Study species and spider maintenance
Argiope versicolor, H. multipuncta, L. decorata, N. pilipes and N. malabarensis are commonly found in the subtropics (Supplementary Figs. 1–5) and tropics. With the exception of Leucauge, all abovementioned four genera are currently catalogued in the family Araneidae30. We collected large juveniles and subadults of the five species in Singapore.
Sexual cannibalism in the genus Argiope takes place before and during copulation44,46,49,50. The cannibalism frequency is reported to be around 36–80%, regardless of the female mating status43,46,47. Courtship begins when the male cuts into the female web to which he connects a mating thread (Supplementary Fig. 1); the latter forms a substrate for vibrational courtship. The female moves from her position on the web to the mating thread, where the male moves down to copulate. Sometimes, males are eaten during the first copulation trial. If the male is not cannibalised, he will repeat his courtship for a second insertion. However, the highly cannibalistic Argiope often sees the males getting cannibalised within the first two insertions43,50. This would render the copulation duration to be very short, about 8 s on average47. Despite the short copulation duration, the fertilisation rate and hatching success are not correlated47. If this was so, the possible reason for males attempting a second insertion could be to increase the number of sperm transferred for success in sperm competition47,51.
Courtship and mating in H. multipuncta occur on the female’s web52. Several adult males can live on the web of a female, even if the female is immature53. The epigynum of the female is snugly blocked as she positions herself on the web (Supplementary Fig. 2), which means that when a male tries to copulate with her, she will, if receptive, raise her body out of the ‘hub-cup’ for access to her epigynum53. This may, however, be the initiating point of her aggressiveness and cannibalism if a male tries to squeeze into her ‘hub-cup’ (personal observations). Post-copulatory palp breakage may serve as a mating plug, and it is observed in aggressive females exhibiting sexual cannibalism, which occurs 50% of time15.
According to Aisenberg (2009), L. mariana was observed not to be sexually cannibalistic54, although a more recent study reports its cannibalism frequency to be 9–30%48. Although the Leucauge species used in this study is different, it is a reasonable prediction that its cannibalism levels are similar. Among the five studied species, L. decorata exhibits the lowest SSD (Supplementary Fig. 3). Females appear to form their own mating plugs by secreting a substance after copulation55,56, which may hinder future copulation. During courtship, the pair uses chelicerae clasping to hold onto each other while the male stretches out his palp for insertion21 (Supplementary Fig. 3). Each insertion appears to be in a form of short and long tappings and occur multiple times, where sperm transfer occurs54. Males also use their legs to beat the females’ abdomen which visibly acts as an enhancement for copulation approval by the female54. If the female is receptive, she will position herself to face the male for the copulation to start.
SSD is the most pronounced in N. pilipes (Supplementary Fig. 4), which could be the possible reason for females failing to detect approaching males or ignoring them since male size is too small for female optimal foraging57. Sexual cannibalism is observed before and during copulation, with the frequency observed to be at around 17% in N. pilipes24. Like Herennia, several males are observed to be on the web of the female57,58,59. Courtship by the male includes a unique mate binding behaviour where it spins silk on the female’s abdomen, legs and carapace24,51.
Females are highly cannibalistic in N. malabarensis17 with sexual cannibalism levels up to 75%. Because of this, male Nephilengys often break off their palps during copulation (Supplementary Fig. 5) to avoid being cannibalised while continuing to transfer sperm into the female’s spermatheca to increase male chances of producing progeny37. Eunuch males also benefit from being lighter and agile: the ‘gloves-off’ hypothesis predicts enhancement of eunuch survival through higher physical stamina and endurance17. Since the advantages of palp severance appear to exceed its cost, most males break off their palps.
We maintained spiders in the controlled environmental conditions following the standard protocols used in earlier studies24,37,60. We reared and monitored them until they matured. We used only virgins (both males and females) in this study and recorded their post-maturity ages from the last moult, which aided in standardising the ages of female spiders used in the experiment as an attempt to eliminate age as a confounding factor. A total of 86 successful mating trials were conducted for this study. The mating trial procedures are summarised below.
