Neuroscience Research

Age of first alcohol use is associated with alterations in brain development and an increased risk of problematic substance use later in life. However, it remains unclear if neurobiological alterations associated with first-use age are present in young adulthood and how these alterations may be associated with different patterns of adult substance use. In this study, we employed SuperBigFLICA, a semi-supervised multimodal imaging data fusion approach, to identify joint structural morphometric and functional MRI (fMRI) brain profiles present in adulthood that are most predictive of age of alcohol first-use in the Human Connectome Project-Young Adult cohort (n=622). Our model achieved moderate performance (r=0.22 in external validation), identifying multimodal structure-function profiles predictive of first-use largely involving frontal regions. We then utilized Canonical Correlation Analysis to determine whether these multimodal brain profiles were differentially linked to patterns of adult substance use behaviors. We identified three significant brain modes (Mode 1: r = 0.2750, p = 0.014; Mode 2: r = 0.2249, p = 0.005; Mode 3: r = 0.1568, p = 0.029) associated with distinct profiles of alcohol and cannabis use. Together, we identified multimodal brain patterns in young adults that predict age of first use of alcohol and demonstrated that resulting multimodal structure-function brain profiles capture distinct multivariate patterns of adult substance use, providing a more integrated view of substance use neurobiology.

Objective. Intermittent theta burst stimulation (iTBS) is a repetitive transcranial magnetic stimulation (rTMS) protocol used in clinical settings to alleviate symptoms of psychiatric disorders. Although the pulse parameters currently used in iTBS have remained unchanged since the first treatments, recent controllable TMS devices allow for the personalization of pulse shape and duration, opening the possibility of testing the efficacy of other pulse parameters in modulating corticospinal excitability (CSE). Methods. In a repeated-measures design, changes in CSE were induced by the administration of iTBS over the left primary motor cortex under three different conditions: long bidirectional pulse (L-Bi), long quasi-unidirectional pulse (L-QU), and a control short quasi-unidirectional pulse (S-QU). We assessed TMS-induced motor evoked potentials before and up to 70 minutes after iTBS. We hypothesized that L-QU pulses would induce greater facilitation than other conditions. Results. A mixed linear model showed that all conditions induced increased MEPs compared to baseline. However, the L-Bi pulse led to a larger and more sustained MEP increase than both unidirectional pulses, while the S-QU pulse appeared less effective in modulating CSE. We observed large within- and between-subject variability in responsiveness to the iTBS protocols. Conclusion. These results suggest that L-Bi pulses in iTBS protocols are more potent than unidirectional pulses in inducing CSE changes in the cortex of healthy adults. The individual variability in response to iTBS and the potential clinical relevance of these findings are discussed.

Background Tobacco use during adolescence is an important public-health concern because early initiation increases the likelihood of continued tobacco use and dependence. This study assessed the prevalence of tobacco smoking and factors associated with ever smoking among high school students in Harar town, eastern Ethiopia. Methods A school-based cross-sectional study was conducted among 467 students in grades 9-12 from selected public and private high schools in Harar town. Data were collected using a self-administered questionnaire adapted from the Global Youth Tobacco Survey. Descriptive statistics and binary logistic regression were used. Variables were entered into a multivariable logistic regression model based on the prespecified analytical approach, and adjusted odds ratios (AORs) with 95% confidence intervals (CIs) were estimated. Statistical significance was declared at p<0.05. Results Among 467 students, 64 (13.7%; 95% CI 10.9-17.1%) reported ever smoking tobacco, while 25 (5.4%; 95% CI 3.7-7.8%) were current smokers. In the SPSS multivariable logistic regression model, grade 11 (AOR 10.61, 95% CI 3.52-31.97) and grade 12 (AOR 12.18, 95% CI 3.95-37.53), alcohol drinking (AOR 2.60, 95% CI 1.25-5.40), khat chewing (AOR 9.57, 95% CI 4.04-22.65), and access to tobacco (AOR 3.18, 95% CI 1.58-6.43) were independently associated with ever smoking. Male sex, age [&ge;]18 years, and grade 10 were not statistically significant in the adjusted model. Conclusion Ever smoking was reported by approximately one in seven students and current smoking by about one in twenty. Alcohol drinking, khat chewing, access to tobacco, and higher grade were independently associated with ever smoking. School- and community-based tobacco prevention should address substance use and the availability of tobacco products around schools.

