Structural Elucidation of cis/trans Dicaffeoylquinic Acid Photoisomerization Using Ion Mobility Spectrometry-Mass Spectrometry

Abstract

Due to the recently uncovered health benefits and anti-HIV activities of dicaffeoylquinic acids (diCQAs), understanding their structures and functions is of great interest for drug discovery efforts. DiCQAs are analytically challenging to identify and quantify since they commonly exist as a diverse mixture of positional and geometric (cis/trans) isomers. In this work, we utilized ion mobility spectrometry coupled with mass spectrometry to separate the various isomers before and after UV irradiation. The experimental collision cross sections were then compared with theoretical structures to differentiate and identify the diCQA isomers. Our analyses found that naturally the diCQAs existed predominantly as trans/trans isomers, but after 3 h of UV irradiation, cis/cis, cis/trans, trans/cis, and trans/trans isomers were all present in the mixture. This is the first report of successful differentiation of cis/trans diCQA isomers individually, which shows the great promise of IMS coupled with theoretical calculations for determining the structure and activity relationships of different isomers in drug discovery studies.

Graphical Abstract

Chlorogenic acids (CGAs) are polyphenolic natural products consisting of esters of quinic acid and hydroxycinnamic acids (i.e. caffeic-, ferulic-, sinapic- or coumaric acid). CGAs are found in a wide variety of food and beverage plants and have been linked to reducing obesity and influenza, and preventing cataractogenesis,^1–3 therefore making them of great interest for drug discovery efforts. Further, caffeoylquinic acids (CQAs) and dicaffeoylquinic acids (diCQAs), which are two main classes of CGAs, also have reported antioxidant, anti-HIV, antihypertensive, and anti-inflammatory health benefits.^4–11 Analyzing CGAs, however, can be very difficult as they normally exist as complex mixtures with various positional and geometric (cis/trans) isomers.^12,13 Additionally, exposure to UV light further complicates their analyses as it has been shown to change the trans double bond orientations to cis, and when multiple double bonds exist, cis/cis and combinations of cis/trans isomers are very common.¹⁴ Thus, the structure and activity relationship for these cis/trans isomers has not been well understood to date, but the bioactivities of CGAs are known to be affected by the cis/trans isomerism.^15,16

As mentioned, diCQAs exhibit anti-HIV-1 activity by binding irreversibly to the HIV-1 DNA integrase (a key enzyme for viral HIV DNA integration) and inhibiting HIV-1 DNA replication.^7–9 Related hydroxycinnamic acid derivatives such as L-chicoric acid (dicaffeoyltartaric acid) also possess the same anti-HIV activities, but both the diCQAs and dicaffeoyltartaric acids activities are highly dependent on the structural isomers present.^16–18 Structural isomerization analyses for other natural or synthetic drugs have also shown either agonistic or antagonistic effects. For example, an isomeric mixture of conjugated linoleic acids has been shown to possess enhanced biological activities when compared to the pure components,¹⁹ indicating biological synergism of the isomers. Also, the unnatural cis isomers of potent proteasome inhibitors have exhibited stronger inhibitory activity than the trans counterparts.²⁰ Thus, understanding the structures and activities of CGAs is essential for their use in drug studies.

The identification and quantification of phytochemicals in plant extracts is challenging due to their structural diversity; therefore, reliable analytical methods capable of differentiating related structures are needed.²¹ Although great efforts have been put forth, commonly used analytical platforms have shown significant deficiencies with these separations. For instance, LC-MSⁿ approaches developed to discriminate positional isomers of CGAs provide limited differentiation of geometric isomers.^22,23 Geometric CGA isomers in previous studies were determined through chromatographic elution orders; however, this can be problematic depending on the chromatographic parameters and their reproducibility.¹⁴ Recently it was reported that cis geometrical isomers of diCQAs preferentially bind to alkali metal ions and can be discriminated from the trans isomers using LC-MS.²⁴ However, this method is only capable of discriminating the group of cis isomers from the group of trans isomers, but not identifying individual isomers.

