FEATURE
PAULA ZHU, Harvard College '24
Recent Advances in Neurotechnologies to Understand and Direct Specific Behavior of Live Animals
THURJ Volume 14 | Issue 1
Abstract
Neuroscience has traditionally faced difficulty in accessing and understanding the behavior of the brain at a single-cell level. Many of these challenges are due to the brain’s complexity; the brain has 100 billion neurons and well over 100 trillion connections and synapses, all operating at a millisecond timescale with functionally distinct circuits crossing over each other within microscale regions. However, starting with breakthroughs in the 2000s, such as fluorescent genetically-encoded calcium ion (Ca2+) indicators (GECIs)—where Ca2+ detection is used as a proxy of neural activity—and optogenetics—where neuron firing is controlled using light-sensitive ion channels in neurons—new advances in revolutionary neurotechnology have made it increasingly feasible for neuroscientists to understand how the brain operates on a cellular level. In the past few years,
four distinct categories of neurotechnology advances have catalyzed this change in understanding, including chemogenic control, electrical recording and control, optogenetics-based control, and optogenetics recording and visualization.
1) Chemogenetic Control
Chemogenetic control is the use of chemical and genetic engineering to control cellular processes. Traditional pharmacology typically manipulates the uptake and degradation of natural neurotransmitters, acts as a direct agonist/antagonist (stimulant/blocker) of naturally occurring receptors or affects natural cell signaling pathways. However, these pharmacological methods lack any specificity beyond the distribution of the endogenous drug target. One major attempt at chemogenetic control sought to address these challenges associated with the lack of pharmacological specificity.
DREADDs and Designer Receptors with Designer Drugs
Chemogenetic designer receptors, known as DREADDs (Designer Receptors Exclusively Activated by Designer Drugs), are artificially engineered G-protein-coupled receptors designed to initiate intracellular signaling pathways that influence neuronal activity. For example, the hM3Dq receptor leverages the native Gq-protein signaling pathway to induce firing, while the hM4Di receptor employs Gi signaling to silence activity. These receptors only respond to designer compounds made to be typically inert and blood-brain barrier permeable, such as recently developed compound-21 (c21), which can be administered in vivo via intraperitoneal injection for use up to several hours or through drinking water for use over several days. While chemogenetic techniques offer precise spatial modulation (control over which cells express DREADDs) and a minimally invasive method of vuse, they do not provide temporal precision superior to traditional optogenetic methods, since their timescale of neural control is dependent on the slow and gradual absorption, distribution, metabolism, and excretion of the designer drug used (Smith et al., 2021).
2) Electrical Recording and Control
In traditional electrophysiology, a sharp electrode is placed near or in neurons to record their activity, which can also lead to tissue damage and an inflammatory response. This invasiveness makes it challenging to study long-term neural activity and limits the number of electrodes that can be implanted in a brain simultaneously. Moreover, electrophysiological recordings are typically limited to a small number of neurons or specific brain regions, making it challenging to understand large-scale network interactions. Ultimately, these limitations restrict traditional electrophysiology to very few recordings at a time, with a high learning curve for quality recordings that do not overly damage the neuron.
Neuropixels and Recording Thousands of Neurons at Once
Many recent neural electrical recording technologies have focused on simultaneously recording a greater spatial coverage and a larger number of neurons. Advancements in lithographic fabrication techniques have allowed semiconductors and electrodes to be scaled down to subcellular dimensions, enabling the recording and stimulation of individual neurons. One common example is the Neuropixels probe, a millimeter-long, micrometer-thick straight silicon shank of high-density electrodes incorporating CMOS circuits for multiplex- ing and amplification, which maps neuron electrical recordings across multiple functional depths (Jun et al., 2017) (Steinmetz et al., 2021). This year, a recent version of Neuropixels recorded over 3000 single neurons per single probe (Trautmann et al., 2023). Moreover, specific neurons can be selected by combining probe implantation with the expression of optogenetic opsins that induce neuron activation with light–in this way, activity spikes from neurons can be identified as part of, or not part of, a genetically specific population via their response to light. Nevertheless, challenges associated with Neuropixels probes include their relative stiffness, that their insertion can lead to tissue displacement and potential damage, and gradual drift of probe location. This may trigger neuroinflammatory responses affecting the quality of neural recordings over time. Finally, combining Neuropixels with optical techniques may risk confounding the photovoltaic effects of the CMOS devices if experiments are not designed to minimize optoelectric crosstalk and overlap.
