A New Parallel High-Pressure Packing System Enables Rapid Multiplexed Production of Capillary Columns

Reversed-phase HPLC is the most commonly applied peptide-separation technique in MS-based proteomics. Particle-packed capillary columns are predominantly used in nanoflow HPLC systems. Despite being the broadly applied standard for many years, capillary columns are still expensive and suffer from short lifetimes, particularly in combination with ultra-high-pressure chromatography systems. For this reason, and to achieve maximum performance, many laboratories produce their own in-house packed columns. This typically requires a considerable amount of time and trained personnel. Here, we present a new packing system for capillary columns enabling rapid, multiplexed column packing with pressures reaching up to 3000 bar. Requiring only a conventional gas pressure supply and methanol as the driving fluid, our system replaces the traditional setup of helium-pressured packing bombs. By using 10× multiplexing, we have reduced the production time to just under 2 min for several 50 cm columns with 1.9-µm particle size, speeding up the process of column production 40 to 800 times. We compare capillary columns with various inner diameters and lengths packed under different pressure conditions with our newly designed, broadly accessible high-pressure packing station.


In Brief
Capillary columns in nanoflow application are a basis of highperformance LC-MS-based proteomics but are difficult or expensive to procure. We present a new high-pressure packing station for capillary columns (1000-3000 bar) together with detailed instructions for reproducing our setup. Compared with previous publications, this produces capillary columns 40 to 800 times faster at undiminished chromatographic performance.
Reversed-phase HPLC is the most commonly applied peptide-separation technique in MS-based proteomics. Particle-packed capillary columns are predominantly used in nanoflow HPLC systems. Despite being the broadly applied standard for many years, capillary columns are still expensive and suffer from short lifetimes, particularly in combination with ultra-high-pressure chromatography systems. For this reason, and to achieve maximum performance, many laboratories produce their own in-house packed columns. This typically requires a considerable amount of time and trained personnel. Here, we present a new packing system for capillary columns enabling rapid, multiplexed column packing with pressures reaching up to 3000 bar. Requiring only a conventional gas pressure supply and methanol as the driving fluid, our system replaces the traditional setup of helium-pressured packing bombs. By using 10× multiplexing, we have reduced the production time to just under 2 min for several 50 cm columns with 1.9-mm particle size, speeding up the process of column production 40 to 800 times. We compare capillary columns with various inner diameters and lengths packed under different pressure conditions with our newly designed, broadly accessible high-pressure packing station.
State-of-the-art MS-based proteomic pipelines typically consist of a sample preparation workflow to digest proteins and harvest pure peptides, an LC system for peptide separation, a mass spectrometer, and a sophisticated bioinformatics pipeline for raw data interpretation and subsequent statistical analysis (1, 2). The LC system plays a central role by partially separating the complex mixture of tens of thousands of peptides in a time-resolved manner according to their physicochemical properties, making them ultimately manageable for the MS system over the course of a gradient (3,4). The most widely applied technique for highperformance applications is reversed-phase separation, originally introduced in the 1970s (5). In essence, chromatographic systems are made of programmable pumps with the ability to form a gradient of a mixture of different agents. In the case of reversed-phase LC, the stationary phase is nonpolar, separating analytes by hydrophobicity over the course of a gradient of an increasing nonpolar mobile phase. The LC system is coupled to the mass spectrometer by electrospray (ES) ionization via an emitter (6). Glass or steel needles are commonly connected to the column. Particlepacked capillaries for chromatography can also be used for ES without being coupled to an additional emitter (7)(8)(9). These basic attributes are shared by most LC-MS systems, and differences are mainly defined by operational flow. Nanoflow LC operates at flow rates of several hundred nanoliters per minute and is the standard in proteomics because of the high sensitivity obtainable.
