Selectivity of hydrophilic interaction liquid chromatography (HILIC) stationary

phases is crucial for choosing a column for separating target polar compounds, which is why it is important to know the effect of each functional group on the chromatographic properties of the material. Revealing the trends in selectivity changes with variations of adsorbent structure at different stages of preparation can help to regulate the interactions of the stationary phase with analytes and create novel materials with improved performance. This article compares several silica-based HILIC stationary phases with different types of coatings, including polar polymers, antibiotic macromolecules, zwitterionic, and amide groups. Synthesized adsorbents were characterized and compared using the Tanaka test for hydrophilic stationary phases. Their chromatographic performance was evaluated using model mixtures of neutral, zwitterionic, and negatively and positively charged polar analytes. The obtained results provided better understanding of the factors affecting stationary phase selectivity.

Synthesis of stationary phases for hydrophilic interaction liquid chromatography (HILIC) utilizes different techniques where a hydrophilic substrate, mostly silica, is covered with hydrophilic coating containing neutral or charged functional groups of different nature. The structure of those groups, as well as their arrangement in the layer relative to each other (as separate functional sites or uniform polymer layer), defines chromatographic performance of stationary phases in HILIC mode and their ability to retain and separate various classes of polar analytes.

Our research group synthesized and described HILIC stationary phases with various coatings including monomeric amide ligands prepared via the Ugi reaction (1,2), linear polymers (3,4), glycopeptide antibiotic (5), and a zwitterion (6) (Table I). Functional groups were attached to the substrate either directly or using 1,4 butanedioldiglycidyl ether (1,4-BDDGE) as a linker.

Using polymers is promising for obtaining stationary phases for HILIC. It leads to both hydrophilizing and shielding the substrate and thus decreasing its effect on the retention of the analytes. One of the suitable polymers is poly(ethylene glycol) (PEG), which is easily synthesized with a pre-defined mass polymer with a linear architecture. The presence of oxyethylene groups may lead to dipole–dipole interactions and the potential formation of hydrogen bonds. Polyethyleneimine (PEI) is another polymer which provides both high hydrophilicity and anion-exchange properties, ensuring the simultaneous determination of polar neutral compounds and anions.

Another way of shielding the substrate is attaching macromolecules. Glycopeptide antibiotics such as eremomycin (ERM) contain a lot of amide-, hydroxy-, and amino groups providing hydrophilicity and thus the prospect for obtaining HILIC phases in addition to chiral ones.

A promising approach for the creation of new functional layers for HILIC adsorbents is using multicomponent reactions. Their advantages are the high yield, ease of execution, “one-pot” reaction without isolating intermediate products, and flexibility of structure variations. The multicomponent Ugi reaction involves the interaction of amines, carbonyl compounds, isocyanides, and acids. Simultaneous incorporation of several types of polar functionalities including amides via this reaction can provide hydrophilicity to the phase and variability in its derivatives. The use of aminated silica as an amino component in this reaction proposed in (1,2) makes sure that the substrate is modified exclusively with the necessary functional groups, while possible byproducts of the competing Passerini reaction are formed only in the liquid phase and can be washed out.

Another promising set of HILIC phases are zwitterionic ones. They tend to be the most universal species allowing separation of analytes from different classes. Such a phase was obtained by covalent attachment of glycine to 3-aminopropyl silica via the linker 1,4-BDDGE (6). The structures of the discussed phases are given in Table I.

To characterize and compare the synthesized adsorbents, the Tanaka test for hydrophilic stationary phases (7) was used. The set of substances is recommended in the test to quantify the different types of selectivity parameters. Hydrophilicity of the adsorbent surface 

(U) is determined as the uridine retention factor. The degree of hydrophobic interaction 

) is represented by selectivity toward uridine and 5-methyluridine 

(5MU). Hydrophilic partitioning property as selectivity to the hydroxy group 

(2dU) using 2’-deoxyuridine (2dU). Hydrophobic ions sodium 

-toluenesulfonate (SPTS) and N,N,N-trimethylphenylammonium chloride (TMPAC) reflect anion- and cation-exchange properties α(AX) and α(CX), respectively. The phases are also characterized by their ability to separate configurational isomers vidarabine and adenosine 

(Tb/Tp) for theobromine and theophylline, which determine surface pH. Conditions are given in Table II.

In this work, model mixtures of sugars, amino acids, and water-soluble vitamins were also used as hydrophilic compounds representing neutral, zwitterionic, and positively or negatively charged polar analytes. Comparison of the chromatographic performance of the prepared stationary phases and discussion of the modification pathways leading to better selectivity and efficiency toward each class of analytes is presented.

According to the data given in Table II, hydrophobic 

) and configurational isomers selectivity 

(V/A) does not differ significantly for the obtained phases. Neutral character for S-ERM is determined since 

(Tb/Tp) = 1.09, while basic nature of the other adsorbents is supported with 

(Tb/Tp) < 1. Other parameters are discussed further in detail.

To reveal which of our modification ways allowed to obtain more hydrophilic adsorbents, neutral hydrophilic compounds of different structures were taken. Uridine retention factor was used from Tanaka HILIC test to reflect this characteristic (conditions are given in Table II and [7]). Hydrophilic retention was also evaluated using the set of neutral polar analytes—sugars and zwitterionic amino acids. An appropriate mobile phase composition was used for each model mixture. Acetonitrile and water were taken for sugars. Acetonitrile and phosphate buffer with 

 pH 6.5 was used for the amino acids, which is close to their isoelectric points (other details are given in Figure 1).

