Biological-function clustering of HFF proteins

Among the 617 factors that the extensive literature search designated as HFF unique and nonambiguous proteins, 337 protein factors (Supplementary Table II) were clustered into five, statistically significant (P < 0.001), main groups by the DAVID enrichment analysis tool according to their GO biological functions (Supplementary Methods and Table II). Protein distribution into these groups was visualised using a five-way Venn diagram by the jquery.venny software (http://bioinfo.genotoul.fr/index.php?id=116) (Fig. 1). Such groups exactly match with those we described in a previous study (Ref. Reference Bianchi17) in which HFF was investigated by a proteomic and bioinformatic approach, such as (i) inflammation/regulation of inflammation/acute phase, (ii) response to wounding, (iii) complement and coagulation cascades, (iv) protein–lipid complex/lipid metabolism and transport and (v) cytoskeleton organisation.

Figure 1. Five-way Venn diagram showing reviewed HFF proteins in common among the five main gene ontology groups in which HFF proteins have been clustered by DAVID.

Inflammation/acute phase and response to wounding are the most representative classes including, respectively, 172 and 136 proteins, that are 28 and 22% of the HFF proteins. Anyway, this is not surprising. Mammalian ovulation is actually assimilated to an acute inflammatory reaction, culminating with the release of mature oocyte, which is followed by coagulation and tissue repair processes to support corpus luteum formation (Refs Reference Angelucci9, Reference Twigt14, Reference Field18).

In this regard, by DAVID analysis we pointed out that 72 proteins out of the 83 factors that were clustered in the complement and coagulation cascades are known to be active also in inflammation and/or wound response (Fig. 1). Therefore, several inflammation, coagulation and tissue repair active proteins are profoundly integrated in the follicle microenvironment, suggesting that balanced reactions among these proteins regulate physiological ovarian functions.

About 13% of HFF proteins we found out in Literature are involved in lipid transport and metabolism: triglycerides, cholesterol esters, phospholipids and nonesterified fatty acids are key molecules in a number of follicle processes, including hormonal responses and oocyte competence acquiring (Refs Reference Ambekar13, Reference Cataldi19, Reference Valckx20). Moreover, approximately half of the proteins belonging to this group interact with inflammation/acute phase, response to wounding, and coagulation groups (as detailed in Supplementary Table II and visualised in the Venn diagram, Fig. 1). In line with these results, well known inflammatory cytokines that control the immune system and the inflammatory response are reported to be able to modulate glucose and lipid metabolism, namely the adipokines. Several of them have been detected in the HFF, such as: IL-6, Tumor necrosis factor (TNF)-α, plasminogen activator inhibitor-1 (PAI-1), retinol-binding protein 4 (RBP4), C-reactive protein, etc. Furthermore, TNF-α and IL-6, as well as other cytokines, also play key roles in the activation of coagulation (Ref. Reference Esmon21). Interestingly, a number of pro- and anti-inflammatory molecules, active in immune cell trafficking and signalling as well in TNF-α activation and prostaglandin synthesis, have been proved to be expressed by theca, granulosa, and/or luteal cells and, in some measure, even by the oocyte (Refs Reference Urieli-Shoval7, Reference Andersen22, Reference Wissing23).

Finally, about 7% of HFF proteins, which were found out by our revision process, are involved in cytoskeleton organisation. Their presence into collected fluids may be because of a certain level of HFF contamination by follicular cells in consequence of physiological cell rupture during the pre-ovulatory phase and of cell damage caused by medical-procedure for oocyte retrieval.

In conclusion, about 24% of known HFF proteins are active in more than one of the above described four main clusters. These shared factors operate as functional ‘molecular-bridge’ among the delineated functional-groups, and their altered expression in HFF may profoundly interfere with fertility and assisted reproduction technology outcome. Moreover, such a tight cluster intersection underlines how the reproductive axis is subjected to systemic and follicular highly balanced integration of energy metabolism, inflammation, and tissue injury and repair.

