Of all the life-history traits that present difficulty in the transition to the laboratory, mating is the most problematic 1 but is essential to any attempt to establish a colony. This character is at the heart of successful SIT operations as well, yet little is known about specific environmental conditions that promote it 2.

Although most major Anopheles malaria vector species have been colonised for laboratory rearing (see additional file 1), for SIT applications it is logical to ask, "Would these males mate with wild females and at what rate relative to wild males?" The quantitative measure of this trait is termed competitiveness 3. Population bottlenecks, genetic drift, deliberate and inadvertent selection, rearing, sterilisation and release methods can all reduce competitiveness. It is generally appreciated that hybrid vigour contributes to improvement of traits of many organisms, and it is possible that crossing and colony maintenance to achieve similar effects in mosquitoes would be useful. Adult size, and the related characters fecundity and wing length, were significantly increased by hybridizing two strains of Anopheles gambiae thus providing promising support for this approach 4. It will now be important to determine whether these observations translate into increased mating competitiveness. While it is widely assumed that maintaining natural genotypes or increasing heterozygosity per se are desirable to obtain competitive males, a literature review identified no experiments with mosquitoes in which this has been explicitly tested. Laboratory experiments in which heterozygosity was incidentally increased did not demonstrate an increase in competitiveness as a result 5.

Differences between mass-produced and natural mosquitoes begin to accumulate during colonisation. Such effects of colonisation and inbreeding on the genetic composition of the colony are commonly viewed from a probabilistic standpoint using founder, drift and selection models. From this perspective, colonies become increasingly homogeneous entities that are genetically very different from wild populations and whose competitiveness is assumed to decline. While this view describes general trends 6, there are countering forces whose genetic bases are poorly understood, particularly the role of lethals and natural recombination suppressors (discussed previously 7). As an example of such effects, two inbred strains of Aedes triseriatus undergoing full-sib mating for at least 12 generations were analysed for isozyme polymorphism 8. In spite of intense inbreeding, a significantly higher-than-expected level of polymorphism was observed at several loci. The authors concluded that heterozygosity itself is a strongly selected character in inbred populations and that the mechanism in this case was the presence of recessive lethals resulting in balancer systems. While the effort to accomplish genetic homogenisation that was conducted in other studies 9 was not as intense, heterozygosity was maintained at a level near, or higher than, the parental material, possibly again reflecting selection for heterozygosity.

A similar natural phenomenon has been observed in anophelines, including members of the Anopheles gambiae complex. Unexpectedly high levels of paracentric inversion polymorphism have been observed in field and laboratory populations of Anopheles gambiae s.s. 10. While this may be co-selected by insecticide resistance alleles linked to the inversions, heterosis independent of such selection is likely. A heterozygous chromosomal rearrangement present in excess of the expected frequency in field and laboratory populations of North American Anopheles quadrimaculatus 11 again suggests that natural balancers may be common. Until the genetic mechanisms and frequency of such balancing systems is determined, a cautious approach to anopheline colonisation will be to maintain and increase polymorphism insofar as is practical. In order to provide experiment-based direction on this subject, a useful research initiative would be to conduct experiments that would specifically address this question. This information would be valuable for those implementing SIT and other genetic control projects.

Concerns about behavioural and genetic incompatibility have resulted in a default approach to select a release strain for colonisation that originated in the target population locale. While entirely defensible, this approach ignores possible beneficial effects of heterosis that might be obtained by producing geographic hybrids and choosing strains with intrinsically superior competitiveness. Several factors of importance to production and mating competitiveness have a genetic component: longevity and fecundity particularly appear to be amenable to manipulation by strain selection and crossing 12. Until laboratory surrogate indicators that accurately predict field performance are devised, it will be necessary to conduct colonisation and competition experiments on a more ambitious scale than has been performed previously.

