Address correspondence and reprint requests to Jörg Hanrieder, PhD, Department of Chemical and Biological Engineering, Analytical Chemistry, Chalmers University of Technology, Kemivägen 10 (Floor 5), SE-41296 Gothenburg, Sweden. E-mail: firstname.lastname@example.org
Amyotrophic lateral sclerosis (ALS) is a devastating, rapidly progressing disease of the central nervous system that is characterized by motor neuron degeneration in the brainstem and the spinal cord. Matrix-assisted laser desorption/ionization (MALDI) imaging mass spectrometry is an emerging powerful technique that allows for spatially resolved, comprehensive, and specific characterization of molecular species in situ. In this study, we report for the first time the MALDI imaging-based spatial protein profiling and relative quantification of post-mortem human spinal cord samples obtained from ALS patients and controls. In normal spinal cord, protein distribution patterns were well in line with histological features. For example, thymosin beta 4, ubiquitin, histone proteins, acyl-CoA-binding protein, and macrophage inhibitory factor were predominantly localized to the gray matter. Furthermore, unsupervised statistics revealed a significant reduction of two protein species in ALS gray matter. One of these proteins (m/z 8451) corresponds to an endogenous truncated form of ubiquitin (Ubc 1–76), with both C-terminal glycine residues removed (Ubc-T/Ubc 1–74). This region-specific ubiquitin processing suggests a disease-related change in protease activity. These results highlight the importance of MALDI mass spectrometry as a versatile approach to elucidate molecular mechanisms of neurodegenerative diseases.
sodium dodecyl sulfate–polyacrylamide gel electrophoresis
superoxide dismutase 1
total ion count
As the worldwide proportion of elderly people above 60 is increasing constantly, the prevalence for age-related diseases is on the rise, particularly with respect to emerging newly industrialized countries (W.H.O 2002). These diseases of affluent societies include cardiovascular diseases, chronic obstructive pulmonary disease, cancer, diabetes and neurodegenerative diseases such as Alzheimer's disease, Parkinson's disease, multiple sclerosis and amyotrophic lateral sclerosis (ALS) (Hirtz et al. 2007). ALS, also referred to as Lou Gehring's Disease or Maladie de Charcot, is characterized by irreversible degeneration of motor neurons in the spinal cord, brainstem, and cortex. This results in increasing muscle weakness and muscle atrophy. Most commonly, people between age 40 and 60 are affected and worldwide prevalence of ALS has been reported to be 2–4 cases per 100 000 a year (Hirtz et al. 2007). ALS is a very progressive, fatal disorder, and patients have an average life expectancy of 3–5 years after diagnosis. Most of ALS cases are idiopathic in nature; however, 10% have been found to be familial cases caused by genetic factors. The main mechanisms underlying ALS are still unknown, although a couple of findings concerning molecular ALS pathology have been reported. For example, a mutation of zinc/copper-superoxide dismutase 1 (SOD1) has been discovered in 10% of the familial ALS cases (Chio 2000; Clement et al. 2003; Cristini 2006).
The limited knowledge of the disease results in further drawbacks such as the inaccuracy and time delay for diagnosis. This is because particularly early disease stages of ALS are shared by other neurological disorders, including multifocal motoric neuropathy (www.alsa.org). As the major biomolecular mechanisms underlying ALS remain largely elusive, no individual clinical or molecular marker for accurate disease diagnosis and prognosis is available. Establishing ALS diagnosis is therefore significantly complicated and involves rather a series of diagnostic tests to rule out other neurological diseases. These facts highlight the need for further investigation of the disease on a molecular level. The limited understanding of ALS pathobiology on a molecular level is a direct consequence of the lack of appropriate analytical technologies that feature the necessary sensitivity and specificity.
