**2. Results**

#### *2.1. Systematic Discovery of Sphingolipids in Almonds and Pistachios*

We modified our standard LC-MS-MS method used for targeted ceramide measurement [9] to allow the systematic identification (discovery) of ceramides and ceramide hexosides with di fferent sphingosine and fatty acid components. Modifications include:



**Table 1.** Structures of the screened long chain bases and of the corresponding reporter ion fragments.

The systematic search for ceramides and hexose-linked ceramides that contain each of the five hypothesized LCBs was performed, for each sample, in four separate chromatographic runs. Each run alternates two 1-s *i*-PI scans of *m*/*z* 264 (reporter ion of d18:1 internal standards and of ceramides with d18:1 sphingosine) and one for each of the other LCBs (*m*/*z* 262, 266, 280, 282). This procedure allowed for an estimation of the amount of unexpected ceramides and carbohydrate-linked ceramides with reference to those of the Cer12:0 and Cer12:0-Glc that show in the *i*-PI scan of *m*/*z* 264.

To verify whether any identified compound is really a ceramide hexoside (there is no possibility to ascertain hexose stereochemistry and configuration with this set of experiments), an iso-energetic Neutral Loss scan (*i*-NL) of the hexose fragment (C6H12O6, MW 180.2 Da) was paired to the Precursor Ion scan that yielded the putative ceramide hexoside.

Figure 1 shows the results of this set of experiments applied to a few interesting samples used to illustrate the discovery procedure. The other samples showed a similar pattern, with only differences in the relative intensity of the chromatographic peaks.

**Figure 1.** Chromatographic traces of the iso-energetic Precursor Ion scans of five long chain bases ((**A**–**E**), see Table 1 for their structures) in the extract of a pistachio cultivar. The dashed trace displays the chromatographic gradient, as volume fraction of acetonitrile (%B; right vertical axis). Numbers on top of prominent peaks are the *m*/*z* of the most intense ion signals. Labelled are the added internal standards (12:0-Cb and 12:0-Cer) and the newly identified hexosyl-ceramides A1 and A2.

Figure 2 illustrates the corresponding integrated mass spectra of each iso-energetic Precursor Ion scan of Figure 1. Main signals that may correspond to likely unexpected ceramide species awaiting identification are highlighted.

**Figure 2.** Integrated Precursor Ion mass spectra of the iso-energetic scan five long chain bases ((**A**–**E**), see Table 1 for their structures) in the extract of a pistachio cultivar. Integration is performed, for all experiments, within the 3.5 to 21 min interval of the chromatographic traces of Figure 1. Labelled are the *m*/*z* values of some main signals.

The measurement of more than 50 chromatographic runs for the 15 examined samples highlighted the presence of at least two main chemical species. A major, earlier-eluting one (A1, retention time, RT ≈ 8.5 min; relative retention time, RRT = 1.18 vs. 12:0-Cb; MH<sup>+</sup> 714) is present in all pistachio and almond samples (Figure 1A is a representative pistachio sample). A minor, later-eluting one (A2, RT ≈ 9.1 min; RRT = 1.23 vs. 12:0-Cb; MH<sup>+</sup> 698) is present only in pistachios, but not in almonds. The two compounds featured prominently in the Par262 trace (Figure 1A) and yielded weaker peaks in the Par280 (Figure 1D) trace.

According to the general fragmentation pattern of protonated ceramides reported in Figure 3 [11], the glycosidic nature of compounds A1 and A2 is warranted by the coelution also in the NL180 experiment (Figure 4A,C) and by the observation of the same precursor ions in the NL180 and Par262 spectra (Figure 4B,D).

**Figure 3.** Overall fragmentation pattern of protonated hexosyl-ceramides, exemplified for those with the "standard" 1,3-dihydroxy-C-18:1 long chain base. Panel (**A**): correspondence of fragments to motifs in the molecular structure. Panel (**B**): development of decomposition processes and triple quadrupole scan experiments used to highlight individual molecules from their fragmentation. Panel (**C**): formal connectivity of the ceramide pivot fragment O".

**Figure 4.** Confirmation of an unexpected hexose-containing ceramide species in the extract of an almond cultivar. The highlighted peak in panels (**A**,**C**) is that of compound A1. *m*/*z* 714 is the molecular mass (MH+) in both experiments (Par262, panel (**B**), and NL180, panel (**D**). *m*/*z* 696 derives from MH<sup>+</sup> by water loss (panels (**B**,**D**)). In Par262 (panel (**B**)) *m*/*z* 552 and 534 derive from the previous ones by the loss of hexose.

