The authors have declared that no competing interests exist.
Current address: The Solomon H. Snyder Department of Neuroscience, Johns Hopkins University, Baltimore, MD 21205, United States of America
A plethora of experimental studies have shown that long-term synaptic plasticity can be expressed pre- or postsynaptically depending on a range of factors such as developmental stage, synapse type, and activity patterns. The functional consequences of this diversity are not clear, although it is understood that whereas postsynaptic expression of plasticity predominantly affects synaptic response amplitude, presynaptic expression alters both synaptic response amplitude and short-term dynamics. In most models of neuronal learning, long-term synaptic plasticity is implemented as changes in connective weights. The consideration of long-term plasticity as a fixed change in amplitude corresponds more closely to post- than to presynaptic expression, which means theoretical outcomes based on this choice of implementation may have a postsynaptic bias. To explore the functional implications of the diversity of expression of long-term synaptic plasticity, we adapted a model of long-term plasticity, more specifically spike-timing-dependent plasticity (STDP), such that it was expressed either independently pre- or postsynaptically, or in a mixture of both ways. We compared pair-based standard STDP models and a biologically tuned triplet STDP model, and investigated the outcomes in a minimal setting, using two different learning schemes: in the first, inputs were triggered at different latencies, and in the second a subset of inputs were temporally correlated. We found that presynaptic changes adjusted the speed of learning, while postsynaptic expression was more efficient at regulating spike timing and frequency. When combining both expression loci, postsynaptic changes amplified the response range, while presynaptic plasticity allowed control over postsynaptic firing rates, potentially providing a form of activity homeostasis. Our findings highlight how the seemingly innocuous choice of implementing synaptic plasticity by single weight modification may unwittingly introduce a postsynaptic bias in modelling outcomes. We conclude that pre- and postsynaptically expressed plasticity are not interchangeable, but enable complimentary functions.
Differences between functional properties of pre- or postsynaptically expressed long-term plasticity have not yet been explored in much detail. In this paper, we used minimalist models of STDP with different expression loci, in search of fundamental functional consequences. Biologically, presynaptic expression acts mostly on neurotransmitter release, thereby altering short-term synaptic dynamics, whereas postsynaptic expression affects mainly synaptic gain. We compared models where plasticity was expressed only presynaptically or postsynaptically, or in both ways. We found that postsynaptic plasticity had a bigger impact over response times, while both pre- and postsynaptic plasticity were similarly capable of detecting correlated inputs. A model with biologically tuned expression of plasticity achieved the same outcome over a range of frequencies. Also, postsynaptic spiking frequency was not directly affected by presynaptic plasticity of short-term plasticity alone, however in combination with a postsynaptic component, it helped restrain positive feedback, contributing to activity homeostasis. In conclusion, expression locus may determine affinity for distinct coding schemes while also contributing to keep activity within bounds. Our findings highlight the importance of carefully implementing expression of plasticity in biological modelling, since the locus of expression may affect functional outcomes in simulations.
Long-term synaptic plasticity is widely thought to underlie learning and memory as well as developmental circuit refinement [
After the discovery by Bliss and Lømo [
Going back several decades, a multitude of highly influential computer models of neocortical learning and development have been proposed, some of them focusing on aspects such as the dependence of induction on firing rates [
However, the expression of plasticity is not always well modelled by this sole change of synaptic weight. This is because presynaptically expressed plasticity leads to changes in synaptic dynamics, whereas postsynaptic expression does not (
Experimentally, it is long known that the induction of neocortical long-term plasticity may for example alter short-term depression [
Here, we use computational modelling to explore the consequences of expressing plasticity pre- or postsynaptically in a single neuron under two simple paradigms (
From a conceptual point of view, a synapse receives the output of a presynaptic neuron and transforms it into an input for the postsynaptic neuron. Most phenomenological models implement this by scaling the signal amplitude by a specific value, or ‘weight’. However, it is known that a presynaptic action potential doesn’t always elicit an output, which means this transmission is unreliable. The probability of transmission is largely determined by the probability of vesicle release from the presynaptic side. We take these two factors in a minimalist model of a synapse (
Postsynaptically expressed plasticity is readily implemented as a simple change in synaptic gain, by adjusting the quantal amplitude,
Presynaptic plasticity, however, has at least two different distinct types of impact on a synapse. First, the reliability and noise levels of neurotransmission are altered by presynaptic plasticity, because vesicle release is stochastic. Assuming release is binomially distributed, increasing the probability of release,
To limit the scope of the study, we focus on early forms of plasticity for which we have detailed experimental data [
We furthermore decompose presynaptic plasticity to distinguish between two distinct types of impact: unreliable transmission without short-term dynamics, or with short-term dynamics. We start with presynaptic expression modelled as direct changes in the probability of vesicle release (
Here, changes in the presynaptic weight
With the latency paradigm, STDP leads early inputs to potentiate and late inputs to depress. In this paradigm, a volley of stimuli arrives at the postsynaptic neuron with varying delays (
Next, we explored the correlation paradigm, in which plasticity selectively potentiates correlated inputs (
To explore this exception in more detail, we ran simulations where all of inputs were correlated, but half of them expressed plasticity only presynaptically, and the other half only postsynaptically. We imposed a limit to the total sum of weights so these two input populations competed, so that one potentiated at the expense of the other, which depressed. With this approach, we systematically explored the correlation-frequency space in distinct simulations where all inputs had a specified correlation and firing rate. We found that postsynaptic expression won for very highly correlated inputs for sufficiently low input frequencies (
We next explored the effects of altering short-term dynamics (see
In the simulations with the timed input configuration, results differed considerably depending on the specific locus of plasticity in the latency configuration. Postsynaptic expression alone provided the largest latency reduction, and also achieved it faster than the other plasticity implementations (
On the other hand, under modulation of short-term plasticity, plasticity rates in the correlated inputs paradigm evolved differently compared to the above stochastic release implementation (
These findings show that the outcome in the latency paradigm was more affected by the locus of expression of plasticity than in the correlation paradigm. In conclusion, computational advantages could be tailored to optimally achieve a specific functional outcome by recruiting pre- or postsynaptic plasticity differentially.
