Our transcriptomic analyses suggest, as expected, that early salmon aSVF culture contains a number of cell types, including vascular cells, macrophages and lymphocytes, in addition to preadipocytes. Similar cell composition of aSVF was reported in mammals . The gene expression profiles at days 1 and 4 implied retention of multipotency of aSVF cells during the early stages of the culture. Decrease in gene expression of markers of osteo/chondrogenic, myogenic, immune and vasculature cell lineages (Fig. 5) and increase in adipogenic markers indicate that the preadipocyte precursors, probably due to their active proliferation, become the preponderant cell type at confluence.
Recent studies in mammals showed that tissue-resident MSCs originate from the smooth-muscle-like pericytes [39, 40], which are laid over junctions of endothelial cells in blood vessel walls . Once liberated from the endothelial cell, pericyte is activated and can be considered a stem cell . Characterisics of the WAT population of stem cells and early events in the determination phase are much less studied than the terminal differentiation phase of adipogenesis. The breakthrough in the understanding of elusive origins of the adipogenic MSCs was made only recently in a study of Tang et al. , which demonstrated that pericytes surrounding blood vessel walls in WAT are precursor cells of preadipocytes. Gradual attenuation of the vasculature specific genes glomulin and LIVE1 and increase of the pericytic marker TAGLN at days 7 and 9 support perivascular origin of salmon preadipocytes as well. Early PPARγ expression could represent an additional evidence of the pericytic identity of adipocytes' precursors in our model since Tang et al. confirmed PPARγ as a specific marker of the perivascular fraction in subconfluent aSVF . Indeed, our finding of high levels of PPARγ mRNA well before confluence contradicts the cell culture studies reviewed in  as the expression of C/EBPβ and C/EBPδ predate the expression of PPARγ and C/EBPα [17, 18], which then activate adipocyte-specific genes during terminal differentiation. This illustrates the advantage of primary cell cultures, which offer possibilities to gain valuable insight into the early molecular events of adipogenic differentiation.
Commitment of multipotent MSCs to the adipogenic lineage involves simultaneous inhibition of other mesenchymal lineages. Starting from day 6 we observed consistently decreased expression of MyoD, an early marker of the myogenic lineage. MyoD regulates the cell cycle in terminally differentiated myocytes by inducing CDK inhibitor p21, which irreversibly arrests proliferation . MyoD may play a similar role in mature adipocytes and in theory this could account for the increased expression at day 30. Activation of BMP4 from day 9 was an additional evidence for the inhibition of myogenesis. This growth factor equally promotes differentiation of the adipose, cartilage or bone lineages depending on the culture conditions [44, 45]. Therefore, activation of BMP4 is necessary but not sufficient for the commitment of preadipocytes. PPARγ is a master adipogenic regulator, which simultaneously inhibits myogenesis, osteogenesis and chondrogenesis [44, 46, 47]. Thus, high expression of PPARγ observed long before confluence with the peak at day 6 probably posed a barrier to the induction of non-adipogenic mesenchymal cell lineages. Non-adipogenic developmental pathways are expected to be strongly suppressed by the addition of hormones. Yet, we did not find any noticeable changes in the markers of different cell types after day 9 and earliest increased level of the anti-osteogenic GAS6 was observed at day 15.
Our PCNA results suggested the highest proliferative activity during the first days of culture followed by a moderate reduction of mitotic activity at day 6 and much greater decrease by day 30. Overall, the expression patterns of genes involved in cell cycle were in concordance with these changes. Highest levels of the key negative cell cycle regulators (e.g. RBL1, p53-like, GADD45γ) were observed during days 15-30. An additional evidence for the anti-proliferative status was the up-regulation of several pro-apoptotic genes, including retinoic acid-regulated apoptosis-related protein 3 (APR3). APR3 arrests cell cycle by suppressing the activity of cyclin D1 , which was also induced at the latest stage. Differentiation of adipocytes in mammals is associated with a reduction in proliferation at confluence followed by subsequent activation or clonal expansion and eventually with cell cycle arrest at terminal differentiation . Our results revealed the diversity of mechanisms potentially employed in the execution of the similar stage of events that occur during Atlantic salmon adipogenesis. Given that WAT is a highly specialized organ that functions to regulate metabolic homeostasis and energy balance, nutrient sensing mechanisms could be opted for a role in the control of proliferation. Growth cycle arrest of salmon adipocytes was in parallel with the up-regulation of the fatty acid binding protein 3 (FABP3) that participates in the uptake, intracellular metabolism and transport of long chain fatty acids. FABP3 seems like a good candidate that could be involved in this type of regulation since its growth arrest activity was previously demonstrated in mammalian epithelial cells . Further, minimizing protein synthesis may also control proliferation as cell cycle arrest is a well established consequence of the general translational arrest, implied from day 15 onwards.
