*4.2. Plant Material*

Three selected almond (A6, A7, A8) and nine pistachio (P7–P15) extracts were analyzed in this study. Their codes and their main characteristics are reported in Table 3.


**Table 3.** Pistachio and almond samples examined in this study. The first column refers to the blinded analysis code and the second column to the sample code.

#### *4.3. Sphingolipid Extraction Procedure*

The total lipid extracts were prepared as described elsewhere [9]. Briefly, after the addition of 10 μL of IS (Cer C12, gluCer C12 20 mM), the powder (250 mg) was extracted in a microtube with an O-ring seal screw with 500 μL of methanol, 100 μL water, and 250 μL chloroform [32]. The samples were sonicated for 30 min and incubated overnight in an oscillator bath at 48 ◦C. The supernatant was evaporated under a stream of nitrogen. The residues were dissolved in 150 μL of methanol and

then centrifuged for 10 min at 13,000 rpm. The clean supernatant (130 μL) was transferred into the autosampler vials, and 10 μL were directly injected in LC-MS/MS.

## *4.4. LC-MS*/*MS Instrumentation*

Two separate sets of mass spectrometry measurements were performed in di fferent computerized integrated systems, each consisting of a liquid chromatograph coupled to a tandem mass spectrometer.

Discovery of ceramides and ceramide hexosides with a diverse long chain base was accomplished in a hybrid triple quadrupole-linear ion trap tandem mass spectrometer. The integrated liquid chromatograph system was an UltiMate ® 3000 LC Systems (Dionex ™, Sunnyvale, CA, USA), with an autosampler, binary pump, and column oven (Thermo Fisher Scientific, Waltham, MA, USA). The tandem mass spectrometer was an AB Sciex 3200 QTRAP LC-MS/MS instrument with electrospray ionization (ESI) TurboIonSpray ™ source (AB Sciex Framingham, MA, USA). Instruments were managed with manufacturers' software (Analyst software (version 1.6.2) and according to manufacturers' instructions. Results exported as .txt files were post-processed in custom spreadsheets, as needed. The principle of the innovative scan modes employed in this instrument is briefly described in Appendix A.

A set of confirmative measurements used a Shimadzu UPLC interfaced to a TripleTOF 6600 (Sciex, MA, USA) high-resolution hybrid quadrupole time-of-flight mass spectrometer equipped with Turbo Spray IonDrive, under essentially the same chromatographic conditions (*2.3*, *v. infra*). The resolution was close to 30,000 according to the manufacturer's software evaluation. The main instrument parameters were: CUR 35, GS1 55, GS2 35, capillary voltage 5.5 kV, source temperature 350 ◦C, declustering potential (DP) 50 eV. Two analysis modes were applied. One ("*targeted*" approach) recorded, within each 1-s measurement cycle, the source spectrum ( *m*/*z* 200–1400; 250 ms accumulation time) and the fragment ion spectra of selected precursors only (100 ms accumulation time). The other ("*untargeted*" approach) recorded, within each 1-s measurement cycle, the source spectrum, and the fragment ion spectra of the ten most abundant precursor ions that occur in the previous source spectrum. Collision energy was set at 30 V on a nitrogen gas target. Data were extracted with the proprietary PeakView data managemen<sup>t</sup> software and exported to spreadsheets for further evaluation.

#### *4.5. Separation and Detection of Sphingolipids by LC-MS*/*MS*

Separation of the lipid extract containing ceramides was accomplished in an ACQUITY UPLC BEH C-8 Column, 130 Å, 1.7 μm, 2.1 mm × 100 mm (Waters, Millford, MA, USA) preceded by a security guard cartridge. The flow rate was 0.3 mL/min; the autosampler and the column oven were kept at 15 ◦C and 30 ◦C, respectively, the operating pressure was 450 Psi.

The two mobile phases were: phase A, 2 mM ammonium formate in water and phase B 1 mM ammonium formate in methanol, both containing 0.2% formic acid (*v*/*v*). A multi-linear extended gradient with a total analysis time of 22 min was programmed: the column was equilibrated with 80% (B), increased to 90% (B) in 3 min, held for 3 min, increased to 99% (B) in 9 min, held for 3 min, back to the initial conditions in 2 min, and kept for 2 min at 80% (B).