The above minimalist toy models had the advantage that they provided full control of several key parameters. However, the relevance of the findings for the intact brain was unclear. To address this shortcoming, we explored the biological plausibility in a model [
We first explored the latency paradigm (
In keeping with experimental results [
In the correlation paradigm (
To better understand the robustness of this property, we quantified the capacity of separation between correlated and uncorrelated populations with a linear separator. It was trained to classify inputs as correlated or uncorrelated according to the average and variance of
In the same way as in the latency paradigm (
In recent years, it has become clear that diversity in LTP expression is both ubiquitous and considerable, depending on factors such as animal age, induction protocol, and precise brain region [
We investigated two different learning paradigms, one with differently timed inputs, in which postsynaptic latency to spike was used as a measure of learning (
Our study showed that the locus of expression of plasticity determined affinity for different coding schemes. Presynaptic plasticity expressed as the regulation of release probability alone did not result in appreciable differences for steady-state postsynaptic activity compared to postsynaptic expression (
Presynaptically expressed plasticity alone was not ideally suited for changing rate coding, because presynaptic short-term plasticity acts as a filter on the presynaptic firing rate. As a consequence, postsynaptic instantaneous firing frequency shows reduced changes (Figs
Frequently, the effect of unreliability of single synapses is considered to simply be one of noise or energy economy [
The same basic properties were observed in the biologically tuned model with simultaneous pre- and postsynaptic plasticity. Learning was dramatically affected by postsynaptic plasticity, while the presynaptic side appeared to act more on the rate of learning and on weight dynamics. It is possible that these results could be modified according to the ratio of pre- versus postsynaptic forms of plasticity, to optimally achieve a specific computational outcome. It is noteworthy that the biologically tuned model was also capable of separating groups of correlated and uncorrelated inputs without the need for a hard competitive mechanism.
Since it is possible to specifically block pre- or postsynaptic STDP pharmacologically [
Here, we have challenged the standard assumption that modelling synaptic plasticity as a weight change is neutral and unbiased. To do so, we relied on two classic STDP studies [
As our collective understanding of the expression of long-term plasticity has improved, it has become clear that the long-held notion that plasticity is expressed predominantly postsynaptically is erroneous [
All of the simulations consisted of one postsynaptic neuron receiving a number of presynaptic Poisson inputs. In the first section, we used a simple leaky integrate-and-fire model defined by
Inputs were received with probability
The corresponding parameters for a pyramidal neuron were
The postsynaptic neuron was in one of two stimulus paradigms. The first one was based on [
The second paradigm was based on [
For the majority of the simulations we opted to implement STDP with the simple additive model proposed by Song and Abbott [
Each increment to the synaptic weights
To enable comparison of convergence rates for different types of plasticity expression, we ensured that the weight change Δ
The initial value of all simulations was the same for
We also kept the same range of total
We compared the results of the straightforward additive model to a slightly more complex STDP model that acts separately over pre- and postsynaptic factors [
The parameter values were taken from [
In the last section, we used a linear least-squares classifier to infer whether presynaptic inputs were correlated or uncorrelated. A linear model was fitted to separate the values of synaptic weight averages and variances from half of the inputs (labelled correlated or uncorrelated), and then used to classify the other half of inputs.
Presynaptic control of the probability of release per stimulus was implemented either as a Markovian process or as short-term plasticity, with presynaptic weight
Depression and facilitation time constants,
(PDF)
We thank Alanna Watt and Mark van Rossum for suggestions, help, and useful discussions.
Dear Dr Sjöström,
Thank you very much for submitting your manuscript "Pre- and postsynaptically expressed spike-timing-dependent plasticity contribute differentially to neuronal learning" for consideration at PLOS Computational Biology.
