For high throughput native mass spectrometry (MS) protein characterization, it is advantageous to desalt and separate proteins by size exclusion chromatography (SEC). Sensitivity, resolution, and speed in these methods remain limited by standard SEC columns.
Trang 1Contents lists available at ScienceDirect
journal homepage: www.elsevier.com/locate/chroma
Col Liquid Chromatography
ES Hechta, ∗, EC Obioraha, X Liub, L Morrisonb, H Shionb, M Lauberb, ∗
a Genentech, Inc South San Francisco, CA, USA
b Waters Corporation, Milford, MA, USA
a r t i c l e i n f o
Article history:
Received 13 August 2022
Revised 2 November 2022
Accepted 4 November 2022
Available online 7 November 2022
Keywords:
Size exclusion chromatography (SEC)
Microflow
Native MS
Noncovalent interactions
a b s t r a c t
Forhighthroughputnativemassspectrometry(MS)proteincharacterization,itisadvantageoustodesalt andseparateproteinsbysizeexclusionchromatography(SEC).Sensitivity,resolution,andspeedinthese methodsremainlimitedbystandardSECcolumns.Moreover,theefficientpackingofsmallborecolumns
isnotoriouslydifficult.SECsensitivityisinherentlylimitedbecausesolutesarenotfocusedinto concen-tratedbandsand lowaffinity native complexesmay dissociateoncolumn.Recent work evaluatedthe suitabilityofcrosslinkedgelmediainsmallboreformats foronlinedesalting.Here,smallbore format onlineSECfornativeMSstudiesisagaininvestigatedbutwithalternativematerials.Wesystematically studiedtheutilityofdiolandhydroxyterminatedpolyethyleneoxide(PEO)bonded1.7μmorganosilica particlesaspackedinto1mmID stainlesssteel(SS) hardwareand hardwaretreatedwithhydrophilic hybrid surfacetechnology(h-HST).Forthe equivalentdiol-bondedparticleand hardware,UVlimitsof detection(LODs)werereduced32to89%withamicroflowseparation(15μL/min)ona1× 50mm col-umnascomparedtoa4.6× 150mmhigh-flowseparation(300μL/min)atthesamelinearvelocity.Run timeswerealsoshortenedby45%.AswitchfromSStoh-HSThardwareledtoasignificantreductionin secondaryinteractionsandacorrespondingimprovementindetectionlimitsfortrastuzumab,myoglobin, IgGandalbumin forbothUVandMS Couplingofthesmallborecolumns tomultichannelmicroflow emittersresulted in10to 100-foldgainsinMSsensitivity, dependingontheanalyte MSLODvalues weresignificantlyreducedintothelowattomoleranges.Columnswerethenevaluatedfortheireffects
onthepreservationofcomplexes,includingconcanavalinA,initsapoandligand-boundstates,andthree therapeuticallyrelevantnoncovalentsystemspreviouslyundetectedonlargecolumnformats.Theresults suggestthatthedetectionoflargecomplexesbySECisnotjustafunctionofsensitivitybutisdirectly affectedbychemicalsecondaryinteractions.Theabilitytodetect0.1to1MDacomplexes,withbetween
1and40micromolardissociationconstants,representsacriticaladvancementforhigh-throughputnative
MSworkflowsasappliedtotheanalysisoftherapeutics
© 2022TheAuthors.PublishedbyElsevierB.V ThisisanopenaccessarticleundertheCCBY-NC-NDlicense (http://creativecommons.org/licenses/by-nc-nd/4.0/)
1 Introduction
The biggest challenges in native mass spectrometry (MS) of
proteins include preserving structures, stabilizing non-covalent in-
teractions, and achieving high signal-to-noise detection When
successful, native MS provides a wealth of information about
the biological properties of intact protein assemblies under near-
physiological conditions [ 1, 2] When used in a biotechnology en-
∗ Corresponding authors
E-mail addresses: hecht.liz@gene.com (E Hecht), matthew_lauber@waters.com
(M Lauber)
vironment, this information can contribute to early stage under- standing of disease states and provide quick insights on the for- mulation, processing, and purification of therapeutic products [3] Industry adoption of native MS remains limited by the accessi- bility of instrumentation, the throughput of the experiments, and the reproducibility of the measurements In the field of intact anal- ysis, solutions exist to meet these challenges, with diverse chro- matography options ever improving to provide compatibility with time of flight (TOF), Orbitrap, and Fourier transform ion cyclotron resonance (FTICR) MS instrumentation [4] In both native and in- tact analysis, proteins or large molecule drug targets are typically over-expressed and then isolated from cells for further study Re-
https://doi.org/10.1016/j.chroma.2022.463638
0021-9673/© 2022 The Authors Published by Elsevier B.V This is an open access article under the CC BY-NC-ND license ( http://creativecommons.org/licenses/by-nc-nd/4.0/ )
Trang 2verse phase liquid chromatography (RP-LC) [4] or capillary elec-
trophoresis (CE) [5]for intact protein analysis show value in giv-
ing reasonable throughput and leading sensitivity by means of on-
column clean-up and focusing While CE methods have emerged
for native mass spectrometry [6], they remain limited by surface
interactions that hinder reproducibility, resolution, and sensitiv-
ity [7] Ion exchange chromatography and hydrophobic interaction
chromatography, when performed with volatile mobile phase com-
ponents, are compatible with MS and have been used in native
applications including the mispairing of bispecific antibodies [8],
antibody-drug-conjugate quantification [9], and mass deconvolu-
tion of heterogeneous targets [10] However, the high salt and el-
evated pH conditions required by those techniques can be desta-
bilizing to some protein aggregates or complexes Size exclusion
chromatography (SEC) is a compelling and widely used option for
these applications, and it separates solutes on the basis of their
hydrodynamic radii That said, it is a low sensitivity technique that
requires concentrated samples, and, as with any instance of sample
handling, it can be difficult to mitigate the disruption of noncova-
