Best Practices and Guidelines (2022) for Scale-up and Technology Transfer in Freeze Drying Based on Case Studies. Part 2: Past Practices, Current Best Practices, and Recommendations.

Scale-up and transfer of lyophilization processes remain very challenging tasks considering the technical challenges and the high cost of the process itself. The challenges in scale-up and transfer were discussed in the first part of this paper and include vial breakage during freezing at commercial scale, cake resistance differences between scales, impact of differences in refrigeration capacities, and geometry on the performance of dryers. The second part of this work discusses successful and unsuccessful practices in scale-up and transfer based on the experience of the authors. Regulatory aspects of scale-up and transfer of lyophilization processes were also outlined including a topic on the equivalency of dryers. Based on an analysis of challenges and a summary of best practices, recommendations on scale-up and transfer of lyophilization processes are given including projections on future directions in this area of the freeze drying field. Recommendations on the choice of residual vacuum in the vials were also provided for a wide range of vial capacities.

Translating dosimetry of Dibenzo[def,p]chrysene (DBC) and metabolites across dose and species using physiologically based pharmacokinetic (PBPK) modeling.

Dibenzo[def,p]chrysene (DBC) is an environmental polycyclic aromatic hydrocarbon (PAH) that causes tumors in mice and has been classified as a probable human carcinogen by the International Agency for Research on Cancer. Animal toxicity studies often utilize higher doses than are found in relevant human exposures. Additionally, like many PAHs, DBC requires metabolic bioactivation to form the ultimate toxicant, and species differences in DBC and DBC metabolite metabolism have been observed. To understand the implications of dose and species differences, a physiologically based pharmacokinetic model (PBPK) for DBC and major metabolites was developed in mice and humans. Metabolism parameters used in the model were obtained from experimental in vitro metabolism assays using mice and human hepatic microsomes. PBPK model simulations were evaluated against mice dosed with 15 mg/kg DBC by oral gavage and human volunteers orally microdosed with 29 ng of DBC. DBC and its primary metabolite DBC-11,12-diol were measured in blood of mice and humans, while in urine, the majority of DBC metabolites were obeserved as conjugated DBC-11,12-diol, conjugated DBC tetrols, and unconjugated DBC tetrols. The PBPK model was able to predict the time course concentrations of DBC, DBC-11,12-diol, and other DBC metabolites in blood and urine of human volunteers and mice with reasonable accuracy. Agreement between model simulations and measured pharmacokinetic data in mice and human studies demonstrate the success and versatility of our model for interspecies extrapolation and applicability for different doses. Furthermore, our simulations show that internal dose metrics used for risk assessment do not necessarily scale allometrically, and that PBPK modeling provides a reliable approach to appropriately account for interspecies differences in metabolism and physiology.

Best Practices and Guidelines (2022) for Scale-Up and Tech Transfer in Freeze-Drying Based on Case Studies. Part 1: Challenges during Scale Up and Transfer.

The freeze-drying process scale-up and transfer remain a complicated and non-uniform practice. We summarized inefficient and good practices in these papers and provided some practical advice. It was demonstrated that using the same process set points/times in laboratory and commercial scale dryers may lead to loss of product quality (collapse or vial breakage). The emerging modeling approach demonstrated practical advantages. However, the upfront generation of some input parameters (vial heat transfer coefficient, minimum controllable pressure, and maximum sublimation rate) is essential for model utilization. While the primary drying step can be transferred with a high degree of confidence (e.g., using modeling), and secondary drying is usually fairly straightforward, predicting potential changes in product behavior during freezing remains challenging.

Automated fast computational adaptive optics for optical coherence tomography based on a stochastic parallel gradient descent algorithm.

