Ultra-Rare Variants in Mitochondrial DNA Identified in Bipolar Disorder

Post by Amanda Engstrom

The takeaway

Mitochondrial dysfunction has been thought to play a role in various neuropsychiatric disorders. In bipolar disorder, there is an enrichment of a small number of ultra-rare, potentially pathogenic mitochondrial DNA variants that could be directly targeted to improve mitochondrial function in the brain. 

What's the science?

Bipolar disorder (BD) is a major psychiatric disorder characterized by recurrent manic and depressive episodes. Despite being highly heritable, genomic studies have not been able to specify the mechanism underlying BD etiology. However, different variants, or changes to the DNA sequence, of mitochondrial DNA have been suggested to be a factor in BD. The mitochondria play a critical role in the function of all cells and have been theoretically linked to neuronal dysfunction due to their high requirement for energy processing. This week in Molecular Psychiatry, Ohtani and colleagues used advanced DNA sequencing technology to investigate the contribution of brain mitochondrial DNA variants to BD’s pathophysiology. 

How did they do it?

The authors analyzed brain DNA from 54 BD patients, 55 schizophrenia patients, and 54 healthy controls to investigate the association between BD and mitochondrial DNA variants. They used duplex molecular barcode sequencing, a high-precision technique that tags each DNA molecule to detect rare mitochondrial DNA variants with single-molecule resolution. This approach allowed for the identification of heteroplasmic variants, where both normal and mutated mitochondrial DNA are present within the same cell. They then calculated the Variant Allele Frequency (VAF), that is the percentage of DNA sequences in a sample that carry the specific genetic variant. Using both bulk analysis and single-molecule analysis pipelines, they could detect variants with either high and moderate-VAF or low VAF, respectively. The authors could then determine which variants are present and the frequency of each variant in healthy controls compared to BP or schizophrenia patients. 

The authors hypothesized two potential models for the mode of association between BD and mitochondrial variants in the brain:

1- A limited number of likely pathogenic variants with a high VAF would cause mitochondrial dysfunction and impair neuronal activity.

2- Accumulation of numerous non-specific variants with low VAF progressively degrades mitochondrial function and impairs neuronal activity. 

What did they find?

Using bulk sequencing analysis, the authors identified 116 heteroplasmic variants with high/moderate VAF. Thirty-six of them were ultra-rare variants (rarely seen in healthy controls). Both BD and schizophrenia brain mitochondrial DNA had more variants compared to controls, but when limiting the analysis to the ultra-rare variants, only BD had an increase in variants. Through further investigation of these ultra-rare variants, the authors identified potentially pathogenic mutations such as 1) the m.3243A>G mutation, which is causative of the mitochondrial disorder MELAS, 2) four loss-of-function mutations, and 3) six rRNA variants with scores for high pathogenicity. To detect low-VAF variants, the authors used single-molecule analysis. They identified a total of 52,312 low-VAF heteroplasmic variants, however, there was no difference in the frequency of these variants between BD, schizophrenia, and controls. There was an association with age across all groups, with the frequencies increasing with age, reinforcing prior evidence of age-associated systematic mutation rate increase. Additionally, when analyzing the mutation patterns at the single nucleotide level, there was no difference in the mutation pattern among the three groups. Together, these data suggest the mitochondrial variants in BD are due to a consistent specific mutagenic mechanism and not due to a general increase in DNA variability, supporting model 1 from the authors' original hypothesis. Thus, a subset of BD patients could harbor ultra-rare, potentially pathogenic, mitochondrial DNA mutations in their brain tissue, which could be impacting the pathophysiology of their disease. 

What's the impact?

This study was the first to examine mitochondrial variant accumulation at single-molecule resolution in mood disorders. Ultra-rare loss-of-function and rRNA variants are novel candidates for further investigation into the role of mitochondrial dysfunction in BD. There are already several compounds targeting mitochondrial function that could be useful therapies for the treatment of patients with BD.

