CAR affinity modulates the sensitivity of CAR-T cells to PD-1/PD-L1-mediated inhibition.

Chimeric antigen receptor (CAR)-T cell therapy for solid tumors faces significant hurdles, including T-cell inhibition mediated by the PD-1/PD-L1 axis. The effects of disrupting this pathway on T-cells are being actively explored and controversial outcomes have been reported. Here, we hypothesize that CAR-antigen affinity may be a key factor modulating T-cell susceptibility towards the PD-1/PD-L1 axis. We systematically interrogate CAR-T cells targeting HER2 with either low (LA) or high affinity (HA) in various preclinical models. Our results reveal an increased sensitivity of LA CAR-T cells to PD-L1-mediated inhibition when compared to their HA counterparts by using in vitro models of tumor cell lines and supported lipid bilayers modified to display varying PD-L1 densities. CRISPR/Cas9-mediated knockout (KO) of PD-1 enhances LA CAR-T cell cytokine secretion and polyfunctionality in vitro and antitumor effect in vivo and results in the downregulation of gene signatures related to T-cell exhaustion. By contrast, HA CAR-T cell features remain unaffected following PD-1 KO. This behavior holds true for CD28 and ICOS but not 4-1BB co-stimulated CAR-T cells, which are less sensitive to PD-L1 inhibition albeit targeting the antigen with LA. Our findings may inform CAR-T therapies involving disruption of PD-1/PD-L1 pathway tailored in particular for effective treatment of solid tumors.

Learning Assurance Analysis for Further Certification Process of Machine Learning Techniques: Case-Study Air Traffic Conflict Detection Predictor.

Designing and developing artificial intelligence (AI)-based systems that can be trusted justifiably is one of the main issues aviation must face in the coming years. European Union Aviation Safety Agency (EASA) has developed a user guide that could be potentially transformed as means of compliance for future AI-based regulation. Designers and developers must understand how the learning assurance process of any machine learning (ML) model impacts trust. ML is a narrow branch of AI that uses statistical models to perform predictions. This work deals with the learning assurance process for ML-based systems in the field of air traffic control. A conflict detection tool has been developed to identify separation infringements among aircraft pairs, and the ML algorithm used for classification and regression was extreme gradient boosting. This paper analyses the validity and adaptability of EASA W-shaped methodology for ML-based systems. The results have identified the lack of the EASA W-shaped methodology in time-dependent analysis, by showing how time can impact ML algorithms designed in the case where no time requirements are considered. Another meaningful conclusion is, for systems that depend highly on when the prediction is made, classification and regression metrics cannot be one-size-fits-all because they vary over time.

Epitope spreading driven by the joint action of CART cells and pharmacological STING stimulation counteracts tumor escape via antigen-loss variants.

BACKGROUND: Target antigen (Ag) loss has emerged as a major cause of relapse after chimeric antigen receptor T (CART)-cell therapy. We reasoned that the combination of CART cells, with the consequent tumor debulking and release of Ags, together with an immunomodulatory agent, such as the stimulator of interferon gene ligand (STING-L) 2'3'-cyclic GMP-AMP (2'3'-cGAMP), may facilitate the activation of an endogenous response to secondary tumor Ags able to counteract this tumor escape mechanism. METHODS: Mice bearing B16-derived tumors expressing prostate-specific membrane Ag or gp75 were treated systemically with cognate CART cells followed by intratumoral injections of 2'3'-cGAMP. We studied the target Ag inmunoediting by CART cells and the effect of the CART/STING-L combination on the control of STING-L-treated and STING-L-non-treated tumors and on the endogenous antitumor T-cell response. The role of Batf3-dependent dendritic cells (DCs), stimulator of interferon gene (STING) signaling and perforin (Perf)-mediated killing in the efficacy of the combination were analyzed. RESULTS: Using an immune-competent solid tumor model, we showed that CART cells led to the emergence of tumor cells that lose the target Ag, recreating the cancer immunoediting effect of CART-cell therapy. In this setting, the CART/STING-L combination, but not the monotherapy with CART cells or STING-L, restrained tumor progression and enhanced overall survival, showing abscopal effects on distal STING-L-non-treated tumors. Interestingly, a secondary immune response against non-chimeric antigen receptor-targeted Ags (epitope spreading), as determined by major histocompatibility complex-I-tetramer staining, was fostered and its intensity correlated with the efficacy of the combination. This was consistent with the oligoclonal expansion of host T cells, as revealed by in-depth T-cell receptor repertoire analysis. Moreover, only in the combination group did the activation of endogenous T cells translate into a systemic antitumor response. Importantly, the epitope spreading and the antitumor effects of the combination were fully dependent on host STING signaling and Batf3-dependent DCs, and were partially dependent on Perf release by CART cells. Interestingly, the efficacy of the CART/STING-L treatment also depended on STING signaling in CART cells. CONCLUSIONS: Our data show that 2'3'-cGAMP is a suitable adjuvant to combine with CART-cell therapy, allowing the induction of an endogenous T-cell response that prevents the outgrowth of Ag-loss tumor variants.

Intermittent tACS during a visual task impacts neural oscillations and LZW complexity.

