Sensory encoding and memory in the mushroom body: signals, noise, and variability.

To survive in changing environments, animals need to learn to associate specific sensory stimuli with positive or negative valence. How do they form stimulus-specific memories to distinguish between positively/negatively associated stimuli and other irrelevant stimuli? Solving this task is one of the functions of the mushroom body, the associative memory center in insect brains. Here we summarize recent work on sensory encoding and memory in the Drosophila mushroom body, highlighting general principles such as pattern separation, sparse coding, noise and variability, coincidence detection, and spatially localized neuromodulation, and placing the mushroom body in comparative perspective with mammalian memory systems.

Compensatory variability in network parameters enhances memory performance in the Drosophila mushroom body.

Neural circuits use homeostatic compensation to achieve consistent behavior despite variability in underlying intrinsic and network parameters. However, it remains unclear how compensation regulates variability across a population of the same type of neurons within an individual and what computational benefits might result from such compensation. We address these questions in the Drosophila mushroom body, the fly's olfactory memory center. In a computational model, we show that under sparse coding conditions, memory performance is degraded when the mushroom body's principal neurons, Kenyon cells (KCs), vary realistically in key parameters governing their excitability. However, memory performance is rescued while maintaining realistic variability if parameters compensate for each other to equalize KC average activity. Such compensation can be achieved through both activity-dependent and activity-independent mechanisms. Finally, we show that correlations predicted by our model's compensatory mechanisms appear in the Drosophila hemibrain connectome. These findings reveal compensatory variability in the mushroom body and describe its computational benefits for associative memory.

Mechanisms underlying homeostatic plasticity in the Drosophila mushroom body in vivo.

Neural network function requires an appropriate balance of excitation and inhibition to be maintained by homeostatic plasticity. However, little is known about homeostatic mechanisms in the intact central brain in vivo. Here, we study homeostatic plasticity in the Drosophila mushroom body, where Kenyon cells receive feedforward excitation from olfactory projection neurons and feedback inhibition from the anterior paired lateral neuron (APL). We show that prolonged (4-d) artificial activation of the inhibitory APL causes increased Kenyon cell odor responses after the artificial inhibition is removed, suggesting that the mushroom body compensates for excess inhibition. In contrast, there is little compensation for lack of inhibition (blockade of APL). The compensation occurs through a combination of increased excitation of Kenyon cells and decreased activation of APL, with differing relative contributions for different Kenyon cell subtypes. Our findings establish the fly mushroom body as a model for homeostatic plasticity in vivo.

Inguinal Lymph Node Dissection Does Not Improve Overall Survival in Anal Cancer Nodal Disease.

BACKGROUND: Anal SCC is a rare disease mainly treated with chemoradiation. Abdominoperineal resection (APR), once the mainstay of treatment for anal cancer, now serves a role as salvage therapy for persistent or recurrent disease after chemoradiation. In addition, clinically positive nodes are currently treated by extending the radiation field to the groin. The role of inguinal lymph node dissection in recurrent or persistent anal SCC is unclear. The aim of the study is to determine the role of inguinal lymph node dissection in the management of inguinal lymph node metastasis for anal squamous cell carcinoma (SCC). METHODS: Retrospective analysis of patients with anal SCC in the National Cancer Database with positive inguinal nodes undergoing salvage APR between 2004 and 2014 was performed. A comparison of overall survival between patients who underwent APR with lymph node dissection versus APR only was analyzed using Kaplan-Meier plot. RESULTS: A total of 3424 patients underwent salvage APR, with 274 (8%) having clinically positive nodes. Within the subgroup that had clinically positive nodes, 195 (71%) underwent APR, whereas 79 (29%) underwent both APR and node dissection. Kaplan-Meier analysis demonstrates no difference in overall survival in the two groups (P = 0.99). Five-year survival for both groups was similar (36% versus 42%; P = 0.987). No significant difference was found when adjusted for age, gender, and Tumor Node Metastasis staging. CONCLUSIONS: Inguinal lymph node dissection does not appear to improve overall survival in patients with advanced-stage anal cancer requiring salvage APR. Proper patient selection for node dissection is essential to spare patients from additional morbid procedures.

