Obesity is a complex multifactorial disease that results from the interplay between environmental and genetic factors (1). Overweight and obesity are characterized by adipose tissue (AT) expansion, attributed either to hyperplasia and/or hypertrophy, which eventually becomes dysfunctional (2). AT expansion is often associated with a chronic low-grade inflammatory state that will increase oxidative stress (3) and may have a significant contribution for obesity-related comorbidities development, including insulin resistance (IR) and dysglycemia (4).

Dysglycemia states are defined as conditions of abnormal glucose homeostasis and include impaired fasting glucose, impaired glucose tolerance or both. Fasting plasma glucose above 100 mg/dl and 126 mg/dl (5.6 mmol/L and 7.0 mml/L) or hemoglobin A1c (HbA1c) above 5.7% and 6.5% are used as biochemical thresholds to define the clinical conditions known as pre-diabetes and diabetes, respectively (5). Type 2 diabetes (T2D) is one of the most prevalent comorbidities among individuals with obesity, being responsible for a great number of adverse health outcomes (6). Dysglycemic conditions are characterized by a continuous spectrum of glycemic imbalance (5). From a pathological perspective, these processes are initiated by increased resistance to insulin action in peripheral organs, such as skeletal muscle, liver and AT (7). Consequent to a decreased insulin-mediated glucose uptake, the pancreas is stimulated to increase insulin secretion in an attempt to overcome resistance (7, 8). When the pancreatic capacity to sustain insulin hypersecretion is lost, circulating glucose levels rise and pre-diabetes or overt T2D arise (7, 8).

Individuals with severe obesity can be defined as metabolically healthy obese (MHO), i.e. if no evidence of abnormal metabolic parameters is found in routine clinical and biochemical assessments (9). However, a considerable number of patients with obesity are also affected by obesity-related metabolic disorders, such as pre-diabetes, T2D, hypertension and dyslipidemia, and therefore are clinically classified as having metabolic unhealthy obesity (MUHO) (10). For the purpose of this systematic review, only studies that sought to evaluate the differences between individuals with overweight or obesity in the absence or presence of dysglycemia, were considered and defined as MHO or MUHO, respectively.

The relationship between increasing body mass index (BMI) above 25 kg/m² or 30 kg/m², which is the anthropometric measurement most frequently used to define overweight and obesity, except for Asian ethnic background populations, and T2D is far from being linear. Obesity is a well-recognized risk factor for T2D. The risk of developing T2D increases with increasing BMI when compared to normal weight individuals (10). However, despite approximately 90% of the patients with T2D presenting concomitant overweight or obesity (11), less than one third of the patients with severe obesity harbor T2D as comorbid condition (12). Therefore, the relationship between obesity and T2D is a complex one. Moreover, the risk of cardiovascular disease and cardiovascular disease mortality, which is the number one cause of death among individuals with obesity, increases not only with increasing BMI, but also with the number of metabolic abnormalities, being the highest among patients with T2D (13). Thus, the presence of T2D renders a more advanced disease stage to obesity, forecasting a poorer prognosis and decreased life expectancy (13). Identifying the triggering factors responsible for T2D development in some individuals with obesity is clinically very relevant, because it would allow to implement targeted intervention strategies with the aim of preventing cardiometabolic complications in those patients at higher risk

T2D is a multifactorial and polygenic disease with hundreds of genetic variants identified as risk factors in genome-wide association studies (GWAS) (14). Despite the fact that obesity and T2D can unquestionably be associated as mentioned above, only a few loci were identified as being related with both conditions, such as FTO, MC4R, ADAMTS9, GRB14/COBLL1 and QPCTL/GIPR (11), suggesting that these genes may have a particularly important role in the pathogeny of both conditions. However, it is important to notice that single nucleotide polymorphisms (SNPs) primarily associated with obesity tend to have a positive correlation between the effect size on BMI and the effect of the same SNP on T2D, yet SNPs primarily associated with T2D have no impact on BMI per se (15). Importantly, both GWAS and gene candidate approaches have led to the identification of genes implicated in the pathogenesis of obesity and/or T2D (14, 16), which are involved in critical pathways for glucose regulatory processes, such as insulin signaling (INSR, IRS1, IRS2) (16–20), beta-cell differentiation and insulin secretion (PDX1, HNF4A, TCF7L2, SLC2A4, GLP1R, KCNQ1) (16, 17, 21–24), adipocyte differentiation (PPARG, PPARGC1A, LEP, ADIPOQ) (17, 19, 20, 25, 26), mitochondrial biogenesis and function (PGC1) (27), lipid and glucose homeostasis (SREBF1) (28) and cytokine signaling and inflammation (ADIPOQ) (29).

