Amyloid protein identification from tissue specimens analyzed by laser capture microdissection and liquid chromatography tandem mass spectrometry (LMD/LC–MS/MS) and data dependent acquisition (DDA) method

Our microdissected and processed tissue specimens were analyzed in DDA mode and identified spectra were filtered for ≥ 95% peptide and ≥ 95% protein probabilities with at least 1 unique peptide and minimum of 3 spectral matches. The amyloid protein identification is based on the proteins list obtained after the protein database search. For example, Figure S1, shows a total number of 342 proteins from tissue specimens collected from Congo red negative controls from bone marrow and heart biopsy versus Congo red positive heart biopsy is reported in a table showing a list of protein names and corresponding total spectral counts for each sample. Transthyretin amyloid protein subtype was identified from the three biological replicates from the Congo red positive heart biopsy sample (ATTR-H-4) together with signature proteins Serum amyloid P component and Apolipoprotein E. The transthyretin amyloid protein was absent in the Congo red negative controls from bone marrow (CRN-B-3) and heart (CRN-H-2). Therefore, these specimens were sub-typed as transthyretin amyloidosis or “ATTR”.

We analyzed 31 tissue specimens, i.e., 5 Congo red negative and 26 Congo red positive with the corresponding number of samples per amyloid protein sub-type identified (Fig. 2A).

Fig. 2

A Amyloidosis sub-types from tissue specimens analyzed by LMD/LC–MS/MS. B Total spectral counts for transthyretin protein from CR (+) ATTR specimens analyzed by ± FAIMS with no dynamic exclusion (DE). C Total number of proteins identified from CR (+) ATTR specimens with and without FAIMS with no dynamic exclusion (DE). D Total number of proteins identified with and without FAIMS (± FAIMS) with and without dynamic exclusion E Total spectral counts of amyloid proteins with and without FAIMS with and without dynamic exclusion (± DE)

Out of 31 specimens analyzed, 46% of the specimens analyzed were sub-typed as transthyretin amyloidosis (ATTR) (Fig. 2A, Table S1, Figure S2A and S2B). Out of the twelve ATTR subtypes, ten specimens were from heart, one from nerve and one from stomach. This indicates that cardiac ATTR is the most prevalent amyloidosis subtype among the specimens we have analyzed. Transthyretin is synthesized in the liver and functions as a transport protein for thyroxine hormone and retinol (Vitamin A). The wild type form (senile ATTR amyloidosis) is observed in older patients over 65 years old. The hereditary form of ATTR has 120 sequence variants of transthyretin [10, 19, 21]. Both the wild type and mutant forms of transthyretin can result in extracellular accumulation of amyloid fibrils that can be localized or systemic resulting in organ failure.

Immunoglobulin light chain amyloidosis (AL) is one of the most common types of amyloidosis. We identified AL-kappa and AL-lambda, each accounting for 19% of the amyloidosis types in our dataset. The organ source varied from lymph node, kidney, heart, intestine, stomach, colon and thigh mass (Fig. 2A, Table S1, Figure S2C and S2D). Our proteomics analysis highlighted that careful differential diagnosis between cardiac AL from cardiac ATTR amyloidosis (wild type and Val122I variants) is possible using our pipeline. This is critical, because ATTR has clonal immunoglobulin abnormality and resembles AL amyloidosis, which may result in misdiagnosis of AL. This could lead to incorrect treatment by unnecessary high-dose chemotherapy or autologous peripheral blood stem cell transplantation, while ATTR requires liver transplantation [21, 59].

Apolipoprotein-A1 amyloidosis (AApoAI) was identified from the lung and colon specimens (Fig. 2A, Table S1). APOA1-associated gastrointestinal and pulmonary amyloidosis have been reported and support our results [60, 61]. Apolipoprotein A-I is synthesized in the liver and small intestine and is the main protein component in high density lipoprotein (HDL) and together they transport cholesterol from peripheral tissues to the liver. Excessive systemic accumulation of misfolded APOA1 fibrils in tissues can result in APOA1 amyloidosis which is rare and can be wild type form or as hereditary variants [16, 62, 63]. There are about 20 variants associated with APOA1 amyloidosis found on hot spots from residues 50–93 and 170–178 [63]. APOA1 is often misdiagnosed or undiagnosed by other methods but the combined Congo red staining and laser capture microdissection with LC–MS/MS has greater sensitivity to enable the identification of the APOA1 amyloid protein.

