Research ArticleKidney Disease

Yin Yang 1 protein ameliorates diabetic nephropathy pathology through transcriptional repression of TGFβ1

See allHide authors and affiliations

Science Translational Medicine  18 Sep 2019:
Vol. 11, Issue 510, eaaw2050
DOI: 10.1126/scitranslmed.aaw2050

A target in diabetic nephropathy

Diabetic nephropathy (DN) is a major cause of kidney failure, but treatments for associated fibrosis and glomerulosclerosis are lacking. Here, Gao and Li et al. showed that Yin Yang 1 (YY1) protein negatively regulates transforming growth factor–β1 (TGFβ1) in human mesangial renal cells (HMRCs) by binding the TGFB1 promoter to repress its transcription. Overexpression and knockdown of Yy1 conversely affected DN progression in diabetic mouse models, and in humans, YY1 expression in glomeruli negatively correlated with TGFβ1 and renal fibrosis. The small molecule eudesmin increased YY1 expression in HMRCs and attenuated renal fibrosis in diabetic mice, demonstrating that YY1 is a potential antifibrotic target in DN.


Transforming growth factor–β1 (TGFβ1) has been identified as a major pathogenic factor underlying the development of diabetic nephropathy (DN). However, the current strategy of antagonizing TGFβ1 has failed to demonstrate favorable outcomes in clinical trials. To identify a different therapeutic approach, we designed a mass spectrometry–based DNA-protein interaction screen to find transcriptional repressors that bind to the TGFB1 promoter and identified Yin Yang 1 (YY1) as a potent repressor of TGFB1. YY1 bound directly to TGFB1 promoter regions and repressed TGFB1 transcription in human renal mesangial cells. In mouse models, YY1 was elevated in mesangial cells during early diabetic renal lesions and decreased in later stages, and knockdown of renal YY1 aggravated, whereas overexpression of YY1 attenuated glomerulosclerosis. In addition, although their duration of diabetic course was comparable, patients with higher YY1 expression developed diabetic nephropathy more slowly compared to those who presented with lower YY1 expression. We found that a small molecule, eudesmin, suppressed TGFβ1 and other profibrotic factors by increasing YY1 expression in human renal mesangial cells and attenuated diabetic renal lesions in DN mouse models by increasing YY1 expression. These results suggest that YY1 is a potent transcriptional repressor of TGFB1 during the development of DN in diabetic mice and that small molecules targeting YY1 may serve as promising therapies for treating DN.


Diabetic nephropathy (DN) has become a leading cause of end-stage renal disease globally (1). About 20 to 40% of diabetic patients have a high risk of developing nephropathic diseases, whereas the remaining diabetic patients are less vulnerable to developing DN, although these two groups have similar durations of diabetes and comparable blood glucose concentrations (2). It is possible that differential expression of protective endogenous factors results in the distinct susceptibility to DN in these two populations.

Hyperglycemia in patients with DN leads to marked glomerular abnormalities, among which oversecretion of transforming growth factor–β1 (TGFβ1) by mesangial cells causes the most severe damage (3). Previous studies have identified TGFβ1 as a leading pathogenic factor in DN (4). Excessive TGFβ1 facilitates the accumulation of extracellular matrix (ECM) proteins, such as fibronectin (FN) and collagen IV (COL4), which consequently lead to glomerular fibrosis (5). The apoptosis of podocytes caused by TGFβ1 further increases the permeability of the glomerular basement membrane and aggravates diabetic albuminuria and renal fibrosis (6). These pathologic alterations promoted by TGFβ1 exacerbate DN progression. Therefore, inhibition of TGFβ1 has been proposed as a promising therapeutic target for treating DN. Despite considerable efforts to develop therapies targeting TGFβ1 (such as neutralizing antibodies), clinical trials have shown that these macromolecule approaches have failed to provide satisfactory outcomes (7).

Some transcription factors that are involved in DN pathology can be overactivated by hyperglycemic or advanced glycation end products (8, 9). Under hyperglycemic conditions, upstream stimulatory factor 2 (USF2) up-regulates the transcription of thrombospondin 1 (TSP1), which, in turn, converts latent TGFβ1 to its active form, accelerating renal fibrosis (8). Other transcription factors such as cyclic adenosine monophosphate response element–binding protein, stimulating protein 1, and nuclear factor κB are also activated by hyperglycemia and contribute to inflammation and glomerulosclerosis (10). Because of the importance of transcriptional factors in the progression of DN, we hypothesized that important unknown TGFβ1 regulators in individuals with mild DN have critical renoprotective functions. Therefore, we surmised that it would be clinically beneficial to determine endogenous transcriptional suppressors targeting TGFB1 (11).


DNA-protein interaction screening identifies Yin Yang 1 as a potential TGFB1 transcriptional regulator

We collected renal biopsy samples from patients with DN and divided them into two groups: diabetic patients with severe DN symptoms, defined by an estimated glomerular filtration rate (eGFR) below 90 ml/min per 1.73 m2, and those without obvious symptoms, defined by an eGFR over 90 ml/min per 1.73 m2. Masson’s trichrome staining showed that renal fibrosis was worse in severe compared to mild DN (Fig. 1A). Immunohistochemical (IHC) analysis confirmed that the group with mild symptoms had much lower expression of glomerular TGFβ1 than the severe group, and as an indicator of active TGFβ1, phosphorylated Smad2 (p-Smad2) abundance was lower in patients with mild compared to severe DN (Fig. 1A).

Fig. 1 YY1 negatively regulates TGFB1 transcription through direct binding to the TGFB1 promoter.

