Integration of Global Lipidomics and Gonad Histological Analysis via Multivariate Chemometrics and Machine Learning: Identification of Potential Lipid Markers of Ovarian Development in the Blue Mussel (Mytilus edulis)

Abstract

Background/Objectives: Gonad histological analysis (GHA) is the traditional method for assessing the gonad maturation status of blue mussels (Mytilus edulis). GHA has some operational disadvantages, such as limited processing outputs, subjectivity in the assessment of transitional stages of gonadal maturation and the need for experienced and trained operators. Lipids could become important indicators of gonadal maturation as they cover many essential functions during such processes in mussels. In this work, blue mussel ovary (BMO) ultrastructure is integrated with liquid chromatography coupled with mass spectrometry (LC-MS) lipidomics fingerprinting to identify suitable markers for ovarian maturation through the application of chemometrics and machine learning approaches. Methods: BMOs are classified here as ripe or non-ripe by means of GHA and the gamete volume fraction (GVF). Receiving operating characteristic (ROC) curves were used to classify the results of the different statistics according to their area under the curve (AUC), and the functional role of important lipids was assessed by lipid ontology enrichment (LiOn) analysis. Results: This approach allowed for the selection of a panel of 35 lipid molecules (AUC > 0.8) that can distinguish non-ripe from ripe BMOs. Ceramide phosphoethanolamine (CerPE) 40:2 was the molecule with the highest classification ability (AUC 0.905), whereas glycerophosphoserine (PS) was the class mostly changing between the two groups. LiOn analysis indicated significant differences in the functional roles of these lipids, highlighting enrichment terms associated with membrane lipids, lysosomes and highly unsaturated triglycerides (TGs) in non-ripe ovaries, whereas terms associated with storage lipids and low-saturated TG characterised ripe BMOs.

1. Introduction

Mussels are important species for marine ecosystems as they are recognised as habitat engineers and provide valuable ecosystem services for coastal areas and their economies [1,2]. Mussels are valuable aquaculture resources characterised by an extremely low environmental footprint [3,4] and high nutritional properties [5], and their demand as food products is projected to increase in the future’s net-zero society [6]. Unfortunately, despite their prospective importance, mussel production in Europe has stagnated since 1999 [7], and the expansion of the mussel aquaculture sector is hampered by various factors, including climate change [8]. Therefore, it is important to increase our knowledge of mussel physiology and biochemistry and to identify further useful tools to monitor and predict the ability of mussels to adapt and thrive in future environmental scenarios.

Gonadal development is the main physiological process that runs through the annual life cycle of temperate mussels [9]. Blue mussels (Mytilus edulis) have an annual reproductive cycle with an accumulation of energy resources in late summer and autumn, a gonadal maturation process that extends through winter, and one or more spawning events in spring and early summer [10,11,12,13,14]. During maturation, gonads proliferate on the mantle, which undergoes a series of ultrastructural and biochemical reorganisations. In mature mussels, the mantle can account for 50% of the total body dry weight [14]. In resting females, the mantle consists of a single-layered cuboidal epithelium that is ciliated on the inner surface and composed of adipogranular (ADG) and vesicular connective tissue (VCT) cells [15]. ADG and VCT cells support oocyte development by providing glycogen and lipids, respectively [12]. The oocytes increase in numbers and mature, accumulating lipid droplets that form the yolk reserves of the embryos [16]. Glycogen is consumed during the development process, while lipids are accumulated in the oocytes [17,18]. This process leads to the annual accumulation patterns of lipids in several bivalve species, including Magallana gigas [19,20,21], Ostrea edulis [22], M. edulis [17,23], Mytilus galloprovincialis [24], Modiolus barbatus [25], Mimachlamys crassicostata [26], Ruditapes decussatus and Ruditapes philippinarum [27].

Detailed information about the maturation of the gonads is important for the industrial production of mussels in hatcheries, where the mussels are artificially conditioned (so-called broodstock conditioning) to maximise gamete production. Mussels do not store their oocytes for prolonged periods of time, as it is their physiology to partially or completely spawn within a few weeks of sexual maturity [28]. Missing the right time window for gonad development can lead to unnecessary resource consumption, uncontrolled spawning or the occurrence of pre-spawning atresia [28,29,30].

Various histological, biochemical and molecular methods have been used to assess gonadal maturation in bivalves. Historical approaches to study gonad maturation in mussels included the use of allometric ratios as a measure of condition. The condition index, which measures the ratio of soft tissue to shell [31], and the gonadosomatic index, which measures the ratio of mantle to soft tissue [32], are two commonly cited examples in the literature. Traditional techniques also include gonadal smears and gonadal histological analysis (GHA) [33,34]. The fastest method to obtain information on the gonadal status of mussels is the gonadal smear, although the loss of tissue structure due to sample preparation makes it difficult to distinguish between different stages of maturity. The ultrastructure of the tissue is well preserved in the GHA, making it the most important technique for analysing the reproductive cycle in bivalves [10,12,30,34,35]. The limitations of GHA are the limited sample size (e.g., the number of samples that can be processed per time unit), the difficulty in distinguishing transitional stages of development, and a certain degree of subjectivity that requires experienced operators in classifying the degree of development of gonads [36,37]. GHA has also been coupled with stereological indices, leading to the development of several quantitative indices of gonadal maturation, such as gamete volume fraction (GVF) or oocyte cytoplasmic area, overcoming some of the limitations of classical GHA [11,38,39,40,41]. Hines and co-workers compared classical histological approaches with metabolic and molecular techniques and showed that the latter two methods are advantageous in distinguishing sex in dormant or spent mussels [42].

Lipids play an important role in mussel gonad maturation, as several authors have reported direct changes in lipid composition as a result of the maturation process [18,21,43,44,45]. Triglycerides (TGs) were the major lipid class that changed during gonadal development in the ovaries of both Pecten maximus and M. gigas, accumulating in mature individuals [21,43]. Martinez-Pita and co-workers found no changes in the composition of lipid classes between the different stages of ovarian development in M. galloprovincialis and Donax trunculus [18,44]. When analysing fatty acids (FA), these authors observed an increase in FA18:4n-3 in the mature females of both species, and of FA18:0, FA18:3n-3 and docosahexaenoic acid (DHA, FA22:6n-3) specifically in mature female mussels. The complexity of observing a direct effect of gamete development on gonad lipid composition may be related to the seasonal behaviour of the gonad development in bivalves, which tends to span over several months [11,12,13,32]. Furthermore, gonads are also a non-uniform tissue, as several developmental stages (or regressions) can occur simultaneously in the same organism [18,46], which greatly increases the difficulty of such a task.

Lipidomics and the use of liquid chromatography coupled with mass spectrometry (LC-MS) could offer several advantages over conventional lipid analysis techniques when analysing a complex tissue such as bivalve gonads. First, the study of lipids at the molecule level combines information from FA and lipid class composition analysis, revealing patterns that are otherwise lost when relying on only one of these two traditional lipid profiling techniques [47]. Second, modern MS platforms are characterised by higher sensitivity than traditional detectors used in lipid class analysis, providing an opportunity to evidence and extend existing knowledge on the role covered by minor, uncommon and poorly understood lipid classes in the ovarian maturation of mussels. To date, rare are the studies investigating gamete development in marine invertebrates using LC-MS. One of the first examples of this type of study comes from Chansela and coworkers who investigated the lipid composition of Paeneus merguirensis ovaries using a mass spectrometry imaging approach [48]. The authors showed how follicles from the same histological preparation, but with different maturation stages, differed in their composition of major phosphocholine (PC) and TG molecular species. The results of this study demonstrate the presence of lowly unsaturated PC species (PC 32:1) and highly unsaturated TGs (58:9) especially in early-development gonadal areas. However, PC and TG molecular species associated with essential polyunsaturated FA such as eicosapentaenoic acid (EPA, FA20:5n-3), arachidonic acid (AA, FA20:4n-6) and DHA were only abundant in mature follicles.

More recently, the lipidome of the gonads of M. edulis has been characterised in detail, focusing on sex-specific differences between testis and ovaries [49]. In the present study, we aimed to build on this previous work and continue here by analysing the blue mussel ovarian (BMO) lipidome by integrating GHA with LC-MS global untargeted lipidomics through a comprehensive statistical approach, incorporating unsupervised and supervised multivariate chemometrics and machine learning (RandomForest—RF [50]), applied here to highlight potential lipid markers that distinguish between histologically ripe (stage III) and non-ripe BMOs (stages I–II).