Ultra-wide field and new wide field composite retinal image registration with AI-enabled pipeline and 3D distortion correction algorithm.

PURPOSE: This study aimed to compare a new Artificial Intelligence (AI) method to conventional mathematical warping in accurately overlaying peripheral retinal vessels from two different imaging devices: confocal scanning laser ophthalmoscope (cSLO) wide-field images and SLO ultra-wide field images. METHODS: Images were captured using the Heidelberg Spectralis 55-degree field-of-view and Optos ultra-wide field. The conventional mathematical warping was performed using Random Sample Consensus-Sample and Consensus sets (RANSAC-SC). This was compared to an AI alignment algorithm based on a one-way forward registration procedure consisting of full Convolutional Neural Networks (CNNs) with Outlier Rejection (OR CNN), as well as an iterative 3D camera pose optimization process (OR CNN + Distortion Correction [DC]). Images were provided in a checkerboard pattern, and peripheral vessels were graded in four quadrants based on alignment to the adjacent box. RESULTS: A total of 660 boxes were analysed from 55 eyes. Dice scores were compared between the three methods (RANSAC-SC/OR CNN/OR CNN + DC): 0.3341/0.4665/4784 for fold 1-2 and 0.3315/0.4494/4596 for fold 2-1 in composite images. The images composed using the OR CNN + DC have a median rating of 4 (out of 5) versus 2 using RANSAC-SC. The odds of getting a higher grading level are 4.8 times higher using our OR CNN + DC than RANSAC-SC (p < 0.0001). CONCLUSION: Peripheral retinal vessel alignment performed better using our AI algorithm than RANSAC-SC. This may help improve co-localizing retinal anatomy and pathology with our algorithm.

Neuroscience: Hacking development to understand sensory discrimination.

Fine sensory discrimination abilities are enabled by specific neural circuit architectures. A new study reveals how manipulating particular network parameters in the fly's memory centre, the mushroom body, alters sensory coding and discrimination.

SpaRCe: Improved Learning of Reservoir Computing Systems Through Sparse Representations.

"Sparse" neural networks, in which relatively few neurons or connections are active, are common in both machine learning and neuroscience. While, in machine learning, "sparsity" is related to a penalty term that leads to some connecting weights becoming small or zero, in biological brains, sparsity is often created when high spiking thresholds prevent neuronal activity. Here, we introduce sparsity into a reservoir computing network via neuron-specific learnable thresholds of activity, allowing neurons with low thresholds to contribute to decision-making but suppressing information from neurons with high thresholds. This approach, which we term "SpaRCe," optimizes the sparsity level of the reservoir without affecting the reservoir dynamics. The read-out weights and the thresholds are learned by an online gradient rule that minimizes an error function on the outputs of the network. Threshold learning occurs by the balance of two opposing forces: reducing interneuronal correlations in the reservoir by deactivating redundant neurons, while increasing the activity of neurons participating in correct decisions. We test SpaRCe on classification problems and find that threshold learning improves performance compared to standard reservoir computing. SpaRCe alleviates the problem of catastrophic forgetting, a problem most evident in standard echo state networks (ESNs) and recurrent neural networks in general, due to increasing the number of task-specialized neurons that are included in the network decisions.

Pigmentary retinopathy associated with immune therapy for advanced cutaneous melanoma.