As with all papers reviewed by the journal, your manuscript was reviewed by members of the editorial board and by several independent reviewers. In light of the reviews (below this email), we would like to invite the resubmission of a significantly-revised version that takes into account the reviewers' comments.
We cannot make any decision about publication until we have seen the revised manuscript and your response to the reviewers' comments. Your revised manuscript is also likely to be sent to reviewers for further evaluation.
When you are ready to resubmit, please upload the following:
[1] A letter containing a detailed list of your responses to the review comments and a description of the changes you have made in the manuscript. Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.
[2] Two versions of the revised manuscript: one with either highlights or tracked changes denoting where the text has been changed; the other a clean version (uploaded as the manuscript file).
Important additional instructions are given below your reviewer comments.
Please prepare and submit your revised manuscript within 60 days. If you anticipate any delay, please let us know the expected resubmission date by replying to this email. Please note that revised manuscripts received after the 60-day due date may require evaluation and peer review similar to newly submitted manuscripts.
Thank you again for your submission. We hope that our editorial process has been constructive so far, and we welcome your feedback at any time. Please don't hesitate to contact us if you have any questions or comments.
Sincerely,
Hugues Berry
Associate Editor
PLOS Computational Biology
Kim Blackwell
Deputy Editor
PLOS Computational Biology
***********************
Reviewer's Responses to Questions
Reviewer #1: The manuscript addresses a relevant and interesting questions, aiming to distinguish the effects of changes in synaptic strength via increase/decrease in presynaptic vesicle release probability versus via an increase/decrease in postsynaptic receptor density. The former process impacts the short-term dynamics of vesicle release, such that long-term changes in synaptic strength are likely to have concomitant changes in the ratio of facilitation/depression. Since these changes that are overlooked in most/all models of plasticity in networks, the question is important.
Overall, the strategy used is clear and good, as the authors address a couple of roles of plasticity in a simple setting and compare models with and without the different contributions to changes in synaptic strength. I do find some of the specifics to be unclear though, in particular, I found insufficient description of the precise model when the authors attempt to isolate the facilitation/depression dynamics from the presynaptic strength-change – that is, how Fig 4 is achieved in comparison to Fig 3.
Fig 2 Caption and throughout: “The learning task” is a misnomer as there is no animal performing a task who can learn anything. Rather the paper is assessing how patterns of inputs give rise to different outputs in a circuit. There is no goal or “right answer”. Moreover, connections do not learn they change even if those changes almost certainly underlying an animal’s learning. Perhaps the paper could be reframed to particular tasks that animals must learn and the simple circuits be proposed to underly such learning. I don’t think that is necessary and just an alteration in the language used would be fine.
l. 156 “We next explored the effects of altering short-term dynamics”. The statement is confusing as the previous section already explored these effects via changes in “p”. Do you mean “altering short-term dynamics *without* any concurrent change in synaptic strength? Or do you mean while the change in synaptic strength is matched to the postsynaptic plasticity? It is really unclear to me, even when looking over the methods. Moreover, however this is done, how doe “postsynaptic alone” in this section (and Fig 4) differ from that in the prior section (and Fig 3) since “postsynaptic alone” could not impact short-term dynamics in either section. The “combined pre and post” becomes even more complicated to think about as distinct from the prior section or from “post alone” if short-term dynamics are present/missing. All rather confusing, so please clarify in the main text as well as the methods!
l. 198-214. There seems to be some confusion in the text here. If “LTD is solely postsynaptically expressed” then why in l.211 does “postsynaptic plasticity lack the ability to depress”?? And those problems resolved by inclusion of “presynaptic LTD” in the biological model if LTD is not presynaptic? I think some switching of “pre” and “post” in the text is needed and ne sure it matches Fig 5.
Fig 6. It is confusing that for the red and the blue traces in A there is essentially no difference between the “unc” and the “corr” results so there is no clear reason as to how inputs are segregated?
And how is this compatible with the classic papers of Song & Abbott showing separation of strengths for “corr” and “uncorr” inputs with postsynaptic plasticity alone? (Blue lines in your figure converge to identical values).
And if weights are kept constant in the red/blue traces then why are they not kept constant in the black trace (far right)? All rather confusing.
l. 317 “considerable impact on the learning outcome”: Maybe I missed this, but I saw no overall difference, just a change in the speed at which changes arose in the correlation paradigm – though a change in latency was a difference, but I think “small impact” with no qualitative change and it is hard to see what “learning outcome” for an animal is particularly impacted here if you wish to match these results to learning.
l.343-4 Eq.2: Why is the conductance only impacted by “q” rather than release probability “p”? Especially since in the rate model the connection strength “W” is proportional to both “p” and “q”.
l. 367-8 Eq.5: Why is the function “F” a constant rather than exponentially decaying? As written the rate of change of W depends on all prior (and future) spikes throughout all of time as they each contribute the same constant amount. This seems unphysical.
l. 376-7 “When the rate convergence rates were compared” It is unclear how you are comparing them when you are forcing them to be identical as you say you “ensure” they were “the same for all simulations”?
