Nanoscaled, Fully Automated Hydrogen/Deuterium Exchange for Analysis of Macromolecular Assemblies Reaching the MDa Scale

Abstract

Hydrogen/deuterium exchange mass spectrometry (HDX-MS) has become an indispensable tool for structural biology, yet conventional microflow configurations limit its application to readily expressible proteins available in microgram quantities. Here, we present nanoscaled HDX, a nanoflow implementation on a commercial automated HDX platform that overcomes these limitations. We demonstrate that rapid, online microflow digestion and trapping can be effectively coupled with nanoflow separation at low temperatures to achieve unprecedented sensitivity. By integrating narrower bore tubing, optimized valve configurations, and appropriate columns, we achieved a >20-fold reduction in gradient delay volume while enhancing sensitivity by over 100-fold, maintaining equivalent performance while using just 10 ng versus 1 μg of peptide mix per injection. We demonstrate exceptional system robustness with chromatographic reproducibility below 1.5% CV for peptides, retention time shifts averaging 0.035 min, deuterium (D)-uptake measurements with SDs of 0.1 Da and improved D-retention compared to conventional microflow HDX. The drastic reduction in the amounts of protein required enabled characterization of challenging macromolecular complexes previously inaccessible to conventional HDX-MS. Using 250 fmol of the 320 kDa polycomb repressive complex 2 per injection, we achieved sequence coverage exceeding 86%, while the 0.9 MDa RNA editing catalytic complex 2, generated following one-step purification, yielded 77% coverage with 600 fmol of RNA editing catalytic complex 2 per injection. The nanoscaled HDX configuration reduces sample requirements to the sub-pmol range per injection without compromising performance relative to conventional HDX-MS, enabling the analysis of previously intractable protein systems with limited sample availability or those available only from rapid, low-yield purification protocols. Our straightforward implementation on a commercial platform eliminates the need for extensive method development, making this enhancement readily accessible and scalable to the broader structural biology community.

Graphical Abstract

Highlights

• •A commercial, automated HDX platform is adapted for nanoflow separation (nHDX).
• •nHDX enhances sensitivity by over 100-fold compared to conventional microflow HDX.
• •The new setup demonstrates improved D-retention, resulting in lower back exchange.
• •High-coverage of large, MDa-scale assemblies is achieved using fmol quantities.
• •Straightforward implementation makes nHDX available to the structural biology community.

In Brief

Conventional HDX-MS requires microgram-level samples, limiting its application to scarce protein amounts or large complexes. In this study, we introduce nano-scaled HDX, a straightforward modification of a commercial system that boosts sensitivity over 100-fold compared to conventional HDX and enhances deuterium retention. This allows robust protein dynamics analysis with fmol-level inputs, while achieving high-coverage of challenging megadalton-scale complexes, including the 320 kDa PRC2 and 0.9 MDa RECC2.

HDX-MS has emerged as a powerful analytical technique for investigating protein structure, dynamics, and interactions in solution (1). By monitoring the exchange of amide hydrogen atoms with deuterium from the solvent, HDX-MS provides valuable insights into protein conformational dynamics (2, 3), binding interfaces (4), and allosteric mechanisms (5) that are often challenging to study using traditional structural biology approaches. HDX has undergone significant technological advancements over the decades and state-of-the-art automated sample preparation workflows have transformed HDX-MS from a niche technique to an indispensable tool in structural biology, drug discovery, and biopharmaceutical development (6, 7).

Despite its strengths, conventional HDX-MS approaches face limitations when applied to challenging protein systems (8, 9). The experiments are typically restricted to one protein or complex at a time. Traditional microflow (10–100 μl/min) configurations typically require substantial amounts of purified protein, often tens to hundreds of micrograms, to achieve adequate sequence coverage (10). This constraints investigating biologically significant proteins and complexes that are difficult to isolate in such amounts. Consequently, transcription factors, membrane proteins, and complexes recalcitrant to heterologous expression remain challenging targets for HDX-MS analysis (11). In addition, while HDX-MS has been extensively applied to proteins and complexes up to 150 kDa, its application to larger systems has presented significant technical hurdles with only a few such reports appear in the literature to date (12, 13, 14, 15, 16, 17). These studies, though successful, demanded extensive optimization or instrument modifications to manage inherent sample complexity, alongside time-consuming data analysis, thereby limiting HDX-MS accessibility for routine characterization of large macromolecular assemblies.

In this study, we adapt the fluidic architecture of a conventional, commercially available, microflow HDX system to implement nanoflow separation. This shift is inspired by the transformative evolution seen in proteomics, where the adoption of nanoflow systems has enabled unprecedented analysis of complex protein mixtures (18, 19, 20) and advanced the field to single-cell analysis (21, 22, 23, 24) and comprehensive posttranslational modification profiling (25, 26, 27, 28). Following optimization of the flow path, we rigorously evaluated system capabilities through extensive benchmarking of reproducibility, sensitivity, and deuterium retention. To demonstrate its practical utility, we applied nanoscaled HDX (nHDX) as a proof-of-principle to two biologically significant macromolecular assemblies using minute sample amounts: the 320 kDa human polycomb repressive complex 2 (PRC2) and the 0.9 MDa, 14-subunit RNA editing catalytic complex 2 (RECC2) multienzyme assembly executing mitochondrial RNA editing in trypanosomes. Our new nHDX configuration represents a significant advancement in broadening HDX-MS application to protein systems that have historically remained beyond the reach of conventional approaches due to analytical sensitivity constraints and, often, limited sample availability.

Experimental Procedures

Chemicals, Reagents, and Liquid Chromatography Components

All chemicals and reagents used were purchased from Sigma, unless otherwise noted. Deuterium oxide (D₂O, 99.9 atom % D; A151882), acetonitrile (ACN, 99.9%; A998), and formic acid (FA, 99.5% purity; A117) were from Fisher and HPLC water from J.T.Baker (JT4218–3). Human recombinant histones (H3.1; M2503S and H4; M2504S) were purchased from New England BioLabs Inc. ΝanoVipers of various lengths were purchased from Thermo (75 μm, 50 μm and 20 μm internal diameter (ID)) and valves (150 μm ID bore size) were purchased from VICI Valco (V2: C82X-6670D-TRJ and V3: C85FX-6673D-TRJ). The capillary (1–10 μl/min, PN 6041.7902A) and nanoflow (0.05–1 μl/min, PN 6041.7901A) for the NCS-3500RS Nano system were purchased from Thermo.

Sample Preparation

Jurkat cells and histones

Jurkat cells were grown, lysed, and digested as previously described (29). Resulting peptides were resuspended either in 3% ACN/5% FA for optimizing different configurations and reproducibility of nHDX and in 20 mM Hepes (pH 7.5), 150 mM sodium chloride (NaCl) for deuterium exchange experiments. H3.1 and H4 were supplied in PBS (pH 7.5, sodium phosphate, dibasic (Na₂HPO₄) 10 mM, potassium phosphate, monobasic (KH₂PO₄) 1.8 mM, NaCl 137 mM, KCl 2.7 mM) at 1 μg/μl and diluted with PBS to the desired concentration for identification runs and full deuteration (FD) controls.

