Scalable synthesis and validation of PAMAM dendrimer‐N‐acetyl cysteine conjugate for potential translation

Abstract Dendrimer‐N‐acetyl cysteine (D‐NAC) conjugate has shown significant promise in multiple preclinical models of brain injury and is undergoing clinical translation. D‐NAC is a generation‐4 hydroxyl‐polyamidoamine dendrimer conjugate where N‐acetyl cysteine (NAC) is covalently bound through disulfide linkages on the surface of the dendrimer. It has shown remarkable potential to selectively target and deliver NAC to activated microglia and astrocytes at the site of brain injury in several animal models, producing remarkable improvements in neurological outcomes at a fraction of the free drug dose. Here we present a highly efficient, scalable, greener, well‐defined route to the synthesis of D‐NAC, and validate the structure, stability and activity to define the benchmarks for this compound. This newly developed synthetic route has significantly reduced the synthesis time from three weeks to one week, uses industry‐friendly solvents/reagents, and involves simple purification procedures, potentially enabling efficient scale up.


| I N TR ODU C TI ON
Drug development for central nervous system (CNS) disorders is highly challenging as compared to other disease targets. The critical obstacle in this endeavor is the delivery of therapeutics across the blood brain barrier (BBB) which controls the passage of molecules from circulatory system into the brain. 1,2 Various invasive and non-invasive techniques are currently being explored to get appropriate exposure of therapeutics into the brain. 1,[3][4][5] Invasive strategies involve the temporary disruption of the BBB or direct injections into CNS tissues, while noninvasive approaches primarily utilize endogenous cellular mechanisms for enhanced delivery across the BBB. These approaches can be valuable but may not be feasible in clinical settings when multiple therapeutic doses are required. In addition, the need to further deliver the drugs to cells far away from the BBB is a challenge. Neuroinflammation, mediated by activated microglia and astrocytes, has been shown to play a key role in various neurological diseases making it a potential therapeutic target. [6][7][8][9] Neuroinflammation causes transient disruption of the BBB through loosening of the tight junctions of the epithelial layer, resulting in increased permeability under certain pathological conditions. 10 This impaired BBB in neuroinflammatory disorders presents an excellent opportunity to transport therapeutic molecules or drug-loaded nanoparticles into the brain via intravenous administration.
Recently, the field of drug delivery has been revolutionized with a plethora of innovative nanotechnology-based products which can effectively diffuse through the biological barricades within the body and deliver payloads of drugs, DNA, proteins or peptides to injured or diseased cells and tissues. [11][12][13][14][15][16] Despite the significant growth in the field of nanotechnology, nanoparticles face challenges during clinical translation. Major challenges include reproducible scale-up of the synthesis, and definition of the benchmarks of the final product and the stability. 17 Maintaining batch-to-batch consistency in physico-chemical, pharmacological and biological properties is critical for enabling translation. The utility of a nano-therapeutic in the clinic greatly depends upon its safety profile which is governed by its components, size, charge, and other physiochemical properties. 18 The risk of failure can be minimized beforehand by carefully selecting rationally designed components that have already been tested and approved and can be incorporated together for the preparation of the particle.
Dendrimers are structurally "monodisperse" with well-defined building blocks and surface functionality, and have garnered significant attention as promising scaffolds for various biomedical applications including drug/gene delivery, targeting, imaging and diagnosis due to their unique physico-chemical properties. [19][20][21][22][23][24][25] These multivalent nanostructures can be tuned conveniently to incorporate various molecules of interest.
Amongst several different types of dendrimers, polyamidoamine (PAMAM) dendrimers have been most widely explored for drug delivery applications due to their commercial availability and aqueous solubility. [26][27][28] We previously reported that non-cytotoxic, hydroxyl terminated generation 4 PAMAM dendrimers (PAMAM-G4-OH, 4 nm size) can cross the impaired BBB upon systemic administration and target activated microglia and astrocytes at the site of injury in the brain several folds more than the healthy control in a neonatal brain injury model without any targeting ligand. 29 We further validated these findings in various small and large animal models of neuroinflammtory disorders. [30][31][32][33][34][35][36][37] The selective uptake and localization of these neutral dendrimers in activated microglia might be attributed to their ability to cross the impaired BBB and diffuse rapidly within the brain parenchyma. 38,39 Due to their neutral surface, hydroxyl PAMAM dendrimers have a relatively low risk of opsonization upon systemic administration and exhibit low non-specific interactions, which are a major drawback associated with cationic amine functionalized PAMAM dendrimers. [40][41][42] Unlike their charged counterparts (PAMAM-NH 2 and PAMAM-COOH), neutral PAMAM-G4-OH dendrimers are nontoxic even at intravenous doses greater than 500 mg/kg. 30,41 D-NAC is a covalent conjugate of PAMAM-G4-OH attached to Nacetyl cysteine (NAC) through a disulfide linker which can selectively release NAC intracellularly in the presence of glutathione. 30 NAC is a clinically approved anti-inflammatory and anti-oxidative agent but needs to be given at high doses due to its poor bioavailability which can cause neuronal toxicity. 43-46 D-NAC selectively delivers NAC to the inflamed areas of the brain, enhancing its efficacy, reducing collateral damage in healthy cells, and minimizing other off-target effects. We have observed 100-fold improved efficacy of D-NAC when compared to equivalent doses of the free drug. 11 D-NAC has shown significant efficacy in a neonatal rabbit model of cerebral palsy, 30 a mouse model of hypoxic-ischemia, 31 and other neuro-inflammation models of various small and large animals when injected systemically. 32,36 Although we have successfully synthesized numerous batches of D-NAC at the lab scale, in order to support our preclinical and clinical demands, we needed a highly optimized, well-established, reproducible, reliable, and robust synthetic protocol which can lead to kilogram scale quantities with a high purity and quantitative yields in few reaction steps. Based on the critical analysis of our prior reported procedure, we here report a scalable improvised process for D-NAC synthesis which can be transferred for cGMP manufacturing. The synthetic protocol discussed here for the construction of D-NAC is highly robust and reproducible, involves industrial friendly solvents, is environmentally friendly, and provides rapid access to the final conjugate with ease. We have also evaluated the reproducibility of D-NAC developed by this synthetic route by generating several batches with a similar drug payload. In addition, we developed another synthetic methodology to construct D-NAC with modifications to the linker on the dendrimer. The resultant compound has been extensively characterized for structure, drug payload, stability and activity.

