Liquid Chromatography Evaluation of a Novel Graphitic Carbon Stationary Phase Using Glucagon-Like Peptide‑1 Analogs

Abstract

Innovations in liquid chromatography stationary phase media can address emerging peptide separation and purification challenges. Herein, we report the synthesis of novel carbon microbeads from natural micrographite flakes as the starting material (referred to as All Carbon microbeads) and their performance as reversed-phase liquid chromatography stationary phase media using the glucagon-like peptide-1 (GLP-1) analogs semaglutide and liraglutide as probe analytes. High-performance liquid chromatography (HPLC) performance metrics were characterized and validated, including column efficiency measured by theoretical plate count (N), as well as linear response, precision, limit of detection (LOD), limit of quantitation (LOQ), and loading capacity. Commercially available silica C18 media, the current industry standard, were used as controls for comparison. The results indicated that HPLC columns packed with All Carbon microbeads consistently separate the GLP-1 analogs with retention times similar to those of the reference silica C18 columns. Their performance metrics are comparable to those of silica C18 in terms of plate count (N), LOD, LOQ, and loading capacity. Additionally, they perform better in the precision of retention time and linear response compared with silica C18. The chromatographic performance of All Carbon microbeads was stable in the presence of ion-pairing agents, extreme pHs (pH 1 and 13), salt concentrations, under field conditions (liraglutide crude), and with 100% aqueous loading conditions. The results present opportunities for further development as a sustainable reversed-phase media for GLP-1 analog analytical characterization and manufacturing.

Introduction

Developments in stationary phase media are advancing separation science to meet the demands of the pharmaceutical industry. New materials with unique separation and purification characteristics aim to facilitate discovery R&D with small molecule and biologic pharmaceuticals, increase production capacity, and lower operational costs. One of the key pharmaceutical market segments driving innovation is peptide therapeutics. − The primary technique for purifying peptides is reversed-phase high-performance liquid chromatography (RP-HPLC). Recently, the increased demand for glucagon-like peptide-1 (GLP-1) analog therapeutics for type 2 diabetes and weight management has burdened peptide manufacturing, including downstream purifying processes. , High-volume production batches translate to more chromatographic runs and a high demand on the stationary phase. Higher-performance chromatographic resins are required to handle the workload of pharmaceutical manufacturers to keep production costs in check and advance the next line of peptide therapeutics.

Silica C18, also referred to as octadecylsilane (ODS), is the most widely used stationary phase media for RP-HPLC. − Silica C18 is the recommended media to use when beginning a new reversed-phase method. However, this material has well-documented issues. It is intrinsically difficult to eliminate or shield polar residual silanol groups present in silica, which contribute to unwanted chromatographic behavior. Silica is stable across a limited pH range. − Retention loss due to pore dewetting (formerly attributed to phase collapse) occurs when attempting to run a 100% aqueous mobile phase on silica C18 columns. Synthetic graphite, inorganic-based, and polymer-based stationary phase media have been explored as alternatives to address these issues.

Recently, we reported a microfluidic-based platform to synthesize porous carbon microspheres using carbon materials with sp² carbon networks such as graphite, carbon nanotubes, graphene, or fullerenes as starting materials. A combinatorial approach utilizing carbon materials, cross-linkers (binders), additives, and processing parameters allowed for custom compositions (tunable hydrophobicity), structures, and functionalities. The carbon materials could be formed into spheres ranging from a few micrometers to several hundreds of micrometers in diameter with a relatively consistent size distribution (less than 25% relative standard deviation). The pore structure and size, varying from tens to hundreds of angstroms, could be adjusted by adding porogen diluents during synthesis. The carbon microbeads synthesized through this process exhibited mechanical stability, withstanding dynamic fluid flow pressures of up to 9000 psi. The results, taken together, indicated the potential of carbon microbeads as reversed-phase media for high-performance liquid chromatography.

In this article, we have synthesized carbon spheres using natural micro graphite flakes as the starting material (hereafter referred to as All Carbon microbeads to differentiate them from other commercially available porous graphitic carbon media) and evaluated their performance as reversed-phase liquid chromatography stationary phase media. We characterized and validated key chromatography performance metrics of this new media as they relate to assaying an active pharmaceutical ingredient (API) in drug formulations using the glucagon-like peptide-1 (GLP-1) analogs semaglutide (the API in Ozempic) and liraglutide (the API in Victoza) as probes. Analytical columns packed with this experimental media were assessed for column efficiency, described by theoretical plate count (N), linear response, precision, limit of detection (LOD), limit of quantitation (LOQ), and loading capacity. Additionally, preliminary durability studies investigating the effects of ion-pairing agents, extreme pH levels (pH 1 and 13), and salt concentrations under field conditions (liraglutide crude) and with 100% aqueous loading were conducted.

Experimental Section

Chemicals

GLP-1 analog probes, semaglutide, CAS No. 910463–68–2, Lot No. B23S04211, and liraglutide, CAS No. 204656–20–2, Lot No. BVS18U08105, were sourced from BOC Sciences, Shirley, NY, USA. Crude liraglutide – Art No. 4161028, Lot No. 1000099343, was procured from Bachem, Switzerland. Ammonium formate (Sigma-Aldrich 156264), ammonia solution (Sigma-Aldrich 499145), and ultrapure water (sourced from a Milli-Q Direct-Q 3UV system) were used in mobile phase A. Acetonitrile (Concord Technology – HPLC grade) was used as mobile phase B.

