Examination of the Impact of Triazole Position within Linkers on Solubility and Lipophilicity of a CDK9 Degrader Series

Abstract

Optimization of degrader properties is often a challenge due to their beyond-rule-of-5 nature. Given the paucity of known E3 ligases and the often-limited choice of ligands with varied chemical structures for a given protein target, degrader linkers represent the best position within the chimeric molecules to modify their overall physicochemical properties. In this work, a series of AT7519-based CDK9 degraders was assembled using click chemistry, facilitating the tuning of aqueous solubility and lipophilicity while retaining their linker type and molecular weight. Using chromatographic logD and kinetic solubility experiments, we show that degraders with similar chemical constitution but varied position of the embedded triazole demonstrate different lipophilicity and aqueous solubility properties. Overall, this work highlights the impact of triazole placement on linker composition through application of click chemistry for degrader synthesis and its ability to be used to promote the achievement of favorable physicochemical properties.

The importance of CDK9 in multiple cancers, including Acute Myeloid Leukemia (AML), has been attributed primarily to its role in regulating global gene transcription, particularly the proto-oncogene Myc and the antiapoptotic gene MCL-1.¹⁻³ Targeted inhibition of Cyclin-Dependent Kinase-9 (CDK9) activity has been explored as a therapy for the treatment of cancer.⁴ Unfortunately, no CDK9 inhibitor has yet been approved for marketing in any indication, a result borne of the toxicity and side effects that are observed with existing small molecule inhibitors.⁵ For this reason, an alternative approach to these small molecules has been explored, namely, Proteolysis Targeting Chimeras (PROTACs).⁶⁻¹¹ Multiple reports have noted improvements in potency and/or selectivity of traditional inhibitors when they are transformed into protein degraders. In particular, it has been shown that the ability of a single degrader molecule to degrade multiple units of a target protein significantly augments its potency relative to its parent target-binding ligand.^12,13 Therefore, degrader technology may offer an advantage over traditional inhibitors for AML, i.e., the potential for more potent and prolonged inhibition of CDK9 activity that can be afforded by the catalytic mode of action of degraders.^13,14

While the selection and incorporation of appropriate E3 ligase and protein of interest (POI) ligands is critical for the successful development of degraders, degrader linkers have also been shown to significantly affect degrader activity.¹⁵ Linkers are known to affect ternary complex formation, determining cooperativity by engaging in specific interactions in the ternary complex.^16,17 In addition, linker length and composition contribute to the overall physicochemical properties of the molecules, including their permeability and aqueous solubility.^18,19 Degraders are considered Beyond-Rule-of-5 (bRo5) compounds due to their tendency toward high lipophilicity and molecular weight, among other chemical properties. These bRo5 characteristics often result in poor permeability, poor aqueous solubility and high plasma protein binding, translating to compounds which would be expected to have low bioavailability.²⁰ Notably, however, more degraders are being reported to have achieved good oral bioavailability. Most of these reports highlight the importance of exploring linker design strategies to optimize overall drug properties.^15,21−24 These advances demonstrate that it is possible to strategically employ linker design to account for both efficacy and good drug properties in the early stages of degrader development.

To date, linker design campaigns for degraders have largely depended on the syntheses and testing of a series of compounds with linkers of varied compositions and lengths.¹⁵ This trial-and-error approach to degrader design is one of the impediments to the progression of degraders into the clinic as it is often a time and resource consuming process. The necessity of such an approach is due in part to the relatively limited structural, biological, and pharmacokinetic data of degraders that have progressed into the clinic, limiting the amount of Structure–Activity–Relationship (SAR) data that could be utilized by medicinal chemists in the early stages of drug design to make other more promising degraders.²⁵ In this work, the degrader design approach is streamlined by choosing a specific set of linker building blocks and leveraging click chemistry to strategically modify linker composition. This was accomplished within a short degrader series in which the position of the embedded triazole ring was varied relative to each protein-binding ligand. While click chemistry has been employed as a synthetic strategy for the generation of various degraders, the impact of the embedded triazole has been largely unexplored.^15,26,27 Examining the role of this rigid moiety is critical given the ability of linkers to engage in intermolecular interactions within ternary complexes that impact degradation potency.²⁸ In this study, variation of the rigid triazole position across the flexible alkyl linker resulted in differences in CDK9 degradation, overall in vitro activity, kinetic solubility (KS), and lipophilicity. Our data shows click chemistry as a potential approach for tuning degrader biological activity while managing chemical parameters such as topological polar surface area (TPSA) and molecular weight for the achievement of favorable physicochemical properties.

