Oogenesis and lipid metabolism in the deep-sea sponge Phakellia ventilabrum: an histological, lipidomic and transcriptomic approach

Background Sponges contain an astounding diversity of lipids which serve in several biological functions, including yolk formation in their oocytes and the embryos. On animal reproduction, lipids constitute one of the main energy storage forms for the adult and the offspring. The study of lipid metabolism during reproduction can provide information on food-web dynamics and energetic needs of the populations in their habitats, however, there are no studies focusing on the lipid metabolism of sponges during seasonal reproduction. The deep-sea sponge Phakellia ventilabrum (Demospongiae, Bubarida) is a key species of North-Atlantic sponge grounds, but its reproductive biology is not known. In this study, we used histological sections, lipidome proling (UHPLC-MS), and transcriptomic analysis (RNA-seq) with goal to i. assess the reproductive strategy and seasonality of this species, ii. examine the relative changes in the lipidome signal, and the gene expression patterns (RNA-seq) of enzymes participating in lipid metabolism in female specimens during gametogenesis. Results P. ventilabrum is an oviparous and most certainly gonochoristic species, reproducing in May and September in the different studied areas. Half of specimens were reproducing, generating two to ve oocytes per mm 2 . Oocytes accumulated both protein and lipid droplets. As oogenesis progressed, the signal of most of the unsaturated and monounsaturated triacylglycerides increased, as well as of few other phospholipids. Most of the other lipids and especially those with > 3 unsaturations showed a decrease in signal during the oocyte maturation. In parallel, we detected upregulated genes in female tissues related to triacylglyceride biosynthesis and others related to fatty acid beta-oxidation. yolk lipid category most marked changes, some other fatty

oxygen, osmolarity, diet, pressure), defence from predators, and antioxidant activity (38,(43)(44)(45)(46). Sponge lipids have also been studied for their potential pharmaceutical applications, e.g. several demosponge fatty acids exhibit antimalarial, antimycobacterial and antifungal activity (1,47) and many glycolipids have anti-in ammatory, anticomplement, antitumoral, and immunomodulatory properties (48)(49)(50)(51). In regard to their role in reproduction, yolk of lipid origin is also one of the main components, together with protein platelets, of oocytes and embryos of sponges (52). Sponge yolk has either a homogeneous (53)(54)(55) or heterogeneous structure with a mix of lipids and proteins in the same yolk platelet (56)(57)(58)(59)(60). But notably, in some demosponges, lipid droplets are the only nutrients accumulated within the oocytes (55). Yolk is formed by i) autosynthesis, in which the egg itself produces the nutrients, and ii) heterosynthesis through phagocytosis of yolk precursors, nutrients and bacteria either provided from accessory cells or taken from the mesohyl of the sponge (58,(60)(61)(62). Currently, the type of sponge yolk is only assessed by electron microscopy observations in which the lipid and protein yolk can be largely recognized by their different electron dense nature: proteins are more electron dense showing crystal structure and lipids appear lighter and relatively more homogeneous (63).
Almost all sponges have lecithotrophic larvae (64,65) and only a handful of deep-sea species are known to have direct development (from embryo directly to young sponge without the larval phase) (66)(67)(68). With the lack of a planktotrophic phase, all of the nutrients needed for the next developmental stages, until settlement and metamorphosis into a sessile adult sponge, are accumulated in the egg during gametogenesis in the oviparous species or during embryogenesis in the viviparous species. Consequently, the quality and quantity of the yolk is crucial to provide the propagule with all the necessary energy.
Few studies have investigated the variations of sponge lipid metabolism with respect to seasonality. Koopmans et al. (69) studied fatty acid (FA) composition in North-East Atlantic/Meditarranean Sea demosponge species of the genera Haliclona, Halichondria and Aplysina and found a strong correlation between the FA composition in the surrounding dissolved organic matter and the sponge FAs during nutrient blooms in spring-summer. On the other hand, Lüskow et al. (70) measuring the lipid content as a fraction of sponge dry weight, found that it remained invariable throughout the year, without being affected by seasonal planktonic blooms or periods of starvation. However, none of these studies linked directly these variations with reproduction. Two other studies, using obsolete methods and rough estimates, have linked the seasonality and nutrient blooms with the physiological status of sponges: Reiswig (71) quanti ed the nutritional resources within the sponge tissue and found a direct link to growth and reproduction in the Jamaican population of Mycale sp., and Elvin (72) quanti ed the tissue growth and reproductive output with respect to various abiotic factors (temperature, light, salinity and nutrient abundance) in Haliclona (Reniera) cf. cinerea (H. permollis in the publication). Finally, Elvin (73) studied the direct relationship of lipid content and reproductive investment in H. (R.) cf. cinerea, nding that both protein and lipid levels increased in female individuals just before the onset of gametogenesis. However, the nature of the lipids used in reproduction i) for energetic demands and ii) for the formation of yolk platelets is still almost completely unknown in sponges. Presently, a single study has investigated the expression levels of lysophospholipids in the homoscleromorph viviparous sponge Oscarella tuberculata (74) in response to seasonality and reproduction, and found that their levels increased towards the end of the reproductive cycle: during embryogenesis and larval development (75). Understanding the lipid level uctuations during the reproductive period is very important from an ecological point of view, in order to understand nutrient availability, food-web dynamics, and energy requirements of sponge populations. Lipid content dynamics can also inform about the ability to survive and reproduce in case of habitat disturbance, including unbalance in lipid sources due to contamination of the habitat, changes in nutrient blooms, sediment resuspension blocking the ltering system and no access to nutrient uptake and lipid storage. Additionally, such studies are pivotal from an evolutionary point of view, to understand the origin and evolution of yolk formation and lipid composition in Metazoa.
Despite being a very common species, from the British Isles to the coasts of Norway and Sweden, the reproduction of this species has never been investigated. Interestingly, some of its FAs have been previously identi ed with GC-MS (36,39). In the present study, we chose to investigate the reproductive strategy of the boreal deep-sea demosponge, Phakellia ventilabrum (85) with histological observations (light and electron microscopy). Furthermore, we compared the lipid pro le (UHPLC-HRMS) and the expression levels (RNA-seq) of genes related to lipid metabolism in reproductive female vs nonreproductive specimens of this species. By doing so, we aimed to determine i) the types of lipids used for yolk formation ii) the molecular routes involved in the production of these lipids and iii) the energetic demands during gametogenesis in this species. Therefore, information on the reproduction of Phakellia species will signi cantly improve our understanding of sponge ground habitats and in the long run contribute to their conservation.

