Nano liquid chromatography columns

Nano liquid chromatography (nanoLC), with columns having an inner diameter (ID) of ≤ 100 μm, can provide enhanced sensitivity and enable analysis of limited samples. NanoLC has become an established tool in omics research, and is gaining ground in other applications as well. There are several variants and formats of nanoLC columns, including packed columns, monoliths, open tubular columns, and the pillar array format. Most applications are done with packed columns, while e.g. the monolith and open tubular columns are still less established as routine tools. The pillar array format is a new variant with excellent resolution and low backpressure, and has recently been commercialized and used for bio-applications. In this minireview, we summarize and discuss recent research on nanoLC column development and uses, focusing on literature between 2016 and medio 2019.


Introduction
Liquid chromatography (LC) is widely used for separating compounds prior to detection, and continues to be a key tool in analytical chemistry. However, the chromatographic process dilutes the analytes from injection to detection, and this can affect sensitivity, a central factor in the analysis of e.g. limited samples. A key factor in the dilution is the inner diameter (ID) of the column 1 . By reducing the column ID, the concentration of eluted compounds becomes higher, and enhanced sensitivity is obtained when coupled with concentration sensitive detectors like the electrospray ionization mass spectrometer (ESI-MS). Columns with reduced IDs include the microscale variant (about 1 mm ID), the capillary LC variant (about 0.2-0.5 mm ID), and the nanoLC variant, herein defined as a column with cylinder formed tube with ID ≤ 100 µm or as a chip format column with ≤ 100 µm channel depth/width. Columns with ID ≤ 100 µm may also be included in the capillary column notation 2 . A recent review paper by Novotny gives a nice historical account on the development of capillary LC all the way to the nanoscale 3 . The nanoLC columns can in both formats (tube and chip) be filled with particles (particle packed) or with a porous continuous monolithic structure (monolith), or have an open structure (open tubular). In addition, pillar array columns can be used in the chip format 4 .
In this mini-review, we focus entirely on nanoLC due to its strong advantages related to sensitivity, but also, in many cases, chromatographic resolution. NanoLC in combination with MS is also a powerful combination, and has proven to be an immensely important tool for proteomics 5 . Also in metabolomics, nanoLC-MS has become an important technique because of its increased sensitivity relative to that obtained with larger ID columns 6 . Thus, much of the work described here also feature MS analysis. However, nanoLC is not limited to "omics", and has furthermore become of increasing interest for e.g. enantiomer separations 7,8 . Due to nanoLC´s compatibility with limited sample amounts, emerging bio-applications include single-cell analyses and increasing focuses on clinical analysis.
In this minireview, we discuss the characteristics, recent developments, applications, obstacles and future opportunities of nanoLC columns in papers published from 2016 until medio 2019. The development of instrumentation is not included; instead, the reader is referred to reviews on miniaturized LC instrumentation 9 and nanoLC platforms 10 . The small format and samples sizes of nanoLC can present challenges to the operator, often requiring add-ons such as online sample handling, but these issues have not been addressed in this review.
The authors reviewed the use of nano columns in proteomics in 2015 1 and several other recent reviews also show the applicability of nanoLC [5][6][7] . Subjects of nanoLC we focus upon are: packed columns, monoliths, open tubular LC columns, pillar array columns, the chip format, and extended-nanoLC.

Commercial columns with nanoLC dimensions and their application areas
At present, packed columns in the tube format are the most common columns in nanoLC. Packed columns are commercially available with inner diameter of 50 and 75 µm from several vendors (e.g. Agilent, Thermo Fisher, Waters, Sigma Aldrich, Dr. Maisch). These columns are packed with particles of 1.7 -3.5 µm, and come in various lengths, and typically with a C18 stationary phase. Very narrow ID columns (10-30 µm), intended for ultra-sensitive analyses, are commercially available from at least one vendor (CoAnn Technologies), and packed with 1.7 µm particles with C18 stationary phase.
The 75 µm ID columns have become more or less a standard in proteomic analyses [11][12][13] , and also for the glycated proteome 14 . The 75 µm ID C18 columns have also been applied in metabolomics 15 and for the determination of drug of abuse 16 .

