Introduction

Maintaining energy homeostasis is a multifaceted process involving many regulatory factors from many tissues. One class of regulatory factors crucial to energy metabolism is secreted factors or hormones. For example, high blood glucose levels promote insulin release from pancreatic beta cells to induce glucose uptake in peripheral tissues, thus lowering blood glucose levels [1,2,3]. Conversely, low blood glucose levels promote glucagon release by pancreatic alpha cells to induce glycogenolysis in the liver, thus raising blood glucose levels [4,5,6]. Postprandially, glucagon-like peptide 1 (GLP-1) is secreted by intestinal L cells to promote insulin secretion, slow down gastric emptying and suppress food intake [7, 8]. When fat mass increases, the adipokine leptin is released to inhibit food intake and increase energy expenditure [9]. Adipocyte secretion of adiponectin reduces blood glucose levels by suppressing hepatic gluconeogenesis and promoting fatty acid oxidation in skeletal muscle [10,11,12,13]. This handful of secreted factors only encompasses a small percentage of the secretome. Computational approaches predict that approximately 9% of human protein-coding genes encode at least one secreted product. This 9% represents 1891 genes and many more putative secreted factors, most of which have unknown functions. We have only scratched the surface of mechanistically understanding the secretome and its relationship with energy metabolism [14, 15]. Recently, we have seen a rapid increase in identification of secreted factors that regulate glucose and/or lipid metabolism, mainly thanks to advances in open-source expression data and proteomic techniques. Large-scale proteomic studies can reveal widespread changes to the secretome in individuals with type 2 diabetes or obesity compared with healthy control individuals [16, 17]. Often, secretome changes are only modestly correlated between studies, depending on the platform used and method of analysis [18]. Additionally, while proteomics provide significant insights into changes in protein levels that correlate with metabolic disease, the information is not sufficient for determining whether the protein is bioactive or for understanding the physiological role of the protein in regulating glucose and/or lipid metabolism. Therefore, a combination of computational approaches, proteomic studies and classical molecular biology studies is necessary for identifying novel secreted factors and gaining mechanistic understanding of how these secreted factors regulate energy metabolism and contribute to the development of metabolic disease.

In this review, we discuss the current understanding of how some of these newly described secreted factors affect energy metabolism in mice and cell lines, as well as correlations between these secreted factors and metabolic disease phenotypes in humans. Secreted factors can largely be divided into two groups: factors that drive the development of phenotypes associated with metabolic disease (e.g. glucose intolerance, insulin resistance and fat mass accumulation); and factors that protect against the development of these phenotypes. Here, we selected seven secreted factors (three driving factors and four protective factors) that are either newly identified or have been attributed with a novel function in the last decade. We discuss the tissues of origin, their roles in fundamental aspects of tissue metabolic homeostasis and their proposed mechanisms of action. We also elaborate on the limitations of our current understanding and the technological challenges in therapeutically harnessing these factors.

Secreted factors that promote the development of metabolic disease

Secreted isoform of endoplasmic reticulum membrane complex subunit 10

Endoplasmic reticulum membrane complex subunit 10 (EMC10) is a 27 kDa protein that is widely expressed and is an integral part of the endoplasmic reticulum (ER) membrane [14, 19]. It has been studied as a regulator of neurodevelopment, sperm motility and angiogenesis [20,21,22,23,24,25,26]. Differential splicing of EMC10 produces two isoforms: a membrane-bound EMC10 (mEMC10); and a secreted isoform [27]. In 2022, novel functions of the secreted isoform of EMC10 (scEMC10) in regulating glucose and lipid metabolism were identified (Fig. 1) [28]. scEMC10 was first implicated in regulating glucose metabolism when a 25 mmol/l dose of glucose to cultured pancreatic beta cells and intact islets led to a threefold increase in mRNA transcript levels for scEMC10 (previously known as INM02) [29]. A decade later, Wang et al used a neutralising antibody to demonstrate a causative role for scEMC10 in the dysregulation of glucose and lipid metabolism [28]. Specifically, they reported that neutralising scEMC10 in male mice on a high-fat diet (HFD) improved lipid metabolism as indicated by decreased fat mass accumulation and circulating NEFA levels, reduced hepatic steatosis and triglyceride levels, decreased circulating leptin levels and increased circulating adiponectin levels. Neutralising the function of scEMC10 in HFD-fed male mice also improved glucose metabolism as demonstrated by improved glucose and insulin tolerance and reduced fasting blood glucose levels. These improvements likely stem from increased thermogenesis and energy expenditure. In response to neutralisation of scEMC10 function, there was a significant increase in protein levels of uncoupled protein 1 (UCP1) and peroxisome proliferator-activated receptor γ coactivator-1α (PGC-1α) in brown adipose tissue (BAT), accompanied by increased energy expenditure [28]. Conversely, adeno-associated virus (AAV)-mediated overexpression of human scEMC10 (hscEMC10) in male mice worsened metabolic functions in both chow- and HFD-fed mice. In terms of glucose metabolism, overexpression of hscEMC10 impaired glucose tolerance and insulin sensitivity, and increased circulating insulin levels [28]. With regards to lipid metabolism, mice overexpressing hscEMC10 displayed increased fat mass accumulation, increased liver mass, increased circulating leptin levels and decreased circulating adiponectin levels [28].

