Closed-book appointment exam · independently graded
Professor — Neurobiology. The candidate agent answered from its own knowledge, closed-book; a second, independent examiner agent graded it adversarially.
vaiu-sci-bio-prof-neuro v1.0.0 — Professor of Biology (Neurobiology)AI-transparency disclosure. Opened as an AI faculty agent, closed-book; no clinical diagnosis/treatment advice; holds the levels-of-analysis distinction and is strict about correlation-vs-causation and necessity-vs-sufficiency; attributes eponyms, flags folklore, no fabricated references.
Preparation: squid giant axon, in vitro, ~6°C. Resting −65/−70 mV; selective permeability + gradients held by the Na/K-ATPase (3 Na out, 2 K in per ATP). Nernst E_ion=(RT/zF)ln([out]/[in]) ≈ (61/z)log₁₀ mV; E_K≈−90, E_Na≈+60, E_Cl≈rest, E_Ca very positive. Resting NOT a single Nernst — Goldman–Hodgkin–Katz permeability-weighted (P_K dominates → V_rest near E_K); moves toward the equilibrium of the dominant conductance. HH 1952 voltage clamp: g_Na∝m³h (3 activation gates m fast, 1 inactivation gate h slow, closes with maintained depolarization), g_K∝n⁴ (4 activation gates n slower, no inactivation classically); fast Na activation vs slower Na inactivation + K activation → a transient spike. Threshold = inward Na exceeds outward K + leak, a regenerative positive-feedback tipping point (not a fixed voltage). Spike: threshold → m open → V toward E_Na (overshoot); h inactivate + n open → back toward E_K; K lags → afterhyperpolarization. Refractory: absolute (Na inactivated, h shut, recovers with repolarization), relative (some Na recovered, K still elevated/hyperpolarized); enforces directional propagation + caps rate. Propagation: local inward current spreads passively to adjacent membrane; cable theory λ=√(r_m/r_i) (high r_m, low r_i); faster via diameter (squid) or myelination (Schwann PNS / oligodendrocyte CNS raise r_m, lower capacitance, jump between nodes of Ranvier where Na is clustered = saltatory). Demyelination degrades conduction (won't extend into clinical territory).
AP invades the terminal → voltage-gated Ca²⁺ channels open → Ca²⁺ enters; Ca²⁺ binds synaptotagmin (sensor) + the SNARE complex (synaptobrevin/VAMP on the vesicle; syntaxin + SNAP-25 on the membrane) drives fusion of docked/primed vesicles at the active zone → exocytosis into the cleft. Quantal (del Castillo–Katz, frog NMJ): transmitter released in quanta = single vesicles; spontaneous minis set quantal size; the evoked response is an integer-like sum of quanta; release is probabilistic (Ca²⁺ raises release probability). Postsynaptic: ionotropic (receptor IS the channel, fast ms — glutamate/AMPA → Na → EPSP; GABA/GABA_A → Cl → IPSP; ACh/nicotinic) vs metabotropic (GPCR, receptor NOT a channel → G protein → second messengers cAMP / IP3-DAG / Ca²⁺, slower tens-of-ms to s, modulatory, can affect channels indirectly + gene expression — mGluR/GABA_B/muscarinic/monoamine, amplifying and long-lasting). Termination: reuptake transporters / degradation (acetylcholinesterase) / diffusion. Electrical synapses = gap junctions (connexons/connexins), direct cytoplasmic channels, no synaptic delay, usually bidirectional, pass current directly, synchronize populations (escape/oscillation); limits — little gain, less plasticity.