Access the original scientific publication here

Neural Responses to Internal and External Signals Predict Coma Recovery

Post by Shireen Parimoo

The takeaway

During wakefulness, the brain simultaneously processes both internally generated signals, such as the heartbeat, and external sensory stimuli, like sounds. Patients in a comatose state who later recover from the coma show preserved regularity of neural responses to cardiac and auditory signals. 

What's the science?

Interoception is the ability to sense signals generated by the body, such as the heartbeat or the sensation of goosebumps. The heartbeat evoked potential (HEP) is a specific neural response to heartbeats and is indicative of the brain’s processing of cardiac signals. The brain also tracks sensory input from the external world, like sounds, showing altered patterns of activity when there is deviation from an expected pattern or regularity. This change in response to the deviation is called a prediction error. Interestingly, during both wakefulness and sleep, neural responses to internally generated signals like the heartbeat track neural responses to externally generated signals like sounds in the environment. However, it is unclear whether neural responses to cardiac signals influence sensory processing during a deeply unconscious state. This week in PNAS, Pelentritou and colleagues used electrophysiological techniques to investigate whether the brain uses cardiac signals to track auditory input in a comatose state, and its relationship to patient outcomes. 

How did they do it?

The authors recorded brain and cardiac activity from 48 patients who had suffered cardiac arrest and entered a comatose state. In an auditory paradigm, patients were exposed to sounds and silences with varying levels of regularity relative to the heartbeat. There were four conditions: (1) baseline or control condition with no sound; (2) synchronous condition in which a sound occurred at a fixed interval after a heartbeat was detected; (3) isosynchronous condition in which a sound occurred at a fixed interval relative to the previous sound, but was not synchronized to the heartbeat; and (4) asynchronous condition in which sounds were presented irregularly relative to other sounds and to the heartbeat. Importantly, sounds were omitted on 20% of the trials, which allowed the authors to determine if the patients’ neural and cardiac responses showed evidence of prediction error (i.e., deviation from regularity) in any of the conditions. 

Electroencephalography was used to record neural responses, which included auditory evoked potentials (AEPs) in response to sound onset and omission-evoked potentials (OEPs) on omission trials, recorded relative to when the sound would have occurred. Cardiac activity was recorded using electrocardiography, including HEPs and omission HEPs during sound-on and omission trials, respectively. Here, OEPs and OHEPs were used as indicators of a prediction error response. The authors compared the regularity of neural and cardiac responses across conditions and for patients with a favorable outcome (i.e., recovery from coma) and an unfavorable outcome. Next, they used a support vector machine classifier (a machine learning technique) on neural data from each trial to predict which condition the brain activity belonged to and whether it could be used to predict patient outcome. Finally, they measured cardiac deceleration – or the amount of slowing between heartbeats – in response to sound omissions in the synchronous condition to predict whether a patient would recover from the coma. 

What did they find?

Patients with favorable outcomes showed a significant difference in OEPs during omissions in the synchronous condition, as compared to the asynchronous and baseline conditions. In patients with unfavorable outcomes, however, there was no difference in OEPs across the four conditions. Moreover, there was no difference in OEPs between the isosynchronous and baseline conditions in either patient group. This means that deviation from the regularity of external sounds relative to the heartbeat, but not to sounds, disrupts cardiac-auditory regularity of neural responses, but only in patients who later recover from the coma. 

Single-trial neural activity was predictive of patient outcomes. Specifically, patients with a favorable outcome showed greater cardiac-auditory regularity in the synchronous condition compared to the baseline condition. Relatedly, sound omissions in the synchronous condition influenced cardiac deceleration, but this effect was only observed in patients with a favorable outcome. The same effect was observed in the isosynchronous condition as well, but to a smaller extent. Thus, deviation from the regularity of auditory input – both in relation to the heartbeat and to previous sensory input – led to temporary slowing in between heartbeats of patients who went on to recover from the coma. 

What's the impact?

These results demonstrate that the brain uses internally generated signals to monitor sensory processing, even in deep unconsciousness, like a coma. Notably, the degree of neural synchronization in response to these signals predicts patient outcomes, offering a promising prognostic marker for coma recovery. 

