Diabetic kidney disease (DKD) remains the most common cause of kidney failure, and the treatment options are insufficient. Here, we used a connectivity mapping approach to first collect 15 gene expression signatures from 11 DKD-related published independent studies. Then, by querying the Library of Integrated Network-based Cellular Signatures (LINCS) L1000 data set, we identified drugs and other bioactive small molecules that are predicted to reverse these gene signatures in the diabetic kidney. Among the top consensus candidates, we selected a PLK1 inhibitor (BI-2536) for further experimental validation. We found that PLK1 expression was increased in the glomeruli of both human and mouse diabetic kidneys and localized largely in mesangial cells. We also found that BI-2536 inhibited mesangial cell proliferation and extracellular matrix in vitro and ameliorated proteinuria and kidney injury in DKD mice. Further pathway analysis of the genes predicted to be reversed by the PLK1 inhibitor was of members of the TNF-α/NF-κB, JAK/STAT, and TGF-β/Smad3 pathways. In vitro, either BI-2536 treatment or knockdown of PLK1 dampened the NF-κB and Smad3 signal transduction and transcriptional activation. Together, these results suggest that the PLK1 inhibitor BI-2536 should be further investigated as a novel therapy for DKD.
Diabetic kidney disease (DKD) is the most common cause of end-stage renal disease in the U.S. (1). Intensive glycemic control, blood pressure control, and blockade of the renin-angiotensin-aldosterone system are considered standard of care for the treatment of patients with DKD (2–4). Recent clinical trials suggest that sodium–glucose cotransporter 2 inhibitors are effective in slowing down the progression of DKD, providing an additional treatment option (5). However, these therapeutic regimens are insufficient because they provide only partial protective effects (6). Therefore, there is an urgent need to develop more effective therapies for DKD.
Over the last decade, omics technologies have been widely applied to study the global molecular changes that occur in diseases such as kidney disease. These omics data sets include, for example, transcriptomics, epigenomics, proteomics, and metabolomics, and are publicly available and ready for reuse and integration (7,8). From these data sets, disease-specific signatures could be identified via computational analyses. Connectivity mapping (CMAP) (9) is a systems pharmacology approach that can be used to identify drugs and small molecules that may reverse or mimic disease-specific signatures. This approach bypasses the need to identify specific drug targets and experimentally screen individual small molecules in the disease context. A seminal contribution to the CMAP approach was a large-scale study conducted at the Broad Institute (10). The CMAP study produced gene expression signatures for almost all U.S. Food and Drug Administration (FDA)-approved drugs using cDNA microarrays where single small molecules were applied to four human cell lines in different concentrations and gene expression was measured after 6 h. This data set enabled the repurposing of many drugs for new indications and therapies. We previously used the CMAP data set to deduce combinational therapy for kidney disease, where we identified an additive beneficial effect for combining ACE inhibitors and histone deacetylase (HDAC) inhibitors (11). More recently, the CMAP resource was expanded to include more than one million signatures (12) under the auspices of the National Institutes of Health (NIH) Library of Integrated Network-based Cellular Signatures (LINCS) Common Fund program (13). Such expansion was made possible due to the L1000 technology, a high-throughput transcriptomic technology that directly measures 978 landmark genes. The initial release of the L1000 data set measured the changes in gene expression before and after treatment of >63 human cells with >20,000 small-molecule compounds including most FDA-approved drugs (12). This unique data set was developed as a resource to assess the global effects of small molecules and drugs on human cells as well as serve as a search engine to identify compounds that could therapeutically reverse a disease state or identify potentially toxic small molecules that mimic a disease state (14). However, whether the connectivity mapping approach works in practice remains questionable for many disease contexts. In querying of a disease signature against a connectivity mapping resource, there are several considerations. For example, does the disease signature reliably represent the disease state, and will the drugs identified to reverse the disease signature work the same way in different cells and tissues? One of the ways this issue can be overcome is by consideration of consensus signatures (15). Therefore, in this study, we first collected gene expression signatures for DKD from independent publications in which transcriptomic signatures were generated by different independent laboratories. Such collection enabled us to arrive at a consistently reproducible consensus DKD signature. Then, using the LINCS L1000 data set, we queried each DKD gene signature to prioritize small molecules and drugs that could potentially reverse the DKD gene signatures. From this computational screen, we identified BI-2536, an inhibitor of polo-like kinase 1 (PLK1), as one of the highly ranked drugs. To test the potential of this compound to induce a positive effect as a potential treatment for DKD, we treated OVE26 mice, a transgenic model of type 1 diabetes, with BI-2536. Indeed, we observed that this compound ameliorated DKD progression.
