Key words

1 Introduction

Mass spectrometry (MS)-based proteomics at single-cell resolution is poised to generate new insights into biological heterogeneity. Clearly, information on the spatial context of a given cell is a powerful addition to proteomics data, as the metabolic niche, cell–cell contacts, and signaling gradients contribute greatly to proteome variability and dynamics [1].

The mammalian liver is an exciting model for biological discoveries related to niche/proteome dependency. In addition, it is a compelling in situ system for benchmarking MS-based single-cell proteomics workflows, due to prior knowledge of spatial markers [2,3,4,5], fairly large cells (about 675 pg of total protein per complete hepatocyte [6]), ease of isolation, and the repetitive nature of liver zonation. The liver consists of approximately 80% hepatocytes whose proteome and therefore function are highly dependent on their position within the organ. Hepatocytes are organized into functional units termed lobules that are histologically defined by portal and central veins. Due to paracrine signaling and changes in nutrient and oxygen composition, half of the hepatocyte proteome significantly changes along the zonation axis within just a few hundred micrometers [7].

To map the spatial proteome of the murine liver at single-cell resolution, we have recently described a workflow termed single-cell Deep Visual Proteomics (scDVP, [7]) to resolve the hepatocyte proteome at a median depth of 1700 proteins per hepatocyte slice, where each cell shape corresponds to a third to one half of a cell. The method builds on our previously described DVP workflow [8] that integrates imaging of immuno-stained frozen mouse liver sections, AI-guided cell segmentation, laser microdissection, and ultrahigh sensitivity mass spectrometry.

In MS-based proteomics, data-independent acquisition (DIA) has become increasingly powerful and popular due to improved instrumentation and algorithms, which now enable very deep, quantitative, and reproducible proteome measurements [9, 10]. To enable scDVP, we integrated multiplexed DIA (mDIA, [11]) using dimethyl labeling to combine two single cells and a reference sample in one MS run. In our approach, we used the lowest mass label (“Δ0”) for a higher abundant reference sample of the same nature as the low input material, while the other two channels (“Δ4” and “Δ8”) contain proteomes of single cells. This maximized proteome depth, quantitative accuracy, and throughput at minimal reagent costs.

In this chapter, we present a step-by-step protocol to obtain deep and high-quality proteomes of single hepatocytes starting from isolated cells. We describe quality control features that harness liver zonation for benchmarking single-cell proteomics experiments, showcasing applications of single-cell proteomics to biology. The protocol is highly modular, and almost all steps can be easily adapted to existing laboratory equipment and extended to other cell types and scientific questions. In particular, we present our current approach for MS data acquisition, using the Evosep One LC system coupled to a timsTOF SCP mass spectrometer (Bruker) and with DIA-NN [12] for data analysis. However, the prepared samples can be measured on a wide range of instrumentation when the step-by-step protocol is adapted accordingly.

2 Materials

Prepare all solutions using ultrapure water (MilliQ-grade or MS-grade water). Prepare all reagents fresh and use them within one day. Diligently follow your local waste disposal routines and regulations, as well as the appropriate safety regulations when handling chemicals.

2.1 General Reagents and Materials

  1. 1.

    1× phosphate-buffered saline (PBS).

  2. 2.

    1 M triethylammonium bicarbonate (TEAB) buffer (pH 8.5, see Note 1).

  3. 3.

    0.5 μg/μL (stock concentration) of LysC in 50 mM ammonium bicarbonate.

  4. 4.

    0.5 μg/μL (stock concentration) of trypsin in 1 mM HCl.

  5. 5.

    10% trifluoroacetic acid (TFA) in water (see Note 2).

  6. 6.

    Buffer A: 0.1% formic acid in water.

  7. 7.

    Buffer B: 0.1% formic acid, 99.9% acetonitrile (ACN).

  8. 8.

    100% acetonitrile.

  9. 9.

    80% acetonitrile in water.

  10. 10.

    pH paper.

  11. 11.

    Vacuum concentrator.

2.2 Reference Channel

  1. 1.