New/in-house packed nanoLC columns
Several research groups prefer to pack their own nanoLC columns for various reasons, e.g. cost and flexibility. Here we provide some examples of this approach. A 75 µm ID column was in-house packed with C18 particles (1.9 µm) and used by the group of M. Mann for high coverage proteomics 17 19 . An in-house packed column (75 µm ID and 3 µm C18 particles) was also used in a study where electron ionization MS was used for elucidation of the free fatty acid profile in mussel samples 20 .
While the latter studies all used C18 stationary phases, an in-house made stationary phase was used for enantiomer separation in the nano format (75 µm ID) by D'Orazio et al. 21 Use of packed columns with ID less than 50 µm is not common. However, Shao et al recently packed 1.7 µm C18 particles into a 22 µm ID column, with a 3 µm ID tip, for single-cell proteomics 22 (Figure 1). A 70 cm long in-house packed PicoFrit column with 30 μm ID and a tip size of 10 μm was used in combination with a 4 cm long 100 µm ID trap column, using 3 µm C18 particles in both, by Zhu et al. in their nanodroplet processing platform for proteome profiling of 10 -100 mammalian cells 23 .

Particles used in nanoLC columns
Totally porous particles are still the most common in packed nanoLC columns, and most of the commercial columns are packed with silica-based totally porous particles. However, a few vendors (e.g. Thermo Scientific and Biotech) sell 75 µm ID columns packed with silica-based superficially porous (core shell) particles. Totally porous particles were used in most of the studies mentioned above [11][12][13][14][15][16][17][19][20][21][22] , and only one used core shell particles 18 . However, the benefits of core shell particles (e.g. high efficiency with reduced backpressure), and the fact that columns packed with such particles now are commercially available should increase their use in the future. For an overview of advantages of coreshell particles, see e.g. the review by Tanaka and McCalley and references therein 24 . Nonporous particles have also been used in the nanoLC column format. The group of Wirth has used submicrometer particles and found that the backpressure was reduced due to slip flow 25 , however, no recent studies using such particles has been published to the authors´ knowledge.

