Completing and experimentally validating the unified calcium-based model of NMDA receptor-dependent bidirectional synaptic plasticity: metaplastic stabilization, non-NMDA calcium sources, and the predicted long-delay pre-post LTD window

The unified model of Shouval, Bear, and Cooper (PNAS 2002) explains NMDA receptor-dependent bidirectional synaptic plasticity - pairing-, rate-, and spike-timing-dependent induction protocols - with a single calcium-based mechanism: postsynaptic calcium below a threshold produces no change, intermediate calcium produces LTD, high calcium produces LTP, with calcium entering predominantly through NMDA receptors gated by back-propagating action potentials that carry a slow after-depolarizing tail. The authors themselves specified what remained open: whether coupling the plasticity rule to activity-dependent regulation of NMDA receptor properties (metaplasticity) stabilizes the dynamics, and what its network and receptive-field consequences are; the role of non-NMDA calcium sources; the existence and width of a predicted LTD window at long positive spike-timing intervals; and the systematic parameter dependence of the quantitative windows. Later work advanced the theoretical side - a companion paper supplied the metaplasticity dynamics, and follow-up studies demonstrated stability, input selectivity and place-field dynamics in specific models - but a rigorous general derivation of the coupled dynamics and their network-level consequences, together with the experimental validation, remains open: whether the proposed coupling to activity-dependent NMDAR regulation actually stabilizes synaptic dynamics, the existence of the long-delay pre-post LTD window under physiological conditions, and the contribution of non-NMDA calcium sources to the predicted phenomena. The problem is the completion and experimental validation of this unified theory as its authors framed it.

Background

Long-term potentiation (LTP) and long-term depression (LTD) are persistent, activity-dependent increases and decreases in synaptic strength. In many brain regions both signs of modification depend on NMDA receptors (NMDARs) - glutamate-gated ion channels that are permeable to calcium and open only when presynaptic glutamate release coincides with sufficient postsynaptic depolarization, so that they act as coincidence detectors. In 2002 Shouval, Bear, and Cooper proposed a 'unified model' that explains three apparently distinct NMDAR-dependent induction protocols with a single mechanism: pairing protocols (constant presynaptic stimulation while varying postsynaptic membrane potential), rate protocols (varying the presynaptic firing rate), and spike-timing-dependent plasticity (STDP), in which the sign of the change depends on the interval $\Delta t$ between pre- and postsynaptic action potentials - classically LTP when the presynaptic spike precedes the postsynaptic one within tens of milliseconds and LTD for the reverse order (Bi and Poo 1998).

The model rests on three assumptions. (i) The calcium control hypothesis: the postsynaptic calcium concentration $[\mathrm{Ca}]$ alone determines the sign and rate of synaptic change - below a lower threshold $\theta_d$ nothing happens, between $\theta_d$ and a higher threshold $\theta_p$ the synapse is depressed, above $\theta_p$ it is potentiated - a mathematical version of earlier kinase-versus-phosphatase threshold proposals (Lisman 1989). (ii) NMDARs are the dominant source of the postsynaptic calcium that drives plasticity. (iii) The back-propagating action potential (BPAP, the action potential that actively travels from the soma back into the dendrites) carries a slow after-depolarizing (ADP) tail with a time constant of order 25 ms. With one fixed parameter set the model reproduces pairing, rate, and timing curves, and it makes distinctive falsifiable predictions: the STDP curve changes shape with stimulation frequency (LTP at all $\Delta t$ near 10 Hz); shifting NMDAR kinetics shifts the LTP/LTD thresholds, a form of metaplasticity - an activity-dependent change in the plasticity rule itself (Abraham and Bear 1996) - that can implement the sliding modification threshold postulated by the BCM theory of cortical plasticity (Bienenstock, Cooper, and Munro 1982); and, most distinctively, a previously uncharacterized LTD window at long positive $\Delta t$, in which a presynaptic spike preceding the postsynaptic one by more than the conventional LTP window produces depression.