Reproductive season
Reproductive activity was found in specimens collected both in March (Swedish coast) and September (Norwegian coast) ( Table 1). P. ventilabrum is an oviparous species, as no further developmental stage, e.g. embryos, was observed within the tissue. Although no male individuals were found in none of the locations, P. ventilabrum is expected to be gonochoristic since all species of Bubarida follow this strategy, but we cannot exclude the possibility of successive hermaphroditism. At the population level, 50% of the collected specimens (two out of four) were female in the Swedish coast in March, while 66.6% (six out of nine) were female in the population of the Norwegian coast, in September (Table 1). There was relatively asynchronous development of the gametes within the populations, with some individuals having previtellogenic oocytes, while others had oocytes in a more advanced developmental stage. Asynchronous oocyte development was also observed within the same individual in a few cases. The density of oocytes was around 2 to 5 oocytes per mm 2 sponge tissue (Table 1).  (Fig. 2B). In some Vi_II oocytes, protein platelets were more abundant (Fig. 1G), while in some others, lipid droplets were equally abundant to the other two types of yolk (Fig. 1H). Altogether, lipid and protein yolk were formed in similar amounts while heterogenous yolk occupied twice the area of the ooplasm, compared to the other two (Fig. 2B).
Although vertical transmission of associated microsymbionts was not observed in this species, a phagocytosis of a bacterium by the oocyte was observed in a single instance (Fig. 1G, insert). From the histological observations, it is clear that P. ventilabrum is a Low Microbial Abundant (LMA) species, as almost no bacteria were observed in the mesohyl.

Lipidomic analysis
To characterize the changes of the lipidome in relation to the reproductive status, we analysed the samples by semitargeted mass spectrometry lipidomics with relative quanti cation. In total, we detected 532 different lipids in the extracts of P. ventilabrum with a remarkable variety of lipids (Additional le 1: Table S1). The main lipid categories were: 96 free fatty acids (FFAs) (Additional le 1: Table S1), 61 phosphatidylcholines (PC) (Additional le 1: Table S1) and 155 triacylglycerides, (TGs) (Additional le 1: Table S1); and nally, 26 sphingolipids and glycosphingolipids (Additional le 1: S1). In most of the above lipid categories, the highest number of lipids detected with our analysis was unsaturated fatty acids (UFA), and most particularly polyunsaturated fatty acids (PUFA) (Fig. 3), which also had the highest signals (Fig. 3).
Exceptions were the phosphatidylglycerols (PGs), lysophosphatidylcholines (LPCs) and sphingolipids for which the number of SFA or MFA detected was higher than PUFA and had higher signal detections (Fig. 3).