    Liver tissue on a glass slide (see Note 3).

  2. 2.

    ref-Lysis buffer: 10% ACN, 0.01% dodecyl-β-d-maltoside (DDM) in 60 mM of TEAB. Prepare a 1% DDM stock solution by dissolving 5 mg DDM in 500 μL ultrapure water. For 1 mL lysis buffer, mix 60 μL of 1 M TEAB, 100 μL 100% ACN, 10 μL of 1% DDM solution, and 830 μL water.

  3. 3.

    Light labeling solution 1: 1.5% light formaldehyde (CH2O, FA-L) in water. Add 18 μL of 37% FA-L into 426 μL water.

  4. 4.

    Light labeling solution 2: 0.23 M light sodium cyanoborohydrate (BA-L) in water. Dissolve 7.23 mg BA-L in 500 μL water (stock solution, see Note 4).

  5. 5.

    C-18 solid-phase extraction cartridges and an appropriate pressure device.

  6. 6.

    Methanol at HPLC grade.

2.3 Single Cell Peptide Preparation

  1. 1.

    Sorted single cells in low-binding 384-well plates (see Note 5).

  2. 2.

    sc-Lysis buffer: 0.013% dodecyl-β-d-maltoside (DDM) in 60 mM of TEAB buffer, pH 8.5. For a final volume of 50 mL, mix 3 mL of 1 M TEAB, 650 μL of 1% DDM, and 46.35 mL water.

  3. 3.

    Digestion mix: 4 ng/μL of LysC, 6 ng/μL of trypsin in 60 mM of TEAB buffer, pH 8.5. To 1.225 mL of 60 mM TEAB buffer (pH 8.5), add 10 μL of 0.5 μg/μL LysC and 15 μL of 0.5 μg/μL trypsin.

  4. 4.

    Intermediate labeling solution 1: 1.5% intermediate formaldehyde (CD2O, FA-I) in water. Add 75 μL of 20% FA-I to 925 μL water.

  5. 5.

    Intermediate labeling solution 2: 0.23 M light sodium cyanoborohydrate (BA-L) in water. Dissolve 14.45 mg of BA-L in 1 mL water (stock solution).

  6. 6.

    Heavy labeling solution 1: 1.5% heavy formaldehyde (13CD2O, FA-H) in water. Add 75 μL of 20% FA-H to 925 μL water.

  7. 7.

    Heavy labeling solution 2: 0.23 M heavy sodium cyanoborohydrate (BA-H) in water. Dissolve 15.15 mg of BA-H in 1 mL water (stock solution).

  8. 8.

    Adhesive PCR sealing foil sheets.

2.4 Loading of StageTips

  1. 1.

    Evotips pure (Evosep).

  2. 2.

    One empty Evotip box.

  3. 3.

    1-propanol.

2.5 Mass Spectrometry and Sample Acquisition

  1. 1.

    timsTOF SCP mass spectrometer or comparable.

  2. 2.

    Evosep One LC system.

  3. 3.

    Column: Aurora Elite CSI third-generation C18, 15 cm, 75 μm ID (IonOpticks, Part No. AUR3-15075C18-CSI).

2.6 Data Analysis

  1. 1.

    FragPipe version 18.0 or later [13] running MSFragger version 3.5, Philosopher version 4.4.0, and EasyPQP version 0.1.32 (or later versions).

  2. 2.

    Species-specific FASTA file, which can be downloaded from https://www.uniprot.org/proteome.

  3. 3.

    DIA-NN version 1.8.1 or later (https://github.com/vdemichev/DiaNN, [12]).

  4. 4.

    RefQuant (https://github.com/MannLabs/refquant, [11]).

  5. 5.

    Python version 3.8 or later.

  6. 6.

    Jupyter notebook.

