Insights into Factors Affecting Ethylene-Vinyl Acetate Copolymer Crystallinity in Islatravir Implant.

Islatravir, a highly potent nucleoside reverse transcriptase translocation inhibitor (NRTTI) for the treatment of HIV, has great potential to be formulated as ethylene-vinyl acetate (EVA) copolymer-based implants via hot melt extrusion. The crystallinity of EVA determines its physical and rheological properties and may impact the drug-eluting implant performance. Herein, we describe the systematic analysis of factors affecting the EVA crystallinity in islatravir implants. Differential scanning calorimetry (DSC) on EVA and solid-state NMR revealed drug loading promoted EVA crystallization, whereas BaSO4 loading had negligible impact on EVA crystallinity. The sterilization through γ-irradiation appeared to significantly impact the EVA crystallinity and surface characteristics of the implants. Furthermore, DSC analysis of thin implant slices prepared with an ultramicrotome indicated that the surface layer of the implant was more crystalline than the core. These findings provide critical insights into factors affecting the crystallinity, mechanical properties, and physicochemical properties of the EVA polymer matrix of extruded islatravir implants.

Deep learning image analysis models pretrained on daily objects are useful for the preliminary characterization of particulate pharmaceutical samples.

Visible and subvisible particles are a quality attribute in sterile pharmaceutical samples. A common method for characterizing and quantifying pharmaceutical samples containing particulates is imaging many individual particles using high-throughput instrumentation and analyzing the populations data. The analysis includes conventional metrics such as the particle size distribution but can be more sophisticated by interpreting other visual/morphological features. To avoid the hurdles of building new image analysis models capable of extracting such relevant features from scratch, we propose using well-established pretrained deep learning image analysis models such as EfficientNet. We demonstrate that such models are useful as a prescreening tool for high-level characterization of biopharmaceutical particle image data. We show that although these models are originally trained for completely different tasks (such as the classification of daily objects in the ImageNet database), the visual feature vectors extracted by such models can be useful for studying different types of subvisible particles. This applicability is illustrated through multiple case studies: (i) particle risk assessment in prefilled syringe formulations containing different particle types such as silicone oil, (ii) method comparability with the example of accelerated forced degradation, and (iii) excipient influence on particle morphology with the example of Polysorbate 80 (PS80). As examples of agnostic applicability of pretrained models, we also elucidate the application to two high-throughput microscopy methods: microflow and background membrane imaging. We show that different particle populations with different morphological and visual features can be identified in different samples by leveraging out-of-the-box pretrained models to analyze images from each sample.

Evaluation of a Self-Supervised Machine Learning Method for Screening of Particulate Samples: A Case Study in Liquid Formulations.

Imaging is commonly used as a characterization method in the pharmaceuticals industry, including for quantifying subvisible particles in solid and liquid formulations. Extracting information beyond particle size, such as classifying morphological subpopulations, requires some type of image analysis method. Suggested methods to classify particles have been based on pre-determined morphological features or use supervised training of convolutional neural networks to learn image representations in relation to ground truth labels. Complications arising from highly complex morphologies, unforeseen classes, and time-consuming preparation of ground truth labels, are some of the challenges faced by these methods. In this work, we evaluate the application of a self-supervised contrastive learning method in studying particle images from therapeutic solutions. Unlike with supervised training, this approach does not require ground truth labels and representations are learned by comparing particle images and their augmentations. This method provides a fast and easily implementable tool of coarse screening for morphological attribute assessment. Furthermore, our analysis shows that in cases with relatively balanced datasets, a small subset of an image dataset is sufficient to train a convolutional neural network encoder capable of extracting useful image representations. It is also demonstrated that particle classes typically observed in protein solutions administered by pre-filled syringes emerge as separated clusters in the encoder's embedding space, facilitating performing tasks such as training weakly-supervised classifiers or identifying the presence of new subpopulations.

Convolutional Neural Networks Enable Highly Accurate and Automated Subvisible Particulate Classification of Biopharmaceuticals.

