Neuroanatomy Journal Papers
Curated papers covering neuronal ultrastructure as observed in electron microscopy. Each paper includes summaries written for three expertise levels.
1. Harris & Weinberg (2012) — Ultrastructure of synapses in the mammalian brain
Citation: Harris KM, Weinberg RJ. Ultrastructure of synapses in the mammalian brain. Cold Spring Harbor Perspectives in Biology. 2012;4(5):a005587. DOI: 10.1101/cshperspect.a005587
Tags: neuroanatomy:synapse neuroanatomy:spine neuroanatomy:postsynaptic-density neuroanatomy:vesicle imaging:electron-microscopy
Summaries
Beginner: This review is a visual guide to what synapses — the connection points between neurons — actually look like under a powerful microscope. It explains the small structures (vesicles, densities, clefts) that make up a synapse and why their size and shape matter for brain function. If you are new to EM, treat this as a reference atlas for recognizing synapses in images.
Intermediate: Harris and Weinberg provide a systematic catalog of excitatory and inhibitory synapse ultrastructure in mammalian cortex and hippocampus. They describe presynaptic vesicle pools (readily releasable, recycling, reserve), the structure of the postsynaptic density, and the cleft components visible at EM resolution. The review also covers perforated synapses, multi-synaptic boutons, and spinule formations that are often misidentified in automated segmentation.
Advanced: This review is essential reading for anyone developing synapse detection algorithms or evaluating connectome edge weights. Key methodological points include the distinction between symmetric (Gray Type II) and asymmetric (Gray Type I) synapses, the quantitative relationship between PSD area and vesicle count, and the caveats around interpreting synapse size from single-section profiles versus serial reconstructions. The discussion of spine apparatus and smooth ER in spines is relevant for compartment-level connectivity models.
Key figures: Fig. 1 (synapse ultrastructure overview), Fig. 3 (vesicle pool organization), Fig. 5 (perforated synapse morphology)
Discussion prompts:
- Which synapse features are robust enough for automated detection versus requiring human adjudication?
- How does fixation quality affect measurements of cleft width and PSD thickness?
- What is the minimum section thickness needed to reliably distinguish Type I from Type II synapses?
Related content: Synapse classification, Dendrite biology
2. Peters, Palay & Webster (1991) — The Fine Structure of the Nervous System
Citation: Peters A, Palay SL, Webster H deF. The Fine Structure of the Nervous System: Neurons and Their Supporting Cells. 3rd ed. Oxford University Press; 1991.
Tags: neuroanatomy:soma neuroanatomy:axon neuroanatomy:dendrite neuroanatomy:myelin neuroanatomy:organelle imaging:electron-microscopy
Summaries
Beginner: This classic textbook is the definitive atlas of what the brain looks like under an electron microscope. It describes every type of cell and structure you will encounter: neuron cell bodies, axons, dendrites, synapses, glia, and myelin. Think of it as a field guide for identifying structures in brain tissue images.
Intermediate: Peters, Palay & Webster remains the gold standard reference for neuronal ultrastructure. Each chapter systematically catalogs the EM appearance of a major structure, with micrographs, drawings, and quantitative parameters. For connectomics trainees, the chapters on axon terminals, synaptic junctions, and myelin sheaths provide the visual vocabulary needed for accurate proofreading decisions.
Advanced: While dated in its functional interpretations, this text’s morphological descriptions remain unsurpassed in rigor. The systematic treatment of organelle distribution across neuronal compartments (Ch. 2-4), the classification of synaptic vesicle types and their relationship to neurotransmitter content (Ch. 5), and the detailed treatment of nodes of Ranvier and paranodal junctions (Ch. 8) are all directly relevant to current automated annotation pipelines. The illustrations remain valuable calibration references.
Key figures: Ch. 1 plates (neuronal soma organelles), Ch. 5 plates (synaptic junction types), Ch. 8 plates (myelin and nodes)
Discussion prompts:
- Which morphological criteria from this text translate directly into automated detection features?
- How has immuno-EM since 1991 refined or contradicted any of the structural assignments?
