How can insects survive without an adaptive immune system?

How can insects survive without an adaptive immune system?

We are searching data for your request:

Forums and discussions:
Manuals and reference books:
Data from registers:
Wait the end of the search in all databases.
Upon completion, a link will appear to access the found materials.

How can insects survive in a world full of pathogens that are able to defeat the innate immune system?

There is not really a definitive explanation for why, although it's important to note that many mammalian pathogens are not adapted to insects and vice-versa. Insects need to survive insect pathogens, and they have a number of defenses for this purpose. Animals and their pathogens co-evolve and exert selection pressures on each other.

Killing the host (or killing the host quickly) is often not in the pathogen's best interest. Pathogens evolve towards their own successful reproduction as much as animals do. Success may ultimately mean evolving towards a less pathogenic state that improves a pathogen's chances of being transmitted.

As animals develop new defenses, they gain an advantage (relative to others of their own species) in surviving various pathogens. Pathogens evolve to escape these defenses, and a new equilibrium is established. An animal lacking this defenses is likely to be more susceptible to dying of infection, as the new equilibrium is dependent on them having these protections.

Here is an interesting paper on this view:

As mentioned by InactionPotential, organisms and their parasites are caught in an arms race. When an organism develops a new defense, the parasites with traits that allow them to survive those defenses excel and vice versa. Parasites must balance their survival and reproduction with that of their hosts or go extinct. Over time they may become commensal/symbiotic, transfer to another species, and/or eradicate its host species.

Invertebrates' immune systems (including insects) as a whole are quite complicated. Because of the diversity of these systems we may not have anywhere near a complete understanding of them for decades. Some species intentionally upregulate their immune systems in anticipation of events that may expose them to pathogens. Social insect species, like ants, have evolved specialized immunities that take advantage of individual sacrifice, situational recognition, and colony health. Fruit flies possess radically expressive immunoglobulin receptors (IgSF), similar in some ways to the vertebrate major histocompatibility complex.

This diversity is key: the world is full of pathogens capable of defeating certain innate immune systems to some degree. However, the variety of challenges presented by even two different species usually prevents a pathogen from killing all of either species. As a pathogen becomes adapted to one species, it may lose its fitness for the other. If a pathogen is limited to one species, it has to make do with what it has or it will disappear with its host.

References and further reading

In the innate immune reponse, pathogens are recognized by a fixed repetoire of cell-surface receptors and soluble effector molecules. These receptors have evolved to recognize pathogens over hundreds of millions of years and provide a formidable defense against a wide range of pathogens. Adaptive immunity has evolved only in vertebrates and complements the innate system with a very different strategy that allows for an almost infinite number of adaptations in the production of B and T cells. That lower life forms survive without this adaptation is probably just a testimony to the efficiency of the innate system. Much of this answer must be credited to Peter Parham's text,The Immune System.

Insects have generally shorter life spans than many vertebrates, and since they are smaller they also have a lesser amounted of invested energy in the generation of individual members of the species. A large number of animals have nervous systems and are capable of movement, and can thus actively avoid environments that might be harmful to them and have predators that are either orders of magnitude greater or smaller in size than the organism in question.

If they have variation among individuals for self recognition and complement, then a greater ability to reproduce more different individuals quickly with a lesser amount of food may help fill the relative deficit in disease resistance from not having adaptive immunity.

For vertebrates that is generally not an option due to the longer life cycles of each individual organism in comparison with either insects or bacteria or viruses.

Only vertebrates have an additional and more sophisticated system of defense mechanisms, called adaptive immunity, that can recognize and destroy specific substances. The defensive reaction of the adaptive immune system is called the immune response. Any substance capable of generating such a response is called an antigen, or immunogen. Antigens are not the foreign microorganisms and tissues themselves they are substances --- such as toxins or enzymes --- in the microorganisms or tissues that the immune system considers foreign. Immune responses are normally directed against the antigen that provoked them and are said to be antigen-specific. Specificity is one of the two properties that distinguish adaptive immunity from innate immunity. The other is called immunologic memory. Immunologic memory is the ability of the adaptive immune system to mount a stronger and more effective immune response against an antigen after its first encounter with that antigen, leaving the organism better able to resist it in the future.

Adaptive immunity works with innate immunity to provide vertebrates with a heightened resistance to microorganisms, parasites, and other intruders that could harm them. However, adaptive immunity is also responsible for allergic reactions and for the rejection of transplanted tissue, which it may mistake for a harmful foreign invader.


An allergy is a disorder in which the immune system makes an inflammatory response to a harmless antigen. It occurs when the immune system is hypersensitive to an antigen in the environment that causes little or no response in most people. Allergies are strongly familial: allergic parents are more likely to have allergic children and those children&rsquos allergies are likely to be more severe. This is evidence that there is a heritable tendency to develop allergies. Allergies are more common in children than adults because many children outgrow their allergies by adulthood.


Any antigen that causes an allergy is called an allergen. Common allergens are plant pollens, dust mites, mold, specific foods (such as peanuts or shellfish), insect stings, and certain common medications (such as aspirin and penicillin). Allergens may be inhaled or ingested, or they may come into contact with the skin or eyes. Symptoms vary depending on the type of exposure and the severity of the immune system response. Two common causes of allergies are ragweed and poison ivy. Inhaling ragweed pollen may cause symptoms of allergic rhinitis, such as sneezing and red itchy eyes. Skin contact with oils in poison ivy may cause an itchy rash. This type of allergy is called contact dermatitis.

Prevalence of Allergies

There has been a significant increase in the prevalence of allergies over the past several decades, especially in the rich nations of the world, where allergies are now very common disorders. In the developed countries, about 20 percent of people have or have had hay fever, another 20 percent have had contact dermatitis, and about 6 percent have food allergies. In the poorer nations of the world, on the other hand, allergies of all types are much less common.

One explanation for the rise in allergies in the developed world is called the hygiene hypothesis. According to this hypothesis, people in developed countries live in relatively sterile environments because of hygienic practices and sanitation systems. As a result, people in these countries are exposed to fewer pathogens than their immune system evolved to cope with. To compensate, their immune system &ldquokeeps busy&rdquo by attacking harmless antigens in allergic responses.

How Allergies Occur

Figure (PageIndex<3>): This diagram shows how the adaptive immune system is activated by an otherwise harmless antigen on ragweed pollen, responding to the allergen as though it was a pathogen.

The diagram in Figure (PageIndex<3>) shows how an allergic reaction occurs. At the first exposure to an allergen, B cells are activated to form plasma cells that produce large amounts of antibodies to the allergen. These antibodies attach to leukocytes called mast cells. Subsequently, every time the person encounters the allergen again, the mast cells are already primed and ready to deal with it. The primed mast cells immediately release cytokines and histamines, which in turn cause inflammation and recruitment of leukocytes, among other responses. These responses are responsible for the signs and symptoms of allergies.

Treating Allergies

The symptoms of allergies can range from mild to life-threatening. Mild allergy symptoms are often treated with antihistamines. These are drugs that reduce or eliminate the effects of the histamines that produce allergy symptoms.

Figure (PageIndex<4>): Anaphylaxis is a rapid, systemic reaction to allergens that may lead to life-threatening symptoms.

Treating Anaphylaxis

The most severe allergic reaction is a systemic reaction called anaphylaxis. This is a life-threatening response caused by a massive release of histamines. Many of the signs and symptoms of anaphylaxis are shown in Figure (PageIndex<4>). Some of them include a drop in blood pressure, changes in heart rate, shortness of breath, and swelling of the tongue and throat, which may threaten the patient with suffocation unless emergency treatment is given. People who have had anaphylactic reactions may carry an epinephrine autoinjector (widely known by its brand name EpiPen®) so they can inject themselves with epinephrine if they start to experience an anaphylactic response. The epinephrine helps to control the immune reaction until medical care can be provided. Epinephrine constricts blood vessels to increase blood pressure, relaxes smooth muscles in the lungs to reduce wheezing and improve breathing, modulates heart rate, and works to reduce swelling that may otherwise block the airways.

Immunotherapy for Allergies

Another way to treat allergies is called immunotherapy, commonly called &ldquoallergy shots.&rdquo This approach may actually cure specific allergies, at least for several years if not lifelong. It may be particularly beneficial for allergens such as pollen that are difficult or impossible to avoid. First, however, patients must be tested to identify the specific allergens that are causing their allergies. As shown in Figure (PageIndex<5>), this may involve scratching tiny amounts of common allergens into the skin and then observing whether there is a localized reaction to any of them. Each allergen is applied in a different numbered location on the skin so if there is a reaction, such as redness or swelling, the responsible allergens can be identified. Then, through periodic injections (usually weekly or monthly), patients are gradually exposed to larger and larger amounts of the allergens. Over time, generally from months to years, the immune system becomes desensitized to the allergens. This method of treating allergies is often effective for allergies to pollen or insect stings, but its usefulness for allergies to food is unclear.

Figure (PageIndex<5>): Skin testing for common allergens is one way to identify the cause(s) of a patient&rsquos allergic symptoms.

How one pathogen evades the immune system

T. brucei causes sleeping sickness. Credit: Meyer-Natus/Siegel

An LMU research team led by Nicolai Siegel has uncovered a mechanism that enables the parasite that causes sleeping sickness in humans to escape the attention of the immune system. The finding may also be relevant to other infectious diseases.

Our immune system is never idle. Their task is to detect and eliminate invasive pathogens, and they have no time to lose. The adaptive immune system identifies infectious organisms by recognizing foreign proteins on the surfaces of bacteria, viruses and unicellular protozoans. The interaction of these antigens with immune cells triggers a series of downstream events, which in most cases leads to the elimination of the pathogen.

But pathogenic organisms have developed strategies that enable them to escape detection by the immune system, and the strategies employed by remotely related organisms are often remarkably similar to each other. One way of confusing the immune system is to increase the structural heterogeneity of the antigens it encounters. In bacteria,pathogenic yeast and parasites this can be done by randomly activating different members of gene families, which code for non-identical versions of the proteins expressed on their surfaces. This strategy essentially allows the infectious agent to duck under the immune system's radar. By doing so, it significantly increases the likelihood that the invader will survive to establish an infection, and has a better chance to be transmitted to new hosts. If pathogens alter their surface proteins rarely—or too often—the white blood cells that are responsible for recognizing them have a much easier task. Nicolai Siegel (Professor of Molecular Parasitology at LMU) and his group, in collaboration with colleagues at the University of Dundee, have now elucidated an important step in the mechanism that controls surface-antigen variation.

The experimental model: Trypanosomes

While Siegel's team is part of the department of Experimental Parasitology and affiliated with the Faculty of Veterinary Medicine at LMU, it makes use of laboratories located in the Physiological Chemistry section of the Biomedical Center (Faculty of Medicine). "This arrangement greatly facilitates scientific discussion and interdisciplinary exchanges," he says.

His team works with the unicellular organism Trypanosoma brucei. There are several reasons for this. T. brucei causes sleeping sickness. It is transmitted by the tsetse fly, and it presents a threat to millions of people in 36 African countries south of the Sahara. From a scientific point of view, however, this species has become a model system for the study of antigen variation in pathogens, and has therefore been widely studied.

The genome of T. brucei includes more than 2000 genes that code for variant forms of the major protein expressed on its surface. In each individual cell, only one of these genes is activated—and it directs the production of a single surface-protein variant. "The pathogen must therefore ensure that only one of these genes ¬– not a few, and certainly not all—of the genes for surface proteins are expressed at any given time," Siegel explains. "We have now identified the mechanism that guarantees that the product of only one of these genes is expressed."

Notably, T. brucei does not possess complex arrays of regulatory genomic sequences—such as enhancers—which are involved in determining the set of genes that are transcribed from the genomic DNA into messenger RNAs (mRNAs) at any given moment. These mRNAs subsequently direct the synthesis of the corresponding proteins. "The control mechanism that we have discovered appears to achieve the required selectivity by differentially regulating mRNA maturation," Siegel says. This in turn is accomplished by chemically modifying specific mRNAs, which prevents them from being rapidly destroyed.

The authors of the new study have identified a three-dimensional structure in the nucleus of T. brucei that serves as a separate compartment, in which the mRNA molecules that encode the cell's single surface protein variant are modified. As a result, they avoid rapid destruction, and therefore survive long enough to produce the protein in the required amount. Conversely, when one of the proteins that contribute to the assembly of this compartment was inactivated, several different surface antigens were synthesized at the same time.

"So we now know why only one surface antigen is successfully expressed," says Siegel. Moreover, these new results have implications that transcend their importance for basic research. "If we could control the process that leads to the switching of surface antigens, it might be possible to inhibit it," he muses. And indeed, in the medium term, he sees in this possibility a new approach to the elimination—by the body's immune system—of pathogens that depend on this form of antigenic variation.

This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.


Boomsma J. J., Schmid-Hempel P.& Hughes W. O. H.

. 2005 Life histories and parasite pressure across the major groups of social insects . Insect evolutionary ecology (eds

, Fellowes M. D. E., Holloway G. J.& Rolff J.

), pp. 139–175. Wallingford, UK : CABI Publishing . Google Scholar

. 1998 Principles of insect pathology. Norwell, MA: Kluwer . Crossref, Google Scholar

. 2009 Review. Analogies in the evolution of individual and social immunity . Phil. Trans. R. Soc. B 364, 129–142. (doi:10.1098/rstb.2008.0166). Link, Google Scholar

Cremer S., Armitage S. A. O.& Schmid-Hempel P.

