Lipidomic and Metabolomic Signatures of the Traditional Fermented Milk Product Gioddu

Abstract

Fermented dairy products such as yogurt, kefir, and traditional cheeses are increasingly consumed worldwide for their nutritional and probiotic properties. Lipidomic profiling provides valuable insights into microbial-driven biochemical changes during fermentation. In this study, we performed a comprehensive untargeted lipidomic analysis of sheep milk and Gioddu, a traditional Sardinian fermented dairy product. Using UHPLC-QTOF-MS platform, we observed that fermentation significantly reshaped the lipidome. Gioddu samples showed higher levels of phosphatidylethanolamines (PE) and lysophosphatidylethanolamines (LPE), together with a pronounced reduction in sphingolipids (glucosylceramides, ceramides, sphingomyelins) and glycerophospholipids (phosphatidylinositols, phosphatidylserines, phosphatidylcholines) compared to sheep milk. These findings align with known enzymatic activities of lactic acid bacteria (LAB), including phospholipases A1 and A2, phosphatidylinositol-specific phospholipase C (PI-PLC), and sphingomyelinase. Fermentation also affected triglycerides, with reduced levels of FA 18:1-containing species, suggesting the selective lipolysis of monounsaturated fatty acids by microbial lipases. Complementary metabolomic profiling revealed reduced levels of simple sugars such as galactose and inositol in Gioddu samples, consistent with their use as primary carbon sources during early fermentation. Conversely, a marked accumulation of carboxylic acids (succinic, malic, hydroxyisovaleric, hydroxyglutaric, glyceric) was revealed, reflecting enhanced microbial fermentative activity. Increased levels of amino acids, including alanine, serine, proline, and ethanolamine, further highlighted active proteolysis and membrane remodeling driven by LAB metabolism. These findings show that LAB enzymes play a key role in modifying the lipidome of fermented dairy products, highlighting their metabolic flexibility and potential impact on nutritional and health properties. This integrated approach sheds new light on the metabolic plasticity of fermentative processes and underscores the value of omics-based tools in understanding traditional food systems.

1. Introduction

Fermentation, one of the oldest food preservation methods, relies on microbial carbohydrate breakdown [1] and enzymatic activity [2,3]. Fermented milk products, such as sour milk, yogurt, and cheese, originated in regions with hot climates and were valued not only for preservation [4] but also for enhanced nutrition and the mitigation of lactose intolerance [5,6].

Milk from different mammal species varies in total solids, lactose, fat, protein, and mineral content [7] and lipid composition [8], and fermented milks can be broadly classified based on their microbiological characteristics [9,10]:

-

Thermophilic fermented milk products typical of yogurt, in which the fermentation process is driven exclusively by lactic acid bacteria (LAB) (Lactobacillus and Streptococcus), that grow optimally at high temperatures (42 °C) and convert lactose into lactic acid without alcohol production. These cultures may be used in combination with mesophilic LAB (e.g., Dhai and Sky).

-

Mesophilic fermented milk products obtained by homo- and heterofermentative microorganisms of the genera Lactococcus, Lactobacillus and Leuconostoc, fermented at room temperature (<30 °C). These products include buttermilk ferments, sour creams, Ymer and stringy fermented milks from Scandinavian countries.

-

Probiotic fermented milks, which contain probiotic microbial species belonging to the Lactobacillus and/or Bifidobacterium genera, with more pronounced beneficial effects on health,

-

Acid-alcoholic fermented milks, which contain both LAB and yeasts that play an essential role in alcoholic acid-fermented milks such as Lben, kefir, and Gioddu [11,12] yeasts contribute to the production of ethanol and carbon dioxide along with lactic acid, resulting in beverages that have a unique combination of acidity and moderate alcohol content.

Among the wide variety of fermented milk products, Gioddu, also called “Miciuratu,” “Mezzoraddu,” or “Latte ischidu” (meaning “acidulous milk”), is an acid-alcoholic fermented beverage traditionally made in Sardinia (Southern Italy) using goat or sheep milk. This fermented dairy product is characterized by a firmer texture compared to cow’s yogurt, translucent white color, and acidic flavor [13]. Gioddu is the only type of fermented milk originating in Italy listed as a traditional product in the official register maintained by the Italian Ministry of Agriculture and Forestry (G.U. Repubblica Italiana no. 168, 22/07/2015 Suppl. Ord. no. 43) [12]. In ancient times, Gioddu was probably obtained through the spontaneous fermentation of milk, or by adding a piece of bread dough or a few drops of fig latex to initiate or maintain the fermentation process.

