We mined the genomes of the 11 newly sequenced Drosophila species for homologs of the D. melanogaster peptide precursor genes Akh (CG1171), Ast (CG13633), Ast-C (CG149199), capa (CG15520), Ccap (CG4910), Crz (CG3302), Dh (CG8348), Dh31 (CG13094),ETH (CG18105), Fmrf (CG2346), hug (CG6371), IFa (CG33527), Leucokinin (CG13480), Mip (CG6456), Dms (CG6440), npf (CG10342), Nplp1 (CG3441), Pdf (CG6496), Proct (CG7105), Dsk (CG18090), sNPF (CG13968), and Dtk (CG14734). We then predicted the encoded neuropeptides; an overview of their numbers is given in Table 1. With the exception of the FMRFa-like peptides (see below), the analyzed genes code for the same number of neuropeptides in each species (43 in total, plus 10-17 FMRFa-like peptides). The translated coding sequences for the prepropeptides and predicted peptides are given as Additional data files 1 and 2.

In Drosophila larvae, the main neurohemal organs that store and release peptide hormones are the ring gland, and the thoracic and abdominal perisympathetic organs. The epitracheal cells (Inka cells) are endocrine glands along the trachea. These tissues represent a rich source of neuropeptides: their peptidome contains about half of all known D. melanogaster neuropeptides 35 36 . To biochemically assess whether the neuropeptides are expressed and processed as predicted, we directly profiled these neurohemal organs in D. sechellia, D. pseudoobscura, D. mojavensis and D. virilis. These species cover main phylogenetic lines within Drosophila. Obtained masses in the range of 850-2,500 Da were matched to the theoretical masses of predicted peptides (Table 2). This - and the observed tissue distribution - revealed that the peptidome of the investigated peptide release sites is identical in all species, at least in the mass range up to 2.5 kDa. In other words, all fruit flies appear to store the same set of (ortholog) peptides as D. melanogaster in the respective neurohemal release sites 35 36 .

Alignment of the prepropeptide sequences showed that the peptide families of each Drosophila species consist of an identical set and number of ortholog neuropeptide copies, with the exception of FMRFa-like peptides (Table 1; Additional data files 1 and 2). For example, in all species the crustacean cardioactive peptide (CCAP) precursor contains one CCAP, and the allatostatin A (ASTa) precursor contains 4 ASTa peptides. The FMRFa precursor, however, encodes 10 FMRFa-like peptides in D. mojavensis and D. virilis, 11 FMRFa-like peptides in D. ananassae, 12 FMRFa-like peptides in D. willistoni, 14 FMRFa-like peptides in D. erecta, 17 FMRFa-like peptides in D. grimshawi, and 13 FMRFa-like peptides in all other species. The fmrf gene contains 2 exons, of which exon II codes for the whole FMRFa prepropeptide. The differences in peptide-coding sequences can thus not be explained by exon duplication. Higher numbers of tandem repeats exist for FMRFa-2 (DPKQDMRFa; for example, 5 copies in D. melanogaster, 7 copies in D. grimshawi) in all species but D. mojavensis and D. virilis. This may suggest that mispairing of template versus replicating nucleotide sequences coding for this peptide has resulted in insertions/deletions during Drosophila evolution and has caused the high number of FMRFa-like peptide copies.

Two prepropeptides contain neuropeptides that are usually not grouped into the same peptide family: the CAPA prepropeptide contains two periviscerokinins and one pyrokinin, and the neuropeptide-like precursor (NPLP)1 prepropeptide contains one MTYamide, one IPNamide and one non-amidated peptide. The CAPA pyrokinin and the NPLP1 peptides have therefore been treated as single copy peptides (but see Discussion).

If the neuropeptide sequences are subjected to stabilizing selection due to their signaling function, it is reasonable to assume that the peptide-coding parts of the prepropeptides are more conserved than the spacers (the parts separating the bioactive peptides), which by existing evidence do not act as signaling molecules in insects. In other words, the sequence similarity between ortholog neuropeptide parts of the prepropeptides is likely to be higher than the sequence similarity of ortholog spacer parts. To test this hypothesis, it is not sufficient to simply calculate amino acid identities, since substitutions of amino acids do not occur randomly but are correlated with their physico-chemical characteristics 42 . We thus calculated the overall average amino acid distance D_so for each set of orthologs (Figure 5) based on the Jones-Thornton-Taylor (JTT) matrix 43 as a more appropriate measure of sequence variation (see Material and methods). The raw values are listed in Additional data file 5. The median D_so between peptide orthologs was 0.041, and thus significantly lower than the calculated 0.408 for the spacers (Figure 5; Mann-Whitney, two-tailed p < 0.0001, U = 211.5), although the sequence of several spacers was quite conserved (for example, in the CCAP or CAPA prepropeptides). In contrast to the peptides (p < 0.01), the spacer distances followed a Poisson distribution.

