Experimental human data regarding the nature of RETN-expressing cells as well as the physiological role of secreted resistin are controversial. In contrast to our own studies 
and also those from several other laboratories 
, McTernan et al. 
reported that resistin mRNA and protein could be recovered from both isolated human pre-adipocytes and mature adipose cells. Moreover, it was stated that RETN is differentially expressed between various fat depots 
. In both studies, a higher signal was found in abdominal WAT suggesting that, similar to rodents, resistin may connect central obesity to insulin resistance and diabetes in man. Also, Degawa-Yamauchi et al. 
detected RETN protein in isolated adipocytes and observed a positive association between serum resistin level and adiposity. Finally, in the study by Janke et al. 
, human preadipocyte differentiation was coupled with a declining gene expression. Although this observation was corroborated by barely measurable mRNA levels in primary cultured mature adipocytes, no link with insulin resistance was found.
In contrast to these discrepant results, various investigations in man consistently revealed the presence of resistin in myeloid cells: both primary monocyte-derived and/or differentiated macrophages 
as well as in primary acute leukemia cells and the myelocytic lines U937 and HL60 
. This leukocyte expression of human resistin was recently discovered to depend on the myeloid-specific nuclear transcription factor CEBPE 
. The action of CEBPE was previously known to be restricted to the haematopoietic tissues, especially myeloid granulocyte/macrophage and lymphoid T-cell lineages in the bone marrow 
. In line with, but independent of the CEBPE research, the highest level of resistin was found in the bone marrow 
. Moreover, human resistin possesses diverse immune activities such as induction of chemotaxis of myeloid cells 
, as an inflammatory cytokine in rheumatoid arthritis 
and in endotoxemia 
, and as a macrophage-derived atherogenic factor inducing endothelial dysfunction and vascular smooth muscle cell migration 
Human RETN and mouse Retn genes are orthologs 
and, consequently, due to their evolutionary genetic relatedness, thought to be homologous traits. Surprisingly, human resistin exhibits characteristics which are clearly distinct from those of its mouse counterpart Retn. Furthermore, RETN shares prominent similarity with Retnlg 
, another resistin family member identified only in rodents. This still unrecognized phenomenon took place during evolution of other mammals, like cow and pig, also expressing RETN in myeloid cells 
rather than in adipose tissue 
. The current understanding of this discrepancy is difficult since mouse and human resistin, nevertheless, exhibit common arthritogenic 
and atherogenic 
features even if they are produced by different cells.
In the present study, we carefully examined in which human cells resistin is expressed and also focused on the CEBPE-RETN transcriptional link and the putative immune/inflammatory action in WAT. Consistent with Chumakov et al. 
, we found that RETN generally follows CEBPE expression. Both transcripts were also determined to be attributes of myeloid cells because of the stringent co-localization with EMR1, which appeared to be a common leukocytic marker () rather than an exclusive macrophage determinant 
. Therefore, it is essential to further define the precise mechanism of RETN expression in particular types of leukocytes present in blood and synovia as well as in adipose tissue. We observed a clear discrepancy in expression of the RETN and CEBPE genes in myeloid samples newly obtained in vivo
vs. the same cells conditioned in vitro
. There was a high and concomitant CEBPE-RETN expression in fresh blood and synovial fluid specimens but a reduced and dissociated transcription in cultured PBMC, macrophages and THP-1 cells. The latter immortalized cells do not express RETN but visibly display CEBPE as also corroborated by Northern blot analysis 
. Nonetheless, very low CEBPE expression in macrophages, seen here and by others 
, was yet accompanied by RETN transcription, albeit at a low level (). Although the RETN expression is under ultimate transcriptional control of CEBPE 
, the activity of CEBPE is modified by phosphorylation 
, heterodimerization with abundant associate proteins 
and a synergistic effect of MYB and ATF4 
. Other transcriptional factors, like PPAR gamma 
, SP1 and SP3 
and also SREBF1 
, could influence the RETN regulation in vitro
. The role and activation of CEBPE is different during granulocytic and monocytic differentiation 
. Consequently, CEBPE-to-RETN regulatory strength can change considerably upon various conditions and in certain cell type (). Our results show a higher expression level of both genes in synoviocytes, presumably granulocytes, than in peripheral monocytes as well as a maintained CEBPE-RETN proportion in these ex vivo
Importantly, human preadipocytes and isolated fat cells were clearly devoid of RETN mRNA ( and ), similar to endothelial cells, smooth muscle and many other non-myeloid tissues, which have been tested 
. Furthermore, the resistin release by adipose tissue is not due to adipose cells 
. Together, these data suggest that human resistin should not be acknowledged as an adipose-secreted hormone but as a specific myeloid-derived cytokine.
