Sleep and Synaptic Homeostasis

We thank Dr. Heller for discussing in a recent editorial in SLEEP¹ the implications of the study by Dr. Aton and colleagues² for the synaptic homeostasis hypothesis³ (SHY). Dr. Heller provides useful context by making three general statements. We beg to differ on all three, and on the interpretation of the results of Aton et al. as well.

1. Before the article by Aton et al., there have been no studies with the potential to disprove SHY.

Far from it, unless one disqualifies the many studies whose results turned out to support rather than disprove SHY.³ In particular, the studies quoted by Heller, by both us^4–6 and others⁷ found molecular, electrophysiological, and structural evidence that net synaptic strength increases with wake and decreases with sleep. Yet, if the opposite had happened, or had there been no net changes, these very same studies would have proven SHY wrong.

2. There have been studies in both animals and humans indicating that selected synapses are strengthened during sleep.

This is a misunderstanding of both SHY and the evidence. The human and animal studies quoted by Heller^8–10 show a strengthening of the memory after sleep, not of the synapses in the specific circuits involved. SHY claims that sleep not only protects brain cells from the cellular consequences of synaptic strengthening in wake, but benefits memory in multiple ways, including acquisition, consolidation, gist extraction, integration, smart forgetting, and protection from interference.³ However, sleep does so not by additional synaptic strengthening, but through the differential downselection of synapses. The relative preservation of relevant synapses in competition with others that are downselected leads to an increase in signal-to-noise ratios and thus to strengthening of the associated memory.

3. It is “inconceivable” that the role of sleep in consolidating long-term memory does not involve synaptic strengthening in specific cells and circuits…

This is in fact quite conceivable, as explained in detail in a recent review,³ predicted by computer simulations using various learning rules for net synaptic depression during sleep,^11–13 and supported by experimental data showing an increase in signal-to-noise ratios after sleep.¹¹ For memory consolidation, all that matters is that synapses supporting new and relevant memories depress less than synapses supporting memories that are weak or less integrated.

4. …and the recent paper by Aton and colleagues has done just that.

Not really, if one analyzes the results carefully. Aton and colleagues² have taken advantage of orientation-specific response potentiation,^14,15 in which the absolute amplitude of visual evoked potentials (VEPs) increases after repeated stimulation with oriented gratings. Despite the title “Sleep promotes cortical response potentiation following visual experience,” the paper does not actually show that VEP amplitude increases after sleep. Instead, it shows that: (1) after visual training cortical responses in primary visual cortex remain elevated for several hours independent of behavioral state; and (2) the selectivity of the visual responses increases after sleep but not after spontaneous wake or sleep deprivation. Taken at face value, such findings not only do not disprove SHY, but actually provide evidence consistent with it.

To understand why, it is again crucial to distinguish clearly between absolute and relative measures. The only absolute measure provided by Aton et al. is in Figure 4 in their article. This figure shows that after training the firing rates of excitatory neurons increased for 6–8 h, independent of whether the mouse spent that time mainly asleep or awake. Because VEPs were not reported, it is impossible to say whether they increased with sleep, with wake, or just with the passage of time. Based on the firing data in Figure 4, the third scenario is the most likely one. If so, this would mean that after intensive learning during waking triggers synaptic strengthening, the process proceeds with its own course, independent of behavioral state, provided that sleep deprivation does not occur. This is exactly what was found after motor learning in adolescent mice:^16,17 an increase in the number of spines (assumed to herald synapse formation) occurred both if training was followed by several hours of sleep and after spontaneous wake (see supplementary Figure 8 in Yang et al.¹⁶), but was blocked by sleep deprivation (coincidentally, the title of the study by Yang et al., “Sleep promotes… formation of dendritic spines after learning” is also misleading, because what is shown is actually that lack of sleep interferes with spinogenesis, and a previous study in the same laboratory¹⁷ found spine formation only at night, when the mice are mainly awake). Moreover, in freely exploring adolescent mice not engaged in an explicit training paradigm, spine formation and elimination have been found to occur at all times, although elimination occurs more prominently in sleep.^5,7

Because absolute VEPs were not reported, what can be inferred based on relative changes? The two measures used by Aton et al. were the orientation index (OI) and the evoked responsiveness index (ERI), both ratios. Therefore, their increase after sleep could be caused by an increase of the absolute response to the trained orientation, to a decrease of the response to orthogonal orientations or to the blank screen, or to a combination of both. The data were not analyzed to distinguish these possibilities, and it may be impossible to reach firm conclusions because (1) visual responses were not recorded immediately after training, but only before training and 12 h later, so the crucial comparison between sleep and spontaneous wake (Figure 3) cannot be performed properly; and (2) very few neurons are available for the comparison between sleep and sleep deprivation (Figure 2). Nevertheless, it is intriguing that ERI increases after sleep, suggesting that sleep enhances the selectivity of visual responses overall, not just for the trained stimulus. If this were caused by sleep dependent synaptic potentiation, one would have to conclude that sleep after training potentiates all responses, including those to untrained stimuli—an undesirable outcome that in the long run would make neurons less selective.³ Thus, it would seem that the combined increase in OI and ERI after sleep reported by Aton et al. is actually more compatible with an overall increase in signal-to-noise ratios caused by synaptic downselection, as envisioned by SHY.

CITATION

Cirelli C, Tononi G. Sleep and synaptic homeostasis. SLEEP 2015;38(1):161–162.

DISCLOSURE STATEMENT

The authors have indicated no financial conflicts of interest.