Research Design and Methods
BI-2536 was purchased through Selleck Chemicals (cat. no. S1109; Houston, TX). According to the manufacturer’s information, BI-2536 inhibits PLK kinase family (IC50 of 0.83 nmol/L for PLK1, 3.5 nmol/L for PLK2, and 9.0 nmol/L for PLK3) and bromodomain 4 (BRD4) with a Kd of 37 nmol/L. Our initial screen of cellular toxicity of BI-2536 in cultured mesangial cells at a range of 1–10,000 nmol/L using the LDH assay did not detect significant cellular toxicity, but we observed a significant biological effect of BI-2536 in mesangial cells at the dose of 10–100 nmol/L. Therefore, doses at this range were used as indicated in each analysis. As for in vivo studies, BI2536 has been used at 50 mg/kg once or twice per week for cancer therapy in mice (e.g., 16,17). Therefore, we initially selected 20 mg/kg daily in the pilot studies and were able to progressively reduce to 10 mg/kg every 2 days in the final experiments. We did not observe any obvious toxicity in mice treated with such doses of BI-2536.
Immortalized mouse mesangial cells were obtained from Creative Bioarray (CSC-I9238L) and cultured according to the manufacturer instructions. Conditionally immortalized mouse podocytes and immortalized mouse glomerular endothelial cells were cultured as described in a previous study (18,19). For high-glucose treatment, cells were first starved for 4 h with serum-free DMEM medium (5 mmol/L glucose) and treated with either high-mannitol (5 mmol/L glucose + 25 mmol/L mannitol) or high-glucose (total of 30 mmol/L glucose [HG]) medium. BI-2536 compound was added to 100 nmol/L final concentration.
Lactate dehydrogenase (LDH) cytotoxicity was measured using a colorimetric assay kit (cat. no. 88953, Pierce LDH Cytotoxicity Assay Kit; Thermo Fisher Scientific) according to the manufacturer’s protocol with addition of vehicle or BI-2536 with their concentrations as indicated in the text.
Cell Proliferation Assay
Cell proliferation was assessed by EdU incorporation and 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium) (MTS) assay. Click-iT Plus EdU Cell Proliferation Kit for Imaging, Alexa Fluor 594 dye (cat. no. C10639; Thermo Fisher Scientific), was used according to the manufacturer’s protocol. The result is quantified by the number of EdU-positive cells divided by DAPI-positive cell number. MTS assay was used for a time course stimulated by HG (30 mmol/L) with or without BI-2536 (100 nmol/L) for 24 h and 48 h. The optical density was measured using a colorimetric assay kit (cat. no. ab197010, MTS Assay Kit; Abcam) according to the manufacturer’s protocol.
RNA Extraction and Real-time PCR
Total RNA was extracted with use of Trizol (Invitrogen, Carlsbad, CA); 500 ng total RNA was reverse transcribed to cDNA with the SuperScript III First-Strand Synthesis System (Invitrogen). Real-time quantitative RT-PCR was performed with the 7500 Real-Time PCR System (Applied Biosystems, Waltham, MA). Gene level was normalized to GAPDH and expressed as fold change. The primer sets used were synthesized by Sigma-Aldrich (St. Louis, MO) and listed as follows: mPlk1 (forward, 5′-GGTTTTCAATCGCTCCCAGC-3′; reverse, 5′-AGGGGGTTCTCCACACCTTT-3′), mGapdh (forward, 5′-TCGTCCCGTAGACAAAATGG-3′; reverse, 5′-AATTTGCCGTGAGTGGAGTC-3′), and mCol4a1 (forward, 5′-TGTTGACGGCTTACCTGGAGAC-3′; reverse, 5′-GGTAGACCAACTCCAGGCTCTC-3′).
Knockdown of PLK1
Lentivectors expressing shRNA for mouse Plk1 (pLKO.1-shplk1; four independent clones) and scrambled shRNA (pLKO.1-shScrambled) were obtained from Sigma-Aldrich. Together with psPAX2 packaging and VSVG envelope plasmids, pLKO.1 plasmids were transfected in the human embryonic kidney (HEK)-293T cells with use of PolyJet transfection reagent (SignaGen Laboratories, Rockville, MD) according to the manufacturer’s manual. Supernatant was collected 48 h posttransfection, harvested, and concentrated.
