Bottom-Up LESA Mass Spectrometry of Human Liver Tissue Sections
Intact proteins were extracted from healthy human liver tissue via automated direct surface sampling by LESA and subjected to automated trypsin digestion using the method described by Martin et al. . In that work, Martin et al. found that 1 h was the optimum digestion time. Three extraction solvents were considered: 70% (v/v) methanol, 50% (v/v) methanol, and 50 mM ammonium bicarbonate. The three extraction solvents were applied for LESA of three serial sections from a tumor resection margin. For each extraction solvent composition, two technical replicates were performed. Two searches of the data against the protein database were performed in order to determine the effect of the short digestion time. In the first, two missed trypsin cleavages were allowed, and in the second, three missed cleavages were allowed. When two missed cleavages were allowed, a total of 475 non-redundant proteins were detected: 428 proteins were detected following the ammonium bicarbonate extraction; 200 were detected following the 50% methanol extraction, and 99 were detected following the 70% methanol extraction. When three missed cleavages were allowed, 549 non-redundant proteins were detected: 464 proteins were detected following the ammonium bicarbonate extraction; 214 were detected following the 50% methanol extraction, and 116 were detected following the 70% methanol extraction. See Figure 1a. Proteins identified are detailed in Supplementary File 1. Single peptide protein identifications are presented in Supplementary File 2. Increasing the organic solvent content of the extraction solution resulted in fewer identifications, and this is perhaps not surprising. This observation has relevance to the top-down analysis and is discussed further below.
The bottom-up protocol used here was originally developed for the analysis of dried blood spots . The number of proteins identified from the liver tissue is over 4-fold greater than were detected in dried blood spots, despite using a much smaller volume for the liquid microjunction (1 μL versus 6 μL). This observation is perhaps expected given that the liver is a center for metabolism within the body and has an abundance of proteins and enzymes . Blood is a small component of liver tissue and abundant proteins such as alpha and beta hemoglobin chains were also detected. It is possible that the numbers of identified proteins could be increased, either by pooling extracted samples or by performing on-tissue digestion . The latter cannot be achieved, at least in our hands, using the TriVersa NanoMate because of the problem of evaporation of the trypsin solution.
Three known liver disease markers were identified in each of the three extractions: the liver fatty acid binding protein (FABP1) (MW 15.1 kDa) , gamma-glutamyltransferase (MW 77.3 kDa) , and aspartate amino transferase . There are two forms of aspartate amino transferase, one found in the mitochondria (MW 47.5 kDa) and the other found in the cytoplasm (MW 46.2 kDa). The homology between the two sequences is 68% (as calculated by BLAST http://blast.ncbi.nlm.nih.gov/Blast.cgi). Both proteins were identified by unique peptides in the current dataset. Proteins gamma-glutamyltransferase and the mitochondrial aspartate amino transferase were detected with greatest sequence coverages (32.90% and 48.37% respectively) following the ammonium bicarbonate extraction. The greatest sequence coverage (52.52%) for cytoplasmic aspartate amino transferase was observed following extraction with 50% methanol.
FABP1 was detected in each of the six experiments with sequence coverage ranging between 53% and 57% (see Supplementary Figure 1). Figure 1b shows the sequence coverages obtained for the two replicates obtained following extraction with ammonium bicarbonate solution. As described above, FABP1 and its variant implicated in NASH differ by a single amino acid residue (i.e., Thr→Ala at position 94). The tryptic peptide containing residue 94, that is [91-96], was not observed in any of the analyses. (It is possible that an SRM approach may identify the relevant peptide more reproducibly; however, the Orbitrap instrument used in this work is not suited to such an approach). One of the analyses following extraction with ammonium bicarbonate resulted in the identification of the missed-cleavage tryptic peptide [81-96] (m/z
meas 1791.9931, m/z
calc 1791.9851, Δ 4.9 ppm) from FABP1. The CID mass spectrum obtained from the [M+3H]3+ peptide ions is shown in Figure 2a. The fragmentation sequence coverage for the peptide is poor with no cleavage around the potential substitution site. Clearly, as the variant is the result of a single amino acid substitution and the sequences of FAPB1 and its variant are virtually identical, with the exception of peptide [81-96], the peptides assigned to FABP1 in the database search could also have originated from the variant. The human SwissProt database against which the database was searched was supplemented with the FABP1 variant (hereafter referred to as FABP1TA). The peptide [90-96] from FABP1TA was not identified in any of the analyses; however, the missed-cleavage peptide [81-96] was identified in one of the ammonium bicarbonate extraction replicates (m/z
meas 1761.9831, m/z
calc 1761.9745, Δ 4.5 ppm). The CID spectrum is shown in Figure 2b. Again, fragmentation sequence coverage is poor and no fragmentation was observed adjacent to the site of substitution. It should be noted that top-down analysis of this tissue sample (see below) revealed the presence of the variant.
Untargeted Identification of Intact Proteins in Human Liver
Four samples were investigated by top-down LESA mass spectrometry: two NASH samples and two healthy (non-diseased) samples. The non-diseased samples were from (1) a tumor resection margin, and (2) a donor liver not suitable for transplantation.