PRC2

Purification of the WT human recombinant PRC2 containing 5 members (EZH2, SUZ12, EED, RBAP48 and AEBP2) occurred as described in detail in (30). pFast-Bac plasmids containing each member of the complex were a gift from T. Cech. All complex members had PreScission-protease cleavable N-terminal maltose-binding protein tags and were co-expressed in Sf9 cells using the Bac-to-Bac baculovirus expression system. Cells were lysed, and the lysate clarified, loaded onto an amylose column, washed, and eluted with 10 mM maltose. Maltose-binding protein tags were cleaved off with overnight incubation with PreScission protease, and then the protein was further purified with a heparin column and size-exclusion chromatography with a Superose 6 Increase 10/300 GL column. Protein was concentrated to 25 μM, and purity was checked by SDS-PAGE and nanodrop to ensure an A260/A280 < 0.7, indicating an absence of nucleic acid contamination. The complex was supplied in 20 mM Hepes (pH 7.5), 150 mM NaCl, 1 mM Tris(2-carboxyethyl)phosphine hydrochloride (TCEP), and diluted at the desired concentration using 20 mM Hepes (pH 7.5), 150 mM NaCl, 2 mM magnesium chloride (MgCl₂), and 1 mM TCEP.

RECC2

The Trypanosoma brucei Lister 427 insect form cell line expressing CTS-tagged (31) KREN2 endonuclease was cultured in 10 to 12 L of SDM-79 medium containing 10% fetal calf serum at 27 °C until reaching a concentration of ∼4 × 10⁷ cells/ml. Mitochondrial vesicles were isolated by hypotonic lysis, DNase digestion, and differential centrifugation. The mitochondrial fraction (∼20 g wet weight pellet) was resuspended in 40 ml of buffer containing 50 mM Tris–HCl (pH 7.6), 60 mM KCl, 10 mM MgCl₂, and 0.4% Triton X-100 (final concentrations). One tablet of Complete EDTA-free protease inhibitors (Roche), 1000 U of DNase I, and 2 mg of avidin were added, and the lysate was incubated with gentle mixing for 30 min. The lysate was clarified by centrifugation at 40,000 RPM for 30 min in a SW41 Beckman rotor and incubated with 0.5 ml of Strep-TactinXT 4Flow high-capacity resin (packed volume, Iba Lifesciences) for 30 min. The lysate was transferred to a 2 ml Pierce disposable column, reloaded twice, and washed five times with 10 ml of buffer (25 mM Tris–HCl, pH 7.6, 60 mM KCl, 10 mM MgCl₂, and 0.1% Triton X-100). Proteins were eluted with three sequential additions of 0.5 ml of the same buffer supplemented with 100 mM biotin and 3 mM CaCl₂. The eluted material was exchanged into 20 mM Tris (pH 7.5), 60 mM KCl, 5 mM octyl glucoside, 10 mM MgCl₂ and diluted at the desired concentration using the same buffer.

Jurkat Peptide Identification at Different Configurations

Jurkat peptide mix (0.5 μg/μl for 1 μg and to 0.005 μg/μl for 0.01 μg injections) was injected using a CHRONECT HDX Extended Parallel sample preparation platform with lipid filtration capabilities (Trajan Scientific and Medical) coupled with an Orbitrap Exploris 480 and the NCS-3500RS Nano system (Thermo). Peptides were loaded and trapped for 1 min at 100 μl/min onto an Acclaim PepMap 300 μ-Precolumn (C18, 1 mm × 15 mm, 5 μm, 163593, Thermo) in micro configuration or onto an Acclaim PepMap 100 μ-Precolumn (C18, 0.3 mm × 5 mm, 5 μm, 160454, Thermo) in capillary and nano configurations. Peptides were separated on a Hypersil Gold C18 (1 mm × 50 mm, 1.9 μm, 25002-051030, Thermo) at 40 μl/min, nanoEase Peptide BEH C18 (0.3 mm × 50 mm, 1.7 μm, 130 Å, 186009257, Waters) at 5 μl/min, and PepSep C18 (0.15 mm × 40 mm, 1.9 μm, 1895619, Bruker) at 0.5 μl/min with gradient of 10% to 30% B (0.23% FA in ACN) applied over 6, 12, or 18 min for micro, capillary and nano configurations, respectively. Both trap and analytical columns were housed in a temperature-controlled box at 0 °C. Peptides were introduced into the mass spectrometer using heated electrospray ionization (HESI, Thermo) ion source coupled with Hi-Flow emitter (100 μm ID, OPTON-30694, Opton) for micro configuration and EASY-Spray Source coupled with EASY-Spray emitter (15 μm ID, ES994, Thermo) for capillary and nano configurations. For the micro configuration, ion source parameters were as follows, spray voltage 3.5 kV, transfer capillary temperature 250 °C, vaporizer temperature 50 °C, auxiliary gas 5, and sweep gas 1 (gas flow rates were optimized to maximize total ion intensity while ensuring robust spray stability). For capillary and nano configurations, spray voltage and transfer capillary temperature were set to 1.9 kV, and 250 °C, respectively, and the auxiliary gas was set to 0, following vendor's recommendation. For peptide identifications, samples were analyzed in data-dependent acquisition mode. Full mass spectrometry (MS) scans were acquired in the 300 to 1200 Th range, with resolution of 120,000 at 200 m/z, and RF lens set at 40%. Tandem MS (MS/MS) acquisition parameters were as follows; resolution 7,500, normalized automatic gain control (AGC) target 50% (micro) and 30% (capillary and nano), higher-energy collisional dissociation (HCD) collision energy 27 (micro) and 30 (capillary and nano), max fill time 20 ms (micro) and 50 ms (capillary and nano), and isolation window 2 m/z (micro), and 0.7 m/z (capillary and nano). Unassigned and singly charged ions were excluded from MS/MS. Data were acquired in profile mode in MS and centroid mode in MS/MS scans.

Peptide Identifications Using Spectrum Mill

Peptides were identified using Spectrum Mill Proteomics Workbench (prerelease version BI.08.02.218, Agilent Technologies). For Jurkat peptides, spectra were searched against the UniProt database (release date December 28, 2017, 65068 protein entries), including a set of common contaminant proteins (264 sequences). Spectra were allowed ±20 ppm mass tolerance for precursor and product ions, 1% protein and peptide false discovery rate, 40% minimum matched peak intensity, and “trypsin allow P” enzyme specificity with up to four missed cleavages. The instrument was ESI QExactive HCD. Carbamidomethylation was added as a fixed modification and variable modifications were acetyl (ProN-term), oxidized methionine (M), pyroglutamic acid (N-termQ), and deamidation (N). For peptide identification of histones, PRC2 and RECC2, spectra were searched using a nonspecific enzyme search with up to 5 missed cleavages. Spectra were searched against a SwissProt FASTA file (release date January 16, 2020, 23099 protein entries) including a set of common contaminant proteins (264 sequences), pepsin and protease type XIII.