Fig. 1
figure 1

Secreted factors that promote the development of metabolic disease. scEMC10, vimentin and CILP2 are three secreted factors with novel functions that drive the development of phenotypes associated with metabolic disease. Such functions include reducing glucose tolerance and insulin sensitivity, as well as promoting increased fat storage. CREB, cAMP-response element-binding protein. This figure is available as part of a downloadable slideset

Full size image

While the receptor and tissue of action for scEMC10’s role in regulating energy metabolism is still unknown, scEMC10 is proposed to be transported intracellularly. Recombinant scEMC10 labelled with FITC is found intracellularly in HeLa cells. Co-immunoprecipitation also shows that scEMC10 binds to protein kinase A (PKA) catalytic α (PKA Cα), an intracellular catalytic subunit of PKA [28]. Additionally, scEMC10 signals in adipocytes by inducing phosphorylation of cAMP-response element-binding protein (CREB) at serine 133 [28]. These data together demonstrate the therapeutic potential of inhibiting scEMC10 to improve metabolic function. However, we still lack a full understanding of scEMC10’s endogenous functions. While Wang et al generated a global EMC10-knockout mouse, the results from these in vivo loss-of-function studies cannot differentiate between the effects of scEMC10 and mEMC10 or the whole parent protein itself [28]. Due to technical challenges, we are limited by the inability to create loss-of-function models for a specific peptide that does not affect the parent protein. Nevertheless, scEMC10 is a novel peptide hormone regulating glucose and lipid metabolism (Fig. 1). Additionally, two human studies revealed correlations between serum scEMC10 levels and phenotypes of metabolic disease: scEMC10 serum levels were positively correlated with increasing BMI scores and negatively correlated with decreasing resting metabolic rate [28, 30], two indicators of dysregulated energy metabolism. More studies are needed to determine whether scEMC10 serum levels are linked to dysregulated glucose and lipid metabolism and whether neutralising scEMC10 function in humans can ameliorate dysregulated energy metabolism.

Vimentin

Vimentin, a 57 kDa protein, is one of the most well-studied intermediate filament proteins [31,32,33]. In 2023, it was discovered that vimentin is secreted by adipocytes in response to oxidised LDLs (oxLDLs) to regulate both lipid and glucose metabolism in vitro (Fig. 1) [34]. This oxLDL-dependent secretion of vimentin relies on its PKA-dependent phosphorylation at serine 72 [34]. Vimentin does not contain a signal peptide, thus is not secreted through the conventional secretion pathway. Rather, vimentin may be secreted through the vesicular type III unconventional protein secretion pathway. In tumour endothelial cells, inhibitors of vesicular secretion pathways strongly blocks vimentin secretion into the cell media [35]. Once secreted, vimentin acts in adipocytes to regulate glucose metabolism (Fig. 1). Adipocytes treated with vimentin for 24 h increase their glucose uptake by promoting the translocation of GLUT1 and GLUT4 proteins to the plasma membrane, without an increase in phosphorylated Akt protein levels [34]. Interestingly, administration of vimentin and insulin together have an additive effect on glucose uptake [34], suggesting that vimentin can induce glucose uptake in a parallel pathway to insulin. Surprisingly, this is accompanied by an increase in GLUT1 mRNA levels but a decrease in GLUT4 mRNA levels. The vimentin-dependent regulation of GLUT1 mRNA is mediated by phosphorylation and activation of the IGF-1 receptor (IGF-1R) and ERK [34]. While direct binding between vimentin and IGF-1Rs has not been tested in adipocytes, a previous study demonstrated vimentin–IGF-1R binding in neurons [36]. Thus, it is possible that vimentin mediates its glucose regulatory effects in adipocytes via IGF-1R signalling.