Two kidney specimens from our cohort were subtyped as AA due to the presence of Serum amyloid A1 protein (Fig. 2A, Table S1). Both Serum Amyloid A1 (SAA1) and A2 (SAA2) isoforms are synthesized in the liver mostly bound to apolipoproteins such as APOA1 or APOE that bind to high density lipoproteins (HDL). SAA is produced in response to pro-inflammatory cytokines such as IL-1, IL-6 and tumor necrosis factor-alpha (TNF-α) [64]. Individuals under prolonged inflammatory conditions due to infectious diseases, hereditary auto-inflammatory diseases, rheumatoid arthritis, granulomatous disease, familial Mediterranean fever malignancies and HIV, have significantly elevated levels of Serum Amyloid A proteins which progress to extracellular deposition of fibrils that cause renal amyloidosis [65]. This is due to the undegraded form of this protein formed from incomplete proteolysis of Serum amyloid A in lysosomes and macrophages that end up bound to proteoglycans and Serum Amyloid P component protein. The kidney is the most commonly affected organ that progress to renal failure under prolonged inflammatory state and if AA is undiagnosed earlier. About 97% of renal amyloidosis is due to glomerular deposition in tubulointerstitium and/or interlobular arteries and arterioles [66].

We confirmed that mass spectrometry has multi-protein detection capability in the analysis of amyloidosis specimens, which is particularly important since amyloidosis is a heterogeneous group of disorders. Having specimens with more than one highly abundant amyloid protein is highly possible and creates ambiguity leading to equivocal results. It is therefore imperative that the test result should contain a full list of all the proteins confidently identified from every biological replicate derived from clinical specimens to help pathologists carefully diagnose the correct amyloid protein subtype(s) (Table S2).

High-field asymmetric waveform ion mobility spectrometry (FAIMS) boosts sensitivity for low sample input collected from amyloidosis specimens

The use of high-field asymmetric waveform ion mobility spectrometry (FAIMS) has proved to increase sensitivity for ultra-low input samples [67]. By using two compensation voltages at − 50 and − 70 V, we demonstrated that FAIMS significantly improved the sensitivity of our analysis. Specifically, the spectral counts (peptide-spectrum matches) for the transthyretin amyloid protein are ten times higher compared to the same biological replicates analyzed without FAIMS on the same LC–MS/MS setup (Fig. 2B).

We also explored the dynamic exclusion of the acquisition method, i.e. the time interval in DDA acquisition to “ignore” a given precursor to prevent its continuous re-selection. Due to the relatively small number of precursor ions present in our sample(s), we demonstrated that the selection of transthyretin amyloid peptides for fragmentation increased in frequency when the dynamic exclusion was disabled (Fig. 2C). The higher number of spectra for transthyretin peptides improves amyloid protein subtyping, which is the major goal of this pipeline (Fig. 2C). However, this did not result in an increased protein identification, regardless of the use of FAIMS when dynamic exclusion is disabled (Fig. 2D and E). On the other hand, with FAIMS and dynamic exclusion enabled, the number of protein identifications doubled (Fig. 2E) but it drastically lowered the total spectral counts for the transthyretin peptides (Fig. 2C). The increase in the number of protein identification only becomes useful for increasing the depth of the patients’ proteome profile that could be valuable for determining personalized treatment options or a cross comparison of patients’ proteome profiles for biomarker discovery that can be translated as diagnostic markers or potential drug targets other than the amyloid proteins.

