C. elegans was cultured as described in 31. The preparation of NGM plates and Escherichia. coli strain OP50 food bacteria and M9 buffer is also described in 31.

Mating plates: 6 cm NGM plates were seeded with 30 μl of an E. coli (OP50) culture such that the plates contained a small round dot of bacteria in the center.

All experiments were done in an air-conditioned room at a temperature of 21 ± 1°C and 40% humidity. To minimize fluctuations of physical conditions plates were kept in boxes, randomized in piles that were placed evenly distributed within the boxes.

N2: Standard laboratory wild type strain, isolated in Bristol, UK

CB4856: Standard polymorphic mapping strain, isolated in Hawaii.

AB1: isolate from Australia

JU258: isolate from Madeira

MY1, MY15, MY18, RC301: isolates from Germany

All strains are available from the Caenorhabditis Genetics Center at the University of Minnesota 32.

This assay served to determine the persistence of males over time in different natural C. elegans isolates and in different population sizes. For each strain, we set up several crosses using a male:hermaphrodite ratio of 2:1. These crosses yielded populations with a gender ratio of approximately 1:1. For each strain, the populations were mixed four days after setting up the crosses and a defined number of individuals (population size) was randomly chosen and transferred onto NGM plates for the experiment (day 0 of the experiment). These experimental populations were all treated as follows: After three days adult males and hermaphrodites were counted (counting, see below). On the next day the population was reduced to the original population size and transferred to a new plate (transfer, see below). Three days later, adult males and hermaphrodites were counted again, followed by population size reduction and worm transfer one day later, as above. This whole procedure was repeated for a total of eight times (equivalent to 32 days). All male maintenance assays were done on 9 cm NGM plates, seeded with 1 ml of E. coli OP50 culture. Two sets of experiments were performed: i) two replica runs per strain and population size were done in parallel for strains N2, CB4856, AB1, JU258, MY1, MY15, MY18 and RC301 using population sizes of 75 and 150; and ii) five replica runs per strain and population size were done in parallel for N2 and CB4856 using population sizes of 40, 70, 100 and 150. In this context, one important objective was to evaluate the effect of different population sizes on male persistance. The exact numbers used (e.g. 75 versus 150) were chosen arbitrarily.

This assay served to evaluate the mating efficiency of males in terms of mated hermaphrodites, total offspring produced per male and also cross- as well as self-progeny produced per mated hermaphrodite. A single male was confronted with an excess of hermaphrodites, so that it could mate as often as possible. In a pilot experiment, 14 hermaphrodites were found to be sufficient to ensure that the male would never come close to mating with all of them. In fact, during the main experiment the highest number of mated hermaphrodites per plate was 9.

Mating plates were prepared four days prior to the experiment. One male and 14 young adult hermaphrodites were placed on mating plates. Every 24 hours the male was moved to a new mating plate with 14 young adult hermaphrodites until no more successful mating was observed. Hermaphrodites, which were exposed to males, were placed individually on NGM plates seeded with 200 μl E. coli OP50. The hermaphrodites were transferred to new plates every 24 h. After three days the progeny was counted or, if the number of plates was too high to be processed immediately, equal numbers of plates from both treatments were put at 4°C and counted within a few days. This step allowed us to do more replicas in parallel and have all plates scored by the same person, in order to avoid possible observer biases. We did not observe any lethality in response to the cooling step. A hermaphrodite was considered to be mated when more than one male was found among the progeny (successful mating event). The number of cross progeny per mated hermaphrodite was estimated as twice the number of male progeny. Since it was not feasible to do all crosses in parallel, we used an incomplete block design, where two different crosses were set up in parallel and different pairs of parallel crosses were assayed in four experimental runs: i) N2 males crossed with N2 and with CB4856 hermaphrodites; ii) CB4856 males crossed with N2 and with CB4856 hermaphrodites; iii) N2 and CB4856 males crossed with N2 hermaphrodites; iv) N2 and CB4856 males crossed with CB4856 hermaphrodites.

These two assays were used to characterize in more detail the time required by males until first contact with a hermaphrodite and first spicule insertion (One-hour assay) and also the number of contacts with hermaphrodites as well as spicule insertions over a nine hour period (Nine-hours assay). In both assays, we used the same general conditions as in the male mating efficiency assay (particularly as to usage of 14 hermaphrodites), in order to permit comparison of results.