Lipid signal variations in different locations/months
The main variation in the lipidome among the specimens was due to the different sampling locations/months or both, independently of the reproductive status (Fig. 4Α). Among the different locations, the total signal of almost all the main lipid categories studied was higher in samples collected in Kosterfjord (Sweden) in March than in Korsfjord (Norway) in September ( Fig. 4Β; Additional le 1: Table S1). Especially the lipid categories of PC, PG, PE, LPE, and PUFA_FFA were two to seven times higher ( Fig. 4Β; Additional le 1: Table S1). Studying the variation of the signal within each lipid category, signi cant differences were observed in several monounsaturated fatty acids (MFA) and polyunsaturated fatty acids  Table S1).

Lipid signal variations dependent on the reproductive statuses
To analyse the variation in the lipidome regarding the reproduction, we correlated the variation in the lipid signal to the area of the sponge tissue occupied by oocytes, focusing mainly on the larger lipid categories detected in our analysis. Female specimens with an increasing area occupied by oocytes are considered those with more mature oocytes because during maturation process, the oocytes get larger (see paragraph 1.2) ( Table 1). The sponge tissue of nonreproductive specimens (NR) did not have any oocytes. In regard to the general signal uctuation of each lipid category, the majority of lipids indicated a tendency to decrease their signal with the oocyte maturation (increasing area occupied by oocytes), with the exception of the TGs, which in their majority increased their signal with oocyte maturation (Figs. 5, 6; Additional le 3: Table S2).
Speci c observations were made on the lipids of each lipid category, based on their number of carbons and unsaturation.
Finally, we studied the regulation of different oxylipins along oocyte maturation (Figs. 5, 6F). They presented a general decreasing trend. Remarkably, HEPEs and HDoHEs presented the highest decreasing trend and only the PGE 2 and PGF 2, as also the 12-HHTrE and 5(6)-EpETrE increased their signal with oocyte maturation (Fig. 6F).

Lipid signal variations in different locations/months
The main variation in the lipidome among the specimens was due to the different sampling locations/months or both, independently of the reproductive status (Fig. 4Α). Among the different locations, the total signal of almost all the main lipid categories studied was higher in samples collected in Kosterfjord (Sweden) in March than in Korsfjord (Norway) in September ( Fig. 4Β; Additional le 1: Table S1). Especially the lipid categories of PC, PG, PE, LPE, and PUFA_FFA were two to seven times higher ( Fig. 4Β; Additional le 1: Table S1). Studying the variation of the signal within each lipid category, signi cant differences were observed in several monounsaturated fatty acids (MFA) and polyunsaturated fatty acids  Table S1).

Lipid signal variations dependent on the reproductive statuses
To analyse the variation in the lipidome regarding the reproduction, we correlated the variation in the lipid signal to the area of the sponge tissue occupied by oocytes, focusing mainly on the larger lipid categories detected in our analysis.
Female specimens with an increasing area occupied by oocytes are considered those with more mature oocytes because during maturation process, the oocytes get larger (see paragraph 1.2) ( Table 1). The sponge tissue of nonreproductive specimens (NR) did not have any oocytes. In regard to the general signal uctuation of each lipid category, the majority of lipids indicated a tendency to decrease their signal with the oocyte maturation (increasing area occupied by oocytes), with the exception of the TGs, which in their majority increased their signal with oocyte maturation (Figs. 5, 6; Additional le 3: Table S2).
Speci c observations were made on the lipids of each lipid category, based on their number of carbons and unsaturation.
Finally, we studied the regulation of different oxylipins along oocyte maturation (Figs. 5, 6F). They presented a general decreasing trend. Remarkably, HEPEs and HDoHEs presented the highest decreasing trend and only the PGE 2 and PGF 2, as also the 12-HHTrE and 5(6)-EpETrE increased their signal with oocyte maturation (Fig. 6F).