HDX Labeling and Analysis

Jurkat peptides

For D-uptake experiments, peptides were resuspended in 20 mM Hepes (pH 7.5), 150 mM NaCl at a final concentration of 0.005 μg/μl (10 ng of peptide mix per injection) and for FD controls, at a final concentration of 0.025 μg/μl (50 ng of peptide mix per injection). Peptides were labeled at 20 °C for 60 s (88% D₂O) and were quenched with guanidine hydrochloride (pH 2.3, guanidine HCl, 4 M) at 1:1 (v:v). FD controls were prepared as previously described (29), transferred to the autosampler, and equilibrated at 20 °C prior to quenching as described above. MS measurements occurred in MS1 using the parameters described in the Jurkat peptide identification section for nHDX over a 6-min gradient. For D₂O measurements, RAW files were imported into HDExaminer (v3.5.0 RC 5, Trajan Scientific and Medical) using 88% D₂O content. Analysis occurred assuming that the first 1 residue of a peptide cannot hold deuterium. This loss of 1 Da parameter was chosen to reflect the rapid trapping step (1 min at 100 μl/min) for these peptides, which did not include an online digestion step. The number of deuteriums measured was not corrected for back exchange (BE).

Histones

FD controls were prepared by incubating histones for 30 min at 50 °C in deuterated PBS (pH 7.5, 83% D₂O). Samples were transferred to the autosampler and equilibrated at 20 °C prior to quenching with guanidine HCl (pH 2.3, 4 M) at 1:2 (v:v). Upon injection (800 fmol (20 ng) of H3/H4 per injection), samples were digested for 2 min at 200 μl/min using immobilized protease type XIII/pepsin column (w/w, 1:1, NBA2014002, NovaBioAssays, Woburn, MA) and analyzed using an inter-scan DIA method as described in (32). RAW files were imported into Mass Spec Studio (version 2.4.0.3667), using the AutoHX module (University of Calgary). The liquid chromatography (LC) and MS parameters were set as follows: MS1 and MS/MS mass tolerance of ±10 ppm with extracted ion chromatogram integration range of 0.2 min and retention time variance of ±0.2 min. Minimum number of fragments was set to 3, and maximum % deuteration SD was set to 5. The deuterium content of the peptides was calculated assuming loss of 2 Da from the N terminus of the peptides (N-P-2 enabled). Only peptides accepted across all replicates were used in the FD evaluation.

PRC2

For HDX analysis of PRC2, samples were labeled at 20, 100, and 1000s (83% D₂O content) and at 20 °C using either 250 fmol (80 ng) or 75 fmol (24 ng) of PRC2 per injection. Samples were quenched with guanidine HCl (pH 2.3, 4 M) at 1:2 (v:v) and digested for 3 min at 200 μl/min using the protease described for histones. MS measurements occurred as described above for Jurkat peptides. RAW files were imported into HDExaminer, using 83% D₂O content and assuming loss of 2 Da from the N terminus of the peptides. The number of deuteriums measured was not corrected for BE.

RECC2

For HDX analysis of RECC2, we rewired the plumbing of nHDX as depicted in Supplemental Figure S1A to enable backflushing of the trap column and performed peptide separation at 1 μl/min over a 12-min gradient (total run time 21 min, backflushing started at 18.3 min). This configuration only slightly increased the void volume of the system from 1.9 μl to 2.5 μl, allowing for peptides starting eluting at 2.5 min. Samples were labeled (600 fmol (550 ng) of RECC2 per injection) and quenched as described above for PRC2 with the exception that samples were incubated for 1 min in the loop to allow for efficient denaturation of the complex. Samples were analyzed in MS1, and RAW files were imported and processed into HDExaminer using 83% D₂O content and assuming loss of 2 Da from the N terminus of the peptides. Chiclet plots were generated using HDgraphiX (Vahidi lab, University of Guelph) and upon correction for 20% BE.

Experimental Design and Statistical Rational

Jurkat peptide samples were analyzed across all tested configurations, including varied gradients and columns, using three processing replicates per condition. For inter- and intra-day reproducibility assessments, Jurkat peptides were analyzed using 10 processing replicates within each day across 3 days. Jurkat (60 s and FD controls) and histone (FD controls) samples were independently labeled and analyzed as five processing replicates each. Biological replicates were not included given the methodological focus of this study. D-uptake data are presented with confidence intervals calculated using Welch's unequal variances t test, considering all available replicates for the optimal charge state (defined as the charge state with the most complete set of replicates). Statistically significant differences in D-uptake between microflow and nanoflow HDX configurations were determined using a paired t test (p < 0.0001). PRC2 and RECC2 samples were labeled independently at each time point and analyzed using four processing replicates.

Results

Modifications of a Fully Automated Micro-HDX System to Enable Capillary and Nanoflow Configurations

The HDX-MS platform utilized in this study integrates the commercially available Chronect HDX Extended Parallel sample handling solution (Trajan Scientific and Medical) (Fig. 1A) with the NCS-3500RS Nano system and an Orbitrap Exploris 480 (Thermo Fisher Scientific). A schematic representation of the main components of the Chronect system is shown in Figure 1A, depicting the temperature-controlled compartments. The LC-box operates at 0 °C and houses three valves: a six-port injector valve (V1) that switches between sample injection and loading through the digestion column, a ten-port switching valve (V2) that facilitates loading of digested sample onto a trap column, and a six-port selector valve (V3) that enables sample separation on the analytical column. The three-valve configuration allows for both forward and reverse directions of flows across all columns, ensuring effective sample processing while preventing carryover through backflushing procedures. The conventional micro configuration commercially available yields a total void volume of ∼63 μl (calculated from the gradient pump-to-MS flow path) during the peptide separation step (Fig. 1B, Supplemental Table S1) and requires high flow rates (typically 20–40 μl/min) for fast elution of peptides to prevent deuterium from exchanging back to hydrogen, a process known as BE. In microflow mode, we utilize a flow rate of 40 μl/min resulting in peptide elution starting in <2 min (Fig. 1C).

Fig. 1.

