Biology subjects

Akrami, A.

Publications and source records attributed to Akrami, A..

15 recordsLinked to original sources

bioRxiv2026

From sequences to schemas: low-rank recurrent dynamics underlie abstract relational representations

A hallmark of intelligent behavior is the ability to extract abstract relational structure from temporal sequences -- recognizing, for instance, that aab, ccd, and eef all follow the same underlying pattern, regardless of the specific elements involved. This capacity, observed across species and sensory modalities, is thought to underlie the formation of cognitive schemas: compressed internal models that support rapid generalization to novel experiences. Yet the neural circuit mechanisms by which such abstract, identity-independent representations emerge from sequential experience remain largely unknown. Here, we investigate this question using Recurrent Neural Networks (RNNs) as mechanistic models of neural circuits, trained to classify sequences based on their latent algebraic patterns (e.g., aab, aad [->] AAB; aba, aca [->] ABA) without supervision on intermediate transitions. We demonstrate that RNNs spontaneously learn low-dimensional representations that mirror the hierarchical generative structure of the sequences, and that this abstraction is mechanistically supported by the emergence of low-rank recurrent connectivity. The leading singular component of the recurrent weight matrix integrates relational transition information -- whether consecutive tokens are the same or different -- across time, driving the formation of a structured, tree-like geometry in the population state space. Through singular vector ablation, we establish a causal role for this component: removing it selectively erases memory for earlier transitions while leaving local, single-step sensitivity intact. Finally, while RNNs trained on next-token prediction do not spontaneously acquire these abstract representations, transferring the low-rank scaffold learned from classification significantly accelerates learning and improves generalization -- an effect specific to the abstract structure of the scaffold rather than generic statistical pretraining. These findings offer a computational account of how task demands shape recurrent connectivity to support temporal abstraction, with direct implications for understanding schema formation in biological brains.

neuroscience↗

bioRxiv2026

The hippocampus enables abstract structure learning without reward

Statistical learning (SL) allows organisms to infer latent structure from sensory input without instruction, feedback, or reward, yet how the brain accomplishes such abstract, unsupervised learning remains unknown. Here we show that mice, like humans, rapidly acquire multiple forms of statistical structure, including event frequency, sequence identity, and abstract structural rules, and that the hippocampus is essential for this capacity. Pupil dynamics provided a cross-species, implicit readout of expectation formation, revealing spontaneous sensitivity to these regularities during passive listening. In mice, pharmacological and temporally precise optogenetic inactivation of dorsal CA1 abolished all learning-related pupil signatures without affecting baseline pupil size, target-evoked responses, or task performance, demonstrating a causal requirement for the hippocampus in forming and updating internal models of sensory structure. High-density recordings further revealed that dCA1 ensembles track evolving statistical contexts while dynamically reorganising population activity into subspaces that separately encode sensory features and abstract rules, enabling generalisation across distinct but structurally equivalent sequences. Together, these results identify the hippocampus as a critical neural substrate for latent abstract structure learning and offer a mechanistic account of how internal models emerge from unsupervised experience.

neuroscience↗

bioRxiv2025

Acetylcholine reflects uncertainty during hidden state inference

To act adaptively, animals must infer features of the environment that cannot be observed directly, such as which option is currently rewarding, or which context they are in. These internal estimates, known as hidden states, guide behaviour but are inherently uncertain. Theoretical models propose that efficient inference requires tracking both the most likely state and the uncertainty of that estimate. While neural representations of state identity have been described in cortical and hippocampal circuits, the origin of uncertainty signals remains unclear. Here we show that acetylcholine (ACh) reflects this uncertainty during hidden state inference. Using fibre photometry in mice performing a probabilistic two-armed bandit task, we found that ACh release in medial prefrontal cortex and ventral hippocampus tracked recent omission history and predicted whether animals would switch choice on the next trial. Blocking muscarinic receptors with scopolamine selectively impaired loss-driven switching without affecting stable performance. A hidden-state inference model in which ACh modulated how uncertainty from past experience shaped future beliefs reproduced both the physiological and behavioural effects, whereas alternative models did not. These results identify ACh as a neuromodulatory signal reflecting uncertainty during hidden-state inference, linking theoretical models of belief updating to their biological implementation in the mammalian brain.