lent interactions [ 11, 12]
There are ways to enhance MS sensitivity independent of
the protein separation Historically, nano-electrospray (nESI) (static
spray) of proteins has achieved the best limits of detection yet does
not facilitate rapid screening [ 13, 14] The application of multichan-
nel emitters that are compatible with microflow flow rates bridges
the gap between throughput and sensitivity Multichannel emitters
most commonly work by splitting a single flow channel into mul-
tiple outlets so as to simultaneously spray liquid from more than
one tip Each tip within this clustered sprayer has a lower onset
voltage and flow rate [15–18] In addition to sensitivity improve-
ments, the reduced onset voltage and an application of sheath
gases helps keep proteins and noncovalent complexes in their na-
tive state [ 13, 14] The degree to which a structure remains native
is correlated with charge state, where denaturation leads to higher
charge states and lower m/z values [19–21] Nearly all commercial
SEC columns have internal diameters (IDs) of 4.6 mm or higher,
and are recommended for use with flow rates of 200 μL/min or
higher, making microflow compatibility difficult Split-flow meth-
ods can be used to access this flow regime, but that comes with
higher sample requirements to see significant sensitivity gains [22]
There is a paucity of work on microflow SEC applications, es-
pecially when compared to the comprehensive literature that can
be found for capillary/microflow intact RPLC analyses [23–27] The
packing of SEC phases into small column dimensions, compatible
with lower flow rates, presents manufacturing challenges brought
on by the disproportionate increase of the wall surface area to par-
ticle sizes and an increasing likelihood of uneven flow and poor
column efficiencies [28] If packed with excessively high flow rates
or pressures, bed compression can also limit resolution, when a
compressible packing material is used Lastly, any gains in sensitiv-
ity from smaller columns must be balanced with refined needs for
sample concentration, where the injection volume is ideally < 5%
of the column volume [29] Likewise, to access reasonable sensi-
tivity, instrumentation considerations must be made to minimize
extra-column dispersion effects, though these issues are bound
to be faced in any miniaturized SEC experiment Microflow na-
tive MS using polyacrylamide beads packed into 300 μm diame-
ter columns was recently demonstrated [30] Very high sensitiv-
ity was achieved, although the columns were not compatible with
pressures greater than 400 psi and did not provide separations
other than the fractionation of protein samples from salts Two mi-
croflow SEC columns are commercially available One, a 4 μm hy-
drophillic diol-bonded phase is available from Tosoh Biosciences as
a 1 × 30 mm column, where the maximum flow rate is 20 μL/min;
no LC-MS applications with this hardware dimension were found
in literature searches Second, the polyhydroxyethyl A column from
PolyLC, Inc is available across a range of IDs and lengths The ideal flow for the 2.1 mm column (various lengths) was 100 μL/min across different studies [ 31, 32], demonstrating significant advan- tages over larger columns, yet no peer-reviewed literature could be found for a 1 mm ID version of these columns
The following work entails the use of 1.7 μm ethylene bridged hybrid (BEH) particles packed into small hardware (1 × 50 mm) dimensions and an investigation of these columns for native MS
on noncovalent complexes The effects of hardware dimensions, column hardware surfaces, and particle surface chemistry are as- sessed over a range of native MS applications and instrumentation
We demonstrate that new capabilities can be had with the use
of microflow columns constructed from hydrophilic hybrid surface technology (h-HST) column hardware [33]and packed beds made from hydroxy terminated polyethylene oxide silanized BEH (HO- PEO BEH) particles [34] These devices make it possible to expand beyond traditional native protein characterization to the study of complexes that otherwise dissociate by high-flow SEC approaches Indeed, the new surface chemistries are seen to minimize chem- ical forces that are disruptive to complexes, such that is possible
to preserve quaternary interactions that have relatively weak dis- sociation constants ranging from 1 to 40 μM In sum, this work provides compelling evidence for the capabilities of small dimen- sion SEC as an MS-inlet technique that provides sufficient desalt- ing and chromatographic resolution, minimal disruption to native complexes and <3 min, highly sensitive analyses
2 Materials and methods
2.1 Materials
Trastuzumab, an RGY antibody hexamer complex described pre- viously in [35], a complement protein hexamer complex (expected mass 443 kDa), and phospholipase b-like 2 (PLBL2) antibody com- plex [36]were produced in-house at Genentech, Inc The 443 kDa protein hexamer was composed of six calcium-mediated non co- valent assemblies, where each assembly was a noncovalent com- plex of three proteins totaling ∼74 kDa each Ammonium acetate, methanol, conconavalin A (ConA) from jack bean, glucose (Glu), and p-nitrophenyl-ar-D-mannopyranoside (PNM) were purchased from Sigma-Aldrich (St Louis, MO) Acetonitrile and water were purchased from Fisher Scientific (Hampton, NH) Five-protein SEC mixture was purchased from Waters Corporation (Milford, MA) All solvents were HPLC grade or > 99.9% purity
2.2 Column manufacturing
SEC packing materials were prepared from or- ganic/inorganic hybrid particles with an empirical formula of SiO 2(O 1.5SiCH 2CH 2SiO 1.5) 0.25 [37] One batch of these BEH TM par- ticles was bonded with a hydroxy terminated (HO) polyethylene oxide (PEO) silane [34]This packing material is referred to herein