With recent technology advancements, ion mobility spectrometry-mass spectrometry (IMS-MS) has become an appealing tool for the separation and structural characterization of biomolecules.^25,26 IMS-MS has also been used in the characterization of positional and geometrical isomers of small molecules^27,28 and those due to photoisomerization.^29,30 Recently, IMS was applied to separate mono- and diCQAs and differentiate their positional isomers,^31,32 but unfortunately the cis/trans isomers were not separated. Here we utilized drift tube ion mobility spectrometry-mass spectrometry (DTIMS-MS) to elucidate the cis/trans isomer structures of diCQAs generated by UV irradiation-induced photoisomerization. Furthermore, a newly developed structures for lossless ion manipulations (SLIM) platform was used for ultrahigh-resolution IMS separations of the cis/trans isomers. SLIM IMS uses traveling waves and a compact serpentine drift path for high ion mobility resolution. A switch can also be utilized in these devices to perform multiple passes in the drift path for even higher IMS resolution.^33–38 The work detailed in this manuscript allowed the structure for each cis/trans isomer to be unambiguously identified by comparing the experimental collision cross section (CCS) values with theoretical CCS values obtained from molecular modeling. This comparison enabled the differentiation and identification of isomers from complex natural bioactive phytochemicals, enabling an understanding of their bioactivity.

In this study, first we characterized the positional isomers for the five diCQA standards (1,3-diCQA, 1,5-diCQA, 3,4-diCQA, 3,5-diCQA, and 4,5-diCQA; Figure 1A) using a 90 cm DTIMS-MS platform (Agilent 6560 IMS-QTOF MS, Santa Clara, California, USA).^39,40 The DTIMS profiles for these diCQA isomers ([M + Na]⁺, m/z = 539.12) are shown in Figure 1B. Prior to UV irradiation, each diCQA isomer displayed a single IMS peak, consistent with the fact that these compounds naturally exist as trans/trans isomers. The arrival time distributions (ATDs) for these positional isomers were similar; however, 4,5-diCQA eluted first (indicating a smaller structure) and was baseline separated, while 1,5-diCQA had the longest arrival time and the largest structure. The CCS value for each diCQA isomer was measured by DTIMS and is given in Table 1. To better differentiate the positional isomers, the SLIM IMS- MS platform was then used to obtain ultrahigh-resolution IMS separations. As shown in Figure 1C, after five passes of the SLIM IMS-MS platform (67.5 m drift length) 4,5-diCQA, 3,4-diCQA, and 1,5-diCQA were well separated, and 1,3-diCQA and 3,5-diCQA were partially separated. Since the SLIM IMS-MS platform is based upon traveling waves, direct CCS measurement is not possible without calibration, so only the CCS values from DTIMS are noted in Table 1.

Figure 1.

IMS characterization of the dicaffeoylquinic acid positional isomers 1,3-diCQA, 1,5-diCQA, 3,4-diCQA, 3,5-diCQA, and 4,5-diCQA. (A) The chemical structures of each isomer and arrival time distributions (ATDs) using a (B) DTIMS platform and (C) ultrahigh-resolution SLIM IMS.

Table 1.

Experimental Collision Cross Sections Measured by DTIMS in Nitrogen Gas (^DTCCS_N2, in Ų) for Each diCQA Positional Isomer^a

	experimental DTCCSN2 (Å2)	experimental CV (%)
1,3-diCQA	250	0.5
1,5-diCQA	252	0.3
3,4-diCQA	249	0.1
3,5-diCQA	251	0.5
4,5-diCQA	236	0.1

^aThe coefficient of variance (CV) determined by triplicate measurements are provided.

Among these positional isomers, the most abundant form and potent HIV-1 inhibitor 3,5-diCQA was selected to establish an analytical method for cis/trans isomer separation since four possible geometric isomers exist: 3-trans,5-trans-diCQA isomer, 3-cis,5-trans-diCQA, 3-trans,5-cis-diCQA, and 3-cis,5-cis-diCQA. Previous molecular docking studies showed that these isomers have different binding activities with the HIV-1 INT enzyme, a key enzyme for viral HIV DNA integration.⁴¹ While these isomers all bind to the catalytic domain of HIV-1 INT enzyme, the cis isomers were found to bind to the metal cofactor of HIV-1 INT, which is related to its antiviral activity. Moreover, 3-trans,5-cis-diCQA interacted with both the LYS156 and LYS159 residues that are significant for viral DNA integration. These docking results also illustrated that different binding activities between the 3,5-diCQA isomers with HIV-1 INT enzyme are synergistic and provide wider inhibition activity than a single isomer. Therefore, understanding the structures of the geometric isomers is important for elucidating their bioactivities.