3 Methods

The entry point to this protocol is a 384-well plate with sorted single hepatocytes (or single hepatocyte shapes; one per well) without buffer, for which we do not describe the protocol here. We have tested two approaches, either laser microdissection (scDVP, described in detail in [7]) or fluorescence-activated cell sorting (FACS) of a primary hepatocyte solution (e.g., obtained with a gentleMACS™ Dissociator, Miltenyi Biotec, see Note 6). Some prior markers for portal and central veins should be included for scDVP or FACS (see Note 7). Alternatively, marker expression, e.g., murine Arg1 and Cyp2f2, is a good proxy for pseudo-spatial alignment of single-cell samples, similar to the concept described in [3]. Store the sample plates at −80 °C prior to peptide preparation.

3.1 Reference Channel Peptides

  1. 1.

    For OCT-embedded and frozen tissue sections on glass slides (see Note 8), dry the slides for 60 s after cutting. Wash the slides twice in PBS to remove the embedding medium. Scratch the tissue off the slides with a scalpel and transfer it into one 1.5 mL safe-lock tube (see Note 9).

  2. 2.

    Add 100 μL of ref-Lysis buffer and centrifuge the sample briefly.

  3. 3.

    Boil the sample at 96 °C in a thermoshaker for 20 min.

  4. 4.

    Sonicate the samples at maximum output for five on/off cycles at 30 s per cycle.

  5. 5.

    Centrifuge at 2000 ×g for 2 min. Transfer the supernatant to a new 1.5 mL tube.

  6. 6.

    Estimate protein concentration on a Nanodrop or similar method, and dilute the sample to less than 10 μg/μL (see Note 10) with ref-Lysis buffer. Ensure that the pH is between 8 and 9 by spotting 2 μL on pH paper.

  7. 7.

    LysC and trypsin are added in a protein-to-enzyme ratio of 1:50. The sample is digested in a thermoshaker overnight (about 18 h) at 37 °C and 700 rpm (see Note 11).

  8. 8.

    The sample is labeled with the light labeling solutions (see Note 12). First, add FA-L, vortex briefly, and then add BA-L directly to reach a final concentration of 0.15% for FA-L and 23 mM for BA-L. Allow labeling at room temperature for one hour (see Note 11).

  9. 9.

    Quench and acidify the reaction with 10% TFA to a final concentration of 1% (see Note 13).

  10. 10.

    Dry the reference channel sample in a vacuum concentrator at 60 °C, and dissolve the resulting pellet in 1 mL of buffer A.

  11. 11.

    Desalt the sample on a suitable C-18 solid-phase extraction cartridge. Activate the column with 2 mL of methanol, then 2 mL of buffer B and equilibrate with 2 mL of buffer A before sample loading. Load the sample solution and subsequently wash the column twice with 1 mL of buffer A. Elute the sample with 1 mL of buffer B. The solutions are manually pushed through the C-18 column with a syringe as a pressure device (see Note 14).

  12. 12.

    Dry the samples in a vacuum concentrator at 60 °C until completely dry. Reconstitute the reference peptide to a working solution of 1 ng/μL in buffer A and shake it in a thermoshaker for 20 min at 1500 rpm. Store at −20 °C.

3.2 Single-Cell Peptide Preparation

In our laboratory, all pipetting steps are currently conducted automatically by an Agilent Bravo pipetting robot executing custom-written protocols. The solutions are either provided in pyramid-bottom reservoirs or in PCR tubes stacked into 96-well plates. The pipetting steps can also be performed with the help of any pipetting robot, or by hand. All centrifugation steps are carried out at 1000 ×g for 1 min, if not indicated differently.

  1. 1.

    Take the 384-well plate containing the single cells from the freezer. Immediately centrifuge the plate for 2 min at 2000 ×g.

  2. 2.

    Optional: For the laser microdissection workflow, rinse each well with 30 μL of 100% ACN on each side and dry the samples in a vacuum concentrator at 45 °C for 20 min (see Note 16). Add 6 μL of sc-Lysis buffer to each well, seal tightly with an adhesive PCR sealing foil sheet (see Notes 6 and 17), and centrifuge.

  3. 3.

    Heat the plate at 95 °C in a 384-well plate PCR cycler (see Note 18) for 30 min (FACS-sorted live cells and fresh frozen scDVP samples) or 1 h (FFPE samples, see Note 19).