(A) Representative images of immunohistochemical (IHC) staining and Masson’s trichrome staining of clinical renal glomeruli healthy controls (Cntr) (n = 3) and patients with mild DN symptoms (eGFR over 90 ml/min per 1.73 m2; n = 20) termed “Mild DN” and severe DN symptoms (eGFR < 90 ml/min per 1.73 m2; n = 40) termed “Severe DN.” (B) Schematic of procedures for mass spectrometry (MS) screening in low glucose–treated (5.5 mM) and high glucose–treated (30 mM) human renal mesangial cells (HRMCs). Strep., streptavidin. (C) Screening identified YY1 bound to the TGFB1 promoter. m/z, mass/charge ratio. (D and E) Protein (D) and mRNA expression (E) of indicated genes in HRMCs transfected with a YY1-expressing plasmid. GAPDH, glyceraldehyde-3-phosphate dehydrogenase. (F and G) Protein (F) and mRNA expression (G) of indicated genes in HRMCs transfected with siYY1. siNC, no target siRNA control. (H) Luciferase reporter activity of TGFB1 and protein abundance in HRMCs transfected with increasing doses of YY1-expressing plasmid. (I) Luciferase reporter assays of TGFB1 promoter fragments fused with a luciferase reporter gene (Luc1 to Luc6) in YY1-transfected HRMCs. (J) Chromatin immunoprecipitation (ChIP) assay–qPCR analysis in the region −3258/−2854 of the TGFB1 promoter. IgG, immunoglobulin G. (K) Three putative YY1 binding sites in the TGFB1 promoter. (L) Electrophoretic mobility shift assay (EMSA) of YY1 binding to the TGFB1 promoter with three probes. (M) An EMSA was performed on the TGFB1 promoter. Ab, antibody. (N) A point mutation of the YY1 binding site on the TGFB1 promoter (−3123/−3115) was constructed (TGFB1 YY1-mut). (O) Reporter activity of wild-type TGFB1 (TGFB1 wt) or mutated TGFB1 (TGFB1 YY1-mut) with YY1 overexpression. Data are means ± SEM. *P < 0.05, one-way ANOVA.

To identify critical factors down-regulating TGFB1 transcription, we designed a DNA-protein screening experiment in human renal mesangial cells (HRMCs) (Fig. 1B). Consistent with our approach focusing on transcription factors that down-regulate TGFB1, we identified three transcriptional suppressors [Yin Yang 1 (YY1), p66-beta, and CTCF (CCCTC-binding factor)] from the promoter DNA-protein precipitates in HRMCs (Fig. 1C and fig. S1, A and B). We chose YY1 as the transcription factor of interest because of its potential involvement in glucose metabolism (12, 13) and its identification exclusively in high glucose–treated HRMCs but not in low glucose–treated cells (data file S1).

YY1 directly binds to the TGFB1 promoter and suppresses its transcription

Immunoblotting and quantitative reverse transcription polymerase chain reaction (qRT-PCR) assays showed that expression of TGFβ1 and downstream proteins COL4 and FN was reduced by overexpression of YY1 and increased by knockdown of YY1 in HRMCs (Fig. 1, D to G). In addition, YY1 overexpression attenuated glucose-induced high TGFβ1 expression, and YY1 deficiency aggravated it (fig. S1C). Overexpression of YY1 also reduced TGFB1 luciferase activity (Fig. 1H). We cloned a series of TGFB1 promoter deletion mutants into the luciferase reporter system (Luc1 through Luc6; Fig. 1I) and found that overexpression of YY1 suppressed TGFB1 activity only for the Luc1 and Luc2 constructs, indicating that the minimal YY1 binding site within the TGFB1 promoter was between −3258 and −2853 base pairs (bp).

We further performed chromatin immunoprecipitation (ChIP) assays and found that YY1 was recruited to the −3258- to −2853-bp region (Fig. 1J). In addition, three putative YY1 binding motif sequences were predicted in the region using promoter analysis tools PROMO and TFSEARCH (Fig. 1K). Electrophoretic mobility shift assays (EMSA) confirmed that the −3123/−3115 region was probably the only YY1 binding site within the fragment of the TGFB1 promoter (Fig. 1, L and M).

We next constructed a mutant luciferase reporter plasmid with a TGFB1 promoter sequence containing a point mutation (AGCCaTAGT→AGCCgTAGT) in the YY1 binding site (residues −3123 to −3115), which we termed TGFB1 YY1-mut (Fig. 1N); the reporter system containing a wild-type TGFB1 promoter was used as control (TGFB1 wt). Overexpression of YY1 decreased TGFB1 luciferase activity when cells were transfected with TGFB1 wt reporter system; this effect was abolished when the YY1 binding site was mutated (Fig. 1O). These results collectively demonstrate that region −3123 to −3115 within the TGFB1 promoter is the effective binding site of YY1.

Expression of YY1 negatively associates with renal TGFβ1 in DN mice

We found that YY1, TGFβ1, and COL4 expression showed no changes in db/dm mice at different ages (fig. S2, A and B). However, YY1 was increased markedly in 4-week-old db/db mice and tended to decrease when mice were 24 weeks old (Fig. 2, A to C). By contrast, TGFβ1 started to increase at the time point when YY1 began to decrease and further enhanced with diabetic conditions extended (Fig. 2, A to C). Consistently, IHC analysis further confirmed that YY1 in glomeruli increased in early diabetes and decreased later (Fig. 2, D and E). TGFβ1, p-Smad2, and FN, however, continuously increased with age, and a surge in TGFβ1, p-Smad2, and FN occurred in conjunction with a decrease in YY1 expression (Fig. 2, D and E), when renal fibrosis was markedly worse (Fig. 2F). Female db/db mice showed similar variation in renal YY1 expression as males, indicating the sex-independent expression pattern of YY1 (fig. S2, C to F). Moreover, a consistent result was found that YY1 significantly decreased in mice bearing type 1 diabetes for 12 weeks (P < 0.05; Fig. 2, G and H), when DN symptoms were obviously detected (Fig. 2, I to M). Together, these results show that YY1 expression negatively correlates with TGFβ1 and the extent of renal fibrosis.

Fig. 2 YY1 expression in glomeruli is negatively associated with TGFβ1 in db/db mice.

(A to F) Kidney samples of db/db mice were collected at indicated time points. Control (Cntr): C57BLKS/J db/dm mice at 8 weeks (n = 3 for each group). (A) Immunoblotting assays of YY1, TGFβ1, and FN in db/db mice of different ages and control mice. (B) Quantification of (A). ***P < 0.001 versus TGFβ1 and versus FN. (C) mRNA expression of Yy1 and Tgfb1 in db/db mice. ***P < 0.001 versus Tgfb1. (D) Representative images of IHC staining of db/db mice of different ages. (E) Densitometric quantitative results of indicated proteins in (D). ***P < 0.001 versus TGFβ1, versus FN, and versus p-Smad2. (F) Representative images of H&E, PAS, and Masson’s trichrome staining of db/db mice of different ages. (G to M) Kidneys of 12-week-old STZ-induced type 1 diabetic mice. Control (Cntr): C57BL/6J mice (n = 10 for each group). (G to K) YY1 protein expression (G), Yy1 mRNA expression (H), blood glucose (I), proteinuria (J), and representative images of PAS and Masson’s trichrome staining (K) of control and STZ-injected mice. (L and M) Densitometric quantification of PAS (L) and Masson’s trichrome staining (M) in (K). Data are means ± SEM. Two-way ANOVA was used for the db/db mouse model (B, C, and E), and Wilcoxon signed-rank test (H to M) was used for STZ-injected mice. *P < 0.05 and ***P < 0.001.

