine thus reducing systemic exposure. Although some compounds we studied were P-gp substrates, the effect on absorption was not definitively assessed. In addition, P-gp is also expressed in distal colon which could limit exposure to this target organ. As a result, P-gp inhibition was not selected as a strategy of focus.
Restricted by the limitations of mouse physiology that preclude both the use of capsulebased formulation (due to size) and pH-modulated release (due to lack of an appreciable pH gradient in the mouse GI tract [34]), the prodrug approach to limit oral absorption in the upper GI tract was selected. Several reviews have highlighted the assortment of prodrug options that may be amenable to lower GI delivery, including dimethylglycine, valine, glucoside, glucuronide, cyclodextrin, and dextran adducts. [35][36][37] Typically, large, polar molecules are poorly absorbed in the small intestine and arrive intact in the colon where active parent may then be liberated by various cleavage mechanisms. In spite of this breadth of historical prodrug work, a comprehensive comparison for different prodrugs conjugated to the same active parent molecule has not been studied. An assessment of the relative effectiveness of each prodrug to selectively deliver active compound to the GI tissue is therefore lacking. An encouraging precedent for colon-specific delivery of a cyclodextrin prodrug is seen with an orally dosed prednisolone-cyclodextrin conjugate that demonstrated efficacy in a rat colitis model. [38] The prodrug gave comparable effects to prednisolone on colonic disease score, and demonstrated a significantly improved thymus:body weight ratio, a measure of side effects induced by systemic exposure.

Results and discussion
To enable the preparation of prodrug derivatives, we selected four parent compounds (1, 4, 5, 6; CSF1R enzyme IC 50 < 0.010 μM). These CSF1R inhibitors afford a classical two-point hinge binding arrangement with the NH and C = O of Cys666 as a consequence of the exocyclic NH and the adjacent triazole N. In addition, a water-mediated interaction from the remaining non-bridging triazole N to gatekeeper Thr663 is proposed (Fig 3). The pyrazole resides in a water-accessible region that tolerates a variety of substituents, and the cyclohexyl group provides an optimal fit under the Gly-rich loop. Each compound exhibited good CSF1R cellular potency (EC 50 = 0.104-0.245 μM) in a CSF-induced mouse macrophage differentiation assay, and generally demonstrated at least 10x selectivity when CSF1R enzyme activity was assessed against our internal kinome screening panel (Fig 4; see Methods for kinases tested). The four compounds exhibited high PAMPA permeability, P-gp efflux ratios <2 (suggesting they are not P-gp substrates), and low to moderate absorption (FaFg). This allowed for an assessment of FaFg on penetration into colon tissue and access to the lamina propria where the majority of inflammatory monocytes and macrophages accumulate in diseased GI tissue. The compounds also exhibited low to high systemic clearance.
In order to evaluate colon exposure, the compound concentration in mouse distal colon homogenate was compared to plasma levels, and a maximal ratio of unbound colon:plasma exposure was favored. Although we were most interested in the compound concentration in the lamina propria region of the distal colon, current limitations in tissue isolation methods with mouse colons and methods to quantify in this specific region led to reliance on whole colon homogenate concentrations as a surrogate for concentrations in the lamina propria. To reduce the possibility of fecal contamination impacting measured levels of drug in colon tissue homogenate samples, the isolated colons were flushed thoroughly with PBS prior to snap freezing. The necessity of using whole colon tissue concentrations as a surrogate has been described in the literature. A recent report studying colon tissue concentrations of raltegravir in rats, eloquently described the challenges faced with total tissue measurements, where complete colon tissue drug levels were used as a substitute measurement of gut-associated lymphoid tissue (GALT) concentrations. [40] Moreover, exposure measurements from total tissue homogenate have been used in the antiretroviral field to compare rectal tissue exposure with blood plasma concentrations in human clinical trials. [41][42][43][44][45] To simulate the release of active parent molecule in the colon after oral dosing of a prodrug, active parent 1 was