Little is known about how transcranial alternating current stimulation (tACS) interacts with brain activity. Here, we investigate the effects of tACS using an intermittent tACS-EEG protocol and use, in addition to classical metrics, Lempel-Ziv-Welch complexity (LZW) to characterize the interactions between task, endogenous and exogenous oscillations. In a cross-over study, EEG was recorded from thirty participants engaged in a change-of-speed detection task while receiving multichannel tACS over the visual cortex at 10 Hz, 70 Hz and a control condition. In each session, tACS was applied intermittently during 5 s events interleaved with EEG recordings over multiple trials. We found that, with respect to control, stimulation at 10 Hz ([Formula: see text]) enhanced both [Formula: see text] and [Formula: see text] power, [Formula: see text]-LZW complexity and [Formula: see text] but not [Formula: see text] phase locking value with respect to tACS onset ([Formula: see text]-PLV, [Formula: see text]-PLV), and increased reaction time (RT). [Formula: see text] increased RT with little impact on other metrics. As trials associated with larger [Formula: see text]-power (and lower [Formula: see text]-LZW) predicted shorter RT, we argue that [Formula: see text] produces a disruption of functionally relevant fast oscillations through an increase in [Formula: see text]-band power, slowing behavioural responses and increasing the complexity of gamma oscillations. Our study highlights the complex interaction between tACS and endogenous brain dynamics, and suggests the use of algorithmic complexity inspired metrics to characterize cortical dynamics in a behaviorally relevant timescale.

Prefrontal Cortex Neuromodulation Enhances Frontal Asymmetry and Reduces Caloric Intake in Patients with Morbid Obesity.

OBJECTIVE: The objective of this study was to test the feasibility of a combined intervention involving transcranial direct current stimulation (tDCS) on the dorsolateral prefrontal cortex (dlPFC) and cognitive training (CT). Short-term effects on food consumption, cognition, endocannabinoid (eCB) levels, and electroencephalogram (EEG) markers of future weight loss were explored. METHODS: Eighteen healthy volunteers with morbid obesity were randomized in a double-blind, placebo-controlled, parallel trial. Participants received sham or active tDCS plus CT for four consecutive days. Cognitive performance, daily food intake, and eCB blood samples were collected before and after the intervention; EEG data were gathered before and after daily training. RESULTS: The active tDCS + CT group reversed left-dominant frontal asymmetry and increased frontal coherence (FC) in the γ-band (30-45 Hz) after the intervention. The strength of the latter predicted BMI reduction. Additionally, a large intervention effect on food intake was shown in the active tDCS + CT group at follow-up (-339.6 ± 639 kcal on average), and there was a decrease of plasma eCB concentrations. CONCLUSIONS: dlPFC modulation through tDCS + CT is an effective tool to restore right dominance of the dlPFC and enhance FC in patients with morbid obesity. Moreover, the effect of the strength of FC on BMI suggests that the interhemispheric FC at the dlPFC is functionally relevant for the efficient regulation of food choice.

Deep Learning With EEG Spectrograms in Rapid Eye Movement Behavior Disorder.

REM Behavior Disorder (RBD) is now recognized as the prodromal stage of α-synucleinopathies such as Parkinson's disease (PD). In this paper, we describe deep learning models for diagnosis/prognosis derived from a few minutes of eyes-closed resting electroencephalography data (EEG) collected at baseline from idiopathic RBD patients (n = 121) and healthy controls (HC, n = 91). A few years after the EEG acquisition (4±2 years), a subset of the RBD patients were eventually diagnosed with either PD (n = 14) or Dementia with Lewy bodies (DLB, n = 13), while the rest remained idiopathic RBD. We describe first a simple deep convolutional neural network (DCNN) with a five-layer architecture combining filtering and pooling, which we train using stacked multi-channel EEG spectrograms from idiopathic patients and healthy controls. We treat the data as in audio or image classification problems where deep networks have proven successful by exploiting invariances and compositional features in the data. For comparison, we study a simple deep recurrent neural network (RNN) model using three stacked Long Short Term Memory network (LSTM) cells or gated-recurrent unit (GRU) cells-with very similar results. The performance of these networks typically reaches 80% (±1%) classification accuracy in the balanced HC vs. PD-conversion classification problem. In particular, using data from the best single EEG channel, we obtain an area under the curve (AUC) of 87% (±1%)-while avoiding spectral feature selection. The trained classifier can also be used to generate synthetic spectrograms using the DeepDream algorithm to study what time-frequency features are relevant for classification. We find these to be bursts in the theta band together with a decrease of bursting in the alpha band in future RBD converters (i.e., converting to PD or DLB in the follow up) relative to HCs. From this first study, we conclude that deep networks may provide a useful tool for the analysis of EEG dynamics even from relatively small datasets, offering physiological insights and enabling the identification of clinically relevant biomarkers.

Algorithmic Complexity of EEG for Prognosis of Neurodegeneration in Idiopathic Rapid Eye Movement Behavior Disorder (RBD).

