Nanoparticle-based method identifies over 2200 proteins in a cardiovascular disease study covering known biomarkers among other differentially expressed proteins

Michael Burgess1, Hasmik Keshishian1, Michelle Dubuke2, Juan Cruz Cuevas2, Laurie Farrell3, Debby Ngo3, Karsten Krug1, DR Mani1, Robert Gerszten3, Steven A Carr1 [1] Broad Institute of MIT and Harvard, Cambridge, MA; [2] Seer Inc., Redwood City, CA; [3] Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA

Overview

Plasma is an ideal sample for clinical proteomics but analysis is confounded by the intrinsic high dynamic range of plasma proteins making confident and reproducible measurement of low abundance proteins difficult.
In a previous cardiovascular disease study, our proteomics workflow involving abundant protein depletion, iTRAQ-labeling, off-line basic reversed-phase fractionation and analysis of 30 fractions by LC-MS/MS quantified over 3500 proteins including the measurement of clinical markers such as troponins I and T, but at the cost of extended labor, cost and low throughput1,2. Specifically, to process 16 patient samples required 7-10 days of lab work followed by 120 LC-MS/MS runs.
Here we compare this strategy with the newly developed ProteographTM Product Suite (Seer) which simplifies sample processing prior to MS analysis by using nanoparticles in a robotic platform with much higher throughput.
Methods: Seer Proteograph Workflow

Figure 1: Seer Proteograph Plasma Sample Processing Scheme3. This plate-based workflow includes the processing of 16 total samples across 5 NPs along with Plasma Controls for digestion, NP- binding and SPE. Samples and reagents are loaded in 30min followed by the 6.5 hour process. Each patient analyses is composed of 5 LC-MS/MS runs or 80 total for all 16 samples.

Initial evaluation of the Proteograph plasma processing system was done using a plasma control sample provided by Seer. Method reproducibility was evaluated across 15 samples processed for 3 consecutive days.
Patient plasma samples included in this study were comprised of peripheral blood draws at 5 time points (baseline-pre-heparin, BL-post heparin, 10min, 1hr and 4hr after alcohol ablation) across 3 patients undergoing a therapeutic septal alcohol ablation heart surgical procedure2. These planned myocardial infarctions (PMI) act as a human model for longitudinal myocardial injury.
Following the standard protocol, 200uLs of each sample is arrayed across 5 wells prior to incubation with Seer’s panel of 5 nanoparticles (NP) (Step 1). After extensive washing to remove nonspecific binders (Step 2), samples undergo an on-bead, 3 hour Trypsin/Lys-C digestion (Step 3) using an adapted EasyPep 96 MS Sample Prep Kit (ThermoFisher).
Each sample was then desalted using an on-board positive pressure manifold for sample cleanup.
Total protein recovered post process ranged from few hundred nanograms to 5 microgram per NP.
Data were acquired on a Thermo Exploris LC-MS/MS system using a 1 hour gradient. Each sample analysis is comprised of 5 LC-MS/MS runs, one for each NP subpopulation. Data were searched with Spectrum Mill software using the label free quantification (LFQ) mode for differential analysis of proteins identified with 2 or more peptides.
Initial Evaluation
Enriching plasma with the NP panel increases depth of protein coverage by 5X relative to direct analysis; proteins span 8 orders of magnitude

Table 1: Proteins identified using each of the 5 individual nanoparticles versus direct analysis of unprocessedPlasmaControlsample. Each individual NP showed a 2 to 3-fold increase in the number of proteins ID’ed with a combined 5-fold increase in distinct proteins (2 or more peptides/protein) using them as a panel (ca. 1400 proteins vs. 300 proteins). Each individual NP captured unique proteins with most showing some overlap with at least another member of the panel.

Figure 2: Dynamic range of total intensities for an example NP particle. Total ion current intensities for all proteins observed (2 or more peptides) using nanoparticle 2 (NP2) for Plasma Control showing a span of 8 orders of magnitude. Those proteins observed to be unique to NP2 only are highlighted in orange. There appears to be no protein abundance bias for any of the NP enrichments.

Reproducibility
Consistent protein ID’s, tight clustering of each NP and median CV of 25%

Inter-day analysis over 3 consecutive days (5 replicates/day)

Proteograph Applied in Planned Myocardial Infarction

Figure 4: PMI deep profile plasma processing workflow. The original iTRAQ study comprised of peripheral blood draws for 4 patients taken at 4 time points: baseline taken pre-administration of heparin block (BLpre), and then 10min, 1hr and 4hr post alcohol ablation heart surgical procedure (figure 4)2.

In the present study employing the Proteograph, we analyzed the same 4 plasma timepoint samples from 3 of the 4 original patients, and we also analyzed a 5th time point sample taken immediately post heparin application but pre- alcohol ablation from these same patients (BLpost).