initially dosed per rectal (PR) and compared to oral dosing (PR versus PO, 30 mg/kg dose). Compound exposure was measured in colon and liver tissue homogenate, and in plasma (Fig 5) at the estimated colon Cmax (1 h for PR dosing, 3 h for PO dosing) and Clast (11 h). The unbound fraction (fu) was measured in each tissue compartment as this value can vary substantially due to lipid and protein makeup. The fu was used to calculate unbound tissue concentrations for all comparisons. Mean colon unbound Cmax was significantly higher with PR than with PO dosing (67 ng/g and 9 ng/g, respectively) and the 1 h colon exposure exceeded the cellular EC 50 while the 1 h plasma exposure did not. The colon:plasma and colon:liver ratios at 1 or 3 h were dramatically increased with PR dosing (colon:plasma -22x versus 0.5x; colon:liver -7x versus 0.34x). Lack of compound detection in the colon at the terminal time point after PR dosing was attributed to the challenge of maintaining a rectally-dosed solution within the colon in active mice for sufficient time to enable the extended absorption that would be anticipated with oral dosing. The ability to achieve high colon and low systemic exposure with PR dosing confirmed that absorption from the colon is indeed less than from the small intestine [30] and built confidence that prodrug delivery of compound to the colon would give compelling colon levels of drug and low levels in circulation.
Multiple prodrugs can be prepared from a single active compound bearing an appropriate synthetic attachment point. Given the favorable preliminary exposure data with PR dosing, secondary alcohol 1 was selected as the prototype parent molecule from which a broad range of prodrug conjugates (2,(7)(8)(9)(10)(11)(12)(13)(14) were prepared (Fig 6). Each prodrug was evaluated for colon and plasma exposure of active parent 1 at 3 and 11 h following oral administration of a dose adjusted to equal a 30 mg/kg dose of 1. The simple dimethylglycine (7) and valine ester (8) analogs afforded no advantage over dosing the active parent directly, resulting in significantly higher plasma versus colon drug levels. These "drug-like" derivatives (molecular weight (MW) 466-480; total polar surface area (tPSA) 102-125) are likely absorbed in the small intestine prior to intestinal or plasma esterase-mediated release of active parent. Prodrug 7 is not stable in plasma and no prodrug levels were measurable in plasma after oral dosing. Prodrug 8 was detected, at~10x lower plasma concentrations than active parent 1. Oral doses of glucoside and glucuronide analogs 9-11 resulted in approximately equivalent amounts of active parent 1 in both colon and plasma. Plasma levels of prodrug 9 were~100x less than active parent, while no measureable plasma levels of 10 and 11 were observed. The moderate increase in molecular weight (543-557) and polarity (PSA 172-189) may account for reduced absorption of 9-11 in the small intestine. For 8-11, relatively low levels of prodrug in feces were measured (3-30% of active parent in feces) suggesting efficient conversion of the prodrug to parent drug. Superior colon:plasma ratios were obtained upon oral administration of succinate-linked cyclodextrin prodrugs 2, 12, and 13 (as inseparable mixtures of the 6-O and 2-O regioisomers) in which unbound colonic levels of 1 at 3 h were at or near the CSF1R cellular EC 50 and colon: plasma ratios of up to 260-fold and liver:colon ratios of 17-30x were achieved with the α-and β-variants. The large, highly polar nature of these compounds (MW 1436-1760; tPSA 580-738) should preclude absorption of prodrug in the small intestine whereas the low levels of active parent 1 that are detectable in the plasma are ostensibly due to colonic absorption of 1 after microflora-mediated cyclodextrin cleavage. High levels of uncleaved cyclodextrin 2 were detected in the feces (~2-fold higher than levels of active parent in feces). The dextran prodrug (14), also attached via a succinate linker, gave low plasma levels of parent drug as expected due to both large size (MW~70,000) and high polarity (tPSA >> 800), however colonic levels were also low and well below the CSF1R cellular EC 50 . The low colon levels may not be attributed to reduced conversion to active parent as~20-300x higher levels of active parent 1 in the feces than in the colon tissue homogenate were observed from 3-11 h. Overall, the cyclodextrin prodrugs were preferred for selective delivery of 1 to the colon, and as such were investigated with further analogs.