Stationary Phase Material

All Carbon microbeads (three grades) were synthesized and characterized as described elsewhere. Silica C18, 10 μm diameter (HxSil), was sourced from the Hamilton Company, Reno, Nevada, USA. Silica C18, 5 μm (SP-100–5-ODS-P), was sourced from the Osaka Soda Co., Ltd., Osaka, Japan. Before column packing, the materials were characterized for particle diameter and overall particle size distribution by the Coulter Principle using a Multisizer 4e (Beckman Coulter). The results of this size analysis informed the subsequent description of the materials as 4 μm silica C18 (control), 9 μm silica C18 (control), 6 μm All Carbon microbeads (experimental), 10 μm All Carbon microbeads (experimental), and 7 μm All Carbon microbeads (experimental). The pore structure and specific surface area were characterized by water intrusion porosimetry and nitrogen adsorption isotherms by using B.E.T. theory. Auapore Water Intrusion Porosimeter: Aquapore-5k-A-1 (Porous Materials Inc., Ithica, NY) and BET Sorptometer: BET 201A-N-SA (Porous Materials Inc., Ithaca, NY).

Column Packing

Equipment

A Teledyne SSI CP-class column packing unit was used to pack the columns. All test columns used in this study were stainless steel with dimensions of 150 mm L × 4.6 mm ID, sourced from Zodiac Life Sciences, India.

Packing Protocol

Each test column was prepared under identical conditions. The packing slurry was formed by adding 3 g of stationary-phase media to a 50 mL centrifuge tube. Next, 100% IPA was added to increase the total volume to 20 mL. The slurry was thoroughly vortexed and sonicated to ensure a homogeneous mixture. 100% IPA was used as the push solvent. The packing pump was set to flow at 24 mL/min but was autoregulated to maintain a constant pressure of 5000 psi. Packing proceeded until 150 mL of push solvent flowed through the column, at which point the pumps were turned off and the system was allowed to return to 0 psi. A top frit and cap were applied to finalize each column.

HPLC Instrumentation

All measurements were carried out on an Agilent 1100 HPLC system equipped with a quaternary pump, column heating compartment, degassing unit, and a variable wavelength detector model G1314A. Control and analysis were conducted with Open Lab CDS ChemStation software application v. c.01.10[201].

LC Method

For all tests, a standardized method (optimized for plate count N on the experimental media – see Supporting Information Section 1) was applied to each column in this study.

Standardized Method: Mobile Phase A – 20 mM HCOONH₄ (ammonium formate) adjusted to pH 8.5 with NH₃ (ammonia solution). Mobile Phase B – Acetonitrile. The mobile phases were pumped at 1 mL/min using a gradient of 20%B to 65%B over 10 min. The column temperature was held fixed at 25 °C. UV detection was conducted at 220 nm (see Supporting Information Section 3 for the rationale on using 220 nm instead of 280 nm, the more commonly used UV absorbance wavelength for monitoring proteins/peptides). Injection volume was the only parameter that varied. In the column loading study, injection volume ranged from 1 to 100 μL.

Results and Discussion

This work aims to characterize and validate key HPLC performance metrics of All Carbon microbeads (a novel porous graphitic carbon media) related to assaying active pharmaceutical ingredients (APIs). We used the GLP-1 analog therapeutics semaglutide (the API in Ozempic and Wegovy) and liraglutide (the API in Victoza and Saxenda) as probes. They were chosen since they readily absorb light in the UV spectrum for easy detection and quantification, are separated by reversed-phase liquid chromatography, and possess ionizable moieties to take both charged and uncharged forms depending on the pH of the solution. Further, our microbead characterization indicated the porosity and pore size most suitable for biomacromolecules. Additionally, these macromolecular probes allow for evaluation of the media under alkaline conditions, wherein the pH stability of silica-based media has limitations. Finally, their importance and demand for the treatment of type 2 diabetes and promoting weight loss have spurred intense research and interest in novel media for the analytical characterization of purity, potency, and assay and process purification to improve yields, purity, and throughput by liquid chromatography. The distinct differences in the composition of these GLP-1 analogs allow for the examination of the variation in chromatography methods and performance metrics for this class of peptides. Analytical HPLC columns packed with All Carbon microbead media were assessed for column efficiency, described by theoretical plate count (N), linear response, precision, limit of detection (LOD), limit of quantitation (LOQ), and loading capacity. Each test was benchmarked against commercially available silica C18 to provide a comparative analysis. Only silica C18 media was applied as a control since it is more widely used than other media (graphitic carbon, polymeric, or inorganic) and is considered the gold standard for separating GLP-1 analogs. This work was not designed to qualify the API probes or validate a method or lab operation. The tests were designed to evaluate how similar or dissimilar experimental columns behave compared with reference silica C18 columns under identical test conditions.

The All Carbon microbead microdroplet synthesis process is described in detail elsewhere and summarized in the Introduction. The All Carbon microbead’s structure is a network of sp² (natural graphite) and intergraphite sp³ (covalent) carbon bonds (via cross-linker). For this work, a blend of styrene and divinyl monomers was used as cross-linkers. However, other cross-linkers such as diacrylates (e.g., 1,4-butanediol diacrylate) or dienes (e.g., 1,7-octadiene) could also be employed. The bridging group (between the two carbon–carbon double bonds) of the cross-linker can be either hydrophobic (e.g., alkyl chains), hydrophilic (e.g., polyethylene glycol), or functional groups (e.g., epoxides), depending on the functionality (e.g., hydrophobic, amphiphilic, affinity) requirements of the microbead. This synthesis process significantly differs from that used to prepare commercially available synthetic porous graphitic carbon (PGC) media. Other PGC media have an sp² graphite layer, but unlike natural graphite, each layer is connected by sp³ carbon bonds. These PGC media are manufactured by infusing carbon-rich molecules or polymers into sacrificial inorganic templates, removing the template, and then treating the remaining material with high temperatures to achieve graphitization in the final product.

Figure A shows a representative bright-field optical image of 6 μM All Carbon microbeads used in the present studies. Figure B presents the size analysis of a representative sample of the All Carbon microbeads. Figure C shows a representative pore size distribution histogram of the All Carbon microbeads. Table summarizes key technical specifications for each media used in this study.

1.

