All mammals studied exhibit some form of sleep, yet the adaptive value of sleeping remains obscure 1. When viewed from this evolutionary perspective, the time spent in this state of reduced environmental awareness and behavioural quiescence is likely to be costly in terms of predation risk 234, competition for resources 245, and reproductive opportunities 6. To outweigh these costs, the benefits of sleep must be substantial.

A wide range of hypotheses have been proposed to explain why this seemingly vulnerable and unproductive state has evolved, including suggestions that sleep conserves energy when alternative activities would bring little advantage 7, is required for the consolidation of memories and learning 8, or plays a role in brain development or repair 910. However, phylogenetically controlled analyses investigating the evolution of mammalian sleep durations have produced mixed support for these explanations 345, leaving the evolutionary significance of mammalian sleep a mystery.

A further idea that has yet to be tested is that sleep evolved through the need to augment immune defences and protect against disease 111213. It is increasingly recognized that the immune system is energetically costly, as evidenced by its impairment under conditions of nutritional stress and when resources are diverted to increased growth or reproductive activity 14. Since sleep is an enforced period of inactivity and physiological down-regulation, energy that will otherwise be expended during waking activity would be available to meet the demands of the immune system 71315.

This immune theory of sleep is supported by clinical studies that provide evidence of an intimate physiological link between sleep and the immune system. For example, mammals spend more time asleep, and particularly in non-rapid eye movement (NREM) sleep, when infected with a range of parasitic agents (e.g. influenza 16, Escherichia coli 13, and Candida albicans 13). Immunomodulatory cytokines – the signalling molecules of the immune system – play a role in both the normal regulation of sleep and its modulation during an immune response 17. Direct evidence of a role for sleep in immunocompetence is provided by studies showing that rabbits that sleep more following infection have an increased chance of recovery 18, and rats that are totally sleep deprived die with a systemic invasion of bacteria 12. Finally, the human antibody responses to vaccination can be halved when subjects are deprived of sleep either before or after vaccination 1920. While the numerous other physiological changes accompanying sleep deprivation or infection make it impossible to isolate sleep per se as a causal factor in experiments 17212223, when taken together these studies highlight the potential importance of disease resistance in the evolution of sleep.

Here, we assess the evolutionary relationship between sleep and immunocompetence across a wide range of mammalian species by examining the extensive variation in sleep times (between 3 and 20 hours a day 1; Figure 1), investment in immune defences, and parasitic infections.

We first assessed the influence of sleep on the immune system. To do so we extracted data on sleeping times for different mammalian species from the published literature and matched these data where possible with white blood cell counts reported by the International Species Information System (ISIS 24; [see Additional File 1]). We use white blood cells as a proxy for immune system investment as they are central to all immune responses and are a measure of immunocompetence 2526. White blood cells originate in bone marrow and are derived from the same hematopoietic stem cells that produce red blood cells and platelets 25. As these latter cells have no direct immunological function, we use them as natural controls to test the specificity of any relationship between sleep and the immune system. If a key selective advantage of sleep is that it allows greater investment in the immune system, then species that sleep for longer should have increased numbers of immune cells in circulation, but there should be no similar relationship with control cells. After matching species values from each database we were able to analyze data for 26 mammalian species while controlling for confounding factors (body size and activity period; see the Methods for details).

As expected if sleep enhances immune defences, species that engaged in more sleep had higher numbers of white blood cells circulating in peripheral blood (coefficient = 0.00976, s.e. = 0.00171, t₂₂ = 5.71, P < 0.001; Figure 2a). Across our dataset, a 14 hour increase in sleeping times corresponded to an additional 30 million white blood cells in each millilitre of blood (a 615% increase). Crucially, no similar patterns were evident with either red blood cells or platelets (red blood cells: coefficient = -0.028, s.e. = 0.113, t₂₄ = -0.24, P > 0.8, platelets: coefficient = -0.00602, s.e. = 0.00576, t₂₁ = -1.04, P > 0.3).

We also tested predictions using phylogenetically independent contrasts to account for the lack of statistical independence in species data 27. These analyses examine evolutionary change and showed that when lineages evolved longer sleep durations, they also increased their white blood cell counts (coefficient = 0.0153, s.e. = 0.00250, t₂₃ = 6.11, P < 0.001). Again, this relationship was specific to immune cells (for red blood cells: coefficient = -0.200, s.e. = 0.0123, t₂₃ = -1.63, P > 0.11, for platelets: coefficient = -0.00463, s.e. = 0.00628, t₂₁ = -0.74, P > 0.45), leading to an increase in the ratio of immune cells to other blood cell types when species evolved longer sleeping durations (coefficient = 0.0639, s.e. = 0.0108, t₂₀ = 5.93, P < 0.001).

