Retro-1-Oligonucleotide Conjugates. Synthesis and Biological Evaluation

Addition of small molecule Retro-1 has been described to enhance antisense and splice switching oligonucleotides. With the aim of assessing the effect of covalently linking Retro-1 to the biologically active oligonucleotide, three different derivatives of Retro-1 were prepared that incorporated a phosphoramidite group, a thiol or a 1,3-diene, respectively. Retro-1–oligonucleotide conjugates were assembled both on-resin (coupling of the phosphoramidite) and from reactions in solution (Michael-type thiol-maleimide reaction and Diels-Alder cycloaddition). Splice switching assays with the resulting conjugates showed that they were active but that they provided little advantage over the unconjugated oligonucleotide in the well-known HeLa Luc705 reporter system.


Introduction
Forty years after the therapeutic potential of synthetic oligonucleotides was first perceived [1], delivery is the one problem that remains not well solved [2,3]. Oligonucleotides can easily be modified to favor hybridization and ameliorate their bioavailability [4,5], but there is no gold standard methodology for effective and complete internalization. Most forms of oligonucleotides are thought to be taken up by endocytosis and accumulate in intracellular endomembrane compartments [6,7]. Only a tiny fraction of the accumulated oligonucleotide spontaneously leaks from the endosomes, although this is sometimes enough to provide a pharmacological effect. There have been many efforts to increase both the cellular uptake and intracellular release of oligonucleotides including various targeting ligands and endomembrane destabilizing polymers or nanoparticles [2,8].
A few years ago, Juliano and cols. described that the compound named Retro-1 enhanced the pharmaceutical potency of antisense and siRNA oligonucleotides [9]. Retro-1 is a member of a group of compounds able to reduce the action of bacterial toxins [10] by interfering with their intracellular trafficking. Subsequently, high throughput screening has allowed identification of additional molecules capable of enhancing the effectiveness of synthetic oligonucleotides [11][12][13]. Some of the hits developed in these studies have been modified to assess their effect on oligonucleotide delivery [14,15]. However, to our knowledge the impact of covalently linking one of these molecules to the oligonucleotide chain has not been evaluated.
Retro-1, the first compound identified to ameliorate both antisense and siRNA oligonucleotides effect, was chosen for this evaluation. In this manuscript we wish to describe the preparation of three [14,15]. However, to our knowledge the impact of covalently linking one of these molecules to the oligonucleotide chain has not been evaluated.
Retro-1, the first compound identified to ameliorate both antisense and siRNA oligonucleotides effect, was chosen for this evaluation. In this manuscript we wish to describe the preparation of three Retro-1 derivatives, their conjugation reactions to a splice switching oligonucleotide, and the results of the biological evaluation. These studies indicate that the efficacy of covalently linking the two moieties is inferior to that achieved by administering the two separate molecules.