2. Materials and Methods

2.1. Mussel Collection and Sample Preparation

Commercial-size blue mussels (Mytilus edulis, average shell size: 5.6 ± 0.6 cm) were obtained from local growers on the West Coast of Scotland (Inverlussa Marine Services—www.inverlussa.com, Loch Spelve, Isle of Mull, UK) in February 2017. The mussel rearing and processing (sizing, feeding, aquarium rearing conditions) used in this study has been described in a previous publication [49], and details are provided in Method Supplement S1. Each adult was opened ‘book-like’ by cutting the hinge and abductor muscle with a scalpel and carefully removing the gills with bow scissors. The left valve mantle lobe was dissected and placed in a 1.5 mL tube (Eppendorf), snap-frozen in liquid nitrogen and stored at −80 °C for lipid analysis. The right mantle lobe was excised and stored in 7 mL Bijoux tubes (VWR) in 4% buffered seawater formalin for GHA.

2.2. Histological Analyses

The sex of each adult animal was determined by GHA after haematoxylin and eosin (H&E) staining, the standard method for sex determination and assessment of gonadal maturation [30,35,42,51]. H&E-stained slides were photographed with an Axioshop 2 inverted microscope equipped with an AxioCam MRc5 camera (CarlZeiss, Cambridge, UK) and acquired with the corresponding AxioVision 4.7 software. GHA scores were assigned according to the ultrastructure of the gonads using a 4-staging classification system [10,12,52]. The reference samples used in the gonad classification system are shown in Method Supplement S2a. Three different sections (anterior, medial and posterior) of the mantle of each mussel were scored to distinguish between ripe (stage III, N = 25) and non-ripe (stages I-II, N = 26) ovaries. A distinction between vitellogenic and atretic/degenerating gonads was made according to the criteria reported in [16,28] as described in Method Supplement S2b. When different stages occurred simultaneously in a single section, the staging decision was based on the condition of the largest part of the preparation [18]. A stereological variable with which to quantify the percentage of mantle area covered by gametes (gamete volume fraction—GVF) was calculated through an image analysis of the GHA preparations using ImageJ software (version 1.52r; www.imagej.nih.gov) via the colour threshold method (see Method Supplement S2c for details). The GVF was calculated as the percentage between the total mantle surface area and the mantle surface area covered with gametes. The relationship between GHA and the GVF is shown in Appendix A (Figure A1). All GHA and GVF scores are provided in Supplementary Data S1.

2.3. Lipid Extraction

Frozen and dissected BMOs were freeze-dried (ẞ18 LO plus, Christ) and ground to a fine powder in liquid nitrogen; ≈0.05 g samples of dry BMO powder were individually homogenised in HPLC water (Fisherbrand) for 1 min in glass potters (1:8 weight–volume) on ice. Aliquots of the homogenate (20 µL) from each gonad were pooled and used to form the biological quality control (BQC) samples. Lipid extraction was performed according to [53] as described in [49]. Total lipids were extracted from 50 µL of BMO homogenate and BQC with 6 mL of 2:1 chloroform–methanol (both Fisherbrand) in pre-combusted (450 °C, 5 h) 20 mL screw-cap vials (Pyrex). The mixture was shaken vigorously, each vial was flushed with nitrogen, and the mixture was then extracted on ice for 1 hr under dark conditions. The organic extract was separated from the pellet by centrifugation (350× g, 5 min, 4 °C) and collected in a clean tube. The pellet was re-extracted with a additional 3 mL of chloroform–methanol. A 0.9% KCl solution (VWR) was added to the organic extract to obtain a final ratio of 2:1:0.8 chloroform–methanol–KCl, and the polar and organic phases were separated after a second centrifugation step (350× g, 5 min, 4 °C). The organic layer was recovered and evaporated under nitrogen flow (NVap, Organomation). The dried lipid extracts were weighed (0.0001 g accuracy) and resuspended in 0.5 mL chloroform, yielding the BMO total lipid extract (TLE). The fatty acid analysis of the BMOs is reported in Supplementary Data S2a.

2.4. LC-MS Untargeted Lipidomics

Global lipidomics analyses of BMO TLE were performed by liquid chromatography coupled with high-resolution mass spectrometry (LC-MS) via a binary ultra-high pressure liquid chromatography (UHPLC) system (1250 Accela, ThermoFisher) connected to an electron spray ionisation (ESI) source and an Orbitrap mass analyser (Exactive, ThermoFisher). Separation was performed on a C18 Hypersil Gold, with a 100 × 2.1 mm column and 1.9 nm particle size (ThermoFisher, Waltham, MA, USA), maintained at 50 °C. The binary solvent system comprised a constant flow rate of 400 μL min⁻¹ with a gradient as described in Supplementary Table S1. Water and acetonitrile were HPLC-grade and purchased from Fisherbrand; isopropanol was LC-MS-grade (Hypergrade LiChrosolv, Merck, Gilligham, UK), while ammonium formate and formic acid were both LC-MS-grade and purchased from Sigma-Aldrich (St. Louis, MI, USA). TLEs were resuspended in 3:1 methanol–chloroform at a concentration of 1 mg mL⁻¹ for a 3 μL injection volume.

All samples were analysed in positive (POS) and negative (NEG) ion mode over the m/z range of 250–2000 Da at a resolution of 100,000 FWHM and a scan speed of 1 Hz. Mass error was kept below 5 ppm by routine calibrations for both polarities using a calibration solution (Pierce™ LTQ ESI calibration solutions, ThermoFisher). Chromatograms and mass spectra were reviewed and integrated using Xcalibur software (version 3.2, ThermoFisher, Waltham, MA, USA).

2.5. Data Processing

Raw LC-MS data were processed using Progenesis QI (Nonlinear Dynamics, Waters). Lipids were identified using the exact mass of the precursor ion (MS’, 5 ppm Δm/z), and with the exception of CAEPs (ceramide aminoethyl phosphonates [54]), lipid IDs were reported according to the LIPID MAPs nomenclature standard system [55,56]. Raw data from LC-MS lipid adducts for POS and NEG were experimentally evaluated using the lipid standard mixture (Supplementary Table S2) purchased from Avanti Polar Lipids (www.avantilipids.com) and Matreya (www.matreya.com). Lipid IDs were obtained by searching for exact masses in the lipid mass spectra libraries LIPID MAPS (www.lipidmaps.org—accessed on 10 October 2024), the Human Metabolome Data Base (HMDB) (www.hmdb.ca—accessed on 10 October 2024) and Metlin (www.metlin.scripps.edu—accessed on 13 March 2020), as well as from an in-house lipid database derived from lipidomics studies on bivalves and other marine invertebrates, which served as a tool for the identification of unusual marine lipids [54,57,58,59,60,61,62,63,64,65]. Isomeric lipid molecules resolved by reverse phase separation (same exact mass of 5 Δppm, but with a different retention time) are distinguished with a lowercase letter after the lipid ID (i.e., PC 38:5a and PC 38:5b). The full set of lipids identified in the BMO and used in this study is included in Supplementary Data S1. All LC-MS raw data files and experimental metadata were deposited in the Metabolights [66] data repository (www.ebi.ac.uk/metabolights/MTBLS8492—accessed on 10 October 2024) and are available under study identifier MTBLS8492.

2.6. Data Analysis

Statistical analysis was performed with the statistical platform R (version 4.2.0). Data are presented as mean ± standard deviation unless otherwise stated. Statistical differences were considered significant at p < 0.05.

Peak intensity tables (PITs), created from the LC-MS data obtained by POS and NEG ionisation modalities, were manually sorted to remove contaminant masses. Further data filtering included removing features with >30% missing values, with the remaining missing values replaced with a small value (half of the minimum intensity value) and characteristics with low repeatability removed by filtering using BQC and the interquartile range. The filtering of the raw data was performed using the filtering functions of the R-based package ‘MetaboAnalystR’ [67]. At this stage, POS and NEG data were merged into a single dataset to avoid redundancy in the analysis. Duplicate lipids (i.e., observed in the POS and NEG data) were manually checked for their retention time between the different samples; if they matched, the ionisation polarity that gave the best results in terms of signal for that specific lipid class was retained, assessed using as reference the lipid standard mixture.

