3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT, ≥98%), acetone (≥99.9%), formic acid (≥99.9%), acetic acid (100%), dipotassium phosphate trihydrate (≥99.9%), glycerol (86%), monopotassium phosphate (≥99%), magnesium chloride (≥98.5%), sodium chloride (≥99.8%), monosodium phosphate monohydrate (≥98%), n-hexane (≥98%), sulfuric acid (96%, p. a.), decanoic acid (C10:0, >98%), octanoic acid (C8:0, >99%), oleic acid (C18:1, >99%), stearic acid (C18:0, >98%), sodium hydroxide (≥98%), 4-methylumbelliferyl-β-d-galactopyranoside (≥99%, for biochemistry), dimethylsulfoxide (≥99.8%), and molecular sieve 3 Å (0.3 nm, type 564, beads) were purchased from Carl Roth (Karlsruhe, Germany). Acetonitrile (≥99.9%), disodium phosphate (≥99%), sodium bicarbonate (≥99.7%), pancreatin from porcine pancreas (8 × USP specifications), bile extract porcine, dioleoylglycerol (diolein, >99%, mixture of 1,3- and 1,2-isomers), glyceryl trioleate (triolein, ≥99%), 1-oleoyl-rac-glycerol (monoolein, >99%; 2-monoolein was rarely available and six times more expensive), caffeine (>99%), linoleic acid (C18:2, 60%−74%), myristic acid (C14:0, >99%), palmitic acid (C16:0, >99%), dodecanoic acid (C12:0, 98%), peptone from casein (for microbiology), Müller–Hinton broth (for microbiology), d-(+)-glucose (99.5%), ampicillin sodium salt, and lysogeny broth (LB) powder (including 5 g/L of sodium chloride) were purchased from Fluka-Sigma-Aldrich (Steinheim, Germany). Methanol (99.9%) was supplied by VWR International (Darmstadt, Germany). Magnesium sulfate heptahydrate (99.5%), potassium chloride (≥99.5%), citric acid monohydrate (≥99.5%), HPTLC plates silica gel 60 RP-18 W, HPTLC plates silica gel 60 RP-18, and HPTLC plates silica gel 60 as cover plates (all 20 cm × 10 cm), and Bacillus subtilis subsp. spizizenii spore suspension (DSM 618) were purchased from Merck (Darmstadt, Germany). Diammonium phosphate (≥99%), diethyl ether (≥99%), and linolenic acid (C18:3, 99%) were purchased from Acros Organics (Morris Plains, NJ, USA). Yeast extract powder (for microbiology), ethyl acetate (≥99.8%), o-phosphoric acid (85%), ethanol (≥99.9%), and dichloromethane (≥99.9%, stabilized with amylene) were purchased from Th. Geyer (Renningen, Germany). Copper(II) sulfate pentahydrate (p. a.) was purchased from Honeywell International (Morristown, NJ, USA). Calcium chloride dihydrate (≥99.9%) was supplied by Bernd Kraft (Duisburg, Germany). The luminescent marine Aliivibrio fischeri (DSM 7151) bacteria were purchased from the DSMZ Leibnitz Institut (Berlin, Germany). Tetracycline hydrochloride (research grade, USP) was purchased from Serva Electrophoresis (Heidelberg, Germany). Bidistilled water was produced by a Heraeus Destamat B-18E (Thermo Fisher Scientific, Dreieich, Germany). Rhodamine 6G (100% ± 3%) was purchased from Alfa Aesar (Kandel, Germany). Salmonella typhimurium bacteria strain TA1535, modified to contain the plasmid pSK1002 (PTM S. typhimurium TA1535/pSK1002, cryostock), was purchased from Trinova Biochem (Giessen, Germany). 4-Nitroquinoline-1-oxide (98%) was purchased from TCI (Eschborn, Germany). Edible vegetable oils (Supplementary Table S1) were purchased from local supermarkets.