Cells were homogenized in NP-40 lysis buffer containing protease and phosphatase inhibitor cocktail. Equal amounts of protein samples were separated on SDS polyacrylamide gel, transferred to polyvinylidene fluoride membranes (Millipore), and probed with primary antibodies phosphorylated (p)-Smad3 (cat. no. 9520; Cell Signaling Technologies), total Smad3 (cat. no. 9523; Cell Signaling Technologies), p-p65 (cat. no. 9420; Cell Signaling Technologies), p65 (cat. no. 4764; Cell Signaling Technologies), PLK1 (cat. no. 05-844; Millipore), and GAPDH (cat. no. 2118; Cell Signaling Technologies).
The SBE4-Luc plasmid and NF-κB(5x)-Luc plasmid were purchased from Addgene (cat. nos. 16495 and 49343, respectively). Renilla luciferase reporter plasmid (pRL) was purchased from Promega. HEK-293T cells seeded in 12-well plates (∼60% confluence) were cotransfected with SBE4-Luc (0.5 mg), NF-κB(5x)-Luc (0.5 mg), and pRL plasmids (0.2 mg) with use of the PolyJet transfection kit according to the manufacturer’s instructions (SignaGen Laboratories); 48 h posttransfection, cells were treated with assigned concentrations of BI-2536 with or without 10 ng/mL TGF-β1 for 16 h. Luciferase activities were measured with the Dual-Luciferase Reporter Assay kit (cat. no. E2910; Promega). Data are expressed as the ratio of firefly luciferase activity to Renilla luciferase activity.
OVE26 mice on FVB/N genetic background were purchased from The Jackson Laboratory. The genotyping primer sets used were as follows: INS2 (forward, 5′-ACTCCAAGTGGAGGCTGA GA-3′; reverse, 5′-TCCTTCCACAAACCCATAGC-3′). Male OVE26 mice were used for the study, and male wildtype littermates were used as controls. Mice in the treatment group received BI-2536 dissolved in 30% polyethylene glycol-400, 0.5% Tween-80, and 5% propylene and diluted in saline in vivo by gavage at a dose of 10 mg/kg body wt every 2 days (n = 6 in each group). Mice in the control group received the same volume of vehicle solution. All mice were treated at the age of 8 weeks for 8 weeks and euthanized at the age of 16 weeks. All mice used in this study were on an institutional animal care and use committee–approved protocol. All mice were treated humanely under the standards described in the Guide for the Care and Use of Laboratory Animals (20).
Monitoring Blood Glucose
Blood glucose was measured with a blood sample from the tail vein using a glucometer (Accu-Chek Aviva). Since the maximum upper limit of the glucometer is 600 mg/dL, in cases when the maximal values were obtained, the actual glucose level may have been higher in mice. Therefore, whether BI-2536 may have lowered blood glucose to ranges beyond the maximal detection in mice cannot be determined with this method.
Urine Albumin and Creatinine
Spot urine was collected by applying pressure to the transabdominal area, and 24-h urine was collected using metabolic cages. Urine albumin was determined with a commercial assay from Bethyl Laboratory, Inc. (Houston, TX). Urine creatinine levels were measured in the same samples with the Creatinine Colorimetric Assay Kit (Cayman Chemical, Ann Arbor, MI) according to the manufacturer’s instructions. Urinary albumin-to-creatinine ratio (ACR) was calculated as follows: ACR (mg/mg) = urine albumin (mg/dL) / urine creatinine (mg/dL).