A closer look at the data (Additional data file 5) shows that high D_so values only occur in multiple copy peptide families. For example, neuropeptide F (NPF) and MTYa show the highest D_so for single copy families (0.134 and 0.093, respectively). The respective maximum values for multiple copy families are 0.748 for FMRFa-7, 0.601 for sulfakinin (SK)-0, 0.267 for ETH-2, 0.208 for FMRFa-7, and 0.205 for CAPA-PVK-2. Yet, orthocopy sets without sequence variation or with low D_so values occur not only in single copy, but also in each multiple copy peptide family: CAPA-PVK-1 (0.042), ETH-1 (0.025), SK1- (0), ASTa-3 and -4 (0), sNPF-1 (0), myoinhibiting peptide (MIP)-3 and -4 (0), Drosophila tachykinin (DTK)-3 (0.02) and FMRFa-2 and -6 (0).

To test for differences in the sequence variability between single and multiple copy peptide families, we computed the average amino acid distance D_af for each amino acid position between all paracopies within a peptide family (Figure 6a, d) and then calculated the mean (Figure 6b). For single copy peptides, we calculated the corresponding average amino acid distance D_ao for the respective orthologs (Figure 6c). The results in Figure 6 show that the mean D_af between paracopies of multiple copy peptide families is typically higher than the D_ao observed between the single copy peptides. Due to a large standard variation, these differences are only significant for amino acid positions 5 and 7 from the carboxyl terminus (paired t-test, p < 0.05). This reflects the spread of sequence variation in multiple copy peptide families. For most amino acid positions there are families that show no variation, and, at the same time, families with considerable sequence variation. The high mean D_ap at position 1 from the carboxyl terminus mostly originates from the sNPFs, which end either RFa (sNPF-1 and -2) or RWa (sNPF-3 and -4). There is no clear tendency that the sequence variation increases from the carboxyl to the amino terminus; a correlation between D_af and copy number is not discernible (Figure 6d).

The distance D_af contains both the sequence variation between individual orthologs (inter-ortholog variation) as well as between individual paracopies (inter-paracopy variation; Figure 1). To test the contribution of the inter-ortholog variation to D_af, we calculated the average amino acid distance D_ao for each amino acid position for each set of orthocopies individually (Figure 7). A comparison of Figures 7c and 6b shows that the mean D_ao for the ten carboxy-terminal amino acids is considerably smaller than the mean D_af and not significantly different between single and multiple copy peptides. This region likely contains the active core of the peptides, which typically consists of the last five to seven carboxy-terminal amino acids and the amidation signal (for example, 44 45 46 47 ). Somewhat higher D_ao values occurred for more amino-terminal amino acids. This shows that the orthologs are strongly sequence-conserved throughout Drosophila, irrespective of whether they belong to a single or multiple copy peptide family - with the exception of FMRFa-7 and SK-0 (see below).

We next calculated the average net amino acid distances (Figure 1) between paracopies D_anp for multiple copy peptide families; results are shown in Figure 8. The mean D_anp was higher than the mean D_ao of single (Figure 8b) or multiple copy (compare Figure 7c) peptides throughout amino acid position 1-11. This was again only significant for positions 5 and 7 due to the high standard variations (paired t-test, p < 0.05). As for D_af, the high variation of D_anp reflects the spread of the degree in sequence variation between multiple copy peptide families: given positions were variable in some peptide families, but imvariantly occupied by the same amino acid in others. The high D_anp of 1.69 at position 1 from the carboxyl terminus is again caused by the carboxy-terminal difference RFa and RWa between the sNPFs. When omitting the RWamides sNPF-3 and -4 - which could not be biochemically detected yet - this value drops to 0.42. With this value, it seems that the amino acids at positions 1-3 and 8 from the carboxyl terminus are the most conserved amino acids between the paracopies of each multiple copy peptide family.

