27-plex Tandem Mass Tag Mass Spectrometry for Profiling Brain Proteome in Alzheimer's Disease

Abstract

Multiplexed isobaric labeling methods, such as Tandem Mass Tags (TMT), remarkably improve the throughput of quantitative mass spectrometry. Here we present a 27-plex TMT method coupled with two-dimensional liquid chromatography (LC/LC) for extensive peptide fractionation and high-resolution tandem mass spectrometry (MS/MS) for peptide quantification, and then apply the method to profile complex human brain proteome of Alzheimer’s disease (AD). The 27-plex method combines multiplexed capacities of the 11-plex and the 16-plex TMT, as the peptides labeled by the two TMT sets display different mass and hydrophobicity, which can be well separated in LC-MS/MS. We first systematically optimized the protocol for the newly developed 16-plex TMT, including labeling reaction, desalting, and MS conditions, and then directly compared the 11-plex and 16-plex methods by analyzing the same human AD samples. Both methods yielded similar proteome coverage, analyzing > 100,000 peptides in > 10,000 human proteins. Furthermore, the 11-plex and 16-plex samples were mixed for a 27-plex assay, resulting in more than 8,000 protein measurements within the same MS time. The 27-plex results are highly consistent with those of the individual 11-plex and 16-plex TMT analyses. We also used these proteomics datasets to compare AD brain with the non-dementia controls, discovering major AD-related proteins and revealing numerous novel protein alterations enriched in the pathways of amyloidosis, immunity, mitochondrial and synaptic functions. Overall, our data strongly demonstrate that this new 27-plex strategy is highly feasible for routine large-scale proteomic analysis.

Introduction

In the postgenomic era, along with the revolution in DNA sequencing, substantial advances have been introduced in proteomics^{1, 2}, which enables the measurement of the vast majority of whole proteome and numerous posttranslational modifications, such as phosphorylation^3-5 and ubiquitination^6-8. With the improvements in throughput, sensitivity, and robustness, proteomics has been emerging as a powerful tool in basic and clinical research^9-11. Moreover, proteomics results are often integrated with other large-scale datasets of genomics, transcriptomics, and metabolomics to define signaling networks and master regulators by systems biology approaches^{12, 13}. For example, many groups applied the multi-omics approaches in several common diseases, such as cancer, Alzheimer’s disease (AD)^{14, 15}, revealing crucial disease-related proteins and pathways in these disorders. These studies offer unpreceded opportunities to elucidate fundamental molecular mechanisms of pathogenesis and discover promising therapeutic strategies.

Mass spectrometry (MS) has become the mainstream approach for proteomics analysis, including label-free quantification and stable isotope labeling methods^{1, 16}. The label-free quantification can be accomplished by the methods of data dependent acquisition (DDA) and data-independent acquisition (DIA)/SWATH-MS^{17, 18}. The quantification in the DDA methods relies on spectral counts¹⁹ or MS1 signals²⁰, while the measurement in the DIA method is derived from MS2 signals. Stable isotope labeling can be categorized into MS1-based quantification such as stable SILAC²¹ and dimethyl labeling²², and MS2/3-based quantification such as iTRAQ²³, TMT²⁴, and DiLeu²⁵. Both MS1- or MS2/3-based stable isotope labeling methods allow differentially labeled peptides to be mixed and extensively fractionated to achieve ultra-deep proteome coverage²⁶. In the MS1-based quantification, the complexity of MS1 spectra is increased, resulting in the reduction of peptide identification efficiency. To overcome this limitation, the MS2/3-based quantification utilizes a strategy to generate isobaric precursor ions and enables multiplexed tagging. After the fragmentation of the isobaric precursor ions, disparate reporter ions are released for relative quantification. However, the MS2-based method often suffers from ratio compression (i.e. underestimation of quantitative ratios) caused by co-eluted interfering ions. Nonetheless, this issue has been addressed by numerous developed strategies, such as extensive peptide fractionation, MS3 data acquisition, and computational correction^26-29.