neuroscience↗

bioRxiv2024

Reduced monitoring of task performance is an effective biomarker of autism

Can reduced flexibility in autistics behavior be a reliable classifier of autism? We asked this question in the context of pitch discrimination, which is performed with similar accuracy by autistic and non-autistic participants, matched for cognitive skills. We found that three reliable, uncorrelated characteristics quantified reduced behavioral flexibility. Most non-autistic participants showed a clear feedback-related negativity (FRN) in their EEG when incorrect, larger EEG responses during feedback following difficult versus easy trials, and increased perceptual bias following correct and easy trials versus difficult ones. All three measures were weaker in autistic participants. To assess reliability, we replicated all three measures in an experiment approximately a year later. These measures were not correlated within participants, indicating that several mechanisms, associated with anterior cingulate cortex, are atypical in autism. Together, these three measures provide group classification with above 80% accuracy, showing promise for their utility as a neurocognitive biomarker of autism. Significance statementA characteristic of autism is cognitive "slowness" in updating expectations in new contexts. Our study shows that this cognitive difference is related to a reduced use of feedback, linked to atypical activity in a brain area involved in error detection and behavioral adjustment. Using EEG, we found that autistic individuals have weaker feedback-related signals and do not adjust their behavior as effectively as non-autistic individuals. Together, these neurocognitive signals could serve as a biomarker for autism. This is the first time such brain responses have been measured in a perceptual task requiring continuous adjustments. Understanding these brain mechanisms can inform strategies to enhance learning and adaptability in autism, providing new insights into the neural basis of behavioral flexibility.

neuroscience↗

bioRxiv2024

Repix: reliable, reusable, versatile chronic Neuropixels implants using minimal components

Neuropixels probes represent the state-of-the-art for high-yield electrophysiology in neuroscience: the simultaneous recording of hundreds of neurons is now routinely carried out in head-restrained animals. In contrast, neural recording in unrestrained animals, as well as recording and tracking neurons over days, remains challenging, though it is possible using chronic implants. A major challenge is the availability of simple methods that can be implemented with limited or no prior experience with Neuropixels probes, while achieving reliable, reusable, versatile high-density electrophysiology. Here we developed, deployed, and evaluated the real-world performance of Repix, a chronic implantation system that permits the repeated re-use of Neuropixels probes. The lightweight system allows implanted animals to express a full range of natural behaviors, including social behaviors. We show that Repix allows the recording of hundreds of neurons across many months, up to a year, with implants across cortical and subcortical brain regions. Probes can be reused repeatedly with stable yield. Repix has been used by 16 researchers in 10 laboratories to date, and we evaluated the real-world performance of Repix in a variety of chronic recording paradigms in both mice and rats with a combined 209 implantations. We found that the key advantage of Repix is robustness and simplicity. Adopters of Repix became proficient at five procedures on average, regardless of prior experience with in vivo electrophysiology. With the companion protocol alongside this article, the performance and user-friendliness of Repix should facilitate a wide uptake of chronic Neuropixels recordings.

neuroscience↗

bioRxiv2024

Zapit: Open Source Random-Access Photostimulation For Neuroscience

Optogenetic tools are indispensable for understanding the causal neural mechanisms underlying animal behavior,providing millisecond-precision control over genetically defined neural populations. Random-access laser-scanning optogenetics is a powerful and underutilized method for transiently manipulating cortical activity in mice. Despite the utility of this technique, no general-purpose open-source implementation currently exists, and potential users need expertise across optics, real-time hardware control, and programming. This represents a major barrier to adoption, particularly for newly established research groups. Here we present 'Zapit', the first open-source general-purpose platform for random-access laser-scanning optogenetic experiments in head-fixed mice. Zapit is fully documented, has a user-friendly GUI, works in stereotaxic coordinates, and comes with easy to build hardware options that extend functionality beyond any published system. We validate Zapit's performance through electrophysiological recordings and cortical photoinhibition in behaving mice. Zapit is a novel and innovative tool with the potential to democratize laser-scanning optogenetics and significantly increase uptake throughout the scientific community.

neuroscience↗

bioRxiv2023

DNA damage and nuclear morphological changes in cardiac hypertrophy are mediated by SNRK through actin depolymerization

