Mosquito ecology and control of malaria



  1. Mosquitoes transmit some of the most important infectious diseases of man including malaria that today kills around 0·6–1·2 million people a year, the majority children in low-income countries.
  2. There is increasing realisation that no single intervention is likely to halt malaria and a multipronged approach is needed including vector control.
  3. Very effective vector control measures are currently available, most involving insecticides, although there is evidence of growing problems with the spread of resistance. A variety of novel genetic approaches to vector control are under active development.
  4. Research on targeting the mosquito has been greatly facilitated by huge investment in molecular resources, including the provision of numerous full-genome sequences.
  5. Vector control is applied population biology, and I argue here that further progress will require as much attention to mosquito ecology as has been paid to mosquito molecular biology.


Many of the most significant infectious diseases of humans are transmitted by arthropods and in particular by mosquitoes (Culicidae). The most important is malaria, vectored by Anopheles mosquitoes (Gilles, Warrell & Bruce-Chwatt 1993), which sickens about 200 million a year and probably kills somewhere between 0·6 and 1·2 million people (WHO 2010; Murray et al. 2011), chiefly children in low income countries. The often spectacularly black-and-white-patterned mosquitoes in the genus Aedes are vectors of the yellow fever virus, which in the 18th and 19th centuries caused devastating epidemics in the New World as far north as Washington DC and in the Old World in Barcelona (D'Antonio & Spielman 2002). Today the most important disease it carries is dengue fever, which is becoming an increasing problem in the tropics and semi-tropics (Gibbons & Vaughn 2002; Simmons et al. 2012). Different genera of mosquitoes are vectors of the nematodes (Wuchereria, Brugia) that cause filarial elephantiasis (Hotez et al. 2007). Less well characterised are a suite of arboviruses (arthropod-borne viruses with different evolutionary origins) that chiefly circulate in animal hosts and that are occasionally transmitted to humans though are not vectored between humans. Perhaps, the best known, due to its recent introduction to the United States, is West Nile Virus which is primarily a bird parasite (Hayes et al. 2005). Others include the viruses that cause Rift Valley fever and Ross River fever, a variety of different encephalitis viruses, and the Chikungunya virus (WHO 1985).

Mosquitoes also vector diseases of livestock and wild animals. For example, different equine encephalitis viruses sicken horses in the USA (Young et al. 2008). Introduced mosquitoes and avian malaria are a major cause of the decline of native bird species in Hawaii (Woodworth et al. 2005).

The importance of mosquitoes in human and animal diseases has made them an important target of medical, veterinary and conservation research since Patrick Manson and Sir Ronald Ross first implicated mosquitoes in the transmission of filarial nematodes and malaria in the closing decades of the nineteenth century (D'Antonio & Spielman 2002). However, the extent to which studies of mosquito ecology have been seen to be central to medical entomology has varied over the years as the focus of disease control has evolved. My impression is that in the first half of the twentieth century, ecology, at least descriptive ecology, was accorded high importance as the life cycle and breeding sites of the major vectors were worked out. After the Second World War, the discovery of synthetic insecticides led to optimism that diseases could be eliminated by blanket spraying of the environment with DDT and its successors. This did indeed lead to major successes, for example the end of endemic malaria in Italy and other parts of Europe (Bruce-Chwatt & de Zulueta 1980). But progress in the tropics was much more mixed (Hay et al. 2004). In 1969, the Garki Project was launched in Northern Nigeria, a project that used a combination of insecticide spraying and prophylactic drug administration to attempt to eliminate malaria from an area of about 150 villages (Molineaux & Gramiccia 1980). The intervention failed to interrupt transmission, a disappointment that supported the view that control of mosquitoes (at least away from the human feeding site) was not the best way to reduce disease burden.

By the 1980s, advances in immunology and vaccination science offered the prospect that vector-borne diseases might be eliminated purely by interventions targeted at the humans. This had largely been the case with yellow fever, and there was enormous optimism about several malaria vaccines that got as far as large-scale field trials. Twenty years ago, Anderson & May (1991) in their magnum opus on human infectious disease epidemiology stated ‘Today, with the intense interest in the development of vaccines and the application of new molecular and biochemical techniques to the study of acquired immunity in humans, the research emphasis has changed from entomology to infections in humans’. But unfortunately, the malaria and filarial pathogens turned out to be very complex antigenic targets (Greenwood et al. 2008), as did dengue virus which is a complex of four strains and where disease results from the immunological response to multiple strain infections (Gubler 1998). Research on human interventions continues unabated, but by the turn of the century, there was less confidence that silver-bullet solutions would emerge, at least in the short- to medium-term. Instead, many workers argued that a more integrated approach was required, targeting both the pathogen in the human and the vector. This has led to renewed interest in the last ten years in vector biology in general and ecology in particular (e.g. Yohannes et al. 2005; Killeen et al. 2006; Mukabana et al. 2006).