As the development of soft ionization technologies that facilitate analysis of large biomolecules such as proteins and peptides, mass spectrometry (MS) has gained an immense popularity in biological and medical research (Karas and Hillenkamp 1988; Fenn et al. 1989). In particularly, proteomics has gained great significance in these fields, as it provides the possibility to get a comprehensive insight into the protein expression profile of a biological sample, such as body fluids or tissue biopsies (Aebersold and Mann 2003; Domon and Aebersold 2006). Proteomic profiling can be employed to detect disease-related changes of protein abundance that might provide a further insight in disease pathology (Aebersold and Mann 2003). Furthermore, these protein regulations can serve as diagnostic markers for improved disease diagnosis and prognosis (Calvo et al. 2005). While proteomics has been employed for protein analysis in various neurodegenerative disorders, including Parkinson's disease (Licker et al. 2009) and Alzheimer's disease (Abdi et al. 2006), only few proteomic studies on ALS have been reported so far. These comprised analysis of different CNS specimen, including CSF (Ramstrom et al. 2004; Ranganathan et al. 2005; Pasinetti et al. 2006) and spinal cord tissue (Ekegren et al. 2006). Two of these studies were based on surface-enhanced laser desorption ionization mass spectrometry profiling of CSF samples (Ranganathan et al. 2005; Pasinetti et al. 2006). Here, a number of protein mass peaks were found to be significantly changed in ALS samples compared with controls including cystatin C, transthyretin, that were decreased as well as neurosecretory protein VGF (Ranganathan et al. 2005; Pasinetti et al. 2006) and secretogranin-5 (7B2) that was found increased, respectively (Ranganathan et al. 2005). Moreover, using a three protein peak model allowed for accurate sample classification at a 95% confidence level (Pasinetti et al. 2006). Other studies employed Fourier-transform ion cyclotron resonance mass spectrometry (FTICR)-based analysis of CSF (Ramstrom et al. 2004) and spinal cord tissue extracts (Ekegren et al. 2006) for sample classification as well as biomarker discovery. Here, LC-FTICR mass chromatogram patterns could be used to correctly identify 80% of all analyzed ALS sample (Ramstrom et al. 2004). FTICR-MS-based analysis on spinal cord tissue extracts and motor neurons excised by means of laser capture dissection, resulted in the detection of several protein species that could only be identified in either ALS samples or controls (Ekegren et al. 2006). These findings include proteins such as amyotrophic lateral sclerosis 2 protein (ALS2), neuregulin 1 (NRG1), neuropeptide B precursor, and alpha enolase that were found exclusively in ALS samples (Ekegren et al. 2006).
While tissue proteomics facilitates protein identification and quantitation, spatial information within the respective tissue compartment is difficult to obtain. Tissue dissection is a particular challenge in analysis of nervous tissue where different nuclei are located in close proximity. Considering the complexity of the human central nervous system, however, spatial information on protein distribution is of major interest to resolve ongoing molecular pathophysiological mechanisms. Matrix-assisted laser desorption/ionization (MALDI)-imaging mass spectrometry (IMS) is a powerful approach for spatial profiling of large molecular species in biological tissue samples (Caprioli et al. 1997). The method is based on discrete application of matrix solution to a thaw-mounted tissue section followed by MALDI MS analysis of the individual matrix deposits. The spatial resolution is predefined by the matrix application pattern. This technology features high molecular specificity that allows spatial intensity profiling of drugs, lipids, peptides, or proteins in situ and enables matching of chemical information to histological features. The technique is also referred to as molecular histology and can be employed in various clinical applications (Stoeckli et al. 1999, 2001; Cornett et al. 2007; Seeley and Caprioli 2008).
Here, we report a study on MALDI-IMS-based analysis of post-mortem human spinal cord tissue samples obtained from ALS patients and controls. Multimodal imaging MS approaches have previously been applied to rat spinal cord samples for lipid and neuropeptide and protein analysis (Van de Plas et al. 2007; Monroe et al. 2008; Landgraf et al. 2009, 2011; Girod et al. 2010, 2011; Tucker et al. 2011; Hanada et al. 2012), but studies on human spinal cord material have not been reported yet. The aim of this study was therefore to topographically elucidate the protein expression profile of human spinal cord and to gain further insight in molecular mechanisms underlying ALS, particular with respect to lower motor neuron degeneration.
Materials and methods
Post-mortem spinal cord of thoracic and high-thoracic level from four individuals with ALS and three controls were used in the study. Patients were diagnosed according to the El Escorial World Federation of Neurology Criteria for the diagnosis of ALS and clinically rated according to the Norris score. None of the ALS patients had the familial form of the disease. The cause of death in the control group was cardiac failure and myocardial infarction. The demographical data of ALS patients and control subjects are presented in Table 1. At autopsy, the spinal cords were dissected, removed from the dura mater, and cut into 5-mm sections. The tissue was then immediately frozen on a metal plate maintained in liquid nitrogen and stored at −72°C. Autopsy was performed with written approval from the family in accordance with Swedish rules and regulations at the time of autopsy.
