Chromatography at -30 °C for Reduced Back-Exchange, Reduced Carryover, and Improved Dynamic Range for Hydrogen-Deuterium Exchange Mass Spectrometry.

For hydrogen-deuterium exchange mass spectrometry (HDX-MS) to have an increased role in quality control of biopharmaceuticals, H for D back-exchange occurring during protein analyses should be minimized to promote greater reproducibility. Standard HDX-MS analysis systems that digest proteins and separate peptides at pH 2.7 and 0 °C can lose >30% of the deuterium marker within 15 min of sample injection. This report describes the architecture and performance of a dual-enzyme, HDX-MS instrument that conducts liquid chromatography (LC) separations at subzero temperature, thereby reducing back-exchange and supporting longer LC separations with improved chromatographic resolution. LC separations of perdeuterated, fully reduced, iodoacetamide-treated BSA protein digest standard peptides were performed at 0, -10, -20, and -30 °C in ethylene glycol (EG)/H2O mixtures. Analyses conducted at -20 and -30 °C produced similar results. After subtracting for deuterium retained in arginine side chains, the average peptide eluted during a 40 min gradient contained ≈16% more deuterium than peptides eluted with a conventional 8 min gradient at 0 °C. A subset of peptides exhibited ≈26% more deuterium. Although chromatographic peaks shift with EG concentration and temperature, the apparatus elutes unbroadened LC peaks. Electrospray ion intensity does not decline with increasing EG fraction. To minimize bias from sample carryover, the fluidic circuits allow flush and backflush cleaning of all enzyme and LC columns. The system can perform LC separations and clean enzyme columns simultaneously. Temperature zones are controlled ±0.058 °C. The potential of increased sensitivity by mixing acetonitrile with the analytical column effluent was also examined.

Introduction

Hydrogen–deuterium exchange mass spectrometry (HDX-MS) is a powerful tool for investigating protein dynamics, including protein–ligand interactions, folding dynamics, and interactions among proteins including antibodies, glycoproteins, lipoproteins, membrane proteins, virus fragments, enzymes, chaperones, amyloids, fibrils, and pharmaceuticals.[1−3] In the biopharmaceutical discovery and development sector HDX-MS data have been used to substantiate and protect intellectual property and to evaluate physicochemical similarity between a biosimilar candidate and the originator product.[4,5] In addition, HDX-MS data are increasingly provided to support biologics license applications (BLAs).[6]

The bottom-up HDX-MS measurement involves immersion of protein in D2O for a specific labeling time tHDX, proteolysis of the protein into peptides, liquid chromatographic (LC) separation of the peptides, mass spectrometric measurement of the isotopic envelope for each eluting peptide, and computation of peptide deuterium content, DUncorrectedPeptide(tHDX) from the mass envelope. Back-exchange of H for D during the analysis process reduces the apparent peptide deuterium content, and either short-term or long-term variations in back-exchange limit the reproducibility of the HDX-MS measurement. Back-exchange rates vary with residue and sequence. Reports of back-exchange in peptides range from 15% to 60%.[7−18] Back-exchange is minimized by conducting analyses under cold, acidic conditions (≈pH 2.7, ≈0 °C) and by limiting chromatographic gradients to short duration (6.0–9.5 min).[19−21] Short gradients chosen to reduce back-exchange may not provide optimal separation of eluting peptides.

Because commercial lifecycles of biopharmaceutical products can span decades, the emergent application of HDX-MS for pharmaceutical quality control (QC) will require minimization of measurement drift. Applications that determine the comparability of biosimilars will also require quantitative HDX-MS of similar precision.[5,22−31] HDX-MS measurement drift can be partially alleviated by reporting DCorrectedPeptide(tHDX)’s that are corrected for H/D back-exchange and scaled to immersions in 100% D2O.

Most commonly, DCorrectedPeptide(tHDX) is computed by an approximate formula given by Zhang and Smith that uses the centroid, ⟨m(tHDX = ∞)⟩Peptide, found for the perdeuterated protein.[21,32,33] Each experimentally determined ⟨m(tHDX = ∞)⟩Peptide captures all factors affecting back-exchange, including perturbations induced by peptide interactions with stationary phases,[34] peptide refolding, exchange acceleration by certain acids in the mobile phase,[35,36] catalysis by histidine side chains in amides beyond nearest neighbors, and polymer end-chain effects.[37,38]

We note that preparations of perdeuterated reference protein samples have occasionally proven impracticable.[21,39] In these cases, the back-exchange correction provided by ⟨m(tHDX = ∞)⟩Peptide is estimated by incorporating residue-specific back-exchange rates into a nonlinear fit of peptide data and by imposing appropriate boundaries for HDX rates.[13,40−43] However, the veracity of computed corrections can suffer from difficulties associated with accurately simulating all the aforementioned exchange rate perturbations. Regardless, the uncertainty of each DCorrectedPeptide(tHDX) is decreased as total back-exchange is decreased. Accordingly, an operationally stable metrology platform that minimizes uncertainties associated with computation of DCorrectedPeptide(tHDX)’s will improve comparability of HDX-MS data.