| Synthesis of intermediates and dendrimers
The detailed synthetic procedures are described in the Supporting Information.

| Nuclear magnetic resonance
Nuclear magnetic resonance (NMR) spectra were recorded on a Bruker 500 MHz spectrometer at ambient temperatures. Proton and carbon chemical shifts (d) are reported in ppm. The resonance multiplicity in the 1 H NMR spectra are described as "s" (singlet), "d" (doublet), "t" (triplet), and "m" (multiplet) and broad resonances are indicated by "b".
Residual protic solvent of CDCl 3 ( 1 H, d 7.27 ppm; 13 C, d 77.0 ppm (central resonance of the triplet)), DMSO-d 6  (v/v) acetonitrile:water mixture at 10 mg/mL concentration. The samples were prepared by mixing 10 mL of conjugate solution with 10 mL of DHB solution and 3 mL of the sample was spotted on a MALDI plate. Laser power used for this purpose was 55-100%

| High-performance liquid chromatography
The purity of intermediates and final dendrimer conjugate were analyzed using high-performance liquid chromatography (HPLC). HPLC

| Calculation of NAC loading by TCEP reduction protocol
The release of free NAC from dendrimer was achieved by the reduction with tris(2-carboxy ethyl) phosphine hydrochloride (TCEP). D-NAC (1 mg in HPLC grade water (600 mL) was incubated with TCEP (780 mg) in 400 mL HPLC grade water for 2 h at room temperature with constant stirring in a glass vial. The overall concentration of D-NAC was 1 mg/mL. A calibration curve of free NAC was made using area under the curves (AUC) obtained from HPLC vs. amount, by injecting different samples of NAC ranging from 1 to 100 mg in HPLC. Hundred micro liters of sample solution from the incubation vial was injected to HPLC and AUC was recorded for NAC peak at retention time (5 min). The NAC content was calculated from the calibration curve.     These issues will be critically evaluated one by one as we discuss the advantages of improved protocol.