To control for hunger possibly affecting sexual cannibalism and palp choice, we fed the test spiders 24 h before the trials. An hour prior to the mating trials, we measured the spiders’ mass, total body length and carapace width, and then returned them back to their housing cages.
As males reach their partners on female webs to perform ritualised vibratory courtships61, we staged all mating trials in the resident female cages. We started a mating trial by releasing a male at the bottom of the cage, at a minimum distance of 10 cm to the female. We encouraged the test male to continue to approach the female every 10 min gently using a paint brush if needed, but not after the male was within 5 cm from the female. The time taken for the male to approach the female varied between and within species; we therefore allowed 4 h from the introduction of the male before terminating an unsuccessful mating trial. We substituted the male if he repeatedly failed to respond to the paint brush stimulus.
We recorded the copulation duration using a stopwatch and noted which palp (left or right) was used first for each insertion (palp choice). Copulation duration refers to the entire time the tip of the embolus remained inserted in the epigynal tract, with or without the male (in the case of remote copulation37). Full palp breakage only occurred in N. malabarensis (Supplementary Figs. 7 and 8); those severed palps were immediately removed (Supplementary Fig. 7) to prevent continuous sperm transfer. Although full emasculation did not occur in the other four species, the males of those species do occasionally engage in lesser genital mutilation, and when they do, a broken embolus may form a mating plug lodged in the female spermatheca (Supplementary Fig. 9)18,19,37.
We also recorded whether the male or the female terminated the copulation. If the former occurred, the trial was categorised as an insertion (or palp breakage in the case of N. malabarensis). If the female terminated copulation, it was either an attempt to cannibalise the male; or a true cannibalism of the male (Supplementary Fig. 10). Thus, the female was classified as cannibalistic or not, and the male was classified as being cannibalised or not. We then calculated the sexual cannibalistic rates of the spiders, and compared them to the literature reports15,17,24,44,45,46,48. In all species, male usually inserted one of its palp in one genital opening of female, and once we observed that a male inserted one of its palp consecutively in both genital openings of females, we discarded that pair in data analysis.
All trials were recorded using a digital HD camcorder (JVC GZ–MG50AG and JVC GX–MG50AG, Japan). Subsequent video playback allowed us to verify the measured copulation duration and observed palp choice.
The following four scenarios occurred during and post-copulation (Supplementary Fig. 10): (1) If the male initiated the end of a successful copulation (i.e., after insertion), we separated the male and the female. (2) If the female attempted to cannibalise the male or displayed aggressive attacks, such as kicking to displace the male from her epigynum during copulation, we immediately forced the aggressive female to retreat to a corner using a wooden stick. We then recorded any broken body parts of the male. (3) If the female cannibalised the male before the start of copulation, we immediately removed the male. We did not reuse the female for any future mating trials to ensure that our sample remained unbiased. (4) If the female cannibalised the male during copulation, we immediately separately the male from the female to prevent his further body damage as the palps were required intact for sperm counting. After these steps, we euthanised the spiders with carbon dioxide (CO2) for 20 min.
Female sexual cannibalism
Sexual cannibalism is broadly defined as any female hostile behaviour toward males, which includes displaying aggressiveness, attacking, grabbing, or killing the male. Aggressiveness is the mildest extent of sexual cannibalism, such as showing signs of reluctance to mate. Attacking refers to the female giving chase or attempting but failing to grab the male. Grabbing occurs when the female moves to catch and hold on to the male, prior to biting him. We define killing as when the female bites and injects her venom into her partner. Although sexual cannibalism in nature often culminates in the actual consumption of the victim, our definition of cannibalism needed to stop with the killing itself (and not consumption) because, as per the experimental design described above, we separated the victim from its aggressor before his likely consumption. Therefore, within this study, we use a restricted definition of sexual cannibalism as females that grabbed and/or killed their mates.