YY1 expression is specifically augmented in mesangial cells under diabetic conditions

Next, in a 5/6 nephrectomized mouse model (5/6 Nx mice), which exhibits albuminuria and renal failure without development of hyperglycemia (14), we detected glomerular fibrosis and elevated glomerular TGFβ1, p-Smad2, and FN expression (Fig. 3, A to I, and fig. S3). However, in contrast to the db/db mice, the 5/6 Nx mice exhibited unchanged glomerular YY1 expression compared with control mice (Fig. 3, D and E), suggesting the specific up-regulation of YY1 in diabetic kidneys.

Fig. 3 YY1 is increased in mesangial cells from kidneys of DN mice but remains unchanged in 5/6 Nx mice.

(A to I) Eight-week-old male C57BL/6J mice were used for a 5/6 Nx mouse model (control group, n = 6 and 5/6 Nx group, n = 7). Control (Cntr): Sham-operated mice. (A) Histopathological staining. (B) Quantification of PAS and (C) Masson’s trichrome staining in (A). (D) Representative images of IHC staining. (E) Quantitative assessment of (D). (F) YY1 and TGFβ1 protein expression and (G) mRNA expression in the kidneys. (H) Tight-slit pore density (I) thickness of the glomerular basal membrane (GBM) from transmission electron microscopy analysis. (J to L) Immunofluorescence analysis of YY1 and specific markers for (J) mesangial cells (α-SMA), (K) podocytes (WT1), and (L) tubular epithelial cells (AQP-1) in kidneys from db/dm (Cntr) and db/db mice (16 weeks). DAPI, 4′,6-diamidino-2-phenylindole. (M) Quantification of immunofluorescence staining in (J). (N to P) mRNA expression in (N) HRMCs, (O) mouse renal podocytes (MPCs), and (P) rat tubular epithelial cells (NRKs) treated with high glucose (30 mM). (Q to S) Protein expression in high glucose (HG)–treated HRMCs (Q), MPCs (R), and NRKs (S). Data are means ± SEM. Wilcoxon signed-rank test (B, C, and I) and Student’s t test (H and M) were used for 5/6 Nx mice, or two-way ANOVA was used for in vitro analysis (N). *P < 0.05, **P < 0.01, and ***P < 0.001.

Furthermore, we found that the increased YY1 fluorescence overlapped well with the signal from the mesangial cell marker α–smooth muscle actin (α-SMA) (Fig. 3J). By contrast, little YY1 fluorescence colocalized with Wilms tumor type 1 (WT1) fluorescence, a marker of podocytes (Fig. 3K), or with Aquaporin-1 (AQP-1) fluorescence, a marker of renal tubular cells (Fig. 3L). Quantification of YY1-positive cells further showed a threefold increase in the expression of YY1 in mesangial cells (P < 0.01; Fig. 3M), suggesting a mesangial cell–dominated up-regulation of YY1 after a high glucose stimulus. Western blot and qRT-PCR analysis in cultured HRMCs, mouse podocytes, and rat tubular cells (NRKs) confirmed an alteration of YY1 expression specifically in mesangial cells under hyperglycemic conditions (Fig. 3, N to S).

Nuclear factor erythroid 2–like 2 interacts with and up-regulates YY1

Nuclear factor erythroid 2–related factor 2 (Nrf2) also increased in HRMCs after short-term high glucose treatment but decreased at prolonged administration (Fig. 4A), indicating a similar dynamic alteration in expression to that of YY1 (Fig. 3Q). Moreover, Nrf2 deficiency blocked high glucose–induced YY1 expression (Fig. 4B). Overexpression of Nrf2 notably increased, whereas knockdown of Nrf2 decreased YY1 at both the protein (Fig. 4, C and D) and mRNA transcript abundance (P < 0.05; Fig. 4, E and F). As a transcription factor, the abundance of nuclear YY1 positively correlates with its activity. As shown in Fig. 4G, the cytoplasmic expression of Nrf2 and YY1 was increased, whereas the nuclear forms were decreased in the kidney of db/db mice (20 and 24 weeks old). These data indicate that YY1 is positively modulated by Nrf2.

Fig. 4 Nrf2 interacts with and up-regulates YY1.

(A) YY1 and Nrf2 protein expression in HRMCs treated with high glucose (30 mM) for indicated periods. (B) Nrf2 and YY1 in the presence or absence of Nrf2 small interfering RNA (siRNA) in HG-treated HRMCs. (C and D) Proteins in HRMCs transfected with Nrf2 (C) or siNrf2 (D). (E and F) mRNA expression in HRMCs transfected with (E) Nrf2 or (F) siNrf2. HO-1 (heme oxygenase 1) was used as control. (G) Cytoplasmic and nuclear Nrf2 and YY1 from the kidneys of db/db mice at different ages (n = 3). Control (Cntr): db/dm mice, 8 weeks. (H) ChIP-qPCR and ChIP-PCR analysis for binding of Nrf2 and YY1 to the TGFB1 promoter in HRMCs. (I) Immunoblotting assays after immunoprecipitation of YY1 from HRMCs. LG, low glucose. (J) Immunoblotting assays after immunoprecipitation of Nrf2 from HRMCs. (K) qPCR and PCR analysis of ChIP-reChIP assays for binding of Nrf2 and YY1 to the TGFB1 promoter in HRMCs. (L) Reporter activity and TGFβ1 expression were determined in HRMCs. Data are means ± SEM. Significance was determined by one-way ANOVA. *P < 0.05. HSP90, heat shock protein 90.