Purpose: To describe a case of bilateral retinal pigmentary changes in the setting of immune checkpoint inhibitor therapy (ICIT). Observations: A 69-year-old man with a history of advanced cutaneous melanoma was started on combination ICIT with nivolumab and ipilimumab and stereotactic body radiation therapy. Soon after, he developed photopsias and nyctalopia with findings of discrete retinal pigmentary changes bilaterally. Initial visual acuities were 20/20 and 20/30 in the right and left eye, respectively. Multi-modal imaging revealed sub-retinal deposits with progressive changes in pigmentation and autofluorescence, associated with decreased peripheral fields on formal perimetry. A full-field electroretinogram revealed attenuated and delayed a- and b-waves. Positive serum retinal autoantibodies were identified. The patient developed left-sided optic nerve edema and center-involving cystoid macular edema which improved after treatment with sub-tenon's triamcinolone. Conclusions: The use of ICIT has greatly expanded in oncologic practice with subsequent increases in immune related adverse events that pose significant systemic and ophthalmologic morbidities. We propose that the new retinal pigmentary changes seen in this case are the sequelae of an autoimmune inflammatory response against pigmented cells. This adds to the rare side effects that may occur after ICIT.

The Sensitivity of Ultra-Widefield Fundus Photography Versus Scleral Depressed Examination for Detection of Retinal Horseshoe Tears.

PURPOSE: Ultra-widefield (UWF) imaging is commonly used in ophthalmology in tandem with scleral depressed examinations (SDE) to evaluate peripheral retinal disease. Because of the increased reliance on this technology in tele-ophthalmology, it is critical to evaluate its efficacy for detecting the peripheral retina when performed in isolation. Therefore, we sought to evaluate UWF imaging sensitivity in detecting retinal horseshoe tears (HSTs). STUDY DESIGN: Retrospective clinical validity and reliability study. METHODS: A single-institutional retrospective analysis was performed on patients at the Shiley Eye Institute, University of California, San Diego. Patients with HSTs seen on SDE who underwent treatment with laser were included in the study. A total of 140 patients with HSTs in the right and/or left eyes met the inclusion criteria. Those with concomitant ruptured globes, retinal detachments, and vitreous hemorrhages were excluded. A total of 123 patients with 135 HSTs were included in the final analysis. The primary outcome was the number of HSTs detected by UWF imaging. A secondary outcome was HST location. Sensitivity was measured with respect to HST location, and statistical significance was calculated by Fisher exact testing. RESULTS: A total of 69 (51.1%) HSTs were visualized on UWF images and 66 (48.9%) were not visualized. The sensitivity of UWF imaging in capturing HSTs was 7 of 41 (17.1%), 8 of 25 (32.0%), 7 of 14 (50.0%), and 47 of 55 (85.5%) for the superior, inferior, nasal, and temporal quadrants, respectively. Sensitivities between HST visibility and location were statistically significant (P < .001). CONCLUSIONS: Nearly half of HSTs were missed by UWF imaging. This study demonstrates that UWF imaging alone is not sufficiently sensitive to exclude the presence of HSTs.

Assessing the Clinical Utility of Expanded Macular OCTs Using Machine Learning.

Purpose: Optical coherence tomography (OCT) is widely used in the management of retinal pathologies, including age-related macular degeneration (AMD), diabetic macular edema (DME), and primary open-angle glaucoma (POAG). We used machine learning techniques to understand diagnostic performance gains from expanding macular OCT B-scans compared with foveal-only OCT B-scans for these conditions. Methods: Electronic medical records were extracted to obtain 61 B-scans per eye from patients with AMD, diabetic retinopathy, or POAG. We constructed deep neural networks and random forest ensembles and generated area under the receiver operating characteristic (AUROC) and area under the precision recall (AUPR) curves. Results: After extracting 630,000 OCT images, we achieved improved AUROC and AUPR curves when comparing the central image (one B-scan) to all images (61 B-scans). The AUROC and AUPR points of diminishing return for diagnostic accuracy for macular OCT coverage were found to be within 2.75 to 4.00 mm (14-19 B-scans), 4.25 to 4.50 mm (20-21 B-scans), and 4.50 to 6.25 mm (21-28 B-scans) for AMD, DME, and POAG, respectively. All models with >0.25 mm of coverage had statistically significantly improved AUROC/AUPR curves for all diseases (P < 0.05). Conclusions: Systematically expanded macular coverage models demonstrated significant differences in total macular coverage required for improved diagnostic accuracy, with the largest macular area being relevant in POAG followed by DME and then AMD. These findings support our hypothesis that the extent of macular coverage by OCT imaging in the clinical setting, for any of the three major disorders, has a measurable impact on the functionality of artificial intelligence decision support. Translational Relevance: We used machine learning techniques to improve OCT imaging standards for common retinal disease diagnoses.

