Multicomponent Synthesis of Polyphenols and Their In Vitro Evaluation as Potential β-Amyloid Aggregation Inhibitors

While plant polyphenols possess a variety of biological properties, exploration of chemical diversity around them is still problematic. Here, an example of application of the Ugi multicomponent reaction to the combinatorial assembly of artificial, yet “natural-like”, polyphenols is presented. The synthesized compounds represent a second-generation library directed to the inhibition of β-amyloid protein aggregation. Chiral enantiopure compounds, and polyphenol-β-lactam hybrids have been prepared too. The biochemical assays have highlighted the importance of the key pharmacophores in these compounds. A lead for inhibition of aggregation of truncated protein AβpE3-42 was selected.


Introduction
Natural polyphenols of plant origin are important elements of our diet and, for this reason, their biological properties have been thoroughly studied [1,2]. Their most renowned characteristic is their antioxidant activity, which is believed to play an important role in preventing age-related diseases such as atherosclerosis. Since long ago, plant polyphenols have attracted much interest in the nutraceutical and cosmetic fields. On the other hand, the pharmaceutical industry has remained so far rather detached regarding the investigation of plant polyphenols as possible leads for drug development [3]. The reasons for this lack of interest are not fully clear but may be related to the poor pharmacokinetic properties of natural polyphenols, to their in vivo instability, and to the challenging synthetic modification of the natural members of this family. Actually, only few studies on the preparation of artificial analogues through total synthesis or semi-synthesis have been reported so far [3][4][5][6][7][8][9][10][11][12]. Nevertheless, several reports have pointed out that polyphenols, apart from anti-oxidant properties, may have a variety of other biological effects, such as anti-inflammatory [13], anti-cancer [12,14,15], anti-microbial [16], and anti-hyperglycemic [17] activity. Last but not least, some natural polyphenols have been demonstrated to be able to inhibit β-amyloid aggregation, thus being promising for the prevention of the Alzheimer's disease [18][19][20][21].
We reasoned that a "natural fragment-based approach" [22] to the combinatorial synthesis of artificial polyphenols starting from small, phenolic, building blocks, would allow a more systematic exploration of their chemical space, allowing to select possible hits for drug discovery,

Synthesis
Scheme 1 shows a specific example of the general synthetic strategy used for the preparation of polyphenols. This strategy was optimized also on the basis of the outcomes of our initial preliminary work. The need, in particular for in vivo experiments, of high purity final polyphenols 8, prompted us to prepare compounds 8 through a high yielding solvolysis of polyacetates 7. In this way, the purification is very simple, and we can avoid an extractive and chromatographic purification of the polyphenols, which, in some cases, although not always, gave insufficiently pure compounds. On the other hand, the acetate protecting group was found to be not fully stable under the conditions of the Ugi reaction. Thus, our standard, optimized, procedure involved the use of phenolic building blocks protected as allyl ethers. The Ugi reaction gave best results when preformation of the imine was implemented and CF3CH2OH (TFE)/ethanol mixture was used as the solvent. Removal of the allyl protecting groups was best performed with a minimum amount of Pd catalyst and ammonium formate as the scavenger, followed by immediate acetylation. Chromatographic purification afforded acetates 7, which were obtained in high purity and fully characterized at this level. Finally,