Liver sections were thaw-mounted onto glass slides and dipped into 80% (v/v) ethanol prior to LESA analysis. This washing step removes lipids and improves protein signal in the mass spectra. The sections were subjected to LESA using extraction solvent comprising 40% (v/v) acetonitrile + 1% formic acid, and the resulting samples introduced directly, via electrospray ionization, into the mass spectrometer. Deconvolution of the resulting mass spectra using the Xtract program suggested that the spectra comprised peaks corresponding to up to 16 proteins in healthy tissue and up to 25 proteins in NASH tissue, with molecular weights ranging between 3000 and 20,000 Da. The results from the bottom-up analyses described above suggest that a greater number of proteins are extracted using ammonium bicarbonate solution; however, the top-down workflow requires that the extraction solution must also be suitable for electrospray ionization. If the extraction and electrospray steps were decoupled, it may be possible to increase the number of top-down identifications.
Figure 3 shows the mass spectrum obtained from a healthy liver sample. Four abundant protein ions were selected for tandem mass spectrometry. Note that the nature of the proteins was not known a priori. Ions at m/z 946.38 (+16 charge state) were selected for collision induced dissociation MS/MS. The data were searched against the simple human database in Prosight PTM 2.0, and the protein was identified as the alpha chain of hemoglobin (MWmeas 15117.9167, MWcalc 15117.8924, Δ 1.6 ppm). Fragments identified are detailed in Supplementary Table 1. The protein sequence coverage following CID was 31%. Ions of the same m/z (m/z 946.38) observed in the analysis of NASH tissue were selected and subjected to electron transfer dissociation MS/MS. Again, the protein was identified as α-hemoglobin. The ETD protein sequence coverage was 13% and the fragments observed are detailed in Supplementary Table 2. The combined sequence coverage from the two analyses was 42%.
Ions at m/z 775.42 (+14 charge state) observed in the mass spectrum from healthy tissue were selected and subjected to both CID MS/MS and ETD MS/MS and the data searched against the simple human database in Prosight PTM 2.0. (A peak at this m/z was also observed in NASH tissue however its abundance was insufficient for MS/MS). The protein was identified as mitochondrial 10 kDa heat shock protein (MWmeas 10836.8056, MWcalc 10836.8485, Δ 3.96 ppm). The protein sequence coverage was 25% (CID) and 21% (ETD), giving an overall coverage of 39%. Details of the fragments observed are given in Supplementary Tables 3 and 4.
The ions at m/z 1084.80 and 1087.04 (both +15 charge state) were each selected for CID MS/MS. The two species were identified as FABP1 (MWmeas 14111.4161, MWcalc 14111.3892, Δ 1.9 ppm) and the variant FABP1TA (MWmeas 14081.3600, MWcalc 14081.3687, Δ 0.6 ppm). Both proteins were detected with the initiator methionine removed and acetylation of the N-terminus. Therefore, the 92nd and 93rd amino acids in each chain are TT and TA, respectively. The CID sequence coverage obtained was 47% for FABP1 and 33% for the variant FABP1TA. Fragments observed are detailed in Supplementary Tables 5 and 6. The site of substitution can be confirmed. The CID MS/MS spectra from both FABP1 and FABP1TA are shown in Figure 4.
Ions of m/z 942.23 and 940.29 (+16 charge state) (corresponding to FABP1 (MWmeas 14111.3943, MWcalc 14111.3892, Δ 2.4 ppm) and FABP1TA (MWmeas 14081.2845, MWcalc 14081.3687, Δ 6.0 ppm) from a NASH sample were subjected to ETD MS/MS. The sequence coverages obtained were 38% and 29%, see Supplementary Tables 7 and 8. The ETD MS/MS spectra of FABP1 and FABP1TA from NASH tissue are shown in Figure 5.
Figure 6 shows expanded m/z regions containing the +16 charge state of FABP1 and FABP1TA of the mass spectra obtained for the four samples analyzed. Figure 6a and b show the mass spectra from healthy (non-diseased) tissue. In one sample, only FABP1 is observed, whereas in the other both FABP1 and the variant are observed. Figure 6c and d show the mass spectra from NASH tissue. Similarly, one sample contains the variant alone, whereas the other reveals the presence of both FABP1 and FABP1TA. As described above, it has been demonstrated recently that individuals with FABPTA have an increased risk of developing NASH . In addition, it was shown that another polymorphism compounds that risk. The observation of the variant in non-diseased tissue could be explained by the risk associated with FABPTA being attenuated by other factors. It is worth noting, however, that the sample containing both FAPB1 and the variant was from a tumor resection margin, whereas the sample containing FAPB1 alone was from a liver rejected for transplant because of the presence of a small lesion. The non-diseased samples are also characterized by reduced signals for α-hemoglobin. In the case of the liver rejected for transplant (Figure 6a), the organ was perfused with buffer (i.e., blood was removed) following removal from the donor. In the case of the tumor resection margin (Figure 6b), it is likely that the surgical procedure resulted in local hemolysis and reduced blood supply.