Discussion
In our study we found that the sponge species P. ventilabrum was gonochoric and most possibly oviparous, reproducing in spring and end of summer/ autumn in Kosterfjord and Korsfjord respectively. Our comparative lipidomic and transcriptomic analysis shows that during oogenesis and at the stage of yolk formation, the signal of most TGs increases and their possible de novo lipogenesis occurs while in parallel the signal of other lipid categories decreases with their possible beta-oxidation. Our results suggest that TGs might be the main component of the energy stock in the female gametes while other lipids undergo degradation to generate energy for the adult during this energetically high consuming process.
1. Seasonality and reproductive strategy of P. ventilabrum In both locations and sampling times, individuals with mature, vitellogenic oocytes were found among the collected specimens (  (92) in boreo-arctic North-Atlantic were correlated to seasonal nutrient blooms (11,86,93). Previously, it has been stated that the seasonal primary production and the subsequent nutrients reaching the sea oor, in uence the reproductive cycle of organisms (94,95) in deep-sea habitats where the rest of abiotic conditions are constant along the year.
P. ventilabrum is an oviparous species with potentially lecithotrophic larvae (as other members of closely related orders) or even direct development, so all the nutrients are accumulated in the egg during vitellogenesis. Here we observed that the oocyte maturation phase in specimens of P. ventilabrum, during which yolk accumulation occurs, was synchronized with the predictable increase of energy stocks in the surroundings, as noted previously for other deep-sea organisms (96).
The yolk within the oocytes of P. ventilabrun was heterogeneous, composed of proteins and lipids. We do not have any information on reproductive strategies of other sponge species from the same genus, with the exception of a study on Phakellia hirondellei Topsent, 1980 (97). They brie y reported only protein platelets within the oocyte of P. hirondellei. The closest phylogenetically sponge species being studied more extensively is Raspaciona aculeata of the order of Axinellida, (98) with heterogeneous yolk, mainly of protein origin described in the oocytes. Protein synthesis has proven more costly energetically in cold waters than in temperate environments for adult isopods (99), but the cost for protein synthesis was very low during the early developmental stages of the Antarctic sea urchin Sterechinus neumayeri (100). There is no study calculating the metabolic cost of protein and lipid synthesis in sponges from cold habitats. However, lipid was the only type of yolk present in the Antarctic sponge Mycale acerata, and was considered an adaptation to cold environments since its Caribbean counterpart M. laevis had mainly heterogeneous yolk (55). Similarly, Geodia spp. from the boreo-arctic deepsea contained much more lipid yolk (87) than Geodia cydonium from shallow temperate waters (101). From the above we could hypothesize that the yolk origin in P. ventilabrum is governed both by phylogenetic constrains and adaptation to its boreal water environment, revealing higher amounts of lipid yolk than its counterparts.