as HO-PEO bonded BEH or HO-PEO BEH The average particle size of this packing material was 1.64 μm in diameter, and the particles were measured to have an average pore diameter of 262
˚A, surface area of 170 m 2/g, surface coverage of 1.15 μmol/m 2and pore volume of 1.26 cm 3/g A second BEH packing material was investigated in this study and it was acquired in the form of bulk manufactured diol-bonded BEH particles, just as they are prepared for commercially available columns (ACQUITY TM UPLC TM BEH200 SEC columns, Waters, Corporation, Milford, MA) The selected batch had an average particle diameter of 1.54 μm, average pore diameter of 193 ˚A, surface area of 225 m 2/g, pore volume of 1.30
cm 3/g and surface coverage of 5.42 μmol/m 2 These SEC packing materials were either packed into 1 × 50
mm stainless steel column hardware or 1 × 50 mm column hard-
Trang 3ware that had been treated to have a hydrophilically modified
hybrid organic/inorganic surface In a previous report, this latter
type of hardware has been referred to as h-HST hardware, which
stands for hydrophilic hybrid surface technology [33] Columns
were slurry packed using constant pressure packing conditions and
to produce columns with mechanical stability to withstand pres-
surization to beyond 6,0 0 0 psi
2.3 LC-UV-MS
All protein samples were diluted or resuspended in 50 mM am-
monium acetate and used within three days of thawing All dilu-
tion curve experiments were injected from least to most concen-
trated, with a minimum of a 10 min wash and re-equilibration
time between each sample for the small dimension hardware to
ensure no carryover Regardless of the concentration, the injection
volume was held constant at 1 μL For ConA experiments, all injec-
tions were 510 ng (5 picomole based on the tetramer mass), with
the exception of split flow experiments, where 153 ng was directly
injected at 15 μL/min, or else 1012 ng was injected at 100 μL/min
For hexamer experiments, 480 ng was injected onto the column
For PLBL2-IgG4 experiments, a 2:1 molar complex, equating to 250
ng of each species, was injected
A Vanquish TM LC (Thermo Fisher Scientific, Waltham, MA) was
operated in isocratic mode with 100% 50 mM ammonium acetate
at appropriate micro or high flow rates Regardless of the flow
rate, the column outlet was connected to a biocompatible column
cooler, a semi-micro bio 2.5 μL flow cell, and then a switching
valve via a 100 μm x 750 mm stainless steel line A 0.005" ID peek
line was used for connecting a HESI (heated electrospray ioniza-
tion) source The HESI source was operated at an inlet tempera-
ture of 275 °C and an electrospray voltage of 4 kV For low-flow
experiments, the sheath gas was set to 15 with no additional heat
and for high flow (50 or 300 μL/min) experiments, the sheath gas
was set to 20 and a 30 °C auxiliary gas was set to 5 Unless spec-
ified, all microflow 1 × 50 mm SEC-MS data was collected with
the multichannel emitters A 75 μm x 550 mm nanoViper TM line
(Thermo Fisher Scientific, Waltham, MA) was used to connect to
the microflow-nanospray Electrospray Ionization (MnESI) source,
equipped with 20 μm, 8 nozzle emitters (Newomics, Inc Berkeley,
CA) The MnESI source was operated at 4 kV ESI voltage and with-
out the use of sheath gases After ten minutes of isocratic flow, the
switching valve changed positions and infused methanol over the
multichannel emitters from an external syringe pump (Chemex,
Chicopee, MA) at 35 μL/min, while the column was simultane-
ously washed for 2.5 min at 40 μL/min, followed by a pressure
re-equilibration to twenty minutes total UV data was collected at
280nm Q Exactive TMUHMR (Thermo Fisher Scientific, Bremen, DE)
mass spectrometer settings were tuned for the sample of interest
2.4 Mass spectrum deconvolution
Byos TM software (Protein Metrics Inc, Cupertino, CA) was used
for intact mass deconvolution For all intact analysis, the intensity
summed across all charge states was used and, if applicable, fur-
ther summed across glycoforms
2.5 UV Peak fitting
MagicPlot software (Magicplot Systems, Saint Petersburg, RUS)
was used to peak fit the UV data spectra Gaussian curves were
provided with an initial set of parameters defining their approx-
imate elution time, intensity, and width The software then per-
formed a simultaneous multi-fit optimization of these parameters
such that the sum of squares was minimized against the A280
trace From these Gaussian fits, peak width and apex elution time were extracted
2.6 Statistics and curve fitting
R was used to fit dilution curves and determine limits of de- tection Dilution curves used linear or quadratic equations as most appropriate to the data The regression fit of each model was evaluated with the Breusch-Pagan test, and where appropriate, a weighted model was used [38] The limit of detection was calcu- lated as 3.3 x the standard deviation of the intercept/slope The gradient end time was selected as the point at which the A280 spectra returned to baseline after the elution of the last peak (uracil), and the start of the thyroglobulin elution, whose size
is above the exclusion limit of the pore size of these SEC par- ticles The theoretical plates (N) reported was calculated as the mean apex protein elution time/standard deviation across the five- protein mix The height was calculated as the average plate/the length of the column The peak capacity (P) was calculated as
P = 1 + sqrt(N) 0.2 The resolution of the UV peaks was evalu- ated as the difference in the retention time between two peaks, divided by the original reference retention time
3 Results and discussion
