Millipedes (Diplopoda) are well known to protect themselves against enemies and microorganisms by defensive secretions (Shear, 2015; Ilić et al., 2018). These fluids are produced by multiple pairs of exocrine glands and show repellent or even toxic properties. Meanwhile an impressive chemical repertoire has been reported from different millipede orders, including various quinones, esters, alcohols, cyanogenic compounds and unusual aromatic alkaloids (Shear, 2015; Bodner et al., 2018). Alkaloids are – with a few exceptions (Bodner et al., 2017) – mainly found in the secretions of two groups of diplopods, Glomerida and Colobognatha. The chemistry of Glomerida appears to be predominated by two unique quinazoline alkaloids: glomerin and homoglomerin (Meinwald et al., 1966; Shear et al., 2011). Most of the published data from Colobognatha refer to the order Polyzoniida, showing that their chemistry comprises monoterpenes such as pinene and limonene as well as heterocyclic alkaloids. Compounds such as polyzonimine (1), nitropolyzonamine (2), buzonamine and various spiropyrrolizidine oximes have been reported (Meinwald et al., 1975; Röper, 1978; Wood et al., 2000; Saporito et al., 2003; Kuwahara et al., 2007). Recently, unique hydroindolizidine alkaloids and deoxybuzonamine isomers were described from the platydesmid colobognathans, Gosodesmus claremontus and Brachycybe lectonii (Hassler et al., 2020; Jones et al., 2022).

The complex chemistry of millipedes prompted researchers to apply increasingly advanced analytical methods for their investigations. The defensive chemistry of European Polyzonium germanicum (Polyzoniida), for instance, was studied in the late 1970’s by Röper (1978), who detected a single compound, polyzonimine (1). At the same time, spirocyclic alkaloids from Polyzoniida were isolated by preparative gas chromatography and further characterized in crystallographic studies (Meinwald et al., 1975; Smolanoff et al., 1975). With the rapid development of analytical techniques, metabolite profiling has been increasingly applied for the analysis of complex natural product extracts, especially in pharmaceutical biology.

Commonly used analytical platforms are either based on NMR or on mass spectrometry (Kellogg et al., 2017). However, in newer publications on millipede chemistry, GC-MS data are already accompanied by NMR data of purified compounds isolated from secretions (Wood et al., 2000; Hassler et al., 2020; Jones et al., 2022). While multi-technique approaches in chemical ecology studies are still rather scarce, and most researchers analyzing gland secretions still focus on GC-MS data which still represent the gold standard when working with thermostable volatiles. As a disadvantage, many other compounds, such as less volatile material and large molecules, may remain inaccessible. NMR spectroscopic evaluations do not underlie these restrictions, but clearly benefit from information by methods of mass spectrometry, e.g., information on exact mass and elementary composition by high resolution mass spectrometry (HRMS). Particularly when dealing with natural compounds, a combination of these methods arises as a most promising strategy to quickly identify novel structures. We here rely on such a multi-technique strategy, expanding the NMR-concept of Schröder and colleagues to analyze crude animal secretions by adding further methods (Taggi et al., 2004; Schroeder et al., 2008). In detail, we re-analyzed the defensive secretion of the millipede species Polyzonium germanicum, combining GC-MS, LC-HRMS and two-dimensional NMR spectroscopy. A high-field NMR spectrometer equipped with a cryo-probe enabled us to record the relatively insensitive HSQC-TOCSY experiment in which the carbon axis resolves signal overlaps in the proton spectrum or proton-proton experiments like COSY and TOCSY. Compared to previous studies, our approach allowed us I) to work with a much lower number of individuals and II) without prior compound isolation. This led to the discovery of novel compounds, and may serve as a model standard for chemical-ecological investigations with arthropods in the future.

In the pooled extract of Polyzonium germanicum seven compounds were detected by GC-MS analysis ( Figure 1 , Table 1 , SI Figure S1–S7 ). The structures of compounds 1–5 could be fully elucidated by NMR spectroscopy ( Figure 2 ). Due to their low abundance in the sample, structure elucidation via NMR was not possible for compounds A and B.