Quantification of subvisible particles, which are generally defined as those ranging in size from 2 to 100 µm, is important as critical characteristics for biopharmaceutical formulation development. Micro Flow Imaging (MFI) provides quantifiable morphological parameters to study both the size and type of subvisible particles, including proteinaceous particles as well as non-proteinaceous features incl. silicone oil droplets, air bubble droplets, etc., thus enabling quantitative and categorical particle attribute reporting for quality control. However, limitations in routine MFI image analysis can hinder accurate subvisible particle classification. In this work, we custom-built a subvisible particle-aware Convolutional Neural Network, SVNet, which has a very small computational footprint, and achieves comparable performance to prior state-of-art image classification models. SVNet significantly improves upon current standard operating procedures for subvisible particulate assessments as confirmed by thorough real-world validation studies.

Flexibility in Drug Product Development: A Perspective.

The process of bringing a drug to market involves innumerable decisions to refine a concept into a final product. The final product goes through extensive research and development to meet the target product profile and to obtain a product that is manufacturable at scale. Historically, this process often feels inflexible and linear, as ideas and development paths are eliminated early on to allow focus on the workstream with the highest probability of success. Carrying multiple options early in development is both time-consuming and resource-intensive. Similarly, changing development pathways after significant investment carries a high "penalty of change" (PoC), which makes pivoting to a new concept late in development inhibitory. Can drug product (DP) development be made more flexible? The authors believe that combining a nonlinear DP development approach, leveraging state-of-the art data sciences, and using emerging process and measurement technologies will offer enhanced flexibility and should become the new normal. Through the use of iterative DP evaluation, "smart" clinical studies, artificial intelligence, novel characterization techniques, automation, and data collection/modeling/interpretation, it should be possible to significantly reduce the PoC during development. In this Perspective, a review of ideas/techniques along with supporting technologies that can be applied at each stage of DP development is shared. It is further discussed how these contribute to an improved and flexible DP development through the acceleration of the iterative build-measure-learn cycle in laboratories and clinical trials.

Quantifying Molecular Mixing and Heterogeneity in Pharmaceutical Dispersions at Sub-100 nm Resolution by Spin Diffusion NMR.

Molecular miscibility and homogeneity of amorphous solid dispersions (ASDs) are critical attributes that impact physicochemical stability, bioavailability, and processability. Observation of a single glass transition is utilized as a criterion for good mixing of drug substance and polymeric components but can be misleading and cannot quantitatively analyze the domain size at high resolution. While imaging techniques, on the other hand, can characterize phase separation on the particle surface at the nanometer scale, they often require customized sample preparation and handling. Moreover, a mixed system is not necessarily homogeneous. Compared to the numerous studies that have evaluated the mixing of drug substance and polymer in ASDs, inhomogeneity in the phase compositions has remained significantly underexplored. To overcome the analytical challenge, we have developed a 1H spin diffusion NMR technique to quantify molecular mixing of bulk ASDs at sub-100 nm resolution. It combines relaxation filtering (T2H and T1ρ) that leaves the active pharmaceutical ingredient (API) as the main source of 1H magnetization at the start of spin diffusion to the polymer matrix. A spray-dried nifedipine-poly(vinylpyrrolidone) (Nif-PVP) ASD at a 5 wt % drug loading was a homogeneous reference system that exhibited equilibration of magnetization transfer from API to polymer within a short spin diffusion time of ∼3 ms. While fast initial magnetization transfer proving mixing on the 1 nm scale was also observed in Nif-PVP ASDs prepared by hot-melt extrusion (HME) at 186 °C at a 40 wt % drug loading, incomplete equilibration of peak intensities documented inhomogeneity on the ≥30 nm scale. The nonuniformity was confirmed by the partial inversion of the Nif magnetization in the filter that resulted in an even more pronounced deviation from equilibration and by 1H-13C heteronuclear correlation (HETCOR) NMR. It is consistent with the observed differential 1H spin-lattice relaxation of Nif and PVP as well as a domain structure on the 20 nm scale observed in atomic force microscopy (AFM) images. The incomplete equilibration and differential relaxation were consistently reproduced in a model of two mixed phases of different compositions, e.g., 40 wt % of the ASD with a 15 wt % drug loading and the remaining 60 wt % with a 56 wt % drug loading. Hot-melt extrusion produced more inhomogeneous samples than spray drying for the samples examined in our study. To the best of our knowledge, this spin diffusion NMR method provides currently the highest-resolution quantification of inhomogeneous molecular mixing and phase composition in bulk samples of pharmaceutical dispersions produced with equipment, procedures, and drug loadings that are relevant to industrial drug development.