Total white blood cell counts are a compound measure of the abundance of different cell types, each of which fulfils specialized immunological roles 25. We therefore also investigated whether the correlated evolution of sleep and white blood cell counts was limited to specific cell types. Neutrophils constitute the largest component of the innate immune system, representing 47% of white blood cells in our sample, and are seen as a first line of defence that responds rapidly upon detection of invading pathogens 25. Analysis of independent contrasts indicated that higher numbers of neutrophils in the bloodstream have evolved in association with elevated sleep durations (bootstrapped coefficient = 0.0948, s.e. = 0.0223, n = 25, P < 0.001). Lymphocytes, which account for 44% of the white cell count in our sample and are mediators of the acquired immune response 25, were also present in greater numbers when evolutionary increases in sleeping durations occurred (bootstrapped coefficient = 0.0103, s.e. = 0.00385, n = 25, P = 0.004). Finally, eosinophils and basophils, which are relatively minor components of the white blood cell count (5% and 1%, respectively) and act primarily against macroparasites 25, showed the predicted association with sleep (eosinophils: bootstrapped coefficient = 0.0246, s.e. = 0.0125, n = 25, P < 0.035; basophils: coefficient = 0.0676, s.e. = 0.0198, t24 = 3.42, P = 0.002). Only monocytes, accounting for 5% of the total white cell count 25, failed to show a significant association with sleep durations (bootstrapped coefficient = 0.0478, s.e. = 0.0323, n = 25, P > 0.14). Thus, evolutionary increases in the abundance of four of the five immune cell types have occurred in association with longer sleep times, suggesting that increased sleep may allow a generalized heightening of immune defences.

Next, we addressed hypotheses regarding the role of the two main sleep stages (NREM and REM) in immune system investment. In clinical studies, increased duration and intensity of NREM sleep during illness are associated with an improved prognosis, but occur at the expense of REM sleep 111318. From this, it is argued that immunological benefits of sleep occur while brain function is down-regulated during the NREM phase 1122. Our results did not support this suggestion, however, as we found that evolutionary increases in both NREM and REM sleep occur in parallel with elevated white blood cell counts (NREM: coefficient = 0.0853, s.e. = 0.0249, t₁₇ = 3.42; P = 0.003; REM: bootstrapped coefficient = 0.0633, s.e. = 0.0112, n = 19, P < 0.001). Thus, evolutionary increases in sleep are associated with increased investment in the immune system regardless of its specific form.

Finally, we assessed whether a role for sleep in enhancing immune defences could translate into improved resistance against parasitic infections. We were able to match sleep times with parasitism for 12 mammalian species from the Global Mammal Parasite Database 28, which details the diversity and prevalence of microparasites (viruses, bacteria and fungi) and macroparasites (helminths, protozoa and arthropods) that infect wild populations of mammals [see Additional File 1]. If sleep is effective in protecting against infection, then species that engage in more sleep should have fewer parasites (measured as a combination of species richness and prevalence, see Methods). After correcting for differences arising from sampling effort 293031, we found this predicted relationship (coefficient = -3.554, s.e. = 0.888, t₁₀ = -4.00, P = 0.003; Figure 2b). This analysis suggests that across the 10 hour range of sleep durations present in the dataset there is a 24-fold decline in levels of parasitism. A significant negative relationship was also evident in analyses of independent contrasts (bootstrapped coefficient = -3.48, s.e. = 1.51, n = 11, P = 0.006). Thus, as species evolved longer sleep durations and enhanced their immune systems, they become less parasitized.

Our results are consistent with parasite resistance having played an important role in the evolution of sleep, and suggest therefore that sleep is of greater immunological significance than is currently recognized. It had been noted that the physiological links between sleep and immunity that have been highlighted by experimental studies (e.g. in rats 12) could arise from a negative impact of sleep deprivation on the brain, leading to an impaired coordination of immune system activity 1117. However, our findings reveal strong relationships between sleep and immune defences in the absence of sleep deprivation, and thus point to a more direct constitutive role for sleep in promoting immunocompetence.

We suggest that sleep fuels the immune system. While awake, animals must be ready to meet multiple demands on a limited energy supply, including the need to search for food, acquire mates, and provide parental care. When asleep, animals largely avoid these energetic costs, and can thus allocate resources to the immune system (sensu 1315). Unlike the energy conservation hypothesis of sleep 7, this reallocation hypothesis predicts little or no overall energy savings during sleep. This appears to be true: in humans, for example, energy savings from eight hours of sleep would be expended within one hour of waking (63 kilocalories 32). Direct estimates of the energetic cost of maintaining immunity are not currently available. However, numerous studies have suggested that these costs are large enough to generate trade-offs with key life history traits, such as growth and reproduction (e.g. 3334). Increased energy requirements when the immune system is upregulated also point toward a substantial metabolic cost associated with immune defence. Even during mild antigenic challenge, basal metabolic rate can be increased by as much as 15 to 30% 14353637. Thus, a generalised elevation of immune defences may come at considerable energetic cost.