There is a second reason why there has been a renewed interest in vector biology. Molecular advances over the last decade have told us a huge amount about the biology of mosquito vectors. Species of Anopheles and Aedes were some of the first insects after Drosophila to be sequenced (Holt et al. 2002; Nene et al. 2007), and currently 13 further Anopheles species alone are being sequenced ( The genomic resources now available on sites such as VectorBase ( and elsewhere provide unparalleled insight into cellular processes and, coupled with the availability of the Plasmodium and other disease organism genomes, allow detailed exploration of the interaction between the vector and pathogen. Genomic resources offer the prospect of the rational design of new insecticides and repellents, a better understanding of mosquito population structure of relevance to the design of control interventions, as well as the more radical possibility of the genetic modification of mosquitoes either to render them incapable of transmitting diseases or directly to affect their mortality or fecundity.

Vector control is applied population dynamics and hence is an aspect of applied ecology. In this article, I argue that a more explicitly ecological approach to mosquito biology may assist in the design of vector control programmes. I illustrate this argument chiefly but not exclusively using examples concerning the mosquitoes that transmit malaria in Africa. In the next section, I provide a brief and simplified account of Anopheles and Plasmodium biology and the main interventions available to reduce disease burden. The following section provides examples of some of innovative ways it may be possible to control mosquitoes utilising modern molecular biology in the next few decades. The final section considers these threats and opportunities and describes a series of challenges to mosquito vector ecology that if addressed may make a real contribution to reducing disease prevalence.

Man, malaria and mosquitoes

The pathogen

Several species of malaria in the genus Plasmodium, single-celled eukaryotes in the phylum Apicomplexa (Cavalier-Smith 1993), infect man of which the most important is Plasmodium falciparum, which is responsible for the majority of human deaths. A second species, Plasmodium vivax, is also widespread and causes a relapsing form of malaria and it is assumed far fewer deaths (Gilles, Warrell & Bruce-Chwatt 1993). The malaria pathogen enters humans through the bites of infectious mosquitoes and immediately colonises a liver cell. Here, it divides many times and eventually ruptures the cell producing a new form of the pathogen that infects red blood cells (Plasmodium vivax is able to relapse as unlike P. falciparum it has a dormant form that can persist in the liver for months to years). Once in the red blood cell, the Plasmodium again multiplies producing pathogen forms that can infect further red blood cells. Through cycles of infection and multiplication, often synchronised and causing bouts of fever, the Plasmodium can maintain itself in the circulatory system. However, some infected cells enter a different developmental pathway with the Plasmodium undergoing meiosis to produce male and female gametes that, still in the red blood cells, can be taken up by a feeding mosquito. The pathogen is relatively protected from the host immune system because it spends most of its time inside human cells. However, it can be removed by the spleen, and to protect itself from this, it expresses proteins on the human cell wall that cause the red blood cell to adhere to capillary sides. Much of the pathogenesis of malaria is caused by malaria-infected cells blocking small capillaries. Despite the protection afforded by the host cell, humans can mount an immune defence against the pathogen and most adults in high-prevalence countries have some immunity. The complexity of the pathogen–immune interaction has so far frustrated the development of an effective malaria vaccine.

Returning to the Plasmodium life cycle, when cells containing male and female gametes are ingested by the Anopheles mosquito, the gametes are released in the insect gut and mating occurs to form a mobile zygote. The zygote penetrates the gut wall where it forms a cyst in which the pathogen divides multiple times. A further mobile form is released from the cyst that migrates to the salivary gland, and it is this that is injected into the human at the time of a blood meal and that infects the liver cell. I should add that every different form of the malaria parasite has a long technical name that those of us who did classical biology courses were made to memorise – full details can be found in Gilles, Warrell & Bruce-Chwatt (1993) and an accessible introduction to malaria biology in Packard (2007).

