Introduction

Among several separation modes in HPLC including normal-phase (NPLC), reversed-phase (RPLC), ion exchange, size exclusion, chiral separation and affinity modes, the RPLC mode, with columns packed with octadecyldimethylsilyl-bonded (C18 or ODS) silica particles, has been practiced most widely because of its high separation efficiency for a wide variety of solutes. However, RPLC has a significant drawback in the separation of polar solutes including derivatives of sugars, carboxylic acids, amino acids, and peptides, though these species are very important in the field of the life science related to genomics, proteomics, metabolomics, pharmacology, agrochemistry and so on. In RPLC, polar solutes are often eluted as unretained peaks without enough separation due to weak interaction between the solute and the stationary phase. To avoid this problem, the use of a ODS silylating reagent containing polar groups [1] or C30 silylating reagent [2] has been proposed.

As another approach, HILIC (hydrophilic interaction liquid chromatography) mode separation was invented to separate polar solutes effectively [3]. This separation mode utilizes polar stationary phases and less polar mobile phases, commonly a mixture of an organic solvent and water, and it belongs to the NPLC modes. Indeed, bare silica can be used for an HILIC mode separation, although it has a limitation in compatibility of the stationary phases with solutes such as amines, peptides and proteins. In many cases, silica particles functionalized with amino [4] and amide groups [5] or polar polymeric particles [6] are used as packing materials. Generally in a HILIC mode, polar solutes are retained strongly in 100% organic solvent as a mobile phase, and they are eluted by mixing water with the mobile phase, which increases the polarity of the mobile phase. Since this system has a separation selectivity orthogonal to that of an RPLC mode, the combination of the HILIC mode and RPLC mode in multidimensional HPLC is promising [7].

HILIC mode LC-MS (mass spectrometry) with ESI (electrospray ionization) or APCI (atmospheric pressure chemical ionization) processes to determine polar compounds have been reported [8–10]. Recently, the successful separation of sugar derivatives and peptides by HILIC was reported [11–21], and carbohydrate separation by HILIC was reviewed [22, 23].

Both bare silica and amino silica columns [20] have problems with chemical stability, i.e., the short lifetime of these columns under aqueous conditions. To improve these circumstances, the preparation of more stable silica columns chemically modified with polar functional groups would be a reasonable approach. Chromatographic use of the monolithic silica columns has attracted considerable attention because they can potentially provide higher overall performance than the particle-packed columns, based on the variable external porosity and through-pore size to skeleton size ratios for faster separation or higher separation efficiency than the conventional HPLC columns [24–27]. Particularly, the monolithic silica capillary columns are known to show high separation efficiency, providing up to 200,000 theoretical plates [28, 29]. Since highly efficient HILIC columns are attractive for the separation of complex mixtures, we investigated the method of surface modification of the monolithic silica columns in capillary. There is an example of the use of a monolithic column in HILIC mode, which used a column coated with didodecyldimethylammonium bromide [21]. However, the use of monolithic silica columns under HILIC conditions is limited. Here we report the chemical modification of monolithic silica capillary columns by an on-column polymerization of acrylamide to prepare an HILIC column. This preparation method is free from the compatibility problem of silylating reagents with some polar functionalities and the difficulties of the column packing process.

Experimental

Materials

Reagent-grade methanol (Nacalai Tesque, Kyoto, Japan) was used after a single distillation. Reagent-grade acetonitrile (ACN) (Wako Pure Chemical Industries, Osaka, Japan) was used after distillations from phosphorous pentoxide and from anhydrous potassium carbonate. Ion-exchanged water was distilled, followed by a filtration using Barnstead E-Pure Model D46 (Thermolyne, Dubuque, IA, USA). For the preparation of the stationary phases, 3-aminopropyltriethoxysilane (APS, Chisso, Tokyo, Japan), methacryloyl chloride, acrylamide, and ammonium persulfate (Wako) were used without further purification. The following compounds were used in this study: thymine, uracil, cytosine, guanine, uridine, adenosine, cytidine, and guanosine (all from Wako); adenine (Aldrich, Milwaukee, WI, USA); arabinose, xylose, glucose, mannose, garactose, lactose, maltose, cellobiose, maltopentaose, maltohexaose, and maltoheptaose (all from Nacalai Tesque); and maltotriose and maltotetraose (Seikagaku Corporation, Tokyo, Japan). Sugar isomers and homologues, from arabinose to maltoheptaose were derivatized using 2-aminopyridine, acetic acid, and borane-dimethylamine complex (all from Nacalai Tesque) by a previously reported method [30–32]. All sample solutions were prepared at a concentration of 1 mg/ml in water.