Binge drinking is characterized by an initial bout of rapid, high-vigor alcohol intake (front-loading) followed by a prolonged period of lower-rate "maintenance" drinking, yet the circuit mechanisms that generate and regulate these phases remain unclear. Here, we develop a systems-level computational model that links the structure of binge drinking to interactions among medial prefrontal cortex (mPFC), insula, striatum, and ventral tegmental area (VTA). The model consists of cortical Seek (mPFC) and Go (insula) populations that form a mutually excitatory loop, a SetPoint population that integrates striatal activity as an internal estimate of effort/consumption, ventral (nucleus accumbens, NAc) and dorsal (dorsolateral striatum, DLS) striatal populations, and a dopaminergic VTA population providing transient reward-prediction signaling. Using a firing-rate formalism with noisy dynamics, we reproduce the canonical two-phase binge pattern and show that: (i) the PFC-insula loop acts as a bistable module that initiates and sustains front-loading; (ii) striato-cortical feedback to the Setpoint population controls the timing of the transition from front-loading to maintenance without altering its vigor; and (iii) the amplitude of the VTA dopamine transient selectively scales front-loading intensity. Sensitivity analyses identify PFC-insula connectivity, VTA-to-NAc modulation, and striato-cortical feedback as key parameters that bifurcate two-phase drinking into regimes with abolished or persistent front-loading. These findings provide a mechanistic framework for interpreting binge-like alcohol intake in terms of identifiable circuit motifs and generate testable predictions for targeting mPFC-insula-striatal-VTA interactions in alcohol use disorder.

Mammals entering torpor or hibernation undergo an abrupt transition from normothermia to hypothermia, yet how thermoregulation enables this switch remains poorly understood. Here, we identify dynamical signatures that precede these transitions and a mathematical principle that can generate them. In fasting-induced torpor in mice, body-temperature fluctuations increased before torpor onset, providing an early-warning signal that tracked proximity to the transition better than temperature decline alone. A heat-balance model showed that reducing how strongly the effective heat-loss coefficient depends on body temperature reorganizes thermoregulatory stability, allowing a low-temperature equilibrium to emerge while the normothermic state remains stable. This organization is consistent with a symmetry-broken pitchfork involving a saddle-node. Similar increases in temperature fluctuations preceded hibernation onset in hamsters. These findings link pre-transition temperature dynamics to changes in the underlying thermoregulatory landscape and provide a framework for detecting and understanding transitions from normothermia to hypothermia.

The basal ganglia (BG), thalamus and cortex form the BG-thalamo-cortical (BGTC) network, which is essential for voluntary movement and centrally involved in Parkinson's disease (PD). In both patients and animal models, BG neuronal activity exhibits exaggerated oscillatory synchronization in the beta frequencies (13-30 Hz). Theoretical studies have proposed multiple network mechanisms for the generation of these abnormal beta oscillations. However, key properties, such as frequency and power distribution across BG nuclei, vary substantially among patients and between animal models (rodents vs. non-human primates). This variability complicates direct comparisons between theoretical models and experimental data, and questions whether distinct neuronal mechanisms may underlie beta oscillations across species. In an experimentally constrained BGTC network model with species-specific parameters (synaptic, neuronal, and network properties), we evaluate the features of abnormal beta oscillatory activity generated by different mechanisms. The network's negative feedback loops serve as potential sources of spontaneous oscillations. Using rodent- or primate-constrained models, we compare the spectral properties of oscillatory activity across network populations for each loop and derive the expected phase relationships between populations to align predictions with existing rodent data and propose testable hypotheses for primates. We also demonstrate how oscillation frequency can be modulated when multiple generation mechanisms interact as coupled oscillators in the full network. Our results, combined with observed cross-species beta oscillations characteristics, suggest that abnormal beta oscillations likely arise from distinct mechanisms in rodents and primates.