Synthesis
Scheme 1 shows a specific example of the general synthetic strategy used for the preparation of polyphenols. This strategy was optimized also on the basis of the outcomes of our initial preliminary work. The need, in particular for in vivo experiments, of high purity final polyphenols 8, prompted us to prepare compounds 8 through a high yielding solvolysis of polyacetates 7. In this way, the purification is very simple, and we can avoid an extractive and chromatographic purification of the polyphenols, which, in some cases, although not always, gave insufficiently pure compounds. On the other hand, the acetate protecting group was found to be not fully stable under the conditions of the Ugi reaction. Thus, our standard, optimized, procedure involved the use of phenolic building blocks protected as allyl ethers. The Ugi reaction gave best results when preformation of the imine was implemented and CF 3 CH 2 OH (TFE)/ethanol mixture was used as the solvent. Removal of the allyl protecting groups was best performed with a minimum amount of Pd catalyst and ammonium formate as the scavenger, followed by immediate acetylation. Chromatographic purification afforded acetates 7, which were obtained in high purity and fully characterized at this level. Finally, deacetylation was performed with a protocol that completely avoided both extractive and chromatographic purification of the final polyphenols 8. As shown in Scheme 1, for compound 8c this procedure was very efficient, and the final HPLC purity was 99%.  [25]. In that paper, we tested the polyphenols on two amyloid proteins: Aβ1-42 and AβpE3-42. For the first one, the best compound was found to be 8a. On the smaller fragment, the best hits contained a residue derived from caffeic acid, but we later found that polyphenols containing the catechol moiety typical of caffeic acid where less stable and, most of all, cytotoxic to neuronal cells [26]. Thus, we decided to keep, as the lead for AβpE3-42, compound 8b, which was the best among those derived from ferulic acid. Other similar polyphenols prepared by us are shown in Figure 2. In designing the new entities to be synthesized, we took as a model the best compounds that emerged from the preliminary studies [25].
In that paper, we tested the polyphenols on two amyloid proteins: Aβ1-42 and AβpE3-42. For the first one, the best compound was found to be 8a. On the smaller fragment, the best hits contained a residue derived from caffeic acid, but we later found that polyphenols containing the catechol moiety typical of caffeic acid where less stable and, most of all, cytotoxic to neuronal cells [26]. Thus, we decided to keep, as the lead for AβpE3-42, compound 8b, which was the best among those derived from ferulic acid. 8f and 8g we replaced p-hydroxybenzaldehyde (that was the aldehyde component for preparation of both 8a and 8b) with an aliphatic aldehyde (isobutyraldehyde). In 8e we replaced t-butyl isocyanide with a phenol containing an aromatic isocyanide, whereas for 8g and 8h a benzyl isocyanide containing a phenol was used. Finally, we also prepared compound 9, which is the analogue of 8a, totally devoid of phenolic groups, in order to check the importance of this moiety for biological activity.  Since the ferulic derived part seemed to be very important, in this study we maintained the carboxylic building block, and varied only the other three components. For example, looking at the structure of 8a, we replaced benzylamine with a phenol containing benzylamine (see 8c and 8d). In 8f and 8g we replaced p-hydroxybenzaldehyde (that was the aldehyde component for preparation of both 8a and 8b) with an aliphatic aldehyde (isobutyraldehyde). In 8e we replaced t-butyl isocyanide with a phenol containing an aromatic isocyanide, whereas for 8g and 8h a benzyl isocyanide containing a phenol was used. Finally, we also prepared compound 9, which is the analogue of 8a, totally devoid of phenolic groups, in order to check the importance of this moiety for biological activity.
As far as it concerns the overall yields of the various syntheses, the most critical step was found to be the Ugi reaction. In general, we can say that aromatic isocyanides (see 6d) and aliphatic aldehydes (see 6e and 6f) bring about a less efficient multicomponent reaction. The high yield achieved for compound 9 demonstrates that also protected ferulic acid, as the carboxylic component, is not ideal for the Ugi reaction. The decrease in the yield is typically due to sluggish reactions and incomplete conversion. In some instances, we have determined the yield considering the recovered starting aldehyde. However, it should be noted that, since our main goal was to assess the biochemical properties of our polyphenols, the individual syntheses have not been optimized.
All the polyphenols prepared in the previous paper and those depicted in Scheme 1 and Figure 2 are racemic. To verify the possible influence of the absolute configuration of the stereogenic center generated during the Ugi reaction, we decided to use, as amine component, enantiopure α-methylbenzylamine 10 (Scheme 2).
As far as it concerns the overall yields of the various syntheses, the most critical step was found to be the Ugi reaction. In general, we can say that aromatic isocyanides (see 6d) and aliphatic aldehydes (see 6e and 6f) bring about a less efficient multicomponent reaction. The high yield achieved for compound 9 demonstrates that also protected ferulic acid, as the carboxylic component, is not ideal for the Ugi reaction. The decrease in the yield is typically due to sluggish reactions and incomplete conversion. In some instances, we have determined the yield considering the recovered starting aldehyde. However, it should be noted that, since our main goal was to assess the biochemical properties of our polyphenols, the individual syntheses have not been optimized.
All the polyphenols prepared in the previous paper and those depicted in Scheme 1 and Figure  2 are racemic. To verify the possible influence of the absolute configuration of the stereogenic center generated during the Ugi reaction, we decided to use, as amine component, enantiopure αmethylbenzylamine 10 (Scheme 2).  As expected, the Ugi reaction was poorly diastereoselective (and slower compared to the ones employing unsubstituted benzylamines). The two diastereomers could be conveniently separated at this stage, and then independently converted into enantiopure polyphenols 13a and 13b.
Finally, we wanted to prepare some rigidified analogues, by applying a post-MCR cyclization. In a previous work within our group [27], we have noticed that Ugi adducts derived from glycolaldehyde could be cyclized to β-lactams, exploiting the isocyanide derived secondary amide as the nucleophile in an intramolecular S N 2 process [23]. This interesting reaction was not fully explored, but we reasoned that in the present case it could offer a possibility to obtain polyphenol-β-lactam hybrids as rigidified analogues of 8a and 8b. Scheme 3 shows this approach.  As expected, the Ugi reaction was poorly diastereoselective (and slower compared to the ones employing unsubstituted benzylamines). The two diastereomers could be conveniently separated at this stage, and then independently converted into enantiopure polyphenols 13a and 13b.
Finally, we wanted to prepare some rigidified analogues, by applying a post-MCR cyclization. In a previous work within our group [27], we have noticed that Ugi adducts derived from glycolaldehyde could be cyclized to β-lactams, exploiting the isocyanide derived secondary amide as the nucleophile in an intramolecular SN2 process. [23] This interesting reaction was not fully explored, but we reasoned that in the present case it could offer a possibility to obtain polyphenol -β-lactam hybrids as rigidified analogues of 8a and 8b. Scheme 3 shows this approach.
Glycolaldehyde dimer 14 has been already used by us and others in Ugi or Ugi-type reactions [27][28][29]. Ugi reactions with 14 are known to proceed only in moderate yields, and, as stated above, protected ferulic acid, like other α,β-unsaturated acids, is not an ideal component for this MCR. Thus, we were not surprised by the low to moderate yields achieved for 15a and 15b. Cyclization was then carried out using sulfonyl diimidazole in the presence of NaH. This method, first introduced by Hanessian [30,31], is a variant of the better renown Mitsunobu reaction, and offers advantages in terms of operational simplicity and atom economy. [32,33] Formation of the β-lactams took place quite fast, but it was much cleaner starting with the more acidic aromatic secondary amine 16b, whereas with the t-butyl derivative 16a several side products were present, lowering the yields.
In this case, we preferred to avoid the final deacetylation. Therefore, after deprotection of the allyl groups, pure final phenols 17a,b were directly obtained by chromatography in highly pure form, being not susceptible to degradation under chromatographic conditions.