The transverse resolution of optical coherence tomography is decreased by aberrations introduced from optical components and the tested samples. In this paper, an automated fast computational aberration correction method based on a stochastic parallel gradient descent (SPGD) algorithm is proposed for aberration-corrected imaging without adopting extra adaptive optics hardware components. A virtual phase filter constructed through combination of Zernike polynomials is adopted to eliminate the wavefront aberration, and their coefficients are stochastically estimated in parallel through the optimization of the image metrics. The feasibility of the proposed method is validated by a simulated resolution target image, in which the introduced aberration wavefront is estimated accurately and with fast convergence. The computation time for the aberration correction of a 512 × 512 pixel image from 7 terms to 12 terms requires little change, from 2.13 s to 2.35 s. The proposed method is then applied for samples with different scattering properties including a particle-based phantom, ex-vivo rabbit adipose tissue, and in-vivo human retina photoreceptors, respectively. Results indicate that diffraction-limited optical performance is recovered, and the maximum intensity increased nearly 3-fold for out-of-focus plane in particle-based tissue phantom. The SPGD algorithm shows great potential for aberration correction and improved run-time performance compared to our previous Resilient backpropagation (Rprop) algorithm when correcting for complex wavefront distortions. The fast computational aberration correction suggests that after further optimization our method can be integrated for future applications in real-time clinical imaging.

Site-Specific Incorporation of N-(2'-Deoxyguanosine-8-yl)-6-aminochrysene Adduct in DNA and Its Replication in Human Cells.

The environmental pollutant 6-nitrochrysene (6-NC) is a potent mutagen and a mammary carcinogen in rats. 6-NC is the most potent carcinogen ever tested in the newborn mouse assay. In mammalian cells, it is metabolically activated by nitroreduction and a combination of ring oxidation and nitroreduction pathways. The nitroreduction pathway yields two major adducts with 2'-deoxyguanosine (dG), one at the C8-position, N-(dG-8-yl)-6-AC, and the other at the exocyclic N2-position, 5-(dG-N2-yl)-6-AC. Here, we report the total synthesis of a site-specific oligonucleotide containing the 6-NC-derived C8 dG adduct, N-(dG-8-yl)-6-AC. Pd-catalyzed Buchwald-Hartwig cross coupling of 6-aminochrysene with protected C8-bromo-dG derivative served as the key reaction to furnish protected N-(dG-8-yl)-6-AC in 56% yield. The monomer for solid-phase DNA synthesis was prepared by its deprotection followed by conversion to the corresponding 5'-O-dimethoxytrityl 3'-phosphoramidite, which was used to synthesize a site-specifically adducted oligonucleotide. After purification and characterization, the adduct-containing oligonucleotide was incorporated into a plasmid and replicated in human embryonic kidney (HEK) 293T cells, which showed that N-(dG-8-yl)-6-AC stalls DNA replication as evidenced by 77% translesion synthesis (TLS) efficiency relative to the control and that the adduct is mutagenic (mutation frequency (MF) 17.8%) inducing largely G→T transversions. We also investigated the roles of several translesion synthesis DNA polymerases in the bypass of N-(dG-8-yl)-6-AC using siRNA knockdown approach. TLS efficiency was reduced in hPol η-, hPol κ-, hPol ζ-, and hREV1-deficient HEK 293T cells to 66%, 45%, 37%, and 32%, respectively. Notably, TLS efficiency was reduced to 18% in cells with concurrent knockdown of hPol κ, hPol ζ, and REV1, suggesting that these three polymerases play critical roles in bypassing N-(dG-8-yl)-6-AC. MF increased to 23.1% and 32.2% in hPol κ- and hREV1-deficient cells, whereas it decreased to 11.8% in hPol ζ-deficient cells. This suggests that hPol κ and hREV1 are involved in error-free TLS of this lesion, whereas hPol ζ performs error-prone bypass.

Coupling Genome-wide Transcriptomics and Developmental Toxicity Profiles in Zebrafish to Characterize Polycyclic Aromatic Hydrocarbon (PAH) Hazard.