A PIT was autoscaled (mean centred and divided by each lipid standard deviation), and the preliminary approach used to highlight lipids that differ between ripe and non-ripe BMOs involved principal component analysis (PCA) obtained from the R package ‘factoextra’ [68] as an unsupervised statistical model [69]. This approach was used to highlight the presence of outliers; a brief discussion of the results can be found in Appendix A. Other important markers were identified with supervised statistical methods on Pareto-scaled data (mean centred and divided by the square root of the standard deviation of each lipid). The Pareto scaling of the data was used to include information regarding molecule abundances that would otherwise be lost by autoscaling [70]. The first approach used in biomarker discovery involved the identification of significant variables via volcano plots. A volcano plot was calculated via the R package ‘MetaboAnalystR’ [67] and plotted via the package ‘EnhancedVolcano’ [71]. For the volcano plot, the lipids were filtered by a fold change (FC) > 1.3 following a Wilcoxon rank-sum test with a false discovery rate (FDR [72]) and adjusted p-value < 0.05. Orthogonal partial least squares discriminant analysis (OPLS-DA, [73]), from the ‘MetaboAnalystR’ package, was the second supervised method used to identify lipids associated with the two gonad maturation conditions. The performance of OPLS-DA was assessed by 10-fold leave-one-out cross-validation (LOOCV), and model fit was tested for using a 2000-fold permutation test (see Appendix A [74]). Important lipids were identified in a Sigma plot, and lipids characterised by a p(corr) < −0.6 and >0.6 were retained and considered important in distinguishing ripe and non-ripe BMOs. RandomForest (RF) [50] was the last method used here. RF is a machine learning algorithm that was trained as shown in [75] and was obtained from the package ‘RandomForest’ [76]. In this study, RF was used as a classification method (RFclass) using the grouping variables (ripe/non-ripe) and as a regression method (RFreg) considering the GVF. In both cases, the best 10% of RF classification (lowest out-of-bag error rate—OOB-ER) and RF regression (highest percentage of variability explained—PVE) were considered important markers. A backward purging approach for the top 5% of lipids for both RFclass and RFreg was used to identify the most important molecules detected by the two approaches (see Method Supplement S2d for details on model training). The results of all four statistical approaches were compared using a Venn diagram and ranked according to their ROC (receiving operating characteristic) curves and AUC (area under the curve) values. ROC and AUC values were calculated using the biomarker discovery function of the MetaboAnalyst package. Details of the full ROC and AUC analyses of all biomarkers identified in this section can be found in

Table A1. Lipid molecule markers of gonad ripeness identified witg the supervised statistical approaches in BMO and ranked according to receiving operating curves (ROCs) and area under the curve (AUC). Isomeric lipid molecules resolved by reverse phase separation (same exact mass of 5 Δppm, but different retention time) are distinguished with a lowercase letter after the lipid ID. Pval: Wilcoxon rank test with p-value adjusted for false discovery rate [72]; FC: fold change; r: Spearman rank correlation coefficient calculated on gamete volume fraction (GVF). Computed via R package ‘MetaboanalystR’. Average calculated on Pareto-scaled data ± 95% CI (N = 26, non-ripe and N = 25, ripe). In Bold: panel of potential lipid biomarkers (AUC > 0.8).