The vector

The genus Anopheles is one of about forty genera of mosquitoes and the only one that can transmit malaria between humans (Harbach 2004). Of the c. 500 described and undescribed species of Anopheles, around 70 have been shown to be competent vectors of the human disease. Concentrating on Africa, the two main vectors are a complex of species referred to as Anopheles gambiae sensu lato and Anopheles funestus which is also probably a complex of species. Anopheles gambiae sensu lato is madeup of seven species that are morphological indistinguishable but which are largely genetically isolated: An. gambiae sensu strict and Anopheles arabiensis are the two most important vectors (Hay et al. 2010b; Sinka et al. 2010). Unfortunately, the complexity does not end there. It has been known for fifty years that An. gambiae consists of a series of chromosomal races amongst which partial barriers to gene flow exist. Chromosomally typing mosquitoes is time-consuming, and today two forms, ‘M’ and ‘S’, defined by molecular markers are routinely recorded, although these overlap with the chromosomal forms (Coluzzi et al. 1979; Turner, Hahn & Nuzhdin 2005). Very recently, a further molecular variant has been found commonly in larval collections made in West Africa, but these mosquitoes do not appear to enter houses like normal An. gambiae (Riehle et al. 2010). Understanding the ecology and evolution of the diversity of An. gambiae, arguably the insect that most impacts upon humans, is one of the ecological challenges I return to below.

Stepping back from the complexities of taxonomy and population structure, the basic biology of the major African vectors (and most other Anopheles) is rather similar (Clements 1992). Females emerge and mate and require one or more blood meals in order to mature eggs. They may also obtain energy by feeding on nectar. Most Anopheles species will feed from several potential host species, and the reason why several competent vectors are not more of a problem is that humans make up only a small fraction of their diet. What makes An. gambiae (and to a lesser extent Anopheles funestus) such an efficient vector, from the point of view of the pathogen, is that they are relatively specialised on humans. Anopheles gambiae in particular seems to have evolved to feed on humans in huts and other buildings.

After taking a blood meal, the laden and aerodynamically challenged mosquito has to rest to begin digestion, frequently on the walls of the building where it has fed. After this, it flies to a breeding site for oviposition. All mosquitoes have aquatic larvae and breed in different habits, An. gambiae sensu stricto typically in small ephermeral water patches such as hoof prints or tyre tracks, related species in slowly flowing water bodies, paddy fields, etc. Less common members of the gambiae complex are specialised in brackish water, or in saline springs. Because of their biology, mosquito numbers and force of infection are highly influenced by seasonal rainfall patterns.


The two main planks of malaria control today are treating the disease in humans and attacking the vector. Anti-malarial drugs have been known since the days of herbalists although many are now ineffective as the pathogen has evolved immunity. Currently, the major drug in use is artemisinin and its synthetic relatives and derivatives, a compound found in Artemisia annua and known to Chinese medicine for many centuries (Klayman 1985). The World Health Organisation recommends it as the primary treatment against P. falciparum malaria but only in combination with other drugs (Artemisinen Combination Therapy, ACT) that can mop up low levels of infection and can help prevent the spread of resistance. There have been a number of reports of artemisinin resistance, and were this to spread widely, it would be a major threat to control programmes as there are no major drugs that could replace it (White 2010).

Currently, the two major strategies for vector control are insecticide-treated (bed) nets (ITNs) and indoor residual spraying (IRS). Most African Anopheles bites occur at night, and so protecting individuals with bed nets obviously reduces disease. From the 1980s onwards, different programmes distributed bed nets treated with pyrethroid insecticide that killed rather than deflected mosquitoes and that under field conditions could remain active for six months. More recently, long-lasting insecticide(-treated bed) nets (LLIN) have been developed that may remain active for five years. A meta-analysis of ITN programmes found substantial reductions in mortality and disease prevalence due to nets (Lengeler 2004). Exactly how best to distribute nets, whether freely or at a low but still real cost, is an active area of economic and social science (Cohen & Dupas 2010). Indoor residual spraying (IRS) results in the death of mosquitoes, often as they rest up after feeding (Pluess et al. 2010). Unlike ITNs, a much broader range of insecticides can be used, and in many areas, DDT is still widely employed, especially where people live in unplastered huts. Sustained IRS campaigns have been shown to have significant effects on disease prevalence (Pluess et al. 2010).

A major worry for vector control is the spread of insecticide resistance (Abilio et al. 2011; Chanda et al. 2011). This is particularly true for bed nets as the insecticides used in ITNs are all pyrethroids and evolved resistance to one typically gives protection from the whole family (Hemingway & Ranson 2000). Mosquitoes encounter insecticides both as adults and as larvae, sometimes as part of deliberate mosquito control programmes but more often as a by-product of their use in agriculture.

