Multiparameter optimisation of a magneto-optical trap using deep learning.

Machine learning based on artificial neural networks has emerged as an efficient means to develop empirical models of complex systems. Cold atomic ensembles have become commonplace in laboratories around the world, however, many-body interactions give rise to complex dynamics that preclude precise analytic optimisation of the cooling and trapping process. Here, we implement a deep artificial neural network to optimise the magneto-optic cooling and trapping of neutral atomic ensembles. The solution identified by machine learning is radically different to the smoothly varying adiabatic solutions currently used. Despite this, the solutions outperform best known solutions producing higher optical densities.

Retroviral transfer of a dominant TCR prevents surface expression of a large proportion of the endogenous TCR repertoire in human T cells.

The latent membrane protein-2 (LMP2) of Epstein-Barr virus is a potential target for T-cell receptor (TCR) gene therapy of Hodgkin lymphoma and nasopharyngeal carcinoma. Here, we modified a human leukocyte antigen-A2-restricted, LMP2-specific TCR to achieve efficient expression following retroviral TCR gene transfer. The unmodified TCR was poorly expressed in primary human T cells, suggesting that it competed inefficiently with endogenous TCR chains for cell surface expression. In order to improve this TCR, we replaced the human constant region with murine sequences, linked the two TCR genes using a self-cleaving 2A sequence and finally, codon optimized the TCR-alpha-2A-beta cassette for efficient translation in human cells. Retroviral transfer of the modified TCR resulted in efficient surface expression and HLA-A2/LMP2 pentamer binding. The transduced cells showed peptide-specific interferon-gamma and interleukin-2 production and killed target cells displaying the LMP2 peptide. Importantly, the introduced LMP2-TCR suppressed the cell surface expression of a large proportion of endogenous TCR combinations present in primary human T cells. The design of dominant TCR is likely to improve TCR gene therapy by reducing the risk of potential autoreactivity of endogenous and mispaired TCR combinations.

WT1-targeted immunotherapy of leukaemia.

Since malignant cells are derived from normal cells, many tumour-associated antigens are also expressed in normal tissues. For examples, WT1 is expressed at elevated levels in most leukaemias, but it is also expressed at reduced levels in normal CD34+ haematopoietic stem cells and in progenitor cells of other tissues. Antigen expression in normal tissues is likely to trigger immunological tolerance and thus blunt T cell responses. This could explain the observation that WT1 vaccination in mice frequently fails to stimulate high avidity cytotoxic T cell responses. In order to circumvent tolerance, we have isolated from HLA-A2-negative donors high avidity CTL specific for HLA-A2-presented peptide epitopes of WT1. These allorestricted CTL efficiently kill HLA-A2-positive leukaemia cells but not normal CD34+ haematopoietic stem cells. However, adoptive cellular therapy with allorestricted CTL could only be performed in leukaemia patients rendered tolerant to the infused CTL by prior allogeneic stem cell transplantation. In order to circumvent this limitation, we propose to exploit the TCR of allorestricted CTL as therapeutic tool. TCR gene transfer can be used to take advantage of the specificity of allorestricted CTL and transfer it to patient CTL, while avoiding the transfer of immunogenic alloantigens from the donor CTL to the patient.

Use of the allogeneic TCR repertoire to enhance anti-tumor immunity.

It is well established that antigen-specific T lymphocytes can inhibit tumor growth in humans and in mice, leading to complete tumor elimination in some cases. However, in many cases T cell immunity is unable to successfully control tumor progression. Since tumors are derived from normal tissues, most antigens are shared with normal tissues, although expression levels are usually elevated in malignant cells. Nevertheless, low-level expression in normal cells can be sufficient to render autologous T cells tolerant and thus unable to mount effective immune responses against tumors. Here, we review how allogeneic T cells can be used to isolate T cells that effectively recognise and kill tumor cells, but not normal cells with low level of antigen expression. The TCR of allogeneic T cells can be introduced into patient T cells to equip them with anti-tumor specificity that may not be present in the autologous T cell repertoire.