Benchmarking vector arthropod culture: an example using the African malaria mosquito, Anopheles gambiae (Diptera: Culicidae)

The global public health significance of mosquitoes is manifested by continuing high mortality and morbidity rates due especially to dengue and malaria. Their public health importance results in them being a common subject of research, usually for the purpose of improving or developing control methods, including insecticides, repellents, genetic and biological methods. In addition, they are used similarly to a reagent for evaluation of disease control approaches that do not control the mosquito directly such as vaccines and drugs.

In the absence of reproducible laboratory mosquito culture, unnecessary and undesirable variability in the data collected due to variation in mosquito characteristics is likely to be introduced. For example, one easily measured characteristic, size, has been reported to affect numerous traits relevant to disease transmission, including host seeking and repellence by DEET [1], susceptibility to arboviruses [2], longevity [3], dispersal [4], and fecundity [5]. It is well established that larval culture is the strongest determinant of adult size, fecundity and longevity [6‐8]. The weight of diet available/larva is particularly important [8].

Although culturing by an experienced and conscientious person can produce predictable outcomes, an alternative is to systematize the culture methods, thus creating a more portable technique that does not depend upon the skill of an individual to such an extent. A standard operating procedure (SOP) that specifies the diet to be used, the amounts, the larval density, containers, amount and type of water, and temperature provides such a method. Accurate use of SOPs results in reproducible outcomes and possibly standardizes outcomes of various research activities. Methods for consistent production of mosquitoes have been reported, usually in the context of mass production, where rearing must necessarily be systemized (for examples see [9, 10]).

Consistency, however, does not provide any measure of the intrinsic quality of mosquitoes. The same mosquito strains consistently produced within various laboratories may differ greatly among them. An objective external standard of quality would allow assessment and comparison of methods used to culture mosquitoes in different laboratories. One such method is to compare laboratory-cultured mosquitoes with the wild-type population. For example, the size of mosquitoes destined for field release has been compared with those collected from the field [11‐13].

A systematic method to assess a rearing procedure, and to standardize arthropod culture methods between facilities aiming to rear the same strain of arthropods with the same diet and obtain comparable results, would be a valuable addition to these efforts. A simple procedure is therefore proposed to assess insect culture, particularly for mosquitoes: benchmarking. The benchmarking method implicitly reflects the assumption that, under fixed extrinsic conditions such as a particular diet, temperature and density, there are genetically and physiologically determined limits to growth rate and size and where survival is assumed to be determined largely by the specific characteristics of the culture methods but not by intrinsic species characteristics.

Insectaries where mosquitoes were held averaged 27.6 °C (±0.01), 81.9 % RH (±0.09), (95 % CIs of the mean). A 12:12 h light: dark schedule was employed with a 30 min simulated dawn and dusk. On the day of hatching, 16 first instar larvae (L1s) of the G3 strain (obtained from the Malaria Research and Reference Reagent Resource Center) were counted into standard polystyrene 90-mm Petri dishes containing 30 ml of a standard rearing water (Milli-Q, Integral Water Purification System, Darnstadt, Germany) containing 0.3 g/l artificial pond salts (Tonic Pond Salts, Aquatics, UK). Each was given 640 μl of either 0.5, 1.0 or 2.0 %w/v diet containing 2:2:1 by weight tuna meal (Progressive Baiting, Dietenhofen, Germany), liver powder (Now Foods, Bloomingdale, IL, USA) and Vanderzant vitamin mix (BioServ, Frenchtown, NJ, USA) [14] in Milli-Q water. Dishes were given 640 μl of diet on alternate days. Diet concentrations were chosen based on previous experience as providing low, medium and high levels for Anopheles arabiensis [8] (Table 1). Four dishes of each combination of larval concentration and diet were established.

Pupae were collected daily, their sex determined by examination of terminalia and then they were pooled by diet level and transferred to insectary cages. Adults were provided with a solution of 10 % sucrose and 0.1 % (both w/v) methylparaben in Milli-Q water. Dead adults were stored, the right wing was removed when not damaged, or alternatively the left wing, and dry mounted on a microscope slide using double-stick transparent adhesive tape. Wing length was measured as the distance from the axial incision to the R4+5 vein excluding the fringe seta using ImageJ software [15].