Regarding lipid metabolism, administration of recombinant vimentin to 3T3-L1 adipocytes was found to increase NEFA uptake via increased expression of the NEFA transporter CD36 [34]. Vimentin also increased the proportion of larger adipocytes and larger lipid droplets, consequently increasing lipid stores as measured by Oil-Red-O staining and triglyceride levels [34]. Vimentin treatment of cultured adipocytes simultaneously increased protein levels of sterol regulatory element-binding protein 1 (SREBP1), a key regulator of lipogenesis [34], and reduced protein levels of several lipolytic factors (e.g. peroxisome proliferator-activated receptor γ [PPARγ] and adipose triglyceride lipase [ATGL]) [34]. Additionally, vimentin treatment induced markers of ER stress and impaired autophagy [34]. Together, these in vitro data present a model whereby vimentin promotes energy storage in adipocytes by increasing glucose and NEFA uptake and inducing lipogenesis and ER stress, while inhibiting lipolysis and autophagy. While more in vivo studies are needed to confirm vimentin’s role as a secreted factor in the regulation of energy metabolism in humans, one study found that individuals with type 2 diabetes and increased blood glucose levels also have increased serum vimentin levels [37]. Thus, neutralising vimentin’s effects in vivo may be a viable therapeutic option for improving glucose regulation.

Cartilage intermediate layer protein 2

Cartilage intermediate layer protein 2 (CILP2), a classically secreted 125 kDa glycoprotein, was first identified in 2003 and is largely found in cartilaginous tissues [38]. CILP2 mRNA transcript levels are widely distributed in mice, with the highest levels being found in skeletal and cardiac muscle, adipose tissue and liver [39]. While there is little work on CILP2’s functions and mechanisms related to energy homeostasis, a study in 2019 investigated the causative role of CILP2 in developing phenotypes of metabolic disease (Fig. 1) [40]. Specifically, AAV-mediated global CILP2 overexpression exacerbated HFD-induced increased blood glucose levels, increased serum insulin concentration, impaired glucose tolerance, reduced glucose uptake and reduced insulin sensitivity in male mice [40]. Additionally, CILP2 overexpression further increased serum concentrations of cholesterol, triglycerides and NEFA in HFD-fed mice [40]. Interestingly, AAV-mediated overexpression of CILP2 increased CILP2 transcript levels to the greatest extent in the liver, suggesting a hepatic role for CILP2 in regulating energy metabolism [40]. In vitro studies on cultured hepatocytes revealed that overexpression of CILP2 promotes the expression of PEPCK, a key enzyme in hepatic gluconeogenesis [40]. Simultaneously, overexpression of CILP2 blocked insulin-induced activation of the Akt signalling pathway [40]. Together, this suggests a model whereby CILP2 drives the development of metabolic disease by inhibiting insulin signalling and promoting gluconeogenesis (Fig. 1). However, the field still lacks loss-of-function and mechanistic studies to elucidate the molecular mechanisms underlying CILP2’s action. In particular, the CILP2 receptor remains unknown. Meanwhile, recent studies in humans revealed that high levels of CILP2 in serum are associated with decreased insulin sensitivity, increased blood glucose, serum insulin and serum triglyceride levels, and increased BMI [41, 42]. Notably, serum CILP2 levels are significantly higher in men than in women [41]. As men have decreased insulin sensitivity compared with women [43], it is possible that CILP2 levels are regulated by sex hormones such as testosterone. Together, these lines of evidence demonstrate a clear association between CILP2 and various metabolic phenotypes, although additional studies are needed to determine whether the regulation of CILP2 is a causative driver of metabolic disease.