Consistent with previous findings (15), ChIP assays revealed that Nrf2 bound to the TGFB1 promoter (Fig. 4H), although there was no direct Nrf2 binding motif detected within the TGFB1 promoter (16). A coimmunoprecipitation assay demonstrated a physical interaction between Nrf2 and YY1 (Fig. 4, I and J). Sequential ChIP (ChIP-reChIP) analysis further confirmed the interaction between Nrf2 and YY1 when they both bound to the TGFB1 promoter (Fig. 4K). In addition, both luciferase assays and Western blot analysis demonstrated that decreases in Nrf2 or YY1 significantly up-regulated the luciferase activity of the TGFB1 promoter, as well as the TGFβ1 and FN proteins (P < 0.05; Fig. 4L). Likewise, the expression of TGFβ1 and FN was suppressed by transfection of siNrf2 alone (fig. S4A) or together with increasing doses of siYY1 (fig. S4B). These results suggest that Nrf2 interacts with YY1 and the YY1/Nrf2 complex reinforces the YY1-mediated suppression of TGFβ1.

Knockdown of YY1 in glomeruli aggravates DN progression

Next, we generated a renal YY1 knockdown (YY1-KD) model using mice carrying homozygous-conditional flox/flox alleles (YY1flox/flox) (13, 17). As shown in Fig. 5 (A and B), the expression of renal YY1 was significantly reduced and that of TGFβ1 was markedly increased in the AAV2 (adenovirus-associated virus 2)–Cre–injected mice (P < 0.05). Immunofluorescence staining assays confirmed that YY1 in glomeruli was diminished in YY1-KD mice (Fig. 5C), without any detectable effects on body weight or blood glucose (Fig. 5, D and E).

Fig. 5 Knockdown of YY1 exacerbates diabetic kidney lesions.

(A to P) An HFD/STZ diabetic model was induced in YY1flox/flox mice. Animals were subsequently in situ renal injected with AAV2-Cre (YY1-KD group, n = 6) or AAV2-GFP [control (Cntr) group, n = 7]. (A) Expression of Yy1, Tgfb1, and Fn in kidneys. (B) Immunoblot analysis of YY1 and TGFβ1 in kidneys. (C) Immunofluorescence of YY1 in renal glomeruli. (D) Body weight and (E) blood glucose of YY1-KD and control mice. (F) Representative images of H&E, PAS, and Masson’s trichrome staining of glomeruli. (G and H) Quantitative analysis of mesangial expansion (G) and mesangial fibrosis (H) in glomeruli in (F). (I to K) Urea albumin creatinine ratio (UACR) (I), urine cystatin C (J) in the urea, and blood urea nitrogen (BUN) (K) in the blood of YY1-KD and control mice. (L) Transmission electron microscopy analysis of glomerular lesions. (M and N) Quantitative analysis of tight-slit pore density (M) and thickness of the glomerular basal membrane (GBM) (N) in (L). Foot process fusion was estimated by measuring the density of tight-slit pore. (O) IHC staining of indicated proteins in renal glomeruli. (P) Quantitative analysis of the IHC staining in (O). Data are means ± SEM. *P < 0.05, **P < 0.01, and ***P < 0.001, two-way ANOVA (D and E), Wilcoxon signed-rank test (G, M, and N), and Student’s t test (H to K).

As indicated by hematoxylin and eosin (H&E), periodic acid–Schiff (PAS), and Masson’s trichrome staining, YY1-KD mice exhibited exacerbated glomerular fibrosis compared to control mice (Fig. 5, F to H). Furthermore, the urea albumin creatinine ratio (UACR), urinary cystatin C, and blood urea nitrogen (BUN) concentrations were significantly higher in the YY1-KD diabetic mice than in the AAV2-GFP (green fluorescent protein)–injected diabetic mice (P < 0.001, Fig. 5, I and J; P < 0.05, Fig. 5K). Consistent with this observation, electron microscopy showed that the YY1-KD diabetic mice exhibited a higher tight-slit pore density (Fig. 5, L and M) and a thicker glomerular basal membrane (Fig. 5, L and N). IHC staining showed that when YY1 was depleted from glomeruli, TGFβ1, p-Smad2, and FN were elevated (Fig. 5, O and P). Collectively, these results reveal that loss of YY1 accelerates DN development.

AAV2-mediated YY1 gain of function in glomeruli ameliorates the progression of DN

To overexpress renal YY1, we used an AAV2 system with renal in situ injection in mice with type 2 diabetes induced by a high-fat diet (HFD) and streptozocin (STZ) [termed YY1-overexpression (YY1-OE)]. As shown in Fig. 6 (A and B), AAV2-YY1 injection up-regulated renal YY1 mRNA and protein abundance. Immunofluorescence staining confirmed that YY1 in glomeruli was markedly up-regulated after AAV2-YY1 injection (Fig. 6C), which did not affect body weight or blood glucose (Fig. 6, D and E). Accordingly, mRNA and protein expression of TGFβ1 and FN was suppressed in the kidneys of YY1-OE mice (Fig. 6, A and B).

Fig. 6 AAV2-mediated YY1 overexpression ameliorates diabetic glomerulosclerosis.

(A to O) An HFD/STZ diabetic model was induced in C57BL/6J mice. Mice were then in situ renal injected with AAV2-GFP [control (Cntr) group] or AAV2-YY1 [YY1-overexpressing (YY1-OE) group] (n = 10 for each group). (A) Relative mRNA expression of Yy1, Tgfb1, and Fn in kidneys. (B) Representative immunoblot analysis of indicated proteins in kidneys. (C) Immunofluorescence of YY1 in renal glomeruli. (D) Body weight, (E) blood glucose, (F) UACR, (G) urine cystatin C, (H) BUN, and (I) kidney/body weight ratios in YY1-OE and control mice. (J) Representative images of H&E, PAS, and Masson’s trichrome staining of glomeruli. (K and L) Quantitative analysis of mesangial expansion (K) and mesangial fibrosis (L) in glomeruli in (J). (M) Transmission electron microscopy analysis of glomerular lesions. (N) Quantification of tight-slit pore density in (M). Foot process fusion was assessed by the density of tight-slit pore. (O) Quantification of thickness of GBM in (M). (P) IHC staining of indicated proteins in renal glomeruli. (Q) Quantitative results of the IHC staining in (P). Data are presented as means ± SEM. One-way ANOVA (A and Q), Wilcoxon signed-rank test (F to I, K, L, N, and O), and two-way ANOVA (D and E). *P < 0.05 and **P < 0.01. NS, not significant.