BACKGROUNDProper nuclear organization is critical for cardiomyocyte (CM) function, as global structural remodeling of nuclear morphology and chromatin structure underpins the development and progression of cardiovascular disease. Previous reports have implicated a role for DNA damage in cardiac hypertrophy, however, the mechanism for this process is not well delineated. AMPK family of proteins regulate metabolism and DNA damage response (DDR). Here, we examine whether a member of this family, SNF1-related kinase (SNRK), which plays a role in cardiac metabolism, is also involved in hypertrophic remodeling through changes in DDR and structural properties of the nucleus. METHODSWe subjected cardiac specific (cs)-Snrk-/- mice to trans-aortic banding (TAC) to assess the effect on cardiac function and DDR. In parallel, we modulated SNRK in vitro and assessed its effects on DDR and nuclear parameters. We also used phospho-proteomics to identify novel proteins that are phosphorylated by SNRK. Finally, co-immunoprecipitation (co-IP) was used to verify Destrin (DSTN) as the binding partner of SNRK that modulates its effects on the nucleus and DDR. RESULTScs-Snrk-/- mice display worse cardiac function and cardiac hypertrophy in response to TAC, and an increase in DDR marker pH2AX in their hearts. Additionally, in vitro Snrk knockdown results in increased DNA damage and chromatin compaction, along with alterations in nuclear flatness and 3D volume. Phospho-proteomic studies identified a novel SNRK target, DSTN, a member of F-actin depolymerizing factor (ADF) proteins that directly binds to and depolymerize F-actin. SNRK binds to DSTN, and DSTN downregulation reverses excess DNA damage and changes in nuclear parameters, in addition to cellular hypertrophy, with SNRK knockdown. We also demonstrate that SNRK knockdown promotes excessive actin depolymerization, measured by the increased ratio of globular (G-) actin to F-actin. Finally, Jasplakinolide, a pharmacological stabilizer of F-actin, rescues the increased DNA damage and aberrant nuclear morphology in SNRK downregulated cells. CONCLUSIONSThese results indicate that SNRK is a key player in cardiac hypertrophy and DNA damage through its interaction with DSTN. This interaction fine-tunes actin polymerization to reduce DDR and maintain proper CM nuclear shape and morphology. Clinical PerspectiveO_LIWhat is new? O_LIAnimal hearts subjected to pressure overload display increased SNF1-related kinase (SNRK) protein expression levels and cardiomyocyte specific SNRK deletion leads to aggravated myocardial hypertrophy and heart failure. C_LIO_LIWe have found that downregulation of SNRK impairs DSTN-mediated actin polymerization, leading to maladaptive changes in nuclear morphology, higher DNA damage response (DDR) and increased hypertrophy. C_LI C_LIO_LIWhat are the clinical implications? O_LIOur results suggest that disruption of DDR through genetic loss of SNRK results in an exaggerated pressure overload-induced cardiomyocyte hypertrophy. C_LIO_LITargeting DDR, actin polymerization or SNRK/DSTN interaction represent promising therapeutic targets in pressure overload cardiac hypertrophy. C_LI C_LI

cell biology↗

bioRxiv2023

Heron: A Knowledge Graph editor for intuitive implementation of python based experimental pipelines.

To realise a research project idea, an experimenter faces a series of conflicting design and implementation considerations, regarding both its hardware and software components. For instance, the ease of implementation, in time and expertise, should be balanced against the ease of future reconfigurability and number of black box components. Other, often conflicting, considerations include the level of documentation and ease of reproducibility, resource availability as well as access to online communities. To alleviate this balancing act between opposing requirements we present Heron, a new Python-based platform to construct and run experimental and data analysis pipelines. Herons main principle is to allow researchers to design and implement the experimental flow as close as possible to their mental schemata of the experiment, in the form of a Knowledge Graph. Heron is designed to increase the implementation speed of experiments (and their subsequent updates), while minimising the number of incorporated black box components. It enhances the readability and reproducibility of the final implementation and allows the use of combinations of hardware and software otherwise impossible or too costly to achieve. Given this, Heron offers itself to sciences whose needs involve experiments with a large number of interconnected hardware and software components like robotics, neuroscience, behavioural sciences, physics, chemistry, environmental science, etc.. It is designed with those experimentalists in mind which: i) Demand full control of their setup. ii) Prefer not to have to choose between hardware and software that run only on a specific chip/operating system combination. iii) Appreciate the ease and speed that high-level languages (e.g. Python) and Graphical User Interfaces (GUIs) offer them. It assumes an intermediate knowledge of the Python language and ecosystem, offering a flexible and powerful way to construct experimental setups. It removes any inaccessible corners, yet keeps implementation costs significantly reduced compared to using lower level languages. Finally, its use results in a much cleaner and easier to understand code base, amicable to documentation and reproducibility efforts.

