Introduction

Mass spectrometry (MS) and tandem MS (MS/MS) play an important role in structure investigations of molecules like proteins or other organic compounds [1,2,3,4]. Several different MS/MS methods are commonly available. For instance, fragmentation can be achieved by collision-induced dissociation (CID) or higher-energy collisional dissociation (HCD), representing collision-based fragmentation activation methods. Here, energy is transferred into the precursor ions by collisions with neutral gas molecules. Other activation methods, such as electron capture dissociation (ECD) and electron transfer dissociation (ETD), creating an instable radical species by electron addition, are also widely spread [5,6,7,8]. In addition, infrared multiphoton dissociation (IRMPD) and blackbody infrared radiative dissociation (BIRD) are often used in Fourier-transform ion cyclotron resonance (FTICR) instruments [9,10,11]. Both methods use infrared (IR) radiation with wavelengths of typically 10.6 μm for IRMPD and ~ 5.5 μm for BIRD to activate the precursor ions. The energy transferred by individual IR photons is low compared to the bond dissociation energy of organic molecules (~ 0.116 eV for IRMPD) [12,13,14,15], so a multiphoton absorption of ~ 10 photons is typically needed.

In the last couple of years, also ultraviolet (UV) light came into focus for fragmentation activation. These approaches typically utilize lasers lines at λ = 213 nm, 193 nm, or 157 nm [16,17,18,19]. The latter belong to the region of vacuum ultraviolet (VUV) light, which allow the radiation with high-energy photons (7.9 eV at 157 nm and 6.4 eV at 193 nm), thus reducing the number of transferred photons necessary for fragmentation to 1 or 2 [18]. Dipeptide and tripeptide fragmentation was then investigated after irradiation with light of 193 nm [19, 20]. Larger sequences followed after the introduction of electrospray ionization (ESI). In comparison to UV light with longer wavelengths (266 nm, 248 nm), a significantly higher fragmentation efficiency was observed for VUV irradiation. Furthermore, complete y-type series were generated, resulting in high sequence coverages of peptides. Due to the high fragmentation rate, increased abundances of a-, b-, c-, x-, y-, and z-type fragments were observed (for nomenclature, please refer to Figure 1) [23, 24]. Peptides with aromatic side chain residues showed a significantly higher dissociation rate due to the aromatic absorption maxima in the UV range [19, 23]. Another predominant reaction pathway observed was the loss of electrons resulting in charge-reduced radical ions [25,26,27]. Similar studies were carried out using the 157 nm activation. For peptides, a charge-dependent fragmentation was observed here. For singly charged peptides containing arginine residues, Reilly [19] observed predominantly a- and x-type fragments, [19]but also fragments generated by side chain dissociation. In comparison, doubly charged peptide precursors were found to typically dissociate into y- and b-type fragments in CID [18]. The formation of atypical fragments like w-, d-, and v-type fragments led to the possible discrimination between leucine and isoleucine [21, 22]. Furthermore, disulfide bridges could be cleaved homolytically by VUV light, resulting in two radicals [28]. When comparing the 193 nm and 157 nm activations, a shift towards larger m/z values of fragments was found for the 157 nm photodissociation of peptides [28, 29].

Figure 1
figure 1

Scheme of the peptide and oligonucleotide fragmentation nomenclature [21, 22]

Besides peptides, also glycan structures have been studied. In these investigations, predominantly B-/Y- and C-/Z-type fragments were observed [30]. While especially sialic acid residues lead to very low–abundance species in carbohydrate fragmentation, Brodbelt et al. [16, 31] detected a loss typical for sialic acid. Other studies also investigated fatty acids besides peptides and glycans using VUV-based MS/MS [29, 32, 33,34,35,36]. The possibility of generating unique fragments, as the photoactivation of precursor and reporter ions creates significantly higher information content, has led to a further development [37].

