Fast liquid chromatography-mass spectrometry reveals side chain oxysterol heterogeneity in breast cancer tumour samples

Oxysterols can contribute to proliferation of breast cancer through activation of the Estrogen Receptors, and to metastasis through activation of the Liver X Receptors. Endogenous levels of both esterified and free sidechain-hydroxylated oxysterols were examined in breast cancer tumours from Estrogen Receptor positive and negative breast tumours, using a novel fast liquid chromatography tandem mass spectrometry method. Multiple aliquots of five milligram samples of 22 tumours were analysed for oxysterol content to assess intra- and inter-tumour variation. Derivatization was performed with Girard T reagent (with and without alkaline hydrolysis) and sample clean-up was performed using a robust automatic on-line column switching system ("AFFL"). Oxysterols were separated isocratically on a 2.1 mm inner diameter column packed with ACE SuperPhenylHexyl core shell particles using a mobile phase consisting of 0.1% formic acid in H2O/methanol/acetonitrile (57/10/33, v/v/v) followed by a wash out step (0.1% formic acid in methanol/acetonitrile, 50/50, v/v). The total analysis time, including sample clean-up and column reconditioning, was 8 min (80% time reduction compared to other on-line systems). Analysis revealed large intra-tumour variations of sidechain oxysterols, resulting in no significant differences in endogenous oxysterols levels between Estrogen Receptor positive and Estrogen Receptor negative breast cancers. However, a correlation between esterified and free 27-hydroxycholesterol was observed. The same correlation was not observed for 24S-hydroxycholesterol or 25-hydroxycholesterol. The oxysterol heterogeneity of tumour tissue is a critical factor when assessing the role of these lipids in cancer.


Oxysterols and breast cancer
Breast cancer (BCa) is the most commonly diagnosed cancer among woman [1] and elevated LDL-cholesterol is a predictor of poor outcomes [2]. Oxysterols are cholesterol metabolites and the different isomers possess diverse biological roles in development and numerous diseases [3][4][5][6][7][8]. Oxysterols can be formed by either enzymatic or nonenzymatic oxidation of cholesterol. In breast cancer, the enzymatically formed 27-hydroxycholesterol (27-OHC) is of particular interest; In BCa, 27-OHC induces cell proliferation and tumour growth via its ability to bind and activate the Estrogen Receptor (ER) [9,10] thus being at least part of the molecular evidence that links elevated cholesterol with BCa [9]. 27-OHC also activates the liver X receptor (LXR) leading to enhanced expression of transcription factors that promote the epithelial to mesenchymal transition [9,11], and modifies immune γδ-T cells and polymorphonuclear-neutrophil function to promote ERnegative BCa metastasis [12]. A related sidechain oxysterol, 25-OHC, has also been found elevated in the circulation of BCa patients under going treatment for metastatic disease [13]. In BCa research, the focus has therefore been mainly on the 27-hydroxycholesterol [10,14,15]. However, the other enzymatically formed sidechain oxysterols (scOHC) can play roles, as e.g. LXR activators. In addition, little is known regarding the levels of oxysterols in ER-positive and ER-negative BCa tissue. Oxysterols can, as cholesterol, exist either as free sterols or bound to fatty acids (esterified) or sulfonated species [16]. In BCa neither the relative abundance, nor the roles of these modifications are well understood. Hence, methods for studying oxysterols in breast cancer tissues are needed.

Measurement of oxysterols
Traditionally, gas chromatography mass spectrometry (GC-MS) has been used for determination of sterols, such as oxysterols, after derivatization with trimethylsilyl (TMS) to increase volatility [17][18][19]. This procedure usually includes an alkaline hydrolysis step, resulting in the determination of total oxysterol, including both free and esterified oxysterols, as well as sulfonated oxysterols. Liquid chromatography mass spectrometry (LC-MS) is increasingly used for oxysterol determination [18]. The neutral nature of the sterol structure makes ionization of oxysterols using electrospray ionization (ESI) difficult, and derivatization is performed to "charge-tag" the analytes (e.g. with picolinyl, N,N-dimethylglycine (DMG) or Girard P and T [18,[20][21][22]. With charge-tagging, LC-MS surpasses GC-MS in sensitivity [23]. In our hands, Girard derivatization has both been highly robust and allowed extremely low limits of detection. Additional sensitivity gains are achieved by downscaling the chromatography system (as demonstrated with capillary LC [24], nano LC [21] and open tubular LC [25]. Regarding BCa tumours, a goal is to use as little sample as possible to allow multiple analysis of the same tumours; both derivatization and LC down-scaling are attractive tools in this regard. Also, in contrast to published GC-MS methods, LC-MS methods with derivatization allow for both free oxysterol determination (derivatization without hydrolysis) and total oxysterol determination (free + esterified and sulfonated oxysterols).
Separation of oxysterol isomers using conventional reversed phase LC particles/stationary phases can be challenging and time-consuming [19]. When samples take upwards of tens of minutes per sample then the high-throughput analyses required in large cohorts is not possible. However, with improvements in column particle technology (e.g. coreshell particles) and alternative stationary phases, the chromatographic process can be speeded up [19]. Shorter times have been achieved previously. For example Pataj et al. [20] have resolved oxysterol isomers using core-shell particles and a biphenyl phase in less than 8 min whilst McDonald et al. [26] have resolved side-chain oxysterols in less than 12 min using core-shell particles and C18 phase. With this in mind, we sought out to combine Girard T derivatization, core-shell particles, alternative stationary phases and downscaled LC for oxysterol determinations. We have combined these features with rugged on-line solid phase extraction [27,28] to reduce manual sample preparation. The final method was applied to tumour samples from 22 BCa patients to examine and compare the endogenous concentration of free and total scOHCs in ER-positive and ER-negative primary breast tumours. Tumour heterogeneity was examined by cutting tumours into multiple aliquots.