Facial and vocal behaviors jointly convey affective and communicative information, yet their cross-modal organization remains difficult to characterize quantitatively. We propose a framework for characterizing facial-vocal organization by aligning facial and vocal dynamics using canonical time warping. Multivariate facial action unit trajectories and vocal prosodic trajectories were extracted from recordings of spontaneous speech and aligned in a shared latent space. From the alignment process, we derived three cross-modal coordination metrics: "scaling factor", reflecting the relative scaling between modalities; "time shift ratio", reflecting the frequency of temporal shifts; and "time shift magnitude", reflecting their size. We evaluated these metrics in verbally fluent adults with autism spectrum disorder (ASD; N = 35) and controls (N = 44). The proposed metrics did not differ significantly between diagnostic groups but showed associations with individual differences in communication and anxiety. Scaling factor was linked to trait anxiety in ASD, whereas time shift ratio was associated with communication characteristics across participants. The proposed metrics also showed little association with conventional vocal pitch variability, suggesting that they capture aspects of expressive organization not reflected by vocal variability alone. Our findings support alignment-derived metrics as a complementary approach for characterizing individual variability in facial-vocal expressive behavior.

Chronic heavy alcohol use shifts brain metabolism from glucose toward acetate utilization. Sudden decreases in acetate following alcohol cessation in individuals with alcohol use disorder (AUD) may create an energetic imbalance that contributes to craving and withdrawal. A recent randomized study found that, compared with a standard American (SA) diet, a ketogenic diet (KD) during alcohol detoxification reduced withdrawal severity and craving and increased ketones and glutamate in the dorsal anterior cingulate cortex. These findings suggest that ketones may support metabolic adaptation during alcohol withdrawal, but the effects of altered brain metabolism on neural dynamics and alcohol cue reactivity remain unknown. Here, we used a control theoretic approach to test whether shifts in cerebral energy substrate utilization influence large-scale network dynamics through altered information flow. We analyzed fMRI cue-reactivity scans collected after one, two, and three weeks of inpatient treatment during presentation of alcohol, food, and neutral cues. Whole-brain effective connectivity was estimated using Shannon transfer entropy applied to whole brain atlas-aligned fMRI time series. Assuming neural activity flow along effective connections, we identified distinct shifts in the state-transition landscape in participants receiving the KD compared with the SA diet. Increases in serum {beta}-hydroxybutyrate (BHB) were associated with changes in state-transition landscape organization, which were in turn associated with reduced cue-induced alcohol craving. These findings link metabolic adaptation to large-scale neural dynamics and clinically relevant behavior and suggest that cerebral energy substrate availability may shape behavior by altering the organization of large-scale neural dynamics.