The SOP is provided as Additional file 1. Briefly, it uses the same diet and a similar number of larvae/ml of water as the Petri dish experiments, one larva/2 ml water. Larval development time was measured from the day of hatching to the day of pupation. The SOP feeding schedule uses 2 %w/v of the same diet as used for the Petri dishes, but is fed in volumes which increase, beginning with 5 ml on day 0 (hatching at a density up to 1000 L1/500 ml in a 35 × 25 × 8-cm plastic tray), 5 ml on day 1 when the larvae are reduced in density to 250/tray, and then given 0, 5, 7, 10, and 10–12 ml on days 3–7, respectively. On day 7 and thereafter, the amount of food is adjusted according to a judgement of the number of larvae remaining (Table 1).

All statistical analyses were performed using R 3.0.1 [16]. Analysis of variance was used for the wing length data to assess the influence of main effects (diet level, sex, experiment) and their interactions. The significance of all terms was assessed by deletion testing from an initial maximal model. Larval duration (the number of days from hatch to pupation) was measured for all individuals reaching pupation. These measures were nested in the dish from which they came, and to avoid pseudoreplication a mixed-effects model using the R package ‘lme4’ was used to account for the random effect of ‘dish’. The proportion of each replicate (dish) pupating was analysed by quasi-binomial general linear model (GLM) to account for the overdispersion found in these data.

The proportions of L1s reaching the pupal stage differed with diet level (Table 2, Fig. 1a, [17]). The proportions of pupae obtained with the SOP did not differ from that of either high (F = 0.45, df = 1,23, p = 0.51) or medium (F = 1.9, df = 1,23, p = 0.17) diets and these did not differ from each other (F = 0.29, df = 1,23, p = 0.58). The low diet level resulted in the lowest pupal production relative to the other three combined (F = 20.5, df = 2,24, p < 0.001). Pupation when using the SOP usually commenced on day 7.

Similar trends were observed in the duration of the larval stage (Fig. 1b, [17]). There was variation in larval duration as a function of diet (L ratio = 40.2, df = 3,6, p < 0.001): the SOP and the high level diet did not differ (L ratio = 1.7, df = 5,6, p = 0.18), nor was the medium diet different from these (L ratio = 2.2, df = 4,5, p = 0.13). Only the low level diet resulted in a longer larval stage (L ratio = 36.2, df = 3,4, p < 0.001).

Wing length was used as an indicator of adult body size. Wing length was continuous and normally distributed. The explanatory variables were categorical: the sex of the mosquito, the experimental runs and the four diet levels (high, medium, low and SOP). Analysis of variance was appropriate and a maximal model was fit to the data to assess the influence of main effects and their interactions. Initially, the maximal model included both sexes, but as male and female wing length are known to differ, a fact confirmed by these data, the sex-specific responses to diet and experiment were examined separately.

The experimental-runs vector was partially confounded with diet levels as the wing measurements for the low, medium and high diet came from two separate experimental runs and those of SOP from standard culturing. A separate ANOVA asked whether the wing length differed between experimental runs. It did not (F = 2.04, df = 1291, p = 0.15) and this factor was excluded from subsequent analyses.

There was a significant interaction between diet level and sex (F = 2.7, df = 286,289, p < 0.05) indicating that the effect varied between males and females. Males responded to, but were less responsive to, the high diet than were females (Fig. 2).

For females there was an overall effect of larval culture, and wing length increased with diet quantity (F = 35.4, df = 3143, p < 0.001). Individuals fed on the low diet had shorter wings than those fed on a medium diet (F = 11.2, df = 1143, p < 0.01), and the high diet led to longer wings than medium (F = 49.3, df = 1143, p < 0.001). The SOP feeding regime gave rise to wing lengths between those of the medium and high diet but distinct from both (SOP: medium, F = 13.4, df = 1143, p < 0.001, SOP: high, F = 18.5, df = 1143, p < 0.001) (Fig. 2a, [17]).