Hebb 1949: if A repeatedly takes part in firing B, the connection strengthens ("fire together, wire together" = later folklore, not his exact words); coincidence detection. LTP at glutamatergic synapses (canonically hippocampal CA1, rodent slice): the NMDA receptor is the coincidence detector — needs glutamate bound (presynaptic activity) AND postsynaptic depolarization to expel the Mg²⁺ block; it is Ca²⁺-permeable, so the postsynaptic Ca²⁺ transient is the trigger. Large/fast Ca²⁺ → LTP (CaMKII → AMPA-receptor phosphorylation/insertion + increased conductance; late LTP needs new protein synthesis/transcription — CREB — and spine enlargement). Modest/prolonged Ca²⁺ → LTD (phosphatases — calcineurin/PP1 — AMPA removal). The sign is set by Ca²⁺ amplitude/timing: one detector, two outcomes. STDP: pre-before-post potentiates, post-before-pre depresses (ms). Evidence: correlational (learning changes synaptic strength — suggestive), loss-of-function (NMDAR block impairs some spatial learning — necessity), gain-of-function/sufficiency (engram tagging + optogenetic reactivation drives the learned behavior; manipulating plasticity alters memory — Tonegawa); necessity + sufficiency earns the causal claim. Limits: LTP is an operationally defined phenomenon in a preparation (often slice + tetanic stimulation), a candidate mechanism not proven identical to natural memory; necessity ≠ sufficiency (NMDAR block affects many things); species/region (CA1 rodent ≠ human in vivo); systems-level memory (consolidation) is a different level, not fully reduced to the synaptic mechanism.
E/I balance: glutamatergic excitation counterpoised by GABAergic inhibition, dynamically matched — keeps networks out of runaway excitation and total suppression, sharpens temporal precision, controls gain, creates spiking windows; inhibition structures computation. Motif: feedforward inhibition (an input excites a principal cell and, in parallel, an interneuron that inhibits it a moment later → narrow temporal integration window, coincidence detection) vs feedback/recurrent inhibition (principal cells excite interneurons that inhibit them back → negative feedback, gain control/normalization, lateral inhibition sharpens contrast). Rate code (frequency averaged over a window; robust, discards timing) vs temporal code (precise spike timing/latency/phase; faster/higher-capacity, needs reliable timing + a reference); which the brain uses is context-dependent and debated (open). Optogenetics: ChR2 to activate (blue light), halorhodopsin/archaerhodopsin to silence; necessity = silence during the behavior (fails → necessary, loss-of-function); sufficiency = activate in the absence of the natural stimulus (behavior appears → sufficient, gain-of-function); necessity AND sufficiency together license "this element drives this behavior"; one alone is weak (necessary-not-sufficient is common; sufficiency without necessity can be an artificial bypass); control off-target/antidromic/non-physiological synchrony; state species/preparation. Correlational calcium imaging (a population "lights up") = a correlate only, good for hypothesis generation, insufficient for causation.
Axon guidance: a growth cone (motile, filopodia/lamellipodia, actin-driven) samples the environment and steers by guidance cues (attractant/repellent, contact-mediated or diffusible): netrins (attractive via DCC, repulsive via UNC5; midline), slits (repulsive via Robo; cross the midline once), ephrins/Eph (contact repulsion, graded → topographic maps, retinotectal — Sperry chemoaffinity, matching molecular labels), semaphorins (largely repulsive via plexins/neuropilins); transduced into Ca²⁺/Rho-GTPase steering, stepwise via intermediate targets/guideposts. Activity-dependent development: Hebbian competition (co-active inputs stabilized, poorly-correlated eliminated), activity-dependent pruning + neurotrophic dependence (NGF/BDNF target-derived, neurons compete for survival; NGF — Levi-Montalcini/Hamburger). Critical periods: ocular dominance plasticity (Hubel–Wiesel; monocular deprivation during the critical period shifts cortical territory toward the open eye; far less after; cat/monkey visual cortex). Hard-wired (activity-independent): gross axon guidance, the coarse topographic scaffold, tract formation, the broad wiring plan (molecularly specified). Experience-dependent: fine-tuning — synaptic weights, column refinement, elimination of exuberant connections. Intermediate: activity-dependent-but-NOT-experience-dependent — spontaneous correlated activity (retinal waves) refining eye-specific segregation before eyes open. Nature builds the scaffold + molecular address system; activity (first spontaneous, then experiential) carves the final circuit; a critical period gates when experience is most instructive.
vaiu-sci-bcs-*); flagged the reverse-inference error (activation ≠ mental process). Taught the underlying mechanism (BOLD reflects neurovascular coupling; LFP/synaptic activity correlates better than spiking; resolution limits mechanistic claims; cellular/circuit tools — ephys/optogenetics — move from correlation to causation) as its own domain.