Previous studies also sought to identify genetic determinants that could explain the diversity in metabolic phenotypes among patients with overweight or obesity. Genetic risk scores derived from GWAS (30) resulted in modest improvements in the accuracy to predict T2D compared to traditional risk scores that rely on patient characteristics, namely age, family history of T2D and BMI, such as the FINDRISK Diabetes Score (31). Few studies focused on MHO and MUHO, mainly due to difficulties in patient characterization. It appears that there is no clear association between susceptibility to develop obesity-associated comorbidities and adiposity related genes, with some variants of these genes having potentially cardiometabolic protective effects (IRS1, COBLL1/GRB14, PLA2G6 and TOMM40) (32). The lack of a clear association may be explained by the observation that within the same gene some variants could have opposing effects on these traits (e.g. COBLL1/GRB14) (32). However, most recently, the first genome-wide cross-phenotype meta-analysis of adiposity–cardiometabolic trait pairs led to the discovery of 62 loci, clustered in three groups, of which the same allele was significantly associated with both higher adiposity and lower cardiometabolic risk (33).

Additional factors known to modulate gene expression, such as epigenetic marks, were hypothesized to play a relevant role in the pathogenesis of obesity-related dysglycemia (34). The term epigenetics refers to reversible alterations in gene activity, which occur without altering the DNA sequence and that are heritable (during cell divisions in somatic cells, but also through the germline and inherited trans-generationally) (35). Epigenetic information is laid upon the genome as epigenetic marks that include DNA methylation, histone modifications (such as methylation, acetylation, etc.), non-coding RNAs and chromatin remodelers, all of which affect the chromatin architecture and provide long-term stability of gene expression patterns (35). Epigenetic states are tissue-specific and are implicated in a range of cellular processes, such as cell differentiation, genomic imprinting and chromosome X-inactivation in females (34). Importantly, epigenetic modifications are able to influence the cellular phenotype and responsiveness to external stimuli (36).

Despite the fact that an association of epigenetics with obesity and dysglycemic disorders has been known since the nineties, the interest in epigenetic alterations associated with obesity and obesity-related dysglycemic disorders has increased exponentially in the last decade (37). Several studies have focused on identifying epigenetic modifications occurring in patients with obesity and dysglycemia states including T2D, when compared to normal weight (NW) and/or individuals with normoglycemia, thus providing considerable advance in our knowledge (38, 39). In fact, numerous DNA differentially methylated regions (DMRs) associated with T2D were identified in targeted or epigenome-wide association studies (EWAS), some of which occurred in genes reported to be of particular interest given the direct involvement in mechanisms known to regulate glucose homeostasis (40). Consistently, increased DNA methylation in PPARGC1A was found to down-regulate gene expression in islet cells (41), skeletal muscle (42) and AT of individuals at high risk for T2D (43), along with decreased mitochondrial content (44). Moreover, PPARGC1A methylation was shown to be highly responsive to exercise (45) and bariatric surgery (42). Additionally, altered DNA methylation in ABCG1 and SREBF1 genes, involved in lipid homeostasis, were demonstrated to be associated with down-regulation of mRNA levels in the liver and skeletal muscle from individuals with T2D (46). These are only a few among the large number of genes in which epigenetic changes were reported to be correlated with glucose and insulin concentrations, BMI and Homeostatic Model Assessment of Insulin Resistance (HOMA-IR) and were identified as being associated with the risk of future T2D (40, 47).

