High-Resolution Mass Spectrometry Driven Discovery of Peptidic Danger Signals in Insect Immunity

The ‘danger model’ is an alternative concept for immune response postulating that the immune system reacts to entities that do damage (danger associated molecular patterns, DAMP) and not only to entities that are foreign (pathogen-associated molecular patterns, PAMP) as proposed by classical immunology concepts. In this study we used Galleria mellonella to validate the danger model in insects. Hemolymph of G. mellonella was digested with thermolysin (as a representative for virulence-associated metalloproteinases produced by humanpathogens) followed by chromatographic fractionation. Immune-stimulatory activity was tested by measuring lysozyme activity with the lytic zone assays against Micrococcus luteus cell wall components. Peptides were analyzed by nano-scale liquid chromatography coupled to high-resolution Fourier transform mass spectrometers. Addressing the lack of a genome sequence we complemented the rudimentary NCBI protein database with a recently established transcriptome and de novo sequencing methods for peptide identification. This approach led to identification of 127 peptides, 9 of which were identified in bioactive fractions. Detailed MS/MS experiments in comparison with synthetic analogues confirmed the amino acid sequence of all 9 peptides. To test the potential of these putative danger signals to induce immune responses we injected the synthetic analogues into G. mellonella and monitored the anti-bacterial activity against living Micrococcus luteus. Six out of 9 peptides identified in the bioactive fractions exhibited immune-stimulatory activity when injected. Hence, we provide evidence that small peptides resulting from thermolysin-mediated digestion of hemolymph proteins function as endogenous danger signals which can set the immune system into alarm. Consequently, our study indicates that the danger model also plays a role in insect immunity.


Introduction
A prerequisite for the evolution of Metazoa was the ability to discriminate between self and nonself. This power of distinction is mediated by a functioning immune system which encompasses all endogenous mechanisms and molecules contributing to host defense against microbes and parasites.
Vertebrate and invertebrate animals share the ancient innate immunity mediating discrimination between infectious nonself and noninfectious self by a limited number of germ line encoded receptors which bind to molecules common among pathogens, but absent from the host such as microbial cell wall components, particularly, bacterial lipopolysaccharides (LPS), peptidoglycans, and fungal ß-1,3-glucans [1]. This binding of receptor proteins to pathogen-associated molecular patterns (PAMPs) differs from the highly specific recognition of a wide range of antigens by a large and diverse spectrum of somatically rearranged receptors of T and B cells which represent an evolutionary novelty of vertebrates. Their adaptive or acquired immunity relies on antibody-based immunological memory of infection [2]. Janeways model claiming that activation of the immune system requires recognition of PAMPs has become a paradigm, particularly in insect immunity [3]. Binding of PAMPs to corresponding pattern recognition receptors results in activation of immune responses including both cellular mechanisms such as phagocytosis or multicellular encapsulation of microbes, and humoral defences such as the rapid synthesis of antimicrobial peptides and proteins [4].
Matzinger has challenged immunological research by introducing her danger model, which is based on the idea that the immune system is more concerned with entities that do damage than with those that are foreign [5]. Her alternative model explains recognition of foreign threat by molecules released from damaged cells or wounded tissues whose binding to corresponding pattern recognition receptors results in activation of immune responses. Such damage-associated molecular patterns (DAMPs) function as danger signals which can set the immune system into alarm [6]. First evidence for the presence of immunity-related danger signals has been reported from insects, particularly in the larvae of the greater wax moth Galleria mellonella [7]. The latter have in between emerged as powerful model hosts for pathogens infecting insects or humans [8]. This study elucidated that activation of innate immune responses in G. mellonella does not necessarily require PAMPs, the presence of microbial enzymes, particularly metalloproteinases of the M4 family with thermolysin as the prototype, in the hemolymph is sufficient to produce DAMPs [7]. Members of the thermolysin-family encompass prominent virulence factors and toxins of human pathogens such as aureolysin, bacillolysin and pseudolysin which have been implicated to be responsible for increase of vascular permeability, hemorrhagic edema, sepsis and necrotic tissue destruction [9]. Thermolysin-mediated digestion of hemolymph proteins in G. mellonella results in formation of small peptidic fragments smaller than 3 kDa which elicit innate immune responses that are qualitatively (spectrum of immune-related proteins secreted within the hemolymph) and quantitatively (expression levels of antimicrobial peptides) comparable with the response to injected bacterial lipopolysaccharide (LPS), a widely used microbial elicitor of innate immune responses [10].