Polycyclic Aromatic Hydrocarbons (PAHs) are diverse environmental pollutants associated with adverse human health effects. Many studies focus on the carcinogenic effects of a limited number of PAHs and there is an increasing need to understand mechanisms of developmental toxicity of more varied yet environmentally relevant PAHs. A previous study characterized the developmental toxicity of 123 PAHs in zebrafish. Based on phenotypic responses ranging from complete inactivity to acute mortality, we classified these PAHs into eight bins, selected 16 representative PAHs, and exposed developing zebrafish to the concentration of each PAH that induced 80% phenotypic effect. We conducted RNA sequencing at 48 h post fertilization to identify gene expression changes as a result of PAH exposure. Using the Context Likelihood of Relatedness algorithm, we inferred a network that links the PAHs based on coordinated gene responses to PAH exposure. The 16 PAHs formed two broad clusters: Cluster A was transcriptionally more similar to the controls, while Cluster B consisted of PAHs that were generally more developmentally toxic, significantly elevated cyp1a transcript levels, and induced Ahr2-dependent Cyp1a protein expression in the skin confirmed by gene-silencing studies. We found that cyp1a transcript levels were associated with transcriptomic response, but not with PAH developmental toxicity. While all cluster B PAHs predominantly activated Ahr2, they also each enriched unique pathways like ion transport signaling, which likely points to differing molecular events between the PAHs downstream of Ahr2. Thus, using a systems biology approach, we have begun to evaluate, classify, and define mechanisms of PAH toxicity.

Systematic developmental neurotoxicity assessment of a representative PAH Superfund mixture using zebrafish.

Superfund sites often consist of complex mixtures of polycyclic aromatic hydrocarbons (PAHs). It is widely recognized that PAHs pose risks to human and environmental health, but the risks posed by exposure to PAH mixtures are unclear. We constructed an environmentally relevant PAH mixture with the top 10 most prevalent PAHs (SM10) from a Superfund site derived from environmental passive sampling data. Using the zebrafish model, we measured body burden at 48 hours post fertilization (hpf) and evaluated the developmental and neurotoxicity of SM10 and the 10 individual constituents at 24 hours post fertilization (hpf) and 5 days post fertilization (dpf). Zebrafish embryos were exposed from 6 to 120 hpf to (1) the SM10 mixture, (2) a variety of individual PAHs: pyrene, fluoranthene, retene, benzo[a]anthracene, chrysene, naphthalene, acenaphthene, phenanthrene, fluorene, and 2-methylnaphthalene. We demonstrated that SM10 and only 3 of the individual PAHs were developmentally toxic. Subsequently, we constructed and exposed developing zebrafish to two sub-mixtures: SM3 (comprised of 3 of the developmentally toxicity PAHs) and SM7 (7 non-developmentally toxic PAHs). We found that the SM3 toxicity profile was similar to SM10, and SM7 unexpectedly elicited developmental toxicity unlike that seen with its individual components. The results demonstrated that the overall developmental toxicity in the mixtures could be explained using the general concentration addition model. To determine if exposures activated the AHR pathway, spatial expression of CYP1A was evaluated in the 10 individual PAHs and the 3 mixtures at 5 dpf. Results showed activation of AHR in the liver and vasculature for the mixtures and some individual PAHs. Embryos exposed to SM10 during development and raised in chemical-free water into adulthood exhibited decreased learning and responses to startle stimulus indicating that developmental SM10 exposures affect neurobehavior. Collectively, these results exemplify the utility of zebrafish to investigate the developmental and neurotoxicity of complex mixtures.

Impact of Natural Variations in Freeze-Drying Parameters on Product Temperature History: Application of Quasi Steady-State Heat and Mass Transfer and Simple Statistics.

Inter- and intra-batch variability in heat and mass transfer during the drying phase of lyophilization is well recognized. Heat transfer variability between individual vials in the same batch arise from both different positions in the vial array and from variations in the bottom contour of the vials, both effects contributing roughly equally to variations in the effective heat transfer coefficient of the vials, Kv. Both effects can be measured in the laboratory, and variations in average Kv values as a function of vial position in the array for lab and production can be calculated by use of the simple steady-state heat and mass transfer theory. Typically, in the laboratory dryer, vials on the edge of the array, "edge vials," run 2-4°C warmer than "center vials," but differences between laboratory and manufacturing temperatures are modest. The variability in mass transfer can be assigned to major variations in ice nucleation temperature (both intra-batch and inter-batch), including major differences between laboratory and manufacturing. The net effect of all random variations, for each class of vial, can be evaluated by a simple statistical model-propagation of error, which then allows prediction of the distribution in product temperatures and drying times, and therefore prediction of percent of vials dry and percent of vials collapsed and proximity to the edge of failure for a given process. Good agreement between theoretical and experimentally determined maximum temperatures in primary drying and percent collapsed product demonstrates the calculations have useful accuracy.