Gioddu is still traditionally produced at farm level and is sometimes sold at local markets. For this reason, the milk is pasteurized in accordance with hygiene food regulations (EU Reg 853/2004). It is then cooled to around 32–35 °C and inoculated with whey (with or without curd pieces) from raw milk cheese production. This product serves as the natural starter culture for subsequent processing, using a method known as back-slopping.

Despite the long history of consumption, only few studies have been conducted on the microbiota of Gioddu [14,15,16] providing initial insights into the microbial ecology of Gioddu, which remains an untapped source of microbial diversity. So far, only a limited number of LAB species and yeast species have been identified, with recent detection of these species using Polymerase Chain Reaction–Denaturing Gradient Gel Electrophoresis (PCR-DGGE) to analyze microbial DNA directly from Gioddu [12]. A metataxonomic analysis, conducted for the first time on Gioddu by Maoloni et al., uncovered a complex kefir-like microbiota of bacteria and yeasts. L. delbrueckii was present at high levels in all samples, along with S. thermophilus, indicating a yogurt-like symbiosis. Interestingly, was the presence of L. kefiri, suggesting the presence of bioactive compounds similar to those found in milk kefir, while the examination of eumycetes population revealed a diverse mycobiota, including potentially probiotic species such as K. marxianus [15].

While numerous studies explored the microbial ecology, nutritional value, and functional properties of fermented dairy products, much of the existing work has focused on either metabolomic or lipidomic profiling in isolation, and largely on widely consumed products such as yogurt or kefir. In contrast, there is a striking lack of integrated lipidomic–metabolomic approaches applied to traditional, region-specific fermented dairy matrices. To the best of our knowledge, no comprehensive multi-omics investigation has yet been conducted on Gioddu, a traditional Sardinian fermented milk product. To address this research gap, the present study aims to explore how the lipidome and metabolome of raw milk change during the fermentation process. Metabolomics and lipidomics are powerful analytical approaches that provide deep insight into the biochemical complexity of fermented dairy products. By profiling small molecules and lipid species, these techniques help to understand fermentation dynamics, monitor product quality, and identify bioactive compounds with potential health benefits. Their application is essential for optimizing production processes, ensuring consistency, and developing functional dairy foods with enhanced nutritional value.

To achieve the objective, samples of sheep milk and Gioddu from three different local producers were analyzed using a combined lipidomic and metabolomic approach. High-performance liquid chromatography coupled with quadrupole time-of-flight mass spectrometry (UHPLC-QTOF-MS) was employed to achieve a broad characterization of intact lipid species, while targeted and untargeted metabolomic workflows were used to capture changes in small-molecule metabolites. This integrated strategy provides a comprehensive overview of the molecular landscape, offering insights into both the lipidome and metabolome, and elucidates the impact of fermentation on the transformation of the milk matrix.

2. Materials and Methods

2.1. Chemicals

Analytical LC grade methanol, chloroform, acetonitrile, 2-propanol, acetic and formic acid and ammonium acetate and formiate, were purchased from Sigma Aldrich (Milan, Italy). Bi-distilled water, (<18 MΩ·cm at 25 °C) was obtained with a MilliQ purification system (Millipore, Milan, Italy).

2.2. Sample Production

Milk samples were collected in April 2024 from Sardinian dairy sheep raised on three agro-pastoral farms located in a homogeneous geographical area in the province of Sas-sari (Sardinia, Italy). As these were small farms, each with fewer than 200 lactating ewes, the farmers used to produce cheese and Gioddu only from the milk of their own flock. The animals grazed on pastures composed of a mixture of clover (Trifolium spp.) and ryegrass (Lolium spp.) and received a commercial concentrate supplement (approximately 800 g/day), almost always divided between the two daily milkings. A total of nine Gioddu samples and nine raw milk samples were analyzed. The samples were obtained from three different producers, with three biological replicates collected from each. Each biological replicate represented a pool of ten individual samples, in order to minimize individual variability and more accurately reflect the characteristic production profile of each producer.