animal behavior and cognition↗

bioRxiv2023

Humans, rats and mice show species-specific adaptations to sensory statistics in categorisation behaviour

Animals must exploit environmental regularities to make adaptive decisions, yet the learning algorithms that enabels this flexibility remain unclear. A central question across neuroscience, cognitive science, and machine learning, is whether learning relies on generative or discriminative strategies. Generative learners build internal models the sensory world itself, capturing its statistical structure; discriminative learners map stimuli directly onto choices, ignoring input statistics. These strategies rely on fundamentally different internal representations and entail distinct computational trade-offs: generative learning supports flexible generalisation and transfer, whereas discriminative learning is efficient but task-specific. We compared humans, rats, and mice performing the same auditory categorisation task, where category boundaries and rewards were fixed but sensory statistics varied. All species adapted their behaviour near-optimally, consistent with a normative observer constrained by sensory and decision noise. Yet their underlying algorithms diverged: humans predominantly relied on generative representations, mice on discriminative boundary-tracking, and rats spanned both regimes. Crucially, end-point performance concealed these differences--only learning trajectories and trial-to-trial updates revealed the divergence. These results show that similar near-optimal behaviour can mask fundamentally different internal representations, establishing a comparative framework for uncovering the hidden strategies that support statistical learning.

animal behavior and cognition↗

bioRxiv2023

Functional gene delivery to and across brain vasculature of systemic AAVs with endothelial-specific tropism in rodents and broad tropism in primates

Delivering genes to and across the brain vasculature efficiently and specifically across species remains a critical challenge for addressing neurological diseases. We have evolved adeno-associated virus (AAV9) capsids into vectors that transduce brain endothelial cells specifically and efficiently following systemic administration in wild-type mice with diverse genetic backgrounds and rats. These AAVs also exhibit superior transduction of the CNS across non-human primates (marmosets and rhesus macaques), and ex vivo human brain slices although the endothelial tropism is not conserved across species. The capsid modifications translate from AAV9 to other serotypes such as AAV1 and AAV-DJ, enabling serotype switching for sequential AAV administration in mice. We demonstrate that the endothelial specific mouse capsids can be used to genetically engineer the blood-brain barrier by transforming the mouse brain vasculature into a functional biofactory. Vasculature-secreted Hevin (a synaptogenic protein) rescued synaptic deficits in a mouse model.

bioengineering↗

bioRxiv2022

From recency to the central tendency bias in working memory: a unifying attractor network model

The central tendency bias, or contraction bias, is a phenomenon where the judgment of the magnitude of items held in working memory appears to be biased towards the average of past observations. It is assumed to be an optimal strategy by the brain, and commonly thought of as an expression of the brains ability to learn the statistical structure of sensory input. On the other hand, recency biases such as serial dependence are also commonly observed, and are thought to reflect the content of working memory. Recent results from an auditory delayed comparison task in rats, suggest that both biases may be more related than previously thought: when the posterior parietal cortex (PPC) was silenced, both short-term and contraction biases were reduced. By proposing a model of the circuit that may be involved in generating the behavior, we show that a volatile working memory content susceptible to shifting to the past sensory experience - producing short-term sensory history biases - naturally leads to contraction bias. The errors, occurring at the level of individual trials, are sampled from the full distribution of the stimuli, and are not due to a gradual shift of the memory towards the sensory distributions mean. Our results are consistent with a broad set of behavioral findings and provide predictions of performance across different stimulus distributions and timings, delay intervals, as well as neuronal dynamics in putative working memory areas. Finally, we validate our model by performing a set of human psychophysics experiments of an auditory parametric working memory task.