Palp detachment procedures
If full-emasculation occurred, we removed the broken palp from the female epigynum immediately. We also removed the unused palp from the male body using soft forceps and scalpel under a stereomicroscope (Supplementary Fig. 11); this procedure ensured the severance of the intact palps between the palp tarsus and tibia, i.e., at the position of natural emasculation62. We stored all palps in 0.2 ml Eppendorf tubes and refrigerated them for sperm counting.
Epigynum dissection and sperm counting procedures
The epigynum refers to a sclerotized plate on female abdomen containing genital openings on the surface and spermathecae and ducts in its interior63. To determine if and how much sperm were transferred to the female genital tract, we dissected the epigynum in its entirety, then cleaned the soft tissue covering internal genitalia using soft forceps, scalpel, and a 27G × ½ inch fine injection needle (BD PrecisionGlide, USA) under a stereomicroscope. This procedure exposed the undamaged spermathecae, copulatory ducts and fertilisation ducts (Supplementary Fig. 12); not damaging the spermathecae was important to prevent sperm loss, particularly in L. decorata with fragile inner organs. We individually stored the isolated female epigyna in 0.2 ml Eppendorf tubes and refrigerated them before sperm counting.
Sperm counting procedures
We counted sperm in the used and unused palps, as well as the epigynum. Since females have paired spermathecae, we counted the sperm in the spermatheca that was used during copulation with a particular palp. The number of sperm in the used palp and spermatheca gave the total number of sperm in the palp chosen by the male for copulation (Supplementary Fig. 13). Sperm clumping is an issue that complicates sperm quantification as it often underestimates the total sperm count, causing inaccuracy and unreliability of the sperm numbers. To ensure that this problem is minimised, the sperm counting procedure was modified by referring to multiple references1,13,28,64 and using ‘mix-and-match’ to derive a best method. Sperm counting was conducted separately for the used palp, unused palp and epigynum. Since females have paired spermatheca, the sperm count was taken from the one spermatheca that was used with a particular palp. The number of sperm in the used palp and the spermatheca gave the total number of sperm in the palp the male has chosen to use for copulation (Supplementary Fig. 14).
We modified the sperm counting procedure from earlier studies37,43,65,66. We added 5 μl of 0.9% saline + 10% triton-X detergent to the Eppendorf tube containing either a palp or spermatheca and used metal forceps to thoroughly grind and break the reproductive organs for sperm to be released into the solution. We then added another 5 μl of solution to wash the forceps of any debris and sperm which might have accumulated. Samples containing the palp were ultrasonificated (Microcomputer heater–timer controller DSC–120TH, Sonicor Instrument Corporation, USA) five times 30 s after which the Eppendorf tube was manually flipped to the other side so that the solution was evenly sonicated to break up sperm clumps. The samples then underwent another 30 s of pulse-vortexing (Type 37600 Mixer, Thermolyne Corporation, USA). Samples containing the spermatheca underwent a similar procedure, except that the period of ultrasonification was extended to five min (sonicated ten times 30 s) as sperm clumping (Supplementary Fig. 14) was observed to occur more frequently.
We extracted two 4 μl samples using a pipette and placed them on both chambers of a 0.1 mm Neubauer improved haemocytometer (Blaubrand, Wertheim, Germany). Using a compound microscope at 400× magnification, we counted sperm in all 25 quadrats (0.05 × 0.05 mm) from each chamber. As only encapsulated sperm could be observed under the microscope, we equated the number of capsules to the number of sperm37. An average was calculated from the quantities in the two chambers, which was then multiplied by the dilution factor to get the total number of sperm of the sample. We also calculated the percentage of sperm transferred to the female during the copulation using the formula from Snow and Andrade (2004)65:
$$P=[T/(U+T)]\times 100 \%$$
Where P is the percentage of sperm transferred to female’s spermatheca, T is the total number of sperm in the female’s spermatheca, and U is the total number of sperm counted in the used palp.
Statistics and reproducibility: comparative study
All statistical analyses were done using R 4.1.167. We checked for data normality using the Shapiro–Wilk tests and transformed the data if it were not normally distributed. We used parametric or nonparametric tests depending on the normality of the data.