The authors state the model's incompleteness themselves. A rule built on assumptions (i)-(iii) alone is expected to be unstable, like other Hebbian rules, so they proposed a fourth assumption - that metaplasticity implemented by activity-dependent regulation of NMDAR properties stabilizes the synapses - but the paper neither demonstrates that this coupling suffices for stability nor writes down the NMDAR plasticity dynamics (a companion paper supplies a proposal). Nor does it work out the network- and receptive-field-level consequences of the coupled equations (the receptive field of a neuron - the input pattern that effectively drives it - is what BCM-type rules were built to explain). They further concede that other postsynaptic calcium sources - voltage-gated calcium channels (VGCCs, dendritic membrane channels that admit calcium when locally depolarized) and release from intracellular calcium stores (the endoplasmic reticulum acting as an internal calcium reservoir) - were omitted and could modify the quantitative predictions (later reviews additionally emphasize calcium-permeable AMPA receptors; AMPA receptors are the fast glutamate-gated ion channels that carry most excitatory synaptic current, and a minority of them also pass calcium); that existing STDP data were too sparse at pre-post delays beyond about 20 ms to test the predicted long-delay pre-post LTD (one report, Nishiyama et al. 2000, was consistent with it); and that the quantitative windows and thresholds must vary across preparations because they depend on measurable but then-unmeasured quantities - the BPAP ADP time constant, the NMDAR kinetic composition, and calcium clearance time constants, including their activity dependence.

Problem Statement

Complete and experimentally validate - that is, test for consistency with the published experimental record - the unified calcium-based theory of NMDA receptor-dependent bidirectional synaptic plasticity that the 2002 model was proposed to found: decide whether the model's assumption set, closed by a dynamical equation for activity-dependent NMDA receptor regulation, is the correct mechanistic account of bidirectional plasticity. Four components jointly serve this single objective - all four being required for an affirmative resolution - as specified by the source paper's own discussion of what remains to be done:

1. Coupled dynamics and stabilization. Analyze the coupled system in which the fast synaptic dynamics

$$\frac{dW_j}{dt} = \eta([\mathrm{Ca}]_j)\,\bigl(\Omega([\mathrm{Ca}]_j) - \lambda W_j\bigr)$$

(with $W_j$ the strength of synapse $j$; $\eta$ a nonnegative learning-rate function that increases monotonically with postsynaptic calcium and sets the timescale of change; $\Omega$ the calcium-control function, negative for $\theta_d < [\mathrm{Ca}]_j < \theta_p$ and positive for $[\mathrm{Ca}]_j > \theta_p$; $\lambda$ a small weight-decay constant, set to 1 without loss of generality in the source; and postsynaptic calcium set by NMDAR currents gated by the BPAP waveform with its slow after-depolarizing tail) is coupled to a slow, activity-dependent evolution of NMDAR conductance and kinetics, as in the companion formulation of Shouval, Castellani, Blais, Yeung, and Cooper (2002). Determine whether this coupling suffices to stabilize the synaptic dynamics - bounded equilibria without Hebbian runaway - across the parameter envelope supplied by the sources, or whether additional mechanisms are required. That envelope is the parameter set of Shouval, Bear, and Cooper (2002) and the companion formulation - the calcium thresholds and learning-rate constants, the BPAP after-depolarization time constant, and the NMDAR kinetics used there to reproduce the published pairing, rate and timing curves - together with the ranges reported for those same quantities in the experimental studies cited below. The analysis must state the envelope it adopts and show that its verdict holds uniformly across it, so that the stability conclusion does not depend on the particular values chosen within those reported bounds.

2. The distinctive timing prediction. Determine, from existing experimental reports together with model computation, whether the predicted LTD region at long positive $\Delta t$ - depression when the presynaptic spike precedes the postsynaptic spike beyond the LTP window, roughly tens of milliseconds - exists, and its width and amplitude, as tested by whether the published experimental record, together with model computation, is consistent with such a window; existing data are relatively sparse at pre-post delays beyond about 20 ms. Because the predicted width depends on model parameters such as the NMDAR time constants, the BPAP ADP width, and the stimulation frequency, examine the window across the dimensions on which its width and amplitude are expected to depend - preparation and region, dendritic location, pairing frequency, postsynaptic spike count, and extracellular calcium concentration (hippocampal CA3-CA1, neocortex, and dissociated cultures being illustrative instances rather than a required checklist) - and assess how the other postsynaptic calcium sources omitted from the model (voltage-gated calcium channels, release from intracellular stores) contribute to the induction windows and thresholds, not only to the timing prediction. This assessment - whether the NMDAR-dominant account of the induction windows, thresholds and timing prediction survives once voltage-gated calcium channels and intracellular stores are included - is what discharges the objective's question about the role of non-NMDA calcium sources.

