High-protein paternal diet confers an advantage to sons in sperm competition

Parental environment can widely influence offspring phenotype, but paternal effects in the absence of parental care remain poorly understood. We asked if protein content in the larval diet of fathers affected paternity success and gene expression in their sons. We found that males reared on high-protein diet had sons that fared better during sperm competition, suggesting that postcopulatory sexual selection is subject to transgenerational paternal effects. Moreover, immune response genes were downregulated in sons of low-protein fathers, while genes involved in metabolic and reproductive processes were upregulated.


Introduction
Parental effects can be triggered by diverse factors and describe non-genetic contributions of parents to offspring developmental phenotypes. Maternal effects are well documented, but less-understood paternal effects can also significantly impact offspring phenotypes [1,2], including sexually selected traits [3][4][5], even when males contribute only sperm [1,5,6]. Paternal diet, in particular, can influence offspring traits, if females choosing sperm from males adapted to the local nutritional environment produce offspring with higher fitness [7]. Molecular mechanisms of transgenerational paternal diet effects remain poorly understood but include altered methylation in metabolism-linked loci (reviewed in [8]), perturbed glucose-insulin homeostasis [9], altered cholesterol biosynthesis [10], and modified chromatin states related to obesity [6]. Here, we examine how high-and low-protein paternal larval diet influences postcopulatory sexual selection and gene expression in sons of Drosophila melanogaster.

Material and methods
Experimental D. melanogaster expressed green fluorescent protein (GFP) in sperm heads and ubiquitously in somatic cells for paternity assignment (focal males) or red fluorescent protein (RFP; females and competitor males) in sperm heads [11]. GFP larvae were reared on high-(HP; 200 g yeast : 50 g sugar) or low-protein (LP; 50 g yeast : 50 g sugar) diet known to yield 80 -96% survival [12]. For each treatment, 10 vials were prepared upon eclosion, each with five CO 2 -collected males and five same-stock females reared on standard diet (SD; 100 g yeast : 50 g sugar), housed in SD vials (see electronic supplementary material for more detailed methods). Virgin focal sons were transferred to SD until mating. Three-day-old virgin SD RFP females were first mated with SD RFP competitor males (day 0) in individual vials and provided 6 h opportunities to remate with a focal son for 4 subsequent days (days 1 -4) under continuous observation. After remating, females oviposited on fresh SD food vials for 4 days. Paternity of adult offspring [13] was determined using a Nikon SMZ18 fluorescent stereoscope. GFP-sired progeny, and data were analysed with logistic regressions with binomial error structure (glm in R v. 3.2.0 [14]).
RNA was extracted from two replicates of 20 7-day-old focal sons per treatment using an RNeasy kit (Qiagen) and quantified using Agilent Bioanalyzer. Illumina TruSeq mRNA stranded libraries were constructed, and 76 bp paired-end sequences were obtained on an Illumina NextSeq 500, replicated across two flow cells, with within-sample replicates pooled for further analysis [15]. We performed RNASeq data analysis using the Tuxedo Protocol in the DNA Subway online platform [16] with quality control using FASTX-Toolkit (v. 0.0.13.2). Reads were mapped to the D. melanogaster transcriptome and genome (Ensembl r76, BDPG5) using TOPHAT (v. 2.0.11, [17]). Differentially expressed (DE) genes were identified using CUFFDIFF (v. 2.1.1, [16]) at a q-value , 0.05 after false discovery rate correction [18]. Results were visualized with CUMMERBUND and CYTOSCAPE (for biological networks, [19]) in R.

Discussion
Sons of fathers reared on LP diet fared worse in sperm competition, with associated downregulation of immune response genes and upregulation of genes involved in metabolism and reproduction. Non-mutually exclusive mechanisms of paternal effects on paternity success include seminal fluid and other ejaculate effects [20] and cryptic female choice [21]. Females may have been able to detect treatment-induced variation in male behaviour and may have allocated more resources into reproduction with descendants of high-diet males. It is well known that high-quality diet positively affects male sexual characters [22], fitness [23] and subsequent female choice [24]. Indeed, the gene paralytic ( para) affects courtship song [25] and male olfaction in response to female pheromones [26] and was downregulated in sons of LP fathers. As downregulation of para reduces neuronal excitability [27], it is conceivable that negative fitness effects include lower-quality courtship song and reduced olfaction ability, which are very important factors in female precopulatory choice [28]. However, while higher latency (willingness) to mate and reduced mating duration for males with low-quality courtship song and reduced olfactory ability may be expected, we did not find an effect of paternal diet regime on mating duration, and we did not investigate more detailed behavioural traits to confirm correlational outcomes with the expression of para. Only few studies have so far reported transgenerational effects in relation to diet quality [29,30]. To our knowledge, this is the first study reporting on postcopulatory advantages conferred by parental diet.
Importantly, DE genes confirm the existence of differences between sons of fathers reared on different diets, enabling further investigations of transgenerationally affected sexually selected traits. Antimicrobial peptides (AMPs) are upregulated by D. melanogaster when challenged by Gram-negative bacteria [31,32]. Downregulation of these AMPs in sons of LP fathers in our study might therefore be a form of immunosuppression, which, according to theory, trades off against sexually selected traits [33]. Thus, reproductive fitness of LP sons might have been even lower if immunosuppression had not occurred. Indeed, sexually selected male D. melanogaster that showed higher competitive mating ability had lowered immune function, compared with control males [34].
The two most upregulated genes in sons of low-diet fathers are YP1 and YP2. While the suggested functional annotation, vitellogenesis, is clearly a female-limited function, effects of YP1 and YP2 in male D. melanogaster [35] and the moth Spodoptora littoralis [36] include yolk protein precursors, which directly interact with spermatozoa. YP2 coats the spermatozoa and might provide protection or aid in gamete recognition. However, the functional significance of these proteins has not been established, and we have no knowledge about how upregulation of YP1 and YP2 may influence reproductive fitness in male fruit flies.
The direction of regulation of proteolysis (CG9377), biosynthesis of chitin-based cuticle (Cpr92F and dp) and gamete generation (mei-P26) is consistent with organismal preparation for a suboptimal nutritional environment, investing less and recycling more. Intriguingly, CG9377 has been also found to be upregulated in brains of male D. melanogaster courting females, compared with non-courting males [37], establishing another link of our paternal diet treatment to precopulatory sexual selection (although the direction of the effect seems to promote courtship, rather than reduce it, as discussed above). Valtonen et al. [38] found substantial transgenerational effects of larval diet on development time and adult body size in D. melanogaster, but not on pathogen resistance. The different findings in immune response between [38] and the presented study may be due to the efficiency of the manipulated media. Diet components and protein : carbohydrate ratios are difficult to compare between studies, owing to use of different protein (P) and carbohydrate (C) sources. Crude estimates of P : C ratios and the withinstudy difference between ratios were much higher in our study (low ¼ 0.4, high ¼ 8; [38] assuming 100 g of sugar/ litre diet: low ¼ 0.07, standard ¼ 0.14), illustrating the need to employ a more exact nutritional framework to determine high-resolution reaction norms of traits of interest [39].