Kidney samples were fixed in 10% formalin, embedded in paraffin, and sectioned to 4-μm thickness. Periodic acid Schiff (PAS) staining for the determination of mesangial matrix expansion was carried out as described by Zheng et al. (21). Images were taken at ×400 magnification with a Zeiss camera system microscope (Carl Zeiss Jena, Toronto, Ontario, Canada). These images were then compared with a set of standard images in which no expansion was scored as 1, scores from 2 to 4 were based on standards with the progressive expansion of PAS-stained matrix, and a score of 5 indicated the presence of Kimmelsteil-Wilson nodules. The same randomly ordered images were then blindly scored from 1 to 5 for the severity of matrix expansion, based on standard images. Mouse glomerular volume (VG) was calculated from the cross-sectional area with the formula VG = β/k (AG)3/2, where β = 1.38 is the shape coefficient for a sphere and k = 1.1 is the size distribution coefficient. Also, mesangial expansion was defined as a PAS-positive and nuclei-free area in the mesangium. Quantification of mesangial expansion was based on a minimum of 30 glomeruli per section in a blinded fashion, under ×400 magnification. The relative mesangial area was expressed as mesangial/glomerular surface area (%). Tissues for electron microscopy were fixed in 2.5% glutaraldehyde with 0.1 mol/L sodium cacodylate (pH 7.4) for 72 h at 4°C. Samples were further incubated with 2% osmium tetroxide and 0.1 mol/L sodium cacodylate (pH 7.4) for 1 h at room temperature. Ultrathin sections were stained with lead citrate and uranyl acetate and viewed on a Hitachi H7650 microscope. Briefly, negatives were digitized, and images with a final magnitude of up to ×12,000 were obtained. ImageJ 1.26t software (NIH) (rsb.info.nih.gov) was used to measure the length of the peripheral glomerular basement membrane (GBM), and the number of slit pores overlying this GBM length was counted. The arithmetic mean of the foot process width (WFP) was calculated as follows: WFP = π/4 × (∑GBM length)/(∑slits), where Σslits indicates the total number of slits counted, ΣGBM length indicates the total GBM length measured in one glomerulus, and π/4 is the correction factor for the random orientation by which the foot processes were sectioned.
Immunofluorescence staining was conducted on paraffin-embedded kidney sections with use of standard procedures. Briefly, deparaffinized sections were incubated with primary antibody diluted 1:100 and incubated at 4°C for 16 h: anti-WT1 (cat. no. 89901; Abcam), anti-PLK1 (cat. no. 05-844; Millipore), anti-collagen IV (cat. no. AB756P; Millipore), anti-F4/80 (cat. no. 14-4801-82; eBioscience), anti-podocalyxin (cat. no. FAB1556A; R&D Systems), anti-CD31 (no. 550274; BD Biosciences), and anti-PDGFR2 (cat. no. 14-1402-82; eBioscience). After being washed, sections were incubated with Alexa Fluor 488 or Alexa Fluor 568–labeled secondary antibody (cat. no. A11001, A11034, A11006, A11011, or A11004; Invitrogen) at room temperature for 1 h. Slides were counterstained with DAPI and mounted with use of Aqua-Poly/Mount (Polysciences, Inc.), and images were acquired with an AxioVision II Microscope with a digital camera (Carl Zeiss, Dublin, CA).
The upregulated and downregulated genes for each DKD study that showed BI-2536 as being a possible reversal drug were used to query L1000CDS2. If BI-2536 was returned as one of the top 50 reversers of DKD, the genes upregulated by DKD and downregulated by BI-2536 were collected. These genes were submitted to Enrichr for enrichment analysis. The Fisher exact test was applied for computation of enrichment P values.
Data are reported as mean ± SD. The unpaired two-tailed t test was used for comparison between two groups, and ANOVA followed by Tukey multiple comparison test was used for comparison between three or more groups. GraphPad Prism software was used for statistical analyses. All experiments were repeated at least three times, and representative experiments are shown. Data were considered statistically significant when P < 0.05.
Data and Resource Availability
The data sets generated during and/or analyzed during the current study are available from the corresponding author upon reasonable request.
Acquisition and Processing of Transcriptomic Data Sets for DKD From the Public Domain
Gene expression data sets from 11 previously published studies that compared diabetic nephropathy disease with normal kidney tissue were identified from two publicly available gene expression repositories: the Gene Expression Omnibus (GEO) (22) and Nephroseq (23) (Supplementary Table 1). The identified data sets from these resources were used to generate 15 DKD gene expression signatures. We analyzed the cDNA microarray data sets by first identifying the control/normal and disease samples and then performing differential expression (DE) analysis to identify diabetic nephropathy disease signatures characterizing the effect of DKD on mRNA expression. These data sets were analyzed with GEO2Enrichr (24), a browser extension that enables the analysis of gene expression samples directly from the GEO website. GEO2Enrichr uses the characteristic direction (CD) (25) algorithm to identify DE genes between the disease and normal samples. The RNA-sequencing (RNA-seq) data sets from Nephroseq were analyzed with our published RNA-seq analysis pipeline (26), which normalizes the gene-level raw counts to counts per million and then applies the CD algorithm to identify the top 500 DE genes. Next, we performed consensus analysis applied to the 15 DKD signatures from the 11 studies to prioritize small molecules that may be able to mitigate these signatures.