For on-surface metabolic reactions, the HPTLC RP-18 W plate was pre-treated as follows. The plate was heated at 120°C for 60 min (TLC Plate Heater III, CAMAG, Muttenz, Switzerland; to fix the binder for the current plate batches used) and pre-washed by development first with methanol and then, after plate drying, with ethyl acetate, both up to 90 mm in a twin-trough chamber. To ensure the pancreatin reaction in the application zone, the acidic plate pH (ca. pH 4.2) was neutralized via piezoelectrical spraying (2.8 ml, ultra-yellow nozzle, level 3, Derivatizer, CAMAG) with phosphate-citrate buffer (6 g/L of citric acid and 10 g/L of disodium hydrogen phosphate, adjusted to pH 12 by sodium hydroxide). Therefore, except for the application zone, the plate was covered by a cut HPTLC plate silica gel 60, with the layer facing upward (Supplementary Figure S1). Then, the plates were dried at 120°C for 10 min.

The digestion fluid stock solution was prepared as described by Minekus et al. (27). The pancreatin solution (200 TAME mU/μl) and the corresponding calcium chloride solution (6 pmol/μl) were prepared according to Morlock et al. (28). Monoolein, diolein, triolein, fatty acids (reference compounds were applied via overspraying to obtain the mixture on the start zone), and samples were weighed via a pipette and dissolved in n-hexane (all 1 mg/ml each), whereby solid fats (i.e., palm oil and coconut oil) were slightly warmed to the melting point before pipetting. All solutions were stored in solvent-tight vials in the dark at 4°C.

Oil sample and FA standard solutions (10 μl/band each; 1 mg/ml) were applied as 8-mm bands, unless stated otherwise, as follows: a track distance of 10 mm, distance from the lower edge of 10 mm and left edge of 10 mm, dosage speed of 150 nl/s, filling speed of 15 μl/s, filling vacuum time of 0 s, and syringe volume of 25 μl (Automatic TLC Sampler ATS 4, CAMAG). The plate was developed with dichloromethane/acetic acid/acetone 2:4:5 (V/V/V) (32) in a twin-trough chamber (20 cm × 10 cm) up to 80 mm. The plate was subjected to a derivatization reagent sequence performed via dipping (immersion time 8 s, immersion speed 3 cm/s, Chromatogram Immersion Device 3, CAMAG), i.e., first in rhodamine 6G reagent solution (0.1% in ethanol) and, after plate drying and detection at FLD 366 nm (TLC Visualizer 2, CAMAG), then in copper sulfate phosphoric acid reagent (25 g copper sulfate pentahydrate in 250 ml o-phosphoric acid/water 4:41, V/V), followed by heating at 150°C for 20 min (TLC Plate Heater III, CAMAG) and detection at FLD 366 nm and white light illumination.

The following workflow was adapted from the nanoGIT^+active methodology (28, 33). The application was performed as mentioned, except for a band length of 6 mm, a track distance of 9 mm, and a distance from the left edge of 14.5 mm. Reference standard mixtures were applied via overspraying. Since the previous solutions were in n-hexane, the cleaning unit of the ATS 4 had to be rinsed with bi-distilled water once and the syringe twice before pancreatin application. The applied sample bands were first oversprayed with pancreatin solution (5 μl/band, 200 TAME mU/μl) and then with calcium chloride solution (1 μl/band, 6 pmol/μl) using a dosage speed of 50 nl/s and a filling speed of 8 μl/s. The application zone of the plate was wetted with 2.5 ml of 0.1 M sodium chloride solution by piezoelectrical spraying (yellow nozzle, level 6, Derivatizer), whereby the plate area for chromatographic separation was covered with a cut HPTLC plate silica gel 60 (Supplementary Figure S2A) (33) to avoid the salt load on this adsorbent area and ensure good zone resolution during the later separation. This plate package was incubated at 37°C in a humid plastic box (26.5 × 16 × 10 cm, ABM, Wolframs-Eschenbach, Germany) for 60 min. After plate drying (120°C, 10 min), the lipids were focused twice by front elution with acetone up to 25 mm in a twin-trough chamber (10 min before, the second trough of the twin-trough chamber was filled up to half with molecular sieve 0.3 nm), and the lower part of the plate was cut off at 15 mm (Supplementary Figure S3) to reduce the influence of the pancreatin matrix on the chromatographic separation. The plate was developed from the cut edge side with either n-hexane/diethyl ether/formic acid 90:25:2 (V/V/V) (34) up to 60 mm for the separation of acylglycerols or acetonitrile/water 4:1 (V/V) up to 50 mm for the separation of FAs in the twin-trough chamber filled with molecular sieve as mentioned. Chromatograms were derivatized via the reagent sequence, whereby the cut 15-mm plate strip was stuck together with adhesive tape on the glass side of the HPTLC plate and detected as mentioned.

