Biological differences between adult and perinatal human mesenchymal stromal cells and their impact on the manufacturing processes.

The biological properties of human mesenchymal stromal cells (hMSCs) have been explored in over a thousand clinical trials in the last decade. Although hMSCs can be isolated from multiple sources, the degree of biological similarity between cell populations from these sources remains to be determined. A comparative study was performed investigating the growth kinetics and functionality of hMSCs isolated from adipose tissue (AT), bone marrow (BM) and umbilical cord tissue (UCT) expanded in monolayer over five passages. Adult hMSCs (AT, BM) had a slower proliferation ability than the UCT-hMSCs, with no apparent differences in their glucose consumption profile. BM-hMSCs produced higher concentrations of endogenous vascular endothelial growth factor (VEGF) compared to AT- and UCT-hMSCs. This study also revealed that UCT-hMSCs were more efficiently transduced by a lentiviral vector carrying a VEGF gene than their adult counterparts. Following cellular immunophenotypic characterization, no differences across the sources were found in the expression levels of the typical markers used to identify hMSCs. This work established a systematic approach for cell source selection depending on the hMSC's intended clinical application.

Can we predict pathology without surgery? Weighing the added value of multiparametric MRI and whole prostate radiomics in integrative machine learning models.

OBJECTIVE: To test the ability of high-performance machine learning (ML) models employing clinical, radiological, and radiomic variables to improve non-invasive prediction of the pathological status of prostate cancer (PCa) in a large, single-institution cohort. METHODS: Patients who underwent multiparametric MRI and prostatectomy in our institution in 2015-2018 were considered; a total of 949 patients were included. Gradient-boosted decision tree models were separately trained using clinical features alone and in combination with radiological reporting and/or prostate radiomic features to predict pathological T, pathological N, ISUP score, and their change from preclinical assessment. Model behavior was analyzed in terms of performance, feature importance, Shapley additive explanation (SHAP) values, and mean absolute error (MAE). The best model was compared against a naïve model mimicking clinical workflow. RESULTS: The model including all variables was the best performing (AUC values ranging from 0.73 to 0.96 for the six endpoints). Radiomic features brought a small yet measurable boost in performance, with the SHAP values indicating that their contribution can be critical to successful prediction of endpoints for individual patients. MAEs were lower for low-risk patients, suggesting that the models find them easier to classify. The best model outperformed (p ≤ 0.0001) clinical baseline, resulting in significantly fewer false negative predictions and overall was less prone to under-staging. CONCLUSIONS: Our results highlight the potential benefit of integrative ML models for pathological status prediction in PCa. Additional studies regarding clinical integration of such models can provide valuable information for personalizing therapy offering a tool to improve non-invasive prediction of pathological status. CLINICAL RELEVANCE STATEMENT: The best machine learning model was less prone to under-staging of the disease. The improved accuracy of our pathological prediction models could constitute an asset to the clinical workflow by providing clinicians with accurate pathological predictions prior to treatment. KEY POINTS: • Currently, the most common strategies for pre-surgical stratification of prostate cancer (PCa) patients have shown to have suboptimal performances. • The addition of radiological features to the clinical features gave a considerable boost in model performance. Our best model outperforms the naïve model, avoiding under-staging and resulting in a critical advantage in the clinic. •Machine learning models incorporating clinical, radiological, and radiomics features significantly improved accuracy of pathological prediction in prostate cancer, possibly constituting an asset to the clinical workflow.

Post-operative KEloids iRradiation (POKER): does the surgery/high-dose interventional radiotherapy association make a winning hand?

PURPOSE: To report the results involving post-operative interventional radiotherapy (POIRT) in a homogenous cohort of patients affected by keloid and treated at a single institution with the same fractionation schedule. PATIENTS AND METHODS: Inclusion criteria were: surgery with a histopathological diagnosis of keloid, subsequent high-dose rate interventional radiotherapy (HDR-IRT)-12 Gy in 4 fractions (3 Gy/fr) twice a day-and follow-up period ≥ 24 months. RESULTS: One-hundred and two patients and a total of 135 keloids were eligible for the analyses. Median follow-up was 64 [IQR: 25-103] months. Thirty-six (26.7%) recurrences were observed, 12-months and 36-months cumulative incidence of recurrence were 20.7% (95% CI 12.2-28.5) and 23.8% (95% CI 14.9-31.7) respectively. History of spontaneous keloids (HR = 7.00, 95% CI 2.79-17.6, p < 0.001), spontaneous cheloid as keloid cause (HR = 6.97, 95% CI 2.05-23.7, p = 0.002) and sternal (HR = 10.6, 95% CI 3.08-36.8, p < 0.001), ear (HR = 6.03, 95% CI 1.71-21.3, p = 0.005) or limb (HR = 18.8, 95% CI 5.14-68.7, p < 0.001) keloid sites were significantly associated to a higher risk of recurrence. CONCLUSIONS: The findings support the use of surgery and POIRT as an effective strategy for controlling keloid relapses. Further studies should focus on determining the optimal Biologically Effective Dose and on establishing a scoring system for patient selection.