Sample introduction; trap column -analytical column combinations
To maintain a high efficiency with direct-injection nanoLC, only a few nLs should be injected. Hence, sample introduction methods which allow for more of the sample to be introduced (=better chance of analyte detection) without being detrimental to the column performance is wanted. Indeed, such systems are widely used in nanoLC. Typically, a trap column, also called pre-column or solid phase extraction (SPE) column, is used, although not always 16,20,22 . The ID of the trap column is typically larger than that of the analytical column. As an example, Levernaes et al. used a 1 mm ID trap column in combination with a 75 µm ID analytical column 12 . To minimize effects of void volumes, the trap column may be packed with larger particles or a material providing less retention relative to that on the analytical column in order to have some phase focusing and thus improved efficiency. Zhang et al.
used a 300 µm ID trap column packed with 5 µm C18 particles in combination with a 75 µm ID column packed with 2 µm C18 particles 14 . In addition, they used an in-house packed boronate affinity enrichment column (1 mm ID) upstream the C18 trap column in an online system for glycated peptides.
The trap column and the analytical column had the same ID (75 µm), but the trap column was packed with larger particles (3 µm vs. 2 µm) in a quantitative metabolomics study 15 . Berg et al. found that a C8 monolithic trap column performed better than a standard C18 packed trap column (both 50 µm ID and the same as ID of the analytical column) for targeted proteomics 18 . Schöbinger et al. 11 , carrying out the separation at ≥ 45 °C, pointed out that loading the sample at high temperature may cause loss of sample. They used a low-temperature mobile phase for trapping of peptides on a 300 µm ID trap column packed with 5 µm C18 particles, in combination with a 75 µm ID analytical column packed with 3 µm C18 particles. Another sample introduction system is the "speLC" system described by Falkenby et al 26 . The speLC uses C18 StageTips (small C18-based SPE microcolumns with a peptide binding capacity of up to 5 μg).
Instead of eluting fractions into the autosampler vial of the LC system, a low-pressure pump passes a 5-10 min gradient through the StageTip and then to the analytical column. The speLC−system eliminates sample-to-sample carry-over by using disposable StageTips. In a recent paper, the same authors have improved their approach by capturing the analytes from the StageTip into a long capillary loop with a pre-formed gradient, allowing subsequent isocratic pump operation for the analytical column 27 . The authors claim that this system, which is now commercialized (Evosep One), provides sensitivity, throughput and robustness, and is applicable for large clinical studies.
Another approach to reduce sample-to-sample carry-over and increase throughput is that by Spencer et al 19 . They developed a trap column-exchanging robot that was equipped with four in-house packed 150 µm ID trap columns, packed with 4 µm C12 particles. They claim that the retention time was sufficiently repeatable using the four different trap columns without the need for rescheduling the selection windows, as long as the columns were repeatable packed.
To achieve high coverage proteomics another novel approach (called "spider fractionator") has been introduced 17 . In this approach, fractions from a 250 µm ID and 30 cm long column packed with 1.9 µm C18 particles and used with a high pH mobile phase, were transferred to the 75 µm ID analytical column also packed with 1.9 µm C18 particles (Figure 2).
In summary, packed columns are still the workhorse of nanoLC, with fully-porous particles dominating applications, but alternatives such as core shell materials are emerging along with decreasing particle sizes. Sample introduction techniques with these familiar materials continue to be developed. Key advantages include robustness and commercial availability. Key disadvantages include limitations on resolution related to back-pressure constraints and difficulties in particle-packing very narrow columns.

Monolithic columns
Monolithic columns can be categorized into three main types depending on their composition; organic polymer-based, silica-based and organic-silica hybrid monoliths 28 .

Commercial nanoLC monoliths
While no longer the case regarding packed nanoLC columns, the number of commercial monolithic nanoLC columns is very limited 29 . Apparently, only one vendor produces (50 and 100 µm ID) polymer-based monoliths, and two produce (50 and 100 µm ID) silica-based monoliths. The PepSwift™ phase intended for peptide separations is based on a poly(divinylbenzene-coethylvinyl-benzene-styrene) copolymer. The ProSwift RP-4H™ phase intended for intact protein separations is based on a poly(divinylbenzene-co-ethylvinylbenzene) co-polymer with similar selectivity as the PepSwift™ phase.
The silica-based Chromolith® CapRod® nanoLC columns have a C18 stationary phase, but is also available with a C8 stationary phase, and in a selection of internal diameters (

