Design, synthesis, and biological evaluation of polyphenol derivatives as DYRK1A inhibitors. The discovery of a potentially promising treatment for Multiple Sclerosis

SUMMARY

Green tea and its natural components are known for their usefulness against a variety of diseases. In particular, the activity of main catechin Epigallocatechin gallate (EGCG) against Dual-specificity tyrosine-(Y)-phosphorylation Regulated Kinase-1A (DYRK1A) has been reported; here we are showing a structure-activity relationship (SAR) for EGCG against this molecular target. We have studied the influence of all four rings on the activity and the nature of its absolute geometry. This work has led to the identification of the more potent and stable trans fluoro-catechin derivative 1f (IC₅₀ = 35 nM). This molecule together with a novel delivery method showed good efficacy in vivo when tested in a validated model of multiple sclerosis (EAE).

Graphical Abstract

A sizeable and increasing body of evidence points to a role for DYRK1A in Alzheimer’s Disease (AD) pathogenesis through various pathways. DYRK1A is a member of the DYRK family, and it affects tau phosphorylation and the formation of tau neurofibrillary tangles¹. In addition, DYRK1A alters APP phosphorylation and induces amyloid-beta (Aβ) production², and DYRK1A expression in the hippocampus is increased in neurodegenerative diseases^3,4. Moreover, DYRK1A is strongly associated with neuroinflammation⁵. These findings support DYRK1A as a potential target for preventing or treating neurodegenerative diseases, particularly AD. Several laboratories have suggested treatments to reduce DYRK1A activity; however, the development of inhibitors is in its infancy. To date, most of the work has focused on developing ATP-competitive DYRK1A inhibitors like harmine, a β-carboline alkaloid⁶. This approach’s drawback is the well-known difficulty in achieving selectivity against the other kinases since the ATP site is highly conserved among kinases. The most advanced compound is SM07883, an orally bioavailable DYRK1A inhibitor that also shows potent inhibition toward other kinases like DYRK1B, CLK4, and GSK3β⁷. No DYRK1A inhibitor is currently in clinical development for AD.

EGCG, the principal component of green tea extract, is a potent inhibitor of DYRK1A⁸. Evidence that EGCG may possess the desired biological activities are the following: 1) it is a selective allosteric inhibitor of DYRK1A; 2) it can improve cognitive functioning and synaptic plasticity in DYRK1A overexpressing mice; 3) it shows efficacy in several animal models for neurodegeneration including AD. EGCG is potent, has a relatively good safety profile⁹ but poor pharmacokinetic properties^10,11. The poor bioavailability of EGCG is due to a combination of low permeability, chemical instability, and metabolic biotransformation, particularly at the gastrointestinal (GI) and hepatic levels (when given orally). EGCG, a BCS class 3 compound, has been shown to have very low oral bioavailability (F) in test animals; in mice, F has been reported to be up to 12%¹² and in rats up to 5%^13,14. EGCG can undergo distinct degradation pathways: in solution, the primary degradation pathway is the oxidation to the correspondent B-ring quinone intermediate leading to the formation of dimers like Theasinensin A¹⁵. Unfortunately, due to its poor chemical/metabolic stability and extremely poor access to the brain, EGCG is not well-suited for use as a drug. Accordingly, we have initiated a discovery project in which we aim to optimize the interaction of EGCG with DYRK1A and its pharmacokinetic properties.

Green tea is a mixture of several catechins, we tested several of them against DYRK1A by using the ELISA assay^8c, and the results are summarized in Table 1. Interestingly, trans catechin derivatives are generally more potent against DYRK1A than the correspondent cis (i.e. (−)GCG > (−)EGCG). Catechins (i.e., molecules that bear only 2 OHs in the B-ring like ECG and CG) are usually slightly less potent than the correspondent pyrogallol derivatives (i.e., molecules that bear 3 OH in the B-ring like EGCG and GCG). However, thanks to the trans geometry, CG is more potent than EGCG (Table 1). The absolute configuration has a small effect on potency since (−)GCG shows only slightly better activity than (+)GCG. Gallic acid (D-ring) is essential for the activity, for example, EGC IC₅₀>>100 μM. Finally, the methylated derivative EGCG-3”-OMe, the main component of the Benifuuki tea, is known to have about 6-fold improved oral bioavailability than EGCG¹⁶ but results in a significant loss in activity. (Table 1).