According to Table II, hydrophilicity k(U) of the obtained adsorbents measured using the Tanaka test decreases in the following order:

S-B-PEI > S-PEG > S-ERM > S-A > S-B-ERM > S-B-G

(U) values for the phases, the highest retention factors for sugars were registered for S-PEG (Figure 1a). Zwitterionic S-B-G phase with the lowest hydrophilicity 

(U) demonstrated strong retention of sugars similar to S-B-PEI with maximum 

’ values were obtained for the phases modified with eremomycin (S-ERM and S-B-ERM) characterized by moderate 

(U) values. These results showed that there is no correlation between hydrophilicity described via 

(U) in the Tanaka test and sugars retention. In fact, retention mechanism for carbohydrates is not only partitioning but also involves adsorption and formation of hydrogen bonds with OH- groups. The latter could have greater impact on retention for the phase where PEG containing H-acceptors is attached directly to aminopropyl silica as compared to PEI containing both H-donors and H-acceptors attached via the additional 1,4-BDDGE linker. Also, Schiff base formation is limited for sugars with linear PEI layer containing substituted amino groups. Therefore, it’s not easy to predict carbohydrates’ retention relying only upon the hydrophilicity parameter 

Amino acid retention showed a slightly better correlation with stationary phase hydrophilicity 

’ values were registered for the phases S-B-G and S-B-ERM with low 

(U) (Figure 1b). S-A showed medium retention values according to its 

’ values were observed for the most hydrophilic polymer-coated phases, but again S-PEG > S-B-PEI was opposite to their 

(U) values. The deviation from the trend could be caused by the effect of functional groups structure and several types of interactions defining the retention of polar analytes.

Therefore, Tanaka HILIC test parameter determining hydrophilic retention 

(U) cannot be exclusively applied for reflecting and prediction the retention of polar analytes from different classes between the columns of diverse functionalities and in the cases when different eluting conditions are used. Additional model analytes representing broader structure variations are useful to complement HILIC columns characterization and comparison.

Electrostatic interactions have a significant impact on HILIC retention for polar charged analytes. This contribution is evaluated by the slope value for the dependences of their retention factors on the counter-ion concentration in mobile phase. It was determined in (1) for ascorbic and nicotinic acids, where their retention factors decreased with increasing concentration of the eluting ion in the mobile phase. For the phases with different anion- exchange selectivity 

(AX), the slope, which reflected electrostatic impact, was identical for the same analyte.

Retention of acidic analytes depend on anion-exchange properties in HILIC mode. To determine anion-exchange selectivity 

(AX) in the Tanaka test, a hydrophobic organic anion sodium p-toluenesulfonate (SPTS) was used. Its retention is divided by 

(U) to account for the influence of hydrophilic interactions. The mobile phase for this test contains ammonium-acetate buffer with wpH 4.7. It is useful to reveal whether it is possible to predict the retention of charged analytes relying upon 

(CX) values, when the elution conditions necessary to ionize target analytes differ from those used in the test. Acidic vitamins were chosen as the test compounds. An ammonium-acetate buffer with wpH 5.8 was used for their separation as it provided the best selectivity within the model mixture of water-soluble vitamins. Also, at this border of the acetate buffer, capacity carboxyl groups of ascorbic and nicotinic acids should be dissociated.

As demonstrated in Table II, all the phases provide anion-exchange selectivity determined as in (8). High 

(AX) values were obtained for the phases containing the linker 1,4-BDDGE. This effect is caused by quarternization of nitrogen atoms on the substrate surface when three molecules of 1,4-BDDGE could be attached per each amino group (Table I). 

(AX) for S-PEG is also high indicating that terminal epoxy-groups of the initial reagent could also quarternize substrate’s nitrogen atoms. Increased 

(AX) values resulted, as expected, in higher retention of the acids. Figure 2 demonstrates the dependence of retention factors for nicotinic and ascorbic acids on 

 for nicotinic acid was about 0.97 which could be used for approximate prediction of its retention. Lower 

 observed for ascorbic acid (0.86) is consistent with involving different mechanisms in its retention shown in (1).

Ascorbic acid demonstrated stronger retention as compared to nicotinic acid on all the phases (Figure 2). This is consistent with its higher hydrophilicity, and thus reflects larger partitioning impact into its retention mechanism. The latter is supported by the results previously discussed in (1), where nicotinic acid showed the predominant impact of the adsorption mechanism in its retention, whereas partitioning balanced adsorption for ascorbic acid for all the described adsorbents.

For complex samples containing analytes of different charge, it is useful to evaluate the conditions for determination of both negatively and positively charged ones. For this purpose, the mixture of water-soluble vitamins was used which included positively charged thiamine and the acids mentioned above.

(CX) is defined in Tanaka test as N,N,N-trimethylphenylammonium chloride (TMPAC) capacity factor divided by 

(U) to account the impact of hydrophilic interactions. Since all columns possess anion-exchange selectivity, positively charged analytes are prone to repulsion effects. So, hydrophobic TMPAC had no retention on the majority of the columns and only S-ERM provided cation-exchange selectivity 

(CX) > 1 (Table II). Obviously, for hydrophilic positively charged compounds, the repulsion is compensated with hydrophilic interactions. For thiamine (B

), as such a cation example, the retention response on 

(AX) of the phases is shown in Figure 2. B

 retention factors decreasing with the increase of 

(AX) obviously confirms the key role of electrostatic repulsion from positively charged groups in the functional layers. Summarizing all the trends for charged analytes from Figure 2, it seems possible to make a choice of the column relying on its 

(AX) value, which provides reasonable retention and better selectivity for separation of both positively and negatively charged polar compounds.