Applications of monolithic nanoLC columns
The extent to which these commercially available columns are used in applications, is not easily found  Organic polymer monoliths: Organic-based monoliths have a long history 39 , and are considered to be rather simple to make, as opposed to their silica-based sibling. An overview of the state-of-the-art and guidelines in tuning the macropore structure of polymer-based monoliths can be found in a recent review by Dores-Sousa et al 37 . Two papers from 2016 describe the trends in the development of porous polymer monoliths 39,40 . The research group of Hanfa Zou has used click polymerization for making both organic monoliths and organic-silica hybrids, and state in their review from 2016 that the efficiency of these is greatly improved compared with organic monoliths prepared by free radical polymerization 41 , which is more commonly used. Table 1 gives an overview of the most recent advances in organic polymer monolithic columns.  and Kennedy) that ultralow aspect-ratio columns generate a markedly lower dispersion than larger aspect-ratio columns 50 . The lowest capillary-to-domain size aspect-ratio column had the best performance, but was still inferior to the open-tubular format. Kobayashi et al. made a 100 µm ID silica monolith according to a prior reported procedure and used octadecyltrimethoxy silane to prepare a low-density octadecyl (ODS) monolith which was additionally modified, to study the effect of acidic mobile phase additives for peptide separations 51 . They found both high peak capacity and sensitivity using cyanoacetic acid as mobile phase additive with these phases.
Hybrid monoliths: It has been argued that silica-based monoliths are both tedious to prepare and have low chemical and mechanical strength, and this has been the reason for investigating other approaches to obtain high efficiency monolithic columns, such as organosilicon-based hybrid columns. A review published in 2017 covers the advances in organic-silica hybrid monoliths up to mid 2016 28 .
In Table 2 we present some recent studies focusing on hybrids. Most of the columns have been made using an one-pot (one-step), either photo-initiated or thermally initiated, polymerization.  ( The recent review paper on the use of monolithic columns for intact proteins 68 , also shows that mostly standards and not real biological samples have been chromatographed. In a review by Shibasaki et al.
on the molecular and physiological study of Candida albicans they refer to a study where an in-house made 470 cm long and 100 µm ID C18 silica-based monolith was used 69 .
In summary, there is undoubtedly a high activity in the field of monolithic separation column development for nanoLC, with a large variety of selectivities. However, monolithic columns do not appear to be widely used in modern applications, as opposed to packed nanoLC columns. Key advantages of monolithic columns include low back pressures that allow for fast separations/long column separations. Key disadvantages include difficulties in reproducibility and commercial availability.

Open tubular columns
Due to their theoretical high efficiency and small sample volume consumption open tubular (OT) columns have been a topic in LC since the late seventies 70 , however, the increased interest in such columns in the last decade stems from the work of Karger's group. They were the first to show the great potential of OT columns using a 10 μm ID column with a porous layer of poly(styrene-codivinylbenzene) for proteomics 71  The injection was carried out using split-flow injection, and detection by on-column fluorescence detection using a confocal microscope. The mesopore size, and hence surface area, could be controlled by the temperature used during the hydrothermal treatment 78 . Recently they reported increased hydrophobicity of 5 µm ID silica-based OT columns by applying hybrid TMOS/methyltrimethoxysilane (MTMS) layers with inserted methyl groups. Due to higher hydrophobicity, thinner porous layers gave similar retention factor (k) as in octadecylsilylated columns synthesized using TMOS only. Since thinner layers have a lower intra-layer mass transfer resistance, superior column efficiencies were obtained compared to that of TMOS-based porous layer OT columns giving the same retention 79 . These columns obviously have a great potential for use in applications where high resolution is needed for samples of limited amount.
OT columns with 2 µm ID have been prepared by Chen et al. 80 , and Yang et al. 81  layer, and report run-to-run retention time repeatability below 1% 83 . In the more recent paper 84 , the same group reports a multi-lumen capillary (also with 126 parallel channels of 4.2 μm ID) with a C18functionalised silica porous layer OT column for both on-capillary preconcentration and separation, followed by MS detection. Following modification, 100% of the channels displayed a homogenous porous silica layer, 257 ± 36 nm thick. They state that the multi-channel structure allowed the capillary to be applied at higher flow rates which simplifies system requirements and increases detection options, however, the separation efficiencies could be improved. The possibility of using a multichannel capillary for increased sample loading has also been explored by Ribeiro da Silva et al. 85 , who used a multichannel capillary with 126 parallel channels of 8 µm each as trap column. The channels were coated with a layer of poly(styrene-co-octadecene-co-divinylbenzene) (PS-OD-DVB).
The trap column was coupled online with a 10 μm × 2 m poly(styrene-co-divinylbenzene) (PS-DVB) OT LC column with nanospray mass spectrometry detection. Compared to using monolithic/particle- monolithic trap column with sample capacity (>2000 ng on a 10 cm column; comparable to the capacity of commercial particle-packed columns) in combination with a 10 μm × 3000 mm OT column functionalized with octadecyl groups bound to a silica skeleton that coats the wall of the column 76 . The approach used by Rodriguez et al. 84 , and Ribeiro da Silva et al. 85 may also be a solution to look into.