P1 P2 P3 P1 P2 P3 P1 P2 P3 P1 P2 P3 P1 P2 P3

Figure 5: Bar graph protein ID’s across 3 PMI patients. On average, 2200 proteins were detected in each individual patient sample with a total of over 4000 observed across all 15 PMI samples processed using the Proteograph.

Correlation of patient samples for every time point is high, PC1 separates patients, and PC2 time points within each patient

Protein ID’s

Fold Increase

Unique ID’s

Nanoparticle

No NP

NP1

NP2

NP3

NP4

NP5

Combined

296

549

625

797

608

930

1376

–

1.9

2.1

2.7

2.1

3.1

4.6

31

51

137

90

254

A

TNNI3

10min 1hr 4hr

10 10 109 108 107 106 105

Pa2t13 Pa2t27 Pa3t36

B

10 10 109 108 107 106 105

TNNT2

10min 1hr

BLpre

BLpost

Time point

BLpre

BLpost

4hr

Time point

Figure 7: Longitudinal intensity changes for 3 troponin markers, TNNI3 (A), TNNT2 (B), TNNC2 (C)

Cardiac Troponin I and T were detected and observed to rise early while skeletal Troponin C2 remained flat as previously observed. Moreover, cardiac Troponin I and T were detected robustly across the patient samples as early as 10 minutes after the procedure in one of the PMI patients and at 1 hour after the injury in two out of three patient samples.

Correlation of Seer Proteograph results to iTRAQ discovery

P23 P27 P36

BLpre

BLpost

4hr

Time point

P23 P27 P36

Pat2P27_10min 4

2

-4-2 24

-2

-4

Seer

r

95% confidence interval R squared

-0.004549
-0.08572 to 0.07668 2.070e-005

Pat2P27_1hr 4

2

-4-2 24

-2

-4

Seer

r

95% confidence interval R squared

0.02399
-0.05265 to 0.1004 0.0005756

PatP227_4hr 4

2

-4-2 24

-2

-4

Seer

r

95% confidence interval R squared

0.03661
-0.03976 to 0.1126 0.001340

Figure 8: Scatter plots of ~ 450 proteins detected in Proteograph and iTRAQ discovery for one of the PMI patients at 10min A), 1hr B) and 4hr C) relative to baseline. More than 150 of the previously observed ca. 300 regulated proteins were detected in the current analysis and showed similar trends.

B

NP1 NP2 NP3 NP4 NP5

C

A

A

B

• • • • •

•

1 2 3

Conclusions

An average of 2200 proteins were detected in each patient plasma sample using the Proteograph system; a total of ca. 4000 proteins across the 15 PMI samples.

Cardiac Troponins, known markers of myocardial infarction, were robustly detected as early as 10 min post ablation using the Proteograph system and increased with time as expected.

Troponins were detected at earlier times post ablation than in the original iTRAQ study and with many more peptides/protein.

More than 150 of the previously observed ca. 300 regulated proteins were detected in the current study and showed similar trends.

Comparison of pre- and post- heparin baseline samples revealed a large number of heparin binding proteins in the Proteograph data, and AEBP1 and FHL1 proteins were detected and showed trends across the time course similar to the iTRAQ discovery study (data not shown).

New potential markers of PMI were detected using the Proteograph that need to be validated.

Figure 3: (A) Proteins identified in all 5 individual NPs across 3 consecutive days for Plasma Control with 5 replicates/day. NP4 showed consistently the lowest number of identifications while NP5 always resulted in the most. (B) PCA analysis of same data showing tight clustering of individual surface chemistries and separation of them on PC1 and PC2. (C) Histogram of %CV for all NPs combined across all three plates. CVs are calculated for the protein intensities across all 5 nanoparticles per replicate. Median CV was 25% while over 50% of the data has a CV of less than 30%.

Histogram of % CV (Sum of all) 500

Median %CV = 25%

400
300
200
100
0
0 20 40 60 80 100120140

Bin Center %CV

Figure 6: (A) Box-Whisker plot of correlation coefficient distribution grouped by timepoints. Each group shows a similar correlation of ca. 0.83 and tight distributions. (B) PCA analysis of the samples shows separation of the patients on PC1 and timepoints on PC2.

Pat1

Cardiac Troponins robustly detected in all 3 PMI patient samples as early as 10min post-injury – much earlier than in our original deepscale iTRAQ experiments

Keshishian et al., Nature Protocols Vol.12 No.8, 2017. Keshishian et al. Molecular and Cellular Proteomics 2015 Blume et al. Nature Communications 11, 3662 2020

References

Disclosure: S.A.C. is a scientific advisor to Seer

C

10 11 10 10 109 108 107 106 105

TNNC2

10min 1hr

Pat2

Pat3

Frequency

iTRAQ

iTRAQ

Log10(total Intensity)

iTRAQ

Log10(total Intensity)

Log10(total Intensity)