More recently, TMT has been commonly used in large-scale proteomics studies due to its high multiplexing capacity and deep proteome coverage^{30, 31}. The TMT reagents consist of a mass reporter, a mass balance group and an amine reactive group, in which the mass reporter and the mass balance group are differentially labeled by stable isotopes, but their summed mass is kept constant for isobaric MS1 ions. In the past decade, the number of TMT channels has increased from 2 to 6, 11 and recently to 16^{24, 32-34}. Compared with TMT11, TMTpro (termed TMT16 hereafter) contains a longer balance group and a proline-based mass reporter, enabling the combination of nine stable C13 and N15 isotopes to achieve 16 reporters. TMT16 has been reported to display similar performance as TMT11 in recent analyses^{34, 35}.

To further enhance the multiplexing capacity, several strategies that combine MS1-based and MS2/3-based quantification methods were reported. For instance, the combination of triplex SILAC labeling with 6-plex TMT permits the quantification of 18 samples in one batch³⁶, but the strategy is not widely adopted due to the restriction of SILAC labeling. Similarly, duplex dimethyl labeling (e.g. heavy and light reagents) can be combined with 12-plex isobaric DiLeu tags to allow 24-plex quantification³⁷, but this method requires stringent conditions to firstly label the peptide N-termini by the dimethyl group, and then label the Lys residue by DiLeu tag.

Here we design a simple and versatile 27-plex method by merging two sets of isobaric reagents with different masses (i.e., TMT11 and TMT16), and have compared the performance of the 11-, 16- and 27-plex methods using human brain tissue specimens of Alzheimer’s disease. The 27-plex method is capable of quantifying more than 8,000 proteins with highly similar results to the individual 11-plex and 16-plex datasets, demonstrating that the 27-plex TMT method is of general application to large-scale proteomics studies.

Methods

Sample preparation and protein digestion

Human postmortem brain tissue samples (frontal cortex) were provided by the Brain and Body Donation Program at Banner Sun Health Research Institute. A total of 15 cases (7 controls and 8 AD cases, ~10 mg tissue/case) were used in this study (Supporting Table S1). A wide-type mouse (C57BL/6J, The Jaxson Laboratory) was used in this study. Mice were bred and maintained in the Animal Resource Center at St. Jude Children’s Research Hospital. All animal protocols were approved by the Institutional Animal Care and Use Committee. The mouse brain tissue was dissected rapidly, frozen in liquid nitrogen, and stored at −80°C. Protein extraction, quantification and digestion were performed as described in Supporting Information.

Optimization of TMT labeling with the mouse brain sample

The mouse brain peptide sample was resuspended in 50 mM HEPES, pH 8.5 (~1 μg/μL). TMT reagent (TMTpro Zero, Thermo Fisher Scientific) was prepared by dissolving in anhydrous acetonitrile (AcN). In the reaction, different TMT/protein ratio (w/w) were carried out with a fixed labeling time of 60 min at 21°C. Time course labeling was then performed from 0 to 60 min, with the selected TMT/protein ratio of 1.5:1 by weight. The reactions were quenched with 0.5% hydroxylamine for 15 min.

TMT labeling, pooling of 11-plex, 16-plex and 27-plex samples

Human peptide samples were resuspended in 50 mM HEPES, pH 8.5 (~1 μg/μL), and divided into two aliquots. One aliquot was labeled with TMTpro reagents (termed TMT16) and the other aliquot was labeled with TMT11 following the procedures described in Supporting Information. After complete labeling, equally pool the TMT11 and TMT16 samples respectively to obtain TMT11 pool and TMT16 pool. Part of the TMT11 pool and the TMT16 pool were mixed for the 27-plex analysis. The detailed procedures were described in Supporting Information.

Optimization of desalting conditions for the TMT16 peptides

The mouse brain peptide sample was labeled with TMT16zero for optimization of desalting conditions. The optimization was performed following the procedures described in Supporting Information. Based on the MS results, the condition of 10 bed volumes of 0.1% TFA plus 5% AcN was selected for desalting of TMT samples.