The last decade has seen much greater efforts put into malaria control from both governments (for example the Global Fund against Aids Tuberculosis and Malaria (GFTAM) and the US President's Malaria Initiative launched in 2005) and non-governmental organisations (in particular the Bill and Melinda Gates Foundation). These efforts have resulted in significant declines in malaria burdens, globally and in Africa (Enayati & Hemingway 2010; O'Meara et al. 2010). The improved economic and governance performance of many low-income countries will also have helped as people above the poverty line who have access to functioning health services are far more likely to be able to withstand the disease. Investment in the ACT, ITN and IRS saves the lives of children today and in the absence of evolution might eventually lead to the eradication of the disease. But the evolution of resistance could render all these tools obsolete hence the need to invest in new strategies to take their place. Developing new drugs and possible vaccines to target Plasmodium is a major international research effort, as is the search for new insecticides and repellents, areas I will not explore further here. Instead, I will sketch some new potential strategies to target the vector before going on to discuss what we need to know about mosquito ecology to make them work.

New strategies to target the vector

Engineering mosquitoes so that they cannot transmit disease

Most diseases that are transmitted by mosquitoes go through a period of replication in the vector, which can thus be thought of as a secondary host. To do this requires an intimate relationship with the insect; for example, in the case of malaria, the mobile zygote must penetrate the gut wall and form a cyst and then produce forms that colonise the salivary gland. The pathogen is thus exposed to the mosquito's humoral and cellular immune system and requires a variety of cell and tissue recognition factors to orchestrate successful transmission. Strains of mosquitoes are known that are very poor at transmitting malaria, and these presumably either attack the pathogen or lack factors essential for Plasmodium.

Anopheles mosquitoes were successfully genetically transformed for the first time in 2000 (Catteruccia et al. 2000), and it has now been shown several times that mosquitoes can be genetically engineered such that their capacity to transmit malaria is removed or severely hampered. To give one early example, Ito et al. (2002) developed a small peptide that bound to the gut wall of the mosquito and prevented the mobile zygotes from invading and forming zygotes. They created an artificial gene with a promoter that caused it to be expressed in the mid-gut epithelium and a molecular signal motif that ensured it was exported to the right location. The gene was injected into mosquito eggs (as part of a suitable construct with additional DNA processing enzymes and marker genes), so that it became incorporated into the germ line. Adult female mosquitoes expressed the peptide in the correct location and were found to be very poor vectors of malaria. Adult female fitness, in as much as it is possible to assess in the laboratory, was not severely reduced by the inserted transgene.

Nearly ten years later, there are a range of possible strategies for reducing transmission competency by inserting genes as well as the possibility of reducing transmission by knocking out genes that are essential for the mosquito. But of course creating beneficial variants in the laboratory is very different from spreading them through field populations. This is obviously true for lethal or null reproductive variants, but is also true for manipulations that target the Plasmodium and are designed to have minimal direct effects on mosquito fitness. Mosquito and Plasmodium fitness are highly aligned, and one of the reasons why it is such an efficient disease is that the pathogen causes relatively little harm to its vector. Knocking out transmission does not lead to a large enough increase in mosquito fitness that the variant will spread naturally through a population by Darwinian selection at a rate useful for disease control (Sinkins & Gould 2006).

In the absence of passive spread, to be effective these variants need to be spread by man. There are two chief ways of doing this. The first is through mass release. There is a long history of the mass release of sterile male insects including mosquitoes (Benedict & Robinson 2003) with the aim of saturating the local population with infertile males that reduce the fecundity of their mates (SIT or sterile insect technique). SIT has worked best where the target insect population is relatively small and isolated although there are examples of its successful use on larger scales. A disadvantage of this technique is the need to sterilise large numbers of individuals using dangerous radiation or chemical sterilents, as well the difficulty of separating males and females before release. Modern variants of SIT have been proposed where the mosquito is engineered to express a female-specific lethal factor except in the presence of a chemical not found in nature (Thomas et al. 2000). The compound can be added to the insect's food allowing it to be reared normally in the laboratory until in the final generation the substance is withdrawn and only males are produced. If the males are released into the wild, their mates will produce no daughters and their sons will carry female-specific lethal gene. Although the construct dies out after a few generations, it is a more efficient form of SIT compared with releasing irradiated males, though it does carry the disadvantage of involving genetic manipulation (Alphey et al. 2010).

Mass release of engineered mosquitoes may have a role in the targeted elimination of some disease or vector populations but is unlikely to be a viable solution for disease control over large areas, particular in the lowest-income countries. An alternative is to spread the new gene or construct through a population using a genetic drive mechanism.