While lasers exhibit advantages like a high brilliance and low divergence, they suffer from two shortcomings for dissociative ion activation: (I) Even 157 nm (7.9 eV) photons do not carry enough energy for single photon dissociation of most bonds, and (II) pulsed lasers lack an efficient duty cycle, and the typical nanosecond pulses do not correspond to the typical trapping times. Since the ions inside an ion trap are continuously available for excitation, a continuous wave light source, preferably with a wavelength between 130 and 70 nm, would be favorable. The lower limit is derived from considerations of conservation of energy. All excess energy transferred into a molecule or ion and not needed for the bond rupture is usually channeled into kinetic energy, hampering a trapping of the nascent fragments. Another, mere technical, consideration is the manipulation of light, since traditional optical elements like lenses are not readily available for wavelengths below 115 nm.

Thus, recently discharge lamps and light-emitting diodes (LEDs) came into focus for continuous radiation of trapped ions. Giuliani et al. [38] presented an approach using 12 LEDs in an Orbitrap by substitution of the HCD cell. Since LEDs are still not available at the wavelengths of interest, Giuliani et al. [38] went one step further and coupled an extreme ultraviolet (XUV) discharge lamp operated at the He 1s2 ← 1s12p1 line of 21.2 eV to the TriWave region of a Waters Synapt G2-Si Q-ToF [38]. Besides neutral loses and the dominant radical cation, they found b-/y-type ion series, besides x- and v-type ions [38].

Like the single photon energy and the temporal characteristics affect the energy dissipation after excitation, and consequently, the fragment distribution, the exact excitation strongly differs between monochromatic and polychromatic irradiations. Consecutive absorption is favored by longer irradiation because the photon densities are usually too low for simultaneous two-photon processes. A subsequent absorption, however, relies on a resonant step, i.e., an energy level corresponding to the exactly available wavelength.

Thus, in the present study, we employed a low-cost continuous polychromatic VUV light source with emission maxima at around 115 nm and 160 nm as a simple and economically meaningful alternative to lasers or synchrotrons. A linear ion trap was employed, and the fragmentation behavior of several substance classes upon VUV irradiation was investigated.

Experimental

Oligonucleotides were purchased from BioTeZ Berlin-Buch GmbH. Peptides were synthesized in house by solid-phase synthesis. Other chemicals were supplied by J.T. Baker, Sigma-Aldrich, and Carl Roth GmbH &Co. KG. All samples were dissolved in water with 30% methanol and 1% trimethylamine for negative ions and 30% methanol and 0.1% formic acid for positive ions with a final concentration of 10 μM.

ESI-MS and MS/MS spectra were obtained using a LIT LTQ (Thermo Fisher Scientific). The instrument parameters were tuned automatically for the maximum intensity of each precursor ion investigated. Fixed instrumental parameters were as follows: 100 °C transfer capillary temperature and 30 arbitrary units of nitrogen sheath gas flow. Samples were directly infused at a flow rate of 10 μL min−1. The precursor ions were isolated for CID and VUV-induced dissociation experiments with an isolation width of Δm/z = 3 and fragmented with an activation time of 30 s for VUV light with 0% CID activation energy. In CID, 35% relative collision energy was applied for 30 ms.