Chemicals and solutions
All reagents were of HPLC grade or higher. Methanol (MeOH) and acetonitrile (ACN) were purchased from VWR (Radnor, PA, USA) and were used to make mobile phases together with formic acid (98%, Sigma-Aldrich, St. Louis, MO, USA) and type 1 water from a Millipore Milli-Q ® type 1 ultrapure water system (Merck KGaA, Darmstadt, Germany).

Derivatization with Girard T reagent
Analytes were "charge-tagged" following the procedure described by Griffiths et al. [29] with modifications as described in [21]. Briefly, aliquots of 200 μL 0.03 mg/mL cholesterol oxidase from Streptomyces sp. (Sigma-Aldrich) in 50 mM phosphate buffer pH 7 were added and the solutions were heated to 37°C for one hour. To each sample/standard solution 15 mg Girard T reagent (Sigma-Aldrich), 15 μL glacial acetic acid (VWR) and 500 μL MeOH were added and reaction took place overnight, in the dark at room temperature. All solutions were stored at 4°C after derivatization.

Breast cancer tumour samples
Breast cancer tumours were obtained from the Leeds Breast Research Tissue Bank (IRAS ID: 170113; Tissue Access Committee approval: LBTB_TAC_1/17). Tumours from ER-positive (n = 11) and ERnegative (n = 11) BCa patients were used in this study (Table 1). For more patient characteristics see Supplementary material table S1. Three consecutive slices from each tumour (approx. 5 mg per slice) were collected and homogenized separately in 500 μL 1.5 nM internal standard mixture solution and 30 μL autoxidation monitoring solution (6 μM cholesterol-25,26,27-13C, Sigma-Aldrich) using an IKA T10 Ultra-Turrax homogenizer (VWR). When sample size was ample, the slices were collected from the middle of the tumours, and not the edges. For free oxysterols analysis, 100 μL sample solution was mixed with 100 μL 2-propanol and applied to a Oasis PRiME HLB 1 cc (30 mg) SPE cartridge (Waters) and the oxysterols were eluted with 200 μL MeOH. Solvents were evaporated in an Eppendorf concentrator plus (Hamburg, Germany) and residues were reconstituted in 20 μL 2-propanol followed by derivatization as described above. For total oxysterol (free and esterified) analysis, alkaline hydrolysis was performed by adding 35 μL 2 M KOH (Sigma Aldrich) in MeOH to 100 μL sample solution. Samples were heated at 60°C for 120 min followed by liquid-liquid extraction (LLE) with n-hexane. To get phase separation, 150 μL type 1 water was added to the sample followed by 150 μL n-hexane (VWR). The samples were vortexed for 1 min and centrifuged at 3000 rpm for 2.5 min.
Hexane layer was removed, and the LLE was repeated twice with 150 μL n-hexane (combining all the hexane phases in the end). Solvents were evaporated in an Eppendorf concentrator plus and reconstituted in 200 μL 2-propanol. Samples were applied to an Oasis PRiME HBL 1 cc (30 mg) SPE cartridge and eluted off with 200 μL MeOH. Solvent was evaporated in an Eppendorf concentrator plus and samples reconstituted in 20 μL 2-propanol followed by derivatization as described above. Sample preparation was performed with and without cholesterol oxidase on one slice from each tumour to identify if any natural occurring 3-keto groups could be present above the detection level.