Certain genes spread in a population because they outcompete alternative alleles even though they reduce the fitness of the individual who carried. Bill Hamilton in his ground-breaking 1967 paper on sex ratio theory (Hamilton 1967) speculated that they might be useful in pest and vector management, and this is likely to be realisable in the next decade. Several different drive mechanisms are now under active consideration (Sinkins & Gould 2006), and I shall describe two that I think are nearest deployment in mosquitoes.


Endonucleases are enzymes that cut DNA at specific sites defined by nucleotide base sequences. They occur naturally in many organisms and have a variety of different functions, in particularly protecting cells from viral and other pathogen attack. A homing endonuclease gene (HEG) codes for a protein that recognises a c. 24 base pair sequence and causes a double-strand cut in the DNA (Stoddard 2005; Taylor & Stoddard 2012). HEGs have a particular trick that allows them to spread from rare. When uncommon, they are to a good approximation always in heterozygotes where one chromosome carries a HEG and the other chromosome lacks the gene but in its place has the recognition sequence (which appears nowhere else in the genome). When the HEG protein cuts the chromosome, the cell repairs it using the other chromosome as a template – in effect it copies the HEG onto the other chromosome and converts a heterozygote into a homozygote. This gives the gene an enormous advantage and modelling (Burt 2003; Deredec, Burt & Godfray 2008; Deredec, Godfray & Burt 2011) and now experiments (Windbichler et al. 2011; Klein et al. 2012) have shown that a HEG will quickly spread to fixation.

HEGs occur naturally in single-celled organisms including bacteria, yeast and algae raising the question of why, if they have such great fitness, they are not more abundant and found in multicellular species. Studies by Goddard & Burt (1999) provide an answer: HEGs have such a fitness advantage that they spread quickly to fixation but then are under no selection pressure to maintain their function. They just decay, and indeed it is possible to detect the ‘fossil’ sequences of these defunct HEGs by careful examination of yeast genomes. HEGs can only persist in the long term by jumping from species to species, their community dynamics determined by the rate of colonisation and decay (parameters estimated by Goddard & Burt by comparing live and dead HEG frequencies across yeast communities). The germ line of multicellular species is more protected than that of single-celled organism making colonisation by HEGs more difficult and their natural persistence in ‘higher’ organisms problematic.

Austin Burt (2003) first argued that HEGs might be used to control vector species. He envisaged this might happen in a number of different ways. First, a HEG could be engineered to recognise a site in a gene essential for the pathogen but of relatively low importance for the host (presuming, of course, that such genes exist) and placed in the genome opposite the recognition site. The HEG would spread to fixation replacing the targeted gene and rendering the mosquito unable to transmit the pathogen. Second, the same strategy could be employed, but with the targeted gene chosen to be essential for mosquito survival or reproduction. This of course has a massive effect on mosquito fitness, but genetic modelling shows that the HEG drive mechanism is so strong that it can spread even if has a crippling effect on mosquito survival and reproduction including causing population extinction (Deredec, Burt & Godfray 2008). A third strategy is not to knock out a gene but to ‘knock in’ a gene: to spread a novel desirable sequence through a population. Here, the target site can be chosen to minimise the effect on mosquito fitness, perhaps a site with little functional significance (natural HEGs often occur in non-transcribed introns). The desirable sequence is inserted beside the HEG and hitch hikes to fixation. In practice, there would be a real risk that the HEG and the new sequence become separated, but it might be possible to devise a construct that loses its HEG activity were this to happen. Finally, Burt suggested a fourth strategy that has no precursor in nature. Suppose a HEG was engineered to recognise a sequence (ideally several) on the X chromosome and then inserted onto the Y chromosome. If the HEG was placed under the control of a promoter that caused it to be expressed only at male meiosis then any X gametes would be destroyed so that the male would only produce Y gametes and thus all its progeny would be sons. In effect, the engineered Y chromosome would be at an advantage compared with the wild-type Y chromosome and would spread through the population even though it causes a substantial male sex ratio bias and a reduction in population growth rates. The exact magnitude of these effects depends on the details of mating physiology and population regulation, but modelling studies suggest they can be substantial.