Detection of features associated with neovascular age-related macular degeneration in ethnically distinct data sets by an optical coherence tomography: trained deep learning algorithm.

BACKGROUND: The ability of deep learning (DL) algorithms to identify eyes with neovascular age-related macular degeneration (nAMD) from optical coherence tomography (OCT) scans has been previously established. We herewith evaluate the ability of a DL model, showing excellent performance on a Korean data set, to generalse onto an American data set despite ethnic differences. In addition, expert graders were surveyed to verify if the DL model was appropriately identifying lesions indicative of nAMD on the OCT scans. METHODS: Model development data set-12 247 OCT scans from South Korea; external validation data set-91 509 OCT scans from Washington, USA. In both data sets, normal eyes or eyes with nAMD were included. After internal testing, the algorithm was sent to the University of Washington, USA, for external validation. Area under the receiver operating characteristic curve (AUC) and precision-recall curve (AUPRC) were calculated. For model explanation, saliency maps were generated using Guided GradCAM. RESULTS: On external validation, AUC and AUPRC remained high at 0.952 (95% CI 0.942 to 0.962) and 0.891 (95% CI 0.875 to 0.908) at the individual level. Saliency maps showed that in normal OCT scans, the fovea was the main area of interest; in nAMD OCT scans, the appropriate pathological features were areas of model interest. Survey of 10 retina specialists confirmed this. CONCLUSION: Our DL algorithm exhibited high performance for nAMD identification in a Korean population, and generalised well to an ethnically distinct, American population. The model correctly focused on the differences within the macular area to extract features associated with nAMD.

Function and regulation of local axonal translation.

An increasing body of evidence indicates that local axonal translation is required for growing axons to respond appropriately to guidance cues and other stimuli. Recent studies suggest that asymmetrical synthesis of cytoskeletal proteins mediates growth cone turning and that local translation and retrograde transport of transcription factors mediate neuronal survival. Axonal translation is regulated partly by selective axonal localization of mRNAs and by translation initiation factors and RNA-binding proteins. We discuss possible rationales for local axonal translation, including distinct properties of nascent proteins, precise localization, and axonal autonomy.

A functional equivalent of endoplasmic reticulum and Golgi in axons for secretion of locally synthesized proteins.

Subcellular localization of protein synthesis provides a means to regulate the protein composition in far reaches of a cell. This localized protein synthesis gives neuronal processes autonomy to rapidly respond to extracellular stimuli. Locally synthesized axonal proteins enable neurons to respond to guidance cues and can help to initiate regeneration after injury. Most studies of axonal mRNA translation have concentrated on cytoplasmic proteins. While ultrastructural studies suggest that axons do not have rough endoplasmic reticulum or Golgi apparatus, mRNAs for transmembrane and secreted proteins localize to axons. Here, we show that growing axons with protein synthetic activity contain ER and Golgi components needed for classical protein synthesis and secretion. Isolated axons have the capacity to traffic locally synthesized proteins into secretory pathways and inhibition of Golgi function attenuates translation-dependent axonal growth responses. Finally, the capacity for secreting locally synthesized proteins in axons appears to be increased by injury.

Asymmetrical beta-actin mRNA translation in growth cones mediates attractive turning to netrin-1.