Nevertheless, the role of epigenetic modifications in modulating the risk for obesity-related dysglycemic disorders is far from being fully understood. The type, strength, and rate of epigenetic changes and how cell-specific these are, or which individuals are more predisposed to develop it, remains to be elucidated. This is partially because human data so far available, with very few exceptions, is mostly derived from small patient cohorts, or limited to whole blood (WB) analysis, although many epigenetic modifications are known to be tissue-specific (48). Therefore, since AT expansion and dysfunction are known to play a crucial role in the development of several obesity-associated metabolic comorbidities, gaining knowledge on the AT-specific epigenetic modifications is paramount for understanding the mechanisms of disease, which could lead to the identification of potential targets for treatment intervention. Importantly, the visceral adipose tissue (VAT) is recognized for having a higher metabolic activity compared to the subcutaneous adipose tissue (SAT), as well as by having a greater impact on systemic metabolism via rapid release of free fatty acids (49, 50). Additionally, the interplay of altered metabolic and physiological states within distinct AT territories, such as VAT and SAT, were demonstrated to be detrimental for whole-body glucose homeostasis (49). This evidence reinforces the pressing need to evaluate and assess the relative contributions of distinct epigenetic modifications observed in each AT depot in the mechanisms associated with IR and dysglycemic states. Therefore, the aim of this systematic review was to summarize the available data on ‘epigenetic fingerprints’ in human AT that allow differentiation between MHO and MUHO associated with dysglycemia states, and to uncover the potential contribution of epigenetic-regulated cellular pathways. In addition, we discuss the major drawbacks of the current human models for the study of epigenetics in AT and the limitations of epigenetic analyses in this tissue and point out to future directions of research.

In this review, we systematically appraised and discussed the published epigenetic changes identified within the AT, namely DNA methylation, histone modifications and lncRNAs, associated with obesity-related IR/dysglycemic traits and T2D. Our integrative approach, based on bioinformatic analyses, highlighted a number of additional avenues for future research, such as new signaling pathways, novel TFs and blood-based DNA methylation biomarkers that can discriminate between MHO and MUHO individuals.

Hypothesis-driven DNA methylation studies on selected genes provided some good, albeit limited, links between dysglycemia and levels of DNA methylation. Across the reviewed studies, three genes (FGF21, INSR and SLC2A4) associated hypermethylation around the promoter region in subjects with MUHO versus subjects with MHO. DNA methylation at the promoter regions of these genes showed a negative correlation with expression of the corresponding genes (56, 58). Importantly, these targeted studies found significant correlations between glycemic parameters (such as fasting glucose or HbA1c), insulin resistance measurements (such as HOMA-IR) or insulin sensitivity measurements (such as the Matsuda index or QUICKI) with levels of DNA methylation for ADIPOQ, TNFA, FKBP5, INSR and SLC2A4 (56, 57, 59). These comparisons draw some parallels between glucose levels and DNA gene methylation changes in several tissue types from in vitro studies (142, 143). Along those lines, all five genes are related to the cellular responses to insulin: INSR encodes the insulin receptor (56), ADIPOQ encodes an insulin sensitizer (144), TNFA encodes a cytokine linked to insulin resistance and was found to be elevated in the serum of subjects with T2D (145), SLC2A4 encodes GLUT4 that is the major glucose transporter in adipocytes (146) and FGF21 encodes a hormone-like protein with insulin sensitizing actions in adipocytes (58).

Most EWAS in the context of obesity performed so far focused on assessing DNA methylation levels in WB. These DNA methylation studies have uncovered several DMRs located in or near genes known to be implicated in body weight regulation or glucose homeostasis, which presented significant correlations with BMI (147). Interestingly, data suggests that these DNA methylation changes associated with BMI are a consequence rather than the cause of obesity (39). In addition, a methylation risk score calculated across loci associated with BMI, was found to be strongly predictive for future T2D risk, while methylation patterns observed in peripheral blood cells and other tissues suggest that altered ‘systemic’ methylation is a signature of T2D (148). Furthermore, a large number of these epigenetic signatures tend to occur in genomic regions where genes that are involved in processes related to glucose homeostasis, such as insulin signalling or glucose/lipid metabolism, are located (149). Our analyses also led to the identification of eight differentially methylated CpGs, associated to eight genes, which have tightly correlated changes between MUHO and MHO individuals in WB and AT. Three of these genes – TBC1D4, TXNIP and PAMR1 – were previously implicated in dysglycemia (150–152). However, the remaining five genes (PCDHGA1/PCDHGA4, PSMD5, SLC9A2, SUMO3 and ZNF251) to the best of our knowledge, have not been mechanistically associated with dysglycemia. We speculate that these genes could highlight new molecular pathways relevant for AT dysfunction and progression to dysglycemia. Additionally, since these DMRs are strongly correlated between AT and WB, it is possible to envisage that these could be potentially useful biomarkers when considering to develop a CpGs panel to predict the risk of future T2D in MHO.