Lipidome and gene expression variations during oogenesis
Reproduction is energetically a very costly process, and many changes on lipid metabolism have been observed in marine invertebrates during this period of time, with some lipids to increase and others to degrade (20,102). Indeed, some lipid categories constitute the nutrients for the future embryo (lipid droplets formed during phase of vitellogenesis) and they increase in female individuals suggesting their de novo synthesis and other lipids provide energy resources for the adult, so they are catabolized during gamete formation (oocyte growth, differentiation, nutrient formation and nutrient transport).
Our lipidomic analysis revealed that sponges remodel their lipidome during oogenesis following a tendency to i) increase TGs with SFAs and MFAs, ii) decrease glycerophospholipids, iii) decrease PUFAs with a high number of unsaturations and iv) increase the beta-oxidation of FFAs.
Almost half of the detected SFA and MFA TGs showed a tendency to increase their signal from NR towards the females with mature oocytes (Figs. 5, 6D) suggesting that this type of lipid could be the main component of yolk in this species.
TGs store much more energy (10 times higher/gram) than any other type of lipids and carbohydrates (103) (114). Furthermore, we observed a decrease of several of the free fatty acids that constitute these TGs, such as FA(16:0) or FA(18:0) (Fig. 6A), which could be explained by a more extensive incorporation of these fatty acids into TGs. Most of the genes regulating the enzymes of the TG biosynthetic pathway (most of which also participate in PL biosynthetic pathway) (115,116) were overexpressed in females with vitellogenic oocytes (either Vi_I or Vi_II or both), con rming the occurrence of de novo lipogenesis in female individuals, likely engaged in yolk formation ( Fig. 8; Additional le 10: Table S7; Additional le 11: Table S8). The gene regulating the penultimate enzyme in the TG biosynthesis: diacylglycerol acyltransferase (dgat) was the only one not signi cantly overexpressed, but still with a tendency of increased expression in females with Vi_I oocytes (Fig. 8A, C). GO enriched categories, KEGG pathways and other upregulated genes related to the long chain FAs and elongation of FAs were also overexpressed in females ( Fig. 8B; Additional le 8: Table S5, Additional le 11: Table S8; Additional le 12: Fig. S2), possibly also related to TG biosynthesis as TGs contain particularly long-chain FAs (117). In general, SFA and MFA FAs can generate a higher equivalent of ATP (118) and this might suggest that the enrichment of TGs with less unsaturated fatty acids is associated with a higher energy accumulation during reproduction.
In our analysis, even though the majority of PLs were decreasing their signal as oogenesis progressed, a ~20% of the PLs exhibited an increased signal (Figs. 5, 6, Additional le 3: Table S2), indicating that they could also have a role in yolk formation /oogenesis but in a more selective way than TGs. Possibly the different lipids participate in the formation of different fractions (homogeneous lipid platelets/ heterogeneous lipid-protein platelets) of the lipid yolk in sponges as it has been previously reported on the lizard Zootoca vivipara (named as Lacerta vivipara, in the publication) (119). Another potential role of PLs could be the formation of lipoproteins for TG transport. TGs have a hydrophobic nature and they are transported extracellularly or intracellularly in animals in the form of lipoproteins (120). Even though there are no studies discussing the presence and role of lipoproteins in sponges, it is known that lipoproteins have appeared early in evolution (120) and have been detected in ovaries of marine invertebrates, playing a role in lipid transport (16). So, we expect that lipid transport occurs in a similar way in sponges, including the TG transport. In our transcriptomic data, GO enriched categories ("very-low-density lipoprotein particle assembly" and "lipoprotein particle binding") and overexpressed genes related to lipoprotein formation (low-density lipoprotein receptor-related protein, lrp; low density lipoprotein receptor adapter protein 1, ldlrap1; low-density lipoprotein receptor, ldlr; apolipoprotein L3, apol3) in females, compared to NR (see paragraph 3 results) (Fig. 8A, C; Additional le 8: Table S5; Additional le 9: Table S6; Additional le 10: Table S7), strengthen our hypothesis. Furthermore, vitellogenin receptors, which are lipoproteins related to yolk formation in egglaying species (121), have been identi ed in other demosponges (122,123). In P. ventilabrum, genes expressing both vitellogenin and its receptor were three times more expressed in the Vi_I female than NR, although they did not qualify as differentially expressed in our analysis.
Lipids with high number of unsaturations (PUFAs with ≥ 3 unsaturations), including PUFA TGs, had their signal decreased towards females with more mature oocytes ( Fig. 6; Additional le 3: Table S2). The largest variation of PUFAs was observed in samples between the different locations/months (Fig. 4), indicating that these lipids are mostly related to nutrients in the environment. During oogenesis, these lipids might be used/oxidized to provide fast energy to the adult because they are more unstable than the SFAs or MFAs. Specimens that do not reproduce, potentially, store them for other future physiological functions and/or consume them at a slower pace than the reproductive specimens, and that is why we observed a higher signal in specimens without oocytes in their mesohyl (Fig. 6D). Even though the mechanism of PUFA oxidation in sponges is not known, based on the mechanism of peroxisomal PUFA oxidation in mammals (124), we could not nd any upregulated gene related to PUFA oxidation. In addition the decrease of FA(22:6), (normally produced after PUFA oxidation in mammals (125)) and of their oxylipins (HDoHEs), do not suggest either an increase of the PUFAs oxidation, at least in a similar way to what happens in mammals. It might be the case that sponges follow a different mechanism of PUFA oxidation, and further studies are needed to investigate this.
While the PUFA oxidation route in sponges might be novel, it is also possible that PUFAs in sponges are oxidized via the regular FA oxidation route in mitochondria. Indeed, we found a higher activity of genes related to FA beta-oxidation in females compared to NR ( Fig. 8B−C, Additional le 12: Fig. S2), including activation and transportation of FA to the mitochondria by acs; carnitine:palmitoyltransferase, cpt; and mitochondrial carnitine/acylcarnitine carrier protein, Slc25a20) (126) and their beta-oxidation to produce acyl-CoA, the main component entering the citric cycle. This overexpression indicates that there was an increase in FA beta-oxidation due to high energetic requirements during oogenesis. Indeed, most of the carbon accumulated in sponges (90%) is spent in generating energy for physiological processes such as growth, pumping and reproduction (Koopmans, 2009), and lipids are a large source used for oxidation in sponges as they consist of more than 50% of the dry weight of the particulate organic matter that sponges feed on (127). Therefore, we cannot discard alternative routes for lipid oxidation to provide energy in sponges. The extra enzymes required for the oxidation of unsaturated FA, either for odd or even number of carbons (Delta(3,5)-Delta(2,4)dienoyl-CoA isomerase, ech1; and 2,4-dienoyl-CoA reductase, decr1) were detected in our analysis, however, without overexpression in females. This suggests either that these FA might be absorbed for production of other lipids, or that maybe, again, there is a different mechanism for oxidising these lipids in sponges.
Lastly, the oxylipins seem to have a strong negative correlation with the female reproduction, and only the prostaglandins PGF 2 and PGF 2a showed an increase in their signal in the tissues with oocytes developing in P. ventilabrum. The hormonal regulation of reproduction has not been studied in sponges so far, but in mammals prostaglandins regulate several processes of female reproduction, including oocyte maturation, ovulation and fertilization (e.g. Sugimoto et al., 2015), and this could be potentially the case here.
One of the major limitations of our study is that we explored the lipidome and the gene regulation of the whole sponge and not isolated cells and so we observed the parallel physiological processes occurring in different cells at the same time. For instance, while in oocytes the generation of TGs might occur, in other cells oxidation of PUFA TGs and other lipids for energy production can probably happen. Further studies with a focus on targeted cells would be necessary to understand in depth these processes. In addition, the characterisation of the speci c structure of lipids and the tracing with isotope labelling would enhance our comprehension on the role of nutrient uptake from the surrounding to the pathways of FA incorporation for synthesis of other lipids, to the yolk formation during oogenesis in sponges, and their route of oxidation.