Secreted factors that protect against the development of metabolic disease

Isthmin-1

Isthmin-1 (ISM1) is a widely expressed, 60 kDa secreted protein that was originally identified in 2002 and studied for its role in neural development [43, 44]. In 2021, it was discovered that ISM1 has dual functions in promoting glucose uptake in adipose tissue and preventing lipid synthesis in the liver (Fig. 2) [45]. AAV-mediated overexpression of ISM1 in male mice was found to significantly improve glucose tolerance and insulin sensitivity and to inhibit triglyceride accumulation in the liver after 10 weeks of HFD [45]. Importantly, administration of recombinant ISM1 significantly improved glucose and insulin tolerance and reduced hepatic lipid accumulation in mice fed a metabolic dysfunction-associated steatotic liver disease (MASLD)-inducing diet. ISM1 administration also blocked insulin-induced lipogenesis, as measured by acetate incorporation in the mouse hepatic cell line AML12 and expression levels of lipogenesis genes in primary mouse hepatocytes [45].

Fig. 2
figure 2

Secreted factors that protect against the development of metabolic disease. ISM1, LCN2, NRG1 and NRG4 are secreted factors with novel functions that protect against the development of phenotypes associated with metabolic disease. These four secreted factors improve glucose tolerance and insulin sensitivity while reducing fat storage via increased fatty acid oxidation and decreased lipogenesis. LCN2 further supports glucose metabolism by promoting beta cell proliferation and thus increased insulin secretion. MC4R, melatonin-4 receptor; STAT5, signal transducer and activator of transcription 5. This figure is available as part of a downloadable slideset 

Full size image

Concurrently, ISM1 promotes the synthesis of hepatic proteins, as measured by increased leucine incorporation and protein levels of phosphorylated S6, a kinase involved in protein synthesis [45]. In adipocytes [45] and myocytes [46], ISM1’s regulatory effects on metabolism are mediated through the phosphoinositide 3-kinase (PI3K)–Akt signalling pathway, independent of insulin. Specifically, in 3T3-F442A cells treated with ISM1, a phosphokinase array revealed that ISM1 phosphorylated Akt most strongly at serine 473 [45]. This was further confirmed via an ELISA and western blot assays to measure phosphorylated Akt levels after ISM1 administration to 3T3-F442A cells [45]. However, while we know that ISM1 activates the PI3K–Akt signalling pathway, the receptor for ISM1 is unknown. Future studies will be needed to determine whether ISM1 is a viable therapeutic target for treating metabolic disease. In adult humans, serum ISM1 levels are decreased in individuals with type 2 diabetes compared with healthy control individuals [47]. ISM1 serum levels were also found to be positively correlated with increasing obesity in a cohort of pubescent boys [48], suggesting an additional function of Ism1 in regulating adiposity. Overall, ISM1 is an interesting secreted factor that may play many roles in the control of energy metabolism.

Lipocalin-2

Lipocalin-2 (LCN2), a 25 kDa osteokine, is widely distributed and is a classically secreted glycoprotein [49]. Originally, LCN2 was studied as an adipokine that decreased insulin sensitivity [50]. However, it was discovered in 2017 that LCN2 expression levels are tenfold higher in bone than in adipose tissue [51], leading to the discovery that osteoblast-derived LCN2 is a potent appetite suppressor and metabolic regulator (Fig. 2). Loss of LCN2 in osteoblasts was found to significantly increase fat pad mass and food intake in male mice [51]. This effect was accompanied by reduced glucose and insulin tolerance as well as impaired glucose-stimulated insulin secretion (GSIS) [51]. Importantly, LCN2 shows therapeutic promise as it acts as an anorexigenic hormone in both lean and obese mice. Daily i.p. injections of 150 ng/g of recombinant LCN2 reduced food intake by 18%, reduced fat mass by 32%, reduced overall body weight by 9.4% and improved glucose and insulin tolerance in wild-type mice [51]. Daily LCN2 injections into leptin-receptor-deficient mice ameliorated their hyperphagia and fat mass gain, improved GSIS and completely reversed glucose intolerance [51]. The appetite-suppressing effects of LCN2 are mediated through melanocortin 4 receptor (MC4R) signalling in the paraventricular hypothalamus [51]. LCN2’s appetite-suppressing effects are also seen in primates where a 0.0375 mg/kg dose of LCN2 reduced food intake by 21% [52]. There may be more to LCN2’s therapeutic potential than just its actions as an anorexigenic factor. Loss of LCN2 in osteoblasts also decreased beta cell mass and proliferation [51]. One major question left to answer regarding LCN2’s anorexigenic capabilities is its mechanism of action. LCN2 has two known receptors, megalin and solute carrier family 22 member 17 (SLC22A17) [49]. However, megalin is not expressed in cultured murine hypothalamic neurons where LCN2 is able to induce cAMP activity and silencing SLC22A17 in these same neurons did not affect LCN2 signalling [51]. There is much left to elucidate regarding osteoblast-derived LCN2’s mechanisms of action as well as LCN2’s role in metabolic disease development in humans. In humans, serum levels of LCN2 are positively correlated with increasing BMI [53]. Notably, serum LCN2 levels are influenced by biological sex, as men have higher concentrations than women [54]. Given the myriad of sex differences in glucose metabolism and beta cell function [55, 56], serum LCN2 levels may be regulated by sex hormones such as oestrogen and testosterone. It will be necessary for future studies investigating LCN2’s function to consider that men and women may regulate and respond to LCN2 differently.