To produce Gioddu (three batches per farm), the milk was collected during the morning milking. Each producer used the same batch of sheep milk for both raw and fermented samples, with each raw milk sample acting as the starting material for the corresponding spontaneously fermented Gioddu. The milk was pasteurized at 71 °C for 15 s and then cooled to approximately 35 °C. An unspecified amount of Gioddu (10–15% w/v) from previous batches was added to the milk as a starter culture (back-slopping). The milk was then incubated at 35 °C for approximately 12 h, which is typically sufficient for the pH to drop to at least 4.2. Finally, the products were stored at 4–6 °C and transported to the laboratory under the same temperature conditions. pH values on raw milk (collected before filtration and pasteurization) and Gioddu were determined by pH-meter (Crison Instruments SA, Barcelona, Spain). For each sample, the values were determined in triplicate.

2.3. Sample Preparation for Lipidomic and Metabolomic Analysis

For the preparation and purification of the lipid fraction, 10 µL of milk and Gioddu samples were extracted using the Folch method [17] which remains a gold standard for untargeted lipidomics due to its high efficiency in recovering a broad spectrum of lipids, including neutral lipids, phospholipids, and sphingolipids, from complex biological matrices [18]. This broad coverage makes it particularly suitable for comprehensive lipidomic profiling of dairy matrices, where both membrane lipids and storage lipids play crucial roles. The lipid extraction was performed adding 375 µL of a methanol and chloroform mixture (2:1, v/v) containing 5 mg/L of 2,2,3,3-D4-succinic acid for the aqueous analysis and Splash lipidomix for the lipid fraction used as the internal standard. The samples were vortexed every 15 min for 1 h, then 350 µL of chloroform and 150 µL of methanol were added. After 1 h, the samples were centrifuged (Eppendorf 5810R Refrigerated Centrifuge, Milan, Italy) at 17,700 rcf for 10 min, and 400 µL of the organic and 200 uL of aqueous layer were transferred into separate glass vials and dried under a gentle nitrogen stream. The dried chloroform phase was reconstituted with 20 μL of a methanol and chloroform mixture (1:1, v/v) and 780 μL isopropanol:acetonitrile:water mixture (2:1:1 v/v). Three technical replicates per sample were extracted. Furthermore, quality control (QC) samples were prepared by pooling equal aliquots (10 µL) from each individual sample. The pooled QCs were then subjected to the same extraction, acquisition, and data processing procedures as applied to the study samples. All samples were injected into the UHPLC-QTOF-MS apparatus and acquired in the negative ionization mode, while for positive ionization mode they were diluted in ratio 1:10 v/v. The dried aqueous phase was subjected to derivatization process and analyzed by GC-MS technique. QC samples were injected regularly throughout the analytical sequence to monitor instrument stability and reproducibility.

2.4. UHPLC-QTOF-MS Lipidomic Analysis

A 6560-LC-QTOF-MS system, coupled with an Agilent 1290 Infinity II LC system (Agilent Technologies, Santa Clara, CA, USA), was used to analyze the chloroform phase of sheep milk and Gioddu samples. For the positive ionization mode, a 2.0 μL aliquot from each sample was injected into a Kinetex C18 chromatographic column (1.6 μm, 100 mm × 2.1 mm; Phenomenex, Bologna, Italy), while for the negative ionization mode, a 10 μL aliquot was used. The column was maintained at 50 °C with a flow rate of 0.5 mL/min. The mobile phase for positive ionization mode consisted of (A) a solution of 10 mM ammonium formate in 60% Milli-Q water and 40% acetonitrile, and (B) a mixture of 90% isopropanol and 10% acetonitrile (9:1) containing 10 mM ammonium formate. For negative ionization mode, the mobile phase was identical except that 10 mM ammonium acetate replaced ammonium formate. Chromatographic separation was achieved using the following gradient: starting with 60% of phase A, decreasing to 50% in 2.1 min, and further reducing to 30% within 10 min. Mobile phase A was further reduced to 1%, and maintained for 1.9 min, and then returned to the initial conditions within 1 min.

An Agilent Jet Stream Technology source was operated in both positive and negative ionization modes with the following parameters: gas temperature at 250 °C, nitrogen gas flow at 5 L/min, nebulizer gas pressure at 20 psig, sheath gas temperature at 275 °C, and sheath gas flow at 12 L/min. The capillary voltage was set to 4000 V for positive mode and 3000 V for negative mode, with a nozzle voltage of 500 V, fragmentor voltage of 400 V, skimmer voltage of 65 V, octapole RF voltage of 750 V, and a mass range of 50–1700 m/z. Collision energy was 20 eV for positive mode and 25 eV for negative mode, with a mass precursor per cycle of 3.