Table 1.

Natural catechin DYRK1A inhibitors.

Catechin	Chiral centers	R1	R2	DYRK1A IC50 (nM)
EGCG	2R,3R	OH	H	232
ECG	2R,3R	H	H	523
CG	2S,3R	H	H	220
(−)GCG	2S,3R	OH	H	121
(+)GCG	2R,3S	OH	H	150
EGCG-3”OMe	2R,3R	OH	Me	429

Potency and metabolic stability are only a few aspects of drug development; the other key for a CNS drug is reaching the target organ, i.e., the brain. Earlier studies about EGCG’s brain exposure showed that this molecule could not reach the brain in a high amount. Despite a few reports that showed some brain exposure, a more in-depth search, and our experimental data showed that EGCG has a very poor tendency to cross the blood-brain barrier (BBB). Despite the bioavailability problem, EGCG has demonstrated some very interesting in vivo activities against various disorders and is the subject of more than 20 ongoing clinical trials. However, because of the high doses needed to obtain therapeutic concentration, activity is sometimes associated with liver toxicity^9b, which led to trials’ early termination¹⁷. Therefore, improving the bioavailability and reducing the metabolism is necessary to develop viable catechin therapeutics. Despite the earlier improvement achieved by our group, molecules like GCG and CG still suffer from low bioavailability and poor brain exposure (data not shown). We have developed a new intranasal (IN) formulation for our catechin derivatives to overcome these problems. This is an ideal solution for this class of molecules since, at the same time, we can bypass GI and first-pass metabolism and, thanks to the absorption through the olfactory region, reach the brain without crossing the BBB. With our proprietary nasal formulation based on the penetration enhancer (2-hydroxypropyl)-β-cyclodextrin (HP-β-CD), we achieved high drug concentration in dosing solution (up to 27% w/w), which is fundamental for intranasal formulations. Using this formulation, we can increase several folds exposure to this class of drugs both systemically and in the brain. For example, in the case of (−)-CG, we achieved a 50-fold exposure increase with intranasal compared to the oral route in the brain and peripheral blood (data not shown). Because of the general applicability of the formula for this class of molecule, its simplicity, and ease of use for patients, we envision using it in animal models and future clinical studies.

Even with the IN delivery, we realized that GCG was still less than optimal since it suffers from low bioavailability and a short half-life. Increasing the absolute potency and/or bioavailability for these drugs against DYRK1A should be achieved due to the nature of the administration route. Intranasal delivery requires potent and bioavailable molecules because of the small surface available, limiting the amount of drug that can be absorbed. Improvement in the overall potency should help us enhance the overall selectivity profile and reduce the amount of drug that must be delivered to the brain and, therefore, will provide us with a better chance of success in the clinic. To achieve the above goals, we systematically explored the catechin scaffold. Here below, we describe our optimization work on the different parts of the original lead GCG.

A-ring SAR:

we explored the importance of the 2 hydroxyl groups in the A-ring of trans catechin GCG. The total synthesis used is described in Scheme 1 and is a modification of a known published protocol¹⁸. With the appropriate choice of starting materials, this synthesis has also been used to make B-ring analogs of GCG. O-Benzyl-protected 3,4,5 trihydroxybenzaldehyde (3) (synthesized from fully protected gallic acid 2 by reduction/oxidation protocol) was condensed with 2’-hydroxy acetophenone (4) in KOH/EtOH at room temperature for 16hrs to yield the chalcone 5. Then, the chalcone was cyclized directly to the Chromene 6 using the following protocol: To a solution of 5 (1.2 g, 2.21 mmol, 1.0 eq.) in THF (20 mL) and EtOH (6 mL) was added anhydrous CeCl₃ (1.36 g, 5.53 mmol, 2.5 eq.) and NaBH₄ (0.21 g, 5.53 mmol, 2.5 eq.) at 0 °C. The mixture was stirred at RT for 16 h and gave the desired intermediate 4 in 69% yield. Hydroboration-oxidation of 6 with BH₃/THF followed by H₂O₂/NaOH gave the trans-3-flavanol 7 (catechin). The trans stereochemistry of 7 was assigned based on the nuclear magnetic resonance (NMR) coupling constants and nuclear Overhauser effect (NOE) experiments. The alcohol intermediate 7 was esterified with 3,4,5- trimethoxybenzoyl chloride and deprotected by hydrogenation to afford the trans rac-A-ring analogs 1a and 1b.