Conclusions
This is the rst study in which the reproductive strategy of the sponge P. ventilabrum, a keystone species of the vulnerable deep-sea sponge grounds of North Atlantic is investigated. It is also the rst application of semi-targeted lipidomics to a sponge, which allowed us to detect a total of hundreds different lipids by retention time and m/z, more sponge lipids than any other previous work. Furthermore, we detected lipids that are absent or in low amounts in mammals, such as phytosphingolipids and glycerophospholipids of methylethanolamine or dimethylethanolamine. This illustrates the outstanding diversity of the sponge "chemical dark matter" often di cult to comprehend, and that new metabolomic methods are just starting to reveal. The ndings of this study have important ecological and evolutionary implications.
Information for the reproductive activity and strategy of P. ventilabrum enhances our understanding on ecosystem services of sponge grounds and can contribute to developing conservation strategies in these areas. Additionally, the understanding of the lipid content dynamics during reproduction could also give an idea on the tness of the propagules and how inbalances in lipid resources in the environment could affect the tness of the species. Finally, this study provides essential information for understanding the origin of the lipid yolk content in Metazoa.  (Table 1). Collections were done either with a Remote Operated Vehicle (ROV) (Kosterfjord) or a triangular dredge (Langenuen/Korsfjord). No special permission was needed for the sponge sampling in Norway while a permit was given by Tjärnö Marine station, Sweden, for sponge collection within the national park at the Kosterfjord. Specimens were identi ed based on external morphology and spicules by P. Cárdenas and H.T. Rapp (slides for spicule preparations are available upon request to PC and AR). On board, three ~5mm 3 pieces were cut with sterile and RNAse-free instruments from different parts of each specimen and xed in 2.5% glutaraldehyde solution for histological analysis. Another three ~5mm 3 pieces, from the same specimens, were xed in RNAlater™ Stabilization Solution (Thermo Fisher Scienti c). For lipidomics, a large piece of each specimen was ash frozen in liquid nitrogen, transported in dry ice to Uppsala University and kept at -80°C until freeze-drying for the lipidomic analysis.

Histological Analysis
Sample preparation for histological analyses was done according to the protocol by Koutsouveli et al. (2020a, b). In brief, the fresh collected samples, once in the lab, were rst rinsed in a solution of 0.6M NaCl and 0.4M PBS and then post xed in 2% osmium tetroxide in 0.4M PBS for two hours and incubated overnight in 4% hydro uoric acid (HF) to remove any silica structures from their skeleton. Afterwards, samples were rinsed with distilled water and dehydrated with ethanol in an ascending series (50-70-96-100%). For light microscopy, samples were then embedded in para n blocks and those were cut with a HM325 microtome (ThermoFisher Scienti c) into sections of 5 μm, which were stained with a standard Harris' Hematoxylin and Eosin (HandE) protocol. Histological sections were observed with an Olympus microscope (BX43) with an attached UC50 camera. For transmission electron microscopy (TEM), samples were embedded in LRW resin blocks (agar Scienti c) (according to the guidelines of the manufacturer) and those cut at 60 nm in a Ultracut Reichert-Jung ultramicrotome. Then, the sections were stained with 2% uranyl acetate/lead citrate (129), and observed with a Hitachi TEM Microscope (H-7650) at 80kV.
We then measured the size and number of gametes within reproductive females with the Olympus Microimaging software CellSens standard, integrated to the Olympus microscope, on the histological sections. Several tissue areas (0.58 mm 2 each area) of each section were surveyed for the measurements, avoiding to outnumber the density of the gametes within the tissue. All the images for measurements were taken at 10x magni cation. To extract the nal measurements, we calculated the average and the standard deviation of the size of the gametes from all the different images. We also conducted quanti cation of the different types of yolk content within the oocytes with ImageJ (130) on the TEM images.
Information extracted from the histological observations of the reproductive specimens, regarding the developmental stage of their oocytes, was further used for the lipidomic and transcriptomic analyses. Speci cally, for the lipidomic analysis the area of sponge tissue occupied by oocytes was taken into account in order to observe the variation of lipid signal. As the number of oocytes did not change in the different developmental stages of oogenesis, the increase of sponge tissue occupied by oocytes was proportional to the maturation stage of the oocytes. For transcriptomic analysis, individuals with reproductive elements were classi ed to Vi_I or Vi_II stages, based on the developmental stage of their oocytes observed in the histological sections, while nonreproductive specimens (NR) did not have any gametes. Vi_II stages contain the most mature oocytes, which are larger in size and have greater yolk content than the Vi_I individuals.

