07. Neuroscience and Quantitative Neuroimaging

A first line of activity, which we intend to complete within a year, is the development of methods for the mitigation of physiological noise. At the same time, we will deepen the dynamic characterization of the signal. The relationship between plastic modulation of networks and behavior is a question of utmost importance in terms of basic knowledge of brain function and its implications for the understanding of major neurological and psychiatric pathologies. Our group is among the first to have studied the issue of the dynamic modulation of brain networks induced by brain function, in multiple experimental models, as shown in the figure.

Left: brain networks during spontaneous pupillary diameter modulation associated with a visual attention task. Right: Granger causality analysis between the rate of variation of the pupillary diameter dP/dT and the brain networks. Although Locus Coeruleus is physiologically associated with pupil diameter, Granger Causality indicates the presence of a complex pattern of interdependence between networks, with LC and pupil diameter separated by numerous stages of cortical processing (From DiNuzzo et al 2019).

In particular, we confirmed that the topology of the resting brain networks is globally conserved during the execution of a continuous cognitive task. However, we have highlighted two phenomena that deserve further study. The first is that a small number of brain network nodes actually change their topological relationships during activity, suggesting that the networks are globally stable, but undergo short-term plastic remodeling phenomena. We plan to investigate this modulation, both in terms of the modalities with which it occurs, and in terms of functional significance.

Correlation matrix at rest (top left) and during a cognitive task (bottom), and t-test (unthresholded) on the difference between the two (top right, from Tommasin et al. 2018).

The second point that we will study is the significance of the amplitude of the modulation of functional connectivity in MRI (fcMRI). As can be clearly seen in the figure, intranetwork connectivity, which is normally higher than connectivity between networks, tends to decrease during tasks, while the opposite occurs for connectivity between networks. With our work, we have shown that the functional connectivity variation ΔFC as a function of resting connectivity FC_R is well-described by a simple linear model ∆FC=βFC_R+β₀. Interestingly, the amplitude of this modulation is irrelevant from a behavioral point of view (percentage of exact responses during the cognitive task), while the slope β shows a significant inverse correlation with cognitive performance in some areas involved in the execution of the task. These results indicate that, without proper normalization, the amplitude of connectivity changes could be an irrelevant parameter from the physiological point of view. This result would have a significant impact, considering the increasing use of this parameter to study neurodegeneration. We will then investigate the physiological origins of stimulation-induced modulation of fcMRI to identify the determinant of the behaviorally irrelevant component. It is important to note that vascular reactivity is a potent modulator of fMRI response, and vascular reactivity has been reported to spatially modulate functional connectivity at rest and the amplitude of BOLD fluctuations. We believe we can hypothesize a role of vascular reactivity and/or the autonomic system in determining the connectivity changes associated with activity. If this hypothesis, subjected to experimental verification, proves to be correct, we will develop methods of renormalization of the signal aimed at excluding the vascular component from the analysis.

The second line is partially linked to the first and, in particular, aims to investigate the mechanisms of plasticity induced by electrical stimulation. This is a study in collaboration with the Santa Lucia Foundation (Prof. Marangolo), aimed in particular at studying the effects of transpinal electrical stimulation in patients suffering from Alzheimer’s dementia (AD). Transpinal direct-current stimulation is a non-invasive stimulation tool that involves the application of a weak electric current (1–2 mA) using electrodes applied to the back. The current modulates neuronal excitability, with protracted effects similar to the mechanisms of long-term enhancement and depression. While there is some fMRI evidence of the plastic effect in the case of cortical stimulation, the effects of transpinal stimulation on brain networks are still unknown. In this series of experiments, we intend first of all to study the focal effect of tsDCS as such, by means of task-based studies, and secondly to determine its plastic-type effects and the relative temporal duration (a relevant characteristic for therapeutic applications). To this end, we will develop a technique capable of taking into account the spatial scale of the modulations. In fcMRI studies, the nodes of the network are identified a priori, and the evolution of the branches is studied, often assuming the spatial stationarity of the network. However, plastic phenomena occur at multiple scales, starting from the minimum functional unit (canonical microcircuit), and can modulate by definition the physiological substrate of the network. We will therefore develop two optimized approaches to characterize plasticity by fcMRI: 1) a method based on the extraction of exemplary dynamic components by constrained Independent Vector Analysis (SED-cIVA), adapted to the very high temporal resolution data that we will acquire through multiband approaches (0.8 s as compared to the ordinary 2–3 s). SED-cIVA is an iterative approach, which in a first phase extracts models with independent component analysis approaches (ICA), and in a second phase determines the spatiotemporal dynamics by means of fit on a sliding window. 2) We will then develop a univariate approach based on the connectivity radius. The techniques to identify fMRI networks are usually of multivariate nature. In this way, it is inherently difficult to define the location and directionality of the modulations. Preliminary tests conducted over the past year suggest that a metric based on the average of the Fisher transformation of the correlation coefficients of a voxel with the voxels included in spherical shells of increasing radius has the potential to capture local modulations of connectivity. We therefore propose to develop this metric and show its adequacy to describe phenomena of neuronal plasticity.