Minor:
Fig 3 caption: two “and”s near the end.
l.136 just “2B” I think
l. 185 “I”
l. 192-3 “relevance … was unclear”
l. 318 “these effects”
l. 377 I assume that the superscripts “f” and “i” denote “final” and “initial” here and elsewhere, but I don’t see where these are defined.
Fig 7. I suggest you make the arrows smaller so that they are separated – it is hard to tell which way they point when they run into each other, and the change in direction of the flow fields is important.
Reviewer #2: Here the authors present simulation results illustrating how the effects of spike-timing-dependent plasticity can vary depending on whether it is expressed pre- or postsynaptically. They focus on two scenarios: 1) a latency paradigm, where early inputs potentiate and late inputs depress to reduce the latency of responses of a postsynaptic neuron, and 2) a correlation paradigm, where a set of correlated inputs potentiates and sets of uncorrelated inputs depress causing the postsynaptic neuron to covary with the correlated inputs. In simulations of stochastic release, short-term plasticity, and a biologically tuned model they find that the locus of expression can alter the speed of potentiation/depression and several other outcomes for each paradigm.
This is a well-structured paper that makes a strong argument for the importance of thinking about locus of expression with plasticity. It develops interesting new ideas, and it will certainly be interesting to many computational and experimental neuroscientists.
I have no major concerns. The logic and arguments are clear. However, I did find the results section somewhat difficult to understand. Some details seemed missing or could otherwise be more clearly described here (see multiple points below).
Minor Issues:
Line 95: “First, in the reliability of transmission due to stochastic vesicle release.” Incomplete sentence?
Line 96: “keeping the standard deviation roughly the same” This seems unnecessarily vague, since you can write down the change in CV analytically. Why not just say that the standard deviation decreases (when initial p>0.5) or increases by a smaller amount than the mean
Line 110: “we model presynaptic expression as changes in short-term plasticity and compare that postsynaptic expression” Missing “to” unclear what “that” refers to.
Line 119: “Here, changes in pj were explored in terms of their impact on stochastic release” … the previous sentences make it seem like p and q will both be varied. I would also suggest spelling out what j is
Line 120: It may be helpful to have add some context here. E.g. “With the latency paradigm, STDP leads early inputs to potentiate and late inputs to depress.”
Line 129: What is “variance of the latency shift” referring to? Variance across simulations at a specific time point? The variance is difficult to see in the figure alone.
Fig 3 caption: “Time evolution of average synaptic weight among early and late presynaptic inputs (i.e., input cells that spiked in the first or the second half of the stimulus) show how post-only expression (blue) is relatively slower.” … This seems incomplete – isn’t postsynaptic expression slower for potentiation, faster for depression?
Fig 3: Suggest maybe using dashed lines to make it clear that pre follows pre+post in panel F
Fig 3H: It would be helpful to add “corr” and “unc” labels here, as well as an indicator that c>0.9 here. I would also suggest adding the value of c for 3G for context.
Fig 3I: a colorbar would be helpful here. Is this the difference in slopes? Weights? The caption seems to suggest it’s related to “which side potentiated faster”, but the text seems to focus on who “won”
Fig 3I caption: typo “and and”
Line 140: “was faster for strongly correlated input firing at certain input frequencies”. Should clarify that it’s faster for both correlated (faster to potentiate) and uncorrelated (faster to depress) inputs? Additionally, it’s not clear what the “input frequency” is here or that it matters.
Line 150: “we systematically explored the correlation-frequency space”. It’s somewhat unclear what this means. It may be useful to spell out that (if I understand correctly?) these are distinct simulations where all inputs have a specified correlation and firing rate.
Line 152: “a scenario that corresponds to fewer inputs spiking synchronously” I’m not sure what exactly this is comparing.
Line 167: “more subtle” I wasn’t completely sure what this meant. Just that the change in latency was smaller?
Fig 4 caption: “between p and q” and “between p and both”. Suggest spelling out pre or postsynaptic expression here to avoid confusion.
Line 182: “STP modulation” I guess this if the first use of the acronym. Could spell it out for clarity here.
Line 185: typo “I”
Line 187: “It thus appears that computational advantages could be tailored to a functional task at hand by recruiting pre- or postsynaptic plasticity differentially.” I’m somewhat confused by this conclusion. Unlike the results from Fig 3G, it seems like there’s minimal difference between the speeds when the dynamics are included. I’m unsure if/how 4G and 3H can be compared directly, but at low firing rates it also appears that the site of expression doesn’t matter.
Line 206: “As with the above simplistic modelling scenarios” I guess this is just the case with dynamics.