The instrument was calibrated prior to analysis using an Agilent tuning solution (Agilent Technologies, Santa Clara, CA, USA) covering the mass range of m/z 50–1700, and a reference mass mix was continuously injected during the run for re-calibration. To improve sensitivity and acquire fragmentation spectra useful for identifying complex lipids, the experimental data were also obtained using the auto-MS/MS technique. This iterative method operated with a mass error tolerance of 10 ppm, a retention time exclusion tolerance of 0.2 min and a collision energy of 20 eV, where precursors selected for MS/MS fragmentation were sequentially excluded in each run.

Data acquisition was performed using the Agilent MassHunter LC/MS Acquisition software (revision B.09.00), and the results were processed with the Agilent Lipid Annotator software (version 1.0) (Agilent Technologies, Santa Clara, CA, USA) as part of a comprehensive lipidomics workflow. Data obtained from the LC-MS platform were pre-processed using Mass Profinder (version 10.0) software (Agilent Technologies, Santa Clara, CA, USA), enabling time alignment and signal deconvolution. This process generated a matrix containing all detected features across samples.

2.5. GC-MS Metabolomic Analysis

The dried residue was subjected to a two-step derivatization process. Briefly, each sample was treated with 50 µL of methoxyamine hydrochloride in pyridine (10 mg/mL) and incubated at room temperature for 17 h. Subsequently, 50 µL of MSTFA were added, and the samples were incubated at 70 °C for 1 h. Finally, 0.2 mL of n-hexane were added, and the samples were filtered through a 0.45 µm nylon membrane syringe filter. One microliter of the derivatized sample was injected in splitless mode into a 6850-gas chromatograph coupled to a 5973-mass selective detector (Agilent Technologies Inc., Santa Clara, CA, USA). The injector was maintained at 200 °C, and helium was used as carrier gas at a constant flow rate of 0.8 mL/min. Chromatographic separation was performed using a HP-5MS Ultra inert (30 m × 0.25 mm i.d., 0.25 μm film thickness; J&W Scientific, Folsom, CA, USA). The GC oven temperature was initially held at 70 °C for 1 min, then ramped at 10 °C/min to 260 °C held for 2 min, then to 280 °C at 30 °C/min held for 2 min followed by an increase to 330 °C at 50 °C/min for 5 min. Solvent delay was set at 5.20 min. The MS transfer line and ion source temperatures were set to 280 °C and 180 °C, respectively. Electron ionization (EI) was performed at 70 eV, and mass spectra were recorded over the m/z 50–550 range at a scan rate of 1.6 scans per sec. Data acquisition was carried out by integrating each resolved chromatographic peak. Metabolite identification was conducted by comparing mass fragmentation patterns and retention indices to those in the NIST 20 Mass Spectral Library (NIST, 2020). When available, identifications were further confirmed by injection of authentic chemical standards, allowing comparison of both retention time and spectral match. All detected metabolites were normalized by dividing each peak area with the area of succinic acid-2,2,3,3-d₄ (internal standard). Linear Retention Index (LRI) were calculated basing on a homologous series of alkanes (C7–C40) as external standards, to support accurate metabolite annotation. The experimental RI values obtained were compared with reference values available from the NIST database.

2.6. Data Analysis

Features were filtered based on their presence in QC samples (threshold = 40%), and the remaining features were compiled into a data matrix. With the aim to minimize the observed instrumental variation that can affect the detection of the biological variation between the different classes of samples, the data matrices were normalized using an algorithm called “Quality Control Samples and Vector Regression Support (QC-SVRC)” [19]. The SVR algorithm uses a radial-based function kernel to fix the instrumental drift within a lot using data acquired from quality control samples. Data processing was carried out using Mass Profinder 10.0 (Agilent Technologies, Santa Clara, CA, USA). Retention time alignment was performed using the internal standard signal as reference, calculating the percentage difference in retention time between the earliest and the latest eluting standard to correct for RT drifts across samples. Mass accuracy tolerance was set to ±10 ppm, and feature extraction was performed using the Recursive Feature Extraction (RFE) algorithm with a signal-to-noise threshold of 5. Normalization was applied to the 75th percentile of total signal intensity, and only features detected in at least 80% of replicates within each group were retained. Missing value imputation was carried out using the NaN (Not a Number) algorithm implemented in MATLAB (version R2022b). After correction, the quality assurance was performed on the matrix to eliminate non-specific information. This filtered matrix was then subjected to multivariate statistical analysis using SIMCA software 15.0 (Umetrics, Umeå, Sweden). Principal Component Analysis (PCA) was performed to visualize sample and variable distributions in multivariate space, followed by Partial Least Squares-Discriminant Analysis (PLS-DA) and its orthogonal extension (OPLS-DA) to classify and evaluate differences between sample groups. The quality of the models was evaluated based on the cumulative parameters R2Y and Q2Y, the latter estimated by cross validation (Figure S1). The variable importance in projection (VIP) scores in the predictive component were analyzed and only those metabolites having VIP values >1 were considered as discriminants between the classes.