Recently, we reported the synthesis of two series of CDK9 degraders that contained alkyl and amide linker moieties.²⁹ These degraders, which were based on the pan-CDK9 inhibitor AT7519 (1),³⁰ demonstrated that compounds of this type can selectively degrade CDK9 and illustrated the direct impact that the linker has on the overall physicochemical properties and in vitro activity of degraders containing the cereblon-recruiting ligands 4-hydroxy thalidomide and pomalidomide. While synthesizing these initial degraders, an alternative strategy utilizing click chemistry to rapidly assemble the linker region was also considered. Due to the modular nature of this approach, the click reaction was expected to facilitate efficient synthesis of a library of degraders with structurally distinct linkers.³¹ For this reason, we set out to prepare a series of compounds with three different chain lengths. Within each of these chains, the position of the triazole would then be varied relative to the thalidomide and AT7519 moieties. The location of the triazole was hypothesized to impact the degradation of CDK9 by affecting differences in the ability of the CDK9 or E3 ligase ligands to engage their respective targets or the ability of the degrader itself to induce ternary complex formation. Given that bRo5 molecules such as degraders have been shown to form intramolecular hydrogen bonds (IMHBs) that allow them to conceal or expose polar and apolar parts of their structure, modifying the relative triazole position was also hypothesized to affect the ease of IMHB formation and, hence, impact observed lipophilicity and aqueous solubility.

To investigate this, simple alkyl linkers were employed to install the requisite azide and alkyne functionality on the 4-hydroxythalidomide and AT7519 ligands, respectively, effectively focusing the impact of the linker exclusively on the relative position of the triazole ring. As shown in Scheme 1, 4-hydroxythalidomide (2) was reacted with readily accessible tosyl azide precursors using mild basic conditions previously reported by Qiu and colleagues to provide azides 3d–6d with good regioselectivity.³² The AT7519-linked alkynes 7a–12a were synthesized through alkylation of the piperidine nitrogen of AT7519. This approach resulted in the formation of four different azide precursors and six different alkyne precursors for the click reaction. Although this small library could theoretically generate a total of 24 distinct products through combination of all of the reaction partners, only ten representative products were synthesized. These products were selected to provide three sets of compounds that would have roughly the same overall chain length but would vary in the triazole positioning relative to the two binding ligands. Therefore, each set, containing 8 or 9, 10, and 14 carbon atoms in the alkyl linkers, would have one compound with the triazole closer to the thalidomide, one with the triazole in the center, and one with the triazole closer to AT7519. An additional example was synthesized to more closely examine the direct impact of chain length in these compounds. In all cases, the click reactions between the alkynes and azides smoothly provided the desired products 13-22 in the presence of CuSO₄·5H₂O and sodium ascorbate.

Scheme 1. General Synthetic Scheme for Thalidomide-Based Alkyl Azides and AT7519-Linked Alkynes as well as Triazole-Containing Degraders.

The obtained triazole-containing degraders were tested in biological experiments to measure their potency in growth inhibition of AML cells and in their degradation of CDK9. Cell cytotoxicity (IC₅₀) and degradation potency (DC₅₀ and D_max) were determined in MV4-11 and MOLM-13 cells, respectively. These values are presented in Table 1, along with pIC₅₀ values [log(M IC₅₀)] to more accurately reflect the differences/lack thereof in in vitro cytotoxicity. Predicted logD values, calculated using the ChemAxon logarithm, are also provided for each degrader.³³ To characterize the physicochemical properties of the series, kinetic aqueous solubility and lipophilicity were assessed. Lipophilicity was measured using the chromatographic logD experiment reported by Caron and colleagues.³⁴ The values derived from these analyses for each degrader are also shown in Table 1.