Biochemical Assays
Then, we investigated the interaction of our candidates with two β-amyloid peptides, Aβ1-42 and AβpE3-42. While the full-length Aβ1-42 is one of the most abundantly identified in the brain deposits, AβpE3-42 is a peptide N-terminal truncated at residue 3 (Glu) and further modified by Glycolaldehyde dimer 14 has been already used by us and others in Ugi or Ugi-type reactions [27][28][29]. Ugi reactions with 14 are known to proceed only in moderate yields, and, as stated above, protected ferulic acid, like other α,β-unsaturated acids, is not an ideal component for this MCR. Thus, we were not surprised by the low to moderate yields achieved for 15a and 15b. Cyclization was then carried out using sulfonyl diimidazole in the presence of NaH. This method, first introduced by Hanessian [30,31], is a variant of the better renown Mitsunobu reaction, and offers advantages in terms of operational simplicity and atom economy [32,33]. Formation of the β-lactams took place quite fast, but it was much cleaner starting with the more acidic aromatic secondary amine 16b, whereas with the t-butyl derivative 16a several side products were present, lowering the yields.
In this case, we preferred to avoid the final deacetylation. Therefore, after deprotection of the allyl groups, pure final phenols 17a,b were directly obtained by chromatography in highly pure form, being not susceptible to degradation under chromatographic conditions.