Local protein synthesis regulates the turning of growth cones to guidance cues, yet little is known about which proteins are synthesized or how they contribute to directional steering. Here we show that beta-actin mRNA resides in Xenopus laevis retinal growth cones where it binds to the RNA-binding protein Vg1RBP. Netrin-1 induces the movement of Vg1RBP granules into filopodia, suggesting that it may direct the localization and translation of mRNAs in growth cones. Indeed, a gradient of netrin-1 activates a translation initiation regulator, eIF-4E-binding protein 1 (4EBP), asymmetrically and triggers a polarized increase in beta-actin translation on the near side of the growth cone before growth cone turning. Inhibition of beta-actin translation abolishes both the asymmetric rise in beta-actin and attractive, but not repulsive, turning. Our data suggest that newly synthesized beta-actin, concentrated near sites of signal reception, provides the directional bias for polymerizing actin in the direction of an attractive stimulus.

How nitric oxide helps update memories.

Some dopaminergic neurons release both dopamine and nitric oxide to increase the flexibility of olfactory memories.

Localized inhibition in the Drosophila mushroom body.

Many neurons show compartmentalized activity, in which activity does not spread readily across the cell, allowing input and output to occur locally. However, the functional implications of compartmentalized activity for the wider neural circuit are often unclear. We addressed this problem in the Drosophila mushroom body, whose principal neurons, Kenyon cells, receive feedback inhibition from a non-spiking interneuron called the anterior paired lateral (APL) neuron. We used local stimulation and volumetric calcium imaging to show that APL inhibits Kenyon cells' dendrites and axons, and that both activity in APL and APL's inhibitory effect on Kenyon cells are spatially localized (the latter somewhat less so), allowing APL to differentially inhibit different mushroom body compartments. Applying these results to the Drosophila hemibrain connectome predicts that individual Kenyon cells inhibit themselves via APL more strongly than they inhibit other individual Kenyon cells. These findings reveal how cellular physiology and detailed network anatomy can combine to influence circuit function.

Multiple network properties overcome random connectivity to enable stereotypic sensory responses.

Connections between neuronal populations may be genetically hardwired or random. In the insect olfactory system, projection neurons of the antennal lobe connect randomly to Kenyon cells of the mushroom body. Consequently, while the odor responses of the projection neurons are stereotyped across individuals, the responses of the Kenyon cells are variable. Surprisingly, downstream of Kenyon cells, mushroom body output neurons show stereotypy in their responses. We found that the stereotypy is enabled by the convergence of inputs from many Kenyon cells onto an output neuron, and does not require learning. The stereotypy emerges in the total response of the Kenyon cell population using multiple odor-specific features of the projection neuron responses, benefits from the nonlinearity in the transfer function, depends on the convergence:randomness ratio, and is constrained by sparseness. Together, our results reveal the fundamental mechanisms and constraints with which convergence enables stereotypy in sensory responses despite random connectivity.

Neuronal mechanisms underlying innate and learned olfactory processing in Drosophila.

Olfaction allows animals to adapt their behavior in response to different chemical cues in their environment. How does the brain efficiently discriminate different odors to drive appropriate behavior, and how does it flexibly assign value to odors to adjust behavior according to experience? This review traces neuronal mechanisms underlying these processes in adult Drosophila melanogaster from olfactory receptors to higher brain centers. We highlight neural circuit principles such as lateral inhibition, segregation and integration of olfactory channels, temporal accumulation of sensory evidence, and compartmentalized synaptic plasticity underlying associative memory.

Inhibitory muscarinic acetylcholine receptors enhance aversive olfactory learning in adult Drosophila.

Olfactory associative learning in Drosophila is mediated by synaptic plasticity between the Kenyon cells of the mushroom body and their output neurons. Both Kenyon cells and their inputs from projection neurons are cholinergic, yet little is known about the physiological function of muscarinic acetylcholine receptors in learning in adult flies. Here, we show that aversive olfactory learning in adult flies requires type A muscarinic acetylcholine receptors (mAChR-A), particularly in the gamma subtype of Kenyon cells. mAChR-A inhibits odor responses and is localized in Kenyon cell dendrites. Moreover, mAChR-A knockdown impairs the learning-associated depression of odor responses in a mushroom body output neuron. Our results suggest that mAChR-A function in Kenyon cell dendrites is required for synaptic plasticity between Kenyon cells and their output neurons.