Islatravir Case Study for Enhanced Screening of Thermodynamically Stable Crystalline Anhydrate Phases in Pharmaceutical Process Development by Hot Melt Extrusion.

The emergence of new active pharmaceutical ingredient (API) polymorphs in pharmaceutical development presents significant risks. Even with thorough polymorph screening, new pathways toward alternate crystal phases can present themselves over the course of formulation development; thus, further improvements in phase screening methods are needed. Herein, a case study is presented of a thermodynamically stable crystalline phase of the HIV drug Islatravir (MK-8591, EFdA) that was not isolated from initial pharmaceutical polymorph screening. In total, five Islatravir phases are identified: one monohydrate and four anhydrate phases. The new phase, anhydrate form IV, was unexpectedly discovered during hot melt extrusion (HME) process development of polymeric implant drug product formulations while probing extreme manufacturing process conditions (elevated shear forces). X-ray diffraction (XRD), differential scanning calorimetry (DSC), and solid-state nuclear magnetic resonance (ssNMR) were utilized as principal tools to identify the new polymorph. The result suggests that HME introduces conditions that may allow a thermodynamically stable crystalline phase to form and these conditions are not necessarily captured by routine pharmaceutical polymorph screening. Subsequent investigations identified procedures to generate the new anhydrate phase without HME equipment by the use of special thermal procedures. It is found that for a crystalline hydrate phase the rate of water loss as well as water entrapment in a heating vessel play a crucial role in phase conversions into different anhydrate polymorphs. Further, the polymer involved in the HME manufacturing process also plays a critical role in the phase conversion, likely by coating the API microparticles and thereby altering the phase conversion kinetics. Strategies presented herein to mimic phase changes during formulation manufacture hold promise for the identification of thermodynamically stable anhydrate forms in earlier stages of pharmaceutical development.

An Imaging Toolkit for Physical Characterization of Long-Acting Pharmaceutical Implants.

In pharmaceutical development alternative drug delivery modalities are being increasingly employed. One example is an implant, which achieves gradual drug release in patients over a period of many months or years. Due to the complexity of these long-acting formulations, advanced physical characterization methods are desirable as screening tools during protracted formulation development. Imaging methods are of particular interest due to their ability to interrogate the structure and composition of implants spatially across multiple length scales (macro, micro, nano). In this work, spatiochemical imaging is shown to interrogate many crucial drug product attributes of solid implants: overall implant structure, drug distribution, micro-domain size and orientation, agglomeration, porosity and defects, drug/excipient interface, dissolution process, and release mechanism. Imaging methods facilitate a detailed understanding of the process/structure correlation to inform on formulation selection, process parameter optimization, and batch consistency. Numerous case studies of implant applications with imaging are discussed. Methods utilized are X-ray computed tomography (XRCT), scanning electron microscopy (SEM), energy-dispersive X-ray spectroscopy (EDS) imaging, and Raman microscopy. The imaging data is complemented with solid-state nuclear magnetic resonance (ssNMR). Altogether, these examples demonstrate that complementary imaging methods are highly effective for analyzing complex and novel pharmaceutical modalities such as solid implants.

Molecular Mechanism of Crystalline-to-Amorphous Conversion of Pharmaceutical Solids from 19F Magic Angle Spinning NMR.

Crystalline and amorphous materials usually possess distinct physicochemical properties due to major variations in long-range and local molecular packings. Enhanced fundamental knowledge of the molecular details of crystalline-to-amorphous interconversions is necessary to correlate the intermolecular structure to material properties and functions. While crystal structures can be readily obtained by X-ray crystallography, the microstructure of amorphous materials has rarely been explored due to a lack of high-resolution techniques capable of probing local molecular structures. Moreover, there is increasing interest in understanding the molecular nature of amorphous solids in pharmaceutical sciences due to the widespread utilization of amorphous active pharmaceutical ingredients (APIs) in pharmaceutical development for solubility and bioavailability enhancement. In this study, we explore multidimensional 13C and 19F magic angle spinning (MAS) NMR spectroscopy to study the molecular packing of amorphous posaconazole (POSA) in conjunction with the crystalline counterpart. Utilizing methods integrating homonuclear and heteronuclear 1H, 13C, and 19F correlation spectroscopy and atomic 19F-to-13C distance measurements, we identified the major differences in molecular packing between crystalline and amorphous POSA. The intermolecular "head-to-head" interaction along the molecule's major axis, as well as the "head-to-tail" molecular packing perpendicular to the major axis in POSA crystals, was recapitulated by MAS NMR. Furthermore, critical intermolecular distances in the crystal lattice were determined. Most importantly, the head-to-tail contact of two neighboring molecules was found to be preserved in amorphous POSA, suggesting localized molecular order, whereas crucial interactions for head-to-head packing are absent in the amorphous form resulting in long-range disorder. Our study, likely one of the first documented examples, provides molecular-level structural details to understand the molecular mechanism of crystalline-to-amorphous conversion of fluorine-containing drug substances occurring in drug processing and development and establish a high-resolution experimental protocol for investigating amorphous materials.

