Annular centrifugal extraction for the recovery of non-sugar compounds from honey Authors: Gabriela Belén Lemus Ringele, Petros Gkiouvetidis, Dimitris Michailidis, Elodie Bossard, Lorène Gourguillon, Stavros Beteinakis, Leandros A. Skaltsounis, Maria Halabalaki Published: 2025 ABSTRACT The high viscosity of honey, due to its carbohydrate content, complicates analysis and hinders the study of low-concentration, yet crucial, honey phenolics. Consequently, many compounds remain unknown or are only tentatively identified. To address these challenges, this study introduces Annular Centrifugal Extraction (ACE) - a continuous-flow liquid-liquid extraction method under centrifugation - as a novel approach for investigating honey phenolics. ACE enables high-yield recovery of extracts while removing carbohydrates. Despite its application in various fields, ACE remains underutilized in food analysis, making this the first report on its application to honey. The method´s efficiency was evaluated qualitatively and quantitatively using nuclear magnetic resonance (NMR) spectroscopy and liquid chromatography coupled to high-resolution mass spectrometry (LC-HRMS). To assess extraction repeatability, ACE was applied to 50 honeys of different botanical origins, with the resulting data analyzed via multivariate analysis (MVA). Furthermore, the method was compared with established honey analysis methods, such as liquid-liquid extraction (LLE), adsorption resins, and solid-phase extraction (SPE), to assess its performance and suitability in both analytical and preparative scales. ACE proved to be superior, offering repeatability, scalability, and compatibility with different scale set-ups and therefore various applications. Its potential extends beyond honey, with possible applications in other matrices.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-025-31250-4. Introduction Honey is a globally significant commodity, recognized as the oldest and only natural sweetener that can be consumed without processing1. It is generally categorized based on its origin as either blossom or honeydew honey. Blossom honey primarily comes from the nectar of flowers, whereas honeydew honey is made by bees from the sugary secretions of aphids. Chemically, honey is a natural product mainly composed of sugars (fructose, glucose, sucrose) and water, along with small amounts of minerals (0.1–0.7%), proteins (0.2–2.9%), free amino acids (approx. 1%), organic acids (0.1–1.1%), phenolics (0.02–0.2%) while vitamins, enzymes and aroma compounds account for less than 0.1%2,3. The high market demand, driven by its nutritional value and health beneficial properties, has made honey a prime target for adulteration. The goal of such fraud is to increase production while reducing costs, which can also potentially compromise human health4. Given the widespread practice of adulterating honey with sweeteners, most methods for verifying quality focus on sugar content, or occasionally assessing volatile compounds. Over recent decades, several analytical methods have been developed and applied to ensure quality, with isotope ratio mass spectrometry (IRMS) being one of the most established techniques5. Apart from carbohydrates, nearly all chemical classes in honey are studied to identify adulteration markers, detect various types of fraud, and verify authenticity and quality. Further information can be found in several related review articles6–8, as well as the very informative book chapter by Santos-Buelga and González-Paramás3. Amongst these lines, increasing attention is given to the more specific group of minor compounds in honey, commonly referred to as “honey phenolics.” Phenolics, although extremely low in concentration, vary qualitatively and quantitatively between different honey types, and their composition depends on botanical and geographical origin9. Also, the phenolic profile can change significantly over time or due to processing and storage conditions, qualifying them as valuable quality markers10. The honey phenolics group is continuously enriched given the available literature, including a wide range of substances from various chemical families, such as phenolic acids, flavonoids, ionones and terpenoids, to name just a few. Additionally, these low-concentration compounds are known to contribute to honey’s biological properties, including its antimicrobial, antioxidant, wound-healing, anti-inflammatory and cancer-preventive effects11. However, one of the major challenges in studying honey phenolics, whether for use as quality markers or for exploring their biological properties, is their identification. These compounds, which originate from both nectar and honeydew, and are further modified by honeybees, form a chemically complex and unexpected mixture. A key difficulty is their extremely low concentration, which requires the use of advanced techniques, primarily liquid chromatography (LC) coupled with UV or mass spectrometry (MS) detectors, for structure determination12. However, accurate identification heavily depends on the availability of standards, which in turn excludes a large number of unknown compounds, most of which remain unidentified in most analyses. In other cases, where reference standards are not used, tentative identification of phenolics is proposed with therefore low identification confidence. Moreover, honey’s intricate chemical composition and viscous nature, due to excessive carbohydrate content, presents also challenges in handling, sample preparation and recovery of phenolics13. A key difficulty in honey and foods assessment in general lies in the analytical process itself, and particularly the sample preparation phase before analysis. This phase involves critical steps such as extraction, evaporation, and enrichment, which can account for approximately 70% of the total analytical workflow14. Sample preparation must be tailored to the specific substrate being analyzed and is highly dependent on the analytical instrumentation used15,16. Therefore, any sample processing prior analysis is decisive for the accuracy and validity of the analytical results. Furthermore, in most cases sample preparation methods are designed to be compatible with analytical scale requirements. This does not allow high extraction yield