Physical Characteristics of the All Carbon microbeads. (A) Bright-field microscopy of All Carbon microbeads shows a spherical conformation with uniform size distribution against a 5 μm scale bar. (B) Particle diameter analysis by Coulter Counter. (C) Particle pore diameter distribution by water intrusion porosimetry.

1. Physical Characteristics of Media Packed in Test Columns .

Column No.	Packing Media	Particle Size Diameter, RSD (μm, %)	Pore Size (Å)	SSA (m2/g)
C1	4 μm Silica C18 (control)	4.23, 45 •	94 ‡	459 ‡
C2	9 μm Silica C18 (control)	9.23, 29 •	70 ≫	788 ≫
E1	6 μm All Carbon microbeads (experimental)	6.19, 22 •	255 ≫	419 ≫
E2	6 μm All Carbon microbeads (experimental)	6.19, 22 •	255 ≫	419 ≫
E3	6 μm All Carbon microbeads (experimental)	6.19, 22 •	255 ≫	419 ≫
E4	6 μm All Carbon microbeads (experimental)	6.19, 22 •	255 ≫	419 ≫
E5	10 μm All Carbon microbeads (experimental)	9.68, 17 •	163 ≫	389 ≫
E6	7 μm All Carbon microbeads (experimental)	7.12, 55 •	284 ≫	177 ≫

^aMethod of measurement: (•) Coulter Counter-Multisizer 4e, (‡) B.E.T., (≫) Water Intrusion Porosimetry.

The experimental All Carbon microbeads are produced with size distributions in line with current analytical-grade and preparative-grade HPLC media (6 μm diameter, 22% RSD, and 10 μm diameter, 17% RSD, respectively). This media’s pore structure (300 Å pore diameter) accommodates large molecule separations, making it a good candidate for separating peptide analytes.

The LC Method

The LC method was optimized to arrive at the standardized protocol presented in the Experimental Section. Multivariate analysis by design of experiments (DOE) was carried out to assess the concurrent influences of changing factors on the defined response variable N (plate count). Optimization of mobile phase conditions to maximize the response variable was performed using a 2⁴-level full factorial experimental design (see Supporting Information Section 1). Four factors were studied in this experiment, each at two effect levels (-, +). The four factors were: 1) Organic Solvent, 2) pH, 3) Buffer, and 4) Gradient. The total number of experiments was 16 to enumerate every possible combination of the 4 factors at each effect level. Each experiment was replicated 5 times. The order in which each experiment was executed was randomized.

Parameter values tested in this method were determined through a combination of trial experience in our lab, as well as from ref . The results and discussion are presented in Supporting Information Sections 4 and 5.

Flow rate analysis was performed with a 150 mm L × 4.6 mm ID column packed with the All Carbon microbead (experimental) media using semaglutide as the probe. Figure S4 (Supporting Information Section 6) shows the data plotted as plate height (μm) vs flow rate (mL/min) fitted with the van Deemter model of chromatographic efficiency with parameters A = 0.1, B = 4.5, C = 3.4. The observed difference in efficiency N using mobile phase flow rates of 0.8 and 1.0 mL/min is −10%. Considering that recommendations for method development with a 150 mm length × 4.6 mm ID (internal diameter) column suggest starting with a flow rate of 1 mL/min and considering the need to balance efficiency with the speed of analysis, which implies faster flow rates, this study performed analysis at a flow rate of 1 mL/min.

Efficiency

Column efficiency, described by the number of theoretical plates (N), primarily depends on the chromatographic system’s kinetic factors, such as molecular diffusion, mass-flow dynamics, the properties of the column packing bed, and the flow rate. Smaller particles packed more uniformly with uniform size distributions produce higher efficiencies. This performance metric, calculated as a function of retention time and bandwidth (eq ), is principally determined by the column and, effectively, the physicochemical properties of the packing material. Given this relationship, efficiency is a suitable metric to use as the basis of comparison between different column packing materials.

$$N=16{\left(\frac{t_{\mathrm{R}}}{w}\right)}^2$$ 1

where t _R is the analyte retention time and w is the peak width measured in time units.

Over a series of n = 5 unique trial injections of the GLP-1 analog probes, using the standardized method described in the Experimental Section, the reference 4 μm silica C18 (control) column (C1) produced a plate count N of 27001 ± 1498, with a retention time t _R of 6.60 ± 0.005 (min) for semaglutide and N of 44311 ± 4323, with t _R of 8.66 ± 0.005 (min) for liraglutide. Similarly, the reference 9 μm silica C18 (control) column (C2) produced N of 6011 ± 487, with t _R of 5.89 ± 0.008 (min) for semaglutide and N of 9507 ± 407 with t _R of 7.60 ± 0.007 (min) for liraglutide (Figure A). Both of these control columns performed within or above the accepted performance limits for analytical columns based on their length and particle diameter. As a reference, the plate number N for a well-packed HPLC column under optimized test conditions is listed as 10,000–12,000 for a column of length 150 mm packed with 5 μm particles. A plate count of 6,000–7,000 is listed for columns of length 150 mm packed with 10 μm particles. Therefore, using plate number N as a response variable to compare the different stationary phase materials, we assume that the system conditions applied are suitably configured across the experimental design, including 1. column packing method, 2. mobile phase, 3. flow rate, 4. analyte concentration, and 5. injection volume.

2.

(A) A comparison of retention time and plate count for GLP-1 analog probes across columns packed with experimental All Carbon microbeads and control silica C18 stationary phases. (B) Plate Count vs Particle Diameter. Trend lines defined by reference silica C18 columns packed with 4 and 9 μm particles. Experimental column response with 6 and 9.7 μm particles falls above or within 1 standard deviation of silica C18 trendlines. This graph illustrates comparable performance for the tested GLP-1 analogs between the experimental and reference silica C18 columns. All columns tested have dimensions of 150 mm L × 4.6 mm ID. The terms col C1, col C2, col E1, and col E6 identify the specific columns used to produce the corresponding chromatogram. Unique column characteristics are listed in Table . Mobile Phase A – ammonium formate (pH 8.5), Mobile Phase B – ACN, Gradient 20%B – 65%B over 10 min, Temp: 25 °C, Flow rate: 1 mL/min, UV: 220 nm.