Application of Deep Learning Convolutional Neural Networks for Internal Tablet Defect Detection: High Accuracy, Throughput, and Adaptability.

Tablet defects encountered during the manufacturing of oral formulations can result in quality concerns, timeline delays, and elevated financial costs. Internal tablet cracking is not typically measured in routine inspections but can lead to batch failures such as tablet fracturing. X-ray computed tomography (XRCT) has become well-established to analyze internal cracks of oral tablets. However, XRCT normally generates very large quantities of image data (thousands of 2D slices per data set) which require a trained professional to analyze. A user-guided manual analysis is laborious, time-consuming, and subjective, which may result in a poor statistical representation and inconsistent results. In this study, we have developed an analysis program that incorporates deep learning convolutional neural networks to fully automate the XRCT image analysis of oral tablets for internal crack detection. The computer program achieves robust quantification of internal tablet cracks with an average accuracy of 94%. In addition, the deep learning tool is fully automated and achieves a throughput capable of analyzing hundreds of tablets. We have also explored the adaptability of the deep learning analysis program toward different products (e.g., different types of bottles and tablets). Finally, the deep learning tool is effectively implemented into the industrial pharmaceutical workflow.

Three-Dimensional NMR Spectroscopy of Fluorinated Pharmaceutical Solids under Ultrafast Magic Angle Spinning.

High-resolution solid-state analysis of multicomponent molecular systems, e.g., pharmaceutical formulations, is a great challenge. Solid-state nuclear magnetic resonance (ssNMR) spectroscopy plays a critical role in the characterization of solid dosage forms due to its capabilities of chemical identification, quantification, and structural elucidation at a molecular level. However, the low NMR sensitivity as well as the high spectral complexity and low drug loading of multicomponent products hinder an in-depth investigation of the active pharmaceutical ingredient (API) at the natural isotopic abundance. Herein, we developed two new three-dimensional (3D) ssNMR methods, including 1H-19F-1H and 19F-19F-1H correlations and successfully applied them to characterize a fluorinated drug molecule, aprepitant, and its commercial nanoparticulate formulation EMEND (Merck & Co, Inc., Kenilworth, NJ, USA). These 1H-detection methods utilize the significantly enhanced sensitivity and resolution of 1H and 19F afforded by 60 kHz ultrafast magic angle spinning (MAS) and enable the analysis of milligram samples. The 3D techniques simultaneously provide homonuclear 1H-1H and 19F-19F, and heteronuclear 1H-19F correlations of the crystalline aprepitant without interferences from other pharmaceutical components in the drug product. Moreover, our results demonstrate that 19F is a highly sensitive spin for probing molecular details of fluorinated drug substances in solid formulations, due to its high isotopic abundance, large gyromagnetic ratio, and absence of signal interference from pharmaceutical excipients, as well as for characterizing structural properties of a broad range of fluorine-containing materials.

Extended-Duration MK-8591-Eluting Implant as a Candidate for HIV Treatment and Prevention.