3.1 Evaluation of microflow small dimension SEC columns for UV and MS analyses
The small dimension SEC columns evaluated in this study were designed to optimize MS sensitivity and speed for the analysis
of native proteins and complexes In this work, we have studied samples ranging from a monoclonal antibody (trastuzumab) to a protein test mixture as well as concanavalin A, which forms a concentration-dependent dimer and tetramer Microflow columns yield smaller electrospray droplets that lead to increases in ioniza- tion efficiency but it is challenging to achieve efficient separations with them Standard stainless steel (SS) columns or hydrophilic hybrid surface technology columns (h-HST) of 4.6 × 150 mm or
1 × 50 mm dimensions were packed with standard diol bonded BEH or HO-PEO bonded BEH particles Using the MS-compatible mobile phase of 50 mM ammonium acetate, the MS sensitivity of these devices was explored followed by a characterization of their
UV capabilities
A control experiment was first done to estimate the ioniza- tion gains that come with moving from a traditional electrospray source (HESI) to multichannel emitters at 15 μL/min (Figure S1) A 3-fold gain in signal-to-noise and a shift to a more native charge state distribution was observed when replacing the HESI source with the multichannel emitter (MnESI) source for the analysis of trastuzumab (Figure S1A) The reduction of 1-2 charges, on aver- age, indicated that the protein remained in a more native state The harshness of the HESI source was further apparent on its ef- fects on the ConA tetramer, which was solely preserved by the MnESI source (Figure S1B) Thus, hereafter, microflow comparisons between small bore columns were made using the multichannel emitters and a HESI-appropriate flow rate of 50 μL/min was used
to compare large and small bore columns of the same chemistries
A seven-point standard curve of trastuzumab from 1- 400 ng was generated across all columns and analyzed by MS ( Fig 1A) With the flow rate controlled at 50 μL/min, the 4.6 × 150 mm column yielded an LOD of 724 attomole, while the 1 × 50 mm SS/BEH Diol column produced an LOD of 253 attomole (Table S1) When operated at 15 mL/min, this same column gave an LOD of
85 attomole, while its h-HST/HO-PEO BEH equivalent produced an LOD with a value of 60 attomole These results suggest that column
Trang 4Fig 1 Standard curves for the MS detection of (A) trastuzumab, (B) IgG, (C) BSA, or (D) myoglobin were generated, with the log 2 summed intensity plotted against concen- tration Confidence intervals (95%) are shown as shaded ribbons
miniaturization, low flow rate, and reduced secondary interactions
can each contribute to gains in MS sensitivity
Standard curves were next built for three of the five proteins
injected from a more complex, five-protein test mixture spanning
molecular weights from 1 to 600 kDa Thyroglobulin could not be
detected by MS due to its large size and heterogeneity, and uracil
fell below the mass cutoff of the instrument For BSA, IgG, and
myoglobin, the 1 × 50 mm microflow h-HST/HO-PEO BEH column
performed the best with LODs of 40, 39, and 14 attomoles, re-
spectively This was a 34%, 58%, and 61% reduction compared to
the 4.6 × 150 mm column, respectively (Table S1) All small di-
mension columns detected proteins over at least three orders of
magnitude At low concentrations (less than 100 ng), the signal
began to plateau ( Fig 1BCD) The Q Exactive TM UHMR is an ion
trapping instrument and all runs maxed out their method injec-
tion time (200 ms) (AGC value not reached) It is possible that
by increasing the injection time, a larger dynamic range could be achieved at the cost of fewer points across the curve Interestingly for the SS/BEH Diol columns, saturation was observed at high con- centrations for trastuzumab, myoglobin, and uracil This suggests that at some point, any potential gains in signal from increased protein loads are mitigated by a corresponding increase in non- specific binding to non-coated surfaces Columns were also com- pared based on changes in the signal at the midpoint of the stan- dard curve, rather than at the limit of detection Increases of up
to 100-fold gains in absolute signal intensity were observed for the three protein mix analytes compared to approximately 10-fold gains observed for trastuzumab ( Fig.1) Similar observations were made upon comparing results from the SS/BEH Diol 1 × 50 and 4.6 × 150 mm columns Thus, it appears that much of the advan-
Trang 5Fig 2 A280 traces for elution of the protein mix are shown in dark blue for the (A) 4.6 × 150 mm column or the (B) SS/BEH Diol (C) h-HST/BEH Diol (D) h-HST/HO-PEO
BEH 1 × 50 mm columns The fitted Gaussian peaks that were used for LC peak resolution and capacity calculations are shown as overlaid traces The spectra shown were generated from an injection of protein mix containing uracil, BSA, myoglobin, IgG and thyroglobulin at concentrations 0.02, 1, 0.04, 0.4, and 0.6 μg, respectively
Table 1
Separation characteristics of columns, determined by UV, injected with a five protein test mixture As described in Section 2.6 , all figures of merit were calculated from Gaussian fits to the LC trace and reported as the average across all proteins For comparison, the theoretical plates calculated from the uracil peak is also provided Dimension (mm) Flow Rate (μL/min) Hardware Particle
Theoretical Plates, Uracil
Theoretical Plates, 5-Protein Average Avg Height
Avg Peak Capacity
tage conferred to complex mixtures on small bore columns derives
from the use of microflow ESI and the related ionization efficiency
gains
To provide a thorough characterization of these devices, we also