Six columns packed with All Carbon microbead experimental media were evaluated in three different grades of packing material. Column Nos. E1, E2, E3, and E4 were packed with the experimental media in the 6 μm, 22%RSD size grade. Column No. E5 was packed with the 10 μm, 17% RSD size grade, and Column No. E6 was packed with the 7 μm, 55% RSD size grade (Table ).

The combined average response of the experimental columns packed with the highest-grade material (E1, E2, and E3) showed a plate number N of 21129 ± 356 with t _R of 5.79 ± 0.023 (min) for semaglutide and N of 29718 ± 919 with t _R of 7.43 ± 0.027 (min) for liraglutide. The experimental column packed with the 10 μm media produced N of 6199 ± 160 with t _R of 5.49 ± 0.004 (min) and N of 8317 ± 84 with t _R of 7.07 ± 0.008 (min) for semaglutide and liraglutide, respectively. To place this efficiency in context and evaluate the response of these columns, a plot of plate number (N) vs particle diameter (μm) was generated (Figure B). The responses of the reference silica C18 columns were used to define trendlines for column efficiency as a function of particle diameter for each probe analyte. The experimental columns performed above, or within 1 standard deviation of, the defined trendlines. This information indicates that the experimental media’s performance characteristics align with the commercially available silica C18 for separating peptide analytes.

The selectivity of the All Carbon microbead columns regarding the analyte retention time is quite similar to that of silica C18. The chemical characteristics of a stationary phase, namely, bonded phase type (alkyl, phenyl, diol, etc.), length, and density, primarily determine the selectivity. Figure A illustrates the close alignment of the retention times for the GLP-1 analog probes across the experimental All Carbon and reference silica C18 columns. This alignment in retention times indicates that well-established methods for silica C18 columns can be quickly adapted to All Carbon microbead columns for these peptides. Studies show that other porous graphitic carbon media exhibit significantly different selectivity to silica C18. Taking our results together with those of these other studies suggests differences in selectivity behavior between the All Carbon microbeads and other porous graphitic carbon media.

We also investigated experimental media with a broader size distribution (7 μm diameter, 55% RSD) and injected the same probes under identical experimental conditions. This alternate column (E6) produced similar retention times but lower plate numbers for the probes than any of the other experimental columns (6 μm diameter, 22%RSD columns E1, E2, E3, and 10 μm diameter, 17% RSD column E5). The efficiency response for alternate column E6 was N of 3217 ± 89 with t _R of 5.61 ± 0.005 (min) and N of 3920 ± 64 with t _R of 7.20 ± 0.004 (min) for semaglutide and liraglutide, respectively. The primary contributing factor to this decreased efficiency could be the marked difference in the particle size distribution (%RSD) of this column compared with the others tested. The reduction in plate count is a function of band broadening resulting from eddy diffusion. As the mobile phase solvent moves through different microscopic flow streams between particles in the column, the velocity of those streams varies with the pathway’s width. Within snapshots of time, analytes traveling at different velocities migrate to different positions down the column, contributing to a broader distribution band. Larger differences in size between particle diameters produce more variation in the flow paths between those particles and thus wider analyte bands and lower chromatographic efficiency. Although the size distribution of the All Carbon microbeads used in this study is tighter than the commercially produced silica C18 media chosen here as controls, further optimization in the manufacturing process for the particle diameter will produce better chromatographic performance.

Linear Response

The linear response of an LC system refers to the concentration range over which the detector signal is directly proportional to the analyte concentration in a sample. In other words, the linearity of a method is a measure of how the plot of response vs concentration approximates a straight line and fits the equation:

$$y=mx+b$$ 2

where y is the response, x is the concentration, m is the slope, and b is the y-intercept. For the functional relationship between analyte concentration and output response, it is ideal for the relationship to be linear with b ≈ 0. This makes calculations easier and improves the precision. UV detector responses for dilute samples are expected to follow Beer’s law and be linear with b ≈ 0. For a chromatographic peak, a linear calibration proves that the system is performing correctly throughout the concentration range of interest.

A standard set of linearity measurements was conducted for the GLP-1 analog semaglutide using evaluation columns packed with the All Carbon microbeads (experimental) as well as with both grades of silica C18 (control). The semaglutide API is available clinically in both injectable and oral dosage forms. A 2024 review found subcutaneous injection doses ranging from 0.25 to 2.4 mg, with oral doses ranging from 2 to 40 mg. Novo Nordisk offers semaglutide in a once-weekly subcutaneous injection as Ozempic in concentrations of 0.68 mg/mL, 1.34 mg/mL, and 2.68 mg/mL. In this linearity study, we set the target concentration of the API at 3 mg/mL and tested concentrations from 50% to 130% of the target concentration. Reference standards at 1.5, 2.1, 2.55, 3, 3.45, and 3.9 mg/mL were injected into each column in triplicate for evaluation. Each injection’s peak area was recorded, and a response factor (peak area/concentration) was calculated.