Lipid ID	AUC	Pval	FC	r	Volcano	OPLS-DA	RFclass	RFreg	Not Ripe	Ripe
									Average ± CI (95%)	Average ± CI (95%)
CAEP.35.3	0.74	<0.01	0.39	−0.48		x		x	1855.96 ± 247.94	1393.62 ± 125.7
CAEP.42.2	0.81	<0.01	1.58	−0.60		x	x	x	47.59 ± 9.06	20.67 ± 7.5
CAEP.42.3	0.8	<0.01	3.13	−0.56	x		x	x	36.95 ± 11.25	11.19 ± 6.08
Cer.36.2a	0.58	0.09	0.21	−0.10			x		267.81 ± 60.62	224.45 ± 23.52
Cer.38.1	0.76	<0.01	−0.99	0.38			x		47.24 ± 13.18	75.37 ± 14.06
Cer.42.0	0.76	<0.01	−0.5	0.51		x			61.06 ± 8.32	87.2 ± 11.11
CerPE.34.1	0.76	<0.01	0.57	−0.43			x		450.48 ± 85.69	277.56 ± 41.11
CerPE.34.2b	0.75	<0.01	0.58	−0.54				x	128.16 ± 22.58	84.28 ± 13.44
CerPE.34.3a	0.84	<0.01	0.98	−0.67		x	x	x	95.62 ± 20.84	48.43 ± 8.13
CerPE.35.3/2	0.87	<0.01	1.08	−0.67		x	x	x	132.82 ± 29.82	55.95 ± 7.87
CerPE.40.1	0.81	<0.01	1.17	−0.57		x	x	x	34.15 ± 7.02	13.71 ± 5.25
CerPE.40.2	0.9	<0.01	1.48	−0.74	x	x	x	x	46.4 ± 7.88	17.69 ± 4.48
CerPE.42.1	0.8	0.01	5.5	−0.53	x		x	x	7.66 ± 5.08	0.23 ± 0.45
CerPE.d44.2	0.79	<0.01	0.73	−0.52			x		52.96 ± 10.36	31.51 ± 4.19
CL.88.21	0.56	0.68	−0.3	−0.05				x	39.63 ± 15.32	44.23 ± 15.59
DG.32.1	0.59	0.2	−0.21	0.17				x	1223.06 ± 217.88	1396.08 ± 272.03
DG.33.2	0.56	0.43	−0.13	0.18				x	254.51 ± 47.85	270.76 ± 52.22
LPC.22.2	0.81	<0.01	−1.63	0.41	x		x		115.47 ± 39.6	308.4 ± 101.03
LPC.22.5	0.76	<0.01	−2.39	0.42	x				65.26 ± 37.91	244.7 ± 114.32
LPC.22.6	0.72	<0.01	−1.86	0.41	x		x		131.28 ± 61.4	497.37 ± 246.18
LPE.22.6	0.61	0.22	−0.73	0.22	x		x		27 ± 10.79	38.01 ± 18.64
PA.38.6	0.82	<0.01	5.66	−0.54	x		x	x	12.45 ± 6.38	1.46 ± 1.54
PA.40.6	0.86	<0.01	2.85	−0.70	x	x	x	x	33.14 ± 12.05	8.31 ± 3.64
PA.44.4	0.81	<0.01	−1.76	0.63			x	x	102.52 ± 41.73	226.25 ± 58.24
PA.O-38.2/P-38.1	0.74	0.01	1.36	−0.48			x	x	32.04 ± 6.9	20.2 ± 7.86
PA.O-38.5/P-38.4	0.72	<0.01	4.13	−0.41	x	x			6.66 ± 3.71	1.22 ± 1.53
PA.O-38.6/P-38.5	0.85	<0.01	2.12	−0.68	x	x	x	x	46.5 ± 13.09	15.64 ± 4.86
PC.28.0	0.84	<0.01	1.35	−0.55	x	x	x	x	426.29 ± 119.56	140.95 ± 36.64
PC.29.0	0.79	<0.01	0.94	−0.42		x			326.9 ± 78.46	150.35 ± 29.08
PC.30.0	0.85	<0.01	1.37	−0.60	x	x	x	x	1149.26 ± 291.11	396.62 ± 108.82
PC.31.0	0.76	<0.01	0.66	−0.36		x	x		706.06 ± 121.33	419.14 ± 60.12
PC.32.0	0.78	<0.01	0.95	−0.51		x	x		1181.4 ± 230.72	572.94 ± 110.41
PC.42.3	0.73	0.1	−0.49	0.40			x	x	435.88 ± 121.3	582.86 ± 82.68
PC.42.4	0.72	<0.01	−0.53	0.30		x	x		76.01 ± 13.45	104.97 ± 12.52
PC.O-29.0	0.8	<0.01	1.57	−0.51	x	x	x		274.36 ± 86.89	97.24 ± 36.48
PC.O-30.0	0.8	<0.01	1.48	−0.54	x	x	x		862.53 ± 256.36	318.65 ± 112.49
PC.O-30.1/P-30.0	0.72	<0.01	0.6	−0.39		x	x	x	479.19 ± 98.68	292.4 ± 56.3
PC.O-30.2/P-30.1	0.69	<0.01	0.68	−0.31			x		489.48 ± 140.93	197.21 ± 62.1
PC.O-31.0	0.8	<0.01	1.37	−0.55		x	x	x	68.9 ± 21.25	41.64 ± 9.75
PC.O-31.1/P-31.0	0.75	<0.01	0.85	−0.45		x			313.33 ± 66.93	170.76 ± 36.29
PC.O-32.0	0.8	<0.01	1.35	−0.56	x	x	x	x	923.78 ± 291.49	351.64 ± 111.32
PC.O-38.1/P-38.0	0.76	<0.01	−0.69	0.51				x	400.67 ± 74.02	634.91 ± 112.22
PC.O-38.5/P-38.4	0.68	0.02	−0.37	0.33				x	3158.87 ± 452.94	4126.82 ± 599.14
PC.O-39.2/P-39.1	0.71	0.02	−0.5	0.41				x	587.31 ± 109.84	964.5 ± 162.84
PC.O-40.2/P-40.1	0.78	<0.01	−0.73	0.51			x	x	587.31 ± 109.84	964.5 ± 162.84
PC.O-42.10/P-42.9a	0.81	<0.01	−1.43	0.14	x	x	x	x	145.61 ± 41.99	341.56 ± 83.65
PE.36.3a	0.65	0.05	0.16	0.26			x		33 ± 6.89	43.8 ± 10.85
PE.38.4a	0.57	0.28	0.1	−0.42			x	x	102.72 ± 14.62	96.79 ± 11.17
PE.37.4	0.72	0.01	1.19	−0.37				x	33.96 ± 9.79	17.15 ± 5.86
PE.39.5	0.76	<0.01	0.72	−0.48			x	x	75.17 ± 13.77	39.58 ± 7.07
PE.41.5	0.64	0.01	0.62	−0.28				x	33.31 ± 12.76	16.49 ± 4.91
PE.42.9c	0.56	0.59	−0.22	0.23				x	185.63 ± 49.07	193.23 ± 41.74
PE.O-34.3/P-34.2	0.83	0.01	0.98	−0.51			x	x	53.51 ± 20.01	24.5 ± 4.84
PE.O-35.3/P-35.2	0.77	0.01	0.97	−0.46				x	40.95 ± 14.52	18.98 ± 4.37
PE.O-36.1/P-36.0b	0.87	<0.01	1.63	−0.74	x	x	x	x	65.23 ± 19.65	21.09 ± 4.88
PE.O-36.1/P-36.0c	0.77	<0.01	0.88	−0.58				x	91.33 ± 25.83	41.6 ± 7.31
PE.O-36.2/P-36.1b	0.85	<0.01	0.96	−0.62		x	x	x	94.97 ± 22.76	47.13 ± 7.8
PE.O-36.2/P-36.1c	0.88	<0.01	0.97	−0.73			x	x	251.77 ± 50.85	121.82 ± 15.8
PE.O-36.3/P-36.2	0.85	<0.01	0.61	−0.58			x	x	308.04 ± 55.87	193.04 ± 19.3
PE.O-37.2/P-37.1	0.86	<0.01	0.99	−0.71		x	x	x	231.1 ± 39.56	121.75 ± 21.86
PE.O-37.3/P-37.2	0.84	<0.01	0.51	−0.55		x	x	x	428.34 ± 49.7	300.34 ± 26.36
PE.O-37.4/P-37.3	0.87	<0.01	0.74	−0.40		x	x	x	84.17 ± 14.05	50.25 ± 6.82
PE.O-38.2/P-38.1	0.8	<0.01	0.63	−0.62			x	x	627.28 ± 92.01	407.74 ± 58.2
PE.O-38.3/P-38.2	0.78	<0.01	0.38	−0.52			x		1411.67 ± 151.08	1084.61 ± 86.66
PE.O-39.2/P-39.1	0.73	<0.01	0.51	−0.47				x	125.68 ± 18.84	92.57 ± 15.07
PE.O-39.3/P-39.2	0.8	<0.01	0.45	−0.56			x	x	360.03 ± 39.86	264.36 ± 23.89
PE.O-39.4/P-39.3	0.73	<0.01	0.38	−0.38			x	x	194.69 ± 27	146.64 ± 13.67
PE.O-40.7/P-40.6b	0.67	0.07	−0.24	0.35				x	886.89 ± 125.56	1034.58 ± 117.48
PE.O-41.4/P-41.3	0.65	0.04	0.25	−0.33				x	110.58 ± 15.74	92.1 ± 9.6
PG.34.1	0.86	<0.01	2.19	−0.70	x	x	x	x	25.61 ± 8.8	7.32 ± 2.83
PI.40.3	0.62	0.05	−0.28	0.21			x		294.18 ± 38.48	350.62 ± 36.02
PI.40.4	0.65	0.04	−0.39	0.32			x		266.66 ± 45.88	331.8 ± 43.8
PI.40.8	0.71	0.01	−0.39	0.47			x	x	118.42 ± 14.83	147.88 ± 16.72
PI.41.5	0.6	0.26	−0.2	0.33				x	102.64 ± 16.24	115.25 ± 13.04
PI.41.6	0.74	<0.01	−0.49	0.54				x	103.73 ± 14.79	140.82 ± 13.95
PI.42.3	0.79	<0.01	−0.95	0.53			x	x	45.08 ± 10.05	78.71 ± 12.29
PI.42.10a	0.71	0.02	−0.71	0.43				x	27.67 ± 6.84	41.65 ± 10.88
PI.44.7	0.61	0.03	−0.61	0.22				x	11.63 ± 4.06	26.72 ± 12.24
PI.O-38.5/P-38.4a	0.76	<0.01	1.03	−0.59				x	55.34 ± 11.27	30.46 ± 8.17
PI.O-40.2/P-40.1	0.73	0.01	−1.37	0.35	x				16.8 ± 6.25	44.17 ± 17.7
PS.34.1	0.8	<0.01	6.82	−0.54	x		x	x	13.17 ± 7.65	0.53 ± 0.75
PS.36.2	0.81	<0.01	7.06	−0.56	x		x	x	16.21 ± 10.11	0.14 ± 0.22
PS.38.2a	0.64	0.1	−0.42	0.32				x	55.04 ± 12.89	68.49 ± 13.29
PS.38.4	0.64	0.03	1.55	−0.35				x	27.56 ± 19.08	5.6 ± 1.94
PS.38.5	0.89	<0.01	1.61	−0.71	x	x	x	x	215.46 ± 51.5	66.38 ± 17.26
PS.38.6a	0.84	<0.01	1.23	−0.60		x	x	x	164.04 ± 40.39	66.94 ± 16.21
PS.39.6	0.82	<0.01	1.21	−0.55			x	x	96.45 ± 35.18	39.88 ± 10.52
PS.40.2	0.83	<0.01	0.75	−0.68			x	x	157.85 ± 32.77	91.97 ± 13.65
PS.40.5	0.82	<0.01	1.52	−0.67				x	41.86 ± 11.05	18.17 ± 5.55
PS.40.6	0.89	<0.01	1.06	−0.74		x	x	x	329.08 ± 66.25	155.36 ± 27.89
PS.40.8a	0.78	<0.01	4	−0.46	x	x			46.14 ± 17.07	9.88 ± 4.45
PS.42.3b	0.6	0.26	0.11	−0.25				x	82.48 ± 17.25	72.17 ± 12.55
PS.42.4	0.61	0.03	0.3	−0.23				x	265.25 ± 70.08	188.69 ± 29.82
PS.O-34.0	0.78	<0.01	1.45	−0.58	x	x			41.47 ± 14.23	15.36 ± 4.76
PS.O-38.1/P-38.0	0.58	0.15	−0.41	0.03			x		49.83 ± 10.01	59.14 ± 5.8
PS.O-38.2/P-38.1	0.7	0.04	0.41	−0.44				x	403.79 ± 54.5	327.41 ± 58.89
PS.O-38.5/P-38.4	0.88	<0.01	0.82	−0.74		x	x	x	163.08 ± 24.44	94.22 ± 13.23
PS.O-38.6/P-38.5	0.86	<0.01	0.99	−0.67		x	x	x	361.75 ± 74.27	168.69 ± 30.8
PS.O-40.0	0.7	0.01	−0.6	0.36				x	66.34 ± 12.97	88.35 ± 13.27
PS.O-40.6/P-40.5	0.75	<0.01	0.36	−0.46			x	x	372.13 ± 29.19	294.14 ± 34.19
PS.O-42.5/P-42.4b	0.79	<0.01	0.44	−0.57		x	x	x	313.53 ± 31.31	233.07 ± 24.54
PS.O-42.6/P-42.5b	0.79	<0.01	0.45	−0.52		x	x		226.36 ± 26.28	165.82 ± 14.05
TG.44.0	0.76	<0.01	−0.77	0.49		x			371.86 ± 72.65	620.47 ± 120.92
TG.46.4	0.78	<0.01	−0.91	0.52			x	x	389.78 ± 87.09	641.69 ± 122.92
TG.47.4	0.78	<0.01	−1.01	0.56				x	233.64 ± 57.24	399.27 ± 78.14
TG.49.4	0.74	0.01	−0.61	0.55				x	898.38 ± 176.02	1280.61 ± 184.12
TG.49.5	0.73	0.01	−0.58	0.51				x	967.3 ± 181.85	1356.58 ± 186.94
TG.50.1	0.73	0.01	−0.56	0.46				x	3679.66 ± 783.32	4970.1 ± 532.03
TG.51.7	0.68	0.05	−0.43	0.46				x	803.73 ± 148.86	1007.24 ± 129.46
TG.52.1b	0.72	<0.01	−0.62	0.45				x	2349.98 ± 497.12	3209.19 ± 324.5
TG.52.2	0.73	0.01	−0.5	0.47				x	5644.84 ± 1125.16	7359.71 ± 769.69
TG.52.3	0.67	0.08	−0.41	0.38				x	5871.8 ± 1225.93	7209.96 ± 967.84
TG.54.1b	0.75	<0.01	−0.69	0.52			x		1119.48 ± 233.54	1650.3 ± 207.51
TG.54.6	0.71	0.02	−0.44	0.51				x	10750.34 ± 1936.83	13813.28 ± 1289.65
TG.55.5	0.69	0.03	−0.41	0.48				x	1918.8 ± 352.6	2479.95 ± 257.98
TG.56.8	0.76	<0.01	−0.98	0.58		x	x		150.5 ± 34.36	263.58 ± 60.64
TG.58.2a	0.74	<0.01	−0.69	0.35				x	350.25 ± 73.99	476.86 ± 48.03
TG.62.13	0.64	0.03	0.34	−0.21			x		3162.55 ± 736.72	2296.29 ± 325.35
TG.63.13	0.64	0.03	0.53	−0.24			x	x	539.74 ± 170.8	348.6 ± 78.85
TG.64.16b	0.7	0.01	0.58	−0.33				x	740.92 ± 199.04	464.71 ± 88.7
TG.66.15	0.61	0.07	1.33	−0.22				x	332.46 ± 148.58	180.21 ± 61.03
TG.66.18	0.62	0.03	0.47	−0.22				x	910.13 ± 314.41	524.17 ± 151.27
TG.O-52.1	0.76	<0.01	−0.93	0.46			x	x	369.81 ± 102.77	637.06 ± 140.8