Table 1. Patient data overview
Post-mortem time [h]
Disease duration [years]
Spinal cord Level
High thoracal (9)
High thoracal (7)
The complete experimental strategy is outlined in Fig. 1, as well as described in detail in the supporting information. Thoracic spinal cord sections (12 μm) were collected with a cryostate microtome and thaw-mounted on conductive glass slides suitable for MALDI-IMS analysis. Additional sections from control patients were collected for protein identification and stored at −80°C. The thaw-mounted sections were washed with ethanol followed by matrix application (sinapinic acid/2,6-dihydroxy-acetophenone) using a chemical inkjet printer (ChIP-1000, Shimadzu, Japan) with a lateral resolution of 350 μm covering the whole tissue section. Two replicate experiments were performed using histology directed analysis on the gray matter region (Cornett et al. 2006). Here, sample spots were manually assigned in accordance with the respective region of interest to reduce analysis time and data size. Data sequence preparation, data acquisition, and visualization were performed using the FlexImaging software (v 2.0, Bruker Daltonics, Bremen, Germany). All mass spectrometry data (mass range of 3–20 kDa) were acquired with an Ultraflex II MALDI TOF/TOF instrument (Bruker Daltonics) running in linear positive mode.
Data processing and statistical analyses
Regions of interest (ROI) were assigned for each spinal cord IMS experiment comprising the (1) white matter (WM), (2) the entire gray matter area, (GM) as well as the gray matter subregions: (3) ventral/anterior and (VH) (4) dorsal/posterior horn (DH) (for details see Fig. 2c). All mass spectra were batch processed by means of baseline subtraction, external calibration followed by peak picking and binning, analysis which was performed in order to reduce the data significantly (Wikland et al. 1985; Chen et al. 2009). Finally, the bin borders were used for area-under-curve peak integration followed by total ion count (TIC) normalization. Average peak areas for each peak bin were calculated for each ROI of every patient sample. These average peak area values were evaluated using the ‘Statistical Analysis of Microarray data’ (SAM, v.3.0) tool for grouped two class unpaired and paired statistical analysis (non-parametric test) in Excel (Tusher et al. 2001). The SAM tool was originally developed for microarray analysis and allows comprehensive and unbiased analysis of significant changes in abundance levels between two groups. First, two class paired analysis was performed by comparing the gray matter with the white matter peak values of each sample. Then, two class unpaired analysis was performed comparing white matter or gray matter data of the ALS samples to the respective ROI data of the controls. Follow-up statistical analysis of individual proteins and comparisons between the groups were performed with non-parametric anova (Kruskal–Wallis) followed by post hoc analysis (Turkey). The null hypothesis was rejected at p < 0.05.
Protein identification strategies
Sodium dodecyl sulfate–polyacrylamide gel electrophoresis
Spinal cord proteins were extracted from previously collected cross-sections by ultrasound homogenization in 2% sodium dodecyl sulfate buffer, followed by centrifugation (RCF = 10 000 g) and overnight protein precipitation with ice-cold acetone. Precipitated protein fraction were denaturated, reduced, and alkylated before sodium dodecyl sulfate–polyacrylamide gel electrophoresis (XT 12% Bis Tris, Bio Rad, Hercules, CA, USA) at 200 V for 45 min in MES running buffer. Nine gel-bands covering the region between 2.5 and 20 kDa were excised from the sodium dodecyl sulfate gel according to the illustration in Fig. 1f, cut in small cubes of approximately 2-mm edge length and subjected to in gel digestion as described elsewhere (Hanrieder et al. 2011).
Intact protein HPLC and MALDI MS
Protein extracts were divided and one part was further fractionated by means of mass weight cut-off (MWCO, 30 kDa) filtration. Three different protein samples including the intact extract, the MWCO flow through, and supernatant were then further fractionated by means of C8 reversed phase HPLC. Gradient elution was followed by manual collection of the eluting proteins in 1.5 mL tubes with a collection rate of approximately 1/min. MALDI MS of the intact protein fractions was performed using a so-called sandwich protocol for target preparation with sinapinic acid as matrix (Hanrieder et al. 2011). MALDI MS was performed in automatic acquisition and processing mode using the WarpLC software (v.1.2, Bruker Daltonics) designed for LC-MALDI analysis. The samples were analyzed by MALDI TOF MS (Ultraflex II, Bruker Daltonics). Fractions of interest were selected and subjected to in-solution digestion as described elsewhere (Ekegren et al. 2006).
Nanoflow liquid chromatography coupled to electrospray ionization Fourier-transform ion cyclotron resonance mass spectrometry (nanoLC-ESI FTICR-MS) was performed on an Agilent 1100 nanoflow system (Agilent Technologies, Santa Clara, CA, USA) hyphenated to a LTQ-FT 7.0 Ultra mass spectrometer (Thermo Scientific, Waltham, MA, USA). The reconstituted digests (both in-gel and in-solution digests) were injected automatically and separated on a C18 reversed phase column using gradient elution. Peptide elution was followed by ESI FTICR-MS and tandem mass spectrometry (MSMS) for peptide sequencing. Protein identification was achieved by comprehensive MSMS search against the uniprot knowledge base (www.uniprot.org) using the Mascot search engine (v. 2.2, Matrix Science, London, UK).