The majority of G protein-coupled peptide receptors of D. melanogaster have been deorphanized, with some still uncharacterized to date 48 . From the obtainable literature 48 49 50 51 , we compiled the number of characterized D. melanogaster G protein-coupled receptors per peptide family. Although these numbers may be subjected to future changes, to date there are either only one or two receptors known for the paracopies of each peptide family. The occurrence of two receptors for some peptide family opens the possibility for receptor-ligand coevolution and subsequent sub- and neofunctionalization of paracopies. To test for this, we plotted the D_anp between multiple copy peptide families with one known receptor against those with two known receptors. Although there are differences (Figure 9), they are neither statistically significant nor do they follow an obvious pattern. This result speaks against a sub- or neofunctionalization of paracopies during the evolution of Drosophila.

The obtained mass fingerprints of the neurohemal organs and endocrine cells were identical: in each species, the obtained masses corresponded to the respective orthologs in D. melanogaster neurohemal organs or peritracheal cells 35 36 . Vice versa, for each peptide characterized in the neurohemal organs or peritracheal cells of D. melanogaster, there was a mass present that corresponded to the respective ortholog in the other species. Similar to the use of peptide fingerprints in proteomics, the exactly matching tissue fingerprints chemically identify the underlying peptides and precursor products with high probability. All fingerprint masses matched the respective theoretical masses calculated for the in silico predicted peptides. In conclusion, the mass spectrometric profiling supports our in silico prediction of the neuropeptidome.

The finding of identical peptide hormone complements in the mass range of 800-2,500 Da in main Drosophila phylogenetic lineages suggests that the peptidome of the major neurohemal organs and the peritracheal cells has been evolutionary stable for at least 40-60 million years since the divergence of the Drosophila species from their last common ancestor. Obviously, all Drosophila species share the same peptidergic hormonal communication possibilities. Even more, our genomic comparisons suggest that the whole peptidome is highly conserved throughout the genus Drosophila. We observed no loss of peptide precursors, or individual peptides as suggested to have occurred, for example, between flies and mosquitoes 55 . Thus, the number of peptide copies in each precursor was identical throughout the species with exception for the FMRFamides, which most likely duplicated by unequal recombination. These recombination events must have occurred independently from each other, since multiple repeats of FMRFa-2 coding sequences are present both in the Sophophora subgroup and the Hawaiian species of the Drosophila subgroup (D. grimshawi), but are lacking in the other Drosophila subgroup species D. virilis and D. mojavensis. Unequal recombination is also the likely mechanism behind the duplication of most other multiple copy peptides, but for them recombination must have occurred prior to Drosophila speciation.

The high conservancy of the peptidome is remarkable and unexpected, since drosophilid flies have undergone several radiations and have adapted to a variety of environments with, for example, a very different supply of water, such as sea shores, forests and deserts 34 56 . In contrast, the genome of the tenebrionid beetle Tribolium castaneum shows a gene expansion for putative diuretic peptides 57 . This has been interpreted as an adaptation to dry conditions in tenebrionids, a beetle family that thrives in deserts and other very dry places. It is, however, unclear whether this is a special tenebrionid or a common beetle feature. At least for fruit flies, our data show that the adaptation to different environments is not paralleled by changes in the number or increased sequence variability of diuretic hormones or other neuropeptides.

In our analysis, spacer sequences showed significantly higher amino acid distances than peptide sequences. This suggests that Drosophila neuropeptides are subjected to stabilizing selection or evolutionary constraint to a much larger extent than spacer sequences. This is further supported by the non-random distribution of peptide distances not observed for spacers. This finding may not be unexpected, but is shown here for the first time on a neuropeptidome level.

In Drosophila, there is a higher proportion of highly constrained codons in essential genes than in any other dispensability class 58 . As hypothesized at the outset, stabilizing selection and the resulting sequence conservation may thus indicate functional importance of neuropeptides, signaling molecules for which single amino acid exchanges can result in drastically altered receptor efficacy, binding or effect (for example, 28 59 ). If this hypothesis is correct, then the observed low inter-orthocopy distances (D_ao, D_so) indicate that the multiple peptide copies are functionally important and not individually dispensable.

The observed higher amino acid distances that reflect a considerable sequence variation for the spacers do not allow us to conclude that structural features of the spacers are unimportant for proper peptide processing and packaging into secretory vesicles. They speak, however, against a general signaling function of the spacer regions ('associated peptides') at the receptor binding site, where single amino acid changes can already result in altered efficacy, effect or specificity (for example, 28 59 ). Nevertheless, this conclusion needs proper physiological testing. Several spacer regions are quite conserved throughout the Drosophila species (for example, in the CAPA and CCAP precursor), and a FMRFa-spacer-derived peptide has been shown to modulate the activity of FMRFa at the receptor in Lymnea 60 .