This assay served to determine the number of offspring produced trough self-fertilization in the absence of males for two natural isolates, N2 and CB4856. Young adult hermaphrodites, which had no developing embryos in the uterus yet, were placed individually onto plates and moved to new plates at least once every day until they stopped laying eggs. Three days after the removal of the mother the progeny on every plate was counted and summed up to total numbers.

This assay was used to test how more frequent mating affects the hermaphrodite's production of self- as well as cross-progeny. Mating plates were prepared four days prior to the assay. One L4 hermaphrodite was placed together with 1, 3, 6 or 12 males onto a mating plate. The hermaphrodite was transferred onto a new mating plate with new young males every day until it stopped laying eggs. The old males were removed from the plates. After one day, additional food bacteria (OP50) were added to the plates and hermaphrodites and males were counted two days later, when they were young adults.

All statistical analyses were done with the program JMP IN version 5.1.2 (SAS Insitute Inc., USA) or SPSS version 14.0 (SPSS Inc., USA). The male proportion over time in populations of the different isolates was evaluated using logistic regression based on a full factorial model with time and either population size or C. elegans strain as fixed predictors. The male proportion at specific time points was additionally examined with a Wilcoxon sign rank test. Variation in the number of mated hermaphrodites, the number of offspring per male, and the self- and cross-progeny per mated hermaphrodite was assessed with a General linear model, using an incomplete block design, including male strain, hermaphrodite strain and the interaction between them as fixed factors and experimental block as a random factor. In case of significant interaction terms, significant differences among groups were evaluated with Tukey HSD posthoc tests. Variation in offspring number per repeatedly mated hermaphrodite was tested with an ANOVA, using a full factorial design with male strain and hermaphrodite strain as fixed factors. Subsequent posthoc tests were performed with Tukey HSD. A similar ANOVA was performed to assess the effect of different numbers of males on offspring numbers of repeatedly mated hermaphrodites. In this case, number of males was used as a fixed predictor in the model. With respect to mating behaviour, the time measurements until first contact or first copula were always terminated after 60 min, resulting in non-continuous data. Therefore, differences between crosses were assessed using the Wilcoxon sign rank test. The variation in the number of copulas during this time frame were examined with the Fisher exact test. Differences in the number of contacts or copulations over 14 observation points within a 9 h interval were assessed with the Wilcoxon sign rank test, since the data were non-parametric.

We first tested in how far there is variation in male maintenance among populations of different natural C. elegans isolates. Eight strains were tested at two different arbitrarily chosen population sizes (75 and 150). The proportion of males was significantly affected by the factor time, the strain studied, and also the interaction of the two. In particular, males disappeared completely from the cultures of some strains, among them N2. In contrast, in other strains, among them CB4856, the cultures appeared to reach a stable frequency of males after about two weeks (Fig. 1 and [see Additional files 1, 2]).

To refine the analysis we examined the persistence of males at four different arbitrarily chosen population sizes (40, 70, 100 and 150) for N2 and CB4856 (Fig. 1C and [see Additional files 3, 4]). The proportion of males was significantly reduced within few generations. In fact, males disappeared almost entirely from all N2 cultures. The loss was much slower in CB4856 populations. Here, male frequencies were significantly affected by population size, whereby larger populations (100 and 150) sustained a higher number of males [see Additional file 3]. For these two population sizes, the final male frequency was significantly different between N2 and CB4856, while this was not the case for the smaller populations [see Additional file 4].

Taken together, our results confirm long-standing anecdotal knowledge available within the C. elegans community: We, and many others, have noticed that it is necessary to deliberately set up crosses with an excess of males every few generations in order to maintain N2 populations with males for genetic analysis. For CB4856 this is not necessary. Our data are also in agreement with previous studies, in which males were rapidly lost in experimental populations of N2 781011 but maintained at constant levels in those of CB4856 11. Since during our experiments the different strains were kept in parallel under identical conditions, the difference in male maintenance between them must have a genetic basis. Thus, our results suggest that C. elegans bears considerable intra-specific genetic variation that affect male frequency, making it a potentially selectable trait. Interestingly, population size differences had a significant effect on male persistence. This finding may be a consequence of the population size itself, e.g. smaller populations may loose males more often due to chance, thus accelerating male decline. A non-exclusive alternative explanation may be density differences among the population sizes. In this case, higher densities in the large population sizes may associate with more male-hermaphrodite contacts, which could result in higher mating rates, thus stabilizing male frequencies. At the moment, our results do not allow to distinguish between these two effects.