Hughes W. O. H., Eilenberg J.& Boomsma J. J.

. 2002 Trade-offs in group living: transmission and disease resistance in leaf-cutting ants . Proc. R. Soc. Lond. B 269, 1811–1819. (doi:10.1098/rspb.2002.2113). Link, ISI, Google Scholar

Hughes W. O. H., Petersen K., Ugelvig L., Pedersen D., Thomsen L., Poulsen M.& Boomsma J. J.

. 2004 a Density-dependence and within-host competition in a semelparous parasite of leaf-cutting ants . BMC Evol. Biol. 4, 45. (doi:10.1186/1471-2148-4-45). Crossref, PubMed, Google Scholar

Hughes W. O. H., Thomsen L., Eilenberg J.& Boomsma J. J.

. 2004 b Diversity of entomopathogenic fungi near leaf-cutting ant nests in a neotropical forest, with particular reference to Metarhizium anisopliae var. anisopliae. J. Invertebr. Pathol. 85, 46–53. (doi:10.1016/j.jip.2003.12.005). Crossref, PubMed, Google Scholar

. 2007 Social learning in insects—from miniature brains to consensus building . Curr. Biol. 17, R703–R713. (doi:10.1016/j.cub.2007.06.012). Crossref, PubMed, ISI, Google Scholar

Rosengaus R. B., Maxmen A. B., Coates L. E.& Traniello J. F. A.

. 1998 Disease resistance: a benefit of sociality in the dampwood termite Zootermopsis angusticollis (Isoptera: Termopsidae) . Behav. Ecol. Sociobiol. 44, 125–134. (doi:10.1007/s002650050523). Crossref, ISI, Google Scholar

. 1998 Parasites in social insects. Princeton , NJ: Princeton University Press . Google Scholar

. 2003 Statistical significance for genomewide studies . Proc. Natl Acad. Sci. USA 100, 9440–9445. (doi:10.1073/pnas.1530509100). Crossref, PubMed, ISI, Google Scholar

Traniello J. F. A., Rosengaus R. B.& Savoie K.

. 2002 The development of immunity in a social insect: evidence for the group facilitation of disease resistance . Proc. Natl Acad. Sci. USA 99, 6838–6842. (doi:10.1073/pnas.102176599). Crossref, PubMed, ISI, Google Scholar


Mosquitoes are susceptible to infections during all stages of their life cycle, and in response have evolved a robust innate immune system [1]. Mosquitos acquire pathogens via ingestion and through breaches in the outer cuticle, such as during injury. For example, larvae acquire Bacillus thuringiensis while feeding in their aquatic environment, and adults acquire Plasmodium sp. while imbibing infected vertebrate blood [10, 98]. Because the adult stage is responsible for the transmission of disease-causing pathogens such as malaria [9, 10], the vast majority of research on mosquito immunology has focused on that life stage, and in particular, on the adult females that feed on blood. In contrast, research on the immune system of the larvae of other insects, such as Manduca sexta, Bombyx mori and Drosophila melanogaster has proven invaluable for uncovering fundamental aspects of the cellular and humoral immune response of insects, and animals in general [99,100,101]. Nevertheless, lacking in Insecta are comparative studies that examine differences in immunity across the different life stages of a given species. As noted by Fellous & Lazzaro, “Almost all studies of the immune system of animals with metamorphosis have focused on either larval or on adult immunity, implicitly assuming that these traits are either perfectly correlated or evolutionarily independent” [41]. Because mosquito larvae live in aquatic environments rife with microorganisms and survival through the larval stages is required to reach sexual maturity, at the onset of this study we hypothesized that larvae have evolved more proficient means of neutralizing infections than adults. To compare larval and adult immunity in A. gambiae, we measured a broad range of immune parameters that were previously unexamined in larvae, and show that immune activity is strongest in larvae and wanes in the adult life stage (Table 1). Furthermore, because immunity phenotypes differ between life stages that are separated by metamorphosis, these findings suggest that adaptive decoupling, or the independent evolution of larval and adult traits made possible by metamorphosis [40], has occurred in A. gambiae.

Evidence for adaptive decoupling has been found in a broad range of animals with complex life cycles, including amphibians, fish, marine invertebrates and insects [40, 41, 102,103,104,105,106,107,108]. Several lines of evidence suggest that adaptive decoupling of immune responses has also occurred in the mosquito lineage. First, we have shown that the strength and composition of the mosquito immune response differs between immature larvae, newly-emerged adults and reproductively active older adults. Secondly, we have previously shown [11, 48, 55], and recapitulate here, that the mosquito circulatory and immune systems display stage-specific functional integration, enabling functionally analogous, yet anatomically disparate, immune strategies in larvae and adults. Thirdly, mosquitoes have complex life cycles with ecologically distinct larval and adult stages that are subject to different selection pressures, due to differences in pathogen exposure and reproductive status. Thus, we expect that a decoupling of key traits between these life stages, which are separated by metamorphosis, is advantageous to both stages in responding independently to differing selection pressures and in the performance of different fitness tasks (for example, food consumption and growth in larvae and reproduction in adults). Taken together, these arguments suggest that adaptive decoupling has occurred in mosquitoes, enabling the independent evolution of key larval and adult immune traits.

Our data show that larvae kill bacteria at a higher rate, possess more circulating hemocytes, have higher lytic and melanization activity in their hemolymph and have higher infection-induced expression of immunity genes than adults, thus strongly supporting our hypothesis that mosquito larvae invest more into immunity than adults. In our view this is not surprising larvae inhabit bacteria-rich environments and display higher gut microbial diversity than adults [44, 109], suggesting that mosquitoes are exposed to a higher density and diversity of bacteria as larvae than as adults, and require more robust immune responses as larvae, as has been hypothesized previously [110]. Thus, given the vastly different ecologies between larvae and adults, coupled with the fact that evolutionary pressures are strongest on younger individuals that have yet to reach their reproductive potential [34, 35], we hypothesize that the stronger selection pressures at work in the larval stage, compared to the adult stage, have over time caused larvae to invest more greatly in immunity than adults.

Our findings on immune proficiency in larvae and adults is also indicative of immune senescence, as immunity in larvae is stronger than in adults, and immunity continues to weaken after metamorphosis, with 5-day-old adults mounting weaker immune responses than adults that were 4 days younger. This finding is consistent with well-attested evolutionary theories of senescence, which state that greater selection pressures are at work in the larval stage to ensure survival to the reproductively mature adult life stage [34, 36]. According to the “Disposable Soma” theory of aging, this occurs, in part, due to immune trade-offs that would not be in play at the larval stage, particularly those related to reproduction [111]. For example, resistance to pathogens in adult mosquitoes has been associated with reproductive costs [112,113,114,115]. Furthermore, numerous studies have documented age-associated declines in immune functions in mosquitoes [13, 26, 46, 47, 110, 116,117,118,119,120], fruit flies [121,122,123], butterflies [124, 125], bees [126,127,128,129,130], and other insects [131,132,133,134,135]. However, unlike these earlier studies, which focused almost exclusively on adults, our findings show that this trend extends into the earlier larval life stage. This suggests that mosquito adults employ a “live fast, die young” life history strategy [136], whereby immunity declines and mortality increases after adults reach reproductive maturity and selection for their survival wanes [34].

Although the data presented in this study conclusively show that larvae mount a stronger immune response than adults against bacteria that are present in the hemocoel, the adult stage has evolved mechanisms to boost the immune system when the possibility of infection is anticipated. Mainly, because many adult infections begin with an infectious blood meal [6, 10, 137], adult mosquitoes boost their immune reserves upon ingestion of blood. For example, in both mosquito subfamilies, acquisition of a blood meal induces the replication of hemocytes even in the absence of infection [138,139,140]. Whether this has a positive impact on the ability of anopheline mosquitoes to respond to an infection in the hemocoel remains unknown, but for A. aegypti (a culicine mosquito), blood feeding improves the ability to combat a low dose E. coli infection but is detrimental when facing a high dose E. coli infection [140].

The age-dependent decrease in immune proficiency correlates with age-related changes in the number of circulating hemocytes present in the hemocoel. We found that mosquito larvae contain more circulating hemocytes than adult mosquitoes, and that the number of larval hemocytes increases in response to infection. This is consistent with earlier studies conducted in Anopheles gambiae and Aedes aegypti adults that showed that immune stimulation or blood-feeding increases the number of circulating hemocytes [26, 45, 86, 138,139,140,141,142], and that the number of circulating hemocytes declines with adult age [26, 46, 47, 116]. Hemocytes are the first cellular responders to infection, phagocytosing pathogens in larvae and adults within minutes of infection [11, 48, 55, 57, 143, 144], and producing factors that limit both bacteria and Plasmodium infection [66, 86, 137, 145]. Thus, developmental changes in hemocyte profiles in part explain the higher bacteria killing efficiency of larvae relative to adults.

Although sessile hemocytes were abundant in both larvae and adults, the spatial arrangement of sessile hemocytes differed as adults aged, with newly-emerged adults displaying a pronounced segmental arrangement of hemocytes, which is more akin to what is observed in larvae than what is observed in 5-day-old adults [11, 46]. Furthermore, whereas hemocytes in adults aggregate at the periostial regions of the heart, hemocytes in larvae aggregate in the tracheal tufts of the 8th abdominal segment, where the sole incurrent openings into the heart is located. These differences are due to the stage-specific functional integration of the immune and circulatory systems, as hemocytes aggregate in the areas of highest hemolymph flow [11, 46, 48, 55]. These differences may also be explained by stage- and age-specific changes in heart physiology [42, 146]. Overall, these data show that sessile hemocyte populations undergo developmentally-related changes in spatial configuration and infection-induced abundance.

The finding that larvae have more circulating hemocytes than adults may account for the seemingly counterintuitive finding that larvae show lower phagocytic indices and capacities than adults. This is because larvae have more hemocytes available to phagocytose pathogens than adults, and based on our finding on the lytic and melanization activity of hemolymph, they also have a greater recourse to other means of pathogen clearance. Thus, an accumulation of phagocytic events in adults may signify a greater need for this immune mechanism, or lower phagocytic turnover rates, which has also been observed in aging fruit fly adults [121]. Interestingly, these differences in phagocytosis are accompanied by changes in hemocyte morphology. Larval hemocytes often had a fibroblast-like spread morphology, which has been observed in other insect hemocytes [27, 147], as opposed to the more rounded spread morphology typical of mosquito adult hemocytes [26, 46, 48, 56, 57, 143]. These fibroblast-shaped hemocytes have the functional characteristics of granulocytes, which are phagocytic, and not the functional characteristics of oenocytoids, which are the major producers of the phenoloxidase enzymes that drive the melanization pathway [3, 56, 57].

In addition to stage-specific differences in hemocyte biology, we detected profound differences in the melanization activity of hemolymph. Melanization is commonly used in the immune response against bacteria, malaria, and filarial worms [2, 7, 59, 148], and we found that mosquito larvae display a far greater capacity for phenoloxidase-mediated humoral melanization than adults. This finding is consistent with our recent observations of rapid and extensive melanin deposits in the abdomen of larvae following a bacterial infection [11], as well as with the decline in phenoloxidase activity that occurs with adult age in Aedes aegypti [117], and across life stage and with adult age in Culex pipiens [110]. In addition to enhancing humoral immunity, larvae may increase the melanization activity of hemolymph in preparation for cuticle tanning during and immediately after ecdysis, and for the rapid melanization of wounds in the more perilous aquatic environment of larvae.

In our final set of experiments, we show that immunity gene induction is generally stronger in larvae compared to adults, with the difference being exacerbated as adults age, and that larvae and adults differentially regulate the expression of some immunity genes. The higher rate of E. coli killing in larvae is consistent with the expression of immune effector genes such as the Gram (−) binding protein gene, GNBPB4, which was expressed 470- and 710-fold higher in naïve larvae compared to naïve 1-day-old and 5-day-old adults, respectively, as well as the higher expression of the phenoloxidase gene PPO1, which was expressed 170- and 110-fold higher in larvae relative to naïve 1-day-old and 5-day-old adults, respectively. The higher rate of melanization we observed in larval hemolymph is also consistent with the higher PPO6 gene induction we observed in larvae, and supports our previous finding of higher infection-induced melanization in the abdomen of larvae when compared adults [11]. Furthermore, the higher melanization activity in the hemolymph of larvae is also consistent with the high level of dual oxidase (DUOX) gene induction detected in this life stage, as increased reactive oxygen species levels have been correlated with increased melanization [93]. In addition to the immune effector genes PPO6 and DUOX, we also found that the Jak/Stat essential pathway member AGSTAT-A and, to a lesser extent, the negative regulator of this pathway, PIAS, were induced in response to E. coli infection in larvae and not in adults, showing that the same infection results in stage-specific induction of signaling pathway genes. The upregulation of Imd pathway member CASPL1 in larvae and 1-day-old adults is consistent with this pathway’s role in combating Gram (−) bacterial infections, however the cause for Jak/Stat pathway involvement in the larval response alone is unknown, though it may have to do with its function in mosquito development [82], as this pathway is involved in Drosophila development [149]. Adults, not larvae, showed distinct upregulation of nitric oxide synthase, an important component of the adult immune response [91, 150], as well as higher upregulation of the pathogen recognition gene FREP13 and the signal modulation gene CLIPB15. These differences in immunity gene induction across life stages could result from changes in the tissues expressing these genes during metamorphosis, as has been hypothesized previously [28, 41]. More broadly, however, these differences in immune gene regulation in larvae and adults are suggestive of adaptive decoupling, which would permit the independent regulation of larval and adult immune gene expression [40, 41].