3. Results and Discussion

Despite the variability of the milk and the unknown microbial composition of the starter culture, all Gioddu samples exhibited an adequate level of acidity, with pH values ranging between 3.71 and 4.13 as reported in

Table 1. pH values reported as a medium value and standard deviation (SD) for the 3 different producers (P, F and M).

	Milk		Gioddu
Samples	pH	SD	pH	SD
P	6.28	0.27	4.13	0.38
F	6.33	0.16	3.78	0.35
M	6.32	0.23	3.71	0.27

. Values below 4.2 ensure product safety and quality, particularly when high numbers of live lactic acid bacteria (LAB) and yeasts are present.

3.1. Lipidomic Analysis

The lipidomic analysis performed on sheep milk and Gioddu samples from three different producers revealed significant changes in the lipid profile as a result of the fermentation process. After deconvolution, UHPLC-QTOF-MS data processing resulted in the detection of 437 features in positive ionization mode (PIA) and 474 features in negative ionization mode (NIA). These molecular features were subsequently analyzed through multivariate statistical approaches. Principal Component Analysis (PCA) was initially employed to assess sample clustering, detect potential outliers, and explore overall patterns and trends.

PLS-DA and OPLS-DA were applied to classify samples and identify variables driving group separation. The strong predictive ability of the OPLS-DA models allowed the selection of discriminant metabolites based on VIP scores. OPLS-DA score plots are presented in Figure 1, while the relative loadings plots are reported in Figure 1C, D.

The strong classification and prediction performance of the OPLS-DA models enabled the identification of discriminant features based on their Variable Importance in Projection (VIP) scores. These key features were then annotated through MS/MS fragmentation at a collision energy of 20 eV and, where possible, confirmed using analytical standards. The list of discriminant metabolites identified in positive and negative ionization modes is reported in Table S1, while the summary graph bar of the different discriminating lipid classes are reported in Figure 2.

The results indicate that Gioddu samples exhibit higher levels of phosphatidylethanolamine and lysophosphatidylethanolamine, alongside lower levels of glucosylceramides, ceramides, sphingomyelins, phosphatidylinositols, phosphatidylserines, and phosphatidylcholines when compared with raw milk samples. Moreover, triglycerides containing FA 18:1 in their molecular structure were found to be reduced in Gioddu samples compared to raw milk, while those lacking FA 18:1 showed a corresponding higher level.

Fermented products undergo a complex biochemical process characterized by several enzymatic reactions. During fermentation, the pH typically drops, and the growth of other microorganisms is suppressed due to the production of various metabolites, including lactic acid, hydrogen peroxide, diacetyl, acetaldehyde, reuterin, and peptides [20]. The alterations observed in the lipid profile of Gioddu relative to raw milk strongly suggest the involvement of microbial metabolism, especially by lactic acid bacteria (LAB) and non-starter yeasts typically present in spontaneous or artisanal fermentation.

Gioddu samples showed elevated levels of phosphatidylethanolamines (PE) and lysophosphatidylethanolamines (LPE). These phospholipids are integral to eukaryotic cell membranes and serve important functions as substrates and intermediates in microbial lipid metabolism. The increased levels of PE and LPE observed in Gioddu samples are likely associated with the activity of phospholipases, particularly phospholipase A1 and A2 produced by LAB [21]. These enzymes hydrolyze ester bonds at the sn-1 or sn-2 position of glycerophospholipids, resulting in the formation of lysophospholipids and free fatty acids (FFA) [22]. The rise in LPE levels may thus be attributed to the enzymatic deacylation of PE.

Throughout the fermentation process, LAB encounters various stress factors, including temperature fluctuations, pH, solute concentration, and water activity. Depending on the stage of fermentation, the nature of stress along with its intensity and duration, LAB may trigger different adaptive responses. These active mechanisms, which enhance bacterial resilience to fermentation-related stress, typically involve alterations in membrane lipid composition, modifications in protein levels, and the regulation of specific gene expression [23,24]. Additionally, LAB, under acidic and fermentative conditions, often activate mechanisms to preserve membrane fluidity, such as increasing PE species, which promote membrane curvature and fluidity [25]. This lipid remodeling aligns with the upregulation observed in Gioddu.