The mechanism of cleavage of the cyclodextrin prodrugs is hypothesized to occur via a twostep process (Fig 7). First, bacterial esterases cleave the ester linkage to the cyclodextrin, releasing the hemi-succinate 15. Although able to inhibit CSF1R enzyme activity, 15 is not stable in mouse plasma in vitro and a PK experiment dosing the hemi-succinate 15 IV or PO (1 mg/kg) confirmed the degradation and release of 1 with no detectable plasma levels of 15 at any time point evaluated (from 0.5 to 12 h). Thus it is unlikely that 15 would contribute meaningfully to any in vivo efficacy. It is conceivable that the order of degradation steps could be reversedester hydrolysis to give 1 directly and subsequent cleavage of succinic acid from the cyclodextrin. Alternatively, hydrolysis of the sugar moiety into smaller saccharide conjugates may occur prior to linker cleavage as was reported from a study on n-butyric acid directly linked to β-cyclodextrin. [46] Cyclodextrin analogs of the primary alcohol 4, tertiary alcohol 5, and carboxylic acid 6 ( Fig  8) were prepared and profiled to compare the utility of these alternative prodrug attachment prodrug derivatives of the tertiary alcohol were not pursued. For prodrugs of the carboxylic acid parent 6, the γ-cyclodextrin 23 was favored, providing colonic exposure well above the cellular EC 50 while maintaining a preferred colon:plasma ratio (~30-70 from 3-11 h). The αand β-cyclodextrin comparators (21 and 22, respectively) gave substantially lower overall exposures in the colon. Plasma concentrations of 6 were generally low despite its low systemic clearance and plasma concentrations tended to increase with colon exposure.
From the first set of prodrugs prepared with secondary alcohol 1 (2, 12, 13), α-cyclodextrin 2 was chosen for in vivo profiling based on acceptable total colon concentration approaching the total cellular EC 50 , sufficient window of colon:plasma exposure, and relative ease of synthesis to access multi-gram quantities as required to enable in vivo studies. A mouse dextran sodium sulfate (DSS)-induced model of colitis was selected due to significant macrophage component and up-regulation of the receptor (CSF1R) and its ligands (CSF1 and IL34) (internal data, not shown). Prodrug 2 was dosed orally twice a day from day 0 to day 14 (30, 100, and 300 mg/kg dose equivalent, Fig 10). The reasoning for providing pro-drug 2 one week prior to initiation of DSS administration was to enable CSF1R-mediated depletion of gut macrophages prior to providing DSS and ensuing inflammatory insult, as the exact contribution resident macrophages have to DSS-mediated disease is unknown. The systemic CSF1R inhibitor 3 was used as a positive control, administered at 100 mg/kg (QD) from day 7-14, as it had historically shown robust efficacy in a mouse DSS model without the need for 1-week pre-treatment dosing regimen. Disease effect on the colon was measured in two ways: a pharmacodynamic readout of the change in macrophage numbers (% of the mucosa that was IBA1+ (ionized calcium-binding adapter molecule 1) [47]) in the colon by image analysis, and a functional readout measuring erosion length in the colon by histology. Systemic effects on liver macrophages were measured by image analysis (% IBA1+ cells) of liver tissue, and on circulating monocytes by fluorescence-activated cell sorting (FACS) to determine depletion of a non-classical monocyte subtype (CD11b+ CD11c+ MHCII-) as well as a classical (CD11b+ CD11c-MHCII-) population. In addition to terminal exposures and PD/efficacy readouts, exposure of 1 in colon, liver, and blood was measured in a separate study after a single dose and after 7 days of BID dosing to evaluate whether drug accumulation occurred prior to DSS administration. Compared to exposures following a single dose, after 7 d of BID dosing of 2 in healthy animals no significant accumulation of 1 in the colon or liver was observed, while a modest accumulation was noted in the blood (~3x) in the 300 mg/kg dose group (Fig 11). The data suggests that maximal absorption of active parent is generally achieved at 30 mg/kg in the presence of intact GI epithelium and no benefit is gained with increased dose. Interestingly, in mice subsequently exposed to DSS for 7 days with BID dosing of prodrug 2, dramatic increases in colon and liver exposures of parent 1 were observed, presumably resulting from loss of epithelial integrity and increased permeability of the GI barrier due to DSS-mediated chemically induced damage. Although blood exposures were higher with DSS treatment than in naïve animals, at 30 mg/kg the mean unbound colon exposure was at cellular EC 50 , while unbound blood exposure was 10x lower than EC 50 (corresponding to a colon:blood ratio of 11 at this dose) and the 100 mg/kg treated group maintained coverage of cellular EC 50 in the colon while remaining below in the blood. These results show colon-restricted exposure and therefore allow for the determination of whether GI-restricted exposure of CSF1R inhibitor could enable local depletion of colonic macrophages in the DSS study, while sparing systemic PD effects such as liver macrophage and blood monocyte depletion.
