The BCM (Bienenstock–Cooper–Munro) theory of visual cortical plasticity stabilizes Hebbian learning through a moving modification threshold Θm\Theta_m — the level of postsynaptic activity separating synaptic depression from potentiation — that slides as a superlinear function of the neuron's slowly averaged activity. Intrator and Cooper's 1992 objective-function reformulation, which casts BCM learning as stochastic gradient descent on a projection-pursuit-like statistical objective (a search for projections of the input with maximal non-Gaussian structure), requires the threshold to be the second moment of the postsynaptic response (Θm∝⟨c2⟩\Theta_m \propto \langle c^2\rangle) rather than the square of its average (Θm∝⟨c⟩2\Theta_m \propto \langle c\rangle^2) as in the original 1982 proposal; the two laws differ by the response variance, and the paper left open which is biologically correct and what physiological mechanism implements the moving threshold at all. Decades of subsequent work established the phenomenon (experience-dependent sliding of the LTP/LTD crossover in visual cortex) and produced a family of candidate mechanisms — NMDA-receptor subunit regulation, CaMKII autophosphorylation, HCN hh-current regulation, slow calcium homeostasis, neuromodulator-dependent expression changes — but no identified implementation. The problem — identify the molecular machinery that integrates prior activity and moves the threshold, and determine the correct threshold law — remains open.

Background

Long-term potentiation (LTP) and long-term depression (LTD) — the persistent strengthening and weakening of synapses following neural activity — are the leading cellular substrate of learning in the cortex. At most excitatory cortical synapses both are gated by the NMDA-type glutamate receptor, a channel that admits calcium only when presynaptic transmitter release coincides with postsynaptic depolarization; modest postsynaptic calcium elevation triggers LTD, whereas stronger elevation triggers LTP. A pure Hebbian rule that only reinforces co-active synapses is unstable — potentiation raises activity, which begets further potentiation — so a viable plasticity rule needs a homeostatic component that counteracts runaway.

The Bienenstock–Cooper–Munro (BCM) theory supplied one for visual cortical development. For a cell with synaptic weight vector mm and postsynaptic response c=d⋅mc=d\cdot m to input pattern dd, the BCM update is m˙=ϕ(c,Θm) d\dot m=\phi(c,\Theta_m)\,d with modification function ϕ(c,Θm)=c (c−Θm)\phi(c,\Theta_m)=c\,(c-\Theta_m): responses below the modification threshold Θm\Theta_m weaken the active synapses, responses above it strengthen them. The crucial postulate is that Θm\Theta_m is a moving threshold — a superlinear, conventionally quadratic function of the slowly time-averaged postsynaptic activity,

Θm≃⟨c⟩2,\Theta_m \simeq \langle c\rangle^2,

so that chronically active cells raise their threshold (LTP becomes harder, LTD easier) while quiet cells lower it. This single postulate accounted for ocular dominance shifts under monocular deprivation (the redistribution of cortical responsiveness between the eyes when one eye is closed during development), for the development of orientation selectivity, and for the effects of dark rearing (raising animals without patterned visual input during development), and it is the theoretical ancestor of metaplasticity — the activity-history-dependent regulation of synaptic plasticity. Experiments later confirmed the phenomenon: in visual cortex of light-deprived rats, the LTP/LTD crossover point, the experimental proxy for Θm\Theta_m, shifts with sensory experience (Kirkwood, Rioult & Bear, 1996).

Intrator and Cooper (1992) recast the BCM rule as stochastic gradient descent on an explicit objective function, in the spirit of projection pursuit — a family of statistical methods that search for low-dimensional projections of multivariate data with maximal non-Gaussian structure. The identification works, and yields fixed-point and stability theorems for the learning dynamics, when the moving threshold is taken as the second moment of the projected activity,