To couple VUV radiation into the instruments’ ion trap, a commercially available deuterium lamp (L10366 series, Hamamatsu) with an emission spectrum ranging from ~ 115 to 400 nm (Supplementary Information) was mounted via the backdoor of the LTQ. A schematic is shown in Figure 2a. For positioning the lamp, the central hatch at the backend of the LTQ was replaced by an aluminum plate with a central boring mounted to a 50 mm KF-pinch screw connection. At the inner side of the plate, two rail posts are mounted, which allow for a manual adjustment of the exact position of two lens holders. The first lens mount (as following the path of the light) housed a custom-made planoconvex MgF2 lens for collecting the light, and the second holder downstream was equipped with a planoconcave counterpart for collimation. The respective lens specifications were 12.7 mm (6 mm) in outer diameter, + 15 mm (− 5 mm) in curvature radius, 1 mm (2 mm) in rim thickness, and 2.41 mm (1 mm) in central thickness for the planoconcave and the planoconcave (values in parenthesis), respectively (Kristalle, Korth, Germany). Both lenses were polished to laser grade surface roughness. The exact position of the lenses were optimized according to a ray tracing simulation (Zemax, OpticStudio, Zemax LLC, USA) to narrow down the beam path through the central aperture in the LTQ’s ion trap end electrode. Since the refractive indices for MgF2 for 115 nm and 160 nm significantly differ and thus with a simple doublet, no ideal focusing of polychromatic light at the given wavelengths can be achieved, and the optimum settings for 160 nm were chosen. This assumption is, however, not without pitfalls since highly energetic VUV photons are prone to cause a photoelectric effect on the electrode surfaces of ion optics. Thus, the observed fragment ion distributions may, up to some content, also contain products of collisions between precursor ions and low kinetic energy photoelectrons. Since the output of the D2 lamp is continuous, no temporal synchronization was necessary. Again, potential artifacts caused by irradiation during the filling of the trap may occur but were not considered in the interpretation of the spectra. The lamp was operated at a driving power of 140 VA, corresponding to a 30 W D2 spectrum. Unfortunately, no values for the exact output in the VUV region are given. As a rough estimate, a comparable lamp provided and calibrated by the PTB (German Metrology Institute) provides a spectral radiant intensity of 10−4 W nm−1 sr−1 to 2 · 10−3 W nm−1 sr−1. Under optimum conditions, roughly 50% of the emitted light could be guided into the ion trap. For all experiments, the activation time was controlled by trapping the ions with 0% relative collision energy for a defined time, while the VUV light was switched on continuously. This continuous exposure of the ion trap volume may potentially introduce experimental artifacts as mentioned above. However, since in the current study, the ion storage and exposure times were long in comparison to the filling times of the ion trap, all the fragmentation was considered to occur while the parent ions were confined inside the trap.

Figure 2
figure 2

Experimental setup. (a) Schematic of the VUV lamp–ion trap setup and respective beam paths. (b) Lamp exit window enclosed in vacuum hatch. (c), (d) MgF2 lenses (convex and concave, respectively), mounted on sliding holders. Pictures taken before eloxation. (e) ray tracing simulation of VUV light path under given configuration. The three different colors represent light of different wavelengths, 115 nm (green), 160 nm (blue), and 200 nm (red), respectively

Results

Multiply charged species of positive and negative ions of biomolecules were investigated. For negative ions, mainly electron losses, resulting in a charge reduction, were observed for oligonucleotides (Supplementary Information). The extend of electron losses depended on the radiation time. After 1 s irradiation with VUV light, a threefold charge reduction to [M−8 H]5− was observed already. With longer irradiation times, the charge reduction increased, resulting in a fivefold electron loss from the precursor [M−8 H]8− to [M−8 H]3−. An increase in irradiation time thus resulted in higher electron losses in both abundance and the number of losses with an increasing number of additional fragments. This phenomenon is called “electron photodetachment dissociation” and is well known for UV radiation [39, 40]. Gabelica et al. [26, 41] could show that this is also a base-dependent phenomenon. They correlated the electron photodetachment to the nucleobase ionization potentials, showing that the radical ion intensity increases with higher guanine content during irradiation with 260 nm [26, 41].

A complementary phenomenon was observed for multiply charged lysozyme with positive ions that stands in contrast to other investigations (Figure 3). Giuliani et al. [38] radiated multiple protonated cytochrome c. They increased the photon energy to 15.5 eV and investigated a radical product ion (9+) from the precursor [M+8 H]8+ [42]. In the present study, the precursor ion [M+11 H]11+ was detected as [M+11 H]10+ with a reduced charge state after radiation with VUV light. This means electron capture by the positively charged precursor ions occurred. While in negative mode the electron losses can be explained by removal of an electron from the precursor ions upon VUV irradiation, the electron attachment in positive mode requires additional electrons. These electrons have to originate from the inside of the MS. In the instrumental setup using a diverging light source, scattered stray light occurs. The photons of VUV light are of high energy (up to ~ 10 eV) allowing electrons to be removed from metallic surfaces, lenses, or even the LIT rods. These generated electrons are mostly slow and may be captured by highly charged positive ions. This capture then leads to the observed charge reduction. This effect cannot be observed for negative ion mode because of Coulomb repulsion, and it stands in contrast to the other works [38]. With increasing irradiation times, also fragments from dissociation of the charge-reduced species were found. The ability to capture or to remove electrons from trapped ions could be used for substances that show low-abundance fragments or a high number of neutral losses. Especially for low-charge-level precursor ions, this could be a possible way to post-ionize fragments that are otherwise uncharged (neutral losses) and so increase the informational content of MS/MS spectra. To further investigate the compound-specific fragmentation behavior, we investigated a number of biocompounds in their VUV fragmentation behavior.