Prioritizing Drugs and Small-Molecule Compounds for Reversing DKD Signatures
To prioritize drugs and small-molecule compounds predicted to reverse the DKD gene expression signatures, we used L1000CDS2 (27), a signature search engine that uses a subset of the LINCS L1000 data for its underlying database (12). L1000CDS2 was queried with each DKD signature (top 500 DE genes) to prioritize matching signatures created from >20,000 drug and small-molecule treatments of multiple human cell lines. Signature similarity was computed with the cosine distance applied to the entire signature or a composite computation of the observed overlap between the sets of DE genes for each signature and for a consensus signature created by counting of the genes in the up and down lists across all signatures. These queries generated predictions for the top 50 small molecules that could reverse the 15 input signatures. These predictions were aggregated by counting of the number of times each drug appeared in the top 50 predicted reversers across all signatures (Fig. 1A and Supplementary Table 2). These counts were also compared with expected counts observed for these drugs in submission of queries of 826 signatures for other diseases extracted from GEO in a similar way (28). P values were calculated using a t test or a Welch t test if the variances were not equal (Supplementary Table 2). It should be noted that an individual drug, or a small-molecule compound, can appear multiple times within the top 50 matching signatures for each query, as these may have been profiled in multiple cell lines and applied in different concentrations and, where gene expression was measured, at different time points. Similarly, we counted the number of times genes occur as DE in the up/down sets for each signature (Supplementary Table 3). We then retained the genes with the most counts and submitted those as a query to L1000CDS2. Upregulated and downregulated genes that consistently occurred in more than four DKD signatures were included in the DKD consensus signature. This resulted in 128 upregulated and 73 downregulated genes that were consistent across the DKD signatures. The top 50 reversing drugs (Supplementary Table 4) include BI-2536, which appeared twice in the top 10 reversing drugs for the DKD consensus signature (Fig. 1B). The entire process is depicted in a workflow diagram (Fig. 1C).
We then explored previously established relationships between DKD and the most common drugs and small molecules from the top candidates that resulted from these integrative analyses. One of the top reversing drugs for DKD is dasatinib, a tyrosine kinase inhibitor that has multiple targets including Abl, Src, and c-Kit. Indeed, we recently showed that dasatinib attenuates kidney fibrosis in mouse models of kidney disease (29). However, a recent study suggests that dasatinib can cause podocyte toxicity (30) and may not be a suitable drug candidate for DKD. The sixth drug was I-BET-151, a bromodomain inhibitor. We have previously shown that a bromodomain inhibitor improves kidney injury in HIV-1 transgenic mice through inhibition of the inflammatory pathway (31). The fifth on the list was PD-0325901, an MEK inhibitor. The MAPK pathway has been shown to play a major role in kidney disease including DKD, and MAPK inhibitors have been shown to improve DKD (32,33). The third on the list was a PLK1 inhibitor, BI-2536. PLK1 is a serine/threonine protein kinase that is known to be involved in several important functions throughout the M phase of the cell cycle (34), as well as other cellular processes (35). However, its role in the setting of DKD has never been investigated. The second most common drug on the list, TG-101348, is a Jak2 inhibitor. Jak2 inhibitors have been shown to improve DKD in both animal models and patients with diabetes (36,37). The most common drug to reverse the 15 DKD signatures was CGP-60474, a cyclin-dependent kinase inhibitor, which has been shown to reduce adverse outcomes in endotoxemic mice by reducing TNF-α/NF-κB (38), and the role of TNF-α/NF-κB is well-documented in the pathogenesis of DKD (39).
From these candidates, we selected the PLK1 inhibitor BI-2536 for further experimental analysis, since its role has not been explored yet as a potential treatment for DKD.