The Bill and Melinda Gates Foundation is funding an interdisciplinary consortium led by Burt to explore whether HEGs can be used to help control malaria by targeting Anopheles (I lead the modelling component). The project includes protein structural chemists redesigning HEGs to target new sequences, bioinformaticians exploring new targets, mosquito molecular biologists inserting HEGs into Anopheles genomes and population biologists and malariologists concerned with deployment. A HEG with its recognition site has been inserted into a mosquito and in cage experiments rapidly increases in frequency (Windbichler et al. 2011). Progress has been made in redesigning HEGs to recognise new sequence targets (Ashworth et al. 2010) and inserting a HEG onto a Y chromosome to cut the X chromosome. Although challenges remain, these studies demonstrate substantial proof of principle and suggest that deployment will be possible in the not too distant future. We thus need to be able to address the ecological consequences of releasing these insects into the field.


In Euripides' drama, Medea killed her children with Jason in revenge for her abandonment. In the drama of flour beetle (Tribolium) reproduction, the children of a mother carrying the Medea element die unless they inherit her Medea element or it is present in the paternal genome (in this context, Medea stands for Maternally Expressed Dominent Embryonic Arrest; Stevens & Wade 1990; Wade & Beeman 1994). As the Medea element increases in fitness, an ever greater proportion of homozygous Medea-negative zygotes die increasing the relative fitness of the element which rises in frequency. Modelling shows that if there is a cost to Medea (reduced fecundity for example), a threshold frequency needs to be exceeded before the element begins to spread, but as Medea increases in frequency, it rapidly accelerates to fixation. Linking a desirable gene to a Medea element would thus be a way to drive it through a population.

How Medea functions in Tribolium is not known, but Chen et al. (2007) designed de novo an artificial Medea system in Drosophila that functions in the same way. For an embryo to develop, a gene called myd88 is essential. The artificial Medea element produces a small RNA molecular that binds to this gene's messenger RNA and causes its expression to be silenced. The RNA molecule has a germ line promoter and silences the expression of myd88 in all zygotes, whether or not they inherit the Medea element. However, the Medea element contains an antidote, a version of myd88 that is immune to RNA silencing. This is expressed in the zygote and if the embryo receives the antidote, either by inheriting it from its mother or through its father's sperm, it will survive. Chen et al. (2007) found that their artificial construct behaved as predicted and spread through cage populations of Drosophila. It also had little effect on the fitness of its host suggesting that there would not be a large threshold frequency to be exceeded before spread would occur.

In principle, this model of an artificial Medea system should work in mosquitoes, and homologues of the Drosophila genes have been identified in the Aedes and Anopheles genomes (Hay et al. 2010a). Work on creating a mosquito Medea system is underway, and this provides a second example (and there are others) of a drive system likely to be deployable in the next decade or so, again underlying the need to understand Anopheles ecology.

Ecological challenges

The renewed political focus on the eradication of malaria and other vector-borne diseases afflicting people in low-income countries, as well as the astounding advances in vector molecular biology, make this an exciting time for those hoping to control disease by targeting mosquitoes (Feachem & Sabot 2008). Although much can be performed using existing technologies such as insecticide-treated bed nets and indoor residual spraying, many of the most innovative ideas involve novel interventions affecting mosquito populations in new ways and/or involve genetic engineering that needs very careful regulatory appraisal and approval before deployment. For these new ideas to be implemented successfully and safely, we need a detailed and comprehensive understanding of the population and community ecology of disease-bearing mosquitoes in the wild. There is some truly wonderful vector ecology being pursued, but I would argue that the resources being invested in this field is not commensurate with those being put into molecular biology. Without vector ecology catching up, we are at risk of going down molecular blind alleys developing interventions that will not work in the field, and where viable control measures are developed, we may not have the knowledge to regulate them properly or deploy them most effectively in the field (see also Ferguson et al. 2010).

In the remainder of this section, I highlight a number of areas of vector ecology where greater insights would be desirable, as well as some new approaches afforded by recent technological advances. The list is somewhat partial and motivated by the ecology of malaria-transmitting mosquitoes, and other vector biologists might highlight different areas.


We are still relatively ignorant about the larval ecology of some of the most important malaria vectors. Consider An. gambiae in Africa. Extensive habitat surveys have shown the types of water bodies where it is most likely to breed (e.g. Fillinger et al. 2004; Minakawa et al. 2004, 2005; Mutuku et al. 2006a,b), but we know far less about the ecological processes that shape its ecology. Laboratory experiments have demonstrated the potential for larval competition, but extrapolating these results to the field must be made with some care – typically mosquitoes are maintained in the laboratory on aquarium fish food (and the microbial communities they support) and protected from most of the biotic and abiotic stresses they would experience in the field. Larvae are filter feeders, but it is still not clear exactly what type of micro-organisms they prefer to feed on. Mosquito densities do tend to be correlated with algal densities (Tuno et al. 2006), and their preference for sunny sites and the observation that shading reduces mosquito performance suggests algae may be favoured (Gimnig et al. 2001). However, the addition of yeast or maize pollen can also increase larval performance (Ye-Ebiyo et al. 2003).

