Synthesis of 5’-fluorophosphate-modified short-interfering RNAs

Abstract We have developed an improved scheme for the synthesis of a mono-fluorinated phosphoramidite for the 5’-modification of nucleic acids using standard phosphoramidite chemistry. We describe the first report of a phosphofluoridate modified siRNA strand and have evaluated C18 HPLC for purification of modified strands from unreacted siRNA strands. Lastly, the biological activity of the high purity siRNA strands, when placed on the sense and/or antisense strand, was evaluated to assess the impact of 5’ phosphofluoridate modifications on siRNA activity.


Introduction
Chemically modified nucleotides, alongside effective delivery platforms, have revolutionized the field of gene-and immuno-therapies. Without the aid of modified nucleotides, the groundbreaking antisense oligonucleotide (ASO; e.g., Spinraza), short interfering RNA (siRNA; e.g., Onpattro), and mRNA (Moderna and Pfizer's COVID-19 vaccines) drug delivery platforms would not be possible. We have previously developed a variety of chemical modifications for siRNAs that add unique functionalities to improve cellular uptake, targeted cell delivery, strand selection, photodynamic therapy, and reversible photocontrol of siRNA activity. [1][2][3][4][5][6][7][8][9] In this paper, we synthesize a fluorophosphoramidite and use this functionality to synthesize and evaluate the impact of a 5′-fluorophosphate group on strand selection and gene-silencing in siRNA.
Short interfering RNA (siRNA) are short, double stranded RNA that have found extensive use in research and, to date, have seen the US FDA approval of four therapeutics (Patisiran, Givosiran, Lumasiran, and Inclisiran). Once within the cytoplasm of a cell, one of the siRNA strands is preferentially incorporated within the RNA guided nuclease, Ago2, to form the RNA induced silencing complex (RISC). This captured strand then guides the RISC to target and cleave complementary strands, resulting in potent gene silencing. As unmodified RNA is highly unstable within the cell and can lead to undesirable immunogenicity, the success of siRNA therapeutics has been made possible through the incorporation of nucleotide analogues, namely the phosphorothioate, 2′-O-methyl, and 2′-fluoro. These impart nuclease resistance and reduce innate immune responses, ultimately leading to improved potency, duration of effect, and safety. [10] Beyond these, several other modifications have been developed to improve cell delivery, impart functionalities, and/or abolish sense strand activity. [11][12][13][14][15] Sense strand activity is an important safety concern for siRNA therapeutics as both the antisense and sense strand carry the potential for off-target effects. It has been shown that the 5′-end of the antisense strand binds within a deep pocket of Ago2, providing an opportunity to explore 5′-modifications for their impact on strand activity. For example, recent developments, such as the incorporation of a 5′-(E)-vinylphosphonate (5′-VP) at the 5′-end of the antisense strand has led to potent improvements in strand selection activity. [16] The synthesis of phosphorofluoridate precursors and dinucleotides have previously been explored; [17][18][19][20][21][22] however, to the best of our knowledge, 5′ terminal phosphofluoridate modified oligonucleotides ( Figure 1) have not yet been characterized. Therefore, we investigated a strategy for the synthesis, purification, and characterization of a 5′-phosphorofluoridate modified RNA. To achieve this, we evaluated and modified synthetic schemes for the synthesis of a fluorinated phosphoramidite based on the substitution of 4-nitrophenol by ionic fluorine, as developed by Dąbkowski and collea gues. [19,20]