BackgroundAlcohol use disorder (AUD) is an established contributor to neurocognitive impairment and a major modifiable risk factor for dementia. However, the extent to which lifetime AUD interacts with aging to accelerate subtle cognitive trajectories prior to overt clinical dementia remains unclear. ObjectiveTo investigate whether a lifetime history of DSM-5 AUD and correlated indices of alcohol consumption are associated with accelerated age-related cognitive decline in middle-aged and older adults. MethodsParticipants (N = 1,421) were drawn from the Collaborative Study on the Genetics of Alcoholism (COGA) prospective longitudinal study. Lifetime AUD status (defined as 2 or more DSM-5 criteria endorsed at peak drinking) and drinking patterns were derived from the Semi-Structured Assessment for the Genetics of Alcoholism (SSAGA). Cognitive performance across multiple domains was evaluated using the NIH Toolbox Cognition Battery to yield Total, Fluid, and Crystallized Composite Scores. Linear mixed-effects models examined main effects of AUD and age, as well as an AUD-by-age interaction, controlling for testing site. ResultsAt the time of cognitive assessment, the overall sample had a mean age of 52.5 years (SD = 14.5) with a mean longitudinal observation period of 25.0 years (SD = 4.7). The sample was 60% female, 40% male, 7% Hispanic, 23% non-Hispanic Black, 68% non-Hispanic White, 2% other race/ethnicity. 64% met lifetime criteria for AUD (n = 908). While baseline mean cognitive composite scores were similar between those with and without AUD, lifetime AUD was significantly associated with steeper age-related decline for both the Total Composite score (interaction estimate = -0.092, 95% CI [-0.173, -0.010], p = 0.03) and Fluid Composite score (interaction estimate = -0.13, 95% CI [-0.220, -0.040], p = 0.005). No significant AUD-by-age interaction was observed for Crystallized Composite scores (interaction estimate = -0.026, 95% CI [-0.092, 0.041], p = 0.449). Based on model projections, the AUD-associated cognitive deficit on Total Composite scores was equivalent to 7.1 additional years of cognitive aging at age 60 (95% CI [-0.5, 14.7]) and 13.8 additional years at age 75 (95% CI [0.2, 27.3]). Measures of current (past-year) drinking status and current drinks per week showed no significant interaction with age. ConclusionsA lifetime history of AUD acts as a catalyst for cognitive aging, significantly steepening the trajectory of age-related decline in total and fluid cognitive functioning well before the onset of clinical dementia. Cross-sectional measures of current drinking failed to capture these effects, demonstrating that evaluating cumulative lifetime drinking history is important for assessing alcohol-related cognitive decline.

Bilateral cochlear implantation (BCI) can improve spatial hearing in adults with bilateral hearing loss, but the early development of spatial hearing following sequential BCI remains poorly understood. This prospective longitudinal study investigated changes in spatial hearing experience in 23 adults undergoing sequential BCI, assessed before and at ~1, 3, and 8 months after bilateral implantation under two conditions: spontaneous and restrained head movements. Using a motion-tracking system combined with virtual reality, sound localization was assessed through absolute and signed errors, along with self-reported localization confidence and listening effort. Subjective spatial hearing was also assessed using the Speech, Spatial and Qualities of Hearing Scale (SSQ). Localization performance improved by 1 month after bilateral implantation, as evidenced by a reduction of absolute error and in the bias toward the first-implanted ear. Participants reported greater localization confidence and lower listening effort at 1 month, although these changes were not consistently maintained at later follow-ups. SSQ scores showed improved spatial hearing at 1 month, speech perception at 3 months, and sound clarity and discriminability at 8 months. Compared with head-restrained condition, accuracy and confidence improved when spontaneous head rotations were allowed. The temporal trajectories of localization bias, confidence, and effort were modulated by the duration of experience with the first CI. These findings indicate that changes in spatial hearing can emerge soon after second CI activation. They further highlight the contribution of spontaneous head movements and previous CI experience, and support the use of complementary measures to monitor the early development of bilateral listening.

Animal pose estimation typically requires manual annotation of body parts to train neural networks for each new recording system or environmental condition. Here we use a general-purpose vision-language model (VLM) to generate annotations in multi-animal recordings of fruit flies (Drosophila melanogaster). Initial annotations follow written anatomical descriptions and are reviewed by an independent VLM session using neighboring video frames. On 120 held-out fly images, AI annotations had a median distance of 2.28 pixels from human labels, with 84.0% agreement within eight pixels, averaged across landmark types. The comparison included 10 landmarks, with head and eyes analyzed separately because of systematic differences in placement. Review improved agreement primarily at the feet, with the largest remaining discrepancies at positions with limited visual evidence. Visual inspection revealed cases in which the AI annotations were more anatomically accurate than those of an experienced human annotator, particularly at the feet. A convolutional network trained using SLEAP on 300 AI-annotated fly images showed similar agreement with human labels and closely reproduced the AI annotations on held-out images, with a median distance of 1.56 pixels. We also applied the procedure to recordings from another laboratory, using new AI annotations to train a separate pose network. Although VLM annotation required minutes per fly, the trained networks enabled pose estimation throughout more than 700,000 frames. These results demonstrate that VLM annotations can support fast animal pose estimation without manually labeled training images.