The energetic costs of immune system maintenance and routine functioning take multiple forms (see 14). These include the relatively short lifespan and so high turnover rate of granulocytes (every 2 to 3 days 25), the cost of sustaining the hypermetabolic rate of immune cells 38, and repairing the immunopathological damage that results when cytotoxic compounds are released by immune cells responding to antigens 39. Thus, an influence of sleep need not be confined to investing in greater numbers of immune cells. Indeed, both antibody responses and natural killer cell activity are reduced following sleep deprivation 1940, showing that sleep could have a far broader influence on immunocompetence.

Our results and interpretation are consistent with experimental studies showing that animals sleep for extended periods when mounting an immune response 1316. If evolved increases in sleep allow animals to channel more energy into their immune defences and so protect against the development of acute infections, then short term increases in sleep may help provide the additional energy required for an acute phase response to an already established infection 131415. The possibility also exists that evolutionary and facultative changes in sleep share a common underlying mechanism. Short term increases in sleep appear to be triggered by immunomodulatory cytokines that are released by white blood cells during immune reactions 17. If larger numbers of white blood cells produce a greater immune response to antigenic challenge, and hence a greater release of sleep promoting cytokines, this could potentially drive evolutionary increases in sleep durations.

Our finding that both sleep phases are associated with immune investment appears to be in conflict with observations that REM sleep is reduced during acute infection (e.g. 13). It should be noted, however, that the advantage gained through evolutionary increases in normal sleep can also differ from the selective advantage of modulating sleep phases during the course of an infection. For example, REM sleep is partly characterized by a loss of thermoregulation 41, and thus it is argued that REM sleep is inhibited during an acute phases response to infection as it would prevent animals from maintaining an elevated body temperature that impedes further microbial proliferation 42. While this explanation is plausible, it cannot be applied to evolutionary changes in normal REM sleep durations and immune investment, which occur in the absence of an acute phase response. As data become available on sleep architecture during the acute phase response of different species, comparative studies may be able to assess the benefits of REM sleep suppression directly. Similarly, analysis of the increased intensity of sleep that occurs during infection, as identified through an elevation in slow wave activity 22, could reveal an additional role of the 'quality' of sleep.

It is commonly suggested that sleep may serve multiple functions (e.g. 1). While our analyses yielded no evidence that sleep influenced cell production in the other physiological systems we assessed, they do not eliminate the possibility that sleep could have an additional function(s) elsewhere in the body. In particular, it has been argued that sleep is 'primarily for the brain' 43, which is a view that has both intuitive merit and considerable experimental support (see review 44). However, from an evolutionary perspective, phylogenetically controlled analyses have yet to produce support for a key expectation of this hypothesis, namely that species with greater cognitive abilities should require more sleep. At best, a recent study has suggested that REM sleep durations may increase with the brain size of mammals (3 but see 4). Since REM sleep usually accounts for less than 20% of total sleep durations [our dataset; see Additional File 1], these comparative findings cannot explain why different mammalian species sleep for as long as they do. Instead, it may be that sleep quality is of greater importance to brain function than the overall duration of sleep per se, which is a possibility that could be assessed when sufficient data have accumulated. In the absence of these data, the presence of characteristic patterns of brain activity in mammals during NREM and REM sleep, alongside the cognitive 'black-out' that is experienced while sleeping, imply that it does perform some important function for the brain. The nature of this function remains hotly debated 1

An important implication of our findings is that ecological factors that impact sleep could indirectly affect immune defences. By sleeping regularly in 'safe' sites such as burrows or dens, for example, individuals may be better protected from predators, and thus able to sleep for longer durations 23445. Conversely, herbivorous species with large foraging requirements could have less time available for sleep than species living on an energy-rich carnivorous diet 245. Trade-offs between time invested in sleep and alternative activities may occur at key life history stages, such as during periods of reproductive competition 6 or parental care 46. Current evidence links these activities to reduced immunocompetence 47, which could be due to reductions in the time available for sleep. However, field studies should also ascertain the relationship between sleeping behaviour and an animal's exposure to parasites, which could reveal important ecological relationships between sleep and parasitic infection that have the potential to influence the analyses we report here 48.

Our results suggest mammalian species that spend more time asleep are able to increase investment in their immune systems, and thus are better protected from parasitic infection. These comparative findings are broadly consistent with experimental evidence showing close physiological links between sleep and the immune system, and point towards a major role for disease resistance in the evolution of mammalian sleep. Given the declines in human sleep durations that have occurred in recent decades 49, there is a clear need for studies that further clarify the immunological significance of sleep. In particular, studies using artificial selection regimes would allow the evolutionary relationship between immunocompetence and sleep to be quantified under controlled experimental conditions, and without potential confounds that are associated with sleep-deprivation studies 50 and the correlational analyses we present here. Similarly, further studies assessing the selective advantages of modulating sleep durations and its sub-phases during infection are clearly warranted. There is also a need to further uncover the physiological mechanisms underpinning the influence of sleep on the immune system, to examine how ecological factors might constrain sleeping durations, and to investigate whether sleep deficits increase susceptibility to disease at key stages of an animal's life history.