Cross-linking mass spectrometry, abbreviated XLMS, works by threading a chemical bridge between amino acid residues that sit close together in a folded protein or a protein complex. After digestion with trypsin, the mass spectrometer hunts for pairs of peptides joined by the cross-linker, and each confirmed pair translates into a distance constraint that can be used to model structures or map protein interaction networks. The reagent at the heart of the method consists of two reactive N-hydroxysuccinimide ester warheads that latch onto lysine side chains and protein N-termini, separated by a spacer arm. In DSS that spacer is a plain, flexible carbon chain whose strong bonds survive collision-induced dissociation. In DSSO, the chain carries two symmetric C-S bonds that snap readily under fragmentation, releasing the two cross-linked peptides as a characteristic pair of doublet ions with a fixed mass difference, a signature that specialized acquisition methods and search algorithms were designed to exploit.

Because the doublet signature promised a way to narrow the astronomically large search space that arises when spectra from complex samples are matched against thousands of candidate peptide sequences, cleavable cross-linkers such as DSSO became the presumed standard for ambitious interactome-mapping projects, while DSS was relegated to simpler structural studies. But this folk wisdom had never been put to a properly controlled test, largely because cleavable workflows come in many variants and demand expensive reagents, mass spectrometers capable of multi-stage fragmentation, and dedicated software. The Beijing-led team set out to change that. Crucially, they refused to compare a polished DSS workflow against a mediocre DSSO one. They systematically tuned the DSSO acquisition parameters on an Orbitrap Fusion Lumos instrument and confirmed that stepped higher-energy collisional dissociation at normalized collision energies of 21, 27, and 33 percent remained optimal, keeping both the signature doublet intensity and the fragment ion coverage of the released peptides at their maximum.

On the analysis side, the group built a new cleavable module into their pLink 3 search engine, which offers false discovery rate control at multiple levels from peptide pairs up to protein pairs. Benchmarked against public datasets of DSSO and the related cleavable reagent DSBU, the new module matched or exceeded the performance of Scout, a recent high-powered search engine for cleavable cross-linkers, identifying between 53 and 368 percent more cross-link spectrum matches and 7 to 54 percent more cross-linked peptide pairs. Only after both workflows had been pushed to their limits did the team stage the actual contest, analyzing samples that ranged from eight purified proteins and small multi-subunit complexes, through the 4.2-megadalton yeast ribosome with its roughly 80 proteins, up to a whole-cell lysate of the bacterium Escherichia coli containing thousands of proteins.

The results were strikingly consistent. Across every sample simpler than a mammalian cell lysate, the DSS workflow identified markedly more cross-linked peptide pairs than the DSSO workflow. On the purified proteins and complexes the advantage ranged from 21 percent for the CNGP complex to an astonishing 714 percent for the PUD1/2 heterodimer, with 1299 nonredundant cross-links found by DSS against 551 by DSSO. DSS covered 81 percent of the residue pairs that DSSO found, while DSSO recovered only 34 percent of the DSS set. On the yeast ribosome DSS still led by 41 percent, and on the E. coli lysate by 31 percent, with 598 cross-links identified. But the margin narrowed steadily as complexity grew, and at the most complex sample of all, a lysate of cultured human HEK293T cells, the trend finally reversed: using the water-soluble BS3 reagent, whose cross-linking products are chemically identical to those of DSS, the non-cleavable strategy yielded 5971 peptide pairs while DSSO delivered 8232, a 38 percent lead for the cleavable reagent.

What explains DSS’s dominance in the simpler samples? The team suspects the answer lies in geometry rather than chemistry. DSS’s spacer arm is only about 13 percent longer than DSSO’s, 11.4 angstroms versus 10.1, but in three dimensions that small difference is amplified enormously. If one end of each cross-linker is anchored and the other sweeps a sphere, DSS can reach a volume more than 40 percent larger. The effect is even bigger when flexibility is considered, because the all single-bond carbon chain of DSS bends freely while DSSO’s spacer contains a rigid sulfur-oxygen double bond. Molecular dynamics simulations, performed by anchoring one end of each reagent and modeling the accessible positions of the free reactive group, revealed that both cross-linkers sample a shell-like volume, but DSS’s shell is larger and more uniform whereas DSSO’s carries a noticeable gap. The computed shell volumes were 1060 versus 620 cubic angstroms, meaning DSS effectively interrogates nearly 72 percent more space. SDS-PAGE experiments on dimeric proteins backed this up, showing that DSS stitched dimers together far more efficiently than DSSO.