HPLC equipment

A chromatographic evaluation of a commercial HILIC column, TSK gel Amide-80 (5 μm, 2 mm ID×15 cm, TOHSOH, Tokyo, Japan, abbreviated as Amide-80) was carried out by an HPLC system, PU611 pump (GL Sciences, Tokyo, Japan), an Injector Model 7125 (Rheodyne, Park Court, CA, USA), a CO630 (GL Sciences) column oven (30 °C), and an UV620 (GL Sciences) detector using mixtures of ACN and water as mobile phases.

A chromatographic evaluation of the monolithic silica capillary columns was carried out by a LC-10AD VP pump (Shimadzu, Kyoto, Japan) and a CE1575 (Jasco, Hachioji, Japan) detector. Samples were injected with a Rheodyne Injector Model 7725 with a split flow injection mode. Capillary columns were evaluated at ambient temperatures using mixtures of ACN-water as a mobile phase. Off-column detection was carried out by using a capillary (50 μm ID) possessing a detection window 10 cm from the column outlet. Data were collected at 210 or 254 nm, and processed by EZChrom Elite (Scientific Software, GL Science) throughout the experiments.

Preparation of columns and the chemical modification

All monolithic silica capillary columns were prepared from tetramethoxysilane (TMOS) by the previously reported method in 100 μm ID fused-silica capillaries [33]. Bare silica columns are described as MS(100) throughout this article, and (100) stands for their ID. Then the columns were modified with anchor groups, N-(3-trimethoxysilylpropyl)methacrylamide [34]. A 1:1 mixture of silane and pyridine was passed through the column for 24 h at 80 °C by N2 pressure (1 MPa) followed by a wash with toluene, and methanol for 24 h, both by N2 pressure (1 MPa). This anchor-bonding step was repeated twice to obtain the monolithic silica capillary columns modified with the methacrylamide functionality. Then a monomer solution shown in Table 1 was charged in the column and allowed to react at 60 °C for 1 h. After the reaction, the column was washed with water and methanol for 24 h, both by N2 pressure (1 MPa). The washing process was repeated twice to obtain the polyacrylamide-modified monolithic silica capillary columns. Throughout this article, PAMS(100)-x is used to refer to the polyacrylamide-coated monolithic silica columns (100 μm ID).

Table 1 Conditions of the polymer-coating step
Full size table

Results and discussion

Preparation of polymer-coated monolithic silica capillary columns

At the first step of modification, an anchor, N-(3-trimethoxysilylpropyl)methacrylamide, was chemically bonded to the silica skeleton at 80 °C. At that temperature, no thermal polymerization of the anchor was detectable by a 1H-NMR measurement of the residue. Porous silica particles (YMC, particle size 5 μm, micro pore 120 Å, specific surface area 330 m2/g) were also treated with the same silane under the same conditions to the capillary columns. The anchor-bonded silica (29.0 g) was obtained from 25.3 g of bare silica, and the surface coverage was estimated as 2.5 μmol/m2 (an increase in weight was attributed to the reaction with the silylating reagent).

In the second step, the silica was treated with acrylamide in the presence of ammonium persulfate under the conditions shown in Table 1. In a similar manner, the anchor-bonded silica particles were also subjected to the reaction in order to provide information on the polymer-coating process for conditions 1 and 2. Increases in weight were obtained: 0.10 g was observed for condition 2 (Table 1) for 2.26 g of silica particles, but in the case of condition 3, viscous agglomerates including silica particles were obtained. The viscosity average molecular weight (M v) of poly(acrylamide) obtained under conditions 1 and 2 was obtained by a viscosity measurement using an Ubbelohde viscometer: aqueous solutions of poly(acrylamide) were examined at 25 °C. The Mark-Houwink-Sakurada equation [Eq. (1)] was applied to obtain the M v of poly(acrylamide), where [η] stands for the intrinsic viscosity with the constants K=0.631×10−4 ml/g and α=0.8 for poly(acrylamide) in water [35, 36].

$${\left[ \eta \right]} = KM^{\alpha }_{{\text{v}}} $$
(1)

The M v values of polyacrylamide obtained under conditions 1 and 2 were found to be 280,000 and 120,000, respectively. When condition 3 was applied to a monolithic silica capillary column for the polymer-coating step, the column was clogged with the polymer, and could not be washed with water.