As mentioned in the previous subsection, one oil sample per plate (10 cm × 10 cm) was applied as a 3-mm band, except for setting the distance from the left edge to 9 mm and using cover plates that covered everything but the applied sample band (Supplementary Figures S1B, S2B). The applied sample was treated with pancreatin and calcium chloride, wetted, incubated, focused (but no plate cut), developed two-dimensionally, detected via a bioassay as follows, or derivatized optionally via the reagent sequence, and detected as mentioned. For 2D development, the apolar mobile phase for acylglycerol separation was chosen as the first dimension, and the polar mobile phase for FAs separation as the second dimension. In between, the plate was dried at 120°C (TLC Plate Heater III) for 10 min and rotated by 90° (Supplementary Figure S4). Before the bioassay application, the plate was freed from residuals of the mobile phase via heating at 120°C for 10 min (TLC Plate Heater III) and neutralization with 2.5% sodium bicarbonate solution (2.8 ml, yellow nozzle, level 3, Derivatizer) followed by drying at 120°C for 10 min.

The European Pharmacopeia Chapter 2.3.2 method on HPTLC RP-18 plates (32) was tested first and extended to include FAs. Differently, the plate was derivatized with the rhodamine 6G reagent. The resulting FLD 366 nm chromatogram showed bands, although weak, only for FAs but not for TAGs (Supplementary Figure S5). Therefore, by exploiting the reagent sequence technique, the plate was subsequently derivatized with copper sulfate phosphoric acid reagent, and in the study, both TAG and FA zones appeared in the HPTLC RP-18 chromatogram at FLD 366 nm (Figure 1), but no charring reaction could be observed on the HPTLC RP-18 plate at white light illumination as intended for this reagent. The currently revealed fluorescence of the rhodamine 6G reagent was explained by the pH dependence of the rhodamine 6G fluorescence and the needed acidic pH for proper visualization, here provided via the copper sulfate phosphoric acid reagent (36).

Next, the plate type had to be changed from RP-18 to a wettable reversed-phase (RP-18 W), and thus, the mobile phase system had to be changed as well since the desired aim for this study was on-surface digestion via the nanoGIT method (28). Before on-surface digestion, the intrinsic acidic pH of the RP-18 W plate (ca. pH 4.2) needed to be neutralized in the application zone with a phosphate-citrate buffer of pH 12. Unfortunately, the buffer salts interfered with derivatization via the rhodamine 6G reagent (Supplementary Figure S6); therefore, everything except the application zone had to be covered as narrowly as possible (Supplementary Figure S1). Unification of the wetting and neutralization processes (after the application) to only one neutralizing wetting step is recommended, as we established in another study. The pancreatin matrix interfered during development by causing a retardation shift in contrast to the reference standards. Thus, it had to be removed by focusing the lipids twice by front-elution with acetone after the lipolysis and cutting off the lower plate part containing the remaining pancreatin matrix. Since the focusing result with acetone was strongly dependent on the relative humidity of the laboratory environment, the second trough was half-filled with a molecular sieve of 0.3 nm within 10 min prior to focusing. When a dry environment (< 15% relative humidity) in the twin-trough chamber was reached, acetone was filled in the opposite trough, and the plate was placed inside as fast as possible for development. The dry conditions during focusing as well as the (second) polar mobile phase development showed reproducibly good zone resolutions.

Due to the amphiphilic properties of the RP-18 W phase (apolar C18 chains and residual silanol groups), two orthogonal mobile phase systems were developed, i.e., one apolar to separate acylglycerols and one polar to separate FAs. Both resulting nanoGIT–HPTLC RP-18 W chromatograms (Figures 2A, C) proved the orthogonality of the mobile phases. Surprisingly, the derivatization with rhodamine 6G directly showed all lipophilic compounds. This variation in the fluorescence response was explained by different initial plate pHs and proven by additional experiments, in which the rhodamine 6G fluorescence showed a strong plate batch dependence due to different plate pHs. Fortunately, the current plate batch pH supported the required acidic milieu for the rhodamine 6G fluorescence (41). To exploit a reagent sequence, the subsequent derivatization of the same plate with copper sulfate phosphoric acid reagent (Figures 2B, D) led to a charring reaction of all unsaturated lipophilic compounds detected as black zones. The combination of both derivatization reagents on the same plate surface made it possible to first visualize all lipophilic compounds and then differentiate saturated and unsaturated zones.