(Appendix A).

Lipid ontology (LiOn) enrichment analysis terms were calculated and analysed via the LiOn web platform [77]. PCA of LiOn terms was visualised via GO-PCA [78] and presented as a heatmap plot using Euclid as the distance matrix and Complete as the clustering algorithm. Enriched LiOn terms were ordered by p-value (calculated with a t-test), and terms were considered significantly different if the FDR-adj p-value was <0.05 [72].

3. Results and Discussion

In this study, a total of 51 BMOs were analysed using LC-MS global untargeted lipidomics. Ovaries were categorised as ripe (N = 25) or not ripe (N = 26) based on their GHA scores and GVF (Supplementary Data S1). Stereological analyses such as of the GVF are less prone to subjectivity bias than GHA, especially when assessing histological preparation, which indicates the presence of different gonadal stages in the same tissue [79]. In contrast, gonadal atresia and gamete degeneration are hardly recognised when using GVF alone [28]. Therefore, the two approaches offer the best performance in combination. GVF increased significantly in ripe BMOs, providing higher confidence in the criteria used for GHA classification (Figure A1).

During ovarian maturation, the area covered by follicles increases due to oocyte proliferation, with the volume covered by energy support cells decreasing [10,15,16]. As the oocytes proliferate, lipids are accumulated in the gonads at the expense of glycogen content [17,18,40]. This usually leads to the reported strong seasonal fluctuations in the major lipid classes of bivalves [19,23,27,43,80]. Here, BMO lipidomics profiles were integrated with GHA and the GVF through various chemometric and machine learning approaches to identify specific lipid molecules associated with the maturation process of BMOs.

3.1. Multivariate Chemometrics

Our investigation started with the application of a multivariate unsupervised chemometric approach such as PCA (Figure A2). As an unsupervised approach (e.g., samples are plotted without ‘a priori’ consideration of grouping variables), PCA does not suffer from overfitting problems; however, PCA can be difficult to interpret when distinguishing highly similar and heterogeneous groups with small multivariate differences, with a large number of features and a small amount of variability explained by the main axes [81]. Selecting the correct orthogonal axes that best explain the differences between ripe and immature BMO was problematic through PCA, because a high percentage of variability is explained by the first four pairs of principal components of the PCA model (>10%, Figure A2A). Indeed, in all the PCA plots showing the combination of the main five principal components, the two BMO groups did not evidence marked differences in their lipidomic profiles, which is an expected observation given the heterogeneous origin of the different gonad samples considered in this study (see Method Supplement S1).

To highlight specific lipids linked to the two BMO groups, we continued the biomarker discovery approach using more powerful supervised statistical models such as volcano plots, OPLS-DA and RF. Through volcano analysis, important lipids distinguishing ripe from undeveloped BMOs were highlighted based on their fold change (FC; here, 1.3 was considered the cut-off value) and their statistical significance (FDR p < 0.05, calculated according to a non-parametric Wilcoxon rank test). The resulting plot is shown in Figure 1A and led to the selection of 25 significant features (Table A1). Volcano plots are a quick and reliable approach for identifying markers that define a particular condition, although this statistical approach is limited to pairwise comparisons and often lacks statistical power, especially when analysing typical omics, datasets where the number of variables is far greater than the number of observations. In these cases, correction factors for the false discovery rate are usually applied, which reduce the statistical power of the test [82,83], and indeed, volcano analysis turned out to be one the most restrictive (lowest number of important feature highlighted) methods discussed in this work.

OPLS-DA is an extension of partial least squares regression and an approach characterised by a fast computation time and a decomposition of the variance of the samples into uncorrelated (orthogonal to the groups, orthogonal T-score) and correlated (predictive, T-score) to the grouping factor [73]. The results of OPLS-DA can be found in Figure 1B and show that 10.2% of the variance within our observation is correlated with the two BMO maturation classes (T-score). OPLS-DA as a supervised method may suffer from overfitting problems, especially when the number of observations is small. Therefore, an assessment of model performance (cross-validation) and permutation tests (time-consuming) are required to obtain a measure of model fit [74]. Model performances and fitting were evaluated, respectively, with a 10-fold LOOCV (predicted: R²x: 0.102, R²y: 0.5, Q²: 0.363; orthogonal: R²x: 0.167, R²y: 0.17, Q²: 0.07; Figure A3A) and 2000-fold permutation (Q²: 0.436 p < 0.001 and R²y: 0.669 p < 0.05, Figure A3B). From the resulting S-Plot, a set of 40 important lipids (Table A1) correlated with the grouping variable (ripe and non-ripe BMOs) were selected, setting an arbitrary threshold for the p(corr) value < −0.6 and >0.6 (Figure 1C).

The final approach used in this study was RF, a machine learning method that creates a series of classification (or regression) trees (a ‘forest’) by continuously bootstrapping the sample set. For each bootstrapping, the dataset is split into a test sample (66% of the samples) and an out-of-bag sample (training data with 33% of the observations). As a bootstrapping method, RF is also robust to several attributes that may constrain other statistical approaches (e.g., zero values, normality and homoscedasticity assumptions), while in classification mode, it assumes a balanced number of observations [50,75], and compared to the previous statistical approaches, RF is far more computationally intensive [75]. From the training data, the algorithm recursively selects a random set of variables whose ability to predict a categorical response (ripe or non-ripe BMO) or its correlation with a continuous response variable (such as the GVF) is evaluated using training data and tested on OOB samples.

