Drosophila Sex Peptide controls the assembly of lipid microcarriers in seminal fluid

Significance Seminal fluid plays a critical role in reprogramming female physiology and behavior to promote male reproductive success. We show, in the fruit fly, that specific seminal proteins, including the archetypal “female-reprogramming” molecule Sex Peptide, are stored in male seminal secretions in association with large neutral lipid-containing microcarriers, which rapidly disperse in females. Related structures are also observed in other Sex Peptide-expressing Drosophila species. Males lacking Sex Peptide have structurally defective microcarriers and exhibit abnormal transfer of many seminal proteins to females. Our data reveal that this key signaling molecule in Drosophila seminal fluid is also a microcarrier assembly factor that modulates transfer of other seminal factors and that this may be a more evolutionarily ancient role of this protein.

250μl of dilute trypsin was added and the solution incubated at 37°C overnight. The resulting peptides were extracted via two repeated ACN replacements before drying and desalting them in Sola SPE columns (Thermo Scientific). Prior to LC-MS/MS, the peptides were suspended in a 2% ACN/0.1% formic acid buffer.
For peptide analysis, we used a LC-MS/MS platform composed of a Dionex Ultimate 3000 and a Q-Exactive HF mass spectrometer (Thermo Scientific). Peptide loading took place in a solution of 0.1% TFA in 2% ACN on a trap column (PepMAP C18, 300μm x 5m, 5μm particle; Thermo Scientific). Peptides were separated using an easy spray column (PepMAP C18, 75 μm x 500 mm, 2 μm particle size; Thermo Scientific) with a gradient of 2% ACN to 35% ACN in 0.1% FA in 5% DMSO. For MS spectra collection, a resolution of 60,000 was used in profile mode on the Q-Exactive HF (ion target = 3x106). The top 12 most intense features were selected for subsequent MS/MS analysis (resolution of 30,000). The following parameters were set: dynamic exclusion = 27 seconds; AGC target = 1x105; isolation width = 1.2 m/z; and maximum acquisition time = 45ms.

MS data processing
The MS data processing pipeline employed has previously been described (1). RAW data were imported into Progenesis QIP (version 4.1.6675.48614), exporting spectra as MGF files using the 200 most intense peaks without deconvolution for searching. For peptide identification, the Drosophila melanogaster UniProt reference proteome was used as a search target, with database retrieval conducted on 30/03/2015 (21361 sequences) in Mascot 2.5.1. The search parameters incorporated the following: Oxidation (M) as a variable modifications; Propionamide (K), Propionamide (N-term) and Propionamide (C) as fixed modifications; one missed cleavage sites; 0.05 Da fragment mass accuracy; 10 ppm precursor accuracy. Prior to importing the search results into Progenesis for quantification via the Top3 method, a peptide-level 1% FDR was applied alongside a further Mascot ion cut-off of 20. The resulting protein abundance data were subsequently normalised using the internal Progenesis algorithm to a set of housekeeping proteins.

Data analysis
As expected, significantly lower abundance of SP was detected in the SP null treatment compared to controls (LM: F1,12=30.726, p=0.0009). However, surprisingly SP peptides were detected in all null male samples at approximately 17% of control abundance, which we attribute to read-through of the stop codon in the null allele, a common event in Drosophila (3). Whether this protein is correctly processed and secreted remains unclear. To confirm that the presence of SP in our null samples was not a result of carry-over from previous control samples run on the MS, we independently re-ran two SP null samples: again, SP was detected, thus verifying our findings. SP was omitted from further analysis to avoid the confound of attributing changes to the SFP proteome to altered SP levels. We further omitted S-Lap7 and Spn28Db due to high between-replicate variability.
Hierarchical clustering analyses were performed on log2 abundances using a Pearson correlation distance metric and graphically displayed using the pheatmap package. For each protein, a mean abundance was taken across replicates. The first five earliest-branching clusters were selected for further analysis. To visualise the general abundance profile across treatments that each cluster captured, the non-log2 transformed abundance for each protein was divided by the mean calculated across all samples and averaged across replicates. Mean centring in this way gives a measure of abundance change that is comparable across the substantial variance range of proteins. We tested the significance of a relationship between the protein abundances within a cluster and our measured variables using linear mixed effects models that modelled protein identity as a random effect and mating and genotype as fixed effects. Protein abundances were averaged across the three replicates for each treatment combination to improve the model fit, which was inferred through visual inspection of diagnostic plots. The statistical significance of factors was assessed by analysis of deviance using the 'drop1' function. Where the interaction term was insignificant, the model was refitted without it.
To identify proteins that are not classified as SFPs, but which have a profile that suggests they might fail to be transferred or be transferred in excess quantities in SP null versus rescue males, we iterated a linear model over every non-SFP protein in our dataset. A hierarchical clustering analysis (as above) was then performed on all proteins that gave a significant (p<0.05) interaction between genotype and mating status.