Table shows the resulting peak areas and response factors at each concentration for all columns evaluated. Each column’s first 1.5 mg/mL observation was omitted as an equilibration run. It should be noted that the peak area normalized by concentration (response factor, RF) should remain relatively constant across all injections if the system is performing with a linear response, and RSD% (RF) should be close to zero to meet the acceptance criteria. The linearity results were strikingly similar when compared across all three columns. The injection series on the experimental 6 μm All Carbon microbead column produced an average RF = 3414.7 with RSD% (RF) = 2.3. The control 9 μm silica C18 column showed an average RF of 3658.3 with RSD% (RF) = 2.3, and the control 4 μm silica C18 column showed an average RF of 3537.0 with RSD% (RF) = 2.6. Figure A depicts peak area vs concentration, modeled with least-squares regression lines. For each graph, the equation of this best-fit line and the coefficient of determination R ² (describing a measure of how well the model predicts area from concentration) is listed. The experimental 6 μm All Carbon microbead column showed a very strong linear response over the 1.5 – 3.9 mg/mL concentration range with R ² = 0.9941 for y = 3408.1x + 25.005, p = 0.0000129. The linear fit using the control 9 μm silica C18 column was R ² = 0.9943 for y = 3638.5x + 60.565, p = 0.0000120, and for the control 4 μm silica C18 column, the resulting R ² = 0.9945 for y = 3474.8x + 170.43, p = 0.0000114.

2. Linearity Results for Column Uniformity: (A) 6 μm All Carbon Microbead (Experimental). (B) 9 μm Silica C18 (Control).(C) 4 μm Silica C18 (Control).

6 μm All Carbon microbeads (experimental)
Conc. (mg/mL)	Peak Area	Average	Response Factor (RF)
1.5	5244		3495.9
1.5	5192	5218.0	3461.4
2.1	6948		3308.5
2.1	7106		3383.9 Average (RF): 3414.7
2.1	6968	7007.3	3318.1
2.55	8528		3344.4 SD(RF): 78.9
2.55	8705		3413.7
2.55	8462	8565.0	3318.3 RSD% (RF): 2.3
3	10409		3469.5
3	10420		3473.3 Slope (RF): −4.4
3	10542	10456.8	3514.0
3.45	12131		3516.1
3.45	12023		3484.8
3.45	12109	12087.4	3509.8
3.9	13006		3334.8
3.9	13071		3351.4
3.9	13071	13049.2	3351.6

9 μm Silica C18 (control)
Conc.(mg/mL)	Peak Area	Average	Response Factor (RF)
1.5	5643		3762.0
1.5	5681	5661.7	3787.0
2.1	7537		3588.9
2.1	7556		3598.1 Average (RF) 3658.3
2.1	7472	7521.4	3557.9
2.55	9162		3592.9 SD (RF) 85.8
2.55	9061		3553.2
2.55	9079	9100.6	3560.5 RSD% (RF) 2.3
3	11145		3715.1
3	11178		3726.0 Slope (RF) −17.8
3	11123	11148.7	3707.5
3.45	12911		3742.4
3.45	12940		3750.7
3.45	12962	12937.7	3757.2
3.9	14054		3603.6
3.9	14003		3590.4
3.9	14029	14028.5	3597.2

4 μm Silica C18 (control)
Conc.(mg/mL)	Peak Area	Average	Response Factor (RF)
1.5	5561		3707.6
1.5	5586	5573.8	3724.1
2.1	7268		3460.8
2.1	7280		3466.6 Average(RF) 3537.0
2.1	7321	7289.3	3486.0
2.55	8782		3444.0 SD (RF) 91.8
2.55	8817		3457.6
2.55	8809	8802.7	3454.5 RSD% (RF) 2.6
3	10576		3525.2
3	10591		3530.4 Slope (RF) −42.8
3	10670	10612.2	3556.6
3.45	12434		3604.0
3.45	12470		3614.6
3.45	12604	12502.6	3653.2
3.9	13509		3463.8
3.9	13517		3465.9
3.9	13703	13576.2	3513.6

3.

(A) Linear response for GLP1–1 analog probe semaglutide. Peak area vs concentration – 1.5, 2.1, 2.55, 3, 3.45, 3.9 mg/mL injections (50% to 130% of target injection) evaluated on (i) 6 μm All Carbon microbeads (experimental), (ii) 9 μm silica C18 (control), and (iii) 4 μm silica C18 (control) columns. (B) Response Factor RF (area/conc.) plotted for each injection for (iv) 6 μm All Carbon microbeads (experimental), (v) 9 μm silica C18 (control), and (vi) 4 μm silica C18 (control) columns. The extent to which an RF plot agrees with zero slope indicates the degree of uniformity in peak responses over the injection series. All columns tested have dimensions of 150 mm L × 4.6 mm ID. Mobile Phase A – Ammonium formate (pH 8.5), Mobile Phase B – ACN, Gradient: 20%B – 65%B over 10 min, Temp: 25 °C, Flow rate: 1 mL/min, UV: 220.

The experimental All Carbon media performed better than both reference silica C18 samples in two measures. 1. The y-intercept of the linear model generated from the experimental media was closest to zero among the columns tested. 2. The slope of the response factor graph was also closest to zero for the data from the experimental media. Each of these metrics is an indicator of a strong linear response.

Precision

Precision refers to the reproducibility of multiple measurements of a homogeneous sample. In this investigation, precision measures the column’s characteristics that define the chromatography system’s analytical response. The International Conference on Harmonization (ICH) divides the term precision into three types: (1) repeatability, or the measurement of instrument variation over a short period of time as with sequential injections of 10 or more; (2) intermediate precision, or the agreement of complete measurements when the same method is applied within the same laboratory; and (3) reproducibility, which examines how well a method transfers between laboratories. Given that this work aims to evaluate the instrumental precision of the experimental column and how this compares with commercially available materials, we focused on definition (1) repeatability and evaluated n = 10 consecutive injections of the same homogeneous mixture. Both standard deviation (SD) and average values (AVE) were computed for each chromatographic parameter to compute the relative standard deviation (RSD%) of all injections.

$$\mathrm{RSD}\%=\frac{\mathrm{SD}·100\%}{\mathrm{AVE}}$$ 3

RSD% should be <1% for an HPLC analysis of an API in a pharmaceutical formulation. Depending on the application and governing body, a precision of 5 to 10 RSD% is acceptable for assessing low-level impurities.