Table 1. Physicochemical Properties and In Vitro Activity of Triazole-Containing Degraders.

ID	N:N*a	LogDb	logk80′c	Solubility (μM)	DC50 (μM)	Dmax (%)d	IC50 (μM)	pIC50
13	5:4	1.81	0.100	12.4 ± 7.4	0.13	67.5	0.24	6.6
14	4:4	1.29	0.084	40.9 ± 4.9	0.07	79.1	0.19	6.7
15	4:5	1.57	0.106	31.3 ± 2.1	0.40	58.6	0.19	6.7
16	8:2	3.21	0.146	5.65 ± 1.7	0.06	88.3	0.05	7.3
17	5:5	2.08	0.112	13.2 ± 4.9	0.95	68.5	0.13	6.9
18	4:6	1.93	0.116	12.9 ± 1.3	0.10	75.3	0.07	7.1
19	8:6	3.78	0.184	2.38 ± 0.4	0.07	86.2	0.07	7.1
20	7:7	3.78	0.176	4.70 ± 0.4	0.06	78.0	0.06	7.2
21	4:10	3.70	0.197	3.42 ± 1.9	0.04	88.6	0.01	8.1
22	7:5	2.97	0.140	5.46 ± 2.9	0.10	70.1	0.05	7.3
AT7519	--	–0.67	0.134	512.4 ± 24.3	--	--	0.12	6.9

^aNotation for number of linker atoms on either side of triazole (THAL:AT7519).

^bCalculated logD derived from octanol/water model.

^cLipophilicity descriptor derived from retention on PLRP-S stationary phase.

^dMaximum percentage of CDK9 degraded.

The cytotoxicity data showed that multiple degraders in the click series (16–22) were able to achieve low nanomolar potency against MV4-11 cells, outperforming AT7519 in the same cell line. Most importantly, certain degraders demonstrated downstream repression of the antiapoptotic gene MCL-1, as determined by Western blots obtained from drugged MV4-11 cells (Supporting Information (SI), Figure S3). To better evaluate the catalytic mechanism of action of this degrader series, a drug washout (WO) experiment was performed. This experiment involved treatment of cells with compounds for six hours after which the culture media was replaced with a drug-free counterpart. After a total incubation time of 48 h, cytotoxicity (Post-WO IC₅₀) was evaluated and compared to the values derived from cells treated under the same conditions but did not undergo a washout procedure (No-WO IC₅₀). The results are provided in the SI (Table S1) and show that degraders 20, 21, and 22 were able to retain their potency after the drug washout. The potency of each of these degraders after drug washout was at least 5-fold the observed activity of the parent warhead, AT7519, under the same experimental conditions. The data suggest that these degraders are more cell-permeable and deliver more potent activity pre- and postwashout relative to the traditional inhibitor.