  4. 4.

    Centrifuge the plate and add 1 μL of 80% ACN. Seal tightly, centrifuge, and heat the samples at 75 °C for 30 min (FACS-sorted live cells and fresh frozen scDVP samples) or 1 h (FFPE samples).

  5. 5.

    Centrifuge the plate and add 1 μL of digestion mixture. Seal tightly, centrifuge, and digest samples overnight (12–18 h) at 37 °C in a 384-well plate PCR cycler.

  6. 6.

    Centrifuge the samples prior to labeling with intermediate (Δ4) or heavy (Δ8) tags (see Note 12). Add 1 μL of either intermediate (FA-I) or heavy (FA-H) formaldehyde, resulting in a final concentration of 0.15%. Seal and centrifuge the plate. Next, add 1 μL of light (BA-L) or heavy (BA-H) sodium cyanoborohydrate, resulting in a final concentration of 23 mM per sample (for pipetting scheme see Fig. 1). The combination of FA-I + BA-L and FA-H + BA-H results in intermediate (Δ4) and heavy labeling (Δ8), respectively. Seal, centrifuge, and incubate the labeling reaction at room temperature for 1 h.

  7. 7.

    The reaction is quenched by adding 1 μL of 10% TFA. Seal and centrifuge the plate. The samples can be stored at −20 °C, or processed directly after.

Fig. 1
A schematic diagram exhibits the colored plate for the sample preparation. The plate has an array of wells, in which the columns are labeled from 1 to 24 and the rows are labeled from A to P. The wells not used, delta 4, and delta 8 are indicated by the different colors.

Layout of the sample preparation plate. We do not sort cells into the outermost wells (indicated by the black filling), as these have a higher likelihood of evaporation during high-temperature incubation steps. For assigning wells to a 384-well plate in the laser microdissection workflow, see Note 15. The different colors indicate the later labeling approach used. The Δ4 and Δ8 labeling is performed in an alternating manner from top to bottom of the plate

Full size image

3.3 Loading on Evotips

All centrifugation steps are carried out at 700 ×g for 1 min.

  1. 1.

    Activate the Evotips in 1-propanol for 3 min (see Note 20).

  2. 2.

    Wash the Evotips twice with 50 μL of buffer B. After each washing step, centrifuge the Evotips (see Note 21).

  3. 3.

    Activate the Evotips again in 1-propanol for 3 min (see Note 22).

  4. 4.

    Wash the Evotips with 50 μL of buffer A and centrifuge. Repeat this washing step once. From here on, keep the disk wet at all times.

  5. 5.

    Add 70 μL of buffer A per Evotip and centrifuge for 15 s (see Note 23).

  6. 6.

    Load the samples into the buffer A present in the Evotips. First, add 10 μL of 1 ng/μL reference peptides (Δ0), followed by the intermediate (Δ4) and heavy (Δ8) labeled single-cell samples into the same Evotip (see Note 24). Rinse each corresponding well of the 384-well plate with 15 μL of buffer A and add it to the respective Evotip. Centrifuge the Evotips and extend the centrifugation step if the samples in buffer A did not completely run through the Evotip.

  7. 7.

    Wash the Evotips with 50 μL of buffer A and centrifuge.

  8. 8.

    Add 150 μL of buffer A to the Evotips, centrifuge for 20 s at 700 ×g, and put the loaded Evotips in an Evotip box (never contained 1-propanol) with fresh buffer A to prevent the Evotips from drying.

3.4 Mass Spectrometry and Sample Acquisition

We present here our current approach using the Evosep One LC system coupled to a timsTOF SCP mass spectrometer (Bruker). Importantly, the prepared samples can be measured on a wide range of instrumentation; adapt the step-by-step protocol accordingly.

  1. 1.

    Ensure your mass spectrometer is set up correctly, and you have checked the performance of your setup (see Note 25).

  2. 2.

    Load samples onto the Evosep One LC system coupled to a timsTOF SCP mass spectrometer and write your sample table.