Label free quantitation based on spectral counting versus MS1 precursor intensities

Even though proteomics applied to basic science is mostly performing quantification via extracted ion chromatography, amyloid protein identification based on spectral counts has remained the recommended format for reporting test results for ease in clinical interpretation of results by clinicians. Spectral counting is today considered a semi-quantitative approach [68], which is highly robust but with limited linear dynamic range for protein quantification compared to extracted chromatography. We compared these two label free quantification methods using a two-group comparison between a Congo red negative controls such as heart biopsy specimen (CRN-H-2) versus Congo red positive heart biopsy specimen (ATTR-H-4) (Fig. 3A–F). The total spectral counts based on the MS2 level which is the total number of MS/MS or peptide-spectral matches (Fig. 3A) versus MS1 based peak area (or extracted ion chromatogram–XIC) of the precursor ion for peptide, GSPAINVAVHVFR (m/z 683.88, + 2), are shown in Fig. 3A and B. The two tables of raw data from all the proteins identified were both log2 transformed and normalized resulting in two volcano plots that looked similar, with transthyretin (TTR) amyloid protein as the most significantly enriched in the Congo red positive heart biopsy specimens (Fig. 3C and D). Both the total spectral counts and protein raw abundance for transthyretin also showed similar patterns (Fig. 3E and F). Based on our results, the two label free quantitation methods are comparable which may be due to the enrichment of amyloid protein fibrils collected by laser capture microdissection resulting in boosting the abundance of the amyloid protein compared to other proteins in the sample.

Fig. 3

Two label free quantitation methods based on A Spectral counts based on peptide-spectral matches at the MS2 level versus B Extracted ion chromatogram (XIC) of peptide GSPAINVAVHVFR (m/z 683.88, + 2) at the MS1 level. C, D Volcano plots of Congo red negative heart biopsy (CRN-H-2) vs Congo red positive heart biopsy (ATTR-H-4) based on C total spectral counts versus D normalized abundances of all the proteins identified. E. Transthyretin total spectral counts versus F raw abundances of transthyretin from Congo red negative heart biopsy (CRN-H-2) versus Congo red positive heart biopsy (ATTR-H-4)

Data dependent acquisition (DDA) versus data independent acquisition (DIA) modes for amyloid protein identification

Unlike the DDA method, the data independent acquisition (DIA) method involves dividing the precursor mass range into isolation windows relatively larger than with DDA for MS/MS fragmentation, resulting in mixed MS/MS spectra but with the advantage of selecting all the signals at every duty cycle of the mass spectrometer. Recently, DIA has become more popular for proteomics due to this aspect of not leaving any ion unfragmented, including for the discovery of biomarker candidates from clinical samples derived from plasma [69], cerebrospinal fluid [70], urine [71, 72], tissues [73], cell lysates [74] and in single cell proteomics analysis with low input sample amounts [75]. Here, we explored the applicability of DIA for amyloid protein identification. Three Congo red negative (CR−) samples (CRN-H1-1 to CRN-H1-3) and three Congo red positive (CR+) for transthyretin samples (ATTR-H8-1 to ATTR-H8-3) were run back-to-back using DDA and DIA methods. DIA from a 15-min gradient remarkably replicated the results obtained from DDA from a 60-min gradient when the Congo red negative (CR−) samples (CRN-H1-1 to CRN-H1-3) and three Congo red positive (CR+) for transthyretin samples (ATTR-H8-1 to ATTR-H8-3) samples were compared (Fig. 4A and B). Transthyretin (TTR) was identified and quantified to be ≥ 6 orders of magnitude higher in the Congo red positive samples vs the Congo red negative ones using both DDA and DIA (Fig. 4C and D). Despite DDA having a higher number of total proteins identified compared to DIA, there is higher variability in protein abundances across samples due to more missing values resulting in higher % coefficient of variation (% CV) in DDA compared to DIA (Fig. 4E). The lower % CV in DIA is due to lower stochasticity in signal extraction resulting in an improved reproducibility of the peptide/protein quantification. With the DIA method we assessed 51 significantly regulated proteins (t-test P-value < 0.05) compared to only 31 assessed by DDA. Twenty (20) of these proteins were found to be significant by both methods (Fig. 4F). The low overlap might mislead to the interpretation that the two methods provide different results; in reality, these differences are due to small changes in the p-value, demonstrated by the very similar volcano plots generated using DDA and DIA data (Fig. 4G and H).