Evaluation of the polyacrylamide-coated monolithic silica capillary columns

The column performance was evaluated in 90% ACN in terms of the permeability K [Eq. (2)], the number of theoretical plates N, the plate height H, and the separation impedance E [Eq. (3)], where u, L, η, ΔP, and t 0 stand for the linear velocity of the mobile phase, the column length, the viscosity of the mobile phase, the column back pressure, and the column dead time, respectively [37]. These parameters are shown in Table 2 for Amide-80, PAMS(100)-1, and MS(100)-4. The three-times larger K value of the monolithic columns than the particle-packed column means greater permeability, which results in a lower column back pressure, suggesting the possibility of using higher flow rate than for the conventional column, or of using longer columns to obtain higher column efficiency. A PAMS(100)-2 column also showed similar column efficiency to that of the PAMS(100)-1.

$$K = \frac{{u\eta L}}{{\Delta P}}$$
(2)
$$E = \frac{{\Delta Pt_{0} }}{{\eta N^{2} }} = {\left( {\frac{{\Delta P}}{N}} \right)}{\left( {\frac{{t_{0} }}{N}} \right)}{\left( {\frac{1}{\eta }} \right)} = \frac{{H^{2} }}{K}$$
(3)
Table 2 Comparison of column efficiencies
Full size table

Generally, “good columns” show E=2,500–3,000 in RPLC. The Amide-80 showed E=14,000, while the PAMS(100)-1 showed E=3,200 for uridine in 90% ACN at 1.0–1.5 mm/s linear velocity (u). The large E value observed for the Amide-80 column might be due to a large extra-column band broadening associated with the small column size. The column efficiency of the Amide-80 column, however, was found to be lower for the late-eluting solutes than for the early-eluting solutes, suggesting that the low performance observed was caused by the column itself. The higher performance of the PAMS column can be explained by the lower column pressure drop and greater permeability due to the large through-pore size-to-skeleton size ratio, and simultaneously large N of the monolithic columns. The E values for the C18 monolithic silica capillary columns were smaller than 1,000 in 80% ACN at u=1.0–1.5 mm/s, and for a C18 particle-packed column, E was ca. 3,000 under the same conditions [28]. It is also possible that the difference in the retention mechanisms may have caused the greater decrease in the column efficiency in HILIC mode than in RPLC. HILIC conditions provide stronger intermolecular, solute-stationary phase interactions such as hydrogen bond and dipole-dipole interactions than dispersion interactions in RPLC [38].

The phase ratios of monolithic silica capillary columns prepared from TMOS were estimated to be ca. 1/6 those of the particle-packed columns, and the small k values were understandable. Total porosity of a monolithic silica column prepared from TMOS is typically ca. 95%, while a particle-packed column usually possesses 70–75% total porosity [39]. The difference in the amount of silica in a column results in a five- to six-fold difference in the phase ratio. This is a disadvantage of the monolithic silica capillary columns. On the other hand, the separation efficiency of the capillary column was better than that of the particle-packed column: the N value for uridine was 24,000 for the 38 cm column as shown in the chromatogram in Fig. 1.

Fig. 1
figure 1

Chromatograms of nucleic base and nucleoside separations on a PAMS(100) (a) and an Amide-80 (b) column. HPLC was carried out under ambient temperatures with a detection λ of 260 nm. a Column: PAMS(100)-1 (100 μm I.D. ×38 cm), mobile phase: 90% ACN, linear velocity: u=1.2 mm/s at ΔP=2.1 MPa. b Column: TSKgel Amide-80 (2 mm I.D. ×15 cm), mobile phase: 90% ACN, flow rate: 0.2 ml/min, linear velocity: u=1.0 mm/s at ΔP=3.6 MPa. Solutes: 1 thymine, 2 uracil, 3 adenine, 4 uridine, 5 adenosine, 6 cytosine, 7 guanine, 8 cytidine, 9 guanosine

Full size image

Separation of nucleosides and nucleic bases by the HILIC systems

The retention behaviors of the monolithic silica capillary columns and the Amide-80 were examined using mixtures of ACN–water as a mobile phase and nucleosides and nucleic bases as samples. The retention factors, k values, are plotted against various ACN concentrations in Fig. 2.