  3. 3.

    Use your LC method of choice. We use the Whisper40 SPD (samples per day) method with the Aurora Elite CSI third-generation 15 cm and 75 μm ID C18 column at 50 °C inside a nanoelectrospray ion source.

  4. 4.

    Use your high-performance and high-sensitivity mass spectrometer of choice. We operate the timsTOF SCP in high sensitivity mode with an optimal dia-PASEF method generated with py_diAID [14]: Use 8 dia-PASEF scans with variable width and 2 ion mobility windows per dia-PASEF scan, covering an m/z range from 300 to 1200 and an ion mobility range from 0.7 to 1.3 Vs cm−2 (Table 1). Additional settings are an accumulation and ramp time at 100 ms, capillary voltage set to 1400 V, and the collision energy as a linear ramp from 20 eV at 1/K0 = 0.6 Vs cm−2 to 59 eV at 1/K0 = 1.6 Vs cm−2. For more details, refer to [11].

Table 1 The optimal eight dia-PASEF window scheme for high-sensitivity tryptic HeLa digest for the acquisition of multiplexed samples with a reference channel and two single cells
Full size table

3.5 Generating a Spectral Library

There are multiple options for generating a spectral library for mDIA (see Note 26). Below, we describe the generation of an experimental spectral library recorded in data-dependent acquisition (DDA) mode with single DDA shots of the reference channel. This approach can be enhanced by prior fractionation of the reference sample and DDA acquisition of individual fractions. Additionally, the spectral library can be predicted, for example, with AlphaPept Deep [15]. Maximum protein numbers will be reached with a combined strategy to first predict a library covering the Δ0 modified proteome that is subsequently used to search experimental DIA data in DIA-NN. The resulting smaller library saved by DIA-NN can be used for single-cell searches (for more details, see [11]). There are advantages and disadvantages to using each of the library approaches. We note that there is an overall trend to generate the spectral library directly from the data as software algorithms improve, obviating the need to separately create the spectral library.

  1. 1.

    Use the same LC-MS setup with the same gradient as used for single-cell sample acquisition.

  2. 2.

    Measure five dda-PASEF single-shots, each from 50 ng of reference channel peptides.

  3. 3.

    Generate a spectral library by MSFragger run in FragPipe against your FASTA reference file of choice including 50% decoys (added by clicking on “add decoys”).

  4. 4.

    Use standard settings of the DIA_SpecLib_Quant workflow with the following exceptions: (a) Set N-terminal and lysine mass shift of 28.0313 Da as fixed modifications, and methionine oxidation as variable modification (Fig. 2a). (b) Unselect N-terminal alkylation ([^, 42.0106) as well as carbamidomethylation (57.02146) from cysteine (the presented protocol does not use peptide reduction and alkylation). (c) Accept only one missed cleavage and a maximum of one variable modification on a peptide. (d) Set the precursor charge range from 2 to 4. (e) Set the peptide mass range from 300 to 1800, and peptide length from 7 to 30.

  5. 5.

    Modify the output directory to your path of choice.

  6. 6.

    For DIA-NN compatibility, remove the column “FragmentLossType” in the output library file (called “library.tsv”) (see Note 27) and rename all “Unimod:36” strings to “Dimethyl” (Fig. 2b).

Fig. 2
A. A screenshot of the variable modifications has max variable mods on a peptide, max combinations, and use all mods in first search. A screenshot of the fixed modifications has enabled, site, and mass delta. B. A screenshot lists precursor, product, annotation, proteinid, gene name, peptide, and modified sequence.

(a) Modification settings in the “MSFragger” tab of FragPipe to generate a spectral library from DDA data. Note that the variable modification “[^” is unticked, and the fixed modifications “N-Term peptide” and “K (lysine)” are changed to 28.0313, and “C (Cysteine)” is changed to 0.0. Note that other settings should resemble the DIA-NN settings in Fig. 3. (b) Example of one precursor fragment in a DIA-NN compatible MSFragger library. Note the replacement of “UniMod:36” with “Dimethyl.” Only some of the relevant columns are shown

Full size image

3.6 DIA-NN Search

  1. 1.