Qualitative and/or quantitative variations of those factors, even if heightened by the hyperstimulation treatment and oocyte retrieval procedure, may impact on follicle and oocyte development and maturation in IVF programs thus justifying their evaluation as biomarkers for oocyte quality estimation (Refs Reference Jarkovska16, Reference Bianchi17). Moreover, some HFF factors may quantitatively change in consequence of different responsiveness of individual patients to hyperstimulation protocols and/or to different causes of infertility. This may be of valuable impact not only on the assessment of oocyte quality, but also on individualised treatment planning for the eventual second cycle of ovarian stimulation.

The complexity of biological interactions occurring among HFF proteins and, in particular, their functional hierarchy have to be clarified in order to decipher how HFF multifunctional factors reciprocally interact and operate into the follicle microenvironment. We attempted to achieve this result applying pathway analysis.

Pathway analysis of HFF proteins

A total of 507 out of the 617 HFF unique proteins were unambiguously supported by the MetaCore network building tool v. 6.9. Their functional processing, by the ‘direct interactions’ algorithm, generated a statistically significant net of 315 elements reciprocally cross-linked by 776 interconnections (Fig. 2). To properly understand the generated net, MetaCore synonyms (applied in the net to name nodes) are reported in Supplementary Table I along with the conventionally accepted gene nomenclature. We also performed pathway analysis applying the ‘shortest path’ algorithm. In this analysis, the program introduces into the net specific molecular factors, which are not listed by the user among proteins to be processed but that are required, according to literature, to cross-link proteins under investigation. By this approach, the statistically significant generated net includes 370 proteins that are linked by 938 interconnections (Fig. 3). However, the core of both direct and shortest path networks integrate approximately the same number of proteins (301 versus 303) and share the same main central hubs, underlining limited variations occurring between them.

Figure 2. MetaCore ‘direct interaction net’ highlighting how HFF proteins establish, according to literature, direct and reciprocal functional-interconnections. Network proteins are visualised by proper symbols, which specify the functional nature of the protein (network legend). Edges define the relationships existing between individual proteins, and arrowheads represent the direction of interactions.

Figure 3. MetaCore ‘shortest path’ net. The main central hubs are circled: MMPs (green), Thrombin (red) and VDR/RXR-α (light blue). To distinguish the functional factors added by the program, reviewed HFF proteins are encircle in blue. Several net nodes are collapsed in a representing protein used by the software to group two or more proteins linked by logical relations or physical interactions. Thus representing factors are marked by yellow little star attached to the factor visualising symbol. For network legend see Figure 2.

Functional pathways, in both nets, essentially converge on matrix metalloproteinases (MMPs), thrombin and micronutrient receptors, in particular vitamin D receptor and retinoid X receptor-alpha heterodimer (VDR/RXR-α) in the shortest path network. These proteins were the principal central hubs of the generated networks. Since the highest complexity of functional-interconnections was generated by the shortest path algorithm application, here we merely investigate the shortest path net.

MMP system in modification and reorganisation of the extracellular matrix (ECM)

The ECM is a highly complex and dynamic mesh of fibrous proteins and glycosaminoglycans. Its degradation and remodelling requires the combined action of several hydrolases, and correlated inhibitors, with different substrate specificities. MMPs are key enzymes in such processes, degrading several ECM and non-ECM components (Ref. Reference Nissinen and Kähäri24). The MMP system, along with the fibrinolytic one, is actually retained the main proteolytic systems in ECM dynamics.

MMPs are mainly synthesised, as ProMMP precursors, by macrophages, mast cells and neutrophils. They are proteolytically activated by intracellular furin or by other MMPs and serine proteases, such as plasmin, thrombin, and neutrophil elastase in the extracellular milieu (Refs Reference Shamamian25, Reference Jackson26, Reference Koo27). ProMMPs may be activated also by conformational changes, without proteolytic processing, that are induced by chemical agents, such as reactive oxygen species (ROS) and reactive nitrogen species (RNS) (Ref. Reference Okamoto28).