and, therefore, the recovery of phenolics in adequate quantity enabling de novo structure identification using techniques such as NMR spectroscopy. In this regard, the recovery of honey phenolics in high yield with the simultaneous elimination of sugars is a particularly complex task. A significant hurdle in extracting non-sugar compounds is the formation of emulsions during honey dilution, typically in a biphasic system of water and organic solvents. These emulsions are sample-dependent and often difficult to prevent, control, or disregard17. Their occurrence leads to reduced repeatability in the extraction process, compromising the overall validity of the analytical results. In literature, different extraction techniques regarding the non-sugar constituents of honey have been reported. Common methods include liquid-liquid extraction (LLE)18, solid-phase extraction (SPE)19, and amberlite sorbents20. Based on the reported applications, SPE is almost exclusively used on characterization of honey phenolics (qualitatively and/or quantitative, analytical scale), resins mainly for phenolics recovery and preparative purposes, while LLE have been used for both. Figure 1 illustrates the main advantages and drawbacks of commonly used honey pretreatment methods. Several issues related to small repeatability, low yields, and the inability to completely remove sugars could be considered as the most important19. Amberlite XAD resins (mainly XAD2 and XAD4) have shown lower recoveries compared to C18 SPE cartridges21, and require an additional clean-up step after application20. SPE is widely used employing various sorbents for honey analysis e.g. C18, HLB (Oasis), polymeric (Strata-X) and copolymers (Poly-Sery PSD)22,23 and, while being more selective for non-polar compounds, demonstrates lower recoveries for polar substances19. Moreover, it is characterized by low capacity or necessary repeated steps when focusing on high recovery for isolation purposes. LLE commonly employs organic solvents (EtOAc, DCM, ACN, MeOH, EtOH) and acidified – or not – water. This technique also faces challenges, particularly with the formation of interfacial emulsions, which leads to low yields and non-repeatable results24. Amongst these lines, the Annular Centrifugal Extractor (ACE) is a single-stage extractor based on the continuous counter-current liquid-liquid extraction principle using centrifugation and gravity separation. It uses two co-axial cylinders to ensure the mixing of two immiscible phases, commonly aqueous and organic25, while simultaneously allowing their separation without transferring one phase into the other26. ACE has been successfully employed across different fields, such as biochemical processes (e.g., biodiesel production, extraction of biophenols from vegetable oils etc.), pharmaceutical processes (e.g., extraction of antibiotics, hydrocortisone, caffeine and L-phenylalanine, etc.), wastewater treatments (e.g., extraction of volatile organic compounds), and the chemical industry (e.g., extraction of rare elements and metals)27. The use of ACE and generally annular centrifugal contactors offers significant advantages in terms of extraction efficiency, scalability, and speed. Their ability to handle continuous processing, gentle separation, and precise control makes them a valuable tool in complex mixtures and high volumes27,28. Despite its versatility, the use of ACE in the food sector remains highly limited with only a handful of applications reported so far27. Fig. 1Advantages and disadvantages of the most common techniques used for recovery of honey phenolics based on literature. Advantages and disadvantages of the most common techniques used for recovery of honey phenolics based on literature. As part of our group’s ongoing research on honey phenolics, we present in the current study a novel method using ACE for optimizing the extraction of honey and the recovery of phenolics. The proposed method is designed to enable the simultaneous recovery of honey phenolics and effective removal of carbohydrates. Also, enables the recovery of honey phenolics in high yield enabling their further separation and isolation in pure form, as well as their accurate identification. Given the unique physicochemical properties of honey and its strong tendency to form emulsions in aqueous environments, ACE was applied in this study in a non-conventional way. The proposed method is based on the intentional and controlled formation of emulsions in a biphasic system, followed by the separation of the aqueous and organic phases using continuous centrifugal force and, therefore, recovery of sugar and non-sugar extracts, respectively. The high capacity of ACE, its continuous nature and the removal of sugars allows the recovery of minor metabolites that are overlooked due to their low concentration. Furthermore, it facilitates the isolation of phenolics in pure form, enabling their accurate identification using HRMS/MS and, primarily, 1D and 2D NMR techniques. To evaluate its applicability also in analytical setups, the ACE-based method was compared with conventional, well-established protocols in honey analysis. Key parameters such as extraction efficiency, recovery, and selectivity were assessed using 1D NMR and LC-HRMS. Materials and methods Chemicals and materials Distilled water (d-H2O), methanol (MeOH; analytical grade; Fisher Scientific, New Hampshire, USA), ethyl acetate (EtOAc; analytical grade; Macron Fine ChemicalsTM, Leicestershire, England), chloroform (CHCl3; HPLC grade; Carlo Erba Reagents S.A.S, Cornaredo MI, Italy) and dichloromethane (DCM; HPLC grade; Carlo Erba Reagents S.A.S), were used for the extraction of honey samples. XAD-type adsorbent resin (XAD-4 and XAD-7 HP; Amberlite; 20–60 mesh) were obtained from Merck SA (Athens, Greece). Deuterated methanol (methanol-d4, 99.8% D; Merck SA) was used for NMR analysis, while hexamethyldisiloxane (HMDSO, NMR grade, ≥ 99.5%; Sigma-Aldrich Corporation; St. Louis, MO, USA) was used as an internal standard (IS). Ultrapure H2O (Millipore Direct-Q®3 UV purification system, Burlington, MA, USA), MeOH (LC-MS grade, J.T. Baker®, Gliwice, Poland), acetonitrile (ACN, Merck SA) and formic acid (FA, Optima® LC-MS, Fisher Scientific) were used for UPLC-HRMS/MS analysis. Honey