    Install the latest version of DIA-NN.

  2. 2.

    Search all files together by adding folders via the “diaPASEF.d” button.

  3. 3.

    Set the path to the spectral library (see Note 26).

  4. 4.

    Set the Precursor FDR to 1.0%.

  5. 5.

    Set the mass and MS1 mass accuracy to 15.0 (see Note 28).

  6. 6.

    Use scan windows of 9 (see Note 28) and activated isotopologues, match between runs (MBR), heuristic protein inference, and no shared spectra.

  7. 7.

    Infer proteins from genes, use single-pass mode as a neural network classifier, “Robust LC (high precision)” as the quantification strategy, “RT-dependent” as cross-run normalization, set library generation as “IDs, RT & IM profiling,” and use “Optimal results” for the speed and RAM usage settings.

  8. 8.

    Specify additional commands in the additional options window (Fig. 3): (a) Set dimethyl labeling at N-termini and lysines as fixed modification at 28.0313 Da: --fixed-mod Dimethyl, 28.0313, nK (see Note 29); (b) Δ4 and Δ8 were spaced 4.0251 Da and 8.0444 Da from the reference Δ0, respectively: --channels Dimethyl, 0, nK, 0:0; Dimethyl, 4, nK, 4.0251:4.0251; Dimethyl, 8, nK, 8.0444:8.0444 (see Note 29); (c) and additional settings:--original-mods --peak-translation --ms1-isotope-quant --report-lib-info.

Fig. 3
A screenshot of the optional window. It includes options for input, output, spectral library, additional options, precursor ion generation, and algorithm.

Settings in DIA-NN for mDIA single-cell experiments

Full size image

3.7 Data Processing with RefQuant

  1. 1.

    Download and install RefQuant.

  2. 2.

    Modify your input path file to your DIA-NN report.tsv output file in the provided Jupyter notebook demo (./tutorial/tutorial.ipynb) with RefQuant.

  3. 3.

    Run the demo notebook provided with RefQuant; it will perform the following steps: (a) Filter for precursor and protein FDR: “Lib.PG.Q.Value” < 0.01, “Q.value” < 0.01. (b) Filter for channel FDR: “Channel.Q.Value” < 0.15 (see Notes 30 and 31). (c) RefQuant: ratio calculation for each precursor between the reference channel and the single-cell channels. (d) Protein quantification using directLFQ (directLFQ, see Note 32).

3.8 Quality Control

  1. 1.

    Use the filtered RefQuant quantification matrix to plot the unique protein IDs per single cell as the first step of quality control (see Note 33).

  2. 2.

    Further filter the quantified protein matrix by only including samples with a number of proteins detected between 1.5 standard deviations below and 3 standard deviations above the sample identification median.

  3. 3.

    The coefficient of variation (CV) per protein should be calculated to indicate the variation within the dataset. The CV is calculated by dividing the standard deviation by the mean protein intensity (see Note 34).

  4. 4.

    Perform a principal component analysis (PCA) to identify potential biases within the dataset, indicated by clustering. A color overlay should be applied to visualize potential batch effects, e.g., different biological replicates, sample preparation batches, and column changes (see Note 35).

  5. 5.

    To visualize the spatial power of single-cell proteomics, the expression levels of known liver zonation markers can be overlayed on top of the PCA (Fig. 4). The respective position of a hepatocyte within the liver lobule is expected to drive variance within the dataset and hence should clearly be resolved in principal component 1 (PC1). Typical murine markers for the portal vein area are argininosuccinate lyase (Asl), arginase-1 (Arg1), argininosuccinate synthase (Ass1), and steroid 21-hydroxylase (Cps1), as well as for the central vein area Cytochrome P450 2E1 (Cyp2e1) and glutamine synthetase (Glul/GS).