Idiopathic rapid eye movement sleep behavior disorder (RBD) is a serious risk factor for neurodegenerative processes such as Parkinson's disease (PD). We investigate the use of EEG algorithmic complexity derived metrics for its prognosis. We analyzed resting state EEG data collected from 114 idiopathic RBD patients and 83 healthy controls in a longitudinal study forming a cohort in which several RBD patients developed PD or dementia with Lewy bodies. Multichannel data from ~ 3 min recordings was converted to spectrograms and their algorithmic complexity estimated using Lempel-Ziv-Welch compression. Complexity measures and entropy rate displayed statistically significant differences between groups. Results are compared to those using the ratio of slow to fast frequency power, which they are seen to complement by displaying increased sensitivity even when using a few EEG channels. Poor prognosis in RBD appears to be associated with decreased complexity of EEG spectrograms stemming in part from frequency power imbalances and cross-frequency amplitude algorithmic coupling. Algorithmic complexity metrics provide a robust, powerful and complementary way to quantify the dynamics of EEG signals in RBD with links to emerging theories of brain function stemming from algorithmic information theory.

Feelings of loss in parents of children with infantile cerebral palsy.

BACKGROUND: Raising a child diagnosed with infantile cerebral palsy is a challenge for families and causes many changes in their lifestyle. When the diagnosis is unexpected, feelings related to loss and hard-to-manage emotions such as uncertainty and bewilderment can arise. OBJECTIVE: To identify how feelings of loss are structured in fathers and mothers of children diagnosed with infantile cerebral palsy. METHODS: A qualitative design with based on a grounded theory approach was used. Twenty-four participants were selected to participate in the research from San Cecilio Clinical Hospital in the city of Granada (Spain). The sampling procedure was purposive based on inclusion and exclusion criteria and ended when data saturation was acquired. The participants were interviewed according to a script developed ad hoc. Data were collected during 2012. The interviews were analyzed with Atlas.ti 6.2 software, using the sequence suggested by Straus and Corbin including open, axial and selective codification. RESULTS: The analysis led to the identification of the main category, "Experiences of loss." The codes contributing to explain these experiences were "Shock," "Hope," "Traumatic Experience," "Feelings related to loss," "Ideal Child" and "Acceptance of the Child." CONCLUSIONS: These parents experience feelings of loss of the ideal child, which are more complex in the first stage of the diagnosis and when the severity of the cerebral palsy is greater. Emotional intervention on the part of health care providers is needed to aid parents in facing the various obstacles encountered throughout their child's up-bringing.

Neuronal oscillations form parietal/frontal networks during contour integration.

The ability to integrate visual features into a global coherent percept that can be further categorized and manipulated are fundamental abilities of the neural system. While the processing of visual information involves activation of early visual cortices, the recruitment of parietal and frontal cortices has been shown to be crucial for perceptual processes. Yet is it not clear how both cortical and long-range oscillatory activity leads to the integration of visual features into a coherent percept. Here, we will investigate perceptual grouping through the analysis of a contour categorization task, where the local elements that form contour must be linked into a coherent structure, which is then further processed and manipulated to perform the categorization task. The contour formation in our visual stimulus is a dynamic process where, for the first time, visual perception of contours is disentangled from the onset of visual stimulation or from motor preparation, cognitive processes that until now have been behaviorally attached to perceptual processes. Our main finding is that, while local and long-range synchronization at several frequencies seem to be an ongoing phenomena, categorization of a contour could only be predicted through local oscillatory activity within parietal/frontal sources, which in turn, would synchronize at gamma (>30 Hz) frequency. Simultaneously, fronto-parietal beta (13-30 Hz) phase locking forms a network spanning across neural sources that are not category specific. Both long range networks, i.e., the gamma network that is category specific, and the beta network that is not category specific, are functionally distinct but spatially overlapping. Altogether, we show that a critical mechanism underlying contour categorization involves oscillatory activity within parietal/frontal cortices, as well as its synchronization across distal cortical sites.

Spike train auto-structure impacts post-synaptic firing and timing-based plasticity.

Cortical neurons are typically driven by several thousand synapses. The precise spatiotemporal pattern formed by these inputs can modulate the response of a post-synaptic cell. In this work, we explore how the temporal structure of pre-synaptic inhibitory and excitatory inputs impact the post-synaptic firing of a conductance-based integrate and fire neuron. Both the excitatory and inhibitory input was modeled by renewal gamma processes with varying shape factors for modeling regular and temporally random Poisson activity. We demonstrate that the temporal structure of mutually independent inputs affects the post-synaptic firing, while the strength of the effect depends on the firing rates of both the excitatory and inhibitory inputs. In a second step, we explore the effect of temporal structure of mutually independent inputs on a simple version of Hebbian learning, i.e., hard bound spike-timing-dependent plasticity. We explore both the equilibrium weight distribution and the speed of the transient weight dynamics for different mutually independent gamma processes. We find that both the equilibrium distribution of the synaptic weights and the speed of synaptic changes are modulated by the temporal structure of the input. Finally, we highlight that the sensitivity of both the post-synaptic firing as well as the spike-timing-dependent plasticity on the auto-structure of the input of a neuron could be used to modulate the learning rate of synaptic modification.