Figure represents ten consecutive injections of the GLP-1 analog probes plotted on the same time axis for each of the media tested. This representation of repeated injections offers a visual overview of the stability of each peak’s height, shape, and retention time. The relative consistency in retention time for both peaks developed by the experimental All Carbon microbead media is noticeable. The silica C18 columns show subtle drifting in retention time, most pronounced with the 4 μm silica C18 (control) column.

4.

Ten consecutive injections of the GLP-1 analog probes to validate precision and repeatability. (i) 6 μm All Carbon microbeads (experimental), (ii) 9 μm silica C18 (control), and (iii) 4 μm silica C18 (control).

Table provides the calculated RSD% for retention time (RT), capacity factor (k’), peak area, peak height, symmetry, peak width (50%), plate count (N), resolution, and selectivity, summarizing the 10 injections across all three columns. The experimental columns produced the most precise measurement for retention time with RSD% = 0.06 and 0.05 for semaglutide and liraglutide, respectively. The 9 μm silica C18 (control) column also produced high levels of precision for retention time with RSD% = 0.09 and 0.11. Of note, for the injection method applied, the 4 μm silica C18 (control) column showed a marked variation in retention time, with RSD% values of 1.15 and 2.58. The high variation exhibited by the 4 μm silica C18 (control) column may be due to the smaller particle size of the media, which would require a more developed flush between analyte injections for these peptide probes to prevent column fouling. Repeated injections of large “sticky” analytes can lead to a subtle accumulation on the column, inducing drifting retention times if the flush sequence between injections is not optimized.

3. Relative Standard Deviation% (RSD%) for Liquid Chromatography Response Parameters of Each Probe .

Column	Analyte	RSD % RT	RSD % k’	RSD % Area	RSD % Height	RSD % SYMM	RSD % Width	RSD % Plates	RSD % RES	RSD % SEL
6 μm All Carbon microbeads (experimental)	SG 0.25 mg/mL	0.06	0.15	6.36	3.55	7.56	1.36	2.73	18.35	0.35
	LG 0.25 mg/mL	0.05	0.00	20.61	5.91	17.82	1.44	2.81	1.06	0.00
9 μm Silica C18 (control)	SG 0.25 mg/mL	0.09	0.14	1.73	1.65	1.58	1.65	3.40	1.20	NA
	LG 0.25 mg/mL	0.11	0.14	1.86	2.51	1.96	1.04	1.93	0.92	0.23
4 μm Silica C18 (control)	SG 0.25 mg/mL	1.15	1.48	1.81	2.83	8.69	3.48	5.01	137.35	107.68
	LG 0.25 mg/mL	2.58	3.08	4.23	7.21	14.94	11.14	22.19	4.04	1.61

^aLegend for abbreviated terms: semaglutide (SG), liraglutide (LG), retention time (RT), capacity factor (k’), peak symmetry aka tailing factor (SYMM), resolution (RES), selectivity (SEL).

The observed general trend for precision RSD% for the tested columns was All Carbon experimental column > 4 μm silica C18 control column > 9 μm silica C18 control column. A clue into the root cause of this trend may be seen in the correspondingly high values for symmetry (also known as tailing factor) RSD% for the experimental media. The symmetry factor measures the general shape of the chromatographic peak. Perfectly symmetrical Gaussian peaks produce a symmetry factor of 1. Fronting peaks produce values less than 1, and tailing peaks produce values greater than 1. The chromatograms illustrated in Figure show signs of tailing. Subtle differences in the peak’s shape (from tailing) will introduce variation in the peak’s area. Further optimization of the All Carbon microbead’s pore structure in future iterations should lead to reductions in peak asymmetry and offer improved precision across most chromatographic parameters.

Limit of Detection (LOD) and Limit of Quantitation (LOQ)

In liquid chromatography systems, there is some minimal concentration of a test analyte, which allows for detection and quantification to occur with reasonable confidence. From a statistical perspective, the limit of detection (LOD) is defined for a peak with a signal-to-noise ratio of 3:1. The limit of quantitation (LOQ) follows a similar definition but for a peak with a signal-to-noise ratio of 10:1. In this work, we take the definition of signal-to-noise ratio (S/N) from the US Pharmacopoeia as

$$\frac{S}{N}=\frac{2H}{h}$$ 4

where H is the height of the peak measured from the apex to a baseline extrapolated over a distance greater than or equal to five times the peak width at half height. h is the difference between the largest and smallest noise values observed over a distance greater than or equal to five times the width of the peak at half height.

Both GLP-1 analog probes were injected into each column at concentration values of 1, 0.5, 0.25, 0.1, 0.075, 0.05, 0.03, and 0.01, measured in mg/mL. From this injection series, signal-to-noise ratios were calculated for each probe on each column. The 6 μm All Carbon (experimental) column was evaluated to have an LOD of 0.01 and an LOQ of 0.01 < X < 0.03 for semaglutide with an LOD of 0.03 and an LOQ of 0.05 < X < 0.075 for liraglutide. The 9 μm silica C18 (control) column produced identical results with an LOD of 0.01 and an LOQ of 0.01 < X < 0.03 for semaglutide, and an LOD of 0.03 and an LOQ of 0.05 < X < 0.075 for liraglutide. The 4 μm silica C18 (control) column offered marginally better values with an LOD of X < 0.01 and an LOQ of 0.01 < X < 0.03 for semaglutide, and an LOD of 0.01 < X < 0.03 and an LOQ of 0.03 < X < 0.05 for liraglutide (Table ).

4. Calculated Values for LOD and LOQ for Semaglutide (SG) and Liraglutide (LG) .

Column	SG LOD (mg/mL)	SG LOQ (mg/mL)	LG LOD (mg/mL)	LG LOQ (mg/mL)
6 μm All Carbon microbeads (experimental)	X = 0.01	0.01 < X < 0.03	X = 0.03	0.05 < X < 0.075
9 μm Silica C18 (control)	X = 0.01	0.01 < X < 0.03	X = 0.03	0.05 < X < 0.075
4 μm Silica C18 (control)	X < 0.01	0.01 < X < 0.03	0.01 < X < 0.03	0.03 < X < 0.05

^aLimit of detection (LOD) is defined for peaks with S/N = 3. Limit of quantitation (LOQ) is defined for peaks with S/N = 10.