compared the separation capabilities for the protein test mixture
by online UV detection ( Fig 2) It should be noted that the com-
parisons described used a single LC instrument without any run-
to-run adaptions The same flow cell and LC capillary lines were
used for both small and wide bore columns to model the prac-
tical use of an LC in an industry lab, where a user often cannot
re-plumb a configuration for a specific application Thus, optimiza-
tion of the LC system for microflow conditions might be an area of
future work that would likely result in improved performance for
1 mm ID SEC analyses
Key metrics for column evaluation, including plate heights, peak
capacity, and the limits of detection, were determined for the
standard five protein test mixture SS/BEH Diol 4.6 × 150 mm
columns were run at 300 μL/min and 1 × 50 mm columns at
15 μL/min, which yields comparable linear velocities Certain fea-
tures were lost entirely in the transition to small columns, includ-
ing the shoulder observed on the IgG protein, which corresponds
to a dimer species ( Fig 2) For all 1 mm and 4.6 mm diameter
experiments, thyroglobulin eluted at 1.75 and 3 min, respectively
This translated to a 41% reduction in elution time A significant loss
in observed plate count was expected and observed for the switch
to 1 mm ID columns The number of average theoretical plates de-
creased from 35100 on the 4.6 × 150 mm column at 300 μL/min,
to approximately 460 0 ( +/- 110 0) on the 1 × 50 mm columns ( Table 1) Peak capacity was reduced by up to 63% on the small bore columns However, the addition of h-HST and then addition- ally HO-PEO BEH particles resulted in small but statistically signif- icant (t-test, p < 0.05) increases in peak capacity and theoretical plates between the small bore devices As noted before, optimiza- tion of pre and post column tubing might help in future work to reduce the dispersion of the small volume chromatographic bands that are generated during microflow SEC
With the addition of h-HST surfaces and HO-PEO particles, the resolution between peaks of proteins < 10 0,0 0 0 Da significantly increased, whereas the separation between the larger proteins was less affected ( Fig 2) Proteins can exhibit unique types of non- specific binding depending on their physicochemical properties The small proteins in the text mixture might be subject to pro- nounced surface interactions as evidenced by their comparatively wider peak widths This effect was further highlighted in a com- parison of columns using a single injection of trastuzumab (Figure S2A) Unlike on the 4.6 × 150 mm column, trastuzumab eluted as two peaks on all 1 × 50 mm columns MS analysis confirmed there
to be no differences in post translational modifications and the charge state distribution was identical, suggesting no perturbations
to structure (Figure S2B) Consequently, when standard curves of trastuzumab were generated on the 1 × 50 mm columns, the main peak showed high nonlinearity compared to the 4.6 × 150
Trang 6Fig 3 The standard curves for UV detection of (A) trastuzumab (B) myoglobin (C) IgG (D) thyroglobulin (E) BSA or (F) uracil are shown with their confidence intervals Area
under the curve was calculated from the Gaussian fits to the raw spectra data for each protein
Trang 7Fig 4 (A) Comparison of the charge summed deconvolved intensity ratio of the tetramer to all peaks for 510 ng ConA on 1 × 50 mm hardware with a 15 μL/min flow
rate (B) Values obtained with the h-HST/HO-PEO BEH column from 153 ng of ConA loaded under microflow conditions (15 μL/min) or 1020 ng loaded and run with a 1:6.7 splitflow (153 ng effective MS detection) (C) Com parison of the percent of glucose bound to ConA across the different 1 × 50 mm columns All experiments were performed with three replicates
mm columns However, when the sum of the main and secondary
peak areas were modeled, the nonlinearity was rescued (Figure S3)
The resolution between the first and second trastuzumab peaks in-
creased with the use of h-HST hardware and the HO-PEO BEH par-
ticles (Figure S2A) Additional work is needed to understand the
behavior of trastuzumab and its split peaks It is possible that sys-
tem effects, including flow rates, column pressures, and injector
processes might also be at play and impacting separation quality
For the microflow h-HST/HO-PEO BEH column compared to the
4.6 × 150 mm column at 300 μL/min, the UV LOD for trastuzumab,
BSA, IgG, and uracil decreased by 32%, 75%, 89%, and 85% (Table
S2) Myoglobin was detected with an approximately equal LOD of 2
picomole, respectively Myoglobin represented the smallest protein
eluting from the mix, and had the greatest peak overlap with other
proteins in the 1 × 50 mm ID columns ( Fig.2), potentially account-
ing for there being no change in LOD Thyroglobulin was detected
with reduced sensitivity on microflow columns Thyroglobulin is
largely excluded from the intraparticle pores of the applied BEH
packing material With a compressed elution time in microflow
columns, this could cause a decrease in sensitivity, and this issue
would likely be solved through use of larger pore size particles
The microflow columns showed high reproducibility, with min-
imal retention time shift, sensitivity loss, and column degrada-
tion over 150 injections Full robustness testing was beyond the
scope of this study, where the focus ultimately was to provide a
base level characterization of the column behavior and demon-
strate their utility for MS experiments Some aspects of the ro-
bustness of the microflow SEC device can be predicted from the
performance of the applied BEH particle and its history of use
in analytical scale applications To this end, it can be noted that