The comparably high D_anp distances show that there is sequence-variation between paracopies (inter-paracopy variation). Assuming that paracopies at some point have arisen from a common ancestor, we have hypothesized at the outset that newly duplicated paracopies can escape selection pressure and may be allowed to drift neutrally. However, the small D_ao distances between orthocopies (inter-orthocopy variation) do not support this scenario. A significant difference in sequence variation between the individual sets of orthocopies was not observed between single and multiple copy peptide families, and inter-orthocopy distances were small compared to the distances found between spacer regions. This suggests that: the inter-paracopy variation originates from a time before divergence of the Drosophila taxa; and paracopies have never fully escaped selection pressure and have never experienced a phase of neutral mutation. Hence, the classic theory for duplicated genes may only apply in a limited sense for paracopies.

At the outset, we reasoned that peptide copies following the 'more-of-the-same' concept will show low sequence variation between paracopies. For paracopies with differential activities, we expected at the same time an increased sequence variation between paracopies. Since the inter-paracopy distances D_anp were similar for all multiple copy peptide families and did not correlate with receptor number, it is not possible to draw conclusions from our data in this respect. For Drosophila, there are also no data regarding different half-lives during peptide degradation, or induction of ligand-selective receptor conformation and activity 61 as has been demonstrated for locust and cockroach AKHs 27 28 .

The CAPA and NPLP1 prepropeptides contain neuropeptides that are usually not grouped into the same peptide family. We have therefore treated the CAPA pyrokinin and the NPLP1 peptides as single copy peptides. However, some sequence similarities can be found between CAPA pyrokinins and periviscerokinins, and between the amino-terminal stretches of the NPLP1 peptides 62 . It has also been shown that the CAPA pyrokinin specifically activates a G protein-coupled receptor (CG9918) that is evolutionarily related to the other Drosophila pyrokinin receptors, but is not activated by CAPA periviscerokinins and the HUG-pyrokinin at physiological concentrations 63 ; data on NPLP1 peptide receptors are not yet available. Thus, it is possible that at least the CAPA peptides are, in fact, an example of multiple copies that have sub- or neofunctionalized by acquiring sequence variation: the CAPA prepropeptide appears to date back at least to the origin of insects, since it contains a few periviscerokinins plus one highly sequence-conserved pyrokinin in all insect taxa investigated so far 64 . If this is the case, this sub- or neofunctionalization must have occurred a long time before the radiation of Drosophila. While this justifies the classification of at least the CAPA pyrokinin as a single copy peptide in this study, it emphasizes the need for further comparisons similar to that reported here for Drosophila on larger phylogenetic units spanning longer evolutionary time frames. Such studies will soon become possible with the increasing number of fully sequenced insect genomes.

Although the inter-orthocopy distance D_so was, in general, very low throughout the peptides, we observed differences in D_so within the CAPA-PVK, ETH, FMRFamide and tachykinin peptide families. Can these differences be correlated to differential activities? A comparison of the calculated D_so values with the available pharmacological data shows that there is at least a correlation to the reported efficacies: the paracopies with lower amino acid distance typically are the ones with higher receptor or pharmacological activity.

For the ETHs, the EC₅₀ of the paracopy with the lowest sequence variation (ETH-1) is nine times more potent in heterologous receptor assays than the more sequence-variable ETH-2 65 66 . Two groups have characterized the FMRFa receptor of D. melanogaster in heterologous expression systems. Both found that FMRFa-6 (PDNFMRFa) has the highest potency to activate the FMRFa receptor, whereas FMRFa-7 has no activity at all 67 68 . Meeusen and colleagues 68 report similar EC₅₀ values for FMRFa-2 to -5. All were only slightly less potent than FMRFa-6. Cazzamali and Grimmelikhuijzen 67 found the following order: FMRFa-6 > FMRFa-2 > FMRFa-3,-5,-8 > FMRFa-1 > FMRFa-4. This pharmacological ranking correlates well with the calculated D_so distances (in brackets): FMRFa-6 = FMRFa-2 (0) < FMRFa-5 (0.04) < FMRFa-8 (0.064) < FMRFa-4 (0.11) < FMRFa-1 (0.127) < FMRFa-3 (0.172) << FMRFa-7 (0.748).