Double-antigen fluorescent immunohistochemistry was performed as described previously using rabbit anti-ubiquitin (1 : 100, ab7780, Cambridge, UK) in combination with mouse anti-NeuN (1 : 100, Millipore, Temecula, CA, USA) primary antibodies (Ljungdahl et al. 2011). The fluorescent secondary antibodies anti-rabbit-Alexa Fluor 488 (1:200, Invitrogen, Carlsbad, CA, USA) and anti-mouse-Alexa Fluor 555 (1 : 200; Invitrogen) were used for visualization. Cell nuclei were stained by 4′,6-diamidino-2-phenylindole. Sections stained without primary antibody (only the antibody diluent, 2.5% horse serum in phosphate buffered saline, in the primary antibody step) served as negative controls.
Gray matter regions were dissected at −20°C from frozen spinal cord, followed by proteins extraction in RIPA buffer (Radio Immuno Precipitation Assay) using manual disintegration (pestle) and sonication. The tissue debris was removed by centrifugation (10 000 g) followed by protein concentration determination in the supernatants (BCA assay, Sigma, St. Louis, MO, USA) and normalization prior reduction, alkylation and sodium dodecyl sulfate–polyacrylamide gel electrophoresis at 150 V for 45 min. Blotting was performed in transfer buffer (25 mM Tris, 192 mM glycine, 20% MeOH) for 90 min at 40 V, using a double layer of hybond-P membrane (GE Healthcare, Uppsala, Sweden). The membrane was washed (Tris-buffered saline-Tween 20), blocked for 1 h (2% blocking reagent in Tris-buffered saline-Tween 20, enhanced chemiluminescence (ECL) GST, GE Healthcare), and incubated with primary antibody overnight (1 : 500, rabbit anti ubiquitin ab7780, abcam, Cambridge, UK). After profound washing, secondary antibody was applied for 1 h (1 : 5000, goat anti-rabbit horseradish peroxidase, P0448, Dako Sweden AB, Stockholm, Sweden), followed by ECL detection (ECL GST, GE Healthcare). The ECL film was scanned using a commercial office scanner and evaluated in image J (v.1.46, http://imagej.nih.gov/ij/). Statistical analysis was performed by means of unpaired Student's t-test (2 tails, p < 0.05) in MS Excel (v. 2010).
Results and discussions
Experimental parameters for MALDI-IMS
The aim of this study was to relate chemical information obtained by mass spectrometry to distinct anatomical regions in human spinal cord. Therefore, we performed MALDI-IMS on 12 μm thick thoracic spinal cord tissue cross-sections. Furthermore, we aimed to identify protein species that exhibited significant changes in regional abundance distribution in between the ALS patient group and controls. As MALDI mass spectrometry on tissue samples is highly dependent on several parameters including molecular heterogeneity, complexity, analyte abundance, ionization behavior, and tissue type inherent matrix effects, routine throughput analysis is far from trivial and requires extensive protocol optimization. Hence, multiple experiments were conducted to optimize the experimental parameters including sample handling, matrix composition, and matrix application. With respect to sample handling, we observed that a stepwise washing protocol with first 70% EtOH for 30 s followed by 95% EtOH for 120 s was found to give the best data with respect to number of detected peaks and signal to noise. For matrix application, we utilized a chemical inkjet printer that enables application of matrix solution as discrete pL-volume droplets, resulting in single isolated matrix spots arranged in a quadratic pattern. This approach is particularly beneficial with respect to restricted sample diffusion and sample extraction efficiency (Aerni et al. 2006). The parameters that had to be optimized included the matrix concentration and matrix solvents. Furthermore, the composition of the matrix solution had to be optimized with respect to organic modifiers, other matrices, and pH. The final matrix concentration per spot could be modified by the actual matrix concentration in the matrix solution, the total number of droplets deposited per spot, as well as by the number of application passes. Here, an increased number of passes allows the deposition of smaller number of droplets per pass resulting in a higher lateral resolution, which on the downside increases the overall throughput time dramatically. In summary, concerning the matrix application protocol one has to compromise between the matrix concentration, the lateral resolution, and the throughput time. For spinal cord samples, we observed the best protein mass-peak signal was observed with a matrix concentration of 10 mg/mL SA and 5 mg/mL DHAP solved in 60% MeOH, 10% 0.15 M AmAc, and 0.3% TFA. The matrix application protocol was optimized to the deposition of 1500 pL (15 drops) per spot during each application pass with a lateral resolution of 350 μm. A number of 30 passes of matrix application resulted in a total volume of 45 nL matrix per spot.