batch-to-batch reproducibility for the HO-PEO BEH packing mate-
rial has been previously reported [34] Particles corresponding to
7 different manufacturing batches were studied in 4.6 mm ID col-
umn hardware and used to separate NIST mAb reference material
8671 with a phosphate buffered saline mobile phase Elution times,
area %, USP resolution and USP tailing values were compared for
the monomer main peak as well as high molecular weight species
RSD% values were all less than or equal to 7% Column lifetime was
also previously investigated for a 4.6 × 300 mm packing of 1.7 μm
Fig 5 (A) The ∼900 kDa intact RGY antibody hexamer (black) elutes from 2.2-
2.5 min and undergoes gas-phase dissociation into monomer, dimer, trimer, and tetramer units (B) The RGY hexamer’s in-solution monomer (red) independently elutes at 2.7 min
particles No change was observed in elution times and area% val- ues after the course of 10 0 0 repeat injections
Trang 8Fig 6 A 443 kDa hexamer complex of protein trimers (18 proteins total) was observed by the 1 × 50 mm h-HST/HO-PEO BEH SEC microflow column Extracted ion
chromatograms of the (A) hexamer, (B) tetramer, (C) dimer, and (D) monomeric species are shown, where the “monomeric unit” is considered to be the protein trimer (E) The average spectra showing the in solution dimer, tetramer, and hexamer from 2.3-2.7 min (F) The average spectra from 3-3.4 min showed the tailing of the hexamer, a gas-phase generated dimer, and the monomeric species
The microflow SEC devices described in this work showed im-
proved LODs for both MS and UV detection Overall, the 1 × 50
mm columns offered a more sensitive platform compared to the
large bore columns, compounding the benefits of reduced sur-
face interactions and improved ionization efficiency The small bore
columns had peak capacities approaching 20 ( Table1), differentiat-
ing the column from alternative desalting columns, such as those
packed with compressible large particle size materials [ 30, 39, 40]
The 1 × 50 mm SEC columns constructed with h-HST hardware
and HO-PEO BEH particles offered a highly sensitive column for MS
analysis with fast LC run times that enabled attomole level protein
detection
3.2 Noncovalent complex stability as a function of particle chemistry
and hardware
The detection of noncovalent complexes is a particular chal-
lenge for native SEC-MS, where the column can cause complexes to
dissociate The stability of a complex can be affected by pressure,
nonspecific interactions, shear, buffer, and pH Protein complexes
of interest are also often found at low relative abundances Accord-
ingly, we studied the effects of the microflow columns for several
well characterized protein-protein and protein-small molecule sys-
tems
ConA is a tetrameric lectin with the capacity to bind up to
four glucose and mannose type sugars [41] Formation of the ConA
tetramer is reversible, with the tetrameric form stabilized at neu-
tral to high pH, and the dimer favored under acidic conditions
[42] Sensitivity to pH makes it potentially more susceptible to gas- phase dissociation due to the acidic nature of electrospray ConA binds glucose and PNM with dissociation constants of 5.7 [43]and 40.9 μM [44], respectively While ConA has been extensively stud- ied, it is well documented that it can present problems when analyzed by chromatography During purification of glycoproteins where ConA crosslinked to sepharose is used in an affinity column, leaching of ConA is a historic and persisting problem [45] Like- wise, while ConA is well studied in the field of mass spectrometry,
to our knowledge, all analyses of the native tetramer by MS has been done via direct infusion [ 46, 47]
As discussed earlier and shown in Figure S1B, the ConA tetramer could not be observed from the high flow setup, even with up to 3 μg injections, due to the harshness of the electro- spray source Across the microflow columns, differences in ConA tetramer detection were observed The ConA tetramer was quan- tified from the SS/BEH Diol, h-HST/BEH Diol, and h-HST/HO-PEO BEH microflow columns at 0.5%, 3.7%, and 7% of the total protein signal, respectively ( Fig 4A) The combination of hydrophilically optimized particles and column surfaces maximized the amount of multimer detected, suggesting that secondary surface interactions could be responsible for complex dissociation
To further investigate effects that can influence tetramer recov- ery, split flow and higher linear velocity experiments were per- formed The same h-HST/HO-PEO BEH column was evaluated for
a 153 ng, 1 μL injection at 15 μL/min and a 1020 ng, 1 μL injection
at 100 μL/min For the latter scenario, post column flow was split
at a 1:6.7 ratio to ensure equal protein concentrations were elec-
Trang 9Fig 7 The PLBL2-IgG4 complex as observed by (A) the h-HST/HO-PEO BEH SEC microflow device (black) or (B) static spray (red) The peaks corresponding to the complex
are shown with stars
trosprayed into the mass spectrometer for both conditions There
was a statistically significant loss of tetramer (29%) observed when
the column flow rate was increased to 100 μL/min ( Fig.3B) ConA
forms in a concentration dependent fashion An increased concen-
tration would theoretically increase the relative percent starting
tetramer in solution So the observed reduction could in fact be an
underestimation of the amount loss Between the 100 μL/min and
15 μL/min flow rates, the pressure increased from 41 to 237 bar,
and the elution time decreased from 3.7 min to 0.6 min As the
split-flow and microflow experiments were performed under lam-
inar flow conditions, the only changes to shear forces would be in
direct correlation to the change in flow rate Therefore, it is pos-
