Biological Evaluation of 5′-(N-Ethylcarboxamido)adenosine Analogues as Grp94-Selective Inhibitors

Abstract

The heat shock protein 90 kDa (Hsp90) family of chaperones is highly sought-after for the treatment of cancer and neurodegenerative diseases. Glucose regulated protein 94 (Grp94) is the endoplasmic reticulum localized isoform that is responsible for the maturation of proteins involved in cell adhesion and the immune response, including Toll-like receptors, immunoglobulins, and integrins. Consequently, Grp94 has been implicated in many different diseases including cancer metastasis, glaucoma, and viral infection. 5′-(N-Ethylcarboxamido)adenosine (NECA) was identified from a high-throughput screen as one of the first molecules to exhibit isoform selectivity toward Grp94, with the ethyl group projecting into a unique pocket within the ATP binding site of Grp94. This pocket has since been exploited by several groups to develop Grp94 selective inhibitors. Despite success in the development of other classes of inhibitors, relatively little work has been done to further develop inhibitors with the NECA scaffold. Unfortunately, NECA is also a potent adenosine receptor agonist, which is likely to confound any biological activity. Therefore, structure–activity relationship studies were performed on the NECA scaffold leading to the discovery of several molecules that displayed similar selectivity and affinity as the parent compound.

The 90 kDa heat shock proteins (Hsp90) represent a family of molecular chaperones that are responsible for the conformational maturation of a variety of client protein substrates.¹ These client proteins are involved in numerous cellular processes, such as cell survival, hormone signaling, cell cycle control, and stress response. Because many of these pathways are dysregulated in cancer and Hsp90 modulates these processes, Hsp90 has emerged as a promising target for the development of anticancer agents.^2,3

While several Hsp90 inhibitors have undergone clinical evaluation, they also encountered both off- and on-target issues that prevent progression through the clinic.⁴ One concern is the lack of selectivity manifested by the compounds, as they exhibit similar affinity toward all four isoforms of Hsp90: Hsp90α, Hsp90β, Grp94, and TRAP1. It appears that inhibition of all four isoforms contributes to clinical detriments and, consequently, necessitates the need for a more focused approach. As a result, much work has sought to develop isoform-selective inhibitors of Hsp90 to overcome these detriments and to elucidate isoform-dependent client protein substrates.⁵⁻⁸

Grp94 is the Hsp90 isoform that is localized to the endoplasmic reticulum and mediates the trafficking and maturation of numerous proteins associated with cell–cell adhesion and cell migration. Toll-like receptors, integrins (α-2, α-4, α-L, and β-4), insulin-like growth factors, immunoglobulins, and mutant myocilin represent proteins that have been shown to be Grp94-dependent.^9,10 Grp94 also interacts with Her2 and is necessary for trafficking the receptor to the cell membrane. Consequently, Her2 overexpressing cancer cells exhibit an increased sensitivity to Grp94 inhibition.⁶ In addition, Grp94 inhibition leads to the disaggregation of mutant myocilin, which presents an exciting therapeutic opportunity to treat primary open-angle glaucoma (POAG).¹¹

Previous studies utilizing a high throughput screen identified 5′-(N-ethylcarboxamido)adenosine (NECA) as one of the first molecules to exhibit isoform selectivity, as it bound Grp94 but displayed no affinity toward the cytosolic isoform, Hsp90α.^12,13 Differential binding to Grp94 is best explained upon analysis of the N-terminal ATP-binding domain. While the ATP-binding pocket is highly conserved across all four isoforms, Grp94 contains a five amino acid insertion (QEDGQ) into the primary sequence that leads to the formation of a hydrophobic binding pocket that is not present in the other Hsp90 isoforms.¹⁴ Elucidation of these interactions by solution of the cocrystal structure of NECA bound to Grp94 revealed the 5′-ethyl group to project into this unique pocket. In contrast, access to this pocket is restricted in Hsp90α, and the cocrystal of NECA bound to Hsp90α confirms the 5′-ethyl substituent to arrange in an energetically less favorable conformation to avoid unfavorable steric interactions.¹⁵

While NECA has been used to investigate the role played by Grp94, its high affinity for the adenosine receptors limits its potential use as a cellular tool. Even though it was one of the first isoform selective inhibitors discovered, few studies have followed up to investigate NECA analogues for Grp94 inhibition.¹⁵ Therefore, we disclose herein a study of NECA derivatives, including those that contain a rigid and/or bicyclic ribose ring system and their evaluation as Grp94 selective inhibitors. Both commercially available and synthetic analogues (Supporting Information, Schemes S1–S3) of adenosine were included. Many but not all the nucleosides, like NECA, were originally reported as adenosine receptor agonists.¹⁶