Although the data presented herein show the formidable strength of the larval immune system relative to that of adults, these data were collected following the intrahemocoelic injection of a facultative pathogen, and not following feeding of an obligate pathogen. The route of administration and choice of pathogen was necessary to ensure that an identical dose with a known pathogen was provided to all groups, and to establish a precise time of infection. While many pathogens, including bacteria and entomopathogenic fungi, invade the hemocoel directly through the cuticle, future studies focusing on infections via the gut should be conducted to complement and refine our understanding of the present work. Furthermore, although this study did not track long-term mosquito survival (the larvae begin to pupate after 24 h), the long-term consequence of larval infection in the adult stage is an important gap in our current understanding and is an ongoing topic of study in our research group. Finally, our experiments on hemocyte biology did not differentiate between granulocytes, oenocytoids, or prohemocytes due to limitations inherent to the cell staining method, but it is well established that approximately 95% of hemocytes are granulocytes [3], and our qualitative observations suggest that life stage does not impact the proportion of hemocyte types.


Defective Immune Responses Against Influenza Virus Infection in High Heat-Exposed Mice.

To assess the effects of ambient temperature in the induction of adaptive immunity to influenza virus infection, mice were kept at 4, 22, or 36 °C for 7 d before influenza virus infection. Cold or high-heat exposure of naïve mice was generally well tolerated (SI Appendix, Fig. S1A). Although cold-exposed naïve mice exhibited significant increase in their food intake compared with room temperature (RT)-exposed group (SI Appendix, Fig. S1B), high-heat exposure of naïve mice significantly reduced their food intake and body weight by 10% (SI Appendix, Fig. S1 B and C). Cold- or high heat-exposed mice were then infected i.n. with a sublethal dose (30 pfu) of A/PR8 influenza virus. After infection with influenza virus, cold- or high heat-exposed mice were kept at 4 or 36 °C, respectively, for the entire duration of the experiments. Remarkably, both frequency (Fig. 1A and SI Appendix, Fig. S2) and number (Fig. 1B) of influenza virus-specific CD8 + T cells in the lung as well as the virus-specific IgG antibody (Fig. 1C) and CD4 + T cell responses (Fig. 1D) were severely impaired in high heat-exposed mice. As a consequence, viral titer in the lung remained significantly elevated in the high heat-exposed mice at 7 d postinfection (p.i.) (SI Appendix, Fig. S3A). Further, freshly isolated mLN DCs from high heat-exposed mice infected with recombinant influenza A virus expressing the MHC-I OVA peptide SIINFEKL in the neuraminidase (NA) stalk of the A/PR8 backbone (PR8–OT-I virus), in the absence of exogenous peptide, induced no differentiation of OT-I CD8 TCR Tg T cells specific for the OVA epitope presented on H-2D b ex vivo (Fig. 1E). This defect was likely due to impairment of migration of antigen-captured lung DCs to the mLN in high heat-exposed mice, because the same mLN DCs were able to differentiate OT-I naïve CD8 T cells after exogenous OVA peptide addition (Fig. 1F). In addition, migration of antigen-captured lung DCs to the mLN was severely impaired in high heat-exposed mice (Fig. 1 G and H), in which MHC-II + CD11c + lung DCs were comparable to that seen in RT-exposed mice (SI Appendix, Fig. S4). These data suggest that migration of antigen-captured lung DCs to the mLN are impaired in high heat-exposed mice. Consequently, proliferation of adoptively transferred naïve OT-I CD8 + T cells was reduced in the mLNs of high heat-exposed mice at 5 d after infection with PR8–OT-I virus (Fig. 1 I and J).

High heat-exposed mice fail to induce adaptive immunity to influenza virus infection. Mice were kept at 4, 22, or 36 °C for 7 d before influenza virus infection (30 pfu per mouse) and throughout infection. (A) Ten days later, lymphocytes were isolated from the lung. Influenza virus-specific CD8 + T cells were then detected using the H-2D b influenza virus nucleoprotein (NP) tetramer. (B) Total numbers of influenza virus-specific CD8 + T cells in the lung are shown. (C) Serum was collected at 10 d postinfection. Influenza virus-specific serum IgG levels were measured by ELISA. (D) CD4 + T cells were isolated from spleen and restimulated with irradiated splenocytes in the presence or absence of inactivated influenza virus for 72 h, and IFN-γ production from CD4 + T cells was measured by ELISA. (E and F) Naïve OT-I CD8 + T cells (2 × 10 5 cells per well) were cocultured with CD11c + DCs (1 × 10 5 cells per well) that were isolated from the mLNs of PR8–OT-I virus-infected animals with (F) or without (E) NP or OVA peptide. Splenic DCs (1 × 10 5 cells per well) from infected animals were used as a negative control. Seventy-two hours later, IFN-γ production was measured by ELISA. (G and H) Mice were inoculated intranasally with DQ-ovalbumin (DQ-OVA). Six hours later, mice were infected with 1,000 pfu of PR8 viruses. Eighteen hours after infection, mLNs were collected. Numbers adjacent to outlined areas indicate percent DQ-OVA + CD11c + DCs (G). Percentages of DQ-OVA + CD11c + DCs are shown (H). (I) Mice were injected with CFSE-labeled OT-I CD8 + CD45.1 + T cells on the day before infection with 1,000 pfu of PR8–OT-I viruses. Five days after influenza virus infection, mLNs were excised to assess T cell proliferation by CFSE dilution. (J) Total numbers of CD45.1 + cells in the mLNs are shown. (K) The BALF was collected from influenza virus-infected animals at 2 d postinfection. IL-1β levels in BALF were determined by ELISA. The data are representative of three independent experiments (A, G, and I) or are from three independent experiments (B, F, H, JK mean ± SEM). *P < 0.05, **P < 0.01, and ***P < 0.001 n.s., not significant (one-way ANOVA and Tukey’s test).

We next examined whether influenza virus replicates in the lung of high heat-exposed mice. We found that high heat-exposed mice sustained high virus burden in the lung until 7 d p.i. (SI Appendix, Fig. S3A). In addition, the extent of infection by the influenza virus in the lung of high heat-exposed mice was similar to that of RT-exposed mice (SI Appendix, Fig. S3B), suggesting that the inability of heat-exposed mice to mount the virus-specific adaptive immune responses was not due to viral replication in the lung tissue. Migration of antigen-captured lung DCs to the mLN and induction of influenza virus-specific CD8 + T cell responses require inflammasome activation and IL-1R signaling in pulmonary DCs (15, 16). This led us to consider the possibility that high heat-exposed mice fail to stimulate inflammasome-dependent cytokine secretion following influenza virus infection. To test this possibility, we measured the levels of secreted IL-1β in the bronchoalveolar lavage fluid (BALF) of cold-, RT-, or high heat-exposed mice infected with influenza virus. Notably, high heat-exposed mice impaired secretion of mature IL-1β in the BALF (Fig. 1K) as well as mRNA expression of pro–IL-1β in the lung tissues and secretion of IFN-α, IL-6, IL-12p40, and TNF-α (SI Appendix, Fig. S5) following influenza virus infection. These data indicated that immune responses in high heat-exposed mice are impaired, and critical pathways known to initiate adaptive immune responses including inflammasome activation and lung DC migration are severely compromised in mice maintained at high-heat ambient temperature.

Effects of High-Heat Exposure of Mice in the Induction of Adaptive Immunity to Vector-Borne Pathogens.

To examine whether the high-heat exposure of mice resulted in general immunodeficiency, we immunized cold-, RT-, and high heat-exposed mice with formalin-inactivated influenza virus vaccine and aluminum adjuvant. Unlike lung infection with influenza virus (Fig. 1), immunization with influenza virus vaccine and alum led to normal antibody (SI Appendix, Fig. S6 A and B) and T cell responses (SI Appendix, Fig. S6 C and D) in high heat-exposed mice. We next examined the effects of high-heat exposure of mice in the induction of adaptive immune response to vector-borne pathogens. To this end, we infected RT- or high heat-exposed mice i.p. with Zika virus (ZIKV) or severe fever with thrombocytopenia syndrome phlebovirus (SFTSV), a tick-borne human pathogenic virus. We observed normal IgG antibodies specific for ZIKV and SFTSV in high heat-exposed mice (SI Appendix, Fig. S7 A and B). In contrast, IFN-γ–producing CD4 + T cells were considerably reduced in high heat-exposed mice following i.p. infection with ZIKV or SFTSV (SI Appendix, Fig. S7 CH). These data suggested that the effects of high-heat exposure of mice in the induction of adaptive immune responses are also impaired against other viral pathogens.

Commensal Bacteria Composition Remains Intact in High Heat-Exposed Mice.

Cold exposure of mice leads to change in the gut microbiota composition (21). In addition, intact commensal microbiota is required for adaptive immune responses to influenza virus infection (17, 18). These observations led us to consider the possibility that high-heat exposure of mice changes commensal bacteria composition, which could dampen adaptive immune responses to influenza virus infection. To test this possibility, we assessed the effects of high-heat exposure on bacterial load and composition in the cecum. Consistent with previous reports (17, 23), antibiotic treatment resulted in significant changes in the composition of commensal bacteria (SI Appendix, Fig. S8 A and B). Both frequency (SI Appendix, Fig. S8C) and number (SI Appendix, Fig. S8D) of influenza virus-specific CD8 + T cells in the lung were diminished in the Abx mice. Consequently, Abx mice enhanced susceptibility to both low (50 pfu) and high (200 pfu) doses of influenza virus infection (SI Appendix, Fig. S8 E and F). Further, rectal inoculation of LPS restored immune responses to influenza virus infection in Abx mice (SI Appendix, Fig. S8 GI).

In contrast to Abx mice, high heat-exposed mice did not change in the amount of 16S rRNA (SI Appendix, Fig. S9A) and the composition of commensal bacteria (SI Appendix, Fig. S9 B and C) present in the cecum. Consistent with a previous report (21), profiling of the microbiota composition by 16S rRNA gene sequencing, followed by principal coordinates analysis (PCoA) revealed significantly different clustering of the microbiota in the cecum of cold-exposed mice that was distinct from that of RT-exposed mice (SI Appendix, Fig. S9D). In contrast, high-heat exposure of mice did not significantly change the bacterial clusters in the cecum (SI Appendix, Fig. S9D). In addition, rectal inoculation of LPS did not restore immune responses to influenza virus infection in high heat-exposed mice (SI Appendix, Fig. S9 EG). These data indicated that rectal TLR stimulation is insufficient to restore immune responses in high heat-exposed mice and further suggested that commensal bacteria composition is unlikely to account for immune defects in high heat-exposed mice.

High-Heat Exposure-Induced Autophagy Regulates Adaptive Immune Responses to Influenza Virus Infection.

Next, we asked how high-heat exposure of mice suppresses the generation of adaptive immunity to influenza virus in the lung. High heat-exposed mice reduced their food intake and body weight by 10% (SI Appendix, Fig. S1). As a consequence, the level of light chain 3 (LC3)-II, a marker specific for autophagosome, was increased in the lung of high heat-exposed mice (SI Appendix, Fig. S10 A and B). In addition, flow cytometric analysis revealed that both CD45.2 − epithelial cells and CD11c + DCs significantly increased autophagy in the lung of high heat-exposed mice (SI Appendix, Fig. S10 CE). Autophagy and mitophagy restrict inflammasome-dependent cytokine release by regulating the amounts of pro–IL-1β and damaged mitochondria, respectively (24, 25). These observations led us to consider the possibility that high-heat exposure of mice induces autophagy in the lung, which inhibits IL-1β secretion following influenza virus infection. To determine the importance of autophagy in the induction of adaptive immune responses against influenza virus infection, we induced autophagy in vivo by rapamycin treatment or nutrient starvation by established methods (26, 27) before influenza virus infection. The level of LC3-II was increased in the lung of mice starved for 24 h or injected i.v. with rapamycin (SI Appendix, Fig. S11), without affecting viral load in the lung at 3 d p.i. (SI Appendix, Fig. S12), the amount of 16S rRNA (SI Appendix, Fig. S13A), the gut microbiota composition (SI Appendix, Fig. S13 B and C), or their clusters (SI Appendix, Fig. S9D). Notably, induction of autophagy by nutrient starvation or treatment with rapamycin impaired frequency (Fig. 2 A and B) and total number (Fig. 2 C and D) of influenza virus-specific CD8 + T cells in the lung as well as the virus-specific IgG antibody (Fig. 2 E and F). Further, secretion of mature IL-1β in the BALF (Fig. 2G) as well as mRNA expression of pro–IL-1β (Fig. 2H) were impaired in the lung tissues of food-restricted or rapamycin-treated groups compared with ad libitum-fed mice. Consequently, migration of antigen-captured lung DCs to the mLN (Fig. 2I and SI Appendix, Fig. S14) as well as proliferation of OT-I naïve CD8 T cells in the mLNs at 4 and 5 d p.i. (Fig. 2 J and K) were reduced in food-restricted or rapamycin-treated groups. Collectively, these data indicated that induction of autophagy in the lung by high-heat exposure impaired adaptive immune responses to influenza virus infection.