Fig. 4
2 scatter plots, a and b, of P C 2 versus P C 1 for A s l and C y p 2 e 1. The y axis ranges from negative 2.5 to 2.5, and the x axis ranges from negative 7.5 to 5. The plots are distributed through the graph. A gradient scale ranges from low to high.

Principal component analysis of 400 murine hepatocytes that were excised from murine liver sections at random positions (data from [7]). The color overlay shows the scaled expression levels of a marker for portal (Asl) and central (Cyp2e1) vein, “n.d.” not detected. Note that the expression of these markers correlates negatively with one another as expected

Full size image

4 Notes

  1. 1.

    The 1 M of TEAB stock solution (pH 8.5) can be stored in the fridge for up to 6 months. Check the pH of all buffers with pH paper prior to using them. Keep TEAB buffers in closed containers since they are volatile.

  2. 2.

    Dilute TFA to a 10% stock to facilitate pipetting. Work under the fume hood when handling TFA.

  3. 3.

    The tissue for the reference channel must represent the biological nature of all single-cell samples. In our liver workflow, we cut additional liver sections on a cryostat-microtome (frozen tissue sections) or microtome (FFPE tissue sections) and process them similarly to single-cell samples. If different conditions are compared, e.g., disease states, all proteomes need to be pooled and present in the reference sample.

  4. 4.

    Weigh and dissolve all cyanoborohydride salts in the fume hood, as their fumes are highly toxic.

  5. 5.

    We have tested several plates from different vendors and found that the twin.tec PCR 384 plates, LoBind (Eppendorf 30129547), performed the best and are thus highly recommended for ensuring optimal results.

  6. 6.

    When using an FACS-based approach, we recommend sorting live single cells directly into 6 μL of cold sc-Lysis buffer and placing the plate on dry ice directly after sorting.

  7. 7.

    For scDVP, we have established a “liver painting” protocol staining for e-cadherin as a portal vein marker, glutamine synthetase as a central vein marker, phalloidin as a segmentation marker to outline single cells, and a nuclear marker. For an FACS-based approach, refer to [4].

  8. 8.

    Many types of samples can be used to create the reference channel peptides. In the scDVP workflow, we cut consecutive sections of frozen or FFPE tissue, but it is also possible to start with small pieces of tissue (e.g., 1 mm3) and lyse them appropriately.

  9. 9.

    Ensure that the tissue pieces are at the bottom of the tube by either pushing down the samples with a pipette tip or rinsing the walls of the tube with 30 μL of 100% acetonitrile, centrifuging the tube for 2 min at 2000 ×g, and subsequently evaporating the ACN in a vacuum concentrator at 45 °C for about 20 min.

  10. 10.

    We aim for a final amount of 100 μg for the reference channel sample prior to digestion, which will be enough for >2000 single cells.

  11. 11.

    For instance, if prior to digestion the volume is at 100 μL and the protein concentration is 1 μg/μL, add 4 μL each of trypsin and Lys-C to reach 2 μg of each. For labeling, add 13.5 μL of 1.5% FA-L, vortex, and then 13.5 μL of 0.23 M BA-L.

  12. 12.

    We currently use the lowest mass label (“Δ0”) for the reference sample, while the other two channels (“Δ4” and “Δ8”) contain proteomes of single cells. However, the reference channel can also be put into Δ8 with single cells in Δ0 and Δ4. Change the analysis pipeline (steps 3.5 onward) accordingly.

  13. 13.

    The pH of the solution is required to be below pH 3. If the solution is too basic, add more TFA until the required pH is achieved.

  14. 14.

    To push the solutions through the C-18 desalting column one can build a self-made syringe adapter out of 1 mL pipette tips stabilized by parafilm (see also [16]). Note that any clean-up protocol can be used.

  15. 15.

    We use a Leica LMD7 for laser microdissection and assign well “A1” to the true well “B2” of the 384-well plate, thus omitting the outermost rows and columns.

  16. 16.

    The addition and subsequent evaporation of acetonitrile ensure that the isolated cells end up at the bottom of the well to secure the success of the following steps. Pipette 4× 7.5 μL of 100% acetonitrile on the four sides of the well to ensure the sliding of the isolated cells to the bottom of the well. The drying step in the vacuum concentrator might need to be prolonged. It is important that no liquid remains in the well.