sible for the equilibrium of the complex to have been affected by
the high flow and >200 bar pressure conditions Additional exper-
iments with controlled flow restriction might help better elucidate
the operational boundaries to consider for these types of microflow
SEC-MS experiments and application of SEC to weakly bound com-
plexes
Small molecule binding was next studied Each ConA pro-
tomer has the ability to bind small ligands Glucose was de-
tected bound to the tetrameric form of ConA by microflow-SEC-MS
only when the h-HST column hardware was employed ( Fig 3C)
Interestingly, there was no difference in the ratio of tetramers
with glucose bound between the HO-PEO BEH and BEH Diol par-
ticles In all cases, a high concentration of ConA ( ∼5 picomol)
was applied and evidence of column overload can be seen in the
form of peak tailing Nevertheless, a comparison of the MS1 to-
tal ion chromatogram (TIC) between control, PNM, and Glu bind-
ing experiments showed clear differences, with new peaks corre-
sponding to multiple binding events detected (Figure S4) In fu-
ture work, it might be of interest to assess the limits of de-
tection for protein-ligand complexes across a range of binding
affinities
3.3 Application of small ID hardware to characterize therapeutic complexes
For therapeutic complexes, reproducibility, sensitivity, and specificity gains must be balanced with the speed of analysis We sought to benchmark the utility of the consistently best performing column, the 1 × 50 mm h-HST/HO-PEO BEH column, across protein therapeutic applications Antibody hexamer structures, for exam- ple, routinely need characterization to qualify higher order struc- ture features, including relative quantification of subunit to intact species, clipped species and glycoforms Yet, due to sensitivity is- sues exacerbated by the challenge to efficiently transmit high m/z
ions, native MS is generally performed with direct infusion and static spray tips even when an SEC-UV or SEC-MALS method has already been established [ 35, 48-50] For the first time, we have detected a 900 kDa RGY antibody hexamer species from online na- tive LC-MS This allows the accurate quantification of the hexamer
to monomer ratio and to look at monomer glycoform enrichment within the hexamer As shown in Figure S5, the hexamer species
is chromatographically resolved from the antibody monomer The most abundant free monomer Ab species was 529.1 Da less in mass than the hexamer-dissociated monomer ( Fig 5) This mass difference corresponds to a HexNAcHex 2 residue, confirming prior work that showed higher-mass glycans are enriched in the hex- amer complex [35]
This online SEC approach extends to hexamers formed from different noncovalent protein subunits In Fig.6, the elution pro- files of a three-protein complex (74 kDa) are shown This protomer structure assembles into a larger hexamer complex to form a 18 protein ternary structure of ∼443 kDa There was a 0.3 min differ- ence between the hexameric protein and monomer elution times, enabling relative quantification and the potential to screen across batches of drug product ( Fig 6ABCD) In Fig.6F, the trimer pro-
Trang 10tomer spectra is clearly observed Comparing the spectra of 6E and
6F, two unique m/z distributions of the dimer are observed, which
provides distinction between the in-solution and gas phase gener-
ated species
Protein-antibody complexes may also be examined with mi-
croflow SEC-MS Lipase-antibody complexes are historically diffi-
cult to analyze due to the extensive glycan heterogeneity of the
lipase and the micromolar dissociation constants of the affinity in-
teractions [ 51, 52] Interrogating these complexes is critical to refin-
ing downstream process parameters for a new drug, because host
cell lipases can bypass purification steps and be carried through
to therapeutic products [53] To design effective purification strate-
gies, the nature of these interactions must be characterized We re-
cently published work showing that certain complexes can be de-
tected by direct infusion static spray, ion mobility, and microscale
thermophoresis (MST), but these approaches are not amenable to
use as high throughput screening techniques [36] In the case of
MST, there can also be issues with labeling artifacts Thus, an SEC
method has long been desired As shown in Fig 7, a PLBL2-IgG4
complex was detected at ∼5% the level observed by static spray
However, SEC enabled the analysis of protein:Ab complexes at a
2:1 ratio, rather than a 10:1 lipase: Ab ratio Microflow SEC seems
to have minimized ion suppression problems encountered with di-
rect infusion This advantage opens up the possibility of generating
concentration dose curves for complex formation, which would not
be achievable by static spray The utility of the small dimension
columns seems therefore to lie in its sensitivity gains, its preserva-
tion of native states, and its subunit-level separations
4 Conclusion
The need to improve the sensitivity and softness of native MS
analyses is particularly pressing in therapeutic areas, where screen-
ing of native protein-ligand binding must be performed in an auto-
mated and high throughput manner SEC-MS is traditionally a slow
technique (at least ten minutes long) but it offers online separa-
tions, reproducibility, and a potentially universal, broadly applica-
ble platform Here, fast elution times of less than 1 minute and
no more than 3.5 minutes are achieved to better meet throughput
demands Moreover, 10-fold and higher increases in signal are re-
ported Limits of detection were driven into the high attomole and
lower ranges by employing the microflow devices Microflow SEC