Line 207: “spike timing” Might specify that it’s “latency”, since other aspects of timing aren’t directly analyzed?
“when both pre- and postsynaptic plasticity were active, the presence of postsynaptic potentiation further reduced the latency compared to presynaptic plasticity alone.” This seems like an odd framing for modelers who traditionally only use postsynaptic plasticity. It may be useful to reframe by focusing on the fact that pre+post expression is much slower than post alone.
Line 225: It’s not clear to my why one would classify based on p rather than w.
Line 283: “It is also interesting to think about the role of presynaptic plasticity if it is not very useful in the context of usual ’coding’ frameworks.” To the extent that STP acts as a filter or mechanism for gain control wouldn’t presynaptic expression of long-term plasticity just reflect changes in the filter/gain control? This seems like a direct impact on coding.
Line 352: typo “Peduction”
Line 364: typo “implemente”
I couldn’t find how many inputs were used here or how many simulations are being averaged over in the results. I would also suggest explicitly describing the error bands/bars in the captions of figures where they appear (S.D., C.I. or S.E.M?).
Reviewer #3: In their paper Mizusaki et al address an important and timely question of how the site of expression of synaptic plasticity affects the outcome of learning. While we do know since work of Tsodyks & Markram that pre- and postsynaptic changes modify cell's responses to irregular inputs in completely different ways, detail of these differences and their computational consequences remain poorly explored. Systematic analysis of such differences would make an important contribution toward understanding specifics of learning in diverse neuronal networks.
However, while the topic of the study is interesting, and the paper could potentially make an important contribution to the field, there are several concerns that should be addressed.
In my opinion, major problems that needed to be addressed are:
1. Description of results (actually methods too) is neither precise nor complete; there are numerous unsubstantiated or unclear statements (specified below). Also, in the present version it is not clear whether and how model settings were changed in different experiments (see below); this should be clearly described.
2. Modeling presynaptic changes that do not lead to changes of synaptic dynamics looks odd. The rationale for using such a model, which contradicts almost everything we know about synaptic transmission, should be clearly articulated.
3. Results of simulations with Latency paradigm and Correlation paradigm should be presented to allow comparison. e.g. there is a lot of description of spiking response in latency paradigm, but not in correlation paradigm experiments.
Specific comments:
l.55 'depletion' – double
l.96-98 "Assuming release is binomially distributed, increasing the probability of release, p, 96
increases the mean of synaptic responses while keeping the standard deviation roughly 97
the same, ..." – this is plainly wrong.
With N=5 (Methods, l.408) and q=1; with p= 0.1, 0.2, 0.3 the mean is 0.083, 0.167, 0.25 and SD is 0.133, 0.197, 0.248. How 0.133, 0.197, 0.248 are "roughly the same"?
l. 108-112 "We start with presynaptic expression modelled as direct 108
changes in the probability of vesicle release and compare that to postsynaptic 109
expression. Subsequent to that, we model presynaptic expression as changes in 110
short-term plasticity and compare that postsynaptic expression."
Not clear what's the difference. If 'direct changes of p' means no short-term plasticity, it's a mathematical abstraction without any biological sense. At least I am not aware of a cortical synapse that does not express STP that depends on p.
Please either provide a clear rationale for using such a construct, or remove.
Fig 2 legend.
"(A) ... The learning task is thus to reduce the
latency and to shorten the duration of the postsynaptic spiking response"
" (B) ... The learning task
is thus to select for inputs that are correlated at the expense of those that are not"
'Learning task' sounds weird here. Whose task?
What is actually described, is how inputs with two fundamentally different types of temporal structure are affected by STDP.
"After learning, the postsynaptic spiking (blue raster at top) was more correlated
with the correlated inputs (pink histograms) than with the uncorrelated inputs (red
histograms), indicating that the former drove postsynaptic activity." Same was true before learning... It's kind of commonplace that more correlated inputs have stronger impact on postsynaptic firing, with or without plasticity.
Also, from this figure and its description it's not clear how two paradigms will be compared; How to compare a decrease of spike number and latency on the one hand, with a change of the correlation between inputs and spiking on the other?
l. 126. " decrease of postsynaptic activity duration and increase of postsynaptic firing frequency" – do not fit together.
Fig. 3
3B: X-scale labels are missing. Why is it different from A, D-F?
3G: what was correlation (strongly, weakly correlated inputs)?
3B,D-F X-scale is in trials; in G,H – in s; is it possible to make them uniform? To somehow compare the rate of synaptic changes?
3I – not sure I understand what and how was going on here. All inputs were saturated but some faster, some slower? Please show examples of synaptic changes in two groups, for a 'red' and a 'blue' case.
(see also below – text describes results in that same figure in a completely different way... )
Statement on faster learning with presynaptic plasticity is not well substantiated. Even in 3F – it depends how learning is defined; if it's defined as a separation of synaptic weights in early vs late groups, then building up of the separation will be about same (at least looks like in this resolution). It also might depend on temporal difference between early and late inputs.