samples For the extraction procedures, three honey samples with different botanical and geographical origin were analyzed: flower honey (FH) from Ouessant (France), heather honey (HH) from Ikaria (Greece) and pine honey (PH) from Evoia (Greece). For principal component analysis (PCA), 50 samples representing different botanical origins were extracted and analyzed to further assess the extraction repeatability of ACE. Samples were stored at 4 °C pending analysis. Sample preparation Extracts obtained from established approaches were compared with the ones from the newly developed method using ACE. The use of Amberlite XAD resins and LLE was already described in our previous work24. Extracts obtained with all four techniques were assessed with 1H NMR spectroscopy to evaluate the efficiency of sugar removal and identification purposes, and LC-HRMS/MS to examine the recovery, selectivity and repeatability of each approach. Specifically: Amberlite XAD-4 and XAD-7 resins 10 g of honey were dissolved in 100 mL of d-H2O and 30 g of Amberlite XAD-7 resin were added. The suspension was stirred for 3 h, filtered and washed three times with 200 mL of d-H2O. The resin was then transferred to a glass container and non-sugar compounds were extracted with 100 mL of MeOH and an ultrasonic bath (20 min). Afterwards, the resin was filtered and washed with 100 mL MeOH. The extraction and washing procedures were repeated in triplicate. Methanolic phases were combined and evaporated to dryness. The procedure was also carried out with acidified water (pH = 2, HCl). Additionally, the extraction was repeated with XAD-4, both with acidified and d-H2O. As a clean-up step, the extract obtained from the resin was dissolved in 10 mL of d-H2O and was extracted with EtOAc (3 × 10mL). Organic phases were combined and evaporated to dryness. Solid-phase extraction (SPE) 3 g of honey were dissolved in 5 mL of d-H2O. SPE cartridges (silica-bonded C18, Discovery®, DSC-18, Supelco) with a 6 mL capacity were preconditioned with 5 mL of d-H2O followed by 5 mL of MeOH and 5mL of d-H2O. The solution was then loaded onto the cartridge and washed with 5 mL of d-H2O. Recovery of target compounds was accomplished with 5 mL of MeOH (A) and 5 mL of MeOH/EtOAc (1/1 v/v) (B). Obtained extracts were evaporated to dryness. Liquid-liquid extraction (LLE) 6 g of honey were dissolved in 15 mL of d-H2O. Then, 15 mL of EtOAc/DCM (80/20 v/v) were added, and the mixture was shaken for 10 min with a vortex agitator. Following centrifugation at 4000 rpm for 15 min, organic phases were collected and evaporated to dryness. Annular centrifugal extraction (ACE) A laboratory scale annular centrifugal extractor, model BXP012, (Rousselet-Robatel Kromaton, Annonay, France) was employed. 12 g of honey were weighed and dissolved in 60 mL of d-H2O. 60 mL of the organic system EtOAc/DCM (80/20 v/v) were added, and the mixture was stirred for 10 min. The solution, under continuous stirring, was pumped into ACE at a flow rate of 8 mL/min (basic Verderplux pump, Castleford, UK), with a rotor speed of 8500 rpm. Among the two obtained separated phases, the organic was evaporated to dryness. Each sample was extracted in triplicate to evaluate the repeatability. Nuclear magnetic resonance (NMR) All obtained extracts were dissolved in 650 µL methanol-d4 and transferred into NMR tubes. As internal standard, HMDSO at a concentration of 0.02% v/v was used. Spectra were recorded using parameters previously described, with slight modifications24. 1H NMR experiments were recorded on a Bruker AVANCE III 600 NMR spectrometer (Bruker BioSpin AG, Faellanden, Switzerland), using a 60-place sample changer (B-ACS-60). The spectrometer had a proton frequency of 600.11 MHz (B0 = 14.1T) and was equipped with a z-gradient inverse detection 5-mm probe and a Bruker Cooling Unit (BCU) for temperature control at 305 K. Acquisition parameters include: number of scans, 64; π/2 pulse, ~ 8 us, time domain (TD), 64 k data points; acquisition time, 3.9 s; relaxation delay, 2 s; and spectral width, 8417.51 Hz. An exponential multiplication window function with a line-broadening factor (lb) of 0.3 Hz was applied on the free induction decay (FID) to obtain the final spectra. Resulting spectra were automatically phase- and baseline-corrected using the Bruker TopSpin software (version 4.2.0). Chemical shifts were reported with respect to methanol-d4 at 3.31 ppm. Key metabolites were identified by comparing their chemical shifts and multiplicities with those of previously isolated honey compounds, using both one- and two-dimensional NMR experiments (data not shown). Ultra-high performance liquid chromatography high resolution mass spectrometry (UHPLC-HRMS) Analysis was performed on a Velos Pro Ion Trap – Orbitrap Elite hybrid mass spectrometer (Thermo Scientific, Bremen, Germany), with a heated electrospray ion source (H-ESI II) hyphenated to a Waters Acquity H-class UPLC (Waters Corporation, Milford, MA, USA) system. Separation was achieved with a Waters Acquity BEH C18 column (2.1 mm × 150 mm, 1.7 μm) heated at 40 °C. Mobile phase consisted of H2O with 0.1% v/v FA (A) and ACN (B) at a flow rate of 0.3 mL/min. Elution started with 95% A which was maintained for 1 min. The system transitioned to 80% A within the next 2 min and reached 40% A in another 17 min. Finally, 100% B was completed in 3 min from the last point and maintained for 4 min. The system returned to initial conditions within 1 min, where it was equilibrated for 2 min. Total runtime was 30 min, injection volume was 10 µL, while the autosampler maintained a temperature of 10 °C. Honey extracts were reconstituted at 1mL of MeOH/H2O (80/20 v/v). Then, 150 µL of the solutions were diluted to a final volume of 1 mL of the same solvent system. Mass spectra were obtained in negative ionization mode, at a mass range of m/z 113–1000, with a resolving power of 60,000. Capillary and source heater temperature were set at 350 °C, while source voltage was 3.5 kV and S-lens RF level 45%. Sheath and auxiliary gas flows were 45 and 15 arbitrary units, respectively. A data-dependent (DDA) method was set up involving two scan events: a full MS scan followed by HRMS/MS acquisition. HRMS/MS experiments were obtained with a collision energy of 35.0% (q = 0.25) and