Scheme 1.

Synthesis of A-ring derivatives.

Using the total synthesis reported in Scheme 1, we removed one (1b) or both (1a) hydroxy groups. As reported, removing both hydroxyl groups leads to a molecule that is about 4-times less potent against DYRK1A while removal of the 7-OH gives us a product that is almost 2-times less potent compared to the di-hydroxy derivatives. The Scheme 1 produces rac products. Since we noticed above that (+) enantiomers are only slightly less active than the (−) ones, we inferred that the activity of the rac compounds is representative of the (−) chiral enantiomers. From this data, we can extrapolate that both OHs are necessary for optimal activity against the enzyme.

C-ring SAR:

We tested the replacement of the O atom with C making the correspondent tetrahydronaphthalene derivatives. Synthesis of rac 1c was obtained in a multistep synthesis starting from dihydronaphthalene (9) as described in Scheme 2. The total synthesis started from 1,4-dihydronaphthalene 9, reaction with mCPBA gave in good yield the epoxide intermediate 10. Introduction of the “B-ring” was obtained by electrophile coupling reaction using aryl lithium derivative of protected pyrogallol derivative 11 and Lewis acid BF₃ˑEt₂O, this gave the desired compound (2R,3S)-3-(3,4,5-tris(benzyloxy)phenyl)-1,2,3,4-tetrahydronaphthalen-2-ol (12). Finally, the standard esterification reaction introduced the D-ring using the acid chloride of protected gallic acid 13. Final deprotection of intermediate 14 by hydrogenation gave the desired product 1c. Unfortunately, the activity of this compound was less than half compared to the correspondent O derivative (1a, Scheme 1, i.e., 422 vs 1055 nM). Furthermore, we observed that this modification does not improve stability in biological media. Because of the better chemical tractability and potency, we focused only on the O-catechin series.

Scheme 2:

Synthesis of tetrahydronaphthalene catechins.

D-ring SAR.

Considering that modification of both A and C rings did not improve these catechins’ overall activity, we focused the project on the D ring (Table 2). We have developed a robust 4-steps semi-synthetic procedure (Scheme 3) from readily available natural EpiGalloCatechin (EGC). Isomerization of EGC using phosphate buffer (pH=7.4) afford in good yield the trans derivative GalloCatechin (GC). The selective benzylation of the phenolic groups gives us the alcohol intermediate 15. The esterification reaction using acid 16 provides the ester intermediate 17. Final deprotection using hydrogenation reaction provide the desired gallocatechin derivative 1d-l with an overall good yield.

Table 2.

New D-ring derivatives. Compound 1d-l (Scheme 3)

Catechin	X	Y	Z	DYRK1A IC50 (nM)
(−)GCG	H	H	OH	121
1d	H	H	H	236
1e	H	OH	H	172
1f	H	F	OH	35
1g	OH	H	H	108
1h	F	F	OH	28
1i	H	F	H	107
1j	Me	H	H	553
1k	H	H	F	145
1l	H	CF3	H	432

Scheme 3.

Synthetic scheme for D-ring derivatives.

