Prior to monitoring the reaction of hexacosanoic acid and phytosphingosine, both starting materials were analyzed. Standard electrospray ionization of phytosphingosine, in which the analyte is dissolved in acetonitrile:water:acetic acid (50:50:0.1% v/v) and is sprayed through a central needle that is sheathed only in nitrogen gas for pneumatic assistance yielded a strong signal for [M + H]⁺ at m/z 318.2994 (theoretical 318.2989) with a signal to noise ratio (S/N) of 7000. Slightly raising the nozzle voltage provided in-source fragmentation with consecutive losses of water m/z 300.2846 and 282.2787, Figure 1. The structure and purity of phytosphingosine were also confirmed by NMR. Proton and carbon chemical shift assignments were made from the analysis of COSY, HMQC and HMBC spectra (using ~1.0 mg phytosphingosine dissolved in 0.65 mL DMSO-d₆. as sample). The aliphatic region (0.8–1.8 ppm) of the proton spectrum integrates for 29 protons, 13 methylene groups, and a single methyl resonance are observed which are consistent with a straight chain (CH₂)₁₃CH₃ alkyl side chain. The one-dimensional proton-decoupled carbon spectrum also confirms the presence of a straight chain alkyl side chain. The 0.7 ppm chemical shift dispersion observed for the central nine carbons of the alkyl chain indicate that no branching is present. The amino NH₂ protons are visible as a very broad resonance at ~5.65 ppm. The ¹H and ¹³C chemical shifts, proton coupling constants, ¹H–¹H COSY correlations and heteronuclear ¹H–¹³C 2- and 3-bond correlations confirm the structure depicted in Scheme 1. NMR and mass spectral data were in agreement, indicating that the phytosphingosine starting material was pure with trace impurities limited to << 5%.

Due to the long aliphatic chain of the hexacosanoic acid, this starting material was not soluble in acetonitrile:water. The sample dissolved in tetrahydrofuran (THF) yielded [M − H + 2Na]⁺ at m/z 441.3699 (theoretical 441.3679) and the cluster species [2M − 2H + 3Na]⁺, m/z 859.7601 (data not shown for brevity).

Attempts were made to monitor the N-[3-(dimethylamino)propyl]-N′-ethylcarbodiimide hydrochloride/1H–1,2,3-benzotriazol-1-ol hydrate (EDC/HOBT) reaction and acid chloride reaction products by standard monospray ESI. Strong signals were observed for phytosphingosine; however, no signal was observed for the hexacosanoylceramide product expected at m/z 696, [M + H]⁺. These samples were only partially soluble in an acetonitrile:water mixture. In order to solubilize the hexacosanoic acid starting material and ceramide, approximately 1 mL of tetrahydrofuran (THF) was added to the 1 mL sample solution until dissolution was complete. The solubilized sample was sprayed by standard single needle design. Despite optimization of gas flow rates and nozzle/skimmer voltages, which were skewed towards higher nozzle voltages, only a minor signal was observed at m/z 696.6891 (theoretical 696.6864) with an S/N of 10 and a ¹³C isotope S/N of 4.0. This signal intensity is insufficient to unambiguously identify the product. Since no analytical methods were available to us to verify the product at the onset of the investigations, we sought alternate methods of ionization that we have at our disposal. The ionization schemes, attributes and pitfalls found for ceramides are discussed below.

Prior experience with highly apolar, water insoluble compounds, such as acrylic polymers, demonstrated that direct laser desorption ionization (DLDI) provides a rapid means of ionizing and characterizing such samples [35,36]. The technique is simply described as “dilute and shoot”, yielding representative data of polymer multicomponent samples. Matrix-assisted laser desorption ionization (MALDI) FTICR MS analysis of phospholipids has been successfully performed to yield high quality results. The identification of a proper matrix for the compounds is essential to the analysis [37]. DLDI does not suffer from matrix compatibility problems because the solubilized sample is placed directly on a cellulose substrate without a chemical matrix. Figure 2 shows that DLDI of the EDC/HOBT reaction yielded a significant signal at m/z 734.7 (theoretical 734.6), the targeted ceramide product, [M + K]⁺, with an S/N of ~200. The cellulose substrate which has strong absorbance at 10.6 µm (CO₂ emission wavelength) functions as our matrix. The reaction mixture of hexacosanoyl chloride with phytosphingosine displayed additional signals at m/z 1113 and 1491, corresponding to the 2nd and 3rd covalent additions of hexacosanoyl to phytosphingosine (Figure 3). Accordingly, the EDC/HOBT coupling reaction was determined to be the best and was monitored further. Direct laser desorption mass spectrometry provided a rapid method to identify products from both reactions in a simple manner that takes less than twenty minutes to perform including sample preparation.

From the data, it is apparent that the solid cellulose substrate that holds the sample can randomly yield significant matrix ions, which could interfere with analysis. Second, the resolving power is much lower than expected for a 3T FTICR MS instrument due to laser desorption’s ionization mechanism and ion capture which yields ions that are not centered in the ICR cell, the homogeneous region of the magnetic field. Without quadrupolar axialization and collisional cooling to center the ions [38,39], the resolving power of DLDI is limited to ~3500 as the transient damps quickly to noise within 0.3 s. Mass measurement accuracy without internal reference ions were approximately 0.1 Da, dependent on mass measured. It is possible to use the matrix ions as an internal reference and achieve mass accuracies of ±0.01 Da; in this case, the observed mass was m/z 734.630 (theor. 734.642), data not shown for brevity. DLDI is also not amenable to online chromatography, which may be required due to sample solubility differences and the concentration ranges of the compounds under study. However, sample fractions could be collected and easily analyzed off-line.