Fig. 4

A Mean total number of proteins identified from Congo red positive CR (+) and negative CR (−) samples analyzed by DDA versus DIA. B Total number of proteins identified from each replicate from each group (shown in Panel A). C Transthyretin protein raw abundances from CR (+) and CR (−) samples by DDA versus DIA. D Normalized abundances of the transthyretin protein from CR (+) and CR (−) specimens by DDA versus DIA. E Violin plot of % coefficient of variation from DDA versus DIA analysis of CR (+) versus CR (−) samples. F Venn diagram of significant proteins identified from DDA and DIA analysis of CR (+) and CR (−) samples. Volcano plots of proteins identified from CR(+) samples versus CR (−) samples analyzed by: G DDA with transthyretin protein as one of the most significantly enriched (p-value = 7.62) among the 31 significant proteins in the CR (+) samples (with p-values > 4.32) versus H DIA with transthyretin protein as one of the most significantly enriched (p-value = 8.33) among the 51 significant proteins identified in the CR (+) samples (with p-values > 4.32)

Sequence variants and post-translational modifications (PTMS) analysis by mass spectrometry

The hereditary ATTR variants have over 90 sequence variants [10], some of which are not present in canonical protein sequence databases such as SwissProt. We therefore performed MS/MS spectra identification analysis using Mascot Error Tolerant Search feature from Matrix Science and the De Novo Sequencing from PEAKS to identify sequence variations and new sequences from a customized Amyloidosis protein database (http://amyloidosismutations.com/). For this, we used DDA runs to search only MS/MS spectra from narrow precursor ion isolations. We detected the transthyretin variant V122I/L (or V142Xle) (Figure S3A). The peptide YTIAALLSPYSYSTTAVVTNPK containing the V122I/L variant was identified using Mascot Error Tolerant Search from 6 out of 12 ATTR positive specimens. Figure S3A shows that a possible assignment of Val- > Xle (v) with a mass shift of [+ 14.0156] was identified more than once. Mascot site analysis also localized the Val- > Xle to the second valine residue in the peptide at 94.40% versus the first valine residue at 5.60% probabilities (Figure S3B). The MS/MS assignment also confirmed the shift in mass by + 14.0156 on the second valine residue (Figure S3C). In addition, the non-variant or wild type form of the peptide YTIAALLSPYSYSTTAVVTNPK was also identified indicative of heterozygous genotype. These results can also be uploaded and re-processed in Scaffold for validation (Fig. S3D). Altogether, the Mascot Error Tolerance search detected not only the wild type transthyretin but also sequence variant form.

Sequence variants analysis can also be performed using PEAKS, a proprietary software with a de novo sequencing feature (https://www.bioinfor.com/peaks-studio/). This means that PEAKS can determine the amino acid sequence from an MS/MS mass spectrum without using a sequence database. PEAKS was unable to identify V122I/L by de novo sequencing, PTM or spider homology search tool, showing complementarity between the two software tools we used. Possibly, PEAKS missed this identification because it converts all isoleucine to leucine in all protein sequences, and so it is possible that PEAKS does not recognize Unimod’s symbol for those containing Isoleucine/Leucine modification as this is denoted by e.g. V- > Xle modification. In fact, PEAKS was able to identify the V122I/L variant peptide when the V- > Leu was specified in the list of variable modifications (Figure S4A). The PTM and spider search tool of PEAKS [76, 77] also assigned the + 14.02 Da mass shift due to methylation on lysine residues (Figure S4B and C). This is due to modification introduced during the formalin fixation of the tissues [23]. This can introduce ambiguity on the peptide RYTIAALLSPYSTTAVVTNPKE for V122I/L or methylation on K. Figure S4D shows an annotated MS/MS spectrum for the peptide YTIAALLSPYSTTAVVTNPKE when the methylation localizes on lysine K(+ 14.02). This match was generated by the PTM search module of PEAKS. However, the C-terminus of the peptide, i.e. VTNKE, does not have sufficient fragments to unambiguously differentiate between V- > 122I/L and a methylated K. Therefore, caution is advised in interpreting such PTM assignment. We utilized the results of this analysis to create our customized Amyloid protein database that includes the known sequence variants (Table S4). This customized Amyloid protein database can also be edited to add potentially new sequence variants.