General methods
Solvents were purchased from Sigma-Aldrich. Tetrahydrofuran and triethylamine were dried on a Pure-Solv 400 solvent purification system (Innovative Technology (China) Ltd.). Column chromatography was performed using Silicycle Siliaflash P60. NMR spectra were obtained on a Bruker Avance III HD 500 MHz NMR Spectrometer ( 1 H: 400 MHz, 31 P: 162 MHz, 13 C: 101 MHz,  and 19 F: 377 MHz). Spectrum analysis was performed using ACD/NMR Processor (Advanced Chemistry Development, Inc.). LC-QTOF was performed on an Agilent 6545 QTOF-MS after separation on a Zorbax Eclipse Plus C18 1 × 100 mm 1.8-Micron Agilent column with an Agilent 1260 Infinity Binary Pump. The mobile phase was 0.6 ml/min of 5% ACN in 5 mM ammonium acetate pH 7 buffer. Tuning was done in negative mode, mass range 3200 m/z, extended dynamic range 2 GHz, high resolution mode. Sample concentration was 0.01 O.D/µl with injection volume of 20 µl. Data was analyzed using Agilent Technologies MassHunter Workstation Qualitative Analysis Software (Qual. 10.0).
For synthesis of phosphorofluoridate modified RNA, manual coupling was performed by preparing 250 µl of 100 mM 3 in anhydrous ACN under argon gas in a flame dried vial. The phosphoramidite was activated using 250 µl of 250 mM 5-ETT in anhydrous ACN, quickly mixed and loaded into a 1 ml syringe. With an empty syringe inserted into the opposite end of the solid support column, the 500 µl activated phosphoramidite solution was syringed through the column several times over a period of 10 minutes. The column was then washed with anhydrous ACN. The coupling and wash process was repeated twice more to achieve highest possible conversion. To oxidize the 5′ phosphorofluoridite to a phosphorofluoridate, 500 µl of oxidation solution was applied to the column for 60 seconds then washed with ACN.
Cleavage of oligonucleotides from their solid supports was performed using 1 ml of 1:1 40% w/v methylamine in H 2 O and 33% w/v methylamine in ethanol for 1 hr at room temperature. This solution was transferred to a 1.5 ml microcentrifuge tube and incubated overnight at room temperature to remove cyanoethyl protecting groups. Solvent was evaporated using a Genevac miVac Centrifugal Concentrator (Fisher Scientific) evaporator and resuspended in a mixture of 100 µl DMSO and 125 µl triethylamine trihydrofluoride and incubated for 3 hours at 65 °C to remove the 2′-O-TBDMS protecting groups. Triethylamine trihydrofluoride was removed using a miVac evaporator overnight. Oligonucleotides were then precipitated with −70 °C 125 mM sodium acetate in EtOH and washed twice with −70 °C ethanol. Desalting of 21mers was performed thrice using Millipore Amicon MWCO 3000 spin columns.

Purification and characterization of siRNA
Oligonucleotides were HPLC purified using a Vydac 218MS 5 μm C18 4.6 mm x 150 mm reverse phase column (HiChrom) on a Waters 1525 binary HPLC pump with a Waters 2489 UV/Vis detector controlled by Empower 3 software. Conditions were 5 to 20% ACN gradient in 0.1 M aqueous triethylammonium acetate pH 7 over 30 minutes. Peaks were collected, dried on a miVac evaporator, suspended in ultrapure H 2 O, and evaluated via electrospray ionization tandem mass spectrophotometry. High purity antisense and sense strand samples with the desired product mass were used for siRNA annealing. These were 6771.92 g/mol (measured 6772.02 g/mol) for the antisense and 6685.86 g/mol (measured 6685.90 g/ mol) for the sense strand. Equimolar quantities of RNAs were annealed by heating RNA to 95 °C for 2 minutes in binding buffer (75.0 mM KCl, 50.0 mM Tris-HCl, 3.00 mM MgCl 2 , pH 8.30) and slowly cooling the samples to 35 °C. For circular dichroism (CD) analysis, 1.0 OD of siRNA was annealed in sodium phosphate buffer (90.0 mM NaCl, 10.0 mM Na 2 HPO 4 , 1.00 mM EDTA, pH 7.00). CD spectroscopy was performed on a Jasco J-815 CD at 25 °C with a screening rate of 20.0 nm/min and a 0.20 nm data pitch. Only samples with a spectrum that corresponds to A-form helixes were used for further analysis.

Preparation of 2-cyanoethyl (4-nitrophenyl) diisopropyl phosphoramidite (2)
In a flame dried 100 ml round bottom flask, a stirred mixture of 2.5 mmol 3-((bis(diisopropylamino)phosphaneyl)oxy)propanenitrile 1 (Toronto Research Chemicals) in 25 ml dry THF was treated with 1.05 equivalents (2.6 mmol) of 4-nitrophenol under argon gas. Upon addition of 4-nitrophenol, the mixture turned yellow, and the production of a soluble black side product slowly formed. The reaction was monitored by TLC (dichloromethane) using UV light and KMnO 4 staining. Once the starting material was consumed (30 minutes), 1 ml of dry triethylamine was added to quench the reaction and dried on a rotary evaporator at 30 °C to yield a yellow oil. The oil was purified by silica gel chromatography, using a mobile phase of neat dichloromethane with a product Rf of 0.7. Solvent removal afforded the colorless oil 2-cyanoethyl (4-nitrophenyl) diisopropylphosphoramidite 2 in 63% yield. Rapid purification during gel chromatography is essential, as the acidic silica degrades the product and the use of triethylamine in the mobile phase is not possible due to co-elution of other compounds. Nuclear magnetic resonance (NMR) spectra of the product agree with previous reports. 1