Mosquitoes clearly mate satisfactorily in the field, but simulating this in the laboratory is difficult, presumably due to differences between the environment of the laboratory and natural one. Strenuous efforts to simulate natural lighting, temperature, humidity, presence of other swarming mosquitoes, and spaces have been made, but it is difficult to conclude how effective many measures are. Marchand concluded that an artificial horizon stimulated swarming behaviour of An. gambiae and Anopheles arabiensis, but that ground markers in the cage had no noticeable effect 13. Gradual light changes to simulate sunset and sunrise are also useful 2, however, knowledge of which of the multitude of factors being simulated is necessary is often lacking during colonisation and stock-keeping, and it is even less clear as to whether they ultimately affect the performance of the mosquitoes that are colonised. Comparison of semi-natural vs. laboratory colonisation of Culex tarsalis was performed in either a constant or variable environment beginning with identical field-derived cohorts 9. In an attempt to simulate a more natural environment, variation in temperature, oviposition containers, cage size, and resting sites was provided to one group while constant conditions were offered to the other. The experimental question was, "Which condition resulted in mosquitoes that were more similar to wild type after 12 generations?" Indicators included fecundity, development rate, size, mating competitiveness and genetic similarity to the founder material. While the authors concluded that the variable environment resulted in a colony genetically more similar to the wild type, both had similar high levels of heterozygosity, and laboratory adaptation still occurred, albeit at a slower rate in the natural simulation. Until experiments identify specific environmental factors that minimize directional selection, mosquito SIT programmes must proceed without clear guidance of which environmental factors must be manipulated to avoid it.

A mosquito species' propensity to mate in confined (i.e. laboratory cages) vs. open spaces is termed stenogamy vs. eurygamy. Preliminary evidence from Anopheles maculipennis complex mosquitoes suggests that this propensity is controlled by the Y chromosome 14, but similar studies to confirm this have not been performed in other Anopheles. If the following two propositions are true, (1) the Y chromosome does in fact control stenogamy/eurygamy, and (2) these characteristics are important for mating competitiveness; then it is critical that genetic sexing strains (GSS) be created using a Y chromosome that carries traits necessary for mating with the target population. Because traditional anopheline sexing strains have been created by linking a dominant marker to the Y chromosome, introgression of Y chromosomes - and stenogamy/eurygamy - from a wild target population into such a sex-separation strain will be impossible because there is no opportunity for recombination between Y chromosomes. Encouraging, but limited, experience in El Salvador with the Anopheles albimanus MACHO sex separation strain confirmed field competitiveness, even though no preliminary tests assured competitiveness of the stock from which it was created 15. Classical sex-separation strains have disadvantages of reduced fertility and breakdown due to recombination, but they are still useful because they are not subject to special regulations imposed upon transgenic insects. The disadvantages listed above are characteristics over which transgenic mosquitoes might have advantages 1617. For example, they would not have reduced fertility, and they would allow introgression of wild genotypes to a greater extent than a classical strain in which a large proportion of the chromosomes experience reduced recombination due to rearrangements.

Beside competitiveness considerations, the introduction or modification of genotypes may be required for introducing specific characters into laboratory and production colonies. Outbreeding is the most apparent solution. This strategy rapidly introduced wild genotypes into Culex nigripalpis 18, however, it required the use of field-collected males crossed to colonised females. Because wild males will mate colonised females more readily than wild females will mate colonised males 2, this strategy cannot be implemented in a single generation as it is inimical to introgression of wild material into classical Y-translocation sexing strains in Anopheles.

Alternatively, however, an intermediate strain produced by repeated crossing of field-collected males to colony females could provide suitable progeny females to which males from the sexing strain could be mated. Repeated introduction of wild material was employed to create a vigorous wild-type An. albimanus strain, CAMPO, but without genetic assessment during the El Salvador SIT releases 15.

A single GSS selection genotype has been used for the control of Mediterranean fruit flies Ceratitis capitata globally in area-wide SIT programmes releasing four billion sterile males per week 19. This is based on the use of classic genetic approaches in which a male-linked chromosomal translocation links wild-type alleles of a pupal colour mutation and normal temperature sensitivity to the male sex chromosome. Females carry a white pupa mutation and are temperature sensitive.