Furthermore, UACR was significantly lower in YY1-OE diabetic mice than control diabetic mice (P < 0.001; Fig. 6F). YY1-OE mice also exhibited markedly reduced urinary cystatin C (Fig. 6G) and BUN (Fig. 6H). Although the kidney/body weight ratios between the two groups were not different (Fig. 6I), glomerular fibrosis was ameliorated in YY1-OE mice, as evidenced by H&E, PAS, and Masson’s trichrome staining (Fig. 6, J to L). Electron microscopy analysis also showed that YY1-OE mice exhibited lower density of tight pores of podocytes and less ECM deposition (Fig. 6, M to O) compared with control mice. Prominently reduced TGFβ1, p-Smad2, and FN, as assessed by IHC analysis, also supported the hypothesis that YY1 overexpression inhibited ECM accumulation in glomeruli (Fig. 6, P and Q). These data suggest that the overexpression of YY1 in glomeruli attenuates DN development by suppressing TGFβ1.

Glomeruli-resident YY1 negatively associates with TGFβ1 expression and DN development in individuals with diabetes

To further validate the role of YY1 in DN, we collected renal biopsies from human patients pathologically diagnosed with DN and divided these patients into mild DN (n = 20) and severe DN (n = 40) based on eGFRs (table S1). Although patients in two groups had comparable durations of diabetic course, age, and blood glucose, the group with mild DN had significantly lower proteinuria values compared to the severe DN group (P < 0.001; table S1). IHC analysis showed that YY1 in glomeruli was significantly elevated in the mild DN group (P < 0.05; Fig. 7, A and B). In contrast to YY1, TGFβ1 and its downstream targets p-Smad2, FN, and COL4 were minimal in the mild DN but markedly up-regulated in severe DN (Fig. 7, A and B).

Fig. 7 Expression of YY1 in renal biopsies of DN patients correlates with DN progression.

(A) IHC staining of YY1, TGFβ1, FN, and COL4 in control patients with minimal change disease [control (Cntr), n = 3], mild DN (n = 20), and severe DN (n = 40). (B) Quantitative results of YY1, TGFβ1, p-Smad2, FN, and COL4 in IHC staining in (A). (C) H&E, PAS, and Masson’s trichrome staining of the indicated human renal biopsies. (D and E) Quantitative analysis of PAS staining (D) and Masson’s trichrome staining (E) of the control and DN individuals in (C). (F to H) Correlation of YY1 expression in human renal biopsies with various parameters, including eGFR (F), proteinuria (G), p-Smad2 (H), and FN expression (I). Red dots represent patients with severe DN; black dots represent patients with mild DN. Data are means ± SEM. *P < 0.05, one-way ANOVA.

Consistent with the elevated TGFβ1, the severity of glomerulosclerosis was worse in the group with severe DN (Fig. 7, C to E). Furthermore, statistical analysis showed that the staining intensity of YY1 positively correlated with the eGFR (P < 0.0001; Fig. 7F) but negatively with albuminuria (P = 0.0015; Fig. 7G), p-Smad2 (P < 0.0001; Fig. 7H), and ECM component FN (P < 0.0001; Fig. 7I). Together, these results suggest that YY1 may play a renoprotective role by down-regulating TGFβ1 in the setting of diabetic pathology.

Eudesmin treatment protects kidneys from diabetic lesions through YY1 activation

We next sought to identify small-molecule compounds that could stimulate YY1 expression and thereby reverse TGFβ1-mediated kidney fibrosis. To this end, we constructed a luciferase reporter containing a 1500-bp fragment of the YY1 promoter. Using a commercial chemical library, we found that the small molecule eudesmin (EDN) increased the transcriptional activity of Luc-YY1 more than 2.1-fold (P < 0.001) and down-regulated Luc-TGFB1 activity by threefold (P < 0.001) in HRMCs (Fig. 8A, fig. S5A, and table S2), which we confirmed by Western blot assays (fig. S5B). Moreover, in YY1-depleted HRMCs, we found that EDN treatment could no longer induce YY1 expression or decrease TGFβ1 (fig. S5C); however, other compounds still inhibited TGFβ1 in the absence of YY1 (fig. S5D). These data suggest that the effect of EDN, but not the other assessed compounds, is dependent on YY1.

Fig. 8 Eudesmin treatment protects mice from diabetic kidney lesions by activating YY1.

(A) Heat map representing the screening hits obtained from a luciferase reporter screening system. (B and C) mRNA (B) and protein expression (C) of YY1, TGFβ1, and FN from HRMCs treated with eudesmin (EDN; 1 μg/ml) in the presence or absence of high glucose (30 mM). (D to K) HFD/STZ-induced diabetic mice were injected with saline (Cntr) or EDN for 8 weeks (n = 10 in each group). (D and E) Representative images (D) and quantitative results (E) of IHC staining of control and EDN-injected mice. (F to H) Immunoblotting (F) and qRT-PCR (G) analysis and H&E, PAS, and Masson’s trichrome staining (H) of control and EDN-injected mice. (I to K) Zand BUN (K) of control and EDN-injected mice. (L to S) HFD/STZ-induced mice were injected with saline (Cntr) or EDN for 8 weeks in the presence or absence of YY1-KD (n = 10 in each group). (L) H&E, PAS, and Masson’s trichrome staining of control or EDN-treated mice with or without YY1-KD. (M to P) UACR (M), urine cystatin C (N), BUN (K), and representative images of IHC staining (P) in control or EDN-injected mice with or without YY1 depletion. (Q) Quantitative results of (P). (R and S) qRT-PCR (R) and representative immunoblotting (S) analysis in kidney samples of control or EDN-injected mice with or without YY1 depletion. Results are means ± SEM. Wilcoxon signed-rank test (I to K), one-way ANOVA (B, E, G, M to O, Q, and R). *P < 0.05.

qRT-PCR and immunoblotting analysis demonstrated that EDN blocked high glucose–induced expression of TGFβ1 and FN in mesangial cells (Fig. 8, B and C). Subsequent assessments of HFD/STZ-induced diabetic mice showed that EDN administration significantly up-regulated Yy1 expression and suppressed the abundance of TGFβ1, p-Smad2, and FN in glomeruli compared to saline-treated mice (P < 0.05; Fig. 8, D to G). Histopathological studies showed that EDN treatment improved glomerulosclerosis (Fig. 8H and fig. S5, E and F) without affecting body weight or blood glucose (fig. S5, G and H). In line with these findings, analyses of biochemical parameters showed that UACR, urea cystatin C, and BUN were lower in EDN-treated mice (Fig. 8, I to K). Furthermore, electron microscopy revealed that EDN-injected HFD/STZ-induced diabetic mice exhibited less foot process fusion and ECM deposition compared to the control mice (fig. S5, I to K).

