Combining laser capture microdissection with liquid chromatography and tandem mass spectrometry is the original spatial proteomics, and it remains the deepest. Its logic sidesteps the fundamental constraint on imaging: rather than accepting whatever a pixel contains, you decide what a sample is by drawing around it, then hand that sample to a conventional workflow. The cost is that the sample is very small.
Key Takeaways
|
The LCM Plus LC-MS Workflow
Five stages, and the difficulty is concentrated in the middle three rather than distributed evenly.
Stage | What Happens | Where the Difficulty Lies |
Sectioning and mounting | Thin sections cut onto specialised slides, either metal-framed membrane or glass, compatible with microdissection | Slide choice constrains both cutting and downstream handling |
Staining and region identification | Morphological or immunohistochemical staining to identify the regions to be collected | Staining reduces protein recovery, so this stage costs depth |
Microdissection and collection | Regions traced and excised by laser, then collected into a tube, plate, or chip | Collection efficiency and avoiding loss of very small samples |
Lysis, retrieval, and digestion | Protein solubilisation, de-crosslinking if the tissue is fixed, and enzymatic digestion | The dominant source of variability. Adsorptive losses matter at these amounts |
Chromatography and mass spectrometry | Low-flow separation followed by tandem MS | Sensitivity, and matching gradient and method to a very small load |
Table 1. The five stages of an LCM proteomics workflow. Stages two to four determine how much of your proteome survives to be measured.
Worth stating plainly why this approach is chosen over direct imaging of proteins on tissue. As the microdissection literature notes, imaging approaches provide spatial localisation but proteome coverage is limited to a few hundred identified proteins and is prone to matrix effects that challenge quantification. Microdissection trades continuous spatial coverage for an order of magnitude more depth, and for compatibility with quantitative workflows already validated in bulk proteomics. The wider comparison of approaches sits in Spatial Proteomics by Mass Spectrometry: LCM, Single-Cell, and Imaging Approaches.
Working in analytical science?
Register for a FREE Separation Science account to subscribe to the Separation Science Newsletter.
Choosing and Cutting Regions
Region selection is a scientific decision disguised as a technical one, and it deserves more deliberation than it usually gets. Three considerations dominate.
- Homogeneity against material. A tightly drawn region containing one cell type gives interpretable results and little protein. A larger region gives depth and averages across cell types. Deciding which matters more is the study design.
- Slide format. Specialised microdissection slides, either metal-framed membrane or glass, are required, and the choice affects cutting behaviour and downstream handling. Published benchmarking has compared both formats directly.
- Section thickness. Thickness multiplies the material collected for a given area. Published low-input work commonly uses 5 micron sections, so an area figure implies a volume only once thickness is stated.
That last point causes real confusion in the literature, because studies report their sample size variously as an area in square microns, a volume in cubic millimetres or nanolitres, or a cell count. These are not interchangeable without knowing thickness and cell density, which is why Table 2 below records the units each study used rather than converting them into a single scale. Treat any single-number rule of thumb for how much tissue you need with suspicion.
Does Staining Cost You Proteins?
Yes, and this is the tension at the centre of the method. You are using microdissection because you need to see the tissue in order to choose regions, and the standard way of seeing it interferes with what you are trying to measure.
Haematoxylin Reduces Protein Detection Work on LCM-MS of formalin-fixed and stained human lung tissue states the position directly: haematoxylin and eosin staining provides critical morphological characterisation enabling researchers to identify anatomical features of interest, but haematoxylin staining has been shown to reduce protein detection by mass spectrometry. The authors note that few LCM-MS studies existed for H&E-stained FFPE sections precisely because of this, and that novel protocols for such material were warranted. Their response is instructive: rather than avoiding the stain, they combined several steps that individually enhance protein yield, beginning with a detergent-based heat retrieval procedure shown to improve protein solubility, then adding two techniques to enhance extraction of extracellular matrix proteins, namely physical disruption and chemical extraction with a urea-based buffer. In other words, the answer to a staining penalty is a more aggressive extraction, not a compromise on morphology. |
The alternative is immunohistochemical identification, which substitutes specificity for general morphology. One published approach stains for a marker of interest, isolates the positive population by microdissection, and analyses it directly, which allows a defined cell population rather than an anatomical region to be the unit of analysis. That has obvious appeal where the population is defined by a marker rather than by position.
A framework for ultra-low-input spatial tissue proteomics published in Cell Systems benchmarked this systematically, comparing heat-induced and protease-induced epitope retrieval against H&E-based preparation across microdissected areas of 1,562, 12,500, and 50,000 square microns from 5 micron sections, and reporting proteome correlations and coefficients of variation between the retrieval strategies along with which proteins were exclusive to each. If you are establishing a protocol, that comparison is the right place to start rather than adopting a single method on trust.
Low-Input Sample Preparation
This is where most LCM proteomics experiments are won or lost, and the reason is unglamorous: at these amounts, protein lost to a tube wall is a measurable fraction of the sample.
Four principles follow from that, and they are consistent across the published protocols.
- Minimise transfers. Every pipetting step and every vessel change loses material. One-pot protocols that perform lysis, digestion, and clean-up in a single vessel outperform multi-step approaches at low input for this reason alone.
- Collect directly into the reaction vessel. Modern workflows collect microdissected material straight into the chip or plate where lysis and digestion will occur, eliminating a transfer entirely.
- Automate where possible. Manual handling at these volumes is both variable and slow, and automation addresses both.