Expression of PLK1 Increases in Mesangial Cells of the Diabetic Kidney
Querying the publicly available transcriptomics data sets from Nephroseq, we observed an increase in Plk1 mRNA in the glomeruli of eNOS-deficient diabetic db/db mice as compared with the nondiabetic controls (Fig. 2A). We further experimentally validated the increase in Plk1 mRNA and protein expression in the glomeruli of eNOS knockout mice with streptozotocin-induced diabetes (Fig. 2B) and diabetic OVE26 mice (Fig. 2C and D). Immunostaining of PLK1 with glomerular cell markers showed a predominant colocalization with the mesangial marker PDGFRβ but not with the endothelial marker CD31 or the podocyte marker podocalyxin (Fig. 2E). Moreover, we confirmed that the HG conditions led to enhanced Plk1 mRNA expression in cultured murine mesangial cells but not in podocytes or glomerular endothelial cells (Fig. 2F). These data suggest that PLK1 expression is upregulated mostly in mesangial cells in the diabetic kidneys.
BI-2536 Reduces Mesangial Cell Proliferation and Collagen IV Production in the Diabetic Condition
Since mesangial cell proliferation is one of the hallmark features of DKD (40), and PLK1 is a key regulator of the cell cycle, we next examined whether PLK1 inhibition by BI-2536 affected the mesangial cell proliferation under HG conditions. As anticipated, 100 nmol/L BI-2536 treatment reduced the proliferation of mesangial cells in HG condition, as assessed by the incorporation of the thymidine analog, EdU, and by the MTS assay (Fig. 3A–C). This reduction was not associated with cellular toxicity and cell death by BI-2536, as increasing the concentration of BI-2536 to 10 μmol/L did not increase the release of cellular dehydrogenase (LDH) (Fig. 3D). By immunostaining and quantitative PCR, we found that BI-2536 also inhibited TGF-β–induced collagen IV expression in mesangial cells (Fig. 3E–G).
BI-2536 Attenuates Diabetic Kidney Injury in Diabetic OVE26 Mice
We next determined the effects of BI-2536 on the progression of DKD in vivo in the type 1 diabetic OVE26 mouse model (19). Since OVE26 mice usually display significant albuminuria by 8 weeks of age (Supplementary Fig. 1A), we randomized the 8-week-old mice into either a treatment or a control group. The mice in the treatment group received BI-2536 at 10 mg/kg body wt every 2 days by oral gavage for 8 weeks (Supplementary Fig. 1B). Age-matched wild-type (WT) littermates and OVE26 mice treated with vehicle served as controls. All mice were euthanized at 16 weeks of age, and a schematic in Supplementary Fig. 1C illustrates the sample size used for each part of the analysis. BI-2536 treatment did not significantly alter blood glucose or blood pressure levels in OVE26 mice (Supplementary Fig. 2A and B). Although no significant change in body weight was observed in any groups, the kidney–to–body weight ratio increased in the diabetic mice in comparison with the WT controls, which was less pronounced in the OVE26 mice with BI-2536 treatment (Supplementary Fig. 2A and B). Notably, BI-2536–treated OVE26 mice had a marked reduction in albuminuria that was detectable starting at 4 weeks of treatment (12 weeks of age) by urine ACR of spot urine collection (Fig. 4A) and by quantification of 24-h urine albumin at the end of 8 weeks of treatment (16 weeks of age) (Fig. 4B). Importantly, the rise in blood urea nitrogen levels in OVE26 mice was also attenuated by BI-2536 treatment (Fig. 4C). Consistent with these changes, the histological analysis showed marked attenuation in mesangial matrix expansion, glomerular hypertrophy, and collagen IV deposition (Fig. 4D and E and Supplementary Fig. 3) and reduced infiltration of F4/80+ macrophages in the kidneys of OVE26 mice (Supplementary Fig. 3). The ultrastructural analysis of glomerular capillary loops by transmission electron microscopy also showed attenuation in the GBM thickening and podocyte foot process effacement (Fig. 5A and B). Quantification of podocyte numbers per glomerular cross section by WT1+ immunostaining showed a reduction in podocyte loss in diabetic mice by BI-2536 treatment (Fig. 5C). Taken together, these data suggest that treatment with BI-2536 attenuates diabetic kidney injury in OVE26 mice.