a resolving power of 30,000. Data recording was performed with Xcalibur v. 3.0.63 software. Metabolite identification using HRMS and principal component analysis (PCA) HRMS data were processed using the MZmine v.2.53 software. Raw data files were imported and subjected to mass detection and chromatogram building, applying an error tolerance of 5 ppm for peak picking and mass list generation. Subsequent steps included chromatogram deconvolution, deisotoping, alignment, and gap filling. Following noise reduction and filtering, a final feature list comprised of 175 features was generated. Annotation of compounds was performed based on literature and using online databases like The Food Database (FooDB), NIST and METLIN. Using MATLAB (version R2024a) and in-house routines, further spectral processing was performed to the NMR raw data. The icoshift tool29 was used for alignment. NMR spectra with spectral width from − 0.2 to 12 ppm were segmented into bins with a size of 0.01 ppm. Data were normalized using total intensity. To build the PCA model using the environment of SIMCA (v. 17.0.2, Sartorius AG, Göttingen, Germany), bins corresponding to solvent signals, baseline noise and IS were removed to avoid interferences with the model. Results and discussions Extraction using ACE As previously mentioned, ACE is an extractor that performs continuous counter-current liquid extraction by combining centrifugation and gravity-based separation. It employs two co-axial cylinders to efficiently mix immiscible phases while enabling their separation without phase carryover (Figs. 2 and S1a). In more detail, the immiscible liquids enter the mixing chamber (between the two cylinders) through the heavy and light phase inlets, which are separated. The rotating impeller imparts power which results in a fine dispersion of the two immiscible phases27. This dispersion starts descending and flows in the region below the rotating cylinder, entering through the rotor inlet to the separation chamber. After entering, the liquid mixture gets deflected towards the wall and gets separated as it moves upwards. Centrifugal forces act on the mixture, causing the denser phase to migrate towards the outer edge of the separation chamber, and the less dense phase to move towards the center, closer to the separation vane. The radial motion of the mixture creates a structure with different layers that corresponds to a density gradient. Once separation is achieved, each phase flows through specific weirs towards the corresponding outlet (heavy phase and light phase outlet)30. This study marks the first attempt to apply the ACE technology to honey for the repeatable extraction and recovery of low concentration level honey phenolics. A distinctive aspect in our extraction method is the unconventional application of ACE. Traditionally, ACE operates by introducing heavy and light phases (more specifically, in this case, aqueous and organic) separately into the system, allowing the equipment to facilitate their interaction and subsequent separation. In contrast, the approach proposed in this study was specifically designed to address the uncontrolled emulsions typically formed during LLE of honey pretreatment, which can affect the consistency and robustness of the extraction. More specifically, it involved the intentional pre-forming and stabilization of emulsion through continuous stirring before its introduction into the extractor. Therefore, rather than feeding the device with two separate non-miscible solvents, a single solution is introduced (Fig. S1b). This controlled pre-emulsification ensures a more regulated phase interaction, allowing the emulsion to break down systematically within the extractor, rather than forming in situ, as seen in conventional LLE. So, once this mixture enters the separation chamber, the emulsion is greatly reduced by centrifugal forces, and the organic and aqueous phases are effectively separated containing the phenolics and sugars, respectively (Fig. S1c). A schematic representation of the ACE workflow is presented in Fig. S2. Fig. 2Graphical representation of an annular centrifugal extractor (ACE) for honey extraction, based on the scheme produced by CINC®27. Graphical representation of an annular centrifugal extractor (ACE) for honey extraction, based on the scheme produced by CINC®27. Based on this principle and for the development of the method, the most suitable solvent system for phenolics’ recovery was chosen and ACE parameters were optimized. Multiple trials were conducted and finally EtOAc/DCM (80/20 v/v) was selected as optimum based on (a) stability of emulsion prior extraction, (b) extraction yield and (c) successful removal of sugars. Starting quantity of honey was also assessed, as well as the matrix/solvent ratio, and was set to 12 g (Table S1). Regarding the ACE parameters i.e. flow rate and rotor speed, they were set to 8 mL/min, and 8500 rpm, respectively based on the effective separation of the two phases and efficient recovery of phenolics (Table S1). For method development and optimization, a flower honey (FH) sample, the most common type, was used. To evaluate the method’s applicability and reliability, two additional samples from different geographical and botanical origins were also analyzed to account for potential variability among honeys. Specifically, apart from FH, from Ouessant (France), heather honey (HH) from Ikaria (Greece) and pine honey (PH) from Evoia (Greece) were treated with ACE. It is important to note that pine honey is honeydew compared to the other two samples belonging to blossom honeys. Each sample was extracted three times. Upon completion of the extraction, two distinct phases were recovered for each sample, with the organic being subjected to further analysis. After evaporation, 1H NMR was used to assess the efficiency of sugar removal and for compositional information. Furthermore, to evaluate the extraction repeatability, 50 honey samples were selected from our in-house honey bank and extracted twice with the proposed method. After the determination of phenolics extracts yield, the extracts were subjected to 1D NOESY NMR analysis and multivariate data analysis (MVA). Specifically PCA was employed as an unsupervised method to depict samples classification