Biochemical Assays
Then, we investigated the interaction of our candidates with two β-amyloid peptides, Aβ1-42 and AβpE3-42. While the full-length Aβ1-42 is one of the most abundantly identified in the brain deposits, AβpE3-42 is a peptide N-terminal truncated at residue 3 (Glu) and further modified by cyclization of Glu (E) to pyroglutamic acid (pE). These structural modifications are known to increase AβpE3-42 aggregation propensity [34].
First, in order to check that no precipitation of our phenols could occur under the assay conditions, we determined the solubility in phosphate buffer solution (PBS) at pH 7.4 containing 1% of DMSO. This percentage of DMSO does not alter the aggregation of β-amyloids, as demonstrated by control experiments carried out with the solvent alone. This analysis was performed by UV-VIS spectroscopic monitoring of the solutions of the compounds at different concentrations, both at the λ max (at around 330 nm) and at 405 nm to evidence the formation of turbidity. Only compound 8e was found to be fully soluble in the whole range of concentrations tested (10-500 µM). In this case the absorbance curve at 333 nm was linear and the line at 405 nm flat. For the other compounds bending of the absorbance curve and increase of turbidity started at the concentrations indicated in Table 1, which were anyway far beyond the one used for the subsequent experiments (25 µM). Interestingly, most of the newly synthesized compounds turned out to be more soluble than our previous lead 8a, and this may represent ad advantage for in vivo tests. The aggregation inhibition experiments were carried out with the thioflavin T methodology. This assay was selected during our previous work as the most helpful in order to predict activity of our polyphenols. We also preliminary tried to investigate the activity of our polyphenols by ANS fluorescence. However, ANS tests were not informative in the case of compounds derived from ferulic acid, since they showed major spectral interferences with ANS. On the other hand, also circular dichroism experiments were not useful, because of interference of 1% DMSO, which was needed to keep our compounds in solution. During our preliminary work [25,26], we have also used electron microscopy and NMR studies, which confirmed the thioflavin data. However, these long studies were carried out only on 8a or other few selected compounds, and are not suited for a fast selection of the here reported second generation compounds, that was made with thioflavin T. β-Amyloid aggregation is known to start when a change of the secondary structure from α-helix (in the membrane environment) or coil (in basic environment) to β-sheet conformation takes place [35].
In the presence of thioflavin-T, β-sheet formation results in a strong increase of fluorescence of the solution. Ideally, inhibitors of aggregation should decrease the maximum plateau of fluorescence obtained in the control experiment. Table 1 shows the relative decrease (or increase) of plateau achieved with compounds 8a,h, 9, 13a,b, 17a,b.
From the results collected we can draw some useful information. First, as long as we consider Aβ1-42, the lead compound 8a remains the best one. In particular, substitution of the t-Bu group of the isocyanide with a hydroxybenzyl group is highly deleterious, leading to a complete loss of the inhibitory effect (see entries 8 and 9). Only substitution with a hydroxyphenyl group is accepted (entry 6) although a decrease of activity is observed compared to the lead compound 8a. Also replacing of p-hydroxybenzaldehyde with an aliphatic counterpart is detrimental (entry 7).
For AβpE3-42, as already experienced in our previous work, even small differences seem to have a significant effect. The best compound, among those prepared in this second campaign, is 8c, which turned out to be slightly superior to our lead 8b. Moreover, this molecule is one of the few, together with 8b, that shows acceptable activity towards both proteins. Therefore, we think that it is worth of further investigation through in vivo assays. The difference, compared to 8b, is the amine component, which is a hydroxybenzyl instead of a hydroxyphenyl group. Also, for AβpE3-42, the compounds derived from iso-butyraldehyde did not inhibit aggregation. On the contrary, they seem, especially 8f, to even favor β-sheet formation.
Among the β-lactams, only the one containing the p-hydroxyphenyl group showed some activity (entry 14). It was equally active on both proteins and may be a good starting point for further refinement.
It was interesting to see if the different configuration of the new stereogenic center created during the Ugi MCR could influence activity. The results achieved with diastereomers 13a and 13b shows a negligible effect on Aβ1-42, but a significant difference in behavior for AβpE3-42. It is worth noting that 13b, which has just an additional methyl, displays a higher activity for the truncated peptide than 8a, but we do not know whether this is due to its enantiomeric purity or for the presence of the methyl group.
Finally, it should be noted that compound 9, lacking any phenolic group, had no inhibitory activity at all. On the contrary, it seems to favor aggregation of Aβ1-42. Furthermore, compounds 8f and 17a, that are just monophenols, behaved poorly, stressing the need for a polyphenolic system. The best candidates found so far (8a, 8b, 8c) are indeed either diphenols or triphenols. . Chemical shifts are reported in ppm (δ scale). Peak assignments were made with the aid of gCOSY and gHSQC experiments. In ABX system, the proton A is considered upfield and B downfield. IR spectra were recorded as solid, oil, or foamy samples, with the ATR (attenuated total reflectance) technique. TLC analyses were carried out on silica gel plates and viewed at UV (λ = 254 nm or 360 nm) and developed with Hanessian stain (dipping into a solution of (NH 4 ) 4 MoO 4 ·4H 2 O (21 g) and Ce(SO 4 ) 2 ·4H 2 O (1 g) in H 2 SO 4 (31 mL) and H 2 O (469 mL) and warming). R f values were measured after an elution of 7-9 cm. HRMS: samples were analyzed with a Synapt G2 QToF mass spectrometer (Waters, Milford, MA, USA). MS signals were acquired from 50 to 1200 m/z in either ESI positive or negative ionization mode. Column chromatography was done with the "flash" methodology by using 220-400 mesh silica. Petroleum ether (40-60 • C) is abbreviated as PE. All reactions employing dry solvents were carried out under nitrogen. After extractions, the aqueous phases were always re-extracted 2 times with the appropriate organic solvent, and the organic extracts were always dried over Na 2 SO 4 and filtered before evaporation to dryness.