Mass spectrometry (MS) is a powerful and universal tool for the analysis and identification of proteins and peptides. Complex biological samples are usually separated by liquid chromatography or gel electrophoresis prior to MS analysis. The MS-based identification of peptides and proteins from G. mellonella is hindered by fact that the genome is not sequenced and thus protein databases are only rudimentary. Traditional database search approaches using SEQUEST [11] or Mascot [12] do not result in satisfactory results. To compensate for this impediment we have recently subjected the immunity-related transcriptome of G. mellonella to next generation sequencing using the Roche 454-FLX platform combined with traditional Sanger sequencing to obtain a comprehensive immune gene repertoire [13]. In addition, we have developed a method for de novo sequencing of peptides isolated from G. mellonella hemolymph [14]. The reliability of peptide identification can be significantly improved by using high resolution and accurate measurements [15,16]. The highest mass resolution is achieved by Fourier Transform mass spectrometers based on orbital trapping [17] or ion cyclotron resonance [18,19]. A method that simplifies de novo sequencing is the composition based sequencing (CBS) approach which takes advantage of accurate mass measurements [20]. An example for CBS analysis of the tree frog H. savigni with unknown genome sequence was recently reported [21].
The application of high-resolution mass spectrometry enabled the identification of a number of hemolymph protein fragments resulting from thermolysin-mediated hydrolysis of hemolymph. Their putative function as danger signals was verified using synthetic analogues which were injected into G. mellonella larvae in order to test their capacity to elicit humoral immune responses. As in our previous studies [7,10] we used freeze-dried or living Micrococcus luteus bacteria as indicator organisms in lytic zone or inhibition zone assays, respectively.

Materials and Methods
Insect rearing and sample preparation G. mellonella larvae were reared on an artificial diet (22% maize meal, 22% wheat germ, 11% dry yeast, 17.5% bee wax, 11% honey and 11% glycerin) at 32uC in darkness. For collecting the hemolymph the larvae were first cooled at 4uC for 15 min and the prolegs were pierced with a sterile needle. Hemolymph melanisation was prevented by the use of few crystals of phenylthiourea. Cell-free hemolymph was degraded with thermolysin (Sigma, Taufkirchen, Germany) with an end concentration of 1 mg/mL in sterile bidistilled water. The mixture was incubated for 1h at 36uC on a rotary shaker. These parameters (moderate enzyme concentration and short incubation time) were chosen in order to mimic the in vivo conditions. Thermolysin-degraded fragments smaller than 3 kDa were obtained gradually by centrifugation at 7000 x g for 4-6h at 4uC using Centricon centrifugal concentrators Centricon-30, Centricon-10 and Centricon-3 (Millipore, Billerica, Massachusetts, USA), respectively.

Fractionation of hemolymph sample
Separation of the thermolysin-digested hemolymph sample (, 3 kDa) was realized by an Agilent 1200 RP HPLC system (Agilent, Waldbronn, Germany) with a Symmetry C18 4.66250 mm column at a flow-rate of 1 mL/min. The injection volume of the sample was 100 mL. Separation was carried out using a 40 min gradient starting at 5% acetonitrile in water (v/v). After 15 min the solvent concentration of acetonitrile was raised up to 50% (v/v). After 40 min the concentration of acetonitrile was increased to 95% (v/v). The resulting fractions were collected manually in 30sec intervals, lyophilized and dissolved in 50 mL H 2 O. The fractions were injected in Galleria larvae and immune stimulation was determined by lytic zone assay, subsequently.