As shown in Fig. 8L and fig. S5 (L and M), YY1-KD mice showed aggravated renal fibrosis compared to control mice in type 2 diabetic kidneys. After knockdown of YY1, EDN treatment showed no additional effects on renal fibrosis when compared with YY1-KD mice (Fig. 8L), suggesting that EDN affected its role mainly through YY1. The body weight and blood glucose remained steady in all groups (fig. S5, N and O). UACR, cystatin C, and BUN were notably increased in YY1-KD mice but remained unchanged with additional EDN injection (Fig. 8, M to O). Histopathology analysis confirmed the above observations that EDN injection caused no additional protection after YY1-KD (Fig. 8, P and Q). Furthermore, EDN could not confer enhancement of YY1 and suppression of TGFβ1 and FN after YY1 ablation from glomeruli (Fig. 8, R and S).

We next examined the biosafety and off-target effects of EDN on regulating renal fibrosis. EDN caused no toxic effects on cell viability in vitro (fig. S6A) or in vivo (fig. S6B). Serum alanine transaminase and aspartate transaminase concentrations remained stable in mice injected with or without EDN (fig. S6, C and D), indicating the biosafety of EDN.

As shown in Fig. 8, EDN treatments promoted the expression of Yy1 in mouse models, suggesting that EDN might stimulate the transcription of Yy1. To further study the mechanism of EDN-driven enhancement of YY1, we examined the effects of EDN on YY1 protein degradation. YY1 was decreased after 72 hours of high glucose treatment and further decreased when cycloheximide, a protein synthesis inhibitor, was coincubated with high glucose–treated HRMCs. EDN treatment retarded long-term high glucose–induced YY1 degradation even with cycloheximide coincubation (fig. S6E), indicating that EDN stabilized YY1 by inhibiting its degradation. Together, we concluded that EDN elevates YY1 at both transcriptional and posttranslational abundance.

Previous reports showed that EDN could inhibit S6 kinase 1 (S6K1) in mesenchymal stem cells (18), inhibit apoptosis in human gastric adenocarcinoma cells (19), and promote apoptosis in lung cancer cells (20). However, our data showed the stable proteins of cleaved caspase 3 and the phosphorylation of S6 in HFD/STZ-induced mice and HRMCs treated with or without EDN (fig. S6, F and G). Other S6K1 inhibitors (H89 2HCl and PF-4708671) or apoptosis inhibitors [apoptosis inhibitor, apoptosis inhibitor II, Bax channel blocker, transformed mouse 3T3 cell double minute 2 (MDM2) antagonist, Rho-associated coiled-coil containing protein kinase (ROCK) inhibitor, and doxorubicin] failed to alter the protective effect of EDN on renal fibrosis (fig. S6, H to J). These results demonstrate that EDN attenuates renal fibrosis by up-regulating YY1, independent of S6K1 activity and apoptosis-related pathway signaling.

To further confirm the therapeutic efficacy of EDN on DN development in vivo, we administered EDN to db/db mice at 12 weeks of age, when the mice already had obvious glomeruli damage. After 2 months of treatment with EDN, db/db mice showed attenuated mesangial expansion and fibrosis as shown by H&E, PAS, and Masson’s trichrome staining (fig. S7, A to C) as well as repressed expression of TGFβ1, p-Smad2, and FN (fig. S7, D to G). In parallel, UACR, cystatin C, and BUN were consistently decreased after EDN treatment, whereas body weight and blood glucose remained steady (fig. S7, H to L). Furthermore, electron microscopy results revealed that db/db mice exhibited less foot process effacement and ECM deposition compared to control mice (fig. S7, M to O). These results collectively suggest that EDN injection ameliorates renal fibrosis compared to control mice by increasing YY1 expression and thus decreasing TGFβ1 pathway.


It has been well documented that the TGFβ superfamily is key to the pathogenesis of DN and end-stage renal disease and that TGFβ1 is the most pivotal fibrogenic cytokine (4). Because of its deleterious role in the development of DN, numerous approaches have been explored to inhibit this profibrotic cytokine, such as the use of neutralizing antibodies (21). The neutralization of TGFβ1 is protective against DN in rodent models (21); however, this strategy has not yielded satisfactory outcomes in human patients (21). For example, in a phase 2 clinical trial with 416 patients with DN, an anti-TGFβ1 antibody failed to delay the progression of DN after 12 months of treatment and follow-up (7). The lack of efficacy of such a strategy may result from low accessibility of the antibodies in target cells due to large molecular weight, low diffusion, and limited stability (22).

Diabetic hyperglycemia promotes TGFβ1 expression through multiple transcription factors. USF1/2 have been linked to glucose-regulated genes in mesangial cells (9, 23). USF1 binds to the TGFB1 promoter between −835 and −406 bp as both a homodimer and heterodimer with USF2 and activates TGFβ1 expression, leading to renal fibrosis (24). Furthermore, high glucose drives activating protein 1 (AP-1)–mediated up-regulation of TGFB1 and AP-1 is also involved in angiotensin-driven and TGFβ1-independent FN transcription (25). However, these transcriptional factors are mainly transactivators of TGFβ1, whereas suppressors of TGFβ1 were rarely reported (6). The present study identified a direct transcription repressor of TGFB1 as a therapeutic target for the treatment of DN. The direct binding site of YY1 on TGFB1 promoter was found between the −3123 and −3115 fragments, located 2280 bp away from USF1/2 (23, 24). Our study might broaden the knowledge of transcriptional regulation of TGFB1 and highlight the importance of endogenous transcription factors in protecting against DN (6, 16).