MMPs are the principal nodes of the network where the gelatinases MMP-2 and MMP-9 represent the two main central hubs (Fig. 3). MMP-2 and MMP-9 respectively establish 95 and 84 distinct interconnections with other net-molecules, which represent more than the 40% of HFF proteins entered into the net. The majority of edges occurring among gelatinases and other HFF proteins consists in ‘outgoing’ connections: from MMPs toward fluid components. Hence, gelatinases impact on ovarian physiology by a direct positive or negative control on HFF factors while, in turn, they are regulated by a very small number of proteins, several of which are other MMPs (Fig. 3).

Also stromelysin-1 (MMP-3) as well as, MMP-1, -25, -13 and -12, with even fewer edges, occur as central hubs of the network (Fig. 3). Several of the proteins cross-linked to these proteases are in common with MMP-2 and MMP-9; hence, the fraction of fluid proteins directly linked to MMPs slightly increase, reaching about 45% of net factors (Fig. 3).

In line with our pathway analysis data, MMPs have been reported as key proteins in follicle bioactivities triggering proteolytic cascades throughout development, maturation, apex rupture, corpus luteum formation and regression, and follicle atresia (Table 1) (Refs Reference Smith29, Reference Curry and Osteen30). Accordingly, MMPs failure or reduced activity has been called into question for ECM derangements in subfertility and infertility, for aberrant follicle growth and deficiencies in ovulation (Refs Reference Smith29, Reference Liu31, Reference Liu32, Reference Ebisch33, Reference Curry and Osteen34, Reference Peluffo35, Reference Goldman and Shalev36). Controlling ECM organisation, MMPs are actually involved in the bioavailability control of hormones, growth factors, and osmotically active molecules by regulating local accumulation and activation of these factors, as well as cell response to extracellular stimuli and cell to cell communications (Refs Reference Ambekar13, Reference Fan37, Reference Visse and Nagase38, Reference Nagase39, Reference Taipale and Keski-Oja40, Reference Armstrong and Webb41). As a result, MMPs ‘escaped from control’ may have severe negative effects on follicular and oocyte development and maturation causing excessive degradation not only of the follicular ECM, but also of resident growth factors and co-factors, as well as cell-surface receptors on follicle somatic cells and, possibly, on the growing oocyte, too (Ref. Reference Nissinen and Kähäri24).

Table 1. Main activities and processes in which MetaCore central hubs are involved in Human Follicular Fluid

MMPs aberrant activity may also impact on cumulus–oocyte complex (COC) and zona pellucida (ZP) integrity. Based on the tight metabolic dependence of the growing oocyte from cumulus cells as well as on the firm control that the oocyte exerts on cumulus cells physiology (Ref. Reference Sánchez and Smitz42), functional integrity of COC gap junctions is fundamental for oocyte development and maturation. MMPs ‘uncontrolled’ excessive activity may hence lead to the degradation of intra-cumulus and cumulus/oocyte junctions, with consequent impairment of oocyte maturation and competence acquiring. The access of MMPs to the ZP may also result in its partial degradation thus affecting sperm–oocyte interaction and fertilisation.

The selective proteolytic activity of MMPs is usually balanced by endogenous inhibitors alpha(2)-macroglobulins, and by tissue inhibitors of metalloproteinases (TIMPs) (Refs Reference Baker, Edwards and Murphy43, Reference Brew and Nagase44). TIMP altered expression results in deregulation of ECM degradation and turn-over, but it may also impact on ovarian cell proliferation, differentiation and vascularisation during folliculogenesis (Ref. Reference Curry and Osteen34). Literature data proved TIMP production by theca and granulosa cells, ovary surface epithelium, corpora lutea, blood vessels and even by the oocyte (Refs Reference Shores and Hunter45, Reference Zhang, Moses and Tsang46, Reference Fedorcsák47). In this respect, a down-regulation of TIMP-1 may have a detrimental effect in the ovulation and an altered MMPs/TIMPs ratio may lead to a precocious regression of the corpus luteum with consequent depletion in progesterone and estradiol synthesis, thus to reduce the negative feedback to LH pituitary secretion (Ref. Reference Goldman and Shalev36).