Partitioning is based on the identification of the predictor that minimises either the within-group variance (categorical response variable) or the error in the regression against the response, in the case of a continuous response variable [50,75]. The optimal predictor becomes the best node of the forest, and the process is repeated for all the number of trees specified in the function. In RF classification (RFclass), the variables with the lowest OOB error rate (OOB-ER) are the best predictors for the categorical variables, with the lower OOB-ER indicating the higher predictive significance of a given molecule. In RF regression (RFreg), the importance of each predictor is evaluated as the proportion of variation explained (PVE), a measure of the reduction in the predictive ability of the tree when that variable is removed. A backward purging (BP) approach is used to select the most important features in the RF. BP consists of performing multiple RFs on the most informative subset of variables (here, the top 5% of features were selected from both RFclass and RFreg, Figure 2) and recording the performance of each variable (in terms of OOB-ER or PVE). The results of the BP process of the RFclass data showed that the saturated PC 28:0 (OOB-ER 0.12) was the best predictor of the ripe/non-ripe condition (Figure 2A). As for RFreg, the BP process yielded unsaturated phosphatidic acid (PA) 44:4 (PVE 0.70, Figure 2B), followed by saturated ether-linked PC such as PC(O-30:1/P-30:0), PC(O-30:0) and PC(30:0), all with a PVE of 0.69 like most important molecules. All the saturated plasmanyl/plasmenyl and diacyl PC molecules and PA(44:4) are included in the panel of 35 lipids reported in

Table 1. Lipid species (AUC > 0.8) markers of gonad ripeness identified by the supervised statistical approaches in blue mussel ovary (BMO) and ranked according to receiving operating curves (ROCs) and area under the curve (AUC). Isomeric lipid molecules resolved by reverse phase separation (same exact mass of 5 Δppm, but different retention time) are distinguished with a lowercase letter after the lipid ID. Cut-off: value of the variable that maximises sensitivity and specificity; Sens.: sensitivity of a variable, true positive ratio at the cut-off value; Spec.: specificity of the variable (true negative ratio) at the cut-off value; p-value: Wilcoxon Rank test with p-value adjusted for false discovery rate [72]; FC: fold change. The complete set of variables highlighted by statistical methods is available in Table A1. Computed via R package ‘MetaboanalysR’ [67].

Lipid ID 1	AUC	95% CI	p-Value 2	FC	Volcano	OPLS-DA	RFClass	RFreg	Not Ripe Average ± CI (95%)	Ripe Average ± CI (95%)
CAEP(42:2)	0.808	0.671–0.925	3.63 × 10−4	1.57		X	X	X	47.59 ± 9.06	20.67 ± 7.5
CAEP(42:3)	0.804	0.683–0.916	9 × 10−3	3.13	X		X	X	36.95 ± 11.25	11.19 ± 6.08
CerPE(34:3)a	0.845	0.729–0.939	2.08 × 10−3	0.98		X	X	X	95.62 ± 20.84	48.43 ± 8.13
CerPE(35:3/2)	0.874	0.771–0.954	2.08 × 10−3	1.08		X	X	X	132.82 ± 29.82	55.95 ± 7.87
CerPE(40:1)	0.806	0.668–0.93	3.62 × 10−3	1.17		X	X	X	34.15 ± 7.02	13.71 ± 5.25
CerPE(40:2)	0.905	0.799–0.981	3.04 × 10−4	1.48	X	X	X	X	46.4 ± 7.88	17.69 ± 4.48
LPC(22:2)	0.806	0.673–0.922	7.42 × 10−3	−1.62			X		115.47 ± 39.6	308.4 ± 101.03
PA(38:6)	0.815	0.69–0.917	0.023	5.66	X		X	X	12.45 ± 6.38	1.46 ± 1.54
PA(40:6)	0.861	0.749–0.958	7.42 × 10−3	2.85	X	X	X	X	33.14 ± 12.05	8.31 ± 3.64
PA(44:4)	0.811	0.685–0.925	0.014	−1.75			X	X	102.52 ± 41.73	226.25 ± 58.24
PA(O-38:6/P-38:5)	0.851	0.735–0.94	2.62 × 10−3	2.12	X	X	X	X	46.5 ± 13.09	15.64 ± 4.86
PC(28:0)	0.842	0.732–0.939	3.62 × 10−3	1.35	X	X	X	X	426.29 ± 119.56	140.95 ± 36.64
PC(30:0)	0.848	0.717–0.943	2.36 × 10−3	1.37	X	X	X	X	1149.26 ± 291.11	396.62 ± 108.82
PC(O-30:0)	0.803	0.672–0.927	0.012	1.48	X	X	X		862.53 ± 256.36	318.65 ± 112.49
PC(O-31:0)	0.8	0.664–0.902	0.014	1.37		X	X	X	68.9 ± 21.25	41.64 ± 9.75
PC(O-32:0)	0.802	0.668–0.919	0.016	1.35	X	X	X	X	923.78 ± 291.49	351.64 ± 111.32
PC(O-46:10/P-46:9)a	0.811	0.68–0.915	2.25 × 10−3	−1.42	X	X	X	X	145.61 ± 41.99	341.56 ± 83.65
PE(O-34:3/P-34:2)	0.828	0.717–0.942	0.059	0.98			X	X	53.51 ± 20.01	24.5 ± 4.84
PE(O-36:1/P-36:0)b	0.868	0.748–0.962	2.66 × 10−3	1.63	X	X	X	X	65.23 ± 19.65	21.09 ± 4.88
PE(O-36:2/P-36:1)b	0.848	0.719–0.949	3.62 × 10−3	0.96		X	X	X	94.97 ± 22.76	47.13 ± 7.8
PE(O-36:2/P-36:1)c	0.875	0.772–0.948	2.08 × 10−3	0.97			X	X	251.77 ± 50.85	121.82 ± 15.8
PE(O-36:3/P-36:2)	0.851	0.737–0.95	7.42 × 10−3	0.61			X	X	308.04 ± 55.87	193.04 ± 19.3
PE(O-37:2/P-37:1)	0.858	0.749–0.948	1.32 × 10−3	0.99		X	X	X	231.1 ± 39.56	121.75 ± 21.86
PE(O-37:3/P-37:2)	0.845	0.719–0.938	2.08 × 10−3	0.51		X	X	X	428.34 ± 49.7	300.34 ± 26.36
PE(O-37:4/P-37:3)	0.868	0.752–0.964	3.26 × 10−3	0.74		X	X	X	84.17 ± 14.05	50.25 ± 6.82
PE(O-39:3/P-39:2)	0.8	0.692–0.905	3.62 × 10−3	0.45			X	X	360.03 ± 39.86	264.36 ± 23.89
PG(34:1)	0.86	0.735–0.96	7.29 × 10−3	2.19	X	X	X	X	25.61 ± 8.8	7.32 ± 2.83
PS(36:2)	0.81	0.65–0.913	3.0 × 10−3	7.06	X		X	X	16.21 ± 10.11	0.14 ± 0.22
PS(38:5)	0.894	0.79–0.962	1.32 × 10−3	1.61	X	X	X	X	215.46 ± 51.5	66.38 ± 17.26
PS(38:6)a	0.84	0.728–0.935	3.62 × 10−3	1.23		X	X	X	164.04 ± 40.39	66.94 ± 16.21
PS(39:6)	0.823	0.695–0.926	0.033	1.21			X	X	96.45 ± 35.18	39.88 ± 10.52
PS(40:2)	0.828	0.72–0.926	7.94 × 10−3	0.75			X	X	157.85 ± 32.77	91.97 ± 13.65
PS(40:5)	0.82	0.691–0.92	5.49 × 10−3	1.52				X	41.86 ± 11.05	18.17 ± 5.55
PS(40:6)	0.889	0.785–0.981	2.08 × 10−3	1.06		X	X	X	329.08 ± 66.25	155.36 ± 27.89
PS(O-38:5/P-38:4)	0.875	0.777–0.957	1.29 × 10−3	0.82		X	X	X	163.08 ± 24.44	94.22 ± 13.23
PS(O-38:6/P-38:5)	0.855	0.732–0.956	2.36 × 10−3	0.99		X	X	X	361.75 ± 74.27	168.69 ± 30.8

¹ Isomeric lipid molecules resolved by reverse-phase separation (same exact mass of 5 Δppm, but different retention time) are differentiated with a lowercase letter after the lipid ID. ² FDR-adj p-value [72].