Over the past decade, several groups have introduced methods for suppressing back-exchange in top-down or bottom-up HDX-MS strategies. Amon et al. reported a top-down method that quenched and cooled protein samples to −15 °C, volatilized peptide ions from a chip-based nanoelectrospray source, and fragmented ions in an electron transfer dissociation (ETD) facility.[44] Similarly, Pan et al. demonstrated a top-down approach involving injection of a thawed, deuterium-labeled protein sample into a LC column at −20 °C that eluted sample into a mass spectrometer equipped with an ETD facility.[45,46]

Several groups have reported strategies for suppressing back-exchange during bottom-up HDX-MS analyses by employing chromatography in aprotic solvents,[15] supercritical fluids,[10] and subzero temperature environments.[47−50] Venable et al. investigated the use of several buffer modifiers (ethylene glycol (EG), dimethylformamide, methanol, and formamide) that depressed the solution freezing points to as low as −30 °C.[47] Using methanol as the buffer modifier, Wales et al. reported LC separations at −20 °C.[48] Zhang et al. conducted HDX-MS analyses of epitopes of birch pollen allergen at −9 °C by adding 4.5% ACN to mobile phase A.[49] Fang et al. evaluated methanol and ACN buffer modifiers during their HDX-MS study of a complex deuterated E. coli lysate.[50] Separations conducted at subzero temperature have facilitated use of lengthy chromatography gradients of 25–90 min.[47−50] The 90 min gradient used by Fang et al. yielded ∼3-fold more peptide identifications than found for 15 min gradients.

As temperature decreases, practical difficulties with system implementation arise. Chromatography fluids must contain an increasing fraction of buffer modifier, which can introduce complications such as reduced electrospray source ion (ESI) production. Venable et al. reported that sequence coverage of BSA digest peptides declined from 64% to 45% as the volume fraction of EG in the LC solvent reached 40%, corresponding to a −20 °C freezing point. Buffer modifiers can affect system performance by inducing shorter retention times and diminished resolution,[47,48] and certain organic buffer modifiers can wash out early eluting peptides.[50] Increased viscosity of subzero LC fluids at subzero temperature can result in backpressures that exceed the capacity of the LC pump. Phase separations, as occurs in ACN/H2O mixtures at <−17 °C,[51] may degrade chromatograph performance, although this complication is removed by adding trace EG.[52]

A distinct challenge to the use of HDX-MS in a QC application is bias from chromatographic carryover, which can adversely affect determinations of Dcorrectedpeptide(tHDX). Chromatographic carryover originates from peptides of prior runs postreleased from the protease, trap, and analytical columns. Because carryover peptides are fully protonated from extended exposure to H2O, their contributions skew the associated centroid to lower D-content.[53] Such contributions can also create false EX1-like kinetics signatures.[54]

Carryover also originates from protein aggregates and agglomerates accumulated at each column entrance. Hamuro and Coales recently reported an HDX-MS valve system that can remove protein particles by backflushing the protease, trap, and analytical columns.[55] The time devoted to system cleanup of protein residue represents lost productivity, and its duration can exceed that of the HDX-MS measurement.

The HDX-MS interlaboratory comparison for the Fab fragment of NISTmAb found that the Dcorrectedpeptide(tHDX > 0) in peptides diminished by ∼3.5% across measurements of three complete data sets (each set comprised 18 to 24 injections).[21] Since the larger cohort of laboratories within the HDX-MS study found a reproducibility of %Ecorrectedpeptide(tHDX) = 6.5 ± 0.6% (all values reported as mean ±1 standard deviation (SD)), elimination of chromatographic carryover will improve the repeatability, intermediate measurement precision, and reproducibility of HDX-MS.[56] These precision characteristics are essential attributes of biopharmaceutical QC based on HDX-MS.

This report describes the design of a dual-enzyme, HDX-MS analysis instrument and reports its performance at subzero temperature. The system houses two distinct protease columns that are perpetually cleaned and conditioned while idle. To minimize back-exchange, protein proteolysis is conducted at ≈0 °C, and peptides are captured on a trap column and separated on an analytical column at subzero temperature. To minimize noise and chromatographic carryover from peptides trapped on the column and from aggregates and agglomerates, quaternary pumps flush and backflush columns with varied cleaning solutions. For improved data curation, operational temperatures and pressures are recorded and archived with data files. The instrument is integrated into a robotic rail, and the entire system is compatible with the software of the HDX-MS system. The apparatus features stringent temperature regulation, which is little-affected by suboptimal laboratory environments. The instrument is designed to provide uniform results for long-term projects.

HDX-MS Analysis Apparatus Description

System Design

The HDX apparatus comprises a fluidic circuit box conjoined with a commercial robotic rail (Trajan Scientific and Medical, Morrisville, NC) that transports samples and provides automated scheduling for the conduct of HDX experiments. The present dual-enzyme, −30 °C, chromatography instrument is encased in an aerogel insulated steel box of similar dimensions to the original equipment manufacturer (OEM) box, and it shares the same injection and exit port positions. Consequently, it is a “drop in” replacement for the OEM component. Much of the apparatus was 3D-printed in bronze steel, aluminum, silver, and polylactide (PLA) plastic by in-house and commercial facilities. Mechanical drawings, electrical schematics, parts lists, and STL and EPS file types are available for construction of the dual-enzyme chromatography platform.[57]

The interior of the dual-enzyme, −30 °C, analysis instrument (Figure ) contains four distinct insulated compartments (referred to as zones), each regulated at a selected temperature. Within each compartment homogeneous temperature is maintained by mounting fluidic and chromatographic components on a 9.6 mm thick aluminum plate. The thermally isolated plates are mounted to a common aluminum frame. The entire system is cooled with a liquid chiller (thermal capacity = 275 W at −30 °C) that circulates solutions containing volume fractions of 30% H2O/60% EG/10% methanol through 9.53 mm outer diameter (7.75 mm inner diameter) stainless tubing that is clamped and thermally anchored to each aluminum plate. A throttle valve apportions coolant flow between the 0 °C preparation chamber (Zone 1) and subzero analytical chamber (Zone 2). The system cooling capacity is conserved by employing a counter flow heat exchanger in the fluid return circuit of the 0 °C coolant circuit.