For the DTKs, our data predict the following ranking of pharmacological activity: DTK-3 > DTK-1 > DTK-6 > DTK4 = DTK-5 > DTK-2. This again corresponds quite well with the efficacy of DTKs on DTKR - one of the two DTK receptors known - in HEK-293 cells 69 : DTK-1 > DTK-3 = DTK-6 > DTK-4 > DTK-5 > DTK-2. For CAPA-PVKs, the copy with the lower sequence variation (CAPA-PVK-1) is more effective in inducing calcium responses and fluid transport in the Malpighian tubules 70 and about 1.5-times more potent in receptor assays than the more sequence variable CAPA-PVK-2 65 71 . In other words, the more potent peptides were typically the more sequence-conserved. This might extend well beyond Drosophila. For example, CAPA-PVK-2 orthologs are much more sequence-variable in their carboxyl terminus than CAPA-PVK-1 orthologs not only in Drosophila, but also in other flies 72 73 . At the same time, the housefly Musca domestica CAPA-PVK-2 shows a ten-times diminished efficacy in fluid secretion assays on housefly Malpighian tubules compared to M. domestica CAPA-PVK-1 45 . The degree of sequence variation appears not to be linked with peptide position along the precursor.

In contrast to the CAPA-PVKs, ETHs, FMRFamides and DTKs, the SK-1 and -2, MIP and ASTa copies all showed a consistently low inter-orthocopy distance. The pharmacological profiles of SK and MIP copies on their respective receptors have not been characterized so far, but such data exist for the two Drosophila ASTa receptors (DARs) expressed in Chinese hamster ovary (CHO) cells 74 75 . The data suggest that DAR-1 has a lower sensitivity for ASTa-4 than for ASTa-1 to -3, whereas DAR-2 is more sensitive to ASTa-3 to -4. It has, however, to be kept in mind that not all peptide receptors have been deorphanized to date, and that signaling properties and specificities of receptors may be changed by modifying proteins such as RAMP or RGS in native cells. It is also possible that ligand-selective receptor conformations may exist 61 , and the ligand properties and activated intracellular pathways in vivo may be different to those in heterologous expression systems. This may explain why - in contrast to the data from heterologous expression systems - all FMRFamides had a similar dose-response effect at the neuromuscular junction 22 . Only FMRFa-7 (SAPQDFVRSa) was inactive in all systems. Clearly, further data, especially from bioassays, will be needed to confirm the observed correlations between sequence variation and efficacy.

The evolution of neuropeptides and their receptors is linked, and neuropeptide receptors are under evolutionary pressure to maintain a high affinity to the authentic ligands 9 59 . The finding that the more sequence-variable neuropeptides typically had a lower pharmacological efficacy does not speak for the occurrence of fast adaptive structural changes of G protein-coupled receptors to maintain a high ligand affinity to sequence-variable peptides during the evolution of Drosophila. Does that mean that peptide copies with high amino acid distances are functionally unimportant? The by far highest distances were found for FMRFa-7 (0.748) and SK-0 (0.601). The carboxyl terminus of FMRFa-7 (FVRSa) is highly deviated and is the likely cause for its lack of receptor activation and bioactivity 22 67 (and see above). The high D_so value of FMRFa-7 is around the mean value found for spacer regions, which suggests that FMRFa-7 has escaped selection pressure to a considerable amount. The high D_so value for SK-0 correlates with its inactivity at the DSK-R1 receptor 76 and its lack of bioactivity at physiological concentrations below 1 μM 77 . Unlike SK-1 and -2, SK-0 has, furthermore, not been found biochemically so far 35 37 . Hugin-gamma, a predicted but obviously not processed peptide 78 that seems to be missing from the genome of D. persimilis 55 , shows a D_so of 0.259. Nevertheless, the synthetic D. melanogaster HUG-gamma is still able to activate the receptor 65 79 . We propose from this as a rough estimate that the peptides with an amino acid distance above 0.6 are nonfunctionalized. Distances below 0.3 and the non-Gaussian distribution of all other peptide copies suggest that they are under stabilizing selection that prevents nonfunctionalization by random or deleterious mutations.

D. virilis was obtained from a colony in Ulm. Genome library strains of D. mojavensis,D. pseudoobscura, and D. sechellia were obtained from the Tucson Drosophila Stock Center. Flies were kept at standard conditions - a light:day cycle of 12 h:12 h and either 18°C or 25°C. D. virilis and D. sechellia were raised on standard cornmeal medium, D. mojavensis and D. pseudoobscura on standard banana-Opuntia medium.