and extraction repeatability. LC-HRMS/MS was also employed to evaluate recovery and selectivity aspects of the extraction based on the detected features and identified compounds. As shown in the1H NMR spectra (Fig. 3) of the three honey samples analyzed first (FH, HH and PH), ACE afforded great results, as it allowed both the recovery of enriched extracts with minor compounds and the simultaneous removal of sugars. Moreover, the chemical profile is clearly different, as expected, given the different botanical and geographical origins. Employing also 2D NMR (data not shown) and literature data, major compounds can be identified in all extracts, such as 5-hydroxymethylfurfural (5-HMF) for FH, 1-(4-methoxyphenyl)-ethane-1,2-diol for HH, and 2-trans-decenedioic acid for PH. It is worth noting the removal of free sugars as indicated by the limited resonances that appeared between 3 and 5 ppm. Fig. 3Representative1H NMR spectra (methanol-d4) of (a) Flower honey-FH, (b) Heather honey-HH and (c) Pine honey-PH extracts obtained with ACE. All extracts present different profiles due to their botanical origin. Major compounds are annotated: 5-hydroxymethylfurfural (FH), 1-(4-methoxyphenyl)-ethane-1,2-diol (HH), and 2-trans-decenedioic acid for (PH). Characteristic resonances for identified compounds are noted. Carbohydrates (sugars) region is also marked. Representative1H NMR spectra (methanol-d4) of (a) Flower honey-FH, (b) Heather honey-HH and (c) Pine honey-PH extracts obtained with ACE. All extracts present different profiles due to their botanical origin. Major compounds are annotated: 5-hydroxymethylfurfural (FH), 1-(4-methoxyphenyl)-ethane-1,2-diol (HH), and 2-trans-decenedioic acid for (PH). Characteristic resonances for identified compounds are noted. Carbohydrates (sugars) region is also marked. For further evaluation of the extracts LC-HRMS/MS analysis was also performed (Fig. 4) and several compounds with different polarities could be detected. Like the NMR spectra, the chromatographic profiles differ according to their botanical source. Based on different chromatographic and spectrometric features in full scan and HRMS/MS level, key compounds could be also annotated such as 4-quinolone (m/z 144.0448) and abscisic aldehyde (m/z 247.1335) for FH, both trans-trans and cis-trans isomers of abscisic acid (m/z 263.1284) for HH, and 2-trans-decenedioic acid (m/z 199.0972) along with sebacic acid (m/z 201.1126) for PH. NMR spectra (Figs. S3, S7, S12) and LC-HRMS chromatograms (Figs. S6, S10, S15) of each sample’s triplicates can be found in the supplementary materials, along with key identified metabolites from NMR (Tables S2, S4, S6) and annotations from LC-HRMS (Tables S3, S5, S7). Chemical structures of metabolites identified via NMR are also provided (Figs. S4, S5, S9, S14). Fig. 4Base peak LC-ESI(−)-HRMS chromatograms of the three extracts obtained with ACE and indicative annotated compounds: (a) 4-quinolone and abscisic aldehyde in Flower honey-FH, (b) trans-trans and cis-trans isomers of abscisic acid in Heather honey-HH and (c) 2-trans-decenedioic acid and sebacic acid in Pine honey-PH. Base peak LC-ESI(−)-HRMS chromatograms of the three extracts obtained with ACE and indicative annotated compounds: (a) 4-quinolone and abscisic aldehyde in Flower honey-FH, (b) trans-trans and cis-trans isomers of abscisic acid in Heather honey-HH and (c) 2-trans-decenedioic acid and sebacic acid in Pine honey-PH. Comparison with other commonly used extraction techniques For the next step, other commonly used methods, i.e. SPE, LLE and XAD-7 resin, were examined by analyzing the same samples and comparing the results obtained with the ones from ACE. The obtained extracts, similarly to ACE extracts, were also subjected to NMR and LC-MS analysis, while the extraction yield was also evaluated. SPE is broadly applied especially for analysis in food matrices, among them honey. Reversed-phase SPE (RP-SPE) is most employed and is widely applied as a clean-up step after extraction in acidic conditions19. Nevertheless, there are several studies where SPE was used as a one-step extraction technique to obtain a range of compounds31–33. In the current study, silica-bonded C18 cartridges were used, together with MeOH and a mixture of MeOH/EtOAc (1/1 v/v) for the extraction as a commonly used method34. Since MeOH eluted most of the constituents, NMR spectra obtained with MeOH/EtOAc were almost empty of compounds (data not shown). Due to its simplicity, conventional LLE is one of the most applied extraction techniques. Over the last years, alternative methods have been developed, such as homogeneous LLE (h-LLE) and salting-out assisted LLE35. Regarding honey, LLE is widely performed using EtOAc or ethanol as extractant solvent12. In one of our previous studies24, different solvent systems and quantity of starting material were tested. It was observed that the presence of EtOAc in each organic system improves the chemical profile obtained, with the successful elimination of most sugars. Nonetheless, as previously mentioned, a significant amount of emulsion is formed during the extraction. In 2010, Schievano et al.36 presented an NMR-based study for the determination of botanical origin in honey, where an LLE with CHCl3 was proposed using mechanical stirring and centrifugation. Based on the latter, after several trials using this method, the solvent system EtOAc/DCM (80/20 v/v) was found to provide an enriched extract, with satisfactory yields. It is important to note that the use of centrifugation decreases the amount of emulsion formed between the two phases. XAD-4 is a sorbent resin made of styrene-divinylbenze commonly reported for the recovery of small phenolic compounds19, while XAD-7 is a polyacrylic resin that has the capacity to absorb polyphenols with high molecular weight37. In the current study, honey samples were treated using both sorbents with and without acidified water. Literature recommends the use of acidified water, as it allows high recovery of certain target compounds such as flavonoid aglycones17. In our case, the use of acidified water did not show differences in the profile of obtained extracts, while XAD-7 provided a richer profile than XAD-4. Moreover, to achieve a satisfactory elimination of sugars