The evolution of a placenta accelerates the evolution of post-copulatory reproductive isolation

The evolution of placentation is predicted to intensify intergenomic conflicts between mothers and offspring over optimal levels of maternal investment by providing offspring opportunities to manipulate mothers into allocating more resources. Parent-offspring conflicts can result in the evolution of reproductive isolation among populations when conflicts resolve in different ways. Postzygotic reproductive isolation is hypothesized to evolve more rapidly following the evolution of placentation due to the predicted increase in conflict. We tested this hypothesis by performing interpopulation crosses within placental and non-placental species of Poeciliopsis to determine if the relationship between genetic distance and measures of postzygotic reproductive success differed as function of reproductive mode. We did not observe any differences in offspring viability or sterility among crosses. Offspring size declined rapidly as a function of interpopulation genetic distance within the placental species, but not among our non-placental species. The decrease in offspring size in the placental species was beyond normal variation, likely representing a major fitness cost, consistent with the prediction that negative epistatic interactions are evolving more quickly among populations in our placental species than the non-placental species. We discuss how our results support the role parent-offspring conflicts play in the evolution of reproductive isolation and reproductive mode.

Methods
We assessed interfertility among populations from three species of Poeciliopsis. We selected Poeciliopsis prolifica and Poeciliopsis infans to represent a paired comparison of closely related placental and non-placental species from the same clade, respectively (Reznick, Mateos et al. 2002, Pollux, Meredith et al. 2014). Poeciliopsis gracilis, was included as an additional non-placental species from a separate clade (Reznick, Mateos et al. 2002). We established lab populations at the University of California Riverside with 15 wild-caught pregnant females and five adult males from 4 populations of P. prolifica, 4 populations of P. nfans, and 2 populations of P. gracilis (Table S1) between 2014-2015. Fish from an additional population of P. gracilis collected in 2004 and maintained in 3 stock tanks were also included (Table S1). To ensure allopatry, we collected each population from a different river (Figure 1; Table S1). Schrader et al. (2013) found in their crosses among populations of H. formosa that differences among populations in offspring size cause reductions in the viability of offspring from hybrid crosses. We exploited existing data for P. infans and P. gracilis (Frías-Alvarez, Macías Garcia et al. 2014) to choose sites with large differences in late-stage embryo size. We recorded water pH, water temperature, canopy cover, and elevation at each collection site (Table S1) and combined that with the annual rainfall data for each locale so that we could assess the potential influence of local adaptation as a contributor to our results. We do not have environmental data for the population of P. gracilis collected in 2004. Wild-caught adults were housed in 19 or 38 L aquaria with clumps of aquatic moss (Vesicularia dubyana) to provide cover for newborn offspring. We removed F1 offspring daily and reared them in group tanks. We used the metamorphosis of the male anal fin into the gonopodium, the intromittent organ, to identify the sex of individuals while they were immature, then moved them into single-sex aquaria to rear them to maturity. Lab-born male tanks were seeded with two females from the same population to stimulate sperm production. As lab-born females approached sexual maturity, we isolated them into individual 2-gallon tanks containing gravel and aquatic moss, and fed them ad libitum. Once individual lab-born females reached sexual maturity, we used them in a single intrapopulation or interpopulation cross.