As observed for the standard mixture (Figures 2A, B, St), the separation of acylglycerols according to polarity in TAGs, DAGs, and MAGs was successful, but all the reference FAs were eluting as one unseparated diffuse zone (hR_F 37–57). This system also allowed the separation of both DAG isomers (hR_F 19 and 26) and TAG isomers, as observed for flaxseed oil (F, hR_F 80 and 86). Comparing the side-by-side separated undigested and digested (⁺) samples, a massive increase in the FAs and decrease in the TAGs amount/signal was observed, indicating the successful simulation of the lipolysis. The undigested samples showed two bands at the DAG zone; however, after the lipolysis, only one band remained, which was caused by the pancreatin enzyme blank (EB). Two further interferences were observed from the pancreatin blank, one intense zone was just below the MAGs, and another weaker zone was at the FAs position (assigned to triterpenoid acids as discussed later), which complicated the evaluation of those in the sample. Nevertheless, the comparison of both derivatization reagents supported the literature-known oil composition of all samples. Oils with a variety of FAs, such as flaxseed, hemp, and coconut oils, showed broader TAG and FA zones, whereas olive oil (mainly containing C18:1) showed comparatively compact zones. Concentrating on the unsaturated FAs in the copper sulfate phosphoric acid chromatogram (Figure 2B), the most intense zones for flaxseed oil (mainly PUFAs) and, in contrast, almost no zones for coconut oil (containing comparatively much more SFAs) confirmed the oil compositions as well.

The orthogonal selectivity selected for the separation of FAs according to lipophilicity showed a successful qualitative separation of nearly all FA reference standards (Figures 2C, D, St). In this system, TAGs and DAGs were retained at the application zone, whereas the MAGs were eluted. Due to their similar lipophilicity, C18:1, C16:0, and MAGs co-eluted as well as C18:2 and C14:0, which was proven in another experiment (Figure 3, framed). In the nanoGIT–HPTLC RP-18 W chromatogram (Figures 2C, D), the undigested samples did not show any noticeable FA zones, but the digested samples did. Thus, the lipolysis of TAGs into FAs was successful. High FA amounts as in the reference track (Figures 2C, D, St) and pancreatin matrix effects on the sample tracks led to a retardation shift; thus, the zone matching between the samples and reference compounds was challenging but nevertheless possible. With the aid of the copper sulfate phosphoric acid reagent chromatogram, C18:2 and C18:3 could be identified as intense black zones. The rhodamine 6G reagent chromatogram helped identify C8:0, C10:0, and C12:0. Due to this assigned pattern of the FAs and the literature data (37, 38), the FAs in the samples could be identified successfully via pattern recognition based on their main components. Intense zones for C18:3 (zone d) in flaxseed oil, C18:2 (zone c) in hemp, walnut, soybean, and sunflower oils, and C14:0 (zone e) in coconut oil were determined after digestion via pancreatin (Figures 2C, D).

As mentioned for the separation of acylglycerols, polar impurities of the pancreatin co-eluting with the FAs hindered their assignment and could also lead to false-positive interpretations. Using automated heart-cut elution of the interesting zones to RP-HPLC–DAD–HESI-HRMS/MS (39), these impurities were assigned as the bile acids ursodeoxycholic acid (UDCA), hyodeoxycholic acid (HDCA), (cheno)deoxycholic acid (CDCA/DCA), and cholenic acid (Figure 4 and Supplementary Table S2). The isomers UDCA, HDCA, and (C)DCA were identified in the negative ionization mode (HESI⁻) in two separate peaks at retention times of 8.11 min and 8.45 min with [M–H]⁻ at m/z 391.2858 and 391.2860, respectively. The HESI⁻ and respective positive ionization mode (HESI⁺) revealed the presence of their dimers ([2M–H]⁻ at m/z 783.5791 and [2M+H]⁺ at m/z 785.5927, Figure 4), identified via fragmentation (Supplementary Figure S7), as well as their tetramer ([4M+2H]⁺ at m/z 785.5909 with its corresponding isotopic pattern, Supplementary Figure S8). Cholenic acid could only be assigned via HESI⁺ as [M+H]⁺ at m/z 375.2885. Since bile acids also show antibacterial properties (40), false-positive results should be considered and, when necessary, confirmed/excluded via HRMS. Fortunately, the following comprehensive separation system solved this coelution issue.