Starvation-induced autophagy impaired influenza virus-specific adaptive immune responses. Mice were treated i.v. with rapamycin daily from day −2 to day 9 during infection or kept on food-restricted or ad libitum-fed condition for 7 d before influenza virus infection (30 pfu per mouse) and throughout infection. (A and B) Ten days later, lymphocytes were isolated from the lung. Influenza virus-specific CD8 + T cells were then detected using the H-2D b influenza virus NP tetramer. Numbers adjacent to outlined areas indicate percent tetramer-positive CD8 + T cells. (C and D) Total numbers of influenza virus-specific CD8 + T cells in the lung are shown. (E and F) Serum was collected at 10 d postinfection. Influenza virus-specific serum IgG levels were measured by ELISA. (G) The BALF was collected from influenza virus-infected animals at 2 d postinfection. IL-1β levels in BALF were determined by ELISA. (H) Total RNAs were extracted from the lung of the mice at 0 and 24 h postinfection. mRNA levels of pro–IL-1β were assessed by quantitative RT-PCR. GAPDH was used as an internal control. (I) Naïve OT-I CD8 + T cells (1 × 10 5 cells per well) were cocultured with CD11c + DCs (1 × 10 5 cells per well) that were isolated from the mLNs of PR8–OT-I virus-infected animals. Splenic DCs (1 × 10 5 cells per well) from infected animals were used as a negative control. Seventy-two hours later, IFN-γ production was measured by ELISA. (J and K) Mice were injected with CFSE-labeled OT-I CD8 + CD45.1 + T cells on the day before infection with 1,000 pfu of PR8–OT-I viruses. Four (J) or 5 (K) d after influenza virus infection, mLNs were excised to assess T cell proliferation by CFSE dilution. Total numbers of CD8 + CD45.1 + OT-I cells in the mLNs are shown. The data are representative of two independent experiments (A and B) or from two independent experiments (CK mean ± SEM). *P < 0.05, **P < 0.01, and ***P < 0.001 n.s., not significant (one-way ANOVA and Tukey’s test).

Signals Coming from ad Libitum-Fed Mice Restore Influenza Virus-Specific Adaptive Immune Responses in Underfed Mice.

Next we investigated whether immune defects in underfed mice could be rescued by signals coming from ad libitum-fed mice. To this end, we surgically joined CD45.1 + ad libitum-fed mice with CD45.2 + underfed mice at the time of influenza virus challenge (Fig. 3A). After the challenge, the parabionts were kept under ad libitum self-feeding condition (Fig. 3A). Self-feeding condition after influenza virus infection was insufficient to restore influenza virus-specific CD8 + T cell responses of underfed mice (SI Appendix, Fig. S15). Strikingly, CD45.2 + underfed parabiotic mice surgically joined CD45.1 + ad libitum-fed mice restored influenza virus-specific IgG antibody (Fig. 3B) and CD8 + T cell responses in the lung (Fig. 3C). In addition, most of the virus-specific CD8 + T cells in the lung of underfed parabiotic mice were found to be of host-derived CD45.2 + cells (Fig. 3 D and E). Further, underfed parabiotic mice secreted IL-1β into the alveolar space to the levels of ad libitum-fed parabiotic mice in response to influenza virus infection (Fig. 3F). These data indicated that signals coming from ad libitum-fed parabiotic mice within 10 d restored immune responses to influenza virus infection in underfed parabiotic mice through systemic circulation.

Signals from ad libitum-fed mice restore immune defects in underfed mice. (A) Mice were kept on food-restricted or ad libitum-fed condition for 7 d before influenza virus infection (30 pfu per mouse). Pairs of age-matched underfed C57BL/6 (CD45.2) and ad libitum B6-Ly5.1 (CD45.1) mice were surgically joined at the time of infection and kept on ad libitum-fed condition for 10 d. (B) Serum was collected at 10 d postinfection. Influenza virus-specific serum IgG levels were measured by ELISA. (C) Ten days later, lymphocytes were isolated from the lung. Influenza virus-specific CD8 + T cells were then detected using the H-2D b influenza virus NP tetramer. Total numbers of influenza virus-specific CD8 + T cells in the lung are shown. (D) Ten days later, the presence of influenza virus-specific host and donor CD8 + T cells in the lung of parabiotic mice was analyzed by flow cytometry. (E) Host-derived and partner-derived tetramer-positive CD8 + T cells (n = 12 pairs) in lung tissue are shown. (F) The BALF was collected from influenza virus-infected animals at 2 d postinfection. IL-1β levels in BALF were determined by ELISA. The data are representative of two independent experiments (D) or from two independent experiments (B, C, E, and F mean ± SEM). *P < 0.05, **P < 0.01, and ***P < 0.001 n.s., not significant (one-way ANOVA and Tukey’s test).

Glucose and SCFAs Restore Influenza Virus-Specific Adaptive Immune Responses in High Heat-Exposed Mice.

Thus far, our data indicated that reduced feeding behavior impaired the virus-specific CD8 T cells and antibody responses in underfed or high heat-exposed mice. A recent study has demonstrated that gavage of glucose protects mice from lethal influenza virus infection (28). In addition, Abx mice fail to mount protective CD8 + T cell responses following influenza virus infection (17) (SI Appendix, Fig. S8). These results led us to focus on the role of glucose or microbiota-derived SCFAs in the induction of adaptive immune responses to influenza virus infection. We hypothesized that glucose utilization or diet-derived SCFAs may be required for induction of adaptive immunity induced by influenza virus infection. To address these possibilities, high heat-exposed mice were given 1 M glucose in drinking water for the entire duration of the experiments. Then, we injected high heat-exposed mice with glucose i.v. daily from day −1 to day 3 during infection. Strikingly, we found that i.v. injection with glucose significantly enhanced secretion of mature IL-1β in the BALF, influenza virus-specific CD8 T cells, and antibody responses in high heat-exposed mice (Fig. 4 A–C). Injection of high heat-exposed naïve mice with glucose i.v. did not enhance secretion of mature IL-1β in the BALF (SI Appendix, Fig. S16). In addition, both secretion of mature IL-1β in the BALF and the virus-specific adaptive immune responses were significantly reduced in RT-exposed mice by blockade of glucose utilization with 2-deoxy- d -glucose (2DG) (Fig. 4 D–F), suggesting that glucose utilization is critical to mount adaptive immune responses following respiratory influenza virus infection.

Glucose utilization is required to mount adaptive immunity against influenza virus infection. (AC) Mice were kept at 22 °C or 36 °C for 7 d before influenza virus infection and throughout infection. High heat-exposed mice were treated with glucose i.v. daily from day −1 to day 3 during infection. (A) The BALF was collected from influenza virus-infected animals at 2 d postinfection. IL-1β levels in BALF were determined by ELISA. (B) Ten days later, lymphocytes were isolated from the lung. Influenza virus-specific CD8 + T cells were then detected using the H-2D b influenza virus NP tetramer. Total numbers of influenza virus-specific CD8 + T cells in the lung are shown. (C) Serum was collected at 10 d postinfection. Influenza virus-specific serum IgG levels were measured by ELISA. (DF) Mice were treated i.p. with 2DG daily from day −1 to day 1 (D) or 9 (EF) during infection. (D) The BALF was collected from influenza virus-infected animals at 2 d postinfection. IL-1β levels in BALF were determined by ELISA. (E) Ten days later, lymphocytes were isolated from the lung. Influenza virus-specific CD8 + T cells were then detected using the H-2D b influenza virus NP tetramer. Total numbers of influenza virus-specific CD8 + T cells in the lung are shown. (F) Serum was collected at 10 d postinfection. Influenza virus-specific serum IgG levels were measured by ELISA. The data are from three (AC) or two (DF) independent experiments (mean ± SEM). Statistical analysis was performed by two-tailed Student’s t test (AC) or one-way ANOVA and Tukey’s test (DF). *P < 0.05, **P < 0.01, and ***P < 0.001.

Finally, we examined whether SCFAs, such as butyrate, propionate, and acetate, can restore immune responses to influenza virus in high heat-exposed mice. A previous report indicates that i.v. injection with butyrate significantly enhances CD8 T cell response against influenza virus (20). Indeed, injection with butyrate enhanced secretion of mature IL-1β in the BALF (Fig. 5A), influenza virus-specific CD8 T cells (Fig. 5B), and antibody responses (Fig. 5C) in high heat-exposed mice. Remarkably, secretion of mature IL-1β in the BALF (Fig. 5A), influenza virus-specific CD8 T cells (Fig. 5B), and antibody responses (Fig. 5C) were partially restored in high heat-exposed mice by i.v. injection of propionate or acetate after influenza virus infection. These data collectively indicated that induction of autophagy by high-heat exposure or nutrient starvation impaired virus-specific CD8 T cells and antibody responses following respiratory influenza virus infection. Under such circumstances, both glucose and SCFAs might be important for efficient priming of inflammasome-dependent cytokine release and adaptive immune responses after infection with influenza virus.

SCFAs restored influenza virus-specific adaptive immune responses in high heat-exposed mice. (AC) Mice were kept at 22 °C or 36 °C for 7 d before influenza virus infection and throughout infection. High heat-exposed mice were given butyrate (200 mM), propionate (200 mM), or acetate (200 mM) in drinking water for 7 d before influenza virus infection and throughout infection and treated with each SCFA i.v. daily from day −2 to day 7 during infection. (A) The BALF was collected from influenza virus-infected animals at 2 d postinfection. IL-1β levels in BALF were determined by ELISA. (B) Ten days later, lymphocytes were isolated from the lung. Influenza virus-specific CD8 + T cells were then detected using the H-2D b influenza virus NP tetramer. Total numbers of influenza virus-specific CD8 + T cells in the lung are shown. (C) Serum was collected at 10 d postinfection. Influenza virus-specific serum IgG levels were measured by ELISA. The data are pooled from four independent experiments (A and B mean ± SEM). Statistical analysis was performed by two-tailed Student’s t test. *P < 0.05, **P < 0.01, and ***P < 0.001.

Adaptive suicide: is a kin-selected driver of fatal behaviours likely?

While several manipulated host behaviours are accepted as extended phenotypes of parasites, there remains debate over whether other altered behaviours in hosts following parasitic invasion represent cases of parasite manipulation, host defence or the pathology of infection. One particularly controversial subject is ‘suicidal behaviour’ in infected hosts. The host-suicide hypothesis proposes that host death benefits hosts doomed to reduced direct fitness by protecting kin from parasitism and therefore increasing inclusive fitness. However, adaptive suicide has been difficult to demonstrate conclusively as a host adaptation in studies on social or clonal insects, for whom high relatedness should enable greater inclusive fitness benefits. Following discussion of empirical and theoretical works from a behavioural ecology perspective, this review finds that the most persuasive evidence for selection of adaptive suicide comes from bacteria. Despite a focus on parasites, driven by the existing literature, the potential for the evolution of adaptive suicidal behaviour in hosts is also considered to apply to cases of infection by pathogens, provided that the disease has a severe effect on direct fitness and that suicidal behaviour can affect pathogen transmission dynamics. Suggestions are made for future research and a broadening of the possible implications for coevolution between parasites and hosts.

1. Introduction

Across all taxa that are involved in parasite–host relationships, a range of exploitative and defensive mechanisms have coevolved in the respective ‘sides’. A key question is whether some of the behaviours displayed by parasitized animals represent adaptations of the host or its parasite [1]. Behavioural changes following parasitic invasion vary greatly in their magnitude [2] and the adaptive significance, if any, is not always clear. One possibility is that altered behaviours may simply be a response to the pathological effects of parasites, and are not necessarily adaptive to either parasite or host. However, Moore [3] warns against explaining altered behaviours as ‘side effects’ of ‘pathology’, arguing that the fitness outcomes for participants in host–parasite associations, including parasite-induced behavioural alterations, will be subject to natural selection and therefore we should expect them to be more likely than not explicitly linked with the evolution of those species involved. Mostly though, studies focus on attributing behaviours to parasite adaptation or host adaptation.

On the one hand, if altered host behaviours are adaptive for parasites, they should facilitate the completion of their life cycle. This is typically achieved either by diverting the host's energy away from its own reproduction to the parasite for growth [3–6] or by rendering intermediate hosts more vulnerable to ingestion by the parasite's definitive host [3,7–10]. Where the life cycle of parasites involves stages that spend some time in a particular external environment, host behaviour can also be manipulated for the successful dispersal of parasite propagules in their most suitable conditions [1,3,11–13]. Interestingly, Poulin et al. [14] suggest that hosts may be capable of opposing some behavioural manipulation by established parasites, but the idea has received little attention. Certain host responses to infection by helminth parasites suggest that some hosts can remain at least partially in charge of their bodies, but lack of data is unsurprising because where infected hosts behave normally opposition to manipulation would not be differentiable from a parasite's failure to manipulate [14].

On the other hand, hosts may benefit from behavioural changes following parasitism. Most obviously, behaviours that serve to minimize damage from an internal parasite may reduce the negative impact of parasitism on a host, such as exhibiting sickness behaviour [15], behavioural fever [16,17] or self-medicating foraging [18,19]. More intriguingly, a host individual may benefit by sacrificing its direct fitness for the sake of increasing its inclusive fitness [20,21]. One fascinating, but controversial, mechanism through which this could occur is so-called ‘adaptive suicide’ behaviours, where post-invasion behaviours function to eliminate the propagation of an established parasite, thus protecting kin [22,23].