neuroscience↗

bioRxiv2020

Extracting the Dynamics of Behavior in Decision-Making Experiments

Understanding how animals update their decision-making behavior over time is an important problem in neuroscience. Decision-making strategies evolve over the course of learning, and continue to vary even in well-trained animals. However, the standard suite of behavioral analysis tools is ill-equipped to capture the dynamics of these strategies. Here, we present a flexible method for characterizing time-varying behavior during decision-making experiments. We show that it successfully captures trial-to-trial changes in an animals sensitivity to not only task-relevant stimuli, but also task-irrelevant covariates such as choice, reward, and stimulus history. We use this method to derive insights from training data collected in mice, rats, and human subjects performing auditory discrimination and visual detection tasks. With this approach, we uncover the detailed evolution of an animals strategy during learning, including adaptation to time-varying task statistics, suppression of sub-optimal strategies, and shared behavioral dynamics between subjects within an experimental population.

neuroscience↗

bioRxiv2019

Orbitofrontal cortex promotes trial-by-trial learning of risky, but not spatial, biases

Individual choices are not made in isolation but are embedded in a series of past experiences, decisions, and outcomes. The effects of past experiences on choices, often called sequential biases, are ubiquitous in perceptual and value-based decision-making, but their neural substrates are unclear. We trained rats to choose between explicitly cued guaranteed and probabilistic rewards in a task in which outcomes on each trial were independent. Behavioral variability often reflected the effects of previous trials, including increased willingness to take risks following risky wins, and spatial \"win-stay/lose-shift\" biases. Electrophysiological recordings from lateral orbitofrontal cortex (lOFC) revealed strong encoding of reward history and receipt, and optogenetic inhibition of lOFC during choice reporting eliminated rats increased preference for risk on subsequent trials following risky wins, but spared other sequential effects. Our data show that different sequential biases are neurally dissociable, and that the lOFCs role in adaptive behavior promotes learning of more abstract, task-specific biases (here, biases for the risky option), but not spatial ones. These data are consistent with proposals that the OFC represents the animals location in an abstract representation of task states, and highlight the lOFCs role in learning.

neuroscience↗

bioRxiv2018

Collicular circuits for flexible sensorimotor routing

Historically, cognitive processing has been thought to rely on cortical areas such as prefrontal cortex (PFC), with the outputs of these areas modulating activity in lower, putatively simpler spatiomotor regions, such as the midbrain superior colliculus (SC). Using a rat task in which subjects switch rapidly between task contexts that demand changes in sensorimotor mappings, we report a surprising role for the SC in non-spatial cognitive processes. Before spatial response choices could be formed, neurons in SC encoded task context more strongly than neurons in PFC, and bilateral SC silencing impaired behavioral performance. Once spatial choices could begin to be formed, SC neurons encoded the choice faster than PFC, while bilateral SC silencing no longer impaired choices. A set of dynamical models of the SC replicates our findings. Our results challenge cortically-focused views of cognition, and suggest that ostensibly spatiomotor structures can play central roles in non-spatiomotor cognitive processes.

neuroscience↗

bioRxiv2017

Posterior parietal cortex represents sensory stimulus history and is necessary for its effects on behavior

Many models of cognition and of neural computations posit the use and estimation of prior stimulus statistics1-4: it has long been known that working memory and perception are strongly impacted by previous sensory experience, even when that sensory history is irrelevant for the current task at hand. Nevertheless, the neural mechanisms and brain regions necessary for computing and using such priors are unknown. Here we report that the posterior parietal cortex (PPC) is a critical locus for the representation and use of prior stimulus information. We trained rats in an auditory Parametric Working Memory (PWM) task, and found that rats displayed substantial and readily quantifiable behavioral effects of sensory stimulus history, similar to those observed in humans5,6 and monkeys7. Earlier proposals that PPC supports working memory8,9 predict that optogenetic silencing of this region would impair behavior in our working memory task. Contrary to this prediction, silencing PPC significantly improved performance. Quantitative analyses of behavior revealed that this improvement was due to the selective reduction of the effects of prior sensory stimuli. Electrophysiological recordings showed that PPC neurons carried far more information about sensory stimuli of previous trials than about stimuli of the current trial. Furthermore, the more information about previous trial sensory history in the neural firing rates of a given rats PPC, the greater the behavioral effect of sensory history in that rat, suggesting a tight link between behavior and PPC representations of stimulus history. Our results indicate that the PPC is a central component in the processing of sensory stimulus history, and open a window for neurobiological investigation of long-standing questions regarding how perception and working memory are affected by prior sensory information.

neuroscience↗