We performed three replicate crosses in all possible interpopulation cross directions, and six intrapopulation crosses within each population as a control. We placed lab-born adult males into the tanks of isolated virgin females for 7 days, after which time they were removed. We monitored the mated females daily in order to record the latency time (in days) between mating and producing their first brood. Once females gave birth to their first brood, daily monitoring continued for a period of 60 days, after which females were euthanized using MS-222 and preserved in 95% ethanol. We removed newborn offspring from the female tanks daily to ensure individuals were less than 24-hours old at the time of collection. The first 10 offspring born to a given female were immediately euthanized in MS-222 and preserved in 95% ethanol. All remaining offspring of both sexes were placed in a 2-gallon stock tank and allowed to reach sexual maturity. We monitored these tanks for 6 months to see if the offspring that resulted from interpopulation crosses produced offspring of their own as a measure of hybrid sterility, recorded as the presence of healthy newborn offspring within the tank. Crosses were deemed unsuccessful if females showed no visible signs of pregnancy within 120 days following the end of their 7-day male exposure period. We took multiple measures from the outcome of each cross to serve as an index of fitness. The total number of offspring born to a given female over the 2-month monitoring period was recorded as a measure of fecundity. Preserved offspring were measured for body length (mm), wet weight (mg), and dry weight (mg) at birth. We dissected females and removed all embryonic tissue to record the total number of developing embryos and the wet weight (g) of all the reproductive tissue. We then treated the wet mass of all reproductive tissue, including embryos divided by the total wet mass as a dependent variable (reproductive allocation) which represents the proportion of total wet mass devoted to reproduction. All species in the genus

Poeciliopsis have superfetation, which means that a reproducing female normally carries multiple litters of developing offspring, each in discretely different stages of development. A consequence of superfetation is that litters of offspring are produced more frequently. Reproductive allocation is thus a composite of number of offspring per litter, offspring size and number of developing litters. Reproductive allocation was analyzed as a proportion of female wet weight to account for differences in the size of individual females.
Embryos were independently scored as viable or inviable by a group of individuals unfamiliar with the working hypothesis, based on whether the embryos exhibited the typical phenotype of a normally developing embryo (Haynes 1995). Inviable embryos exhibited the characteristic markers of their developmental stage, but were typically smaller, duller in color, and more semisolid than viable embryos. Embryos were scored as ambiguous whenever there was disagreement between individual scorers or when they only exhibited a subset of the characteristics of inviability. We measured the focal females for body length (g) and wet weight (g) to control for the influence of female size on offspring size and number. All procedures were approved and done in compliance with the guidelines set by the Institutional Animal Care and Use Committee (IACUC) at the University of California Riverside (Animal use protocols #20110007 and #20140003).

Sequencing and Bioinformatics
To assess genetic distance among the populations, we extracted DNA from the tail tissue of five male and five female wild-caught individuals from each population. Extractions were performed using a Qiagen® DNeasy Blood & Tissue Kit, with two modifications to the spin-column protocol; we extended the duration of proteinase K treatment to 10 hours then incubated each sample in 8 μL of RNAse for 30-minutes at 37 O C. We quantified DNA concentrations with a Qubit 2.0 Fluorescence Reader, and each sample was adjusted to a concentration of 10ng/ μL using a Zymo Research DNA Clean & Concentrator™ kit. ddRAD sequencing was performed by the University of Texas at Austin Genomics Sequencing and Analysis Facility (GSAF), including enzyme digestion, size selection, adaptor ligation, and sequencing. We selected the EcoRI-MspI enzyme pair for digestion, and 200-300 base-pair fragments were retained for sequencing. 2x150bp reads of the digested samples were obtained using the Illumina HiSeq 2500 system. We performed all demultiplexing, de novo assembly, and genetic distance calculations using the STACKS pipeline (Catchen, Hohenlohe et al. 2013). Each species was processed independently and all individuals within a species were processed and analyzed simultaneously. We used the process_radtags program to filter out low quality reads with raw phred scores below 10 (< 90% probability of being correct) and uncalled bases. The STACKS core modules (‘ustacks’, ‘cstacks’, ‘sstacks’, and ‘populations’) were executed through the denovomap.pl program with a minimum stack depth of 3 reads, a maximum number of mismatches allowed between loci within an individual (M) of 2, a maximum number of mismatches between loci in a catalog (m) of 1, and the deleveraging and highly repetitive stack removal algorithms enabled. We used the ‘populations’ function to estimate intrapopulation heterozygosity and pairwise F ST between all population combinations (Table S2).