Fig. 2
figure 2

Plots of k values of nucleic bases and nucleosides in various ACN–water mixtures on a a PAMS-(100)-1 and b an Amide-80 column. See Fig. 1 for solutes

Full size image

The PAMS(100)-1 retained nucleosides well in 90% ACN, and the degree of retention decreased drastically in 80% ACN. As the concentration of ACN was decreased, k decreased gradually until it reached 20% ACN. The tendency was very similar to that for the Amide-80 column. In the water-rich mobile phases, k values became larger again, which may be caused by the hydrophobic interactions between solutes and the main chains of the polymer. Judging from these results including those of the same experiment using the Amide-80, the use of 100–60% ACN should be appropriate for the HILIC mode separation of the nucleosides and nucleic bases.

The k values for these solutes on the PAMS(100)-1 and MS(100)-4 capillary columns were plotted against k values of the Amide-80 in Fig. 3. The k(Amide-80) vs. k[PAMS(100)-1] plot showed a linear correlation which suggests very similar retention mechanisms of these two columns. In contrast, the k(MS(100)-4) showed poor correlation against the k(Amide-80). The retention features of the bare silica columns were much smaller than the k[PAMS(100)-1] values. The monolithic amide column gave much smaller k values than the particle-packed column. The high column efficiency of the PAMS(100)-1, however, allowed the separation of these samples as the Amide-80 column as shown in Fig. 2.

Fig. 3
figure 3

Comparison of k(monolith columns) vs. k(Amide-80) in 90% ACN for the separation of the nucleic bases and nucleosides. See Fig. 1 for solutes on a PAMS-(100)-1 (open squares), and a bare silica MS(100)-4 (filled diamonds)

Full size image

The present method of modifying monolithic silica columns was found to be effective for producing a stationary phase for HILIC mode, maintaining the higher permeability and the separation efficiency of the monolithic columns. Ihara et al. have reported the polymer-immobilized silica gels, but they attached the polymer chains to silylating reagents before bonding them onto silica particles [40]. This strategy will not be easily applicable for the modification of silica surface with water-soluble polymers due to the moisture-sensitive nature of silylating reagents.

Separation of sugar derivatives by the HILIC systems

HPLC separation of carbohydrates has been attracting attention recently from the biochemical point of view. Various stationary phases have been used for the separation of sugar derivatives, including APS bonded silica column, anion-exchange polymer-particulate columns at high pH, and cation-exchange polymer-particulate columns containing Pb2+ or Ca2+ [22]. Difficulties in HPLC separation of sugars include the problem of the anomer separations, low or no UV absorption of the sugar derivatives, and the existence of a huge number of isomers. The high efficiency of the monolithic silica capillary columns is expected to contribute to better separations. For easy detection, the use of 2-aminopyridine-labeled sugars (PA sugar) [22, 41] was examined. Figure 4 shows chromatograms for the separation of the sugar derivatives by a gradient mode. The PAMS(100)-1 column gave a good enough chromatogram.

Fig. 4
figure 4

Chromatograms of PA sugar derivative separations on a PAMS(100) (a), and an Amide-80 (b) column. HPLC conditions were ambient temperature, detection λ of 245 nm, mobile phase in ACN–water (0.2% formic acid), linear gradient of ACN 90% to 50 % (t G=30 min). a Column: PAMS(100)-1 (100 μm I.D. ×38 cm), linear velocity u=1.1 mm/s at ΔP=2.1 MPa. b Column: TSKgel Amide-80 (2 mm I.D. ×15 cm), flow rate: 0.2 ml/min, linear velocity: u=1.0 mm/s at ΔP=3.6 MPa. Solutes: 1 2-aminopyridine, 2 PA-xylose, 3 PA-arabinose, 4 PA-glucose, 5 PA-mannose, 6 PA-galactose, 7 PA-maltose, 8 PA-lactose, 9 PA-cellobiose, 10 PA-maltotriose, 11 PA-maltotetraose, 12 PA-maltopentaose, 13 PA-maltohexaose, 14 PA-maltoheptaose