Preparation of 3-(((diisopropylamino)fluorophosphaneyl)oxy) propanenitrile (3)
In a flame dried 50 ml round bottom flask, a stirred mixture of 1.85 mmol 2-cyanoethyl (4-nitrophenyl) diisopropylphosphoramidite 2 in 10 ml dry THF was treated with 1 equivalent of 1.0 M TBAF in THF (1.85 ml; 1.85 mmol) under argon gas. The solution rapidly turned rich yellow with the removal of 4-nitrophenol, and after several minutes, the tetra-n-butylammonium para-nitrophenolate salt crashed out of solution. The reaction was monitored by TLC (dichloromethane) under UV light and staining with KMnO 4 . Once the starting material was consumed (10 minutes), the solution was filtered, rinsed with dry THF, and dried on a rotary evaporator at 30 °C. The residue was purified by silica gel chromatography, with a mobile phase of neat dichloromethane with a product Rf 0.70 to afford the clear oil 3-(((diisopropylamino) fluorophosphaneyl)-oxy)propanenitrile 3 in 80% yield. Rapid purification during gel chromatography is essential. Nuclear magnetic resonance (NMR) spectra agree with previous reports. 1

Preparation of bis(4-nitrophenyl) diisopropylphosphoramidite (5)
In a flame dried 100 ml round bottom flask, a mixture of 6 equivalents 4-nitrophenol (24 mmol) and 10 equivalents of triethylamine (5.57 ml) was added to 25 ml THF under argon gas. In a separate flame dried 20 ml vial, a mixture of 4 mmol 1,1-dichloro-N,N-diisopropylphosphanamine 4 (TCI Chemicals) in 10 ml dry THF was prepared. This solution was slowly added to the 4-nitrophenol/TEA solution over 60 seconds (highly reactive). The reaction was monitored by TLC (45:5 acetone:ethyl acetate) under UV light and stained with KMnO4. Once the starting material was consumed (within 2.5 h), the solution was filtered to remove the triethylammonium hydrochloride salt and dried on a rotary evaporator at 30 °C. The residue was purified by silica gel chromatography with a mobile phase of 2:3 hexanes:dichloromethane and product Rf of 0.6. Solvent removal afforded white crystals at 73% yield. Rapid purification during gel chromatography is essential. Proton and phosphorous NMR agreed with previous reports. 1

Attempted preparation of 1,1-difluoro-N,N-diisopropyl phosphanamine (6)
This product was not successfully isolated due to poor stability. Several attempts to make this product indicated a successful reaction, and NMR spectra of crude reaction mixtures suggest the presence of 6 (Appendix C4); however, any attempts at purification were unsuccessful. The following describes reaction conditions that afforded a crude product.
Prepare 2.2 equivalents of 1 M TBAF in THF by stirring 3 Å molecular sieves under argon gas for 1 hour. A stirred mixture of 0 °C 150 nmol bis(4-nitrophenyl) diisopropyl phosphoramidite 5 in 2.5 ml dry THF was prepared in a 25 ml flame dried round bottom flask. To this mixture, the dry TBAF was added dropwise over 60 seconds. After 10 minutes, the reaction was filtered and dried on a rotary evaporator. Quantitative proton and carbon NMR spectra suggested a low relative abundance of 6 compared to the tetrabutylammonium salt suggesting low product conversion. No aromatic protons were identified, meaning that filtration effectively removed 4-nitrophenol, but several phosphorous shifts were identified. Fluorine NMR did not identify a significant peak associated with tetrabutylammonium fluoride in CDCl 3 (-129 ppm). Instead, two doublets were present at −73.2 ppm and −70.1 ppm, as well as a triplet in the 31 P spectrum, suggesting the presence of two fluorine atoms bound to a phosphorous. Purification using silica gel, liquid-liquid extraction, or precipitation proved unsuccessful due to rapid degradation. Notably, no literature reports exist for the isolation of 6.