BI-2536 Reduces NF-κB and Smad3 Phosphorylation
To further determine the downstream pathways affected by PLK1 inhibition through BI-2536, we first performed an enrichment pathway analysis using Enrichr (41) of the differentially expressed genes that are induced by diabetes but reversed by BI-2536 treatment. Among such genes, NF-κB/Rel-A were highly enriched when querying the ChEA (the top term is RELA  ChIP-Seq FIBROSARCOMA Human P value = 6.757e−24, Fisher exact test) and TRRUST libraries (top term NFKB1 human P value = 1.045e−22). IgA nephropathy and nephropathy were the top enriched term when we queried the GWAS catalog library (IgA nephropathy P value = 0.000001851, nephropathy P value = 0.000005656), while TGF-β was the third top term from the Ligand Perturbations from GEO up library (TGF-β1 mouse renal mesangial MES-13 cells GDS1892 ligand:15 P value = 1.701e−27) This analysis can be accessed from https://amp.pharm.mssm.edu/Enrichr/enrich?dataset=b71ff72905c2c95dc42ff60f936e4909. We also submitted each gene set that had BI-2536 ranked in the top 50 based on L1000CDS2 (14 total) to Enrichr to perform enrichment analysis with the transcription factor gene set libraries ChEA and TRRUST (Fig. 6). We observed, similar to the observations for the consensus list, enrichment for NF-κB/RelA.
Since both NF-κB and Smad3 pathways are implicated in the pathogenesis of DKD (43,44), we next determined whether PLK1 inhibition affected the activation of NF-κB or Smad3. Inhibition of PLK1 by BI-2536 inhibited TNF-α–induced NF-κB p65 phosphorylation in mesangial cells in a dose-dependent manner (Fig. 7A). Similarly, the shRNA-mediated knockdown of PLK1 reduced the effects of TNF-α on NF-κB p65 phosphorylation (Fig. 7B). NF-κB luciferase assays confirmed the effects of BI-2536 on TNF-α–induced NF-κB phosphorylation (Fig. 7C). Similar to NF-κB, both BI-2536 and shRNA-mediated knockdown of PLK1 inhibited TGF-β–induced Smad3 phosphorylation (Fig. 8A and B). BI-2536 also inhibited the TGF-β–induced Smad3 luciferase activity in a dose-dependent manner (Fig. 8C). Taken together, these results suggest that the renoprotective effects of BI-2536 are mediated in part through its attenuation of NF-κB and Smad3 activation in mesangial cells.
Systems biology approaches have been used to predict drug targets, identify combinatorial therapies, and discover the mechanism of drug toxicity, as well as guide drug repurposing (11,30,45–47). In this vein, connectivity mapping approaches can help us to identify potential new drugs in an unbiased manner. Here, we present the first study that integrated most publicly available transcriptomic data sets from DKD to identify a consensus signature. Encouragingly, studies from different laboratories under different conditions and settings showed some level of reproducibility. Integration of available human and mouse DKD transcriptomic data sets helped us to identify small-molecule compounds for repurposing (46). We identified BI-2536, a PLK1 inhibitor, as one of the five top-ranked drugs. BI-2536 is a drug that is provided freely by Boehringer Ingelheim Open Innovation Portal opnMe. Interestingly, we found PLK1 to be expressed in mesangial cells and dysregulated in the kidney of patients and mice with DKD. We confirmed that BI-2536 was able to attenuate albuminuria and kidney injury in diabetic mice. Mechanistically, we found that BI-2536 inhibited mesangial cell proliferation, extracellular matrix synthesis, and NF-κB and Smad3 phosphorylation. Therefore, through this unbiased approach, we were able to identify a PLK1 inhibitor as a potential new therapy for DKD.
However, there are several limitations to our approach. The fact that the L1000 data were not obtained from kidney cells raises questions regarding the cell-type specificity and thus the relevance to kidney disease (46). Future studies are required to establish that this approach can reliably predict potential drugs for kidney disease. Also, as dysregulated genes identified from such a screen may be protective (48), the reversal of their expression may not be necessarily beneficial. Due to these limitations, it is critical to validate these computational predictions with biological experiments. In this study, we were able to validate that BI-2536 indeed attenuates kidney injury in diabetic mice.