2. Adaptive suicide

The host suicide hypothesis [22] proposes that a host may use its own death to increase its inclusive fitness [20,21]. Where a parasitic infection effectively causes sterility or death, the host will be unable to improve its own reproductive fitness suicidal behaviour could enhance its inclusive fitness by preventing the maturation of its parasite and lowering the risk of parasite infection for relatives [22]. The fitness cost associated with death becomes negligible when a host's own expected reproduction approaches zero [23]. Provided that the host's death (and that of its parasite) reduces the level of subsequent parasitism in its kin relative to that in non-kin, there should be a positive selection value on the behaviour. Smith [22] argued that natural selection should drive the evolution of suicidal behaviour even when increases in inclusive fitness are very small, provided that: (i) the host's individual fitness is zero, (ii) upon emergence from the host, parasitoids are more likely to infect the host's kin than non-kin, and (iii) the kin's reproductive success is increased owing to the subsequent lowered risk of parasitism.

In order to satisfy these requirements, adaptive suicide was predicted to be most prevalent in colonial or social host species, or in members of host populations with low dispersal rates and a relatively high degree of inbreeding [22]. Conversely, parasitoid species with relatively small search ranges or areas of discovery would be particularly vulnerable if their hosts adopted this behaviour [22]. Suicidal behaviour can include activity that makes the individual more conspicuous to predators or easy to capture [22], or causes great costs in terms of energy spent, lost feeding opportunities and probability of death [23].

3. Empirical work

(a) Aggregating insects: aphids

McAllister & Roitberg [23] reported what they believed to be the first convincing evidence in support of Smith's [22] host suicide hypothesis, following their observations of pea aphids (Acyrthosiphon pisum) from different regions parasitized by the braconid wasp Aphidius ervi apparently exhibiting suicidal behaviour to different extents. In response to both aphid alarm pheromone and approaching coccinellid predators, aphids for whom the risk of death due to heat stress and desiccation was thought to be higher dropped more frequently when parasitized whereas aphids from cooler coastal regions behaved no differently when parasitized form when unparasitized [23]. From this, McAllister & Roitberg concluded that in a habitat where alternative escape tactics result in significant differences in mortality risk (interior regions), parasitized aphids chose the riskiest behaviour. Meanwhile in the habitat where alternative escape tactics result in no apparent difference in mortality risk (coastal regions), parasitized aphids behaved no differently from unparasitized aphids. Curiously, though, in both situations, parasitized aphids did not drop from plants without mediation by predation [23]. This study received a number of criticisms from Latta [24] and Tomlinson [25] which McAllister & Roitberg [26] addressed as ‘misunderstandings’ in a rebuttal, but the fact that adaptive suicide in this system was predator-mediated arguably suggests that the adaptation concerns more the survival of the parasitoid rather than a benefit to the aphid. If this is not the case, it makes little sense why aphids should not allow themselves to be consumed by predators. Indeed, we find in some more recent cases—discussed in greater detail later—that increasing mobility following invasion by a parasite may increase an aphid's likelihood of being consumed by a predator [27–29].

McAllister et al. went on to examine adaptive suicide in parasitized pea aphids of varying reproductive potential [30]. When aphids are parasitized at the second instar stage, they have no reproductive future and will not produce any offspring prior to mummification. However, aphids parasitized in their fourth instar can expect to produce seven to eight offspring before dying, directly increasing their own fitness, and so the cost of any altruistic behaviours upon parasite invasion may increase relative to the payoff for these individuals. Aphids parasitized at the second instar were found to use dangerous escape behaviour (dropping) when approached by a predator, while aphids parasitized at their fourth instar behaved no differently from unparasitized individuals [30]. This result was consistent with the prediction of McAllister & Roitberg that, as the cost of altruistic behaviour increases relative to inclusive fitness payoff, suicidal behaviour should disappear however, the escape behaviours were again elicited by the presence of a predator, weakening any support for a host-benefitting adaptation [30].

Many aphid species disperse away from their colony mates and mummify elsewhere following parasitism, but there is not always evidence to suggest that this behaviour is host- or parasitoid-mediated [31,32]. It is also possible that both parasitoid and host benefit to an extent. Perhaps the host gains indirect fitness benefits by transporting the parasitoid away from kin, while the parasite does not suffer from this so long as it is no more challenging to find some (non-kin) aphids. Moreover, the parasitoid might actually benefit if the move is to a safer microclimate [33]. Considering other potential evidence for altruism in non-eusocial parthenogenetically reproducing aphids, Wu et al. [34] looked at the smearing of cornicle secretions by cereal aphids (Sitobion avenae) onto parasitoids (Aphidius rhopalosiphi). Cornicle secretions of aphids were concluded to be altruistic against parasitoids, as they provided no direct fitness benefits to secretion-releasing individuals, only indirect fitness benefits through negatively impacting the parasitoid's subsequent foraging time and offering some protection to neighbouring clone-mates [34]. Smearing also occurred more frequently when a greater number of clone-mates were present, increasing inclusive fitness benefits [34]. This appears to be a case of kin-directed altruistic defence outside eusocial animals. Interestingly, non-social aphids also appear to possess surprising kin-recognition abilities, varying in aggregation and defensive abilities depending on the relative presence of clone-mates and non-kin [35].

With an increased awareness of the potential for aphids to recognize kin, it is interesting to consider that adaptive suicide following parasitism in the presence of predators need not involve dropping from, or otherwise leaving, an area to altruistically remove a parasite. Meisner et al. [27] demonstrated that pea aphids at earlier stages of parasitism suffer higher predation by the coccinellid predator Harmonia axyridis than unparasitized aphids. Durán Prieto et al. [28] proposed that if the behaviour of parasitized aphids was the cause of their more intense predation, it should be expected that parasitized aphids will suffer greater predation from predators other than coccinellids, especially if their behaviour has an adaptive value. Those authors explored predation of recently-parasitized pea aphids by the hemipteran Macrolophus pygmaeus, obtaining a similar result to Meisner et al. [27]. As the predation rate was not affected by the ratio of parasitized to unparasitized aphids, the energy and nutrition obtained from both prey types can be assumed to be equal and therefore prey preference was likely down to aphid behaviour rather than physiology. Higher mobility after being parasitized was evident. It is therefore plausible that the suicidal behaviour seen in pea aphids following parasitism [23,30] can function by increasing the rate of encounter between the predator and the parasitized prey [28]. By behaviourally offering themselves up, as well as removing the parasite from the area parasitized, hosts may help satiate predators in order to protect unparasitized kin. This hypothesis is supported by the observation of Meyhofer & Klug [29] that a lacewing predator, Chrysoperla carnea, took significantly less time to capture a parasitized black bean aphid (Aphis fabae) as its next victim than an unparasitized one.

(b) Eusocial insects: bees and ants

Schmid-Hempel & Müller [36] reported that worker Bombus lucorum bumblebees parasitized by conopid flies remain outside the nest longer than unparasitized workers during foraging hours and may abandon the nest altogether. They suggested that this would benefit the parasitoid pupae as they might be less subject to the infections that can develop on abandoned combs in bumblebee colonies. However, Poulin [37] suggested that these changes in behaviour are more plausibly an adaptive response of the host, resulting in inclusive fitness, and therefore an example of the adaptive host suicide as proposed by Smith [22]. Fritz [38] pointed out that natural selection should favour parasitoids that manipulate the host in ways that reduce its mortality likelihood before the parasitoid pupates, but bumblebees would in fact be more susceptible to predation, starvation and superparasitism outside the nest [37].

However, the conopid–bumblebee association does not meet the conditions of adaptive host suicide as laid out by Smith [22] for two reasons: (i) by the time adult conopids emerge from pupae, the host's kin have dispersed or died (ii) adult conopid females spread widely away from their site of emergence and so would not preferentially infect the bumblebee's kin even if the bumblebee allowed it to live [37]. McAllister et al. [30], though, pointed out that early death of a parasitized host will be adaptive as long as the cost of decreased reproductive success are outweighed by the benefits of increased inclusive fitness. Poulin [37] argued that the cost of death for a parasitized bumblebee worker is in fact very low as its reproductive potential approximates to zero following infection, as does its use as a forager in the colony. On the other hand, leaving the nest could increase a bee's inclusive fitness as parasitized workers are susceptible to further attack from conopid flies, and so leaving the colony may attract fly attacks away from non-parasitized kin [37]. Additionally, by leaving a nest, a parasitized bee with lower foraging efficiency might avoid depleting the colony's food stores for its own, unproductive survival, thus leaving more available for its kin [37]—a behaviour we here dub the ‘Captain Oates Effect’.

Poulin's [37] interpretation, though, was criticized in the response by Müller & Schmid-Hempel [39]. Bumblebees tend to intermingle with foragers from many different colonies when outside their nests, and so staying outside the colony and acting as a target for fly attacks is very likely to protect kin and non-kin from parasitism to similar degrees the benefits would not be disproportionately routed towards kin to the extent that the kin-selection hypothesis requires [39]. Müller & Schmid-Hempel [39] also argued that, from their observations, there is no evidence that parasitized bumblebees are not able to feed for themselves on flowers and so they would not necessarily depend on food stores in the hive anyway. Müller & Schmid-Hempel [40] subsequently found evidence of parasitized bumblebees exploiting cold temperatures as a defence against parasitoids. Parasitized workers stayed in the field overnight instead of their nest the cold temperatures outside the nest could retard the maturation of the parasite, reducing its chance of successful development. In choice experiments, parasitized bees were also demonstrated to actively seek out cold temperatures [40] although not supportive of adaptive suicide, these findings did suggest a larger role for host advantage rather than pure parasite manipulation.

Unrelated to parasitism, apparently altruistic self-removal from the hive has been reported in health-compromised honeybees (Apis mellifera), whose presence may be harmful to their colony [41]. Other studies have previously suggested that different eusocial insects permanently leave their colonies when infected [42,43], but it is difficult to pick apart host adaptation from potential parasitic manipulation, or indeed pathological trauma. Through artificially compromising honeybee foragers, Rueppell et al. [41] provided experimental evidence that self-removal need not be caused directly by parasitic manipulation or related to stress-induced foraging [44] or loss of orientation abilities [43] altruistic self-removal could be a host adaptation to increase inclusive fitness. Further, a simple model suggested that altruistic self-removal by sick social insect workers, in order to prevent disease transmission to kin, is expected under most biologically plausible conditions [41]. When occurring after infection from a parasite, self-removal from a colony might in some cases qualify as adaptive suicide.

However, colony desertion following parasitism certainly does not always come from altruism. Hughes et al. [45] describe a fascinating behavioural change in the paper wasp Polistes dominulus following infection by the strepsipteran parasite Xenos vesparum which culminates in colony desertion and the formation of extranidal groups in which up to 95% of occupants are parasitized females. While altruistic desertion to reduce infection of kin would generally be a good strategy for infected social insects, this is untenable in this case because female X. vesparum parasites are only infective if inseminated and wasp copulation does not occur on the nest owing to occupants vigorously attacking free-living males. The nest desertion and aggregation by infected wasps is most likely a case of adaptive parasite manipulation of host behaviour in order to facilitate parasite mating [45].

As in aphids [27–29], however, adaptive suicide in eusocial insects may not always involve spatial separation of a host from its kin selective predation on parasitized hosts could also help hosts altruistically protect their unparasitized kin from a parasitoid. Mathis & Tsutsui [46] studied the rove beetle Myrmedonota xipe, which associates with—typically highly aggressive—Azteca sericeasur ants. Rove beetles were found to selectively locate and prey upon ants parasitized by phorid parasitoid flies. Parasitized ants acted less aggressively towards the beetles than healthy ants, meaning that rove beetles can eat them alive without interruption [46]. Unable to access the aggressive, unparasitized ants as a food resource, M. xipe appeared to almost exclusively prey on parasitized ants, but this could also benefit the infected ants as being consumed would reduce the phorid fly population free to infect their kin. On the one hand, this system seems a good candidate to meet the criteria for the host-suicide hypothesis, as A. sericeasur is a polygynous and polydomous social insect that forms wide-spanning territories and so emerging mature parasitoids are far more likely to encounter their host's kin than non-kin [46]. On the other hand, it may be that not all phorid fly larvae successfully mature, and so selective predation of ants that would survive parasitism would ultimately cost the colony as a whole [46]. Parasitized workers may also be active colony members during the development of the parasitoid, and in these cases the benefit of eliminating the larvae via predation may be offset by the costs to the colony incurred from losing productive parasitized workers [46]. Further work exploring the true costs and benefits of selective rove beetle predation to parasitized ants will certainly shed more light on the evolution of this system but, as Mathis & Tsutsui conclude, beetle predation may indirectly benefit ants where parasitized ants can reduce the numbers of developing parasitoids by increasing their appeal as prey. Selective predation on parasitized hosts, beyond aphids, has been demonstrated in several studies, including in lepidopterans [47] and non-eusocial hymenopterans [48] (also see review by Rosenheim et al. [49]) exploring the possibility of this as a pre-emptive adaptive suicide strategy across different taxa will also be useful in advancing understanding of responses to parasitism.