Approximately 30% of U.S. adolescents have consumed alcohol, and earlier onset is associated with faster escalation and greater risk of adult alcohol use disorder. Predicting not only who will initiate drinking but also when could enable more timely and targeted prevention. Existing machine-learning approaches have largely classified drinking status or predefined trajectory classes, requiring fully observed outcomes and often excluding adolescents who remain abstinent at last follow-up. Many also rely on neuroimaging or small, hand-selected feature sets, limiting scalability. We developed a stacked-encoder survival framework for predicting drinking-onset timing from baseline clinical data alone. The framework compresses 1,582 clinical features using an unsupervised denoising autoencoder and re-embeds them with FocalTab, a TabPFN encoder fine-tuned with focal loss; the encoders are then frozen for survival modeling using a neural Cox or random survival forest head. Among 661 NCANDA adolescents who were non-drinkers at baseline, 495 initiated drinking within six years. The stacked representation outperformed single-encoder and raw-feature models, achieving an Uno C-index of 0.7654 and time-dependent AUC of 0.8274. Focal-loss encoding outperformed cross-entropy encoding, while SHAP identified alcohol expectancies, substance access, and community prevention engagement among the most influential predictors. Unsupervised clustering further identified five biotypes with a monotonic gradient in mean time to onset (4.30 to 2.86 years) and differential limbic and thalamic neuroimaging phenotypes. These findings support the feasibility of scalable, clinically based prediction of drinking-onset timing and data-driven adolescent risk stratification.

Cell type designations label uniquely-identifiable groups of related cells that typically function in similar ways. Cell types divide the Drosophila melanogaster central nervous system into functional units, but remain challenging to target and identify. Previously we generated a driver line and image resource for targeting many neuron types (Meissner et al., 2025). Here we improve on identifying the targeted neurons by automated searching against electron microscopy datasets and manual scoring of search results to add 6,368 annotations across 1,772 driver lines in the collection. These annotations provide a potential training resource for future search algorithms. We then estimate the number of cell types and thus the fraction of the central nervous system labeled by the collection as a whole.

The pure-tone audiogram remains the cornerstone of hearing assessment, yet it captures only part of the functional consequences of cochlear damage. Most phenotyping schemes rely solely on the audiogram, which cannot detect hidden suprathreshold dysfunction when thresholds are normal. How best to represent audiological data - as discrete phenotypes or as a continuous space - remains unresolved. Using a retrospective database from the Copenhagen Hearing and Balance Centre, Rigshospitalet (Denmark; 1995-2022; 251986 ear-session units from 81883 adults with normal or sensorineural audiograms), we compared linear (PCA) and non-linear (UMAP) embeddings of the audiogram alone and with word recognition in quiet (WRSmax-Q). Here we show that a linear embedding suffices for the audiogram, but a non-linear one becomes advantageous once suprathreshold measures are added. PCA compressed the audiogram onto two interpretable axes, overall level and slope, explaining 91.5% of the variance, with ears forming a continuous cloud that published reference schemes tiled into overlapping regions. For the audiogram alone, UMAP recovered the same organisation. WRSmax-Q varied widely among ears with similar audiograms; adding it left the PCA unchanged but reorganised the UMAP. A synthetic experiment illustrated this: ears with identical audiograms but different WRSmax-Q separated under UMAP, not PCA. Combining thresholds with WRSmax-Q, UMAP yielded a continuous representation - a prototype Hearing Loss Map - where each ear's position reflects threshold and suprathreshold measures. Representing hearing loss as a location in a continuous space rather than as discrete categories may better reflect its graded nature, with clinical relevance for individual patients.