HDX valve configurations and respective TICs.A, schematic of the CHRONECT HDX extended parallel with lipid filtration; separate compartments are labeled. The LC box operates at 0 °C and contains three valves (V1, V2, and V3). B, micro-, capillary-, and nano-configurations within the LC box and gradient flow rates employed. Only V2 and V3 are shown; plumbing of V1 remained consistent across all flow configurations. Void volume refers to the total volume calculated from the Dionex gradient pump-to-MS flow path. V, valve; T, trap; and C, column (Table S1). C, total ion chromatograms and gradients employed for the different configurations (6 min effective gradient). The dashed line indicates V2 and V3 switching that enables backflushing of the C-T in micro configuration or T in capillary configuration. For microflow, an Acclaim PepMap trap cartridge (1 mm (ID) × 5 mm (length), 5 μm, Thermo Fisher Scientific) was coupled with a Hypersil GOLD C18 (1 mm × 5 cm, 1.9 μm, Thermo Fisher Scientific). For capillary and nanoflow, a PepMap trap cartridge (5 mm × 300 μm, 5 μm, Thermo Fisher Scientific) was coupled with a nanoEase BEH C18 (5 cm × 300 μm, 1.7 μm, Waters Corporation) and PepSep Series (4 cm × 150 mm, 1.9 μm, Bruker Daltonics), respectively. D, precursor extracted ion chromatograms of selected peptides across different HDX configurations. The average full-width-half-maximum (FWHM) calculated from all peptides identified within a run is indicated for each configuration. ID, internal diameter; LC, liquid chromatography; HDX, hydrogen/deuterium exchange; MS, mass spectrometry.

To enable capillary (1–10 μl/min) and nano (0.1–1 μl/min) flows, we made several modifications: First, we bypassed the precooler and connected the pump outlet directly to V3 (capillary) and to V2 (nano) using narrower ID nanoVipers (50 μm and 20 μm, respectively). While analytical columns used in micro configurations are typically compatible with backflushing procedures, which is not the case for the columns we employed here for the capillary and nano configuration, therefore we placed those downstream of V3 to only allow for forward flow direction and prevent their damage. Next, we replaced the microbore valves (250 μm ID bore size) used in the micro and capillary configurations with narrower ID valves (150 μm ID bore size). The gradient transfer lines were also reduced from 75 μm in micro and 50 μm in capillary to 20 μm ID in nano configuration. Additionally, the trap inlet and outlet lines in the nano configuration were substituted with 50 μm ID tubing. The postanalytical column transfer line was also reduced from 50 μm × 750 mm (ID x length) in micro to 20 μm × 350 mm in both capillary and nano configurations, resulting in a drastic reduction of the postanalytical column volume (2.5 μl for the micro versus 0.12 μl for the capillary and nano, Supplemental Table S1). We transitioned from the HESI to the EASY-spray ion source (Thermo Fisher Scientific) for improved ionization efficiency in both capillary and nanoflow regimes.

Collectively, these modifications decreased the void volume from ∼63 μl in the micro to 7.8 μl in capillary and 1.9 μl in nano configuration, corresponding to ∼5-fold (4.1 μl) and 20-fold (1.1 μl) reductions in the gradient delay volume (21.8 μl in micro), respectively.

Performance Evaluation of the Different Configurations

To evaluate the three configurations, we coupled commercially available columns optimized for their respective flow rates and injected 1 μg of tryptic digest from Jurkat cells. In all configurations, peptides were loaded and desalted on the trap column at 100 μl/min for 1 min and separated over a short gradient (6-min) typical for HDX-MS analysis. The gradient conditions and resulting total ion chromatogram (TIC) traces for each configuration are depicted in Figure 1C. Peptide elution in the micro (40 μl/min) and capillary (5 μl/min) began at ∼1.8 min, while the nano (0.5 μl/min) configuration showed a delayed start (∼3.9 min), still well within the acceptable range for HDX applications. Micro and capillary configurations demonstrated similar separation performance with an average full-width-at-half-maximum of ∼9 to 10 s, while nanoflow separation provided significantly sharper peaks with an average full-width-at-half-maximum of ∼5 s (Fig. 1D, Supplemental Fig. S1B, Supplemental Table S2).

The micro configuration identified ∼2700 unique peptides, nearly double the number identified using an Orbitrap HF under identical conditions (29). The capillary and nano configurations exhibited progressively greater sensitivity, identifying 4000 (1.5-fold more) and 4950 (1.8-fold more) unique peptides, respectively (Fig. 2A). This performance advantage at lower flow rates increased substantially with longer gradients; at 12 min, peptide identifications increased by 2- and 2.5-fold, respectively, and at 18 min, both configurations achieved a ∼4-fold increase (∼9500 peptides). The superior sensitivity of the nano configuration becomes even more apparent at lower sample loads. For instance, a 50 ng of peptide mix per injection yielded nearly three times the peptide identifications of the capillary configuration (3250 unique peptide in nHDX versus 1130 unique peptides in capillary HDX, data not shown). Notably, when injecting 100-fold less material (10 ng of peptide mix per injection), the nano configuration matched or exceeded the peptide identification performance of the micro configuration across all gradient lengths tested, indicating that experiments can be conducted with remarkably reduced sample requirements.

Fig. 2.

Peptide and protein metrics at different configurations.A, number of unique peptides identified from a Jurkat digest (1 μg of peptide mix per injection for all configurations, with an additional 10 ng injection for the nHDX setup) across different HDX configurations and gradient lengths (6, 12, and 18 min). Error bars represent mean values ± SD (n = 3) (Supplemental Table S2). Trap columns used: Acclaim PepMap trap cartridge (5 mm × 1 mm, 5 μm) for micro and PepMap trap cartridge (5 mm × 300 μm, 5 μm) for capillary and nano configurations. Analytical columns used: Hypersil Gold (5 cm × 1 mm, 1.9 μm) for micro, nanoEase (5 cm × 300 μm, 1.7 μm) for capillary and PepSep Series (4 cm × 150 μm, 1.9 μm) for nano configurations. B, protein intensities (log10) identified using a 6-min gradient, ranked from highest to lowest within each configuration (left panels). Proteins identified in each configuration are superimposed on proteins identified using a 2-h gradient and a standard nanoLC-MS/MS method on the same MS (Exploris 480) (right panels). Protein intensities are ranked from highest to lowest for the nanoLC-MS/MS results and are shown here for reference (green, 6000 proteins). Proteins with unique peptides ≥2 are shown (Supplemental Table S23). C, number of unique peptides identified in nHDX using different columns and a 6-min gradient (10 ng of peptide mix per injection). Error bars represent mean values ± SD (n = 3). HDX, hydrogen/deuterium exchange; nHDX, nanoscaled HDX; MS, mass spectrometry.