Modification of Retro-1 for Conjugation
One alternative to prepare oligonucleotide conjugates, that is, to covalently link oligonucleotides to other molecules, involves the use of a phosphoramidite derivative that can be incorporated into the chain using standard oligonucleotide elongation methodologies. Another possibility is making use of a click reaction [16] that chemoselectively links the two components of the conjugate. Irrespective of the synthesis approach, Retro-1 had to be derivatized so as to incorporate an additional functional group.
The structure of Retro-1 (6), which is shown on Scheme 1, has many points in common with that of benzodiazepine psychoactive drugs. It is also composed of two fused rings, where two atoms of the benzene ring and carbons 6-7 of the diazepine function as bridgehead, and the 1,4-diazepine ring contains a carbonyl group at the 2 position and an aromatic ring appending from carbon 5. The main difference is that atoms 4 and 5 of the diazepine are not linked by a double bond but by a single one. As a result, carbon 5 is a stereocenter.
The synthesis of racemic Retro-1 has been described [17], as well as the separation of the two enantiomers using a chiral HPLC column (conformers, which are observed at the NMR spectra, could not be separated). Both enantiomers were shown to be biologically active, and the difference between their EC50 values was small (the S isomer was a bit more active than the R one, but both were within the same 1-10 μM range as the racemic mixture). Therefore, we have worked with the racemic mixture.
The only information available on the biological activity of Retro-1 analogs [14] indicates that compounds with no substituent at the 4 position (see Scheme 1) favor oligonucleotide activity, whilst N-alkylation seems best suited for reducing bacterial toxins effect. Since Retro-1, in which N-4 is acylated, exhibits both effects, we decided to prepare two N-4 acylated analogs (Scheme 2) and an additional one modified at the 3 position (Scheme 3).
As shown in Scheme 1, the diazepine ring is made from three building blocks: the aminobenzophenone, the acylating reagent and ammonia. Acylation of the aromatic amine links the two main components and imine formation closes the cycle. Scheme 1. Synthesis scheme described [17] for the preparation of Retro-1, 6. DCM = dichloromethane; NBS = N-bromosuccinimide; TEA = triethylamine. The wavy bond indicates that the stereochemistry is not defined (in other words, the compound is a mixture of isomers).
Based on this scheme, which we could perfectly reproduce, changing the reagents of the two acylation reactions appeared to be a simple option to adapt Retro-1 for conjugation. Replacement of bromoacetyl bromide with an activated amino acid would modify position 3 of the diazepine ring, and use of a suitably protected trifunctional amino acid would incorporate an additional group suitable for further derivatization. Alternatively, substitution of propionyl chloride with another acylation reagent would allow position 4 to be modified.
We decided that the three Retro-1 analogs would be differently derivatized for conjugation. One would include a phosphitylatable hydroxyl group, and could be attached to the 5' end of a resin-linked chain. The other two would contain functional groups suitable for click conjugations in solution. For this purpose we chose a thiol and a 1,3-diene, which were intended to react with a maleimide moiety. Maleimido-oligonucleotides can be easily assembled on a solid support making use of the appropriate maleimide protection [18].

Retro-1 Derivatives Modified at the 4 Position
Two of the three analogs were synthesized from the amine precursor of Retro-1 5 (see Scheme 1). Replacement of propionyl chloride with acryloyl chloride gave 7 (Scheme 2a), which could be easily modified by means of Michael-type reactions.
Reaction between 7 and 4-hydroxypiperidine afforded 8, from which phosphoramidite 9 was prepared by reaction with a chlorophosphine and a base (Scheme 2b). The other aza-Michael reaction was carried out with S-trityl cysteamine, which furnished 10, and removal of the trityl group under acidic conditions gave 11 (Scheme 2c). To minimize the extent of thiol oxidation to disulfide the thiol-containing Retro-1 analog was kept protected, and thiol deprotection was carried out not long before the conjugation reaction.
Molecules 2018, 23, x 3 of 17 Scheme 1. Synthesis scheme described [17] for the preparation of Retro-1, 6. DCM = dichloromethane; NBS = N-bromosuccinimide; TEA = triethylamine. The wavy bond indicates that the stereochemistry is not defined (in other words, the compound is a mixture of isomers).
Based on this scheme, which we could perfectly reproduce, changing the reagents of the two acylation reactions appeared to be a simple option to adapt Retro-1 for conjugation. Replacement of bromoacetyl bromide with an activated amino acid would modify position 3 of the diazepine ring, and use of a suitably protected trifunctional amino acid would incorporate an additional group suitable for further derivatization. Alternatively, substitution of propionyl chloride with another acylation reagent would allow position 4 to be modified.
We decided that the three Retro-1 analogs would be differently derivatized for conjugation. One would include a phosphitylatable hydroxyl group, and could be attached to the 5' end of a resinlinked chain. The other two would contain functional groups suitable for click conjugations in solution. For this purpose we chose a thiol and a 1,3-diene, which were intended to react with a maleimide moiety. Maleimido-oligonucleotides can be easily assembled on a solid support making use of the appropriate maleimide protection [18].
2.1.1. Retro-1 derivatives modified at the 4 position Two of the three analogs were synthesized from the amine precursor of Retro-1 5 (see Scheme 1). Replacement of propionyl chloride with acryloyl chloride gave 7 (Scheme 2, a), which could be easily modified by means of Michael-type reactions.
Reaction between 7 and 4-hydroxypiperidine afforded 8, from which phosphoramidite 9 was prepared by reaction with a chlorophosphine and a base (Scheme 2, b). The other aza-Michael reaction was carried out with S-trityl cysteamine, which furnished 10, and removal of the trityl group under acidic conditions gave 11 (Scheme 2, c). To minimize the extent of thiol oxidation to disulfide the thiol-containing Retro-1 analog was kept protected, and thiol deprotection was carried out not long before the conjugation reaction. Scheme 2. Synthesis of the Retro-1 analogs modified at position 4 of the benzodiazepine ring. The first step (a) afforded the common precursor, which was subsequently modified to obtain either phosphoramidite derivative 9 (b) or the thiol-modified compound 11 (c). CNE = 2-cyanoethyl; DCM = dichloromethane; DIPEA = N,N-diisopropylethylamine; rt = Scheme 2. Synthesis of the Retro-1 analogs modified at position 4 of the benzodiazepine ring. The first step (a) afforded the common precursor, which was subsequently modified to obtain either phosphoramidite derivative 9 (b) or the thiol-modified compound 11 (c). CNE = 2-cyanoethyl; DCM = dichloromethane; DIPEA = N,N-diisopropylethylamine; rt = room temperature; TEA = triethylamine; TFA = trifluoroacetic acid; TIS = triisopropylsilane; Trt = trityl. The wavy bond indicates a not defined stereochemistry. The starting material for the preparation of the third Retro-1 derivative was the first intermediate in the synthesis of Retro-1, 2 (see Scheme 1). Acylation of the amine with a suitably protected lysine derivative was found not to be straightforward. Attempts to activate the carboxyl group with either a combination of a carbodiimide and 1-hydroxybenzotriazole, 1-[(1-cyano-2-ethoxy-2-oxoethylideneaminooxy)-dimethylaminomorpholinomethylene)] methanaminium hexafluorophosphate (COMU), or mesitylenesulfonyl 3-nitro-1,2,4-triazole (MSNT) in the presence of N-methylimidazole were unsuccessful. Gratifyingly, the mixed carboxylic carbonic anhydride methodology did work (Scheme 3). Carboxyl group activation with isobutyl chloroformate in the presence of N-methylmorpholine afforded 12, and overnight treatment of 12 with 7 M ammonia in methanol quantitatively removed the Fmoc group and promoted cyclization to 13. These two steps were also carried out without isolating 12, and 13 was obtained in a similar yield.