Over much of its range An. gambiae has to withstand long dry seasons, which it does very well with populations rebounding very quickly when the rains come. But how they do this is somewhat of mystery (though see Omer & Cloudsley-Thompson 1970) – do they aestivate as adults or breed in the few water bodies that remain and then disperse widely? A recent study that had very carefully marked individual An. gambiae found, remarkably, a single individual that had clearly survived 7 months between wet seasons (Lehmann et al. 2010). This has led to a reappraisal of adult Anopheles biology, and this lone individual can probably claim to be the single most influential mosquito in the history of vector biology!

The effectiveness of HEGs and other novel interventions (see above) that are designed to knock down mosquito densities depends on the capacity of the mosquito population to increase when rare. The critical parameter is the population's intrinsic rate of increase, rm, which is difficult to estimate as it will vary both temporally and spatially. A greater availability of long-term population time series of vectors populations would greatly assist in estimating, rm. We found, when exploring HEG spread (Deredec, Godfray & Burt 2011), that the Garki project (Molineaux & Gramiccia 1980) now over 30 years old still provided some of best data (though see Russell et al. 2011).

Simple models of mosquito and malaria dynamics (most based on the Ross–Macdonald equations; Ross 1911; Macdonald 1957) often fail to explain observed epidemiological patterns. Many vector biologists believe that the assumption that all mosquitoes behave identically and that all hosts are effectively equivalent is at the route of this mismatch (e.g. Smith et al. 2010). If we understood these heterogeneities and incorporated them in our models, so the argument goes, we might do a better job of understanding the disease. We do know some exquisite details of mosquito behaviour and how they locate hosts, as well as differences amongst humans and animals in their attractiveness to mosquitoes (e.g. Takken & Knols 1999; Takken 2005). But more information is required about actual behaviours in the field, for example exactly how the mosquito responds to contact with an impregnated bed net. We also need new innovation in modelling: how can we incorporate more relevant biological detail without ending up with models that are so complex and highly parameterised that they fail to provide insight and understanding.

Population regulation

In designing vector control interventions, it is particularly important to understand where in the life cycle density dependence occurs, and what form it might take (Killeen, Fillinger & Knols 2002). Similarly models of mosquito population dynamics must include some density dependence if they are to represent a persistent interaction. Often the density dependence is implicit: the ur-model of malaria dynamics is the Ross–Macdonald equation that assumes mosquito recruit to the adult stage at a constant rate, which implies perfectly compensating larval density dependence (Macdonald 1957; Smith & McKenzie 2004). Other models that explicitly include density dependence also assume it occurs at the larval stage (e.g. Hancock & Godfray 2007). In the laboratory, larval mosquitoes do experience density-dependent competition, particularly manifest in a reduction in adult size (Lyimo, Takken & Koella 1992; Timmermann & Briegel 1993; Schneider, Takken & McCall 2000; Ng'habi et al. 2005). But what evidence do we have from the field that this is the case?

Working with An. gambiae larvae in their actual larval habitats is quite challenging, but much more progress can be made in artificial breeding sites that can be placed in the field where they are subject to most of the same biotic and abiotic stresses experienced by wild mosquitoes. Ned Walker's group have used this technique to demonstrate larval density dependence (Gimnig et al. 2002), and using similar methodology, we report in this issue further evidence that larval crowding can affect survival, size and development time (Muriu et al. 2013). White et al. (2011) quote unpublished data by Njunwa (1993) that also finds larval density dependence. Thus, there is mounting evidence that density dependence at the larval stage can at least in principle be critical for An. gambiae ecology. Related experiments in Aedes aegyptii have also demonstrated the importance of larval density dependence (Legros et al. 2009; Walsh et al. 2011). Experiments with this species are slightly more straightforward as they today naturally breed in artificial containers associated with man.