We decided to further characterize the proximate processes that account for variation in male persistence between two of the extremes, namely the strain N2 and CB4856.

The factors underlying variation in male maintenance were systematically assessed by reciprocal crosses. The analysis was based on all possible mating combinations between CB4856 and N2, using an incomplete block design (i.e. not all combinations were assayed at the same time; see methods). The main experiment focused on the consequences of repeated mating for individual males, which is likely to be realistic in most populations, where male frequency is usually less than 0.01 24. Three parameters were simultaneously evaluated: i) The number of successful copulations per individual male, which was offered an excess of virgin hermaphrodites (14 hermaphrodites) every day over a period of six days, ii) the number of sired offspring per male with – accordingly – virtually unlimited access to mates, and iii) the number of cross- and self-progeny per successfully mated hermaphrodite (for experimental details see Methods). Since early reproduction should have a stronger influence on population dynamics than late reproduction, we also performed a separate analysis of the data from only the first two days (Table 1B). The results lead to essentially identical conclusions like the results for the full reproductive period (Figure 2 and Table 1). Therefore, in the following we focus our discussion on the data for the full reproductive life-span.

CB4856 males had significantly more successful copulations (Figure 2A and Table 1) and significantly more offspring than N2 males (Figure 2B and Table 1). These two traits were not significantly affected by the hermaphrodite strain used. Therefore, the difference in progeny production by males is most likely a consequence of the differences in mating rates. It does not seem to be caused by variations in the fertilization rates males achieve per successful mating event: The number of cross-progeny per mated hermaphrodite was not significantly affected by any factor of the model or the overall model as a whole (P > 0.05; Figure 2C and Table 1). At the same time, it is interesting to note that N2 males appear to produce more cross-progeny with N2 rather than CB4856 hermaphrodites (Fig. 2C; see also similar results obtained after repeated mating of hermaphrodites in Fig. 4A). This effect may account for the trend of a difference produced by the factor male strain in this context (Table 1). It is responsible for the significant interaction term, which was inferred for the data from the first two days (Table 2). One possible explanation for this result is a certain degree of genetic incompatibility, which only becomes visible in one type of cross between the two strains (male N2 and hermaphrodite CB4856) and which may be related to the recent report of genetic incompatibilities among different natural C. elegans isolates 33.

Interestingly, N2 hermaphrodites always had significantly more self-progeny than CB4856 – irrespective of the type of the male and irrespective of the cross-progeny produced (Figure 2C and Table 1). This can be explained by their generally higher fertility. In agreement with other authors 1534 we found in a separate experiment (self-brood-size assay; see methods) that unmated N2 hermaphrodites produced more self-progeny than unmated CB4856 hermaphrodites (284.3 ± 5.7 and 245.3 ± 6.1, respectively; t-test, t₃₇ = -4.65, P < 0.001). In general, our results on total cross-progeny sired by N2 males (Table 1) as well as total self-progeny by N2 hermaphrodites (see above) are within the range of previously published data 1534.

In a separate set of experiments we measured different aspects of mating behavior within either the first one or nine hours. As in the above experiments, we combined an individual male with 14 hermaphrodites, although in this case both were always of the same strain (i. e. 1 N2 male × 14 N2 hemaphrodites and 1 CB4856 male × 14 CB4856 hermaphrodites; for further details see Materials and Methods). Within the first one hour, the different crosses did not show any significant variation, neither regarding the time until first male-hermaphrodite contact (253.3 ± 51.4 sec for N2 and 171.5 ± 28.4 sec for CB4856; Wilcoxon test, Z = 0.95, N = 20 per strain, P = 0.343), nor in the time until first successful spicule insertions (2148.5 ± 280.6 sec for N2 and 2212 ± 299.7 sec for CB4856; Wilcoxon test, Z = 0.10, N = 20 per strain, P = 0.923), nor in the number of replicates with successful spicule insertions (12 out of 20 for N2 and 13 out of 20 for CB4856; Fisher exact test, P > 0.999). We conclude that the CB4856 males are neither generally better in finding their first mates nor in achieving their first spicule insertion. This result is consistent with the previous finding that N2 and CB4856 show similar behavioral responses to a hermaphrodite-derived cue 121314.