Upon UV irradiation at 245 nm for 3 h, the photo-isomerization products of 3,5-diCQA were examined by DTIMS-MS (Figure 2). Without UV irradiation, 3,5-diCQA mainly illustrates a single feature which corresponds to the 3-trans,5-trans isomer (Figure 2A). An additional peak at a shorter arrival time however was also observed, indicating a small fraction of 3,5-diCQA exists in other conformations prior to UV irradiation. After 3 h of UV irradiation, several additional features arose with shorter arrival times (Figure 2B). The feature with the shortest arrival time (left) displayed a very narrow distribution similar to the feature with the longest arrival time (right). The middle feature however was twice as broad as the other features and could be fitted with two features, indicating the presence of two conformations with similar abundances.

Figure 2.

ATDs for the 3,5-diCQAs (A) before UV irradiation and (B) after 3 h UV irradiation using the DTIMS-MS, and (C) the SLIM IMS-MS platforms. The dashed lines in panel B represent the IMS peak shapes expected for single structures.

To better differentiate the 3,5-diCQA cis/trans isomers, SLIM IMS-MS measurements were performed for the conformers resulting from UV irradiation. As shown in Figure 2C with only a 14.7 m SLIM IMS separation, the middle feature that was unresolved in Figure 2B is clearly separated into two features, confirming that there are four isomers for 3,5-diCQA after photoisomerization. To assign these cis/trans isomers, theoretical modeling was performed for each isomer allowing the calculation of theoretical CCS values (Figure 3 and Table 2). By comparing the experimental and theoretical CCS values, the four ATD features were assigned (left to right) as 3-cis,5-cis, 3-trans,5-cis, 3-trans,5-cis, and 3-trans,5-trans (Table 2). The theoretical structures also revealed that the 3-cis,5-cis isomer has the most compact structure with the two caffeic acid (CA) groups/moieties collapsed toward each other, while the 3- trans,5-trans isomer was the most extended with the two CA groups widely open and extending in opposite directions. The 3-cis,5-trans and 3-trans,5-cis isomers both adapt a partially open conformation with one CA group collapsing toward the center and the other group extending to the side. The 3-trans,5-cis isomer, however, does have a slightly more open structure than the 3-cis,5-trans isomer resulting in a larger CCS value. Thus, the cis or trans orientation has an important impact on the molecule’s conformation. Moreover, the intensity of the 3-cis,5-trans isomer is comparable to that of the 3-trans,5-cis isomer, indicating the chance of formation for these two isomers appears to be similar. To investigate this occurrence and the conversion pathways between the 3,5-diCQA isomers, we monitored the products formed after 2, 5, 10, 20, and 30 min of UV irradiation. Our analyses found that the cis,cis isomer only forms through the 3-cis,5-trans and 3-trans,5-cis isomers not directly from the trans,trans isomer (see SI Figure S1). To estimate the reaction barriers associated with the transformation pathways from the 3-trans,5-trans isomer to the 3-cis,5-cis isomer, we determined the minimum energy paths using the “string method”⁴² similar to Yoon et al.⁴³ The reaction energies (ΔE) along the minimum energy paths are depicted in Figure S2. The results illustrate that the conversion from 3-trans,5-trans to 3-trans,5-cis or 3-cis,5-trans is very high, explaining why diCQA exists as trans,trans until it is irradiated. They also suggest that, statistically, the 3-trans,5-cis isomer will form first; however, the energy barriers are similar and the photoisomerization process appears to result in the equivalent formation of 3-trans,5-cis and 3-cis,5-trans. Following the formation of the 3-trans,5-cis and 3-cis,5-trans isomers, the reaction barriers show that the 3-cis,5-cis isomer will preferentially arise from the 3-cis,5-trans isomer. Thus, by coupling the experimental IMS approach and theoretical calculations, we were able to monitor the photoisomerization process and understand the conversions occurring between isomers.