Mass spectrometry
NanoHPLC-ESI FTMS measurements. The measurements of the samples were performed on a binary nanoHPLC system consisting of the units Switchos, Famos and Ultimate (LCPackings/Dionex, Idstein, Germany). Solvent A was water (HPLC grade, Fluka, Neu-Ulm, Germany) containing 2% acetonitrile (v/v) (UvasolH grade, Merck KGaA Darmstadt, Germany) and 0.1% formic acid (v/v) (puriss p.a. for ms, Fluka Neu-Ulm, Germany). Solvent B was acetonitrile containing 20% water (v/v) and 0.08% formic acid (v/v). Separation of the hemolymph sample was carried out using a 77 min gradient. First B was increased from 0% to 10% in 10 min, subsequently increased to 30% in 30 min and then increased to 100% in 5 min and maintained for 11 min. The injection volumes were 0.3 mL for the hemolymph bulk sample and 1.2 to 5 mL for hemolymph fractions, respectively. After pre-concentration on a C18 PepMap trap column (5 mm6300 mm i.d.) the samples were separated on a fused silica C18 PepMap100 capillary column (150 mm675 mm i.d.) (Dionex, Idstein, Germany) at 200 nL/min. The separation was monitored by a UV detector at 214 nm.
The nanoHPLC system was coupled to the mass spectrometer by a nanospray source. Pico-TipH Emitter (New Objective, Woburn, MS, USA) were used as nanospray needles. The separated peptides were measured on a tandem mass spectrometer (LTQ FT Ultra, Thermo Fisher Scientific GmbH, Bremen, Germany) consisting of a linear ion trap (IT) and a Fourier Transform Ion Cyclotron Resonance (FTICR) mass spectrometer. For database searches high-resolution (R = 100.000 at m/z 400) survey spectra were measured on the FTICR with high mass accuracy (, 2 ppm). Fragment ion spectra were measured both with high resolution and in the ion trap. For analyses with manual de novo and the composition-based sequencing method precursor and fragment ions were solely measured with high resolution and high mass accuracy. Collision induced dissociation (CID) was used for fragmentation. MS/MS data was obtained on the FT also using the wide scan range (WSR) method with three microscans in order to compensate the loss of ions due to the time-of-flight effect of the FTICR mass spectrometer.
In addition, coupling of nanoHPLC with a linear ion trap / Fourier transform orbital trapping (IT-FTOT MS) mass spectrometer (LTQ Orbitrap Discovery, Thermo Scientific GmbH, Bremen, Germany) equipped with a nanospray ion source, was used. This mass spectrometer allowed, in addition to CID-fragmentation, the detection of low mass fragments by higher-energy collisional dissociation (HCD), providing additional information for peptide sequencing. Full scan and MS/MS spectra on the FTOT instrument were acquired with high resolution (R = 30.000 at m/z 400) and high mass accuracy (, 4 ppm). External calibration of both FT instruments was performed immediately before measurements according to the instructions of the manufacturer. All MS measurements were performed using the positive ion mode.
Offline ESI FTMS measurements. Offline nano ESI MS measurements of synthetic peptides were performed for validation of identified peptides using the orbitral trapping mass spectrometer. Pico tip emitters (5 mm i.d., DNU-MS, Berlin, Germany) were used for this task. The synthesized peptides were dissolved in 200 mL water/acetonitrile 50:50 (v/v) with 0.1% formic acid and diluted 1:10 before measurement.
MALDI FTMS measurements. For matrix-assisted laser desorption/ionization (MALDI) measurements the commercial MALDI source of the orbital trapping mass spectrometer, equipped with a N 2 -Laser (337 nm), was used. Alpha-cyano-4hydroxycinnamic acid (10 mg/mL) dissolved in acetonitrile/water 70:30 (v/v) with 0.1% trifluoroacetic acid was used as matrix. Two mL matrix solution were mixed with 1 mL sample on a stainless steel sample plate. The laser energy was set to 20 mJ and the crystal positioning system (CPS) was activated. MS and MS/ MS experiments were performed with collision induced dissociation (CID) and higher-energy collisional dissociation (HCD) to obtain low-mass fragment ions.