Apart from sequence variants and the lysine methylation discussed, mass spectrometry is capable of an unbiased detection of a range of post-translational modifications (PTMs) that can be found on the amyloid protein fibrils. It can tell the type and the site of PTMs based on observed mass shifts and fragmentation patterns. By varying and controlling conditions for sample preparations for LC–MS/MS, MALDI-TOF or MALDI with Imaging, PTMs such as nitration, isomerization, racemization, phosphorylation were identified from beta-amyloid proteoforms in Alzheimer’s disease [78] while oxidation, pyroglutamylation, glycosylation were found in Alzheimer’s and AL amyloidosis [78,79,80,81]. Conserved disulphide bonds are common in lambda light chain AL fibrils [79, 82] while S-sulfonation (S-Sulfo), S-glycinylcysteinylation (S-CysGly), S-cysteinylation (S-Cys) and S-glutathionylation (S-GSH) on Cys-10 are found in transthyretin ATTR fibrils [83]. We also found some N-glycosylation by performing the error tolerant search on Mascot but this modification was not consistently observed in all samples. The presence of mutations in the encoding gene and PTMs may play a role in inducing conformational changes that favour the formation of the amyloid fibrils that are resistant to degradation [78, 79].

However, it is important to note that mass spectrometry can also detect artifact modification introduced during formalin fixation of tissues [78]. In our case, we used Mascot Error Tolerant for unknown PTMs search. In addition to the sequence variants and glycosylation, methylation and formylation on lysine residues were the modifications were commonly observed due to formalin fixation while carbamidomethylation on cysteine and deamidation on asparagine due to the reduction and alkylation steps performed under basic pH condition. Perhaps we only had 31 specimens analysed and some PTMs as described previously by others are patient specific where it was observed in some patients [79, 83] but not in others [79, 82].

Discovery proteomics: applications beyond the amyloid protein identification pipeline

We aim to show in this section some of the tools that can be used for gene ontology analysis that can provide information on biological processes and pathways that pertain to the amyloid protein as well as the other proteins identified around its environment. The analysis of all the proteins detected in the proteomics acquisition, and not just the predicted biomarkers, revealed that extracellular matrix organization, wound healing and fibrinolysis are some of the top biological processes enriched in the amyloidosis proteome (Fig. 5A). These results were obtained by comparing the heart specimen ATTR-H-8 with transthyretin (TTR) amyloidosis versus negative control CRN-H-1. This was expected, being amyloidosis a disorder that involves deposition of amyloid fibrils in the extracellular matrix. A complete list of the identified proteome can be found in Table S3. Transthyretin in its stable form has a tetrameric structure that transports hormones in plasma and has neuroprotective and oxidative-stress-suppressing factor functions. Destabilization of TTR due to aging factors, mutation and changes in post-translational modifications, proteostasis and Ca+2 result in cytotoxic transthyretin amyloid fibrils observed and linked in the pathogenesis of senile and hereditary transthyretin amyloidosis (ATTR) [84, 85]. Figure 5B is a Venn diagram that shows the number of proteins that are common among the different amyloid protein sub-types. For transthyretin amyloidosis, angiogenin (ANG) was found to be a protein common in four ATTR specimens from heart and stomach (ATTR-H-8, ATTR-S-11, ATTR-H-9 and ATTR-H-12).