Fast LC: choice of stationary phase
Various types of reversed phase stationary phases were examined and compared to our previous used ACE C18 column (3 μm porous particles), with an overall goal to achieve baseline LC separations (Rs > 1.5) of 22R-OHC, 25-OHC, 24S-OHC and 27-OHC within 5 min. For each column, the mobile phase composition was optimized with regards to ACN and MeOH content. Good resolution was achieved using the Hypersil Gold column C18 (Rs = 1.4-4.2) by decreasing the column temperature to 15°C. However, the analysis time was not considered to be acceptable (13 min). An ACE SuperC18 core-shell column provided shorter retention times, but not sufficient resolution (Rs = 0.8-5). Two more "exotic" stationary phases, namely Torous 2-PIC and Torous 1-AA, did not provide acceptable liquid chromatography of the oxysterols, resulting in either several minute-broad peaks (Torous 1-AA) and/or no separation of the oxysterol isomers (Torous 2-PIC). The best performing column was the ACE SuperPhenyl Hexyl column (core-shell, 2.1 mm ID x 150 mm) using a mobile phase consisting of a mixture of ACN, MeOH and water, together with a high flow rate (900 μL/min) and temperature (45°C). All the Girard derivatized side-chain oxysterols were separated with R s > 1.5 in 4 min (Fig. 1A), meeting our abovementioned goal.

Fast LC coupled with an on-line sample clean-up system
To remove excess derivatization reagent (which can e.g. contaminate the LC-MS system), we wanted to include the ACE SuperPhenyl Hexyl column in an automatic filtration and filter backflush solid phase extraction (AFFL-SPE) system (Fig. 1C). SPE columns with ACE SuperPhenyl Hexyl material was not commercially available, hence a standard C 18 SPE column was used instead. Incorporation of the AFFL-SPE system allowed for robust reagent removal and analyte enrichment, but also resulted in some loss of resolution (R s = 1.1-1.8) however still considered sufficient for quantitation. The total method run time was 8 min, including sample clean-up, separation, and washing/conditioning of the column. With this method, we could inject up to 60 μL of sample, resulting in calculated quantification limits (cLOQ) of > 31 pM, without chromatography deterioration (Fig. 1B). With larger injection volumes, an interfering peak began to co-elute with 27-OHC (data not shown).
Narrower columns of 0.1 mm ID and 0.3 mm ID were also explored. In experiments with commercial and self-packed columns, the resolution of these columns was "under-par" compared to the larger 2.1 mm ID columns used above (data not shown). However, the 2.1. mm ID columns set-up described above was found to have sufficient sensitivity for our present task, as described below.

Method evaluation
The method was evaluated in the range of 50-500 pM using pooled plasma samples (biologically complex, but more available compared to limited tumours samples which was the final matrix), see supplementary material table S3. The limit of quantification was calculated (cLOQ) from linearity curves and was in the range of 15 pM-31 pM. Apparent recovery was acceptable, in the range from 80 to 97 %. Intra-day precision was below 20% for all concentration levels. Inter-day precision (calculated using single factor ANOVA) was between 10-23 % at lower LOQ (LLOQ), 5-10 % at medium LOQ (MLOQ) and 5-11 % at high LOQ (HLOQ). No carry-over was observed.

Oxysterol heterogeneity in breast cancer tumours
Free and total scOHCs in three separate slices of 11 ER-positive and 11 ER-negative tumours were determined. By analyzing three consecutive slices (5 mg) of each tumour sample revealed a large intertumour variability (RSD > 20%) in scOHC concentrations. This variability was not always uniform across all target analytes, as some scOHCs were similar across intra-tumour triplicates, whilst others showed higher variability, even within the same tumours. While some samples showed minor concentration variations ( Fig. 2B and C), other tumours had alterations as to which scOHC had the highest concentration ( Fig. 2A, cut 1 and 2 have higher concentration of 27-OHC, while cut 3 has a higher concentration of 24S-OHC). In general, this high intertumour variability meant there was extensive overlap in scOHCs concentration between all individual tumours irrespective of ER subtype status for both total and free scOHC ( Table 2). 27-OHC was the most abundant OHC measured. Interestingly, in ER-positive tumours we observed a strong correlation between free and esterified 27-OHC (R 2 = 0.89, p < 0.0001) and 25-OHC (R 2 = 0.82, p < 0.01) but in ER-negative tumours these correlations were weak (27-OHC; R 2 = 0.61, p < 0.05) or absent (24S-OHC and 25-OHC) (Fig. 3).