Figure 3
figure 3

VUV-based MS/MS spectrum of the DNA oligonucleotide 5′-AAA TAA GGC CGC TAT-3′. The signal m/z 572 was isolated and irradiated for (a) 1 s, (b) 5 s, (c) 10 s, and (d) 20 s. The precursor ion signals are highlighted

As first substance class, we investigated oligonucleotide ions that hold multiple positive or negative charges and form a number of fragments. As the smallest compound, adenosine triphosphate was irradiated with VUV light and compared to CID activation (Supplementary Information). With VUV irradiation, the number of fragments almost doubled in comparison to CID. Multiple cleavages along the pyrophosphate units and combined fragments from those cleavages and base loss were found. The resulting spectra were completely different for both activation techniques. With VUV light, higher fragment intensities were observed.

Next, we analyzed a 4-mer oligonucleotide. Besides the primary reaction pathway of electron loss with VUV light, also a different fragmentation behavior in comparison to CID was observed here (Figure 4).

Figure 4
figure 4

MS/MS spectra of the DNA 4-mer 5′-TGAC-3′ in the twofold negative charge state. (a) CID MS/MS spectrum. (b) VUV-based MS/MS spectrum. The precursor ion is fully fragmented in (a) and is highlighted in (b)

In both methods, predominantly w-type fragments were generated. However, significantly more fragments were observed with VUV light. In the VUV-derived spectrum, additional fragments of the charge-reduced species were found alongside fragments resulting from backbone cleavages like d3/d2 that were not present in CID. The fragment d3 represented the second most intense signal here. Fragments from backbone cleavages without further base losses are a typical fragment class in HCD activation and need more energy [6]. Hence, VUV activation has a potential to fill the gap between CID and HCD. While the most abundant signals in VUV fragmentation originated from charge reduction, consecutive water losses were found exclusively with VUV light. The preferences concerning backbone cleavages including a base loss also shifted from CID to VUV light. In CID, a3-A was the most abundant fragment signal, whereas with VUV, a3-T/z3-C was the most dominant in this fragment class.

Duplex DNA 15-mers were also investigated in the charge state 5−. Again, much more individual fragments were formed with VUV light (Supplementary Information). VUV fragment ions from single strands were found in higher charge states, up to threefold negatively charged, while only one- and twofold negatively charged CID products were detected.

The fragmentation of peptides was investigated starting from a small 2-mer peptide (Glu-Glu). When activated by CID, Glu-Glu yielded typical y- and b-type fragments for negative ions, besides water and CO2 losses (Figure 5a). VUV activation resulted in a-/b-/x- and y-type fragments, and also a v-type fragment resulting from a side chain cleavage was observed (Figure 5b). Similar results were reported for UV laser radiation [16]. For longer peptides, the negative ion CID spectrum was not informative as mainly water and CO2 losses occurred (Supplementary Information). For a better comparison, we analyzed Glu-fibrinopeptide B as it was investigated before with various activation methods, such as ECD, UV photodissociation (UVPD), XUV, and CID [38]. In the negative ion mode, we found significantly more fragments in the VUV fragment spectra in comparison to the CID spectra. CID generated, as expected, preferentially b- and y-type fragments, whereas in VUV light, we detected a large number of a-/b-/c-/x-/y- and z-type fragments (Supplementary Information), similar to previous observations. Thus, with VUV light, a great fragment diversity was found that potentially enables negative ion mode sequencing in contrast to the typical b-/y-type fragments in CID. This stood in contrast to investigation in the positive ion mode. Antoine et al. fragmented a 5-mer peptide in the positive ion mode with a nanosecond laser showing less fragments than in CID. They found two b-type fragments and one v-type fragment, besides radical formation by hydrogen loss [43]. Furthermore, Bari et al. [44] used a synchrotron radiation source for peptide fragmentation in the positive ion mode. They found, as mentioned before, several b- and y-type fragments, besides internal fragmentation and radical formation by electron loss [44].