Fig. 5

A A tree map of biological processes enriched in the Congo red positive specimens B Venn diagram of common proteins from heart (ATTR-H-12); heart (ATTR-H-8); stomach (ATTR-S-11); heart (ATTR-H-9); C Protein interaction network by STRING and Cytoscape; D Gene ontology of the top biological processes

The list of proteins found in the immune system process was re-analyzed by STRING with k-means clustering and found transthyretin and angiogenin to have genetic interaction with other known universal protein markers such as serum amyloid P component, Apolipoprotein A-IV, Clusterin, Complement C9, etc. (Fig. 5C and D) which are pro-inflammatory markers. Aggregated forms of transthyretin that form the amyloid fibrils are linked with immune system response that promotes matrix remodeling [85, 86]. On the other hand, stable transthyretin is a negative marker for inflammation and angiogenin may have anti-inflammatory activity previously reported to have an increase in protein concentrations in serum during the inflammatory response [87]. Angiogenin promotes homeostasis with its dual function of regulating cell growth when it is in the nucleolus and gets translocated to the cytoplasm in response to stress for damage repair and cell survival. However, it had been reported that attenuation of angiogenin anti-stress activity by aggregate deposition of amyloidogenic variant of APOA1 can induce cell death [88]. Overexpression of mutant APOA1 in cells decreased angiogenin expression and altered the cellular location of angiogenin. Under stress conditions, it did not translocate in the cytoplasm but remained in the nucleus promoting rRNA transcription resulting in ribotoxic effects and abrogated its pro-survival function. It is interesting to note that exogenous addition of angiogenin rescued L75P-ApoA-I-induced apoptosis. Perhaps, a similar situation may be happening when transthyretin amyloid protein forms extracellular protein fibrils. Discovering angiogenin as a potential key player in the pathogenesis of amyloidosis, through gene ontology analysis, could be interesting to explore. However, this requires further investigation and is beyond the scope of this paper.

Amyloidosis subtyping by matrix assisted laser desorption-mass spectrometry imaging (MALDI-MSI)

MALDI-Mass Spectrometry Imaging (MALDI-MSI) is another promising tool for pathologists, as it enables in situ peptide spatial mapping skipping laser capture microdissection. We subjected to MALDI-MSI, 9 tissue sections from different organs with AL (kappa or lambda) and ATTR amyloidosis subtypes using an Ultraflextreme MALDI-TOF/TOF mass spectrometer (Fig. 6A). The purpose was to show how MALDI-MSI can complement the current technique of laser capture microdissection and tandem mass spectrometry (LMD/LC–MS/MS) for amyloid protein identification. The transthyretin peptide GSPAINVAVHVFR (m/z 1366.7) was the most consistently detected from samples analyzed by LMD/LC–MS/MS and was also detected in MALDI-MSI (Fig. 6A). Because this is an on-tissue trypsin digestion, there is also the probability of missed cleavage and the peptide GSPAINVAVHVFRK (m/z 1494.8) with one missed cleavage was also mapped. Similarly, we could also map the Serum amyloid P component (SAP), a universal protein marker that is associated with the presence of amyloid proteins using the peptides VGEYSLYIGR (m/z 1156.6) and ERVGEYSLYIGR (m/z 1441.9) with 0 and 1 missed cleavage, respectively.

Fig. 6

A MALDI-MSI analysis of transthyretin peptide (ATTR) GSPAINVAVHVFR (m/z 1366.7) on FFPE heart tissue section from autopsy and biopsy samples using the Ultraflextreme mass spectrometer. Presence of serum amyloid P component peptides VGEYSLYIGR (m/z 1156.6) and ERVGEYSLYIGR (m/z 1441.9) were also mapped on the same tissue sections. B Mass mobility resolved interfering ion observed from the Congo red positive peptide extract spiked with the peptide digests of the transthyretin standard; C three overlapping peptides observed from complex FFPE heart tissue matrix resolved by TIMS; D peptide ion heat maps from transthyretin, serum amyloid P component and Apolipoprotein E; E segmentation analysis

Next, we repeated the analysis with a more recent MALDI-MSI model, the Bruker timsTOF flex. Digested peptides from in-solution trypsin digestion of pure transthyretin standard were used as a positive control and analyzed by MALDI-TOF using the timsTOF flex (Figure S5C). Peptide extracts obtained from on-tissue trypsin digestion of Congo red positive (CR+) and Congo red negative (CR−) FFPE heart biopsy (Figure S5A and S5B) and autopsy specimens (Figure S5D and S5E). Pure transthyretin peptide digests was analysed (Figure S5C) and were spiked in the peptide extracts to test the capability to differentiate isobaric peptides. The timsTOF flex (with mass resolution of 60,000 at m/z 400) was able to detect the well resolved transthyretin peptide GSPAINVAVHVFR (m/z 1366.7) (Figure S6A, S6C and S6D) only from the Congo red positive biopsy, the pure transthyretin standard and autopsy heart specimens while it is absent in the Congo red negative heart specimens (Figure S6B and S6E).