Translesion Synthesis of 2'-Deoxyguanosine Lesions by Eukaryotic DNA Polymerases.

With the discovery of translesion synthesis DNA polymerases, great strides have been made in the last two decades in understanding the mode of replication of various DNA lesions in prokaryotes and eukaryotes. A database search indicated that approximately 2000 articles on this topic have been published in this period. This includes research involving genetic and structural studies as well as in vitro experiments using purified DNA polymerases and accessory proteins. It is a daunting task to comprehend this exciting and rapidly emerging area of research. Even so, as the majority of DNA damage occurs at 2'-deoxyguanosine residues, this perspective attempts to summarize a subset of this field, focusing on the most relevant eukaryotic DNA polymerases responsible for their bypass.

Mutagenicity of a Model DNA-Peptide Cross-Link in Human Cells: Roles of Translesion Synthesis DNA Polymerases.

DNA-protein cross-links are formed upon exposure of cellular DNA to various agents, including antitumor drugs, UV light, transition metals, and reactive oxygen species. They are thought to contribute to cancer, aging, and neurodegenerative diseases. It has been proposed that DNA-protein cross-links formed in cells are subject to proteolytic degradation to the corresponding DNA-peptide cross-links (DpCs). To investigate the effects of DpCs on DNA replication, we have constructed plasmid DNA containing a 10-mer Myc peptide covalently linked to C7 of 7-deaza-dG, a hydrolytically stable mimic of N7-dG lesions. Following transfection in human embryonic kidney cells (HEK 293T), progeny plasmids were recovered and sequenced. Translesion synthesis (TLS) past DpC was 76% compared to that of the unmodified control. The DpC induced 20% targeted G → A and G → T plus 15% semitargeted mutations, notably at a guanine (G5) five bases 3' to the lesion site. Proteolytic digestion of the DpC reduced the mutation frequency considerably, indicating that the covalently attached 10-mer peptide was responsible for the observed mutations. TLS efficiency and targeted mutations were reduced upon siRNA knockdown of pol η, pol κ, or pol ζ, indicating that they participate in error-prone bypass of the DpC lesion. However, the semitargeted mutation at G5 was only reduced upon knockdown of pol ζ, suggesting its critical role in this type of mutations. Our results indicate that DpCs formed at the N7 position of guanine can induce both targeted and semitargeted mutations in human cells and that the TLS polymerases play a critical role in their error-prone bypass.

Controlling the Graphene-Bio Interface: Dispersions in Animal Sera for Enhanced Stability and Reduced Toxicity.

Liquid phase exfoliation of graphite in six different animal sera and evaluation of its toxicity are reported here. Previously, we reported the exfoliation of graphene using proteins, and here we extend this approach to complex animal fluids. A kitchen blender with a high-turbulence flow gave high quality and maximum exfoliation efficiency in all sera tested, when compared to the values found with shear and ultrasonication methods. Raman spectra and electron microscopy confirmed the formation of three- or four-layer, submicrometer size graphene, independent of the serum used. Graphene prepared in serum was directly transferred to cell culture media without post-treatments. Contrary to many reports, a nanotoxicity study of this graphene fully dispersed to human embryonic kidney cells, human lung cancer cells, and nematodes (Caenorhabditis elegans) showed no acute toxicity for up to 7 days at various doses (50-500 µg/mL), but prolonged exposure at higher doses (300-500 µg/mL, 10-15 days) showed cytotoxicity to cells (∼95% death) and reproductive toxicity to C. elegans (5-10% reduction in brood size). The origin of toxicity was found to be due to the highly fragmented smaller graphene sheets (<200 nm), while the larger sheets were nontoxic (50-300 µg/mL dose). In contrast, graphene produced with sodium cholate as the mediator has been found to be cytotoxic to these cells at these dosages. We demonstrated the toxicity of liquid phase exfoliated graphene is attributed to highly fragmented fractions or nonbiocompatible exfoliating agents. Thus, low-toxicity graphene/serum suspensions are produced by a facile method in biological media, and this approach may accelerate the much-anticipated development of graphene for biological applications.