The improved peptide detection translated to significant increases in protein identifications across configurations, with a corresponding enhancement in overall protein intensity (Fig. 2B, left panels, Supplemental Table S3). Nevertheless, a direct comparison of the measured protein intensity distributions from the HDX-length gradients can be misleading. Due to less efficient ionization at higher flow rates, the entire measured intensity range for the microflow analysis is compressed and shifted downward (log10 values of ∼6–11) compared to the nanoflow analysis (log10 values of ∼8–12). This artifact gives the false appearance of more proteins in the microflow's lower-intensity tail. To provide a more accurate assessment of sampling depth, we compared the intensity of proteins identified in each HDX configuration to those identified using a proteomics “gold standard” 2-h gradient with nanoLC-MS/MS on the same instrument (Fig. 2B, right panels). When mapped onto this “true” abundance distribution, we observed that proteins from the micro configuration (535 proteins) correspond almost exclusively to the most abundant proteins in the cell, spanning ∼2 orders of magnitude in intensity. This finding aligns with our previous observations using the same micro configuration paired with an Orbitrap HF (29). Notably, decreasing the flow rate while simultaneously increasing sensitivity in the capillary and nano configurations enabled the detection of a higher number of proteins (678 and 899, respectively) across a substantially wider dynamic range of protein abundances—extending beyond 3 orders of magnitude—suggesting significantly enhanced sampling depth and more comprehensive proteome coverage.

Building on these findings, we next assessed whether further gains in sensitivity could be obtained in the nHDX by testing different column geometries across varying gradient lengths (6, 12, and 18 min). Based on the pump pressure limitations (900 bar) and nanoflow controller constraints (0.1–1 μl/min), we selected three columns with different geometries suitable for our system (PepSep Series, 1.9 μm, Bruker Daltonics, Fig. 2C, Supplemental Table S4). We evaluated these columns by injecting 10 ng of peptide mix per injection and assessing key metrics including TIC and pressure profiles, starting elution time, peak widths, and peptide identifications (Supplemental Fig. S2). For the 6-min gradient, the shortest column identified the highest number of peptides (∼2700 unique peptides), while longer columns yielded slightly fewer (∼2300 unique peptides). With the 12-min gradient, all columns performed similarly, yielding ∼4000 unique peptides. Interestingly, extending to an 18-min gradient did not improve results for the 4 cm column but increased peptide identification for the longer columns, yielding ∼4500 to 5000 unique peptides. The highest number of identifications was achieved with the narrower ID column (15 cm × 75 μm) at a 0.3 μl/min flow rate on an 18-min gradient. Extending the gradient to 30 min only marginally increased peptide identifications (∼5300 unique peptides, data not shown) primarily due to peak broadening that resulted in lower signal-to-noise ratio, demonstrating no additional benefits. Notably, the operating pressure during analytical separation reached ∼680 bar, with elution beginning at 6.3 min (compared to ∼4 min with the other columns), resulting in a total elution time of ∼23 min (Supplemental Fig. S2). This approaches the limit (∼25–30 min) for gradient length in HDX applications necessary to maintain acceptable BE levels. Based on these results, we selected the 4 cm × 150 μm column for downstream HDX experiments using shorter gradients (6 and 12 min).

Assessing nHDX Reproducibility and BE

To assess intra- and inter-day reproducibility of our system, we then reconstituted Jurkat peptides in Hepes (pH 7.5, 150 mM NaCl) and quenched them with guanidine HCl (4 M, 1:1 v:v) prior to injecting them for analysis to mimic conditions employed in a standard HDX experiment. We performed 10 consecutive injections of 10 ng of peptide mix per injection within a day, across 3 days. Peptides were loaded for 2 min at 200 μl/min which are our typical conditions for the digestion/desalting steps and separated using the 6-min gradient shown in Figure 1C with the following modification: at the end of the “wash” gradient at 10 min, we increased flow from 0.5 to 1.5 μl/min to accelerate analytical column re-equilibration. This allowed 6 μl of mobile phase (∼8 column volumes) to flow through the analytical column (0.7 μl column volume). An additional 1.5 μl passed through during quenching-injection-desalting steps of the next sample, totaling >10 analytical column volumes before starting the next gradient. This rapid re-equilibration enabled 10 consecutive injections within a 3-h timeframe while maintaining exceptional system stability, with pressure increasing by only 200 bar at the higher flow rate during re-equilibration (reaching a total of 320 bar), demonstrating remarkable robustness with minimal pressure fluctuation throughout repeated analytical cycles.

An average of ∼2550 peptides were identified across the 3 days (using a standard 6 min gradient, Experimental Section, Fig. 3A), which is slightly lower than the ∼2700 shown in Figure 2A as a result of the increased desalting volume employed here (400 μl compared to 100 μl), which resulted in the loss of hydrophilic peptides to waste during trapping. Intra-day CVs were calculated based on peptide identifications and were 1.1%, 0.7%, and 1.4% for days 1, 2, and 3, respectively (n = 10), and the inter-day CV was 1.1%, demonstrating excellent reproducibility, corroborated further by overlapping TICs and pressure profiles (Fig. 3B, Supplemental Fig. S3, A and B). The chromatographic reproducibility was assessed by calculating the average SD of retention times for 1488 peptides consistently identified across all 10 replicates on day 1 (Supplemental Table S5A). The resulting average standard deviation was 0.035 min, demonstrating highly reproducible chromatographic performance (Supplemental Table S5B).

Fig. 3.

nHDX reproducibility and back exchange.A, unique peptides identified from a Jurkat digest (10 ng injection, 10 injections per day). Peptides were reconstituted in Hepes (pH 7.5, 150 mM NaCl) and quenched with guanidine hydrochloride (2 M final concentration, final pH 2.5) to mimic HDX conditions. Error bars represent mean values ± SD (n = 10). B, total ion intensities of replicate injections 1, 5, and 10 (day 1). Intensities and pressure profiles for all replicate injections on day 1 are given in Supplemental Fig S3. C, D-uptake of a representative subset of the 290 most intense Jurkat peptides identified to precisely assess measurement variance (labeled for 60 s at 20 °C, 10 ng of peptide mix per injection). Back-exchange corrections have not been applied. Error bars represent mean values ± SD (n = 5) (Supplemental Table S6). D, comparison of full deuteration (FD) control values of common Jurkat peptides reported between our previous study (29) using the micro configuration (1 μg of peptide mix per injection) and in this study using nHDX (0.05 μg of peptide mix per injection). Statistically significant differences (paired t-test) are detected between the two configurations (p < 0.0001). FD values for H3 and H4 are also shown. For Jurkats, a loss of 1 Da and for H3/H4 a loss of 2 Da is considered from the N terminus. Whiskers extend to the 2.5th and 97.5th percentiles, encompassing 95% of the data distribution. Any data points beyond these whiskers are plotted individually as outliers. E, volcano plot showing peptides measured with the microHDX and nHDX. Significant differences are determined by multiple unpaired Welch t-tests (p < 0.05) and multiple testing correction using the two-stage linear step-up procedure of Benjamini, Krieger, and Yekutieli. HDX, hydrogen/deuterium exchange; nHDX, nanoscaled HDX.