Loading Capacity – Saturation Capacity

A generally accepted rule of thumb recommends that the injection volume should not be more than 1–2% of the total column volume (for sample concentrations ∼ 1 μg/μL, which is a standard value in analytical liquid chromatography). For the tested 150 mm L × 4.6 mm ID columns, 1–2% volume injections would load ∼15 μg – 30 μg of analyte onto the columns, assuming 0.6 column porosity (see Supporting Information Section 7 for calculations). Chromatographers developed this guidance on sample loading to optimize the resolution and avoid column overload. When sample sizes become too large (from excessive volume or concentration), peak shape changes (tailing and fronting) and widens, deteriorating efficiency and resolution.

The saturation capacity of a column w _s determines the maximum weight of a sample that can be injected for acceptable separation. Values for w _s are functions of the column surface area, pore size, and sample molecular size. For separations where the column’s packing material pore size is matched properly to the size of the target analyte, w _s is proportional to the total surface area available. However, w _s values decrease if sample molecules are precluded from pore entry due to their size (steric hindrance). Saturation capacity is typically more important in preparative separations where the objective is to maximize load for a given resolution target.

The test columns were loaded with increasing sample sizes over consecutive injections. The GLP-1 analog probes were injected onto the test columns in 10, 50, 100, 200, 500, and 1000 μg samples. The parameters of the developed peaks were used to calculate saturation capacity w _s using eq

$$W^2=\frac{16}{N}{t_0}^2{(1+0.5k^{*})}^2+0.9{t_0}^2{k^{*}}^2\frac{m}{w_{\mathrm{s}}}$$ 5

where W is baseline peak width, N is the plate count, t ₀ is the column dead time, k* is the effective retention factor for gradient elution, and m is the sample mass injected (μg).

The series of overlaid chromatograms describing each sample mass loaded (Figure A) portrays a similar picture across all tested columns. As expected, as the sample mass increases, the response of the chromatogram grows proportionally taller and wider. Visually, the experimental column’s chromatograms appear to be in line with those produced by the reference silica C18 columns. Parameters of each unique sample injection were used to calculate saturation capacity w _s for the given column at each sample mass. These individual w _s values were then averaged over all w _s values calculated per column and presented as the column’s saturation capacity. Figure B shows each tested column’s saturation capacity mean and standard deviation. Table offers all of the raw data and calculated values for each column at each injection mass. The 6 μm All Carbon (experimental) column produced a mean saturation capacity of w _s = 615 ± 214 μg. The 9 and 4 μm silica C18 (control) columns produced mean saturation capacity w _s = 1232 ± 949 μg and w _s = 883 ± 702 μg, respectively. The 6 μm All Carbon (experimental) column showed the most consistent mean saturation capacity, as indicated by the standard deviation, compared to the control silica C18 groups. A one-way ANOVA applied to this data showed p = 0.33, indicating no statistically significant differences between these means at the 95% confidence level.

5.

(A) Overlay of injection series of increasing analyte mass load shows a similar response across all tested columns. (i) 6 μm All Carbon microbeads (experimental) column, (ii) 9 μm Silica C18 (control), and (iii) 4 μm Silica C18 (control) columns. (B) Mean saturation capacity w _s for each column. The trend for w _s shows: 6 μm All Carbon microbeads < 4 μm Silica C18 < 9 μm Silica C18, following the corresponding trend in the column’s surface area.

5. Data for Saturation Capacity Calculation .

mass (μg)	W (baseline)	Plates	t0	k*	ws	ws AVE	ws SD
6 μm All Carbon microbeads (experimental)
10	0.593	32581	1.279	2.97	293	615	214
50	1.024	11924	1.279	2.97	491
100	1.1894	5737	1.279	2.97	732
200	1.505	3100	1.279	2.97	918
500	2.9557	1335	1.279	2.97	589
1000	3.9319	642	1.279	2.97	668
μm Silica C18 (control)
10	0.728	11708	1.123	2.97	172	1232	949
50	0.8249	6878	1.123	2.97	673
100	1.0095	4310	1.123	2.97	900
200	1.4125	3118	1.123	2.97	912
500	1.514	2045	1.123	2.97	1998
1000	1.8299	1328	1.123	2.97	2739
4 μm Silica C18 (control)
10	0.7086	48842	1.207	2.97	192	883	702
50	1.1744	14014	1.207	2.97	350
100	1.378	6380	1.207	2.97	511
200	1.5095	3816	1.207	2.97	855
500	1.9026	2125	1.207	2.97	1349
1000	2.1888	1390	1.207	2.97	2044

^aPeak width at baseline (W), column dead-time (t0), and effective retention factor for gradient elution (k*).

Figure S5 in Supporting Information Section 8 presents the observed trend of saturation capacity vs specific surface area across all columns tested. The difference in saturation capacity observed between the experimental and control columns was not statistically significant. The All Carbon microbead (experimental) columns show a lower surface area than silica C18. Thus, their average overall saturation capacity value is lower than that of the silica C18 control columns, suggesting possible loading capacity limits under high peptide loading concentrations. Given the unique fabrication method of the All Carbon microbead media, which utilizes a combinatorial library of inputs, variations on the cross-linker composition and process parameters would increase surface area and enhance saturation capacity. A strength of this novel medium is the ability to uniquely tune its composition by selectively varying inputs to elicit the desired physicochemical properties.

