Industrialization of an Academic Miltenyi Prodigy-Based CAR T Process

INTRODUCTION We have previously described AUTO1, a CD19 CAR with a fast off-rate binding domain, designed to reduce CAR T-cell immune toxicity and improve engraftment. Clinical testing in two academic studies in relapsed/refractory (r/r) paediatric [NCT02443831; CARPALL] and adult B-ALL, B-NHL and B-CLL [NCT02935257; ALLCAR19] confirmed the intended function of the receptor, with low levels of cytokine release syndrome (CRS) and immune effector cell-associated neurotoxicity syndrome (ICANS) and long-term engraftment of CAR T-cell 1,2. Based on data in adult B-ALL, we initiated a phase Ib/II registration study in r/r adult B-ALL [NCT04404660; FELIX]. To facilitate this study and future commercialization, industrialization of the manufacturing process was required. METHODS & RESULTS The original process in the academic studies was based on the Miltenyi CliniMACS Prodigy T cell transduction process. Leukapheresis was performed at the same site as manufacture. T cells were isolated from pheresis by CD4/CD8 positive selection and seeded onto the Prodigy to be activated using the Miltenyi CD3/CD28 targeting activating reagent, TransAct. The following day, transduction was carried out using a lentiviral vector. Cells were cryopreserved after an expansion phase of up to day 10 of the process. To facilitate industrialization of the AUTO1 manufacture in the multi-center, multi-regional FELIX study, we first explored the use of cryopreserved pheresis (81.3% median viability pre-selection (range 71.9 - 94.3), 1.0 days median doubling time (range 0.9 - 1.5) and 47.6% median CD19 CAR expression (range 19.1-62.1)). We concluded that optimal manufacture includes the use of fresh pheresis and the initiation of manufacture within 72 hours (99.0% median viability pre-selection (range 92.5 - 99.7), 1.3 days median doubling time (range 1.1 - 2.1) and 69.3% median CD19 CAR expression (range 22.9-86.2)). To further simplify the process, we explored removal of the pre-selection step. Full-scale runs using starting material from 4 healthy donors were conducted to compare CD4/CD8 selected with unselected cells. On the day following activation, selected cells displayed a higher percentage of viable cells, defined as cPARP-FVS780- (median: 76.1%, range: 84.5-66.4) as compared to unselected cells (median: 52.2%, range: 43.6-59.0). In addition, selected cells demonstrated a median of 23-fold expansion (range: 20.0 - 29.1) compared to a 13.3-fold expansion for unselected cells (range 6.1-17.4). Median transduction efficiencies of viable CAR+ T-cells were 53.9% (range: 43.2-56.9) and 78.0% (range: 64.5-81.1) in selected and unselected cells, respectively. CD4/8 pre-selection was determined to be a critical part of the process. A comparison of phenotype between 18 batches manufactured using the academic process and 5 batches produced from fresh material using the industrial process was carried out. No significant differences, as determined by 2-way ANOVA, were observed between the percentage of CAR+ CD3+ cells, the memory phenotype (% TSC/naive, % TCM, % TEM and % TEMRA) and the percentage expression of PD1 (figure 1). The CD4/CD8 ratio was also comparable between products of the two processes. Data from the initial 6 fresh in patients show that engraftment in the FELIX study is consistent with ALLCAR19 engraftment results. Additional patients, updated clinical data and longer follow-up will be presented at the conference. CONCLUSION Industrialization of an autologous Miltenyi CAR T process is feasible, leading to a comparable product to that manufactured in an academic setting. We have now opened the pivotal multi-center phase II part of the FELIX study in r/r adult B-ALL patients. REFERENCES Ghorashian S et al. (2019) Enhanced CAR T cell expansion and prolonged persistence in pediatric patients with ALL treated with a low-affinity CD19 CAR. Nat Med, 25(9):1408-1414.Roddie C et al. (2021) Durable responses and low toxicity after fast off-rate CD19 CAR-T therapy in adults with relapsed/ refractory B-ALL. J Clin Oncol, in press Figure 1. Comparison of phenotype between 18 CAR T cell batches manufactured using the academic process and 5 batches produced using the industrial process. Boxes represent median, 25th and 75th percentiles and whiskers represent minimum and maximum. Figure 1 Figure 1. Disclosures Culshaw: Autolus Ltd.: Current Employment. Arce Vargas: Autolus Ltd.: Current Employment. Santiago Toledo: Autolus Ltd.: Current Employment. Roddie: Novartis: Consultancy; Celgene: Consultancy, Speakers Bureau; Gilead: Consultancy, Speakers Bureau. Shaughnessy: BMS: Honoraria, Speakers Bureau; Sanofi: Honoraria, Speakers Bureau; Kite: Honoraria, Speakers Bureau. Cerec: Autolus Ltd.: Current Employment. Duffy: Autolus Ltd.: Current Employment. Perna: Autolus Ltd.: Current Employment. Brugger: Autolus Ltd.: Current Employment. Merges: Autolus Ltd.: Current Employment. Pule: Autolus Ltd: Current Employment.