Microarray analyses suggested changes in nutrient metabolism in the late phase of adipogenesis (day 30). As a consequence of the up-regulation of pyruvate dehydrogenase kinase that inactivates pyruvate dehydrogenase, the Krebbs cycle most likely relied on the preferential use of acetyl-CoA from FA oxidation rather than pyruvate to cover the cellular energy requirements. Also, the expression of genes involved in FA β-oxidation, acyl-Coenzyme A dehydrogenase and acyl-Coenzyme A binding protein, steadily increased towards the end of adipogenesis. The up-regulation of G6PD and PGD pointed to the importance of the PPP, which generates NADPH required for accumulation of lipids. The PPP can be seen as an alternative to glycolysis, because in addition to generating NADPH and R5P, it can also provide cells with G3P, the phosphorylated glycerol backbone suitable for TAG synthesis. Induction of glycolytic PFK was also in line with the increased production of G3P and TAG biosynthesis. Biosynthesis of carbohydrates in carnivourous fish species could be important due to the very low carbohydrate dietary level in their natural diets . Indeed, we have seen regulations in all major glucose metabolic pathways in our model (Fig. 8); however, differentiation of adipocytes did not involve induction of HK, one of the key enzymes of glucose metabolism that only showed slight up-regulation at day 30. This indirectly suggested that the transformation of gluconeogenic amino acids into glucose is important in adipocytes of Atlantic salmon, at least in vitro. In addition, late activity of two key glyceroneogenic genes, MDHC and PEPCKC implied that glyceroneogenesis, which can be fed with mitochondrial intermediates derived from lactate and gluconeogenic amino acids, in addition to pyruvate, was active in developing fish adipocytes. To conclude, these findings illuminate potentially novel aspects of nutrient metabolism in adipocytes, which might take advantage of the coupling of glycolysis, the PPP and glyceroneogenesis, in order to fulfil shifting demands for the three major products of these pathways: G3P, R5P and NADPH.
NADPH produced by PPP is also crucially important for the cellular antioxidant defence since it is required for the regeneration of oxidized glutathione by glutathione reductase (GSR). Up-regulation of GSR and a suite of genes involved in metabolism of glutathione, thioredoxin and iron was a prominent feature of late salmon adipocyte differentiation (days 15-30). Indeed, a recently introduced concept [52–54, 4] clearly demonstrated ROS as anti-adipogenic molecules that inhibit preadipocyte proliferation and differentiation. A reduced redox state in WAT is now recognised as a distinct characteristic of visceral obesity in mammals, with overexpression of GPX peroxidases and high content of hydrophilic antioxidant gluthathione associated with pro-adipogenic processes. Induction of the number of GPXs, known to be sensitive to lipid peroxidation, implied ROS-provoked response in our culture and the need to maintain highly reduced state of the intracellular environment. However, contrary to what could be expected in situations of high ROS production, neither gene expression nor activity of SOD increased. In fact, the activity of this important ROS scavenger progressively declined during adipocyte differentiation. Thus, our data corroborated the indespensible role of glutathione-based antioxidant system in the maintenance of the reduced intracellular state in white adipocytes of fish.
Accumulation of lipids in adipocytes coincided with the activation of genes coding for secreted proteins, such as adipokines adipsin and visfatin. Together with the previous finding of leptin secretion by Atlantic salmon adipocytes , this supports the notion of WAT as an active endocrine organ in fish. Enhanced secretion may impose an additional load on the ER actively engaged in lipid droplet synthesis . The ER compartment is suspected to exert a great deal of control over adipogenesis and association of obesity with ER-stress is firmly established . Up-regulation of XBP1 and ATF6 at day 15 in salmon adipocytes implied activation of UPR in response to perturbations in the ER homeostasis . Moreover, endoplasmic GRP78, which is located downstream from ATF6 in the UPR signalling cascade was down-regulated throughout the whole studied period but increased in expression at day 30. A number of changes that are typical of UPR  was a hallmark of the latest stages of adipogenesis in our model. UPR as a collection of pathways aimed at restoring ER function may serve different adaptive roles and is observed both under pathological and normal physiological situations ; for example, it is part of the developmental program in highly specialized secretory B and T cells  and ß-pancreatic cells . The revealed UPR in developing fish adipocytes appears as a highly tailored homeostatic mechanism that is activated when increased number and size of LDs meat the limited ER capacity. Although ER is mainly considered to be a protein-folding factory, all proposed models of LD formation emphasize the engagement of the ER compartment in the process .