Scalable manufacturing of gene-modified human mesenchymal stromal cells with microcarriers in spinner flasks.

Due to their immunomodulatory properties and in vitro differentiation ability, human mesenchymal stromal cells (hMSCs) have been investigated in more than 1000 clinical trials over the last decade. Multiple studies that have explored the development of gene-modified hMSC-based products are now reaching early stages of clinical trial programmes. From an engineering perspective, the challenge lies in developing manufacturing methods capable of producing sufficient doses of ex vivo gene-modified hMSCs for clinical applications. This work demonstrates, for the first time, a scalable manufacturing process using a microcarrier-bioreactor system for the expansion of gene-modified hMSCs. Upon isolation, umbilical cord tissue mesenchymal stromal cells (UCT-hMSCs) were transduced using a lentiviral vector (LV) with green fluorescent protein (GFP) or vascular endothelial growth factor (VEGF) transgenes. The cells were then seeded in 100 mL spinner flasks using Spherecol microcarriers and expanded for seven days. After six days in culture, both non-transduced and transduced cell populations attained comparable maximum cell concentrations (≈1.8 × 105 cell/mL). Analysis of the culture supernatant identified that glucose was fully depleted after day five across the cell populations. Lactate concentrations observed throughout the culture reached a maximum of 7.5 mM on day seven. Immunophenotype analysis revealed that the transduction followed by an expansion step was not responsible for the downregulation of the cell surface receptors used to identify hMSCs. The levels of CD73, CD90, and CD105 expressing cells were above 90% for the non-transduced and transduced cells. In addition, the expression of negative markers (CD11b, CD19, CD34, CD45, and HLA-DR) was also shown to be below 5%, which is aligned with the criteria established for hMSCs by the International Society for Cell and Gene Therapy (ISCT). This work provides a foundation for the scalable manufacturing of gene-modified hMSCs which will overcome a significant translational and commercial bottleneck. KEY POINTS: • hMSCs were successfully transduced by lentiviral vectors carrying two different transgenes: GFP and VEGF • Transduced hMSCs were successfully expanded on microcarriers using spinner flasks during a period of 7 days • The genetic modification step did not cause any detrimental impact on the hMSC immunophenotype characteristics.

Can Ki-67 predict radiotherapy response in neuroendocrine tumors? Retrospective analysis of a monocentric series of patients.

BACKGROUND: The impact of radiotherapy (RT) in neuroendocrine neoplasms is still unknown, and outcomes could be improved by a better insight in RT response predictors. This retrospective analysis investigates the potential correlation between Ki-67 and RT response to evaluate its role as biological marker of radiosensitivity. MATERIAL AND METHODS: Data from patients treated at an Italian NET-referral center between 2015 and 2020 were retrieved. Inclusion criteria included: histologically-proven diagnosis of NEN, Ki-67 status, indication (symptomatic and/or ablative) and at least one post-RT radiological assessment. RESULTS: Forty-two patients and 63 different treatment lines were included. Primary tumors presented Ki-67 values < 3% in 21% of cases, between 3 and 20% in 45% and >20% in the remaining 33%. Almost all patients were metastatic at the time of RT, which was performed with symptomatic purpose in 43% of cases. At a median time of three months, a complete response on the target lesion was observed in nine cases (14%), a partial response in 17 (27%), stability in 23 (37%) and local progression in 14 (22%). With median FU of 22.8 months, OS does not show statistically significant differences among three Ki-67 groups. Considering all lines of therapy, the relationship between ORR and Ki-67, did not show statistically significant differences, even following adjustments for drug types and delivered RT doses. CONCLUSION: No association between Ki67 and local tumor response to RT could be observed in the present cohort, regardless of whether the evaluation was performed on a categorical or continuous scale.