The inhibitory activity of TIMPs, is also decreased by peroxynitrites, which are generated by nitric oxide (NO) reaction with superoxide anion radical (O₂ ⁻) (Ref. Reference Donnini48). RNS as well as ROS have been described to be active in folliculogenesis, designation of the dominant follicle, meiosis progression, ovulation, as well as corpus luteum formation and regression (Ref. Reference Agarwal49). NO is known to act in follicular maturation and in controlling blood flow of the follicle (Refs Reference Anteby50, Reference Rosselli, Keller and Dubey51). Here, it is produced by theca, granulosa/luteal, and capillary endothelial cells and significant high levels of NO have been detected in HFF (Ref. Reference Rosselli, Keller and Dubey51). NO and ROS ‘acceptable’ threshold values are index of healthy metabolic state of the preovulatory follicle and they may positively impact on IVF outcome (Ref. Reference Pasqualotto52). On the contrary, inappropriate levels of ROS/RNS or inadequate total antioxidant capacity in ovary and follicle, have been suggested to adversely influence female fertility (Refs Reference Agarwal49, Reference Appasamy53, Reference Tamura54). In several tissues, RNS and ROS have cytotoxic effects determining membrane protein oxidation, lipid peroxidation and DNA damage (Ref. Reference Valko55). Interestingly, in other biological systems, some evidences identify in ROS-dependent lipid peroxidation a ‘cofactor’ of MMP-1, MMP-2 and MMP-9 up-regulation (Refs Reference Kim10, Reference Polte and Tyrrell56). Moreover, intracellular activation of MMP-2 by reactive oxygen/nitrogen species and consequent intracellular protein digestion has recently been elucidated (Refs Reference Jacob-Ferreira and Schulz57, Reference Soslau58). Being MMP-2 constitutively expressed in the follicle, a similar modulation of MMP activity may also occur in the ovary, with some implications in reproduction failure. Consistent ROS levels in HFF have been correlated to decreased fertilisation, embryo quality and pregnancy rate (Refs Reference Das59, Reference Chattopadhayay60, Reference Jana61).

ECM remodelling and complement/coagulation, response to wounding and inflammation pathways are functionally integrated by thrombin

Thrombin, as properly stressed by our network analysis, has emerged as a basic integration factor in HFF operating, at various levels, in the complement/coagulation system, response to wounding and inflammation (Refs Reference Lane and Bailey62, Reference Chen and Dorling63, Reference Amara64), as well as in ECM remodelling by inducing expression and activation of MMPs (Refs Reference Duhamel-Clérin65, Reference Galis66, Reference Orbe67, Reference Huang68) (Fig. 4). Accordingly, thrombin resulted, with its 29 interconnections, one of the main central hubs of the generated net (Fig. 3).

Figure 4. Schematic representation of major functional-interconnections occurring among key hubs in the HFF net. The schema provides a direct and simplified overview on the complex crosstalk exerted by MMPs, Thrombin and VDR/RXR-α in controlling, modulating and balancing inflammation, coagulation, healing process and lipid metabolism in the follicular microenvironment, as highlighted by pathway analysis.

Thrombin acts a key role in coagulation cascade by processing fibrinogen to generate fibrin and being the most potent platelet activator. However, thrombin is mainly generated after clotting and it has been indeed hypothesised to operate also after coagulation during tissue repair. It actually control even indirectly, through MMPs, the fibrinolytic process and regulate fibrin invasion, as proved in animal models, thus contributing to healing processes (Refs Reference Duhamel-Clérin65, Reference Hotary69, Reference Green70). By protease-activated receptor (PAR) signalling, thrombin induces proliferation, expression of growth factors, and ECM deposition (Ref. Reference Schrör71). Thrombin indeed modulates ECM-remodelling determining matrix degradation by MMP activation, but also inducing ECM formation. Accordingly, it may have implications in corpus luteum organisation and in IVF outcome. To date, fibrinolysis impairment has been correlated to implantation failure and recurrent pregnancy loss (Ref. Reference Kutteh and Triplett72).