We next evaluated the reproducibility of D-measurements by subjecting the same mixture of peptides to deuterium (88% D) for 60 s at 20 °C and measured the D-content for a representative subset of 290 peptides (10 ng of peptide mix per injection, 5 replicates, Fig. 3C, Supplemental Table S6). Under these labeling conditions at pH 7.5 D-values were between 60 and 80% for the majority of peptides, with the full range spanning from 12% (WNTDNTLGTEITVEDQLAR) to 86% (GILNGQK), providing a broad spectrum for assessing reproducibility; the average SD of D-uptake was 0.1 Da.

To next evaluate the extent of BE upon D-retention in nHDX, we generated fully deuterated tryptic peptides (50 ng of peptide mix per injection) prepared similarly as previously described (29). We filtered our identified peptides based on the list for which we had previously measured FD values using the microHDX configuration (564 peptides) (29). Of these 564 peptides, 367 were identified in the current study. Following stringent manual curation for D-retention, we successfully detected and quantified FD values for 245 peptides common between nHDX and microHDX (Supplemental Table S7), with the remainder excluded due to poor or overlapping signals. This comparison is notably conservative, as the nHDX analysis used 20-fold less material (50 ng) than in the microHDX study (1 μg). Furthermore, this same 50 ng nHDX run yielded robust FD values for an additional 1252 unique peptides, highlighting the system's enhanced sensitivity. Overall, we obtained robust FD values for 52% of all peptides identified in nHDX, a substantial increase over the 35% obtained in the microHDX study.

Interestingly, D-retention was higher in nHDX, with average retention values of 76% versus 70% in microHDX (accounting for loss of 1 D from the N terminus, Fig. 3D). A clear pattern emerged with peptides in the nHDX system exhibiting up to 10% higher D-retention, though these general differences did not reach statistical significance. However, significant differences were observed for 73 peptides, some demonstrating up to 30% higher D-retention in nHDX (Fig. 3E), whereas only two peptides exhibited statistically significant higher D-levels in the microHDX configuration.

To assess system-wide D-retention with a realistic substrate, we used FD controls of intact histone H3.1 and H4. As these monomers contain both structured histone-fold domains and disordered tails (33), they provide a more representative measure of BE across the entire workflow—including the online digestion step—than a predigested peptide mix. We injected 400 fmol (10 ng) per histone per injection (5 replicates, Fig. 3E, Supplemental Table S8). Following online digestion with pepsin:protease XIII (1:1, NovaBioAssays) for 2 min at 200 μl/min, we measured 97 peptides across both proteins, achieving nearly complete coverage (96% for H3 and 99% for H4). Average D-retention levels reached 75% (73% for H3 and 77% for H4) when accounting for a loss of 2 D and 68% when considering loss of 1 D (66% for H3 and 70% for H4) from the N terminus.

nHDX Enables the Analysis of the Five-Member PRC2 (320 kDa) Using Nanogram Quantities of Protein

To benchmark the performance of nHDX on a large, multisubunit assembly, we used the PRC2 complex to define the material requirements for achieving comprehensive protein coverage. PRC2 catalyzes methylation of histone H3 lysine 27, an epigenetic marker associated with gene silencing whose dysregulation is implicated in cancers and developmental disorders (30). The PRC2 complex comprises five proteins (EZH2, SUZ12, EED, RBBP4, and AEBP2), with a unique sequence of 2859 amino acids and a molecular mass of ∼320 kDa, exceeding the conventional size range for HDX applications (<150 kDa). Samples of the complex were quenched and digested as described above for histones without deuterium exchange, and peptides were separated using a 6-min gradient resulting in the identification of 1617 unique peptides. We next performed deuterium labeling across three time points (20, 100, and 1000 s) at 20 °C, utilizing 250 fmol (80 ng) of PRC2 per injection and analyzed samples in quadruplicate.

After rigorous inspection and comprehensive manual validation of the data, we applied filtering criteria to select only peptides with D-uptake measured consistently across all time points, yielding 942 peptides with complete deuterium incorporation data throughout the experimental time course (Fig. 4, Supplemental Table S9). Our analysis yielded an exceptional 86% sequence coverage across all PRC2 subunits with an average redundancy of 5.6. Coverage for four of the five subunits ranged from 86 to 97%, whereas AEBP2 exhibited reduced coverage (67%) due to limited detection of its highly acidic N-terminal region. The PRC2 subunits displayed diverse D-uptake kinetics: EZH2 and SUZ12 demonstrated high D incorporation from early time points, indicating their inherently dynamic nature, whereas EED and RBBP4 showed prolonged protection from exchange, consistent with more stable, globular conformations. For the complete experiment, a remarkably low total protein requirement of just 5 pmol (1.6 μg) was sufficient; this minute amount encompassed all identification and labeling runs as well as the void volumes necessary during automated sample aliquoting and transfer steps.

Fig. 4.

Dynamics of PRC2 using nHDX. PRC2 cryo-EM structure (PDB: 6c23) (41) with colored individual subunits. D-uptake (%) of peptides measured for each subunit (subunit labels match the color on structure) at three time points using a 6 min effective gradient. The complex has a total mass of 320 kDa with a unique sequence of 2859 amino acids. HDX-MS analysis was performed in quadruplicates (250 fmol or 80 ng per injection), yielding 942 unique peptides, 86% sequence coverage, and 5.6 redundancy. Only peptides that were detected across all time points are shown. D-uptake values are not corrected for back exchange (Supplemental Table S9). HDX, hydrogen/deuterium exchange; nHDX, nanoscaled HDX; PRC2, polycomb repressive complex 2.

To further investigate whether comparable results could be achieved with even smaller sample quantities, we conducted an identical experiment using ∼3-fold less material (88 fmol or 28 ng of PRC2 per injection), resulting in a total protein consumption of ∼1.5 pmol (0.5 μg) for the entire experiment (identification and labeling runs as well as the void volumes necessary during automated sample aliquoting and transfer steps). Following manual validation, we quantified D-uptake across 693 peptides at all time points, achieving 81% sequence coverage with an average redundancy of 3.7 (Supplemental Fig. S4, Supplemental Table S10). Consistent with our previous findings, four of the five subunits maintained high coverage ranging from 80% to 96%, while AEBP2 coverage decreased to 56%. The deuterium incorporation patterns for each subunit closely paralleled those from the higher-concentration experiment; a direct statistical comparison using multiple unpaired t-tests confirmed that only a minor fraction of peptide pairs (∼1.9%) showed significant differences, substantiating the high, long-term reproducibility of our results across different sample amounts, especially given the 4-month interval between the experiments.

All samples were run consecutively without blank injections between replicates or time points, with complete sample preparation and data acquisition accomplished within 5 h using the CHRONECT system. Notably, despite the absence of backflushing capabilities in the nHDX configuration, no sample carryover was observed.