Figure 1

Fluidic circuits of the −30 °C, dual-enzyme HDX-MS instrument.

Chromatography valves are in thermal contact with their respective aluminum base plates. At solvent entry points and where fluidic circuits pass between Zones 1 and 2, capillaries are clamped to aluminum blocks mounted on a compartment base plate. Capillaries passing from a warm zone to a colder zone have a steady-state temperature profile along their length. Heat is introduced by axial thermal conduction along the capillary and heat advection from liquid flowing through the capillaries. Heat is extracted by forced air convection across unenclosed capillary and conduction through the block. Prior to contacting Valves #1, #3, and #4, thermal modeling estimates that the solvent supply capillaries reach steady state temperature within 0.2 °C of TZone 1 or TZone 2. The same model finds that the termini of capillaries passing between Valve #2 and the trap column differ from the local zonal temperature by <±0.02 °C. Interestingly, the forced convection from the compartment fans accounts for a large fraction of total heat transfer.[58]

All other fluidic components (e.g., mixers, analytical columns) are encased in 3D-printed aluminum housings that are mounted to the compartment base plate. Contacts with the thermal mass of a base plate ensure that fluidic components reside at the compartment set point temperature. This isothermal environment suppresses development of a substantial longitudinal temperature increase along columns due to frictional heating between the solvent and column packing.[59,60]

Two chambers house enzyme columns (Zones 3 and 4). To ensure uniform column temperature, each protease column is mounted into a 3D-printed aluminum or silver collet that fits the enzyme column profile exactly. The collet exterior presents an 8 mm diameter cylinder that is clamped into a holder. Both holders are temperature-regulated independently.

Every 2 s, platinum element resistance temperature detectors (RTDs) and thermistors report to a process controller four compartment temperatures and entrance and exit temperatures of the coolant circulation system. Each coolant circuit has sufficient capacity to chill its aluminum plate below the temperature set point. To maintain each plate at its desired temperature precisely, a proportional–integral–derivative (PID) controller uses a standard negative feedback algorithm to switch DC current to thin-film polyamide encapsulated heaters. In Zones 1 and 2 uniform compartment air temperature is maintained by convective air flows produced by mini-fans.

Compartment temperature profiles recorded during each HDX-MS run are archived with the HDX-MS data. Table lists typical temperature characteristics of the apparatus during operation. Stringent temperature regulation is demonstrated by small standard deviation, ΔTSD, from TAverage in each compartment. Extreme temperature deviations from TAverage of at least 2 s duration, ΔT+ and ΔT–, are also modest. The same system performance was obtained for laboratory temperatures between 20 and 30 °C.

Table 1

Typical Temperature Characteristics of the HDX System, Observed during an Operation Period of 0.5 h

zone	Tavg, °C	ΔTSD,a °C	ΔT+,b °C	ΔT–,b °C
1, Preparation	0.000	0.058	0.164	–0.124
2, Analytical	–30.001	0.020	0.054	–0.031
3, Protease A	0.000	0.002	0.004	–0.004
4, Protease B	0.000	0.002	0.004	–0.004

Uncertainty is 1 standard deviation (SD).

Deviation from the average for a minimum duration of 2 s.

To inhibit water condensation, the box encasing the sample processing apparatus is sealed and purged with dry nitrogen. Two circuits within the six-channel PID controller maintain the heated valve extension shafts above the dew point temperature.

Analysis of a protein sample is initiated by injection of protein solution into a 50 μL sample loop in Zone 1 (Table S1A in the Supporting Information). Actuation of Valve #1 allows Pump 1 (50 μL/min) to propel the protein sample through protease column A or B, as selected by Valve #4 (Table S1B or S1G). In this valve state peptides pass from the selected protease column (Zone 3 or 4), through Valve #4, and into Mixer 1, where Pump 3 enriches the EG concentration. Peptide solution passes through Valve #2 and into Zone 2, where the peptides are retained on Trap Column T at subzero temperature. Upon subsequent actuations of Valves #2 and #3 (Table S1C), Pump 4 (50 μL/min) initiates an EG/H2O/ACN gradient, which elutes peptides from the trap column and through the analytical column at subzero temperature. As an option, analytical column effluent passes into Mixer 2, where ACN from Pump 5 (∼50 μL/min) increases the total flow to 100 μL/min.

The instrument is a flexible platform. The preparation chamber contains a bracket that holds a commercial electrochemical reduction cell.[61] Minor modifications to the fluidic circuits can allow incorporation of an immobilized glycosidase column or immobilized microfluidic enzyme reactor (IMER).[62−64]

As is required of all chemical kinetic apparatus, system cleanliness is essential for good quality measurements. Accordingly, enzyme columns are stored in place, and a reduced flow of solvent passes through them perpetually. This arrangement also saves start-up time by rendering unnecessary the customary, initial column-conditioning process. During all valve states one or more fluidic components are cleaned (Table S1). Valve #4 selects the enzyme column that is backflushed by Pump 2 (Tables S1B and S1G). While chromatographic gradients are in progress, it is feasible to select either enzyme column for flushing or backflushing (Tables S1C and S1E). Actuations of Valves #2 and #3 allow Pump 4 to backflush the analytical and trap columns (Tables S1D and S1F). Backflushing processes can remove protein particles from column entrances. The use of quaternary pumps allows enzyme column cleaning procedures to include gradients containing chaotropic agents and detergents.