(c) Bacteria

An extreme defensive immune strategy in bacteria against phages is the deployment of abortive infection (Abi) systems that abort phage infection but also lead to the death of the infected bacterial cell [50]. Abi systems protect neighbouring bacteria at the expense of the individual expressing the trait [51]. Altruistic deployment of Abi systems is particularly likely to be selected for where a bacterium's neighbouring cells are kin emerging from clonal expansion or, additionally or alternatively, cells have other factors that favour cooperation, such as aggregation as part of a biofilm [52,53]. Makarova et al. [52] hypothesized that immunity and suicide systems in bacteria are coupled and that complex decision-making involving sensing the course of a viral infection may determine whether the response to a virus involves induction of dormancy, an immune response, or suicide in the face of immune system failure. Works investigating recently discovered Class 2 CRISPR-Cas (Clustered Regularly Interspaced Palindromic Repeats and CRISPR-associated genes) systems [54–56] have since found the most direct link between immunity and programmed cell death in microbes discovered yet [57]. It is thought that immunity–suicide coupling is favoured in situations where a system includes dual function components that are involved both in immune and in suicidal activities [58] this could be the case for some Cas proteins [57].

The ‘decision’ to commit adaptive suicide in bacteria likely involves diverse signal transduction pathways [52]. In eukaryote yeast cells (Saccharomyces cerevisiae), natural programmed cell death is thought to hinge on the degree of damage to genetic material, with its critical value determined by quorum-sensing machinery [59]. Quorum-sensing is also an important process in prokaryote bacteria cell–cell communication, wherein extracellular signalling molecules are produced, detected and responded to [60,61]. Quorum sensing has been found to be important in the sporulation-competence decision in Bacillus subtilis [62,63], and Hazan et al. [64] recently described a novel quorum-sensing-regulated bacterial mechanism that controls self-poisoning of the respiratory chain in Pseudomonas aeruginosa, providing a fitness benefit to the microbial collective. A mechanism involving quorum-sensing is likely to be an important element in the mechanics of adaptive cell death following infection in bacteria [52].

Beyond sensing population levels by quorum-sensing, proteins that can sense damage and ‘predict’ the outcome of infections will also be important in mediating Abi systems and toxin–antitoxins [65,66] that colocalize with immunity genes [57]. The exact mechanisms and structures that forecast the course of virus infections remain to be fully elucidated, but it is thought that whenever dedicated sensor molecules indicate an attack is manageable the cell mobilizes its immune system, while if the indications of attack are dire then self-afflicting programmes are triggered [57]. Switching from the immune mode to the suicidal mode of defence may be in part governed by sensors determining the level of damage inflicted on a cell [57]. Intriguingly, though, type VI-A CRISPR-Cas systems appear to take a short-cut in the cell's usual response relay by simplifying—or even skipping—the damage-sensing step and employing the main immune effector as the suicide effector as well, but these systems are rare in bacteria, perhaps suggesting that foregoing damage-sensing is costly [57]. Predictive and damage-sensing signals read and responded to by various sensors likely differ between defence systems (see [57] and references therein for details).

Several studies suggest that spatial structure and migration are important to the evolution of bacterial suicide upon infection as they impact relatedness and therefore the relative benefits of kin selection [67–71]. For example, Fukuyo et al. [69] competed altruistic Escherichia coli with an artificially engineered suicide mechanism against wild-type bacteria in the presence or absence of the phage λ. They found that in a spatially structured soft agar environment, altruistic suicide had a selective benefit for the bacteria, but this was not the case in a well-mixed liquid environment. Using the naturally-occurring Abi mechanism ‘Lit’ in E. coli, Berngruber et al. [67] varied the amount of mixing in environments more continuously and found again that spatial structuring was needed for the evolution of altruistic suicide but also that too little mixing might prevent the evolution of abortive infection owing to the reduced parasite spread under those conditions. A further study by Refardt et al. [70] confirmed these findings using the best characterized Abi system, ‘Rex’ in λ-lysogenic E. coli strains [50]. Refardft et al. [70] demonstrated that adaptive suicide can evolve even when genetic similarity between neighbouring strains is relatively low in their study of E. coli responding to the attack of an obligately lytic phage.

4. Theoretical work and evolutionary predictions

As discussed above, interpretations of empirical data that support the host suicide hypothesis [23,28,30,37] have often been criticized [24,25,39]. The adaptive significance of host suicide in particular has been challenged because the main supporting studies involved clonal aphids that aggregate [23,30] or eusocial Hymenoptera [36], where complex life histories have made it difficult to exclude alternative explanations or carry out rigorous analysis of fitness [70]. Even in Euphydryas phaeton caterpillars and their parasitoids—suggested by Smith [22] as an appropriate system for testing the hypothesis of host suicide—adaptive suicide has not yet been demonstrated to be more plausible than the behavioural changes serving to increase the parasitoid's chance of escaping predation and parasitism itself [72]. Yet theoretical work has convincingly revealed the conditions required for host suicide to evolve [22,30,71].

Smith's original hypothesis [22] logically suggested that adaptive suicidal behaviours would increase the inclusive fitness of a parasitized host if the following conditions were met: (i) suicidal behaviour prevents the parasite's maturation and emergence (ii) the mature parasite is more likely to infect the host's kin than non-kin and (iii) the benefit to the host, in terms of the increased fitness of the kin, is greater than the cost of the suicide, measured in terms of the loss of the host's own reproductive fitness. If not all of these conditions are met, early death of the host may still be adaptive as long as the costs of decreased reproductive success are outweighed by the inclusive fitness benefits [30]. One of the key points here is perhaps that the parasite infection must have a severe, if not lethal, consequence for the host's future reproductive success to ensure that swapping direct fitness benefits for kin-selected benefits would result in a net gain for the infected host. Debarre et al. [71] illustrated via modelling how suicide upon infection can be an adaptation, but only in response to extremely harmful parasites and in spatially structured environments.

Shorter & Rueppell [73] suggested that eusocial insects, rather than just aggregated clonal insects, may provide the best test systems for adaptive suicide owing to the high relatedness and relative strength of kin selection. While this may prove to be true, empirical work on bacteria appears to lend the greatest support for adaptive suicide so far, even in conditions of relatively low relatedness. Where suicide carries very low cost for committers in structured environments, because infected cells are moribund with no opportunities for further reproduction, apparent altruism can evolve if such an act provides a large benefit to survivors that then avoid extinction [70]. Conversely, in unstructured environments self-sacrificial suicide would be futile as it would not preferentially protect relatives and so in these situations individual-based resistance is the best tactic for bacteria to combat phages [68].

Selection for adaptive suicide in bacteria will likely be affected by ecological factors too. Refardt & Kümmerli [68] found that in structured environments suicidal host defence was slightly less efficient than individual-based resistance in withstanding phages. They proposed that the putative lower efficiency of abortive infection might be compensated by a lack of pleiotropic costs compared with those usually associated with individual-based resistance mechanisms. Lion & Gandon [74] further suggest that selection for altruistic suicide should be maximized at low host dispersal and at intermediate parasite dispersal, owing to their roles in spatial structuring. Horizontal transfer of altruistic suicide Abi systems may also play an important role in their evolutionary success [74]. However, it remains unclear how adaptive suicide can outcompete simpler bacterial defence strategies preventing initial infection [74] and selection for adaptive suicidal behaviours is yet to be convincingly demonstrated in more complex organisms.

If there are instances where suicidal behaviours will be selected for in infected hosts, this raises the question of how adaptive suicide persists evolutionarily if the parasite species would consistently lose out. One of Tomlinson's [25] issues with McAllister & Roitberg's first study concerning adaptive suicide in aphids [23] was that natural selection on parasites would favour the subversion of such suicidal behaviour that benefitted their hosts, and that ‘in any ensuing “arms race”, asymmetries of selection should favour the parasite’. Blower et al. [75] describe a fascinating means by which a bacteriophage counter-evolved to avoid having its replication blocked by an infected cell's premature suicide. Here, they found the bacteriophage evolving sequences that mimicked the cell's antidote to its own toxins, allowing it to continue replicating without being destroyed by its host's defensive system. However, there are in fact some conditions in which selection on a parasite might not be able to override selection for host-benefitting suicidal behaviour. Firstly, if suicidal behaviour is triggered by a complex set of stimuli then the likelihood that selection for variation in parasite traits could occur just in order to subvert such a complex behaviour is perhaps very low [26]. Secondly, in situations where the costs of maintaining such strong control over hosts would be high relative to the payoff, parasitoids may not be selected to overcome host behaviours. McAllister & Roitberg [26] give the example of parasites with exceptionally high fecundity, for whom the cost of providing each offspring with sufficient neurotoxins to alter the behaviour of every host would be exceedingly high.

It is also worth considering whether host–parasite interactions may have coevolved over time such that suicidal behaviours in hosts may sometimes benefit both the host and its parasite. As mentioned earlier, it seems plausible to us that there may be cases where a parasite is either neutral towards or may benefit from an infected host dropping or otherwise moving away from its kin. So long as it is possible to encounter hosts of some sort, kin or not, after its emergence, the parasitoid does not need to lose out from the host's behaviour, while the host still gains inclusive fitness benefits from protecting its kin. In fact, if the move away from the host's kin also moves the parasitoid offspring to a safer microclimate for maturation and emergence then perhaps both ‘sides’ of the interaction benefit from the altered behaviours. It is also not much of a stretch to consider that some of the instances where infected hosts make themselves more vulnerable to predation, through either conspicuous behaviour or movement to particular locales, might aid parasites with particular life histories that require transmission from intermediate to definitive host while also sating predators to protect the host's kin. While it would be difficult to parse out whether a host's kin truly benefit from these sorts of scenarios, given that the parasite evidently succeeds in being transmitted to its definitive host, we consider it likely that the benefit of a host's behaviour to either the host or the parasitoid is context-dependent. As an example, nest abandonment by bumblebees could benefit the host more than the parasite in cases where the parasite is highly abundant and virulent and nest cleaning behaviours will be overwhelmed this is discussed further in the next section. It is important that the full population dynamics at play are considered where possible.

5. In which situations might adaptive suicide evolve?

While the focus of this paper has been on host adaptations following infection by parasites, because previous work in this field has focused on parasitism, we see no reason why cases of infection by disease should not also lead to the evolution of suicidal behaviours that benefit the hosts. The key aspect to both diseases and parasites that can potentially provoke the evolution of adaptive host suicide is that they must have a severe effect on direct fitness otherwise it is unlikely that a behaviour will evolve to compromise direct fitness in order to boost indirect fitness. A major means by which pathogens or parasites can impact direct fitness is by being highly virulent. If virulence is not high, then a behaviour that sacrifices a host's direct fitness to favour enhancing indirect fitness would not evolve. If virulence is high, the evolution of this behaviour is more likely, but the host behaviour must also be able to affect the transmission dynamics of the pathogen or parasite—this rules out some parasites, but also some highly virulent pathogens. It is easy to imagine how the transmission of helminths and the like can be affected by host behaviours, but for biting insects that just collect a blood meal and do not lay their eggs in or on a host, host behaviours will not influence their transmission.

Considered as an example, then, adaptive suicide should not develop in humans as a response to parasites like tsetse flies because they are not virulent enough. Nor would it develop in response to the protozoa that use the tsetse fly as a vector and cause sleeping sickness [76]. Even though sleeping sickness is highly virulent, killing virtually anyone untreated, the pathogen is spread only when another biting insect takes a blood meal from the infected person [76]. An infected individual could kill themselves as soon as they realized they were infected, thereby reducing their appeal to further tsetse flies as their body cools. However, it is not obvious that this reduction in pathogen prevalence would benefit kin in any meaningful way because the life cycle of the pathogen in the fly takes three weeks (from feeding on one person to being able to be spread to another), during which time the tsetse fly will have travelled a long distance there is little likelihood that the tsetse fly would spread the pathogen from you to your kin.

On the other hand, other taxa may be expected to evolve host suicidal tendencies when infected with particular pathogens and infections, as well as with certain parasites. We have already touched upon earlier cases where ants infected by fungal disease isolate themselves from their colonies [42], but while entomopathogenic fungi may experience increased transmission from their host's dispersal [41,77], this behavioural manipulation by the disease may be a co-option of host adaptive suicide. While nest hygiene behaviours in ants—e.g. removal of infected individuals or sequestering of individuals within the nest before they reach the infective stage—are typically a more effective fitness-enhancing strategy in the face of infections, adaptive suicide could evolve where infection rates are rapid and so extensive that the hygienic response is overwhelmed [78,79].

6. Conclusions and suggestions for future research

There is a lack of consensus on adaptive suicide. On the one hand, the behaviour seems theoretically very plausible as a highly effective host adaptation given an extremely harmful parasitized state and fate of significantly reduced direct fitness opportunities. On the other hand, empirical work has so far received much criticism, and teasing host adaptation apart from alternative explanations has proven difficult to do definitively. The best evidence, theoretical and empirical, for the selection of adaptive suicide in infected individuals originates in studies of bacteria and Abi systems.

One useful approach for future research—highlighted by Müller & Schmid-Hempel [39] in relation to parasitized bumblebees but true of any study on behavioural alterations upon parasitic invasion—would be for detailed measurements of costs and benefits for both the host and its parasitoid to be carefully analysed, along with any influence physiological stress may have. Including a consideration of the wider population dynamics and ecological context may be an important component of weighing up the net benefits to host and parasitoid. Elucidating the proximate mechanisms underpinning alterations of host phenotype [80], wherever possible, would also be valuable where they could help identify parasite manipulation—or indeed rule it out in favour of host adaptation or pathology. More behavioural studies on generally self-destructive behaviours in social insects, including cost–benefit analyses and mechanistic studies, are also needed [73] and comparisons between disease-related, condition-related and parasite-related behavioural changes may then shed more light on the potential for adaptive suicide upon infection relative to other explanations.