Θm=E[(x⋅m)2]=⟨c2⟩,\Theta_m = E[(x\cdot m)^2] = \langle c^2\rangle,

rather than the square of its mean; the two candidate laws differ by the variance, ⟨c2⟩=Var(c)+⟨c⟩2\langle c^2\rangle=\mathrm{Var}(c)+\langle c\rangle^2. The reformulation reproduced the qualitative outcomes of the classical rearing manipulations (normal rearing, monocular and binocular deprivation, reversed suture — reopening the deprived eye while closing the previously open one — and strabismus, a surgically induced misalignment of the two eyes), but the authors noted that the plasticity experiments available at the time were consistent with either threshold form, and left open what microscopic physiological mechanism actually computes and implements the moving threshold. Subsequent decades verified the sliding-threshold phenomenon and proposed a family of candidate molecular implementations — experience-dependent NMDA-receptor subunit composition (Philpot et al., 2001), calcium-control and NMDAR-kinetics models (Shouval, Bear & Cooper, 2002; Yeung et al., 2004), and regulation by the hyperpolarization-activated nonspecific cation (hh) current carried by HCN (hyperpolarization-activated cyclic nucleotide-gated) channels (Narayanan & Johnston, 2010) — without settling which machinery operates in vivo.

Problem Statement

Identify the microscopic physiological mechanism that implements the BCM moving modification threshold in postsynaptic neurons. Concretely: (i) which molecular machinery integrates prior postsynaptic activity on the slow time scale (seconds to hours) and shifts the LTP/LTD crossover point Θm\Theta_m — candidate substrates proposed in the literature include experience-dependent NMDA-receptor subunit composition and kinetics, autophosphorylation of CaMKII (Ca2+\mathrm{Ca}^{2+}/calmodulin-dependent protein kinase II, a postsynaptic kinase whose self-sustaining active state can hold a biochemical trace of recent calcium elevations) and its phosphatase-dependent reversal, the HCN-mediated hh current, calcium-buffer and dendritic-spine structural regulation, and neuromodulator-dependent changes in the expression mechanisms of LTP/LTD, but an unanticipated substrate is equally admissible; and (ii) what is the correct functional law for the threshold — Θm∝⟨c⟩2\Theta_m\propto\langle c\rangle^2 (the square of the time-averaged postsynaptic response, as in the original BCM proposal), Θm∝⟨c2⟩\Theta_m\propto\langle c^2\rangle (the time-averaged second moment, as required by the objective-function formulation of Intrator & Cooper 1992), or another superlinear dependence on the activity history.

The two parts are distinct and both are required: fixing the functional law does not by itself identify the machinery, and naming a substrate does not by itself fix the law, so answering only one of them does not resolve the task. Resolving each of the two parts requires arguing for a specific mechanism or functional law rather than restating a candidate substrate or threshold law already named above, and supporting that choice against the existing experience-dependent crossover-shift evidence.