Figure 3.

Theoretical modeling structures for the cis/trans photoisomerization products of 3,5-diCQA and their theoretical CCS_N2 values compared to the experimental ^DTCCS_N2 values. The error associated with each measurement is noted in Table 2.

Table 2.

Theoretical and Experimental CCS Values (in Ų) for Each cis/trans 3,5-diCQA Isomer

3,5-diCQA	theoretical CCSN2 (Å2)	theoretical CV (%)a	experimental DTCCSN2 (Å2)b	experimental CV (%)
3-cis,5-cis	210	0.2	208	0.3
3-cis,5-trans	226	0.2	228	0.1
3-trans,5-cis	232	0.1	232	0.4
3-trans,5-trans	252	0.2	251	0.3

^aThe theoretical CV is the deviation from the mean resulting from multiple calculations.

^bDTCCS_N2 is the collision cross sections measured by DTIMS in nitrogen buffer gas

This manuscript tackles the difficult problem of identifying and quantifying cis and trans isomers of phytochemicals in natural product and plant extracts, which to date has been extremely challenging due to the vast structural diversity of isomers present in the mixtures. Here IMS-based approaches were successfully utilized to separate complex diCQA isomers and identify the cis/trans isomers with the aid of theoretical modeling. The ultrahigh-resolution SLIM IMS-MS technology enabled baseline separation of the isomers and was a powerful tool for analyzing the complex natural products and elucidating structure-activity relationships. Further, coupling the experiment IMS separations and theoretical calculations provided insight into isomeric conversions, which will be extremely important for drug discovery studies and optimizing specific bioactivities.

METHODS

Materials and Sample Preparations

Authentic standards of trans,trans-dicaffeoylquinic acids (1,3-diCQA, 1,5-diCQA, 3,4-diCQA, 3,5-diCQA, and 4,5-diCQA) were purchased from Phytolab (Vestenbergsgreuth, Germany). Analytical grade methanol was purchased from Romil Ltd. (Cambridge, UK). To obtain the cis isomers of the dicaffeoylquinic acids, a 1 mg/ mL stock solution was prepared in 100% methanol. The samples were then irradiated using a UV lamp (Spectroline, USA) operating at 245 nm with an intensity of 390 μW/cm². The lamp was not covered with any notch filter. The products of UV irradiation were analyzed by liquid chromatography-mass spectrometry,²² and no other species were observed other than those peaks corresponding to the 3,5-diCQA isomers (LC chromatograms shown in Figure S3). For the time-dependent photoisomerization study, the 3,5-diCQA was UV irradiated for 2, 5, 10, 20, 30, 60, and 180 min and the products were analyzed with IMS-MS.

DTIMS-MS

The chemicals were analyzed using an Agilent 6560 ion mobility-quadrupole time-of-flight mass spectrometry (IM-QTOF MS) platform.^39,40 Briefly, for ion mobility measurements, after electrospray ionization, ions were passed through the inlet glass capillary, focused by a high pressure ion funnel, and accumulated in a lower pressure ion funnel trap (IFT). Ions were then pulsed into the 90 cm-long IMS drift tube filled with ~4 Torr of nitrogen gas, where they travel under the influence of a weak electric field (10–20 V/cm). Ions exiting the drift tube were refocused by a rear ion funnel prior to QTOF MS detection, and their arrival time (t_A) was recorded. The reduced mobility (the mobility scaled to standard temperature and pressure) can be determined from instrument parameters by plotting t_A versus p/V ⁴⁴

$$t_{\mathrm{A}}=\frac{L^2}{K_{\mathrm{o}}}\left(\frac{273.15}{760T}\right)\left(\frac{p}{V}\right)+t_0$$

Here L is the drift length, V is the drift voltage, t₀ is the time ion spending outside of the drift cell, T is the drift gas temperature, and p is the drift gas pressure. The reduced mobility can be related to the collision cross sections of the analyte using kinetic theory:⁴⁴

$$\mathrm{\Omega }=\frac{3q}{16N}{\left(\frac{2\pi }{\mu k_{\mathrm{B}}T}\right)}^{1/2}\frac{1}{K_0}$$

Here, q is the ion charge, N is the buffer gas number density at STP, μ is the reduced mass of the ion-nitrogen collision, and k_B is the Boltzmann constant. All the CCS values were measured using seven-stepped field voltages. Each measurement was performed in triplicate, and the coefficient of variance (CV) was below 0.5% in all cases.