Related content: Soma ultrastructure, Axon biology, Myelin and nodes
3. Bhatt, Zhang & Bhatt (2009) — Dendritic Spine Dynamics
Citation: Bhatt DH, Zhang S, Bhatt AN. Dendritic spine dynamics. Annual Review of Physiology. 2009;71:261-282. DOI: 10.1146/annurev.physiol.010908.163140
Tags: neuroanatomy:dendrite neuroanatomy:spine neuroanatomy:synapse neuroanatomy:cytoskeleton methodology:experimental-design
Summaries
Beginner: Dendritic spines are tiny protrusions on neurons where most excitatory connections form. This review explains that spines are not static — they change shape, appear, and disappear over time. This matters for connectomics because a wiring diagram captured at one moment is a snapshot of a dynamic system.
Intermediate: This review covers the molecular and structural dynamics of dendritic spines including formation, elimination, and morphological plasticity. The authors discuss how spine shape (mushroom, thin, stubby, filopodial) correlates with synapse maturity and strength. For connectomics, the key insight is that spine morphology and density are not fixed — they change with experience, development, and pathology, meaning static EM reconstructions capture one state of a dynamic substrate.
Advanced: Bhatt et al. synthesize live-imaging and EM data to describe spine turnover rates across brain regions and developmental stages. Methodologically relevant points include the disconnect between spine presence and functional synapse presence (not all spines have active zones), the implications of fixation timing for spine morphology in EM, and the question of how spine dynamics affect the interpretability of single-timepoint connectomes. The review provides quantitative spine density and turnover data useful for estimating connectivity completeness.
Key figures: Fig. 1 (spine morphology classification), Fig. 3 (spine turnover rates across regions), Fig. 5 (activity-dependent remodeling)
Discussion prompts:
- How should connectomics studies account for spine dynamics when reporting connection strengths?
- What fraction of spines in a typical cortical EM volume might lack functional synapses?
Related content: Dendrite biology, Synapse classification
4. Shepherd & Harris (1998) — Three-Dimensional Structure and Composition of CA3→CA1 Axons
Citation: Shepherd GM, Harris KM. Three-dimensional structure and composition of CA3→CA1 axons in rat hippocampal slices: implications for presynaptic connectivity and compartmentalization. Journal of Neuroscience. 1998;18(20):8300-8310. DOI: 10.1523/JNEUROSCI.18-20-08300.1998
Tags: neuroanatomy:axon neuroanatomy:bouton neuroanatomy:vesicle neuroanatomy:mitochondria imaging:serial-section case-studies:hippocampus
Summaries
Beginner: This paper used serial electron microscopy to reconstruct the fine branches of axons in the hippocampus in 3D. The authors showed that the small swellings (boutons) along axons, where synapses form, are highly variable in size and contents — some have mitochondria, some don’t, and this affects how they work. It’s an early example of the kind of detailed wiring analysis that modern connectomics does at larger scales.
Intermediate: Shepherd and Harris performed serial section EM reconstruction of Schaffer collateral axons in rat CA1, providing quantitative data on bouton volumes, vesicle counts, mitochondria presence, and multi-synapse bouton frequencies. They found that only ~40% of boutons contain mitochondria, with important implications for sustained transmission. The 3D reconstructions revealed that inter-bouton spacing and bouton complexity vary systematically along the axon.
Advanced: This study established many quantitative priors that remain useful for validating automated reconstructions: mean bouton volume (~0.1 μm³), vesicle density (~200 per bouton), and the fraction of en passant versus terminal boutons. The finding that multi-synapse boutons synapse onto different postsynaptic neurons (not the same neuron twice) has implications for how edge weights should be computed in connectome graphs. The methodology — serial section TEM with manual alignment — provides a baseline for assessing automated pipeline accuracy on similar structures.
Key figures: Fig. 2 (3D axon reconstructions), Fig. 4 (bouton composition quantification), Fig. 6 (multi-synapse bouton examples)
Discussion prompts:
- How would you validate whether modern automated segmentation correctly identifies the bouton features measured in this study?
- What does the multi-synapse bouton finding imply for how we weight edges in a connectome graph?