Earlier, we observed that capping the para OH or having only 1 OH in the D-ring will lead to high micromolar activity. Here we can see that removal of one OH in meta position (1d) results in a loss of activity (Table 2), and moving the meta OH to the ortho position (1e) still results in reduced activity. Overall we can conclude that two hydroxyls groups in the D-ring are essential for the activity against DYRK1A; however, three hydroxyls are best tolerated. Initially, we thought that introducing fluorine atom(s) should help improve the molecule’s overall metabolic stability. As such, we undertook the synthesis of new fluorinated catechin derivatives. To our surprise, the introduction of a fluoro in position 2” led to a dramatic increase in activity against DYRK1A: 1f (GCG-2” F) showed about 3-fold improvement in activity compared to GCG. The introduction of 2 fluorine atoms yielded the most potent compound to date 1h (28 nM, Table 2), although the improvement over 1f is marginal. Replacing a hydroxyl group with fluorine like in 1k did not improve potency. Finally, introduction in meta position of a more lipophilic group like methyl (1j) or trifluoromethyl (1l) was detrimental for the activity of this class.

B-ring SAR:

Based on the D-ring optimization results, we have designed and synthesized several new B-ring derivatives (Table 3) following the synthesis described in Scheme 1 using 4’,6’-bis(benzyloxy)-2’-hydroxy acetophenone in place of 2’-hydroxy acetophenone (4). Earlier in the project, we noticed that B and D rings are interchangeable in the interaction with the enzyme. This agrees with our observation that the same substitution on the B or D ring leads to similar activity, and (−) trans isomers are only slightly better than the corresponding (+) trans isomers against DYRK1A. Because of the easier access to the D-ring derivatives and based on our working hypothesis, we usually first tested new gallic acid derivatives as D-ring. If the activity is promising, we then embark on synthesizing the correspondent B-ring. Our theory of the B and D-rings’ interchangeability was also confirmed in this case: introducing the fluorine atom in the B-ring generated a product, 1n (GCG-2’F), that shows a potency like 1f. Unfortunately, the simple introduction of fluorine in the B ring did not translate into a stability improvement in biological media. When fluoro atoms are introduced in both rings, like in 1o (2’,2” diFGCG), no improvement is observed compared to the mono fluorinated analog 1f. The most interesting compound in this series is the catechin derivative 1r, which showed good potency and improved stability in biological media (plasma and hepatocytes media), typical of the catechin series (2 OH in the B-ring) compared to the correspondent epicatechin one. Thanks to introducing a fluorine atom, 1r is about 3-fold more potent than the natural trans analog catechin gallate (CG, Table 3).

Table 3.

New B-ring derivatives.

Catechin	X	Y	Z	DYRK1A IC50 (nM)
CG	H	H	H	220
1m	F	H	H	153
1n	F	H	OH	60
1o	F	F	OH	54
1p	F	F	H	115
1q	Me	H	OH	296
1r	H	F	H	74

In Figure 2 we have pictured the complete SAR developed in our laboratories against the molecular target DYRK1A. Trans catechins are more potent than corresponding cis analogs, and the introduction of the Fluorine atom in the B and or D-ring significantly increases this class’s potency. Most of the hydroxyls group in EGCG are essential for the activity, especially in the para position of the B and D-rings (i.e., EGCG-4”OMe DYRK1A IC₅₀ = 25,086 nM; EGCG-4’OMe DYRK1A IC₅₀ = 5,796 nM). Having 3 OHs in the B and D-ring gives the best activity, but compounds maintain nanomolar activity also with the removal of 1 OH in the meta position. We have chosen compound 1f¹⁹ for the in vivo efficacy studies based on these results.

Figure 2.

Catechin’s SAR as DYRK1A kinase inhibitors.

We examined the inhibitory potency of GCG and compound 1f with varying concentrations of ATP by the ELISA assay as previously described⁸. The potency was only marginally affected by [ATP] up to 0.8 mM. Therefore, we conclude that these catechin analog functions as a non-ATP competitive inhibitor like the parent compound, EGCG.

Neuroinflammation is a common feature across neurodegenerative disorders and is implicated in the progression of neurodegeneration. Dysregulated microglia activation causes neuroinflammation and has been highlighted as a treatment target in therapeutic strategies²⁰. It has been shown that DYRK1A regulates inflammatory signals like STAT²¹, GFAP, and NFAT3²². More recently, is has been proposed a novel mechanism of neuroprotection in which DYRK1A increases neuronal survival by enhancing the p21-Nrf2 pathway in glial cells²³. DYRK1A inhibitors can rescue neurodegeneration caused by central and peripheral inflammation; thus, they can be broadly applied to inflammatory conditions.