References

  1. Intrator N, Cooper LN (1992). Objective function formulation of the BCM theory of visual cortical plasticity: Statistical connections, stability conditions. Neural Networks 5(1):3–17. DOI 10.1016/s0893-6080(05)80003-6. https://doi.org/10.1016/s0893-6080(05)80003-6
  2. Bienenstock EL, Cooper LN, Munro PW (1982). Theory for the development of neuron selectivity: orientation specificity and binocular interaction in visual cortex. Journal of Neuroscience 2(1):32–48. DOI 10.1523/JNEUROSCI.02-01-00032.1982. https://doi.org/10.1523/JNEUROSCI.02-01-00032.1982
  3. Bear MF, Cooper LN, Ebner FF (1987). A physiological basis for a theory of synapse modification. Science 237(4810):42–48. DOI 10.1126/science.3037696. https://doi.org/10.1126/science.3037696
  4. Gold JI, Bear MF (1994). A model of dendritic spine Ca2+ concentration exploring possible bases for a sliding synaptic modification threshold. Proceedings of the National Academy of Sciences USA 91(9):3941–3945. DOI 10.1073/pnas.91.9.3941. https://doi.org/10.1073/pnas.91.9.3941
  5. Bear MF (1995). Mechanism for a sliding synaptic modification threshold. Neuron 15(1):1–4. DOI 10.1016/0896-6273(95)90056-x. https://doi.org/10.1016/0896-6273(95)90056-x
  6. Kirkwood A, Rioult MC, Bear MF (1996). Experience-dependent modification of synaptic plasticity in visual cortex. Nature 381(6582):526–528. DOI 10.1038/381526a0. https://doi.org/10.1038/381526a0
  7. Abraham WC, Bear MF (1996). Metaplasticity: the plasticity of synaptic plasticity. Trends in Neurosciences 19(4):126–130. DOI 10.1016/s0166-2236(96)80018-x. https://doi.org/10.1016/s0166-2236(96)80018-x
  8. Philpot BD, Sekhar AK, Shouval HZ, Bear MF (2001). Visual experience and deprivation bidirectionally modify the composition and function of NMDA receptors in visual cortex. Neuron 29(1):157–169. DOI 10.1016/s0896-6273(01)00187-8. https://doi.org/10.1016/s0896-6273(01)00187-8
  9. Abraham WC, Mason-Parker SE, Bear MF, Webb S, Tate WP (2001). Heterosynaptic metaplasticity in the hippocampus in vivo: a BCM-like modifiable threshold for LTP. Proceedings of the National Academy of Sciences USA 98(19):10924–10929. DOI 10.1073/pnas.181342098. https://doi.org/10.1073/pnas.181342098. PMID 11517323.
  10. Shouval HZ, Bear MF, Cooper LN (2002). A unified model of NMDA receptor-dependent bidirectional synaptic plasticity. Proceedings of the National Academy of Sciences USA 99(16):10831–10836. DOI 10.1073/pnas.152343099. https://doi.org/10.1073/pnas.152343099
  11. Shouval HZ, Castellani GC, Blais BS, Yeung LC, Cooper LN (2002). Converging evidence for a simplified biophysical model of synaptic plasticity. Biological Cybernetics 87(5–6):383–391. DOI 10.1007/s00422-002-0362-x. https://doi.org/10.1007/s00422-002-0362-x
  12. Yeung LC, Shouval HZ, Blais BS, Cooper LN (2004). Synaptic homeostasis and input selectivity follow from a calcium-dependent plasticity model. Proceedings of the National Academy of Sciences USA 101(41):14943–14948. DOI 10.1073/pnas.0405555101. https://doi.org/10.1073/pnas.0405555101
  13. Castellani GC, Quinlan EM, Bersani F, Cooper LN, Shouval HZ (2005). A model of bidirectional synaptic plasticity: From signaling network to channel conductance. Learning & Memory 12(4):423–432. DOI 10.1101/lm.80705. https://doi.org/10.1101/lm.80705
  14. Narayanan R, Johnston D (2010). The h current is a candidate mechanism for regulating the sliding modification threshold in a BCM-like synaptic learning rule. Journal of Neurophysiology 104(2):1020–1033. DOI 10.1152/jn.01129.2009. https://doi.org/10.1152/jn.01129.2009
  15. Cooper LN, Bear MF (2012). The BCM theory of synapse modification at 30: interaction of theory with experiment. Nature Reviews Neuroscience 13(11):798–810. DOI 10.1038/nrn3353. https://doi.org/10.1038/nrn3353
  16. Huang S, Treviño M, He K, Ardiles A, del Pasquale R, Guo Y, Palacios A, Huganir R, Kirkwood A (2012). Pull-push neuromodulation of LTP and LTD enables bidirectional experience-induced synaptic scaling in visual cortex. Neuron 73(3):497–510. DOI 10.1016/j.neuron.2011.11.023. https://doi.org/10.1016/j.neuron.2011.11.023
  17. Lee HK (2022). Metaplasticity framework for cross-modal synaptic plasticity in adults. Frontiers in Synaptic Neuroscience 14:1087042. DOI 10.3389/fnsyn.2022.1087042. https://doi.org/10.3389/fnsyn.2022.1087042. PMID 36685084.
  18. Leet MP, Bear MF, Gaier ED (2022). Metaplasticity: a key to visual recovery from amblyopia in adulthood? Current Opinion in Ophthalmology 33(6):512–518. DOI 10.1097/icu.0000000000000901. https://doi.org/10.1097/icu.0000000000000901