However, over 14 observation points within the first 9 hours, the CB4856 × CB4856 crosses produced significantly more male-hermaphrodite contacts (Fig. 3 and [see Additional file 5]; Wilcoxon test, Z = 6.59, N = 45 for CB4856, N = 47 for N2, P < 0.001) and spicule insertions (Fig. 3 and [see Additional file 5]; Wilcoxon test, Z = 2.98, N = 45 for CB4856, N = 47 for N2, P = 0.003). From these observations, we conclude that over time CB4856 males achieve a higher rate of mate contacts and spicule insertions than N2 males. These results are in excellent agreement with the results presented in Fig. 2 and Table 1. Consequently, an overall higher mating frequency could contribute to the observed higher male persistence in CB4856 relative to N2. We cannot explain why N2 males mate less frequently than CB4856. One possibility would be that N2 males require a longer time to replenish their sperm stocks. If so, they could produce a smaller number of sperm over their life time which would explain the reduced number of progeny sired.

Variation in male sperm competitiveness (size) could have also affected the observed differences in male persistence. We have not as yet measured this trait. However it is unlikely to have significantly influenced the results on mating efficiency in Fig. 2. We followed the mated hermaphrodites until they ceased to produce progeny, presumably because they had used up their supply of sperm (including both male and hermaphrodite sperm). In this case, all male sperm transferred during mating should have contributed to offspring production irrespective of their competitiveness.

In the experiments leading to Figure 2 and Table 1 it is likely that individual hermaphrodites mated only once or very few times since they were in excess and they were removed from the males after one day. Given the low number of males in natural C. elegans populations, mating only once, if at all, might be realistic. However, in our male maintenance assays (see Figure 1), the male frequency was initially set to approximately 0.5, thus allowing for repeated mating interactions. Therefore, we asked whether repeated mating could influence the number of self- as well as cross-progeny. For this purpose, we set up all reciprocal crosses between N2 and CB4856 and, for each cross, single hermaphrodites were mated each day with a virgin male until the hermaphrodite ceased to produce progeny (see Materials and Methods, hermaphrodite outcrossing efficiency assay with one male).

In agreement with earlier literature 34, we found that repeated mating (repeated mating assay with one male, see Methods) increased the total number of progeny and the proportion of cross-progeny (Figure 4A and [see Additional file 6]). N2 hermaphrodites produced significantly more progeny than CB4856 hermaphrodites (ANOVA, hermaphrodite strain effect, F₁ = 107.69, P < 0.001) and the proportion of self-progeny was small. The origin of the male did not make any significant difference (ANOVA, male strain effect, F₁ = 2.93, P = 0.108). Analysis of the first two days only produced essentially identical results [see Additional file 7]. Consequently, in the case of repeated mating of hermaphrodites, N2 would produce a significantly larger number of cross-progeny and thus a larger proportion of males than CB4856.

At first sight, this result appears to contradict our findings from the male maintenance assays, where N2 populations loose males rapidly in contrast to CB4856. However, hermaphrodites in the repeated mating experiment were exposed to "new" virgin males every day (most likely with high mating efficiency), whereas hermaphrodites in the male maintenance assay encounter the same males over consecutive days, which are likely to show reduced mating efficiency over time (due to the likely high energy demand per mating as well as sperm depletion). The reduction in mating efficiency is particularly pronounced in N2 (see results from the mating behaviour experiments above). Therefore, we expect fewer matings, a relatively larger number of self-progeny, and thus a continuous decrease of males in the N2 populations over time.

A possible higher cost of multiple mating may further contribute to more rapid initial male decline in N2 in the male maintenance assays. Repeated mating with increasing numbers of males caused a significant reduction in offspring number in both strains (Figure 4B and [see Additional file 8]). Importantly, this reduction was more pronounced for N2, for which it caused a loss of up to 53% progeny compared to a maximum of about 40% for CB4856 (comparison between repeated mating with 1 versus 12 males). Consequently, if repeated mating of hermaphrodites did occur in the experimental populations, then the higher costs (i.e. offspring loss) of copulations with multiple males for N2 should decrease male frequencies and thus outcrossing rates to a larger extent in N2 than CB4856.

The finding of a cost of multiple mating is likely a consequence of sexual conflict, as reported for a large diversity of organisms 3536. One possible explanation for this observation could be increased intra-sexual male-male competition or otherwise detrimental male-male interactions, which were shown in the past to decrease C. elegans male life-span 37. Alternatively, it could result from inter-sexual antagonisms such as those mediated by male manipulative substances that are transferred during copulation, in order to enhance male fertilization success 3536. In C. elegans, the possible relevance of such inter-sexual conflict was previously indicated by reduced hermaphrodite longevity after mating 38.