Thrombin may influence reproductive performance also because of its involvement in regulation of inflammation. Theca, granulosa, and cumulus cells are, through PARs, thrombin responsive (Refs Reference Hirota73, Reference Osuga, Hirota and Taketani74, Reference Cheng75). In target cells, proteolytic activation of PAR receptors results in pro-inflammatory cytokine secretion (e.g. IL-1, IL-6, IL-8 and TNF-α), anti-inflammatory cytokine down-regulation, and in recruitment of leucocytes also by monocyte chemotactic protein-1 (MCP-1) production (Refs Reference Chen and Dorling63, Reference Strande and Phillips76, Reference O'Brien77). A proper time-dependent and balanced activity of cytokines is absolutely required in physiological maturation and rupture of the follicle. On the other hand, aberrant cytokines activity may profoundly affect follicle dynamics, as in the case of TNF-α. Elevated intra-follicular TNF-α concentrations have been correlated to reduced development and inhibited maturation of the oocyte as well as to decreased embryo competence and quality (Refs Reference Lee78, Reference Jackson, Farin and Whisnant79). TNF-α can also induce apoptosis and follicular atresia in presence of proper receptors, specific for the apoptotic pathway, that are exposed on granulosa cells and even on the oocyte plasmalemma (Refs Reference Naz, Zhu and Menge80, Reference Hussein81). Moreover, TNF-α also controls NO concentration by the induction of NO synthase 2 activity (Refs Reference Lirk, Hoffmann and Rieder82, Reference Obermajer83). High levels of NO in HFF have been associated to embryo lower grading and reduced implantation and pregnancy rate (Ref. Reference Vignini84). Finally, TNF-α and ILs induce MMP expression and activation (Refs Reference Nissinen and Kähäri24, Reference Kothari85, Reference Gearing86). While proper-operative MMP pathways are crucial in the ovary, MMP excessive activity negatively correlates, as above stated, with reproductive potential.

In conclusion, as highlighted by our functional analysis, Thrombin absolves key functions in folliculogenesis by performing, controlling and triggering several biochemical and molecular processes that are known to be essential in the physiological functioning of the ovary. Indeed, thrombin dysregulation, with consequent detrimental effects on inflammation, coagulation and healing process balancing, may negatively impact on fertility.

Micronutrients: key factors in follicular central processes

According to the DAVID cluster analysis, proteins that are active in lipid metabolism constitute one of the major classes in HFF. In the net, several of those factors are centred in the VDR/RXR-α receptor complex, hence suggesting the existence of a close functional correlation among proteins involved in lipid metabolism and micronutrients. VDR/RXR-α receptors belong to the steroid/thyroid hormone/retinoid nuclear receptors (NRs) super-family and their heterodimerisation regulates the expression of genes coding for proteins involved in numerous cellular functions, such as proliferation, differentiation, apoptosis and angiogenesis. Intriguingly, as highlighted by our pathway analysis, VDR/RXR-α directly or indirectly controls MMPs/TIMPs balancing, the plasminolytic, fibrinolytic and cathepsin/cystatin systems, vasodilatation, interleukins, TNF-α and complement activity (Refs Reference Alroy, Towers and Freedman87, Reference Alvarez-Díaz88, Reference van Greevenbroek89, Reference Jablonski90, Reference Hakim and Bar-Shavit91) (Fig. 4). Such a high functional integration of the VDR/RXR-α dimer in follicular dynamics has been further clarified visualising the shortest-path net according to the ‘subcellular localisation’ of entered nodes, as shown in Figure 5. Here, the VDR/RXR-α operates as a molecular ‘puppeteer’ that directly or indirectly controls several cellular and extracellular proteins in the follicle. Indeed, vitamins D and A, emerge to have a pivotal role in follicular processes influencing reproductive performance (Refs Reference Zile92, Reference Cetin, Berti and Calabrese93, Reference Pauli94, Reference Rudick95).