A Multimethodology for Supporting the Boundary Selection of System Dynamics Models

System dynamics models can produce knowledge for decision-makers and, consequently, provide better choices. To be effective in its purpose, a model must reproduce an observed problem situation effectively. Hence, the compatibility between the observed problem situation and the created model is essential and represents a considerable challenge. In this context, this paper aims to describe an adaptation of the problem structuring method ‘Strategic Options Development and Analysis’ (SODA), used in the Problem Articulation (Boundary Selection) step of the system dynamics modelling process. In summary, this adaptation consists of: (1) Selecting of stakeholders; (2) Capturing, aggregating and interpreting the insights using cognitive and causal maps, and (3) Using the interpretation of the causal maps for building a system dynamics model. The method proved to be satisfactory since it was able to direct the construction of a System Dynamics model based on a problem situation perceived by stakeholders acting in the native forests of the state of Santa Catarina, Brazil.

A Multimethodology for Supporting the Boundary Selection of System Dynamics Models

System dynamics models can produce knowledge for decision-makers and, consequently, provide better choices. To be effective in its purpose, a model must reproduce an observed problem situation effectively. Hence, the compatibility between the observed problem situation and the created model is essential and represents a considerable challenge. In this context, this paper aims to describe an adaptation of the problem structuring method ‘Strategic Options Development and Analysis’ (SODA), used in the Problem Articulation (Boundary Selection) step of the system dynamics modelling process. In summary, this adaptation consists of: (1) Selecting of stakeholders; (2) Capturing, aggregating and interpreting the insights using cognitive and causal maps, and (3) Using the interpretation of the causal maps for building a system dynamics model. The method proved to be satisfactory since it was able to direct the construction of a System Dynamics model based on a problem situation perceived by stakeholders acting in the native forests of the state of Santa Catarina, Brazil.

Diagnosis of Leaf Surface Disease Using Two Datasets of Tomato and Rice Obtained from Image Processing Techniques

It is of a great importance in modern agriculture to study fast, automatic, inexpensive and accurate method of diagnosing plant diseasesTherefore, timely and accurately diagnosis of the disease in the fields is one of the most important factors in dealing with plant diseases. For this reason, in the present study, the image processing method study, has been examined for diagnosing the two important diseases of rice and tomato, brown spots and leaf blasts. In this study, firstly the data section is treated using improved k-means segmentation, after preprocessing. Secondly, comprehensive features are extracted and the disease areas are demarcated. An improved genetic algorithm is used in the feature selection step. Finally, images are categorized using the k-nearest neighbor’s algorithm (k-NN) classifier. The accuracy of the proposed method for the rice data set is 99.12 and for the tomato data set is 97.29, which shows a very good performance compared to other methods.