The one-dimensional separation systems showed some limits, such as interferences with the pancreatin used and no separation of acylglycerols and FAs at once, which could not be overcome by method optimization or modification. Hence, the combination of both orthogonal separation systems into a comprehensive HPTLC method (HPTLC×HPTLC) was evaluated (Figure 5). The normal phase separation mechanism separating according to polarity (apolar acylglycerol-separating mobile phase) was chosen to be the first dimension, whereas the reversed-phase separation mechanism separating according to lipophilicity (polar FA-separating mobile phase) was selected as the second dimension. The orthogonality was given by the very different selectivity of the first dimension in contrast to the second dimension. The orthogonal separation was first tested with reference standards (Figure 6A), and successful separation of acylglycerols and FAs could be achieved, in particular the separation of the previously co-eluted MAGs and FAs. The zone assignment of the FAs in the nanoGIT–HPTLC×HPTLC–FLD chromatogram was more difficult than via one-dimensional separation. Using the co-development of reference standards for each dimension on a separate plate (Supplementary Figure S4), a retardation shift could be observed for the FA zones. Since both mobile phases were prone to changes in relative humidity, the co-development of reference standards on the same plate was recommended to verify a retardation shift and proper assignment. The dominance of the C18:2 and C18:3 fatty acids (zones c and d, respectively) was also helpful for proper assignment.

Next, the on-surface digestion of an oil sample and the subsequent lipid separation of its lipolysis products on the same surface were demonstrated (Figure 6B). Therefore, flaxseed oil rich in C18:3 and C18:2 was selected for the proof of principle. A retardation shift was observed between reference standards (Figure 6A) and samples (Figure 6B). Using automated heart cuts to RP-HPLC–DAD–HESI-HRMS/MS (39), the highest eluting FAs (Figure 6B, zone d) were identified as C18:3, C18:2, C16:0, and C18:1 (Supplementary Figure S9, zone d, Supplementary Table S3). The most intense signal for this zone was from C18:3. Since it was stamped perpendicular to the band due to an accidentally 90°-rotated plate (positioned incorrectly) in the autoTLC interface (42), several FA signals were derived from and assigned to the neighboring bands. Some additional FAs were identified that could not be associated with the flaxseed oil sample: in zones c and d, oxidized C9:0 and oxidized C12:1 were identified, which could be explained as degradation products of linoleic acid and linolenic acid, respectively. In zone d, C14:0, and zone e, C10:0, C11:0, and C12:0 were found. No fragmentation pattern was evident via HRMS/MS recording. The previously interfering pancreatin matrix was presently successfully separated from the FAs since most pancreatin impurities did not elute in the first dimension (in contrast, the FAs did) but did elute first in the second dimension. Thus, the FAs could be identified easily, in contrast to the one-dimensional separation of the reference standards (Figure 6B). Additionally, the eluted FA zones were fully separated from the bile acids, and their mass signals were not detected in the HRMS spectra anymore. By doing so, the nanoGIT–HPTLC×HPTLC–FLD method was proven to be successful in its application to real-life samples and in the detailed study of the lipolysis of complex samples. The whole sample was studied in all aspects on the same surface, and no sample part was lost.

After a successful proof of principle and implementation of the nanoGIT–HPTLC×HPTLC–FLD hyphenation, it was extended to bioassays to evaluate the antibacterial activity of the lipolysis products of digested flaxseed oil (Figures 7A, B) and coconut oil (Figure 7C) against Gram-negative A. fischeri and Gram-positive B. subtilis bacteria. The A. fischeri bioautogram revealed antibacterial effects for all seven FA reference standard zones as well as for the MAG, DAG (both isomers of diolein), and TAG reference standard zones (Figure 7A). In the B. subtilis bioautogram, the antibacterial detection was comparable, apart from the weaker response for two FA reference zones (d and g, Figure 7B), which was proven and confirmed in another experiment (Figure 5). These findings of antibacterial activity were consistent with the literature, which confirmed the antibacterial effect of FAs and MAGs (8, 41, 43, 44) and DAGs (9) against Gram-negative and Gram-positive bacteria. Usually, no antibacterial effect for TAGs (three ester groups but optionally double bonds) would be expected due to the lack of reactive functional groups (41, 45). However, the studied TAG, DAG, and MAG had one double bond in each acyl chain, which could induce a genotoxic or cytotoxic effect, as discussed later.