Intraoperative optical coherence tomography for soft tissue sarcoma differentiation and margin identification.

BACKGROUND AND OBJECTIVE: Sarcomas are rare but highly aggressive tumors, and local recurrence after surgical excision can occur in up to 50% cases. Therefore, there is a strong clinical need for accurate tissue differentiation and margin assessment to reduce incomplete resection and local recurrence. The purpose of this study was to investigate the use of optical coherence tomography (OCT) and a novel image texture-based processing algorithm to differentiate sarcoma from muscle and adipose tissue. STUDY DESIGN AND METHODS: In this study, tumor margin delineation in 19 feline and canine veterinary patients was achieved with intraoperative OCT to help validate tumor resection. While differentiation of lower-scattering adipose tissue from higher-scattering muscle and tumor tissue was relatively straightforward, it was more challenging to distinguish between dense highly scattering muscle and tumor tissue types based on scattering intensity and microstructural features alone. To improve tissue-type differentiation in a more objective and automated manner, three descriptive statistical metrics, namely the coefficient of variation (CV), standard deviation (STD), and Range, were implemented in a custom algorithm applied to the OCT images. RESULTS: Over 22,800 OCT images were collected intraoperatively from over 38 sites on 19 ex vivo tissue specimens removed during sarcoma surgeries. Following the generation of an initial set of OCT images correlated with standard hematoxylin and eosin-stained histopathology, over 760 images were subsequently used for automated analysis. Using texture-based image processing metrics, OCT images of sarcoma, muscle, and adipose tissue were all found to be statistically different from one another (P ≤ 0.001). CONCLUSION: These results demonstrate the potential of using intraoperative OCT, along with an automated tissue differentiation algorithm, as a guidance tool for soft tissue sarcoma margin delineation in the operating room. Lasers Surg. Med. 49:240-248, 2017. © 2017 Wiley Periodicals, Inc.

DNA polymerases κ and ζ cooperatively perform mutagenic translesion synthesis of the C8-2'-deoxyguanosine adduct of the dietary mutagen IQ in human cells.

The roles of translesion synthesis (TLS) DNA polymerases in bypassing the C8-2'-deoxyguanosine adduct (dG-C8-IQ) formed by 2-amino-3-methylimidazo[4,5-f]quinoline (IQ), a highly mutagenic and carcinogenic heterocyclic amine found in cooked meats, were investigated. Three plasmid vectors containing the dG-C8-IQ adduct at the G1-, G2- or G3-positions of the NarI site (5'-G1G2CG3CC-3') were replicated in HEK293T cells. Fifty percent of the progeny from the G3 construct were mutants, largely G→T, compared to 18% and 24% from the G1 and G2 constructs, respectively. Mutation frequency (MF) of dG-C8-IQ was reduced by 38-67% upon siRNA knockdown of pol κ, whereas it was increased by 10-24% in pol η knockdown cells. When pol κ and pol ζ were simultaneously knocked down, MF of the G1 and G3 constructs was reduced from 18% and 50%, respectively, to <3%, whereas it was reduced from 24% to <1% in the G2 construct. In vitro TLS using yeast pol ζ showed that it can extend G3*:A pair more efficiently than G3*:C pair, but it is inefficient at nucleotide incorporation opposite dG-C8-IQ. We conclude that pol κ and pol ζ cooperatively carry out the majority of the error-prone TLS of dG-C8-IQ, whereas pol η is involved primarily in its error-free bypass.

Automated computational aberration correction method for broadband interferometric imaging techniques.