Protein peak identification
Matrix-assisted laser desorption/ionization imaging of post-mortem spinal cord resulted in the detection of multiple protein mass peaks throughout the tissue sample. The observed protein peaks were preliminary assigned according to their intact mass. Direct on tissue protein characterization can be achieved by micro-spotting of trypsin followed by imaging MS and MSMS of the proteolytic peptides. However, this approach is significantly hampered for low abundant analytes (Groseclose et al. 2007). Therefore, bottom-up proteomic analysis of tissue homogenates was performed in order to validate the mass matched protein identities. A number of 197 proteins were identified using a 1 DGE/nanoLC–MSMS approach (Table S1). For targeted identifications of proteins of interest, we employed a complementally top-down approach. Here, spinal cord protein extracts were prefractionated by means of mass cut filtration (MWCO 30 kDa) to prevent bias from larger protein species. Intact proteins (M < 30 kDa) were then separated by C8 reversed phase LC followed by fraction collection and MALDI TOF MS-based inspection of intact protein fractions. Individual protein fractions that contained protein mass peaks of interest were subjected to tryptic digestion and LC-ESI FTICR-MSMS-based protein identification. These targeted LC × LC experiments resulted in 142 protein species (Table S2). Interestingly, only 27 protein identities were found with both proteomic strategies highlighting the orthogonal character of the two protein fractionation techniques. In total, 18 protein mass peaks could be identified through mass matching and MSMS-based sequence validation (Table 2). Here, 17 protein peak assignments were verified through 1DGE/nanoLC-MSMS, while 8 were found by targeted top-down analysis (2DLC). However, it is important to emphasize that the 2DLC experiments served as a targeted strategy for identifying single protein peaks of interest rather than global spinal cord proteomic profiling. The here-reported proteins assigned and verified with both proteomic strategies have also been reported by other groups (Table 2). Although identified in human tissue samples, these clinical specimens (typically tumor tissue) vary dramatically from spinal cord tissue, a fact one has to keep in mind when putatively assigning proteins according to intact mass values found in the literature. This highlights the need for independent validation experiments such as LC-MSMS, immunohistochemistry and western blotting.
Table 2. Protein Identification. Protein species identified by mass matching followed by top-down and bottom-up proteomics
With MALDI-IMS analysis of human post-mortem spinal cord sections, distinct protein localizations were observed. The selected ion images for the corresponding proteins presented in Fig. 2 show a characteristic intensity distribution throughout the tissue sample that are in accordance with anatomical regions and histological features. These included different histone proteins: histone H4 (m/z 11308), H2A1C (m/z 14014) and histone H2B1 (m/z 13778) and their post-translational modified isoforms that were found to localize to the whole gray matter (Fig. 2a) (Hardesty et al. 2011). In contrast, ubiquitin (Ubc) localizes to the dorsal horn of the gray matter (Fig. 2b) (Pierson et al. 2004; Reyzer et al. 2004). Interestingly, a putative truncated version of ubiquitin (Ubc-T, m/z 8451) corresponding to removal of the two C-terminal glycine residues shows distinct distribution to the whole gray matter similar to the histone proteins (Fig. 2b) (Goncalves et al. 2008). On the other hand, myelin basic protein (m/z 14124, Fig. 2b) showed localization to the white matter. This is well expected, as white matter is composed of axon bundles, sheathed with myelin.
Unbiased comprehensive peak analyses were performed to find additional proteins that where predominately localized to gray or white matter regions, or specifically distributed to the ventral or dorsal horn. Here, all peak integration data (binning results, area under curve) of spectra from the respective anatomical region of interest were collected. This comprised in between 94–81 spectra for the GM and 333–857 spectra for the WM depending of the tissue size (Figure S1). Average value of all peak bin integration results for each ROI (GM, WM) of each individual sample was determined. First, pairwise SAM analysis was performed for the white matter compared with the gray matter to determine significant differences between these regions for all samples irrespective to the sample group (ALS and controls). The comparison was also performed for solely control samples and ALS samples, respectively. However, these results did not differ in between the two groups indicating that there is no disease-related change in protein localization to either white or gray matter.