Materials for Biocompatible and Long-Term Electrodes
The mechanical mismatch between conventional electrode metals and neural tissue has made it challenging to avoid physical damage and chronic inflammatory responses when using electrodes for neural control. On the other hand, neuro-electronics need low interface impedances (effective output circuit resistance) and physiologically relevant CICs (the charge that can be injected into brain tissue without inducing any irre- versible chemical reactions at the electrode surface) to achieve high signal-to-noise ratio and low voltage sig- nal loss measurements. Advances using biocompatible, conductive polymers have been employed to increase both conductivity and material flexibility. For example, a system of 16-micrometer-thick flexible probe mod- ules integrating electrodes of the conductive and biocompatible polymer PEDOT:PSS has shown recordings from up to 1,024 discrete channels in freely moving animals. This technology supported both extended re- cordings spanning months and tracking of well-isolated neural units across multiple brain regions for over a week (Chung et al., 2019). Likewise, developed in the
same year, Neurotassels, an array of up to 1024 flexible, 100-micrometer diameter microelectrode filaments have also been used to record from animals for several months. These filaments spontaneously assemble into implantable fibers upon withdrawal from tissue-dissolvable, molten polyethylene glycol (PEG) polymer (Guan et al., 2019).
3) Optogenetics-Based Control
First developed in a landmark paper in 2005, optogenetic control of cell activity involves the introduction of light-sensitive proteins called opsins in target cells. When exposed to specific wavelengths of light, opsins can activate or inhibit neurons, depending on the type used. The revolution of optogenetics is the establishment of causality between the activity of specific neurons and subsequent animal behavior. Since then, light-controlled systems for manipulating various other aspects of neurons have been developed. Of note, ad- vances in widefield scanning and holographic illumination for large-scale, volumetric access further buttress the advantages of these light-dependent tools.
Targeting Behavior-Specific Neurons
A recombinase, such as Cre or Flp, is an enzyme that will cut a section of DNA based on markers around the gene. They can be used to activate target genes only in cells expressing the recombinase, such as by excising stop codons placed before the target gene. In Targeted Recombination in Active Populations (TRAP), neurons express a drug-dependent recombinase (such as tamoxifen-dependent CreERT2) downstream of a neural-activity-dependent promoter (such as Arc or Fos), leading to increased expression of the recombinase following neural activity (Guenthner et al., 2013). In this way, only when something activates the neuron, such as a behavioral stimulus, and the drug is injected is the recombinase translated and allowed to trigger recombi- nation and activation of target genes. These target genes can include optogenetic proteins or any other tools for neural control. This has famously been used to target neurons to manipulate or create specific memories in live mice (Ramirez et al., 2013). However, the reliance on slow drug kinetics prevents this approach from effectively capturing neurons specific for faster behaviors.
More novel approaches have attempted to condition gene expression to Ca2+ rises triggered by action potentials, such that expression is initiated when both Ca2+ and light are present. For example, CalLight-ST from 2022 uses 1) a split-tobacco etch virus protease (TEVp) connected on either end to domains that bind together with Ca2+ is present and 2) a transcriptional activator tethered by a cleavage sequence hidden by a light-dependent domain to the membrane (Hyun et al., 2022). When Ca2+ is present, TEVp fuses together and becomes active. When light is present, the cleavage sequence becomes accessible. Thus, only when both Ca2+ and light are available is the cleavage sequence cut by TEVp, and the transcriptional activator is released for target gene expression in those specific neurons.
Artifical Neurotransmitter Release and Receptor Binding
Neurotransmitter photo-uncaging enables the precise release of neurotransmitters with both high spatial
and temporal control. In this strategy, neurotransmitters are chemically modified by attaching a light-sensitive "cage," rendering it inactive (Ellis-Davies, 2019). When exposed to a focused laser beam, the cage is photo-cleaved and removed, allowing for near-instantaneous release of the neurotransmitter. This has recently been used to mimic natural neurotransmission and cause strengthening or weakening of specific, singular synapses (Noguchi et al., 2019).
Photo-switchable ligands, on the other hand, are tethered to specific target receptors in such a way that their
binding is determined by light. Azobenzenes, which switch between cis and trans when exposed to specific light wavelengths, are often incorporated in such a strategy. Upon illumination, the azobenzene-ligand switches configuration and binds the receptor (Kienzler & Isacoff, 2017).
Controlling Specific Intracellular Signaling Pathways and Proteins
Various tools for modulating other cellular pathways using light have also been developed. Recent tools include those for the artificial production of proteins, such as adenylyl cyclase for cAMP (Stierl et al., 2011) or guanylyl cyclase for cGMP (Gao et al., 2015). There are also systems for degradation, such as the auxin-inducible-degron system which activates an expressed plant ubiquitin-ligase complex following exposure to the plant hormone auxin that then degrades all endogenous proteins tagged with degron (Yesbolatova et al., 2020).