  17. 17.

    Tight sealing of the plate is essential to avoid evaporation of samples during the heating steps. In our hands, the adhesive PCR sealing foil sheets (Thermo Scientific, AB-0626) work well. Seal the plate in a double manner to prevent evaporation of the little volume used.

  18. 18.

    Set the lid temperature 15 °C higher than the incubation temperature to prevent condensation on the lids.

  19. 19.

    Depending on the origin of your tissue sample, the duration of the heating step can vary. For a fresh-frozen sample, the mentioned 30 min duration has proven to be sufficient. However, for FFPE samples, we advise increasing the duration to 60 min. The prolonged incubation time and heat promote de-crosslinking of proteins for high recovery and reduction of formalin-induced lysine methylation [17].

  20. 20.

    We use old Evotip boxes to conveniently store the 1-propanol and activate the Evotips. Ensure that the C-18 disk in the tip is covered with 1-propanol. After 3 min, the tip should appear pale white. We reuse 1-propanol up to five times, and the boxes with 1-propanol can be stored at room temperature.

  21. 21.

    Use an empty Evotip box for centrifugation. The entire liquid should have passed through the disk in these washing steps; otherwise, repeat the centrifugation with the same settings.

  22. 22.

    The same boxes with 1-propanol as in the first activation step can be used.

  23. 23.

    It is important that buffer A has entered but did not completely pass through the C-18 disk. The tip should still contain buffer A.

  24. 24.

    We use the same pipetting tip for all three channels, starting with the reference channel, to minimize peptide loss due to plastic adhesion.

  25. 25.

    We usually acquire 5 ng HeLa digest in DDA in triplicates for performance and mass error evaluation (<5 ppm), and we also check the performance at the DIA level.

  26. 26.

    DIA-NN 1.8.1 supports multiplex-DIA data and only requires the Δ0 (reference channel) generated library. With the command --channels Dimethyl, 0, nK, 0:0; Dimethyl, 4, nK, 4.0251:4.0251; Dimethyl, 8, nK, 8.0444:8.0444, it automatically calculates the other channels (Δ4 and Δ8) from the provided Δ0 library.

  27. 27.

    Please check that the column “Fragment.LossType” is completely empty before deleting it.

  28. 28.

    Check your optimal scan window, mass, and MS1 mass accuracy by running some files with the additional “unrelated runs” option checked and evaluate the recommended settings by DIA-NN.

  29. 29.

    Ensure that “Dimethyl” is named the same way as the modification in your MSFragger library.

  30. 30.

    The RefQuant demo notebooks filter on a channel-FDR “Channel.Q.Value” < 0.15. This value was empirically determined by a count-based FDR with our setup and HeLa peptides but also confirmed by mouse liver peptides. The “Channel.Q.Value” cutoff can be modified directly in the RefQuant demo notebook. If you do, please handle this carefully and refer to the mDIA publication to calculate your own count-based FDR [11].

  31. 31.

    We are currently working on a software solution directly implementing the reference channel for identification with an AI-based FDR model and improved quantification. Look out for a publication termed “AlphaDIA.”

  32. 32.

    You can also use other protein quantification strategies. By default, the demo notebook uses directLFQ, but we also have an example of MaxLFQ.

  33. 33.

    For murine hepatocytes excised by laser microdissection with a thickness of 10 μm, we identified 1726 proteins per shape (median) after filtering.

  34. 34.

    Larger cells have smaller protein CVs. For randomly excised murine hepatocyte shapes with volumes of about 6000 μm3, expect a median CV of 30–50% across all detected proteins. Biological heterogeneity also impacts the CV value. In the case of higher biological heterogeneity, the CV values will be higher as well.

  35. 35.

    PCA reduces a high-dimensional dataset and highlights the axes of highest variance within the data in principal components. Batch correction can be performed to remove non-biological experimental variation.