We used two GLM with a Gamma error and log-link function to test for the effects of sexual cannibalism (whether a female is cannibalistic) and other five exploring factors (sexual dimorphism [the ratio of female mass to male mass], male post-maturation age, female post-maturation age, species, and taxonomy [family to which a species belongs to]) on the two mating strategies (palp used with more sperm and percentage of sperm transfer) separately. For all GLMs, we used an information-theoretical framework68 to assess a series of hypothesis-driven models. We then compared the models using Akaike’s information criterion corrected for small sample sizes (AICc) to identify the best fitting model that predicts the effects of explorative factors on each male mating strategy68 by ranking them based on the AICc values using model.sel, and averaged the models using model.avg, if ΔAICc < 2, in the R package MuMIn69.
We coded the ratio of the number of sperm in the used and unused palp and percentage of transferred sperm as response variables. In these two GLM models, the six main factors and their interactions were predictors, but we included copulation duration as additional predictor when testing for the effects on the rate of sperm transferred to control for its effect. We excluded male and female size from the model because body size measurements are unrelated to sperm transfer70. Instead, we used their mass ratio as SSD. Regression estimations were also conducted on the males of the five individual species to test for a correlation between male carapace width and the total number of sperm it had. The correlations were very weak in all five species, with the highest correlation in H. multipuncta (R2 = 0.470, F1,9 = 7.984, N = 11, P = 0.02). Therefore, we excluded male size from the models.
All five species differed greatly in copulation duration and percentage of sperm transfer (see “Results”), which might reflect species specific behaviours. We ran a GLM for each individual species to identify factors which might have species specific effects, but the GLM was not feasible due to a small sample size (smallest sample in H. multipuncta = 11). As this study focused on sexual cannibalism as the hypothetical driving factor behind the three male mating strategies, combining species into groups may not be the best solution to solve the problem of small sample sizes because of the dilution of cannibalism effects. We thus did not report the results from tests for individual species here.
Hypotheses and predictions for the palp choice mechanisms
The following elaborates how the formulated hypotheses may form the basis underlying palp choice during copulation. The hypotheses are grouped according to two ways a male spider may fill up its palps with sperm. First, mechanism A assumes that a random amount of sperm may be loaded into each palp and the spider is then able to detect minute differences in sperm quantity of each palp to make a choice of palp during copulation. Second, mechanism B assumes that the male may intentionally load one palp with more sperm than the other and thus prefer to use the palp containing more sperm. In this latter case, memory may be involved in palp selection.
(1) Volume detection. The male spider is able to detect differences in amount of sperm in each palp and thus prefers to choose the palp with more sperm for copulation. The male is either able to detect the absolute difference in the sperm amount or it is unable to detect the absolute difference in sperm amount, and instead detects the ratio of sperm amount or proportional difference. The latter refers to the proportional processing of information in decision making, where the traits are judged as a relative to the presented stimuli. This cognitive mechanism is magnitude-dependent, where at high magnitudes of a trait value a larger increase is needed to obtain the same proportional difference than at low magnitudes38,39,42. This could have evolutionary consequences, that when a trait is at a high magnitude it will require more investment to exaggerate it further to achieve a proportional difference sufficient for detection. Therefore, due to cognitive constraints, males may select the palp containing more sperm not by evaluating the absolute, but instead, the proportional difference or the ratio of sperm amount between the two palps. (2) Mass of palp. The male spider may select the palp with greater mass for copulation. The male spiders of sexually dimorphic and cannibalistic species are suspected to be sensitive to differences in mass; studies have shown a change in behaviour of male spiders subsequent to palp severance, thus subsequent to mass loss17,60. The mass of a palp may be linked to the volume of sperm it contains, or to palp dominance or palp capacity. However, the mass of palp is difficult to quantify accurately. (3) Size and capacity of palp. The male spider may have a tendency to select the palp with a larger palpal bulb (part of the palp where sperm is stored) for copulation. A larger palpal bulb is possibly positively correlated to a higher capacity for sperm and this may account for the tendency to select the palp with more sperm for copulation. Size and capacity of palp may be linked to mass of palp. However, size and capacity of palp is difficult to quantify accurately, especially because palps are damaged after emasculation. (4) Palp dominance (subconscious). A dominant palp may naturally tend to draw up more sperm during sperm induction. The male spider may then subconsciously and consistently use the dominant palp for copulation regardless of most circumstances.