At the lowest dose of prodrug tested (2, 30 mg/kg), a trend in reduction of IBA1+ cells was observed in the colon and the liver; however, this was not statistically significant (Fig 12). At higher doses, with concomitantly increasing systemic exposure, more substantial reductions in colon macrophages were observed, as were significant depletion of IBA1+ cells in the liver. The high dose (300 mg/kg) depleted colonic macrophages to a similar extent as systemic CSF1R inhibitor 3 (66% versus 76%, respectively). The impact of prodrug on colonic erosions was less significant, and a statistical decrease in erosion length with comparable effect to the positive control 3 was observed only at the 300 mg/kg dose (Fig 13). No significant effect on circulating CD11b+ CD11c-MHCII-monocytes nor on the more sensitive CD11b+ CD11c+ MHCII-monocytes was detected at the 30 or 100 mg/kg doses (Fig 14), which corresponded to the blood exposures below cellular EC 50 at these doses. At the top dose (300 mg/kg), where the largest decrease in colonic macrophages was observed, there was nearly complete depletion of non-classical monocyte population to a level comparable to the effect with systemic inhibitor 3 (97% and 94%, respectively). Prodrugs for colon-restricted delivery: Design, synthesis and in vivo evaluation of CSF1R inhibitors Overall, although colon-restricted drug exposure at the 30mpk dose was demonstrated, there was a lack of compelling efficacy on erosion length at this dose. Higher efficacy and more significant pharmacodynamic effects of IBA1 reduction in the colon were seen only at 100 and 300 mg/kg doses where significant IBA1 reduction was also observed in liver, signifying lack of a gut-restricted pharmacodynamic effect. If higher exposure in the colon could be achieved, while maintaining similar (or improved) colon:plasma or colon:liver exposures, it may be feasible to achieve a more substantial local effect, but the apparent sensitivity of circulating monocytes and liver IBA1+ cells to CSF1R inhibition made this approach significantly less attractive. Nonetheless, the ability to practically achieve high colonic and low systemic exposure by utilizing cyclodextrin prodrugs could be a promising approach for other biological targets that would benefit from colon-restricted delivery.

Chemistry
The synthesis of primary alcohol CSF1R inhibitor 4 is shown in Fig 15. Displacement of bromide 24 with 4-aminopyrazole occurred with excellent regioselectivity at C8. We explored direct introduction of the cyclohexenyl substituent at C6 via Negishi coupling, however the yields were generally moderate and scalability was poor. Better overall yields were typically obtained by stepwise introduction of the cyclohexenyl substituent via Suzuki coupling and hydrogenation. Using this approach, cyclohexene 25 was prepared in moderate yield from the 6-bromo-8-aminopyrazolo-triazolopyrazine intermediate. The subsequent alkylation of the pyrazole with ethyl bromoacetate only partially went to completion but the product was readily separated from starting material 25 and underwent reduction of the alkene and ester functionalities to give primary alcohol 4.