| Cell toxicity and lactate dehydrogenase activity
To improve the scalability of the protocol, we divided the entire D-NAC synthesis process into two halves. We elected to commence the synthesis in a semi-convergent fashion by the construction of two main intermediates: (a) bifunctional dendrimer (8, Figure 1C PAMAM-G4-OH is received in methanol solution, which is evaporated completely using a rotary evaporator, followed by the dissolution in water and lyophilization. It is very important to remove the traces of methanol and water completely as these interfere in the coupling reaction. We then reacted PAMAM-G4-OH (1, Figure 1C in DMF for 36h at room temperature to yield BOC-protected bifunctional dendrimer (7). The exact stoichiometry is described in the supporting information. The order of addition of reagents in the reaction should be followed exactly to obtain uniform product distribution with the desired loading of GABA-BOC. The dendrimer, BOC-GABA-OH and DMAP are first dissolved in DMF under Argon atmosphere, followed by the addition of EDC.HCl in portions at the end. The progress of the reaction is monitored by HPLC to ensure the formation of product with the desired loading by comparing it with reference material. Upon completion, the precipitation of the crude product is achieved by adding an excess of hexanes, which is then decanted off.
The precipitates are further washed with ethyl acetate to remove excess reagents and side products. The residue is then diluted with water and dialyzed using a 2 kDa membrane against ultra-pure water for 24 h to further remove trace impurities. We observed that the diluted reaction mixture (30 mL water for 1 g of conjugate) provided the best yield during the dialysis process. In this step, we do not perform any DMF dialysis as all the reagents and side-products are watersoluble, which was not the case in previous protocol using Fmoc-based reagents. The lyophilized product is then characterized by NMR and HPLC as further discussed in the characterization section. The BOCprotected bifunctional dendrimer (7) is water soluble. If the bifunctional dendrimer at this stage is water insoluble, that is an indication of high loading of GABA BOC. The next step involves the deprotection of BOC to obtain bifunctional dendrimer 8, which is another critical phase of the synthesis. In short, the improved protocol provides the following advantages: (a) shortens the over-all synthesis time from three weeks to one week, (b) cuts down the production cost drastically, (c) provides the desired and reproducible narrow range of NAC loading, (d) produces a highly pure, white product, and (d) promotes a green synthesis using industryfriendly reagents and solvents.

| Characterization of intermediates and final D-NAC conjugate
All of the intermediates and the final D-NAC conjugate were characterized using NMR, HPLC, MALDI, or LCMS. Figure 2A Figure 2B). The purity of D-NAC is evaluated using HPLC ( Figure 2C). The HPLC traces showed a clear shift in retention time at every stage. We have validated several batches of D-NAC using this improved protocol with more than 98% purity. We further determined the size of D-NAC to be around 5.8 nm using dynamic light scattering ( Figure 2D, Table 1). Zeta potential measurement provided the value around 6.5 mV (Table 1).