With regards to non-eusocial species that tend to aggregate with clone-mates, Durán Prieto et al. [28] propose a convincing explanation of how suicidal behaviours may lead to increased predation of parasitized aphids. Further studies should seek to investigate whether predation rates on unparasitized kin decrease thanks to parasitized aphids substantially increasing their own personal risk of predation by performing particular behaviours. It would be of great interest whether further studies could prove that, at an early stage of parasitism, greater susceptibility of parasitized aphids to predation is a common phenomenon [28].

From a different perspective, it would be interesting to explore whether there are any host–parasite systems that result in an infected individual decreasing its own fecundity in order to prevent parasites producing infectious units that could then infect its kin. This would perhaps be considered adaptive ‘reproductive suicide’, wherein all future reproduction and direct fitness is cut off, but perhaps where an individual could continue to assist kin without infecting them thus, it need not dispose of itself entirely. The reduction of host fecundity following parasitic invasion has previously been suggested as an adaptive strategy for damage limitation in some cases [81]. Hurd [82] describes how female host fecundity reduction in the association between metacestodes of the rat tapeworm (Hymenolepis diminuta) and a beetle intermediate host (Tenebrio molitor) can benefit both parasite and host. Here the host's rate of egg production is slower upon infection but this is traded off with a longer lifespan which might ultimately allow lifetime fecundity to equal or exceed that of uninfected females. The parasite can also gain from this if greater lifespan increases the probability of the beetle being predated, thus increasing the parasite's transmission [82]. Beyond a merely reduced host fecundity, if there are cases where a host ends its fecundity rather than increasing its mortality, a shutting down of reproductive effort could represent an entirely host-benefitting adaptation that might act to protect its kin from multiplied infectious units. Any exploration into such ‘reproductive suicide’ could give a further perspective on extreme kin-selected adaptations in the face of parasitism.

Modelling work exploring the precise relationship of costs and benefits involved in adaptive suicide in social insects could also be of great use in trying to understand in which situations the evolution of suicidal behaviours as a host adaptation could be more plausible than parasite manipulation and/or pathology. In the case of bacteria, future work developing understanding of how altruistic suicide can outcompete simpler defences that prevent infection in the first place would be hugely valuable [74]. Further details on the nature of the switching signals in immunity–suicide coupling in bacteria, the relevant threshold values, and the determinants of these are all intriguing avenues open for future studies [57]. The longer-term effects of adaptive suicide in bacteria on the complexity [83] and evolution of microbial populations will also be interesting to further explore. Broadening the theoretical framework to include awareness of spatial structuring and the diversity of host and parasite life cycles would allow the production of more informative models, and further empirical studies to validate theoretical predictions regarding selection under different spatial structures could also be hugely valuable [74]. The coevolutionary implications of adaptive suicide by bacteria to avoid population-wide infection in spatially structured environments remain ripe for empirical testing [84]. Greater consideration across taxa of where some behaviours could potentially benefit both host and its parasite—and explorations of where this may apply to cases of infection by pathogens too—could also yield interesting results.

Materials and methods


S. littoralis larvae were reared on artificial diet (47.3 g/l wheat germ, 67.3 g/l brewer’s yeast, 189 g/l corn meal, 6.8 g/l ascorbic acid, 0.75 g/l cholesterol, 0.5 g/l propyl 4-hydroxybenzoate, 3 g/l methyl 4-hydroxybenzoate, 1.3 g/l wheat germ oil, 33.8 g/l agar and 3 g/l vitamin mix (1.2 g/Kg vitamin B1, 2.6 g/Kg vitamin B2, 2.5 g/Kg vitamin B6, 40 g/Kg choline, 10 g/Kg pantothenic acid, 32 g/Kg inositol, 0.25 g/Kg biotin, 2.5 g/Kg folic acid, 5 g/Kg 4-aminobenzoic acid, 0.5 mg/Kg vitamin B12, 10 g/Kg glutathione, 2.1 g/Kg vitamin A, 0.25 g/Kg vitamin D3, 24 g/Kg vitamin E, 0.25 g/Kg vitamin K, 25 g/Kg vitamin C in dextrose)), at 25 ± 1°C, 70 ± 5% R.H., and under a 16:8 h light/dark period.

Tissue sample collection and RNA/DNA extraction

S. littoralis larvae were anaesthetized on ice and surface-sterilized with 70% ethanol prior to dissection. Larval haemolymph was collected from a cut of the leg and haemocytes were separated from plasma by centrifugation for 5 min, at 500 × g, at 4°C. Midgut and fat body were isolated after cutting the larval body lengthwise, and the remaining body carcass separately collected. These samples (i.e., haemocytes, midgut, fat body and carcass) were immediately transferred into TRIzol reagent (Life Technologies, Carlsbad, CA, USA) and kept at -80°C until total RNA extraction, that was performed according to manufacturer’s instructions. DNA was extracted from haemocytes using the protocol described elsewhere [59], with minor modifications. The concentration of extracted RNA or DNA was assessed by measuring the absorbance at 260 nm, with a Varioskan Flash Multimode Reader (Thermo Scientific, Waltham, MA, USA), and sample purity was evaluated by assessing 260/280 nm absorbance ratio. RNA quality was checked by electrophoresis on 1% agarose gel.

Sl gasmin cloning and bioinformatics analysis

A partial Sl gasmin cDNA (FQ973054.1) was identified by BLAST analysis [60, 61] in a public database of expressed sequence tags (EST) from S. littoralis female antenna, using as query the full length cDNA sequence of S. exigua gasmin (KP406767.1).

To isolate Sl gasmin ORF, total RNA extracted from haemocytes of S. littoralis 6 th instar larvae was subjected to retro-transcription (Ambion RETROscript ® kit—Life Technologies). Given the very high level of sequence identity to gasmin (Fig 1), a cDNA was obtained by PCR, using Phusion High-Fidelity DNA Polymerase (Fisher Scientific, Pittsburgh, PA, USA) and primers designed to amplify the whole gasmin ORF (gasmin ORF forward primer ATGTTGCCTATTACCATACTAACG, gasmin ORF reverse primer ATACTGGAATTGGACATATTTGAGC). PCR conditions were programmed for 30 s at 98°C 40 cycles of [10 s 98°C, 30 s 60°C, 1 min 72°C] and 15 min at 72°C. After amplification, the obtained PCR product was separated by gel electrophoresis and the visible band of the expected size was purified with a Quick gel extraction & PCR purification COMBO Kit (Invitrogen, Carlsbad, CA, USA). The PCR product was cloned into Zero-Blunt TOPO vector (Zero-Blunt TOPO PCR Cloning Kit, Invitrogen), according to the manufacturer’s instructions. After transformation of One Shot TOP10 chemically competent E. coli cells (Invitrogen), the transformants were incubated overnight at 37°C on LB plates containing 50 μg/ml kanamycin. Bacterial colonies containing the fragment of the appropriate size were selected by colony PCR, using Phusion High-Fidelity DNA Polymerase (Fisher Scientific) and M13 Forward (-20)/M13 reverse primers (Invitrogen), and grown overnight in LB medium containing 50 μg/ml kanamycin. The plasmid DNA was extracted from 4 ml of bacterial culture using a Charge Switch-Pro plasmid miniprep Kit (Invitrogen), as instructed by the manufacturer and sequenced.

The presence of an intron into Sl gasmin sequence was determined by PCR amplification of the total DNA extracted from S. littoralis haemocytes using Phusion High-Fidelity DNA Polymerase (Thermo Fischer Scientific) with specific primers (gasmin ORF forward primer ATGTTGCCTATTACCATACTAACG gasmin intron reverse primer CAGGTGTCCGCATTCCACTGA). The length of PCR products was checked by electrophoresis on 1% agarose gels, before sequencing.

Extensive similarity searches of complete and high-throughput genome sequence databases hosted by NCBI using TBLASTN, as well as of non-redundant protein databases (BLASTP), using the inferred S. exigua and S. littoralis protein sequences, allowed the identification of numerous potential homologues. These were manually annotated to identify probable coding and intronic regions. Preliminary alignments of inferred amino acid sequences (351 aa) were performed using Muscle [62] and manually filtered to identify homologues, which could be aligned essentially contiguously with the Spodoptera protein sequences. Alignments were manually refined, and ambiguously aligned regions were excluded using the program GBlocks [63], leaving XX amino acid positions for phylogenetic reconstruction using PhyML [64] as implemented in the program SEAVIEW [65], under the JTT amino acid substitution matrix, with 4 gamma distributed substitution rate categories (see S2 File). Bootstrap proportions were estimated using parameters optimized on the original ML tree (rate distribution parameter (alpha) = 1.45).

To identify signatures of selection the FEL (Fixed Effects Likelihood) method was used [66], implemented at the DataMonkey website ( [67]. Nucleotide sequences of S. exigua, S. littoralis and C. congregata virus circle 25, including the intron, were aligned and the alignment was manually curated to maintain in-frame codon alignment. Selection was tested in the S. exiguaS. littoralis branches.

Sl gasmin expression analysis by qRT-PCR

Total RNA used for transcriptional analysis was isolated as described above. Relative expression of studied genes was measured by one-step qRT-PCR, using the SYBR Green PCR Kit (Applied Biosystems, Carlsbad, CA, USA), according to the manufacturer’s instructions. S. littoralis β-actin gene (Z46873) was used as endogenous control for RNA loading. Primer Express 1.0 software (Applied Biosystems) was used to design the primers used (S3 Table). Relative gene expression data were analyzed using the ΔΔCt method [68–70]. qRT-PCR for measurement of Sl gasmin expression was carried out using specific primers (S3 Table), designed to detect a region of Sl gasmin mRNA not included in the sequence targeted by the dsRNA (see “dsRNA synthesis” paragraph). For validation of the ΔΔCt method, the difference between the Ct value of Sl gasmin and the Ct value of β-actin transcripts [ΔCt = Ct (Sl gasmin)-Ct (β-actin)] was plotted versus the log of ten-fold serial dilutions (2000, 200, 20, 2 and 0.2 ng) of the purified RNA samples. The plot of log total RNA input versus ΔCt displayed a slope less than 0.1 (Y = 1.149+0.0133X, R 2 = 0.0493), indicating that the efficiencies of the two amplicons were approximately equal.

Expression profiles of Sl gasmin in response to different pathogens

To analyze Sl gasmin expression in response to microbial challenge, S. littoralis 5 th instar larvae, surface-sterilized with 70% ethanol and chilled on ice, received an intra-haemocoelic injection of 2 × 10 7 E. coli, 3 × 10 8 S. aureus or 2 × 10 7 S. cerevisiae cells, suspended in 5 μl of PBS (Phosphate Buffered Saline: 137 mM NaCl, 2.7 mM KCl, 10 mM phosphate buffer, pH 7.4). Injections were performed through the neck membrane, using a Hamilton Microliter 1701RN syringe (10 μl, gauge 26s, length 51 mm, needle 3). At the injection and at different time points after injection, experimental larvae (n = 12 for each experimental treatment) were dissected and haemocytes were collected and processed for total RNA extraction as described above the relative expression of Sl gasmin was assessed by qRT-PCR.

DsRNA synthesis

Total RNA extracted from haemocytes of S. littoralis 6 th instar larvae was retro-transcribed (Ambion RETROscript kit, Life Technologies) and a 789 bp long Sl gasmin cDNA fragment was obtained by PCR, using the Sl gasmin dsRNA forward primer (GCCGGCATGTTGTCTATTACC) in combination with the Sl gasmin dsRNA reverse primer (TCCTTCCAGCTTCTGAGTCA). This cDNA fragment was used as template for a nested-PCR reaction, performed with primers containing at their 5’ ends the T7 polymerase promoter sequence (T7-Sl gasmin forward TAATACGACTCACTATAGGGAG-TTCGAGGATACAAGCAGAG T7-Sl gasmin reverse TAATACGACTCACTATAGGGAG-GGATGCTCAGGATATCTGTTAC). The resulting PCR product was used as template to synthesize Sl gasmin dsRNA (522 bp long), using the Ambion MEGAscript RNAi Kit (Life Technologies), according to the manufacturer’s instructions. Control dsRNA, 500 bp long, was obtained from a control template supplied by the kit used. dsRNA preparations were quantified by measuring their absorbance at 260 nm with a Varioskan Flash Multimode Reader, and purity was evaluated by assessing 260/280 nm absorbance ratios. Products were run on 1% agarose gels to confirm their integrity.

Administration of dsRNA to S. littoralis larvae and silencing of Sl gasmin

S. littoralis 4 th instar larvae (1 st day) were anaesthetized on ice and 1 μl of Sl gasmin dsRNA or control dsRNA (see above), dissolved in PBS, was poured into the lumen of the foregut by means of a Hamilton Microliter 1701RN syringe (10 μl, gauge 26s, length 51 mm, needle 2). dsRNA treatments consisted of one oral administration of 150 ng per day, for 3 days (from 4 th to 5 th instar). After the last dsRNA administration and prior to any experiment, haemocytes from 3–4 treated larvae were used for qRT-PCR analysis, to confirm the occurrence of gene silencing.