For separation of water-soluble vitamins, ammonium acetate buffer was used in mobile phase according to (2). Its solubility in water-acetonitrile media allows the use of gradient mode and increases concentration to elute strongly retained acidic vitamins from the phases with high anion-exchange selectivity. Selectivity comparison for weakly retained water-soluble vitamins showed more similarities for the phases containing 1,4-BDDGE as the linker and more significant differences for other adsorbents (Figure 3). For the phases prepared without the linker, thiamine demonstrated notably higher retention, which resulted in its elution after riboflavin (B

 referred to nicotinamide was demonstrated by S-ERM phase. Using 1,4-BDDGE as a linker resulted in high anion-exchange selectivity of the respective phases (Table II) and repulsion of positively charged B

 with its low retention on S-B-PEI, S-B-G, and S-B-ERM phases. Its highest retention was observed for S-ERM phase where eremomycin (ERM) macromolecule could not fully shield the silica substrate, which is seen from cation-exchange selectivity 

(CX) > 1. In this case, dissociated silanol groups are available for electrostatic interactions and attraction of B

To compare the chromatographic characteristics of the presented homemade columns with commercially available ones (7,8), 

(OH) values for all columns were plotted against respective k(U) values (Figure 4a). S-B-PEI provided much stronger hydrophilic retention 

(U) and a little higher OH-selectivity than Hypersil GOLD HILIC with the same PEI functionality. Polymer-coated phases S-B-PEI and S-PEG demonstrated the highest OH-selectivity and hydrophilic retention 

(U) as compared to the majority of commercial columns and also home-made monomeric ones. Their 

(OH) > 1.8 matches up with the conclusion made in (9) for polymer-modified stationary phases. That could be due to thicker and more uniform coating and a variety of functional sites and groups provided by polymers as compared to the monomers which resulted in increased adsorbed water layer and thus hydrophilic partitioning property of the phase.

Among amide-containing phases, S-A demonstrated the highest OH- selectivity and k(U) close to XBridge Amide, where S-B-ERM, XBridge Amide and Amide-80 columns provided the same 

(OH) value. S-A also demon- strated hydrophilic selectivity α(OH) and k(U) retention close to those for Hypersil GOLD HILIC (PEI-type phase [8]) and amino phase NH

-MS. Low OH-selectivity was observed for S-ERM providing good hydrophilic retention k(U) matching up with k(U) for XBridge Amide and NH

-MS also containing amide or amino groups, respectively. S-ERM is the only phase in the set which is based on glycidoxypropyl silica as a substrate (Table I), and it is located in another region of the plot (Figure 4b). For the phases obtained using 3-aminopropyl silica good correlation of OH-selectivity and 

(OH) parameter represents the phenomenon of hydrophilic partitioning, and it could be used to estimate the thickness of the adsorbed water layer on the surface of the stationary phase. Based on the data given in (9), 1.0–1.2 nm water layer thickness could be adsorbed on the surface of our amide- and polymer-containing homemade phases. Thus, introducing highly hydrophilic amide groups or polar polymers provided the highest OH-selectivity, whereas polymer-coated phases demonstrated improved hydrophilic retention 

(U) among the homemade and commercially available columns.

For the homemade set of adsorbents, the selectivity for the analytes differing in OH-group in their structures was compared to α(OH) values obtained from the Tanaka test. For this purpose, we used the pairs of amino acids: Tyr/Phe and Ser/Ala. The mobile phase for amino acids separation required potassium phosphate buffer solution, wpH 6.5 and acetonitrile, 15:85 v/v. As it is presented in Table II, 

(Ser/Ala) and these pairs of acids had similar trends but no full correlation with 

(OH) of the adsorbents when different from Tanaka test conditions were used. However, the highest OH-selectivity for amino acids was obtained with polymer-coated phases characterized by high 

(OH). Remarkably, for S-A phase the reversed elution order was observed for Ser/ Ala when less hydrophilic amino acid retained stronger despite the moderate 

) for this phase. Some specific interactions probably define higher retention of alanine on S-A phase.

It was shown in (3) that, when comparing efficiency for branched and linear polymers in the functional layer, the highest values were obtained for the linear ones, both PEG and PEI. In (1), one can see that using 1,4-BDDGE as a linker resulted in lower efficiency for S-B-A as compared to S-A. The same linker introduced in stationary phase containing ERM provided significantly higher efficiency for S-B-ERM as compared to S-ERM (5). Therefore, the way of attaching ERM with spacing of functional groups resulted in high efficiency for the phases with macromolecules in the functional layer (up to 5 times higher 

 for S-B-ERM as compared to S-ERM). The reason could be better shielding of the silica substrate with 1,4-BDDGE because three linkers can be attached per each amino group on its surface.

Among all phases discussed here, the highest efficiency (35,000–45,000 N/m) for all the tested classes of analytes was demonstrated by S-A obtained with monomeric amide ligands via the Ugi reaction. The lower range of 15,000–40,000 

 was provided with phases modified with macromolecules, namely, linear polymers (both PEI and PEG) and eremomycin attached via the linker. The phases containing branched PEI and PVP demonstrated efficiencies below 20,000 

, probably due to conformational mobility of such functional layers and mass transfer limitations (3).

Through our examples, it could be concluded that using linear polymers for creating functional layers is advantageous for preparing high efficiency columns as compared to branched ones. Also, attaching ERM macromolecule with the branched linker 1,4-BDDGE allowed to obtain efficient phase S-B-ERM for vitamins separation. While creating amide functionality resulted in the highest efficiency for all the tested classes of compounds for S-A phase.

For sugars where hydrogen bonding has significant effect on the retention, polymers in functional layer (such as PEG) provided the highest retention factors (Figure 1). Despite this advantage, separation efficiency for sugars was lower in case of polymer coatings as compared to those modified with monomers. The best separation of sugars was achieved on amide phase S-A (Figure 6), which provided resolution of maltose and lactose having the same log P value (-5.03, calculated via Epiweb 4.1) with the highest efficiency (up to 35,000 

Polymer-coated phases also demonstrated high selectivity for amino acids. S-PEG phase provided the maximum 

’ values for amino acids, but required about 40 min for separation. S-B-PEI or S-A provides faster separation with higher efficiency and Tyr/Pro resolution (Figure 7). Zwitterionic S-B-G is also a good choice for amino acid analysis. Its unique feature is to separate Leu, Ile, and Val in the same run with other amino acids.