(1) First palp preference. A male spider conducts sperm induction by spinning a sperm web, placing a drop of sperm on the web from its genital pore on the abdominal venter, then filling its palps with sperm. It was implied that male spiders may fill up their palps one at a time71. We have made an assumption that the male spider may place a limited amount of sperm on its sperm web during sperm induction. Therefore, the palp that the male spider loads with sperm first has the potential to be filled to its maximum capacity, while the second palp can only take up the remaining sperm on the sperm web. The male may then remember to use the first palp for copulation later on. The preference for first palp may be linked to palp dominance, whereby the first palp is the dominant palp. (2) Palp dominance (conscious). The male spider may intentionally fill its dominant palp with more sperm and memory may or may not be involved in selecting the palp containing more sperm.
For this experiment, we focused on N. malabarensis, a highly sexually size dimorphic nephilid/araneid29,62 that is common in Singapore and easily recognised in juvenile stages due to its highly characteristic web built on trees (Supplementary Fig. 9)72. We collected juveniles and subadults in Singapore and reared them to maturity (for rearing details, see the “comparative study” section). Only virgin spiders were used in the mating trials.
We followed the procedures as presented in Supplementary Fig. 15, including mating trial 1 and subsequent mating trials up to 5. If the male broke its palp, the male was not used in the subsequent mating trials, and the sperm of the used palp for its first insertion and the unused palp were counted. If the male did not break its palp, it was used for the subsequent trials until it broke one of its palps, and the sperm of its palps were counted. For detailed procedures, see the descriptions below.
First mating trial
Newly matured virgin males (mean ± s.e.m. post-maturation age: 39.5 ± 5.1 days) were used in the first mating trials. The first mating trial is the first step to changing the quantity of sperm in one of the male’s palps, such that the male’s palp choice in subsequent mating trials can give an indication of the basis underlying palp choice.
We randomly selected a virgin male and a virgin female for a mating trial. We placed the male at least 10 cm from the female on the female’s web to allow it to proceed with courtship and copulation, whereby the embolus of the palp is inserted into the epigynum. We recorded which palp the male used for copulation, the copulation duration and whether the male severed its palp(s) with a HD video camcorder (HDR-SR8E, SONY, Japan). If copulation failed to occur within 5 h of the start of the mating trial, the male and female were separated and left to rest for a day before being experimented upon again.
After a successful mating trial, we euthanised the female with CO2 for 20 min to record its body mass and size, and dissect its spermathecae (Supplementary Fig. 12d) for sperm counting. An experimental eunuch male was euthanized similarly, but a male with intact palps after a mating trial was reused. The size and mass of euthanized males were measured and their palp(s) were stored for sperm counting.
Sperm counting 1
After the first mating trial, we counted the amount of sperm in the spermatheca of the mated female and the palps of the male that had lost at least one palp. The experimental procedure for sperm counting was as in the comparative study above.
Subsequent mating trials
Following a successful first mating trial, we promptly sealed the gonopore of experimental males with intact palps with acrylic paint to prevent further sperm induction. This experimental modification is a precautionary measure to block the genital pore, which males use to release sperm from their testes onto a sperm web and then into the palps. The male was then left to rest for at least 15 min to allow the paint to dry before being subjected to the second mating trial with a different virgin female. The mating trial procedures were as in the first mating trial. We recorded the palp the male used for copulation, whether the palp was severed, and the duration of copulation.
Each male underwent a maximum of five mating trials, each with a different female, or until the male finally severed his palp(s) or was killed. The multiple mating trials aimed to change the amount of sperm in the palps of the male after each successful copulation with a different female. After mating, we euthanised the female, measured her body mass and size, and dissected her epigynum for sperm counting. We also euthanised the males that underwent the mating trials and cut the non-severed palps using a scalpel for sperm counting. The males that refused to mate after five trials were likewise euthanised, dissected and measured.