Comparison of automated segmentation techniques for magnetic resonance images of the prostate.

BACKGROUND: Contouring of anatomical regions is a crucial step in the medical workflow and is both time-consuming and prone to intra- and inter-observer variability. This study compares different strategies for automatic segmentation of the prostate in T2-weighted MRIs. METHODS: This study included 100 patients diagnosed with prostate adenocarcinoma who had undergone multi-parametric MRI and prostatectomy. From the T2-weighted MR images, ground truth segmentation masks were established by consensus from two expert radiologists. The prostate was then automatically contoured with six different methods: (1) a multi-atlas algorithm, (2) a proprietary algorithm in the Syngo.Via medical imaging software, and four deep learning models: (3) a V-net trained from scratch, (4) a pre-trained 2D U-net, (5) a GAN extension of the 2D U-net, and (6) a segmentation-adapted EfficientDet architecture. The resulting segmentations were compared and scored against the ground truth masks with one 70/30 and one 50/50 train/test data split. We also analyzed the association between segmentation performance and clinical variables. RESULTS: The best performing method was the adapted EfficientDet (model 6), achieving a mean Dice coefficient of 0.914, a mean absolute volume difference of 5.9%, a mean surface distance (MSD) of 1.93 pixels, and a mean 95th percentile Hausdorff distance of 3.77 pixels. The deep learning models were less prone to serious errors (0.854 minimum Dice and 4.02 maximum MSD), and no significant relationship was found between segmentation performance and clinical variables. CONCLUSIONS: Deep learning-based segmentation techniques can consistently achieve Dice coefficients of 0.9 or above with as few as 50 training patients, regardless of architectural archetype. The atlas-based and Syngo.via methods found in commercial clinical software performed significantly worse (0.855[Formula: see text]0.887 Dice).

Quality assurance for automatically generated contours with additional deep learning.

OBJECTIVE: Deploying an automatic segmentation model in practice should require rigorous quality assurance (QA) and continuous monitoring of the model's use and performance, particularly in high-stakes scenarios such as healthcare. Currently, however, tools to assist with QA for such models are not available to AI researchers. In this work, we build a deep learning model that estimates the quality of automatically generated contours. METHODS: The model was trained to predict the segmentation quality by outputting an estimate of the Dice similarity coefficient given an image contour pair as input. Our dataset contained 60 axial T2-weighted MRI images of prostates with ground truth segmentations along with 80 automatically generated segmentation masks. The model we used was a 3D version of the EfficientDet architecture with a custom regression head. For validation, we used a fivefold cross-validation. To counteract the limitation of the small dataset, we used an extensive data augmentation scheme capable of producing virtually infinite training samples from a single ground truth label mask. In addition, we compared the results against a baseline model that only uses clinical variables for its predictions. RESULTS: Our model achieved a mean absolute error of 0.020 ± 0.026 (2.2% mean percentage error) in estimating the Dice score, with a rank correlation of 0.42. Furthermore, the model managed to correctly identify incorrect segmentations (defined in terms of acceptable/unacceptable) 99.6% of the time. CONCLUSION: We believe that the trained model can be used alongside automatic segmentation tools to ensure quality and thus allow intervention to prevent undesired segmentation behavior.

Design and development of a new ambr250® bioreactor vessel for improved cell and gene therapy applications.