Their connectivity was investigated by recording their Fragment Ion spectra in the EPI mode (Q3 used as a LIT) in separate chromatographic runs (A1, A2, Figure 5).

According to the fragmentation pattern summarized in Figure 3, compounds A1 and A2 thus consist of:


The occurrence of a linked hexose unit at C-1 was confirmed by performing an NL-180 scan concurrent to the Par262 scan, observing peaks in both traces at the retention times of the two compounds and MH<sup>+</sup> and [MH − H2O]+ ions (Figure 3).

Final evidence of the identity of the precursor and fragment ions detected in the triple quadrupole instrument was achieved by re-analyzing two almond and two pistachio extracts under comparable chromatographic conditions in a quadrupole-ToF instrument that measures *m*/*z* at a resolution close to 30,000. Figure S1 and Table S1 summarize the results. High-resolution measurements match the elemental composition of the examined precursor ions and of the expected fragments within 2–26 ppm, thus confirming the interpretation of the LIT fragment ion spectra of Figure 4 and the connectivity of the two compounds.

**Figure 5.** Enhanced Product Ion (EPI, CE 30 DV) spectra of unexpected ceramide species in the extract of a pistachio cultivar. (**A**) Compound A1 is that eluting at 8.5 min in the chromatographic trace A of Figure 1; (**B**) compound A2 is that eluting at 9.1 min.

Previous researchers already extracted and identified similar compounds in almonds [12]. The authors assigned the connectivity of the additional long chain base double bond as z-Δ8 and identified the stereochemistry of the appended hexose with the extensive use of Nuclear Magnetic Resonance experiments.

#### *2.2. Levels of the Discovered Sphingolipids in Almonds and Pistachios*

The analysis of three almond and nine pistachio samples highlighted that A2 (which contains a saturated C16 fatty acid) occurs in both almonds and pistachios, while A1 (which contains a hydroxylated C16:0-h fatty acid) is specific of pistachio. Their abundance in the sample could be estimated in the absence of authentic purified reference ceramides, by comparing the peaks areas to that of C12:0 cerebroside (peak at 7.3 min in panel B of Figure 1) added as an internal standard. In pistachios, components A1 and A2 were estimated on the average at 19.3 and 5.2 μg/g of extracts, respectively. In almonds, compound A1 is estimated on the average at 24.2 μg/g of extracts. The levels of the two characteristic ceramides in the individual nut samples are reported in Table 2.


**Table 2.** Levels of two specific ceramides in three almond and nine pistachio samples.

A1: RT 8.5 min; MH<sup>+</sup> 715; A2: RT 9.1 min; MH<sup>+</sup> 699; Results in μg/g (expressed as Cer12:0-Glc equivalents); n.d.: not detected.

While Reisberg and collaborators [12] did not perform an analytical assessment, but rather isolated and characterized component A1, they could purify approximately 120 mg from 14 kg of almonds. This amount corresponds to 8.6 μg/g and is close to that measured in our samples.

#### *2.3. Interference of Triglycerides in the Discovery of Ceramides in Almonds and Pistachios*

As apparent from the chromatographic traces and integrated mass spectra of Figures 1 and 2, an intense signal was detected for materials that eluted later than 15 min and that contained arrays of signals up to the *m*/*z* 1000 limit of the Q1 quadrupole scan. The time limit of 15 min is the elution time of the ceramide with the longest (C24:0) fatty acid of the standard series. Grounding on the specificity of the even-*m*/*z* O" reporter fragments for nitrogen-containing molecules, this region of the chromatograms was explored to detect possible sphingoid compounds with more complex structures.

In particular, the Par264 trace (Figure 2B) featured even-*m*/*z* molecular signals with a molecular mass around 900 Da (*m*/*z* 898, 900, and 902). Possible connectivity at this molecular size that is compatible with a ceramide structure might belong to epidermosides, a class of very large ceramides with a ω-hydroxylated fatty acid, to which a further fatty acid is connected through an ester bond. Such compounds are found in skin, as components of the lipid mixture of the stratum corneum [13]. However, the Fragment Ion spectra recorded for the main *m*/*z* signals did not match the expected behaviour for the anticipated structures. In particular, the observed fragmentation did not yield the fragment pair spaced by 18 Da that derives from the two fragmentation modes of the long chain ester bond. In addition, the compounds eluted earlier than expected for their molecular size based on the behaviour of the standard series of saturated and unsaturated sphingosine ceramides [14].