The best resolution of water-soluble vitamins was provided with phase S-A. Polymer-coated S-PEG demonstrated some limitations in selectivity for weakly retained B

, while S-B-PEI was inferior in both efficiency and selectivity. Introducing the linker into functional layer resulted in decreased resolution for weakly retained vitamins which required low water content in mobile phase for their separation. At the same time, introduction of the linker was advantageous for ERM-coated phases bringing higher efficiency and better separation of water-soluble vitamins and amino acids. As shown in (5), both ERM-coated phases also demonstrated chiral selectivity.

The best separations of polar analytes with various stationary phases are demonstrated in Figures 6–8 (1–5).

Tanaka HILIC test represented a good algorithm for general characterization of hydrophilic stationary phases. However, it does not always reflect structural differences in the functional layer for retention and separation of various classes of polar substances. By changing conditions and including analytes providing specific interactions with particular surfaces one can complement column characterization to simplify the prediction of retention and selectivity parameters and find the utility of a given phase in contrast with a standard test result.

From the set of described stationary phases, polymer-coated and amide columns demonstrated the best selectivity and efficiency. Specific hydrophilic interactions with amides of S-A phase provided advanced performance and improved separation of polar analytes as compared to polymer-coated adsorbents. Additional advantage of amide phase is the ease of structure variation resulting from flexibility of the Ugi reaction used for the synthesis (1).

The potential for efficiency improvement for polymer-coated phases could include further modification using hydrophilizing or crosslinking agents. Optimizing the structure of a zwitterion and the way of its attachment to the substrate could bring advantages in performance for zwitterionic adsorbents.

Further development of new phases for HILIC with improved selectivity should be focused on the creation of novel chemistries with diversity of functional groups providing multiple types of interactions. Also, flexible synthetic ways look promising for fine tuning of chromatographic performance of the resulting adsorbents.

(1) Chikurova, N.; Shemyakina, A.; Shpigun, O.; Chernobrovkina, A. Multicomponent Ugi Reaction as a Tool for Fast and Easy Preparation of Stationary Phases for Hydrophilic Interaction Liquid Chromatography, Part I: The Influence of Attachment and Spacing of the Functional Ligand Obtained via the Ugi Reaction. 

(2) Chikurova, N.; Shemyakina, A.; Bryskina, D.; Nuriev, V.; Komarov, A.; Statkus, et al. A Novel Adsorbent for Hydrophilic Chromatography Based on Silica Modified by the Ugi Reaction. 

(3) Popov, A.; Maksimov, G.; Shpigun, O.; Chernobrovkina, A. Adsorbents with a Covalently Bonded Polymer Layer for Hydrophilic Interaction Liquid Chromatography. 

(4) Popov, A.; Maksimov, G.; Smolenkov, A.; Shpigun, O.; Chernobrovkina, A. Novel Adsorbents for Hydrophilic Chromatography Based on Silica Gel Covalently Modified with Poly(ethylene glycol). 

(5) Chikurova, N.; Prosuntsova, D.; Stavrianidi, A.; Staroverov, S.; Ananyeva, I.; Smolenkov, A.; Chernobrovkina, A. Novel Mixed-Mode Adsorbents for HPLC Based on Different Substrates Modified with Eremomycin. 

(6) Chernobrovkina, A.; Sedov, Y.; Shpigun, O.; Zatirakha, A.; Smolenkov, A.; Kovalenko I. Influence of Zwitterion Nature on Chromatographic Performance of HILIC Stationary Phases. 

(7) Kawachi, Y.; Ikegami, T.; Takubo, H.; Ikegami, Y.; Miyamoto, M.; Tanaka, N. Chromatographic Characterization of Hydrophilic Interaction Liquid Chromatography Stationary Phases: Hydrophilicity, Charge Effects, Structural Selectivity, and Separation Efficiency. 

(8) Dolci, M. On the Characterization of Hydrophilic Interaction Luid Chromatography Stationary Phases. 

(9) Ikegami, T.; Taniguchi, A.; Okada, T.; Horiec, K.; Arasea, S.; Ikegami, Y. Functionalization Using Polymer or Silane? A Practical Test Method to Characterize Hydrophilic Interaction Chromatography Phases in Terms of their Functionalization Method. 

 is an associate professor in the Department of Chemistry at Lomonosov Moscow State University, in Moscow, Russia. Direct correspondence to: 

Events

This comparison of various silica-based HILIC stationary phases with different types of coatings provides a better understanding of how stationary phase selectivity can be influenced.


Chromatographic Comparison of HILIC Stationary Phases Prepared Using Various Hydrophilization Techniques

In this article, the ability of porous graphitic carbon (PGC) to discriminate between structurally similar compounds is described. After a brief overview of the particle technology and its use in liquid chromatography (LC), two case studies are presented using small-molecule probes to illustrate the importance on how molecular shape and the ability to interact with the graphitic surface lead to the resolution of structural isomers. Finally, an application of this technology in the area of glycan analysis is showcased where released N-glycans from human immunoglobulin G (IgG) were resolved without derivatization, enabling the glycan profile of a biotherapeutic to be determined.

The separation of highly polar compounds by high performance liquid chromatography (HPLC) is still a challenging task. These separations can be approached using ion pair chromatography (IPC), reversed-phase LC (RPLC) separations with a very low content of an organic modifier, or hydrophilic interaction chromatography (HILIC). IPC has the disadvantages of requiring an additional mobile-phase additive, and that the columns must be dedicated for IPC after the first run. Generally, RPLC with a very polar mobile phase is not considered to be very stable (column dewetting). With HILIC, users often lament the lack of robustness and reproducibility. In addition, all of the abovementioned modes utilize silica-based HPLC stationary phases, which are limited in pH range and temperature stability. These modes also have the potential for secondary interactions from active surface silanol groups.

Another promising approach for the retention and analysis of polar compounds by HPLC is using porous graphitic carbon (PGC) as the stationary phase. Carbon materials have been investigated since the 1970s, but only after improving the synthetic route did a carbon stationary phase with satisfying chromatographic and stability properties come to fruition (1,2). Recently, a new carbon particle was developed that has a smaller particle size (2.7 μm) with a narrower particle-size distribution and improved mechanical stability, leading to improved injection-to-injection reproducibility and enhanced efficiency in ultrahigh-pressure liquid chromatography (UHPLC) (3).