Semi-targeted Lipidomics
The samples from the different locations (Langenuen/Korsfjord and Kosterfjord) were processed in different batches.
Frozen subsamples were freeze-dried and grinded in a falcon tube; 52±2 mg and 22±1.2 mg of powder was extracted for the samples of Langenuen /Korsfjord and Kosterfjord respectively (Additional le 13: Table S9A). The extraction was done with chloroform, based on Bligh and Dyer protocol (131). Brie y, we incubated the samples in 1:2 chloroform/methanol (v/v) overnight. The organic phase (lower phase) was collected after centrifugation, mixed with 375 μL of CHCl 3 and incubated for another 24 hours. The organic phases from the two extractions were pooled (1 mL) together and dried under vacuum. The samples were resuspended in 200 µL of ACN/IPA 50:50. We also prepared the quality control (QC) in which we mixed 10 µL from all the extracted specimens. The separation was slightly modi ed from (132). Brie y, lipids were separated by Ultra High-Performance Liquid Chromatography (UHPLC) with a BEH C18 column (1.7 µm, 2.1x150 mm) on an Acquity chromatographer hyphenated to a Synapt G2S QToF (Waters, Manchester UK). The mobile phases were A) water/acetonitrile/isopropanol 40:30:30 (v/v/v) with 5 mM of ammonium formate, and B) acetonitrile/isopropanol 40:60 (v/v) with 5 mM of ammonium formate. The gradient ( ow 0.275 mL min -1 ) changed linearly as in Additional le 13: Table   S9B. The ionization was carried out by electrospray in positive and negative modes. The injections of QC were four at the beginning and at the end of the injections as well as between every ve injections of the samples. During the extraction protocol and injection, the order of the samples was randomized.

Oxylipin quanti cation
For the extraction of oxylipins we followed the protocol by (133). In detail, 30 ± 6 mg of tissue for each specimen (Additional le 13: Table S9A) were mixed with 2 mL of methanol and sonicated for 30 min on ice to enhance the extraction. After centrifugation (15 min at 3,000 g), the supernatant (1.5 mL) was transferred into Pyrex extraction tubes.
We evaporated the solvent with nitrogen (TurboVap LV) until reaching an approximate volume of 300 μL and then we added 2.7 mL of Solid Phase Extraction (SPE) buffer (pH 5.6). SPE was performed in order to eliminate any interferences that could add background signals to our analysis using an Extrahera automated system. After loading the column with the extract, we washed with 2. with 0.1% of acetic acid and B, acetonitrile/isopropanol 90:10. Column temperature was 60ºC and ow rate was 0.5 mL min -1 with a gradient initiated at 90% of A that changed linearly to 65% in min 3.5, to 60% in min 5.5, to 58% in min 7, to 50% in min 9 and to 35% in min 15. Further details about the mass spectrometry parameters can be found in (134).

Data pre-treatment
For lipidomics data-pretreatment we followed (132). Brie y, Water's raw mass spectrometry les were transformed with Data Bridge into CDF les and processed with XCMS package in R (135,136). To identify the lipids, a database of lipids from Lipid Maps was used (137). For this database, the m/z of the different adducts was generated in R by package Rdisop in R (138). For every lipid, the family the adduct, the total number of carbons and the total number of unsaturations were identi ed from the combination of m/z and retention time of the rst and the second isotopologues (Additional le 1: Table S1). When possible, because of the intensity, the fatty acids were identi ed by the fragmentation patterns (Additional le 13: Table S10). The lipid signal was quanti ed as the area under the chromatographic curve of the peaks.
For every specimen, the lipid signal was normalized by the weight of sample extracted in mg.
Regarding oxylipins, for the quanti cation, a calibration curve with 11 external points, spiked with internal standards was used (133,134). Concentration calculations were performed in TargetLynx (Waters, Milford, USA). Finally, the amount of oxylipins were normalized as ng of oxylipin per g of dry sponge extracted.

