Production and analysis of multiply charged negative ions by liquid atmospheric pressure matrix‐assisted laser desorption/ionization mass spectrometry

Rationale Liquid atmospheric pressure matrix‐assisted laser desorption/ionisation (AP‐MALDI) has been shown to enable the production of electrospray ionisation (ESI)‐like multiply charged analyte ions with little sample consumption and long‐lasting, robust ion yield for sensitive analysis by mass spectrometry (MS). Previous reports have focused on positive ion production. Here, we report an initial optimisation of liquid AP‐MALDI for ESI‐like negative ion production and its application to the analysis of peptides/proteins, DNA and lipids. Methods The instrumentation employed for this study is identical to that of earlier liquid AP‐MALDI MS studies for positive analyte ion production with a simple non‐commercial AP ion source that is attached to a Waters Synapt G2‐Si mass spectrometer and incorporates a heated ion transfer tube. The preparation of liquid MALDI matrices is similar to positive ion mode analysis but has been adjusted for negative ion mode by changing the chromophore to 3‐aminoquinoline and 9‐aminoacridine for further improvements. Results For DNA, liquid AP‐MALDI MS analysis benefited from switching to 9‐aminoacridine‐based MALDI samples and the negative ion mode, increasing the number of charges by up to a factor of 2 and the analyte ion signal intensities by more than 10‐fold compared with the positive ion mode. The limit of detection was recorded at around 10 fmol for ATGCAT. For lipids, negative ion mode analysis provided a fully orthogonal set of detected lipids. Conclusions Negative ion mode is a sensitive alternative to positive ion mode in liquid AP‐MALDI MS analysis. In particular, the analysis of lipids and DNA benefited from the complementarity of the detected lipid species and the vastly greater DNA ion signal intensities in negative ion mode.

particular, liquid atmospheric pressure matrix-assisted laser desorption/ ionisation (AP-MALDI) has been shown to produce multiply charged positive ion species while also providing the advantage of low sample consumption and high analytical sensitivity. 3,5 Examples of potential application areas include bottom-up proteomics 6,7 with the advantages of off-line sample preparation, high-speed mass spectrometry (MS) analysis, long-term sample storage and higher tolerance to trifluoroacetic acid (TFA) when compared with ESI, and the analysis of lipids, 8 enabling new fragment ion measurements, and, thus, the detection of diagnostic ions for elucidating fatty acid double-bond locations even from crude biofluids like milk. The latter was achieved by employing an extremely simple sample preparation method and conventional collision-induced dissociation (CID) and in-source fragmentation as implemented on standard commercial instrumentation. Many of the advantages described above have not been demonstrated with any other laser-based ionisation technique while producing multiply charged analyte ions at the same time. In addition, one of the main and unique advantages of liquid MALDI is its extremely stable ion production and sample longevity, enabling analyte ion production for hours or tens of thousands of laser shots from as little as 1 μL of sample with very little analyte ion signal variability and sample consumption. [9][10][11] Using an AP heated ion source, this extremely high ion signal stability and low sample consumption can also be achieved for the production of predominantly multiply charged ESI-like analyte ions. 3,7 The mechanisms behind the production of multiply charged ions using laser-based desorption techniques and the link to ESI have been discussed in previous publications. 5,12,13 Although many future application areas have been explored for liquid AP-MALDI MS in the positive ion mode, no studies have explored its potential by switching to the negative ion mode. Here, it should be emphasised that liquid AP-MALDI MS should not be confused with MALDI MS employing ionic liquid matrices in conventional vacuum ion sources as the latter still employs solid MALDI samples in many cases and does not lead to ESI-like charge states.
This contribution presents an exploratory study of the differences and potential advantages of employing the negative ion mode in liquid AP-MALDI MS as used for the production of ESI-like negatively charged biomolecules. The focus will be on three classes of molecules.
Using peptides and proteins, overall liquid matrix optimisation was undertaken as well as a comparison of the detection response between a basic and acidic protein in both the positive and negative ion mode. Second, DNA was studied as an obvious analyte class with respect to the potential gain by switching to the negative ion mode.