Brain activity is mainly based on oxidative metabolism. For this reason, the measurement of the metabolic oxygen-consumption rate (CMRO₂) is an excellent biomarker for the quantification of brain activity and the physiological state of tissues, with potential applications in the early diagnosis of carcinomas, strokes, neurological, and neurodegenerative diseases. Currently, positron emission tomography techniques based on oxygen isotopes are the gold standard for obtaining CMRO₂ maps of the whole brain. However, the technical complexity of the tests and the level of invasiveness of the same constitute a huge limit to their use.

There are several MRI methods for measuring CMRO₂, based on different technological approaches and physiological characteristics. For example, we cite the exploitation of the magnetic field differences associated with tissue differences between the superior sagittal sinus or main veins and the surrounding parenchyma, the T₂ oxygenation calibration curves with speed selective techniques, or the quantification approaches of venous oxygen saturation through the T₂ of venous blood.

Davis and Hoge introduced in the late 1990s another group of techniques, based on BOLD calibration methods, which aim to estimate CMRO₂ from BOLD and ASL (Arterial Spin Labeling) signals, exploiting respiratory tasks (hypercapnia and hyperoxia) and mathematical models that describe the complex relationship between oxygen metabolism, BOLD signals, and cerebral blood flow (CBF). CBF is a direct biomarker for cerebrovascular function and neurovascular health, and the correlation between CBF, local neuronal activity, and metabolism, known as neurovascular coupling, is a surrogate marker for brain function. Detecting CBF at rest without carrying out complex cognitive tasks is a fairly easy operation to conduct in clinical practice. For example, CBF can be measured with ASL, perfusion, or phase-contrast MRI techniques. Recently, an extension of these techniques has made it possible to record, in the same experiment, the changes in CBF, induced by hypercapnia, and in BOLD induced by hyperoxia (increased concentration of oxygen in the blood), and to exploit them to characterize the cerebral metabolic state through the estimation of various parameters, including the concentration of deoxyhemoglobin in venous blood, the fraction of oxygen extracted (OEF), the absolute CMRO₂, the CBF, and the vascular reactivity (CVR). This innovative approach is called quantitative oxygen imaging or dual calibrated fMRI (dcfMRI) and overcomes one of the basic problems of the method initially proposed by Davis and Hoge, namely, the dependence on an unknown parameter, the signal baseline M. The dcfMRI techniques have technological requirements that can be matched with those of vascular reactivity techniques (described in the paragraph Dynamics of Brain networks), therefore the development of both techniques has clear synergistic aspects. In this part, the tissue-segmentation work based on AI methods developed in collaboration with the University of Kuopio will also be exploited, to obtain structurally homogeneous areas with which to average the signals.

The quantification of CMRO₂ will enable us to continue the study of the metabolic dynamics that has characterized the activity of our group from the beginning. Many of the problems we have successfully studied actually concern the functional characterization of the cells (neurons and glial cells) of the nervous system in biophysical and biochemical terms. Neurons are the components of neural circuits, whose task is to process specific types of information, whose integration represents the substrate of brain function. Glial cells, which are in a 10 : 1 ratio with neurons, have functions that are not yet completely clear, but which certainly include the modulation of nutrient transport, ionic homeostasis, and the modulation of cellular excitability.

From the experimental point of view, we have in the past focused attention on lactate. The importance that lactate plays in the functional metabolism of the brain derives from the paradigm shift that has invested this scientific field after the introduction of the hypothesis of the lactate shuttle, Astrocyte–Neuron Lactate Shuttle (ANLSH), proposed by Pellerin and Magistretti in 1994. This hypothesis has revolutionized the concept that neurons exclusively use glucose, while providing a nutritional role for astrocytes (a type of glial cell). According to ANLSH, astrocytes would couple the increase in electrical activity of neurons to the absorption of glucose from the blood for production of energy. In other words, ANLSH affirms that the primary energy substrate of neurons would not be glucose, but the lactate produced by astrocytes at activity-dependent rate.

We have repeatedly challenged the ANSL hypothesis on experimental and modeling grounds, associating the increase in lactate that is observed during stimulation with a generalized increase in metabolic intermediates during the rapid increase in aerobic metabolism. However, there is no direct experimental demonstration that the change in lactate in vivo is associated with oxygen consumption. We therefore intend to carry out a combined spectroscopy and CMRO₂ experiment aimed at functionally characterizing the link between CMRO₂ and lactate.

Left: brain metabolic network, including astrocytic and neuronal subcellular compartments and extracellular compartments (extracellular matrix, vessels). The part subject to the latest steady-state modeling studies (DiNuzzo et al. 2017) is highlighted in gray and amplified on the right.

From the modeling point of view, the basic idea is to continue past efforts by capitalizing on the research of the last ten years in the field of kinetic and stoichiometric models, research that has been the focus of strong interest in the scientific community and has recently led to important strategic collaboration with Yale School of Medicine.