We, like others, reasoned that investigating epigenetic signatures in peripheral tissues that are metabolically relevant for the pathogenesis of obesity-related metabolic complications, and specifically in the AT, could provide further insights into the role of epigenetic modifications for the mechanisms of the disease (153). Indeed, the limited number of studies performed so far, and described here, highlight distinctive epigenetic signatures in AT of individuals with MUHO compared to AT of individuals with MHO. In contrast to the limited scope of targeted studies, EWAS studies that compare groups of individuals with MUHO and a control population tend to paint us a fuller picture of the epigenetic landscape. Therefore, our analytic approach was to focus on collating all the data, mostly from the EWAS studies reported to date, with the aim of generating potentially meaningful data, in spite of issues related to the high variability between studies. Indeed, despite the heterogeneous nature of the studies reviewed, particularly in terms of selection criteria of the cohorts and tissue types (total SAT/VAT or isolated adipocytes from SAT/VAT, or pre-adipocytes derived from VAT), we catalogued over two thousand DMRs at the cutoff 5%. Interestingly, the majority of the strongest DMRs (>30%) were located at genes that have not been previously linked with adipogenesis or dysglycemia. Among these, three DMRs were located at genes encoding components of the human leukocyte antigen (HLA) system, classically associated with the risk for type 1 diabetes (154), such as HLA-DBR1 and HLA-DQA1, both hypermethylated in MUHO. Other genes that associate strongly hypermethylated DMRs in MUHO include HLA-DBR6, STARD13, GALNT6, NKRF, INIP and CAPN8. Genes that associate strongly hypomethylated DMRs in MUHO include ARMC3, MOB3A, KCNC3 and FCGBP. These genes are implicated in a wide range of activities, from ion transport, to signal transduction and transcriptional regulation (86, 90, 100, 104). Their study in the context of AT biology may provide new clues about the mechanisms leading to MUHO.

Although each individual DMR is unlikely to have a big influence on gene expression, particularly those that associate smaller DNA methylation differences, it is possible that their cumulative effects could have a significant impact on progression to dysglycemia (34). Since most of the DMRs catalogued in our review were identified using the Infinium HumanMethylation450 BeadChip platform (63, 65), we tested whether they showed a similar distribution to that of the probes included on this array. We observed a significantly smaller proportion of DMRs around the gene promoters in cases with MUHO compared to controls. Whether or not this is a general feature of AT from MUHO subjects representing for example specific demethylation events at promoter regions or defects in targeting methylation to specific promoter regions, this will need to be established in multiple and larger cohorts. It will be important also to establish if the relative ‘loss’ of promoter DMRs has an impact on gene expression. Promoter DNA methylation is traditionally associated with repression of gene expression. Nonetheless, there are exceptions to this rule, and hypermethylation induced transcriptional activation has been document in a range of conditions and cell types (155). Although gene body methylation is highly conserved across eukaryotic species, the understanding of its function is still incomplete, and may play a role in alternative splicing or as a “fine tuning” mechanism of gene expression (156, 157). Because the data on gene expression in the studies included in this review was incomplete, we could not assess directly the relationship between the DMRs and gene expression. Nonetheless, we set out to perform bioinformatics analyses on the set of genes that associated DMRs above the cutoff of 5%. Both DAVID and IPA analyses highlighted alterations in several key intracellular signaling pathways. Among these, we identified a significant enrichment of genes implicated in signal transduction mediated by small GTPases. Several small GTPases, such as those from the Rho family, many of which associate DMRs in our study, have been already linked with glucose uptake in AT and with IR (158). Calcium homeostasis, enriched in genes associating DMRs in our study, has been also implicated in a myriad of cellular and subcellular dysfunctional networks found in the context of obesity and diabetes (159). ERK1/2 signaling (109, 111) and intracellular protein trafficking (101) were also linked previously with insulin signaling. Our search also identified a network of TFs predicted to regulate transcription of a large fraction of genes that associated DMRs and seven of these TFs were associated with DMRs themselves. While some of these TFs are strongly linked with glucose homeostasis, others, such as ELF1, IRF8 and RUNX1, are yet to be studied in this context. It has been reported that the binding site for ELF1 is overrepresented in the promoters of genes up-regulated during adipogenesis of human adipose-derived stromal cells (160). Both IRF8 (161–163) and RUNX1 (163) have been implicated in the pro-inflammatory immune response. It is known that AT inflammation plays an important role in the development of insulin resistance (3). Understanding the mechanisms by which the seven TFs identified in our analysis could target DNA methylation changes to specific regions requires further studies. Emerging evidence suggests that TF occupancy can mediate active turnover of DNA methylation by local recruitment of DNMT and TET enzymes implicated in the addition and erasure of DNA methyl groups, respectively (164, 165). As these TFs are involved in complex regulatory pathways and impact a great number of genes, more attention to their potential role in the pathogenesis of obesity and obesity-associated dysglycemia is warranted.