General Information
Due to a tendency to partially degrade, the free phenols (8c-h, 13a,b, 18a,b) were fully characterized and stored in the acetylated form (7c-h, 12a,b, 17a,b) and then deprotected shortly before use through procedure B, checking the purity by 1 H-NMR and HPLC. HPLC analyses were carried out on a HP-1100 system (Agilent, Santa Clara, CA, USA) equipped with a Phenyl C6 reverse phase column (150 × 3 mm, 3 µm) at 25 • C with flow = 0.34 mL/min. Gradient from CH 3 CN/H 2 O 40:60 (time 0) to pure CH 3 CN (time 10). detection was done with a DAD detector at 330 nm. Compounds 2, 3 and 5 and 4-(allyloxy)phenyl isocyanide were prepared as previously described [25]. Known [36] 2-(allyloxy)benzylamine was prepared as previously described [32].  (100 mg, 160 µmol) is treated with such solution (4.8 mL, 920 µmol, 2 eq. for each acetyl group). After stirring for 2 h at rt, the solution is treated with dry Amberlyst ® 15 (4.7 mmol/g) (freshly thoroughly washed with dry methanol) so that the resulting pH is around 4-5 (about 200 mg of dry resin). The resin was filtered off, washing with methanol and the resulting filtrate evaporated to dryness, to afford 8c, pure enough for biochemical assays (80 mg, quantitative). The purity by HPLC (for conditions see the general remarks) was 99%. R

Thioflavin Experiments
One milliliter of DMSO was added to 1 mg of lyophilized synthetic peptide (Aβ1-42, AβpE3-42, AnaSpec, Fremont, CA, USA), to reach a final concentration of 1 mg mL −1 . Aliquots of 75 µL were lyophilized and stored at −20 • C until being used. For all experiments, stock peptides were reconstituted as reported [37]. For the preparation of the working samples, a stock solution of each peptide was divided into two or more aliquots. One was diluted to 5 µM in PBS containing 1% (v/v) DMSO to have a reference sample, and the others were diluted in PBS containing the appropriate quantity of polyphenol stock solution in DMSO in such a manner that each sample contains 1% of DMSO. The final pH was measured and eventually corrected to 7.4 using a few µL of 1 M HCl. Aβ peptides (5 µM) were incubated at 37 • C in the presence/absence of polyphenols as previously described and analyzed in parallel. ThT fluorescence was followed in time during aggregation. For this purpose, 47.5 µL of Aβ with and without test compounds were mixed with 2.5 µL ThT (400 µM) in a 3 mm path length fluorescence cuvette. ThT fluorescence was measured by using a luminescence spectrometer (LS50B, PerkinElmer, Waltham, MA, USA) at excitation and emission wavelengths of 440 nm (slit width = 5 nm) and 482 nm (slit width = 10 nm), respectively. ThT fluorescence data were plotted as a function of time and fitted by a sigmoidal curve described by the following equation: [38] where y i and y f are the initial and final ThT fluorescence, respectively and k fib is the fibril growing rate, t is time and t 0 is the time to 50% of maximal fluorescence. The lag time (T) is derived as t 0 -2/k fib .

Conclusions
In conclusion, the present study has demonstrated that replacement of the various pharmacophores in our previous leads may be critical, only small changes being permitted. A new possible hit has been selected from the second-generation library produced during this work. From the synthetic point of view, the feasibility of the overall strategy was again demonstrated, as well as the possibility to use the Ugi reaction with glycolaldehyde dimer for the synthesis of β-lactam-polyphenol hybrids. Due to the wide range of biological activities of polyphenols, we plan to investigate the many compounds synthesized through this strategy on other biological targets or in phenotypic assays. Studies towards this goal are in progress. Funding: This study is supported by Fondazione Cariplo, under the "Integrated Biotechnology and Bioeconomy" programme.