OT columns with larger ID
Even though inferior efficiency is expected, several papers report the use of larger ID capillaries to prepare columns with various chemistries to obtain the desired selectivity. Peng

OT capillaries without coating
Liquid chromatography has also been performed with micelles in OT capillaries 93  However, Duan et al. performed protein separations in 300 and 500 nm ID cylindrical self-enclosed nanocapillaries, although they point out that the separation cannot be explained by hydrodynamic chromatography alone 97 . In their paper, they also present normal phase, reversed phase and ionvalence chromatography in the nanochannels, which are integrated in a chip, and also can be classified as belonging to the extended-nano LC category (see below).
In summary: the open tubular format is still under development, but columns are not commercially available. Despite several significant advances and applications, the OT format continues to be a niche format. Key advantages include high sensitivity and resolution. Key disadvantages include lack of commercial availability and high demands on the operator.

Chip-format
As Although monolithic beds are rather easily made in microchip channels, particle packed channels are more used.

Commercial columns
Vendors (e.g. Agilent, Waters, Sciex) offer packed column microchips tailored for proteomics, small molecules, and other applications, as well as custom chips. Typically, a 40-150 mm long analytical column is used in combination with a short trap column, which may be of different chemistry for increased selectivity. Information on the bed width and depth is not always easily obtained; however, 75 µm and 85 µm appear to be a common bed width in the nano format. Some challenges of pressuredriven chip LC have been reported, and Lotter et al. have studied various approaches to connect pressure-resistant glass chips with HPLC pumps up to 500 bar 102 .

Applications of chip-format columns
Due to the commercial availability and ease of use, microchip LC-MS has been widely applied, in proteomics [103][104][105] , but also for other applications [106][107][108] . However, because the technique is often used as an aid in solving a research problem, the use of chip LC is not necessarily revealed in literature searches.

New chip-format columns
The need for specialized equipment for making the chip format, has led to that much of the development is now carried out by companies, and less has been done by individual groups lately.
A glass chip with a 35 mm long column segment (with width 90 μm and depth 40 μm packed with C18 In summary, the chip format continues to be developed, especially commercially but also noncommercially, and is applied in a variety of contexts. Key advantages include ease of use. Key disadvantages include limited flexibility.