High flow rates in the μl to ml range, applied to columns with large inner diameters (IDs), are typically used in highthroughput or industrial-scale analysis and analytical MS application areas. Although these microflow and analyticalflow systems limit sensitivity, recent work has demonstrated robust and reproducible performance (10,11). Reproducibility and stability of those systems are high, but drawbacks are lowered sensitivity and a need for high sample amounts. Compared with developments in sample preparation, MS instrumentation, scan modes, and software, the LC apparatus has been largely unchanged in cutting-edge MS-based proteomics. Although identifications in proteomics experiments have doubled in single-shot experiments, this can mainly be traced to improvement on the MS instrumentation and software (12)(13)(14)(15)(16)(17). Current trends in LC developments aim rather toward systems for higher throughput and increasing robustness required for clinical applications (18), whereas the race for better separation in single-shot high performance runs with increasingly higher pump pressures has been comparatively abandoned. Consequently, a typically used setup for maximum sensitivity and performance for most experiments still consists of columns around 75-μm ID with a length of 20 to 50 cm, packed with sub-2-μm particles. Although, better performance could be reached by longer columns or smaller particles, both conditions would result in higher operational pressures that tend to make the LC systems unstable (4,19). For example, very high pressures can lead to leaks in the LC flow paths, resulting in poor reproducibility and subsequently a loss of measurement time.
Commercially available capillary columns in the aforementioned dimensions are expensive, especially considering how frequently they must be replaced (e.g., in our laboratories, a 50 cm column with 75-μm ID has an average turnaround time from 10 to 14 days). Therefore, many high-throughput laboratories produce packed capillaries in-house. Empty glass capillaries, ready to be packed and used, can be either purchased or produced from cheap polyimide-coated capillaries using a laser puller. Typically, a gas pressure system is deployed to pack such columns with particles in the low μm range, and instructions on the manufacturing process can be found online with open access (https://proteomicsresource. washington.edu/docs/protocols05/Packing_Capillary_Columns. pdf). However, this process is inherently slow, and interesting methods have recently been established with the aim of speeding up the packing process with high pressure (20) or dense bead slurry, as in the FlashPack method (21).
Combining these principles, we here present a highpressure packing system for capillary columns using a high-concentration bead slurry that has previously been described as beneficial for column performance (22). These high slurry concentrations and packing pressures of 1000 to 2000 bar allow us to achieve packing times for 50 cm columns in the minute range with our system, compared with hours for traditional procedures. Deploying a manifold system and a pump capable of high flow rates further multiplexes packing to up to ten columns simultaneously and makes column production 40 to 800 times more time efficient than in previous systems. We observe consistently good column performance for packing pressures at over 1000 bar with no adverse effects on the column backpressure and lifetime, while packing times continued to decrease at higher pressures. We provide a detailed blueprint of the system so it can readily be set up in interested laboratories (supplemental Table S1).