further treatment was required. Certain studies performed in honey mention the application of LLE as a clean-up step38,39. Thus, an LLE was also performed as a clean-up step here using EtOAc as extraction solvent. Sugar removal and extraction yield Apart from the visual inspection of spectra, sugar removal was evaluated across all tested extraction protocols by a relative quantification of sugars. Specifically, α-D-glucose (αD) was selected as indicative sugar which along with β-D-glucose accounts for 25–40% of honey. In more detail, for each extract, the ratio of the peak area of the anomeric proton of α-D-glucose (resonating at 5.05 ppm) to that of the IS (HMDSO) peak area (resonated as a single peak) was calculated. The anomeric proton peak of α-D-glucose was selected due to its distinct signal and absence of overlapping in the spectra, facilitating accurate integration. By examining all three selected honeys, Table S8 shows that the XAD-7 extracts exhibited an αD/IS ratio more than three hundred times higher than that of ACE, indicating a substantially higher amount of sugars in the XAD-7 extracts. This can also be observed based on the extraction yields obtained (Table S9). After the LLE clean-up step, the yields were significantly reduced, since a high quantity of sugars were removed. Nevertheless, the αD/IS ratio remained close to two times higher than that observed with ACE. A similar trend was observed with the SPE extract. While SPE appeared to achieve higher extraction yields compared to both LLE and ACE, it failed to sufficiently remove sugars from honey, since it exhibited a ratio of more than eight times higher than that of ACE. Interestingly, LLE alone appeared to remove sugars as effectively as ACE. Furthermore, both LLE and ACE provide satisfactory yields for further isolation. Evaluation of extraction efficiency and selectivity To assess the efficiency of ACE compared to adsorbent resins, SPE, and LLE, additional criteria were considered. These included selectivity, ease of use, and the number of steps and time required. To ensure greater coverage in terms of physicochemical behavior, as already mentioned, we selected honeys of different botanical origin and consistency, and each one will be examined separately below. Starting with flower honey (FH), adsorbent resin XAD-7 presented limited selectivity towards phenolics and is both labor-intensive and time-consuming, with incomplete carbohydrate elimination. The1H NMR spectrum (Fig. 5a) exhibits several peaks in the carbohydrate area (3–5 ppm), as well as strong signals of α- and β-D-glucose. This is also evident in the LC-MS chromatogram (Fig. 6a), where sugars are eluted with the solvent front as a single broad peak, together with gluconic acid, a substance produced by the oxidation of glucose40. After treatment of the XAD-7 extract with LLE (Fig. 5b), phenolics were recovered providing a spectrum comparable to that obtained with ACE. Signals corresponding to compounds such as 5-HMF, 4-quinolone, 1-(4-methoxyphenyl)-1,2,3-trihydroxypropane, 2-trans-4-trans-desoxyabscisic acid, benzoic and 4-hydroxybenzoic acid could be observed. Some other detected compounds are observed in the LC-HRMS chromatogram (Fig. 6b) such as 4-quinolone (m/z 144.0448), 3-phenyllactic acid (m/z 165.0555), 2-trans-decenedioic acid (m/z 199.0972), sebacic acid (m/z 201.1127), isorhamnetin (m/z 315.0501), abscisic aldehyde (m/z 247.1340) and a derivative of abscisic aldehyde (m/z 257.1394). Fig. 5Representative1H NMR spectra (methanol-d4) of flower honey-FH extracts obtained with different techniques tested: (a) XAD7, (b) LLE following the treatment with XAD7, (c) SPE, (d) LLE, (e) ACE. Identified compounds are annotated: (1) 5-HMF, (2) 4-quinolone, (3) 1-(4-methoxyphenyl)-1,2,3-trihydroxypropane, (4) trans-trans-desoxyabscisic acid, (5) benzoic acid, (6) 4-hydroxybenzoic acid, (7) α-D-glucose and (8) β-D-glucose. Carbohydrates (sugars) region is also annotated. Representative1H NMR spectra (methanol-d4) of flower honey-FH extracts obtained with different techniques tested: (a) XAD7, (b) LLE following the treatment with XAD7, (c) SPE, (d) LLE, (e) ACE. Identified compounds are annotated: (1) 5-HMF, (2) 4-quinolone, (3) 1-(4-methoxyphenyl)-1,2,3-trihydroxypropane, (4) trans-trans-desoxyabscisic acid, (5) benzoic acid, (6) 4-hydroxybenzoic acid, (7) α-D-glucose and (8) β-D-glucose. Carbohydrates (sugars) region is also annotated. For SPE, performance was more affected by sample variability and complexity than other methods, impacting its selectivity, as also reported in literature21. Even though SPE provides a rich profile, as can be seen across the NMR spectrum (Fig. 5c), higher signals corresponding to sugars are observed compared to other techniques. Moreover, while it presents a similar profile to that of ACE, there are certain qualitative, but mainly quantitative differences that could potentially be attributed to the retention of compounds on the solid stationary phase of the cartridge. Certain examples were the higher concentration of 4-quinolone, the lower concentration of 5-HMF and desoxyabscisic, and the absence of 4-hydroxybenzoic acid. In the LC-HRMS chromatogram of the SPE extract, compounds of high and medium polarity seem to be in higher abundance (0–11 min) compared to less polar constituents (11–23 min) (Fig. 6c). The same compounds, as previously, are observed, in addition to two other glycosylated compounds: junipediol A 2’-O-β-D-glucopyranoside (m/z 405.1398, [M + CHO2]−) and phenethyl rutinoside (475.1819, [M + CHO2]−). Regarding LLE, the most significant limitation is the formation of emulsions, as previously mentioned41. This leads to issues with repeatability, making it difficult to implement in a more automated setup. As shown in Fig. 5d, the NMR spectrum obtained from LLE exhibits a richer profile compared to those from SPE and LLE treatment of the XAD-7 extract, revealing additional peaks not only in the aromatic region, but also upfield. Moreover, as previously discussed, most of the sugars were successfully eliminated and the resulting spectrum resembles that obtained using ACE (Fig. 5e). Similarly, the LC-HRMS chromatograms of both extracts (Fig. 6d,e) show comparable profiles, with ACE showing an additional peak corresponding