Related content: Axon biology, Synapse classification
5. Fiala & Harris (2001) — Cylindrical Diameters Method for Calibrating Section Thickness
Citation: Fiala JC, Harris KM. Cylindrical diameters method for calibrating section thickness in serial electron microscopy. Journal of Microscopy. 2001;202(3):468-472. DOI: 10.1046/j.1365-2818.2001.00926.x
Tags: imaging:serial-section imaging:electron-microscopy methodology:benchmark neuroanatomy:microtubule infrastructure:alignment
Summaries
Beginner: When scientists cut brain tissue into very thin slices for electron microscopy, they need to know exactly how thick each slice is to build accurate 3D reconstructions. This paper describes a clever method for measuring slice thickness by looking at tube-shaped structures (microtubules) that run through the tissue at known diameters. Getting slice thickness right is important because errors here make reconstructions stretched or squished in 3D.
Intermediate: Fiala and Harris present a calibration method exploiting the known ~25 nm diameter of microtubules to estimate section thickness from their elliptical profiles in oblique sections. This addresses a chronic problem in serial section EM: nominal microtome settings often differ from actual section thickness by 10-30%. The method is applicable to any serial EM dataset containing microtubule-rich processes (dendrites, axons).
Advanced: This paper is important for anyone working on alignment and voxel calibration in serial section datasets. The statistical framework (fitting observed ellipse aspect ratios to expected distributions given true thickness) provides a principled alternative to fold-based or interference-based calibration. Modern connectomics pipelines should incorporate analogous calibration steps, particularly for ATUM-SEM where section thickness variation can be significant. The method’s assumptions (circular microtubule cross-section, uniform orientation) and their failure modes deserve scrutiny.
Key figures: Fig. 1 (geometry of oblique microtubule sections), Fig. 2 (calibration curve)
Discussion prompts:
- Does your automated pipeline account for section-to-section thickness variation, or does it assume constant z-spacing?
- What other structures could serve as internal calibration references?
Related content: EM principles, Reconstruction pipeline
6. Kasthuri et al. (2015) — Saturated Reconstruction of a Volume of Neocortex
Citation: Kasthuri N, Hayworth KJ, Berger DR, Schalek RL, Conchello JA, Knowles-Barley S, et al. Saturated reconstruction of a volume of neocortex. Cell. 2015;163(3):633-647. DOI: 10.1016/j.cell.2015.06.054
Tags: neuroanatomy:synapse neuroanatomy:dendrite neuroanatomy:axon imaging:ATUM imaging:serial-section case-studies:mouse case-studies:visual-cortex methodology:ground-truth case-studies:dense-reconstruction
Summaries
Beginner: This paper reconstructed every single neuron, synapse, and glial cell in a tiny cube of mouse brain — about the width of a hair in each direction. By mapping everything, the authors could count how many synapses each neuron makes and see patterns that would be invisible if only a few cells were traced. It showed both the incredible richness of brain wiring and how much work is needed to map it completely.
Intermediate: Kasthuri et al. performed “saturated” (dense) reconstruction of a 1,500 μm³ volume of mouse somatosensory cortex using ATUM-SEM. They identified all neuronal and glial profiles, mapped all synapses, and quantified connectivity patterns including the finding that axonal boutons form synapses on diverse targets rather than concentrating on single partners. The dataset became an important benchmark for segmentation algorithms and demonstrated that even small volumes reveal non-random connectivity motifs.
Advanced: This paper established methodological standards for dense reconstruction: the concept of “saturation” (all objects identified, all synapses mapped) and the quantitative reporting of completeness. Key methodological contributions include the systematic comparison of human annotation consistency, the use of synapse size as a proxy for strength, and the finding that a small number of strong connections dominate the connectivity distribution. The volume size limitations and the question of whether observations generalize to larger volumes remain active discussion points. The dataset (available through BOSS/BossDB) continues to serve as a ground truth benchmark.
Key figures: Fig. 1 (saturated volume overview), Fig. 3 (synapse size distribution), Fig. 5 (connectivity matrix), Fig. 7 (dense wiring diagram)
Discussion prompts:
- What does “saturated” reconstruction mean operationally, and how would you verify saturation in a new dataset?
- How do the connectivity patterns observed in this small volume compare with findings from larger datasets like MICrONS?
- Which of the biological conclusions might be artifacts of the small volume size?
Related content: Synapse classification, MICrONS visual cortex, Error taxonomy