PGC phases are superior in some ways to silica-based stationary phases, thanks to PGC’s stability over the complete pH range (1–14) and under temperatures up to 250 °C. Compared with C18 phases, PGC has some similarities with C18, but for many analytes, increased retention is typically observed. Knox and Ross were the two main pioneers who developed the chromatographic characterization of PGC, and during its early development they attempted to explain its unexpected polar retention behavior (4).

It was originally believed that an entirely carbonaceous stationary phase would behave solely as a super-hydrophobic support, but soon it became clear that this was not the case. PGC showed a much higher affinity to retain polar analytes than the standard C18 support. Hence, the term 

polar retention effect on graphite (PREG)

 was coined to explain their observations.

 is defined as electronic interactions between polarizable groups of an analyte with the surface of graphite. PREG is hypothesized to stem from some form of charge induction or electron ion pair donating/accepting interactions with the pi cloud of graphite (4).

Ross attempted to quantify the contribution of PREG on PGC retention; however, the work was unsuccessful (4). Nevertheless, the term stuck around, and continued research into the retention mechanisms of PGC by other groups gave some substantiation of the claims within the PREG theory (5,6). Notably, the work by De Matteis and others confirmed, using quantitative structure retention relationships (QSRR) in conjunction with multivariate linear regression analysis, that retention was governed by a hydrophobic factor and two electrostatic descriptors related to polarizability and the lowest unoccupied molecular orbital energy (6). Certainly, a more predictive model is warranted for PGC; however, it is obvious that PGC offers alternative retention properties compared to C18 for polar compounds.

Generally, polar analytes are more retained on PGC than on C18, but how much more firmly depends on the size, structure, and shape of the analyte. Figure 1 demonstrates this concept: Small polar molecules are not well retained on either C18 and PGC, but with polar hydrophilic ring structures, PGC exhibits strong retention, whereas C18 exhibits no retention. The retention difference between uracil and dihydrouracil is quite striking considering their similarities, but the contrast in three-dimensional (3D) structure results in uracil being more strongly retained because of its planar structure.

The example shown in Figure 1 alludes to significant analyte molecular shape criteria that are necessary for polar compounds to be adequately retained on PGC. This shape-selective property is not just the case for polar analytes, but for all compounds in general. An instance of this is shown in Figure 2 where a set of hydrophobic analytes (benzene, naphthalene, diphenylmethane, and triptycene) are separated. For the C18 column, broadly speaking, the analytes are eluted in order of their hydrophobicity; however, the PGC column has no similarities in elution pattern to the C18. The reason for these retention changes is best explained by the different shapes of the test compounds and the ability for shape selective interactions to take place with the graphitic surface of the PGC stationary phase. This interaction becomes stronger when there are multiple aromatic systems and the analyte shape or size allows for multiple contact points. In the case of naphthalene, the double ring system results in strong pi–pi interactions with maximal molecular surface area contact with the graphitic surface. On the other hand, triptycene, a much larger and more hydrophobic molecule, is not able to have all of its surface area interact with the stationary phase surface because of its rigid structure.

Although the previous examples shows analytes with different amounts of aromatic rings, in this example, all compounds are isomers containing one aromatic ring. Figure 3 shows a separation of toluic acid isomers. For the ring-substituted isomers, the elution order is ortho, para, then meta. Because of the methyl group being ortho to the carboxylic acid group, this forces the carboxylic acid group to twist away from the plane of the benzene ring, resulting in a slightly bent structure. Similar to the previous examples, the ramifications here are a smaller molecular surface area that can overlap with the graphitic plane compared to the meta and para isomers, which are completely flat. Why the para isomer is eluted before meta is not as straightforward. Herein, we must consider other orientations besides face-to-face stacking. In the edge-to-face orientation (molecule is flipped on its edge), the carboxylic acid group slightly hinders the pi–pi interactions, whereas the meta isomer has an orientation that is free from the steric effects of the acid group. These findings are in-line with the molecular modeling studies by De Matteis and others, where the retention behavior of structural isomers was probed in finer detail, highlighting the importance of analyte geometry for pi–pi stacking interactions with PGC (7).

Although the examples shown indicate that PGC is not all-encompassing for polar analytes, PGC still can be a promising option for the analysis of polar compounds, if the target analytes have enough molecular size to interact strongly with the graphitic surface. This observation supports the initial findings on PGC for the analysis of carbohydrates, where PGC lacked retention for monosaccharides, but for disaccharides, oligosaccharides, and larger polysaccharides, acceptable retention times were observed (8). Figure 4 shows this point where a single unit sugar does not have enough size for strong interactions with a model graphite surface, but with multiple sugar subunit molecules, there are more contact points to take advantage of the shape selective properties of the graphitic surface.

Glycans are polysaccharides linked by glycosidic bonds. Glycans can be bonded to proteins or lipids. For therapeutic monoclonal antibodies (mAbs), the glycosylation pattern plays a pivotal role in efficacy and lifetime. The glycan profile of a biotherapeutic drug is a critical quality attribute (9).

Because glycans consist of different monosaccharides that can be ordered randomly or even branched, numerous glycans exist, and analysis of these polar compounds is challenging. HILIC with derivatization is the method of choice for glycan analysis. As an alternative, PGC can be successfully used to separate glycans released from human immunoglobulin G (IgG) under RPLC conditions without derivatization (Figure 5, Table I). In this case, two commercial PGC columns were compared for this analysis.

Both columns were able to separate the glycans; however, the performance of PGC B enabled easier identification and accurate mass quantitation. In comparison with the PGC A column, the sensitivity was increased by more than a factor of two, and the separation efficiency was improved, as indicated by the reduction of peak widths by a factor of two. Furthermore, not all glycans were eluted from the PGC A column, which might be the result of too strong of an interaction with the stationary phase. The reasons why PGC B performs remarkably better in this example is an area of further study.