For the correlation paradigm, 3I clearly shows that it's a matter of input parameters.
Which scenario is more 'typical' depends on which combination(s) of input frequency and correlation are more typical for actual neuronal activity.
l. 130-133. "This can be framed as a consequence of 130
potentiation requiring glutamate release [50], so that in a more reliable synapse, with 131
a high p value, there is a greater propensity for potentiation. Conversely, depression 132
was slower with presynaptically expressed plasticity, again because lowered probability 133
of release effectively also led to less plasticity"
Not clear what glutamate release has to do with plasticity in the model. As far as I understand, no mechanisms specific to glutamate release or glutamate receptors were implemented.
l. 143-145. "...in a scenario in 143
which there is competition due to e.g. limited resources, inputs with presynaptic 144
plasticity would be expected to overcome inputs with postsynaptic plasticity"
What if presynaptic resources are limited? e.g. Small ready-to-release pool or slow replenishment of vesicles?
l. 148 " Because of normalization, these two input populations competed, so 148
that one potentiated at the expense of the oher, which depressed."
First, what normalization? Methods do not say a word about it.
Second, how to reconcile "one population potentiated one depressed" with "which side potentiated faster" (Legend Fig 3I)? Slower potentiation is quite different from a depression.
l. 154 and further – "Presynaptic expression modelled as changes in short-term 154
plasticity".
Does this mean that 'presynaptic' model described so far did not include short-term plasticity? If yes, what is rationale for using such a model?
Fig. 4.
4E – what determined different steady-state of potentiated synapses in models with presynaptic (red) compared to postsynaptic or mixed (blue, black) expression?
4A-E, latency paradigm – would the rate of change (and the difference between models) depend on specifics of input patterns?
l. 176. "... synaptic efficacy was still potentiated faster and depressed 176
slower in the presynaptic case (Fig. 4E). " Not sure this is correct description. Red/early do potentiate more in the end, but rate of potentiation is about same as in black or blue, esp in the beginning.
Depression of 'late' looks pretty much the same in all three cases.
l. 185 In the simulations...
l. 198. "... in this model, LTP is expressed both pre- and postsynaptically, 198
whereas LTD is solely postsynaptically expressed."
l. 211. " postsynaptic plasticity in the tuned 211
model lacked the capacity to depress."
How the above two statements fit together?
and next, l. 214 "the inclusion of presynaptic LTD"...
Please explain clearly which model included what; and if 'biologically tuned' model does not include presynaptic LTD, why is it still included here?
l. 218. "groups of correlated and uncorrelated inputs 218
clustered (Fig. 6A) without the need for added competition through weight 219
normalization" .
Does this mean that simulations were run with and without weight normalization? Please clarify in the methods, how normalization was implemented, and how it was turned on and off. In Results and Figures, please clearly indicate which results were obtained with model(s) with normalization, and which in models without normalization.
l. 223-225 (and Fig 6C). Unclear, please explain.
l. 226. "The presynaptic frequency range for optimal 226
separation was between 50 and 80 Hz (Fig. 6C). At the other end of the range..."
What is 'the other end' of 50-80Hz range ???
l. 234. to to
l. 235,236 "In combination with postsynaptic plasticity, presynaptic plasticity provided a degree of output control..." - not clear, please explain.
Fig 5 Legend does not correspond to panel lettering.
l. 248 that that
l. 262. "Presynaptic plasticity expressed as the regulation of release 262
probability alone did not result in any differences over average postsynaptic activity 263
measurements compared to postsynaptic expression."
Fig 3 or related text do not show any data on average postsynaptic activity in correlation paradigm.
Methods.
Please define ALL parameters and terms in equations.
l.375-386. synaptic weight changes.
May be I am wrong (please correct/explain then), but unless you define qi= (0,1], qmax=Pmax =sqrt(qi) could be nonsense (say with qi=4).
l. 384. "The largest possible change for P or q separately was tot = 1 − qi."
This implies that q close to 1 limits changes in P; Why P can't change if qi=1?
l. 385-386 reads like bounds on P and q were set individually for different simulations; if so, please state this clearly and provide values of these limits for each simulation.
l. 405. "Presynaptic control of the probability of release per stimulus was implemented either 405
as a Markovian process or as short-term plasticity."
STP can be implemented using a Markovian process too, also at 5 release sites with initially uniform P. I am not suggesting the authors to do so; but if this should mean that no STP was implemented, it should be clearly stated.
**********
The
Reviewer #1: None
Reviewer #2:
Reviewer #3: Yes
**********
PLOS authors have the option to publish the peer review history of their article (
If you choose “no”, your identity will remain anonymous but your review may still be made public.
