This study found intra-specific variation of sperm length in both A. gambiae and A. stephensi (Figure 1). Differences among A. gambiae males accounted for a significant portion of the variance in sperm length in all three experiments (Table 1). In experiment 2, there was no genetic variation in sperm length in the Keele population of A. gambiae (Table 1), but sperm length was genetically correlated with male wing length (Figure 3). According to the GLM (Table 3, Figure 4), the three sire fitness traits: insemination success, sperm motility, and oviposition success all decreased significantly with the sire family's mean sperm length, but there was no effect of the sire family's mean wing length. One critique of experiment 3 is that the sire family is confounded with the dam family. Hence, variation in the three measures of reproductive success may be caused by differences among sire families, differences among dam families, or interactions between sire and dam families. Regardless, it is still true that for each unique combination of sire and dam family, reproductive success decreased significantly with the sperm length of the sire family. The mean sperm length of A. gambiae was very similar among experiments 1, 2, and 3 (197, 202, and 198 μm, respectively) and was considerably shorter than the estimate from the Klowden and Chambers  study (~280 μm). The two studies are difficult to compare because Klowden and Chambers  measured ~100 sperm on three to five males whereas in the present study 12100 sperm on 217 males were measured.
This is the first study to report intra-specific variation in sperm length among males in a species of mosquito. Klowden and Chambers  showed that A. gambiae males had more variable sperm than other anopheline species, but their study did not estimate variation among males. For A. gambiae in the present study, the mean sperm length among males ranged between 100 and 250 μm (Figure 1). However, most of the variance in sperm length occurred within males (i.e. 89%, 91%, and 97% in experiments 1, 2, and 3). This means that each male has sperm with a great variety of lengths. By contrast, in Drosophila mojavensis, the variance in sperm length among males is over three times greater than that within males . The power to detect genetic variation in mean sperm length among families was low (20%). The estimate of the full sib heritability of mean sperm length in this study was much lower (h2 = 0.18) than what has been reported in the literature [19, 20]. If the heritability of mean sperm length in A. gambiae was similar to G. bimaculatus [h2 = 0.52; 20] or S. stercoraria [h2 = 0.67; 19], this study would have had sufficient power (Figure 2). One obvious explanation for the low heritability of sperm length in A. gambiae is the large variance within males. Another explanation is polyandry (i.e. families had a mixture of full-sibs and half-sibs) which, although rare in the field , is more common in laboratory populations [~24%; 30]. For future work, the power analysis suggests quintupling the number of families (~240 hours of work; Figure 2). The variance component analysis (Table 1) suggests shifting the sampling effort from factors that did not influence variation in sperm length (tissue culture plates and fields of view) to those that did (males and individual sperm). Sperm clumping was a problem in this study (see methods) and sampling bias might have occurred, if sperm of a certain length were more likely to clump. Future work on sperm length in A. gambiae should search for a chemical or technique that can reduce clumping. Other possible explanations for the low heritability of mean sperm length include founding effects and laboratory selection. During the colonization process, there is an inevitable loss of genetic variation in Anopheles colonies as many individuals do not mate under laboratory conditions . The Keele strain used in this study, despite being outbred [see ], might have less genetic variation for sperm length than wild populations, even if the colonization process maintained the genetic variance for wing length. Future quantitative genetic experiments on sperm length should therefore focus on wild populations of A. gambiae.
The importance of the environment on the heritability of quantitative traits is shown by the 10-fold difference in the full-sib heritability of wing length between experiments 2 and 3 (0.65 versus 0.06). Differences in sample size between experiments 2 and 3 (32 versus 16 full-sib families) cannot account for this discrepancy because the variances on which the heritabilities are based are independent of sample size. One explanation is that in experiment 3, the larvae spent two days at high density (see materials and methods) and heritabilities are often lower in sub-optimal environments [31, 32]. The full-sib heritability of wing length in experiment 2 (h2 = 0.65) was higher than that in a field-captured population of A. gambiae (h2 = 0.35; ), illustrating that quantitative genetic variation in a laboratory colony is not always lower than that in a wild population.
The discovery that sperm length was negatively correlated with male reproductive success in A. gambiae suggests that sperm length is an important measure of male fitness and represents a novel contribution to understanding the reproductive biology of this medically important vector. The results in this study are consistent with two other species of insect: G. bimaculatus  and O. taurus , where males with shorter sperm have higher fertilization success. Why do males with shorter sperm have higher reproductive success in A. gambiae ? Life-history theory suggest that there are trade-offs between sperm quantity (number of sperm) and sperm quality (length, viability, swimming speed), although only a few studies have found evidence for such trade-offs . For example, of the 11 studies investigating a negative correlation between sperm number and sperm length , only three found the expected trade-off in Drosphila fruit flies , the snail, Vivaparus ater , and the yellow-pine chipmunk, Tamias amoenus . Klowden  recently showed in A. gambiae that innervations from a sperm-filled spermatheca cause females to switch to the mated state (rather than male accessory gland fluids [37, 38] which are well known to inhibit female multiple mating in the yellow-fever mosquito, Aedes aegypti [39, 40]). This suggests that the volume of sperm transferred to fill the female spermatheca is important for inducing female oviposition behavior . Hence, males producing lots of short sperm may be better at filling a female's spermatheca and inducing oviposition than males producing a few large sperm.