Preparation of Fused Silica
Fused silica from Polymicro (TSP075365 for 75-μm ID, TSP100365 for 100-μm ID, or TSP150365 for 150-μm ID) was cut to 140 cm. Polyimide coating was removed by a Bunsen burner and the silica surface was polished with an ethanol-soaked tissue in the middle of the cut capillary at a width of 2 cm. An ES emitter tip was pulled with a laser puller (Sutter P2000) at the polished part of the capillary resulting in two empty capillary columns ready to be packed.

Sample Preparation: Protein Digestion and in-StageTip Purification
HeLa cells were cultured in high-glucose Dulbecco's modified Eagle's medium with 10% fetal bovine serum and 1% penicillin/streptomycin (all from Life Technologies, Inc). Cells were counted using a countess cell counter (Invitrogen), and aliquots of 1 × 10 6 cells were washed twice with PBS (Life Technologies, Inc), snap-frozen, and stored at −80 • C. Sample preparation was carried out with the PreOmics iST kit (www.preomics.de). We used one HeLa pellet with one million cells per cartridge, determined the peptide concentration after peptide cleanup via NanoDrop, and adjusted the peptide concentration to 0.2 mg/ml.

Ultra-High-Pressure LC and MS
Samples were measured using LC-MS instrumentation consisting of an EASY-nLC 1200 ultra-high-pressure system (Thermo Fisher Scientific), coupled to an Orbitrap Exploris 480 instrument (Thermo Fisher Scientific) using a nano-ES ion source (Thermo Fisher Scientific). Purified peptides were separated on high-pressure packed columns as described in the Results and Discussion section. For each LC-MS/MS analysis with 75-μm ID columns, 500 ng peptides were used. For 100-μm ID columns, 888 ng peptides were used, and for 150-μm ID columns, 2000 ng peptides were used to adjust for the higher column volume. Peptides were loaded in buffer A* (2% acetonitrile (v/v), 0.1% trifluoroacetic acid (v/v)) and eluted with a linear 105 min gradient of 5 to 30% of buffer B (0.1% formic acid, 80% (v/v) acetonitrile), followed by a 10 min increase to 95% of buffer B and a 5 min wash of 95% buffer B. For the 75-μm ID columns, the flow rate was 300 nl/min, 535 nl/min for 100-μm ID columns, and 1200 nl/min for 150-μm ID columns to adjust for linear flow velocity. The column temperature was kept at 60 • C by an in-house developed oven containing a Peltier element, and parameters were monitored in real time by the SprayQC software. MS data were acquired with a Top15 datadependent MS/MS scan method. MS1 automatic gain control target was set to 300% in the 300 to 1650 m/z range with a maximum injection time of 25 ms and a resolution of 60,000 at m/z 200. Fragmentation of precursor ions was performed by higher-energy C-trap dissociation with a normalized collision energy of 30 eV. MS/MS scans were performed at a resolution of 15,000 at m/z 200 with an automatic gain control target of 100% and a maximum injection time of 28 ms. Dynamic exclusion was set to 30 s to avoid repeated sequencing of identical peptides.
Each column was equilibrated with two 120 min HeLa runs before the representative run for column cross-comparison.

Data Analysis
MS raw files were analyzed by MaxQuant software, version 1.6.11.0, and peptide lists were searched against the human Uni-Prot FASTA database (release 2019_01, 188441 entries). A contaminant database generated by the Andromeda search engine was configured with cysteine carbamidomethylation as a fixed modification and N-terminal acetylation and methionine oxidation as variable modifications. We set the false discovery rate to 0.01 for protein and peptide levels with a minimum length of seven amino acids for peptides, and the false discovery rate was determined by searching a reverse database. Enzyme specificity was set as Cterminal to arginine and lysine as expected using trypsin and LysC as proteases. A maximum of two missed cleavages were allowed. Peptide identification was performed with an initial precursor mass deviation up to 7 ppm and a fragment mass deviation of 20 ppm. All proteins and peptides matching to the reversed database were filtered out.

Experimental Design and Statistical Rationale
The overall experimental design was focused on making different capillary columns for proteomics experiments as comparable as possible. To achieve this, statistical analysis was performed from triplicate experiments for the packing time and pressure performance experiments. Experimental conditions for column cross-comparisons were chosen to eliminate outer influences, including measurements on similar LC and MS systems and equilibration procedures.