Additionally, a notable reduction in phosphatidylinositols (PI), phosphatidylserines (PS), and phosphatidylcholines (PC) were detected in fermented samples. Rodriguez et al. (2001) [26] examined the activity of phosphatidylinositol-specific phospholipase C (PI-PLC) across various LAB strains, including members of the genera Lactobacillus, Weissella, and Enterococcus. Their findings revealed that 44% of the tested strains exhibited PI-PLC activity, indicating that LAB can hydrolyze phosphatidylinositol (PI) during fermentation, which may contribute to the observed decrease in PI content in fermented products. Moreover, bacterial phospholipases, including phospholipase D (PLD), play a role in the breakdown of phospholipids such as PC and PS in raw milk. This enzymatic action may result in reduced PC and PS levels in fermented dairy products [27]. This process could indicate microbial adaptation to the fermentation environment, facilitating membrane synthesis or energy production.

In contrast, sphingolipids, such as glucosylceramides, ceramides, and sphingomyelins, were notably reduced in Gioddu. These sphingolipids, which serve structural and signaling functions in mammalian systems, are typically present in high concentrations in raw milk. However, they undergo microbial degradation during fermentation. Many LABs possess glycosidase activity, which allows them to break down glycosphingolipids and potentially release bioactive metabolites with antimicrobial or immunomodulatory functions [28], as well as sphingomyelinase activity [29]. Ceramides are typically formed from glucosylceramides and sphingomyelins through the action of glucosylceramidase (GCase) and sphingomyelinase (SMase), so their levels might be expected to rise after LAB are added to raw milk. However, several studies reported lower ceramide concentrations in fermented milk products compared to raw milk, a difference that may be attributed to the specific environmental conditions and duration of the fermentation process, which can influence microbial enzymatic activity and lipid metabolism [30]. Acid sphingomyelinase activity can be influenced by pH. Studies have shown that ASM activity associated with a recombinant protein (rhAC) exhibits an optimal activity at pH 4.5, indicating that acidic environments can affect the enzyme’s effectiveness [31]. Additionally, research on recombinant sphingomyelinase-D production in Escherichia coli demonstrated that controlled pH conditions at 7.5 promoted the formation of smaller and more stable protein inclusions compared to uncontrolled pH conditions, indicating that pH can influence SMase conformation and enzymatic activity [32].

Changes in the triglyceride (TG) profile were also observed. Gioddu samples exhibited a reduction in TGs containing FA 18:1 (oleic acid), indicating selective hydrolysis or microbial adaptation of unsaturated fatty acids. The lipases produced during fermentation may preferentially hydrolyze TGs enriched in monounsaturated fatty acids, thereby contributing to alterations in the sensory attributes (e.g., flavor), rheological properties (e.g., viscosity), and overall nutritional composition of the product. The observed reduction in triglycerides containing oleic acid (FA 18:1) and the corresponding increase in saturated and medium-chain TGs are consistent with lipolytic activity by LAB lipases. There is considerable variability among strains and enzymes. Microbial lipases and esterases exhibit a wide range of substrate specificities: many can act on fatty acids with double bonds (such as oleic or linoleic acid), but their quantitative preference depends strongly on the enzyme’s structural features as well as on environmental factors including pH, cofactors, and the food matrix. As a result, some LAB strains display marked activity toward unsaturated fatty acids, while others are only weakly lipolytic or show greater activity toward saturated substrates [33]. Species like Lactobacillus delbrueckii subsp. bulgaricus and Streptococcus thermophilus are known to display ester bond specificity, preferentially releasing unsaturated fatty acids such as oleic acid for the synthesis of microbial membranes or for signaling functions [34,35]. The selective hydrolysis of unsaturated fatty acids, such as oleic acid, by LAB lipases provides further evidence that these enzymes are involved in microbial membrane biosynthesis or signaling during fermentation processes. Additionally, other microorganisms involved in fermentation, including yeasts, produce lipases that can modulate triglyceride composition. For instance, Saccharomyces cerevisiae expresses the Tgl1p lipase, which hydrolyzes both steryl esters and TGs, with activity influenced by pH. At pH 7.4, Tgl1p exhibits TG lipase activity, indicating that environmental conditions can modulate lipase activity in yeasts [36].