Analysis of MDa-Size RECC2 by nHDX

To challenge our nHDX and expand the boundaries of mass analysis at high coverage, we then investigated RECC2 complex, which performs an enzymatic cascade of U-insertion/deletion mitochondrial RNA editing in T. brucei. This process converts pseudogene transcripts into mature mRNAs that can be translated into proteins essential for mitochondrial function and parasite survival (34). This macromolecular assembly consists of 14 subunits including four RNase III–like proteins (active endonuclease N2 and pseudoenzymes B4, B5, and B7), the terminal uridylyl transferase T2, RNA ligases L1 and L2, and X2 exonuclease, and six OB-fold proteins (A1–A6), as shown in Figure 5A (35). With a total mass of ∼0.9 MDa and a unique sequence of 6319 amino acids, RECC2 provides an ideal test case for nHDX.

Fig. 5.

Dynamics of RECC2 using nHDX.A, metrics for the 14 RECC2 subunits identified in this study. B, chiclet plots for all subunits identified in this study (see Supplemental Table S11 for detailed information). HDX-MS analysis was performed in quadruplicates (1 pmol per injection), yielding in total 851 unique peptides, 77% sequence coverage and 2.1 redundancy. Only peptides that are detected across all three time points are shown. Peptide profiles for X2 are shown in two lines due to space restrictions. D-uptake values were corrected for 20% back exchange.

We maintained the same digestion and desalting conditions described above for histones and PRC2, with the following modification: upon quenching, we incubated the complex for 1 min in the injection loop to enable more efficient denaturation and subsequent digestion. Peptide identification yielded 2169 peptides belonging to RECC2 subunits. In addition, 971 peptides were identified from methylcrotonyl-CoA carboxylase, a known contaminant in RECC2 purification with strong binding affinity for Strep-Tactin. Taking the ratio of RECC2 peptide spectrum match intensities to the peptide spectrum match intensities of all peptides identified, we calculated that RECC2 comprised 61% of the total protein amount. Therefore, we report injected amounts based on both total protein content and RECC2-specific content, accounting for its 61% fraction in the sample.

We performed labeling experiments for the same 3 time points reported for PRC2 in quadruplicates (900 ng of total protein content or 600 fmol (550 ng) of RECC2 per injection). Upon manual inspection and validation of the D-uptake, we filtered peptides that were consistently identified across all time points. The number of peptides that successfully passed our filtering criteria and yielded complete D-uptake profiles is illustrated in Figure 5B, alongside comprehensive sequence coverage maps for each individual protein subunit. In total, we measured 851 unique peptides, yielding an impressive 77% overall sequence coverage with 2.1 redundancy across the complex (Supplemental Table S11). The D-uptake profile at the peptide level for each subunit is comprehensively depicted in Figure 5B. Core subunits appear to possess both highly stable and remarkably dynamic regions, with the C-terminal domains of all subunits exhibiting particularly dynamic behavior throughout the experiment. The D-uptake profiles of the wing subunits reveal extensive dynamic regions that incorporate high deuterium content from the earliest measured time point. Although the apparent molecular weight on native gel (∼1 MDa) consistently appeared too large to assume one subunit per complex, suggesting shared copies, no bimodal distributions were detected throughout the time course. This observation suggests that under the experimental conditions employed in our study, either the different subunit copies are interchanging faster than the H/D exchange rate, or the H/D exchange itself was too rapid to capture such distinct populations. The remaining functional subunits proceeded with mixed patterns of exchange that varied considerably depending on their structural context and position within the complex.

Discussion

Implementing nanoflow LC-MS for HDX analysis has historically faced significant barriers, including concerns about chromatographic performance under quench conditions, deuterium retention, and overall system robustness (13, 36, 37). In this study, we have successfully overcome these technical challenges by modifying a commercial HDX platform to enable nanoflow separation. By implementing narrower bore transfer lines, optimizing valve configurations, and selecting appropriate columns, we achieved a >20-fold reduction in gradient delay volume while preserving critical HDX workflow elements. We maintained wide bore tubing in the digestion/desalting flow path (except for 50 μm ID nanoVipers before and after the trap in nHDX) to ensure efficient salt and denaturant removal, with the system demonstrating exceptional stability over a 6-month operational period despite the consistent use of high salt, TCEP and guanidine HCl concentrations in sample buffers. While demonstrated here on a specific commercial platform, the underlying methodology—combining low-volume fluidics with a nanoflow meter and a nanoESI source—is broadly applicable to virtually any nanoLC-MS setup. The implementation is straightforward as it relies on universal fluidic principles and commercially available components; for labs with LC expertise, swapping valves and capillaries is a routine task. While a dedicated system is optimal for routine nHDX, some modern commercial platforms are now designed with multi-flow capabilities, simplifying the process of switching between microflow, capillary-, and nanoflow-regimes. Beyond the technical implementation, the practical cost of conversion should also be considered. While the required fluidic components like low-volume valves and capillaries are a minor expense, the major investment is in the core hardware. For labs not already equipped for nanoflow proteomics, a dedicated nanoflow pump and a nano-electrospray ion source are essential for robust performance and represent the most significant costs.

The optimized nHDX system enhances sensitivity by over 100-fold compared to conventional microflow configurations, maintaining equivalent performance with just 10 ng versus 1 μg of peptide mix per injection and excellent chromatographic reproducibility. This dramatic reduction in sample requirements, combined with fast re-equilibration that enables processing up to 80 samples per day, makes nHDX ideal for high-throughput analysis of previously inaccessible protein systems with limited availability and could potentially enable alternative protein production and enrichment strategies that currently remain underexplored (17, 38).

Our evaluation of D-retention demonstrated that nHDX exhibits enhanced efficiency compared to conventional microHDX in preserving deuterium. Using Jurkat peptides, we observed an average D-retention increase of 6% in nHDX compared to our previous microHDX configuration, with some peptides showing up to a remarkable 30% improvement (Fig. 3D). For histones, our average BE rate was 24% for structured peptides, notably better than the average 32% BE reported in an interlaboratory study when using HDX automation (39), demonstrating improved performance. Of note, peptides from the intrinsically disordered N termini showed higher BE values (∼35%) due to increased solvent accessibility and lack of stable hydrogen bonding networks, consistent with our previous observations (3, 33, 40). The enhanced deuterium preservation in nHDX is likely attributable to the gentler ionization conditions of the EASY-spray source compared to the HESI source. The HESI source employs a heated nebulizer gas to enhance desolvation, which is effective for ionization at higher flow rates. However, this process introduces significant heat; the HESI source maintains a minimum temperature of ∼30 °C even when set to 0 °C. This warm environment, combined with high gas flow, can expose peptides to residual water vapor and increase gas-phase BE. In contrast, the EASY-spray source operates at a lower temperature and under tightly controlled low-flow conditions, which minimizes these sources of deuterium loss. While this hypothesis aligns with our observations, it has not yet been systematically tested. These findings challenge the prevailing concern that nanoflow separation might compromise D-retention due to longer chromatographic run times, instead demonstrating that well-optimized nanoflow systems can provide enhanced D-preservation.