An optimized route for the synthesis of secondary alcohol 1 is shown in Fig 16. We knew from earlier studies that separating cis-and trans-1 alcohols at a late stage would be difficult. We therefore focused on a route that would allow for the relative stereochemistry to be set at an early stage, targeting nitropyrazole 30 as a key intermediate. Following literature precedent with the corresponding iodopyrazole, [48] 4-nitropyrazole smoothly displaced sulfonate 27 to give acetal 28 which was deprotected using acidic conditions to provide ketone 29. Reduction with sodium borohydride gave a 4:1 mix of trans:cis alcohols, which could be recrystallized from toluene to furnish trans-alcohol 30 with greater than 99:1 trans:cis selectivity on greater than 40 g scale. Reduction of the nitro group cleanly yielded primary amine 31 which underwent regioselective S N Ar displacement at the C8 position of dibromide 24 to give bromide 32. Final introduction of the cyclohexyl group was achieved by Suzuki coupling and hydrogenation of the resultant alkene to furnish secondary alcohol 1 in excellent yield.
Tertiary alcohol 5 was prepared from pyrazole intermediate 25 over three steps. Ketone intermediate 33 was accessed via a straight-forward Michael addition of pyrazole 25 with but-3-en-2-one. Addition of methylmagnesium bromide to the ketone moiety generated the tertiary alcohol in moderate yield, along with recovered starting material. Subsequent hydrogenation of the cyclohexene ring provided the desired tertiary alcohol 5 (Fig 17).
Carboxylic acid 6 was prepared without incident via a closely related sequence (Fig 18). Conversion of commercially available cis-alcohol 34 to the corresponding tosylate followed by displacement with 4-nitropyrazole gave pyrazole 35. Pd-catalyzed reduction of the nitro substituent and addition of the resulting aminopyrazole to C8 of dibromide 24 gave triazolopyrazine 36 as the only regioisomer. Suzuki coupling and hydrogenation introduced the cyclohexyl substituent, and a final saponification provided the desired acid 6.
Ester prodrugs 7 and 8 were prepared via straightforward coupling with EDC (Fig 19), followed by Boc-deprotection in the case of ester 8. Preparation of glycosides 9 and 10 was initially attempted unsuccessfully via displacement of a tetrapivaloylglucopyranosyl bromide without success. In an alternative approach, Lewis acid mediated displacement of a tetrabenzyl trichloroacetimidate displacement proceeded to generate the protected glycosides excellent yield. Pd-catalyzed deprotection afforded anomeric products 9 and 10 which were readily separated by HPLC. Glucuronide 11 was prepared in low yield via displacement of a triacetyl 2-bromotetrahydropyran followed by base-catalyzed acetate deprotection. The poor yield was attributed to a competing acetyl transfer from the reagent to alcohol 2.
Cyclodextrin prodrugs of carboxylic acid 6 were obtained via saponification of ester 37, isolation of the sodium salt 38 and displacement with commercially available 6-O-sulfonylcyclodextrins (Fig 20). The reactions each proceeded to approximately 50% conversion by LCMS, at which point the crude reaction mixtures were directly subjected to reverse-phase HPLC. The desired prodrugs 21-23 were isolated as single regioisomers in modest yields. Prodrugs for colon-restricted delivery: Design, synthesis and in vivo evaluation of CSF1R inhibitors Cyclodextrin conjugation of the primary alcohol was performed using a modification of the literature procedure for prednisolone. [49] Formation of hemisuccinate ester 15 via reaction with succinic anhydride (Fig 21) proceeded in excellent yield. In contrast to literature reports, CDI coupling with cyclodextrins were ineffective. Instead, EDC coupling with an excess of α, β or γ-cyclodextrin was used to provide the conjugates 2, 12 and 13. Precipitation with acetone gave solids which could be purified via reverse-phase HPLC purification, albeit as inseparable mixtures of 6-O and 2-O cyclodextrins. The poor yields were again attributable to incomplete conversion. The ratio of products was quantitated by 1 H-NMR, with the triazole proton being the most clearly resolved peak despite being remote to the point of attachment. The regiochemistry was also determined by NMR, with a significant downfield shift observed for the cyclodextrin C-H protons at each point of attachment. The cyclodextrin prodrugs of primary alcohol 4 and tertiary alcohol 5 were prepared similarly (Fig 22 and Fig 23). The dextran conjugate of alcohol 1 was prepared using a protocol developed for dexamethasone. [50] Hemisuccinate 15 was treated with CDI followed by dextran (MW~70,000) from Leuconostoc spp ( Fig  21). After trituration of the resultant solid which removed any unconjugated starting material, the extent of substitution was determined via hydrolysis with aqueous NaOH and HPLC quantitation of the released parent. The solid was determined to contain 11.5 wt% 1.