Additionally, across the three sets of isomers, an increase in potency, hydrophobicity (according to aqueous solubility), and lipophilicity with increase in linker length is observed, as has been reported multiple times in degrader development studies.^10,35 Lipophilicity was assessed using the logk₈₀^′ experiment due to its efficiency as well as the demonstrated ability to capture the impact of IMHB formation on overall lipophilicity of ionizable compounds.³⁴ Importantly, across the three sets of degraders, differences in logk₈₀ stack are observed, most notably within the 16–18 set. These data indicate that relative positioning of the rigid triazole group results in variation in lipophilicity even when linker length is maintained, although this data does not necessarily confirm the presence of IMHB’s in the molecule. Overall, a curve fit analysis (Figure 1A) shows that within the click series, IC₅₀ and logD were moderately correlated (R² = 0.76), when 14 and 21 are excluded, given that both degraders show higher activity than expected relative to their logk₈₀^′s (SI, Figure S1A–B). The same curve fit analysis performed with logk₈₀ stack shows that chromatographic logD has a stronger relationship to cellular potency (R² = 0.95) than predicted logD. These data demonstrate that despite the correlation between the calculated and observed lipophilicities (Figure S1C), the use of bRo5-relevant methods to assess physicochemical properties of degraders can better facilitate understanding of degrader SAR relationships, particularly for linker design. Interestingly, as shown in Figure 1B, DC₅₀ is weakly correlated to both predicted logD and measured logk₈₀^′ stack. The lack of a strong relationship in either case suggests that degradation potency is more closely related to the unique mechanism of action of degraders, specifically, the catalytic mode of action. This is because the correlation between overall potency (IC₅₀) and logk₈₀ stack is likely based on permeability, which affects both traditional inhibition and degradation activity.^18,20,35,36 However, with degraders, the impact of permeability is attenuated due to the ability of a single degrader molecule to degrade multiple units of a POI. Therefore, while logk₈₀^′ stack may correlate to the number of degrader molecules that successfully cross the cell membrane, the value does not dictate the efficiency of each molecule in degrading the target protein. This may explain the weak correlation between lipophilicity and the measured DC₅₀s.

Figure 1.

Correlation coefficients (R²) and graphical representations of curve fit for all triazole-containing PROTACs (except compounds 14 and 21) between both LogD and logk₈₀^′ stack (10×) and (A) −Log(μM IC₅₀) (−1* Log10 (μM IC₅₀s in MV4-11). (B) −Log(μM DC₅₀) (−1* Log10 (μM DC₅₀s in MOLM-13).

Within the 16–18 and 19–21 sets of constitutional isomers, some significant changes in potency upon varying the position of the triazole are observed, particularly with regard to cytotoxicity. It is also important to note that the compounds within each set that possess the highest observed logk₈₀^′ stack demonstrated the highest cellular potency. This may be explained by differences in permeability that would ultimately translate to higher potency. However, the minimum difference in logk₈₀ stack between 21 and its constitutional isomers is 0.013, while the corresponding value for 16 is 0.030. Despite this, 21 outperforms its isomeric counterparts in cytotoxicity by almost 10-fold while 16 had a modest 1.5- to 2.7-fold improvement in activity over 17 and 18. These results suggest that the exceptional performance of 21 is likely due to factors beyond its lipophilicity or even its linker length. The data therefore support the proposed hypothesis that variation of the triazole position within a linker of defined composition can impact the potency of these AT7519-based degraders.

A notable correlation was found between KS and both predicted and observed lipophilicities across all degraders (SI, Figure S1D). However, the impact of the triazole is still observed when examining the KS within the 13–15 set. A notable 2-fold difference in aqueous solubility is observed between 13 and 15 from simply shifting the position of the rigid triazole one methylene unit closer to the thalidomide as opposed to AT7519. The presented data also show that at linker lengths beyond that of the 16–18 set, the aqueous solubility of the degraders fall below 5 μM. By contrast, 14 is able to demonstrate the highest aqueous solubility within the series, 40 μM, likely due to its low predicted logD and low measured logk₈₀^′ stack. However, it is important to note the significant reduction in KS of the degraders relative to the parent molecule, AT7519. Despite the presence of the tertiary amine within the degraders, the ionization of the nitrogen atom at physiological pH did not sufficiently increase polarity enough to counteract the increase in overall lipophilicity that occurs from transforming the AT7519 small molecule into a larger chimera. These data emphasize the challenges of degrader design, i.e., the difficulty of achieving drug-like properties when molecular weight is significantly increased.²⁵