The benefits of negative DNA ion analysis have been recorded in many articles, including some of the first articles published on DNA analysis using conventional solid MALDI MS. 14-16 Third, lipids from simple milk extracts were investigated with respect to the complementary information gained by switching to the negative ion mode. Previous studies on MALDI MS analyses of lipids in the negative ion mode have pointed out the opportunities of detecting different types of lipids compared with the positive ion mode. 17,18 Although liquid AP-MALDI does not increase the charge state of lipids as it does for larger biomolecules, its simple sample preparation is highly convenient for the MS analysis of crude liquids and their constituents, providing an extremely stable ion beam. Mass spectra were processed with MassLynx 4.1. Where ion signal intensity is reported, the spectra were first subjected to MassLynx' automatic peak detection and analysis with automatic smoothing and centroiding. Protein spectra were processed with UniDec (University of Oxford, Oxford, UK), which was primarily used for binning and background subtraction. 19  for CCS (applicable to the latter database only). Additional CCS values following the same criteria are cited from the literature. 21

| RESULTS AND DISCUSSION
In the first set of experiments, two established liquid MALDI sample preparations from earlier positive ion mode studies were used as well as two preparations that were slightly different by swapping the matrix chromophores with compounds which promised to be more suitable for negative ion mode analysis. The former two preparations employed DHB and CHCA/3-AQ, respectively, while the latter two employed 3-AQ (without CHCA) and 9-AA as the chromophores; otherwise the MALDI sample preparations were identical.
In positive ion mode it was evident that out of the four different liquid MALDI sample preparations the one with DHB outperformed all of the other three with respect to the highest mean signal intensity for virtually all peptides and charge states (see Table 1). In fact, many charge states, particularly 3+, 4+ and 5+, were only detected with the DHB-based liquid matrix. Interestingly, in negative ion mode multiply charged peptide ions were far less observed (mainly for the ACTH clips) and no ion signal at all was obtained for somatostatin (see Table 2). This poor performance was somewhat surprising since instrument parameters were tuned for negative ion mode prior to data acquisition. However, the two peptides substance P and bombesin were detected, although they have no acidic sites (not even at the C-terminus) that could readily deprotonate. More importantly, substantially lower negative ion signal intensities have previously been reported for basic peptides like the ones used here. 22 Contrary to the positive ion mode, the liquid MALDI matrices The same four liquid matrices used for peptide analysis were also investigated for protein analysis. In addition, a fifth liquid matrix that was prepared with CHCA as the only chromophore was also tested, since this was previously found to be superior to the DHB-based liquid matrix for the detection of larger proteins. 12 For ubiquitin, both the CHCA-and DHB-based liquid MALDI samples performed best with respect to positive ion signal intensities. In negative ion mode, the best performing liquid MALDI samples were clearly the ones containing 9-AA while the DHB-, CHCA-and CHCA/3-AQ-based MALDI samples hardly produced any negative protein ions (see Figure 1). Interestingly, for ubiquitin in the positive ion mode, the recorded protein ion charge states for the DHBand CHCA-based samples covered a range of 5-14, displaying a bimodal distribution in the case of DHB. In all other cases the charge state distribution covered a smaller and lower range of 3-7, which was similar to the range when negative protein ion signals were obtained.
As ubiquitin is a basic protein with an isoelectric point of around 8.5, another protein, trypsin inhibitor, with an isoelectric point of ≤5 was also analysed. Commercial trypsin inhibitor is typically provided as a mixture of three different proteoforms at low purity. In addition, its molecular weight is almost three times as much as that of ubiquitin.
Therefore, compared with ubiquitin the overall quality of the mass spectra was lower, i.e. lower protein ion signal intensities amongst higher background ion signals. As anticipated the CHCA-based liquid MALDI samples now outperformed the DHB-based liquid MALDI samples with respect to ion signal intensity in positive ion mode. In negative ion mode, neither of these two liquid matrices facilitated the detection of trypsin inhibitor while the 9-AA-based liquid MALDI samples allowed its detection in both ion modes albeit at a muchreduced signal intensity (see Figure S1, supporting information).