Data analysis / Identification of peptides
Database search with standard database. Data obtained from HPLC-MS measurements were analyzed with SEQUEST database search using the BioWorks V3.3.1 software (Thermo Fisher Scientific, Bremen, Germany). For bulk samples 5 HPLC-MS measurements were carried out and at least 2 measurements for each bioactive fraction. Protein sequence database entries for G. mellonella are very limited. Therefore all available database entries for Galleria were downloaded from the server of the ''National Center of Biotechnology Information'' NCBI (http://www.ncbi.nlm.nih.gov/). This database contained 162 protein sequence database entries. The database searches of all HPLC-MS measurements were carried out with following settings: precursor ion tolerance 5 ppm; fragment ion tolerance 1 Da. The following variable modifications were selected: phosphorylation on serine, threonine, and tyrosine, hydroxylation on proline, threonine and methionine and carbamidomethylation on cysteine. These modifications are considered in the software with a mass increment of the respective amino acid, for hydroxylation Dm = +15.99492, for phosphorylation Dm = +79.96633 and for carbamidomethylation Dm = +57.02146, respectively. Thermolysin cleaves at the N-terminal side of amino acids with hydrophobic or bulky side chains [22]. The amino acids alanine, isoleucine, leucine, phenylalanine, valine and methionine were selected as cleavage sites. The number of allowed missed cleavage sites was set to 10. Peptides with a Peptide Probability below 0.001 were considered as identified. False discovery rates (FDR) were calculated according to Nesvizhskii et al. [23].
Database search with experimental database. An additional database search was applied with a newly implemented G. mellonella database obtained from transcriptome analysis [13]. The transcriptomic data were translated into the single letter amino acid code, manually implemented in a FASTA file and used the same way as standard database entries for the SEQUEST database searches with BioWorks. The database included 96756 entries, mainly short sequences (64% of all entries had a chain length of less than 50 amino acids).
De novo sequencing. Manual de novo sequencing was performed with accurately measured masses. Additionally, the composition-based de novo sequencing approach was applied using the computer program Peptide Composer 1.0 [20,24]. Calculations of amino acid compositions of peptides were performed using accurately measured mass spectra with a tolerance of 2 ppm for FTICR and 4 ppm for FTOT mass spectra, both for precursor ions and fragment ions. (The isobaric peptides leucine and isoleucine could not be differentiated by exact mass measurements due to their identical elemental composition.)

Synthesized peptides
After sequence determination, peptides were custom-synthesized by GL Biochem (Shanghai Ltd., China) with a purity of over 85%.

Bioactivity tests
Last instar G. mellonella larvae, weighing between 250-350 mg, were used in this study for determining the immune stimulation. TenmL of collected fractions or 20 mL of synthesized peptide solutions (20 mM), respectively, were injected dorsolaterally into the hemocoel of the larva using 1 mL disposable syringes and 0.4620 mm needles mounted on a microapplicator. After injection, the larvae were incubated at 32uC in darkness. 24 hours post injection, hemolymph was collected to investigate the immune stimulation.
As a simple read-out system to determine the immunestimulatory activity of samples, we used the inhibition zone assay using freeze-dried Micrococcus luteus [7,25]. The inhibition zone assay against living bacteria such as Micrococcus luteus is a standard assay which is widely used in microbiology to qualitatively and quantitatively determine the activity of soluble antimicrobial compounds. The latter are pipetted into wholes which have been punched in the agar inoculated with bacteria. The sample defund into the agar around thereby inhibiting the growth of the bacteria. The diameter of the clear zones without growing bacterial colonies can be measured and subsequently quantified by using a calibration curve with a standard antibiotic such as gentamycin. Accordingly, the anti-bacteria activity in samples can by quantified in gentamycin equivalents. Briefly, seven milliliter of M. luteus agar consisting of 1% highpurity agar-agar (Carl Roth, Karlsruhe, Germany), 5 mg/mL freeze-dried M. luteus (Sigma, Taufkirchen, Germany), and 67 mM potassium phosphate buffer (PBS) pH 6.4 were placed onto petri dishes (Ø100 mm). Holes with a diameter of 4 mm were filled with 3 mL hemolymph and incubated for 24h at 37uC. Lysozyme activity was quantitatively determined (units/ mL) by establishing a calibration curve from standard chicken lysozyme (Sigma, Taufkirchen, Germany).