Modification of the 3 Position of Retro-1
The starting material for the preparation of the third Retro-1 derivative was the first intermediate in the synthesis of Retro-1, 2 (see Scheme 1). Acylation of the amine with a suitably protected lysine derivative was found not to be straightforward. Attempts to activate the carboxyl group with either a combination of a carbodiimide and 1-hydroxybenzotriazole, 1- methan-aminium hexafluorophosphate (COMU), or mesitylenesulfonyl 3-nitro-1,2,4-triazole (MSNT) in the presence of N-methylimidazole were unsuccessful. Gratifyingly, the mixed carboxylic carbonic anhydride methodology did work (Scheme 3). Carboxyl group activation with isobutyl chloroformate in the presence of N-methylmorpholine afforded 12, and overnight treatment of 12 with 7 M ammonia in methanol quantitatively removed the Fmoc group and promoted cyclization to 13. These two steps were also carried out without isolating 12, and 13 was obtained in a similar yield. Diene-derivatized 16 was finally obtained after the following series of steps: i) Reduction of the imine of compound 13 with NaBH3CN as previously reported, which gave 14; ii) Removal of the Boc group with an acidic treatment, which yielded 15; iii) Carbodiimide-promoted acylation of the primary amine with (4E)-4,6-heptadienoic acid. We were initially concerned with the possible acylation of the secondary amine (N-4 of the diazepine ring), but we found that reaction took place exclusively on the lysine ε-amine, furnishing 16. In fact, all attempts carried out to acylate N-4 failed. Therefore, derivative 16 differed from the other Retro-1 analogs in the presence of a substituent appending from carbon 3 and no acyl group on nitrogen 4. Diene-derivatized 16 was finally obtained after the following series of steps: (i) Reduction of the imine of compound 13 with NaBH 3 CN as previously reported, which gave 14; (ii) Removal of the Boc group with an acidic treatment, which yielded 15; (iii) Carbodiimide-promoted acylation of the primary amine with (4E)-4,6-heptadienoic acid. We were initially concerned with the possible acylation of the secondary amine (N-4 of the diazepine ring), but we found that reaction took place exclusively on the lysine ε-amine, furnishing 16. In fact, all attempts carried out to acylate N-4 failed. Therefore, derivative 16 differed from the other Retro-1 analogs in the presence of a substituent appending from carbon 3 and no acyl group on nitrogen 4.