Neuropixels and Recording Thousands of Neurons at Once
These systems generally involve a genetically engineered system of light-sensitive proteins that allow the expression of a recombinase to specific wavelengths of light. For example, one system developed in 2020 uses a split-Cre recombinase that would dimerize only when given blue light (Morikawa et al., 2020). By directing light to specific neurons, patterning the expression of specific proteins, including optogenetics, can be done with remarkable precision.
Infrared Optogenetics and Remote Control of Neurons
The penetration depth of visible light is limited by scattering and tissue absorption. As a result, visible light can only czontrol superficial neurons and requires implanting optical fibers for deeper regions. Near-infrared light can overcome these limitations and allow remote control at greater depths within the brain. Combining this with upconverting nanoparticles (UCNPs), which convert infrared to shorter wavelength light, transcra- nial control of seizure silencing and memory recall was first demonstrated in 2018 (Chen et al., 2018)
4) Optogenetics Recording and Visualization
With the first version developed in 2001, GECIs (Genetically Encoded Calcium Indicators) are a widely used tool in neuroscience for monitoring intraneuronal calcium levels, a proxy for neuronal activity and neu- rotransmitter vesicle release. Since then, many optical sensors for measuring other facets of neural activity have been developed.
GEVIs and Non-Invasive Optical Visualization of Electrical Potentials
While modern GECIs can report activity from thousands of cells simultaneously at high signal-to-noise ratios, they are still insensitive to precise spike timing and cannot report subthreshold activity. Genetically encoded voltage indicators (GEVIs) report changes in membrane potential with differences in cellular fluorescence, allowing for the optical detection of fast action potentials and subthreshold changes in membrane potential. Because voltage changes in neurons happen rapidly and therefore require high millisecond temporal precision and extremely high signal-to-noise ratios, these tools have historically been inadequate for precise measurements of membrane potential at cellular resolution. Nevertheless, recent progress has resulted in new GEVIs with much-improved sensitivity and temporal resolution. The first modern GEVIs were fusions of voltage-sensitive domains with fluorescent proteins (FPs). These include the currently only known GEVIs to work under 2-photon microscopy, the ASAP family (Evans et al., 2023), based on fusing a phosphatase domain with a circularly permuted green fluorescent pro- tein (GFP) in between.
Alternatively, rhodopsin-based GEVIs, such as the Arch family, are derived from microbial rhodopsins with voltage-dependent near-infrared (NIR) fluorescence. This shift away from the blue light spectrum allows for usage combinations with other blue/green light-dependent tools, including blue-light activated channelrhodopsin (ChR2) which stimulates action potentials in neurons. Nevertheless, opsin-based GEVIs typically have low fluorescence and quantum yield, meaning they require much higher intensity light for absorption, which combined with the high temporal sampling necessary in voltage recording, can result in significant photobleaching. Photobleaching is when photons, by chance, excite electrons to slightly too stable high energy states, gradually resulting in the formation of extremely reactive species, such as free radicals, that irreversibly damage the fluorescent protein’s molecular structure. Thus, FRET (Fluorescence Resonance Energy Transfer)-based variants of opsin GEVIs exist to require less light delivery. FRET is when energy from a higher energy “donor” fluorophore (e.g. a more blue-colored one) excites an electron in a lower energy “acceptor” fluorophore (a more red-colored one), resulting in increased fluorescence of the acceptor over the donor. Examples of FRET-based variants include Ace-mNeon and VARNAM (Kannan et al., 2022) which utilizes FRET between the rhodopsins with other red and green FPs, respectively, and the Voltron family (Abdelfattah et al., 2020) which adds HaloTag, a protein domain for capturing dyes, for FRET with exogenously injected and bound fluorescent dyes.
Both voltage-sensitive domain-based and FRET rhodopsin-based GEVIs traditionally detected depolarizations as fluorescence decreases. Very recently, to require less light illumination intensity during periods of voltage responses and enable easier separation of signals from the low-intensity auto-fluorescence background, positively tuned sensors have also been developed. These include Positron (Abdelfattah et al., 2020), based on Voltron, and ASAP4 sensors (Evans et al., 2023), from the ASAP family.