Monosubstituted derivatives of adenosine (1) were the first analogues to be examined. The compounds were evaluated for their affinity and selectivity against Grp94 and Hsp90α, using a competition fluorescence polarization assay.¹⁷ Compounds were evaluated by their ability to displace FITC-tagged geldanamycin in a dose-dependent manner, and apparent K_d values obtained are provided in Table 1. For compounds in which the apparent K_d was above 50 μM, the highest concentration used, percent binding at 50 μM is reported. At the C2 position, the presence of a Cl (2) increased affinity for Grp94, but a 2-thiol (3) completely abolished affinity. The 6-hydrazino (4) or 6-methylamino (5) modification prevented Grp94 binding. At the 5′ position, some substitutions were tolerated. 5-Amino-5′-deoxyadenosine (6) exhibited no affinity for Grp94; however, N-acetylation (7) reestablished Grp94 binding. 5′-Phosphoramidate 8 did not bind Grp94 and was completely inactive. Nicotinate ester 9 was inactive, which is consistent with the existence of a sterically restricted subpocket in this region of the protein. The cyclized anhydro-8-oxoadenosine (10) exhibited modest affinity for Grp94, consistent with the requirement for an anti-conformation about the glycosidic bond.

Table 1. Fluorescence Polarization Affinity Data of Compound Library^a.

^aAll values in μM. Data are presented as the average of at least two experiments ± SEM.

5′-(N-Alkylcarboxamido)adenosines 11–23 manifested a similarly moderate affinity for Grp94, except when the substitution contained a charged moiety (24) or contained a carboxamide (22). Consistently, the adenosine 5′-carboxylic acid (25) and its primary carboxamide (16) were inactive or lacked significant affinity for Grp94. Like NECA (11), the small alkyl carboxamides tended to be selective for Grp94 as compared to Hsp90α. The methyl (12) and cyclopropyl (14) analogues were 46- and >42-fold selective, respectively. However, the substitution of fluorines for hydrogens resulted in altered selectivity. Difluoro-cyclopropyl analogue 15 exhibited significantly lower selectivity for Grp94 versus Hsp90α as compared to 14. Cyclobutyl analogue (17) also shifted in the same direction, and the inclusion of an ether (18) enhanced Hsp90α affinity. Fluoro-NECA derivatives 19 and 20 exhibited considerably lower affinity for Grp94 than NECA, with the difluoro (20) displaying higher affinity among the pair. 3-Fluoro (21) or dihydroxy (23) substitution of the N-propyl carboxamide (13) led to a minor enhancement of affinity for Grp94 (21) or Hsp90α (23). A combination of 5′ and C2 substitution in the ribose series (26) was also tolerated.

Replacing the ribose ring with a rigid (N)-methanocarba (bicyclo[3.1.0]hexane)¹⁶ was tolerated by Grp94. Compound 27 displayed similar Grp94 affinity as compared to the ribose analogue, 2. In some cases, the (N)-methanocarba derivative displayed higher affinity and greater selectivity than its corresponding ribose analogue, e.g., 34 cf. 15, or maintained affinity, e.g., 30 cf. 20. Interestingly, the (N)-methanocarba analogue of NECA (32) displayed lower affinity for both Grp94 and Hsp90α while still maintaining good selectivity for Grp94. In contrast to the ribose series, the 2-chloro substitution in the (N)-methanocarba series did not affect Grp94 affinity, e.g., 29 cf. 28 and 33 cf. 32. Consistent with the ribose series, the N-6 modification (38) was found to be deleterious to Grp94 affinity.

Grp94 is responsible for the maturation and localization of numerous extracellular proteins, including several integrin subtypes that contribute to cell migration.¹⁸ Integrins are essential for cellular adhesion to the extracellular matrix by facilitating interactions with the intracellular actin cytoskeleton.¹⁹ Inhibition or knockdown of Grp94 impairs integrin localization to the cell surface, and consequently, inhibits cell migration.^20,21 To correlate in vivo binding of these analogues to inhibition of Grp94, compounds 14 and 33 were evaluated for their ability to inhibit migration in a wound healing scratch assay. These compounds were chosen as representatives of the two scaffolds.