, and will be discussed later in the text. In the next sections, we compare the different roles of important lipids (through LiOn enrichment analysis) and the results of the four different statistical approaches here described. In aiming to maintain the maximum amount of meaningful information [75], it was agreed on to select the top 10% of molecules (with a threshold over which model performances substantially improved; see Method Supplement S2d) as important for RF (≈83 molecules for RFclass and RFreg, Table A1).

3.2. Comparison of the Outputs Obtained with the Different Statisatical Approaches

The four different statistical models applied in this study allowed for the selection of a total of 123 relevant lipid molecules (respectively, 25, 40, 72 and 92 via volcano plot, OPLS-DA, RFclass and RFreg, Table A1) differentiating ripe from non-ripe BMOs from 838 molecules present in the PIT (Supplementary Data S2b). It should be noted that none of the PCs, glycerophosphoethanolamines (PEs), glycerophosphoinositols (PIs) and triglycerides (TGs) detected by the present statistical approach was a major component of the blue mussel gonad lipidome [49], supporting also the previous observation made on the PCA plots.

Key features highlighted by the different approaches are compared using a Venn diagram (Figure 3). RFreg and RFclass yielded the largest number of unique molecules (respectively, 41 and 12 molecules, i.e., 33.3% and 9.8% of the total), while only four variables were exclusive to OPLS-DA (3.3% of the total), and two variables were detected only by volcano analysis (1.6%). Ten variables were common for all four approaches (8.1%). The comparison between classification (volcano, OPLS-DA and RFclass) and regression approaches (RFreg) shows that 51 (41.5% of the total) molecules (out the 92 considered important and highlighted by RFreg) were common with one of more of the models here evaluated. This, again, indicates that there was a large consensus between the GVF and the classification criteria used to define the two BMO groups.

The functional role of the lipids that differentiated undeveloped from ripe BMOs was investigated via LiOn enrichment analysis (Figure 4). Figure 4A shows the PCA for LiOn terms, which reveals a clear distinction of the function of those molecules since 84% of the samples (43/51) could be correctly clustered according to the functional role of the highlighted lipids. Terms related to storage lipids, lipid bodies and TGs were enriched in ripe BMOs, which could be attributed to the vitellogenesis and accumulation of lipid bodies in the yolk granules [16]. Interestingly, TG species, identified by the various statistical methods, showed an inverse pattern according to their unsaturation levels: TG molecules bearing highly unsaturated acyl residuals (DB > 10) were found to be inversely correlated with the GVF and also more abundant in undeveloped BMOs, with the opposite pattern observed for TGs with <10 DB (Table A1). This observation could be the result of the rearrangment and/or resorbition of PUFA during ovarian maturation/regression [84]. Although increased content of unsaturated TG was reported in ripe ovarian follicles of P. merguensis [48], others have reported the dominance of saturated and monounsaturated acyl moyeties in TGs of mud crab ovaries, Scylla paramamosain [85] and sea sponges [86].

In the non-ripe BMO group, significant up-regulation of ‘membrane components’, an enrichment term associated with membrane lipids, was observed instead (Figure 4B). The previous term is mainly linked to the larger abundance in undeveloped ovaries of diacyl and ether-linked (associated with the term ‘plasmalogen’) PC and PE molecules. This observation could be related to the greater surface area of the gonad covered by support cells, such as ADG and VCT, that are characteristic of early maturation stages of bivalve ovaries [15], which is also supported by the reduction in the GVF found in non-ripe BMOs compared to mature ones (Figure A1).

At the same time, many of such membrane lipids included saturated and short-chain molecules, features connected with the up-regulation of terms including ‘below average transition temperature’ and ‘below average bilayer thickens’. On the contrary, longer and more unsaturated PCs and PEs were more abundant in ripe BMOs, leading to the up-regulation of ‘PUFA’ and ‘fatty acids with more than 18C’ terms for this group.

In addition, terms related to ‘lysosomes’, the ‘Golgi apparatus’ and ‘mitochondrion’ were also up-regulated in the immature BMO group. The former two of these terms are connected with sphingolipids (ceramides—Cer, CAEP and CerPE), since these classes of lipids are produced in the Golgi apparatus and act in autolysis and atretic processes [87]. The higher content of such lipids in undeveloped BMOs (Table A1) could be related to gonadal degeneration and atresia, as the presence of lysosomes and vacuoles together with macrophage infiltration into follicles characterises atresia in mussel ovaries [28]. Interestingly, opposite patterns could be observed between Cer, which is higher on ripe BMOs and correlated with the GVF, CAEP and CerPE. Such a pattern could be related to the role of Cer as stabilisers in adipose tissue, and hence could be again related to the accumulation of yolk reserves in ripe ovaries [88]. The mitochondrion term is associated with the detection via the RFreg of a cardiolipin (CL) 88:21, which was also found to be weakly, inversely correlated with the GVF (Spearman rank r = −0.05, Table A1). The reason behind this observation is unclear, since the same molecule did not exhibit significantly different content between the two BMO groups (FDR p > 0.05). CLs are fundamental lipids found in the inner mitochondrial membrane, participating in cellular respiration [89]. Highly unsaturated CLs were found predominant in mussel gonads [49] and generally distinguish mussels from other groups of bivalves such as clams [90]. CL(88:21) must contain four 22C acyl residuals, since 24C FA detected in blue mussel gonads were saturated or monounsaturated species [49]. Therefore, it is likely that this molecule included three DHA and one FA22:3 non-methylene-interrupted dienoic (NMID). Nevertheless, the connection of the aforementioned molecule with mussel ovarian development is unclear if we exclude the structural role of NMID FA in mussel gonads [91]. An increase in CL(88:24) has been observed in quiescent scallops (Placopecten magellanicus), whereas larger extents of CL containing 18C FA (FA18:0 and FA18:1) were found in ripe (pre-spawning) individuals [92]. Given the different biology of the two species (dioecious for mussels and hermaphrodite for scallops), it becomes more difficult to directly compare our results with those of the previous study.

3.3. Ranking the Panel of Potential Biomarkers

ROC curves were generated for all important variables selected by each statistical method to distinguish ripe from non-ripe BMOs, and these were ranked using the AUC (Table A1). ROC curves are the graphical representation between the sensitivity (true positive) and specificity (true negative) of a biomarker [93]. AUC can range between 0.5 (diagonal line, random effect, 50% specific and 50% sensitive) and 1 (perfect biomarker, 100% specific and sensitive). The optimal cut-off value (maximum specificity and sensitivity) of a specific biomarker is calculated using the maximum distance between the ROC curve and the diagonal line (Youden index [94]). Molecules with AUC values between 0.8 and 0.9 have been considered potential biomarker candidates for various human diseases in the past [95,96]. Out of all the important variables highlighted by the statistical approach used in this study, 35 had an AUC value > 0.8 and are listed in Table 1. From the AUC > 0.8 variables (35/123), 10 were detected by all four statistical approaches, 17 by a combination of three of these approaches, 6 by two models and just 2 by a single model. This observation emphasises the importance of combining multiple orthogonal statistical approaches, as any biomarker discovery method will inevitably have some limitations if unaccompanied.