Preparation of the Retro-1 Conjugates
The three Retro-1 derivatives (9,11,16) were linked to oligonucleotide r5'GTTATTCTTTAGAATGGTGC3' (all 2'-O-methyl and phosphorothioate, T = ribothymidine). This sequence is complementary to that of a mutated intron from thalassemic hemoglobin that is introduced into the firefly luciferase gene used for the splice switching experiments [19]. Compounds 9, 11 and 16 were also linked to a control oligonucleotide with a scrambled sequence, namely r5'TGTGTACTGATGTAGTTATC3' (all 2'-O-methyl and phosphorothioate; T = ribothymidine).
introduced into the firefly luciferase gene used for the splice switching experiments [19]. Compounds 9, 11 and 16 were also linked to a control oligonucleotide with a scrambled sequence, namely r5'TGTGTACTGATGTAGTTATC3' (all 2'-O-methyl and phosphorothioate; T = ribothymidine).
All conjugates were purified by reversed phase HPLC and characterized by MALDI-TOF MS. All conjugates were purified by reversed phase HPLC and characterized by MALDI-TOF MS.

Splice-Switching Assays
Oligonucleotides conjugated to Retro-1 derivatives were examined for splice correction activity. The experiments utilized the HeLa Luc 705 cell line [19], as described in Materials and Methods. Oligonucleotides having the known splice correcting sequence r5'GTTATTCTTTAGAATGGTGC3' ' , previously termed SSO623 [19], were used in conjugated or unconjugated form and were compared to control oligonucleotides with the inactive sequence r5'TGTGTACTGATGTAGTTATC3'. In Figure  1 oligonucleotide 19a, which is the immediate conjugate of Retro-1, was compared to its control conjugate 19b. Furthermore, we examined unmodified SSO623, as well as SSO623 followed by treatment with the small molecule UNC7938 as a positive control. As seen in the figure, oligonucleotide 19a produced a dose-dependent luciferase induction while the control oligonucleotide 19b did not. However, SSO623 itself also gave rise to a progressive induction of luciferase that was somewhat greater than that produced by 19a. The dual use of SSO623 and UNC7938 provided the largest degree of splice correction and luciferase induction as expected. This set of experiments demonstrated that coupling of the Retro moiety did not prevent the splice correcting activity of the active oligonucleotide. However, conjugate 19a did not provide an advantage, in terms of potency or efficacy, over SSO623 itself. We also examined two additional conjugates 22a and 23a. However, neither of these conjugates displayed an advantage over SSO623 in the HeLa Luc 705 induction assay.

Splice-Switching Assays
Oligonucleotides conjugated to Retro-1 derivatives were examined for splice correction activity. The experiments utilized the HeLa Luc 705 cell line [19], as described in Materials and Methods. Oligonucleotides having the known splice correcting sequence r5'GTTATTCTTTAGAATGGTGC3', previously termed SSO623 [19], were used in conjugated or unconjugated form and were compared to control oligonucleotides with the inactive sequence r5'TGTGTACTGATGTAGTTATC3'. In Figure 1 oligonucleotide 19a, which is the immediate conjugate of Retro-1, was compared to its control conjugate 19b. Furthermore, we examined unmodified SSO623, as well as SSO623 followed by treatment with the small molecule UNC7938 as a positive control. As seen in the figure, oligonucleotide 19a produced a dose-dependent luciferase induction while the control oligonucleotide 19b did not. However, SSO623 itself also gave rise to a progressive induction of luciferase that was somewhat greater than that produced by 19a. The dual use of SSO623 and UNC7938 provided the largest degree of splice correction and luciferase induction as expected. This set of experiments demonstrated that coupling of the Retro moiety did not prevent the splice correcting activity of the active oligonucleotide. However, conjugate 19a did not provide an advantage, in terms of potency or efficacy, over SSO623 itself. We also examined two additional conjugates 22a and 23a. However, neither of these conjugates displayed an advantage over SSO623 in the HeLa Luc 705 induction assay. In summary, conjugation of the Retro moiety to a splice switching oligonucleotide did not provide a major enhancement of splice correction activity in the widely used HeLa Luc705 reporter system. At this point it is unclear why the conjugates displayed slightly lower activity than the unmodified oligonucleotide. One could hypothesize that the presence of the bulky Retro group might affect either cell uptake of the oligo or might affect the interaction of the oligo with the splicing machinery. However, detailed investigation of these possibilities, especially the splicing aspect, would involve a very substantial amount of new biological investigation and is beyond the scope of this work.