Why Is Our Memory Gist-Like?

Post by Lila Metko

The takeaway

Engram cells are neurons that activate when a memory is formed and reactivate when a memory is recalled. Recollection is not always perfect, and sometimes these cells are activated under similar, but not identical, contexts to those in which the memory was formed. Formation of new cells in the hippocampus is necessary for this gist-like type of memory processing.  

What's the science?

The hippocampus (HC) is a brain region that consolidates and retrieves memories. For our survival, our memories must be more gist-like rather than very precise, so we can flexibly adapt to changing circumstances. Previous theory suggests that interactions between the prefrontal cortex and HC drive gist memory formation. This week in Nature, Ko and colleagues used optogenetic silencing, eGRASP visualization, and other methods to eliminate and accelerate neurogenesis to understand how gist memory formation can occur within the HC. 

How did they do it?

Experiment 1: The authors used a contextual fear conditioning paradigm to test memory in mice at 1 day (recent), 14 days (intermediate), and 28 days (remote). After mice are placed in a new environment (‘context A’), they will typically freeze if placed in an environment they associate with the stimulus again. One benefit of this paradigm is the ability to make a second environment (‘context B’) similar to context A so that they could test for gist memory. During the fear learning session, they labeled active neurons (engram neurons) with a fluorescent protein, and then quantified them for activity at each time point. Engram neurons were also silenced at the timepoints to determine effects on memory.

Experiment 2: The authors visualized engram cell synapses using the eGRASP technique to gain a better understanding of which subparts of the HC were involved in the engram reactivation and which neuron types played a role.  

Experiment 3: The authors then did a tracing experiment to label newborn neurons in the dentate gyrus region of the HC, to examine if they synapsed on a nearby engram cell. Finally, they used gamma irradiation and voluntary wheel running, respectively, to eliminate and boost neurogenesis in different cohorts of mice and examined memory in the contextual fear conditioning paradigm in each group. 

What did they find?

Experiment 1: The authors found that initially, the mice froze mostly in response to context A (the context in which they received the aversive stimulus), but by the 28th day, froze equally to the two similar contexts, indicating a decrease in precise memory. In some areas of the HC, engram cell activity mirrored the freezing patterns, showing high activity for A at timepoint one and equal activity for A and B at timepoint three. Silencing engram cells that project from one specific area of the HC to another suppresses freezing behavior in both context A and B at the 28-day timepoint, which indicates that these specific cells are responsible for gist memory. 

Experiment 2: In the experiment that labeled the engram cells, they found that over time, outputs from the dentate gyrus region of the HC to inhibitory neurons in the CA3 region decreased, and inputs to the CA1 region of the hippocampus from the CA3 increased. This indicates that there may be complementary feed-forward inhibition and excitation processes at play to facilitate gist memory. 

Experiment 3: The tracing experiment showed that newborn neurons from the dentate gyrus do synapse onto CA3 engrams. Importantly, CA3 engram cells that received inputs from newborn neurons were around three times more likely to be activated in context B (a similar but not identical context). When newborn neurons were eliminated, precise memory (more freezing to context A) increased at later timepoints, and the hippocampal connectivity patterns associated with later timepoints in experiment 2 were not seen at 28 days. Conversely, when neurogenesis was promoted with voluntary wheel running, precise memory went away at earlier timepoints (14 days), and the gist memory hippocampal connectivity changes that typically need 28 days to develop were seen as early as 14 days. This demonstrates that hippocampal neurogenesis likely facilitates gist memory. 

What's the impact?

This research shows that hippocampal neurogenesis can actively reshape memory circuits, shifting detailed event memories into flexible gist representations. It suggests that forgetting may not always be a bad thing, but more of an adaptive generalization of past experiences to new situations. This insight could influence strategies for education, mental health therapies, and age-related memory care by targeting neurogenesis or circuit remodeling to fine-tune the balance between precision and generalization in memory.

Access the original scientific publication here.