The search for genetic-based risk factors is particularly affected by the high degree of clinical heterogeneity that exists within each group of healthy or unhealthy individuals, but also because metabolic markers of ‘health’ that define these two groups vary greatly across current human studies and cohorts. In spite of these pitfalls, major insights into the identification of genetic pathways that may help uncoupling adiposity to cardiometabolic comorbidities have come from GWAS studies (32, 33). Huang et al. (33) identified 62 loci of which the adiposity-increasing allele was also associated with a favourable effect on cardiometabolic risk factors. Interestingly, we found that 10 of these genes (ADCY9, ARAP1, CLIP1, CREBBP, GFI1, IRS1, JAZF1, NCOR2, PPARG and PSORS1C1) are differentially methylated between subjects with MUHO and subjects with MHO in our systematic review lists. Another approach that may lead to the identification of gene variants that link increased adiposity with reduced risk of cardiometabolic is a more indirect one (32). Accordingly, GWAS studies aimed to discover genes involved in body fat % identified variants in or near IRS1, COBLL1/GRB14, PLA2G6, TOMM40 loci of which the body fat % increasing allele had protective effects on cardiometabolic outcomes. Interestingly, GWAS studies for obesity or T2D show relatively little overlap, with a screen using the GWAS Catalog (https://www.ebi.ac.uk/gwas/) leading to the identification of 89 gene variants common to both. Of these 89 genes, 8 (SUGP1, MACROD2, RPTOR, FAIM2, LEPR, HS6ST3, TMEM1 and KCNQ1) were found differentially methylated between subjects with MUHO compared to subjects with MHO in our review lists.

Histone modifications are important epigenetic marks, with a regulatory role in numerous cellular events. These modifications can either promote or repress DNA transcription, and act through two main mechanisms, by affecting chromatin structure or protein binding (140, 166). Moreover, histone lysine methylation may activate or repress transcription depending on its position and methylation state (mono, di or trimethylation) (167). The scarcity of histone modifications studies in AT is related to technical difficulties. Indeed, the high lipid content renders AT as an extremely challenging tissue to work with. Clear associations were found between H3K4me3 enrichment in key genes and their expression (E2F1, LPL, SREBF2, SCD, PPARG and IL6) (71), illustrating that histone methylation could have an important role in obesity and T2D, by affecting various pathways associated with metabolic disease. The preliminary work discussed in this review indicate a great potential, but also highlight the need for genome-wide characterization of histone modifications in human AT to better understand their role in MUHO.

Unique lncRNA expression profiles in AT of individuals with obesity were found to correlate with distinct glycemic states (73, 74). These lncRNAs intervene in the glucose regulation pathways and are suggested to be implicated in the pathology of insulin resistance and T2D by modulating pathways involved in adipogenesis, energy metabolism, inflammation or insulin sensitivity (168, 169). GAS5, comprised of 12 exons, belongs to the 5’-terminal oligopyrimidine class and codifies not only the respective lncRNA, but also miRNAs, small nucleolar RNAs and PIWI-interacting RNAs (170, 171). The lncRNA transcript was suggested to interact with the promoter of the INSR gene as a riboactivator, allowing transcription factors to bind more easily, leading to overexpression of this gene (74). GAS5 lncRNA was also previously reported to interact directly with the chromatin in other contexts, such as being a riborepressor of the glucocorticoid receptor in starvation or growth arrest (172). ENSG00000235609.4 and CATG00000111229.1 are novel intragenic lncRNAs and their role in the adipose physiology is largely unknown. Different AT depots and diferent cell populations may have specific patterns of lncRNA expression (173). Hence, it is critical to perform more studies on the role of differentially expressed lncRNAs in the different AT depots and on different cell populations comprised within the AT, in order to understand their functions. It is equally necessary to obtain more data on how these expression patterns vary with time and/or after clinical interventions, such as exercise, diet, drugs and even bariatric surgery, which could prove to be useful therapeutical targets for controling IR and T2D.