Reviewer #1: No
Reviewer #2: No
Reviewer #3: No
While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool,
Please note that, as a condition of publication, PLOS' data policy requires that you make available all data used to draw the conclusions outlined in your manuscript. Data must be deposited in an appropriate repository, included within the body of the manuscript, or uploaded as supporting information. This includes all numerical values that were used to generate graphs, histograms etc.. For an example in PLOS Biology see here:
To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at
Submitted filename:
Dear Dr Sjöström,
Thank you very much for submitting your manuscript "Pre- and postsynaptically expressed spike-timing-dependent plasticity contribute differentially to neuronal learning" for consideration at PLOS Computational Biology. As with all papers reviewed by the journal, your manuscript was reviewed by members of the editorial board and by several independent reviewers. The reviewers appreciated the attention to an important topic. Based on the reviews, we are likely to accept this manuscript for publication, providing that you modify the manuscript according to the review recommendations.
Please take care to take into account the suggestions of reviewer#3 regarding careful proofreading of the text, in particular regarding its logical flow and consistency when referring to the literature.
Please prepare and submit your revised manuscript within 30 days. If you anticipate any delay, please let us know the expected resubmission date by replying to this email.
When you are ready to resubmit, please upload the following:
[1] A letter containing a detailed list of your responses to all review comments, and a description of the changes you have made in the manuscript. Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out
[2] Two versions of the revised manuscript: one with either highlights or tracked changes denoting where the text has been changed; the other a clean version (uploaded as the manuscript file).
Important additional instructions are given below your reviewer comments.
Thank you again for your submission to our journal. We hope that our editorial process has been constructive so far, and we welcome your feedback at any time. Please don't hesitate to contact us if you have any questions or comments.
Sincerely,
Hugues Berry
Associate Editor
PLOS Computational Biology
Kim Blackwell
Deputy Editor
PLOS Computational Biology
***********************
A link appears below if there are any accompanying review attachments. If you believe any reviews to be missing, please contact
[LINK]
Reviewer's Responses to Questions
Reviewer #1: The paper is much clarified now.
Reviewer #2: The authors have addressed all of my concerns.
Reviewer #3: In the revised paper the authors addressed many of the concerns from previous comments.
However, I think a number of concerns remains not fully addressed.
Some parts of the text are still confusing (or have ambiguous meaning, implying what the authors probably did not mean to say), or plainly not correct.
Just few examples.
Introduction (which is by the way lengthy and lacks clear logics)
"In the 1990s, this led to a heated debate on the precise locus of expression of LTP, with 12
some arguing for postsynaptic expression, whereas others were in favour of a 13
presynaptic locus of LTP [8](Fig. 1A). Beginning in the early 2000’s, this controversy 14
was gradually resolved by the realisation that plasticity depends critically on several 15
factors, notably animal age, induction protocol, and precise brain region [9–11]. Indeed, 16
this resolution has now been developed to the point that it is currently widely accepted 17
that specific interneuron types have dramatically different forms of long-term 18
plasticity [12, 13], meaning that long-term plasticity in fact depends on the particular 19
synapse type [14]."
In the above, 1st sentence is about locus of expression;
2nd says controversy resolved, but talks about plasticity in general, without mentioning induction or expression or other features
3rd – all of a sudden talks about interneurons, while most of the debate on expression locus was about excitatory transmission to pyramidal neurons, mostly in the hippocampus...
Also, why interneurons where some peculiar forms of plasticity were reported in 2007 and later, and not a really classical paper from 1990 on pre and postsynaptic forms of plasticity in CA3 neurons (Zalutsky, Nicoll 1990), or clear demonstration of presynaptic mechanisms in neocortical neurons (Markram, Tsodyks 1995)?
I mean, the intro makes an impression that authors care about history of the problem.
"This is because presynaptically expressed plasticity leads 38
to changes in synaptic dynamics, whereas postsynaptic expression does not (Fig.1B). 39
For instance, during high-frequency bursting, as the readily releasable pool of vesicles in 40
a synaptic bouton runs out, leading to short-term depression of synaptic efficacy [29], 41
while short-term facilitation dominates at other synapse types [30]."
1st sentence – yes, OK.
2nd – this is not an elaboration of the statement in the previous sentence; also, readily releasable pool is also not infinite but is eventually depleted in synapses with PPF.
"As a corollary, it follows that presynaptic expression of 50
plasticity may change the computational properties of a given synaptic connection."
Why postsynaptic increase (or decrease) is NOT a change of computational property of a synapse? Yes, it does not change the filtering properties as presynaptic changes would, but it will change the probability of that input inducing a postsynaptic spike (well, on background of other synaptic activity) – why is this NOT a computational property of a synapse?
"In 51
this case, increasing the probability of release by the induction of LTP"
This implies that p can change only one way – increase, which is not true.
"Experimentally, it is long known that the induction of neocortical long-term 56
plasticity may alter short-term depression [16, 41]."
Can it also change short-term facilitation? Yes, many neocortical synapses show PPD, but some do show PPF (and in the hippocampus PPF is typical for many synapses).