Materials for Solution Organic Synthesis
(E)-Hepta-4,6-dienoic acid was prepared as described by Baillie et al. [20], and S-trityl cysteamine as reported by Naumiec et al. [21]. Fmoc-L-Lys(Boc)-OH was from Novabiochem (Merck, Spain); all of the other chemicals were either from Sigma-Aldrich (Millipore, Spain) or Across Organics (Millipore, Spain) (7 M ammonia in methanol), and were used without further purification. Water was obtained from a MilliQ system (Millipore, Spain).

Analysis and Characterization Techniques for Small Organic Molecules (2-16)
TLC was carried out on silica gel plates 60 F254 from Merck (Millipore, Spain). IR spectra were recorded in a Nicolet 6700 FT-IR spectrometer (Thermo Scientific, Millipore, Spain). 1 H and 13 C NMR spectra were recorded on either Varian Mercury 400 MHz or Brucker 400 MHz spectrometers (reference: TMS or residual solvent signals). HR-ESI mass spectra were obtained using an LC/MSD-TOF instrument (Agilent Technologies, Millipore, Spain). HPLC/MS analyses were recorded in an Alliance Waters 2690 separation module with a Waters micromass ZQ4000 MS detector (Waters, Millipore, Spain). Aqueous solutions were lyophilized in either Labconco (Vertex, Millipore, Spain) or Christ freeze dryers (Inycom, Millipore, Spain). In summary, conjugation of the Retro moiety to a splice switching oligonucleotide did not provide a major enhancement of splice correction activity in the widely used HeLa Luc705 reporter system. At this point it is unclear why the conjugates displayed slightly lower activity than the unmodified oligonucleotide. One could hypothesize that the presence of the bulky Retro group might affect either cell uptake of the oligo or might affect the interaction of the oligo with the splicing machinery. However, detailed investigation of these possibilities, especially the splicing aspect, would involve a very substantial amount of new biological investigation and is beyond the scope of this work.

Materials for Solution Organic Synthesis
(E)-Hepta-4,6-dienoic acid was prepared as described by Baillie et al. [20], and S-trityl cysteamine as reported by Naumiec et al. [21]. Fmoc-L-Lys(Boc)-OH was from Novabiochem (Zaragoza, Spain); all of the other chemicals were either from Sigma-Aldrich (Zaragoza, Spain) or Across Organics (Zaragoza, Spain) (7 M ammonia in methanol), and were used without further purification. Water was obtained from a MilliQ system (Zaragoza, Spain). (2-16) TLC was carried out on silica gel plates 60 F254 from Merck (Zaragoza, Spain). IR spectra were recorded in a Nicolet 6700 FT-IR spectrometer (Thermo Scientific, Zaragoza, Spain). 1 H and 13 C NMR spectra were recorded on either Varian Mercury 400 MHz or Brucker 400 MHz spectrometers (reference: TMS or residual solvent signals). HR-ESI mass spectra were obtained using an LC/MSD-TOF Instrument (Agilent Technologies, Zaragoza, Spain). HPLC/MS analyses were recorded in an Alliance Waters 2690 separation module with a Waters micromass ZQ4000 MS detector (Waters, Zaragoza, Spain). Aqueous solutions were lyophilized in either Labconco (Vertex, Zaragoza, Spain) or Christ freeze dryers (Inycom, Zaragoza, Spain).
Final deprotection conditions: conc. aq. ammonia, 3 h, room temperature. After this treatment the resin was filtered and washed with water (3×), the filtrates pooled and concentrated in a SpeedVac apparatus (Inycom, Zaragoza, Spain) (removal of ammonia), and the resulting crude lyophilized.