Mass spectrometry was performed in the positive ion mode (ESI+). The ion spray voltage was set at 5.5 kV, and the source temperature was set at 300 ◦C. Nitrogen was used as a nebulizing gas (GS 1, 45 psi), turbo spray gas (GS 2, 50 psi), and curtain gas (25 psi). Source spectra were recorded in separate experiments in the Enhanced MS (EMS) mode at a scan speed of 1000 Da/s that yielded baseline separation of unit mass peaks in the *m*/*z* range 450–1000. The collision-activated dissociation (CAD) MS-MS experiment used nitrogen as collision gas at the low pressure setting (1.2 × 10−<sup>5</sup> Torr).

#### *4.6. Untargeted Discovery LC-MS*/*MS Analysis by Iso-Energetic Precursor Ion and Neutral Loss Scan in a Triple Quadrupole*

An untargeted discovery method was developed to investigate the presence of unexpected sphingolipids in the samples. Briefly, the method employs Precursor Ion (PI) scans of the analysing quadrupole (Q1), while holding the selecting quadrupole (Q3) to transmit the reporter fragment ions generated by CID of protonated ceramides with different long chain bases (fragment O").

Those selected are at *m*/*z* 264.4 (d18:1D4 sphingosine, [11]) and isomeric d18:1D8 [6], *m*/*z* 262.4 (d18:2D4,8); *m*/*z* 280.4 (t18:1D8, [6]).

Another set of experiments employed Neutral Loss scans, whereby both Q1 and Q3 mass filters are simultaneously scanned at the same rate, with a fixed offset of transmitted *<sup>m</sup>*/*<sup>z</sup>*, which corresponded to the mass of a specific molecular unit. In this case, the neutral fragment was the C6H12O6 unit (MW 180.2) of hexose.

The upper scan range of Q1 was initially *m*/*z* 450–700 in 1 s (measurement of the standard ceramide panel), and later, the upper *m*/*z* value was raised to *m*/*z* 1000 to allow analysis of ceramides with as many 60 carbon atoms. During each 1-s scan of Q1 (or of linked Q1 and Q3), the CAD potential (q2-Q1) was synchronously ramped in order that, at each *m*/*z* value of the precursor ion transmitted by Q1, the effective collision energy was held constant (iso-energetic, or *i*-CID). The particular selected value of the effective collision energy was selected following a spectroscopic study of model ceramides. Actual employed values are of 1.6 eV for the Precursor Ion scans (generation of O" fragment from MH<sup>+</sup>) and 1.0 eV for the Neutral Loss scan (loss of the hexose unit from the MH<sup>+</sup>). Conversion of centre-of-mass collision energy to instrument voltage was accomplished with a re-arranged form of the standard equation (a more detailed explanation is reported in Appendix A). Typical ranges of collision voltage for the Precursor Ion scans were 27.3–44.5 DV for the low mass range (*m*/*z* 450–750) and 27.3–55.9 DV for the extended mass range (*m*/*z* 450–1000). The corresponding values for the Neutral Loss scan were 27.3–44.5 DV for the low mass range (*m*/*z* 450–750) and 27.3–55.9 DV for the extended mass range (*m*/*z* 450–1000).

#### *4.7. Relationship of Molecular Structure to Chromatographic Retention*

To assist and to confirm the identification of unexpected ceramide species, a descriptor of the relationship of molecular structure to chromatographic retention was established. A standard mixture of ceramides ranging from C 12:0 to 24:0 (2.6 μM each) was injected under described conditions. Natural logarithm of capacity factor *k'*, defined as the number of column void volumes necessary to elute each compound, was plotted against the natural logarithm of both the total number of carbon atoms in the ceramide and of the sole number of carbons of the FA (graph reported in Figure S5).

The void volume was estimated from the geometric parameters of the column and from the mobile phase flow. A correction factor for fractional column packing was identified from technical literature, as 0.51 of the physical column volume. Relative retention times (RRT) were calculated with reference to the elution of C12 glucosyl ceramide (cerebroside, 12:0-Cb for short).

#### *4.8. Isotope Pattern Calculation*

Isotope pattern calculation was accomplished with an online freeware calculator (https://www. envipat.eawag.ch/index.php; last accession 4 November 2019). The software allows modulation of the mass resolution to envision the profile of the isotope cluster in the employed instrumental conditions. In this instance, the resolution was modulated at two different values, according to the employed mass spectrometer. A value of 500 (DM/M), slightly lower than that effective in the triple quadrupole instrument, but sufficient to match the measured ion profiles, was employed to simulate the isotopic envelopes for the first tier of measurements. A value of 10,000 (DM/M) was employed to calculate the abundances and accurate *m*/*z* of precursor and fragment ions anticipated for the identified compounds, and to compare with high-resolution measurements.