To examine this class of molecules’ potential application in neurodegenerative disorders, we used the MOG_35–55-induced murine model of chronic progressive Experimental Autoimmune Encephalomyelitis (EAE), an inflammation model predictive for multiple sclerosis (MS). In this model, we tested 1f both orally (PO) and IN and compared its activity against two reference compounds, namely EGCG and Fingolimod (FTY720). As shown in Figure 3, twice daily intranasal administration of 1f (15 mg/kg) resulted in a reversal of the disease scores similar to that observed with FTY720 (1 mg/kg QD). Oral administration of 1f (15 mg/kg BID) had no significant amelioratory effect until the last two days of the study, demonstrating the IN route’s superiority. Despite literature reports²⁴, we did not see any significant activity using EGCG in this model using our dosing regimen (15 mg/Kg, BID, PO).

Figure 3.

Efficacy in the EAE therapeutic model

Pro-inflammatory cytokines IL-17, TNFα, and IFNγ have been important in the development of inflammation and neurological damage in EAE and in the pathogenesis of other autoimmune diseases²⁵. TNF-α, a Th1 cytokine, promotes EAE symptoms and pathological characteristics^26,27. All these cytokines are up-regulated in human MS and murine EAE^28,29. Analysis on the brain of the animals showed that intranasally 1f-fed mice compared with the control mice had lower levels of these cytokines than control mice; in particular, we noted a significant reduction in brain IFNγ (90%) and IL-17 (79%) (Figure 4). These results suggest that 1f-induced improvement in EAE may be partly mediated by its effect on these inflammatory cytokines.

Figure 4.

Compound 1f decreases pro-inflammatory cytokine production in the brain. *P<0.05; **P<0.01;***P<0.001, ****P<0.0001 one-way ANOVA. Data are presented as the mean ± standard error of the mean (n=10).

To further examine the consequences of drug interference in the chronic EAE model, we performed the histopathological analysis of spinal cord sections from the mice included in the experiment shown in Figure 3. We focused on the effects of the different therapy regimes in preventing inflammation, demyelination, and axonal neurodegeneration of the spinal cord by light microscopy examination after the standard hematoxylin and eosin (H&E), Luxol Fast Blue (LFB), and Bielschowsky silver staining, respectively (Figure 5). Semi-quantitative assessment for each parameter (scale 0–5: 0 = absent, 1 = minimal, 2 = mild, 3 = moderate, 4 = marked, 5 = severe) was done by scoring quadrants of each of the eight transverse sections.

Figure 5.

Effect of therapy on histological markers. *P<0.05; **P<0.01;***P<0.001, ****P<0.0001 one-way ANOVA. Data are presented as the mean ± standard error of the mean (n=10).

The vehicle-treated mice showed lesions in spinal cord sections, including inflammatory cell infiltration in the perivascular and meningeal spaces, inflammatory cell infiltration into the white matter, demyelination, and axonal degeneration.

Histologically, FTY720 therapy resulted in mild to substantial reductions in lesion severity scores in all three parameters examined when compared to the vehicle control (Figure 5). Treatment with EGCG resulted only in slight but not statistically significant decreases in mean degeneration revealed by the H&E staining and demyelination scores compared to the vehicle treatment. Mice treated intranasally with 1f exhibited significant reductions in lesion severity scores in all three parameters when compared to the other treatment groups and is similar to that of FTY720 (Figure 5). It is also clear that the intranasal administration of 1f was more effective at reducing lesion severity than oral administration. Overall, the histological analyses closely reflect what is observed for the therapeutic outcomes, as shown in Figure 3.

Compound 1f appears to have a peripheral immunomodulation effect and an overall neuroprotective effect, which is very promising and could have important implications in treating neurodegenerative disorders like AD, Parkinson’s Disease, and Multiple Sclerosis (MS). Nowadays, new strategies in treating MS aim to target both the immune system and neurodegeneration, areas in which 1f seems to have an important role. Furthermore, its inherent antioxidant activity will help reduce MS onset risk and inhibit its irreversible progression. This study’s strong signal could open the road for developing 1f as a breakthrough therapy in MS.