The examples shown here do not prove or disprove the PREG theory, but they highlight that polar retention on a PGC stationary phase is primarily driven by analyte shape and how molecules can orient themselves with the graphitic surface. These factors should be considered early on in method development when evaluating whether PGC is an appropriate stationary phase for the types of compounds to be separated. Nevertheless, when the unique shape-selective properties of PGC are appropriately identified, PGC stationary phases can be harnessed for fine separations of structurally related compounds, which continue to be a challenge for many practicing chromatographers.

(1) Knox, J. H.; Unger, K. K.; Mueller, H. Prospects for Carbon as Packing Material in High-Performance Liquid Chromatography. 

(2) Knox, J. H.; Kaur, B.; Millward, G. R. Structure and Performance of Porous Graphitic Carbon in Liquid Chromatography. 

(3) Corman, C.; Muraco, C.; Ye, M.; Frantz, C.; Maule, W. Applications Employing New 2.7 μm Porous Graphitic Carbon Particles for U/HPLC. 

https://www.sigmaaldrich.com/deepweb/assets/sigmaaldrich/marketing/ global/documents/264/610/analytix-report-10-br7874en-ms.pdf

Carbon-Based Packing Materials for Liquid Chromatography: Structure, Performance, and Retention Mechanisms. In Advances in Chromatography

, vol. 37; Brown, P. R.; Grushka, E., Eds.; M. Dekker, 1997; pp 73–119.

(5) Lepont, C.; Gunatillaka, A. D.; Poole, C. F. Retention Characteristics of Porous Graphitic Carbon in Reversed-Phase Liquid Chromatography with Methanol–Water Mobile Phases. 

(6) De Matteis, C. I.; Simpson, D. A.; Euerby, M. R.; Shaw, P. N.; Barrett, D. A. Chromatographic Retention Behaviour of Monosubstituted Benzene Derivatives on Porous Graphitic Carbon and Octadecyl-Bonded Silica Studied Using Molecular Modelling and Quantitative Structure-Retention Relationships. 

(7) De Matteis, C. I.; Simpson, D. A.; Euerby, M. R.; Shaw, P. N.; Barrett, D. A. Chromatographic Retention Behaviour of n-Alkylbenzenes and Pentylbenzene Structural Isomers on Porous Graphitic Carbon and Octadecyl-Bonded Silica Studied Using Molecular Modelling and QSRR. 

(8) Koizumi, K. High-Performance Liquid Chromatographic Separation of Carbohydrates on Graphitized Carbon Columns. 

(9) Bertozzi, C. R.; Freeze, H. H.; Varki, A.; et al. 

Glycans in Biotechnology and the Pharmaceutical Industry

, 2nd ed.; Varki, A.; Cummings, R. D.; Esko, J.D., et al., Eds. Cold Spring Harbor Laboratory Press; 2009. Chapter 51.

(10) National Library of Medicine, Symbol Nomenclature for Glycans (SNFG), 

https://www.ncbi.nlm.nih.gov/glycans/snfg.html

 are with MilliporeSigma, in Bellefonte, Pennsylvania. 

 is with Merck KGaA, in Darmstadt, Germany. Direct correspondence to: 

When the unique shape-selective properties of PGC are appropriately identified, PGC stationary phases can be harnessed for fine separations of structurally related compounds.

Insights into the Shape-Selective Retention Properties of Porous Graphitic Carbon Stationary Phases and Applicability for Polar Compounds

To improve the selectivity of challenging separations, a possible solution is the modulation of the stationary phase through serial coupling of different columns having different chemistries. Such an approach is interesting when each column used for a tandem setup results in distinct critical peak pairs. For impurity profiling methods, gradient elution mode is preferred. However, in gradient mode, the order of the coupled columns may result in different selectivities. In addition, the length of the individual columns may also impact the separation quality. The aim of this work was to study the possibilities of serial column coupling for impurity profiling methods. Various C8, C18, cyano, and phenyl phases were tested and compared. An algorithm was used to investigate peak migration through tandem columns, considering both the length and order of the individual column segments. As an example, amlodipine impurity profiling was considered to illustrate the benefit of serial column coupling. We also draw attention to the fact that common column tests are not always helpful because they do not declare anything about the compounds of our interest. Oftentimes, it is more informative and less time-consuming to run one or a few linear gradient experiments by injecting the sample of interest.

The idea of coupling columns to separate complex mixtures appeared early in the study of chromatography (1–5). In general, the purpose of column coupling can be either to improve the kinetic performance of a separation by increasing the column length, or to adjust selectivity by combining different stationary-phase chemistries. In unidimensional chromatography, there are two ways to combine two or more columns, namely parallel and serial arrangements (6).

The serially coupled columns (SCC) approach is often referred to as stationary phase optimized selectivity liquid chromatography (SOSLC), or discontinuous stationary phase gradient (4,7,8). The serial column coupling approach has been commercialized under the name of phase optimized liquid chromatography (POPLC), which was provided by Bischoff Chromatography. Studies have reported improved selectivity for coupled columns compared to the use of a single column (9–12). Coupled systems have an intrinsic advantage—there is an additional separation factor (the column length) that has no consequence in the experimental effort. In practice, each serial combination of columns of different chemical natures and lengths operates as a new column with its own selectivity. This significantly increases the wealth of various columns available in a laboratory, from which the most appropriate ones can be selected for a given application in a tandem setup (6).

A few approaches have been proposed to optimize SCC systems (13,14). An important difference compared to mobile phase optimization is that stationary phase is a discrete factor and cannot be varied arbitrarily. Those works demonstrated that serial coupling of columns introduce new degrees of freedom, such as the type, number, relative length, and the order of the individual columns (4,15). However, the full benefit of SCCs is only taken through the interpretive optimization of both the columns’ nature and length, along with the elution conditions (1,2,4,16,17). The SCC approach is mostly applied in isocratic elution mode; however, linear gradient optimizations using different algorithms have already been reported (4,11,14,18–20).

