SPOTTER: Automated Tissue-Barcoding Platform for Spatial Proteomics and Phosphoproteomics

Background:

Modern mass spectrometric (MS) approaches allow the accurate detection and identification of up to 10,000 proteins from complex biological samples. However, the homogenization required for sample preparation eliminates spatial context and masks regional heterogeneity. Subcellular fractionation, IP-MS, or proximity-labeling approaches can provide topological information on protein origin, but often require genetic engineering or larger input quantities posing a challenge for rare clinical samples. In addition, spatial sequencing-based approaches (like Visium, MERFISH) enable measuring the abundance of thousands of transcripts without loss of the anatomical structure of tissues.

Recently, several spatial proteomics approaches have emerged, allowing for the accurate measurement of protein abundance and their PTMs while retaining information about the location and tissue environment the sample was extracted from. Label-free MS imaging approaches like MALDI or LAESI allow for multiplexing but achieve less proteome depth compared to bulk proteomics. DVP (deep visual proteomics) uses laser-dissection microscopy to cut out single cells or tissue sections for downstream MS analysis, achieving deep proteome coverage, but the sequential, labor-intensive workflow and the low sample input per dissected region pose technical challenges for PTM analysis, such as phosphoproteomics.

The Zhang lab introduced SPOT in 2024, where proteins of (clinical) samples were labeled on-site, meaning in their native anatomical spatial arrangement, using TMT for accurate measurement of proteomes between tumor regions. However, SPOT relied on manual, pipette-based labeling, which limited its throughput and the precision of the regions that could be targeted.

In their latest preprint, the Zhang lab introduces an automated robotic platform for barcoding samples for spatial proteomics. Retrofitting a 3D printer with a nano-flow deposition nozzle, TMT labels can be deposited onto tissues at micron scale.

Schematic of SPOTTER workflow as presented in the preprint. This figure was made available under a CC-BY 4.0 international license.

Key Findings:

Using fresh-frozen murine brain sections, the authors TMT-labeled 10 anatomical regions via SPOTTER, followed by sample processing and DDA-proteomics. The technique is compatible with phospho-enrichment workflows like IMAC, allowing detection of ~8,000 proteins and ~2,500 phosphosites from the same sample. Given the spatial barcoding, these protein and PTM abundance measurements can be backmapped onto the anatomical context of the input sample. Since the whole-specimen based preparation is in principle more efficient than LMD-based approaches, the higher input allows for better enrichment and increased coverage of phosphosites.

As a proof-of-principle, the authors recovered known anatomical markers (ICMT and the glutamate transporter Slc1a6/EAAT4 in cerebellum), confirming that the spatial signal reflects real biology rather than technical artifacts. The authors then exemplified the deep dual coverage by measuring kinase activity in different brain regions, showing that phosphorylation patterns can diverge substantially from the underlying protein abundance; information that would be invisible from global proteomics alone.

What gap does this preprint address & what I like about this preprint:

I like the preprint, since it tries to solve some technical obstacles in the spatial proteomics space. First, the direct on-site labelling of tissue avoids labor-intensive LMD workflows, with the added benefit that less tissue is lost while cutting or due to ROI placement on the specimen. Given that regions of interest are defined by morphological markers, this allows for direct comparison between anatomical features and their proteome. Second, the method is, in principle, compatible with fresh-frozen and FFPE samples (as previously shown for the preceding method, SPOT), which would make implementation into clinical workflows with archival samples feasible (though this remains to be directly validated for SPOTTER itself).

Last, the biggest added benefit of SPOTTER is the phospho-dimension of the technique. Signaling pathways often underpin diseases such as cancer and this added layer of phosphoproteomic information allows for better understanding of sample pathology. Moreover, direct statistical comparisons are easier given that all samples are in the same plex, and the additional background channel allows for clean-up of unlabeled background peptides for downstream bioinformatic analysis.

Future directions & open questions:

The spatial labeling approach presented in the preprint represents a promising implementation of spatial proteomics. While previous efforts aim to extract the protein samples based on a priori spatial information, SPOTTER embeds the spatial context into the sample itself, before any extraction and abundance measurement take place. The possible integration into pre-existing imaging or spatial transcriptomics workflows is another possible advantage. For example, if regions of interest could first be defined on a Xenium (or similar spatial transcriptomics) slide, these ROI coordinates could in principle be directly exported and used to drive the printer, painting TMT labels onto the regions already characterized transcriptomically. This would allow direct deconvolution of transcript and protein/phosphosite signal from the same physical regions, and thus extend to TMA arrays run on such platforms.

Questions for the authors:

• The current resolution limit is ~90 µm with the pin-based configuration. Is this limited primarily by the mechanics of deposition (droplet formation, wetting, lateral spreading), or by the amount of peptide input needed per spot for reliable MS detection?
• Is the technique FFPE-compatible, like SPOT?
• Since SPOT was previously applied to TMA cores by manual pipetting, is SPOTTER’s automated printer compatible with TMA-format samples? If so, how would individual labeled cores be harvested and kept separate from one another after printing?
• With the emergence of 35-plex TMTpro kits, would it be feasible to hyperplex further, for instance combining several TMAs or additional specimens for in-plex comparison?
• Is it possible to use additional, unused channels for trigger peptides to enable PRM-based validation across larger sample cohorts?
• Given that the labeling is done on the tissue level, how good is the labeling penetrance for different subcellular compartments (nucleus, mitochondria, Golgi etc.)?
• Related to this, it would be interesting to see the z-depth of labeling penetrance directly, for instance using an NHS-ester-conjugated fluorophore followed by confocal imaging, or by ultrathin sectioning of a SPOTTER-labeled section followed by MS with a variable modification search for TMT.
• For the TMTpro workflow, how many phosphopeptides are recovered and what is the labeling efficiency?

Tags: barcoding, mass spectrometry, phosphoproteomics, proteomics