For validating the potency of identified peptidic danger signals to elicit immune responses we used living M. luteus as an indicator organism in the inhibition zone assay because bacterial growth inhibition indicates the presence of induced antibacterial peptides, whereas the lytic zone assay using freezedried M. luteus indicates enzymatic digestion by lysozyme. Synthetic peptides were dissolved in sterile saline solution (120 mM NaCl) with an end concentration of 20 mM and 20 mL were injected into larvae. For each peptide 6 individual larvae were used. The anti-M. luteus-activity was calculated to gentamicin equivalents (inmg/mL) using a calibration curve with gentamicin as previously described [10,26]. Saline solution (peptide solvent) was used as control, and injected in the larvae following the same procedure as for the tested peptides. Average value and standard deviation for 6 individual animals were calculated for each peptide. Statistically significant differences between activities of larvae injected with peptides and control were determined using Students t-test. Figure 1 shows the workflow for the combined approach used for the identification of bioactive peptides. The thermolysindigested hemolymph sample was measured directly by nanoLC-FTMS. Second, the hemolymph sample was fractionated and resulting fractions showing lysozyme inducing activity were analyzed in more detail by nanoLC-FTMS and MALDI MS. Data analysis was performed by SEQUEST database search and de novo sequencing. Identified peptides were tested individually for immune stimulation by synthetic analogues.

Fractionation of hemolymph sample and initial bioactivity tests
Initial HPLC-MS measurements revealed that the thermolysindigested sample (, 3 kDa) was too complex for the direct determination of bioactive compounds. For this reason the bulk sample was fractionated by HPLC and collected fractions were tested with a lytic zone assay to screen for elevated bioactivity. Details on chromatographic fractionation and activity tests can be found in the supporting information (File S1). Two fractions, termed A and B, in which the highest immune-stimulatory activity was detected were subjected to a second round of HPLC fractionation in order to further decrease the complexity. The resulting sub-fractions of these separation steps were tested again for immune-stimulatory activity. This procedure resulted in 4 subfractions for fraction A and 2 sub-fractions for fraction B. The tests of these sub-fractions showed that sub-fraction A1 and B1 had the highest activity and were therefore subjected to more detailed analysis.

Peptide identification by mass spectrometry
The hemolymph samples were measured by nanoHPLC coupled to FTMS instruments. The acquired data was analyzed using standard database searches as well as experimental databases derived from transcriptome data. Bioactive fractions were analyzed in detail by nanoHPLC-FTMS and additionally by MALDI Orbitrap measurements. Peptide sequence analysis was finally extended to database-independent de novo sequencing methods. Identification results are reported for the bulk sample and for bioactive fractions separately, in the following. A list of all identified peptides can be found in the supporting information (Table S1).
Database search with standard database. One of the standard procedures in proteomics for characterization of peptides is high performance liquid chromatography coupled to mass spectrometry followed by protein database searches. In this work HPLC-FTMS measurements of the thermolysin-digested hemolymph sample were performed with database search using SEQUEST. For this purpose all rudimentary available protein sequence entries from the public database NCBI (162 protein entries) of G. mellonella were used.
Database search of the bulk sample with the NCBI database led to the identification of 75 peptides. The majority of these identified peptides were obtained by HPLC-MS/MS measurements where the precursor ion was measured in the FT cell with high resolution, and the five most intense ions of this survey scan were fragmented and detected in the ion trap.
Most of the identified peptides could be assigned to abundant hemolymph proteins such as hexamerin, apolipophorin and transferrin precursor. The number of identified peptides was Figure 1. Workflow for the identification of bioactive peptides. On the one hand the thermolysin-digested hemolymph sample was analyzed directly by nanoHPLC-FTMS. On the other hand the thermolysin-digested sample was pre-fractionated using a HPLC-system and the collected samples were tested for bioactivity for a more detailed analysis of immuno-relevant peptides. doi:10.1371/journal.pone.0080406.g001 relatively low, especially when compared with database searches of fully-sequenced species where in one run several hundred peptides can be identified. This was due to the limited number of protein entries in the database for G. mellonella and showed the necessity of further identification methods. The determination of a false discovery rate (FDR) that is common in database search studies was only partly possible, because of the small size of the used databases. An FDR of 2,7% was determined by searching a decoy database that was generated by reversing the sequences of the original NCBI database [27].
SEQUEST database search was also performed with na-noHPLC-MS measurements of the bioactive fractions A, B and the sub-fractions A1 and B1 (which had the highest bioactivity in the initial activity tests). Five peptides could be identified in this manner in the fraction B. These peptides were VV-9, IE-8, FN-9, IN-10 and LY-11 (see Table 1 for sequence information). Of these peptides IE-8 and LY-11 were also identified in the hemolymph bulk measurements discussed above. The NCBI database searches of fraction A and sub-fractions A1 and B1 showed no peptide identification.