Experimental Section

Reagents and Materials

D2O (99.96 mol % D) was acquired from Cambridge Isotopes (Andover, MA). Sodium phosphate dihydrate, sodium phosphate monohydrate, sodium chloride, and formic acid (FA) were purchased from Sigma-Aldrich (St. Louis, MO). Ethylene glycol ReagentPlus (>99%) was purchased from Alfa Aesar (Ward Hill, MA). The analytical sample used for these experiments was a tryptic digest of fully reduced and iodoacetamide alkylated BSA (Thermo Scientific Pierce BSA Protein Digest Standard, LC–MS grade, catalog no. 88341).

Samples

Undeuterated and fully deuterated peptides were prepared by reconstituting lyophilized BSA peptides in H2O and D2O (99.96 mol % D), respectively, with 1% ACN to aid solubility. Peptides with D2O were kept at 4 °C for 1 h to fully exchange. Aliquots of each stock were stored at −80 °C. Prior to analysis, samples were diluted in either H2O or D2O to 0.05 pmol/L and equilibrated at 1 °C.

Chromatography

Each LC–MS analysis was initiated by injection of a 2 pmol sample. The peptides were trapped on a C18 column (Phenomenex, Inc.; Torrance, CA; Model Kinetex EVO C18, 100 Å pore, 2.6 μm particle size, 20 mm long × 2.1 mm dia.) and separated on an analytical column (Thermo Fisher Scientific; Waltham, MA; Accucore C18, 80 Å pore, 2.6 μm particle size, 30 mm long × 2.1 mm diameter, catalog no. 17126-032130).

Solvent A comprised mixtures of H2O/EG and 0.1% FA in volume fractions determined by the working temperature (Table ). Solvent B comprised volume fractions of 99.9% ACN and 0.1% FA. Solvents A and B combine for a 50 μL/min flow rate. Two LC gradients were used. The short (8 min) gradient was 3–10% B for 0.5 min, 10–50% B for 7.5 min, 50–95% B for 1 min, 95% B for 4 min, 95–3% B for 0.5 min, 3–95% B for 2 min, 95% B for 3 min, and 95–3% B for 0.5 min. The long (40 min) gradient was 3–40% B for 40 min, 40–95% B for 1 min, 95% B for 4 min, 95–15% B for 0.5 min, 15–95% B for 3 min, 95% B for 2 min, and 95–3% B for 1 min.

Table 2

LC Pump Pressures during Operation of the HDX-MS Analysis System

TZone 2, °C	solvent A H2O/EG volume fractions,a	PPump 1, MPa	PPump 4, MPa
0	100/0	7	24
0	55/45	8	70
–10	74/26	10	40
–20	63/37	20	55
–30	55/45	24	95

Mixtures contain a volume fraction of 0.1% FA.

Mass Spectrometry

Mass spectra were measured by a Thermo Orbitrap Elite (Thermo Fisher Scientific; Waltham, MA). The instrument settings were spray voltage, 3.7 kV; sheath gas flow rate, 25 (arbitrary units); capillary temperature, 275 °C. In the Orbitrap stage, MS spectra were acquired with the resolution set at 60000, which has been shown to yield accurate measurements of hydrogen and deuterium composition.[65] Extracted ion chromatograms (XICs) of undeuterated samples were used for automated peak area calculation in Xcalibur. From mass spectra obtained during HDX-MS experiments, the centroid of each deuterated peptide envelope and the relative deuterium uptake by each peptide was calculated by HDX WorkBench.[66]

Temperature Calibrations

Each RTD and thermistor temperature sensor was calibrated in situ against a colocated, K-type thermocouple (Marlin Manufacturing Corp., Cleveland, OH), which had reference junctions immersed in water/ice bath. The vendor calibrated the thermocouples for service between 0 and −40 °C per procedures recommend by the National Voluntary Laboratory Accreditation Program (https://www.nist.gov/nvlap/about-nvlap).

Estimation of Amide Back-Exchange Rates

Deuterium contents of unstructured peptides in H2O were estimated by a LabVIEW 7.1 (NI, Austin TX) script. The script reproduced main chain intrinsic hydrogen exchange rates obtained by Excel spreadsheets (http://hx2.med.upenn.edu/download.html) containing the 2018 reference parameters.[7,38,67]

Results and Discussion

Retained Deuterium in Peptides

The sample in these experiments was perdeuterated tryptic digest of fully reduced and iodoacetamide-treated BSA protein. Except for passage through a protease column, the peptides experienced the solution conditions found in standard bottom-up HDX-MS measurements. This measurement procedure simplifies data interpretation by avoiding possible conflation of chromatographic and proteolytic performance. During these experiments, the apparatus did not use Mixer 2, i.e., the ESI source of the mass spectrometer was directly connected to the analytical column output via Port 6 on Valve #3.

The measurement procedure defined the kinetics model of the analyses. Samples injected into the apparatus sample loop at t = −3.33 min resided in Zone 1 (0 °C) for 2 min as Pumps 1 and 3 evacuated the sample loop and loaded peptides onto the trap column in Zone 2 (TZone 2 = (0, −10, −20, −30) °C). Peptides resided on the trap column for 1.33 min during the desalting process. At t = 0 min, Valves #2 and #3 changed states, enabling Pump 4 to elute peptides from the trap column into the analytical column. An 8 min LC gradient was employed, as it is typical of contemporaneous HDX-MS studies. Experiments conducted with 40 min LC gradients enabled demonstration of the enhanced capabilities offered by operation at subzero temperature.