Pre-selection against a lethal recessive allele in breeding schemes with optimum-contribution selection or truncation selection

Background We tested the hypothesis that breeding schemes with a pre-selection step, in which carriers of a lethal recessive allele (LRA) were culled, and with optimum-contribution selection (OCS) reduce the frequency of a LRA, control rate of inbreeding, and realise as much genetic gain as breeding schemes without a pre-selection step. Methods We used stochastic simulation to estimate true genetic gain realised at a 0.01 rate of true inbreeding (ΔFtrue) by breeding schemes that combined one of four pre-selection strategies with one of three selection strategies. The four pre-selection strategies were: (1) no carriers culled, (2) male carriers culled, (3) female carriers culled, and (4) all carriers culled. Carrier-status was known prior to selection. The three selection strategies were: (1) OCS in which $$\Delta {\text{F}}_{{{\text{true}}}}$$ Δ F true was predicted and controlled using pedigree relationships (POCS), (2) OCS in which $$\Delta {\text{F}}_{{{\text{true}}}}$$ Δ F true was predicted and controlled using genomic relationships (GOCS), and (3) truncation selection of parents. All combinations of pre-selection strategies and selection strategies were tested for three starting frequencies of the LRA (0.05, 0.10, and 0.15) and two linkage statuses with the locus that has the LRA being on a chromosome with or without loci affecting the breeding goal trait. The breeding schemes were simulated for 10 discrete generations (t = 1, …, 10). In all breeding schemes, ΔFtrue was calibrated to be 0.01 per generation in generations t = 4, …, 10. Each breeding scheme was replicated 100 times. Results We found no significant difference in true genetic gain from generations t = 4, …, 10 between breeding schemes with or without pre-selection within selection strategy. POCS and GOCS schemes realised similar true genetic gains from generations t = 4, …, 10. POCS and GOCS schemes realised 12% more true genetic gain from generations t = 4, …, 10 than truncation selection schemes. Conclusions We advocate for OCS schemes with pre-selection against the LRA that cause animal suffering and high costs. At LRA frequencies of 0.10 or lower, OCS schemes in which male carriers are culled reduce the frequency of LRA, control rate of inbreeding, and realise no significant reduction in true genetic gain compared to OCS schemes without pre-selection against LRA.

Medical Image Classification Using the Discriminant Power Analysis (DPA) of Discrete Cosine Transform (DCT) Coefficients

Medical imaging systems are very important in medicine domain. They assist specialists to make the final decision about the patient’s condition, and strongly help in early cancer detection. The classification of mammogram images represents a very important operation to identify whether the breast cancer is benign or malignant. In this chapter, we propose a new computer aided diagnostic (CAD) system, which is composed of three steps. In the first step, the input image is pre-processed to remove the noise and artifacts and also to separate the breast profile from the pectoral muscle. This operation is a difficult task that can affect the final decision. For this reason, a hybrid segmentation method using the seeded region growing (SRG) algorithm applied on a localized triangular region has been proposed. In the second step, we have proposed a features extraction method based on the discrete cosine transform (DCT), where the processed images of the breast profiles are transformed by the DCT where the part containing the highest energy value is selected. Then, in the feature’s selection step, a new most discriminative power coefficients algorithm has been proposed to select the most significant features. In the final step of the proposed system, we have used the most known classifiers in the field of the image classification for evaluation. An effective classification has been made using the Support Vector Machines (SVM), Naive Bayes (NB), Artificial Neural Network (ANN) and k-Nearest Neighbors (KNN) classifiers. To evaluate the efficiency and to measure the performances of the proposed CAD system, we have selected the mini Mammographic Image Analysis Society (MIAS) database. The obtained results show the effectiveness of the proposed algorithm over others, which are recently proposed in the literature, whereas the new CAD reached an accuracy of 100%, in certain cases, with only a small set of selected features.

Addressing cluster-constant covariates in mixed effects models via likelihood-based boosting techniques

Boosting techniques from the field of statistical learning have grown to be a popular tool for estimating and selecting predictor effects in various regression models and can roughly be separated in two general approaches, namely gradient boosting and likelihood-based boosting. An extensive framework has been proposed in order to fit generalized mixed models based on boosting, however for the case of cluster-constant covariates likelihood-based boosting approaches tend to mischoose variables in the selection step leading to wrong estimates. We propose an improved boosting algorithm for linear mixed models, where the random effects are properly weighted, disentangled from the fixed effects updating scheme and corrected for correlations with cluster-constant covariates in order to improve quality of estimates and in addition reduce the computational effort. The method outperforms current state-of-the-art approaches from boosting and maximum likelihood inference which is shown via simulations and various data examples.