These changes indicate that microbial communities in fermented dairy products can significantly impact the lipidome through lipolytic and phospholipolytic activities. Such modifications may affect the bioavailability of lipid-derived nutrients, interact with gut microbiota, and potentially influence immune responses in the host.

3.2. Metabolomic Analysis

Untargeted metabolomic profiling of sheep milk and Gioddu showed clear clustering between raw and fermented samples, indicating a significant metabolic shift during fermentation. Multivariate analyses (PLS-DA and OPLS-DA) confirmed this separation, with the OPLS-DA model displaying good classification and predictive performance (Figure 3A), while the relative loadings plots are reported in Figure 3B.

Discriminant metabolites were identified based on Variable Importance in Projection (VIP) scores.

Features with VIP values greater than 1 were subsequently subjected to univariate analysis using ANOVA followed by Tukey’s post hoc test to confirm statistical significance among groups and the summary graph bar of the different discriminating metabolites are reported in Figure 4. While the list of metabolites identified by GC-MS is reported in Table S2.

As reported, Gioddu samples exhibited lower levels of sugars such as galactose and inositol compared to sheep milk samples, while showing higher concentrations of carboxylic acids including hydroxyisovaleric, succinic, malic, hydroxyglutaric, and glyceric acids. In addition, amino acids such as alanine, proline, serine, and ethanolamine were identified as discriminant features between the two sample types, showing significantly higher levels in Gioddu compared to raw sheep milk. The metabolomic differences observed between Gioddu and sheep milk samples reveal distinct biochemical profiles likely influenced by species-specific metabolism and microbial fermentation processes.

The observed reduction in galactose and inositol levels in Gioddu may be attributed to microbial metabolism during early fermentation, as LAB are known to utilize simple sugars as primary carbon sources. Galactose reduction likely reflects LAB metabolism during the initial acidification (estimated pH 3.8–4.2) in Gioddu production [15]. Inositol, often associated with prebiotic and membrane-modulating functions, can also be degraded by fermentative microbes, influencing its availability in the final product [37].

Conversely, the increased concentrations of organic acids such as succinic, malic, hydroxyisovaleric, hydroxyglutaric, and glyceric acids in Gioddu suggest a more active fermentative metabolism. Succinic and malic acids are typical intermediates of the tricarboxylic acid (TCA) cycle and are frequently elevated during microbial growth in fermented dairy systems [38]. In yogurt fermentation, Lactobacillus delbrueckii subsp. bulgaricus converts added fumaric and malic acid into succinic acid. Malic acid addition similarly boosted succinate levels, demonstrating that the reverse TCA cycle is active during fermentation [39].

Hydroxyisovaleric acid is a byproduct of leucine catabolism, often associated with proteolysis driven by microbial peptidases [40], whereas the role of hydroxyglutaric acid in fermented dairy systems remains less defined; its occurrence may reflect broader microbial metabolic adjustments during fermentation, including potential alterations in oxidative pathways [41]. The presence of glyceric acid may originate from microbial glycolytic activity and the transformation of glycerol, a known intermediate in lipid and carbohydrate fermentation [42].

In addition, amino acids such as alanine, serine, and proline were identified as discriminant markers. The elevated levels of alanine, serine, and proline highlight enhanced proteolysis by LAB. For instance, proteolytic activity by L. delbrueckii and S. thermophilus, identified as dominant species in Gioddu, supports the release of these amino acids during fermentation [13,43]. Alanine and serine were identified as among the most abundant free amino acids in fermented milk products inoculated with S. thermophilus in co-culture, as reported by Li et al. [44]. Their accumulation is indicative of microbial enzymatic activity, specifically transamination reactions and acidogenic metabolism during fermentation. At the end of the fermentative process, elevated levels of these amino acids serve as functional markers of active microbial metabolism, validating their role as key intermediates in amino acid catabolism and organic acid production. Moreover, caseins are known for their high proline content, making up around 16% of their amino acid composition, which sets them apart from many other milk proteins. The lack of structural rigidity makes caseins particularly susceptible to enzymatic breakdown. In particular, strains of Lactococcus lactis and Lactococcus cremoris, widely employed in dairy fermentations, produce specialized enzymes, proline-specific peptidases, that can effectively cleave at sites near proline residues. The release of free proline during fermentation is particularly relevant to the sensory profile of fermented dairy products, as proline itself is associated with a mild sweetness, while the degradation of proline-rich peptides helps to reduce bitterness; together, these processes contribute to balancing taste perception and enhancing flavor complexity in the final product [45]. As a result, these microorganisms play a key role in liberating and metabolizing proline during the fermentation process, contributing to both the proteolytic profile and the sensory characteristics of the final product [46].