A striking majority of detected proteins were found localized to the gray matter, which was indicated by significantly higher peak intensities in that region compared with the white matter (Fig. 3a, b). To ensure that these GM-specific intensity increases are not a result of higher ionization efficiencies as a result of GM centric sample preparation optimization, the TIC of the gray matter spectra was compared with white matter data (Student's t-test). Here, no significant difference was observed. However, TIC normalization was previously demonstrated to lack robustness and can lead to artifacts as a result of over normalization (Deininger et al. 2011). We therefore performed additional statistical analysis on median normalized data as suggested in a recent study by Deininger et al. (2011). However, we obtained the same results for both normalization techniques, which strengthen the confidence in our findings. In this study, certain protein peaks were found significantly increased in the white matter as well (Fig. 3c). In total a number of 68 out of 294 protein peaks were found to be significantly different in between the two regions of interest (SAM p < 0.05, Fig. 3d). A number of 50 proteins were significantly higher in the gray matter, of which 36 protein peaks showed more than 30% higher peak area values compared with evenly distributed proteins (Fig. 3a, b). In contrast, some protein peaks showed distinct localization to the white matter (Fig. 3c). In total a number of 18 protein masses, of which 17 showed more than 30% higher peak area levels were found to be significantly higher in the white matter compared with the gray matter (p < 0.05, Fig. 3c, d).
The most prominent mass peak detected in the gray matter was corresponding to the mass of thymosin beta 4 (Tyb4, m/z 4963, Fig. 3a, Figure S1a) (Reyzer et al. 2004). Furthermore, ubiquitin, histone proteins (H4, H31, H2A, H2B), hemoglobin (HBA, HBB) macrophage inhibitory factor, protein S100-A6 (S100-A6), and acetyl-CoA-binding protein (ACBP) were found to predominantly localized to the gray matter (Fig. 2, Fig. 3a, Table 2) (Reyzer et al. 2004; Oppenheimer et al. 2010; Schwamborn and Caprioli 2010). Interestingly, free ubiquitin (m/z 8565; Fig. 2b, Fig. 3a) although significantly higher in the gray matter was predominantly localized in the dorsal horn of the gray matter as described above. This is of particular interest, as ubiquitin-mediated degradation in the proteasome pathway was observed previously for mutant SOD1 but not wild-type SOD1 in an animal model of familial ALS (Rosen et al. 1993; Miyazaki et al. 2004). Furthermore, a direct correlation of increased ubiquitin-binding to mutant SOD1 and disease severity was observed (Rosen et al. 1993; Miyazaki et al. 2004). Proteasomal dysfunction and formation of ubiquitin positive inclusion bodies have previously reported to be significantly involved in ALS pathology, as well as in other neurodegenerative diseases (Petrucelli and Dawson 2004; Ross and Poirier 2004). Most interestingly, a recent study in a large family with ALS-identified mutations in the ubiquilin 2 gene, a transporter for ubiquitinated proteins that resulted in proteasomal dysfunction and formation of protein aggregates (Deng et al. 2011). Further investigations revealed that UBQLN2 mutation and formation of protein aggregates occurred in every ALS case studied and are a common factor in both familial and sporadic ALS (Deng et al. 2011). Moreover, mutant UBQLN2 was demonstrated to cause proteasomal dysfunction and impaired Ubc-mediated protein degradation in transfected neuroblastoma cells. These data strongly suggest that UBQLN2 mutation and inferred proteasomal dysfunction plays a prominent role in ALS pathology and is a common cause in different forms of ALS (Deng et al. 2011). Another protein of great interest that was localized to the gray matter is S100 protein A6, a Cu2+/Zn2+ binding protein (m/z 10092; Fig. 3a) (Oppenheimer et al. 2010). Previous studies reported an over-expression of S100-A6 in astrocytes that were adjacent to axons from degenerated motor neurons in an animal model of ALS as well as in post-mortem spinal cord tissue from patients with sporadic ALS (Rosen et al. 1993). These findings might be particularly relevant, as the animal model of ALS is based on a genetic mutation of Cu/Zn SOD1, which is associated with familial ALS (Rosen et al. 1993).
For the obtained IMS results, not all detected mass peaks could be assigned to corresponding protein identities. However, some of these unknown protein peaks that were observed in the SAM analysis results showed distinct distribution patterns in spinal cord that are well in line with histological features. These included, for example, protein peaks with m/z 5361, m/z 5652, m/z 8429, and m/z 12307 that localized significantly to the gray matter (Fig. 3b) as opposed to other unidentified protein peaks that localized to the white matter including m/z 11613, m/z 12643, m/z 16371, m/z 16452, and m/z 16504 (Fig. 3c, Figure S1b).