Figure 5. MetaCore ‘shortest path’ net visualised according to the subcellular localisation layout. VDR/RXR-α, which is cycled in light blue, evidently emerges as a key transcriptional factor with pivotal role in controlling numerous downstream proteins. MMPs and Thrombin are highlighted in green and red, respectively. For network legend see Figure 2.

Vitamin A is a generic term to indicate a large class of related compounds that cannot be de novo synthesised by animals and whose intake can occur only through nutrition. Retinoic acid (RA) is vitamin A active form that controls gene transcription through specific NRs: all-trans-RA binds retinoic acid receptor (RAR-α, β and γ) and 9-cis-RA binds retinoid X receptor (RXR-α, β and γ) (Ref. Reference Lane and Bailey62). Unliganded RXR may dimerise not only with RAR, but also with thyroid hormone receptors (THR) and with VDR (Ref. Reference Lane and Bailey62). Liganded RXR is instead a heterodimeric partner of other NRs, such as peroxisome proliferator-activated receptor (PPAR) and liver-X receptor (LXR), both involved in lipid metabolism and homeostasis (Refs Reference Sugden and Holness96, Reference Nakamura, Yudell and Loor97).

Vitamin D is a secosteroid hormone whose precursor is an intermediate of the cholesterol metabolism. Its active form, the 1,25-dihydroxyvitamin D₃ (1,25(OH)₂D₃), also known as vit D3, interacts with the NR VDR to perform its own functions. VDR has been hypothesised to control several human genes, even up to 5% of the entire genome (Refs Reference Ramagopalan98, Reference Hossein-nezhad, Spira and Holick99). Consequently, vitamin D aberrant endogenous production or dietary up-take may profoundly influence the physiological state of the organism. Besides its well-known involvement in calcium and phosphorus homoeostasis, vitamin D has been suggested to influence physiological processes in several different tissues, including male and female reproductive systems (Refs Reference Lerchbaum and Obermayer-Pietsch100, Reference Wang, Zhu and De Luca101).

In addition, vitamin D supplementation has been recently demonstrated to increase the serum soluble form of receptor for advanced glycation end-products (sRAGE) (Ref. Reference Irani102). These soluble receptors exert a ‘scavenger’ activity for the advanced glycation end-products (AGEs), which are potent cytotoxic metabolites. sRAGE have been also detected in HFF and their increased concentration has been associated to positive IVF outcomes (Refs Reference Fujii and Nakayama103, Reference Malicková104, Reference Bonetti105). Thus, by inducing the expression of sRAGE through VDR/RXR-α, vitamin D may counterbalance the AGE/RAGE system. Actually, the scavenger activity of sRAGE reduces AGEs detrimental effects by preventing their interaction with cellular specific receptors (cRAGEs), which are included in all the generated nets (RAGE node in Figs 2, 3 and 5) (Refs Reference Hudson106, Reference Tatone107).

AGEs are implicated in inflammation, ovarian aging, PCOS pathogenesis, and metabolic syndrome effects (Refs Reference Tatone107, Reference Yan and Boyd108, Reference Diamanti-Kandarakis109, Reference Kellow and Savige110, Reference Vlassara and Uribarri111). They are potent pro-inflammatory molecules that cause the generation of ROS along with an enhanced expression of IL-1, IL-6 and TNF-α in macrophages (Ref. Reference Singh112). Ovary has interestingly emerged to be a target tissue of AGE deposition and the interaction of these compounds with cRAGE, expressed by theca and granulosa cells (Ref. Reference Diamanti-Kandarakis113), determine cell dysfunctions and anomalous follicular growth (Ref. Reference Merhi114). The intrafollicular accumulation of AGEs was described to negatively impact on embryo growth and pregnancy rate (Ref. Reference Jinno115).