Figure 5
figure 5

MS/MS spectra of the twofold negatively charged peptide 2-mer Glu-Glu. (a) CID MS/MS spectrum. (b) VUV-based MS/MS spectrum. The precursor ion is fully fragmented in (a) and is highlighted in (b)

As a saccharide example, an 8-mer maltodextrin was analyzed using CID MS/MS and VUV-based MS/MS. In both techniques, a y-type series and multiply water losses were observed (Figure 6) [18]. In CID, a simple water loss was observed from the precursor ion, whereas VUV light generated also double water losses. The fragment class 2,4An was found in both spectra, but the class 0,2An was exclusively present in VUV-derived fragment spectra. Furthermore, again a higher number of atypical fragmentation pathways were observed with VUV light.

Figure 6
figure 6

MS/MS spectra of the 8-mer maltodextrin in the onefold negative charge state. (a) CID MS/MS spectrum. (b) VUV-based MS/MS spectrum. The precursor ions are highlighted

Melezitose and the raffinose dimer were also assessed (Supplementary Information). In CID, melezitose fragmented within the sugar units, whereas with VUV light, this was not the case. In both methods, complete sugar units and water losses were found. The CID spectrum of raffinose only contained fragments of the monomer units, whereas with VUV light, different fragment classes were detected. Fragments of the monomer units, resulting from water losses and dissociation of separate sugar units, as well as typical A- and X-type fragments, were observed.

Double bonds efficiently absorb UV light [45]. Therefore, VUV activation could represent a new MS/MS method not only for biomolecules but also for substances like organic dyes, containing several double bonds. All tested dyes (xylenol orange, meso-tetraphenylporphin-4,4′,4″,4′″-tetrasulfonic acid, chlorophyllin) resulted in fragment-rich MS/MS spectra after irradiation with VUV light (Supplementary Information). The number of fragments was higher with VUV light in comparison to CID in all cases. In the spectra of xylenol orange, for example, VUV light generated a high number of intense fragment signals originating from CO2 losses, side chain dissociations, and combinations of these, whereas in CID, only one fragment from one side chain cleavage was detected (Figure 7).

Figure 7
figure 7

MS/MS spectra of the singly negatively charged precursor ion of xylenol orange. (a) CID MS/MS spectrum. (b) VUV-based MS/MS spectrum. The precursor ion is fully fragmented in (a) and is highlighted in (b)

Similar results were found for chlorophyllin (Supplementary Information). In CID, only single and double CO2 losses occurred, while with VUV light, single, double, and triple CO2 losses and additional fragments from side chain cleavages were detected.

Conclusion

In this study, we present a new VUV-based MS/MS method using a commercially available low irradiance continuous deuterium lamp instead of a previously employed UV laser [16, 19]. This simple instrumental setup enables to investigate the dissociation behavior of different substance classes using VUV activation. All tested compounds show a high fragmentation efficiency upon VUV irradiation. The observed phenomena of electron loss in the negative ion mode and electron capture in the positive ion mode form the most abundant signals in VUV-based MS/MS. This change of charge can be exploited for post-ionization of fragments not holding a charge. Once charged, the radical species subsequently fragmented to atypical fragments. For almost every compound class, a significantly broader distribution of fragments with VUV light compared to CID was observed. More information was provided in addition to commonly known dissociation pathways, for example concerning the v-type fragment in peptides.

VUV fragmentation is possible and efficient not only for negative ions but also for positive ions. However, the VUV light ionizes and fragments so efficiently; in the positive ion mode, high signals from chemical background in the mass range m/z 0~400 were observed in all spectra and, thus, only higher m/z values could be reasonably assessed.

In summary, VUV light yielded fragment-rich spectra and therefore generated significantly higher information content as CID. Especially the loss and capture of electrons could lead to a better ionization of hardly ionizable fragments. Furthermore, we have shown that a simple deuterium lamp generates photons with enough energy to dissociate even stable molecules like DNA. VUV light may be used as stand-alone technique or, for example in a linear ion trap, in combination with other MS/MS methods alongside CID to unravel more structure information. Currently, the necessary exposure time of 30 s makes the method too slow to compete with existing techniques. A more effective coupling of the VUV light into the trap volume, however, bears the potential to dramatically shorten this time, enabling the technique to a useful complementary addition to conventional MS/MS fragmentation instrumentation.