Statistical Analysis
First, to study the factor location/month, we used a principal component analysis (PCA) and we applied multiple t-tests for To isolate the effect of the reproduction status from location, we performed multivariate regression on the logarithm of the signal of the lipids of all the identi ed lipid categories by using the percentage of the area of sponge tissue occupied by oocytes as predictor. Consequently, the coe cient for the percentage of the area of the sponge tissue occupied by oocytes represented the upregulation (positive coe cient) or downregulation (negative coe cient) of the lipid with the development of oocytes (Additional le 3: Table S2). From nonreproductive to female individuals with most mature oocytes in their tissue, the area of the tissue occupied by oocytes increased. For the analysis, each lipid category was analysed (Additional le 1: Table S1; Additional le 3: Table S2).
All the statistical and graphical analyses were conducted in R (R Core Team 2017).

RNA extraction and library preparation
Given logistic and sampling limitations, only four samples of P. ventilabrum from Langenuen/ Korsfjord were used for the transcriptomic analysis (Table 1, with asterisks): two females and two nonreproductive specimens. The protocols used for RNA isolation, mRNA puri cation and cDNA library preparation were described in Koutsouveli et al. (123). Brie y, total RNA extraction was conducted with a standard TRIzol™ Reagent (ThermoFisher Scienti c) protocol, according to the guidelines of the manufacturer. Further mRNA puri cation was performed with the Dynabeads mRNA DIRECT kit (ThermoFisher Scienti c), applying the nal stage of the protocol, 'Elimination of rRNA contamination'. The quantity and quality of mRNA were assessed by NanoDrop 2000 (ThermoFisher Scienti c). Then, cDNA libraries were prepared with Scriptseq v2 kit (Illumina) (according to the manufacturer's instructions), using an initial mRNA quantity of 50 ng. The amount of cDNA was then assessed with Qubit™ dsDNA HS Assay kit (ThermoFisher Scienti c) and the quality with an Agilent Tapestation 2200 system (Agilent Technologies). The sequencing was done in an Illumina NextSeq 500 platform at the Natural History Museum of London sequencing facility (Molecular Core Labs).

Assembly, Differential Gene Expression Analysis, Annotation
Filtering of reads based on quality was performed with Trimmomatic (139), and the de novo assembly was done with Trinity v2.8.4 (140). Completeness of the assembly was calculated with Benchmarking Universal Single-Copy Orthologs (Busco V2/3) against metazoan cassettes (141). For gene expression analyses, mapping of the reads to the reference assembly was performed with Bowtie2 (142), transcript quanti cation was done with RSEM (143), and differential gene expression (DGE) analysis was conducted with edgeR (144,145). We did pairwise comparisons of each female specimen vs the two nonreproductive specimens, as the two female specimens were in different developmental stages (Vi_II containing more mature oocytes than Vi_I). Therefore, due to lack of replication for the reproductive condition, we used dispersion 0.1 for the conduction of DGE analysis with edgeR.

Gene Ontology and KEGG enrichment analysis
For the transcriptome annotation, we did a blastx search (146) of the transcriptome using the swissprot database (147) containing only metazoan proteins (accessed in 2020), using Diamond (148) with a cut-off evalue of 1e-5. The sequences with blast hits were further annotated by Blast2GOPRO (149) to retrieve the functional information from the Gene Ontology (GO) terms.
We then performed a GO enrichment analysis using a Fisher's Exact Test in Blast2GOPRO (149) with a p-value threshold of ≤ 0.05. This analysis was conducted using as "test dataset" the upregulated genes of each female developmental stage separately (Vi_I, Vi_II) vs nonreproductive individuals and as "reference dataset" the total annotation le of the reference transcriptome. The percentage of sequences contained in each GO term was extracted and used for the depiction of a bubble graph in R (R Core Team, 2018). In some cases, more than one sequence was linked to the same process, so we added the relevant sequences to have a nal summary percentage.
In addition, we performed a KEGG analysis (150) in Blast2GOPRO (149)        sphingolipids; F. oxylipins. The x-axis represents the number of unsaturations of the lipids and the y-axis the coe cient that relates the signal of the lipid with the surface of oocytes. Lipids above the dashed line were upregulated with oocytes, lipids below this line were downregulated with oocytes. To distinguish major from minor species in the same family, the area of the points is proportional to the signal of the lipid in the control group. The number next to the points indicates the number of carbons of the lipid.