Spatial protein dynamics in ALS
To determine region-specific differences in protein intensities that are ALS-related, unpaired statistical analysis of averaged peak bin integration data of the respective ROI (GM, VH, DH, and WM) of the ALS and control samples was performed. The number of spectra varied for each ROI in each sample, respectively. In addition to the number of GM and WM spectra that is described above, the different regions of the gray matter comprised 39–125 spectra for the ventral horn and 23–68 spectra for the dorsal horn region of the GM. Unpaired statistical analysis of the different regions of interest (GM, VH, DH, and WM) in between the two sample groups showed significant changes in the gray matter; however, no consistent change for white matter proteins was observed. Therefore, histology directed experiments that were performed in duplicates for each sample were focused on the gray matter region (GM) and its subregions (DH, VH). The number of acquired spectra for each respective ROI was similar to the imaging data. Here, a number of 82–148 spectra per patient were collected for the GM, 29–42 for the dorsal horn, and 47–106 for the ventral horn, respectively. Although no consistent differences were observed for the subregions (DH, VH), for two protein peaks including m/z 8429 and m/z 8451, a consistently significant lower peak intensity (p < 0.05) was observed in the whole gray matter (GM) of ALS patients compared with the whole GM of controls in all three replicate experiments (Fig. 4). Interestingly, none of the proteins reported previously to be changed in ALS were found changed in the present study (Ekegren et al. 2006). However, this might be because of the different experimental strategies employed, as the present study was performed on proteins with a molecular weight below 20–30 kDa (corresponding to the MALDI mass range) as a result of mass cut filtration or selective gel range excision.
The data analysis results were followed up a series of validation experiments aiming to determine the correct identity of these two protein species. First, the spatial ion intensity distributions were investigated closely to clarify whether m/z 8451 is solely a sodium adduct peak of 8429 (mass shift from protonated peak to sodium adduct peak: 22 Da) which in turn should result in identical localization. However, the ion distribution patterns were found to be different suggesting that both peaks originate from different protein species (Fig. 4a). Another aspect of this is that cation adduct peaks (M+Na+, M+K+) are typically significantly smaller than the corresponding main analyte peak. Furthermore, we performed C8 reversed phase HPLC of intact proteins extracted from spinal cord followed by fraction collection. The individual fractions were spotted onto a MALDI target plate for protein LC-MALDI analysis allowing for targeted follow-up analysis of intact protein peaks of interest. Again, assuming the two protein peaks were originating from the same protein and are a sole cationic adduct observed in MS would result in identical peak retention during liquid chromatography. Here, the two protein peaks were not found to exhibit a similar retention behavior further strengthening the hypothesis previously deduced from the ion images that these two protein peaks originate from different protein species. For protein identification, the corresponding fractions containing the two protein peaks of interest were subjected to tryptic digestion followed by LC-MSMS analysis of the enzymatic cleavage products. Detection and identification of the protein peak m/z 8429 have not been reported previously, and the here-reported identification experiments using bottom-up proteomics did not yield any results for this particular mass peak. The mass peak m/z 8451, however, has been previously reported to correspond to C-terminally truncated ubiquitin (Meistermann et al. 2006; Hardesty et al. 2011). Indeed, database analysis of the LC-MSMS data of the respective protein fraction showed identification of ubiquitin as one of the major protein matches. The other hits comprised myelin basic protein and ACBP. This is well in line with other major peaks observed in the MALDI data of the respective HPLC fraction including m/z 18418 and m/z 9964 previously assigned to myelin basic protein and ACBP, respectively (Table 2). Interestingly, Ubc-T is not an artifact due to unspecific cleavage, as its intensity distribution pattern differs significantly from intact ubiquitin, and points rather to a disease-related processing pattern of ubiquitin in human spinal cord (Fig. 2, panel b). Ubc-T has previously reported as potential biomarker of breast cancer as revealed by surface enhanced laser desorption/ionization mass spectrometry (Goncalves et al. 2008). During Ubc-mediated proteasomal protein degradation, C-terminal removal of two glycines leads to inactivation of Ubc. However, differential regulation of Ubc-T in contrast to steady-state Ubc levels as reported by Goncalves et al. points toward a distinct biological function of Ubc-T and Ubc processing, respectively, that evidently require further investigation (Goncalves et al. 2008). Previous studies have demonstrated that the bioconversion of ubiquitin 1–76 to ubiquitin 1–74 (Ubc-T) is predominantly mediated by cathepsin B, a lysosomal cysteine protease and member of the peptidase 1 family (Herring 2009). Interestingly, microarray-based gene expression studies in human ALS, as well as on experimental ALS in mice, have identified significant changes of cathepsin B mRNA levels in spinal cord tissue, however with contradicting results (Kikuchi et al. 2003; Offen et al. 2009). For example, a decrease of Ubc-T could consequently point toward reduced