Like the low frequency of recombination between the insecticide resistance marker and the Y chromosome seen in the MACHO strain of An. albimanus (which results in resistant females), C. capitata GSS are not absolutely stable and require elimination of unwanted recombinants while not interfering with production. The filter rearing system (FRS) is a management strategy that was designed to remove the products of genetic recombination and hence maintain the integrity of the GSS 20. The FRS is based on the use of a relatively small colony from which recombinants are removed and from which eggs are collected. Through a process of amplification for several generations, sufficient male insects are produced for sterilisation and release. Depending on the stability (i.e. frequency of recombination events), the FRS colony should be purified each generation, and it must be large enough to produce sufficient offspring to support itself and to initiate the colony amplification. Because the FRS is a unidirectional system of production, it also provides a generic system to control additional quality aspects. For example, a new small colony that preserves sexing characteristics refreshed with wild genotypes can be introduced on a regular basis to replace the broodstock to prevent accumulation of "lab-adapted genotypes" or to otherwise modify the genetic composition of the colony. The FRS can also be maintained under less stressful conditions in order to maintain important quality traits.

Effective management systems are absolutely critical for the success of production and distribution of millions of sterile mosquitoes. The statement 22 that, "...the failure of area-wide programmes is usually due to poor management and inadequate public support and not to poor technology." is a widely recognized truth. Good management requires methods similar to those used in business and industrial mass-production - disciplines in which few scientists who are charged with this responsibility have any training. Notably, the two largest release programmes quickly adopted such management systems and carefully controlled production levels and quality analysis. Most notably the World Health Organization/Indian Council of Medical Research (WHO/ICMR) programme attributed its successes to reliance on the Program Evaluation and Review Technique (PERT) 23. This method is still widely used and provides a useful model for programme management. The production of An. albimanus in El Salvador required careful planning and analysis of production systems which have been well described 24. However, no overall management system of that project has been described. Of great value to wider application of mosquito SIT will be development of a standard model of management system for the operation of facilities worldwide.

Because previous mass-rearing efforts have essentially increased the scale of laboratory methods, integrated production systems are an area in which rapid dramatic improvements can be expected. Several infrastructure improvements could significantly improve the quality of rearing: circulating water systems with continuous quality monitoring and waste removal, continuous larval/pupal separation and sterilisation systems, measures to reduce and prevent the development of pathogens, and feeding systems that maintain optimal diet availability.

In spite of the demands and scale of mosquito release programmes, it is surprising that the infrastructure differed mainly in number from that used in laboratories for small scale rearing. While the WHO/ICMR used a centralized aquatic stage collection system, the El Salvador An. albimanus SIT project relied on individual rack-mounted trays. Both programmes utilized pupal separation methods and adult cages nearly identical to those used in research laboratories. Neither mechanized nor continuous methods for blood feeding, egg collection, larval feeding or pupal separation were implemented. Useful models exist however: on a small scale, a circulating system that included biological water conditioning and automatic feeding for rearing of Anopheles stephensi provides an example of how such systems can be created 25.

The economic considerations that limited the development of mechanized specialized systems decades ago are now a smaller obstacle because of the availability of inexpensive personal computers, software, control devices, and solid-state sensors. In the final analysis of the utility of SIT, economics of production will make developing these systems essential. The low cost of labour in disease endemic countries should not be used to justify a lack of technology development. Investing in the training and infrastructure of the persons working in the factory and the host country will be better served by transferring efficient advanced technical skills rather than instruction in menial labour. Furthermore, strikes, absenteeism, and inconsistent performance are less likely to interfere with a production system that does not depend excessively on numerous personnel. Finally, some programmes have found it difficult to decrease staff sizes when increased efficiencies and demand resulted in diminished labour needs. Starting with as small a staff as possible reduces these problems.

Larval rearing conditions have a direct, and often irreversible, effect on adult traits therefore a clearly defined diet is a priority. Developing such a diet is not as straightforward as it might appear and is usually accomplished by trial and error. Because mass rearing has not been performed under aseptic conditions and mosquito larvae are opportunistic feeders, microbes constitute a large portion of their diet, and no truly defined diet for mosquitoes has been used in a large-scale system. Regardless of the fact that controlled conditions are desirable for mass production, and aseptic defined-diet rearing of mosquitoes would help accomplish this, it will likely not be practical on a large scale.