To isolate the variables cleanly, the researchers also tested two additional non-cleavable reagents engineered as surgical controls. Disuccinimidyl heptanoate, or DSH, is essentially DSS minus one carbon atom, while bis(succinimidyl) mono(ethylene glycol), or BSMEG, matches DSSO in spacer length and hydrophobicity but cannot cleave. Comparing DSS to DSH isolates the effect of arm length, DSH to BSMEG isolates flexibility and hydrophobicity, and BSMEG to DSSO isolates cleavability. Across the eight purified samples, DSH, BSMEG, and DSSO yielded only 590, 459, and 655 cross-links respectively, versus 1580 for DSS. Because mono-linked, non-cross-linked peptides appeared at similar rates for DSS and DSSO, the reactive warheads label proteins with comparable efficiency, leaving the longer, more flexible spacer as the decisive advantage.

The team then dismantled the most popular justification for cleavable reagents: that signature doublet ions narrow the search space and boost sensitivity in complex samples. In a series of elegant tests, they computationally spiked K-linked alpha- and beta-peptide ions into simulated DSS spectra and built a filtering workflow, pLink3-PepIon, that uses these ions to pre-filter candidate sequences. Filtering produced essentially no benefit, whether against a 100-protein database or the full human proteome of 20,405 sequences. They also built pLink3-doublet, which exploits DSSO’s 32-dalton doublet signature, and found that using it consistently and slightly reduced identifications compared with not filtering at all, because up to 10 percent of genuine DSSO spectra simply lack doublet ions and get wrongly discarded. Even in the scenario most favorable to filtering, searching E. coli inter-protein cross-links against a database swollen with the entire human proteome, the filtered strategy identified fewer cross-links. These findings confirm and extend a 2022 study from the Rappsilber laboratory that had first challenged the doublet dogma.

The true culprit behind DSS’s relative decline in complex samples turned out to be something subtler: fragment ion coverage, the fraction of each peptide backbone explained by observed fragment ions. Using 100,000 simulated spectra binned by coverage, the team showed that identification sensitivity climbs from about 21 percent at 24 percent coverage to essentially 100 percent at 98 percent coverage. Side-by-side tests on 705 pairs of synthetic peptides cross-linked with both reagents revealed that DSSO spectra achieved a median coverage of 100 percent for both peptides, while DSS spectra reached only 93 and 88 percent. When E. coli cross-link datasets were re-searched with tens of thousands of human protein sequences added as decoys, spectra above 85 percent coverage barely lost sensitivity, whereas those at or below 60 percent coverage collapsed dramatically, with inter-protein identifications dropping to rates as low as 72 percent for BS3 and 63 percent for DSSO. Fragment ion coverage, not cleavability itself, therefore determines how well a workflow survives search-space explosion.

The practical guidance is refreshingly concrete. For anything no more complex than a bacterial lysate, the authors recommend DSS, which is cheaper, needs simpler instrumentation, and captures more cross-links thanks to its superior reach. For mammalian whole-cell interactome mapping, DSSO currently outperforms DSS, but neither reagent achieves deep coverage, and the field will need cross-linkers that can physically separate inter-protein cross-links from the flood of uninformative intra-protein links, mono-links, and ordinary peptides. Until such ideal reagents arrive, the study points to two clear engineering goals: extend DSSO’s short, rigid spacer to access more sites, and push DSS’s fragment ion coverage higher, for instance by testing stepped collision energies against the single energy now used. It is a reminder that in proteomics, as in so much of science, the flashiest diagnostic trick is no substitute for a longer ruler and a sharper signal.

Subject of Research: Comparative evaluation of cleavable and non-cleavable cross-linkers for cross-linking mass spectrometry

Article Title: To cleave or not to cleave: a systemic evaluation of DSS versus DSSO for cross-linking mass spectrometry analysis

Article References: To cleave or not to cleave: a systemic evaluation of DSS versus DSSO for cross-linking mass spectrometry analysis. (n.d.). https://doi.org/10.1038/s44320-026-00222-9

Image Credits: AI Generated

DOI: 10.1038/s44320-026-00222-9

Keywords: cross-linking mass spectrometry, DSS, DSSO, proteomics, structural biology, protein interactions, spacer arm, fragment ion coverage, pLink 3, search space, interactome mapping, mass spectrometry