A High-Pressure Packing Chamber for High-Density Bead Slurries
A central challenge of nano-flow chromatography in proteomics laboratories is the constant demand for new capillary columns. Owing to their costs, commercial columns cannot be treated as a disposable item. However, in our hands, we frequently observe peak performance only for a short life span for ultra-high-performance applications. Therefore, to reach the needed quantity and cost requirements, we and many other laboratories produce own capillary columns. However, the throughput of production is limited, especially for columns with a small ID and extended length such as the 50 cm 75-μm ID columns used in most applications in our laboratories. We produce pulled or fritted capillaries and pack them with solid phase material, typically sub-2-μm C18 beads. A skilled person can pull hundreds of empty columns within a day, and fritted columns are also easy to produce. However, the packing process is inherently low-throughput and error-prone, which makes high-performance columns prized items in MS laboratories. In particular, the use of longer column lengths is -in our experience-a precondition for ultra-highperformance.
We hypothesized that high-throughput packing of capillary columns could be achieved by highly concentrated bead slurries (21) in combination with very-high-pressure packing (>1000 bar) (20). However, an increased packing pressure and bead slurry concentration can lead to column blocking, slowing down and eventually halting the packing procedure. Chloroform as a bead solvent was reported as an approach to avoid this issue because it can solvate higher bead concentrations. However, in combination with our bead particles, we observed poor chromatographic performance during proteomic experiments. Instead, we combined elevated packing pressure with the FlashPack system (21), which prohibited bead aggregation at the column entrance via stirring.
To test our concept, we constructed a custom-made chamber for high-pressure packing, where the pressure derives from a conventional HPLC system (EASY-LC 1000 in our case). The device consists of a central chamber, containing the bead slurry and magnetic stirring bar, and has three openings. A large-bore access allows filling the chamber with the bead slurry, a microbore fitting holds the capillary entrance into the chamber, and a nanoviper connection is used as an inlet for the pressure from the HPLC system (supplemental Fig. S1). The slurry applied to pack columns in this system can be highly concentrated. To prepare the slurry, we mixed about 100-μl of bead particles with 500-μl of methanol. After brief vortexing and 1 minute of sonication in a sonication bath, we let the slurry settle for 5 min, whereupon we loaded 200-μl of slurry into the chamber with a 500-μl Hamilton pipette. The prototype packing chamber enabled us to fill single capillaries within minutes using the HPLC high-pressure pumps (950 bar). However, this system was not suited for highthroughput column production, and moreover, the low pump volume of the HPLC system resulted in noncontinuous packing as the pump had to be refilled several times until a column was filled with beads.
Encouraged by aspects of our newly devised packing system, we set out to further streamline column production. We replaced the small-volume HPLC pump with a Maximator HDpump (Experimental Procedures). This high-flow continuous system converts driving gas from a standard laboratory gas supply line at a pressure ratio of 1:660 to a fluid outlet with a maximal pressure rating of 4000 bar and maximal flow capacity of 140 ml/min (Fig. 1). To use the FlashPack principle, we used methanol as the packing medium, which settles C18 beads at the chamber bottom (supplemental Fig. S2). The high flow capacity allowed us to implement multiple pump outlets for multiplex packing of up to ten columns with our station. We redesigned the original packing chamber to fit high-pressure connections (supplemental Fig. S3). For optimal stirring, we further created a rack system with magnets mounted on electric motors via 3D printed components to fit directly underneath the packing stations (detailed in Experimental Procedures and supplemental Fig. S4). Moreover, we connected a high-pressure range manometer to monitor packing pressure and added a pressure relief valve for efficient and controlled depressurization of the system, a notoriously timeconsuming process. Although the system is typically running at 1500 bar in our laboratory, the relief of pressure takes only 60 s, without flowback from the running beads from the capillary. In addition, the system is secured from capacity exceeding driving gas pressure by a control valve, which prevents the pump to be exposed to a higher input than 6 bar. As with conventional packing systems, the weakest connection is the sealing of the capillary to the high-pressure chamber. We used a standard polyether-ether-ketone ferrule used in HPLC applications in combination with a newly designed, reinforced polyether-ether-ketone screw cap (supplemental Fig. S3D) to pin the column under very high pressure. Nevertheless, if the system pressure exceeds the durability of the material, the column is ejected. Owing to the low compression capabilities of methanol, this is dangerous if one has body parts directly above the fitting when a rupture occurs and hence this must be prevented. Compared with gas, which can compress much more than liquid, no explosion risk should arise from our new packing station. To pass health and safety standards, we set up the packing system in a chemical hood with air circulation to pump off any methanol or bead particle aerosols and minimize the possibility for physical contact.