A chemically defined diet was used to successfully rear several culicines 26, but was not successful for Anopheles freeborni 27, though small-scale sterile anopheline rearing is possible 28. In an interesting example of the hunter becoming the hunted, Munderloh et al. 29 were able to rear An. stephensi larvae on cultured cells of a mosquito that itself preys on mosquito larvae, Toxorhynchites. An ideal diet would provide adequate nutritional components that positively influence rearing conditions and insect quality, is inexpensive, globally available, and of consistent quality. Those that are used for smaller scale culture such as TetraMin™ Flake Food are of high quality but of high cost and would not be desirable for mass-rearing. As an indicator of the amount of diet required - and the significance of cost - we have estimated that approximately 2-3 kg of larval diet per day will be required for production of one million males per day. Mass rearing diets have consisted of inexpensive ground dried animal foods (sometimes defatted to prevent 'scumming'), cereals, yeast, beef liver powder - and of course the happenstance unidentified microorganisms. Conversely, cat and monkey chow are often used, but are neither designed from mosquito larvae nor can they be obtained in the same form globally. Successful combinations were determined empirically, but when used under septic conditions, it is impossible to say to what extent the diets were useful because either they fed the larvae directly, supported the microorganisms on which the larvae fed or both. In response to these concerns, the IAEA mosquito SIT programme is currently testing various ingredients that we feel meet the necessary requirements described above e.g. beef liver powder, yeast, fish meal and squid liver meal (unpublished data).

In some laboratories pupae are not separated from larvae but rather adults are collected directly from larval trays, a method that will likely not be practical for mass rearing. However, unless rearing methods are devised that result in synchronous pupation of a large proportion of larvae in a single day or two, we assume that in mass-rearing systems, pupae will be separated from larvae prior to adult emergence. Fortunately, size, behaviour, and buoyancy differences between larvae and pupae allow selection by means that are readily amenable to mass production and mechanization. To stimulate thinking toward this goal, a conceptual separator is shown that could operate continuously rather than in a batch-wise manner (Figure 1). The basis of the separation is the greater diameter of pupae relative to larvae: since female pupae are larger than males in some species, it might also be suitable for sex separation. This device would have several advantages over previous methods: larvae and pupae could be automatically counted in the stream for redistribution to rearing containers and release, partial or complete water and food changes could be performed automatically, and separations could be performed several times daily from a cohort to provide precisely staged pupae for irradiation.

Sucrose, glucose and fructose solutions are common adult diets, though fruits are sometimes provided, such as raisins. The use of anti-oxidants in diets to increase longevity and fecundity e.g. 3637 should be considered and implemented as a routine measure. One report described a complex vitamin mixture which was offered to mosquitoes during colonisation, but no information was provided about its effectiveness 38. The simplicity of designing and conducting survival and fecundity experiments to test various supplements should lead to greatly improved adult diets. A common food preservative for human diets, methylparaben, has been shown to increase adult longevity of anophelines 39 and was adopted by that programme for routine use. For mass production though, additional information on fecundity would be needed prior to implementation.

For blood-feeding a large number of female mosquitoes, defibrinated blood or blood treated with an anti-coagulant, is provided via a membrane feeding system. Fecundity is reduced in these systems, and the absence of human pathogens in the blood meal must be assured. An ideal blood meal would be totally synthetic, but continued reliance on treated natural blood should not prevent production of competitive adults. For relatively small amounts (e.g. several litres per week), blood can be collected sterile in veterinary bags from cattle and used fresh or after refrigeration. Larger volumes would be problematic as measures to sterilise the blood after collection in an abattoir might be necessary. If this is required, methods for preparation and storage of bovine blood have been developed and tested for tsetse rearing 4041. Commercial companies will supply rather large volumes of aseptically collected animal blood but at a relatively high cost.