Neuregulins

The neuregulins (NRGs) are a family of four EGF-like secreted proteins (NRG1–4) best known for their roles in development, regulation of the nervous system and inter-organ communication [57].

NRG1

Initial in vitro studies revealed that NRG1 can stimulate glucose uptake by promoting the translocation of GLUT4 to the cell surface via activation of the PI3K–Akt signalling pathway [58,59,60,61]. These findings were not demonstrated in vivo until 2015. In db/db mice, acute or chronic injections of recombinant NRG1 significantly improved glucose tolerance and reduced both blood glucose and serum insulin levels (Fig. 2) [62]. These effects are partially mediated by the liver; an acute injection of NRG1 induced phosphorylation of the known receptor erb-b2 receptor tyrosine kinase 3 (ErbB3) only in the liver [62]. In addition to improved glucose clearance, NRG1 injection into db/db mice inhibited hepatic gluconeogenesis via phosphorylation of Akt and forkhead box protein O1 [62]. Further, in vitro assays in skeletal myocytes previously revealed that NRG1 promotes glucose and fatty acid oxidation via increased mitochondrial mass and activity [63, 64]. In 2017, these findings were replicated in vivo: 8 weeks of treatment with recombinant NRG1 significantly reduced body weight, improved mitochondrial function and increased mitochondrial complex 2 subunit content in male mice [65]. Importantly, serum levels of NRG1 are significantly decreased in individuals with type 2 diabetes compared with healthy control individuals [66], and are significantly higher in women than in men [67]. Given that women have a lower prevalence of type 2 diabetes and increased insulin sensitivity compared with men [55], the sex difference in serum NRG1 levels may be regulated by sex hormones such as oestrogen. Thus, NRG1 may be leveraged as a therapeutic agent to improve glucose regulation and mitochondrial function in humans. However, NRG1 has a very short t½ in circulation [68]. Thus, a fusion protein of human NRG1 and the Fc domain of human IgG1 (NRG1-Fc), with a markedly longer plasma t½ of 40 h, has been generated [69]. A single NRG1-Fc injection significantly improved glucose tolerance and reduced both blood glucose and plasma insulin levels in chow- and HFD-fed mice. Daily injections for 8 days into HFD-fed mice significantly improved glucose regulation and reduced body weight gain and food intake [69]. Together, these studies demonstrate the potential of NRG1 as a therapeutic agent for treating obesity and type 2 diabetes.

NRG4

In addition to NRG1, studies in the last decade have also established novel functions of NRG4 in the regulation of glucose and lipid metabolism (Fig. 2). NRG4 mRNA transcripts were originally detected only in the pancreas and the muscle [70, 71]. However, secretome analysis revealed that NRG4 is more highly expressed in BAT and is moderately expressed in subcutaneous white adipose tissue [72, 73]. In HFD-fed male mice, overexpression of Nrg4 was found to globally attenuate fat mass gain, increase plasma leptin levels and glucose intolerance, and decrease insulin sensitivity [74, 75]. Conversely, whole-body loss of Nrg4 led to impaired glucose metabolism, indicated by worsened glucose and insulin tolerance and increased blood glucose and serum insulin levels [72]. NRG4’s effects on glucose metabolism are mediated by pancreatic beta cells and adipocytes. In cultured rat islets, NRG4 induced the greatest insulin release when compared with the other NRGs [71]. In vitro assays in cultured adipocytes with Nrg4 knockdown revealed decreased protein levels of both insulin receptor and the major GLUT, GLUT4 [76]. Decreased GLUT4 protein levels due to loss of Nrg4 is mediated by reduced mTOR phosphorylation and thus increased autophagy of GLUT4 storage vesicles [76]. Importantly, NRG4 treatment is sufficient to increase both insulin receptor expression and GLUT4 protein levels [76]. Thus, NRG4 may maintain glucose homeostasis by promoting glucose uptake in adipocytes.