The impact of the improved aqueous solubility of 14 over the other analogues within the click series is seen when tested in human serum (HS), as opposed to fetal bovine serum (FBS). Figure 2A shows a retention of potency in human serum, with 14 actually demonstrating near 3-fold improvement in cytotoxicity against MV4-11 cells. This result is peculiar because the most common observation when compounds are tested in HS compared to FBS is either a retention of potency or a loss thereof - rarely is an improvement observed. Nevertheless, these results highlight the importance of balancing lipophilicity with permeability in degrader development. This is due to the fact that the difference in activity in HS vs FBS (known as a “serum shift”) is usually associated with the binding of drugs to the plasma proteins that are present to a significantly higher degree in HS, as opposed to FBS.³⁷ The likelihood of binding has been attributed to high drug lipophilicity.³⁸ Therefore, the low lipophilicity (according to logk₈₀^′ stack) and higher aqueous solubility of 14 is likely the reason for the lack of a serum shift for the degrader. This contrasts with 16, which loses its potency by almost 20-fold in human serum (Figure 2C). Interestingly, 21 only loses its potency in human serum by 3-fold, while still maintaining superior potency over other analogues within both FBS and HS (Figure 2B).

Figure 2.

In vitro cellular potency (against MV4-11 cells) in human serum (HS) versus fetal bovine serum (FBS) for compounds (A) 14, (B) 21, and (C) 16.

In this work, we show that tuning degrader physicochemical properties and potency can be done in the early stages of drug design without requiring dramatic changes in structure. Importantly, the retention of potency of the most lipophilic degraders in the WO experiment indicates that cell permeability was likely impacted by the particular design of this degrader series. The high logk₈₀^′ stack values that were observed for these degraders are correlated with the increased carbon content in the linker chain but may also indicate an increased ability to form IMHBs and thus adopt folded conformations that allow better permeability. This type of folding has previously been noted by Poongavanam and colleagues who reported that when degraders were analyzed in chloroform, the population of IMHB-driven folded conformations was predictive of cellular permeability.³⁹ Therefore, the degrader design strategy utilized in this study could positively impact both aqueous solubility and permeability. The discovery of a degrader, 14, with good cellular potency, aqueous solubility and potentially low plasma protein binding is reported. We show that click chemistry allows structural modifications that can significantly improve activity while conserving molecular weight and other chemical properties that are crucial for favorable drug-like properties. It is evident that subtle changes in the connectivity of linker groups can have significant impact on both potency, lipophilicity, and solubility. The achievement of simultaneous optimization across multiple parameters would be expected to lead to optimal in vivo candidates. In addition, the combinatorial nature of click chemistry, the ability to generate libraries of E3 ligase ligands functionalized with alkyne or azides, can facilitate the rapid generation of degraders for different protein targets. While the alkyl linker units used in this study may not be optimal, particularly regarding metabolic liability, the functional group serves as an effective way to manage extraneous factors that would interfere with the ability to investigate the specific impact of the triazole ring.⁴⁰ The continued development and application of this click chemistry strategy is expected to attenuate the strenuous nature of the costly trial-and-error approach that medicinal chemists currently adopt for the design of degrader linkers.

Glossary

Abbreviations

AML

Acute Myeloid Leukemia

CDK9

Cyclin Dependent Kinase 9

PROTACs

Proteolysis TArgeting Chimeras

POI

Protein of Interest

bRo5

beyond Rule-of-Five

SAR

Structure–Activity Relationship

KS

Kinetic Solubility

IMHBs

Intramolecular Hydrogen Bond Donors

TPSA

Topological Polar Surface Area

IC₅₀

Half-Maximal Inhibitory Concentration

DC₅₀

Half-Maximal Degradation Concentration

D_max, WO

Washout

HS

Human Serum

FBS

Fetal Bovine Serum.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.3c00082.

• Full synthetic procedures, spectroscopic characterization of select compounds and details of kinetic solubility and chromatographic logD experiments (PDF)

Author Present Address

^† Biomedical Graduate Education, Georgetown University, Washington, DC 20057

Author Present Address

^‡ Charles River Laboratories, Ashland, Ohio 44805.

Author Contributions

O.R.A and J.R.F. wrote the manuscript. Research was executed by O.R.A., C.S., and E.S. All authors approved the final revision of the manuscript.

Funding from the OSU James Comprehensive Cancer Center, the OSU Pelotonia Fellowship Program and the American Society of Hematology Minority Hematology Graduate Award is acknowledged with gratitude.

The authors declare no competing financial interest.