cathepsin B activity, which is supported by data on post-mortem human spinal cord reported by Kikuchi et al. 2003). where decreased levels of cathepsin B mRNA were detected in ALS patients compared with controls (Kikuchi et al. 2003). Alternatively, decreased Ubc-T levels can also be a result of decreased Ubc levels and unspecific enzymatic cleavage. Decreased ubiquitin levels can be directly related to formation of Ubc-positive protein inclusions. Here, recruited Ubc is incorporated in the protein aggregates by forming stable covalent bonds, which in turn decreases the level of free Ubc molecules that can be detected by MALDI mass spectrometry. Either way, ubiquitin regulation and activity seem to be significantly affected in ALS. As second means of validation of MALDI-IMS results, double-antigen immunohistochemistry for the neuronal nuclear antigen NeuN (Fig. 5c, arrows) and ubiquitin (Ubc) confirmed neuronal loss and a decrease in Ubc immunoreactivity in the gray matter of ALS patients (n = 3 in each group). NeuN was chosen to monitor neuronal loss in ALS as well as target neuronal ubiquitin concentration specifically. However, the Ubc antibody recognizes both free and conjugated ubiquitin. In addition, the immunogen was the full-length human ubiquitin, which raises the possibility that this antibody also recognizes the short (-GlyGly, −114 Da) form of recycled ubiquitin. Lipofuscin autofluorescence was abundant in all sections and is seen as yellow granular staining, sometimes surrounding the nucleus. Due to these interferences, a quantitative statement was impossible to deduce from the immunohistochemistry results. We performed western blot experiments on gray matter protein extracts of all patient samples to quantify the degree of changes in abundance of ubiquitin species. The results show indeed that ubiquitin species are decreased by 57–53% in ALS gray matter samples compared with controls (Fig. 5 a, b; p < 0.05; Figure S2). These results can be regarded as further confirmation that there is an overall decrease of Ubc-related species in ALS as observed in MALDI-IMS, probably related to impaired protein turnover. However, due to lack of antibody specificity, differentiation between the intact and truncated Ubc is not possible. In addition, the 114 Da difference between Ubc and Ubc-T was undetectable by western immunoblotting. This makes it difficult to draw conclusions whether or not C-terminal Ubc processing is related to ALS, although the MALDI results show that Ubc-T but not Ubc peak values were reduced in ALS, further highlighting the superiority in terms of molecular specificity of this technology. Selective decrease in Ubc-T but not Ubc would support the theory that decreased cathepsin B activity is involved in ALS pathology (Fig. 4) as previously suggested (Kikuchi et al. 2003). However, one has to keep in mind that MALDI MS as a semi-quantitative technique is characterized by a certain variance due to its susceptibility to matrix effects, which could explain why intact Ubc 1–76 was not found consistently decreased in ALS. These include, for example, ion suppression as a result of impurities and changes in relative analyte concentration This is significantly aggravated when patient material is used where large individual variations are inherent as well as other factors including post-mortem time, sample retrieval, sample age and sample storage contribute to biological sample variation significantly. As shown in the present study, only a very small number of protein species was identified that showed consistent and significant changes in peak intensity in all replicate experiments. Hence, these results perhaps detect only the protein peaks that exhibited the most consistent changes caused by ALS and might thus represent only an excerpt of all changed protein species including ubiquitin 1-76. This also highlights the need for sufficient sample size and well-controlled experimental parameters for sample handling to minimize biological variation. For the present study, these factors are compromised due to the fact that human post-mortem material is used and sample access is limited due to the low prevalence of ALS. However, the here-presented results demonstrate that profound optimization of sample preparation and data acquisition parameters for minimizing technical variation allows for MALDI-IMS-based investigation of such diverse biological samples with reasonable and reproducible outcome. In summary, the observed results suggest that regulation of ubiquitin and regional ubiquitin metabolism is potentially connected to pathophysiological mechanisms underlying ALS. These data support previous observations that impaired protein turnover is a hallmark in ALS pathology.
In the present study, a MALDI imaging MS-based strategy for spatial protein profiling in post-mortem spinal cord is reported. In addition, at least two protein species including a C-terminally truncated form of ubiquitin were found decreased in ALS spinal cord samples compared with controls. These findings were further validated by western blot and immunohistochemistry experiments. These findings highlight the great potential of MALDI-IMS as a powerful tool in neuroscience research.
The authors thank Prof. Håkan Askmark (Department of Neuroscience, Neurology, Uppsala University Hospital) for the contribution of post-mortem material. The Swedish Research Council (Grant 623-2011-971 (JH), 342-2004-3944 (JB), 621-2008-3562 (JB), 621-2011-4423 (JB), 522-2006-6416 (MA) and 521-2007-5407 (MA); The Royal Swedish Academy of Sciences KVA (MA, JH), the Wenner-Gren Foundations (JH), and the Swedish Chemical Society (JH) are gratefully acknowledged for financial support. The authors declare no conflict of interest.