As a serine-threonine kinase, PLK acts by binding and phosphorylating proteins that are already phosphorylated on a specific motif recognized by the POLO box domains (34). A large body of evidence suggests that PLK1, as a cell-cycle regulator, is an oncogene, and its expression increases in multiple cancers (49,50). Therefore, inhibitors of PLK1 have been developed for cancer therapy (51). However, recent studies suggest that PLK1 also acts as a tumor suppressor (52,53). Thus, the exact molecular mechanisms of PLK1 are complex and context dependent. Interestingly, we found that PLK1 expression is increased in the glomeruli of both human and mouse diabetic kidneys and localized mostly in mesangial cells. Because mesangial cell proliferation is a major feature of DKD (40), we suspected that a PLK1 inhibitor might attenuate DKD through the inhibition of mesangial cell proliferation. Indeed, in vitro results confirmed the reduced mesangial proliferation and in vivo results showed marked attenuation of diabetic glomerular injuries. To further dissect potential molecular mechanisms mediating the renoprotective effects of PLK1 inhibitors in DKD, we performed a pathway analysis of the gene signatures that were predicted to be reversed by BI-2536 treatment. We selected TNF-α/NF-κB and TGF-β/Smad pathways for experimental validations due to their enrichment in our pathway analysis and because of their critical role in DKD (39,43,44,54). We found that PLK1 regulates phosphorylation of both NF-κB p65 and Smad3. This is likely not the off-target effect of BI-2536 because knockdown of PLK1 expression also reduced phosphorylation of NF-κB and Smad3. However, future studies are required to determine whether this is a result of direct interaction and phosphorylation by PLK1 or indirectly through phosphorylation of an intermediate kinase.
Although our data suggest that PLK1 expresses mostly in mesangial cells, we showed that in vivo BI-2536 significantly attenuates podocyte injury and loss. This could be secondary to the improvement of mesangial cell proliferation because it is known that mesangial cells can interact with podocytes (55,56). However, a direct effect of BI-2536 in podocytes could not be completely ruled out. We chose OVE26 mice as a DKD model because these mice develop proteinuria and glomerular cell injury more consistently than db/db mice. However, validation of our findings in a type 2 diabetic model would further support the efficacy of BI-2536 in DKD. Additionally, as the detection of PLK1 protein by immunostaining on formalin-fixed paraffin-embedded human kidney sections was not optimal for analysis, future studies are also required to determine the specific cellular location of PLK1 expression and its level in human kidneys from patients with DKD.
Interestingly, BI-2536 is a dual PLK1 and bromodomain-containing protein 4 (BRD4) inhibitor with IC50 of 0.83 and 25 nmol/L, respectively (17). Since bromodomain inhibitors are also among the top five ranked drugs that are predicted to reverse the gene signatures in DKD, it is also plausible that the renoprotection observed with BI-2536 in OVE26 mice may be through attenuation of both PLK1 and BRD4 pathways, thereby impacting a larger network of signaling molecules in diabetic kidneys. BI-2536 has been used in phase II clinical trials for lung cancer but has potential sides including neutropenia, fatigue, nausea, vomiting, and constipation (57,58). Bromodomain inhibitors have been also used in clinical trials, and potential side effects include thrombocytopenia, anemia, dysgeusia, fatigue, and nausea (59–61). Therefore, when we design future clinical trials for CKD, careful consideration of monitoring these potential side effects is warranted. However, we believe that the dosages of BI-2536 required for CKD treatment might be lower than those for cancer therapy, and therefore we may see fewer side effects in the CKD patients.
In summary, we used a connectivity mapping approach to identify that the PLK1 inhibitor, BI-2536, is a potentially novel therapy for DKD. We validated this computational hypothesis experimentally to show the renoprotective effects of BI-2536 in DKD, uncovering a new role of PLK1 in mesangial cells through its regulation of cell proliferation and NF-κB and Smad3 phosphorylation. Together, the results of our study suggest that PLK1 inhibitor BI-2536 could potentially be developed as a novel therapy for DKD.
See accompanying article, p. 326.
This article contains supplementary material online at https://doi.org/10.2337/figshare.13061864.
Funding. J.C.H. is supported by National Institute of Diabetes and Digestive and Kidney Diseases, NIH, grants R01DK078897, R01DK088541, R01DK109683, and P01DK56492 and Veterans Affairs Merit Award IBX000345C. K.L. is supported by NIH R01DK117913-01. A.M. is supported by NIH grants U54HL127624 and U24CA224260. L.Z. is supported by the National Natural Science Foundation of China (grant 81900657) and Natural Science Foundation of Fujian Province China (grant 2020J011243).
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
Author Contributions. L.Z., K.L., A.M., and J.C.H. designed the study. L.Z., R.L., Z.L., J.L., and J.H. conducted the experiments and/or analyzed the data. Z.W., M.L.W., and A.M. performed the computational analysis. K.L., A.M., and J.C.H. drafted and revised the manuscript. All authors approved the final version of the manuscript. J.C.H. and K.L. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.