LC/LC-MS/MS analysis

The pooled peptide samples of 11-plex, 16-plex and 27-plex were fractionated into 40 concatenated fractions by an offline basic pH RPLC (Supporting Information). Each fraction was analyzed with acidic pH reverse-phase LC-MS/MS with a self-packed column (75 μm × 15 cm with 1.9 μm C18 resin from Dr. Maisch GmbH, heated at 65°C to reduce back pressure) coupled with a Q Exactive HF Orbitrap MS (Thermo Fisher Scientific). Peptides were eluted in a 60 min gradient (buffer A: 0.2% formic acid, 5% DMSO; buffer B: buffer A plus 65% AcN) in a 90 min run. MS settings included MS1 scans (60,000 resolution, 410-1600 m/z scan range, 1 x 10⁶ AGC, and 50 ms maximal ion time) and 20 data-dependent MS2 scans (60,000 resolution, starting from 120 m/z, 1 x 10⁵ AGC, 120 ms maximal ion time, 1.0 m/z isolation window with 0.2 m/z offset, HCD, specified normalized collision energy (NCE), and 15 s dynamic exclusion). The protein identification and quantification were performed by the JUMP search engine (Supporting Information). During the quantification, the identified PSMs with missing values in any TMT channels were removed. The remaining PSMs were used for peptide and protein quantification. This filtering step resulted in the loss of < 2% of the identified proteins.

Results and Discussion

Experimental design of the 27-plex TMT method

We designed a 27-plex TMT method (referred as TMT27 hereafter) that combines multiplexed capacities of TMT11 and TMT16. The principle of this method is that the TMT11- and TMT16-labeled peptides have different masses and can be separated in MS1 scans, which allows simultaneous quantification of two sets of labels in one MS experiment (Fig. 1A).

Figure 1.

Experimental design for evaluating TMT27 method. (A) Schematic comparison of TMT11 and TMT16 reagents. The reagent structures, masses, mass shifts after labeling, and the masses of the reporter ions of TMT11 and TMT16 are shown. (B) Experimental design and procedures for comparing the TMT11, TMT16 and TMT27 by profiling human AD proteome. Sixteen samples including AD cases (n = 8), control cases (n = 7), and internal reference (n = 1) were used for the proteomics analysis. Proteins from different cases of human brain tissue samples were digested into peptides and divided into two aliquots. One aliquot was labeled with TMT16 and the other aliquot of 11 samples (5 AD, 5 controls, and the internal reference) was labeled with TMT11. After labeling, the individual channel was equally pooled to obtain TMT11 pool and TMT16 pool. Each pool was further divided into two aliquots, one aliquot was used for individual TMT11 or TMT16 analysis and the other aliquot of each labeling was mixed for TMT27 assay. Three sets of pooled mixtures (TMT11, TMT16 and TMT27) were desalted and subjected to LC/LC-MS/MS analysis.

To evaluate the performance of the TMT27 method, we determined to compare the TMT11, TMT16 and TMT27 using the same sets of post-mortem brain tissues (Fig. 1B). In total, sixteen samples including Alzheimer’s disease (AD) cases (n = 8), control cases (n = 7), and internal reference (n = 1) were used (Table S1). In the TMT27 assay, the TMT16 and TMT11 results may be viewed as two different batches in one single experiment, the internal reference (i.e. a mixture of human brain samples) was implemented for batch normalization³⁸. The samples were lysed, trypsinized into peptides, and divided into two peptide aliquots. One aliquot was labeled with TMT16 reagents for the individual TMT16 experiment; the other aliquot of 11 samples (5 AD, 5 controls, and the internal reference) was labeled with TMT11 reagents for the individual TMT11 experiment. Furthermore, a portion of the pooled TMT11 and TMT16 peptides were mixed to generate the starting sample for the TMT27 experiment. The three sets of pooled peptides (TMT11, TMT16 and TMT27) were desalted and subjected to LC/LC-MS/MS analyses, and the results were compared with respect to quantified peptides/proteins as well as detected protein changes between AD and control cases.