Numerical correction of optical aberrations provides an inexpensive and simpler alternative to the traditionally used hardware-based adaptive optics techniques. In this Letter, we present an automated computational aberration correction method for broadband interferometric imaging techniques. In the proposed method, the process of aberration correction is modeled as a filtering operation on the aberrant image using a phase filter in the Fourier domain. The phase filter is expressed as a linear combination of Zernike polynomials with unknown coefficients, which are estimated through an iterative optimization scheme based on maximizing an image sharpness metric. The method is validated on both simulated data and experimental data obtained from a tissue phantom, an ex vivo tissue sample, and an in vivo photoreceptor layer of the human retina.

Comparative Error-Free and Error-Prone Translesion Synthesis of N(2)-2'-Deoxyguanosine Adducts Formed by Mitomycin C and Its Metabolite, 2,7-Diaminomitosene, in Human Cells.

Mitomycin C (MC) is a cytotoxic and mutagenic antitumor agent that alkylates DNA upon reductive activation. 2,7-Diaminomitosene (2,7-DAM) is a major metabolite of MC in tumor cells, which also alkylates DNA. MC forms seven DNA adducts, including monoadducts and inter- and intrastrand cross-links, whereas 2,7-DAM forms two monoadducts. Herein, the biological effects of the dG-N(2) adducts formed by MC and 2,7-DAM have been compared by constructing single-stranded plasmids containing these adducts and replicating them in human embryonic kidney 293T cells. Translesion synthesis (TLS) efficiencies of dG-N(2)-MC and dG-N(2)-2,7-DAM were 38 ± 3 and 27 ± 3%, respectively, compared to that of a control plasmid. This indicates that both adducts block DNA synthesis and that dG-N(2)-2,7-DAM is a stronger replication block than dG-N(2)-MC. TLS of each adducted construct was reduced upon siRNA knockdown of pol η, pol κ, or pol ζ. For both adducts, the most significant reduction occurred with knockdown of pol κ, which suggests that pol κ plays a major role in TLS of these dG-N(2) adducts. Analysis of the progeny showed that both adducts were mutagenic, and the mutation frequencies (MF) of dG-N(2)-MC and dG-N(2)-2,7-DAM were 18 ± 3 and 10 ± 1%, respectively. For both adducts, the major type of mutation was G → T transversions. Knockdown of pol η and pol ζ reduced the MF of dG-N(2)-MC and dG-N(2)-2,7-DAM, whereas knockdown of pol κ increased the MF of these adducts. This suggests that pol κ predominantly carries out error-free TLS, whereas pol η and pol ζ are involved in error-prone TLS. The largest reduction in MF by 78 and 80%, respectively, for dG-N(2)-MC and dG-N(2)-2,7-DAM constructs occurred when pol η, pol ζ, and Rev1 were simultaneously knocked down. This result strongly suggests that, unlike pol κ, these three TLS polymerases cooperatively perform the error-prone TLS of these adducts.

Magnetomotive Optical Coherence Elastography for Magnetic Hyperthermia Dosimetry Based on Dynamic Tissue Biomechanics.

Magnetic nanoparticles (MNPs) have been used in many diagnostic and therapeutic biomedical applications over the past few decades to enhance imaging contrast, steer drugs to targets, and treat tumors via hyperthermia. Optical coherence tomography (OCT) is an optical biomedical imaging modality that relies on the detection of backscattered light to generate high-resolution cross-sectional images of biological tissue. MNPs have been utilized as imaging contrast and perturbative mechanical agents in OCT in techniques called magnetomotive OCT (MM-OCT) and magnetomotive elastography (MM-OCE), respectively. MNPs have also been independently used for magnetic hyperthermia treatments, enabling therapeutic functions such as killing tumor cells. It is well known that the localized tissue heating during hyperthermia treatments result in a change in the biomechanical properties of the tissue. Therefore, we propose a novel dosimetric technique for hyperthermia treatment based on the viscoelasticity change detected by MM-OCE, further enabling the theranostic function of MNPs. In this paper, we first review the basic principles and applications of MM-OCT, MM-OCE, and magnetic hyperthermia, and present new preliminary results supporting the concept of MM-OCE-based hyperthermia dosimetry.