UPR utilises two broad strategies to relieve stress in the ER: increased clearance of misfolded proteins from the ER and reduction of new protein influx into the ER. We found ample evidence for the activation of the 26S proteasome and autophagy-lysosomal proteolytic pathways, which degrade misfolded proteins. Components of the membrane protein complex, derlin 1 and p97-interacting cofactor p47, which mediate transport from the ER lumen into the cytosol were induced at day 30, while other proteolytic pathways were up-regulated even earlier. UPR-induced proteolysis may be employed to initiate the breakdown of superfluous proteins and organelles that must be replaced by more highly specialised cellular components during the adipocyte development. Without a doubt, considerable remodelling of the cellular architecture occurs during terminal differentiation of Atlantic salmon adipocytes.
UPR also activates genes involved in protein folding. Hsp40s and other J-domain-containing proteins act as co-chaperones for Hsp70s helping to restore homeostasis in the ER and in addition may have pro-degradation roles. Down-regulation of translational machinery is commonly observed during UPR , and in our study, it involved a dramatic reduction in expression of ribosomal proteins.
Alteration of the ribosomal composition is most certainly involved in the multiple regulation of the ER-stress response. The curious up-regulation of the 40S subunit 18S rRNA through most of the terminal differentiation against the majority of down-regulated ribosomal proteins is bound to have important consequences given the high level of 18S constitutive expression. This could signify a switch to a less efficient protein translation as increase in 18S negatively affects the ribosomal subunit ratio that favours protein production . On the other side, reduced expression of 18S at the very end of differentiation could improve the effectiveness of protein synthesis in adipocytes that acquire secretory phenotype. In conclusion, fully matured adipocytes seem to reprogram the pattern of gene expression to sustain a certain level of protein production, congruent with the now recognised endocrine functions of WAT.
Secreted adipsin and visfatin have multiple immune functions as most other adipokines . Overall, expression changes of a large number of immune genes were characteristic of Atlantic salmon aSVF culture. Dissimilar temporary profiles suggest different regulatory and physiological roles of these genes. Additionaly, high initial levels and subsequent decrease of the chemokine receptors, TNFα and a panel of TNF-related genes could be explained by the changes in relative abundance of non-adipogenic cells that possess pronounced immune properties. In mammals, TNFα has high anti-adipogenic activity , which was recently shown to be conserved in fish . Therefore, down-regulation of the TNF axis at days 7-9 was probably important for the onset of adipocyte differentiation. However, a number of immune genes showed either stable increase since day 4 (complement component C5) or biphasic regulation with the second activation after hormone induction (LECT2). Highest expression levels of many genes with diverse roles in different inflammatory pathways were observed during terminal adipocyte differentiation. Numerous independent studies have provided evidence that a set of inflammatory and stress-response genes is activated in obesity in mammals (reviewed in ). Some of these genes likely have important roles in the coordination of homeostasis in WAT, similar to the established roles of TNFα in modulating proliferative abilities in preadipocytes or metabolic activities in mature adipocytes.
An intriguing example is a group of virus responsive genes including TLR3, a receptor of double stranded RNA and a group of galectins, previously not reported to have roles in any aspect of adipocyte biology. The latter includes galectin 9, which showed highly specific responses to viruses in previous studies from our group [67, 68]. However, a large portion of immune genes, including complement component C5, JAK1, the key actor of interferon signaling and FLAP, the regulator of eicosanoid metabolism, decreased expression by the time of AP1 establishment. Finally, our results do not permit for major conclusions as to whether immune genes are important for adipogenesis or change expression as a consequence of differentiation and/or stress responses. However, it is clear that Atlantic salmon WAT possesses high potential for immune activity.
It is tempting to explain the activation of immune genes in differentiating adipocytes as a side effect of adaptation to the oxidative and ER stress, since stress and immune responses share common regulatory pathways. Ozcan et al.  showed that ER-stress-induced JNK-AP1 axis is the central link between TAG overload in liver and diabetes. Most interestingly, this is in agreement with our study, which shows that the same players are activated in fish adipocytes in response to increased lipid deposition. Continuous administration of lipids led to the maximal observed TAG loading at the end of the studied period, when the collective and highest expression of the AP1 complex components, including all c-Jun and JunB members was also observed. This additionally coincided with the induction of POH1, a regulatory subunit of the 26S proteasome that has a specific de-ubiquitinase activity towards c-Jun leading to its accumulation and subsequent increase in AP1-mediated gene expression . A suite of highly expressed immune genes at this time supported an inflammatory type reaction in adipocytes. Hence, an unrelieved ER-stress could likely be the major cause of chronic inflammatory responses in lipid overloaded white adipocytes in fish.