Fast LC of derivatized oxysterols
Sufficient sensitivity and resolution of the target oxysterols were achieved with the fast LC set-up using 2.1 mm ID column coupled in the on-line AFFL sample clean up system. Some resolution was lost compared to direct injection on the columns. Possible reasons for a loss of resolution may be additional extra-column contributions to band broadening, or that an ACE SuperPhenyl Hexyl trap column was not available, so a conventional C18 trap column was used instead. Hence, sufficient refocusing of the oxysterols on the column was not achieved due to larger relative retention on the SPE column compared to that on the separation column.
A fast separation (8 min including sample clean-up) was possible with high flow rate and temperature and core-shell particles. With coreshell particles a higher flow rate is possible without increased band broadening compared to that of fully porous particles. This is mainly due to decreased Eddy dispersion (A in Van Deemter equation), not lower resistant to mass transfer (C/u in Van Deemter equation) [30]. By using the SuperPhenyl Hexyl column, we did not observe any peak splitting of the derivatized oxysterols, which usually is observed for the Girard P/T hydrazone derivates [21,22]. In addition to the analytes, several other peaks of unknown identity were observed (see e.g. early eluting peaks in 2B); this serves as an example of the need for chromatography, as non-separating interferences with similar MS features could have interfered with identification/quantification under nonoptimized conditions. Downscaling the chromatographic system (e.g. to 0.1 mm ID column) to achieve better sensitivity, hence use smaller sample size was not achieved as commercial columns in these dimensions are not  available, and the in-house packed columns lacked efficiency. Although we may be able to optimize packing procedures for the ACE SPH solid core particles, reduced column efficiency of solid core particles in narrow ID is not too surprising [31]. Nevertheless, the 2.1 mm ID column with large volume injection (60 μL) was sufficiently sensitive to allow determination of BCa related oxysterols in 5 mg tumour tissue. For some tumours, the sensitivity was too good and dilution of the samples (10 x) was performed.

Total vs. free oxysterols
Oxysterols can be present both in their free from, but also as sulfonated and esterified versions in our body. In these samples, the measured oxysterols after alkaline hydrolysis are most likely to be free and esterified sterols, not sulfonated, due to lack of expression of the SULT2B1 enzyme in ER-BCa tumours (Data not published). The esterified version is connected to lipid storage and transport. Historically, GC-methods for oxysterols analysis has mostly been quantifying total oxysterols (e.g. after alkaline hydrolysis), whilste LC-MS method have mostly been quantifying free oxysterols. This has often led to confusion and trouble when comparing results and concentration levels of the different oxysterols in e.g. plasma (see [18] for table with comparison of free and total oxysterols in plasma samples). Hence, we wanted to compare the two methods, by splitting each sample cut in two (after homogenization) and subject one part for alkaline hydrolysis before derivatization into Girard T derivates. As expected, the total oxysterol concentrations were higher than the free oxysterols in the samples.

Oxysterol heterogeneity in breast cancer tumour
The huge intra tumour variation makes comparison of oxysterol concentration challenging. The reason for variation cannot be addressed to sample preparation, e.g. extraction efficiency, as the variation is both in absolute and relative concentration of the oxysterols. Changes in extraction efficiency of the oxysterols from the tumours should only yield changes in absolute concentration, hence the variation are reflecting tumour heterogeneity. The variations could possibly be due to differences in tumour invasion of oxysterol synthesizing macrophages, fibroblast and/or adipocytes. We could however observe a strong correlation between free and esterified 27-OHC and 25-OHC in ER-positive tumours suggesting intact oxysterol metabolism pathways in these tumours. The loss of correlation in ER-negative suggests alternative usage of esterified or free oxysterols in this more aggressive breast cancer type. 27-OHC is a selective estrogen receptor modulator and is linked to metastasis and proliferation of Estrogen Receptor positive BCa. 27-OHC can also activate the LXR receptor, a possible key player in metastatic and/or chemotherapy resistant triple negative BCa. Currently the relative contributions of free and esterified sidechain oxysterols to LXR signalling in breast cancer remains unexplored. The oxysterol heterogeneity of tumours is a critical factor when assessing the role of these lipids in cancer.

Conclusion
A fast LC-MS method for determination Girard T derivatized oxysterols was developed to examine BCa related oxysterols in tumours from both ER-positive and ER-negative patients. With large volume injection and automatic sample clean-up (AFFL) oxysterols could successfully be quantified from 5 mg sample size with a total LC-MS analysis time of 8 min per sample. Both free and total (after alkaline hydrolysis) scOHS concentration were examined. Multiple slices from same tumours revealed huge intra-tumour variations, hence revealing oxysterols heterogeneity and making comparisons challenging. Future studies could include assessing plasma vs. tumour levels, and expanding the method to measure other LXR-active sterols, e.g. 24S,25-epoxycholesterol.

Declaration of interest
None.

Table 2
Side-chain OHC concentrations in breast tumour samples. Total, esterified and free oxysterols were measured in three slices from each of 22 tumours (ERpositive n = 11; ER-negative n = 11). The minimum, maximum and mean concentrations are shown. There were no significant differences in oxysterol concentrations when comparing ER-positive and ER-negative tumours.