RP-HPLC Analysis and Purification of Oligonucleotides and Conjugates
Reversed-phase HPLC analysis and purification was performed using a Shimadzu system (Izasa, Zaragoza, Spain). Linear gradients were always used.

Tert-Butyl (S)-[4-(7-bromo-2-oxo-5-phenyl-2,3-dihydro-1H-benzo[e][1,4]diazepin-3-yl)butyl]carbamate
One-pot synthesis of compound 13 from 2. A solution of Fmoc-L-Lys(Boc)-OH (1.87 g, 4.00 mmol) in anh. DCM (15 mL) was cooled in an ice bath. N-Methylmorpholine (1.37 mL, 9.09 mmol) and isobutyl chloroformate (0.81 mL, 3.64 mmol) were added. After 10 min, a solution of 2 (1.00 g, 3.64 mmol) in anh. DCM (2 mL) was added, and the mixture was stirred for 30 min at 5 • C and 5.5 h at room temperature. Afterwards, the solvent was removed under vacuum and the crude dissolved in 7 M NH 3 (in MeOH, 40 mL), and the solution was left to react overnight at room temperature. Subsequently, the solvent was removed under vacuum and the resulting crude dissolved in EtOAc (100 mL) and washed with 10% aq. HCl (3 × 30 mL). The organic phase was dried over anh. MgSO 4 , filtered and the solvent evaporated under vacuum. The crude material was further purified by silica gel flash column chromatography eluting with a 70:30 hexanes/EtOAc mixture. The title compound (13) (14). To a solution of 13 (800 mg, 1.65 mmol) and NaBH 3 CN (155.4 mg, 2.48 mmol) in MeOH (10 mL), AcOH (500 µL, 8.24 mmol) was added, and the mixture was stirred at room temperature for 2.5 h until complete reduction of the imine as assessed by TLC. Afterwards, the solvent was removed under low pressure, the crude was dissolved in EtOAc (50 mL) and washed with aq. NaHCO 3(sat) (2 × 20 mL). The organic layer was dried over anh. MgSO 4 , filtered and the solvent removed under low pressure. The title compound (14) was obtained as a pale yellow solid (799 mg, 99%).     2905, 2355, 1730, 1658, 1599, 1571, 1497, 1425,  1437, 1391, 1254, 1248, 1173, 1158, 1071, 1043, 757 63 (m, 1H)        Conjugates 19. Oligonucleotide-resins were automatically assembled at the 1 μmol-scale as stated above (Materials and Methods section). After oligonucleotide elongation was completed and the 5' end deprotected, phosphoramidite 9 was incorporated (0.1 M solution in anh. acetonitrile, 2 × 10 min coupling) and the resulting phosphite triester sulfurized using the same procedure as in all the other synthesis cycles, which afforded 18. Final deprotection was carried out under standard conditions (see above).     Conjugates 19. Oligonucleotide-resins were automatically assembled at the 1 μmol-scale as stated above (Materials and Methods section). After oligonucleotide elongation was completed and the 5' end deprotected, phosphoramidite 9 was incorporated (0.1 M solution in anh. acetonitrile, 2 × 10 min coupling) and the resulting phosphite triester sulfurized using the same procedure as in all the other synthesis cycles, which afforded 18. Final deprotection was carried out under standard conditions (see above).   Conjugates 19. Oligonucleotide-resins were automatically assembled at the 1 µmol-scale as stated above (Materials and Methods section). After oligonucleotide elongation was completed and the 5' end deprotected, phosphoramidite 9 was incorporated (0.1 M solution in anh. acetonitrile, 2 × 10 min coupling) and the resulting phosphite triester sulfurized using the same procedure as in all the other synthesis cycles, which afforded 18. Final deprotection was carried out under standard conditions (see above).   Conjugates 19. Oligonucleotide-resins were automatically assembled at the 1 μmol-scale as stated above (Materials and Methods section). After oligonucleotide elongation was completed and the 5' end deprotected, phosphoramidite 9 was incorporated (0.1 M solution in anh. acetonitrile, 2 × 10 min coupling) and the resulting phosphite triester sulfurized using the same procedure as in all the other synthesis cycles, which afforded 18. Final deprotection was carried out under standard conditions (see above).