as shown here also affords minor, but reproducible, separations be-
tween subunits As such it is possible to detect multiple types of
quaternary structures Coupled to multichannel emitters, we have
also demonstrated an improvement in maintaining the native state
of protein complexes such that it has become possible to detect
micromolar affinity complexes
Clearly, there is a penalty to miniaturizing SEC without simul-
taneously making wholesale changes to the LC flowpath The extra
column tubing and dispersion effects are of significant influence to
the apparent performance of the 1 mm ID column Nevertheless, a
device that is capable of achieving half the effective peak capacity
of an optimized 4.6 mm ID SEC column represents a step forward
in downsizing size exclusion chromatography and creating increas-
ingly powerful hyphenated analytical approaches
Declaration of Competing Interest
The authors declare the following competing financial inter-
est(s): Several of the authors are employed by Waters Corporation,
the manufacturer of the prototype columns used for this work and
several are employed by Genentech, Inc., which develops and mar-
kets drugs for profit BEH TM, ACQUITY TM, and UPLC TM are trade-
marks of Waters Technologies Corporation Vanquish TM is a trade-
mark of Dionex Softron GmbH Byos TM is a trademark of Protein
Metrics, Inc Q Exactive TM is a trademark of Thermo Fisher Scien- tific
Data availability
The authors do not have permission to share data
Acknowledgements
The authors would like to thank Yeliz Sarisozen and Nicole Lawrence for providing SEC packing materials, Mathew DeLano for helping to procure different types of column hardware, and Steven Byrd for the preparation of packed columns For assistance
in obtaining protein samples, we thank Bingchuan Wei and Shrenik Mehta We thank Wayne Fairbrother for scholarly conversations
Supplementary materials
Supplementary material associated with this article can be found, in the online version, at doi: 10.1016/j.chroma.2022.463638
References
[1] E Boeri Erba, C Petosa, The emerging role of native mass spectrometry in characterizing the structure and dynamics of macromolecular complexes, Pro- tein Sci 24 (8) (2015) 1176–1192, doi: 10.1002/pro.2661
[2] A C Leney, A J.R Heck, Native mass spectrometry: what is in the name? J Amer Soc Mass Spectrom 28 (1) (2017) 5–13, doi: 10.1007/ s13361- 016- 1545- 3
[3] G Chen, L Tao, Z Li, Recent advancements in mass spectrometry for higher order structure characterization of protein therapeutics, Drug Discov Today 27 (1) (2022) 196–206, doi: 10.1016/j.drudis.2021.09.010
[4] J.F Kellie, J.C Tran, W Jian, B Jones, J.T Mehl, Y Ge, J Henion, K.P Bateman, Intact protein mass spectrometry for therapeutic protein quantitation, phar- macokinetics, and biotransformation in preclinical and clinical studies: an in- dustry perspective, J Am Soc Mass Spectrometry 32 (8) (2021) 1886–1900, doi: 10.1021/jasms.0c00270
[5] V Kuzyk, G Somesen, R Haselberg, CE-MS for proteomics and intact protein analysis, in: A Colnaghi (Ed.), Separation Techniques Applied to Omics Sci- ences Advances in Experimental Medicine and Biology, Springer, 2021 [6] J Sastre Toraño, R Ramautar, G de Jong, Advances in capillary electrophoresis for the life sciences, J Chromatogr B 1118-1119 (2019) 116–136, doi: 10.1016/j jchromb.2019.04.020
[7] M Dawod, N.E Arvin, R.T Kennedy, Recent advances in protein analysis by capillary and microchip electrophoresis, Analyst 142 (11) (2017) 1847–1866, doi: 10.1039/c7an00198c
[8] L Schachner, G Han, M Dillon, J Zhou, L McCarty, D Ellerman, Y Yin,
C Spiess, J.R Lill, P.J Carter, W Sandoval, Characterization of chain pairing variants of bispecific IgG expressed in a single host cell by high-resolution native and denaturing mass spectrometry, Anal Chem 88 (24) (2016) 12122–
12127, doi: 10.1021/acs.analchem.6b02866 [9] J.F Valliere-Douglass, W.A McFee, O Salas-Solano, Native intact mass deter- mination of antibodies conjugated with Monomethyl Auristatin E and F at in- terchain cysteine residues, Anal Chem 84 (6) (2012) 2843–2849, doi: 10.1021/ ac203346c
[10] K Muneeruddin, C.E Bobst, R Frenkel, D Houde, I Turyan, Z Sosic, I.A Kaltashov, Characterization of a PEGylated protein therapeutic by ion ex- change chromatography with on-line detection by native ESI MS and MS/MS, Analyst 142 (2) (2017) 336–344, doi: 10.1039/C6AN02041K
[11] S Tamara, M.A den Boer, A.J.R Heck, High-resolution native mass spectrome- try, Chem Rev (2021), doi: 10.1021/acs.chemrev.1c00212
[12] M Zhou, C Lantz, K.A Brown, Y Ge, L Paša-Toli ´c, J.A Loo, F Lermyte, Higher- order structural characterisation of native proteins and complexes by top- down mass spectrometry, Chem Sci 11 (48) (2020) 12918–12936, doi: 10.1039/ D0SC04392C
[13] M.S Wilm, M Mann, Electrospray and Taylor-Cone theory, Dole’s beam of macromolecules at last? Int J Mass Spectrom Ion Processes 136 (2) (1994) 167–180, doi: 10.1016/0168- 1176(94)04024- 9
[14] E.W Chung, D.A Henriques, D Renzoni, C.J Morton, T.D Mulhern, M.C Pitkeathly, J.E Ladbury, C.V Robinson, Probing the nature of interactions
in SH2 binding interfaces–evidence from electrospray ionization mass spec- trometry, Protein Sci 8 (10) (1999) 1962–1970, doi: 10.1110/ps.8.10.1962 [15] P Su, X Chen, A.J Smith, M.F Espenship, H.Y Samayoa Oviedo, S.M Wil- son, H Gholipour-Ranjbar, C Larriba-Andaluz, J Laskin, Multiplexing of elec- trospray ionization sources using orthogonal injection into an electrodynamic ion funnel, Anal Chem 93 (33) (2021) 11576–11584, doi: 10.1021/acs.analchem 1c02092
[16] W Kim, M Guo, P Yang, D Wang, Microfabricated monolithic multinozzle emitters for nanoelectrospray mass spectrometry, Anal Chem 79 (10) (2007) 3703–3707, doi: 10.1021/ac070010j