Adult rearing usually consists of a larger number and size of cages than those used in numerous research laboratories with little consideration of adult biology. A multitude of concepts should be reconsidered for the next generation of adult rearing cages and tested by experimentation. These are the needs for mating arenas, sanitation, egg collection, blood-feeding, and adult resting behaviour. A concept for such a cage is shown in Figure 2. Several features of such a cage might meet these requirements and be consistent with large-scale production. Similar considerations to those that resulted in this concept cage should be applied to other prototypes that are capable of holding large numbers yet require little maintenance. Preliminary testing and modification of this cage has indeed demonstrated mating at the artificial horizon (Figure 3), but we have found (unpublished data) that extensive modification to the egg collection device is needed. The removable paper bottom also functions well to prevent waste accumulation, and we have retained this feature in a second prototype, which is stocked with 40,000 adults. Like the first version, it retains extensive adult daytime-resting sites, which An. arabiensis prefer.

Interventions for disease control have consisted largely of routine hygiene, bleaching trays and equipment, use of disposable items etc. No air and water filtration sufficient to prevent contamination has been used. If selection for greater competitiveness or ease of production in the factory occurs at the expense of immuno-competence, this will become of greater concern.

In previous mosquito release programmes, measures for disease control have been conducted without knowledge of potential pathogens in the colony. Given the paucity of information, past experiences with pathogens in mosquito mass-rearing systems are encouraging. For instance, mass rearing of An. albimanus in the early stages in El Salvador was fraught with a high number of "bad trays" (10%-16%) from which few or no pupae were harvested. Often these trays exhibited lack of synchrony in development and foul-smelling water suggesting that improper amounts of diet were provided. Later, when the rearing was increased to one million male pupae per day, losses were reduced to around 1%. In both circumstances, eggs were cleaned simply by washing with water, which apparently sufficiently reduced Nosema and/or other surface contamination.

As genes related to growth and reproduction may be independent of genes related to the immune system, the routine mass-rearing process in artificial conditions could lead to a decrease of immune fitness. In shrimp aquaculture, where disease resistance is a crucial component of productivity and consistency, domestication and breeding programmes consider the immune effectors as markers of individual selection. The selection of individuals with the highest immune capacity is based on real-time RT-PCR or microarray analysis of mRNA of the main immune genes such as penaeidins, hemocyanin, phenoloxidase, superoxide dismutase, LPS-binding protein and β-glucan-binding protein. Such a strategy aims not only at avoiding the decrease of immune response but also at increasing the general non-specific resistance against various types of pathogens.

In mosquitoes, several types of viruses have been described which are related to the baculoviruses 4243, iridoviruses 444546, and parvoviruses particularly in Anopheles and Aedes spp. 47. It must be considered that founder field-collected mosquito specimens may be infected - yet healthy - carriers of mosquito pathogens, some of which might be vertically transmitted. As in shrimp aquaculture, the founders could be analysed through e.g. nested-PCR based on highly conserved sequences in order to eliminate asymptomatic carriers. Routine analysis could also be performed in order to confirm that the production continues to be virus-free.

In Anopheles specifically, several intracellular pathogens have been described, such as Rickettsia 48 and Spiroplasma 49. Recent information in shrimp aquaculture has shown that such pathogens are efficiently transmitted vertically and horizontally, and when these occur in high prevalence can result in mortality and severe losses of fecundity and vigour. Consequently, control of intracellular bacteria is a priority.

Large size of individuals is a well-proven indicator of general nutritional status including energy reserves available for dispersion. Adult male size has also been correlated with successful mating in Cochliomyia hominivorax, being even more important in one study than strain type 50. Evidence is inconclusive to support its prediction of mating success in mosquitoes. Large females of An. gambiae have been observed to be preferred mates 51, but it has not been demonstrated whether females preferentially mate larger males: observations for An. gambiae suggest male size is irrelevant 52. Until the general observation that male size is relevant in Diptera is discounted for anophelines, production methods must consider that not merely consistent size, but a particular size itself may be a valuable goal.

For fruit flies standard quality control protocols have been established for mass-reared flies e.g. for pupal size and weight, emergence, flight ability, longevity, sex ratio, and mating propensity 195354. Comparison of results of the same tests from different mass-rearing facilities has allowed the establishment of standards for each of the parameters for several fruit fly species with, and without, irradiation. Though such quality control tests do not directly measure the competitiveness of the insect, they do provide standard indicators of the consistency of the production process. When linked to male releases that are suppressing populations, this is a valuable quality control measure.