Ultra-Fast Column Packing
The time required to fill a capillary column with beads depends on two variables, the bead concentration of the packing slurry and the flow rate through the capillary. Empty capillaries with a pulled ES emitter have high flow rates in the μl/min range even for conventional gas-based packing bombs with lower pressure (<100 bar). However, as the bead bed grows, the flow rate through the column decreases drastically. Hence, the high-density bead slurry of FlashPack enables short packing times especially for shorter columns (21). We anticipated that combining this principle with the potentially high flow rates of our extremely high-pressure system would significantly reduce packing times.
To quantify the production throughput of our system, we consecutively packed 50-cm capillaries with 75-μm ID at different pressures (1000-2500 bar) and measured the time required. With a freshly filled bead reservoir, packing at the lowest tested pressure took on average 4.7 min. Increasing pressure to 2000 bar results in packing times just over a FIG. 1. High-pressure packing station. The scheme of the high-pressure packing station with detailed description of the crucial parts. The high-pressure pump is powered by a driving gas inlet and increases the pressure of a packing medium that is provided in a large volume flask by 660-fold. The compressed packing medium is channeled to ten packing chambers and placed on top of a magnetic stirring rack. A manometer is installed to monitor the system pressure and a pressure-release valve to facilitate time-efficient system depressurization. The inset depicts a packing chamber in detail, including high-pressure fittings, a stirring bar, and a capillary column. minute. Even higher pressure did not result in faster packing. Overall, our system decreased the time for making a single column 10-to 100-fold compared with previous packing procedures (20,21) (Fig. 2, A and B). Of note, the total production throughput is even higher due to multiplex packing and the option to quickly exchange capillaries and bead slurries. This results in a speed-up factor of 40 to 800 (Fig. 2, C). Once filled with bead slurry and mounted on the high-pressure system, the packing chambers can be used to pack several columns consecutively. This merely requires depressurizing the system via the pressure relief valve and exchanging the filled columns with empty capillaries. Consecutive packing of several columns from the same reservoir will decrease the packing speed because of the removal of beads from the reservoir. To fully restore packing speed, the bead chamber has to be opened and refilled, which takes about 10 min for all ten chambers together. Typically, we refilled the reservoir after five capillary exchanges. The average turn-around cycle for producing ten columns is thus 20 min, allowing the production of hundreds of columns in a working day (Fig. 2, D). An additional advantage of the high-throughput system is that it allows us to discard improperly packed columns, which occur in approximately 10% of cases.
The high-pressure system faces the same two main challenges as usual packing stations, which are particle clogging within the capillary and bead aggregation at the column entrance. Particle clogging can only be avoided by clean working conditions. This means dust-free storage and clean cutting of fused silica and the use of filtered fluids and dust-free particles for bead slurry preparation. Bead aggregation from dense slurry can be circumvented by optimized stirring conditions according to the FlashPack principle (21).

Influence of Packing Pressure on Column Performance
To evaluate the effect of packing pressure on column performance on realistic samples, we analyzed three of our laboratory standard HeLa digests on each column. Across all packing conditions, we observed no significant variation in the number of identified peptides and protein groups (Fig. 3, A/B). Moreover, the median peak widths of identified peptides were comparable for all conditions (Fig. 3, C). Correlation between the noncorrected retention times of peptides analyzed using columns produced at varying pressures was remarkably high (Pearson correlation coefficient >0.996) and not significantly altered from replicates packed with similar pressure conditions (Fig. 3, D).
Another factor often used to characterize column performance is the tailing factor that can be calculated as depicted in Figure 3, E (23). Usually, the peak width at 5% peak height is used for peak width calculation but in proteomics experiments where tens of thousands of peaks are investigated, the baseto-base peak width is typically calculated, although full width at half maximum is also often given. We decided to calculate the peak tailing at baseline as a metric. In general, the distribution of peak shapes was wider than what would be expected from an analysis run of few analytes, but the median typically centered around the optimum of 1. The median of the peak tailing at baseline was below 1.0 for the lower and shifts above 1.0 for higher packing pressures up to a median of 1.2 (Fig. 3, F). In the literature, tailing factors in the range between 1 and 1.2 are often described (24). The shift towards this range with the higher packing pressures could result from denser compressed bead beds. As described above, the general performance was not altered for the proteomics metrics, which leads us to the conclusion that the minor change in peak tailing at baseline with higher packing pressures is not FIG. 2. Comparison of packing times. A, packing times of single columns as described in previous efforts and for different packing pressures (data collected in triplicates, displayed with SD) with a detailed view of the tested pressure conditions (B). C, production time for ten columns considering multiplexing (2× multiplexing for the system of Kovalchuk et al. and 10× for the system presented here) (20,21). D, times of a packing cycle of 10 × 5 columns, taking a total of 100 min with filling of the reservoir and changing of capillaries between the actual packing steps. Error bars indicate the standard deviation from triplicate measurements. C, median peak widths at baseline of identified peptides. D, distribution changing the LC-MS performance. This manifests in an only slightly altered distribution of peak widths between representative experiments of columns packed at different pressures (Fig. 3, G). From the correlation of peptide retention times, it is visible that for all representative comparisons, the peptides elute in a narrow and reproducible time window that is not influenced by the applied packing pressure. This retention time stability is accompanied by similar separation properties of the different columns, which can be visualized directly by the peak width at baseline of analyzed molecules. Figure 3, G shows bulk analysis of all identified peptides with nearly overlapping peak width at baseline distributions, whereas the minor differences do not constitute a significant trend toward a better performance for lower or higher packing pressures of capillary columns. We did not observe a significant change in column backpressure from the different packing conditions. Based on these results, it seems that the packing pressure has no or only minimal effect on the column performance.