Each compound was evaluated for its ability to inhibit migration of the highly metastatic prostate cancer cell line, PC3, at 50 and 25 μM (Table 2). While 14 exhibited higher affinity for Grp94, it had no effect on the migration of PC3 cells, whereas 33 was able to slow the migration of PC3 cells after both 24 and 48 h. This unexpected result suggests that while both scaffolds bind Grp94, the more rigid methanocarbo scaffold may be more efficient. An additional member of this scaffold manifests similar selectivity to 33; compound 34 was also evaluated for its ability to inhibit the migration of two metastatic cancer cell lines, A549 and MDA-MB-231, as Grp94 has been implicated in the progression and aggressiveness in both cancer cell lines.^22,23 As shown in Tables 3 and 4, at 25 μM, 33 and 34 were able to impair wound closing against both cell lines for 24 h. Against A549 cells, 33 and 34 resulted in 46% and 35% wound closure, respectively, as compared to 59% for the vehicle control (Figure 1). At later time points, the compounds manifested stronger antimigratory activity, and at 36 h the control was 88% closed, as compared to 56% for 33 and 67% for 34. Against MDA-MB-231 cells, the ability of 33 and 34 to alter the rate of wound healing in a dose dependent manner was as also investigated (Table 5). At 10 μM, only 33 was able to impair wound healing, while at 5 μM neither compound exhibited any effect on the rate of wound closure. In addition, both compounds were evaluated for antiproliferative activity, and neither exhibited any such activity at concentrations up to 50 μM. This confirms that the antimigratory activity exhibited by these compounds is not linked to cell viability. Complete wound healing scratch assay images for both cell lines available in the Supporting Information (Figures S2 and S3).

Table 2. Wound Healing Scratch Assay Results against PC3 Cells^a.

		14		33
time (h)	DMSO	50 μM	25 μM	50 μM	25 μM
24	47%	45%	44%	21%	32%
48	76%	65%	72%	39%	62%

^aPercent closed.

Table 3. Wound Healing Scratch Assay Results against A549 cells^a.

time (h)	DMSO	33 (25 μM)	34 (25 μM)	39 (2.5 μM)
24	59%	35%	46%	77%
36	88%	56%	67%	91%

^aPercent closed.

Table 4. Wound Healing Scratch Assay Results against MDA-MB-231 Cells^a.

time (h)	DMSO	33 (25 μM)	34 (25 μM)	39 (2.5 μM)
24	64%	41%	43%	62%
48	93%	70%	63%	90%

^aPercent closed.

Figure 1.

Wound healing scratch assay results after 24 h treatment with 33, 34, or 39 against A549 cells (N = 3).

Table 5. Dose Dependent Wound Healing Scratch Assay Results against MDA-MB-231 Cells^a.

		33		34
time (h)	DMSO	10 μM	5 μM	10 μM	5 μM
24	64%	48%	63%	61%	62%

^aPercent closed.

In addition, compound 39 (Figure 2) was also evaluated for its ability to modulate the migration of both cancer cell lines. Compound 39 was used as a nonselective adenosine receptor agonist,²⁴ as adenosine receptor agonists such as nonselective NECA or more selective A₁ and A_2A receptor agonists have been observed to accelerate the rate of wound closure in wound healing scratch assays.²⁵ Compound 39 was chosen because the N⁶-2,2-diphenylethyl group was expected to prevent any interaction with Grp94/Hps90α, which was confirmed via a fluorescence polarization assay wherein 39 displayed no affinity for either isoform at 50 μM.

Figure 2.

N⁶-Diphenylethyladenosine, nonselective adenosine receptor agonist, 39.

Interestingly, 39 had no effect on the rate of wound closure against MDA-MB-231 cells at 20 μM after both 24 and 48 h. A possible explanation is that MDA-MB-231 cells only expresses the A_2B subtype of the adenosine receptor and that activation of multiple adenosine receptor subtypes more effectively promotes wound healing.²⁸ When 39 was evaluated against A549 cells, which express more than just the A_2B subtype, 2.5 μM 39 induced an accelerated rate of wound healing when evaluated at both the 24 and 36 h time points (Table 2).²⁷

To validate this hypothesis 39 was evaluated against the hA_2B adenosine receptor. The affinity of 39 for other adenosine receptor subtypes has been previously reported, and are (K_i, nM): human (h) A₁, 49.9, hA_2A, 510, hA₃, 3.9.²⁶ Compound 39 was found to exhibit an EC₅₀ of 485 ± 65 nM in an adenylate cyclase hA_2B-expressing CHO cell line as previously described.^27,29 This was then compared to NECA, which produced an EC₅₀ of 94.8 ± 15.1 nM (Figure S1). Thus, we confirmed that 39 is a pan-adenosine receptor agonist when applied at >1 μM. This supports the hypothesis that activation of multiple adenosine receptors can effectively promote wound healing.