Regimen adherence remains a major hurdle to the success of daily oral drug regimens for the treatment and prevention of human immunodeficiency virus (HIV) infection. Long-acting drug formulations requiring less-frequent dosing offer an opportunity to improve adherence and allow for more forgiving options with regard to missed doses. The administration of long-acting formulations in a clinical setting enables health care providers to directly track adherence. MK-8591 (4'-ethynyl-2-fluoro-2'-deoxyadenosine [EFdA]) is an investigational nucleoside reverse transcriptase translocation inhibitor (NRTTI) drug candidate under investigation as part of a regimen for HIV treatment, with potential utility as a single agent for preexposure prophylaxis (PrEP). The active triphosphate of MK-8591 (MK-8591-TP) exhibits protracted intracellular persistence and, together with the potency of MK-8591, supports its consideration for extended-duration dosing. Toward this end, drug-eluting implant devices were designed to provide prolonged MK-8591 release in vitro and in vivo Implants, administered subcutaneously, were studied in rodents and nonhuman primates to establish MK-8591 pharmacokinetics and intracellular levels of MK-8591-TP. These data were evaluated against pharmacokinetic and pharmacodynamic models, as well as data generated in phase 1a (Ph1a) and Ph1b clinical studies with once-weekly oral administration of MK-8591. After a single administration in animals, MK-8591 implants achieved clinically relevant drug exposures and sustained drug release, with plasma levels maintained for greater than 6 months that correspond to efficacious MK-8591-TP levels, resulting in a 1.6-log reduction in viral load. Additional studies of MK-8591 implants for HIV treatment and prevention are warranted.

Redox Isomeric Surface Structures Are Preferred over Odd-Electron Pt1.

The formation of metal-ligand coordination networks on surfaces that contain redox isomers is a topic of considerable interest and is important for bifunctional metallochemistry, including heterogeneous catalysis. Towards this end, a tetrazine with two electron withdrawing pyrimidinyl substituents was co-deposited with platinum metal on the Au(100) surface. In a 2:1 metal:ligand ratio, only half of the platinum is oxidized to the +2 oxidation state, with the remainder coordinating to the ligand without charge transfer, as Pt0 . The resultant Pt0 /PtII mixed valence structure is thought to form due to the aversion of the ligand towards a four-electron reduction and the strong preference of Pt towards 0 and +2 oxidation states. These results were confirmed through a series of experiments varying the on-surface metal:ligand stoichiometry in the redox assembly formed: added oxidant does not oxidize the already complexed Pt0 . Scanning tunneling microscopy reveals irregular chain structures that are attributed to the mixture of Pt valence states, each with distinct local coordination geometries. Density functional theory calculations give further detail about these local geometries. These results demonstrate the formation of a mixture of valence states in on-surface redox assembly of metal-organic networks that extends the library of single-site metal structures for surface chemistry and catalysis. Redox-isomeric Pt0 versus Pt2+ surface structures can coexist in this ligand environment.

Two- and Three-Electron Oxidation of Single-Site Vanadium Centers at Surfaces by Ligand Design.

Rational, systematic tuning of single-site metal centers on surfaces offers a new approach to increase selectivity in heterogeneous catalysis reactions. Although such metal centers of uniform oxidation states have been achieved, the ability to control their oxidation states through the use of carefully designed ligands had not been shown. To this end, tetrazine ligands functionalized by two pyridinyl or pyrimidinyl substituents were deposited, along with vanadium metal, on the Au(100) surface. The greater oxidizing power of the bis-pyrimidinyltetrazine facilitates the on-surface redox formation of V(3+), compared to V(2+) when paired with the bis-pyridinyltetrazine, as determined by X-ray photoelectron spectroscopy. This demonstrates the ability to control metal oxidation states in surface coordination architectures by altering the redox properties of organic ligands. The metal-ligand complexes take the form of one-dimensional polymeric chains, resolved by scanning tunneling microscopy. The chain structures in the first layer are very uniform and are based on the same quasi-square-planar coordination geometry around single-site V with either ligand. Formation of a different, dimer structure is observed in the early stages of the second layer formation. These systems offer new opportunities in controlling the oxidation state of single-site transition metal atoms at a surface for new advances in heterogeneous catalysts.

Redox-active on-surface polymerization of single-site divalent cations from pure metals by a ketone-functionalized phenanthroline.

Metallic iron, chromium, or platinum mixing with a ketone-functionalized phenanthroline ligand on a single crystal gold surface demonstrates redox activity to a well-defined oxidation state and assembly into thermally stable, one dimensional, polymeric chains. The diverging ligand geometry incorporates redox-active sub-units and bi-dentate binding sites. The gold surface provides a stable adsorption environment and directs growth of the polymeric chains, but is inert with regard to the redox chemistry. These systems are characterized by scanning tunnelling microscopy, non-contact atomic force microscopy, and X-ray photoelectron spectroscopy under ultra-high vacuum conditions. The relative propensity of the metals to interact with the ketone group is examined, and it is found that Fe and Cr more readily complex the ligand than Pt. The formation and stabilization of well-defined transition metal single-sites at surfaces may open new routes to achieve higher selectivity in heterogeneous catalysts.