Male wing length also varied with diet level (F = 39.1, df = 3143, p < 0.001). Males fed the medium diet had longer wings than those fed the low diet (F = 23.3, df = 1143, p < 0.001), and those fed the high diet had longer wings still (F = 25.8, df = 1143, p < 0.001). The SOP gave rise to wing lengths greater than those of the medium diet (F = 28.5, df = 1143, p < 0.001) and indistinguishable from the high diet (F = 0.67, df = 1143, p = 0.41) (Fig. 2b, [17]).

Differences, if supported by further experiments, illustrate that the conditions experienced in small containers differ in some way not explained simply by larval and food density. The amount of food provided in the SOP corresponds to providing 3.8 mg of food per larva during the first seven days, as compared to the 3.2 mg per larva provided during the same period in the high diet Petri dishes. Even though the SOP provided more total diet during the first seven days than the high-level Petri dishes, the females produced were smaller, though female size was more consistent under the SOP conditions, possibly due to the more uniform conditions resulting from increased scale. Even considering that the diet amounts reported in Methods were based on the starting number of larvae rather than those that actually survived to pupation, which differed between the high and SOP experiments, the amount of food provided was still higher (4.2 and 3.7 mg/pupa for the SOP and high Petri dish, respectively). Thus, the reduced female size in the SOP experiment may reflect an effect of the relatively smaller amount of food provided early in development. Possibly, the larvae cultured under the SOP would have benefitted from a larger proportion of the total food upon hatching or on day 1 than was provided. This notion, that not simply the total amount of diet but the timing of its provision affects development, may be supported by the preliminary observations made when scaling up the Petri dish experiments to trays.

Using development rate as an outcome in the absence of wing length or some measure of size for mosquitoes is not advised. In nature, larvae may develop in temporary water sources under conditions of limited nutrition, leading to the evolution of strategies to balance increased size and shorter development time, a trade-off which is likely sex-specific. The relationship between different developmental parameters is, therefore, not always linear (discussed in detail in [18]). Gilles et al. ([8] Fig. 3) demonstrated that larval development rates are maximal at diet levels above which significant increases in wing length are still possible. Therefore, development rate alone could be maximized under conditions that still gave significantly smaller adults than could be obtained.

The method used here does not provide an absolute benchmark. It is possible that other diets or culture conditions could produce a wider range of outcomes, and variation between rearing conditions in different facilities will alter the results. For example, Kivuyo et al. [19] generally observed larval durations much longer than those observed in these benchmarking experiments, and a much greater disparity in pupation rates, when testing new larval diets for An. gambiae. This could be explained by the different temperature and relative humidity in which the experiments were conducted. In contrast, similar feeding regimes have led to a similar range of male wing lengths in different hands [20]. Phelan and Rotiberg [18] recorded very similar values for these two parameters to the current study, in an experiment to determine the effects of food availability, water depth and temperature, despite the differences in culture methods and diet used. The mean days to pupation reported here for the medium, high and SOP experiments are similar to that of the best diet used at the lowest larval concentration by Kivuyo et al. [19] and Damiens et al. [14] and Gilles et al. [8] for An. arabiensis, seven to eight days. Similarly, the wing lengths observed at the medium and high levels were similar to those observed using the same diet by Damiens et al. [14] who reported means of 2903 and 2783 μ for females and males, respectively.

When alternative diets or culture conditions are tested, they can in turn be benchmarked using the method described here and used to expand the array of measured responses. Providing a fixed amount of diet over the course of larval development may be more effective, or different larval densities may result in faster development or larger size. In contrast, the specific diet used was developed for culture of the An. gambiae sibling species, An. arabiensis, by an optimization protocol [8] and is believed to be of high quality, so further increases in development rate, survival and size in this case may be difficult to achieve by using a different diet alone.

By culturing insects under a range of conditions that achieve the intrinsic upper and lower limits of development rate and size, one can determine this range for a given diet, temperature and strain. By comparing the outcomes from an SOP with the full range of measured parameters, an estimate of the quality of the insects produced by the SOP can be determined. It would be useful for a rearing or research facility to benchmark their SOP using a given strain, diet and temperature. Culture methods could thus be objectively and quantitatively evaluated and compared.