LC-MS Performance of Columns With Different Lengths and IDs
The length and ID of capillary columns allow their adaptation to a plethora of sample materials and LC systems, specifically regarding separation power and backpressure. In MS-based proteomics, 75-μm ID columns in combination with flow rates in the range of 200 to 400 nl per minute are typical. Hence, we packed such capillary columns with different lengths (20, 30, 50 cm) with our high-pressure system and compared their performance. Packing time for the shorter columns was even faster and in the range of 30 s. The longest columns produced the smallest peak widths and subsequently resulted in the highest numbers of identified peptides and proteins (Fig. 4, A and B). Interestingly, the distribution of peptide intensities did not change significantly, and the peak tailing at baseline also remained unaffected (Fig. 4, C and D).
Over the last years, the demand for high-throughput analysis has become apparent for the analysis of clinical samples, especially blood plasma as we have described before (25). This has been addressed by a novel HPLC principle with preformed gradients and slightly higher flow rates (18) and by higher-flow systems operating in the high microliter per minute range (10,26). As these strategies require columns with a higher ID to maintain acceptable pressure during analysis, we produced columns with 75-μm, 100-μm, and 150-μm ID and tested their performance.
When comparing column IDs, the experimental setup has to be adapted to the conditions. To enable direct comparison of capillaries between different IDs, we scaled the flow rates to reach the same linear velocities and the amount of input material to the column volume (Experimental Procedures). For the 100-μm ID columns, this results in a flow rate of 535 nl/min and 888 ng of peptides for loading, whereas for the 150-μm ID column, 1200 nl/min and 2 μg of peptide material was loaded to be comparable to the 300 nl/min and 500 ng used for the 75-μm ID columns. This requirement of higher sample amount already limits the applicability of larger column diameters for samples with limited accessibility. The 1400 μl of pump volume from the Easy-LC 1200 used for the experiment was sufficient to run a 2 h gradient with the 150-μm ID column, but longer gradients or higher flow rates would exceed the capabilities of the LC system and require lower flow rates. The higher column IDs led to slightly broader peak widths, but peptide and protein identifications were not affected. Owing to the correction of the sample input amount, we did not see a difference in the peptide intensity distributions, and the peak tailing at baseline was also not affected by the column ID (Fig. 4, E-H). CONCLUSION Here, we aimed to increase the throughput and to streamline the production of capillary columns for MSbased proteomics. We provide a detailed list for the commercial parts and blueprints describing the construction of our high-pressure packing station. The setup can be built at relatively low costs (<$10,000), compared with the cumulative expenses for high-performing commercial columns. We designed this new station to fill multiple columns simultaneously within a few minutes, which accelerates the packing process of capillary columns more than a 100-fold compared with traditional gas pressure-driven stations. In this way, we hope our system helps researchers by streamlining the often work-intensive and fragile column production process. In addition, the extreme high pressures enable the packing of long, high-performing columns (>50 cm). The ability to produce high-performing columns at high-throughput allows for the possibility of only using capillary columns at the peak of their performance, replacing them as soon as peak broadening or decreased ionization is observed. Reassuring in terms of robustness of the packing process itself and the stability achieved at exceedingly high pressures, we have not observed variation in the performance characteristics over a wide range of packing pressure from 1000 to 3000 bar. We hope the technology described here will enable laboratories of any of Pearson correlation coefficients calculated on peptide retention times between columns packed at the same pressure and columns packed at different pressures (p-value of unpaired t-test for difference: 0.6). E, visualization of the tailing factor calculation, this is typically done at 5% peak height. F, Peak tailing at baseline for all identified peptides from runs with 75-μm ID columns and different packing pressures. G, correlation of peptide retention times across packing conditions. The density of peptides is color-coded. The histograms show the peak widths at baseline distribution of five representative runs. ID, inner diameter. size to mass-produce high-performance long capillary columns.

DATA AVAILABILITY
The MS-based proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository and are available via ProteomeXchange with identifier PXD024296.
Supplemental data -This article contains supplemental data (21).