Database search with experimental database. A recently established comprehensive transcriptome of G. mellonella [13] was used as a database in order to identify additional peptides. Database search of the bulk sample with this experimental database resulted in 38 identified peptides. The decoy database searches of three HPLC-MS runs using the reversed database resulted in a FDR rate of 29%. These high values are most likely due to the fact that most of the database entries in the experimental database were short and therefore tended to lead to more hits with the decoy database.
The HPLC-MS measurements of the bioactive fractions were also analyzed by database search using the experimental database. In fraction B1 only one peptide could be identified by this approach: APPSGPAAPPAKTP (AP-14). In fraction B peptide VV-9 could be identified that was also identified with the standard database approach. No peptides were identified in fraction A and sub-fraction A1. The total number of peptides identified with the experimental database was 40.
De novo sequencing. De novo sequencing of peptides allows identifying peptides without prior knowledge about possible amino acid sequences. Therefore this approach was used in order to identify peptides which are not included in the NCBI or transcriptome database. Data analysis by de novo sequencing was always performed using high resolution and high mass accuracy measurements, both for precursor and fragment ions. In doing so, it was possible to determine amino acid-specific increment masses within the MS/MS spectra. For example lysine and glutamine have the same nominal mass, but can easily be differentiated by accurate mass measurements due to their mass difference of 36 mDa.
Twenty-two peptides were identified using the de novo sequencing approach in the bulk sample. Nine of this set were identified by manual de novo sequencing of non-significant SEQUEST database search hits (Peptide Probability above threshold of 0.001).
The de novo sequencing approach led to no identification of peptides in the bioactive fractions. Therefore additional experiments were performed that implemented matrix-assisted laser desorption/ionization (MALDI) and the higher-energy collisional dissociation device (HCD) as a fragmentation technique. This fragmentation method allows the fragmentation in a dedicated quadrupolar collision cell, which results in a lower mass cut-off for fragmentation ions. An example of a peptide identified by this approach is discussed in the following. In sub-fraction A1 two peaks, m/z = 517.28360 and m/z = 631.38722, could be sequenced by the composition based sequencing approach (CBS) (see [20,24] for more details). The MS/MS spectrum of precursor ion m/z = 631.38722 using HCD fragmentation is shown in Figure 2.  therefore be determined as (A,K,K,E,R). In the second step of the composition-based de novo sequencing the peptide sequence was determined by scoring the agreement between observed and expected fragment ions of permuted sequence propositions. For that all fragment ion signals of the MS/MS spectrum ( Figure 2) were used. The results of the de novo (CBS) sequencing showed that the sequence KAERK (abbreviation 'KK-5' in Table 1) was matched with the highest CBS score value. The peptide sequence of precursor ion m/z = 517.28360 was identified using the same approach as ERRG (EG-4). Three peptides were identified by de novo sequencing in the bioactive sub-fractions. Peptide SP-8 was identified in sub-fraction B1 in addition to the two peptides EG-4 and KK-5 discussed above. In total 25 peptides were identified by de novo sequencing.
Comparison of identification methods. The different approaches for the identification of peptides were summarized in a Venn diagram shown in Figure 3. Nineteen peptides were exclusively identified using the de novo sequencing (CBS) approach, 63 peptides could only be identified using the standard NCBI protein sequence databases of G. mellonella. The database search with the experimental database resulted in the identification of 30 additional peptides. Nine peptides were found in the database searches both with the NCBI and experimental databases. Five peptides were identified with de novo and NCBI database search. Consequently most peptides were identified by only one method. Only one peptide (IKIPAPYE) was identified with all three approaches. These results show that each sequencing approach has specific advantages and limitations. The combination of the used methods led to the identification of 127 peptides in total. Most of these peptides could be identified using the NCBI database search. In contrast, the number of peptides identified with the de novo approach in the bulk sample was lower. It has to be mentioned, however, that identification with the de novo (CBS) sequencing approach is not finally completed. Due to the fact that highly resolved MS spectra were analyzed manually, there is still the possibility of identifying additional peptides. Furthermore, reliability of the de novo sequencing results was higher compared to the database search results, because fragment ions were all  assigned based on accurate mass measurements. The FDR rate of the search with the experimental database was high, indicating that there were also false positives among these.