Unlike catalyzing error-free bypass of 8-oxodGuo, DNA polymerase λ is responsible for a significant part of Fapy·dG-induced G → T mutations in human cells.

8-OxodGuo and Fapy·dG induced 10-22% mutations, predominantly G → T transversions, in human embryonic kidney 293T cells in four TG*N sequence contexts, where N = C, G, A, or T. siRNA knockdown of pol λ resulted in 34 and 55% increases in the level of mutations in the progeny from the 8-oxodGuo construct in the TG*T and TG*G sequences, respectively, suggesting that pol λ is involved in error-free bypass of 8-oxodGuo. For Fapy·dG, in contrast, the level of G → T mutations was reduced by 27 and 46% in the TG*T and TG*G sequences, respectively, suggesting that pol λ is responsible for a significant fraction of Fapy·dG-induced G → T mutations.

Mutational analysis of the C8-guanine adduct of the environmental carcinogen 3-nitrobenzanthrone in human cells: critical roles of DNA polymerases η and κ and Rev1 in error-prone translesion synthesis.

3-Nitrobenzanthrone (3-NBA), a potent mutagen and suspected human carcinogen, is a common environmental pollutant. The genotoxicity of 3-NBA has been associated with its ability to form DNA adducts, including N-(2'-deoxyguanosin-8-yl)-3-aminobenzanthrone (C8-dG-ABA). To investigate the molecular mechanism of C8-dG-ABA mutagenesis in human cells, we have replicated a plasmid containing a single C8-dG-ABA in human embryonic kidney 293T (HEK293T) cells, which yielded 14% mutant progeny. The major types of mutations induced by C8-dG-ABA were G→T>G→A>G→C. siRNA knockdown of the translesion synthesis (TLS) DNA polymerases (pols) in HEK293T cells indicated that pol η, pol κ, pol ι, pol ζ, and Rev1 each have a role in replication across this adduct. The extent of TLS was reduced with each pol knockdown, but the largest decrease (of ∼55% reduction) in the level of TLS occurred in cells with knockdown of pol ζ. Pol η and pol κ were considered the major contributors of the mutagenic TLS, because the mutation frequency (MF) decreased by 70%, when these pols were simultaneously knocked down. Rev1 also is important for mutagenesis, as reflected by the 60% reduction in MF upon Rev1 knockdown, but it probably plays a noncatalytic role by physically interacting with the other two Y-family pols. In contrast, pol ζ appeared to be involved in the error-free bypass of the lesion, because MF increased by 60% in pol ζ knockdown cells. These results provide important mechanistic insight into the bypass of the C8-dG-ABA adduct.

Fine-tuning TrailMap: The utility of transfer learning to improve the performance of deep learning in axon segmentation of light-sheet microscopy images.

Light-sheet microscopy has made possible the 3D imaging of both fixed and live biological tissue, with samples as large as the entire mouse brain. However, segmentation and quantification of that data remains a time-consuming manual undertaking. Machine learning methods promise the possibility of automating this process. This study seeks to advance the performance of prior models through optimizing transfer learning. We fine-tuned the existing TrailMap model using expert-labeled data from noradrenergic axonal structures in the mouse brain. By changing the cross-entropy weights and using augmentation, we demonstrate a generally improved adjusted F1-score over using the originally trained TrailMap model within our test datasets.

Fine-tuning TrailMap: The utility of transfer learning to improve the performance of deep learning in axon segmentation of light-sheet microscopy images.

Light-sheet microscopy has made possible the 3D imaging of both fixed and live biological tissue, with samples as large as the entire mouse brain. However, segmentation and quantification of that data remains a time-consuming manual undertaking. Machine learning methods promise the possibility of automating this process. This study seeks to advance the performance of prior models through optimizing transfer learning. We fine-tuned the existing TrailMap model using expert-labeled data from noradrenergic axonal structures in the mouse brain. By fine-tuning the final two layers of the neural network at a lower learning rate of the TrailMap model, we demonstrate an improved recall and an occasionally improved adjusted F1-score within our test dataset over using the originally trained TrailMap model.