I mean, the paper is not specifically about depressing, nor about specifically neocortical synapses, right?
(Discussion).
"Presynaptically expressed plasticity alone was not ideally suited for rate coding, 288
because it did not impact the average summed input effectively. "
This would be kind of true ONLY for the case of exclusively-bursting activity; It is definitely NOT true for the case of low-frequency spikes; also, in case of irregular firing changing p would change steady-state amplitude.
Also, plasticity and coding are two different 'dimensions'; plasticity can change coding, or mediate encoding of learned information, but plasticity is NOT coding;
"As a consequence, 289
postsynaptic firing frequency remained relatively unchanged after presynaptically 290
expressed plasticity (e.g., Figs. 4D, 5D, 6B)."
4D: about 15-20% increase of frequency, and about 40% decrease of burst duration; so kind of yes, total number of spikes in all bursts changed little.
This does not say much about firing frequency, calculated using all spikes.
5D: about 50% decrease of frequency in bursts, and about 60-80% decrease in burst duration;
-> about 5 – 10 fold (!) decrease of spikes in bursts. Unless there is a compensatory increase of single spike (I did not find such data in the Results), this does not look to me as "relatively unchanged".
6B – yes.
Thus, in the three groups of results, one example supports the claim, one does not, and one may be...
These are just some examples, there are many more in the text.
Bottom line: I would strongly recommend the senior authors to carefully read/edit the text, paying attention to preserving logical connections between sentences, overall consistency when referring to results of prior experimental and theoretical work, and consistency of interpretation of and conclusions from their own results.
**********
The
Reviewer #1: Yes
Reviewer #2: Yes
Reviewer #3: Yes
**********
PLOS authors have the option to publish the peer review history of their article (
If you choose “no”, your identity will remain anonymous but your review may still be made public.
Reviewer #1:
Reviewer #2:
Reviewer #3:
Figure Files:
While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool,
Data Requirements:
Please note that, as a condition of publication, PLOS' data policy requires that you make available all data used to draw the conclusions outlined in your manuscript. Data must be deposited in an appropriate repository, included within the body of the manuscript, or uploaded as supporting information. This includes all numerical values that were used to generate graphs, histograms etc.. For an example in PLOS Biology see here:
Reproducibility:
To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at
References:
Review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript.
Submitted filename:
Dear Dr Sjöström,
We are pleased to inform you that your manuscript 'Pre- and postsynaptically expressed spike-timing-dependent plasticity contribute differentially to neuronal learning' has been provisionally accepted for publication in PLOS Computational Biology.
Before your manuscript can be formally accepted you will need to complete some formatting changes, which you will receive in a follow up email. A member of our team will be in touch with a set of requests.
Please note that your manuscript will not be scheduled for publication until you have made the required changes, so a swift response is appreciated.
IMPORTANT: The editorial review process is now complete. PLOS will only permit corrections to spelling, formatting or significant scientific errors from this point onwards. Requests for major changes, or any which affect the scientific understanding of your work, will cause delays to the publication date of your manuscript.
Should you, your institution's press office or the journal office choose to press release your paper, you will automatically be opted out of early publication. We ask that you notify us now if you or your institution is planning to press release the article. All press must be co-ordinated with PLOS.
Thank you again for supporting Open Access publishing; we are looking forward to publishing your work in PLOS Computational Biology.
Best regards,
Hugues Berry
Associate Editor
PLOS Computational Biology
Kim Blackwell
Deputy Editor
PLOS Computational Biology
***********************************************************
Reviewer's Responses to Questions
Reviewer #3: all my concerns addressed. no further comments.
**********
The
Reviewer #3: None
**********
PLOS authors have the option to publish the peer review history of their article (
If you choose “no”, your identity will remain anonymous but your review may still be made public.
Reviewer #3:
PCOMPBIOL-D-21-01572R2
Pre- and postsynaptically expressed spike-timing-dependent plasticity contribute differentially to neuronal learning
Dear Dr Sjöström,
I am pleased to inform you that your manuscript has been formally accepted for publication in PLOS Computational Biology. Your manuscript is now with our production department and you will be notified of the publication date in due course.
The corresponding author will soon be receiving a typeset proof for review, to ensure errors have not been introduced during production. Please review the PDF proof of your manuscript carefully, as this is the last chance to correct any errors. Please note that major changes, or those which affect the scientific understanding of the work, will likely cause delays to the publication date of your manuscript.
Soon after your final files are uploaded, unless you have opted out, the early version of your manuscript will be published online. The date of the early version will be your article's publication date. The final article will be published to the same URL, and all versions of the paper will be accessible to readers.
Thank you again for supporting PLOS Computational Biology and open-access publishing. We are looking forward to publishing your work!
With kind regards,
Anita Estes
PLOS Computational Biology | Carlyle House, Carlyle Road, Cambridge CB4 3DN | United Kingdom