YY1 is a ubiquitously expressed and evolutionarily conserved zinc finger protein that can act as a transcriptional repressor or activator (26). Our study revealed the protective effects of YY1 against DN using both in vitro and in vivo experimental approaches and provided insights into YY1 function. First, our conclusions were made in a diabetic state. YY1 expression was elevated by diabetic hyperglycemia in DN mouse models but not by kidney failure in 5/6 Nx mice, which exhibit similar albuminuria but without hyperglycemia. In particular, YY1 was only up-regulated in mesangial cells, which play pivotal roles in production of ECM and subsequent glomerulosclerosis in DN (27). Second, the distinct role of YY1 across disease types may also be explained by the fact that different cofactors were recruited to YY1 in response to stimulation, which consequently modulated YY1 functionality as a suppressor or activator in transcriptional regulation (26). Previous reports have documented that YY1 interacts with a variety of transcription factors (28). In this study, Nrf2 formed a complex with YY1, bound to the TGFB1 promoter, and served as a cofactor of YY1. Nrf2 not only suppressed TGFB1 with YY1 but also promoted YY1 transcription and translation. It has been reported that Nrf2 is up-regulated by high glucose–induced reactive oxygen species (6) and that long-term high glucose treatment also down-regulates Nrf2 by increasing kelch-like ECH-associated protein 1 (Keap1)–mediated ubiquitination of Nrf2 (29). Given that Nrf2 regulates the mRNA and protein expression of YY1, the loss of Nrf2 in patients with severe DN could at least partially contribute to the reduction of YY1 in the kidneys.

We used a small-molecule library to screen for naturally occurring compounds that activate the expression of YY1 to protect against DN. The small molecule EDN increased YY1 expression in mouse glomeruli and mesangial cells and functionally ameliorated DN progression. EDN is a natural product isolated from various plant families—such as Apiaceae, Rutaceae, Ochnaceae, and Magnoliaceae (30)—and has broad bioactivity, including regulating apoptosis and adipogenic differentiation (18, 19). Our study might extend the application of EDN to the treatment of various diseases such as DN, where we showed that EDN prevented renal fibrosis through elevation of YY1 independent of apoptosis and S6K1. We also found that EDN retarded DN progression without any alterations to food intake, body weight, or detectable cellular and tissue toxicity in vitro and in vivo. Moreover, EDN is a small-molecule compound that has a higher diffusion rate in kidneys compared to that of macromolecules, which could make it a candidate for development as therapeutic option for patients with DN. In addition, TGFβ1 is also involved in the development of other diseases. For instance, excessive hypothalamic TGFβ1 is closely associated with the development of diabetes and aging (31). Thus, it merits further investigation whether EDN, a potent suppressor of TGFβ1, could aid in treating other TGFβ1-mediated pathogenic diseases.

One limitation of this study is that we could not specifically knock out YY1 in renal mesangial cells. YY1 was up-regulated in mesangial cells in the setting of diabetic pathology; however, no specific marker of mesangial cells was identified, and therefore, no specific Cre-mice were generated. In addition, while conducting a genome-wide screen for TGFβ1 transcriptional suppressors with mass spectrometry analysis, we also detected other two transcription factors, CTCF and p66-beta, in the protein precipitation pulled down by the fragment of TGFB1 promoter. However, these factors were not further investigated here and may be studied in the future.

In summary, the current study identifies and characterizes YY1 as a critical transcription suppressor of TGFβ1, and a small molecule, EDN, which has the potential to reverse glomerulosclerosis and slow DN progression primarily through the activation of YY1 (fig. S8), indicating a promising new avenue for ameliorating DN development.


Study design

This study aimed to explore the mechanism of how renal YY1 suppresses TGFβ1 and to identify an agonist of YY1 that relieves symptoms in mouse models of DN. Initially, we identified the mechanism underlying the regulatory network between YY1 and TGFβ1 in human mesangial cells. We then performed renal-specific overexpression and depletion of YY1 in DN mouse models by renal in situ injection of AAV2 to investigate the effect of YY1 on the development of DN. Subsequently, renal biopsy samples from patients with mild or severe DN symptom were collected. We studied the clinical association of YY1 expression and DN progression by analyzing these human samples. Each patient was diagnosed with DN and gave full consent to participate in our study. Sample sizes are listed in figure legends. We then performed drug screening for a small molecule up-regulating YY1 with YY1 and TGFB1 promoter–luciferase systems and assessed the efficacy of the identified compound, EDN, in human mesangial cells and three DN mouse models. Investigators and data analyzers were blinded for mouse and human samples. All of the data were included in analyses except for data points that were further than ±2 SDs from the group mean. Replicates for each experiment are given in figure captions or data file S2.

Human renal specimens

Human samples were from renal biopsy patients admitted to Shanghai Jiao Tong University Affiliated Sixth People’s Hospital during 2012–2018 (approval no.: 2017-KY-005). Informed consent was obtained from each patient, and patients’ clinical biochemical data were extracted from medical records. Assessment of the histopathology was evaluated in a blinded manner by two experienced pathologists and classified according to the Renal Pathology Society’s Pathologic Classification of Diabetic Nephropathy on the basis of glomerular pathology. Patients diagnosed with minimal change disease were considered as control group (n = 3). Because patients with DN might present with a normal eGFR (≥90 ml/min per 1.73 m2), we classified the cases into two groups: mild DN (DN with normal eGFR, n = 20) and severe DN (DN with eGFR < 90 ml/min per 1.73 m2, n = 40).

Mouse models

Animal studies were conducted in accordance with the guidelines from the Ethics Committee of Shanghai Jiao Tong University Affiliated Sixth People’s Hospital (approval no.: 2017-0039). Mice were housed with ad libitum access to food and water in light- and temperature-controlled environments (12-hour/12-hour light/dark, 22°C). The present study established the following mouse models.

C57BLKS/J db/db mice. Four-week-old male C57BLKS/J-LepR (db/db) mice with genetic type 2 diabetes and healthy control mice (db/dm) were purchased from the Model Animal Research Center of Nanjing University. Three of the mice were randomly selected and euthanized at 4, 8, 12, 16, 20, 24, 28, and 32 weeks, respectively. EDN was intraperitoneally injected into db/db mice (10 mg/kg) every other day for 8 weeks.

HFD/STZ-induced diabetic mice. Three-week-old male C57BL/6J mice were purchased from the Model Animal Research Center of Nanjing University. After adaptation for 1 week, mice were fed an HFD for 6 weeks followed by low-dose STZ injection (intraperitoneal, 40 mg/kg) three times with 1-day intervals. Type 2 diabetes was considered to be successfully established when the random blood glucose was more than 16.7 mM for two consecutive days. EDN was intraperitoneally injected into HFD/STZ-induced diabetic mice (10 mg/kg) every other day for 8 weeks.