There are a few limitations to this systematic review that should be acknowledged. First, the number of studies characterizing epigenetic marks in the context of obesity-associated dysglycemia is very limited, and there are even fewer studies that used AT for such analyses. This limitation is particularly striking for studies on histone modifications and lncRNAs. Second, there is a great heterogeneity in patient characteristics analyzed across the studies, with high variability of anthropometric features, surrogate measurements used to define dysglycemia and different ethnic backgrounds. Third, the methodological approaches used by these studies were diverse and the magnitude of DNA methylation changes is overall small. Fourth and most importantly, gene expression was not assessed in the majority of these studies, hampering the possibility to evaluate the putative biological consequences of the DMRs.

In summary, the question on why some individuals with obesity are metabolically ‘healthy’, while others are metabolically ‘unhealthy’ is a complex one, mainly because there are a number of risk factors that have to be present or absent to define an individual as healthy or unhealthy. Those risk factors include environmental factors (e.g. physical activity, diet and smoking), demographic factors (e.g. sex, ancestry, age) as well as genetic factors. Based on our current analysis we propose that genome-wide epigenetic studies in AT have considerable promise as an additional approach to identify alleles that can be associated with protective and/or detrimental risk of IR and dysglycemic states, thereby contributing to the uncoupling of adiposity from its cardiometabolic comorbidities. Additionally, we propose that careful comparison between DNA methylation signatures in WB and AT in individuals with MUHO may lead to the development of new biomarkers with predictive power for progression of MHO towards metabolic complications. The advantage of using epigenetic marks for these studies is that they are powerful readouts of environmental influences on gene expression, thus potentially acting as links between genetic risk factors and environmental ones. However, the work performed in this systematic review also highlights the need to conduct more robust studies in AT of larger cohorts of individuals with MHO and MUHO. In addition, and to our knowledge, genome-wide transcriptome analyses have not been performed in AT in these two groups of interest, and epigenome screens at the level of regulatory elements, such as enhancers, are also missing. In our view, these are important future directions.

In summary, despite several major limitations, our systematic review led to the identification of a catalogue of epigenetic modifications, particularly DMRs, which highlighted several signaling pathways that are dysfunctional in AT of individuals with MUHO compared to AT of individuals that have MHO. Our approach also led to the identification of a network of TFs, some of which are novel in the context of AT biology. Additionally, we identified a small panel of individual CpGs that associate DNA methylation changes in WB and AT in individuals with MUHO versus MHO. Further studies are required to validate these findings and to assess whether they could lead to a better prediction of individuals with obesity at higher risk of developing T2D. This review also underscores the knowledge gap concerning histone modifications and lncRNA in obesity related-dysglycemia. Therefore, additional studies are needed, particularly those focusing on epigenetic mechanisms other than DNA methylation, and most importantly addressing how these affect gene expressions and dysregulate metabolic pathways leading to T2D.

The original contributions presented in the study are included in the article/ Supplementary Material . Further inquiries can be directed to the corresponding author.

SA, ALS, and TM conducted the literature search, screened the papers, and drafted the manuscript. IS performed the bioinformatic analyses and drafted the manuscript. MC reviewed and edited the manuscript. MPM verified the literature search, conceived the research idea, and reviewed and edited the manuscript. All authors contributed to the article and approved the submitted version.

This work is supported by the Portuguese Foundation for Science and Technology: “EPIADIPO - Exploring the role of epigenomic and adipocyte-associated signals in metabolic dysfunction” (PTDC/MEC-MET/32151/2017). Unit for Multidisciplinary Research in Biomedicine (UMIB) is funded by the Foundation for Science and Technology (FCT) Portugal (UIDB/00215/2020 and UIDP/00215/2020). TM has a grant from FCT, Portugal (SFRH/BD/123437/2016).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.