to spiculisporic acid (m/z 327.1810), along with several small peaks attributed to non-polar compounds. Following the same steps, NMR spectra and LC-HRMS chromatograms of heather and pine honeys were also compared. As previously, the NMR spectra of XAD-7 extracts of both heather and pine honeys displayed strong signals corresponding to α-D-glucose and β-D-glucose (Figs. S8a and S13a). Additionally, in both LC-HRMS chromatograms (Figs. S11a and S16a), gluconic acid was present along with erlose (m/z 549.1668). Specifically, for heather honey, following the clean-up step of XAD-7 extracts, minor compounds became visible in the NMR spectrum (Fig. S8b). However, this profile was poorer compared to the spectrum obtained with ACE. In the LC-HRMS chromatogram, the primary quantitative difference between these two extracts (Fig. S11b) was the fluctuation of some peaks that corresponded to compounds like 2-trans-decenedioic acid (m/z 199.0972), sebacic acid (m/z 201.1130) and pinobanksin (m/z 271.0609). The NMR spectrum of the SPE heather extract (Fig. S8c) resulted in the absence of signals corresponding to 4-methoxybenzoic acid and oleic acid, indicating a loss of valuable information. For example, 4-methoxybenzoic acid has been previously reported as a biomarker for Erica arborea honeys24. The corresponding LC-HRMS chromatogram (Fig. S11c) further suggests that SPE exhibits high selectivity for polar compounds, potentially leading to the absence of less polar compounds such as (9Z,12E)-15,16-dihydroxyoctadeca-9,12-dienoic acid (m/z 311.227). Conversely, the NMR spectrum of LLE (Fig. S8d) yielded a richer profile closely resembling that of ACE (Fig. S8e). This similarity is further supported by the comparable profiles observed in the LC-HRMS chromatograms of LLE (Fig. S11d) and ACE (Fig. S11e). Finally, a similar observation was made with the pine honey extracts. The NMR spectrum of LLE after XAD-7 (Fig. S13b) showed a similar profile to ACE, although with lower intensity. The LC-HRMS chromatogram (Fig. S16b) revealed the absence of specific compounds like suberic acid (m/z 173.0816), pinobanksin (m/z 271.0610), and pinocembrin (m/z 255.0661). SPE did not seem to suit pine honey very well, as it resulted in a very poor spectrum (Fig. S13c), primarily displaying signals corresponding to a fatty acid, as well as α- and β-D-glucose. Despite this, the LC-HRMS chromatogram (Fig. S16c) appeared to have a similar overall profile to the other techniques, with the notable presence of more polar substances. Moreover, consistent with the findings for heather and flower honeys, the pine LLE extract (Fig. S13d) showed a very similar NMR spectrum to that obtained with ACE (Fig. S13e), which could also be observed in the respective LC-HRMS chromatograms (Figs. S16d,e). Fig. 6Base peak LC-ESI(−)-HRMS chromatograms of flower honey FH extracts obtained with different techniques tested: (a) XAD7, (b) LLE after XAD7, (c) SPE, (d) LLE, (e) ACE. Key metabolites are annotated: (1) gluconic acid, (2) 4-quinolone, (3) junipediol A 2’-O-β-D-glucopyranoside, (4) phenethyl rutinoside, (5) 3-phenyllactic acid, (6) 2-trans-decenedioic acid, (7) sebacic acid, (8) isorhamnetin, (9) abscisic aldehyde, (10) abscisic acid derivative and 11) spiculisporic acid. The elution region of high, medium and low polarity compounds is also highlighted. Base peak LC-ESI(−)-HRMS chromatograms of flower honey FH extracts obtained with different techniques tested: (a) XAD7, (b) LLE after XAD7, (c) SPE, (d) LLE, (e) ACE. Key metabolites are annotated: (1) gluconic acid, (2) 4-quinolone, (3) junipediol A 2’-O-β-D-glucopyranoside, (4) phenethyl rutinoside, (5) 3-phenyllactic acid, (6) 2-trans-decenedioic acid, (7) sebacic acid, (8) isorhamnetin, (9) abscisic aldehyde, (10) abscisic acid derivative and 11) spiculisporic acid. The elution region of high, medium and low polarity compounds is also highlighted. Since LLE and ACE exhibited similar profiles, it was decided to delve more into efficiency and selectivity aspects towards the different chemical classes of phenolics detected via LC-HRMS/MS analysis. For these reasons, based on one of our previous studies42, after classification into different chemical groups abundant in honey, the total peak areas of the annotated compounds were calculated and compared across samples. As clearly shown in Fig. 7, ACE exhibited the highest overall recovery and the broadest metabolite coverage across all analyzed honey samples, followed by SPE, the combined XAD7–LLE method and LLE, while XAD7 alone appeared significantly less effective. In all three analyzed honeys (Figs. 7 and S17), XAD-7 alone presented the lowest total peak area, with relatively minimal extraction of all metabolite classes. This could suggest limited affinity or retention for a broad range of honey-derived metabolites, highlighting the need for complementary clean-up steps. LLE treatment after XAD-7 demonstrated improved recovery across all classes found in the different samples, indicating that the clean-up step enhances metabolite recovery by reducing sugar interference or matrix effects. LLE, despite being an easy and simple technique, achieved a moderate total peak area along with a balanced coverage of compounds categories. However, it fell short in the concentration of recovered compounds compared to ACE. ACE yielded the highest total peak area by recovering a broader and more abundant range of compounds, including terpenoids, organic acids and fatty acids. This highlights ACE’s ability to extract diverse categories of compounds, including both polar and non-polar metabolites, underscoring its broad selectivity. The substantial recovery of terpenoids and derivatives is especially significant, as these compounds are often poorly extracted by conventional extraction methods. In this comparison, SPE followed as the second most efficient method in total peak area of annotated compounds for flower honey (fourth for heather and pine honey, Fig. S17), but not as well balanced compared to the liquid-liquid approaches (LLE, ACE). This imbalance could be observed also in heather and pine honeys (Fig. S17). The presence of a solid phase during the extraction could potentially increase the retention of compounds on it, affecting their actual