Furthermore, ethanolamine is a key amine derived from membrane lipids, primarily phosphatidylethanolamines (PE), which are major components of microbial cell membranes. During fermentation, microorganisms often remodel their membrane composition to adapt to environmental stresses, such as changes in pH, osmotic pressure, or nutrient availability [47]. A decrease in PE together with a concomitant increase in free ethanolamine is consistent with microbial enzymatic cleavage of PE. Extracellular or released phospholipases (PLA/PLD/lysophospholipases) deacylate and cleave the phospholipid headgroup to give glycerophosphoethanolamine (GPE), and glycerophosphodiester phosphodiesterases (GDPD/GlpQ-type enzymes) hydrolyze GPE to glycerol-3-phosphate and ethanolamine [48,49]. Such GDPD/GlpQ activities are well characterized in bacteria and are known to liberate small alcohol/amine headgroups (including ethanolamine) from glycerophosphodiesters, providing a direct mechanistic link between PE loss and ethanolamine gain [49]. During complex fermentations, lactic acid bacteria together with yeasts release lipolytic and phospholipolytic enzymes, particularly during growth and autolysis. These activities enhance the breakdown of membrane phospholipids, providing substrates that can lead to ethanolamine accumulation in fermented dairy products [50].

A scheme summarizing the proposed microbial lipidomic-metabolomic remodeling is shown in Figure 5.

Such dynamic changes might contribute to the distinctive metabolic profile of fermented products like Gioddu, underscoring a possible link between lipid metabolism and microbial activity.

Taken together, the metabolic fingerprint of Gioddu is characterized by enhanced microbial fermentation of carbohydrates, lipids, and proteins, distinguishing it from raw sheep milk and providing potential markers for product authentication and quality assessment.

4. Conclusions

The comprehensive lipidomic and metabolomic analyses performed in this study provided valuable insights into the biochemical transformations that occur during the fermentation process of ovine milk into Gioddu. By comparing raw sheep milk and Gioddu samples, we were able to characterize the metabolic shifts driven by microbial activity and elucidate the impact of fermentation on the milk’s molecular composition. Microbial fermentation profoundly remodels the lipidome and metabolome of Gioddu compared to raw sheep milk. Lipidomic changes include increased phosphatidylethanolamines and lysophosphatidylethanolamines, alongside reductions in sphingolipids and other glycerophospholipids, reflecting adaptive membrane remodeling. Metabolomic shifts involve the depletion of sugars such as galactose and inositol, with elevated levels of organic acids and amino acids, highlighting active microbial metabolism through carbohydrate fermentation and proteolysis. This highlights the importance of integrating lipidomic and metabolomic approaches with microbiological and sensory analyses to gain a comprehensive understanding of the biochemical transformations occurring in traditional fermented dairy products like Gioddu.

Moreover, this study sheds light on the dynamic interplay between microbial communities and lipid metabolism during fermentation, offering valuable insights into the molecular basis of artisanal dairy fermentations. These results suggest that specific lipid and metabolite changes could influence product quality, flavor development, and potential health benefits, highlighting the value of these analyses for artisanal dairy production and traditional food preservation. Future research should aim to correlate lipid and metabolite alterations with specific microbial strains and detailed sensory attributes, providing deeper insights into the molecular mechanisms driving fermentation and enabling the optimization of fermentation processes for improved quality and functionality of fermented dairy foods.

Limitations. While the present study provides novel insights into lipid and metabolite transformations in Gioddu, it is not without limitations. The sample size was relatively small, limiting the statistical power for detecting subtle changes. Additionally, microbial community composition was not profiled in parallel, which prevented direct correlation between specific taxa and observed lipid remodeling. Future studies integrating high-throughput microbial profiling with lipidomic and metabolomic analyses would strengthen the mechanistic understanding of these processes.

Practical Implications. The findings of this study have practical implications for artisanal dairy production. Understanding microbial lipid remodeling can help producers optimize fermentation conditions to enhance nutritional and sensory qualities, while maintaining traditional characteristics. Moreover, identifying bioactive lipid and metabolite profiles could inform consumers about potential health benefits of fermented dairy products, supporting informed dietary choices.