Chromatographic separations of perdeuterated tryptic digest of BSA were conducted with EG/H2O mixtures that inhibited freezing in the fluidic circuit. Measurements indicate that charge state distribution did not change with EG concentration; though, the total intensity of the charge state envelope is sensitive to %EG (vide infra).

Table lists the percent deuterium content in peptide amide groups, %DTObs, which were computed using eqs and 2where ⟨m(tHDX)⟩Peptide is the centroid mass; tHDX is the immersion time of the sample in D2O; is the estimated deuterium content retained by side chains; tkRT is the retention time (RT) of the peptide, which are numbered k = 1, 2, 3... in the order of their elution; j is the running index for side chains in the peptide; aj is the number of identical exchangeable sites within a side chain; RiSC is the rate expression for deuterium occupancy of the side chain exchangeable site, which is calculated at pH 2.7 and over the temperature history between sample injection into the analysis system (t = −3.33 min) and elution, tPeptideRT; and mH and mD are the proton and deuteron masses.

Table 3

Percent Deuterium Content in Amide Groups, %DTobs, of Peptides Chromatographically Separated on the HDX Analysis System for TZone 2 = 0, −10, −20, and −30 °C

	8 min gradient					40 min gradient
volume fraction of EG, %	0	45	26	37	45	37	45
TZone 2, °C	0	0	–10	–20	–30	–20	–30
peptide sequencea	%D0°CObs	%D0°CObs	%D–10°CObs	%D–20 °CObs	%D–30°CObs	%D–20 °CObs	%D–30°CObs
AEFVEVTK	69.9 ± 0.1	75.9 ± 0.8	79.5 ± 0.3	85.0 ± 0.2	85.1 ± 0.1	83.4 ± 0.1	84.3 ± 0.1
DDSPDLPK	59.8 ± 0.3	67.0 ± 0.2	70.2 ± 0.2	71.0 ± 0.7	69.7 ± 0.6	71.0 ± 0.3	68.2 ± 1.2
DLGEEHFK	42.7 ± 0.2	55.3 ± 0.1	62.8 ± 0.4	70.1 ± 0.2	69.5 ± 0.2	68.0 ± 0.4	67.4 ± 0.4
KQTALVELLK	75.2 ± 0.1	78.1 ± 0.1	82.9 ± 0.1	85.3 ± 0.1	86.0 ± 0.4	83.9 ± 0.2	85.3 ± 0.5
LGEYGFQNALIVR	72.6 ± 1.3	82.9 ± 1.3	84.1 ± 0.9	91.9 ± 0.8	88.6 ± 2.5	87.5 ± 1.2	88.6 ± 1.4
LVNELTEFAK	69.0 ± 0.1	73.2 ± 0.03	80.5 ± 0.1	88.6 ± 0.1	89.4 ± 0.1	84.7 ± 0.2	88.7 ± 0.5
LVVSTQTALA	79.7 ± 0.1	83.9 ± 0.1	88.6 ± 0.2	93.9 ± 0.1	94.8 ± 0.4	92.3 ± 0.1	93.5 ± 0.5
QNCCAMDQFEK	68.0 ± 0.2	81.8 ± 0.3	84.2 ± 0.7	88.6 ± 0.9	84.3 ± 11.1	89.4 ± 0.2	89.7 ± 1.7
QTALVELLK	78.6 ± 0.1	81.4 ± 0.1	84.1 ± 0.1	86.5 ± 0.01	86.7 ± 0.5	85.4 ± 0.2	86.4 ± 0.1
TVMENFVAFVDK	83.0 ± 0.2	85.8 ± 0.6	88.1 ± 0.2	92.7 ± 0.2	90.1 ± 0.4	90.6 ± 0.1	90.5 ± 0.7
YICCAMDNQDTISSK	69.9 ± 0.2	83.9 ± 0.2	91.2 ± 0.3	101.0 ± 0.2	99.6 ± 0.4	97.9 ± 0.4	97.8 ± 1.3
YLYEIAR	58.5 ± 1.2	64.6 ± 1.1	68.0 ± 0.9	73.9 ± 0.8	69.4 ± 0.9	72.5 ± 0.9	71.3 ± 0.8
(avg)	68.9 ± 0.5	76.1 ± 0.6	80.3 ± 0.5	85.7 ± 0.5	84.4 ± 3.7	83.9 ± 0.5	84.3 ± 0.9
(avg)	73.5 ± 0.1	76.8 ± 0.4	83.6 ± 0.3	87.5 ± 0.1	88.8 ± 0.1	86.0 ± 0.1	88.4 ± 0.1
Improvement,		7.2 ± 0.4	11.2 ± 0.3	16.3 ± 0.4	14.9 ± 3.6	14.8 ± 0.4	14.9 ± 1.2
Improvement,		13.5 ± 0.2	19.2 ± 0.3	26.4 ± 0.4	25.7 ± 0.2	24.9 ± 0.3	24.8 ± 0.4
MAD, %:	11.3 ± 0.6	11.1 ± 0.8	9.3 ± 0.6	9.7 ± 0.5	10.9 ± 2.1	9.5 ± 0.5	11.3 ± 1.2

Cysteines are carbamidomethylated (CAM).