Of the 127 peptides that were identified in total, 9 peptides were identified in the fractions with increased bioactivity. These peptides are summarized in Table 1. Within this set, 4 peptides were identified in fraction B using the standard NCBI database search, whereas one peptide (VV-9) was identified both with the standard and experimental database. Two peptides each were identified in sub-fraction A1 and B1. Three peptides (EG-4, KK-5 and SP-8) were identified by the de novo approach, whereas one peptide (AP-14) was determined by database search with the experimental database. Using the standard database search alone would have led to no results in the sub-fractions. Inclusion of MALDI and HCD was also necessary in order to identify the peptides in the sub-fractions. These results show that different ionization techniques and different approaches for peptide analysis had to be used for the identification of a broad set of peptides. All of the identified peptides in the bioactive fractions were of particular interest and had to be validated and tested for bioactivity as discussed in the following section.
Confirmation by synthetic peptides. To validate the sequence of the peptides identified in the bioactive fractions summarized in Table 1, synthetic analogs were synthesized. Figure  4 shows the comparison of the FT-MS/MS spectra of peptide VV-9 detected in the hemolymph fraction and its synthetic analogue. Fragmentation experiments for validation were performed the same way as for the hemolymph sample. The fragment ions of the identified peptide and of the synthetic analogue, matched all with high accuracy thus confirming the initial identification.
Seven peptides could be validated by accurate mass measurements for precursor ion and fragment ions. Two peptides identified by the database search, FN-9 and AP-14, were confirmed by accurate precursor ion mass and fragment ions measured in the ion trap. MS/MS spectra of these peptides can be found in the supporting information (File S2). Consequently all peptide sequences identified in bioactive fractions shown in Table  1 were confirmed by MS/MS measurements of synthesized peptides.

Bioactivity tests of identified peptides
The peptides identified by our MS approaches have been postulated to represent putative danger signals resulting from digestion of hemolymph by microbial metalloproteinases. In order to validate their immune-stimulatory activity we produced synthetic analogues which were injected into last instar larvae of G. mellonella to determine their capacity to induce immune response in vivo. As a simple read-out system we used the inhibition zone assay against living M. luteus to determine antibacterial activity in the hemolymph [25].
Interestingly, not all synthetic peptides turned out to exhibit immune-stimulatory activity when injected ( Figure 5). In comparison with control injections with saline alone, the peptides VV-9, IE-8, FN-9, IN-10, LY-11 and AP-14 induced significantly higher anti-M. luteus-activity. On the other hand peptides KK-5, EG-4 and SP-8 did not induce a significant response. Therefore, we conclude that only particular peptidic hemolymph protein fragments can function as danger signals when released upon metalloproteinase-mediated hydrolysis.
The larvae of the lepidopteran G. mellonella have been established as a powerful model host for pathogens of insects or humans and as a source for novel anti-infective therapeutics [28]. In addition, they were used to demonstrate for the first time that the presence of microbial metalloproteinases alone is sufficient to induce potent innate immune responses [7]. Thermolysin is the prototype metalloproteinase belonging to the M4 family which encompasses many virulence factors of human pathogens [9]. Thermolysin-mediated hydrolysis of hemolymph proteins from G. mellonella results in formation of peptidic fragments which, when injected, can elicit immune responses that are qualitatively and quantitatively comparable with those observed upon injection of bacterial cell wall components such as LPS [10].
Peptide analysis and identification in our study was more complicated in comparison to standard proteomics experiments due to several factors. Firstly, lacking a sequenced genome of G. mellonella it was difficult to characterize these putative danger signals by standard proteomics approaches. However, fully sequenced insect systems such as Drosophila and Tribolium are very small and therefore it is difficult to obtain hemolymph samples in sufficient quantities (for mass spectrometric analysis and bioactivity tests). Furthermore we have discovered the first specific peptidic inhibitor of metalloproteinases in G. mellonella. This so-called IMPI is currently being developed as new second generation antibiotic to cure symptoms caused by thermolysin-like metalloproteinases during infections in humans [28]. Secondly, peptides produced by thermolysin are more difficult to ionize than those resulting from other enzymes. Trypsin, as a digestion enzyme which produces ions which are much easier to ionize, would not be representative for the activity of microbial pathogens. It was shown previously that trypsin does not induce immune response in Galleria mellonella [7]. Finally, thermolysin digestion was performed at moderate enzyme concentration and for 1 hour only, which results in incomplete digestion of hemolymph proteins. A longer incubation time and/or higher thermolysin concentrations would result in a higher number of peptides. But this situation would not be comparable with in vivo conditions during infection of G. mellonella where only limited amounts of thermolysin-like metalloproteinases are produced, plausibly to avoid activation of host immune responses [29]. In addition it is known from our previous studies that stronger or total digestion of hemolymph proteins does not necessarily increases their immune-stimulatory activity [7].