In particular, the aim is to formalize compartmentalized models (neuron, astrocyte, extracellular space, perivascular space, vessels) in which the cellular components are described by detailed metabolic networks on the genomic level to implement simulations of cerebral metabolism (with significant impact also on the interpretation of MRS and fMRI signals) with a level of detail never achieved before (side figure). This effort is aimed among other things at clarifying the role of astrocytic glycogen, for which our group first hypothesized and modeled a modulatory role. This research activity would also allow to acquire general skills in the study of metabolic networks from a more purely bioinformatic point of view (in particular, systems biology). In turn, this would foster the birth of new international collaborations and the expansion of the research line towards biotechnology.

The model development will initially focus on the determination of metabolic fluxes at steady state (Flux Balance Analysis). Our methodology proved to be qualitatively superior to the standard algorithms used to estimate the distributions of metabolic flows. On the other hand, the computational efficiency of our method is lower, and the first step will be aimed at optimizing and implementing strategies, many of which are already known, to reduce the computational load of our method.

On brain networks

• DiNuzzo et al. 2019. Brain Networks Underlying Eye’s Pupil Dynamics. Front Neurosci. doi: 10.3389/fnins.2019.00965
• Tommasin et al. 2018. Scale-invariant Rearrangement of Resting State Networks in the Human Brain under Sustained Stimulation. Neuroimage. doi: 10.1016/j.neuroimage.2018.06.006
• Mascali et al. 2018. Disruption of Semantic Network in Mild Alzheimer’s Disease Revealed by Resting-State fMRI. Neuroscience. doi: 10.1016/j.neuroscience.2017.11.030
• Mascali et al. 2015. Intrinsic Patterns of Coupling Between Correlation and Amplitude of Low-Frequency fMRI Fluctuations Are Disrupted in Degenerative Dementia Mainly Due to Functional Disconnection. PLoS One. doi: 10.1371/journal.pone.0120988

On brain metabolic dynamic, modeling

• DiNuzzo et al. 2017. Computational Flux Balance Analysis Predicts that Stimulation of Energy Metabolism in Astrocytes and their Metabolic Interactions with Neurons Depend on Uptake of K+ Rather than Glutamate. Neurochemical Research. doi: 10.1007/s11064-016-2048-0
• Massucci et al. 2013. Energy metabolism and glutamate-glutamine cycle in the brain: a stoichiometric modeling perspective. BMC Systems Biology. doi: 10.1186/1752-0509-7-103.
• Mangia et al. 2011. Response to ‘comment on recent modeling studies of astrocyte-neuron metabolic interactions’: much ado about nothing. J Cereb Blood Flow Metab. doi: 10.1038/jcbfm.2011.29
• DiNuzzo et al. 2010b. Glycogenolysis in astrocytes supports blood-borne glucose channeling not glycogen-derived lactate shuttling to neurons: evidence from mathematical modelling. J Cereb Blood Flow Metab. doi: 10.1038/jcbfm.2010.151
• DiNuzzo et al. 2010a. Changes in glucose uptake rather than lactate shuttle take center stage in subserving neuroenergetics: evidence from mathematical modeling. J Cereb Blood Flow Metab. doi: 10.1038/jcbfm.2009.232

On brain metabolic dynamic, experimental

• Bednarik et al. 2018. Neurochemical Responses to Chromatic and Achromatic Stimuli in the Human Visual Cortex. J Cereb Blood Flow Metab. doi: 10.1177/0271678X17695291
• Bednarik et al. 2015. Neurochemical and BOLD Responses During Neuronal Activation Measured in the Human Visual Cortex at 7 Tesla. J Cereb Blood Flow Metab. doi: 10.1038/jcbfm.2014.233.

• University of Minnesota, Center for Magnetic Resonance Research (CMRR), Minneapolis. (Prof. S. Mangia)
• Yale University, Magnetic Resonance Research Center, New Haven. (Prof. D. Rothman)
• University of Eastern Finland, A.I. Virtanen Institute for Molecular Sciences, Kuopio. (Prof. O. Gröhn, Prof. J. Tohka)
• Fondazione Santa Lucia, Roma (Prof. M Bozzali, Prof. P Marangolo)
• Università di Chieti-Pescara, Dipartimento di Neuroscienze, Chieti (Prof. R. G. Wise)
• Sapienza Università di Roma, Dipartimenti di Ingegneria dell’Informazione Elettronica e Telecomunicazioni (Prof. F. Frezza) e di Fisica (Prof. S. Giagu)
• IMT Lucca (Dr. T. Gili)
• CNR, Istituto dei sistemi complessi, (Dr. S. Capuani) e Istituto di Nanotecnologia (Dr. M. Fratini)

• 2015–2019 H2020 MSCA-RISE 691110 “MICROBRADAM: Advanced MR methods for characterization of microstructural brain damage”.
• 2015–2018 Regione Lazio POR-FESR 2014-2020 RU-2014-1092, “PAMINA: Piattaforma per l’Analisi Multimodale Integrata in Neuroscienze Applicate”.