Parkinson's disease (PD) gait is characterized by reduced walking speed, short shuffling steps, increased double stance time, and increased step time variability. The most debilitating symptom in PD is freezing of gait (FOG), a context-dependent locomotion arrest, mostly occurring while encountering turns, obstacles, and narrow pathways. FOG is more pronounced during turning, a complex task that challenges motor planning and execution. Here, we simulate this phenomenon using a computational model inspired by the cortico-basal ganglia circuitry, based on reinforcement learning (RL) principles. In the model, the agent was trained to learn a value profile while navigating a path that involved turning. To simulate PD conditions, we clamped the temporal difference (TD) error, typically associated with dopaminergic signaling. Our model results show that compared to healthy controls (HC), PD non-freezers and PD freezers exhibit a reduction in turning velocity and step length, alongside an attenuated compensatory step-width widening at the sharpest turn, more pronounced in PD freezers, who additionally show increased step time variability. These kinematic impairments worsen during tighter turns. Ultimately, this model implicates FOG as a problem of impaired estimation of value over space and could potentially represent a generalized deficit in gait impairments in PD.

Changes in pupil size have been widely used as an indirect measure of attention. However, attention engages on a faster timescale than the pupillary response. Therefore, the slow time course of the pupil response makes it difficult to distinguish attentional responses to stimuli presented in close temporal succession. Although signal deconvolution has been used in previous research to estimate latent attentional responses from the pupil signal, existing methods primarily model dilation without considering the visual stimulus evoked constriction phase of the pupil. Here, we introduce a dual-kernel deconvolution model that models constriction and dilation using separate impulse response functions. Using an RSVP paradigm with emotional faces as target stimuli, we demonstrate that the proposed model estimates latent attentional response despite substantial overlap in the measured pupil signal. Importantly, the recovered attentional response resembles the time course of attentional allocation previously predicted by several models of Attentional Blink. Together, these findings demonstrate that the dynamics of latent attentional episodes can be recovered from the pupil response with a temporal resolution previously obscured by the sluggishness of the pupillary response. The proposed method is also provided as a usable software package that serves as a general tool for investigating the temporal dynamics of attention in rapid visual paradigms.

Background: High-definition transcranial alternating current stimulation (HD-tACS) is favored by the neurostimulation community for its precision and ability to influence neuronal dynamics. Yet, the exact mechanism by which the underlying brain structures are being affected remains unclear. We believe that the investigation of the aperiodic nature of the electroencephalograph (EEG) could shed light on the modulatory effects of HD-tACS. Methods: We analyzed the EEG of 9 participants during a compensatory tracking task (CTT) in two sessions, each with different HD-tACS protocols. Every session consisted of an initial period of no stimulation, followed by 30 Hz HD-tACS in the left motor (M30) or frontal (F30) cortex. We then isolated the aperiodic component of the EEG and calculated its spectral slope {beta}. Results and Discussion: {beta} decreased during F30 mainly in the right frontal cortex, indicating a shift towards higher frequencies and an increase of the excitatory/inhibitory balance. Additionally, we found that despite the long monotonus task the accuracy of the participants did not decrease, which might be attributed to the ability of both M30 and F30 to sustain attention for prolonged time. Finally, the change of CTT accuracy during the stimulation correlated with the {beta} of specific channels before the stimulation. This indicates the potential of {beta} to be used as a screening biomarker in future studies. In conclusion, we showed the ability of HD-tACS to alter EEG aperiodic dynamics and paved the way for future exploration of such dynamics in the field.