According to our previous studies, extensive pre-fractionation of peptides not only reduces the complexity in each run to deepen the proteome coverage^{39, 40}, but also alleviates ratio compression caused by co-eluted interfering ions²⁶. In this experiment, we applied extensive fractionation of collecting 40 concatenated fractions during the basic pH reverse phase liquid chromatography, followed by 90-min runs (60-min gradient time) during the acidic pH reverse phase LC-MS/MS. To validate our selection, we compared different basic pH LC strategies, from no fractionation (n = 1) to 3, 10 and 40 fractions, and found that the proteome coverage was dramatically increased with more fractions (Supporting Information, Fig. S1), further demonstrating the effective role of pre-fractionation to improve the proteome coverage.

Optimization of TMT16 protocol for proteomic analysis

Before we performed the TMT27 experiment to profile AD brain proteome, we systematically optimized the TMT16 protocol, as TMT16 uses a newly developed set of TMT reagents, which have a distinct structure from the TMT11 reagents. First, we examined the TMT/protein (w/w) ratios by titrating the TMT16zero reagent to label mouse brain proteins. The protein concentration was estimated by a Coomassie-stained short SDS gel with BSA as the standard⁴¹ (Supporting Information, Fig. S2A). After trypsin digestion, we further quantified the peptide concentration by UV absorbance at 205 nm using a synthetic peptide as the standard⁴² (Supporting Information, Fig. S2B). Then seven different TMT/protein (i.e. digested protein) ratios from 0:1 to 3:1 were tested to label peptides in 60 min, followed by quenching, desalting and LC-MS/MS analysis. We monitored the decrease of unlabeled peptides in the LC-MS/MS runs, and used their MS1 intensities to calculate the percentage of labeled peptides, for instance, the ~100% of labeled peptides was evidenced by complete elimination of unlabeled peptides. In the analysis, the labeling percentage reached 97% at the ratio of 1:1 and arrived at the 100% plateau at the ratio of 1.5:1 (Fig. 2A). Thus, TMT/protein ratio of 1.5:1 was used for the following experiments, which was 50% higher than the ratio suggested for TMT11 labeling⁴³. This small discrepancy may be due to the mass difference of TMT16 and TMT11 reagents, as the mass of TMT16 reagents is 1.2-fold of TMT11 reagents. Next, we optimized the TMT16 labeling time in a time course experiment (Fig. 2B). The percentage of labeled peptides increased rapidly to 99% within 15 min and reached a platform of ~100% at 30 min. So we suggest the peptide sample is labeled by TMT16 with a TMT/protein ratio of 1.5:1 in 30 min. Considering the variations of protein concentration measurements, we highly recommend to validate the completeness of TMT reaction in every experiment⁴⁴.

Figure 2.

Optimization of labeling conditions for TMT16. (A) Titration of TMT16 reagents. The percentages of identified TMT16-labeled peptides in all identified peptides are shown with different TMT/protein (w/w) ratios. The reaction time was fixed at 1 h. (B) Time course labeling of TMT16. TMT/protein ratio was fixed at 1.5:1 by weight. All labeling experiments were replicated (n = 2) with the standard deviation of the mean shown.

During TMT16 labeling, we observed two major byproducts: TMT16-NHOH from hydroxylamine quenching reaction, and TMT16-OH from TMT hydroxylation. The two byproducts were readily detected as the dominant singly charged ions in LC-MS/MS before desalting (Fig. 3A). So it is important to remove these byproducts by the desalting process. Given that TMT16 reagents have the longest hydrophobic chains among all TMT reagents, which may result in strong binding of the byproducts to the solid-phase extraction (SPE) column, we attempted to increase the wash stringency by adding a low concentration of AcN in the buffer and increase the wash volume during desalting. We tested the combinations of different wash buffers (0.1% TFA plus 0%, 2.5% or 5% AcN) with distinct wash volumes (5 bed or 10 bed volumes), and recorded the ion intensities of [TMT16-OH+H]⁺ and [TMT16-NHOH+H]⁺ in LC-MS/MS runs (Fig. 3A). The base-peak chromatograms of these runs showed that the addition of AcN or the increase of the wash volume effectively reduced the byproducts, and the combination of 5% AcN with 10 bed volume wash led to the best outcome (Fig. 3A, 3B). However, it might be possible that some peptides were lost using the stringent desalting condition. We then counted the numbers of peptides and proteins identified in these LC-MS/MS runs, but found no significant difference in identification (Fig. 3C). Therefore, we suggest that the TMT16-labeled samples are desalted with 5% AcN and 10 bed volumes of wash buffer. This condition is more simplified than the reported two-stage cleanup method, which requires the combination of reverse-phase solid-phase extraction and strong cation exchange chromatography³⁴.