The emergence of cell and gene therapies has generated significant interest in their clinical and commercial potential. However, these therapies are prohibitively expensive to manufacture and can require extensive time for development due to our limited process knowledge and understanding. The automated ambr250® stirred-tank bioreactor platform provides an effective platform for high-throughput process development. However, the original dual pitched-blade 20 mm impeller and baffles proved sub-optimal for cell therapy candidates that require suspension of microcarriers (e.g. for the culture of adherent human mesenchymal stem cells) or other particles such as activating Dynabeads® (e.g. for the culture of human T-cells). We demonstrate the development of a new ambr250® stirred-tank bioreactor vessel which has been designed specifically to improve the suspension of microcarriers/beads and thereby improve the culture of such cellular systems. The new design is unbaffled and has a single, larger elephant ear impeller. We undertook a range of engineering and physical characterizations to determine which vessel and impeller configuration would be most suitable for suspension based on the minimum agitation speed (NJS) and associated specific power input (P/V)JS. A vessel (diameter, T, = 60 mm) without baffles and incorporating a single elephant ear impeller (diameter 30 mm and 45° pitch-blade angle) was selected as it had the lowest (P/V)JS and therefore potentially, based on Kolmogorov concepts, was the most flexible system. These experimentally-based conclusions were further validated firstly with computational fluid dynamic (CFD) simulations and secondly experimental studies involving the culture of both T-cells with Dynabeads® and hMSCs on microcarriers. The new ambr250® stirred-tank bioreactor successfully supported the culture of both cell types, with the T-cell culture demonstrating significant improvements compared to the original ambr250® and the hMSC-microcarrier culture gave significantly higher yields compared with spinner flask cultures. The new ambr250® bioreactor vessel design is an effective process development tool for cell and gene therapy candidates and potentially for autologous manufacture too.

Demonstrating the Manufacture of Human CAR-T Cells in an Automated Stirred-Tank Bioreactor.

Chimeric antigen receptor T-cell (CAR-T) therapies have proven clinical efficacy for the treatment of hematological malignancies. However, CAR-T cell therapies are prohibitively expensive to manufacture. The authors demonstrate the manufacture of human CAR-T cells from multiple donors in an automated stirred-tank bioreactor. The authors successfully produced functional human CAR-T cells from multiple donors under dynamic conditions in a stirred-tank bioreactor, resulting in overall cell yields which were significantly better than in static T-flask culture. At agitation speeds of 200 rpm and greater (up to 500 rpm), the CAR-T cells are able to proliferate effectively, reaching viable cell densities of >5 × 106 cells ml-1 over 7 days. This is comparable with current expansion systems and significantly better than static expansion platforms (T-flasks and gas-permeable culture bags). Importantly, engineered T-cells post-expansion retained expression of the CAR gene and retained their cytolytic function even when grown at the highest agitation intensity. This proves that power inputs used in this study do not affect cell efficacy to target and kill the leukemia cells. This is the first demonstration of human CAR-T cell manufacture in stirred-tank bioreactors and the findings present significant implications and opportunities for larger-scale allogeneic CAR-T production.

Establishing the scalable manufacture of primary human T-cells in an automated stirred-tank bioreactor.

Advanced cell and gene therapies such as chimeric antigen receptor T-cell immunotherapies (CAR-T), present a novel therapeutic modality for the treatment of acute and chronic conditions including acute lymphoblastic leukemia and non-Hodgkin lymphoma. However, the development of such immunotherapies requires the manufacture of large numbers of T-cells, which remains a major translational and commercial bottleneck due to the manual, small-scale, and often static culturing systems used for their production. Such systems are used because there is an unsubstantiated concern that primary T-cells are shear sensitive, or prefer static conditions, and therefore do not grow as effectively in more scalable, agitated systems, such as stirred-tank bioreactors, as compared with T-flasks and culture bags. In this study, we demonstrate that not only T-cells can be cultivated in an automated stirred-tank bioreactor system (ambr® 250), but that their growth is consistently and significantly better than that in T-flask static culture, with equivalent cell quality. Moreover, we demonstrate that at progressively higher agitation rates over the range studied here, and thereby, higher specific power inputs (P/M W kg-1 ), the higher the final viable T-cell density; that is, a cell density of 4.65 ± 0.24 × 106 viable cells ml-1 obtained at the highest P/M of 74 × 10-4 W kg-1 in comparison with 0.91 ± 0.07 × 106 viable cells ml-1 at the lowest P/M of 3.1 × 10-4 W kg-1 . We posit that this improvement is due to the inability at the lower agitation rates to effectively suspend the Dynabeads®, which are required to activate the T-cells; and that contact between them is improved at the higher agitation rates. Importantly, from the data obtained, there is no indication that T-cells prefer being grown under static conditions or are sensitive to fluid dynamic stresses within a stirred-tank bioreactor system at the agitation speeds investigated. Indeed, the opposite has proven to be the case, whereby, the cells grow better under higher agitation speeds while maintaining their quality. This study is the first demonstration of primary T-cell ex vivo manufacture activated by Dynabeads® in an automated stirred-tank bioreactor system such as the ambr® 250 and the findings have the potential to be applied to multiple other cell candidates for advanced therapy applications.