Chronic social isolation induces persistent behavioral and physiological changes across species. In humans, prolonged social isolation is associated with insomnia and increased risks of cognitive decline, cardiovascular disease, compromised immunity, and premature mortality. Yet whether these consequences are modified by the environment in which social isolation is experienced remains unclear. In particular, can access to social sensory information from conspecifics alleviate the effects of isolation and can various ambient environmental settings modify them? A related question arises during behavioral measurement because behavioral assays themselves impose a new environmental context. Whether these assays simply capture the enduring effects of prior social experience or further modify those effects remains unknown. Using social isolation-induced sleep loss in Drosophila as a model, we systematically examined how the social, spatial, and photoperiodic environments during isolation and behavioral monitoring influence the isolation phenotype. Providing isolated flies with visual, olfactory, or tactile social cues from conspecifics living behind a divider, either individually or in combination, was insufficient to alleviate sleep loss. Spatial confinement exacerbated isolation-induced sleep loss, while winter-like photoperiod in spatially confined Drosophila Activity Monitor assays revealed the clearest progressive decline in sleep during sleep monitoring. Together, these findings demonstrate that social isolation-induced sleep loss is both enduring and dynamic: it persists despite exposure to social cues yet remains sensitive to the environmental context in which isolation is experienced and behavior is measured. More broadly, our study establishes a tractable framework for investigating the environmental modulation of chronic social isolation.

Over the course of evolution, many species have adapted to the environmental conditions found at higher latitudes. This is likely also true of the circadian system, which regulates daily behavior and helps measure day length to facilitate adaptation to the seasons, though how the circadian clock adapts to latitude remains largely unknown. Here, we investigated daily activity patterns under long photoperiods and in constant darkness, as well as the underlying circadian clock network, in 52 Drosophila species from around the world. We found a strong, significant positive correlation between the clock's plasticity, measured as the phase difference between morning and evening activity, and the average geographical latitude from which the species originate. The mechanisms underpinning this increase in plasticity differed between species: in the subgenus Siphlodora, the clock weakened with increasing latitude, accompanied by a reduction of the neuropeptide Pigment-Dispersing Factor in the small ventrolateral clock neurons (s-LNv), whereas in the subgenus Drosophila the opposite was true. The subgenus Sophophora, by contrast, showed a negative correlation between circadian plasticity and PDF levels in the s-LNv axons. These results suggest that Drosophila species have evolved diverse, independent strategies to adapt their circadian clock to higher latitudes.

Amacrine cells (ACs) are the most diverse neuron in the retina, with more than sixty different subtypes, each contributing to the detection of specific visual features. Despite their importance in visual feature detection, the genetic basis for this diversity is still not well understood. Recent single-cell transcriptomic datasets have identified several potential genes that may regulate cellular diversity, but the underlying mechanisms have not been investigated. For example, the transcription factor Meis2 is selectively expressed in the GABAergic subclass of ACs. In this study, we focused on how Meis2 regulates the development of GABAergic ACs. We hypothesized that Meis2 is required for specification and survival of GABAergic ACs. To test this, we used the Meis2 conditional knockout mouse (Meis2-CKO) where Meis2 was deleted from either the entire retina or specifically from AC precursor cells. The deletion of Meis2 led to a ~60% reduction in total GABAergic ACs in the retina. This loss of GABAergic ACs occurred only after postnatal day 5 (P5), suggesting Meis2 is not required for early specification but may be required for survival. To further investigate cell death and survival, we quantified cleaved caspase-3 positive cells in the retina at several developmental timepoints. In Meis2-CKO mice, we found an increase in cleaved caspase-3 immunolabeling at P5 and P6 timepoints coinciding with the reduction of GABAergic ACs. Further analysis of GABAergic AC subtypes demonstrated a subtype-specific cell death in Meis2-CKO retinas. For example, the total number of dopaminergic (TH) ACs were significantly reduced, whereas starburst (ChAT) ACs remained unchanged. These findings suggest that Meis2 is important for the survival of a subset of GABAergic AC subtypes in the mammalian retina. In addition, Meis2 may control the expression of specific survival factors that prevent apoptosis during developmental cell death in the retina.