Pillar array columns
Desmet and co-workers have, based on earlier work by Regnier and co-workers, developed an exciting new format of nanoLC separation columns, which they call micropillar array columns (PACs). Ultra-high efficiencies are obtained with an optimized pillar diameter (5 μm) and interpillar distance (2.5 μm) 113 .
Such columns have been used in combination with MS for peptide mapping of monoclonal antibodies (mAbs) and antibody-drug conjugates (ADCs) 4 .
Such columns are now commercially available as PharmaFluidics' µPAC™ columns in the chip format.
The perfected, and importantly, reproducible order of the separation bed virtually eliminates axial peak dispersion, resulting in a high efficiency. The freestanding nature of the pillars leads to much lower backpressure allowing the use of very long columns and shorter conditioning. Commercial columns come in length of 50 cm and 2 m with C18 end-capped stationary phase, with a trap column with the same stationary phase, in the cylindrically shaped pillar form. µPAC™ columns have been shown to provide excellent proteomics capabilities, e.g. consistent identifications of more than 5,000 proteins using 10-hour long gradients 114 (Figure 4). The only drawback of these columns seems to be the price. They describe the application of a sol-gel procedure on radially elongated pillars (REPs) using tetramethoxysilane and methyltrimethoxysilane, with subsequent octadecylsilylation 115 . An increase in accessible specific surface by a factor of 112 compared to a nonporous REP was observed. Plate heights as low as 0.4-0.8 µm (k = 0-1. 97) and kinetic plot analysis demonstrated that the column will deliver more theoretical plates per unit of time than a 5 µm core shell packed bed when plate number higher than 1.0 × 10 4 is required. They have followed up this study which focused only on the on-chip performance with a study on attainable performance under practical conditions (i.e. a sufficiently long column in commercial LC hardware with external detection) 116 . Separation of alkylphenones and peptides was studied in a 16.5 cm long, 1 mm wide channel (three lanes of 5.5 cm long channels connected by turns). The minimum plate height of 1.4 μm for octanophenone (k = 2.21) observed in isocratic mode was 5 times smaller than the smallest off-column plate height previously reported for porous pillar array columns for a retained component. This advantage is related to the earlier introduced shape of the radially elongated pillar bed that outperforms the cylindrically shaped pillar bed in terms of the plate height.
Furthermore, they have adjusted the preparation conditions to make a 1.2-fold thicker layer on the porous layer REP array column 117 . The mesoporosity of the layer was controlled by changing the hydrothermal treatment temperature from 105 °C to 80 °C. When performing a 180 min gradient elution on a 16.5 cm long column, the peak capacity for an alkylphenone mixture was 315 and 365 for the combination of thin layer and large mesopores, and thick layer and small mesopores, respectively.
For peptides, the thicker layer was still favorable, providing a conditional peak capacity of 245 for a commercially available peptide mixture. However, lager mesopores were more advantageous for large molecules (> 1000), because of less content of small pores which hinder the diffusion of large molecules in pores in the layer.
In summary: The pillar array format, carefully developed over a number of years for optimizing chromatographic traits, is now commercially available, and has been shown to be a powerful tool in e.g. proteomics. Key advantages include very high chromatographic performance at low pressures. Key disadvantages include high costs for commercial products (as of today).

Extended-nanoLC
Kitamori and co-workers have developed a technique which they call "extended-nano LC" 118 . They perform pressure-driven chromatography in channels which are down to  100 nm wide and deep.
The separation column is an extended-nano fluidic channel which is fabricated on a glass chip.
Advantages of extended-nanoLC are the use of extremely small sample volumes, the speed and the high separation efficiencies (plate numbers of up to 1.4 x 10 4 ) 119 . In their review paper from 2017 118 , fundamentals of the extended-nano chromatography technique are summarized, as is the instrumentations used to realize attoliter sample injections and sensitive detection methods. The application of the extended-nanoLC system for analysis of a small sample (39 fL) from a single living human cell has been demonstrated in combination with the femtoliter sampling interface 120 ( Figure   5). Gradient elution is also possible with the extended-nano LC format, which Kitamori and co-workers also call femtoliter LC. Shimizu et al. have developed a flexible gradient system using standard HPLC pumps and an external mixer with a simple sample injection system, and showed its potential for separation of intact proteins 121 . The separation nanochannel size was of 950 nm depth, 5.0 µm width and 10 mm length in this case, and the inner surface of the channels was modified with octadecylsilyl (ODS) groups.
In summary, although in its infancy, extended-nanoLC is a technological approach that may push the separation technology towards subcellular analysis. Key advantages include a potential for analyzing very small samples. Key disadvantages include a lack of commercial products and high demands to the operator (as of today).

Conclusions
Even though it has existed for decades, the nanoLC column continues to be developed and be a key tool in a wide range of cutting-edge research areas. The traditional packed column is being refined and is still the most popular choice, especially in the common tube format, but also in the chip format.
However, it is beginning to see a contender in the pillar array column, which is now commercially available. There is great enthusiasm for this variant, especially in proteomics environments. Although older than the pillar array column, the monolith and the open tubular variants are prone to being bypassed in popularity, perhaps due to their reputations as being difficult to reproduce or operate. On the other hand, the monolith shows an enormous versatility, and may be applied for a number of challenging applications. However, it is therefore important that column developers demonstrate their innovations with actual samples and "killer applications" rather than relying on standards and simple protein mixtures.

Conflicts of interest
There are no conflicts of interest to declare.