The major compound 1 was the already described polyzonimine (Röper, 1978) ( Table 1 ). The EI-mass spectrum of the compound (RI 1207) showed a base peak at m/z 82 (C₅H₈N⁺) and a molecular ion (M⁺) at m/z ( Supplementary Figure S1 ), corresponding to published data of polyzonimine (Kuwahara et al., 2007). Subsequently, the molecular formula was determined by HESI-HRMS as C₁₀H₁₇N ( Table 1 ). In the proton NMR spectrum, the major component 1 was easily identified as polyzonimine due to its high signal intensity and its characteristic proton and carbon resonance values for methine group C-1. Its ¹³C NMR data ( SI Ta ble S2 ) are in perfect agreement with NMR data published for synthetic polyzonimine (Mori and Takagi, 2000) indicating that compound 1 has the same relative configuration.

From the correlations observed in the HSQC and HSQC-TOCSY spectra, compounds 2–4 occurring in the extract at medium abundance, were expected to be compounds of the nitropolyzonamine type too. The skeletons of nitropolyzonamine-like molecules comprise three isolated spin systems, the first is C-3, C-4, and C-5, the second C-2’ and C-3’, and the third C-5’, C-6’, C-7’ and C-7a’. The proton and carbon resonances belonging to each of the nine spin systems were identified in the HSQC-TOCSY spectrum ( Figure 3 ) and assigned in the COSY and HSQC spectra.

The EI-MS spectrum of compound 2 (RI 1857) with a weak M⁺ ion at m/z 238 and a base peak at m/z 82 showed full correspondence with nitropolyzonamine (Meinwald et al., 1975; Saporito et al., 2003; Kuwahara et al., 2007), which has already been described from other polyzoniidan species ( Supplementary Figure S2 ). In general, spiropyrroline polyzonimine and spiropyrrolizidine alkaloids are closely related structures, and EI fragmentation results in a common base peak at m/z 82 (C₅H₈N⁺) (Daly et al., 1999). For compound 2, a neutral molecular formula of C₁₃H₂₂O₂N₂ was deduced from its LC high resolution mass spectrum ( Table 1 ). According to the HSQC and HSQC-TOCSY spectra, the first spin system of this compound consisted of three methylene groups, the second of two methylene groups and the third of two methylene and two methine groups. C-1 and C-3 were identified by HMBC correlations with the two methyl groups attached to C-2 ( Figure 4 ). The proton resonances of the ring C-5’ to C-7a were identified at the carbon resonance C-7’ (88.3 ppm) and are also observed at carbon resonances C-5’ (52.4 ppm), indicating that both carbon resonances belong to the same spin system. The four proton resonances of the methylene groups C-2’ and C-3’ are observed at carbon resonance C-3’ (53.3 ppm) and at the carbon resonance C-2’ (35.2 ppm). The two spin systems were then connected by an HMBC correlation between H-7a’ and C-3’ (black arrow in Figure 3 ). Then, protons H-2’ and H-3’ were correlated in the HMBC spectrum with the spiro carbon C-1 and carbons C-5 and C-7a’. This indirect approach was necessary because the proton resonance H-7a’ did not show a two-bond HMBC correlation to the spiro carbon C-1 ( Table 2 ). Compound 2 had the same molecular constitution as nitropolyzonamine, and its ¹³C NMR data were in very good agreement with published primary NMR data for synthesized nitropolyzonamine (Mori and Takagi, 2000). Therefore, it was assigned with the same relative configuration as the synthetic compound.

Comparably, compounds 3, 4, and A (RI: 2200, 1842, 1809) provided a base peak at m/z 82 too. They exhibited weak molecular ions in EI-MS at m/z 296, 236 and 236, respectively. For compound 3, which had a neutral molecular formula of C₁₅H₂₄O₄N₂, the HSQC-TOCSY spectrum indicated that one position in the first spin system comprising C-3, C-4 and C-5 was oxidized. As the two methyl groups at C-2 showed correlations to a methine group with a ¹³C shift value of 81.5 ppm, it had to be assigned as C-3. In the HMBC spectrum, the proton H-3 was correlated with a carbonyl resonance (δ_C 171.0 ppm).

The same correlation was also detected for the additionally observed methyl protons of the side chain at 2.03 ppm, indicating an ethanoyloxy moiety ( Table 3 ). The chemical shift values for the other two rings were very similar to compound 2. Therefore, compound 3 was identified as acetylated 3-hydroxynitropolyzonamine. The configuration at C-3 remains unclear as no useful NOEs were observed for the protons in this ring.