Sheath flow electrospray ionization was developed in the mid-1990s to improve ionization of intractable compounds by single needle ESI and to provide a simplified means to interface chromatographic methods such as gel permeation chromatography [34,40,41]. Prior attempts to improve electrospray ionization efficiency by adding sodium or potassium to the sample before separation showed that it can degrade the chromatography and compromise the column [42]. In sheath flow ESI, a triple needle design allows one to infuse the sample in an appropriate solvent or the chromatographic eluent through the center needle. The first sheath contains the ionization modifier solution such as sodium iodide or potassium iodide in methanol, all contained within the outer sheath of the nitrogen carrier gas. Separation of the ionizing agent and the analyte provides enhanced solubility of both components without the need to identify a solvent mixture that can solubilize both components in one vial. Ionization takes place in the Taylor cone through evaporation of solvents and mild ion/molecule association. Using this triple needle strategy, Ramjit et al. demonstrated improved analysis of apolar compounds [33].

Figure 4 shows the sheath flow ES FTICR mass spectrum of the EDC crude reaction mixture. Strong signals are observed for [M + Na]⁺ and [2M + Na]⁺. Improved resolving power (~10,000) and mass accuracy (2 mDa) is achieved by this technique with FTICR because the collected electrosprayed ions are now centered in the ICR cell.

With improved mass spectral methods and a more thorough understanding of the sample complexity, the reaction chemistry was re-evaluated. In all cases, the combination of DLDI, Sheath Flow ESI-FTICR MS, and high field NMR analyses provided data consistent with the starting materials and desired product. NMR data acquired on THF solubilized samples generated complementary results. The ¹H NMR spectrum recorded on the crude sample obtained from the EDC coupling revealed the presence of the two starting materials, hexacosanoic acid (~28%) and phytosphingosine (~4%) in addition to the desired product. Although proton integration was unreliable for quantifying the hexacosanoylceramide, the reaction product was easily identified. The amide proton is visible as a doublet at 6.93 ppm. There are two-bond HMBC correlations from the amide proton (6.93 ppm) and the methylene resonance (2.12 ppm) to the carbonyl carbon at 173.3 ppm, confirming the amide linkage. Three hydroxyl resonances are observed at ~4.20, ~4.11 and ~3.76 ppm. Specific assignments for each hydroxyl resonance were not possible due to the broad line widths resulting from chemical exchange broadening. Mass spectral data did not show the presence of starting material seen in Figure 4. Both sheath flow and monospray positive ion electrospray ionization of the hexacosanoic acid starting material yielded the expected [M − H + 2Na]⁺. In the presence of phytosphingosine or hexacosanoylceramide, only the molecular ions of these two compounds were observed, demonstrating preferential positive ionization over hexacosanoic acid. Negative ion sheath flow electrospray ionization yielded signals for both hexacosanoic acid and hexacosanoylceramide along with the cluster species (Figure 5). Molecular ionization efficiencies aside, data yields ~35% of residual hexacosanoic acid starting material, which is in close agreement with NMR results of the crude sample. Subsequently, the sample was recrystallized in THF (95% pure). The resolved proton chemical shifts were assigned from the results of COSY, HMQC and HMBC experiments (Scheme 2 and Table 1). Integration of the proton spectrum is consistent with the desired hexacosanoylceramide product (C₄₄H₈₉NO₄). Mass spectral data of the purified sample was also in agreement.

Samples for NMR measurements contained approximately 1 mg of phytosphingosine dissolved in DMSO-d₆ (Isotec 99.96 atom % D) or 1.5 mg of hexacosanoylceramide dissolved in THF-d₈ (Isotec 100.0 atom % D) in a final sample volume of 0.650 mL. ¹H and ¹³C chemical shifts were referenced to tetramethylsilane (TMS) at 0.00 ppm. One-dimensional carbon spectra were acquired with 72 K data points, a sweep width of 28 kHz and 25,000 transients using a combined recycle delay and acquisition time of 2.3 s. Broadband proton decoupling was achieved by the application of a globally optimized alternating phase rectangular composite pulse (GARP) modulation scheme [43] during the acquisition period and recycle delay.

Proton chemical shift assignments were obtained with pulsed-field gradient correlated spectroscopy (PFG-COSY) [44]. PFG-COSY spectra were acquired with 1 K complex points in t₂ and 1024 points in t₁ consisting of 1–8 transients per increment. The data was multiplied by a sine-bell apodization function in both dimensions and zero-filled to 4 K by 2 K points prior to Fourier transformation.

Carbon chemical shift assignments were made with proton-detected ¹H–¹³C heteronuclear multiple quantum coherence spectroscopy (HMQC) [45] and pulsed field gradient heteronuclear multiple bond correlation spectroscopy (PFG-HMBC) [46]. Broadband carbon decoupling in the HMQC spectra was accomplished with a GARP 1 [43] composite pulse decoupling scheme. HMQC spectra were recorded with 1 K complex points in t₂ and 256 increments in t₁ with 4–8 transients per increment. The fixed delays were optimized for a ¹H–¹³C 1-bond coupling constant of 140.0 Hz. The data were zero-filled to 2 K complex points in t₂ and 1 K complex points in t₁ and were multiplied by a Gaussian function in both dimensions prior to Fourier transformation. PFG-HMBC spectra were acquired in the absolute value mode with 1 K complex points in t₂ and 1024 increments in t₁ with 8–48 transients per increment. The heteronuclear multiple bond fixed delay was set to 65.0 ms to optimize observation of 2- and 3-bond heteronuclear correlations. All two-dimensional experiments were acquired in the absolute value mode with the exception of the HMQC experiment which was acquired in the hypercomplex mode for phase-sensitive presentation.