Encouraged by the wound healing scratch assay data, and to validate that 33 and 34 were exhibiting their antimigratory activity through Grp94 inhibition, the ability of 33 to induce the degradation of the Grp94-dependent client protein, integrin α2, was investigated. As shown in Figure 3, 33 induced the degradation of integrin α2 in SKBr-3 cells at 25 μM, the same concentration at which 33 demonstrated antimigratory activity, providing evidence that the decrease in cellular migration is due to the inhibition of Grp94. Consistent with the dose-dependent data, at 5 μM, 33 no longer manifested any effect on integrin α2 levels, suggesting that the ability of this compound to affect the rate of wound closure is related to its ability to induce the degradation of integrin α2. In addition, Akt levels were investigated, as Akt is a well-known client protein of Hsp90α and Hsp90β and does not require Grp94 for maturation.³⁰ If the compounds selectively inhibit Grp94 over Hsp90α/β, as suggested by the fluorescence polarization data, which showed 33 to exhibit ∼17-fold selectivity for Grp94 versus Hsp90α, then Akt levels should be unaffected upon treatment with 33. Consistent with the affinity data, 33 did not induce the degradation of Akt, suggesting that 33 selectively binds to, and inhibits Grp94 in the cellular environment. One of the issues encountered by pan-inhibitors is that they induce the heat shock response, which leads to the upregulation of Hsp90 and Hsp70 and requires increased concentrations of compound. Ultimately, this can lead to dose escalating toxicity. One way to measure induction of the heat shock response is to observe Hsp70 levels. Consistent with other isoform selective inhibitors, treatment with 33 did not increase Hsp70 levels, suggesting that this compound is able to inhibit Grp94 without inducing the heat shock response.^5,20

Figure 3.

(A) Western blot analysis of SKBr-3 cells treated with varying doses of Grp94-selective NECA analogue 33. Shown with representative actin loading control. (B) Ratio of integrin α2, Akt, and Hsp70 to vehicle treated control normalized to actin (N = 5).

Grp94 has emerged as an important target in a variety of diseases including cancer, glaucoma, neurodegenerative disorders, as well as viral and parasitic infections due to Grp94-dependent clients that contribute to disease progression. Despite being one of the first molecules discovered to selectively bind Grp94 over the cytosolic Hsp90 isoforms, NECA has not received the same attention as other scaffolds. This study investigated 38 NECA and adenosine analogues for Grp94 affinity and selectivity versus Hsp90α. The perception that NECA selectively binds Grp94 over Hsp90α was challenged by investigating the inhibitory activity manifested by NECA against the ATPase activity of several members of the heat shock protein family.³¹ However, the data presented in this study provides evidence that NECA and structurally related analogues are Grp94-selective inhibitors. In fact, affinity data provide clear structure–activity relationship (SAR) trends to support the further development of more selective and potent derivatives. In addition to affinity data, functional characterization studies demonstrated that this class of molecules are selective inhibitors and 33 can induce the degradation of Grp94-dependent client proteins, without affecting Hsp90α/β-dependent client protein maturation

As the 5′-ethyl group of NECA projects into the unique binding pocket of Grp94, and explains the preferential binding toward this isoform, much of this study was aimed at probing SAR for this moiety. However, this study shows that this area of the pocket does not tolerate much deviation from the parent compound, as only those compounds that contained either methyl, ethyl or cyclopropyl substituents retained affinity and selectivity for Grp94. This study also showed for the first time that alterations to either the ribose or the purine regions of the molecule are tolerated. Current efforts are underway to explore whether other halogenation patterns are similarly tolerated as well as to provide additional insights into the selective nature of these compounds. One of the other factors that has hindered the development of NECA analogues as Grp94 selective inhibitors is the affinity of the parent compound toward the adenosine receptors. Intriguingly, cyclization of the sugar ring and the purine base that leads to formation of 10 has been shown to reduce adenosine receptor affinity by up to 1000-fold.³² Since 10 exhibits similar affinity and selectivity toward Grp94 as the lead compounds used in this study, 10 represents a promising step toward deconvoluting adenosine receptor and Grp94 affinity.

Glossary

Abbreviations

CHO

Chinese hamster ovary

Grp94

glucose regulated protein 94

NECA

5′-(N-ethylcarboxamido)adenosine

Hsp90

heat shock protein 90 kDa

TRAP1

TNF receptor-associated protein 1

FITC

fluorescein isothiocyanate

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.0c00509.

• Synthetic experimental details, characterization of compounds, and biological data (PDF)

Author Contributions

^§ D.K.T. and C.M.B.contributed equally to this work. The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

NIDDK Intramural Research Program (ZIADK31117). NIH Extramural (CA213566).

The authors declare no competing financial interest.