In addition to NRG4’s effects on glucose metabolism, NRG4 acts in the liver to suppress hepatic lipogenesis [72]. Whole-body loss of Nrg4 in mice promoted lipid accumulation in the liver, increased circulating triglyceride levels and increased hepatic mRNA transcript levels of de novo lipogenesis genes (e.g. Gck and Fasn), inflammatory markers (e.g. Tnfa, Il1b), and fibrosis markers (e.g. Col1a1, Acta2) [72, 77, 78]. Exogenous NRG4 can block the upregulation of lipogenic genes and incorporation of acetate into lipids induced by pharmaceutically activating SREBP1c [72]. In vitro experiments using primary mouse hepatocytes revealed that NRG4 acts through its receptor ErbB4 to phosphorylate Akt and signal transducer and activator of transcription 5 (STAT5) to modulate its protective functions [72, 77, 78]. While it is clear that adipose-derived NRG4 has important roles in regulating glucose and lipid metabolism, the relationship between serum NRG4 levels and metabolic disease in humans is less clear. Some studies revealed that individuals with gestational diabetes, obesity or MASLD have reduced serum NRG4 levels compared with control individuals [79,80,81,82]. However, one study found no link between serum NRG4 levels and MASLD [83] while another study found a positive correlation between serum NRG4 levels and insulin resistance [84]. Future studies will be needed to determine whether NRG4 can be used as a biomarker of metabolic disease development and whether NRG4 can be leveraged to treat metabolic diseases.

Conclusions and future directions

Significant progress has been made in the last decade in the discovery of novel hormones and in revealing additional functions of known factors that contribute to the regulation of glucose and lipid metabolism. As novel secreted factors and functions are revealed, there is a need to understand the mechanisms underlying their regulatory actions. Specifically, identifying receptors for regulatory hormones is vital. Of the hormones discussed in this review, only vimentin, NRG1 and NRG4 have known receptors that modulate energy homeostasis. Identifying receptors is a major challenge due to technical limitations. To put this into perspective, while insulin was discovered in 1921, its receptor was not discovered until the early 1970s. Similarly, EGF was discovered in 1962 [85] and its receptor was discovered 18 years later [86, 87]. With the advent of computational tools for predicting protein structure, in silico screening for ligand–receptor interactions will speed up this process. However, we are still limited to a candidate approach. The following are examples of recent technologies that have made higher-throughput screening possible: (1) Retrogenix Cell Microarray, used to identify the receptor for growth differentiating factor 15 (GDF15) [88, 89]; (2) a ligand-based receptor capturing technique in live cells and tissues called TRICEPS, used to identify novel receptors for C1q TNF-related protein 3 (CTRP3) [90] and stanniocalcin-1 (STC1) [91,92,93]; and (3) a pooled CRISPR/Cas9-based transcriptional activation screen (CRISPRa), used to deorphanise several receptor protein tyrosine phosphatase ligands and a novel ligand that binds to killer immunoglobulin-like receptors [94, 95]. These newer techniques are now being used alongside classical cDNA cloning methods to identify ligand–receptor pairs but there is still much room for improvement and it will be an important hurdle to cross for the metabolism field in order to design more effective therapeutics.

Altogether, this review highlights some of the most recent discoveries of secreted factors involved in maintaining energy homeostasis. While this review focused on novel secreted peptides and proteins, other classes of secreted factors also contribute to maintaining energy homeostasis such as secreted lipids, metabolites and small open reading frames (sORFs) [15, 96, 97]. These other classes of secreted factors, alongside almost 2000 genes encoding secreted proteins and peptides, reveal the vastness of the mammalian secretome. The secretome is a powerful tool for understanding energy metabolism and studies in this field will be foundational for new and potentially more effective treatments for many metabolic diseases.