For peptide-containing arginine ≈ 0.96–1.87 Da, due to deuterium retained at the ε-NH (aj = 1) and η-NH2 (aj = 4) exchangeable sites. Table S2 lists the deuterium mass within the arginine side chains, which were computed with the kinetics model, knowledge of pH, TZone 1, TZone 2, peptide retention times (Table S3), and side-chain exchange rate coefficients determined from NMR studies.[7,68] The known exchange rates of side chains in tryptophan and glutamine predict essentially complete deuterium loss during the proteolysis and desalting processes.[7] However, because the rate coefficients for acid-catalyzed exchange in some side chains are unreported, %DTObs should be regarded as the upper limit of deuterium content in peptide amide groups.

The normalizing term, (n – 1)mD, contains n, the number of exchangeable amides, and 1, which represents the N-terminal amide that is exchanged completely. Accordingly, n does not contain contributions from prolines and chemically blocked cysteines.

Table S4 lists estimated percent retained deuterium content in peptide amide groups, %DTEst, which is computed by the kinetics model with knowledge of pH, TZone 1, TZone 2, peptide retention times, amide exchange rates,[7,38,67] and the (n – 1)mD normalizing term. We note that %DTEst does not account for possible deuterium loss in the ion source nor does it contain deuterium mass retained by side chains.[17,69]

Table lists and , which are averages of %DTObs and %DTEst across all peptides. These averages show that peptides retain more deuterium as TZone 2 decreases below 0 °C. For TZone 2 < 0 °C, Improvement ranges between (7.2 ± 0.4) % and (16.3 ± 0.4) %. The reference for this calculation is , which is the computed for the k peptides measured during the 8 min gradient at 0 °C. Inspection of %DTObs for all peptides reveals that improvements of retained deuterium vary as a function of peptide sequence. Thus, Table also reports the Improvement measured for the three peptides most affected by chromatographic conditions, which is (26.4 ± 0.4) % for an 8 min LC gradient at −20 °C.

Table also reports Relative Mean Absolute Deviation % (MAD), which is the mean of the absolute value differences between %DT,kObs and %DT,kEst divided by %DT,kEst across all peptides. The ratios reveal that %DT,kObs and differ by (9.3 ± 0.6) % to (11.3 ± 1.2) % across all temperatures, which is within the expected accuracy of the calculations.

%DTEst overestimates %DTObs for 57% of peptides. Greater discord was expected. Previously, Woodward et al. reported that addition of EG to water decreased exchange rates in folded proteins at 20 °C.[70] No model for effects of EG on exchange rates, validated against data, was developed; however, it was proposed that exchange rates in EG/H2O mixtures at a given pH and temperature are affected by the volume fraction of water, the water equilibrium constant (kw), and the dielectric constant. Since these effects may be offsetting, this ostensive small bias in may be fortuitus.

This study found essentially the same and for BSA peptides observed during chromatography at −20 and −30 °C for 8 min (EW ≈ 3.5 min) and 45 min gradients (EW ≈ 22 min), where EW is the elution window defined by the time interval between elution of the first (t1RT) and last (t12RT) LC peaks. Simulations predict that separations conducted using longer gradients will show greater differences. Table S5 lists estimated for the same BSA peptides separated at −20 °C and −30 °C with 90 and 120 min gradients. For these estimates the BSA peptide retention times t are stretched to fill the gradient duration in proportion to retention times listed in Table S3 for 40 min gradients at −20 °C and −30 °C, i.e., t1RT,new = t1RT,40 min and t12RT,new = t1RT,40 min + EW. For lengthy gradients, the improvement for conducting the separation at −30 °C vs −20 °C is ≈5.5% for a 90 min gradient and ≈7% for a 120 min gradient.

Solution %EG and Temperature Affect Chromatographic Performance

Adding EG to solvent A of analytical Pump 4 resulted in earlier elution of peptides (Figure ). For example, when conducting 8 min gradients at TZone 2 = 0 °C, solutions containing volume fractions of 45% EG eluted peptides approximately 1–3 min earlier than observed with 0% EG (Table S3).

Figure 2

Chromatographic peaks of peptide YLYEIAR2+ recorded during 8 min gradients: (A) 0 °C, 45% EG and 0% EG, (B) −10 °C, 26% EG, (C) −20 °C, 37% EG, and (D) −30 °C, 45% EG.

Contrariwise, a decrease in temperature is known to increase retention time for most analytes. Such retention time shifts were observed as TZone 2 was sequentially decreased from 0 to −30 °C (Table S3). In a volume fraction of 45% EG solution peptide elution times shift by ∼−0.2 min to ∼+2.2 min as TZone 2 decreases from 0 to −30 °C.

While retention times shifted with respect to volume fraction of EG and temperature, peptide separation was not negatively impacted. Under all conditions evaluated, chromatographic peak shapes during 8 min gradients exhibited minimal change (Figure ). Figure A shows representative peaks observed for YLYEIAR2+ during separation in volume fractions of 45% EG and 0% EG at TZone 2 = 0 °C. The LC peak eluted with 45% EG had similar width and greater peak area than the LC peak eluted with 0% EG. At TZone 2 = −30 °C, the full-width half-maximum (fwhm) peak width was (14 ± 2) s and retention time was comparable to 0% EG at TZone 2 = 0 °C (Figure D).