Further work is needed to explore the how different mosquito age classes interact (possible asymmetric competition) and in particular how density dependence may vary over the year as breeding site densities wax and wane. One still sometimes reads in the vector literature that mosquito populations are not subject to density dependence but instead are driven by availability of breeding sites, which is a misunderstanding though does reflect the importance of rainfall as a driver of vector dynamics. Finally, we need to understand more about the consequences of adult size for disease transmission. Because a mosquito must bite twice to transmit the disease, it is only relative old mosquitoes that are responsible for transmission (Lyimo & Koella 1992). Were larger individuals more likely to live longer, we might have a situation where mosquito control reduces vector densities, but this relaxes density dependence leading to fewer but larger adults that are better vectors.

Spatial processes

Vector biologists have always been interested in mosquito dispersal as it is a critical process determining disease spread as well as the success of any local mosquito eradication. But getting a good understanding of mosquito spatial ecology has been hampered by the difficulties of surveying larval breeding sites over large areas as well as estimating adult dispersal. The traditional way of monitoring dispersal is through mark-recapture experiments (e.g. Touré et al. 1998), but these are technically difficult (most marks affect mosquito fitness), very laborious (large numbers must be marked) and when they involve females have ethically issues (releases increase the risk of malaria transmission).

Modern technologies may provide new means of exploring dispersal and spatial ecology. Remote sensing techniques are now sufficiently sensitive that mosquito breeding habitats can be estimated in real-time with high accuracy (Machault et al. 2011). Marking of very large numbers of mosquitoes may be possible using very low concentrations of harmless elements. Rubidium and other elements spayed as salts onto breeding habitats can be taken up by larvae and retained through to the adult stage (Wilkins et al. 2007). This is not a new technique and has been used successfully to study other insect movements (van Steenwyk et al. 1978), but the drop in detection costs using modern mass spectroscopy may now make it feasible to do this at scale and so obtain estimates of the important tails of the distribution kernels.

Community ecology

Mosquitoes interact with many other organisms, which may affect their dynamics: they are eaten, parasitised and infected by natural enemies, and they compete with other animals and in particular other mosquito species. Regulatory agencies considering novel ways to target specific vector will want to know what the consequences might be of removing the species from the ecological web. Ecologists are interested in the role of competitors and natural enemies in population regulation while applied vector ecologists may look to diseases as biopesticides.

Mosquitoes are eaten by many generalist predators, but it seems unlikely that disease vectors constitute a sufficiently high proportion of their diet that their loss will make a great difference. Of more concern is the possibility that knocking out one species might increase the density of another through competitive release. This would be a particular problem if that species was another mosquito vector, perhaps more efficient that the first. In Africa where An. gambiae sensu lato and to a lesser extent An. funestus are responsible for the greater share of transmission, the emergence of a better vector (from the Plasmodium's standpoint) is unlikely. In other parts of the world where there are a greater range of vectors, the risks may be higher. Numerous field surveys demonstrate that different mosquito species are found in the same habitats, and laboratory studies show the potential for competition (Schneider, Takken & McCall 2000; Koenraadt & Takken 2003; Koenraadt et al. 2004), but more field and semi-field experiments are urgently required to explore this issue.


In terms of the number of people, it kills and sickens An. gambiae can claim to be the most dangerous animal on earth. Control of malaria through measures targeted at its vector is today seen as a critical component of the multipronged attack on malaria. This has been recognised by funding agencies through substantial investment in mosquito molecular biology and genomics that strengthen existing and provide novel interventions. But vector control is applied ecology, and although vector ecology is in rude health and advancing rapidly, it is a Cinderella compared with its molecular sisters. There is a real danger that our lack of knowledge of vector field ecology will hamper our efforts to capitalise on molecular advances and hinder the control of malaria and other vector-borne diseases.


Among the many people who have taught me important things about vector biology and control are Luke Alphey, Austin Burt, Flamminia Catterucia, Andrea Crisanti, Anne Deredec, Chris Dye, Gerry Killeen, Janet Hemingway, Penny Hancock, Ralph Harbach, Simon Hay, Ary Hoffmann, Jacob Koella, Richard Lane, Steve Lindsay, Charles Mbogo, Ellis McKenzie, Simon Muriu, Ace North, Scott O'Neill, Andrew Read, Scott Ritchie, David Rogers, Tom Scott, Steve Sinkins, Dave Smith, Willem Takken, Matt Thomas, Jeff Waage and the members of RAPIDD (Research and Policy in Infectious Disease Dynamics Programme of the Science and Technology Directorate, Department of Homeland Security, and the Fogarty International Center, NIH). I am grateful to all of them whether they agree with what I have written here or not! I would also like to thank the Wellcome Trust, the Foundation of the National Institutes of Health (Vector-Based Control of Transmission, part of Grand Challenges in Global Health initiative) and the European Community for funding.