3. Parameter map. Determine, from existing experimental reports together with model computation, how the induction windows and thresholds depend on the BPAP ADP time constant, the NMDAR kinetic composition, and calcium clearance and buffering - and on the activity dependence of these quantities - so that the theory's quantitative predictions are fixed and testable across preparations, as judged by whether the published experimental record, together with model computation, is consistent with them.

4. Network consequences. Work out and test the network- and receptive-field-level consequences of the coupled synaptic-plus-metaplasticity equations: whether they yield stable receptive-field development (BCM-like sliding modification thresholds), input selectivity and synaptic competition, and bounded recurrent network dynamics, as tested by model computation against the published experimental and modelling literature.

The completion boundary is the one the source authors specified: the model's key assumptions and previously uncharacterized consequences 'can and should be tested experimentally', and only if they pass does it 'serve as a basis for a unified theory of synaptic plasticity'. That test is conjunctive: the four components are jointly required for an affirmative resolution, whereas a decisive negative result on any single one of them - the coupling shown not to stabilize across the parameter envelope set out in component 1, the predicted long-delay window excluded by the published experimental record or the omitted non-NMDA sources shown to overturn the NMDAR-dominant account of the windows and thresholds, the parameter map shown not to fix the quantitative predictions across preparations, or the network- and receptive-field-level consequences shown not to follow - identifies the specific failing assumption and resolves the task by refutation.