Historically, HDX-MS has been extensively applied to proteins and complexes up to 150 kDa, but its application to larger systems has presented significant technical hurdles. Despite these challenges, several pioneering studies have successfully applied HDX-MS to larger assemblies, including the 320 kDa PRC2 chromatin remodeling complex (16), the 400 kDa proteasome (15), the 469 kDa DNA-PKc (13, 17), the 525 kDa membrane-bound FoF1 ATPase (12), and the ribosome in complex with GCN2 (14). These applications, though successful, demanded extensive optimization or instrument modifications to manage inherent sample complexity, alongside time-consuming data analysis, thereby limiting HDX-MS accessibility for routine characterization of large complexes. We address these limitations by developing a more streamlined approach that eliminates the need for extensive method development while maintaining high-analytical throughput.

Approximately 30 to 40% of the PRC2 complex was either unresolved or represented at substantially lower resolution in the pioneering cryo-EM study (41). These missing regions included large portions of the AEBP2 zinc finger domains, multiple flexible loops within EZH2, and significant regions of SUZ12's C-terminal domain. Using typical quenching conditions and a short gradient, we achieved exceptional 86% sequence coverage by nHDX, while our highly reproducible chromatography enabled streamlined data analysis requiring less than 2 days of manual validation. Our data reveal that previously unresolved regions often correspond to intrinsically disordered segments and dynamic interfaces critical for allosteric regulation and substrate recognition, providing valuable complementary insights into this macromolecular assembly's functional mechanisms (41, 42). Our approach required only 250 fmol (80 ng) of PRC2 per injection, representing a 10-fold reduction from a previous study that used ∼0.8 μg per injection with a custom-designed nanospray system (16). The nHDX platform's improvement over our previous, conventional microflow is even more striking, generating nearly three times as many peptides (1617 in nHDX versus 595 in microHDX, data not shown) from 25-fold less material (0.25 pmol in nHDX versus 6 pmol in microHDX). Remarkably, we maintained excellent coverage (81%) even when further reducing the sample amount to 88 fmol (28 ng) of PRC2 per injection, resulting in total sample consumption of ∼0.5 μg for the entire experiment. This significant enhancement in sensitivity while achieving high coverage directly addresses a major limitation in studying large macromolecular assemblies, which are often available only in limited quantities.

The ultra-low sample requirements are particularly valuable for RECC2, which is challenging to purify from parasites in large quantities. Crystallographic (43, 44), negative stain electron microscopy (45) and molecular (46) studies suggested that RECC complex has a slim triangle-like architecture which encompasses conserved enzymatic domains, while a substantial part (40–50%) of the protein content is disordered. Using just 12 pmol (∼11 μg) of RECC2 for nHDX analysis, we achieved 77% sequence coverage, revealing detailed dynamics across the subunits. Our analysis confirms that most of the complex is inherently disordered and exposed to solvent and demonstrates nHDX's complementary value for understanding dynamic behavior in such large macromolecular assemblies (43, 44, 45, 46). This sensitivity will allow us to probe the structure and dynamics of editosome variants that have thus far been inaccessible due to sample complexity and limited availability.

The practical implications of these low sample requirements become even more apparent when considering a typical comparative experiment. For the single-state analyses reported here (three time points in quadruplicates), we consumed a total of ∼5 pmol of PRC2 and 12 pmol of RECC2. This suggests that a comparative study, such as comparing an apo and ligand-bound state with the same number of replicates, would require only ∼10 pmol of PRC2 or ∼24 pmol of RECC2. Such modest sample requirements place functional structural studies of large, challenging macromolecular assemblies firmly within the reach of nHDX. Another practical implication of these low sample requirements is the mitigation of sample carryover. In conventional microflow HDX-MS, carryover often originates from the immobilized protease column. However, by operating in the subpicomole range, the risk of column saturation is significantly reduced, simplifying the workflow by obviating the need for routine backflushing.

During optimization of nHDX, we encountered several practical challenges worth noting. When working with nanogram quantities of protein, we observed signal loss due to protein adsorption on vial surfaces during labeling, particularly for longer time points (>2 h). This issue should be effectively mitigated by utilizing low-bind vials throughout the workflow, highlighting the importance of surface chemistry considerations when handling limited sample amounts. Regarding chromatographic performance, column selection proved challenging. We found PepSep series columns (4 cm and 15 cm) to be the only commercially available options during optimization that delivered acceptable chromatographic separation under our specific flow rates at 0 °C. Additionally, we were not able to find suitable ultra-short (<5 mm) trap cartridges for the desalting step to eliminate the void volume further. The availability of intermediate length analytical columns or short traps would be crucial for further improving signal-to-noise ratios and overall chromatographic performance by minimizing peptide dispersion during trapping and transfer to the analytical column. As nHDX gains adoption, we anticipate the development of additional column options specifically optimized for this application, further enhancing performance and flexibility. Looking forward, while the development of commercially optimized nHDX traps and analytical columns would be a welcome advancement, their current absence does not detract from the overwhelming advantages of the nHDX platform. The ability to work with subpicomole quantities of protein fundamentally changes the scope of tractable biological questions, making the sensitivity gains of the nanoflow regime indispensable. The ultra-sensitive nature of nHDX has already proved instrumental in characterizing transient, low-affinity interactions, such as the mSWI/SNF SWIFT domain-PU.1 complex (dissociation constant 3 to 5 μM), a system where traditional HDX-MS was precluded by limited sample availability (47). Furthermore, nHDX enabled the interrogation of the conformational dynamics of human DNMT3A and variants that are difficult to produce recombinantly in high amounts and concentrations, providing key insights into the process of aberrant macro-oligomerization (48).

Overall, by overcoming current sample requirement limitations, nHDX addresses a critical need in structural biology by enabling the analysis of previously inaccessible protein systems that are difficult to express or purify in large quantities. This innovation holds significant potential to advance structural proteomics. Importantly, our implementation of nHDX on a commercial platform offers straightforward adaptation of existing systems rather than requiring specialized equipment, thereby enabling widespread adoption across the scientific community.

Data Availability

The original mass spectra, spectral library, and the protein sequence databases used for searches have been deposited in the public proteomics repository MassIVE (http://massive.ucsd.edu) and are accessible at ftp://massive-ftp.ucsd.edu/v09/MSV000097574/.

Supplemental Data

This article contains supplemental data.

Conflict of Interest

Y. S. is an employee of Thermo Fisher Scientific. B. B. L. receives research funding from Ono Pharmaceuticals and is a shareholder and member of the scientific advisory board of Light Horse Therapeutics. The other authors declare no competing interests.