A multi-parameter, high-content, high-throughput screening platform to identify natural compounds that modulate insulin and Pdx1 expression

Abstract

Diabetes is a devastating disease that is ultimately caused by the malfunction or loss of insulin-producing pancreatic beta-cells. Drugs capable of inducing the development of new beta-cells or improving the function or survival of existing beta-cells could conceivably cure this disease. We report a novel high-throughput screening platform that exploits multi-parameter high-content analysis to determine the effect of compounds on beta-cell survival, as well as the promoter activity of two key beta-cell genes, insulin and pdx1. Dispersed human pancreatic islets and MIN6 beta-cells were infected with a dual reporter lentivirus containing both eGFP driven by the insulin promoter and mRFP driven by the pdx1 promoter. B-score statistical transformation was used to correct systemic row and column biases. Using this approach and 5 replicate screens, we identified 7 extracts that reproducibly changed insulin and/or pdx1 promoter activity from a library of 1319 marine invertebrate extracts. The ability of compounds purified from these extracts to significantly modulate insulin mRNA levels was confirmed with real-time PCR. Insulin secretion was analyzed by RIA. Follow-up studies focused on two lead compounds, one that stimulates insulin gene expression and one that inhibits insulin gene expression. Thus, we demonstrate that multi-parameter, high-content screening can identify novel regulators of beta-cell gene expression, such as bivittoside D. This work represents an important step towards the development of drugs to increase insulin expression in diabetes and during in vitro differentiation of beta-cell replacements.

Figure 1. Fluorescent protein expression in living beta-cells infected with dual reporter lentivirus.

(A,B) A representative group of MIN6 cells is imaged in the EGFP or RFP channels. (C) Merged image shows the heterogeneity in insulin and pdx1 promoter activity between single cells. Hoechst was used to counterstain the nuclei. (D) Differential interference contrast microscopy image of the MIN6 cells from panels A–C.

Figure 2. Correction of systematic row and column biases in 96-well format high-throughput data using a B score transformation.

(A) Overall column and row effects on medians of multiple parameters. The overall bias is shown for measurements of nuclear intensity (Hoechst; blue box plots), insulin expression (green), pdx1 expression (red) and cell number (purple). Box plots represent intraquartile range (IQR) with the dark line representing the median. Whiskers represent 1.5*IQR and outliers are any values exceeding that. (B) These systematic biases are corrected using a B score transformation. Overall column and row effects on cell measured after B-score transformation. Box plots for nuclear intensity (Hoechst; blue box plots), insulin (green), pdx1 (red) and cell number (purple). Box plots represent IQR with the dark line representing the median. Whiskers represent 1.5*IQR and outliers are any values exceeding that.

Figure 3. Results from multi-parameter high-content screening of marine invertebrate extract library.

B-score normalized values are shown for insulin activity (green), pdx1 activity (red), nuclear DNA staining intensity (blue), and cell number per well (purple). In each panel, the extracts are ranked according to their effect on cell count (bottom panel). Data are from five independent screens conducted on the MIN6 beta-cell line. Black lines around the center median indicate the median +/− 2*(Median Absolute Deviation; MAD), a measure of the spread of the data, analogous to the standard deviation. Hits (numbered circles and arrows) were defined as any extract whose SEM fell outside of the range of the median +/− 2*MAD. DMSO controls have been omitted from the data. Images with obvious artifacts (e.g. fibers) were omitted from data analysis.

Figure 4. Identification of Bivittoside D as a positive regulator of insulin gene expression.

(A–C) MIN6 cells were treated for 18 hours with multiple doses (0.2 to 2000 ng/mL) of the three purified fractions (F1, F2, F3) from the sponge extract identified as hit #3. (D) Cells were also treated with Holothurin A, which is found in similar sponges. RNA was isolated and quantitative qRT-PCR was conducted for Ins1 and Pdx1. Results were normalized to housekeeping genes (cyclophilin and beta-actin). n = 3. * P<0.05 compared to DMSO treated controls.