The pharmaceutical industry is one of the main users of liquid chromatography (LC). Gradient elution must be used in most cases when examining the impurity profile or degradation pathways of active pharmaceutical ingredients (API). 

International Council for Harmonization (ICH) Q3A (R2) 

provides guidance on the content and qualification of impurities in new drug substances (21). Baseline separation of adjacent peaks is often required; thus, routine methods are expected to achieve a critical resolution value (

) of at least 1.5, which is sometimes challenging (22). Therefore, combining two or more columns possessing different selectivities might improve the chance to attain baseline separation for a complex sample.

As mentioned above, a coupled system of columns having different chemistries works as a new column. Therefore, one possibility for method development is to optimize a separation on one given coupled setup. However, in this situation, the contribution of the individual columns to the overall apparent retention will remain unknown. Instead, a better approach is to measure the retention parameters of the different individual column chemistries, and then the overall retention of a solute and expected selectivity can be predicted for any column combination (different lengths or orders).

The aim of this work was to study and compare the selectivities of state-of-the-art reversed-phase (RP) columns of various chemistries (C8, C18, cyano, and phenyl) and to find the most appropriate column combinations for challenging separations. A common column test (that is, Tanaka) was performed first to characterize the individual columns. Then, the most diverse phases were selected. Based on linear input gradient experiments, retention model parameters were derived for amlodipine impurity profiling, and then model calculations were done to find the most promising coupled systems. Finally, the proposed column combinations were experimentally tested, and the impact of the order of coupled columns was also studied.

For column characterization, the general test proposed by Tanaka and coworkers was performed. Based on this test, seven selected substances were used to characterize hydrophobicity, hydrophobic selectivity, steric selectivity, silanol activity, and ion-exchange capacities (at low and high pH) of the stationary phases (23). The test compounds were uracil, pentylbenzene, butylbenzene, o-terphenyl, triphenylene, caffeine, phenol, and benzylamine (0.1 mg/mL).

As a real-life separation example, the tested substance was a spiked sample of amlodipine. The sample contained 1 mg/mL amlodipine and all of its impurities at a level of 1–2% (10–20 μg/mL).

Various column chemistries were tested, namely Acquity HSS fully porous (FP) 2.5 μm C18, C18 SB, T3, fluorophenyl, cyano, and Kinetex superficially porous (SP) 2.6 μm C8, C18, phenyl-hexyl, biphenyl, and F5 phases. Column diameter was 2.1 mm for all columns, and the length was either 3 or 5 cm.

The instrument used in this study was a UHPLC system with a binary delivery pump. The gradient delay volume (VD) was measured as 0.1 mL and the extra column volume (Vec) was ~11 μL. The system contained 500 nL photo-diode-array (PDA) flow cell. The volume of the column connector unit was 0.8 μL.

Chromatographic Conditions and Column Tests

For the Tanaka test, the mobile phase compositions were adopted from the protocol (23), the column temperature was 25 °C, the flow rate was set to 0.4 mL/min, and the injected volume was 0.1 μL.

For the application example (amlodipine impurity profiling), the mobile phase A contained 0.1% v/v perchloric acid solution, and mobile phase B was acetonitrile. With such acidic mobile phases, the residual silanol groups of silica-based columns were expected to be in ion-suppressed form. Therefore, no electrostatic interaction is expected with basic solutes. The flow rate was 0.4 mL/min, and the detection wavelength was 237 nm. The injection volume was 0.5 μL.

Measuring Model Parameters for Amlodipine and Related Impurities

Based on the results of the Tanaka tests, four columns were selected for the application example (FP 2.5 μm C18, cyano and SP 2.6 μm C18, biphenyl). Various linear gradients (with different gradient steepness) were performed by injecting the spiked amlodipine sample. Then, a procedure based on fitting was applied to obtain the model parameters (24–26).

Studying Solute Migration Through Coupled Columns

During the work, a linear solvent gradient was assumed. The amount of organic solvent in the mobile phase (

) can be described by the following relation:

is the initial organic solvent content of the eluent; 

is the time spent adsorbed on the surface of the stationary phase.

) of a band in the chromatographic column as a function of the variable 

can be expressed by the following expression:

) is the retention factor of a solute as a function of 

A simple finite difference algorithm was used to solve equation [2] and to determine the distance traveled by a solute through a coupled system. Very briefly, for a serial system, one can divide the time elapsed during the development of a mobile phase gradient into very short Δ

time intervals. Then, the mobile phase volume fraction at 

describes the change of mobile phase volume fraction in a Δ

time interval. Assuming the linear solvent strength (LSS) model, the retention factor (

th time interval—in general—can be expressed as:

are the model parameters. Then, the migration velocity of solute 

One can consider a gradient as a sequence of small isocratic steps. This allows the determination of the intermediate migrated distance through a column at each time interval during the analysis. If a solute band passes the end of a column, then the algorithm takes the model parameters of the next column, and so on. Then, the actual distance traveled, 

For the first time interval, the distance is simply calculated as:

Because the solute molecules travel only in the mobile phase, the total time spent in the mobile phase corresponding to the 

Accordingly, the total time elapsed during the travel of a band to position 

Based on these calculations, migration plots (distance traveled by the solute vs time elapsed) were constructed to study the elution behavior of various solutes on different coupled systems.

In isocratic elution mode, the order of coupled columns does not play a role in the overall selectivity. However, when running gradients on serially coupled systems, one might have the impression that the less retaining column should be placed first and followed by the more retaining column to benefit the most from the selectivity of both columns. Otherwise (that is, the more retentive column is followed by the less retentive one), the less retentive second column will experience a stronger mobile phase. As a result, the residence time of the solutes will be much shorter on the second column. In extreme cases, it may happen that the compounds will not be retained on the second column at all.

Despite the intuition and general concept that for gradient separations, the order of the columns should follow the order of their retentivity. Here, we illustrated that such column order does not necessarily result in the highest selectivity and resolution for a complex mixture.