A factor affecting LC peak widths is temperature inhomogeneity within the analytical column.[59,71−74] Installation of columns in high conductivity metal cases in contact with the TZone 2 thermal reservoir establishes nearly isothermal longitudinal temperature profiles. However, heat produced by friction between the solvent and the porous column medium produces a radially symmetric temperature profile with its maximum on the column axis, TAxis > TZone 2, and its minimum at the wall boundary, TWall = TZone 2. The radial temperature profile gives rise to inhomogeneous transport and release kinetics between the analyte and column packing. This amount of inhomogeneity appears to have little effect, as the YLYEIAR2+ peptide peak width is rangebound between 14 and 24 s. The estimated radial temperature difference is (TAxis – TWall) ≤ 0.1 °C, assuming incompressible solvents.[60]

As TZone 2 decreases to −30 °C, the viscosity of EG/ACN/H2O mixtures increases and pressures at the trap and analytical columns increase proportionately (Table ). Consequently, the pressure capacity of Pump 4 largely governs choice of analytical column dimensions and particle size. During this study, the maximum pressure at Pump 4 was 95 MPa, which is within the pressure capacity of some commercial LC pumps. Pump 4 can drive 151.8 MPa; however, for chromatography this pressure must be derated slightly to accommodate pressure surges from valve switching. This pump has successfully driven LC gradients at −30 °C with a 5 μm particle size, 4 mm × 2.1 mm diameter trap column and a 1.9 μm particle size, 50 mm long x 1 mm diameter analytical column. However, the present configuration appeared to provide the best combination of resolution and sensitivity.

Effects of EG and Added ACN on MS

EG is significantly more viscous than water and not a common solvent for MS. To evaluate the effect of EG on ion intensity during ESI-MS analyses, the integrated area of the extracted ion chromatogram (XIC) for each peptide separated on the analytical column at TZone4 = 0 °C was tabulated (Table S6). Triplicate measurements were conducted. A0% EG is the peak area of a peptide eluted using solvent A that contains volume fractions of 99.9% H2O/0.1% FA, which is a mixture used for conventional LC–MS. A45% EG is the peak area of a peptide eluted using solvent A that contains volume fractions of 45% EG/54.99% H2O/0.1% FA. The ratio of peak areas for each peptide eluted under the two solvent A conditions is

Unshaded bars in Figure plot the R3 ratios determined from the XIC data. Ratios greater than 100% indicate greater detected ion intensity for peptides eluted in solvent A solutions containing volume fractions of 45% EG. For the 12 selected peptides, nine peptides showed improved performance and three peptides showed reduced performance with EG.

Figure 3

Peak areas of 12 peptides observed from an electrospray ion source that is sampling analytical column effluent (50 μL/min) comprising H2O/EG/ACN (unshaded bars) and analytical column effluent that is supplemented with 50 μL/min ACN for a combined flow of 100 μL/min (shaded bars). Peak areas for both conditions are normalized to the peak areas of the conventional condition without EG. Peptides are listed left to right in the order of retention time. Error bars indicate SD of triplicate measurements.

Since ESI is often improved for solutions containing large fractions of ACN and ACN reduces viscosity of H2O/EG mixtures, Mixer 2 was added to the fluidic circuit, and the effect of supplemental ACN on ESI yield was measured. At Mixer 2, the 50 μL/min analytical column effluent flow was combined with the 50 μL/min flow from Pump 5, which comprised volume fractions of 99.9% ACN and 0.1% FA. The resulting 100 μL/min flow entered the ESI source. The resulting XIC yields A45% EG+ACN, which is normalized by A0% EG:

In Figure , shaded bars labeled “Column effluent with ACN added” present the R4 ratios. Since most ratios are <100%, we conclude that supplementing 50 μL/min column flow with 50 μL/min ACN provides no net increase of signal intensity. This minimal impact may be due to the relatively low flow rate of 50 μL/min, which suitably desolvated ions during ESI even with EG (Table S6).

Many HDX-MS users operate with flow rates around 200 μL/min, which may benefit from an addition of ACN. A common approach to improve ESI efficiency is to use a heated probe at higher flow rates. This is not advisible for HDX-MS, as the excess heat applied can increase back-exchange.[17,69] Addition of ACN prior to ESI may serve as an alternative strategy for probes at ambient temperature and for MS instrumentation with different source configurations.

Previously, Venable et al. reported that ion intensity declined as the EG concentration in the ESI solution increased to 45%.[47] In contrast, the present data indicate that BSA peptide ion intensities mainly remain stable or increase as the EG concentration is increased. The discord between the present and prior reports does not reside in differences of chromatography apparatus, as the present peptide ion intensity trends are also observed for BSA peptides in EG/H2O/FA solutions infused directly by a syringe pump. We speculate that the generational differences between the older and present ESI source configurations account for the stable performance reported in this work.

Subzero TZone 2 Expands the LC Elution Window

Performance parameters of HDX-MS apparatus include , the estimated average deuterium content in peptides at the onset of the LC gradient; LC Peak Width, the temporal width of eluting LC peaks (fwhm); LC retention time of a peptide (defined above); , the estimated average deuterium content in all peptides at the start of the elution window; EW, the elution window; and , the estimated average deuterium content change of the peptide set during the elution window, which is, effectively, the bias across the set of DUncorrectedPeptide(tkRT). Table lists these properties for the reported solution conditions.