For compound 4, a neutral molecular formula of C₁₃H₂₀O₂N₂ was calculated. In comparison to nitropolyzonamine (2), very different carbon shift values of C-2 and C-3 suggested a different chemical structure at C-2 and C-3. This was confirmed by analysis of HMBC correlations which indicated a shift of one methyl group from C-2 to C-3 and the presence of a double bond between C-2 and C-3 ( Table 4 ), perfectly explaining the molecular formula of compound 4, that indicated the absence of two hydrogen atoms compared to compound 2. As the ¹³C chemical shift values values ( Tables 2 – 4 ) of C-1 to C-7a’ in all three compounds 2–4 were very similar and the stereochemistry of nitropolyzonamine had been established by X-ray crystallography and synthesis (Smolanoff et al., 1975; Mori and Takagi, 2000), the same relative configuration was also assumed at C-7’ and C-7a’ of compounds 3 and 4 in the investigated P. germanicum extract.

The EI-spectra of compounds 5 and B (RI: 2184, 2159) displayed pronounced molecular ions at m/z 222 and m/z 246, respectively. For compound 5, HRMS indicated a neutral molecular formula of C₁₁H₁₄N₂O₃ ( Table 1 ). The assignments of resonances C-3 to C-9 were straightforward. COSY and HMBC correlations indicated the presence of a spirocyclic terpenoid compound with a keto group at C-7, two methyl groups at olefinic positions C-8 and C-9, and three methylene groups at C-3, C-4 and C-6, respectively. The high carbon resonance values of the methylene group C-3 suggested a nitrogen at position 2. Protons H-6 were correlated with both, 166.0 ppm (C-9) and 164.9 ppm (C-1). The remaining carbon resonance belonged to a methine group (δC 106.3 ppm, δH 6.27 ppm). The proton resonance corresponding to this methine group showed an HMBC correlation with medium intensity to C-1 (164.9 ppm) and one with a very low intensity to C-5 (56.5 ppm). In addition, a labile proton was observed at 9.13 ppm ( Table 5 ). In the HESI-MS/MS spectrum of the compound, the loss of the NO₂ group was clearly observable ( Supplementary Figure S13 ). The observed chemical shift values ( Table 5 ) for C-1 and C-10 are consistent with all three tentative structures 5, 5b and 5c depicted in SI Figure S23 . However, NOEs observed between H-10 and methylene protons H-6 as well as methyl group at C-9, the NOEs between the labile proton and H-3 ( SI Figure S21B ), the high proton NMR shift value of the labile proton, and the observed loss of NO₂ in the HESI-MS/MS can be explained much more convincingly by structure 5. The NOEs also support the Z-configuration of the exocyclic double bond.

With DFT calculations we aimed to provide additional arguments for the structure of compound 5. We expected a low energy structure. Surprisingly, DFT studies using dispersion corrected hybrid functionals and continuum solvent modelling revealed that 5 – in comparison to the constitutional isomers 5b and 5c ( SI Figure S23 ) – is thermodynamically significantly less favorable by more than 100 kJ/mol. This is confirmed using different density functionals, supporting the notion that this is not a computational artifact but illustrates that the biochemical pathway does not necessarily produce the lowest energy molecule. Other constitutional isomers – not in agreement with the NMR data – that are thermodynamically equally favorable are reported in the supporting information. Nonetheless, the exocyclic nitro group can be confirmed in the depicted cis-conformation being the thermodynamically more favorable (by ca. 35 kJ/mol) conformer showing that only certain aspects of the biogenesis might be thermodynamically controlled. Hence, the structure of 5 was deduced as 8,9-dimethyl-1-(nitromethylidene)-2-azaspiro[4.4]non-8-en-7-one, while structures 5b and 5c were discarded.