STZ mice. Seven-week-old male C57BL/6J mice were purchased from the Model Animal Research Center of Nanjing University. After adaptation for 1 week, mice received consecutive daily intraperitoneal injections of STZ (50 mg/kg) after a 6-hour fast for 5 days; meanwhile, the control mice were intraperitoneally injected with only buffer (control group, n = 10). Blood glucose and UACR concentrations were monitored during the experiments. The mice were euthanized at week 12.

YY1-KD mice. YY1flox/flox mice (C57BL/6J background) were gifts from X. Li from the Zhongshan Hospital affiliated to Fudan University. An HFD/STZ-induced diabetic model was induced in YY1flox/flox mice. In situ renal injection of AAV2-Cre (YY1-KD group) or AAV2-GFP (control group) was performed at the cortex of the left kidney at three independent points (n = 7 for control group and n = 6 for YY1-KD group, respectively). Blood glucose was monitored once per week, and murine urea samples were collected every 2 weeks. Twelve weeks after surgery, the mice were euthanized to examine the severity of their diabetic kidney lesions.

YY1-OE mice. C57BL/6J mice were fed an HFD and injected with STZ as described above. Afterward, mice received in situ renal injection with AAV2-GFP (control group) or AAV2-YY1 (YY1-OE group) at three independent points (n = 10). Blood glucose was monitored once per week, and murine urea samples were collected every 2 weeks. Twelve weeks after surgery, all mice were euthanized to harvest tissues.

5/6 Nx mice. Eight-week-old male C57BL/6J mice were subjected to 5/6 Nx (n = 6 control group and n = 7 5/6 Nx group) in two stages. In the first stage, the left kidney was decapsulated, the upper and the lower poles were partially resected, and a gelatin sponge was used immediately to prevent bleeding. In the second stage, the entire right kidney was removed by renal arteriovenous ligation. Sham-operated mice were included as controls. Eight weeks after surgery, kidneys were harvested.

Statistical analysis

Data are presented as means ± SEM. Wilcoxon signed-rank test, Student’s t test, or one-way or two-way analysis of variance (ANOVA) was used to compare means of numerical variables where appropriate. The normality of the data was analyzed by the Sharpiro-Wilk test and quantile-quantile plots. Student’s t test was used to compare means of the initial data after normal distribution, and Wilcoxon signed-rank test was used to study data after non-normal distribution. If any statistically significant difference was detected, post hoc comparisons were performed using the least significant difference test. Chi-square or Fisher’s exact tests were used for comparisons of proportions of categorical variables. Statistical analysis was performed with SPSS (version 25.0) or GraphPad Prism (version 5.0) software. P < 0.05 was considered significant.


Materials and Methods

Fig. S1. Transcription factors binding to the promoter of TGFB1 identified by mass spectrometry–based DNA-protein interaction screening.

Fig. S2. Expression of YY1 in male and female db/db mice at different time periods.

Fig. S3. Transmission electron microscopy analysis of glomerular lesions in the kidneys of control and 5/6 Nx mice.

Fig. S4. The effects of siNrf2 alone or in combination with siYY1 on the transcriptional activity of TGFB1 in HRMCs.

Fig. S5. Pharmacological activation of endogenous YY1 protects mouse kidneys from diabetic lesion development.

Fig. S6. Off-target effects and toxicity analysis of EDN in the regulation of DN progression.

Fig. S7. Intervention of EDN in db/db mice ameliorates DN progression.

Fig. S8. A graphical abstract of this study.

Table S1. Clinical characteristics of DN in biopsy samples.

Table S2. List of the compounds that increased the luciferase activity of YY1 but decreased that of TGFB1.

Table S3. The sequences of siRNAs used.

Table S4. Primers used in DNA pull-down assays.

Table S5. Primers used in qRT-PCR.

Table S6. Primers used in ChIP assays and site-directed mutagenesis assays.

Table S7. Primers used in EMSA.

Data file S1. Raw data from mass spectrometry analysis.

Data file S2. Data plotted in figures.


Acknowledgments: We thank D. Wan for editing the figures. We appreciate X. Li from Zhongshan Hospital Fudan University for providing the YY1flox/flox mice. Funding: This work as supported by the following grants received by J. Liu: National Science Fund for Excellent Young Scholars (31722028), Training Program of the Major Research Plan of the National Natural Science Foundation of China (91857111), National Key R&D Program of China (2018YFA0800600), National Natural Science Foundation of China (81770797), National Science and Technology Major Project (2018ZX09711002-001-008), Shanghai International Cooperation and Exchange Program (16410723000), Thousand Young Talents Program of China (2016), and SJTU-CUHK Joint Research Collaboration Fund. This study was also supported by the Shanghai Sailing Program (17YF1414200 to P.G.) and the National Natural Science Foundation of China (31500667 to C.P.). Author contributions: J. Liu designed all the experiments. P. Gao, L. Li, and J. Liu performed all of the experiments with assistance from D. Gui, J. Zhang, J. Han, J. Wang, N. Wang, J. Lu, S. Chen, L. Hou, H. Sun, L. Xie, J. Zhou, C. Peng, Y. Lu, X. Peng, C. Wang, and Y. Huang. J. Liu, P. Gao, and C. Peng provided funding. J. Liu wrote the paper with assistance from P. Gao, L. Li, L. Yang, W. Jia, U. Ozcan, and J. Miao. Competing interests: J. Liu, P. Gao, L. Yang, and W. Jia are inventors on a patent application (201810736778.7) submitted by Shanghai Jiao Tong University Affiliated Sixth People’s Hospital that covers “The protective effects of eudesmin on prevention and treatment of renal fibrosis.” P. Gao, J. Liu, L. Yang, and W. Jia are inventors on a patent application (201810735931.4) submitted by Shanghai Jiao Tong University Affiliated Sixth People’s Hospital that covers “A novel effect of recombinant virus on prevention of renal fibrosis.” Data and materials availability: All data associated with this study are present in the paper or the Supplementary Materials. The medical records of clinical patients are archived in the Archive Center of Shanghai Jiao Tong University Affiliated Sixth People’s Hospital, Shanghai, China.

Stay Connected to Science Translational Medicine

Navigate This Article