Conclusions
Colon-restricted delivery has the potential to be a powerful tool for delivering efficacy for GI targets while reducing on-target or off-target side effects resulting from systemic exposure achieved by traditional oral delivery methods. Prodrugs could be of particular use as tools to generate preclinical in vivo proof-of-concept in mouse models of GI disease. We described the synthesis of a series of prodrugs appended to the same active parent molecule, and for each evaluated colon versus systemic exposure of active parent upon oral dosing of prodrug. Of a broad range of prodrugs evaluated, the cyclodextrin-conjugated analogs proved optimal for GI-restricted delivery when coupled to an active parent molecule with low-moderate absorption (FaFg). Carboxylic acids, primary alcohols, and secondary alcohols all serve as viable synthetic attachments for cyclodextrin appendage points, enabling the demonstration of restricted delivery to the colon, with up to >100x unbound colon:plasma exposure ratios. Optimal synthetic routes to synthesize cyclodextrin prodrugs were developed and the Prodrugs for colon-restricted delivery: Design, synthesis and in vivo evaluation of CSF1R inhibitors secondary alcohol α-cyclodextrin derivative 2 was prepared on multi-gram scale to enable in vivo proof-of-concept experiments. In a mouse colitis model, target colon exposures with minimal systemic exposure were successfully achieved with 2 when dosed at 30 mg/kg (active parent-dose equivalent) BID, however only a moderate pharmacodynamic effect, measured as a reduction of IBA1+ macrophages in the colon, was observed. Optimal depletion of colon macrophages was achieved only at higher doses, and these doses resulted in attendant systemic exposure, depletion of circulating monocytes, and reduction of liver macrophages.
Although an increased colon:systemic exposure margin was achieved, the DSS colitis model results suggest a gut-restricted CSF1R inhibitor is only partially effective. This is likely tied to the biology of monocyte recruitment from the blood to the gut both in healthy and inflamed settings, suggesting that systemic CSF1R inhibition is required to elicit full colon macrophage depletion and colitis disease reduction, and perhaps CSF1R was not an ideal target to test proof-of-concept for gut-delivery strategies.
Although CSF1R protein expression is mainly discussed in literature as linked to myeloid lineage cells, there are reports of receptor expression in gut epithelial cells. [51,52] Upon investigation of CSF1R expression patterns by immunohistochemistry (IHC) in mouse and human colon tissue, our group was unable to corroborate CSF1R epithelial staining patterns, observing only myeloid/macrophage cell staining with little to no epithelial cell staining. To note, the differences in IHC expression observed by our group and others could be attributed to different anti-CSF1R monoclonal antibody clones used for staining. A more comprehensive Prodrugs for colon-restricted delivery: Design, synthesis and in vivo evaluation of CSF1R inhibitors evaluation assessing CSF1R expression in multiple tissues and species would be needed to better understand the role of CSF1R inhibition in modulating gut epithelium homeostatic mechanisms. Prodrugs for colon-restricted delivery: Design, synthesis and in vivo evaluation of CSF1R inhibitors On the whole, the pro-drug strategies and learnings employed herein may be better-suited for alternative targets of GI-disease, and these efforts have at least provided a comprehensive frame-work in which to pursue additional gut-restricted delivery strategies for future GI targets.