Based on data by NMR spectroscopy, it was possible to annotate some of the major fragments observed in HESI-MS/MS spectra. MS/MS fragments provided by HRMS analysis in the HESI positive mode ( Supplementary Information Figures S9-S15 ) confirmed the presence of [C₅H₈N]⁺ (m/z 82.066) in compounds 1, 2, 3, 4 and A, corroborating that these compounds possess a similar skeleton. However, m/z 82.066 was the main fragment only in the MS/MS spectra of compounds 2 and 3. In contrast to GC-MS, MS/MS spectra generated by QExactive MS provided a higher range of informative MS/MS fragments ( SI Table S1 ). For example, a fragment indicating the loss of NO₂ (Egsgaard and Carlsen, 1996) was detectable in the MS/MS spectra of compounds 2, 3, 4, 5, A and B, suggesting the presence of a NO₂ group in these molecules, while it was absent in compound 1. Compounds 2, 3, 4 and B also shared the loss of CH₃NO₂ (originating from position 7´) and 2, 3 and 4 the loss of C₃H₅NO₂ (originating from positions 7´, 6´and 5´), which is likely due to the degradation of the pyrrolidine ring bearing the NO₂ group ( SI Table S1 ). Compound 3 displayed additional major fragment ions at m/z 192.175 [M+H-CH₃COONO₂]⁺ and m/z 176.143 [M+H-CH₃COOH-CH₃NO₂]⁺, indicating the presence of an acetyl function in the molecule. Compound 4 differed from compound 2 by the presence of highly abundant low molecular weight fragment ions in the MS/MS spectrum, e.g. at m/z 107.086, 84.081, and 69.058. This may be due to differences in the spirocyclopentane ring (i.e. the presence of a double bond and different positions of the methyl groups in compound 4). The MS/MS fragmentation pattern of compound 5 strongly deviated from those of the latter compounds. This is likely due to the fact that in this compound, the NO₂ group is attached to an exocyclic double bond absent in the skeleton of the latter compounds ( Figure 2 ). The base peak was a fragment at m/z 91.055. An indicative fragment at m/z 205.097 was formed by the loss of H₂O. This is likely possible due to migration of the proton at the nitrogen in position 2, which is not present in all the other structures (Egsgaard and Carlsen, 1996). The HRMS data indicated that the neutral molecular formula of A was C₁₃H₂₀O₂N₂, suggesting that the compound may be an isomer of compound 4. However, despite from showing the presence of a nitro group, the MS/MS fragmentation of A strongly differed from that of 4. The neutral molecular formula of compound B was calculated as C₁₃H₁₄O₃N₂. Its MS/MS fragmentation pattern was devoid of a fragment at m/z 82.066, but showed the presence of a major fragment at m/z 91.955, indicating potential similarities with compound 5. All carbon NMR resonances with high or medium intensities could be assigned to one of the five described alkaloids, therefore, the compounds A and B observed in GC-MS ( Figure 1 ) and LC-HRMS ( Supplementary Figure S8 ), are presumed to have a much lower concentration than compounds 1–5.

We here present a more detailed analysis of extracts from a colobognathan millipedes by a combination of advanced methods, including NMR and different methods of mass spectrometry. Our approach allowed us to directly identify the already known alkaloids polyzonimine (1) and nitropolyzonamine (2) and, in addition, two new compounds (3, 4) as well as a further new alkaloid (5). For two further compounds A and B, at least the molecular formulas could be provided. This analytical strategy feasibly clearly offers the opportunity to identify new chemical compounds, in the best case even at the level of relative configuration. The absolute configuration of the molecules has to be checked via synthesis in future studies.

The complexity of alkaloid chemistry from animal sources has been the topic of several publications, especially regarding the rich repertoire of dendrobatid frogs that sequester toxic alkaloids from their arthropod prey, such as oribatid mites and polyzoniid millipedes, respectively (Saporito et al., 2003; Daly et al., 2005; Takada et al., 2005). The biosynthetic machinery behind alkaloids in arthropods is largely unknown. While our data do not yet allow a detailed discussion of biosynthetic pathways, compound 3 ((2,2-dimethyl-7’-nitro-2’,3’,5’,6’,7’,7a’-hexahydro-spiro[cyclopentane-1,1’-pyrrolizin]-3-yl) acetate) could be a structural link from nitropolyzonamine (2) to compound 4 (2,3-dimethyl-7’-nitro-2’,3’,5’,6’,7’,7a’-hexa-hydrospiro[cyclopentane-1,1’-pyrrolizin]-2-ene) arising via the loss of acetic acid. Furthermore, the exocyclic nitro group of bicyclic compound 5 is in the same position as the second exocyclic nitrogen in common millipede tricyclic pyrrolizidine oximes (Kuwahara et al., 2007) and in nitropolyzonamine 2. Hence, compound 5 ((1Z)-8,9-dimethyl-1-(nitromethyliden)-2-azaspiro[4.4]non-8-en-7-one) could be the structural link between polyzonimine (1) and compounds of the nitropolyzonamine-type (2, 3, 4). However, to get a clearer picture on the formation of these compounds and their biosynthesis, it would be necessary to perform labelling experiments and to analyze additional species.