A separate study of all reference standards (Figure 3) showed in more detail the differences in their antibacterial effects against both Gram-negative and Gram-positive bacteria. The FAs C18:0 and C16:0 showed only a very light antibacterial response, whereas a metabolism-enhancing effect (detected as a halo surrounding an antibacterial effect in the center) was detected for C14:0. If co-eluted with C18:2 as in the standard track, this enhancing effect was weakened since the antibacterial effect of C18:2 was stronger (Figure 7, zone c). Compared to previous bioautograms on HPTLC plates silica gel 60 (39, 46), C16:0 showed no metabolism-enhancing effect on RP-18 W plates, which was explained by the doubled amount (10 μg/band vs. 5 μg/band) since such enhancing effects are dose-dependent and, in addition, also time-dependent (bioluminescence images monitored for 30 min). The antibacterial response for C8:0–12:0 was very intense against both Gram-negative A. fischeri and Gram-positive B. subtilis bacteria (Figure 3). In the A. fischeri bioautogram, a strong antibacterial effect of unsaturated FAs (C18:1–C18:3) against A. fischeri was observed, whereas an increase in the antibacterial effect with increasing double bonds was not observed. In the B. subtilis bioautogram, the antibacterial effect of unsaturated FAs against Gram-positive bacteria was weaker, which was directly evident since the same reference standard amounts were applied. Further research is needed to understand the mechanism of the observed biological responses of the FAs and acylglycerols. On the one hand, the biological response may derive from the acid head group and/or altered membrane permeability and thus be an antibacterial effect (as one example of the many different antibacterial mechanisms). On the other hand, the biological effect may derive from trace impurities (e.g., co-eluting epoxidized longer-chain fatty acids, Figure 3) in the reference standards (only up to 99% pure) or oxidative degradation and thus be a cytotoxic effect. Unfortunately, the separation power of HPTLC is too weak to chromatographically differentiate all of them. Nevertheless, the powerful hyphenation with the bioassay provides the first evidence of harmful compounds present.

Using automated heart cuts of the interesting zones to RP-HPLC–DAD–HESI-HRMS/MS, the Gram-positive antibacterial zones of the reference standard track (Figure 7B) could be identified as the corresponding FAs (Figure 8, Supplementary Table S4). The assignments for zones a (C18:0) and b (C16:0/C18:1) were reached through pattern recognition. The latter assignments were more challenging since these FAs can also be HRMS system signals, which must be excluded first. Zones c and e were the most intense, containing co-eluting C14:0/C18:2 and C12:0/C18:3 (from adjacent zone d), respectively. Zone d (C18:3) was too close to the surrounding zones for an elution head-based analysis (too narrow for an additional elution head imprint). The C8:0 (zone g) was not separated in the 2D bioautogram but co-eluted with the C10:0 (zone f).

Considering the information obtained about the antibacterial behavior of the reference standards, the assignment of the lipolysis products of flaxseed oil was possible. Flaxseed oil, which mainly consists of C18:3, C18:1, and C18:2 and small amounts of C16:0 and C18:0 (38), showed after on-surface digestion and effect-directed analysis of four zones in the 2D A. fischeri bioautogram but only one zone in the 2D B. subtilis bioautogram (Figures 7A, B). The four zones in the 2D A. fischeri bioautogram were identified as C18:0 (zone a), C18:1/C16:0 (zone b), C18:2 (zone c), and C18:3 (zone d) in comparison to corresponding reference standards. No metabolism-enhancing effect was detected, and thus, the presence of C14:0 was excluded, which could have co-eluted with C18:2. The antibacterial zone d in the B. subtilis bioautogram (Figure 7B) was not so clear in the assignment, and thus identified via HRMS as C18:3 ([M–H]⁻, m/z 277.2175, Δ ppm −0.71) and trans-4,5-epoxy-(E)-2-decenal (C₁₀H₁₆O₂, [M–H]⁻, m/z 167.1078, Δ ppm −0.28). The latter is a typical genotoxic marker of linoleic acid oxidation (47–49).