Sperm counting 2
Previously, we counted the sperm from the spermathecae of females that had successfully mated, and of males that had not severed their palps. In addition, we counted the sperm within the male’s palp only after the male had severed at least one palp and was no longer able to exhibit palp choice in mating trials. Here, sperm counting was conducted for both the palps of the remaining males and in the spermathecae of the last female it mated with. The males’ palps were dissected and the same experimental procedures were followed, except that the solution containing sperm from palps were ultrasonicated for 30 s for five times only.
Statistics and reproducibility for the data on palp choice mechanisms
The amount of sperm found in each female and the male’s palps can be used to deduce the amount of sperm present in each palp before each mating trial. Therefore, the amount of sperm that the male holds can be tracked. We performed all analyses in R. Checking for data normality, data transformation, as well as the use of parametric or nonparametric tests, were done as above in the comparative study.
Results from our comparative study showed that significantly more males used the palp with more sperm for the first insertion. Here, we used a paired samples t-test and Chi-square test for goodness-of-fit to determine if males generally selected the palp with more sperm during the first copulation. We then used Wilcoxon signed rank test to test if there was a significant difference in amount of sperm between the used versus not used palp in subsequent mating trials (in those males that mated more than once). We also used Chi-square test for goodness-of-fit to analyse whether males in subsequent mating trials had a preference for using the palp with more sperm for the first and subsequent insertions.
We performed Chi-square test for goodness-of-fit to test whether males had a tendency to change their choice of palp in subsequent mating trials. The preference for consistently using the same palp in all mating trials would support the hypothesis that palp dominance influences palp choice. We also performed binary logistic regression to assess the impact of alternation of palp on the ability to choose the palp containing more sperm. We included the male identity as a random effect in the model because individuals were used repeatedly. The Chi-square test for goodness-of-fit and binary logistic regression model only applied to males that did not sever their palp in the first mating trial because males that had lost at least one palp in their first mating trial were unable to show alternation of palps in subsequent trials.
To test the palp choice based on volume, we used the mixed-effects model with binomial error and logit-link function using the R package lme473 to assess the impact of several factors on the ability of a male spider to choose the palp containing more sperm. We included male identity as a random effect, and included positive difference in the amount of sperm between the two palps of a male, sperm ratio (i.e., the ratio of the amount of sperm present in the palp with more sperm to that in the palp with less sperm), SSD (the ratio of female to male mass) and emasculation frequency as fixed effects. The positive difference and sperm ratio showed the magnitude of the difference between the quantity of sperm present in each palp and did not take into account whether the male had used the palp for copulation. Positive difference and sperm ratio were derived from the amount of sperm present in both palps before each mating trial, detected through sperm counting. As described above in the comparative study, we also used an information-theoretical framework to assess a series of hypothesis-driven models68. We then compared the models using AICc to identify the best fitting model that predicts the effects of explorative factors on palp choice. We identified the best by ranking them based on the AICc values using model.sel, and averaged the models using model.avg, if ΔAICc < 2, in the R package MuMIn69.
As no factors in the models had significant effects on palp choice (see “Results”), we checked the values of the positive difference and sperm ratio in relation to the ability to choose the palp containing more sperm (for all mating trials). According to proportional processing, we assumed that a low positive difference or a low sperm ratio might result in random palp choice, since males might not be able to distinguish between small differences. However, a large positive difference or a high sperm count ratio resulted in choosing the palp with more sperm. We manually set the 75th percentile of the boxplots as a cut-off point to distinguish high and low values of positive difference and sperm count ratio. We then performed four Chi-square tests for goodness-of-fit to test for a significant difference in the occurrence of high and low (according to the set cutoff) positive differences and sperm count ratios, depending on the male choice of palp.
Further information on research design is available in the Nature Research Reporting Summary linked to this article.