Since the occurrence of these likely spurious signals may bias the discovery of high-mass, unexpected sphingosine derivatives, a brief investigation was undertaken to understand their origin. In particular, the mass spectral features of the cluster at 18–20 min elution time in the Par264 experiment (Figures 1B and 2B) matched the loss of the elements of fatty acid-ammonium ion pairs in the Product ion experiment (diglyceride fragments at *m*/*z* 599.5, 601.6, 603.5, data not shown), which is the well-documented behaviour of triglycerides [15]. High-resolution measurement in the Q-ToF mass spectrometer on the same extract confirmed the postulated identity of the observed signals as ammonium ion adducts of triglycerides (Figure S2). In particular, TG ammonium adducts at *m*/*z* 900.8 (TG 54:4) were mainly composed of dioleoyl-linoleoyl-glycerol (OOL). This identification is confirmed by the occurrence of the expected losses of fatty acids in the tandem mass spectra, with the generation of diglyceride fragments at *m*/*z* 601 and 603 in a 3:1 ratio (Figure S2). Under the adopted collision conditions, the acylium fragments do not form.

To investigate whether co-extracted triglycerides may be the source of the high-mass signals erroneously identified as putative complex ceramide (Figures 1B and 2B), a source scan at baseline unit resolution in Enhanced Mass Scan (EMS) mode was recorded (data not shown). Most *m*/*z* signals recorded at the retention times of the considered chromatographic peaks correspond to a cluster of eluting triglycerides (TG) with 12C isotopomers at *m*/*z* 896.6, 898.7, and 900.5. The observed ion clusters corresponded to ammoniated triglycerides with a total of 54 fatty acid carbon atoms and a number of unsaturations that range from six to four, or three, two, and one units of linoleic acid (*m*/*z* 896.6 TG 54:6, *m*/*z* 898.7 TG 54:5, *m*/*z* 900.5 TG 54:4). The precursor ions detected in the Par264 scan (Figures 1B and 2B) at *m*/*z* 898.5, 900.5, and 902.6 correspond to the [M + 2] isotopomers of each ammoniated triglyceride (TG 54:6, TG 54:5, TG 54:4; Figure S2C,E), which mainly contain two 13C carbon atoms in the molecule. Therefore, the observed signal (Figures 1B and 2B), that interferes with the O" reporter fragment ion of sphingosine ceramides, may correspond to the 13C1-isotopomer (*m*/*z* 264) of the acylium fragment of linoleic acid *all*-12C C18H31O+ (*m*/*z* 263) that is present as a minor process in the spectra of triglycerides ionized as ammonium-adducted species [16].

To test whether this hypothesis holds, the same sample was analyzed with a simultaneous recording of the Precursor Ion spectra of *m*/*z* 263 (the acylium fragment of linoleic acid, *all*-12C C18H31O+) and of *m*/*z* 264 (13C1-isotopomer). As apparent in the two superimposed graphs of Figure S3, the chromatographic profiles of *m*/*z* 263 and 264 closely match in the time frame later than approximately 9 min, and their intensities are in an approximately 4:1 ratio. Only in the time frame between 7.5 and 8.5 min, where the abundant 12:0 ceramides elute, the *m*/*z* 264 signal (the O" fragment of S18d:1) predominates over the background signal and yields the expected molecular precursors of the S18d:1 species. In the time frame later than approximately 9 min, the two Precursor Ion spectra record signals of ammoniated triglyceride and diglyceride from the bulk of the un-fractionated lipid extract (spectra not shown). In particular, between 18 and 20 min, the Par263 scan (Figure S3) detects domes of molecular signals, among which those at *m*/*z* 896.9, 898.8, and 900.9 (TG 54:6, TG 54:5 and TG 54:4). The corresponding Par264 scan detects domes of molecular signals (*m*/*z* 897–902, spectra not shown) with an intensity that is approximately 25% of that of the corresponding signals of the Par263 scan. This intensity ratio closely corresponds to that expected for the occurrence of the 13C isotopomers of the fatty acid in the fragment signals of the precursor triglycerides (Figure S4).