In the present study, no lipid was found to be a perfect marker of the two different gonadal stages (e.g., AUC = 1). One molecule showed an AUC > 0.9; CerPE(40:2), which had an AUC of 0.905 (95%CI: 0.799–0.981, Figure 5A) and was more abundant in undeveloped ovaries (Figure 5B). CerPE(40:2), was characterised by a sensitivity of 80% (true positive rate) and a specificity of 100% (true negative rate). Three other CerPEs ranked among AUC > 0.8 molecules (Table 1): CerPE(35:3/2), CerPE(34:3)a and CerPE(40:1), with AUC values of 0.874, 0.845 and 0.806, respectively. Other ceramide lipids important for distinguishing between ripe and immature BMOs included CAEP 42:2 and 42:3 with AUC values of 0.808 and 0.804, respectively. All the previous CerPE and CAEP molecules were minor components of these classes detected in Mytilus gonads [49] and were all found mainly in undeveloped BMOs; the role of these sphingolipid classes in marine invertebrates is not yet fully understood. CerPE and CAEP are mainly found in invertebrate membranes [54,97,98,99,100] and are synthesised from ceramide precursors through the action of a specific CDP ethanolamine:ceramide phosphotransferase localised in the Golgi apparatus [101]. In the scientific literature, there are scarce reports on the role of CerPE and CAEP in bivalve ovarian maturation. Pauletto and coworkers [102] reported a differential expression for genes involved in complex ceramide metabolism between ovarian oocytes and spawned oocytes in the clam R. decussatus. Other authors hypothesised a similar function for the CerPE and CAEP of the vertebrate analogue sphingomyelin (SM), which can form specific domains of the cell membrane defined as membrane rafts [100]. The decrease in CerPE and CAEP observed in ripe females could also be a consequence of the gamete maturation process, as a decrease in SM was observed in Xenopus laevis during oocyte maturation under the influence of progesterone [103].

Glycerophosphoserines (PSs) represented the lipid class that, overall, changed mostly between the two BMO groups (Table 1). PS(38:5) was the second most important biomarker of BMO maturity (AUC 0.894), and several of the other nine PS molecules with AUC > 0.8 (36:2, 38:5, 38:6a, 39:6, 40:2, 40:5, 40:6 and O-38:5/P-38:4 and O-38:6/P-38:5) were between the main diacyl and plasmanyl/plasmenyl PS species found in Mytilus gonads [49]. All AUC > 0.8 PSs were more abundant in undeveloped BMOs. PSs are secondary components of cell membranes, found mainly on the cytosolic side of the lipid bilayer [104]. Their relevance as markers of BMO development could be linked to different reasons. PSs may be converted into different glycerophospholipids, through the action for phosphotransferases, during ovarian maturation, as observed in largemouth gudgeon (Coreius guichenoti) ovaries [105]. On the contrary, the larger content of PSs in undeveloped BMOs could also be linked to apoptotic processes in degenerating gonads, since the exteriorisation of PSs is considered to be one of the ‘triggers’ for the phagocytosis of apoptotic cells [106,107], as observed in UV-exposed O. edulis haemocytes [108] and in apoptotic ovarian cells of Drosophila melanogaster [109,110].

The third molecule for prerelevance as a biomarker evidenced in this study is the ether-linked PE O-36:2/P-36:1c (AUC: 0.875; Table 1). All PE molecules with an AUC > 0.8 were plasmanyl/plasmenyl forms (O-34:3/P-34:2, O-36:1/P-36:0b, O-36:2/P-36:1b, O-36:3/P-36:2, O-37:2/P-37:1, O-37:3/P-37:2, O-37:4/P-37:3 and O-39:3/P-39:4) but did not include any of the principal PE forms described in blue mussel gonads [49], increasing the complexity of explaining the physiological motivation behind our observation. All PEs with an AUC > 0.8 were found to be mainly abundant in undeveloped BMOs, and this could possibly be related to the synthesis of PCs during ovarian development, thorugh the sequential methylation of PE [111]. A total of seven PC molecules had an AUC > 0.8; six of which were saturated molecules (28:0, 30:0 and the plasmanyl/plasmenyl O-30.0, O-31:0 and O-32:0; Table 1) and found to be distincly abundant in non-ripe BMOs. The long-chain and highly unsaturated ether-linked PC(46:10/P-46:9)a characterised, instead, ripe BMOs. This trend is in agreement with that reported by other authors on the PC composition of ovaries in the shrimp P. merguiensis: saturated PC molecules were found in undeveloped regions of the ovary, whereas highly unsaturated PCs (containing 20-22C PUFAs) had a higher coverage of ripe ovarian parts [48]. PCs and PEs are the two major membrane lipids in bivalves [21,49,65,112,113]. Given the low unsaturation of PCs and PEs with an AUC > 0.8 and abundance in non-ripe BMOs, it is likely that these molecules carried saturated or monounsaturated FA rather than long-chain PUFA. Acyl residues of membrane lipids represent energy sources during embryonic development in invertebrates [114,115], so their enrichment with PUFA at later developmental stages may be important for the proper development of the offspring. To this extent, PUFA enrichment patterns have been observed for PC and PE fractions in maturating scallop ovaries [43].

Lastly, four PA molecules (38:6, 40:6, 44:4 and the plasmalogen O-38:6/P-38:5), one glycerophosphoglycerol (PG) 34:1 and lyso-glycerophosphocholine (LPC) 22:2 completed the panel of 35 molecules with an AUC > 0.8 in distinguishing ripe from non-ripe BMOs (Table 1). All these lipid species were more abuntand in undeveloped ovaries, with the exception of PA(44:4) and LPC(22:2). There are not many reports on the role of such minor glycerophospholipid classes in bivalve or invertebrate ovarian maturation, due to their low abundance, which is hardly observable through traditional lipid analysis methods [47]. Therefore, we can only speculate about their role in the development of BMOs. PAs and PGs are intermediates of glycerophospholipid biosynthesis, each forming the precursors of the major membrane lipid classes and of cardiolipins [104]. PA concentration increases in Xenopus oocytes as a response to insulin treatment, as a result of phospholipase D, which produces lyso-glycerophosphate (LPA), which participates in the lipid signalling of oocyte maturation [111]. A reduction in most of PG was instead observed in sea sponges in reproductive individuals [86], which is an observation that could be in agreement with data discussed in the present study. LPCs are metabolites of PC formed by the Lands cycle through the action of a phospholipase that cleaves one of the FA residues [116]. LPC(22:2) is a molecule that is likely to bear a 22:2 NMID FA [49]. NMID FAs are endogenous lipids of mussels that are preferentially incorporated into the mantle, competing with C20 and C22 PUFA under reproductive demands [91], which can suggest the a role for LPC(22:2) in the restructuring of membrane lipids in ripe BMOs. Nevertheless, especially for these minor lipids, further evidence is needed to obtain a definitive overview of the system.

4. Conclusions

Rapid and informative tools for measuring the maturity stage of female mussels are useful both for industrial applications, as they can be used to monitor gamete maturation in hatcheries to reduce problems with uncontrolled spawning and pre-spawning atresia, and for monitoring the reproductive status of native mussel populations. The high resolution and ability to work at the lipid molecule level make LC-MS lipidomics an important platform for studying lipid biomarkers that can assess maturation status in BMOs. LC-MS analysis has produced a very complex and convoluted dataset that requires experienced data analysis skills to extract meaningful information from it. To fulfil this task, in this work we applied different supervised statistical approaches that were more powerful than classical unsupervised approaches (PCA) to highlight the differences between the studied BMO groups. In combining the four supervised statistical approaches used here (volcano, OPLS-DA, RFclass and RFreg), it was possible to highlight a panel of 35 lipid species that differ to a high degree (AUC > 0.8) between ripe and undeveloped BMOs. These potential markers include CerPE, CAEP, LPC, PA, PC, PE, PG and PS. In many cases, however, the exact role and reason for the importance of AUC > 0.8 lipids for classification could only be speculated upon, based on similarities reported by others in different model organisms. Future studies should aim to better understand the role played by lipids during gamete development in marine invertebrates such as bivalves, focussing on less studied lipid classes.

The set of potential lipid biomarkers should be validated and possibly quantified using further independent gonad samples, and if confirmed, the protocol should be improved and refined to develop a non-lethal assay. The method should be more fully developed if the lipid panel presented here is validated by moving to a targeted lipid quantification approach, which will significantly reduce the complexity of the dataset and the amount of data analysis required to obtain information on these important lipids. GHA is considered a traditional approach for monitoring gonadal development in marine mussels [117] and is an important tool for investigating the reproductive physiology, toxicology, life cycle and population structure of bivalves [10,11,16,28,32,34,36,118]. In the future, GHA will continue to be a necessary diagnostic tool, but technological advances and the application of state-of-the-art high-throughput omics techniques should support and complement traditional techniques to increase our knowledge of the physiology of marine organisms and their ability to adapt to a changing environment.