High-fidelity self-assembly of crystalline and parallel-oriented organic thin films by π-π stacking from a metal surface.

Organic semiconductor applications will significantly benefit from atomically precise, cofacial stacking of extended π-conjugated molecular systems for efficient charge transport. Surface-assisted self-assembly of poly(hetero)cyclic molecules via donor-acceptor type π-π stacking is a promising strategy to organize functional, many-layered architectures. We have employed tris(N-phenyltriazole) as a model system to achieve molecular-level structural ordering through more than 20 molecular layers from its own metal-templated monolayer. Effective charge transport through such layers enabled molecular-resolution imaging by scanning tunneling microscopy. The structure and chemical composition of the films, grown on Ag(111) or Au(100), were further analyzed by noncontact atomic force microscopy and X-ray photoelectron spectroscopy, revealing a cofacial stacking geometry of the molecular layers. Scanning tunneling spectroscopy measurements show a decrease of the band gap with increasing film thickness, consistent with π-π stacking and electron delocalization. The present study provides new strategies for the fabrication of normally inaccessible structural motifs, atomic precision in organic films, and the effective conduction of electrons through multiple organic molecular stacks.

Redox-active on-surface assembly of metal-organic chains with single-site Pt(II).

The formation and stabilization of well-defined transition-metal single sites at surfaces may open new routes to achieve higher selectivity in heterogeneous catalysts. Organic ligand coordination to produce a well-defined oxidation state in weakly reducing metal sites at surfaces, desirable for selective catalysis, has not been achieved. Here, we address this using metallic platinum interacting with a dipyridyl tetrazine ligand on a single crystal gold surface. X-ray photoelectron spectroscopy measurements demonstrate the metal-ligand redox activity and are paired with molecular-resolution scanning probe microscopy to elucidate the structure of the metal-organic network. Comparison to the redox-inactive diphenyl tetrazine ligand as a control experiment illustrates that the redox activity and molecular-level ordering at the surface rely on two key elements of the metal complexes: (i) bidentate binding sites providing a suitable square-planar coordination geometry when paired around each Pt, and (ii) redox-active functional groups to enable charge transfer to a well-defined Pt(II) oxidation state. Ligand-mediated control over the oxidation state and structure of single-site metal centers that are in contact with a metal surface may enable advances in higher selectivity for next generation heterogeneous catalysts.

Robust surface nano-architecture by alkali-carboxylate ionic bonding.

Ionic bonding in supramolecular surface networks is a promising strategy to self-assemble nanostructures from organic building blocks with atomic precision. However, sufficient thermal stability of such systems has not been achieved at metal surfaces, likely due to partial screening of the ionic interactions. We demonstrate excellent stability of a self-assembled ionic network on a metal surface at elevated temperatures. The structure is characterized directly by atomic resolution scanning tunneling microscopy (STM) experiments conducted at 165 °C showing intact domains. This robust nanometer-scale structure is achieved by the on-surface reaction of a simple and inexpensive compound, sodium chloride, with a model system for carboxylate interactions, terephthalic acid (TPA). Rather than distinct layers of TPA and NaCl, angle resolved X-ray photoelectron spectroscopy experiments indicate a replacement reaction on the Cu(100) surface to form Na-carboxylate ionic bonds. Chemical shifts in core level electron states confirm a direct interaction and a +1 charge state of the Na. High-temperature STM imaging shows virtually no fluctuation of Na-TPA island boundaries, revealing a level of thermal stability that has not been previously achieved in noncovalent organic-based nanostructures at surfaces. Comparable strength of intermolecular ionic bonds and intramolecular covalent bonds has been achieved in this surface system. The formation of these highly ordered structures and their excellent thermal stability is dependent on the interplay of adsorbate-substrate and ionic interactions and opens new possibilities for ionic self-assemblies at surfaces with specific chemical function. Robust ionic surface structures have potential uses in technologies requiring high thermal stability and precise ordering through self-assembly.