Figure 3.

Optimization of desalting conditions for TMT16-labeled peptides. (A) TMT16-labeled samples were desalted at different wash conditions (5 or 10 bed volumes of distinct wash buffers (0.1% TFA plus 0%, 2.5% and 5% AcN)). Desalted and non-desalted samples were analyzed by LC-MS/MS. Base-peak chromatograms of the samples with various wash conditions are shown. The dashed line indicates the base-peak of [TMT16-OH+H]⁺ and [TMT16-NHOH+H]⁺. (B) The relative ion abundances of [TMT16-OH+H]⁺ and [TMT16-NHOH+H]⁺ under distinct wash conditions are shown. (C) Comparison of the peptide and protein identifications after desalting at different conditions. All the experiments were replicated (n = 2) with the standard deviation of the mean shown.

As TMT16 has a longer hydrophobic linker than TMT11, TMT16-labeled peptides are expected to be more hydrophobic than TMT11-labeled peptides, resulting in different retention time (RT) in reverse-phase liquid chromatography (RPLC), we examined the impact of TMT16 on RT compared with TMT11. The results showed that TMT16 has a strong impact on RT to the peptides of intermediate hydrophobicity but shows a small impact on the peptides with extremely high or low hydrophobicity (Supporting Information, Fig. S3). Thus, the starting and ending buffer B concentrations in the LC gradient are not significantly affected by the selection of different TMT labeling reagents.

In addition, comparing with the TMT11 reagents, the TMT16 reagents have a different molecular structure and may require different energy for MS2 fragmentation. We then optimized the collision energy for TMT16-labeled peptides to maximize protein identification and obtain high intensity of reporter ions. Distinct normalized collision energy levels (20-40%) on the Orbitrap HF instrument were used in LC-MS/MS runs (Supporting Information, Fig. S4). To balance the protein identification and reporter ion intensity, we found the setting of 30-32.5 is optimal for TMT16 experiments, which is slightly lower than the setting of 35 in TMT11 analysis, consistent with the conclusion in recent reports^{14, 34}.

Comparisons of the performance of the TMT11, TMT16, and TMT27 methods

After optimizing the TMT16 protocol, we carried out the entire comparative experiments of TMT11, TMT16 and TMT27 to profile human AD brain proteome using the same protocol of LC/LC-MS/MS (Fig. 1B). The three desalted, pooled samples were individually fractionated by offline basic pH RPLC to collect 40 concatenated fractions. Then each fraction was analyzed by nanoscale acidic pH LC-MS/MS in a 90 min run, totaling 60 h (2.5 days) for one TMT analysis.

To evaluate the performance of the TMT11 and TMT16 methods, we analyzed the overlapped 11 samples and used the internal reference to normalize the two datasets. TMT16 performed almost as well as TMT11, both methods identified > 270,000 peptide-spectrum matches (PSMs) and > 110,000 unique peptides, corresponding to > 10,000 proteins (Fig. 4A). Although more peptides distributed in later fractions in basic pH RPLC in TMT16 than in TMT11 (Supporting Information, Fig. S5), after concatenating from 160 to 40 fractions by combining early, middle, and late RPLC fractions, the identified PSM numbers in acidic pH LC-MS/MS distributed evenly along the 40 fractions (Fig. 4B). The TMT16 method identified 15% less peptides and 6% less proteins than the TMT11 method, which is consistent with the slightly lower performance of TMT16 in the previous report^{34, 35}. Nearly 90% of proteins were identified in both methods (Fig. 4C). After plotting the protein fold change between AD and control cases, the TMT11 and TMT16 methods showed excellent correlation (R = 0.975), demonstrating the consistency and reproducibility of the isobaric labeling methods in large-scale proteomic analysis (Fig. 4D).