Determination of siRNA activity
Biological analysis of siRNAs was performed in vitro using HeLa (human epithelial cervix carcinoma) cells. They were kept in 250 mL vented culture flasks with 25.0 mL of DMEM supplemented with 10% fetal bovine serum and 1% penicillin-streptomycin (Sigma) in an incubator set for 37 °C @ 5% CO 2 humidified atmosphere. They were grown to 80−90% confluency then washed 3 times with 10 mL of phosphate buffered saline (NaCl 137 mM, KCl 2.70 mM, PO4 3-10.0 mM, pH 7.40) and incubated with 3.00 mL of 0.25% trypsin (SAFC bioscience) for 4 min @ 37 °C to detach the cells. Cells were harvested by adding 10.0 mL of DMEM and centrifuged at 300 xg for 5 minutes and resuspended in 5.0 mL DMEM with 10% FBS. Cells were either seeded (1 x 10 7 ) into a new flask for passaging or used in siRNA experiments by seeding 1 x 10 5 cells 12 well plates (Falcon®) with 1 mL of DMEM (10% FBS, 1% penicillin-streptomycin) and incubated at 37 °C with 5% CO 2 . After 24 hours, the cells were inspected and transfected with siRNA, pGL3-Control (200 ng), and pRLSV40 (50 ng) using Lipofectamine 2000 (Invitrogen) in Gibco's 1X Opti-Mem reduced serum media (Invitrogen) according to the manufacturer's instructions (total 100 µL transfection mixture). After 24 hours, The Dual-Reporter Luciferase Assay (Promega) was performed according to manufacturer's instructions. Briefly, cells were incubated at room temperature in 1X passive lysis buffer (Promega) for 20 minutes and lysates transferred into a 96 well, white plate (Costar). Lar II and Stop & Glo® luciferase substrates and bioluminescence were sequentially added/measured on a Synergy HT (Bio-Tek) plate luminometer. The ratio of firefly/Renilla luminescence is expressed as a percentage of reduction in firefly protein expression to siRNA efficacy when compared to unmodified siRNA controls. Each value is the average of at least 3 different experiments with standard deviation indicated.