As a further example, coconut oil was digested on-surface and analyzed for any antibacterial effects (Figure 7B). Coconut oil was selected as a quite different oil sample compared to flaxseed oil since it consists of comparatively much more SFAs of shorter chain lengths (C8:0–14:0, mostly C12:0 and C14:0) (37, 45). In contrast to the flaxseed oil (one antibacterial zone), the 2D B. subtilis bioautogram showed five antibacterial zones. Using the standard track, the FAs were identified as C18:0 (zone a), C18:1/C16:0 (zone b), C18:2/C14:0 (zone c), C12:0 (zone e), and C10:0 (zone f); however, C18:3 (zone d) was not present. The zone was heart-cut eluted to RP-HPLC–DAD–HESI-HRMS/MS (Figure 9, Supplementary Table S5) and revealed C14:0, C12:0, and C10:0 signals for the zones c, e, and f, respectively, but no significant signals for C18:0 and C18:1/C16:0. The A. fischeri bioautogram of the on-surface digested coconut oil (Figure 7C, Co⁺) was used to confirm zone c to be C14:0 via its metabolism-enhancing effect as further proof.

On-surface digested flaxseed oil revealed four genotoxic substance zones in the 2D bioautogram after the genotoxicity bioassay using the S. typhimurium TA1535/pSK1002 strain (Figure 10A). Two genotoxic substance zones did not migrate/elute at all in the second dimension, indicating apolar substances. One zone was assigned as TAGs via comparison with the standard track, and the second more apolar compound (marked^* close to the solvent front of the first dimension) could be a genotoxic aromatic mineral oil contaminant; however, the latter assumption still needs proof. The genotoxic effect of TAGs was explained by the epoxidized fatty acid bond in the TAG molecule. Only two weak signals for the FAs were detected in the flaxseed oil and reference standard mixture (second dimension), which were assigned to C18:2/C14:0 and C18:3/C12:0 or C10:0. The digestion of the TAGs did not eliminate genotoxicity but showed that the FAs produced have strongly different genotoxic potentials. Both FAs were not natively fluorescent, which was expected (Figure 10B); however, native blue fluorescence was observed for the genotoxic TAGs zone of flaxseed oil, which indicated any impurities, e.g., of aromatic structure, as mentioned.

For adequate signal intensity via the genotoxicity bioassay, the amount of flaxseed oil was doubled (20 μg/band). In contrast to our previous very sensitive screening method (10), the amount of oil needed to be increased 200-fold due to the (I) enzymatic metabolization with a 60-min on-surface incubation known to lead to diffusion at the application zone (33), (II) interference by the buffer salts (Supplementary Figure S6), (III) 2D separation known to cause signal loss (50) by the 2-fold diffusion of the substances (as for C18:0, Figure 3), (IV) usage of RP-18 W plates known to be possibly less sensitive in the response, though dependent on the molecule, compared to silica gel 60 plates (51, 52), and (V) purchased oils opened just before analysis (assumedly, comparatively fewer oxidized degradation products). These reasons also explained why HRMS analysis was challenging since oxidized species present at the trace level were not found. In contrast to Morlock and Meyer (19), in which a six-fold concentrated genotoxic compound zone was directly transferred to the HRMS, only one weaker genotoxic compound zone was eluted from the 2D bioautogram, passed through an HPLC column via a prior desalting unit and diode-array detector, and finally reached the HRMS. The presence of highly potent genotoxic FA in oxidized and epoxidized forms at the trace level in various plant oils (10, 49, 53), and its potential sources, such as the unsaturated FAs C18:2 (54) and C18:3 (54, 55), were already reported. If safely delivered to a healthy liver, detoxification may be expected, as was recently shown via simulated on-surface S9 liver metabolization (10, 56). Furthermore, synergistic effects can occur (57), which can be detected and studied via the latest multiplex planar assays (51).

The advantages and disadvantages of this quite new methodology against reported conventional methods (Supplementary Table S6) (58, 59), including the further ability of an effect-directed analysis after separation, strongly highlight the ability to illuminate every facet of the sample.