Full size image

A mixture of aminopyridine, PA-arabinose, PA-glucose, PA-maltose, PA-maltotriose, PA-maltotetraose, PA-maltopentaose, PA-maltohexaose, and PA-maltoheptaose was separated by the PAMS(100)-1, bare monolithic silica capillary column, and the Amide-80 column under isocratic mode. The effects of concentrations of ACN or the retentions are shown in Fig. 5. Water mixed in mobile phases contained 0.2% formic acid throughout the separations of sugars. The MS(100)-4 column gave k=0.8–2 for the PA-oligo sugars in 90% ACN. In this mobile phase, PA-maltotetraose and larger analogues were not eluted within 20 min, while all tested sugar derivatives were eluted at less than k=0.5 in 80% ACN, and they were not retained in 70% ACN. The PAMS(100)-1 column eluted only 2-aminopyridine, PA-arabinose, and PA-glucose in 90% ACN within 20 min, while PA-maltotriose and all larger analogues tested well separated in 70% ACN. In 50% ACN, the column did not retain these PA sugar derivatives. Similarly, the Amide-80 also separated PA-maltotriose and all larger analogues in 70% ACN. The k value 42 was obtained for PA-maltoheptaose on the Amide-80 column, while it was 1.9 on PAMS(100)-1. Figure 6 shows k vs. k plots for the Amide-80 and the PAMS(100)-1 obtained by the measurements in 70% ACN. The linear correlation suggests that the retention behaviors of both columns were very similar for the sugar derivatives.

Fig. 5
figure 5

Plots of k values of PA sugar derivatives at various ACN concentrations on a an MS(100)-4, b a PAMS-(100)-1, and c an Amide-80 column. Water mixed in mobile phases contained 0.2% formic acid. See Fig. 4 for solutes

Full size image
Fig. 6
figure 6

Comparison of k[PAMS(100)] vs. k(Amide-80) plots in 70% ACN-30% water (0.2% formic acid). See Fig. 4 for solutes

Full size image

Further study on LC-ESI-MS application of these capillary columns for the separation and detection of carbohydrates in metabolites is in progress.

The glucose unit selectivities of the HILIC columns

To study the contribution of one glucose unit to the retention factor k, the glucose selectivity α values [Eq. (4)], where n is the number of glucose units), were plotted in Fig. 7.

$$\alpha = \frac{{k_{{n + 1}} }}{{k_{n} }}$$
(4)
Fig. 7
figure 7

The dependence of an α(glucose) value on the size of a solute observed on a PAMS(100)-1 (open shapes) and an Amide-80 (filled shapes) column in 60% (diamonds) and 70% (squares) ACN. Water mixed in mobile phases contained 0.2% formic acid

Full size image

The α values are regarded as α(glucose), the ratio of k based on the difference in a solute structure by one glucose unit number in sample compounds, as α(CH2) discussed in RPLC studies [42]. The average α(glucose) was 1.68 in 70% ACN, while it was 1.30 in 60% ACN for PAMS(100)-1. The results of the same treatment with the Amide-80 were very similar to those of PAMS(100)-1: the average α(glucose) was 1.70 in 70% ACN, while it was 1.33 in 60% ACN. As the glucose number increased from PA-glucose to PA-maltoheptaose, α values gradually approached constant values. The generally decreasing α(glucose) values indicate the decrease in the contribution of one glucose unit to retention. This is in contrast to the α(CH2) value in RPLC, where α(CH2) values are almost constant in one mobile phase irrespective of the size of molecules or the end group functionality. These results imply the contribution of adsorption type interaction rather than partition. If the results are caused by a decrease in the influence of the pyridylamino group on the whole sample molecules, the tendency would be the opposite of the present result because the PA group would contribute to similar retention. As shown above, these columns provide similar retention tendencies and selectivities for the separation of sugar derivatives, though k values on the capillary column were about 20 times smaller than those on a conventional HILIC column.

Conclusions

The monolithic silica capillary columns, PAMS(100), were found to be useful for HILIC mode separation due to the high column efficiency and higher permeability than a particle-packed HILIC column. Separation of nucleosides, nucleic bases, and sugar derivatives using the PAMS(100) was possible at 70–90% ACN concentration, while a bare silica column did not give good separation under these conditions. Generally, a method for increasing the k values on the monolithic silica columns is required, and the polymer-coating can be one of the solutions for this.