| Calculation of NAC loading by tris(2-carboxy ethyl) phosphine hydrochloride reduction method
In addition to the NAC loading by proton integration method using 1 H NMR, we further confirm NAC loading by the tris(2-carboxy ethyl) phosphine hydrochloride (TCEP) reduction method ( Figure 3A). We reduced the disulfide linkages in D-NAC using an excess amount of TCEP to obtain free NAC and thiol-terminating dendrimer (D-SH).   cultures. 52 In this study, we analyzed the results from three different batches of D-NAC synthesized using new route and compared them for reproducibility ( Figure 5A,B, Table 2). The cells were treated with NAC, D-NAC and PAMAM-G4-OH for a brief period of 6 h to ensure maximum uptake 52,53 and to mimic the in vivo blood circulation of the conjugate. At concentrations of 500 and 1,000 lg/ml, both D-NAC (three batches) and PAMAM-G4-OH did not induce cytotoxicity to the cells. The percent cell viability was greater than 98% for both D-NAC  both NAC and D-NAC were effective in suppressing TNF-a production. All the three batches of D-NAC were significantly more effective (2.5-fold, p < .001) than NAC ( Figure 5C, Table 2). NAC at 1 and 10 lg/mL did not attenuate production of TNF-a, whereas D-NAC demonstrated significant suppression. Interestingly, the TNF-a levels from the cells treated with 1lg/mL D-NAC were similar to that of 100 lg/ mL NAC ( Figure 5C). There was no significant difference in TNF-a suppression among the three batches at all the treatment concentrations demonstrates reproducible efficacy of D-NAC. This is due to the fact that all the three batches have similar drug loading and purity. Antioxidant activity was evaluated by measuring the nitric oxide (NO) levels released by the cells in to the medium using Griess reagent kit. Activated microglial cells (LPS exposed) showed significant NO release (3.0-fold) compared to non-LPS treated controls ( Figure 5D, Table 2).

| Reproducibility of D-NAC synthesis
D-NAC treatment at 10, 100 and 200 lg/mL concentrations significantly decreased NO production. In contrast, only 200 lg/mL dose of NAC was effective in decreasing the level of NO release ( Figure 5D). This enhanced efficacy can be attributed to enhanced cellular uptake of dendrimers. 47 We have previously demonstrated that the dendrimer uptake is increased when the microglial cells are activated. 55 D-NAC treatment improves intracellular glutathione (GSH) levels while suppressing glutamate release significantly compared to NAC. 52

| Fluorescently labeled D-NAC localizes with activated microglia at white matter areas of CP rabbit kits
We conjugated a fluorescent tag cyanine 5 (Cy5) to D-NAC to study brain uptake upon systemic administration in a neonatal rabbit model of CP with significant microglial activation. 30 Figure 6A presents the synthesis route of Cy5-D-NAC (15). Bifunctional dendrimer (8) was first reacted with Cy5 NHS ester at pH 7.5 for 2 h followed by the addition of NAC-SPDP (9) in the same pot. The product formation was confirmed with NMR and HPLC (Supporting Information).
The CP kits (n 5 3) received an intravenous administration of Cy5-D-NAC on PND1 and were sacrificed 24 h post-injection. We found that the Cy5-D-NAC conjugate co-localized with activated microglia, indicated by ameboid soma with shortened processes, in the periventricular white matter region (PVR), including corpus callosum, and in the lateral ventricle in the cortex ( Figure 6B). These results indicate that D-NAC crossed the blood-brain barrier, reached the injured white matter region, and targeted activated microglia in a cell-specific manner consistent with our many previous findings. 30,32,52 These properties make D-NAC an ideal therapeutic reagent in the treatment of neonatal brain injuries.

| Another synthetic approach for conjugation of NAC on dendrimer surface
While we were redesigning the synthesis of D-NAC, we also developed another synthetic route which can yield dendrimer-NAC conjugate. In this route, we generated ether bound linker on the dendrimer surface.
The ether linkages are robust, do not undergo hydrolysis and are not substrates of esterase. For this purpose, we reacted PAMAM-G4-OH (1) with allyl bromide in the presence of sodium hydride to conjugate 22-24 allyl arms on the dendrimer surface by ether bonds (11, Figure   7A). The alkene groups on dendrimer surface were then reacted with BOC-aminoethane thiol under UV light (365 nm) using highly efficient and orthogonal photocatalyzed thiol-ene click reaction to afford BOC protected dendrimer 12. Thiol-ene click is a highly robust and scalable reaction; and has been tremendously used to fabricate several