Peptide precursor genes were identified by tblastn homology searches against the respective D. melanogaster coding sequences using the PAM30 matrix of the Drosophila species BLAST site 80 . The coding sequences were identified and translated with GENSCAN 81 and compared with the GLEAN-predicted sequences in the databank. Amino acid sequences were aligned with the ClustalW algorithm implemented in MEGA3.1 82 and plotted using GeneDoc 83 . Signal peptides were predicted by SignalP 3.0 84 . Mono-isotopic masses of the predicted bioactive neuropeptides were calculated with Data Explorer 4.0 software (Applied Biosystems, Darmstadt, Germany).

We predicted the processed bioactive peptides based on cleavage site consensus sequences 85 and comparison with the chemically characterized processing products from D. melanogaster 35 36 37 . Mono-isotopic masses were calculated for all peptides and listed for each species. Peptide designations were inferred from prepropeptide alignment with the ortholog D. melanogaster sequence, for example, the myoinhibiting peptide encoded on the Mip orthologs that aligned with D. melanogaster MIP-3 was also named MIP-3. This allows easy identification and reference to ortholog peptides. In the fmrf-precursor of D. melanogaster, several paracopies are sequence-identical and named either FMRFa-2 or -3. These paracopies and their orthologs were designated according to their position on the gene as FMRFa-2', FMRFa-2", and so on.

The parts of the prepropeptides between the signal peptide and the bioactive peptide sequences flanked by the mono- or dibasic cleavage sites were assigned as spacers. For distance calculations of peptides, sequences were aligned from their carboxyl terminus. Gaps that occurred due to variable peptide copy length were deleted pairwise. Spacers were aligned by the ClustalW algorithm prior to distance calculation. Average distances were calculated as absolute values in MEGA3.1 for pairwise comparisons as outlined in Figure 1 by iterative procedures under a maximum likelihood formulation using the JTT matrix 43 82 . The JTT matrix was calculated from data of the Swiss-Prot protein sequence database and gives a measure of the probability that a given amino acid i is being replaced by residue j per occurrence of j 43 . Variable mutation rates among sites were assumed. Since the peptide sequences are too short to reliably estimate gamma parameters, we adopted a gamma distance with α = 2.4, which is very close to the true distance for sequence divergence under the JTT model 86 . Data were processed and plotted using Microsoft Excel and GraphPad Prism 4.0 (GraphPad Software, San Diego, CA, USA).

The central nervous system was dissected free from all surrounding tissue in standard Drosophila saline. Ring glands of L3 larvae (selected after size: D. virilis >3 mm; D. mojavensis, D. sechellia >2 mm; D. pseudoobscura >2.5 mm) were punched out using pulled glass capillaries and spotted directly onto the MALDI target and left to dry. For isolation of the thoracic and abdominal neurohemal sites, the thoracic or abdominal part of the ventral ganglion of adults was cut out and the lateral parts were removed using fine scissors. The dorsal neural sheath was then isolated and freed from cells using tungsten micro-needles. The isolated sheaths were transferred to the MALDI target using pulled glass capillaries and left to dry. This method results in clean spectra from the neurohemal endings 35 36 . For direct profiling of the peritracheal cells, the main branches of the trachea from L3 larvae were dissected free from other tissue and transferred directly onto the MALDI target using fine insect needles. The peritracheal cells were targeted by directing the laser beam to the obtuse angle between the main trachea and the diverging first order trachea.

To remove salts, a small droplet of ice-cold water was added onto the dried tissues, and aspirated off after about 1 s. Small nanoliter volumes of matrix (saturated solution of re-crystallized α-cyano-4-hydroxycinnamic acid in 30% MeOH/30% EtOH/0.1% trifluoroacetic acid for neurohemal organs, or 60% acetonitrile/0.1% trifluoroacetic acid for peritracheal cells were added to the samples using a manual oocyte injector (Drummond Scientific, Broomall, PA, USA). In prior tests MeOH/EtOH/H₂O (30:30:40) resulted in improved mass spectra compared to 60% MeOH as previously described for D. melanogaster 35 36 .

MALDI-TOF mass spectra were acquired in positive ion mode on a Voyager DE RP mass spectrometer (Applied Biosystems, Darmstadt, Germany) equipped with a pulsed nitrogen laser emitting at 337 nm. Samples were analyzed in reflectron mode using a delayed extraction time of 400 nsec and an accelerating voltage of 20 kV. To suppress matrix signals, the low mass gate was set to 850 Da. Laser power was adjusted to provide optimal signal-to-noise ratios. Data were analyzed using Data Explorer 4.0 software (Applied Biosystems), with a mass tolerance of 0.5 Da.