Figure 4.

Deep proteome comparisons of the TMT11 and TMT16 methods. (A) The histogram plotted the number of accepted PSMs, unique peptides and unique proteins in individual TMT11, TMT16 and internal TMT11 and TMT16 within the TMT27 method. (B) The distributions of PSMs in the 40 basic RPLC fractions from TMT11 and TMT16 experiments. (C) The Venn diagram showing the overlapped proteins among TMT11 and TMT16 datasets. (D) Correlation analysis of protein fold change (AD/control) between TMT11 and TMT16 datasets by scatter plot. Proteins with intensity > 5x10⁴ and PSM >1 were used for analyzation.

Next, we compared the TMT27 results to those from individual TMT11/16 experiments. Using the same MS time, the TMT27 analysis identified > 410,000 PSMs (> 210,000 TMT11-labeled PSMs and >200,000 TMT16-labeled PSMs), which is approximately 1.5-fold of the individual TMT11 or TMT16 analysis. Finally, about 80,000 peptides (> 8,000 proteins) were identified from each of the TMT11- and TMT16-labeled samples within the TMT27 dataset (Fig. 4A), totaling 98,449 peptides and 9,146 proteins. The internal (referring to within the TMT27 experiment) TMT11-labeled proteins were ~20% less than those in individual TMT11 dataset, and similar result was found for the internal TMT16-labeled proteins. This is likely due to doubled sample complexity in MS1 survey scans (i.e., the same peptides labeled by both TMT11 and TMT16 reagents). In either internal TMT11 or TMT16 dataset, 95% of the identified proteins overlapped with the individual TMT11 or TMT16 dataset, respectively (Fig. 5A, 5B). In addition, about 90% of the proteins were overlapped between internal TMT11 and TMT16 datasets (Fig. 5C). The peptides identified in all of these datasets were also compared, showing a slightly lower overlapping percentage (Supporting Information, Fig. S6). Importantly, the protein fold change between AD and control cases in the TMT27 experiment is not only well correlated with that from the individual TMT11 or TMT16 analysis, but also well correlated between internal TMT11 and TMT16 datasets (Fig. 5D, 5E, 5F), indicating the robustness and reproducibility of the TMT27 analysis.

Figure 5.

Large-scale proteome comparisons of individual TMT11 and TMT16 datasets and internal TMT11 and TMT16 datasets within the TMT27 experiment. (A) The Venn diagram indicates overlapped TMT11-labeled proteins identified in the TMT11 and TMT27 methods. (B) The Venn diagram shows overlapped TMT16-labeled proteins identified in the TMT16 and TMT27 methods. (C) The Venn diagram presents overlapped internal TMT11- and TMT16-labeled proteins in the TMT27. (D-F) Correlation of fold change (AD/control) of the overlapped proteins in (A-C), respectively. The overlapped proteins with intensity > 5x10⁴ and PSM >1 were used in the correlation analyses.

Ratio compression is a common effect in isobaric labeling quantitative methods due to the interference of co-isolated labeled ions (e.g. peptides), which may become a significant issue when combining TMT11- and TMT16-labeled peptides increases sample complexity. We applied multiple strategies to reduce this effect²⁶, including pre-MS extensive fractionation (40 basic pH LC fractions), application of narrow isolation window (1 m/z) in the MS setting, and post-MS computational correction, because the reporter ion noise in MS2 scans can be evaluated by contaminated y1 ions²⁶. Moreover, different TMT11 and TMT16 peptides might be co-eluted and co-fragmented to generate mixed MS2 spectra. Indeed, these mixed spectra were found in database search, but occupied a very small portion (<1%) of whole matched spectra, and removed during data processing. When examining the correlation curve of protein fold change between individual TMT11 dataset and the internal TMT11 dataset from TMT27, the slope was very close to 1, indicating that the ratio compression in TMT27 was not visibly higher than that in TMT11 under our experimental condition (Fig. 5D). The similar result was obtained when comparing individual TMT16 dataset and the internal TMT16 dataset from TMT27 (Fig. 5E). In summary, our data clearly show the feasibility of TMT27 to enhance multiplex capacity without exacerbating ratio compression in our protocol.