Using k-mer Embeddings Learned from a Skip-Gram Based Neural Network as Effective Feature Representation for Building a Cross-Species Prediction Model to Identify DNA N6-Methyladenine Sites in Plant Genomes

Identification of DNA N6-methyladenine sites has been a very active topic of computational biology due to the unavailability of suitable methods to identify them accurately, especially in plants. Substantial results were obtained with a great effort put in extracting, heuristic searching, or fusing a diverse types of features, not to mention a feature selection step. We considered DNA, the human life book, as a book corpus for training DNA language models. K-mer embeddings then were generated from Skipgram neural networks and input into several ensemble tree-based algorithms. We trained the prediction model on Rosaceae genome dataset and performed a comprehensive test on 3 plant genome datasets. Our proposed method shows promising performance with AUC performance approaching an ideal value on Rosaceae dataset (0.99), a high score on Rice dataset (0.95) and improved performance on Rice dataset while enjoying an elegant, yet efficient feature extraction process.

A New Breeding Strategy towards Introgression and Characterization of Stay-Green QTL for Drought Tolerance in Sorghum

Several marker-assisted selection (MAS) or backcrossing (MAB) approaches exist for polygenic trait improvement. However, the implementation of MAB remains a challenge in many breeding programs, especially in the public sector. In MAB introgression programs, which usually do not include phenotypic selection, undesired donor traits may unexpectedly turn up regardless of how expensive and theoretically powerful a backcross scheme may be. Therefore, combining genotyping and phenotyping during selection will improve understanding of QTL interactions with the environment, especially for minor alleles that maximize the phenotypic expression of the traits. Here, we describe the introgression of stay-green QTL (Stg1–Stg4) from B35 into two sorghum backgrounds through an MAB that combines genotypic and phenotypic (C-MAB) selection during early backcross cycles. The background selection step is excluded. Since it is necessary to decrease further the cost associated with molecular marker assays, the costs of C-MAB were estimated. Lines with stay-green trait and good performance were identified at an early backcross generation, backcross two (BC2). Developed BC2F4 lines were evaluated under irrigated and drought as well as three rainfed environments varied in drought timing and severity. Under drought conditions, the mean grain yield of the most C-MAB-introgression lines was consistently higher than that of the recurrent parents. This study is one of the real applications of the successful use of C-MAB for the development of drought-tolerant sorghum lines for drought-prone areas.

Using Unique ORFan Genes as Strain-specific Identifiers for Escherichia Coli

BackgroundBacterial identification at the strain level is a much-needed, but arduous and challenging task. This study aimed to develop a method for identifying and differentiating individual strains among multiple strains of the same bacterial species. The set used for testing the method consisted of 17 Escherichia coli strains picked from a collection of strains isolated in Germany, Spain, the United Kingdom and Vietnam from humans, cattle, swine, wild boars, and chickens. We targeted unique or rare ORFan genes to address the problem of selective and specific strain identification. These ORFan genes, exclusive to each strain, served as templates for developing strain-specific primers.ResultsMost of the strains to be deployed experimentally (14 out of 17) possessed unique ORFan genes that were used to develop strain-specific primers. The remaining three strains were identified by combining a PCR for a rare gene with a selection step for isolating the experimental strains. Multiplex PCR allowed the successful identification of the strains both in vitro in spiked faecal material in addition to in vivo after experimental infections of pigs and recovery of bacteria from faecal material. In addition, primers for qPCR were also developed and quantitative readout from faecal samples after experimental infection was also possible.ConclusionsThe method described in this manuscript using strain-specific unique genes to identify single strains in a mixture of strains proved itself efficient and reliable in detecting and following individual strains both in vitro and in vivo, representing a fast and inexpensive alternative to more costly methods.