Neurotransmitter Sensors and Watching Neurotransmission Over Space and Time
Starting in 2013 with iGluSnFR for detecting glutamate (Marvin et al., 2013), the development of genetically encoded fluorescent sensors for various specific neurotransmitters and neuropeptides has greatly increased. These sensors allowed for the real-time, direct observation of endogenous neurotransmission with millisecond and micrometer precision. Two main categories
have been used for neurotransmitter sensors: 1) those utilizing bacterial periplasmic binding proteins (PBPs), which bacteria use to sense small molecules and neurotransmitters, as scaffolds, and 2) those using G-protein Coupled Receptors (GPCRs), which are endogenous transmembrane receptors for neurotransmitters and neuropeptides that eukaryotes use for cell signaling, as scaffolds.
PBP scaffolds are appealing because they undergo significant and consistent structural changes when they bind to specific ligands. PBPs consist of two parts connected by a hinge region with a ligand-binding site located between them. Then, intensity-based sensors for monitoring ligand changes in real-time are made by connecting the two PBP parts with a circularly permuted fluorescent protein (cpFP) in between the parts (Marvin et al., 2013). Using this design approach, sensors have been developed for glutamate, acetylcholine, GABA, nicotine, ATP, glucose, and, most recently, serotonin.
On the other hand, GPCRs are the most extensive and varied category of membrane receptors within eukary- otes and are naturally expressed in the human/mouse brain. Current knowledge of GPCR structures suggests that the most significant structural change upon ligand binding occurs in the intracellular loop 3 (ICL3). Neurotransmitter sensors created by leveraging this conformational change and fusing cpFPs in this region include those for acetylcholine, norepinephrine, serotonin, adenosine, and ATP, in addition to neuropeptides endocannabinoids, somatostatin (SST), cholecystokinin (CCK), corticotropin releasing factor (CRF), neuropeptide Y (NPY), neurotensin (NTS), and vasoactive intestinal
peptide (VIP) (Wu et al., 2022).
Fluorescence Lifetimes and Absolute, Quantitative Measurements
Unlike traditional fluorescence microscopy, which relies on the intensity, number of photons of emitted light, Fluorescence Lifetime Imaging Microscopy (FLIM) measures the time it takes for fluorophores to return to their ground state after being excited by a light source. While intensity-based fluorescence therefore depends on, and must be normalized for, fluorescent sensor expression and density, photon lifetimes do not depend on sensor expression or photobleaching. In this way, fluorescent sensors whose distributions of photon lifetimes depend on the absolute, quantitative concentration of their target molecule can be used to measure and compare across different sessions over any duration of time in any sample, regardless of how genetic transcription of the sensor or sensor degradation may have dramatically changed. Conversely, because shorter lifetimes will contribute to perceptions of higher intensity, many FLIM sensors can likewise be used as intensity-based sensors. Two markers of cellular respiration and metabolism, FAD and NADH, are endogenously auto-fluorescent, albeit at extremely short wavelengths of 450 and 340 nm, respectively. Interestingly, versions of the red Ca2+ indicator, RCaMP, already have Ca2+ concentration-dependent functions of fluorescence lifetime, despite not being originally developed with lifetime sensitivity in mind. Other deliberately engineered FLIM sensors have been made for Ca2+ (a turquoise one) (Van Der Linden et al., 2021), glucose (Díaz‐García et al., 2019), and cAMP (Tewson et al., 2016), a secondary messenger for receptors.
Another common usage of FLIM is with FRET. The fact that FRET needs two fluorophores of different colors makes using these sensors with other sensors difficult (since two sensors of the same color cannot be used for intensity measurements), and intensity depends on relative fluorophore expression. However, as it turns out, donor fluorescence lifetime decreases when FRET occurs. FRET efficiency depends on how far away the acceptor and donor fluorophores are and therefore can be easily adapted for sensing either conformational changes in one protein or multi-protein-to-protein interactions. FLIM-FRET sensors include those for secondary messenger PKA and CREB, a transcription factor for plasticity (Laviv & Yasuda, 2021).
One limitation with FLIM is that since photon lifetimes are a probabilistic distribution, this must be fit or estimated to recorded photons for a measurement. FLIM therefore requires high numbers of accumulated photons to be accurate, and long periods of high-intensity light exposure are not ideal for cells, especially at UV ranges.
Conclusion
In recent years, the landscape of neuroscience has undergone a transformative shift, with strategies for chemogenic control, electrical recording and control, optogenetics-based control, and optogenetics recording and visualization creating a diverse and robust toolset for understanding and driving functionally specific neural circuits in live animals. Basic metal electrodes are still used for clinical deep-brain stimulation (DBS), and early optogenetic opsins are just now being incorporated into clinical trials for vision restoration. Thus, advances in modern neurotechnology not only represent the potential for a greater understanding of neurobiology and behavior in animals but also the possibility for better, targeted therapeutics for more complex circuit-based neurological and psychiatric diseases in humans.
References