Cleaner synthesis of preclinically validated vaccine adjuvants

We developed synthetic glycophospholipids based on a glucosamine core (FP compounds) with potent and selective activity in stimulating Toll-Like Receptor 4 (TLR4) as agonists. These compounds have activity and toxicity profiles similar to the clinically approved adjuvant monophosphoryl lipid A (MPLA), included in several vaccine formulations, and are now in the preclinical phase of development as vaccine adjuvants in collaboration with Croda International PLC. FP compound synthesis is shorter and less expensive than MPLA preparation but presents challenges due to the use of toxic solvents and hazardous intermediates. In this paper we describe the optimization of FP compound synthesis. The use of regio- and chemoselective reactions allowed us to reduce the number of synthesis steps and improve process scalability, overall yield, safety, and Process Mass Intensity (PMI), thus paving the way to the industrial scale-up of the process.


Introduction
Vaccine introduction in 1798 rapidly decreased the morbidity and mortality of several deadly diseases, and their widespread use has been the reason for the eradication or attenuation of several pandemic diseases, including smallpox and the recent COVID-19 (Hilleman, 2000;Stewart and Devlin, 2006;Kayser and Ramzan, 2021;Zheng et al., 2022).
Many modern vaccines, the so-called subunit vaccines, only include parts of the pathogen, normally protein antigens, instead of the entire pathogen.They are therefore safer but less immunogenic than the vaccines containing the whole attenuated pathogen, thus requiring the addition of adjuvants (Delany et al., 2014).
Molecular adjuvants are chemical entities able to induce a strong, but controlled, immune response, thus increasing the efficacy of the vaccine in terms of the quality, intensity, and duration of immune response (Shah et al., 2017;O'Hagan et al., 2020).
The use of adjuvants contributes to the reduction of the amount of antigen required in a vaccine formulation.As normally the antigen is the most expensive component, adjuvants also have the potential to decrease the cost of vaccines, making them more accessible in developing countries (Delany et al., 2014;O'Hagan et al., 2020).
There is high industrial interest in this field: the market size of vaccine adjuvants has been valued at 895 million USD in 2021.This value is expected to double by 2027, with a forecast compound growth rate of 10.6% year-on-year, due to the involvement of companies such as GlaxoSmithKline PLC, Merck KGaA, and Croda International PLC (Industry Research, 2022).
However, the rate of innovation in the field of vaccine adjuvants has been extremely low in the last 20 years, and a formulation of aluminium salts (Alum) has been the only clinically approved adjuvant for years and even today very few compounds have been approved for human use (Li et al., 2008;Lambrecht et al., 2009;Shah et al., 2017).
Synthetic MPLA (Avanti Lipids, United States) is a complex molecule, with a disaccharide core and linear and branched fatty acid chains with a stereogenic center at C-3, whose synthesis is long (>25 steps) with elevated production costs (the present cost of MPLA is ~230 USD/mg in the American market).Furthermore, MPLA synthesis (699800P Avanti MPLA (PHAD ® ); Reed and Carter, 2014) is based on a massive use of "red" or environmentally undesirable solvents such as Pyridine or DMF (Alfonsi et al., 2008;Byrne et al., 2016;Joshi and Adhikari, 2019).
We recently synthesized FP11 and FP18 compounds (Figure 1), that showed to have similar activity to MPLA in inducing innate immune response in animal models of vaccination (Facchini et al., 2021).Their mechanism of action has been studied and it is based on the selective stimulation of the Toll-Like Receptor 4 (TLR4), one of the most important molecular switches of innate immunity (Peri and Minotti, 2019;Facchini et al., 2021).FP compounds possess however a simpler molecular formula and a shorter synthesis than MPLA.FPs retain most of the proinflammatory properties of MPLA when tested in vitro as well as the adjuvancy in vivo (Facchini et al., 2021).Due to their potential as cheaper substitutes for MPLA, we are developing FPs in collaboration with Croda International PLC as efficient substitutes for MPLA as vaccine adjuvants.
Albeit definitively more convenient than MPLA, FP compound synthesis (Scheme 1) still presents some hurdles that may prevent industrial scalability and cause a significant environmental impact, having a Process Mass Intensity (PMI) of 3.0 × 10 4 (Facchini et al., 2021).Furthermore, it employs a large amount of undesirable solvents (e.g., Pyridine, DMF, DCM), as defined by the Pfizer solvent bundle book, a widely accepted guideline for medicinal chemistry (Alfonsi et al., 2008;Byrne et al., 2016;Joshi and Adhikari, 2019).Here, we report a new, improved synthesis for FP designed with industrial scalability and environmental indications as guidelines.While the new synthesis still requires limited amounts of hazardous solvents, it is significantly shorter than the original one, translating in higher overall yields and lower PMI, a very important result due to the forecasted launch on the market of FP.
The published synthesis requires 10 steps with an overall yield of 7%.A chromatographic purification is required for 8 reaction steps out of 10, directly impacting the PMI of the process, calculated to 3.0 × 10 4 (Peri and Minotti, 2019;Facchini et al., 2021).Some synthetic steps have high safety and environmental hazards.For example, the first reaction requires the formation of a potentially explosive low molecular weight azide using the highly toxic pyridine as co-solvent; and high amounts of toxic and pollutant solvents such as DMF and DCM are abundantly used throughout the process.
Finally, the absence of chemical orthogonality between the protective groups does not allow for an easy selective deprotection, in the perspective of selectively functionalizing the C-6 hydroxyl group.
A new, versatile synthesis has been designed (Scheme 2) with a reduced number of synthetic steps (7) and purifications and less toxic solvents involved.The overall yield is 18% and a PMI of 9.8 × 10 3 .This synthesis can be applied for both FP11 and FP18 by employing the correct lipid chain: reaction yields are very similar with a very narrow error range.
The first step is the acylation of the glucosamine on the 2-NH, exploiting its higher reactivity, so that it is not necessary to protect it anymore.
The second reaction is a silylation on the 6-OH of compound 11, the only protection step in the synthesis: it is possible to regioselectively protect the more reactive primary alcohol over the other hydroxyls.However, the protecting group has to be carefully selected: a small one (e.g., TMS, TES) would not be selective enough; and a larger one (e.g., TBDPS, Trt) would prevent phosphorylation for sterical reasons.An additional challenge in this reaction was the choice of the solvent (Table 1), due to the poor solubility of the substrate both in aqueous and organic solvents.Initially, diluted pyridine was used, with a yield of 50%, but its extreme toxicity prompted us to search for a better medium.Several solvents were screened (i.e., MeCN, tBuOH, DMF) to no result, as the substrate failed to dissolve and the product was obtained only in traces.Finally, we managed to dissolve the substrate in DMSO at a low concentration (0.05 M) and to perform the reaction with a yield of 90%: therefore, we managed both to avoid pyridine and to reduce the PMI.
The third step of the pathway is the acylation of 12 on C-3 and C-4 hydroxyls.The reaction was first carried out in pyridine, which was eventually replaced with THF maintaining the high yield (80%) and reducing the hazards.The reaction stereochemistry at the anomeric carbon is dependent on the reaction conditions: short reaction time, high temperature and catalyst loading favor the formation of the thermodynamic α anomer, while longer reaction time, low temperature, and catalyst loading favor the formation of the kinetic β product (Romerio et al., 2023).As we have to remove SCHEME 2 Cleaner synthesis for FP compounds, charcaterized by a reduced number of reactions, a limited use of red solvent, reduced PMI and increased yield. the anomeric lipid chain (v.infra), the anomeric configuration is not relevant and it is possible to choose the protocol most suitable to one's needs.The fourth step is a regioselective deacylation of 13 with cleavage of the lipid chain in the anomeric position using a mixture of acetic acid and ethylenediamine.Interestingly, the anomeric acyl group acted as a leaving group in the presence of acids, and configuration seems to be retained (Zhang and Kováč, 1999).
Subsequent phosphorylation of 14 was performed using the phosphite to phosphate strategy, in which the compound undergoes a phosphitylation followed by one-pot oxidation to phosphate, similar to the previously published synthesis.This reaction is highly stereoselective: it always results in pure α configuration, independently from the starting configuration, as shown in several previous publications (Cighetti et al., 2014;Facchini et al., 2018;2021;Peri and Minotti, 2019).
The 6-OH of 15 was then deprotected in mild conditions to avoid phosphate cleavage.Optimal cleavage conditions without concomitant reaction of protected phosphate consisted of the use of IRC 120 H + resin in acetone.The reaction proceeded with 55% yield, but recycling of unreacted 15 allowed to further enhance yield.
Benzyl groups on the phosphates of compound 16 were removed by catalytic hydrogenation, as in the original synthesis (Peri and Minotti, 2019;Facchini et al., 2021).
The new synthesis is scalable for industrial production, with higher overall yield, lower PMI, and minimum use of "undesirable" or "red" solvents (Alfonsi et al., 2008;Byrne et al., 2016;Joshi and Adhikari, 2019).