Compared with other high multiplexing techniques (e.g. SILAC combined with TMT/iTRAQ, and dimethyl labeling coupled to DiLeu/iTRAQ/TMT), The TMT27 method is much simpler as we use the similar chemical labeling tags and only need one-step labeling (Supporting Information, Fig. S7). Such a strategy can be extended to other combinations of different sets of isobaric labeling reagents with different precursors or reporter ions (Supporting Information, Fig. S8). For example, TMT11 and TMT16 can be further combined with 12-plex DiLeu to achieve a 39-plex quantification, although the combination will further increase the sample complexity. Moreover, even with the similar (isobaric) precursor ion masses of labeled peptides, TMT16 and the 8-plex iTRAQ can be combined due to distinct reporter ions, and the TMT16- and iTRAQ-labeled peptides are expected to exhibit different retention times. Some of these novel combinations to improve multiplexity may be worth further investigation.

Dysregulation of AD disease proteins in the brain by the TMT profiling

To explore the proteome alteration of human brain samples with AD pathology, we performed a differential expression (DE) analysis to define DE proteins in AD brain samples compared to control cases. We combined all four datasets (the individual TMT11/16 and the internal TMT11 and TMT16 in the TMT27 analysis). Individual protein intensities in each dataset were normalized using the internal reference, as the internal reference samples were technically repeated in all datasets (Fig. 6A). All the proteins quantified from 15 samples (7 control and 8 AD cases) were filtered with false discovery rate < 0.1 and Z-score > 2, accepting 253 DE proteins (Supporting Table S2). The DE proteins were categorized by functional annotation analysis, enriched in amyloid pathway, immunity, mitochondrial and synaptic function, consistent with our recent deep proteome dataset¹⁴ (Fig. 6B). Interestingly, ~30% of downregulated proteins are localized in synapse, indicating profound synapse dysregulation in AD, in which multiple GABA-A receptor complex proteins are reduced in the AD brain, supporting GABAergic dysfunction in AD^{45, 46}. Thus, TMT-based quantitative proteomics effectively re-identified reported alterations in AD-related pathways¹⁴. Globally, the DE protein pattern in the four datasets is highly similar, demonstrating the reproducibility of these TMT methods.

Figure 6.

Bioinformatic analysis of differential expression (DE) proteins. (A) The illustration of the process for functional annotation analysis of DE proteins. All the proteins from TMT11, TMT16 and TMT27 were first normalized by the internal reference and the redundancy proteins were removed. The normalized data were corrected for the confounding factor of cell type changes with cell-type-specific protein markers (Supporting Information). The remaining proteins with FDR < 0.1 and Z-score > 2 were accepted as DE proteins and further explored by functional annotation analysis. (B) Heatmap of DE proteins clustered in different pathways in all datasets. Each protein is represented by a colored box after Z-score conversion.

Conclusion

In this study, we designed a TMT27 method to significantly enhance the throughput for large-scale proteomic profiling, via combining two sets of distinct isobaric tags (TMT11 and TMT16 reagents) based on their different mass shifts on modified peptides. Using well-characterized human brain tissue of Alzheimer’s disease and control cases, we systematically compared the outcomes of the individual TMT11, TMT16, and the TMT27 methods. Whereas the experimental protocols and resulting proteome coverages of the TMT11 and TMT16 methods are similar, the TMT27 method achieves a 50% increase in PSM identification efficiency, enabling the identification of more than 8,000 TMT11- and TMT16-labeled proteins within the same MS instrument time. Importantly, the TMT27 dataset is also highly consistent with the individual TMT11 and TMT16 datasets, demonstrating the feasibility of this new strategy. In addition, such a strategy can be applied to other combinations of different sets of isobaric labeling reagents.

Abbreviations:

AD

Alzheimer’s disease

MS

mass spectrometry

TMT

tandem mass tags