For both proteins 9-AA provided substantially higher ion signal intensities in the negative ion mode. However, in all cases the negative ion mode resulted in much higher adduct ion formation, which was identified as the addition of sodium in the cases where the resolution and signal intensity were sufficiently high. Interestingly, DHB has always been preferred to CHCA for the analysis of proteins in solid MALDI, arguably due to its greater 'softness' . In liquid MALDI MS, other aspects such as proton affinity might have greater influence as this and previously published data suggest. 12 Another group of analytes that was investigated consisted of three single-stranded DNA/oligodeoxyribonucleotide molecules of various length. Their sequences were ATGCAT (DNA1), ATGCATGCA (DNA2), and ATGCATGCATGC (DNA3). As can be seen in Figure S2 (see supporting information) positive ion signals were generally weak and the DHB-and CHCA/3-AQ-based matrices clearly outperformed the basic chromophores 9-AA and 3-AQ. In negative ion mode, once more the latter two resulted in far greater ion signal intensities compared with the former, with a clear tendency to produce higher levels of multiply charged ions. For ESI MS analysis of oligonucleotides, negative ion mode generally outperforms positive ion mode. 27 Thus, given the apparent commonalities with ESI processes, it is not surprising that this was also observed for liquid AP-MALDI MS using basic chromophores. In particular, the MALDI samples using 9-AA showed extremely strong negative ion signals, which were superior to the positive ion signals of all tested MALDI samples with respect to the number of charges, the individual DNA ion signal intensities (apart from the singly charged ion signal of the short DNA1 analyte) and the overall DNA ion signal intensity summed up over all charged states.
Comparing the spectra with the greatest DNA ion signal intensities for the positive and negative ion mode, i.e. DHB-based samples in positive ion mode vs 9-AA-based samples in negative ion mode, respectively, the difference in spectral quality and signal-to-noise becomes evident (see Figure 2). It appears that in the positive ion mode cation adducts are produced with a preference for potassium adduct ion formation ( Figure 2A) while sodium adduct ions were observed in the negative ion mode (Figures 2D-2F). In general, cation adduct formation was lower the higher the charge state was, again favouring the negative ion mode.
For DNA analysis in the negative ion mode the limit of detection (LOD) was around 10 fmol for the doubly charged DNA1 analyte using 9-AA as the chromophore (see Figure 3).
The final analyte class tested were lipids, which were extracted from milk. The extract of milk lipids was first analysed in positive ion mode, using the above liquid MALDI sample preparation, and thus similar to the methods of previously published work, 8 though without the addition of metal salts. As before, the DHB-based MALDI samples provided the highest analyte ion signal intensities in positive ion mode.
In negative ion mode, the 9-AA-based MALDI samples proved again to be the superior choice for obtaining the strongest analyte ion signal.
These samples resulted in spectra with the lowest chemical background noise due to MALDI matrix ion signals (Figure 4).
Advantageously, the spectra from the positive and negative ion mode are highly complementary, and thus the good analytical performance of liquid MALDI in negative ion mode is a great benefit, able to add an additional dimension of diagnostic information. Table S1 (   factors such as the sample surface tension and the type of matrix seem to have an important influence on the ionisation. 12 However, in many cases, the use of small concentrations of additives, e.g. other matrix compounds or acids such as TFA, do not seem to influence the overall analytical performance, allowing greater flexibility in sample composition than what is possible with ESI. 7,28 The data of the work presented here suggest that the type of matrix and its basicity, and thus the overall basicity of the liquid MALDI sample, is arguably another factor that needs further consideration and future investigations for establishing optimal liquid MALDI sample preparations, in particular for analytes that might benefit from negative ion mode analysis.

| CONCLUSIONS
Mass spectrometry analysis in negative ion mode is frequently inferior to positive ion mode analysis, which can be partly attributed to the mass and mobility of the electrons (compared with protons and other cations) and the technical challenges arising from this in the negative ion mode. However, for many analytes and MS sample environments/conditions, negative ion mode analysis has its benefits.
In this study, data has been collected for liquid MALDI MS, producing multiply charged analyte ions in both ion modes, that is in general agreement with this difference between negative and positive ion mode analysis. While no real benefit was found for the analysis of peptides and proteins by switching to negative ion mode analysis, the analysis of other analytes was significantly enhanced by the negative ion mode, in particular in combination with a switch to a different MALDI matrix chromophore such as 9-aminoacridine with a higher gas-phase basicity. The liquid MALDI MS analysis of DNA in negative ion mode showed significantly better performance with regard to the individual DNA ion signal intensities and the overall DNA ion signal intensity summed up over all charged states, enabling the detection of higher charge states that were not detectable in the positive ion mode. This improvement was particularly evident for larger DNA. So far the recorded limit of detection (LOD) is around 10 fmol, using low nanomolar concentrations for MALDI sample preparation. Arguably, the greatest benefit of negative ion mode analysis was found in the analysis of lipid extracts from milk, where the two ion modes provided highly complementary datasets of lipids without any overlap.   Table S1 (supporting information) Jeffery M. Brown http://orcid.org/0000-0001-8569-7174 Rainer Cramer http://orcid.org/0000-0002-8037-2511