Already Schröder and colleagues (2008) described a combination of structure elucidation of compounds in mixtures and a more conventional structure elucidation strategy of purified compounds. In this study, structure assignments of venom compounds of spiders mainly relied on data acquired with 500 MHz and 600 MHz NMR spectrometers equipped with room temperature probes. The recorded NMR data sets comprised proton, DQF-COSY, HSQC, and HMBC spectra. We improved this approach by using a 700 MHz NMR spectrometer equipped with a cryoprobe. With our improved hardware we gained a higher spectral resolution and, more importantly, an around 10-fold increase of sensitivity. Therefore, we were not only able to record the aforementioned NMR experiments but also 1D carbon and the 2D HSQC-TOCSY experiments. With the latter, the sampling of resonances for many individual spin systems becomes a much easier task due to an additional carbon axis because ambiguities due to overlapping proton resonances are dramatically reduced. For a mixture of similar compounds, the NMR spectroscopic assignment strategy is in principle the same as used for purified complex natural compounds which consist of several similar building blocks, e.g., sugar units in larger saponins. Proton and carbon resonances of individual spin systems are identified mainly in the 2D HSQC-TOCSY experiment and assigned within the respective spin system by 2D DQF-COSY and HSQC experiments. As a last step, the individual spin systems are combined by HMBC correlations or NOEs to the final structure.

The cryoprobe and the analysis of whole body extracts dramatically reduces the demand for a large sample size of individuals for de novo structure elucidation. Based on this work, it is conceivable that by restricting the NMR measurements to the essential and higher sensitive set of proton-, COSY-, HSQC-, and HMBC-experiments would even allow investigation of more challenging sample quantities, e.g., from smaller arthropods such as oribatid mites (Raspotnig et al., 2023).

Our NMR-spectroscopic data led to the complete assignment of the structure, and in case of mass spectrometry to a tentative annotation of the structures of the detected analytes. This allows so-called dereplication, i.e., the annotation of already known structures without the need for their isolation (Wolfender et al., 2019). NMR spectroscopy has the advantages of simultaneous structural assignment and quantification of the major metabolites in a given mixture, and of being nondestructive to the samples. A useful by-product of the NMR spectroscopic approach is a better quantification of the compounds in the extract. Compared to NMR, MS-based techniques offer a much greater sensitivity, enabling the detection of minor constituents. Mass spectrometers can be easily hyphenated with chromatographic separation techniques like liquid chromatography or gas chromatography, thereby facilitating the in-depth analysis of complex mixtures. However, although mass spectrometric data are highly useful to confirm the presence of known compounds in a mixture or to deduce fragments of a structure, the technique alone does not allow the complete de novo elucidation of unknown compounds. Another limitation of MS-based analysis is the fact that not all analytes are ionized to the same degree. Therefore, the intensity of the signals is not directly proportional to the concentration of the metabolites, and compounds not ionized with the applied technique may remain undetected (Wolfender et al., 2015; Kellogg et al., 2017; Wolfender et al., 2019).

Overall, NMR and MS analysis can be regarded as complementary techniques for high-quality compound identification and quantification in natural product mixtures like in the secretions of P. germanicum.

For future studies on complex mixtures of novel compounds from arthropods, we suggest to rely on a similar approach combining different MS techniques with NMR equipped with a cryo-probe to record 2D experiments (e.g. HSQC-TOCSY). This setup will allow: I) to work with crude extracts from a small sample size, II) to characterize structures in a moderate time, and III) to better quantify the compounds in the extract for further ecological experiments.

The original contributions presented in the study are included in the article/ Supplementary Material . Further inquiries can be directed to the corresponding author.

Ethical approval was not required for the study involving animals in accordance with the local legislation and institutional requirements because millipedes are not listed in the species conservation law. We applied for a permission to perform collections as stated under “Acknowledgments”.

MB designed the study and performed collection and determination of Polyzonium germanicum. Acquisition and analysis of data was done in case of GC-MS by MB, and in case of HRLC-MS by EP-W. NMR experiments and resonance assignments were done by OK. DFT calculations were done by AO. All authors contributed to the manuscript, read, edited, and approved the final version.