References

1. Harel Z. Shouval, Mark F. Bear, Leon N. Cooper (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
2. Harel Z. Shouval, Gastone C. Castellani, Brian S. Blais, Luk C. Yeung, Leon N. Cooper (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
3. Luk Chong Yeung, Harel Z. Shouval, Brian S. Blais, Leon N. Cooper (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
4. Yidao Cai, Jeffrey P. Gavornik, Leon N. Cooper, Luk C. Yeung, Harel Z. Shouval (2007), 'Effect of stochastic synaptic and dendritic dynamics on synaptic plasticity in visual cortex and hippocampus', Journal of Neurophysiology 97(1):375-386. DOI: 10.1152/jn.00895.2006, https://doi.org/10.1152/jn.00895.2006
5. Xintian Yu, Harel Z. Shouval, James J. Knierim (2008), 'A biophysical model of synaptic plasticity and metaplasticity can account for the dynamics of the backward shift of hippocampal place fields', Journal of Neurophysiology 100(2):983-992. DOI: 10.1152/jn.01256.2007, https://doi.org/10.1152/jn.01256.2007
6. G. M. Wittenberg, S. S.-H. Wang (2006), 'Malleability of spike-timing-dependent plasticity at the CA3-CA1 synapse', Journal of Neuroscience 26(24):6610-6617. DOI: 10.1523/JNEUROSCI.5388-05.2006, https://doi.org/10.1523/JNEUROSCI.5388-05.2006
7. Makoto Nishiyama, Kyonsoo Hong, Katsuhiko Mikoshiba, Mu-ming Poo, Kunio Kato (2000), 'Calcium stores regulate the polarity and input specificity of synaptic modification', Nature 408(6812):584-588. DOI: 10.1038/35046067, https://doi.org/10.1038/35046067
8. Harel Z. Shouval, S. S.-H. Wang, G. M. Wittenberg (2010), 'Spike timing dependent plasticity: a consequence of more fundamental learning rules', Frontiers in Computational Neuroscience 4:19. DOI: 10.3389/fncom.2010.00019, https://doi.org/10.3389/fncom.2010.00019
9. Michael Graupner, Nicolas Brunel (2012), 'Calcium-based plasticity model explains sensitivity of synaptic changes to spike pattern, rate, and dendritic location', Proceedings of the National Academy of Sciences USA 109(10):3991-3996. DOI: 10.1073/pnas.1109359109, https://doi.org/10.1073/pnas.1109359109
10. Yanis Inglebert, Johnatan Aljadeff, Nicolas Brunel, Dominique Debanne (2020), 'Synaptic plasticity rules with physiological calcium levels', Proceedings of the National Academy of Sciences USA 117(52):33639-33648. DOI: 10.1073/pnas.2013663117, https://doi.org/10.1073/pnas.2013663117
11. Yanis Inglebert, Dominique Debanne (2021), 'Calcium and spike timing-dependent plasticity', Frontiers in Cellular Neuroscience 15:727336. DOI: 10.3389/fncel.2021.727336, https://doi.org/10.3389/fncel.2021.727336
12. J. L. Hellier, D. R. Grosshans, S. J. Coultrap, J. P. Jones, P. Dobelis, M. D. Browning, K. J. Staley (2007), 'NMDA receptor trafficking at recurrent synapses stabilizes the state of the CA3 network', Journal of Neurophysiology 98(5):2818-2826. DOI: 10.1152/jn.00346.2007, https://doi.org/10.1152/jn.00346.2007
13. Thomas Nevian, Bert Sakmann (2006), 'Spine Ca2+ Signaling in Spike-Timing-Dependent Plasticity', Journal of Neuroscience 26(43):11001-11013. DOI: 10.1523/JNEUROSCI.1749-06.2006, https://doi.org/10.1523/JNEUROSCI.1749-06.2006
14. Anne-Kathrin Theis, Balazs Rozsa, Gergely Katona, Dietmar Schmitz, Friedrich W. Johenning (2018), 'Voltage gated calcium channel activation by backpropagating action potentials downregulates NMDAR function', Frontiers in Cellular Neuroscience 12:109. DOI: 10.3389/fncel.2018.00109, https://doi.org/10.3389/fncel.2018.00109
15. Toviah Moldwin, Menachem Kalmenson, Idan Segev (2023), 'Asymmetric voltage attenuation in dendrites can enable hierarchical heterosynaptic plasticity', eNeuro 10(7):ENEURO.0014-23.2023. DOI: 10.1523/ENEURO.0014-23.2023, https://doi.org/10.1523/ENEURO.0014-23.2023
16. Georgios Kalantzis, Harel Z. Shouval (2009), 'Structural plasticity can produce metaplasticity', PLoS ONE 4(11):e8062. DOI: 10.1371/journal.pone.0008062, https://doi.org/10.1371/journal.pone.0008062
17. Toviah Moldwin, Li Shay Azran, Idan Segev (2025), 'A generalized mathematical framework for the calcium control hypothesis describes weight-dependent synaptic plasticity', Journal of Computational Neuroscience 53(2):333-357. DOI: 10.1007/s10827-025-00894-6, https://doi.org/10.1007/s10827-025-00894-6
18. Toviah Moldwin, Li Shay Azran, Idan Segev (2025), 'The calcitron: A simple neuron model that implements many learning rules via the calcium control hypothesis', PLoS Computational Biology 21(1):e1012754. DOI: 10.1371/journal.pcbi.1012754, https://doi.org/10.1371/journal.pcbi.1012754
19. Wickliffe C. Abraham, Mark F. Bear (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
20. E. L. Bienenstock, L. N. Cooper, P. W. Munro (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
21. Guo-qiang Bi, Mu-ming Poo (1998), 'Synaptic modifications in cultured hippocampal neurons: dependence on spike timing, synaptic strength, and postsynaptic cell type', Journal of Neuroscience 18(24):10464-10472. DOI: 10.1523/JNEUROSCI.18-24-10464.1998, https://doi.org/10.1523/JNEUROSCI.18-24-10464.1998
22. J. Lisman (1989), 'A mechanism for the Hebb and the anti-Hebb processes underlying learning and memory', Proceedings of the National Academy of Sciences USA 86(23):9574-9578. DOI: 10.1073/pnas.86.23.9574, https://doi.org/10.1073/pnas.86.23.9574
23. Kevin Kadak, Davide Momi, Zheng Wang, Sorenza P. Bastiaens, Mohammad P. Oveisi, Taha Morshedzadeh, Minarose Ismail, Jan Fousek, John D. Griffiths (2025), 'Alpha rhythm subharmonics underlie responsiveness to theta burst stimulation via calcium metaplasticity', eLife (reviewed preprint). DOI: 10.7554/eLife.108563.1, https://doi.org/10.7554/eLife.108563.1