Trapped ion mobility spectrometry (TIMS) separation adds another dimension of identification when combined with MALDI-TOF. Mass and mobility resolved the transthyretin peptide m/z 1267.7 (SYSTTAVVTNPK) from an interfering ion from the Congo red positive (ATTR-H-4) peptide extract spiked with the peptide digests of the transthyretin standard (Fig. 6B) without the tissue matrix.

TIMS is a useful way to screen for candidate peptide markers as overlapping or isobaric peptides are further resolved in the gas phase based on their collision cross sections (Table S5). In the presence of tissue matrix, the acquired MALDI-TOF spectra becomes more complex and there are now three overlapping peptides around m/z 1267.7 resolved by TIMS. Ion maps revealed that the three ions were observed in both Congo red positive and negative heart tissue specimens. This suggests that the peptide ion with m/z 1267.7 is not an ideal marker peptide while the peptide m/z 1366.7 is a potential peptide marker for transthyretin.

Peptide ion heat maps for transthyretin (TTR) peptides (m/z 1366.750) and (m/z 1406.667) were observed to be highly abundant in the Congo red positive (CR+) heart tissue biopsy tissue but not in the Congo red negative (CR−) tissue (Fig. 6B). In addition, peptides from universal markers, serum amyloid P component (SAP) (m/z 1156.590) and Apolipoprotein E (APOE) (m/z 968.542) co-localize with amyloid proteins in the CR+ tissue (Fig. 6C). Segmentation analysis (Fig. 6D and 6E) clustered statistically similar regions and the extracted average mass are used to distinguish between CR- and CR+ tissue biopsies which is further illustrated by the principal component analysis (Fig. S7) showing how different the CR+ from the CR− biopsy samples are from each other. Altogether, we demonstrated that MALDI-MSI is a suitable alternative to LMD/LC–MS/MS for quantifying the main biomarkers in Amyloidosis.

Cost analysis for amyloidosis sub-typing by LMD/LC–MS/MS versus MALDI-MSI

Tables 1, S8 and S9 provide a comparison and breakdown of the cost per sample for the two techniques. The materials used in LMD/LC–MS/MS costs 2.3 times more than MALDI-MSI. Each method required a special type of slide such as PEN membrane slide for LMD/LC–MS/MS versus ITO slide for MALDI-MSI. Both types of slides are commercially available and can be easily purchased. While both methods have deparaffinization steps, the extra step of Congo red staining is an added cost to the materials and sample preparation time for the LMD/LC–MS/MS method. Both methods required the heat induced epitope retrieval for the de-crosslinking step. However, the LMD/LC–MS/MS has an extra step of sonication for efficient protein extraction. While it is a time-saver, it could be an added equipment and maintenance cost (~ $100K) and by contrast, the MALDI-MSI does not require the additional sonication step. The laser capture microdissection step is an additional time consuming step as this can take up to 1 h or more for collecting Congo red stained amyloid fibrils. The reduction, alkylation and overnight trypsin digestion times take about 16–17 h for the LMD/LC–MS/MS. On the other hand, there are no reduction/alkylation steps in MALDI-MSI but instead requires on-tissue trypsin spraying and overnight incubation also for 16–17 h. Prior to mass spectrometry analysis, the LMD/LC–MS/MS samples require the desalting step while MALDI-MSI requires matrix spraying. Our standard acquisition times are 75 min for LC–MS/MS and 90 min for MALDI-MSI. However, the acquisition times for LC–MS/MS can be shortened to 15 min gradient if DIA approach can be implemented