The finding that males with short sperm have higher reproductive success appears to contradict the work of Klowden and Chambers . They found that the mean sperm length in the female reproductive tract and spermatheca was significantly longer than that in the male testes. This suggests that males with longer sperm have higher reproductive success. However, these two results are not necessarily mutually exclusive. In A. gambiae, each male has a great variety of sperm lengths. Males that produce lots of short sperm also produce long sperm. As observed in other insects [41, 42], it may be that short sperm act as filler that signal to the female that she is mated whereas long sperm fertilize the eggs. For example, Drosophila pseudoobscura males produce both long (300 μm) and short (75 μm) sperm but only the long sperm fertilize the eggs . Similarly in the Lepidoptera (butterflies and moths), males produce long nucleated (eupyrene) sperm that fertilize the eggs and short anucleated (apyrene) sperm that are believed to act as filler to prevent female re-mating . In contrast to the two distinct sperm lengths in D. pseudoobscura  and the apyrene sperm in the Lepidoptera , A. gambiae males produce a continuous distribution of sperm lengths (this study) and Klowden and Chambers  showed that all of these sperm are nucleated and should therefore be able to fertilize the egg. Finally, it is difficult to compare the present study to that of Klowden and Chambers  due to differences in methodology. They used different males to estimate the distribution of sperm lengths in the female reproductive organs and the male testes. They did not show that females with longer sperm had higher oviposition success. Their study suggests selection on sperm length within a male's ejaculate whereas this study measured selection on sperm length among groups (full-sib families) of males.
Selection experiments with Drosophila melanogaster have shown that male fertilization success is determined by an interaction between sperm length and the length of the female sperm storage organ . Males selected for long sperm had higher fertilization success than males selected for short sperm when mating with females selected for long sperm storage organs, but there was no difference in fertilization success between long-sperm and short-sperm males when mating with females selected for short-sperm storage organs . Conversely in O. taurus, males with short sperm had the highest fertilization success when mating with females with medium- to large-sized spermathecae whereas males with intermediate-sized sperm had the highest fertilization success when mating with females with small spermathecae . Across five different anopheline species, sperm length appears to be positively correlated with spermatheca volume  suggesting that female morphology is exerting selection on sperm length as shown in other taxa .
Selection on sperm length was not caused by correlated selection on body size. Sperm length and male wing length were positively correlated (Figure 3), but selection on these two traits was in opposite directions and was not statistically significant for wing length (Table 4). In this study, male body size was not important for male reproductive success, perhaps because all the males were reared under the same conditions, so that there was little variation in male body size. In contrast, other laboratory studies have found that A. gambiae males that captured a female during swarming were slightly larger (mean wing lengths = 2.82 versus 2.76 mm) than those that did not , although a more recent study found that intermediate-sized males were the most successful at capturing females during swarming . In a wild population of A. freeborni, large males mated more often than smaller ones as revealed by examining their accessory glands . Male body size and sperm length most likely influence different components of male reproductive success. Large males may be more successful at acquiring mates, whereas males with many short sperm are more successful at inducing females to oviposit.
The present study does not support the conclusion that sperm length is causal with respect to oviposition success in A. gambiae. Males with short sperm may produce more sperm or more accessory gland fluids than males with long sperm. Female factors such as the size of her reproductive tract or spermatheca, the nutritional value of the male accessory gland fluids and the blood meal (i.e. the level of resources available for egg production), or her tendency to oviposit in the laboratory environment could all change the relationship between sperm length and oviposition success. Furthermore, in experiment 3, sperm length was measured in the males four days after they were separated from the females. Hence an alternative explanation for the negative relationship between male sperm length and oviposition success is that mated males transferred their long sperm to the females and had not replenished their long sperm supplies by the time they were assayed for sperm length. Previous work on a multiply mated laboratory population of A. culicifacies found that male reproductive success peaked at day-3 and day-7 post-emergence separated by four days of rest . In A. stephensi, males that were given continuous access to virgin females had "rest periods" of up to three days in between fertilizing females . Similarly, Voordouw et al , using lines of A. gambiae derived from this study's Keele population, found that male reproductive success was correlated between two days of mating separated by one day of rest. These three studies suggest that the males in the present study were given sufficient time (four days) to replenish their sperm supplies. Unfortunately, it is not possible to measure male sperm length before mating, because the procedure kills the male. However, it is possible to measure sperm length in the spermatheca of the female and future experiments should investigate whether A. gambiae females that oviposit are more likely to have long sperm in their spermathecae than females that do not.