Conclusion
Here we reported a new synthesis for FP compounds: a class of chemically simplified analogues of the known vaccine adjuvant MPLA, whose synthesis is significantly long and expensive.Regio-and chemoselective reactions allowed a drastic reduction in the use of protecting groups.Consequently, we managed to reduce the number of steps needed for the synthesis, which increased the overall yield (from 7% to 18%) and reduced the PMI (from 3.0 × 10 4 to 9.8 × 10 3 ) of the process.Furthermore, we eliminated the first hazardous intermediate and greatly decreased the use of red solvents replacing them with green or yellow solvents (Acetone, DMSO, or THF).
The described optimized synthesis will be further adapted to safety requirements and employed for industrial upscaling and production of the new immunostimulating agents FP11 and FP18.

Materials and methods
All reagents and solvents were purchased from commercial sources and used without further purifications, unless stated otherwise.
Reactions were monitored by thin-layer chromatography (TLC) performed over Silica Gel 60 F254 plates (Merck ® ).Flash chromatography purifications were performed on silica gel 60 60-75 μm from a commercial source.Solvent removal by rotavapor was carried out at 40 °C for most solvents and 55 °C for toluene and water, unless otherwise stated. 1 H, 13 C, and 31 P NMR spectra were recorded with Bruker Advance 400 with TopSpin ® software, or with NMR Varian 400 with Vnmrj software.Chemical shifts are expressed in ppm with respect to Me 4 Si; coupling constants are expressed in Hz.The multiplicity in the 13 C spectra was deducted by APT experiments.Exact masses were recorded with Agilent 6500 Series Q-TOF LC/MS System.The purity of the final compounds was about 95% as assessed by quantitative NMR analysis.Optical rotation values were acquired with Anton Paar MCP 100 polarimeter with a Type II cell (l = 100 mm; Ø = 5 mm) operating at 20 °C.

TABLE 1
Solvent screening for the silylation reaction.