Results & discussion
The most practical approach to prepare siRNA with 5′-phosphofluoridate modifications would follow standard solid phase nucleic acid coupling protocols via phosphoramidite synthesis. Therefore, we aimed to prepare fluorinated phosphoramidite derivatives similar to those used for the addition of a 5′ natural phosphate in solid phase oligonucleotide synthesis.
We first endeavored to synthesize the monofluorinated phosphoramidite derivative (3). To accomplish this, we first synthesized the intermediate 2 using the synthetic strategy used by Dąbkowski and Tworowska (Scheme 1). [19][20][21][22] Here, the substitution of diisopropylamine with 4-nitrophenol was accomplished via the addition of 5-ethylthio-1H-tetrazole (5-ETT), which activates the diisopropylamine group for nucleophilic substitution; however, TLC analysis revealed considerable side product of higher hydrophilicity. As this side-product also displayed the characteristic yellow color of 4-nitrophenol, we suspected a di-substituted product was limiting the product yield at 48%. We therefore investigated two approaches to limit production of the disubstituted side-product (Table 1). First, we evaluated 4,5-dicyanoimidazole, a less acidic yet more nucleophilic activator; [23] however, 4,5-dicyanoimidzaole did not substantially improve product yield (50%). Therefore, we investigated the reaction in the absence of activator. Importantly, the pKa of 4-nitrophenol (7.15) suggests that it is sufficiently acidic to activate the diisopropylamine group and facilitate its own substitution. Indeed, in the absence of an activator, the desired product was formed. The reaction was quenched with triethylamine and an improved yield of 63% was obtained. Notably, when an activator is used, a salt forms with diisopropylamine, which is easily removed by filtration. Purification of 2 was performed via silica gel chromatography using 100% DCM. Although triethylamine is commonly added to the mobile phase to prevent on-column degradation of these class of compounds, it resulted in co-elution of products and was therefore omitted. NMR reports of the product agree with previous reports. [24] Finally, compound 3 was Scheme 1. synthesis of monofluoridate phosphoramidite (top) and difluoridate phosphoramidite (bottom). synthesized and purified as previously described by Dąbkowski and Tworowski [20] via substitution of 4-nitrophenol with ionic fluorine (1 equivalent TBAF in THF) in 80% yield. Overall, this phosphoramidite was made from 1 over two steps in 50% yield.
As the impact of a 5′-difluorinated phosphofluoridate on siRNA was also of interest, we also attempted to prepare 6, following protocols developed by Dąbkowski and Tworowski (Scheme 1). [19] Under basic conditions, chlorine is readily replaced by 4-nitrophenol and solid white crystals of 5 was obtained in 73% yield. Unfortunately, while 6 was identified in crude reactions, as confirmed by NMR, it was unstable and any attempts at isolation were unsuccessful. Notably, previous reports for the synthesis of 6 have used it as a short-lived intermediate and, to the best of our knowledge, no reports for the isolation of this compound exist. As this did not afford a practical pathway for our synthetic approach, we instead focused on the synthesis and characterization of a 5′-monofluorinated phosphofluoridate modified siRNA from compound 3.
With the monofluorinated phosphitylating reagent 3, we aimed to prepare a 5′-phosphofluoridate modified siRNA via standard solid phase oligonucleotide synthesis. To accomplish this, we coupled 3 to the 5′ hydroxyl group of the antisense and sense strand 21-mers. Coupling was achieved using 0.25 M 5-ethylthiotetrazole (5-ETT) in dry ACN to activate 3, immediately followed by 10 minutes of on-column coupling and washing with ACN. The resulting phosphorofluoridite was then oxidized using H 2 O/I 2 /pyridine then cleaved and deprotected according to routine oligonucleotide methods. However, purification of the desired oligonucleotide via reverse phase purification on a C18 column proved to be challenging due to poor resolution ( Figure S1a). Such columns are routinely used by our lab for oligonucleotide purification but are largely ineffective for the purification of a 5′-phosphorofluoridate 21mer from a 5′-OH 21mer. While multiple rounds of purification and prolonged runs did provide enough purified product for the production and evaluation of siRNA ( Figure S1b), the charge difference between the two strands may have provided improved purification and recovery via ion exchange chromatography. Alternatively, while PAGE purification methods typically provide low yield, the charge difference may also be sufficient to adequately purify the strand via PAGE. Nevertheless, mass spectrometry analysis revealed samples of pure product. These samples were used to prepare siRNA that carry a 5′ phosphorofluoridate on the antisense, sense, or both strands ( Table 2).
The biological utility of these siRNAs was characterized via the dual luciferase assay, which measures knockdown of the bioluminescent luciferase gene. Regardless of its placement, a trend of reduced potency, compared to wildtype siRNA, was observed ( Figure 2). The impact that this modification has may be due to two factors. Firstly, cellular enzymes maintain a balance of 5′-phosphate and 5′-hydroxyl groups on siRNA via phosphate hydrolysis (phosphatases) and addition (kinases). It is possible that the 5′ phosphofluoridate is readily hydrolyzed by cellular phosphatases, leading to its removal and replacement with a natural phosphate, thereby eliminating the modification. If this is occurring, modification of the phosphofluoridate to prevent cleavage may still prove beneficial. Alternatively, the fluorine modification may be sufficiently tolerated by primary RNAi effector protein, Ago2. If the latter is true, the phosphofluoridate modification would not benefit strand selection.

Conclusions
Several 5′-modifications have the ability to improve the therapeutic safety and window for siRNA. Therefore, this study aimed to establish a practical  path for the synthesis, purification, and evaluation of 5′-phosphofluoridate modified siRNA. The chemical synthesis of these siRNAs provided three lessons. Firstly, nucleophilic substitution of diisopropylamine groups with 4-nitrophenol can be easily achieved without the need of an activator and, in this case, can improve yield. Secondly, fluorinated phosphoramidite compounds suffer from significant instability, and the purification of a difluorinated phosphoramidite is not practical; however, Dąbkowski and Tworowska have reported its use as an intermediate without purification. [19] Lastly, traditional oligonucleotide purification methods such as reverse phase C18 HPLC purification is impractical for these modifications. Instead, it is recommended that an ionic exchange column may provide a practical method for high purity, high yield product as it carries an additional negative charge compared to the 5′-hydroxyl 21mer. The biochemical characterization of the 21mer siRNA has revealed that a 5′-phosphofluoridate modification is not an effective modification to limit sense strand activity. However, whether this is due to cellular phosphatase activity or tolerability of Ago2 for the modification is currently unclear. In conclusion, this study improved our understanding of phosphofluoridate chemistry and identified practical methods for 5′-phosphofluoridate modified oligonucleotides.