To account for these specific challenges, we applied in this study high resolution mass spectrometry combined with de novo sequencing of peptides generated by metalloproteinase-mediated digestion of hemolymph proteins from G. mellonella. This approach led to the identification of several peptides in bioactive fractions. Following a manual inspection of LC-MS/MS data, we estimate that the two sub-fractions A1 and B1 contain about 10 peptides each (see supporting information File S3 for details). The full sequence of 2 peptides each was identified in these fractions and partial sequences were obtained for other peptides. In order to validate the postulated function of the de novo sequenced peptides as danger signals we produced synthetic analogues which were injected into G. mellonella larvae to assess their capacity to induce innate immune responses. The latter are characterized by the release of antimicrobial peptides among which some exhibit activity against living Micrococcus luteus [28]. Strikingly but in accordance with our expectation, it turned out that not all peptidic fragments exhibit immune-stimulatory activity. Only the peptides designated as VV-9, IE-8, FN-9, IN-10, LY-11 and AP-14 were confirmed as elicitors of innate immune responses when injected.
We therefore postulate that the release of metalloproteinases produced by invading pathogens and parasites results in generation of these danger signals which, in turn, induce innate immune responses [29]. The formation of these peptidic danger signals can only be detected by direct (mass spectrometric) analysis on the peptide level. Transcriptomic approaches would only detect the endogenous precursor peptides/proteins of these digestion prod-ucts. In this context it is important to note that G. mellonella is the only known animal which is capable to produce a specific peptidic inhibitor against microbial metalloproteinases [30]. This insect metalloproteinase inhibitor (IMPI) was originally discovered in and purified from hemolymph of immune-stimulated larvae [31], and mediates reportedly feedback-loop regulation of harmful microbial metalloproteinases [32].
Consequently, we have identified peptides which are generated in G. mellonella under presence of microbial metalloproteinases, and which mediate the activation of immune responses encompassing the synthesis and release into the hemolymph of both antimicrobial peptides and the IMPI. Regarding the high in vivo toxicity of thermolysin-like metalloproteinases, the characterized peptidic products of their activity can be considered as danger signals because they can set the immune system into alarm. This efficient mechanism providing the ability to sense invading pathogens or parasites by the activity of their virulence-associated enzymes [29] adds to the sophisticated strategies determining the tremendous evolutionary plasticity of insect immunity [33].

Conclusions
In this study we used high-resolution mass spectrometry to identify peptidic hemolymph protein fragments which are generated in the hemolymph of the model host G. mellonella when harmful microbial metalloproteinases belonging to the M4 family, with thermolysin as the prototype, are present. Addressing the lack of a genome sequence we complemented the rudimentary NCBI database with a transcriptome database and de novo sequencing approaches for peptide identification. This approach led to the identification of 9 potentially bioactive peptides. These (tentative) results were validated by studying synthesized peptide analogues. Detailed MS/MS experiments confirmed the amino acid sequence of all 9 peptides. Six out of 9 peptides identified in the bioactive fractions exhibited immune-stimulatory activity when injected in Galleria larvae. These six peptidic elicitors of immune responses are postulated to function as danger signals indicating the presence of microbial metalloproteinases and mediating their regulation by specific inducible metalloproteinase inhibitors. Our results suggest that the validity of the immunity-related danger model proposed by Matzinger can be expanded beyond mammals to include insects. The detailed molecular pathways and importance of this process will have to be investigated in follow-up studies. File S1 Chromatographic fractionation and activity tests of hemolymph. Peptide KK-5 shows no activity compared to the solvent control, in contrast. EG-4 and SP-8 show slightly elevated immune activity. Statistically significant differences between activities of larvae injected with peptides and control were determined using Students t-test and are indicated by * (p,0.05) and ** (p,0.005), (n = 4-6 for each peptide). doi:10.1371/journal.pone.0080406.g005