I'm a neuroscientist interested in understanding how different brain regions coordinate their activity to implement adaptive behaviours in mammals. Due to technical limitations, in neuroscience we have traditionally zoomed into recording the activity of single or small groups of brain structures during specific behaviours. However, even the simplest behaviours rely on the coordination of activity between multiple brain areas located far apart and distributed across distinct functional sub-systems. Recent advances in functional imaging and electrophysiological methodologies are enabling neuroscientists for the first time to probe distributed information processing at cellular resolution.
In my postdoctoral work, I am combining calcium imaging and electrophysiological techniques to study high-order thalamocortical networks in behaving mice. I am also conducting methodological experiments with Neuropixel probes to enhance the interpretability of information we obtain from dense extracellular recordings. I did my first postdoc with Beatriz Rico (KCL), and my PhD in Oxford, with Simon Butt and Zoltán Molnár. Prior to that, I did my undergrad in Psychology at Universidade do Minho in Portugal, where I'm from. I worked with Armando Machado on debunking the Libet Clock task and with Óscar Gonçalves doing neuropsychological assessment and intervention.
Advancing Extracellular Recordings
Extracellular recordings are a stalwart methodology in systems neuroscience. They allow us to probe the activity of neurons located virtually anywhere in the brain at sub-millisecond resolution in freely-moving animals. The past 60 years have seen extraordinary developments in technology, enabling neuroscientists to go from recording a handful of neurons with tungsten electrodes in the 1950s to isolating hundreds of units with CMOS probes in 2018. The information we now obtain from Neuropixels and other ultra-dense probes is extremely rich but also overwhelming. How can we resolve the cacophony of several hundreds of neurons into well-isolated units? Can we derive information about neuronal orientation and morphology from extracellular spike waveforms? And how can we link units to cellular and molecular properties of neurons in 3D space? I'm exploring these and other questions by performing "ground-truth" experiments where the same neuron is recorded with patch-clamp and extracellular probes (see Marques-Smith et al., 2018, biorXiv), followed by modern histology and molecular biology protocols. Ultimately the goals are to 1) provide data that theoretical neuroscientists can use to improve spike-sorting algorithms and 2) link extracellular recordings of neurons to their cellular, biophysical and molecular properties.
The thalamus is often described as a gatekeeper that modulates and relays sensory information to the primary sensory cortices, a perspective born out mainly of research on the dLGn. However, a significant majority of thalamic nuclei receive little to no input from the sensory periphery and target multiple cortical areas. Some thalamic nuclei don't even project to the cortex and use GABA as a neurotransmitter instead of glutamate. I am studying the so-called high-order thalamic nuclei, which receive their dominant input from cortical Layer 5 and project widely across the cortex and striatum. I believe these nuclei may function in fundamentally different ways from their first-order counterparts such as the dLGn.
GABAergic interneurons are critical nodes in neural information processing systems. They can gate sub-networks of neurons, separate individual cells into multiple compartments and regulate the their dynamic firing range. In my PhD, together with Paul Anastasiades, we comprehensively mapped for the first time the development of translaminar circuits for fast-spiking and non-fast-spiking interneurons across all layers of cortex (Anastasiades, Marques-Smith et al., 2016). In other work, myself and Dan Lyngholm found a pioneer SST+ interneuron linking together thalamic input-recipient layers (Marques-Smith, Lyngholm et al., 2016). With colleagues at the Rico lab, we showed how ErbB4 regulates the synaptic development of CCK basket cells (Del Pino, Brotons et al., 2017), how a perineuronal net protein fine-tunes synergy between membrane and synaptic properties in PV interneurons (Favuzzi et al., 2017) and how developmental processes are coupled together in Martinotti cells to optimise for their mature function and architecture (Lim et al., 2018). Looking into the future, I am interested in extending my work on inhibition to the midbrain and thalamus, where complex dis-inhibitory loops are frequent and could have powerful effects on whole-brain dynamics.
Neuroscientists are acquiring rich, high-dimensional data at a faster pace than ever before. Methodologies such as single-cell RNA sequencing, CMOS probe recordings and large field-of-view 2-photon imaging are providing a deluge of data that individual labs will struggle to fully digest in reasonable time. I believe we need to adopt new paradigms for data sharing, collaboration and publication not just to keep up with the pace at which we acquire data, but to ensure our practises are transparent and reproducible. Moreover, I and others believe well-funded labs in prestigious institutions are in a position of immense privilege and have a moral duty to share data after they publish with colleagues in less fortunate positions. New approaches to publishing are indeed emerging, and we highlight some of them in Marques-Smith et al 2018 (Prologue and Epilogue). Currently, I am testing a model for scientific publication with 3 components: a manuscript in an open-access journal or preprint server, the raw dataset behind it and a repository with all the code used to analyse data and generate figures. Importantly, this repository is bi-directional and welcomes collaboration from the community. The idea is to engage in dialogue and collaboration with our peers, with a view to either improving the original work or deriving follow-up publications from it. GitHub tracks contributions transparently so credit assignment is not an issue. This approach makes science more inclusive to scientists from a variety of backgrounds whilst ensuring transparency and reproducibility.
Recording from the same neuron using Neuropixel probes and patch-clamp
Extracellular recordings have long been one of the workhorses of neuroscience. As our ability to record from increasing numbers of neurons grows, so does the challenge involved in sorting and analysing these recordings. "Ground-truth" datasets are fundamental for benchmarking and improving the algorithms we use for spike sorting and analysing extracellular recordings. In this project at the Kampff Lab we sought to acquire one such ground-truth dataset, by recording the activity of the same cortical pyramidal neuron with Neuropixel probes and patch-clamp in anaesthetised rats.
The experiments and dataset for this project are explained in Marques-Smith et al. (2018, biorXiv).
The full dataset is available here. I have set up a companion GitHub repository here. The repo contains all the code used to analyse data and generate figures for Marques-Smith et al (2018, biorXiv) and instructions on how to get started with and navigate the dataset.
If you use the dataset for your own work, we ask that you kindly cite both the preprint and the dataset:
André Marques-Smith, Joana Pereira Neto, Gonçalo Lopes, Joana Nogueira, Lorenza Calcaterra, João Frazão, Danbee Kim, Matthew G. Phillips, George Dimitriadis, Adam Kampff (2018). Recording from the same neuron with high-density CMOS probes and patch-clamp: a ground-truth dataset and an experiment in collaboration. bioRxiv 370080; doi: https://doi.org/10.1101/370080
André Marques-Smith, Joana P. Neto, Gonçalo Lopes, Joana Nogueira, Lorenza Calcaterra, João Frazão, Danbee Kim, Matthew G. Phillips, George Dimitriadis and Adam R. Kampff (2018); Simultaneous patch-clamp and dense CMOS probe extracellular recordings from the same cortical neuron in anaesthetized rats. CRCNS.org
In the repo we have suggested a number of follow-up analysis projects that we'd like to open collaboration on. The idea is to use GitHub as a platform to coordinate the efforts of multiple scientists spread out over the world and interested in addressing these questions. GitHub enables accurate and transparent credit assignment, communication and reviewing of peer contributions, making it a very interesting platform to conduct Open Science on. Check out our proposed workflow here.