Colon tissue preparation for determination of drug exposure. Due to the inherent risk of fecal or PR-dosed material contaminating true levels of colon tissue exposure, careful flushing of colon tissue for drug analysis was required. Upon necropsy, a section of mouse colon was removed, and flushed with PBS until no visible fecal material was observed. Subsequently, a 1-2 cm section of colon was blotted dry, weighed and snap-frozen in liquid nitrogen.
Homogenization of colon tissue. Colon tissue (~50 mg) was diluted 10x in Veterinary Sterile Water for Irrigation (USP by Abbott), then 1.4 mm zirconium oxide beads were added and tissue was homogenized using an Omni Bead Ruptor 24 (Omni International, Kennesaw, GA). Aliquots of colon homogenate samples and a set of colon standards were pipetted in 96-well plates and subjected to protein precipitation extraction (with acetonitrile + internal standard) on a Microlab Star robot (Hamilton Robotics, Reno, NV). Plates were then mixed on a MixMate (Eppendorf, Hamburg, Germany) for 1 min at 1600 rpm and centrifuging for 5 min at 4000 rpm in 5810 R Centrifuge (Eppendorf, Hamburg, Germany). Supernatant was transferred into new plates and diluted with LC-MS/MS mobile phase.
Quantification of drug. Spiked standards were prepared in the appropriate matrices. All unknown samples and spiked standards were combined with acetonitrile containing internal standard, vortexed, and centrifuged to pellet the proteins. Supernatants were used for subsequent analysis by LC-MS/MS. The mobile phase consisted of acetonitrile with 0.1% formic acid, and water with 0.1% formic acid at a flow rate of 0.8 mL/min. Analytes were separated using reverse phase chromatography with a fast gradient on a 30 x 2.1 mm Fortis C18 Pace 5 mm column, prior to analysis on a Sciex API5500™ mass spectrometer. Peak areas were determined using Sciex Analyst™ 1.6 software. Actual concentrations were calculated by regression analysis of the peak area ratio (parent / internal standard) of the spiked standards versus concentration.
Chronic dosing study in healthy mice. On day 0, healthy C57BL/6 female mice (Taconic) were dosed BID with compound 2 and continued dosing until Day 7. Compound 2 was administered in a formulation of 10:90 (w/w) Cremophor EL: 1% sodium carboxymethyl cellulose. On day 1,~12 h after a single dose, a cohort of healthy mice was sacrificed for blood, liver, and colon exposure. On Day 7, 12 h post last dose, the remaining cohort was sacrificed for blood, liver, and colon exposure, representing 7 d BID dosing.
Mouse DSS model of colitis. On day 0, super prophylactic dosing of compound 2 began on C57BL/6 female mice. On day 7 DSS was administered via 3% DSS (Dextran Sulfate Sodium, MP Biomedicals cat# 160110) in drinking water. At this time compound 3 dosing initiated. On day 14, animals were sacrificed for PK and efficacy measurements (~11 h post-dose for compound 2, 3 h post-dose for compound 3).

Statistical analysis
Statistical analyses of in-vivo pharmacodynamic and efficacy data were performed as follows. D'Agostino & Pearson normality test was first used to test data for Gaussian distribution. If all groups passed normality test, a One-way ANOVA was used with Dunnett's multiple comparisons test to compare treatment groups to Vehicle group. If not all groups passed normality test, a Kruskal-Wallis nonparametric ANOVA was used, with Dunn's multiple comparison test to compare treatment groups to Vehicle group. Levels of significance compared to Vehicle are annotated as follows: Ã = p<0.05, ÃÃ = p<0.01, ÃÃÃ = p<0.001, ÃÃÃÃ = p < 0.0001. Optimal number of mice per group was determined using power analysis; N = 10 mice per group at 80% power predicts statistically significant detection of !~40% reduction in colon IBA1+ and !~75% reduction in colon erosion length comparing treatment groups to Vehicle group. Naïve groups were omitted from statistical analyses.

Ethics statement
Animal studies were conducted under a program accredited by the Association for Assessment and Accreditation of Laboratory Animal Care. All animal studies were reviewed and approved by AbbVie Bioresearch Center's Institutional Animal Care and Use Committee.