Durability

Preliminary studies investigating the durability of All Carbon microbead (experimental) media were conducted. Experiments included assessing their performance when exposed to ion-pairing agents, extreme acidic and alkaline pH levels (pH 1 and 13), and high salt concentrations under field conditions in the presence of biological matrices (liraglutide crude synthesized by solid-state peptide synthesis) and with 100% aqueous loading. The results indicated that the chromatography performance (efficiency or column backpressure) of the All Carbon microbeads did not significantly change under these various conditions.

During method development, the ion-pairing agent trifluoroacetic acid (TFA) was used extensively as a mobile phase additive on All Carbon microbead columns. No adverse effects were observed for up to 500 injections.

Tables S.7.1, S.7.2, and S.7.3, and Figure S6 in Supporting Information Section 9 present the results of the column backpressure measurements and corresponding chromatograms of the All Carbon microbead columns under acidic and alkaline pH conditions and salt challenge tests. Under repeated washes of over 100 column volumes, the All Carbon microbead (experimental) and silica C18 (control) columns were exposed to 0.1 M NaOH at an alkaline pH of 13 (Table S.7.1), 0.1 M HCl at an acidic pH of 1 (Table S.7.2), and 0.25 M potassium phosphate salt at pH 7 (Table S.7.3). The All Carbon microbead columns did not show any statistically significant increase in column backpressures under any conditions. Conversely, the Silica C18 columns showed a significant change in backpressures with acidic and salt exposure. The chromatograms of All Carbon microbead columns that underwent alkaline washes did not show significant deterioration in the chromatogram profiles (Figure S.6A). Conversely, Silica C18 columns showed the complete loss of GLP-1 analog peak resolution (Figure S.6B). The chromatograms of All Carbon microbeads and silica C18 columns exposed to acidic and salt challenge tests did not show any deterioration in the profiles (Figure S.6C–F).

The performance of the All Carbon microbeads under field conditions in the presence of biological matrices was tested on liraglutide crude samples synthesized by commercial solid-state peptide synthesis. Figure S.7 in Supporting Information Section 10 shows representative chromatograms of the crude samples before and after passing them through the All Carbon microbead column. The chromatogram of the original crude, evaluated to have 60.8% liraglutide purity, showed a large number of impurity peaks. After a single pass through the All Carbon microbead columns, the purity of the liraglutide peak fraction increased to 96.3%.

Exposing silica C18 to 100% aqueous conditions leads to a well-established problem called dewetting during the loading step. , The consequence of dewetting is that water, the mobile phase carrying the analyte (peptides), does not facilitate sufficient interaction between the analyte and the stationary phase. Thus, the analyte does not separate efficiently after passing through the column. Gradient separation experiments were conducted by loading liraglutide onto the All Carbon column under 100% aqueous mobile phase conditions. Figure S.8 in Supporting Information Section 11 shows a representative chromatogram that indicates normal retention for liraglutide under the tested conditions.

These preliminary studies on the durability of All Carbon microbeads are encouraging. Extensive long-term systematic studies investigating their durability are currently underway. These additional studies are testing the reproducibility of the chromatographic separation performance beyond 500 injections, 100 column washes, multiple cycles of peptide crude exposures, and 100% aqueous loading of the peptides to identify the upper limit for each of these conditions.

To the best of our knowledge, this is the first report that systematically investigates the response of different LC stationary phase support media while performing a method validation study for a major component assay for GLP-1 analogs. The aim of a method validation process is to challenge the method to find the limits of allowed variability. Performing first-principles testing on a novel medium designed for LC applications offers key insights into the experimental medium’s properties and the ruggedness of the applied method. It also provides an understanding of the attributes of the GLP-1 analogs. The subtle differences observed in the results across columns can inform future modifications in GLP-1 analog methods to improve specific characteristics and reveal information about the probes.

We initially designed this study to get baseline data on the performance metrics for the experimental media and see how it compares to industry-standard silica C18. However, this work signals greater significance than introducing new reversed-phase liquid chromatography media. The similarity in response to GLP-1 analogs is remarkable between columns of such strikingly different stationary phase material composition.

Additional work is needed to confirm whether these observations can be generalized to other peptides.

Given that the constituents of the All Carbon microbeads do not degrade under alkaline conditions, are free of surface silanols, and can run 100% aqueous mobile phases, this novel material could specifically address key shortcomings of silica C18. These features that mark durability for stationary phase materials are intrinsic to the All Carbon microbeads and could advance sustainable peptide manufacturing. Reversed-phase chromatographic purification, a crucial unit operation during synthetic peptide production, is a resource-intensive section of pharmaceutical manufacturing. Advances in the liquid chromatography stationary phase media used to purify drugs, such as GLP-1 analogs, could reduce costs for patients, increase access, and improve operational costs for manufacturers.

Conclusions

All carbon microbead-packed experimental reversed-phase HPLC columns reproducibly separate the GLP-1 analog test probes semaglutide and liraglutide. Their retention times are similar to those of reference silica C18 control columns. The All Carbon microbead (experimental) and silica C18 (control) columns exhibited equivalent liquid chromatography performance metrics, plate count (N), limit of detection (LOD), limit of quantitation (LOQ), and loading capacity under identical testing conditions. The experimental media offered better performance in the precision of retention time and linear response compared to the control columns. The All Carbon microbead’s chromatographic performance, along with its potential stability under alkaline conditions and 100% aqueous mobile phases, opens avenues for its further development as a reversed-phase media for sustainable GLP-1 analog peptide manufacturing.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsomega.5c02172.

• Method optimization design, experiments and analysis, flow rate analysis, calculations used in the loading capacity – saturation capacity section, pH and salt challenge tests, liraglutide crude analysis, and 100% aqueous loading study (PDF)

M.J.P. conceptualized, designed, and performed the experiments, analyzed the data, and wrote the paper; B.S. conceptualized the experiments, analyzed the data, and wrote the paper.

The authors declare the following competing financial interest(s): Millennial Scientific and the investigators have filed patents. They are developing commercial products related to the technology reported in this article.