SLIM IMS-MS

To achieve ultrahigh-resolution IMS separation of the cis/trans isomers, the UV irradiated products of 3,5-diCQAs were also measured using a structures for lossless ion manipulations IMS-MS platform (SLIM IMS-MS). SLIM IMS uses traveling waves and a compact serpentine ion drift path for efficient ion selection, trapping, and accumulations.^33–38 The SLIM IMS-MS platform has a 13 m long serpentine drift path and multipass capability, as recently described.^38,45

Theoretical Calculations

Molecular modeling was performed using NWChem (v6.6),⁴⁶ a high-performance computational chemistry software.^47,48 Briefly, 2D structure files (.mol) were analyzed using the Marvin pK_a plugin (Marvin 15.9.14, 2015, ChemAxon) for prediction of protonation, deprotonation and adduction sites.⁴⁹ Initial geometry relaxation was performed using the Merck molecular force field (MMFF94)⁵⁰ implemented in Avogadro (v1.1.1).⁵¹ Density functional theory (DFT) based ab initio molecular dynamics (AIMD),⁵² as implemented in NWChem, were used to sample 100 conformers for each molecule.⁵² AIMD calculations were each run for at least 10.2 ps, with conformers sampled every 101.6 fs. The AIMD temperature was maintained using a stochastic velocity rescaling thermostat.⁵³ This was followed by DFT-based frequency calculations for determining the Gibbs free energy of each conformer. The B3LYP exchange-correlation functional was used for all calculations^54–57 and Pople basis sets at the 3-21G level (a double-ζ split-valence potential basis set)^58,59 and 6-31G* level (a double-ζ valence potential basis set having a single polarization function),^60–62 were used for the AIMD and DFT frequency calculations, respectively. All basis sets were obtained from the Environmental Molecular Sciences Laboratory (EMSL) Basis Set Exchange (bse.pnl.gov).^63,64 NWChem output files were processed using custom-written Python scripts. Python (v2.7.10), with the NumPy package (v1.9.2),⁶⁵ was implemented using WinPython (v2.7.10.1, http://winpython.github.io), a free, open-source, and portable full-featured Python-based scientific environment. IPython (v3.2.0),⁶⁶ an enhanced Python shell, was used within the Scientific Python Development Environment (Spyder v2.3.5.2) for NWChem output data processing. Aided by supercomputers, our molecular modeling approach calculated CCS values with high accuracy, enabling the identification of the geometric isomers. The use of a first-principles theory (DFT) based AIMD approach enables consideration of the electronic structure of each isomer and its role in the conformer geometries. CCS values were calculated for all molecular conformers using the MOBCAL software, modified for the room temperature N2-based trajectory method.^67–69 The atom coordinates, radius, and charge distribution of the optimized geometry structures were used as input to the MOBCAL calculations. Final CCS values were obtained by a clustering approach using histogram distributions to determine the most abundant, low-energy, stable conformers. Scatter plots were then generated to show the relative energy of the 100 structures for each molecule (determined by ab initio calculations) versus their corresponding collision cross sections (determined by MOBCAL). An example plot and frequency diagram are shown in Figure S4 for 3-trans,5-cis diCQA. NWChem was also used to calculate the minimum-energy paths using the DFT-based zero temperature “string method”. We employed 40 beads to represent each reaction path, for a total of 80 beads from 3-trans,5-trans diCQA to 3-cis,5-cis diCQA. The B3LYP exchange-correlation functional with the Pople basis set 3-21G was used for the string method with a stepsize of 0.1 and a maximum of 100 iterations.