Cry1Ca bioassay and assessment of bacterial load

Purified Cry1Ca protein was produced in a recombinant B. thuringiensis strain EG1081 (Ecogen Inc.). Prior to use, Cry1Ca was dialyzed overnight, at 4°C in 50 mM sodium carbonate buffer, pH 9.0. After dialysis, toxin concentration was determined by the Bradford assay [71], using bovine serum albumin as standard.

Silenced and control larvae were singly isolated in multi-well plastic trays (Bio-Ba-32, Color-Dec, Italy), containing artificial diet, covered with perforated plastic lids (Bio-Cv-4, Color-Dec), and maintained under the rearing conditions reported above. For the first 3 days, the upper surface (1 cm 2 ) of the artificial diet (0.3 cm 3 ) was uniformly overlaid with 50 μl of purified Cry1Ca toxin, dissolved in 50 mM sodium carbonate buffer at pH 9.0. Control larvae were reared on artificial diet overlaid with 50 μl sodium carbonate buffer. Experimental larvae were maintained on artificial diet, replaced every 24 h, and daily inspected for survival, until pupation. To determine the 50% lethal concentration (LC50) of Cry1Ca toxin, the bioassay was carried out at 5 different concentrations of toxin and using 16 larvae for each experimental condition and control. Probit analysis [72], to determine LC50 values at day 10, 90% fiducial limits and toxicity increase ratio (TI) for each experimental treatment, was performed with the POLO-PC program (LeOra Software, Berkeley, CA).

The assessment of the bacterial loads was performed as previously described [32]. Briefly, 6 h after the last Sl gasmin or GFP dsRNA administration, newly molted 5th instar larvae of S. littoralis, were exposed for 3 days to 2.7 μg/cm 2 (corresponding to the LC50 of Cry1Ca in Sl gasmin silenced larvae). Both experimental groups included internal controls maintained on a toxin-free diet. On day 7, larvae were transferred to an untreated diet and, 24 h later, the midgut and haemolymph were separately collected under a horizontal laminar flow hood as described above. Experimental samples were obtained by pooling 10 larvae. The experiment was repeated 3 times. Changes over time in the relative bacterial load in the midgut (n = 7 for each sampling point) and haemolymph (n = 8 for each sampling point) samples were determined by qRT-PCR, measuring the transcript level of 16S rRNA (AJ567606.1) to assess the impact of Bt toxin and Sl gasmin silencing on bacterial proliferation. The qRT-PCR was performed as described above, using S. littoralis β-actin as reporter gene (primers used are reported in S3 Table).

Cellular and humoral immune assays

The impact of gene silencing on cellular immune responses was assessed by scoring its effect on encapsulation, nodulation and phagocytosis. Encapsulation and nodulation responses were assessed as previously described [31, 32]. CM Sepharose fast flow chromatography beads (Pharmacia), suspended in PBS, were injected into the haemocoel of S. littoralis larvae using a Hamilton Microliter 1702 RN syringe (25 μl, gauge 22s, length 55 mm, needle 3). After 24 h, beads were recovered upon larval dissection and scored to evaluate their encapsulation rate, which was expressed with the encapsulation index (E.I. = [Σ (encapsulation degree × total beads of this degree)/ total beads × 4] × 100), that takes into account both the encapsulation degree of each recovered bead (0—no cells adherent to the beads, 1—up to 10 adherent cells, 2—more than 10 adherent cells but no complete layer around the bead, 3—one or more complete layers without melanization, 4—one or more complete layers with melanization) and the relative abundance of beads with a given encapsulation degree [73]. For the nodulation assay, 12 h after the last dsRNA administration, S. littoralis larvae, surface-sterilized with 70% ethanol and chilled on ice, received an intra-haemocoelic injection of 5 μl of a PBS suspension of 2 × 10 6 E. coli cells, or 2 × 10 7 S. cerevisiae cells. Injections were performed through the neck membrane, using a Hamilton 1701 RN SYR (10 μl, 26s gauge, 55 mm long, point style 3). A thoracic leg was cut 18 h after injection, and the exuding haemolymph was collected and immediately diluted into an equal volume of ice-cold MEAD anticoagulant buffer (98 mM NaOH, 145 mM NaCl, 17 mM EDTA, and 41 mM citric acid, pH 4.5). The haemocyte nodules occurring in the haemolymph samples were counted under a light transmitted microscope at 400× magnification (Axioskop 20, Carl Zeiss Microscopy, Germany), using a Bürker chamber. When an intense immune response gave rise to large aggregates of nodules difficult to count separately, the number of distinct nodules observed was arbitrarily doubled, because the percentage of nonwhite pixels measured (ZEN software Carl Zeiss Microscopy) on the large aggregates was on average twice that measured on a bright-microscopy field containing discrete nodules and free haemocytes.

To measure phagocytosis competence of S. littoralis haemocytes, an in vitro assay was performed as described in [32] with minor modifications. Briefly, haemolymph samples were collected from a cut of the leg into ice-cold PBS (1:1 v/v) and added with an equal volume of a PBS suspension of 2 × 10 6 fluorescein conjugated E. coli cells (K-12 strain BioParticles, fluorescein conjugate, Invitrogen) or 2 × 10 7 S. aureus (Wood strain, BioParticles fluorescein conjugate, Invitrogen). After incubation with E. coli (10 min) or with S. aureus (30 min), samples were loaded into a Bürker chamber, where total and fluorescent haemocytes were counted under a fluorescence microscope (Axioskop 20). Prior starting incubation experiments, vital staining with trypan blue was used to routinely check the viability of collected haemocytes. A haemolymph aliquot was mixed with 0.4% (w/v) trypan blue (2:1 v/v), prior to count viable and dead cells under a light transmitted microscope (Axioskop 20), using a Bürker chamber. Haemocyte samples with a viability rate lower than 98% were discarded.

For rescue experiments with haemocytes from silenced larvae, haemolymph samples were extracted from S. littoralis larvae chilled on ice, 24 h after the last dsRNA administration (Sl gasmin dsRNA or control dsRNA). Samples were centrifuged 5 min at 500 × g, at 4°C. The plasma was kept on ice, haemocytes were resuspended in PBS and centrifuged as previously described. PBS was then removed and haemocytes from larvae treated with Sl gasmin dsRNA were resuspended in the plasma isolated from larvae treated with control dsRNA, while haemocytes from larvae treated with control dsRNA were resuspended in the plasma isolated from larvae treated with Sl gasmin dsRNA. Then, the phagocytosis by haemocytes was evaluated as described above.

The humoral immune response, as affected by gene silencing, was assessed by measuring the transcript level of genes encoding antimicrobial peptides and lysozyme, in response to injections of different microorganisms, as previously described [32]. Briefly, 6 h after the last dsRNA administration, S. littoralis larvae, surface-sterilized with 70% ethanol and chilled on ice, received an intra-haemocoelic injection of 2 × 10 7 E. coli or S. aureus cells, or 3 × 10 8 S. cerevisiae cells, suspended in 5 μl of PBS. Injections were performed through the neck membrane with a Hamilton 1701 RN SYR (10 μl, gauge 26s, length 55 mm, needle 3). At the time of injection and 18 h after injection, larvae (n = 8 for each experimental sample) were dissected and haemocytes, midgut, and fat body were collected and processed for total RNA extraction, as described above. The relative expression of attacin 1 (FQ971100.1), gloverin (FQ965511.1), and lysozyme 1a (FQ961692.1) were thus assessed by q-RT-PCR as described above. Primers used are reported in S3 Table.

Detection of actin filaments in haemocytes

Newly moulted 5 th instar larvae of S. littoralis, treated with Sl gasmin dsRNA or control dsRNA, as described above, were surface-sterilized with 70% ethanol and chilled on ice. Larval haemolymph from individual larvae was collected from a cut of the leg and placed on glass slides for 10 min, to allow the haemocytes to settle and attach to the glass. Haemolymph was then carefully removed and haemocytes rinsed 3 times with PBS. Attached cells were fixed for 10 min in 4% paraformaldehyde in PBS, washed 3 times in PBS and permeabilized for 4 min with 0.1% Triton-X100 in PBS. Haemocytes were washed 3 times in PBS and then incubated for 20 min with 4 μg/ml TRITC-phalloidin (Tetramethylrhodamine B isothiocyanate-phalloidin). After 3 rinses in PBS, the samples were mounted in Vectashield Mounting Medium with DAPI (Vector Laboratories) and examined under a fluorescence microscope (ZEISS Axiophot 2 epifluorescence microscope). The observations have been performed in 3 different experiments and considering at least 10 randomly selected microscopic fields for each experimental condition.

Multiple reaction monitoring (MRM) targeted proteomic approach

To detect the presence of Sl gasmin in the plasma, haemolymph was centrifuged as described above to remove the haemocytes, and the supernatant (plasma) was stored at -80°C. Samples were then dissolved in denaturant buffer (6 M urea, 10 mM EDTA, 300 mM Tris, pH 8.0) containing dithiothreitol (10-fold molar excess on the Cys residues) at 37 °C for 2 h, before the addition of iodoacetamide (IAM) to perform carboamidomethylation, using 5-fold molar excess of alkylating agent on thiol residues. The mixture was incubated in the dark at room temperature for 30 minutes and the product was purified by Chloroform/Methanol/H2O precipitation. Supernatants were removed and the pellets were dried. Digestion of the protein mixture was carried out in 10 mM ammonium bicarbonate (AMBIC), using trypsin at a 50:1 protein:enzyme mass ratio. The samples were incubated at 37°C for 16 h and dried after acidification (10% HCOOH in water). To eliminate any impurity, samples were suspended in 200 μl of 100 mM AMBIC, filtrated by centrifugal filter units (0.22 μm) and dried in a speed-vac concentrator. Samples were evaporated and suspended in 50 μl of 0.1% HCOOH in water. Peptide mixtures were analyzed by LC-MRM/MS analysis using a Xevo TQ-S (Waters, Milford, MA, USA) with an IonKey UPLC Microflow Source coupled to an UPLC Acquity System (Waters), using an IonKey device. For each run, 1 μl peptide mixture was separated on a TS3 1.0 mm × 150 mm analytical RP column (Waters) at 60°C, with a flow rate of 3 μl/min using 0.1% HCOOH in water (LC-MS grade) as eluent A, and 0.1% HCOOH in acetonitrile as eluent B. Peptides were eluted (starting 1 min after injection) with a linear gradient of eluent B in A, from 7% to 95% in 55 min. The column was re-equilibrated at initial conditions for 4 min. The MRM mass spectrometric analyses were performed in positive ion mode using a MRM detection window of 0.5–1.6 min per peptide the duty cycle was set to automatic and dwell times were minimal 5 ms. Cone voltage was set to 35V. The selected transitions and the collision energy for each Sl gasmin peptide are reported in S2 Table.

To determine whether Sl gasmin is able to bind to the surface of bacteria, plasma samples obtained from S. littoralis 5 th instar larvae (n = 20) were added to an equal volume of MEAD and incubated with an equal volume of E. coli suspension in PBS (4 × 10 6 cells for each μl of haemolymph) for 1 h. The suspension was then centrifuged for 10 min, at 12,000 × g, at 4°C and the pellet resuspended in 2 ml of 10 mM phosphate buffer, 45 mM NaCl, pH 7.4. Centrifugation and resuspension were repeated and the bacterial pellet as well as supernatants were frozen in liquid nitrogen and stored at -80°C until use. In control experiments bacteria were incubated with PBS and MEAD (1:1:1 v/v/v). Samples were submitted to reduction, alkylation and tryptic digestion as described above. After the preparation step they were processed and analyzed by LC-MRM/MS, as previously described. LC-MRM/MS analyses were performed on 3 technical replicates for each biological replicate and the average of these multiple measurements was used for data analysis. The data obtained represent the average value of total ion current associated to each transitions for the selected peptides.


Unless differently indicated, all reagents were provided by Sigma-Aldrich, Italy.

Statistical analysis

Data were analyzed using Prism (GraphPad Software Inc. version 6.0b, San Diego, CA, USA) and SPSS (IBM SPSS Statistics, Version 21, Armonk, NY) software. The comparison between 2 experimental groups was done using the unpaired Student’s t test, while in the case of more than 2 experimental groups, One-Way ANOVA. Two-Way ANOVA was carried out on AMP and lysozyme 1A immune induction experiments, with RNAi treatment and bacterial injection as factors, while a Three-Way ANOVA was carried out for bacteria relative quantification, with dsRNA treatment, time and Cry1Ca toxin exposure as factors. When necessary transformation of data was carried out, to meet the assumption of normality. Levene’s test was carried out to test the homogeneity of variance. When significant effects were observed (P value<0.05), Bonferroni’s post-hoc test was used. When one of the assumptions was not met, even after the transformation of the data, Kruskal-Wallis one-way ANOVA (non-parametric ANOVA) test was employed.

Measles’ superpower: extreme contagiousness

As well as being able to cripple your immune system, the infectivity of the measles virus is legendary: up to 9 out of 10 susceptible people will develop measles if they come into contact with an infected person.

Coughing and sneezing sends the measles virus hurtling out of your body and into the air. Image adapted from: James Gathany / CDC CC0

A measles-induced cough can easily and efficiently launch lots of viral particles into the air, where they can linger on surfaces for a long time (up to two hours, in some cases). This can be especially dangerous in aeroplane toilets, doctors’ waiting rooms or other places where many people share a small air space.