We also studied the murine genes Cebpe, Retn and Retnlg, because the latter was found as a regulatory and functional equivalent of human RETN 
. Only Retn was detected in adipose tissue of ob/ob mice and 3T3-L1 cells, whereas, in agreement with other results 
, Cebpe and Retnlg expression was observed in mouse peritoneal macrophages. However, very little is known about similarities and differences between rodent Retnlg and human RETN since only one study exists in the field 
. Obviously, lack of recognition of this fact is the basis of a recent proposal of murine Retnlg and the non-existent human RETNLG () as a hormonal link between the digestive tract and insulin resistance in both species 
. Such conclusions only create confusion. Also a presumed arthritogenic Retn-Retnlg cross-function in mice 
needs a further clarification.
Despite many RETN mRNA evaluations by real-time PCR, different problems related to bioinformatic errors and test designs exist 
. In addition, conflicting Northern and SYBR Green RT-PCR results 
and lack of initial Ct values in most reports make firm conclusions impossible. With the aim of having a gene expression analysis as decisive as possible, we included the impartial inter-laboratory reference TaqMan RT-PCR assays from Applied Biosystems in parallel to our custom tests (). They all demonstrated comparable outcomes over the entire investigation also in the single-shot trial of three RETN assays (), including the one that was reported to be successful in the quantification of resistin mRNA in human adipocytes 
. All three constructs detect a common sequence in full-length RETN mRNA () and the multiplexed assays have similar detection efficiency (). A higher level of RETN transcript was detected in inflamed synoviocytes in comparison to normal PBMC. In contrast, no valid quantification could be made in adipose and muscle samples by any of these optimized tests, as indicated by both slope and R2
values in . To obtain a RETN signal in the adipose analysis, McTernan et al. 
overloaded the cDNA to 115 ng in which a proportion of 18S ribosomal RNA cannot be linearly distinguished and, hence, the Ct of the internal reference can not normalize the Ct of the tested gene. Under these conditions, the dCt figures and the final quantity evaluations are, at best, conjectural 
. Taken together, we conclude that RETN is not expressed in human isolated fat cells and that reports claiming positive results have methodological shortcomings which have led to spurious conclusions.
A proinflammatory nature of resistin has been suggested, but its precise action in hepatocytes 
and adipose tissue 
is unclear. We previously noticed that, in PBMC, resistin can induce both gene expression and secretion of inflammatory cytokines like IL6, IL8 and TNF of which only TNF, but not resistin, rapidly up-regulates RETN 
. Interestingly, lipopolysaccharide, a bacterial endotoxin, controls RETN in the same, presumably TNF-mediated, manner (unpublished data). Here we demonstrate that, like in PBMC, resistin can also enhance the gene and protein expression of IL6 and IL8 in human WAT, while the effect on TNF mRNA expression was small. However, in contrast to the well-known TNF-induced suppression of different adipose-specific markers such as CEBPA, FABP4 and SLC2A4, these genes were not affected by resistin at the concentration of 50 ng/ml. These data concur with the unchanged insulin-stimulated glucose uptake by human differentiated adipocytes treated with high concentrations of recombinant resistin (58.5–5850 ng/ml) 
. However, McTernan et al. 
reported that 0.1–50 ng of resistin impaired the glucose incorporation into adipocytes. The cause for this discrepancy is unclear.
We also found that resistin, similar to TNF, induces other proinflammatory genes in human adipose tissue. Chemokine C-C motif ligand 2 (CCL2) is a cytokine that displays chemotactic activity for monocytes. Matrix metallopeptidase 3 (MMP3) may affect the extracellular milieu of adipose tissue akin to that in arthritis. Also PBEF1, pre-B-cell colony enhancing factor 1 or visfatin, was activated by these cytokines. Both the basal and resistin-stimulated gene expression profiles in WAT cultures bear a remarkable resemblance to that seen in PBMC 
, but with clear cell-type differences.
In conclusion, human fat cells do not express or secrete RETN but WAT is evidently a target tissue for resistin. This finding, together with the effect of resistin on PBMC, makes it clear that resistin is an immune-derived systemic and locally acting proinflammatory cytokine in man. However, in contrast to TNF, it does not induce a rapid “dedifferentiation” of the adipose cells. This variation is intriguing and obviously implicates that the resistin intracellular signaling pathway is distinct from that of TNF although both activate NFκB.