- Reverse crosslinks if the tissue is fixed. FFPE material needs a de-crosslinking step, which modern protocols integrate into the same vessel as lysis and digestion rather than performing separately.
The gains from automation are documented rather than assumed. Work coupling microdissection to a fully automated proteomics workflow reported, when benchmarked against filter-aided sample preparation, roughly twofold lower variance in quantification and more than fivefold faster throughput. A more recent automated protocol for laser microdissection guided ultrasensitive proteomics uses a robotic sample handling platform capable of processing 192 samples in three hours, collecting microdissected material directly into low-volume chips and performing lysis, formalin de-crosslinking, and tryptic digestion in place, with integration to a low-flow chromatography system allowing sample clean-up during transfer.
Two points from that automation work are worth carrying into any protocol. Formalin de-crosslinking sits inside the automated sequence rather than as a separate manual step, which reduces both handling and variability. And the throughput figure matters for study design: 192 samples in three hours makes cohort-scale spatial proteomics feasible in a way that manual preparation does not.
How Much Tissue Do You Actually Need?
Less than most people assume, and the published figures are specific enough to plan against. The caveat is that they are not directly comparable, since studies report sample size in different units and use different instruments and workflows.
Material Analysed | As Reported | Depth Achieved | Conditions |
Archival tonsil microregions | 4,000 square microns | Approximately 2,000 proteins | B-cell, T-cell and epithelial regions, high cell type specificity, automated preparation |
FFPE tissue, general | Not specified as area | Up to 2,000 proteins reproducibly quantified | Reported for improved preparation and LC-MS workflows on archival material |
Alveolar tissue | 4,000 cells | More than 3,400 proteins | Automated workflow, characterising protein changes in lung development |
Collected tissue with fractionation | 175 nanolitres, 5 to 7 micrograms peptides | 3,600 to 4,400 proteins | Additional peptide fractionation step before LC-MS |
Sorted marker-positive population | 32 nanolitres | 674 high-confidence proteins | Immunohistochemically identified cells, FDR below 0.01, earlier-generation instrument |
FFPE H&E-stained lung regions | 0.082 to 0.094 cubic millimetres | 1,252 uniquely expressed proteins | Three unique peptide threshold; 892 proteins differentially expressed between region types |
FFPE H&E-stained, sensitivity limit | Down to 0.0125 cubic millimetres | Method shown to remain applicable | Demonstrated across a range of dissected volumes |
Table 2. Published sample amounts against achieved proteome depth. Units are as each study reported them and are not interchangeable without knowing section thickness and cell density. Instruments and workflows differ between rows, so read these as documented outcomes rather than as a calibration curve.
Three readings of that table are useful. Depth rises with material, as expected, but not proportionally: the difference between 32 nanolitres and 175 nanolitres is roughly a fivefold increase in material for a fivefold increase in proteins, while much larger volumes do not scale nearly so well. Workflow matters as much as amount, since the 4,000 square micron result and the 0.08 cubic millimetre result are within a factor of two of each other on depth despite very different sample sizes. And fixation is less limiting than reputation suggests, with FFPE results in the same range as frozen material.
For scoping purposes, the honest guidance is to expect between roughly one and three thousand proteins from a carefully collected microregion with a modern workflow, to run a pilot at two or three region sizes before committing a cohort, and to treat any single figure from the literature as conditional on its workflow rather than transferable.
What Are the Real Limitations?
Five, and none is a reason to avoid the method, though two are reasons to design around it.
Limitation | Why It Arises | Mitigation |
Discrete rather than continuous sampling | You measure the regions you chose, so anything between or outside them is unmeasured | Design region selection as sampling, and consider imaging for a continuous survey first |
Region selection is a prior decision | Regions are chosen from morphology or markers before proteomics, so the analysis inherits that framing | State the selection criteria as part of the method, since they shape the result |
Staining penalty on protein recovery | Haematoxylin reduces protein detection, yet morphology is needed to choose regions | Compensate with enhanced retrieval and extraction rather than abandoning staining |
Contamination is proportionally large | At nanolitre volumes, keratin and polymer contamination from handling and membrane slides is a significant fraction | Strict handling discipline, and blank controls processed identically |
Throughput | Manual dissection and preparation limit cohort size | Automation, with reported capacity of 192 samples in three hours |
Table 3. Limitations and responses. The second row is a scientific caveat rather than a technical one and belongs in the methods section of any paper using this approach.
The second row deserves emphasis because it is easy to overlook. Microdissection produces a proteome of the regions you decided were interesting, identified by the markers or morphology you decided to use. That is not a flaw, but it does mean the result is conditional on the selection, and reporting the selection criteria as part of the method is what allows a reader to judge it. This is the same structural point made about imaging-guided approaches in the section hub: in MS-based spatial proteomics, the targeting frequently happens at the cell or region rather than at the analyte.
Where the aim is to select cells by phenotype rather than by anatomical position, and to automate that selection, the imaging-guided extension of this approach is covered in Deep Visual Proteomics: Imaging-Guided Mass Spectrometry. Upstream sectioning, mounting, and fixation decisions common to all these workflows are treated in Sample Preparation for Spatial Analysis: From Tissue to Data. For how the depth achieved here compares with direct imaging of peptides on tissue, see Mass Spectrometry Imaging: Principles, Techniques, and Applications, and for the wider spatial landscape, Spatial Analysis in Analytical Science: Mass Spectrometry Imaging and Spatial Omics.
This article was produced under Separation Science's AI Editorial Guidelines.