concentration in the obtained extract, meaning that the observed profile, either with NMR or LC-HRMS, may not reflect as well the actual one found in each sample. This could extend to implications in the detection and determination of potential marker compounds for authenticity assessment, as we mentioned earlier with heather honey and the compound 4-methoxybenzoic acid. Fig. 7Total peak area distribution of annotated compounds categories in flower honey (FH) extracts obtained using different extraction techniques: XAD-7, LLE after XAD-7, SPE, LLE and ACE. Stacked bars represent the cumulative signal intensities of each. Total peak area distribution of annotated compounds categories in flower honey (FH) extracts obtained using different extraction techniques: XAD-7, LLE after XAD-7, SPE, LLE and ACE. Stacked bars represent the cumulative signal intensities of each. Evaluation of ACE’s repeatability One critical parameter in the extraction process is its repeatability, which is essential in analytical setups. This is particularly important in metabolomic approaches, whether for metabolite profiling or fingerprinting, both of which have become widely used in recent years for quality control purposes. Therefore, beyond developing an alternative and efficient extraction method for honey, we also assessed the repeatability of the ACE method to determine its compatibility with metabolomics studies. The idea behind this is to enable a dual-purpose approach with a single method, supporting both metabolomic workflows and fulfilling the requirements for the isolation and accurate identification of phenolics. Therefore, as mentioned already fifty honey samples (Table S10) were extracted with the ACE proposed method and analyzed by 1D NOESY NMR analysis. After data treatment, the PCA statistical method was employed to visualize the clustering of replicates and assess extraction variability. In, the resulting PCA scores scatter plot (Fig. 8), an extensive statistical space is covered due to the different metabolites, mainly phenolics, detected in the different honey samples due to variant origin. Most importantly, tight clustering of replicates of the same honey sample is observed in all cases indicating minimal variability and high repeatability of the ACE method. This results show that ACE is also well-suited for metabolite profiling studies. Furthermore, the repeatability of ACE was also supported by the LC-MS data. Specifically, as commonly employed in metabolomic studies and, in particular, LC-MS based methods validation, three peaks with different m/z value, polarity (Rt) and peak area were selected and evaluated across replicates43. The compounds were different for each honey sample due to their different composition. The RSD% values calculated (Table S11) were found < 20% and within acceptable analytical requirements44. Additionally, ACE’s ability to handle a large number of samples in an automated, high-throughput manner makes it particularly suitable for metabolite profiling and metabolomics approaches, regardless of the analytical method employed. Moreover, this technique is easily scalable for the isolation of specific compounds, expanding the potential of this method to diverse applications e.g. biological evaluation of minor compounds. On the other hand, the main disadvantage of ACE is the need for specialized equipment, as a certain apparatus is required, which can be costly. Additionally, its setup could often be more complex compared to conventional LLE methods. Specifically, like many other techniques, careful optimization of operational parameters is required to ensure repeatability. This optimization process might be time-consuming before the method is fully established for routine analysis. Fig. 8PCA scores scatter plot of honey samples extracted ACE. Repeatability of extraction is shown. PCA scores scatter plot of honey samples extracted ACE. Repeatability of extraction is shown. Conclusions In the current study, a novel method based on Annular Centrifugal Extraction (ACE) was employed for the first time in honey to simultaneously recover phenolics and remove carbohydrates. This method was specifically tailored to honey’s unique physicochemical properties by deliberately forming an emulsion in a biphasic solvent system. By applying ACE, phenolic compounds present in very low concentrations within complex mixtures were successfully recovered, allowing further analysis. The high extraction yield of honey phenolics, depleted of carbohydrates, enables accurate identification and addresses the challenge of numerous unknowns or only tentative identifications within honey phenolics. LC–MS and NMR analysis revealed the identification of several minor marker compounds, which are critical for quality control, especially considering the high incidence of honey fraud. Additionally, the proposed method is compatible with both analytical and preparative setups, as demonstrated by comparisons with commonly used honey extraction techniques such as solid-phase extraction (SPE), liquid-liquid extraction (LLE), and resin-based methods. This compatibility broadens its potential applications, allowing for parallel analyses such as metabolomics and metabolite profiling, alongside the isolation and biological evaluation of honey constituents. The latter is particularly significant given honey’s long history of use due to its health benefits. Overall, the ACE method offers several key advantages, including versatility, scalability, and suitability for a wide range of applications. The primary limitation of ACE is the cost of the equipment and the need for adaptation to different matrices. However, it is highly recommended for high-throughput, automated analysis of large sample sets, making it well-suited for metabolite profiling regardless of the analytical approach used. It should be emphasized that this method also holds significant potential for use on other materials or matrices, such as various food commodities or biological samples. Potential future validation studies, such as recovery experiments using spiked standards and inter-laboratory comparisons, could be conducted to further strengthen and confirm the robustness of this newly proposed technique. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1