General Column Tests to Compare Different Chemistries

Column tests based on a specific set of known compounds are often performed in the pharmaceutical industry. The test compounds are ideally selected to quantify one specific interaction occurring between the solute and the stationary phase. Those tests can be a good starting point of a column screening procedure. Please note that several column databases are also publicly available (27,28).

In this work, we applied the Tanaka test, which is established as the most popular standard for evaluating the selectivity of different HPLC columns (29). The parameters measured for the 10 columns were illustrated as hexagons with six axes corresponding to different physico-chemical properties. Figure 1 shows the obtained selectivity plots, and Table I summarizes the results of the Tanaka test. The more diverse the shape of the two hexagons, the more different the selectivity. FP and SP column families were examined, corresponding to significant differences between their morphology. For the serial column setup, our goal was to select two diverse FP and two diverse SP phases. Among the FP columns, the C18 and cyano phases showed the highest difference (based on similarity factor, 

). The cyano phase showed very different hydrophobicity (

 (pH = 7.6)) compared to the C18 phase. Among the SP phases, the C18 and biphenyl ones were selected as the most distinct columns considering the 

 values and the specific conditions applied for the application (amlodipine impurity profiling).

In silico Preliminary Work to Study Serially Coupled Columns (Amlodipine Impurity Profiling)

Scouting gradients were performed to find suitable initial and final gradient composition to elute all amlodipine related compounds in a reasonable elution window from all columns. Retention model parameters were determined for each compound on all of the four selected columns (FP C18, FP cyano, SP C18, and SP biphenyl). The apparent retention and migration plots were then calculated for various column combinations and gradient programs.

Among all solutes, compound “Imp-G” showed the most significant retention changes at different coupled setups. And indeed, Imp-G was always one member of the critical peak pairs, regardless of the column combination. Figure 2 shows some representative examples of Imp-G peak migration for different column configurations. Two column orders were assumed (A to B and B to A). Figure 2a shows that the FP C18 phase is much more retentive than a FP cyano one. Despite this huge difference, when combining identical column lengths, it seemed that column order did not change the overall retention (the blue and red curves just intersect at the sum of individual column lengths—10 cm). On the other hand, the SP C18 and biphenyl phases were more similar in terms of retentivity, but when combining 3 and 5 cm segments, the overall retention was slightly different depending on the order of coupled columns (Figures 2b and 2c). A 1–2% difference was observed in the overall retention time based on the column order. (In Figure 2b, the blue and red curves do not intersect, whereas in Figure 2c, the two curves intersect before reaching the end of the second column [7.6 cm compared to 8 cm]).

Figure 3 illustrates the real benefit of column coupling. When combining 5 cm and 3-cm long columns (made of SP C18, SP biphenyl, or both), the retention of Imp-G can be fine-tuned as desired. The 8-cm SP biphenyl setup resulted in the lowest retention (

= 3.14 min), followed by the combination of SP biphenyl 5 cm and SP C18 3 cm (

= 3.21 min). Because the SP C18 phase is more retentive than the SP biphenyl, the highest retention was obtained with a 8-cm SP C18 setup (

= 3.42 min), whereas a combination of SP C18 5 cm and SP biphenyl 3 cm yielded slightly decreased retention (

In the amlodipine mixture that used SP columns on SP C18 phase, the critical peak pair was ImpG-Imp-H, whereas the ImpB-ImpG peak pair were resolved well. On the contrary, for the SP biphenyl phase, ImpB-ImpG was the least resolved peak pair while ImpG-ImpH were baseline separated. One might have the impression that a good solution is probably to combine SP C18 and biphenyl phases to attain suitable resolution between both ImpB-ImpG and ImpG-ImpH pairs. Model calculations were performed to study the possibilities of combining SP C18 and SP biphenyl phases for such peak pairs. As shown in Figure 4, as the length fraction of the SP biphenyl phase increases (compared to SP C18) the critical pair is shifted from ImpG-ImpH towards ImpB-ImpG. Our model calculations suggested that both a 5 cm SP C18 and 3 cm SP biphenyl and a 5 cm SP biphenyl and 3 cm SP C18 combinations resulted in baseline resolution (

Experimental Verification of Model Calculation: Applications of Serially Coupled Columns

To validate our prediction, the same conditions were set as stated in the caption of Figure 4. The experimentally observed chromatograms are shown in Figure 5 for all four column combinations. As suggested by our model, the resolutions between ImpB, ImpG, and ImpH were the most sensitive for the changes of the stationary phase chemistry. The experimental results were in excellent agreement with the predictions—as the length fraction of the biphenyl phase increased (from Figure 5a–d), the critical pair shifted from ImpG-ImpH toward ImpB-ImpG. The average error of retention time prediction was lower than 1.5%.

It also worth noting that the retention of ImpB, ImpG, ImpH, and ImpA was higher on the SP C18 phases compared to SP biphenyl. However, peaks ImpD and ImpF were more retained on the SP biphenyl phase. Therefore, for complex real-life samples, it is not always obvious to see which stationary phase is the more retentive one. Consequently, the common practice of placing the columns in the order of their increasing retentivity was not applicable in this case: both a SP C18 and SP biphenyl and a SP biphenyl and SP C18 combinations resulted in baseline separations.

Serial coupling of different column types can provide several important benefits to the chromatographer, including fine-tuning selectivity to separate complex mixtures.

Advances in LC Column Technology | LCGC Supplements

Chromatographic Comparison of HILIC Stationary Phases Prepared Using Various Hydrophilization Techniques

Insights into the Shape-Selective Retention Properties of Porous Graphitic Carbon Stationary Phases and Applicability for Polar Compounds

Combining Different Stationary-Phase Chemistries to Improve the Selectivity of Impurity Profiling Methods

Pressure–Enhanced Liquid Chromatography as a Suitable Approach to Improve Selectivity for Large Molecule Separations