Table 4

HDX-MS Characteristics at TZone 2 ≤ 0 °C

TZone 2 (°C)	volume fraction of EG (%)	LC Grad (min)	(0)a (%)	LC Peak Width (fwhm) (s)	t1RT (min)	b (%)	EW (min)	c (%)
0	0	8	85.6	21 ± 2	5.52 ± 0.03	75.2 ± 0.1	3.30 ± 0.02	–4.6 ± 0.2
0	45	8	85.6	21 ± 2	2.54 ± 0.06	80.3 ± 0.2	4.80 ± 0.04	–7.7 ± 0.4
–10	26	8	88.1	24 ± 2	4.42 ± 0.14	84.7 ± 0.1	3.50 ± 0.10	–2.3 ± 0.1
–20	37	8	89.0	15 ± 2	3.48 ± 0.02	88.1 ± 0.1	4.40 ± 0.02	–1.0 ± 0.1
–30	45	8	89.2	14 ± 2	3.30 ± 0.11	89.0 ± 0.1	5.30 ± 0.07	–0.3 ± 0.1
–20	37	40	89.0	17 ± 2	3.60 ± 0.25	88.1 ± 0.1	22.2 ± 0.18	–5.7 ± 0.2
–30	45	40	89.2	22 ± 2	3.27 ± 0.32	89.0 ± 0.1	20.4 ± 0.23	–1.3 ± 0.2

Estimate assumes no temporal dispersion prior to LC chromatography (t < 0 min).

Uncertainty includes contributions from uncertainty of t1EW.

Uncertainty includes contributions from the sum of EW uncertainty and one LC Peak Width.

Figure plots vs LC Gradient Time for 12 peptides at TZone 2 = (0, −10, −20, −30) °C. Plots of during negative time estimates the average deuterium content as perdeuterated protein undergoes proteolysis and desalting. After protein proteolysis for 2 min at 0 °C, the kinetics model estimates (−1.33 min) = 89.3%. When operating at TZone 2 = −30 °C, peptides enter the analytical column containing (0) = 89.2% of the initial deuterium content, which is 3.6% more deuterium than retained by peptides desalted at 0 °C (Table ).

Figure 4

vs LC Gradient Time (min) computed for the 12-peptide set separated by liquid chromatography at 0, −10, −20, and −30 °C, at pH 2.7, and conditions listed in Table . The shaded area along each plot of marks the observed elution window for all peptides. The right column lists the average Improvement, referenced to 8 min gradient data observed at 0 °C, for TZone 2 (LC gradient duration): −10 °C (8 min), −20 °C (40 min), and −30 °C (40 min).

Additional deuterium content is lost before elution of the first LC peak. The estimated loss is ≈ 10.4% when TZone 2 = 0 °C and ≈ 0.2% for TZone 2 = −30 °C. Consequently, HDX-MS instruments operating at subzero temperatures feature greater dynamic range for determination of %DTObs than conventional instruments operating at 0 °C.

HDX-MS experiments have used short LC gradients to limit losses of deuterium due to back-exchange. For example, the 15 laboratories that contributed data to the NIST HDX-MS interlaboratory study employed LC gradients of (6.5 to 9.5) min.[21] The present 8 min gradient used to separate perdeuterated BSA digest peptides exhibited an elution window of 3.3 min at 0 °C (Figure , Table ). Short LC elution windows are often congested with overlapping peaks. This congestion can hinder peptide identifications. Temporally coincident LC peaks with overlapping isotopic envelopes can corrupt determination of centroid masses and sequences. Overlapping peaks can suffer from ion suppression of eluents. These issues can limit the mass of proteins accessible to HDX-MS experiments.[19]

Operation of a chromatograph at subzero temperature allows the operator to expand LC elution windows with no back-exchange penalty (Table ). As shown in Figure , analyses of BSA tryptic digest with 40 min gradients at TZone 2 = −20 and −30 °C have LC elution windows of 20.4 and 22.2 min, respectively. For TZone 2 = −30 °C the change in average deuterium content of the peptide set is = (−1.3 ± 0.2) % during a 40 min gradient, which is ∼28% of deuterium loss suffered by the peptide set during an 8 min gradient at TZone 2 = 0 °C (Table ). Thus, when separated at −30 °C in 45% EG using a 40 min chromatographic gradient, peptides exhibiting rapid back-exchange can exhibit ≈25% greater deuterium content than the same peptides separated at 0 °C, 0% EG using an 8 min gradient (Table ).

Conclusion

The present report describes the design of a dual-enzyme HDX-MS analysis instrument and exploration its performance at subzero temperature. This instrument offers greater dynamic range for determinations of %DTObs with reduced back-exchange. The system is a versatile metrology platform for HDX-MS of proteins. Dual enzyme columns can support investigations requiring multiple proteases that produce overlapping peptide sequences[75−77] and the study of glycoproteins requiring protease and glycosidase columns. Plans for this instrument and its electronics are archived in a public repository.[57]

The design and operational capabilities minimize effects of bias, variance, and drift during determination of each Dcorrectedpeptide(tHDX). The fluidic architecture supports system cleanliness by flushing and backflushing columns and by cleaning cycles with quaternary pumps. The ability to conduct cleaning cycles during analytical chromatography gradients will shorten the duration of measurement campaigns, optimizing valuable instrument use. Rigorous temperature control will enable an HDX-MS analysis system to provide the same fluidic environment for extended timeframes, enhancing repeatability and intermediate measurement precision.[56] Temperature logging with each HDX-MS measurement provides a record that will support computational modeling of measurements. These features will enable the dual-enzyme, subzero, HDX-MS instrument to become a reliable platform for future biopharmaceutical QC programs and protein similarity studies based on HDX-MS.